The Routledge Handbook of the Economics of Ageing 0367713322, 9780367713324

Citation preview

THE ROUTLEDGE HANDBOOK OF THE ECONOMICS OF AGEING

Ageing populations pose some of the foremost global challenges of this century. Drawing on an international pool of scholars, this cutting-edge Handbook surveys the micro, macro, and institutional aspects of the economics of ageing. Structured in seven parts, the volume addresses a broad range of themes, including health economics, labor economics, pensions and social security, generational accounting, wealth inequality, and regional perspectives. Each chapter combines a succinct overview of the state of current research with a sketch of a promising future research agenda. This Handbook will be an essential resource for advanced students, researchers, and policymakers looking at the economics of ageing across the disciplines of economics, demography, public policy, public health, and beyond. David E. Bloom is Clarence James Gamble Professor of Economics and Demography at Harvard University, USA. Alfonso Sousa-Poza is Professor of Economics at the University of Hohenheim, Germany. Uwe Sunde is Professor of Economics at the University of Munich, Germany.

THE ROUTLEDGE HANDBOOK OF THE ECONOMICS OF AGEING

Edited by David E. Bloom, Alfonso Sousa-Poza, and Uwe Sunde

Designed cover image: Getty Images/corradobarattaphotos First published 2024 by Routledge 4 Park Square, Milton Park, Abingdon, Oxon OX14 4RN and by Routledge 605 Third Avenue, New York, NY 10158 Routledge is an imprint of the Taylor & Francis Group, an informa business © 2024 selection and editorial matter, David E. Bloom, Alfonso Sousa-Poza, and Uwe Sunde; individual chapters, the contributors The right of David E. Bloom, Alfonso Sousa-Poza, and Uwe Sunde to be identified as the authors of the editorial material, and of the authors for their individual chapters, has been asserted in accordance with sections 77 and 78 of the Copyright, Designs and Patents Act 1988. All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. Trademark notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging-in-Publication Data Names: Bloom, David E. (David Elliot), 1955– editor. | Sousa-Poza, Alfonso, editor. | Sunde, Uwe, editor. Title: The Routledge handbook of the economics of ageing / edited by David E. Bloom, Alfonso Sousa-Poza, and Uwe Sunde. Description: Abingdon, Oxon; New York, NY: Routledge, 2023. | Series: Routledge international handbooks | Includes bibliographical references and index. Identifiers: LCCN 2022030807 (print) | LCCN 2022030808 (ebook) | ISBN 9780367713324 (hardback) | ISBN 9780367713454 (paperback) | ISBN 9781003150398 (ebook) Subjects: LCSH: Aging—Economic aspects. Classification: LCC HQ1061 .R5947 2023 (print) | LCC HQ1061 (ebook) | DDC 305.26—dc23/eng/20220716 LC record available at https://lccn.loc.gov/2022030807 LC ebook record available at https://lccn.loc.gov/2022030808 ISBN: 978-0-367-71332-4 (hbk) ISBN: 978-0-367-71345-4 (pbk) ISBN: 978-1-003-15039-8 (ebk) DOI: 10.4324/9781003150398 Typeset in Bembo by codeMantra

CONTENTS

Figures Tables Contributors 1

xix xxiv xxvii

Introduction to the Handbook David E. Bloom, Alfonso Sousa-Poza, and Uwe Sunde 1.1 Why This Handbook? 1.2 A Brief Overview 1.2.1 Health 1.2.2 Pensions and Social Security 1.2.3 Income and Economic Growth 1.2.4 Work and Employment 1.2.5 Data and Measurement 1.2.6 Ageing and Personality 1.2.7 Regional Developments 1.3 Acknowledgments

1 1 4 4 6 8 9 11 13 15 17

PART I

Health

19

2

21

Modeling the Impact of Population Ageing on Future Fiscal Obligations Jay Bhattacharya 2.1 Introduction 2.2 Accounting Identity/Macroeconomic Approach 2.2.1 Description 2.3 Microsimulation/Markov Transition Approach 2.3.1 Description 2.3.2 Evaluation 2.4 Overlapping Generations/Microeconomic Approach 2.4.1 Description 2.4.2 Evaluation v

21 22 23 25 25 27 28 28 31

Contents

2.5 2.6 3

4

5

Selected Review of Papers and Results Suggestions for Future Work

31 37

Medical Innovations and Ageing: A Health Economics Perspective Volker Grossmann 3.1 Introduction 3.2 Trends in Health Expenditure, Morbidity, and Life Expectancy 3.3 Health Effects and Cost-Effectiveness of Medical Innovations 3.4 R&D Costs and Patent Values 3.5 Life-Cycle Considerations and the Value of Life 3.5.1 Lifetime Utility, Budget Constraint, and Optimization 3.5.2 Value of a Statistical Life 3.6 Morbidity, Healthcare Demand, and Medical R&D 3.6.1 Health Deficit Approach 3.6.2 Health Capital Approach 3.6.3 Non-Path-Dependent Mortality 3.6.4 Market Size Effects: Empirical Evidence 3.7 Effect of Health Innovations on Health Inequality 3.8 Avenues for Future Research 3.9 Conclusion

40

Medical Progress, Ageing, and Sustainability of Healthcare Finance Michael Kuhn 4.1 Introduction 4.2 Ageing and Medical Progress as Drivers of Health Expenditure Growth 4.2.1 Ageing: A Mixed Bag 4.2.2 Medical Progress, Income Growth, and Welfare State Institutions: Will the Culprit behind Spending Growth Please Stand Up? 4.3 Healthcare and Medical Progress as Drivers of Longevity 4.4 What Is It Worth? Value of Medical Progress 4.5 Endogenous Medical Progress: Empirical Evidence 4.6 Putting Things Together: Integrated Macroeconomic Modeling of Healthcare 4.6.1 Ageing in Macroeconomic Models with a Healthcare Sector 4.6.2 Medical Progress in Macroeconomic Models 4.6.3 Macroeconomic Models with Endogenous Medical Progress 4.7 Conclusions 4.7.1 The Ageing–Medical Progress–Healthcare Spending Nexus 4.7.2 Role of Economic Development 4.7.3 Role of Institutions 4.7.4 Type and Implementation of Medical Innovation 4.7.5 Distributional Concerns 4.7.6 Future Development

61

Technologies to Mitigate Cognitive Ageing Neil Charness 5.1 Introduction 5.2 Age-Associated Changes in Cognition

84

vi

40 41 45 46 47 47 48 49 49 50 51 51 53 54 54

61 63 63 64 65 67 68 68 69 70 72 73 73 73 74 74 74 74

84 86

Contents

5.3

5.4

5.5 6

7

8

The Role of Technology in Mitigation 5.3.1 Rehabilitation 5.3.2 Augmentation 5.3.3 Substitution Promising Technologies and Domains for Intervention 5.4.1 Prevention: Boosting Cognitive Health to Support ADLs and IADLs 5.4.2 Rehabilitation and Augmentation for Work 5.4.3 Augmentation and EADL Support Conclusions

87 88 89 91 92 93 94 95 96

Gender, Ageing, and Health Xiaoyan Lei and Yuqi Ta 6.1 Introduction 6.2 Gender Differences in Health among Older Adults 6.2.1 Life Spans 6.2.2 Physical Health 6.2.3 Mental Health 6.2.4 Cognitive Health 6.2.5 Subjective Well-Being 6.2.6 Health Impact of Birth Control 6.3 Explanations for Gender Health Gaps among Older Adults 6.3.1 Epidemiological Explanations 6.3.2 Biological Factors 6.3.3 Social Explanations 6.3.4 Contextual Factors 6.3.5 Methodological Explanations 6.4 Conclusion

102

Economics of Disease Prevention in the Elderly JP Sevilla 7.1 Introduction and Summary 7.2 The Economic Challenges of Ageing Populations and the Role of Disease Prevention in Addressing Those Challenges 7.3 Burden of Disease in the Elderly 7.4 Scope Setting 7.5 The Intrinsic and Instrumental Value of Elderly Health 7.6 Rationale for the Government’s Role in Optimal Investment in Elderly Health Promotion 7.7 Evidence on Value for Money of Health Promotion and Disease Prevention Policies 7.8 Does Government Invest Enough in Elderly Disease Prevention? 7.9 R&D 7.10 Conclusion

123

The Economics of Long-Term Care David N. F. Bell and Elizabeth Lemmon 8.1 Introduction 8.2 The Market for Long-Term Care

139

102 103 103 105 107 108 109 109 110 110 111 112 114 115 115

123 124 124 125 125 127 129 131 134 135

139 140 vii

Contents

8.3 8.4 8.5 8.6

8.7 9

The Contrast between Health Economics and the Economics of Long-Term Care Key Issues for the Economics of Long-Term Care Systematic Search for Articles Relating to the Economics of Long-Term Care Issues Identified by Systematic Search 8.6.1 Unpaid Care 8.6.2 Demand/Insurance 8.6.3 Expenditure 8.6.4 Labor Market Conclusion and Discussion

In Good and Bad Times—Associations between Spousal Health and Assortative Matching on Early-Life Factors in Europe Iris Kesternich, Bettina Siflinger, and James P. Smith 9.1 Introduction 9.2 Literature 9.3 Data 9.4 Associations in Spousal Adult Health and Its Determinants across Regions 9.5 Assortative Matching on Childhood Background and Adult Health 9.5.1 The Matching Model 9.5.2 Results: Matching Attributes and Health Outcomes 9.6 Conclusion

10 Mental Health and Illness in Ageing Sherry Glied, Carolyn D. Gorman, and Richard Frank 10.1 Introduction 10.2 Common Characteristics of Mental Illness Relevant to Ageing 10.2.1 Compounded Consequences of Mental Illness for Social and Financial Well-Being due to Early Onset, Chronic Persistence, and Episodic Recurrence 10.2.2 Responsiveness to Economic, Social, and Personal Circumstances 10.3 Epidemiology of Mental Illnesses in a Life Course Context 10.3.1 Life Satisfaction 10.3.2 Suicide 10.3.3 Mortality and Morbidity 10.4 Policy Challenges Posed by Mental Illness in the Context of Ageing 10.4.1 Financial Security, Savings, and Investments 10.4.2 Social Capital, Social Networks, and Caregiver Burden 10.4.3 Implications for Health Systems 10.5 Concluding Observations

viii

142 143 143 145 145 151 152 153 153

158 158 160 161 167 169 169 170 174 178 178 179

179 181 182 183 183 185 186 187 188 189 190

Contents

PART II

197

Pensions and Social Security 11 Social Security Reforms in Heterogeneous Ageing Populations Miguel S´anchez-Romero and Alexia Prskawetz 11.1 Introduction 11.2 Redistribution in Pension Plans: The Internal Rate of Return 11.3 Mortality Gradient by SES and the Impact on the IRR 11.4 Correction of Pension Plans for the Mortality Gradient by SES 11.4.1 Contributions 11.4.2 Benefits 11.5 Directions of Future Research 11.5.1 Estimation of the Mortality Gradient by SES 11.5.2 Empirical Evidence on the Variation of the IRR across SES Groups 11.5.3 Impact of Pension Reforms on Inequality (Behavioral Reactions) 11.5.4 Transition Costs 11.5.5 Multi-Pillar Approach 11.5.6 Dealing with the Source of the Inequality Problem 11.6 Conclusion 11.7 Appendix 11.7.1 Determinants of the IRR 11.7.2 Independence of the IRR to Different Income Levels

199

12 Economic Preparation for Retirement Michael D. Hurd and Susann Rohwedder 12.1 Introduction 12.2 The Income Replacement Rate 12.2.1 Measurement of Replacement Rates 12.2.2 Shortcomings of the Income Replacement Rate 12.2.3 Financing of Consumption Out of Savings 12.2.4 Differential Mortality 12.2.5 The Role of Children 12.2.6 Life-Cycle Consumption Path Is Not Flat, Varies by Observables 12.3 Preparation for Retirement in the Context of a Structural Life-Cycle Model 12.4 Consumption-Based Measure of Economic Preparation for Retirement 12.5 Comparison with Income Replacement Rates 12.6 Summary and Conclusions 12.7 Future Research

217

13 Pension Policy in Emerging Asian Economies with Population Ageing: What Do We Know, Where Should We Go? George Kudrna, Philip O’Keefe, and John Piggott 13.1 Introduction 13.2 Ageing and Informality 13.2.1 Demographic Change and Population Ageing 13.2.2 Informality and the Elderly Population 13.3 Pension Policy ix

199 202 203 204 204 205 206 206 207 208 209 209 210 210 214 214 216

217 219 220 220 221 221 221 222 223 224 226 229 231

234 234 235 236 238 240

Contents

13.3.1 Multi-Pillar Pension Taxonomy 13.3.2 Pension Policy in Emerging EA and SEA Economies 13.4 Policy Directions—Expanding Social Pensions 13.4.1 Policy Objectives and Economic Analysis 13.4.2 Features of Social Pensions 13.4.3 Fiscal Costs of Targeted Social Pensions—An Example 13.5 Concluding Comments 14 Trends in Pension Reforms in OECD Countries Herv´e Boulhol, Maciej Lis, and Monika Queisser 14.1 Introduction 14.2 Extent, Limit, and Impact of Pension Reforms Over the Last Decades 14.2.1 Greater Recognition of Challenges Ahead but Uneven Pension Policy Achievements 14.2.2 Pension Reforms and the Political Process 14.2.3 More Efforts Needed to Assess the Impact of Enacted Reforms on Old-Age Income Inequality 14.2.4 The Shrinking Pension Gender Gap 14.3 New Avenues for Pension Policies 14.3.1 Policy Options to Increase Effective Retirement Ages and Their Limitations 14.3.2 The Question of Greater Sustainability of Pension Systems through Flexible Retirement 14.3.3 Pension Issues Related to Self-Employment 14.3.4 Impact of COVID-19 on Pensions 14.3.5 Consideration of Inequality in Life Expectancy in the Design of Pension Policies 14.3.6 News in the Debate between PAYG and Funded Pensions

241 242 249 250 252 254 256 262 262 263 263 264 266 267 268 268 271 272 273 274 275

PART III

285

Income and Economic Growth 15 Economic Growth, Intergenerational Transfers, and Population Ageing Ronald Lee 15.1 Introduction 15.2 The Recent Literature 15.3 The Age-Distributed Economy 15.4 Population Ageing, Economic Output, and Its Primary Distribution 15.5 Population Ageing and the Intergenerational Redistribution of Income 15.6 Insights from National Transfer Accounts 15.7 Summary Measures of Imbalances Created by Population Ageing 15.8 Estimated Impact of Population Ageing on Economic Flows Over the Next 50 Years 15.9 Summary Indices: General Support Ratio and Transfer Load 15.10Other Aspects of Population Ageing in Relation to Economic Growth 15.11Research Directions x

287 287 289 290 291 292 292 294 295 297 299 300

Contents

16 Consumption, Saving, and Wealth Accumulation at Old Age: Comparing Evidence from Developed and Developing Countries Marco Angrisani, Jinkook Lee, and Giacomo Rebellato 16.1 Introduction 16.2 Consumption, Saving, and Wealth in Developed and Developing Countries 16.3 Consumption, Saving, and Wealth at Older Ages 16.3.1 Developed Countries 16.3.2 Longevity Risk 16.3.3 Bequest Motives 16.3.4 Medical Expenditures 16.3.5 Developing Countries 16.4 Advanced Research on Consumption, Savings, and Wealth Accumulation Decisions at Old Ages: New Data from Developing Countries 16.5 Conclusion

303 303 304 306 306 308 308 309 310 312 313

17 Automation and Ageing Ana L. Abeliansky and Klaus Prettner 17.1 Introduction and Background 17.2 The Role of Automation in Compensating for Ageing: A Simple Theoretical Illustration and the Current State of Empirical Research 17.3 Potential Displacement of Workers by Robots 17.4 The Effects of Automation on Health 17.5 Important Future Research Questions 17.6 Conclusions

317

18 Working Life—Labor Supply, Ageing, and Longevity Andrew J. Scott 18.1 Introduction 18.2 Cross-Country and Historical Trends 18.2.1 Fact 1. Labor Force Participation Shows an Inverted U-Shaped Relationship with Age 18.2.2 Fact 2. Labor Force Participation of Men and Women Show Broadly Similar Variations with Age 18.2.3 Fact 3. Since the 1990s Labor Force Participation Rates in HighIncome Countries Have Been Rising, Reversing a Century-Long Period of Decline 18.2.4 Fact 4. Increased Employment in Older Age Groups Has Become an Important Source of Overall GDP Growth 18.2.5 Fact 5. While Employment at Older Ages Has Increased, It Is Highest and Increasing Fastest among Those with Higher Levels of Education 18.2.6 Fact 6. Older Workers Are More Likely to Be Engaged in Part-Time Work and Be Self-Employed 18.2.7 Fact 7. Older Age Groups Remain Active and Productive Outside of the Labor Market 18.3 The Age of Retirement 18.3.1 Optimal Retirement 18.3.2 The Link between Retirement and Longevity 18.4 Beyond Retirement

329

xi

317 319 321 322 323 325

329 331 331 331

332 332 332 333 335 335 336 337 340

Contents

18.4.1 Health 18.4.2 Wages 18.4.3 Hiring and Labor Demand 18.4.4 Education 18.4.5 The Nature of Work 18.5 Toward a Theory of Ageing 18.5.1 Different Concepts of Ageing 18.5.2 The Role of Time in Ageing 18.5.3 Are Older Workers Different? 18.6 Conclusions 19 Education and Ageing: Human Capital Investments and Ageing Andries de Grip and Raymond Montizaan 19.1 Introduction 19.2 Production of Human Capital through the Life Cycle 19.2.1 Lifetime Effects of General versus Vocational Initial Education 19.2.2 Learning by Doing 19.2.3 Learning from Peers in the Workplace 19.3 Skills Obsolescence 19.4 Lifelong Learning: Returns on Human Capital Investments of Older Workers 19.5 Training and Retirement of Older Workers 19.6 Research Agenda

340 341 341 342 342 343 343 344 344 345 349 349 350 351 352 352 353 354 356 357

PART IV

361

Work and Employment 20 The Employment of Older Workers Hippolyte d’Albis 20.1 Introduction 20.2 Changes in the Labor Supply 20.2.1 Health and Mortality 20.2.2 Education and Qualifications 20.2.3 Women’s Labor Force Participation 20.3 Technological and Structural Developments 20.3.1 The Productivity of Older Workers 20.3.2 Sectorial Transformations 20.3.3 The Transformations of Working Conditions 20.4 Social Security and Public Policies 20.4.1 Retirement Systems 20.4.2 Health Insurance 20.4.3 Labor Market Regulations

363

21 Retirement and Health Jan C. van Ours 21.1 Introduction 21.2 Setting the Stage—Cross-Country Comparison of Retirement-Related Statistics 21.3 How Retirement Affects Mental Health and Physical Health

381

xii

363 364 364 365 366 367 367 369 370 371 371 372 373

381 383 386

Contents

21.4 21.5 21.6 21.7 21.8

21.3.1 Cross-Country Studies 21.3.2 Single-Country Studies 21.3.3 Overview Studies Effects of Retirement on Mortality Cross-Partner Effects of Retirement What Have We Learned? What Can We Do? Where Do We Go from Here?

386 389 390 390 391 391 392 393

22 The Relevance of Cognition in the Context of Population Ageing Bernt Bratsberg, Ole Røgeberg, and Vegard Skirbekk 22.1 Background 22.2 Data 22.3 Education, Employment, and Earnings 22.4 Marriage, Cohabitation, and Fertility 22.5 Health and Mortality 22.6 Cohort Trends 22.7 Conclusion

396

23 Productivity in an Ageing World Axel B¨orsch-Supan and Matthias Weiss 23.1 Introduction 23.2 Methodological Challenges 23.2.1 Age-Productivity Profile of Individuals 23.2.2 Methodological Challenges at the Country Level 23.3 Age and Productivity at the Micro Level 23.4 Ageing and Productivity at the Country Level 23.5 Summary and Conclusions

410

396 398 399 401 403 404 406

410 415 415 417 418 424 428

24 Population Ageing and Gender Gaps: Labor Market, Family Relationships, and Public Policy 437 Paola Profeta 24.1 Introduction 437 24.2 The Relationship between Female Employment and Fertility 439 24.3 The Labor Market 441 24.3.1 The Labor Supply of Elderly Workers 441 24.3.2 The Labor Demand of Elderly Workers 442 24.3.3 The “Double Burden” of Age and Gender 443 24.3.4 Gender Gaps from the Labor Market to Pensions 443 24.4 Family Relationships 445 24.5 Public Policy 445 24.6 Conclusions 447

xiii

Contents

PART V

453

Data and Measurement 25 Measuring Ageing Holger Strulik 25.1 Introduction 25.2 The Force of Mortality 25.3 The Gompertz-Makeham Formula 25.4 Human Life Span and the Strehler–Mildvan Correlation 25.5 Reliability Theory 25.6 The Frailty Index 25.7 Individual Ageing 25.8 Ageing of Populations 25.9 Discussion and Conclusion

455

26 The Health and Retirement Study John W. R. Phillips and David R. Weir 26.1 Introduction 26.2 Phase 1: HRS Origins 26.3 Phase 2: The Steady State 26.3.1 The Steady-State Sample Design 26.3.2 Ancillary Studies 26.3.3 Internet 26.3.4 Dementia 26.4 Phase 3: Expanding Scope in the Core Survey 26.5 Phase 4: New External Initiatives

474

27 National Transfer Accounts and the Economics of Ageing Andrew Mason 27.1 The Generational Economy and NTA Foundations 27.2 NTA Illustrated 27.3 NTA around the World 27.4 Use of NTA to Understand the Economics of Ageing 27.4.1 Youth and Old-Age Deficits 27.4.2 The Demographic Dividend 27.4.3 A Longitudinal Perspective 27.5 Future Directions 27.6 Conclusion

486

28 Ageing and Dependency Warren C. Sanderson and Sergei Scherbov 28.1 Introduction 28.2 Economic Dependency 28.3 Noneconomic Dependency 28.4 Measures of Population Ageing without Dependency 28.5 Usefulness of the Old-Age Dependency Ratio Despite Its Problems 28.6 Conclusion

506

xiv

455 456 456 459 461 463 466 467 468

474 474 477 477 478 480 480 481 482

486 489 491 494 494 495 498 503 503

506 507 512 515 515 516

Contents

29 Patterns of Time Use among Older People Maddalena Ferranna, JP Sevilla, Leo Zucker, and David E. Bloom 29.1 Introduction 29.2 Literature Review 29.3 Data Resources 29.4 Patterns of Time Use by Age 29.5 Patterns of Time Use among Older People by Sociodemographic Characteristics 29.6 Research Needs and Opportunities

520 520 521 522 525 530 532

PART VI

537

Ageing and Personality 30 Ageing and Economic Preferences Thomas Dohmen, David Huffman, and Uwe Sunde 30.1 Introduction 30.2 Age and Economic Preferences 30.2.1 Measures of Preferences 30.2.2 Risk Attitudes 30.2.3 Patience 30.3 Identifying the Age Gradient 30.3.1 The Age-Period-Cohort Problem 30.3.2 Age and Longevity 30.4 Mechanisms Linking Preferences to Age 30.5 Outlook

539

31 Financial Literacy and Financial Behavior at Older Ages Olivia S. Mitchell and Annamaria Lusardi 31.1 Introduction 31.2 Retirement Planning, Debt, and Financial Fragility at Older Ages 31.3 Financial Literacy at Older Ages 31.4 Financial Literacy and Financial Behaviors in Later Life 31.5 Limitations and Extensions 31.6 Policy Implications and Next Steps

553

32 Age and the Value of Life Maddalena Ferranna, James K. Hammitt, and Matthew D. Adler 32.1 Introduction 32.2 Age and the Value per Statistical Life 32.3 The Social Welfare Function Approach 32.4 Age and the Social Value of Mortality Risk Reduction 32.5 Discussion and Conclusion

566

33 Happiness and Ageing in the United States David G. Blanchflower and Carol Graham 33.1 Introduction 33.2 The Existence of a U-Shape in Happiness and Life Satisfaction 33.3 Empirical Evidence

578

xv

539 540 540 541 542 543 543 544 544 546

553 554 555 559 560 561

566 568 570 572 574

578 582 585

Contents

33.3.1 Behavioral Risk Factor Surveillance System 2005–2017 (n=2,405,840) 33.3.2 Gallup U.S. Daily Tracker Poll, 2008–2017 (n=2,436,798) 33.3.3 General Social Surveys, 1972–2018 (n=60,054) 33.4 The Differences between the Married and the Non-Married 33.5 The Elderly 33.6 Conclusion 34 Ageing and Foreign Policy Preferences Mark L. Haas 34.1 Introduction 34.2 Population Ageing and Increased Preferences for International Peace 34.3 Ageing and Decreased Preferences for Internationalism 34.3.1 Ageing and Reduced Capabilities 34.3.2 Ageing and the Empowerment of Nationalist Leaders 34.4 Conclusion 35 Behavioral Science and Noncommunicable Diseases in Low- and Middle-Income Countries Nikkil Sudharsanan, Michael R. Eber, and Margaret McConnell 35.1 Introduction 35.2 Key Care Differences between NCDs and Infectious Diseases and Their Implications for Clinician and Patient Behavior 35.3 Behaviors of Health System Actors Needed for Delivery of High-Quality NCD Care 35.3.1 Knowledge 35.3.2 Patient Norms and Beliefs 35.3.3 Inertia, Attention, and Habit 35.3.4 Health System Barriers and Design Frictions 35.4 Patient Behaviors Important for NCD Management 35.4.1 Biased Beliefs and Mismatched Mental Models 35.4.2 Present Bias 35.4.3 Forgetting, Inattention, and Salience 35.5 How Might Interventions Improve Compliance with NCD Care? 35.5.1 Provider-Focused Interventions 35.5.2 Provider Interventions to Address Knowledge Gaps 35.5.3 Interventions to Address Clinical Inertia and Low Attentiveness 35.5.4 Provider Interventions to Address Incentives 35.5.5 Patient-Focused Interventions 35.5.6 Patient Interventions for Forgetting or Inattentiveness 35.5.7 Patient Interventions to Address Present Bias 35.5.8 Patient-Focused Interventions 35.6 Conclusions 36 The Implications of Population Ageing for Immigrant- and Gender-Related Attitudes Andreas Irmen and Anastasia Litina 36.1 Introduction 36.2 Ageing and Attitudes toward Immigration xvi

585 586 588 590 595 598 604 604 605 607 607 611 612

616 616 617 619 619 620 620 620 621 621 622 623 623 624 624 624 625 625 625 627 630 630 639 639 641

Contents

36.2.1 Individual Ageing 36.2.2 Population Ageing 36.2.3 Policy Implications 36.3 Ageing and Attitudes toward Women 36.3.1 Individual Ageing 36.3.2 Population Ageing 36.3.3 Policy Implications 36.4 Concluding Remarks

643 644 645 646 648 648 650 651

PART VII

655

Regional Developments 37 Global Ageing and Health Anna Reuter, Till B¨arnighausen, and Stefan Kohler 37.1 An Ageing World 37.2 Ageing Populations 37.2.1 Demographic Transition 37.2.2 Population Growth 37.2.3 Population Aged 65 or Over 37.3 Ageing Individuals 37.3.1 Life Expectancy 37.3.2 Healthy Life Expectancy 37.3.3 Morbidities and Mortality among the Elderly 37.4 Ageing and Other Megatrends 37.4.1 Chronic Conditions and Ageing 37.4.2 Universal Health Coverage and Ageing 37.4.3 Pandemics and Ageing 37.4.4 Climate Change and Ageing 37.4.5 Migration and Ageing 37.4.6 Family Structures and Ageing 37.4.7 Ageing in a Digital World 37.5 Opportunities and Challenges for an Ageing World 38 Social Protection and Population Ageing: A Comparative Analysis of India and Indonesia Mukul G. Asher and Chang Yee Kwan 38.1 Introduction 38.2 Demographic Dynamics in India and Indonesia 38.2.1 National Comparisons 38.2.2 Intra-Country Observations 38.3 Key Characteristics and Recent Initiatives 38.3.1 Overview 38.3.2 India: Recent Initiatives 38.3.3 Service Provision Component 38.3.4 Managerial Component 38.3.5 Co-Payment Component 38.3.6 Technological Component 38.3.7 Indonesia: Recent Initiatives xvii

657 657 658 658 658 659 661 661 662 663 664 664 664 666 666 666 667 667 668

671 671 673 673 675 679 679 680 680 682 683 683 684

Contents

38.4 Sustainability and Fairness: Major Challenges 38.5 Research Agenda 38.5.1 Actuarial Projections 38.5.2 Financial and Investment Management Literacy 38.5.3 Risk Management in the Payout Phase 38.5.4 Modernization of Social Protection Organizations 38.5.5 The Political Economy of Social Protection Reforms

685 686 686 687 688 688 688

39 Ageing in China Peng Nie and Yaohui Zhao 39.1 Introduction 39.2 Demographic Characteristics of Population Ageing 39.3 Economics of Ageing 39.3.1 Social Security 39.3.2 Retirement 39.4 Health Status and Healthcare of the Ageing Population 39.4.1 Health Status 39.4.2 Health Behaviors 39.4.3 Health Insurance and Healthcare Utilization 39.5 Family Support 39.5.1 Financial Support 39.5.2 Long-Term Care 39.5.3 The Future Role of Family 39.6 Discussion and Conclusions

691

40 Ageing in Latin America Bernardo L. Queiroz and B. Piedad Urdinola 40.1 Introduction 40.2 Measuring Ageing in the Region 40.3 Population Ageing and Economic Consequences 40.4 Overview of Social Support Systems in Latin America 40.5 Demographic Changes, Intergenerational Transfers, and Inequality 40.6 Demographics, Gender, and Time Use in Latin America 40.7 Migration, Gender Differences, and Ageing 40.8 Conclusion and Policy Implications

713

41 Population Ageing and Migration Panu Poutvaara 41.1 Introduction 41.2 Previous Literature 41.2.1 Self-Selection of International Migrants 41.2.2 Effects of Immigration in Destination Countries 41.2.3 Welfare Effects of Emigration in Origin Countries 41.2.4 Demographic Change and Global Migration 41.3 Theoretical Framework 41.4 Empirical Evidence 41.5 Research Agenda for the Future

735

Index

757 xviii

691 692 692 693 694 695 695 700 703 707 707 707 707 708

713 715 716 718 724 726 729 730

735 739 740 741 742 743 744 745 751

FIGURES

2.1 3.1 3.2 3.3 5.1 5.2 5.3 6.1 6.2 6.3 8.1 8.2 8.3 12.1 12.2 12.3 12.4 13.1 13.2 14.1 14.2

Schematic overview of FEM simulation Evolution of total health expenditure as percentage of GDP, 1970–2018 Evolution of remaining period life expectancy at age 65 (in years), 1960–2018 Year-of-birth effects in the relationship between the log of health deficits and age in the United States, birth cohorts 1910–1959 World population projections (millions) by older adult age groups age 65+ and age 85+ (millions) The Prevent-Rehabilitate-Augment-Substitute framework (Charness, 2020) outlining most (bottom) to least preferred technology interventions Percent Internet use by age group (18–29 years, 30–49 years, 50–64 years, 65+ years) and year (2000–2021) Sex ratios in more and less developed regions, males per 100 females Trends in life expectancy at age 60 by gender in more and less developed regions, years Trends in gender differences in life expectancy at age 60 and at age 80 in more and less developed regions, years Long-term care expenditure (health and social components) by government and compulsory insurance schemes, as a share of GDP, 2017 (or nearest year) Total number of documents per year (excludes 2021) Number of documents published in the top 11 sources Differential survival, females Fitted life-cycle consumption paths, single females Cumulative distributions (percent) of replacement rates, single and married persons Percent adequately prepared according to consumption-based measure, single and married persons, by income replacement rate Income sources for people aged 60–85 in EA and SEA Multi-pillar retirement income designs Average effective age of labor market exit and normal retirement age, OECD average 1970–2018 Replacement rates will fall in most OECD countries. Change in theoretical replacement rate between the 1940 and 1996 birth cohorts, in percentage points, full-career average-wage workers xix

27 42 43 44 85 87 89 104 105 106 141 144 145 222 223 227 229 240 241 265

267

Figures

14.3 Theoretical pensions of the self-employed are lower than those of employees. Theoretical pensions of a self-employed worker relative to an employee having both a taxable income (net income or net wage before taxes) equal to the average net wage before taxes, for individuals with a full career from age 22 in 2018 based on mandatory contributions 15.1 Total per capita income received by individuals by age and by source and uses made of it, United States (2015) 15.2 Projected changes in population age distributions interact with economic age profiles to generate differing budgetary stresses across nations 15.3 Projected changes in private, public, and total transfer load from baseline zero in 2020 to 2070 for selected countries, not all labeled 17.1 Worldwide stock of operative industrial robots in millions according to the International Federation of Robotics (2016, 2017, 2018a) 17.2 Average robot density, 2018 17.3 Old-age dependency ratio, 2018 18.1 Labor force participation rates (%) by age, 2019 18.2 Labor force participation rates (%) for men aged over 65 years 18.3 Percentage of population employed at older ages by education (Europe 65–74 years, United States over 65 years) 18.4 Distribution of average weekly hours worked by age group across the OECD 18.5 Proportion of time spent caring for others 18.6 Current and future retirement ages across the OECD 18.7 Dynamics of life expectancy increases and GDP growth 18.8 Modeling improvements in frailty by elongation and compression 18.9 Optimal retirement age and life expectancy 21.1 Life expectancy at age 65, effective retirement age and life expectancy after retirement – females and males; average over 11 OECD countries 1970–2018 22.1 Fraction of men with valid IQ score and educational attainment in analysis population, by birth year 22.2 Life-cycle education and employment, by CA and birth cohort group 22.3 Life-cycle earnings by CA and cohort group, unconditional (top panels) and conditional (bottom panels) on employment 22.4 Marriage and children over the life cycle, by CA and cohort group 22.5 Marriage rates by age for those married (top panels) and not married (bottom panels) at age 41, by CA and cohort group 22.6 Mortality and disability by age, by CA and cohort group 22.7 Relative socioeconomic, demographic, and health outcomes of high and low vs. medium CA groups at the age of 50 by birth cohort 23.1 Basic physiological parameters 23.2 Three estimates of the age-IQ relationship 23.3 Crystallized versus fluid intelligence 23.4 Interactions in the age-productivity relationship 23.5 Age and productivity over the life cycle 23.6 Age-productivity profiles in German plants by econometric methodology 23.7 Age profile for the number and severity of errors 23.8 Age and productivity in the assembly line of a German automotive plant 23.9 Age-specific productivity in the insurance industry xx

272 292 296 298 319 319 320 331 332 334 335 335 336 338 339 340 384 398 400 400 402 402 403 405 412 412 413 414 415 421 422 423 424

Figures

23.10 Old-age dependency ratio, GDP per person employed, and TFP, 1960–2019. Labor productivity and TFP smoothed 24.1 Female employment and fertility 24.2 Expenditure on family policies and pensions (as percentage of GDP) 25.1 Age-specific mortality rate: U.S. American men 2010–2019 25.2 Ageing according to Gompertz-Makeham vs. probabilistic death 25.3 The compensation effect of mortality 25.4 Ageing within and across 10 European countries 25.5 Health deficit transitions 26.1 Annual peer-reviewed journal publications, by topic 26.2 Evolution of the HRS sample design 27.1 The U.S. life cycle, 2015 27.2 Age reallocations in the United States in 2015 27.3 Per capita consumption and labor income by age, in purchasing power parity dollars, in 20 countries grouped by income class 27.4 Reallocation share for those 65 and older (family transfers, asset-based reallocations, and public transfers) for 27 countries, circa 2010 27.5 Youth (GAP0–24) and old-age (GAP65+) deficits as percentages of total labor income by income group, 2000, 2020, 2040, and 2060 27.6 Support ratio and first dividend in percent, 1950–2060, Brazil, China, Germany, Nigeria, and the United States 29.1 Patterns of time use by age in MTUS countries, India and China 29.2 Incidence rate and intensity rate of paid work, selected countries 29.3 Gender gap in time use by age in the MTUS countries and in China and India 29.4 Differences in time use among older people (60+) by sociodemographic characteristics, MTUS countries 2000–2010 (minutes per day) 30.1 Age and willingness to take risks 30.2 Age and patience 30.3 Stability of preferences: long-run correlations by cohort (Germany) 30.4 Age and preferences: global heterogeneity 31.1 Debt concerns and financial fragility in the U.S. population 31.2 Americans’ financial literacy by age 31.3 Americans’ financial literacy by age and sex 33.1 Life satisfaction in the United Kingdom 33.2 Mortality from drug overdose 33.3 BRFSS limited and with controls, 2005–2017 – limited controls include age dummies, state and year dummies and full controls adds personal controls for marital status, labor force status and education 33.4 BRFSS, life satisfaction, limited controls by marital status, 2005–2017 33.5 Cantril ladder, Gallup U.S. Daily Tracker, 2009–2017 33.6 Cantril ladder, Gallup U.S. Daily Tracker 2007–2017 33.7 Three-step happiness, GSS, 3-year smoothed averages, 1972–2018 33.8 Three-step happiness, GSS, with limited controls married and unmarried, 1972–2018 33.9 Three-step happiness with marriage (Q4), GSS, with marriage (married only), 1972–2018 33.10 Average life satisfaction by age 35.1 Key differences between infectious and noncommunicable diseases xxi

426 440 446 457 458 460 465 467 475 479 489 491 492 493 495 497 527 528 529 531 541 543 545 547 555 556 557 579 581

587 587 589 589 590 592 595 597 618

Figures

35.2 35.3 36.1 36.2 36.3 36.4 36.5 37.1 37.2 37.3 37.4 39.1 39.2 39.3 39.4 39.5 39.6 39.7 39.8 39.9 39.10 39.11 39.12 39.13 39.14 39.15 39.16 40.1 40.2 40.3 40.4 40.5 40.6 40.7 40.8 40.9 40.10 41.1

Clinician behaviors needed to deliver high-quality NCD care Patient-side behaviors important for NCD management Percentage of population aged 60 years or over by region, 1980–2050 Global population by broad age group, in 1980, 2017, 2030, and 2050 Selected European attitudes toward immigrants Attitudes toward women Organisation for Economic Co-operation and Development differences in daily time spent working in paid and unpaid work Past, present, and future population profiles, by World Bank country groups Life expectancy around the world, 1770–2100 Life expectancy and healthy life expectancy at age 65 by World Health Organization region, 1990–2019 HIV/AIDS and life expectancy in the world and South Africa Share of older people in China, 1950–2050 Retirement rates in China Self-reported health among older people aged 60+ Fractions of bad self-reported health by age, gender, and Hukou Rate of self-reported chronic diseases diagnosed by a doctor by age and gender Rate of reporting having difficulty/needing help for at least one activity of daily living or instrumental activity of daily living by age and gender Rate of reporting having difficulty/needing help for at least one activity of daily living or instrumental activity of daily living by age and Hukou Average mental intactness score by age, gender, and Hukou Average episodic memory score by age, gender, and Hukou Rates of severe depressive symptoms by age, gender, and Hukou Smoking status among older people aged 60+ by gender Smoking status among older people aged 60+ by Hukou Alcohol drinking among older people aged 60+ by gender Alcohol drinking among older people aged 60+ by Hukou Participation in vigorous/moderate/light physical activities among older people aged 60+ by age Participation in vigorous/moderate/light physical activities among older people aged 60+ by Hukou Demographic dependency ratios, Latin America, 2000–2050 First demographic dividend (% growth), Latin America, 1950–2100 Second demographic dividend (% growth), Latin America, 1950–2100 Public pension coverage (% elderly receiving benefits), Latin America, 2000–2020 Health (left-hand panel) and pension coverage (right-hand panel) in % by educational level, selected LA countries, 2000–2020 Working-age individuals contributing to public pension programs by age groups, Latin America, 2000–2020 Economic life cycle, Colombia, 2014 Public transfers age profile, Latin American countries in the NTA Project Time spent in care (measured in hours per week) in Colombia, Costa Rica, United States, and Uruguay Paid labor market work (measured in time) in Colombia, Costa Rica, United States, and Uruguay World population by region: Estimates and projections 1950–2100 xxii

619 621 640 640 642 647 649 660 662 663 665 693 695 696 696 697 698 698 699 700 701 701 702 702 703 704 704 715 718 719 721 722 723 724 725 727 728 736

Figures

41.2 41.3 41.4 41.5 41.6 41.7

Net migration rate by regions Migrant stock by region as percentage of population Population by age group: Estimates and projections 1950–2100 GDP per capita (2015–2019) and net migration rate (2015–2020) Relative cohort size (1995) and net migration rate (2015–2020) Population distribution by age groups: World population vs. migrant population 2020 41.8 Population distribution by age groups in Europe: Total population vs. migrant population 2020 41.9 Age group distribution: European population vs. incoming immigrant population from non-EU28 countries in 2019 41.10 Age group distribution of native Germans without migration background compared with first- and second-generation immigrants 2019

xxiii

737 738 746 747 748 748 749 749 750

TABLES

2.1 3.1 5.1 8.1 8.2 9.1 9.2 9.3 9.4 9.5 9.6 9.7 9.8 10.1 10.2 12.1 13.1 13.2 13.3 13.4 13.5 14.1

Key contributions Healthcare utilization and healthcare resources (selected indicators), 2018 or nearest year Information-processing parameters for young and old adults Main groupings of articles relating to the economics of long-term care Articles from search results ordered by number of citations Literature overview Sample selection: women of age 65–74 years Variable definitions Descriptive statistics by region and gender Regression coefficients of own health outcomes on partner’s health outcomes, controlling for both partners’ ages and ages squared Regression coefficients of own attributes on partner’s attributes, controlling for both partners’ ages and ages squared, female sample Estimates of the affinity matrix for three European regions, quadratic specification Correlations in matching characteristics and health outcomes with actual preferences and with counterfactual preferences Ages at which 25, 50, 75, and 99 percent of respondents are first affected by select mental disorders Prevalence of select mental disorders by country and age groups (including both males and females), 2017 (%) Percent adequately prepared, defined as 95–100 percent chance of dying with positive wealth Demographic drivers and population ageing in EA and SEA Share in percentage of informal employment in total employment (including agriculture) Pension policy in emerging EA and SEA countries Key indicators for social pensions in emerging EA and SEA countries Projections of social pension expenditures for Indonesia (percentage of GDP) Ten-year government bond and nominal GDP growth rates, selected countries, average by periods xxiv

32 42 86 146 149 162 164 165 166 167 168 171 172 180 184 226 236 238 242 244 255 276

Tables

17.1 Total fertility rate (TFR), life expectancy at birth (LEXP), and the old-age dependency ratio (OADR) of people older than 64 in relation to the working-age population in the G7 countries plus China and Russia in 1960 and 2018 318 18.1 Percentage of total employment growth explained by age group 50 years plus 333 18.2 Part-time employment as percentage of full-time employment by age, 2018/2019 333 19.1 Mincer’s earnings function 352 19.2 Causes of skills obsolescence 353 21.1 Life expectancy at age 65 (2018) and educational gap in life expectancy at age 30 (2015) 385 21.2 Labor market at old age, OECD countries, 2018 387 22.1 Average change across 1950–1969 birth cohorts in outcomes at age 50 of highand low-CA groups relative to reference group (medium score) 405 27.1 Examples of age reallocations 488 27.2 Life-cycle deficit in trillions of dollars and as a percentage of total labor income, United States, 2015 490 27.3 Adjustment in per capita consumption, relative to baseline consumption, to fund consumption for ages 45 and older 501 27.4 Steady-state consumption level relative to baseline with life-cycle old-age deficit funded entirely by transfers 501 27.5 Per capita life-cycle pension wealth by age, steady-state, U.S. survival rates for 1933 and 2019, population growth of zero and productivity growth rate of 0.015 per year 502 27.6 Pension wealth as a share of total labor income 503 28.1 Proportion of 65–74-year-olds with one or more IADLs in 2015 513 28.2 Equivalent ages based on the sum of 92 age-related DALYs in age-standardized populations, selected locations, 2017 514 28.3 Percentage increase from 2020 to 2050 in the proportion elderly and the traditional OADR, selected countries 517 29.1 Main sources for time use data 523 29.2 Definition of main time use categories 524 29.3 Characteristics of the countries included in the analyses 526 29.4 Demographic characteristics of the sample 531 31.1 Changes in financial literacy by age: panel data analysis 557 31.2 Regression of financial literacy index on socioeconomic factors 558 33.1 Drug overdose deaths by age 582 33.2 U.S. antidepressant prescription rates, August 2020–June 2021 583 33.3 OLS regressions of four-step life satisfaction, BRFSS, 2005–2011, 2013–2017 586 33.4 OLS regressions of Cantril’s 11-step ladder, Gallup U.S. Daily Tracker Poll, 2009– 2017 588 33.5 OLS regressions of three-step happiness, GSS, 1972–2018 591 33.6 Marriage and divorce rates 593 33.7 OLS regressions of seven-step life satisfaction, ISSP, 2017 594 35.1 Examples of patient-focused interventions to improve NCD care 626 36.1 Summary statistics for attitudes toward immigrants in low/high ageing countries 646 36.2 Summary statistics for attitudes toward women in low/high ageing countries 650 37.1 Common health metrics 661 38.1 Select demographic indicators 673 38.2 Select demographic indicators for India – states and union territories 676 38.3 Select provincial demographic indicators for Indonesia 677 xxv

Tables

39.1 Coverage rate in percentage of different health insurance schemes by Hukou 39.2 Rates of healthcare utilization in percentage by gender, Hukou, and type of medical insurance 40.1 Main features of pension programs in Latin America, circa 2010 41.1 Continental population shares in percent by age groups 41.2 Foreign-born population in selected countries

xxvi

705 706 720 746 751

CONTRIBUTORS

Ana L. ABELIANSKY is an Assistant Professor of Economics (especially macroeconomics and digitalization) at the Vienna University of Economics and Business (WU). Matthew D. ADLER is Richard A. Horvitz Professor of Law and Professor of Economics, Philosophy, and Public Policy, Duke University. His scholarship lies at the intersection of normative ethics and welfare economics and currently focuses on the use of social welfare functions as a framework for distributionally sensitive policymaking. Marco ANGRISANI is an economist at the Center for Economic and Social Research at the University of Southern California. Angrisani is a team member of the Understanding America Study and the Gateway to Global Ageing Data. Mukul G. ASHER is an economist specializing in public financial management, pension, and social protection reforms. He was with the National University of Singapore for 43 years, retiring in 2018. His research continues to focus on Asia. He is an Associate Editor of The Journal of the Economics of Ageing. ¨ Till BARNIGHAUSEN is an Alexander von Humboldt University Professor and the Director of the Heidelberg Institute of Global Health in the Faculty of Medicine at Heidelberg University. He is also senior faculty at the Wellcome Trust’s Africa Health Research Institute and fellow at the Harvard Center for Population and Development Studies. ORCID: https://orcid.org/0000-0002-4182-4212 David N. F. BELL is a Professor of Economics at the University of Stirling, Scotland. ORCID: https://orcid.org/0000-0002-4538-6328 Jay BHATTACHARYA is a Professor of Health Policy at Stanford Medical School and a Research Fellow at the National Bureau of Economics Research. He directs the Stanford Center on the Demography and Economics of Health and Ageing. He is an Associate Editor of The Journal of the Economics of Ageing. ORCID: https://orcid.org/0000-0003-3867-3174 David G. “Danny” BLANCHFLOWER is the Bruce V. Rauner Professor of Economics at Dartmouth, a part-time Professor of Economics at the Adam Smith Business School at the xxvii

Contributors

University of Glasgow and a Research Associate at the National Bureau of Economics Research. He is the author of “Not Working: Where Have All the Good Jobs Gone?” from Princeton University Press. ORCID: https://orcid.org/0000-0002-2856-7039 David E. BLOOM is Clarence James Gamble Professor of Economics and Demography in the Department of Global Health and Population at the Harvard T.H. Chan School of Public Health. He is a co-founding Editor of The Journal of the Economics of Ageing. ¨ Axel BORSCH-SUPAN is Director of the Munich Center for the Economics of Ageing (MEA) at the Max Planck Institute for Social Law and Social Policy. He is a member of the German and Austrian Academies of Sciences. He is an Associate Editor of The Journal of the Economics of Ageing. https://orcid.org/0000-0003-0470-8850 Herv´e BOULHOL is Senior Economist in the Pensions and Ageing unit of the Directorate for Employment, Labour and Social Affairs at the Organisation of Econmic Co-operation and Development. He graduated as a statistician, economist, and actuary from ENSAE (1989) and holds a PhD in economics from Universit´e Paris Panth´eon-Sorbonne (2007). Bernt BRATSBERG is Senior Research Fellow at the Ragnar Frisch Centre for Economic Research and Researcher at the Centre for Fertility and Health, Norwegian Institute of Public Health. His main research interests are labor economics and the economics of migration. Neil CHARNESS is William G. Chase Professor of Psychology, an FSU Distinguished Research Professor, and Director of the Institute for Successful Longevity at Florida State University. He received his BA from McGill University (1969) and MSc and PhD from Carnegie Mellon University (1971, 1974) in psychology. ORCID: http://orcid.org/ 0000-0002-1002-3439 Hippolyte D’ALBIS is a senior researcher at Centre National de Recherche Scientifique ´ ´ and Professor at the Paris School of Economics, Director of the Ecole des Hautes Etudes en D´emographie, and Associate Editor of the Journal of Demographic Economics and The Journal of the Economics of Ageing. ORCID: https://orcid.org/0000-0002-6409-4320 Andries DE GRIP is Professor of Economics at the Research Centre for Education and the Labour Market (ROA) of Maastricht University in the Netherlands. He is affiliated with IZA and Netspar. ORCID: https://orcid.org/0000-0002-1975-4553 Thomas DOHMEN is Professor of Applied Microeconomics at the University of Bonn. He is Spokesperson of the Cluster of Excellence ECONtribute: Markets & Public Policy, Adjunct Professor at Maastricht University, and Project Coordinator at IZA. He is a fellow of CESifo and DIW. ORCID: https://orcid.org/0000-0002-9321-0319 Michael R. EBER is a PhD candidate in health policy and decision science at Harvard University. Eber studies how behavioral insights can improve the design of policy interventions. His work has been supported by the U.S. Agency for Healthcare Research and Quality. Affiliations include Interfaculty Initiative in Health Policy, Harvard University, Cambridge, MA, and Center for Health Decision Science, Harvard T.H. Chan School of Public Health, Boston, United States. ORCID: https://orcid.org/0000-0002-1507-4876 xxviii

Contributors

Maddalena FERRANNA is Research Associate at the Harvard T.H. Chan School of Public Health. Her research focuses on the development and implementation of distributionally sensitive economic evaluation methods, with applications in global health and climate change policy. Richard FRANK is a Senior Fellow and the Director of the University of Southern California – Brookings Schaeffer Initiative on Health Policy. He is the Margaret T. Morris Professor of Health Economics Emeritus at Harvard Medical School and served as the Assistant Secretary for Planning and Evaluation at the U.S. Department of Health and Human Services. ORCID: https://orcid.org/0000-0003-1551-7072 Sherry GLIED is Dean and Professor of Public Service at New York University’s Robert F. Wagner Graduate School of Public Service. She previously served as Assistant Secretary for Planning and Evaluation at the U.S. Department of Health and Human Services. ORCID: https://orcid.org/0000-0001-9432-1662 Carolyn D. GORMAN is an Adjunct Fellow at the Manhattan Institute, where her research examines how policy changes in the United States impact individuals with serious mental illness. She is also an Associate Data Scientist at the JPMorgan Chase Institute and previously served as staff on the U.S. Senate Committee on Health, Education, Labor and Pensions. Carol GRAHAM is a Senior Fellow at Brookings, College Park Professor at University of Maryland, and a Gallup Senior Scientist. She is a world-recognized expert on wellbeing economics and has received awards for her pioneering research. She has degrees from Princeton. Johns Hopkins, and Oxford, and has published eight books and journal articles in Science and Social Science and Medicine, among many others. ORCID: https://orcid.org/ 0000-0001-6088-1960 Volker GROSSMANN is Professor of Macroeconomics at the University of Fribourg/Switzerland and research fellow of the Center for Economic Studies (CESifo) in Munich, Germany, and the Institute of Labor Economics (IZA) in Bonn, Germany. He is an Associate Editor of The Journal of the Economics of Ageing and the Swiss Journal of Economics and Statistics. ORCID: https://orcid.org/0000-0001-8917-3503 Mark L. HAAS is the Raymond J. Kelley Endowed Chair in International Relations and Professor of Political Science at Duquesne University, United States. He was formerly a National Security Fellow at the Olin Institute for Strategic Studies and an International Security Fellow at the Belfer Center for Science and International Affairs, both at Harvard University. James K. HAMMITT is Professor of Economics and Decision Sciences, Harvard T.H. Chan School of Public Health, Boston, MA, United States, and Associate, Toulouse School of Economics, Toulouse, France. He holds degrees in applied mathematics and public policy from Harvard University. David HUFFMAN is Professor of Economics at the University of Pittsburgh and a Faculty Affiliate at the Behavioral Economics Design Initiative, University of Pittsburgh, leading the Labor Markets Focus Area. He holds a PhD from the University of California, Berkeley, and is a fellow of CESifo and IZA. ORCID: https://orcid.org/0000-0002-4714-5560. xxix

Contributors

Michael D. HURD is Senior Principal Researcher, RAND Corporation, Santa Monica, USA and a member of NBER. His research interests include retirement, pensions, Social Security, consumption and saving, the cost and consequences of dementia, survey methods, and the properties and uses of subjective probabilities. He is a co-investigator of the Health and Retirement Study. ORCID ID: https://orchid.org/0000-0001-9167-8869 Andreas IRMEN is Professor of Macroeconomics and Applied Microeconomics at the University of Luxembourg and a fellow of the Center for Economic Studies (CESifo) in Munich, Germany. His research interests include growth theory, population, and cultural economics. ORCID: https://orcid.org/0000-0003-0410-8855 Iris KESTERNICH is Professor of Economics at the Universities of Hamburg and Leuven. Her research focuses on nonmonetary incentives in health and labor markets. She is Research Associate at the Leibniz Centre for European Economic Research (ZEW) and Associate Editor of the Journal of Economic Behavior and Organization. Stefan KOHLER leads a global health and economics research group at the Heidelberg Institute of Global Health of Heidelberg University. He holds a doctorate in economics from the European University Institute and a medical degree from the Charit´e – Universit¨atsmedizin Berlin. ORCID: https://orcid.org/0000-0003-1365-7506 George KUDRNA is a Senior Research Economist at the ARC Centre of Excellence in Population Ageing Research (CEPAR), University of New South Wales, Sydney. His research interests include pension economics, economics of ageing, and computational macroeconomics. He currently leads an Australian Research Council linkage project with the World Bank on modeling policy for ageing in emerging Asia. Michael KUHN is Director of the newly (2021) established Economic Frontiers Program at the International Institute for Applied Systems Analysis. He is an Associate Editor of The Journal of the Economics of Ageing. ORCID: https://orcid.org/0000-0003-0424-3221 Chang Yee KWAN is Non-Resident Fellow at the Center for Southeast Asian Studies, National Chengchi University. His research includes macroeconomics and international trade policy under imperfect competition, public sector economics, and the design and analysis of public and social policies. ORCID: https://orcid.org/0000-0003-3966-4843 Jinkook LEE is a Research Professor of Economics and the Director of the Program on Global Ageing, Health, and Policy, at the Center for Economic & Social Research, University of Southern California. She is an Associate Editor of The Journal of the Economics of Ageing. Ronald LEE is an economic demographer who taught in the departments of Demography and Economics at Berkeley in 1979–2014 and now does research on macroeconomic consequences of population ageing. He holds honorary doctorates from Lund University in Sweden and the University of Montreal in Canada. He is an Associate Editor of The Journal of the Economics of Ageing. Xiaoyan LEI is Professor of Economics in the National School of Development at Peking University and the Director of PKU Center for Healthy Ageing and Development Studies in xxx

Contributors

China. She is Research Fellow of IZA and an Associate Editor of The Journal of the Economics of Ageing. Elizabeth LEMMON is a Research Fellow in Health Economics at the University of Edinburgh, Scotland. ORCID: https://orcid.org/0000-0002-3564-6106 Maciej LIS is an economist in the Pensions and Ageing Team in the Employment, Labour and Social Affairs Directorate at the Organisation for Economic Co-Operation and Development. He holds a PhD in economics from the Warsaw School of Economics. Anastasia LITINA is an Assistant Professor at the University of Macedonia, Thessaloniki, Greece and a Research Fellow at the University of Luxembourg. Her research interests include the study of long-run determinants of growth, cultural, and historical economics. ORCID: https://orcid.org/0000-0001-9575-5489 Annamaria LUSARDI is University Professor of Economics and Accountancy at the George Washington University School of Business (United States) and directs the Global Financial Literacy Excellence Center. She also chairs Italy’s Financial Education Committee. She holds a PhD in economics from Princeton University. Andrew MASON is Emeritus Professor of Economics, University of Hawaii at Manoa. Mason and Ron Lee co-founded the National Transfer Accounts network. Mason is an Associate Editor of The Journal of the Economics of Ageing. ORCID: https://orcid.org/0000-0003-4578-1800 Margaret MCCONNELL is Associate Professor of Global Health Economics at the Harvard T.H. Chan School of Public Health. Her research combines behavioral economics with field and laboratory experiments to understand and evaluate policies designed to change health behaviors and improve health systems. Olivia S. MITCHELL is Professor of Business Economics/Policy and Insurance/Risk Management at the Wharton School of the University of Pennsylvania, United States, where she directs the Pension Research Council. She studied economics at Harvard University and the University of Wisconsin. ORCID: https://orcid.org/0000-0002-6419-8314 Raymond MONTIZAAN is Associate Professor at the Research Centre for Education and the Labour Market (ROA) of Maastricht University in the Netherlands. He is affiliated with IZA and Netspar. ORCID: https://orcid.org/0000-0002-6758-2522 Peng NIE is an Associate Professor of Economics at Xi’an Jiaotong University, China. He is an IZA Fellow and an Associate Editor of The Journal of the Economics of Ageing. ORCID: https://orcid.org/0000-0002-5322-6324 Philip O’KEEFE is Professor of Practice at University of New South Wales John Piggott Business School and Director of the Ageing Asia Research Hub at the ARC Centre of Excellence in Population Ageing Research (CEPAR). Until 2021, he was an economist at the World Bank, most recently as Practice Manager for Social Protection and Jobs in the East Asia and Pacific region. xxxi

Contributors

John W. R. PHILLIPS serves as Chief of the Population and Social Processes Branch of the National Institute on Ageing Division of Behavioral and Social Research. He is the federal Program Official for the Health and Retirement Study and has managed extramural programs on the economics of ageing and the development of international comparators to the U.S. Health and Retirement Study to support ageing research. John PIGGOTT is Director of the ARC Centre of Excellence in Population and Ageing Research (CEPAR) at the University of New South Wales, where he is Scientia Professor of Economics. He has published widely on retirement and pension issues and in public finance more generally; his research has appeared in leading international economics and actuarial academic journals. He is an Associate Editor of The Journal of the Economics of Ageing. Panu POUTVAARA is Professor of Economics at the Ludwig-Maximilians-Universit¨at M¨unchen and the Director of the ifo Center for International Institutional Comparisons and Migration Research. He is also member of Germany’s Expert Council on Integration and Migration, Managing Editor of the CESifo Economic Studies, a member of editorial boards of the European Journal of Political Economy and of the Leadership Quarterly, and CESifo, IZA, and CReAM fellow. ORCID: https://orcid.org/0000-0003-2070-784X Klaus PRETTNER is a Professor of Economics (especially macroeconomics and digitalization) at the Vienna University of Economics and Business (WU) and project leader at the Wittgenstein Centre for Demography and Global Human Capital. He is an Associate Editor of The Journal of the Economics of Ageing. ORCID: https://orcid.org/0000-0002-8971-8750 Paola PROFETA is Full Professor of Public Economics at Bocconi University and Director of the AXA Research Lab on Gender Equality. She is research affiliate of CESifo, President of the European Public Choice Society, member of the board of management of the International Institute of Public Finance, Associate Editor of European Journal of Political Economy, International Tax and Public Finance, and CESifo Economic Studies. ORCID: https: //orcid.org/0000-0002-1493-8564 Alexia PRSKAWETZ is Professor of Mathematical Economics at the Institute of Statistics and Mathematical Methods in Economics, TU Wien, Austria, Deputy Director at the Vienna Institute of Demography, Austrian Academy of Sciences, Austria, and research associate at the International Institute of Applied Systems Analysis, Laxenburg, Austria. She is an Associate Editor of The Journal of the Economics of Ageing. ORCID: https://orcid.org/0000-0002-2850-6682 Bernardo L. QUEIROZ is an Associate Professor of Demography at Universidade Federal de Minas Gerais (Brazil). He specializes in economic demography and demographic methods. His research centers on how demographic changes relate to the changes in the labor market and small-area mortality estimation with defective data. He is an Associate Editor of The Journal of the Economics of Ageing. ORCID: https://orcid.org/0000-0002-2890-1025 Monika QUEISSER is Senior Counsellor to the Director of Employment, Labour and Social Affairs Directorate and the Head of Social Policy Division at the Organisation for Economic Co-Operation and Development, where she supervises and coordinates the work on social protection, pensions, affordable housing, gender equality, diversity, and family policies. She is an Associate Editor of The Journal of the Economics of Ageing. xxxii

Contributors

Giacomo REBELLATO is a Project Specialist at the Program on Global Ageing, Health and Policy within the Center for Economic and Social Research at the University of Southern California. Anna REUTER is a researcher at the Heidelberg Institute of Global Health at Heidelberg University. She holds an doctorate in development economics from the University of G¨ottingen. ORCID: https://orcid.org/0000-0002-3087-7415 Ole RØGEBERG is a Senior Research Fellow at the Ragnar Frisch Centre for Economic Research working on topics related to intelligence and mental health. Susann ROHWEDDER is Senior Economist and Associate Director of the RAND Center for the Study of Ageing, Santa Monica, USA. Her research centers on the financial security of older households, examining household consumption, saving, and retirement behavior, Social Security, long-term care, dementia risk, and individuals’ expectations. ORCID ID: https:// orchid.org/0000-0002-9482-5286 ´ Miguel SANCHEZ-ROMERO is a Researcher at the Vienna Institute of Demography/Austrian Academy of Sciences and Guest Researcher at the International Institute of Applied Systems Analysis, Economics Frontier program. He is an Associate Editor of The Journal of the Economics of Ageing. ORCID: https://orcid.org/0000-0002-5999-6522 Warren C. SANDERSON is Emeritus Professor of Economics at Stony Brook University, Stony Brook, New York, and Visiting Scholar at the Center for Demographic Research, Baruch College, New York, and at the International Institute for Applied Systems Analysis, Laxenburg, Austria. ORCID: https://orcid.org/0000-0002-2205-948X Sergei SCHERBOV is Principal Research Scholar and Project Leader, Population and Just Societies program at the International Institute for Applied Systems Analysis (IIASA), Laxenburg, Austria. He is also a Director of Demographic Analysis at the Wittgenstein Centre for Demography and Global Human Capital and he is an Affiliated Professor at the College of Population Studies, Chulalongkorn University, Thailand. ORCID: https://orcid.org/ 0000-0002-0881-1073 Andrew J. SCOTT is a Professor of Economics at London Business School and a Research Fellow at the Centre for Economic Policy Research. He is the co-author of “The 100-Year Life” and co-founder of The Longevity Forum. ORCID: https://orcid.org/0000-0002-7785-1263 JP SEVILLA is a Research Associate at the Harvard T.H. Chan School of Public Health and Health Economist at Data for Decisions, LLC. He researches quantifying the full health and social value of vaccines and other health technologies. Bettina SIFLINGER is Assistant Professor at the Department of Econometrics & OR at Tilburg University, the Netherlands. Her primary research area is applied microeconometrics with a focus on health and human capital formation during childhood and youth. Vegard SKIRBEKK is a Norwegian population economist and social scientist specializing in demographic analysis. He is Principal Investigator at the Centre for Fertility and Health, Professor at Columbia University, New York, and Researcher at the Norwegian National Advisory Unit on Ageing and Health. xxxiii

Contributors

James P. SMITH joined Rose Li and Associates in 2020 as a Senior Research Associate (Economist). He served as a key advisor to all Health and Retirement Study sister surveys around the world. He was elected to the National Academy of Medicine in 2011 and earned honorary doctorates from the University of Stirling and Trinity College Dublin. Alfonso SOUSA-POZA is a Professor of Economics and Director at the Institute for Healthcare & Public Management at the University of Hohenheim in Germany. He is a co-founding Editor of The Journal of the Economics of Ageing. ORCID: https://orcid.org/ 0000-0002-6524-9654 Holger STRULIK is a Professor of Economics at the University of Goettingen in Germany where he holds the chair for Macroeconomics and Development Economics. He is an Associate Editor of The Journal of the Economics of Ageing and the Journal of Demographic Economics. ORCID: https://orcid.org/0000-0002-5615-2412 Nikkil SUDHARSANAN is a Rudolf M¨ossbauer Assistant Professor of Behavioral Science for Disease Prevention and Healthcare at the Technical University of Munich’s Institute for Advanced Study. ORCID: https://orcid.org/0000-0003-1710-4634 Uwe SUNDE is a Professor of Economics and Chair of the Seminar for Population Economics at the Ludwig-Maximilians-Universit¨at M¨unchen. He holds a PhD in economics from the University of Bonn and is a Fellow of CESifo, IZA, CEPR, and Research Professor at ifo and DIW. He is a co-founding editor of The Journal of the Economics of Ageing. ORCID: https: //orcid.org/0000-0002-2110-7822 Yuqi TA is a doctoral candidate in the National School of Development at Peking University in China. B. Piedad URDINOLA is an Associate Professor in the School of Economics at Universidad Nacional de Colombia-Bogot´a and Director of the Andean Demographic and Epidemiological Observatory. Urdinola is a leader in research on population issues for Latin America and co-directs the Latin American Human Mortality Database. ORDIC: https://orcid.org/ 0000-0003-0273-8706 Jan C. VAN OURS is Professor of Applied Economics, Erasmus School of Economics, Erasmus University Rotterdam, the Netherlands, Fellow of the Tinbergen Institute, Rotterdam, the Netherlands, and Honorary Professorial Fellow, University of Melbourne, Australia. He is an Associate Editor of The Journal of the Economics of Ageing. ORCID: https: //orcid.org/0000-0002-0144-9956 David R. WEIR is a Research Professor in the Survey Research Center of the Institute for Social Research at the University of Michigan. He has been with the Health and Retirement Study since 1999 and its principal investigator since 2007. He is an Associate Editor of The Journal of the Economics of Ageing. ORCID: https://orcid.org/0000-0002-1661-2402 Matthias WEISS is Professor of Economics and Statistics at Regensburg University of Applied Sciences. He is Research Fellow of Munich Center for the Economics of Ageing at the xxxiv

Contributors

Max Planck Institute for Social Law and Social Policy and of the Research Centre for Education and the Labour Market at the University of Maastricht. ORCID: https://orcid.org/ 0000-0001-7973-4725 Yaohui ZHAO is Professor of Economics at Peking University, China. She received her PhD from the University of Chicago in 1995. She is an Associate Editor of The Journal of the Economics of Ageing. ORCID: https://orcid.org/0000-0002-9252-9715 Leo ZUCKER is a Research Assistant at the Harvard T.H. Chan School of Public Health, where he participates in research on the economics of demographic change and the value of health interventions. ORCID: https://orcid.org/0009-0009-5633-9680

xxxv

1 INTRODUCTION TO THE HANDBOOK David E. Bloom, Alfonso Sousa-Poza, and Uwe Sunde

1.1

Why This Handbook?

Population ageing—driven by declining fertility, increasing longevity, and the progression of large-sized cohorts to older ages—is the 21st century’s dominant global demographic trend and one of its most urgent global challenges. An emerging focus of researchers and policymakers, population ageing will invariably touch every facet of economics, healthcare, politics, and social security. The pervasive nature of this development and its attendant changes demands that it be extensively studied, fully apprehended, and directly reckoned with. Never before in history has the proportion of older people in the global population been as high as it is today: The share of those aged 60+ now stands at approximately 14 percent (1.1 billion people), up from 8 percent (200 million) in 1950 (United Nations, 2022). Population ageing will continue to accelerate, with the share of those aged 60+ expected to reach 22 percent (2.1 billion) by 2050 (United Nations, 2022). “Sixty is the new forty” and similar quips are not merely bons mots designed to make people feel better about their age; they reflect the view that people are living (and expecting to live) healthier, longer lives that do not sacrifice vitality as the years accumulate. Advances in physiology, medicine, pharmacology, nutrition, sleep science, and workplace safety, as well as reductions in tobacco use and other harmful behaviors, have contributed to and combined with the aforementioned demographic shifts to redefine the meaning of old age. This “compression of morbidity” hypothesis, while powerfully argued in the literature (Fries et al., 2011; Cutler et al., 2014; Chernew et al., 2017), is not without critiques, including that of its robustness to alternative definitions of morbidity (Crimmins and Beltr´an-S´anchez, 2010). Additionally, pre-COVID life expectancy trends in the United States call into question the monotonic improvement of longevity (Venkataramani et al., 2021). Thus, global policy responses to ageing populations must necessarily be redefined in this rapidly evolving context. More than 99 percent of human history passed before the total population reached 1 billion in roughly the year 1800; we now expect to add 1 billion older individuals in just the next 30 years (Bloom, 2019; United Nations, 2022). By 2050, older people will outnumber adolescents and young adults and represent more than triple the number of children under five (United Nations, 2022). The cohort aged 85+—the so-called “oldest old,” whose needs and capacities tend to differ significantly from those who are merely old—is growing especially fast and is projected DOI: 10.4324/9781003150398-1

1

David E. Bloom et al.

to surpass half a billion by the end of this century (United Nations, 2022). Adding to the complexity of this shift, low- and middle-income countries (particularly in Asia, Latin America, and Eastern Europe) are now getting old before getting rich. This situation is a dramatic turnaround from recent decades when ageing occurred mostly in high-income countries. Population ageing poses numerous microeconomic concerns that policymakers, business leaders, and family units will be forced to grapple with. Do individuals who live longer also work longer or save more for retirement? Do they invest more in education and training? Or do they rely more on family support and, if so, how does that added economic burden affect families? To what extent do new patterns of labor change familial dynamics, especially in cultures where tradition dictates that elders provide childcare? Conversely, who will provide care for the parents and grandparents living longer than ever before, and how will that affect family budgets? And how do these changes affect individual businesses, as older workers extend their stay in the job market? Older workers possess vast institutional knowledge and experience, but how do their presumed diminished productivity, more frequent illnesses, and reduced ability to adapt to new circumstances offset that? Population ageing likewise introduces several macroeconomic concerns. Are older workers truly less productive than younger workers to the point that an older labor force entails diminished aggregate output and surplus? If the typical retirement age does not change, relatively fewer younger workers will be available to replace what could be a wave of retirees; will there be enough workers to produce all the goods and services a society demands? As the COVID-19 pandemic has demonstrated, societies and individual consumers are ill-prepared for such potentially large disruptions in the supply chain. Will these possible labor shortages bring about rising wages, and would that be a net positive outcome in light of rising prices? How would these changes affect savings, the lifeblood of investment? Older people naturally save less and may, in fact, dis-save; will aggregate savings decrease as populations age and older people spend down their savings? How would that affect asset values? Will ageing-induced changes in savings and interest rates drive capital flows from more to less rapidly ageing countries and lead to a global redistribution of economic power? Still other ramifications of population ageing could be categorized as both microeconomic and macroeconomic concerns. For instance, financial gaps already characterize many social security systems around the world, and as large cohorts move into retirement, filling these gaps will become an urgent priority for governments. Pay-as-you-go systems are structurally dependent on a thriving economy, tenuously relying on contributions from current workers to fund the pensions of older individuals. But fully funded systems are not a panacea, as they are susceptible to mismanagement, unrealized investment returns, and other systemic volatilities. Fully funded systems could also disappoint via the close kinship between ageing and its demographic cousin, population decline: A shrinking labor force would plausibly lower returns on physical capital investments, thus bringing down interest rates and depressing the rate of return of fully funded systems. At a time of rapid demographic change, ensuring that resources are sufficient to support retired workers can be challenging, especially as that cohort balloons. Meanwhile, one area most likely to experience increased financial pressure due to population ageing is the health system. Longer life spans mean the time period during which individuals need quality healthcare will expand. Governments, which pay for or facilitate access to much healthcare in many countries, will certainly be pressed to respond to the healthcare needs of older populations. The fiscal consequences of adequately doing so may be enormous. One mechanism by which fiscal stress is likely to increase is that diseases of old age—including cancer, chronic obstructive respiratory disease, heart disease, diabetes, and Alzheimer’s disease and other dementias—are very expensive to treat, not just medically but also in terms of formal and 2

Introduction to the Handbook

informal care needs. Additionally, the value of lost production due to morbidity and mortality from noncommunicable diseases, coupled with the effect of diverting a portion of savings to cover treatment costs, is equivalent to a roughly 3–10 percent tax on gross domestic product (GDP) based on a macroeconomic model calibrated for selected countries out to 2050 (Bloom, 2019). Likewise, education and training systems must prepare for an influx of older people. Human capital embodied in the population is the single most important economic resource in many countries, and with ageing workforces and improved health at older ages, the possibility of continually upgrading and accumulating human capital gains importance for both individuals and societies. Preparing those systems for looming change presents another urgent challenge for policymakers. While recasting all challenges as opportunities may be reductive, a comprehensive economic analysis of population ageing rejects the notion that “demography is destiny.” The benefits of an ageing populace, including the economic benefits, are often undervalued or overlooked. These benefits include older people’s involvement in numerous nonmarket activities that create value, such as caring for grandchildren, undertaking volunteer work, or performing duties crucial to running a household. These benefits must be accounted for and explicitly sought in policy responses to ageing. Central among these responses are institutional adaptations and reforms related to health and long-term care (LTC), including significantly higher public funding for health than in the past, improved targeting of households with elderly members for health outreach and education, and support for seeking and scaling innovations that promote active ageing. Equally necessary are policy responses targeting increased human capital investment throughout the life cycle; social insurance and pension reform; and modernization of business and human resource practices, including initiatives to combat the negative stigma of ageing. There will be no one-size-fits-all approach: Countries age at different points in time and at different rates, with varying initial levels of wealth and infrastructure and with asymmetrical baseline policies that address differing country-specific concerns. Nonetheless, the challenge of population ageing is indeed an opportunity for all countries to reimagine the role of older people in society and the economy, how to leverage their talents and strengths to individual and societal benefit, and how to fulfill and finance their healthcare needs. Can population ageing go beyond simply adding years to life and also add life to years? The goal of this handbook is to provide readers with a contemporary overview of these crucial topics—namely, the current state of the field of the economics of ageing and its main areas of research. Our aim is to create a valuable and comprehensive reference volume that provides an up-to-date compendium of the most important topics on the economics of ageing. Unlike conventional handbooks, which comprise lengthy reviews of subfields, the contributions in this handbook combine a succinct survey of the state of research in the field with an outline of the research agenda that the authors view as particularly urgent and auspicious. Ours is a “translational” goal, in terms of making the content of this handbook accessible to a broad range of analysts and practitioners and not just to specialists or those who are willing to bear a sizable fixed cost of mastering a narrow line of inquiry. Our readers may include graduate students and researchers in economics departments, schools of education, and schools of public policy and public health. We also expect this book to be valuable to policymakers in the fields of social security, health, and finance, among others, and to journalists and the business community. An endeavor like this handbook is many years in the making, and the COVID-19 pandemic has dominated the last several in virtually every conceivable way. The disease and its ubiquity have imparted immeasurable lessons, some inarguably positive, like the application of mRNA methodology to vaccine delivery and the capacity of global public-private partnerships to accelerate vaccine development at astonishing speed; those advances could truly revolutionize our 3

David E. Bloom et al.

capacity to fight disease. The pandemic has also shown us how much we value older people by highlighting the multifaceted vulnerabilities of the aged, who have suffered severe illness and death at tremendously greater rates than younger cohorts (CDC, 2022; Yanez et al., 2020; Ioannidis et al., 2021). Indeed, that most individual and policy responses to COVID-19 acknowledge this vulnerability is telling. While this handbook is predicated on the idea that the world will continue to see increases in the number and longevity of old people, this idea is itself predicated on the assumption that old people’s health will continue to durably improve. Like so many facets of our lives, the COVID-19 pandemic has complicated that assumption, and so, appropriately, many of our chapters discuss its ramifications. That reality—COVID-19 upending conventional wisdom—is the context in which this handbook was prepared. Contributions to this effort come from some of the world’s foremost experts in economics and demography, with many drawn from the editorial board of The Journal of the Economics of Ageing. This handbook comprises seven parts: health, pensions and social security, income and economic growth, work and employment, data and measurement, ageing and personality, and regional developments. Within each part, chapters further explore these broad topics.

1.2

A Brief Overview 1.2.1

Health

The Health part opens with a chapter entitled “Modeling the Impact of Population Ageing on Future Fiscal Obligations” by Jay Bhattacharya. With the world’s population rapidly ageing, governments across the world will face increasing demands—and related financial obligations— for social security protections in the form of pensions and health insurance. The author employs three approaches to model that looming risk: (1) a macroeconomic approach that uses metrics such as the old-age dependency ratio (OADR) and other key demographic measures to examine the outcome of interest (e.g., healthcare spending); (2) a microsimulation approach that uses longitudinal data, paired with a comprehensive health questionnaire, to construct a model of how the given population’s health changes over time; and (3) a microeconomic approach that uses an overlapping generations framework to incorporate the anticipated demographic changes of the given population into “a neoclassical equilibrium model of consumption, savings, production, technology, and growth.” Each approach necessarily has trade-offs—often centered on balancing the model’s required dataset with how policy is practically applied—and future research avenues should strive to incorporate the supply side of healthcare markets. Three chapters then follow that examine the role of technology and innovation in the economics of health. Technological advances have been the lifeblood of improving the human condition throughout recorded history (the development of writing perhaps being the quintessence of societal advancement). Unlike ecological changes, which typically take place over centuries, technological advances—like vaccines and breakthroughs that stabilized food and fresh water supplies—can evolve within decades, years, or even months to facilitate longer and more productive and enjoyable lives. “Medical Innovations and Ageing: A Health Economics Perspective” by Volker Grossmann finds that innovative health treatments, which are driven by demand in advanced countries, are generally cost-effective while still raising health expenditures. The market power of pharmaceutical companies, which plays a ubiquitous role in healthcare financing, could be addressed through policy change, as their profits far exceed their research and development (R&D) costs—even when considering the failures that never come to market. However, the standard health deficit model suggests that healthcare rationing could have negative ramifications on health and longevity by disincentivizing R&D. Finally, any technological 4

Introduction to the Handbook

progress must be implemented thoughtfully, as innovations that increase longevity could also raise inequality: Those with more education, higher income, and better health status may reap greater benefits. “Medical Progress, Ageing, and Sustainability of Healthcare Finance” by Michael Kuhn agrees that further examination of how medical progress exacerbates health inequities, as measured by differences in life expectancy, is warranted. While ageing and medical progress both drive health expenditure, the total impact of ageing on expenditure depends on whether the additional years are spent in good health. If a medical innovation leads to additional healthy years and the innovation substitutes for, rather than complements, standard healthcare usage, then the innovation can “decouple the ageing process from increases in healthcare spending.” Regardless, increases in life expectancy generally lead to additional treatment—and additional costs—and further R&D into medical innovation, which are magnified by the market size effects of a large and growing elderly population. Some novel developments, like medical applications of automation and artificial intelligence, may actually reduce costs despite increasing lifetime years of medical care consumption. The findings here determine that “the simultaneous growth of longevity and healthcare spending can be viewed as a welfare improvement,” notwithstanding some inefficiencies in standalone aspects of healthcare provision. “Technologies to Mitigate Cognitive Ageing” by Neil Charness examines the great promise technology holds for addressing cognitive decline, one of the most intractable problems of ageing. Cognitive decline begins fairly early in life, as “fluid abilities,” such as those for abstract problem solving, generally begin to slip when individuals are in their 20s. “Crystallized abilities”—knowledge acquired culturally—typically begin to decline in people’s 50s or 60s. Reduced memory and thought-processing speeds are other examples of common cognitive declines. Driver-assistance technology, mobile technology that aids memory, and video technology that enables social connectivity are examples of developments that may enhance independence in older age, though hurdles invariably exist in adapting new technology. The next chapter looks at population ageing and health through the spectrum of gender differences. “Gender, Ageing, and Health” by Xiaoyan Lei and Yuqi Ta finds that, across all countries, females have longer life expectancy but poorer functional health. In less-developed nations, women generally score worse on depression, cognition, and self-assessed health. The picture is considerably murkier in developed nations, where the scale and direction of gender health gaps in old age fluctuate depending on country, time, and age group of interest. This speaks to the complexity of pinpointing the source(s) of gender inequities, which epidemiological, biological, social, contextual, and methodological arguments explain inadequately. While studies have consistently indicated that socioeconomic status is a significant covariate, what is most needed is an interdisciplinary approach, including public-private partnerships, to determine the heretofore elusive element of causality. This approach can and should be bolstered by the inclusion of more women and elderly in research studies. The following chapter, by JP Sevilla, is entitled “Economics of Disease Prevention in the Elderly.” While disease prevention intrinsically promotes well-being, its instrumental benefits offer tremendous value: Enabling work, both paid and unpaid; encouraging full participation in leisure; reducing the economic burden of healthcare for individuals and the health system; and decreasing the strain on households to provide healthcare are but a handful of such benefits. To take advantage of this value, governments must assume an active role, as older individuals are unlikely to select socially optimal levels of disease prevention investment on their own. Universal healthcare, “sin” taxes, vaccination programs, and other preventive clinical services and interventions offer high returns on investment and are associated with reduced morbidity and increased functioning. Such efforts could delay retirement (in full or in part) and reduce the formal 5

David E. Bloom et al.

and informal costs of elderly healthcare expenditure—tangible benefits that standard economic analyses often omit, leading to underinvestment in elderly disease prevention initiatives. “The Economics of Long-Term Care” by David N. F. Bell and Elizabeth Lemmon surveys the most-cited articles from economics journals and determines that the topic lacks focus in the literature, so much so that they find it “difficult to develop an overview of the main issues and to explore gaps.” Unpacking causal relationships in the economics of LTC was particularly challenging, which makes it particularly difficult to formulate policy recommendations. Data gaps provide a hurdle for researchers, as LTC data published by international organizations are thematically limited and based on incomplete data provided by member countries. The data inadequacy was particularly resonant during the COVID-19 pandemic, as countries struggled to grasp the magnitude of the threat inside LTC centers. That said, LTC insurance demand, LTC financing, unpaid care, child–parent interactions, the market for LTC workers, and projections of future spending were the most frequently cited topics that emerged from Bell and Lemmon’s literature survey. They conclude that the field would benefit from its topics coalescing into a new standalone journal (or a satellite offshoot to an existing periodical). While the Organisation for Economic Co-operation and Development (OECD) covers important facets of LTC in its “Ageing and Long-term Care” hub, no single publication is the go-to source for the economics of LTC, which currently appears in various medical and health economics journals, a scattershot approach that may hinder more in-depth LTC research. “In Good and Bad Times—Associations between Spousal Health and Assortative Matching on Early-Life Factors in Europe,” authored by Iris Kesternich, Bettina Siflinger, and James P. Smith, analyzes the economics of ageing through the lens of assortative matching, that is, the selection of a life partner based on pre-marriage characteristics, such as health, socioeconomic status, and education. The authors use the Survey of Health and Retirement in Europe to see how those spousal associations manifested in later-life spousal health. Choosing two measures of physical health (major and minor conditions) and one measure of mental health (depression) as outcomes of interest, they employ the focal age of 70 to determine that assortative matching on health and socioeconomic status in early life may lead to positive associations in spousal health later in life. They also find that spousal health behaviors may have spillover effects on their partner’s health. Strong regional differences in European spousal associations were revealed, which they conclude were explained in part by the assortative matching on early-life factors, while speculating that culture, regional institutions, and systemic inequalities also drive the variations. The final chapter in this part is “Mental Health and Illness in Ageing” by Sherry Glied, Carolyn D. Gorman, and Richard Frank. As mental illness often originates early in individuals’ lives, their wage-earning profile, attachment to the labor market, and social connectedness may all suffer. Such outcomes are serious obstacles to healthy ageing. Additionally, increasing life expectancy equals a longer window in which more people risk developing Alzheimer’s disease and related dementias. If an individual suffers those outcomes as they age, they are less likely to have savings, employee benefits, or a network of friends and family that could provide necessary monetary and emotional support in the face of hardships like homelessness and poverty. Social welfare systems that provide such individuals with financial support and stable housing could mitigate their struggles, yet this is a cohort that governments often leave behind.

1.2.2

Pensions and Social Security

The Pensions and Social Security part opens with “Social Security Reforms in Heterogeneous Ageing Populations” by Miguel S´anchez-Romero and Alexia Prskawetz. Focusing on public pensions, the authors examine the role of differential mortality on the internal rate of return 6

Introduction to the Handbook

of the pension system and determine the reforms necessary to make the internal rate of return between different socioeconomic groups equitable. They demonstrate that, in effect, the pension system as currently constituted “redistributes income from short-lived, poor individuals to long-lived, richer individuals” and offer the parametric components of the pension system that can be adjusted. As any policy changes would cause a market reaction from individuals, the authors propose their suggested reforms should be studied in behavioral models “to account for general equilibrium effects, the transition costs of such reforms, and the possibility of using a multi-pillar approach.” Extending this line of inquiry could incorporate other social security systems—such as disability pensions; widowhood pensions; and informal safety nets in the form of children, spouses, and families—and other systemic reforms aimed at ensuring the viability of these schemes. The following entry, “Economic Preparation for Retirement” by Michael D. Hurd and Susann Rohwedder, describes the income replacement rate approach to retirement planning and juxtaposes it with a life-cycle model optimization approach and a consumption-based approach. While the income replacement rate is popular with financial advisors, the latter two approaches address the structural inadequacy of the former—namely, that the income replacement rate is ill-suited to serve as a retirement planning guide for new retirees. If a household has dual earners who retire, or partially retire, at different times, quantifying their income becomes very complicated, thus defeating the simplicity that forms the main draw of the income replacement rate approach. The consumption-based approach, which benefits from the addition of some life-cycle model attributes like survival risk and typical spending trends for people in their demographic group, is nonetheless simple and adaptable and reflects the expected retirement advantage of couples over individuals. This method straightforwardly assesses whether “the individual’s economic resources can support the current level of consumption with some adjustments for eventual widowing, for observed spending reductions at advanced old age, and for personalized life expectancy,” which are key considerations for most people as they approach retirement. “Pension Policy in Emerging Asian Economies with Population Ageing: What Do We Know, Where Should We Go?” by George Kudrna, Philip O’Keefe, and John Piggott focuses on East and Southeast Asian economies, where approximately 30 percent of the world’s population resides. While this region is subject to some of the same demographic shifts occurring worldwide—increasing life expectancy and decreasing fertility rates amid underdeveloped social safety nets—the authors characterize the issue of ageing as urgent due to additional area-specific concerns. These worrisome trends include high rates of informal employment and co-residency, family structures straining under the weight of major rural-urban migration, and the speed of the demographic transition. Of particular focus are informal workers who, while living longer, may not amass the resources needed to fully finance their lives in extended old age. The authors believe that enhanced welfare spending can substantially affect this cohort’s quality of life at low cost relative to GDP. The last chapter in this part, “Trends in Pension Reforms in OECD Countries” by Herv´e Boulhol, Maciej Lis, and Monika Queisser, traces salient developments over the preceding decades and describes how the legislative choices around financing pensions interact with the demographic challenge of increased longevity. Many proposed remedies, such as raising the retirement age, are politically unpopular, though improved communication of the reasons for reform combined with enhanced savings incentives could help to address some aspects of the problem. Yet the issue remains extremely complicated, as inequalities of life expectancy, differential retirement ages, gender discrepancies, occupational risks, and coverage gaps for the self-employed must be factored into any policy response. Often framed as a debate between 7

David E. Bloom et al.

pay-as-you-go and funded pensions, the authors note that combining both pension systems can “help raise the risk-adjusted returns of contributions to the retirement system, at both the individual and aggregate level,” as Canada, Finland, and Sweden have done. Acknowledging that the potential consequences of a complete shift away from pay-as-you-go pensions are unclear, the authors suggest that the economic efficiency of these schemes requires further study before population ageing’s full impact on pension reform design can be assessed.

1.2.3

Income and Economic Growth

The Income and Economic Growth part opens with Ronald Lee’s “Economic Growth, Intergenerational Transfers, and Population Ageing.” As countries age, labor force growth slows, which raises capital intensity per worker and boosts productivity, thereby increasing wages and lowering interest rates. At the same time, citizens are also splitting an already established—and likely shrinking—pie of old-age benefits; that is, fewer people are paying into the system while more people are taking benefits out. The end result is that inputs must increase (workers pay more taxes) or outputs must decrease (retirees receive less). Likewise, while lower fertility rates allow time potentially spent parenting to be time spent working, producing at home, or participating in leisure, a cohort of older workers may lead to less technological advancement and, thus, slower productivity growth. “Consequently,” the author writes, “population ageing has positive and negative effects on economic well-being at the individual level and alters market and nonmarket distributions of income.” Future avenues for research include assessing how ageing affects technological progress and, now that population ageing is happening around the globe, how it influences international markets. “Consumption, Saving, and Wealth Accumulation at Old Age: Comparing Evidence from Developed and Developing Countries,” authored by Marco Angrisani, Jinkook Lee, and Giacomo Rebellato, contrasts trends in these three dimensions across national income categories. Older people in developed countries tend to retain their wealth unless they experience health shocks, which contradicts the accepted wisdom of the standard life-cycle model. Per the authors, the elderly’s reluctance to dis-save has three key drivers: their desire to pass an inheritance to their children, uncertainty about their own life expectancy (and thus about their remaining years of consumption to be financed), and the specter of sudden medical expenses. Comparatively fewer studies examine these issues in developing countries, though rapid population ageing and the availability of richer data may change that situation. The speed and scale of demographic change in developing nations seem to be shifting the focus from the household to the individual, which may spawn new patterns in savings and wealth accumulation among their ageing populaces. Ana L. Abeliansky and Klaus Prettner’s “Automation and Ageing” is the next entry in this part. While worldwide population ageing has engendered concerns over labor shortages, burgeoning automation has sparked fears of “technological unemployment.” Automation— characterized here as industrial robots, three-dimensional printers, and algorithms based on machine learning—may offset the labor shortages wrought by widespread population ageing, and population ageing itself is likely a powerful driver of automation investment and adoption. While such projections always carry an element of uncertainty, the number of jobs that automated technologies could replace is substantial. Automation may also shape the health needs of the future elderly via multiple avenues: Workers fearing their replacement by robots face a persistent mental health threat, yet many of these workers would enjoy better physical health if robots made their physically demanding manual jobs a thing of the past. This field is obviously ripe for future and ongoing study, particularly on how automation affects older adults’ healthcare 8

Introduction to the Handbook

(outcomes, delivery, and costs), their labor force participation, and the migration or reshoring of jobs. “Working Life—Labor Supply, Ageing, and Longevity” by Andrew J. Scott describes how the increased proportion of older people in the populace and increased longevity mean more older people in the workforce. Accounting for older workers is complicated by the diversity of their working arrangements (full-time vs. part-time vs. flexible) and the consequent variance in their work intensity. In the past, the available pension scheme and an individual’s health largely dictated retirement; with social security changes and the improved health of modern aged cohorts, “retirement becomes more of a transition than an event.” This raises the question of whether older workers should be folded into standard economic analyses, or whether a separate strain of research that delves into the unique properties of older workers should be developed. The author concludes that a deeper comprehension of ageing workers, one that separates their specific labor supply and demand issues, is crucial to understanding this uniquely heterogeneous group. Our final chapter in this part is Andries de Grip and Raymond Montizaan’s “Education and Ageing: Human Capital Investments and Ageing.” As various demographic changes that have added an influx of older workers to the labor market intersect with the rapid pace of technological innovation, older workers must continually update and upgrade their skills to remain employable. While human capital investments in older workers have traditionally seen less return than investments at younger ages for obvious reasons, the aforementioned changes have nevertheless increased demand for these investments. As older workers may resist such training amid struggles with retirement-age changes, pension reform, and new technology itself, the most prudent path to understanding how to raise human capital investment in this cohort is to increase their participation in random control trials. Studying their skill training preferences, evaluating what style of training most connects with this cohort (on-the-job? classroom setting? informal instruction? a hybrid of these modes?), and developing new data sources are the best ways to predict which policy interventions can realize the greatest return on investment. The direction of future research must also go beyond understanding the preferences of the cohort to understand where innovations are headed so new skills that are complementary to emerging technologies are developed while other skills are rendered obsolete.

1.2.4

Work and Employment

The Work and Employment part begins with “The Employment of Older Workers” by Hippolyte d’Albis. Not only has the workforce of most wealthy countries grown older in recent decades, but those older workers are also healthier, better trained, and increasingly female. Analyses show that individuals spend more of their lives in good health, which equates to an opportunity to work additional years. Complementing that trend, countries where retirement commences at a later age have better-trained older workers. And increased female employment rates are a boon to older workers’ employment rates; for example, work-life balance policies that induce mothers to rejoin the workforce in the years following childbirth “minimize the effects of career interruptions and make women’s employment likelier after age 55.” While fortified health insurance and unemployment insurance systems have combined with retirement and pension reform to achieve these gains, further modifications are needed to ensure their viability. The success of delaying retirement age, for example, relies on the ability of older people to find jobs. The author also explores the idea that fears about older workers being less productive and unable to adapt to technological change are rooted in ageism—that is, the stereotyping of individuals based on their age and the use of these stereotypes to justify prejudice, discrimination, and adverse treatment. 9

David E. Bloom et al.

The next entry is entitled “Retirement and Health” by Jan C. van Ours. While retiring can have sizable effects on physical and mental health, the direction of the association is far from clear. Depending upon various factors—whether retirement was mandatory or voluntary, the extent of retirement benefits in terms of pension and health insurance, the personal situation of the retiree, and the demands of the particular job—retirement can have positive or negative effects on health. These variables and their outcomes make studying the issue difficult and creating effective policy extremely complex. Yet a few certainties emerge, such as the fact that mandatory retirement amounts to codified age discrimination. The other major conclusion the author arrives at is that flexibility in retirement timing “may be the only policy measure that is unambiguously beneficial for the health of all workers.” The next chapter is “The Relevance of Cognition in the Context of Population Ageing” by Bernt Bratsberg, Ole Røgeberg, and Vegard Skirbekk. While it is generally accepted that cognitive decline accompanies ageing, the reality is more complex. “Fluid intelligence,” such as abstract reasoning, does indeed generally peak in early adulthood. “Crystallized intelligence,” such as learned skills and knowledge, however, may continue to rise for far longer. Of course, cognition varies markedly among people, and those differences persist over time. Crucially, one’s cognitive abilities at a young age likely affect educational attainment and human capital investment, which translates to the age-related outcomes of earnings and occupational status, among others. That relationship between cognition and such outcomes is likely explained in part by feedback effects wherein those who display strong cognitive abilities are afforded capital investments that may further strengthen said abilities. The authors use Norwegian registry data to look at cognitive scores of males at age 18 and outcomes up to 60+ years later and find that population ageing amplifies cognition-related differences in income, partnering, and health. The following chapter, “Productivity in an Ageing World” by Axel B¨orsch-Supan and Matthias Weiss, challenges the conventional wisdom of its title subject. Measuring the productivity of older workers is a complicated prospect, as so many of the relevant covariates—work environment, type of job, and the specific characteristics of individuals—are dependent. Assuming a natural, across-the-board decline due to age is not accurate, as experience, people skills, and enriching social networks can persist far into old age. The age at which one’s collection of advantageous work attributes peaks is extremely variable, and microeconomic analyses like job performance reviews tend to be biased. One line of inquiry that looks at the accomplishments of high achievers like athletes and artists usually finds that productivity peaks when people are in their 30s, but that analysis is less relevant to the everyday workflow of typical employees. For instance, studies of factory workers generally find an initial learning effect that boosts productivity, which then remains constant until retirement. Macroeconomic studies could add the missing element of improved work environment to productivity studies, but they too often become mired in reverse-causality issues. That said, macroeconomic studies typically suggest that population ageing, via the channels of innovation and entrepreneurship, negatively affects productivity. Microeconomic analyses demonstrate that an age-related productivity decline generally does not occur in standard jobs and point the way for future microeconomic studies to investigate the effects of innovation and entrepreneurship on labor force productivity. The final chapter in this part, Paola Profeta’s “Population Ageing and Gender Gaps: Labor Market, Family Relationships, and Public Policy,” covers the complex intersection of ageing, gender, and labor. Gender issues are strongly tied to population ageing, as (to present just one dimension) research reveals that fertility rates are higher in countries with higher female workforce participation, signifying that gender equality in the labor market could offset some degree of population ageing. Gender gaps and ageing are also intertwined outside the labor market, as older women who extend their stay in the workforce necessarily forgo their roles as care 10

Introduction to the Handbook

providers to grandchildren (and their improved health means they need less care from their own children). The COVID-19 pandemic has further roiled shifts on the labor and domestic fronts, contributing to gender equality stagnation in the labor market, declining birthrates, and increased stress in familial dynamics. Additional demographic shifts such as migration have multifaceted effects, not least because migration is often heavily female. Much more study is necessary to produce effective policy responses, as most studies do not present the causal impact of population ageing and gender gaps on the studied outcomes.

1.2.5

Data and Measurement

The Data and Measurement part opens, appropriately, with “Measuring Ageing” by Holger Strulik, which introduces two measures of ageing: the force of mortality (the increasing probability of dying) and the frailty index (worsening health, to put it simply). These measures grow exponentially with chronological age, as differentiated from physiological age; in fact, the process of ageing “can be understood as the deterioration of a complex and highly redundant system experiencing stochastic damage at the subcellular level.” The health deficit model introduced by the frailty index incorporates the self-productive nature of health deficits, namely that having health deficits leads to more health deficits. The self-productivity concept postulates that ageing is an “intrinsic, cumulative, progressive, and deleterious loss of function that eventually culminates in death.” This characterization directly contradicts the foundation of the health capital model, which assumes that a person in better health has greater health capital and is predicted to lose more of that capital—that is, experience a decline in health—in the next observed period than someone in poorer health/possessive of less health capital. Employing the frailty index, however, allows the testing of life-cycle models by directly measuring health deficits. “The Health and Retirement Study,” authored by John W. R. Phillips and David R. Weir, traces the history of the groundbreaking survey that has spawned a family of sister studies that, in many ways, forms the backbone of contemporary research on the economics of ageing. Founded at the University of Michigan by the National Institute on Ageing and the Social Security Administration in 1992, the initial survey conducted 12,652 interviews. The study produced a nationally representative sample of adults aged 51–61 (and their partners regardless of age) and, crucially, linked the survey with administrative data from Social Security and Medicare programs. Aggregate data were de-identified and made publicly available, while researchers could apply to use additional administrative data. Adding blood collection (first dried blood spots and then whole blood), saliva collection for DNA analysis, and important measures of dementia via the Ageing, Demographics, and Memory Study add-on to later iterations, the Health and Retirement Study has fulfilled its original mission of providing multidisciplinary insight into the economic, sociological, psychological, epidemiological, demographic, and biomedical aspects of ageing. Approximately 3,800 papers have been published using Health and Retirement Study data to date. National Transfer Accounts (NTA), an index designed to capture how demographic changes drive macroeconomic effects, are the focus of the next chapter, “National Transfer Accounts and the Economics of Ageing” by Andrew Mason. NTA aim to describe how gender and generational equity, the public financing of health and pensions, and all individuals (regardless of age) interact with the economy. NTA estimates provide data for essential economic models, like overlapping generation models and life-cycle models, to explore quantitative analyses of the economics of population ageing. NTA also can be employed to analyze the demographic dividend (the concept that an economy’s growth can be promoted by age structure shifts that 11

David E. Bloom et al.

elevate the working-age population share), how funding social security systems affects consumption, and how population ageing increases the demand for wealth to fund said systems. The author uses contemporary estimates from the United States to compare results with nations across the globe that use NTA, finding that consumption age profiles and generational transfers are highly variable. Constructing NTA can be a painstaking process but doing so provides a wealth of information on the effects of population ageing on macroeconomic performance across countries. “Ageing and Dependency,” by Warren C. Sanderson and Sergei Scherbov, suggests a replacement for the OADR as a measure of population ageing. The OADR compares people in a country aged 65 and older relative to those aged 20–64. While this measure can be easily and consistently produced for most countries over a long period of time (and straightforwardly projected outward), its origins as a measure of burden betray its flaws: The OADR, in absolute terms, defines old age as 65+ while simultaneously characterizing everyone aged 65 and older everywhere as dependents. Such an unqualified characterization ignores the fact that 65-yearolds in one country can differ significantly from 65-year-olds in another and discounts salient aspects of ageing like health status and life expectancy; the OADR does not actually incorporate measures of dependency like financial flows or activities of daily living. The authors reject the notion inherent in the OADR that dependency is tied to chronological age and suggest the adoption of a replacement measure based on the prospective old-age threshold (POAT). The POAT uses indicators like activities of daily living, instrumental activities of daily living, and the Global Activity Limitations Indicator, along with self-reported health and age-specific mortality rates, to define old age as the point when remaining life expectancy drops to 15 years. This measure thereby recasts the idea of ageing as years left to live instead of years already lived and defines the threshold of old age based on individually varying demographic characteristics in lieu of a fixed chronological number. The authors suggest as a corollary measure to the POAT the proportion elderly—the proportion of people older than the POAT—which would replace the OADR. This part of the handbook closes with “Patterns of Time Use among Older People” by Maddalena Ferranna, JP Sevilla, Leo Zucker, and David E. Bloom. The authors examine the time allocation of the aged, exploring how much of a person’s life is devoted to work (paid and unpaid), leisure, and personal care (attending to biological needs like eating, sleeping, and medical care). This chapter also considers time use activities comparatively, looking at how older people’s allocation differs vis-`a-vis that of younger people and by country, gender, health status, marital status, and education. Much of the data are drawn from the Multinational Time Use Study (participant countries analyzed include Austria, Canada, France, Hungary, Italy, Netherlands, Republic of Korea, Spain, the United Kingdom, and the United States) and time diary data from China and India. Informing this area of research is Becker’s “A Theory of the Allocation of Time” (1965), which holds that the optimal time allocation between work and nonwork activities depends on several variables—household preferences, activity cost, available time, and income—that help to explain labor choices and retirement decisions, as well as on how public policy interacts with those determinations. Unsurprisingly, time devoted to paid work declines precipitously as individuals move toward old age (though that allocation remains economically significant until one’s 70s). This analysis also found that while people in China and India spend more time on paid work than their Multinational Time Use Study counterparts (including among older adults, where the difference is almost two hours per day), they also spend more time on personal care, which may indicate a deficit in personal care technology. Future paths of research include exploring the relationship between time allocation and well-being and documenting evidence of active ageing to combat the stigma of the elderly as unproductive burdens. 12

Introduction to the Handbook

1.2.6

Ageing and Personality

The Ageing and Personality part opens with “Ageing and Economic Preferences” by Thomas Dohmen, David Huffman, and Uwe Sunde. This chapter reviews the literature on economic preferences—especially risk and patience—and ageing. Economic risk taking and patience change over the course of people’s lives, which may signal that anticipation and experience of biological change (like cognitive decline) and evolving economic circumstances related to physical changes (like life expectancy) factor into people’s decisions about these behaviors. Of course, real-life behavior patterns may also reflect naivet´e concerning one’s economic circumstances or health decline. Nevertheless, recent studies find that a pattern linking age and risk taking holds throughout the world, though a person’s position in the wealth distribution likely also affects the magnitude of that relationship. In addition, an individual’s cognitive status affects the association of ageing and risk. Declining cognition cannot be assigned causality, however, in part because the surveys from which these data derive may not capture the intricacy of behavioral reporting, older and younger respondents may have different risks in mind when answering survey questions, and the survey instruments are less sensitive than alternatives like lottery choice measures. A person’s familial situation and health considerations like morbidities surely factor into economic risk-taking decisions, too. Regardless of the multifarious determinants of ageing and risk aversion, economic ramifications will escalate as population ageing accelerates: Changes in saving patterns and investment choices may affect international capital flows, while increasing life expectancy may amplify these channels and intensify their outcomes. The following chapter, “Financial Literacy and Financial Behavior at Older Ages” by Olivia S. Mitchell and Annamaria Lusardi, finds that financial literacy can be protective against the fiscal missteps that currently plague the older generation. Older people have more debt than ever before and are often financially unprepared for retirement or unexpected emergency expenses. Reviewing the literature on financial behavior at older ages, the authors suggest that teaching adolescents—that is, future older people—about financial planning is an imperative, especially as the trend of growing student loan debt quickly saddles them with a deficit that compromises their ability to invest or save. Education is the key to arming the older cohort, too; their financial education can be supplemented while still in the workplace—an approach that aligns with the interest of employers. Cognitive decline often accompanies ageing, but here again financial education could be the solution: Research demonstrates that financial literacy has stronger positive effects on decision-making among those with lower cognitive function. Additionally, financial literacy could impart the importance of designating someone to help older people make financial decisions before cognitive decline sets in, which could be especially appropriate as financial mistakes (including falling for scams targeting the elderly) made late in life may be harder to recover from. As in so many facets of life, the COVID-19 pandemic has had deleterious effects on the elderly, from lingering health issues and job loss to evictions and food insecurity. In parallel with financial literacy, financial regulations are important safeguards for the economic status of older people, with governments, financial advisors, investment firms, bank personnel, and medical professionals all having a role to play. “Age and the Value of Life,” by Maddalena Ferranna, James K. Hammitt, and Matthew D. Adler, reviews the literature on age and measures of well-being. The authors examine this field through the prism of two approaches to assessing value: benefit-cost analysis (BCA), which employs the value of a statistical life metric, and the social welfare function (SWF). SWF operates via a utilitarian function, which relies on remaining life expectancy and its expected quality, or a prioritarian function, which assigns greater weight to policies that benefit younger people (with the rationale that younger people tend to be on the lower end of society financially 13

David E. Bloom et al.

and have also not yet reached the “fair innings” threshold). The authors argue that the SWF approach offers advantages over BCA in that the former incorporates the notion that individual health measures may have different values to different people while avoiding what they deem the “ethically objectionable result that benefits accruing to the well-off are more valuable than similar benefits accruing to the less well-off ” of BCA formulas. Furthermore, the self-evaluation of what matters in life can—and very likely does—change with age, and SWF incorporates these shifting preferences. The bulk of the literature in this area focuses on mortality risk reductions, which the authors argue overlooks the significance of reducing nonfatal risks and should be integrated into future SWF frameworks. Life satisfaction and ageing are the focus of the next chapter, “Happiness and Ageing in the United States” by David G. Blanchflower and Carol Graham. Reviewing the literature on ageing and measures of happiness, the authors reveal a consistent U-shaped pattern of happiness across countries, as insomnia, depression, and deaths of despair reliably peak in middle age. The authors analyzed multiple data sources for the United States and discovered a strong relationship between well-being and ageing, with marked differences when sorted by marital status. The differences between the wed and the unwed include larger gaps and steeper drops in happiness among the latter when compared with peers in other countries—likely a reflection of the socioeconomic differences between the two groups—though U.S. uniqueness in its high marriage and divorce rates and lower mean age at first marriage are worth noting. Analysis of the U.S. Health and Retirement Survey shows that satisfied people have a greater life expectancy, though life satisfaction begins to decline when individuals are in their 70s. Happiness scores have fallen significantly since the COVID-19 pandemic, a descent akin to the drop typically seen in midlife. Lamenting that some recent reports dismiss the association between happiness and ageing, the authors conclude, “Given that the midlife dip is associated with behaviors that result in premature mortality and/or compromised health and quality of life, we believe its causes and costs merit further scientific inquiry that in turn could yield insights into alleviating it, rather than the continuation of a debate over whether it exists.” “Ageing and Foreign Policy Preferences” by Mark L. Haas looks at how governments’ international preferences are likely to respond to population ageing. As the trend of population ageing affects virtually every country in the world, its consequences for policymaking are inescapable, and this well beyond the demographically sensitive domains of health, labor, and social security. One subtle yet crucial dimension of the interaction between ageing and politics is how foreign policies are impacted. Pointing to survey data that show older citizens are more opposed to the use of military force and that declining fertility rates lead to a heightened sensitivity to the casualties of war, the author concludes that population ageing will result in more peaceful nations. The other foreign policy outcome likely to emerge from population ageing is increased isolationism, as the economic challenges of population ageing may compromise nations’ abilities to fulfill international commitments and wage costly wars. Older individuals, growing as a proportion of the populace of a given country, will have the political capital to exert their preferences for more peaceful and inward-looking agendas. The challenge of noncommunicable diseases (NCDs) outside the world’s high-income countries is the subject of our next chapter, “Behavioral Science and Noncommunicable Diseases in Low- and Middle-Income Countries” by Nikkil Sudharsanan, Michael R. Eber, and Margaret McConnell. Rather than focusing on particular NCDs, the authors explore the commonalities of disease management and care that are pertinent to many of the conditions, highlighting how NCDs—which are powerfully driven by age—differ from infectious diseases and the behavioral challenges therein. Those challenges confront both healthcare providers, who must shift their models of care from a curative model addressing symptoms to a proactive prevention model, and 14

Introduction to the Handbook

patients, who must take a similar long-range view of their health and make important lifestyle changes rather than abandoning treatment upon remission of their most acute symptoms. Yet in low- and middle-income countries, the research being carried out and the policies being enacted remain overwhelmingly concentrated on infectious diseases, whose treatment relies on patient and provider behaviors that may hinder the treatment of NCDs. The plight of patients who suffer from NCDs in low- and middle-income countries is unlikely to ease until fresh behavioral perspectives are introduced to treatment models. The concluding chapter in this part, “The Implications of Population Ageing for Immigrantand Gender-Related Attitudes” by Andreas Irmen and Anastasia Litina, examines how population ageing interacts with opinions about women and immigrants in the workforce. Using the results of the World Values Survey, this chapter observes that the economic pressure from population ageing—the proportional loss of working-age people and the demand for wealth to finance the social pensions and healthcare needs of the elderly—will result in a search for additional workers. Immigrants and women are obvious and available options to heed that call, with the hope that acceptance of their necessary role in the labor market will lead to greater social acceptance and equality. An OECD report from Japan projects that if the significant increase in the female employment rate seen over the last decade were to continue (leading to parity with that of Japanese men), real output per capita would see an almost 7 percent boost by 2050 (Jones and Seitani, 2019). While incorporating more immigrants in the workforce is seen as a straightforward influx of added workers, the traditional role of women as caregivers and homemakers complicates appraisal of their economic impact. However, as the Japanese example shows, some of the female employment rate increase was due to formalization of what had been informal work, as women entered the formal labor market and other women were officially hired to perform their previous child- (and elder-) care duties. Thus, some women entering the labor market create an additive effect on employment rates, production, and presumably GDP. This chapter proposes that the labor market additions of immigrants and women will be internalized, with population ageing serving as “a driver of social change toward a more inclusive society.”

1.2.7

Regional Developments

The final part of the handbook, Regional Developments, opens with “Global Ageing and Health” by Anna Reuter, Till B¨arnighausen, and Stefan Kohler. This chapter looks at how rapid population ageing interacts with other demographic trends and movements, such as the proliferation of certain diseases (including pandemics), the desire for universal health coverage, climate change, population migration, the evolving family structure, and widespread digitalization. The authors employ data from the World Population Prospects and the Global Burden of Disease Study to examine ageing (both in the population and in the individual) and its effects on health. While health data on the elderly remain sparse and often rely on extrapolation and modeling, they find tremendous global gains in life expectancy (rising, for example, in Africa from 36.5 years in 1950 to 63.2 years in 2019), largely due to declines in child mortality, though life expectancy has also increased among the elderly. Healthy life expectancy has also climbed, resulting in a gain of 6.8 healthy years since 1990. Increasing Internet access may bolster those gains further, as digital technologies can keep older people connected to their families and loved ones for emotional and psychological support; to their doctors via telemedicine for physical and mental health support; and to the marketplace for banking, shopping, and working. Expanding digital access is just one part of the essential health promotion that healthcare systems must undertake to respond to the looming challenge of ageing populations. 15

David E. Bloom et al.

“Social Protection and Population Ageing: A Comparative Analysis of India and Indonesia,” by Mukul G. Asher and Chang Yee Kwan, looks at how Indonesia’s and India’s moderately rapid ageing intensifies the urgency of the social security reforms these countries have undertaken (exacerbated by the COVID-19 pandemic, which hit both countries particularly hard). The large heterogeneous populations of India and Indonesia, which together comprise more than one-fifth of the world’s population, require a unique assemblage of integrated social protection schemes rarely seen in the literature; for example, the provision of clean cooking fuel in India has played a critical role in improving health—an initiative not conventionally seen in formal healthcare plans. Indonesia passed a wide-ranging national social security system law in 2004, though implementation took a decade, while India began its overhaul in 2014. Both nations share the priorities of sustaining economic growth in the face of population ageing while modernizing and upgrading the services of their government organizations, their internal coordination, and their data collection capacities. Both are also attempting to navigate these issues while moving toward universal health coverage. More specifically, both aim to enhance their actuarial estimates, account for their fiscal contingent liabilities, increase the citizenry’s financial literacy, improve risk management analysis, encourage the adoption of targeted technologies, and enhance their messaging to realize the essential political economy necessary for public approval of social security reform. The countries have attacked these similar problems in slightly different ways, as Indonesia emphasized healthcare initially, while India took a more varied approach that includes tax benefits and liability assumption to finance pensions. The next entry, “Ageing in China” by Peng Nie and Yaohui Zhao, looks at the largest ageing population in the world. China also has one of the world’s fastest ageing populations, multiplying the degree of difficulty in responding to the multifaceted challenges of population ageing. In fact, the pace of ageing in China suggests that it has gotten old before it has gotten rich, resulting in the financial pressures of its current situation. With a decline in contributors and an increase in pensioners, China must address its fiscal problems, which are inextricably tied to its health problems. While the nation could institute a policy of postponing retirement age— particularly among urban residents, where that milestone is age 60, which essentially squanders the potential of a large and growing cohort to continue working—that change would necessitate significant health improvements. Many older Chinese experience physical and cognitive decline, exacerbated by poor health choices. While population ageing disrupts family-centered caregiving, the relaxation of China’s one-child policy has clearly not ameliorated the deficit in the working-age population. While China has a built-in remedy for its worker shortage— extending the retirement age—it is not well positioned to address its health problems, which, again, are fundamentally inseparable from the former because workers must be healthy enough to work. The authors suggest that China should further relax its family planning policies while introducing measures that encourage childbearing, such as affordable childcare and maternal insurance; reform the pension system to include a mixture of familial, community, and government support; establish a long-term care system that is not based on the family; and roll out interventions that promote healthy ageing and health literacy, which would not only enhance the health of all its people, but would perhaps allow it to extend the retirement age beyond 60. “Ageing in Latin America,” by Bernardo L. Queiroz and B. Piedad Urdinola, is the next chapter. While Latin American nations exhibit differences in their pension and healthcare systems, not to mention where they stand in the process of reforming those systems, the countries of the region largely share important traits that population ageing magnifies, such as socioeconomic inequality and labor market pressure to fund social security systems. While recent studies show that health among older people in Latin America is improving, gains in life expectancy are not translating to additional time in the labor market. So, while Latin Americans living 16

Introduction to the Handbook

longer, healthier lives is good news, some of those resulting extra years must be spent working lest the effects of more people drawing benefits from a diminished tax base will impact the already strained pension systems of the region. Significant uncertainty remains over the pace of ageing, as some studies claim nations are effectively “older” than expected, while others say the demographic changes are slow enough to afford Latin America sufficient time to make necessary policy adjustments. The authors argue that adopting a cognitive capacity measure would provide a truer measure of health and vitality in old age than the OADR and thus could clarify the picture. Indeed, such contradictions point to the need for more detailed and nuanced data on ageing, particularly separated life tables and time use surveys that can isolate the heterogeneity of a cohort that displays significant differences in the ageing process by chronological age, gender, region, and income. Latin American countries need to invest resources in studying retirement trends, savings, time use, public transfers, social security incentives, working conditions for older people, and the gender and socioeconomic gaps that proliferate in those areas. For one thing, the data clarify that increased life expectancy is also increasing inequality. Some evidence indicates that Latin America is squandering its demographic dividend and that the introduction of policies that encourage women to enter the labor market—such as increasing the supply and subsidization of childcare services and formalizing informal work—could halt or reverse this trend. The final chapter in the handbook is Panu Poutvaara’s “Population Ageing and Migration.” Immigration predominantly reflects economic conditions, which themselves reflect demographic trends. That is, the economic pressures of population ageing can produce increased migration patterns, the economic effects of which depend on whether investments in education can compensate for the loss of skilled workers and whether immigrants are welcomed into the workforces of their new home countries. Over the last century, North America and Australasia saw the highest immigration flows, Latin America saw high emigration, and Europe experienced a reversal from emigration to immigration. Africa presents a fascinating case, as its predicted population boom is expected to include more than 1 billion people under age 20 by 2050; while to date Africa’s migration flow has lagged behind its population share, “even modest increases in emigration from Africa would generate major increases in immigration pressure in the rest of the world, mostly in Europe,” the author writes. That potential development could catastrophically intersect with climate change (which disproportionately disrupts agricultural bases, which proliferate in Africa) to produce “climate refugees.” Various steps could mitigate these pressures: Investments in education could alter migration trajectories via multiple channels, while technological advancements, particularly around communication, could ease migration in several dimensions. The governments of ageing countries could also pass regulations to facilitate hiring of immigrants to provide care—a practice often relegated to the “shadow economy”—which would benefit the long-term healthcare sector and the economy at large. Instead of pursuing initiatives to limit immigration, governments could integrate the skilled labor of migrants into their labor markets and encourage education-finance partnerships, a mutually beneficial scenario wherein destination countries invest in education in the countries from which they derive the new source of labor.

1.3

Acknowledgments

The editors of this volume are extremely grateful to its contributors, whose diligence and care are reflected in the chapters here. These authors represent some of the finest minds in the fields of economics, demography, healthcare, and population ageing, and their expertise is on display in each entry. We are delighted to present their work. 17

David E. Bloom et al.

As editors, we followed a rigorous process of external review, with two external reviews sought, and in almost all cases delivered, for each chapter. This process, while time-consuming, led to substantially improved chapters as many authors explicitly acknowledge. We wish to thank our distinguished panel of reviewers: Dilba Alp, Elif Aydinlandi, Nick Barr, David N. F. Bell, Fiona Carmichael, Giuglielmo Cavalli, Courtney Coile, Haodi Cui, Nicolas Dauer, Karen Eggleston, Robin Fanger, Ross Gordon, Martin Haas, Yiming Hu, Mark Jit, Martin Kahanec, Arie Kapteyn, Markus Knell, Rainer Kotschy, Vadim Kufenko, Michael Kuhn, Nicole Maestas, Ajay Mahal, Andreas Mair, Anastasia Matveeva, Fabrizio Mazzonna, Salih Nas, Ole Frithjof Norheim, Man Ou, Sabrina Pfeffer, John Piggott, Klaus Prettner, Ziming Qin, Lucia Reisch, Andrew J. Scott, Holger Strulik, Madeleine Trattler, Oytun T¨ukenmez, Zhunan Wang, and Leo Zucker. At our publishing company Routledge, we would like to thank Christiana Mandizha and Natalie Tomlinson. We would like to thank Jeffrey R. Adams, Alicia Brandt, Tora Estep, and Simran Makkar for outstanding editorial and writing assistance. We also acknowledge financial support for this handbook provided by an award from the Carnegie Corporation of New York and a grant from the National Institute on Ageing of the National Institutes of Health (P30AG024409).

References BECKER, G. (1965): “A theory of the allocation of time.” Quarterly Journal of Economics 75(299): 493–517. BLOOM, D. E. (2019): “Sailing into unchartered demographic waters.” VoxEU, 14 October 2019. Available at https://voxeu.org/article/sailing-uncharted-demographic-waters. CDC (CENTERS FOR DISEASE CONTROL AND PREVENTION). (2022): “Provisional death counts for coronavirus disease 2019 (COVID-19).” Available at https://www.cdc.gov/nchs/nvss/vsrr/covid weekly/index.htm#SexAndAge [accessed on March 25, 2022]. CHERNEW, M., CUTLER, D. M., GHOSH, K., LANDRUM, M. B., AND SKINNER, J. (2017): “Understanding the improvement in disability-free life expectancy in the US elderly population.” In: Wise, D. A. (ed.), Insights in the Economics of Ageing. Chicago: University of Chicago Press, Chapter 5, 161–204. ´ -SANCHEZ ´ CRIMMINS, E. M., AND BELTR AN , H. (2010): “Mortality and morbidity trends: Is there a compression of morbidity?” The Journals of Gerontology, Series B 66(1): 75–86. CUTLER, D. M., GHOSH, K., AND LANDRUM, M. B. (2014): “Evidence for significant compression of morbidity in the elderly US population.” In: Wise, D. A. (ed.), Discoveries in the Economics of Ageing. Chicago: University of Chicago Press, Chapter 1, 21–80. FRIES, J. F., BRUCE, B., AND CHAKRAVARTY, E. (2011): “Compression of morbidity 1980–2011: A focused review of paradigms and processes.” Journal of Ageing Research 2011: Article 261702. IOANNIDIS, J. P. A., AXFORS, C., AND CONTOUPOULOS-IOANNIDIS, D. G. (2021): “Second versus first wave of COVID-19 deaths: Shifts in age distribution and in nursing home fatalities.” Environmental Research 195 (April 2021). JONES, R. S., AND SEITANI, H. (2019): “Labour market reform in Japan to cope with a shrinking and ageing population.” Organisation for Economic Co-operation and Development Economics Department Working Papers no. 1568. UNITED NATIONS. (2022): World Population Prospects 2022, Online Edition, Rev. 1., United Nations, Department of Economic and Social Affairs, Population Division, New York. Available at https:// population.un.org/wpp/. VENKATARAMANI, A. S., O’ROURKE, B., AND TSAI, A. C. (2021): “Declining life expectancy in the United States: The need for social policy as health policy.” The Journal of the American Medical Association 325(7): 621–622. YANEZ, D. N., WEISS, N. S., ROMAND, J.-A., AND TREGGIARI, M. M. (2020): “COVID-19 mortality risk for older men and women.” BMC Public Health 20: Article 1742.

18

PART I

Health

2 MODELING THE IMPACT OF POPULATION AGEING ON FUTURE FISCAL OBLIGATIONS Jay Bhattacharya

Abstract Decades of increasing longevity and declining fertility worldwide have resulted in the ageing of populations. With fewer young workers per older retired individual and a rapidly rising number of long-lived elderly persons, governments will likely face sharply increasing pension and healthcare spending obligations over the coming years. Forecasting the extent and timing of these obligations is essential so that policy choices regarding these issues can proceed on an informed basis. In this chapter, I survey three different approaches that modelers have adopted for this task. An accounting/macroeconomic approach decomposes the outcome of interest—health or pension spending, for instance—into economically meaningful components. The decomposition typically features objects like the old-age dependency ratio, labor force participation rates, and other important economic and demographic variables readily available from the macroeconomic literature. The second, a microsimulation approach, constructs a detailed model of the evolution of the population’s health over time. The approach’s data needs are substantial, requiring high-quality longitudinal data with a detailed health questionnaire to populate its parameters. The third, an overlapping generations/microeconomic approach, maps anticipated demographic changes in the population onto a neoclassical equilibrium model of consumption, savings, production, technology, and growth. I describe each approach, discuss strengths and weaknesses, discuss substantive results about the effects of population ageing on future fiscal prospects from papers that feature each approach, and conclude with suggestions for future work motivated by gaps (better modeling of healthcare suppliers and a focus on the effects of the COVID-19 pandemic) in this literature.

2.1

Introduction

In 1950, a child born in the United States could expect to live 68 years, which was incredibly long by historical standards and much higher than the life expectancy of a child born at the beginning of the century. Life expectancy has continued to climb throughout the 20th century and into the early years of the 21st century. By 2009, life expectancy at birth in the United States had risen to 79 years, a stunning 11-year increase in six decades. DOI: 10.4324/9781003150398-3

21

Jay Bhattacharya

The story is similar throughout the developed world. In the United Kingdom, life expectancy at birth rose from 69 years to 80 years over the same period. Life expectancy leaped from 59 to 83 years in Japan. Similar rapid growth has occurred in most developing countries, although life expectancies still lag behind those in developed countries. Over this same period, fertility rates have been falling worldwide (Lesthaeghe, 2014). This worldwide decline in fertility, combined with the worldwide increase in life expectancy, is upsetting the familiar demographic of more children than adults and more adults than elderly that has existed unchanged for millennia. In 1950’s Europe, the modal age groups for the population were less than 16 years old. By 2010, the bulk of the European population was between 20 and 50 years old. In 2050, the most common age groups are forecasted to be 60 years and older (Roser, 2019). These massive changes will irrevocably disrupt the financing of public health and pension systems. Because an understanding of the fiscal impacts of ageing is so vital to the finances of nations, the research effort to model the effects has been robust. This chapter provides an overview of three popular approaches to this task: (1) a macroeconomic/accounting approach that decomposes outcomes of interest into accounting identities for each element with good forecasts available; (2) a health microsimulation approach that pays attention to a detailed model of transitions in population health status; and (3) a microeconomic/overlapping generations modeling approach that relies on a neoclassical model of consumers, firms (including general technological change), and the government sector. The following three sections describe each approach in turn, including an assessment of their strengths and weaknesses. My approach is to provide a detailed description of a particular paper that exemplifies the approach. My choice of papers is idiosyncratic. The selection is not the result of exhaustive cataloging of publications or a meta-analysis of this literature, which is beyond my scope. I chose exemplar papers that provide a clear exposition of the approach in sufficient detail so that the strengths and weaknesses of each approach—rather than of the papers in particular—can be addressed fruitfully. My goal is to introduce readers to these different approaches and my subjective analysis of their strengths and weaknesses, as well as to some of the substantive conclusions about the fiscal consequences of ageing that papers in this literature reach. I have limited this survey to standard methods that many analysts use to forecast the effect of demographic change on health, economic, and fiscal outcomes. As such, I do not discuss the empirical literature that isolates the causal effect of demographic change on economic outcomes (e.g., Acemoglu and Johnson, 2007; Bloom et al., 2009), though this literature is important. The final section provides some thoughts on the gaps in the literature and some suggestions for future modeling efforts.

2.2

Accounting Identity/Macroeconomic Approach

One common approach to forecasting the effects of population ageing involves decomposing the outcome of interest—health or pension spending, for instance—into economically meaningful components. The decomposition typically features objects like the old-age dependency ratio, labor force participation rates, and other important economic and demographic variables readily available from the macroeconomic literature. The decomposition may be complex or straightforward, but the key feature is that the components must yield the outcome exactly, based on an accounting identity. Despite this simplicity, analysts have used the approach to produce surprisingly powerful forecasts of how ageing may influence future government financing of healthcare and pension obligations. 22

Modeling the Impact of Population Ageing on Future Fiscal Obligations

2.2.1

Description

The starting point of the macroeconomic or accounting approach is a population age forecast generated by demographers, which modelers take as given. Let pop(age, t) represent the number of people of each age in year t predicted by this population forecast. This forecast has embedded assumptions about future fertility and future age-specific mortality, which this accounting approach takes as given. As such, authors who adopt this approach typically do not explicitly model the processes that might alter fertility or mortality rates. Clements et al. (2018) provide an excellent example of the techniques used in this approach.1 The authors project the consequences of population ageing on pension and healthcare spending between 2015 and 2100 for 38 developed and 65 less developed countries. Their method relies on demographic projections of the old-age dependency ratio (OADR). Other model inputs include forecasts of labor force participation, gross domestic product (GDP), and other economic variables. Finally, they rely on cross-sectional data on how health spending varies with age in each country. The authors take United Nations forecasts of each of these as given, and they translate these projections into forecasted levels of pension benefits and health spending as a fraction of GDP. The four key building blocks of their accounting approach to forecasting future pension spending as a fraction of GDP are •

The old-age dependency ratio, OADR(t), which is the number of people above the retirement age to the number of adults below it. P∞ age=65 pop(age, t) (1) OADR(t) = P64 age=15 pop(age, t)

•

The labor force participation rate, LFP(t). Let worker(t) be the number of people in the labor force forecasted for year t. Many papers assume this quantity to be a constant fraction drawn from the historical literature of the number of adults between age 15 and 64 who work. worker(t) LFP(t) = P64 (2) age=15 pop(age, t)

•

The pension participation rate, PPP(t). This quantity represents the fraction of the elderly population eligible for government pensions, typically very high in most developed countries. Let pens elig(age, t) be the number of people forecasted to be pension eligible at each age in year t. P∞ age=65 pens elig(age, t) PPP(t) = P∞ (3) age=65 pop(age, t)

•

Pension generosity. Let pens gen(age, t) be total spending on pensions per eligible older adult at each age in year t. P∞ age=65 pens gen(age, t) ∗ pens elig(age, t) pens gen(t) = (4) GDP(t)/worker(t) The product of these four elements produces a forecast of pension spending as a fraction of GDP in future years. Notice that this equation is an exact accounting equality. pens gpd(t) = pens gen(t)PPP(t)LFP(t)OADR(t) 23

(5)

Jay Bhattacharya

The forecasting of government health expenditures as a fraction of GDP takes a similar approach. The idea is to use a model of fixed age-specific health spending and forecast changes in health spending due to demographic changes that alter the age distribution in the population. A single parameter, HEPC 0(t), modulates the level of health spending per capita, equivalent to health spending per capita at age 0. Health spending per capita for people of other ages is expressed as a multiple of the expenditure for people of that age, α(age, t). So average health spending for a 65-year-old, for instance, is α(65, t) HEPC 0(t), with α(0, t) = 1. So the forecasting model for health spending as a fraction of GDP, HE GDP(t), is given by HE GDP(t) =

∞ HEPC 0(t) X α(age, t)pop(age, t) GDP(t) age=0

(6)

To model changes in health technology, Clements et al. (2018) build in an “excess cost 0(t) growth” factor such that HEPC GDP(t) increases over time at a fixed rate. The idea is that medical technology makes health spending more expensive as a fraction of GDP. The authors use this simple framework to forecast the effects of many possible changes in the population, including changes in fertility and longevity, migration policies, labor force participation rates by women and the elderly, changes in the retirement age, and changes in the excess cost growth rate of medical care. Their primary conclusions are as follows: “Nevertheless, shrinking populations pose a grave fiscal threat. Absent further reforms, agerelated spending is expected to rise because of higher longevity and lower fertility in both more and less developed countries. Without the implementation of further reforms of public pension and healthcare systems, age-related outlays are expected to increase over 2015–2100 from 16.5 percent to 25 percent of GDP in the more developed economies, and from 5.6 percent to 16–17.5 percent of GDP in the less developed economies. The fiscal consequences of this outcome are dire: spending increases of such a magnitude could lead to unsustainable increases in public debt, sharp declines in other spending, and large increases in tax rates that could stymie economic growth” (Clements et al., 2018).

2.2.1.1

Evaluation

While I have labeled this an “accounting” approach, in fact, each piece of this accounting identity requires its own forecasting infrastructure, including demographic and economic forecasts. The great strength of this approach is to allow an exploration of the implications of these forecasts on the outcomes of interest (e.g., future pension obligations) with almost no additional assumptions. Because the forecast relies on a necessarily true accounting identity, the primary drivers of the overall forecast are the forecasts of each individual part. The primary weakness of this approach is using it to address outcomes where a simple decomposition is unavailable is difficult. The Clements et al. (2018) paper simplifies many details about pension systems, retirement, health, and healthcare systems to produce an identity where data are available for each element from many countries. For instance, though the paper sets the retirement age at 65, in many countries, eligibility for public retirement benefits can start either before or after age 65, at the worker’s option. In the United States, workers can decide to start retirement benefits anywhere between 62 and 70. It would be difficult, using this approach, to model this complexity while retaining the simplicity of the accounting framework, which is the great strength of this approach. Finally, using this approach to forecast the productivity of health investments on outcomes of interest is difficult. In principle, at least, a healthier workforce will be more productive. A 24

Modeling the Impact of Population Ageing on Future Fiscal Obligations

healthier young workforce might then produce more tax revenue and also be healthier in old age. However, using this approach to model these facts is complex. In Clements et al. (2018), investments in health are a pure burden on government spending, with no offsetting benefit. Given the limitations of the accounting framework and the necessity of using readily available macroeconomic data, relaxing this assumption would be challenging.

2.3

Microsimulation/Markov Transition Approach

The microsimulation/Markov approach is, in a sense, an accounting identity approach that permits much more detailed modeling of health and health spending, retirement transitions, medical technology, and many other outcomes. The consequence of allowing such detail is that the data needs for the approach are considerably greater than the accounting identity approach. This detail permits a broader range of counterfactual questions to be addressed than is possible with other approaches. In this section, I describe the basic structure of the future elderly model (FEM)—a microsimulation-based approach to forecasting how changes in population age structure and health investments may influence many health, labor, and fiscal outcomes.

2.3.1

Description

The most prominent microsimulation approach to forecasting the fiscal impacts of ageing focuses most of its modeling effort on the evolution over time of the population’s health status. This approach contrasts the other two I discuss, which tend to model health in much less detail. The primary justification for this different modeling focus is that it permits forecasts of the effect of interventions that other approaches cannot readily handle. For instance, analysts have used this approach to measure the impact of new medical technologies on health outcomes and health spending. However, the microsimulation approach is also often used to measure governments’ forecast future spending on both health and pensions. To illustrate the basic modeling strategy, I will discuss the FEM, a microsimulation model initially developed by economist Dana Goldman and his group at RAND to examine health and healthcare spending among the U.S. elderly Medicare population. The FEM was first developed to support policy decisions related to Medicare and Medicaid, the public health insurance programs for the elderly, disabled, and poor in the United States. The FEM framework has been extended to address similar policy questions worldwide. Goldman et al. (2004) describe the original FEM. The FEM’s core mechanic is the evolution of a simulated population over time, whose health may change each period and who may live or die, producing other health and economic outcomes that flow from health status. The approach adopts a dynamic Markov microsimulation model to predict future health status and medical spending. The key data input for nearly every version of the FEM is a high-quality longitudinal survey of the older population, such as the Health and Retirement Study (HRS) in the United States. The FEM simulation starts at time T0 with an initial population, typically aged 50 or older, though some versions of the model start at younger ages (Goldman et al., 2016). Individuals in the model are characterized by a vector of demographic and health characteristics, healthi (t), where i indexes individuals drawn with replacement from the longitudinal dataset from the start year of the simulation. The vector of health characteristics includes important chronic disease indicators such as diabetes, heart disease, and chronic pulmonary disease. There is also a vector, Xi (t), of demographic variables like age and sex and other risk factors that predict health status (such as obesity and smoking history). 25

Jay Bhattacharya

The model inputs include four moving parts: •

Health transition functions. These specify the probability of transitioning into a given health state at time t + 1, given that an individual starts in another health state at time t. Analysts estimate these transition probability functions with an econometric model that accommodates the fact that the health indicators are binary variables (like a system of logistic regressions). The data needs are significant, for example, a high-quality, longitudinal dataset with a detailed health status and chronic disease module like the HRS. Chronic diseases are treated as absorbing states, so each transition model is estimated on a population of people who do not have the chronic illness of interest in year t. P(healthi (t + 1)|healthi (t), Xi (t)) = f (healthi (t), Xi (t))

•

Appropriate deterministic transition functions for the demographic variables are also specified (e.g., agei (t + 1) = agei (t) + 1). Mortality risk estimated based on health and demographic characteristics. This function is also estimated using an HRS-like longitudinal dataset. Many FEMs permit an additional parameter, β, to be calibrated such that the mortality function predictions match census mortality predictions. Values of β ̸= 1 mean that the HRS-like dataset is not population representative. P(i dies between t and t + 1|healthi (t), Xi (t)) = β ∗ g(healthi (t), Xi (t))

•

(8)

Health expenditure functions and other outcomes. Typically, FEMs include functions that translate the health and demographic variables of interest, such as per capita health spending, pension obligations, labor force participation, and other outcomes of interest. These models are estimated using an HRS-like longitudinal dataset, other datasets, or literature sources. Two-part models are often employed for outcomes with a skewed distribution and an appreciable number of zero outcomes like healthcare spending. E[HEi (t)|healthi (t), Xi (t)] = h(healthi (t), Xi (t))

•

(7)

(9)

Replenishment cohort. Each simulated individual is aged based on the probabilities determined by the health transition and mortality modules to project outcomes for future years. This necessarily means that if, for instance, the youngest simulated individuals in the model in year t are 50 years old, there will be no 50-year-olds in the model in t + 1. The FEM solves this problem by introducing new 50-year-olds into the model at each simulation, drawn again from the longitudinal database. The simplest version of this module assumes that 50-year-olds in year t will have the same distribution of the health outcomes as 50-year-olds in year t+1 for all t. However, some versions of the FEM relax this assumption to permit health trends in the incoming replenishment cohort.

Figure 2.1 provides a schematic overview of the model. The FEM simulation starts at t = T0 with an initial population obtained by drawing a population with replacement from the HRS-like longitudinal dataset. Each simulated individual i in the population is endowed with a health vector healthi (t) and a demographic vector Xi (t). The health and spending outcomes in T0 are calculated using the health expenditure/other outcome functions. After that, the FEM moves to the following time period T1 in the simulation cycle, with each individual alive at T0 transitioning according to a random draw from the health transition function to a new health state healthi (t + 1) and demographic state, Xi (t + 1). Each simulated individual also faces a 26

Modeling the Impact of Population Ageing on Future Fiscal Obligations

Figure 2.1 Schematic overview of FEM simulation.

random probability of dying, according to the mortality model. The current cohort and the replenishing cohort form the new population of T1 . This process of replenishing new cohorts and transiting across various health states is repeated until the final year of the simulation. This microsimulation framework has been put to many forecasting purposes. The first paper using the FEM framework asked how the introduction of anticipated new technological advances in medicine, such as more effective heart disease and cancer treatments, would alter future trends in health, disability, and healthcare spending on the elderly (Goldman et al., 2005). One key conclusion is “society faces its greatest spending risk not from demographic and health trends, but rather from medical technologies.” The model has been used to forecast the effect of many other medical, social, and demographic policies on outcomes. One paper using the framework found that because of the competing risks problem—the fact that everyone will eventually die of something—“reducing chronic illness in future elderly cohorts will have only modest effects on Medicare’s financial stability” (Joyce et al., 2005). For the same reason, another paper found that even dramatic improvements in the treatment and prevention of cancer would have only modest effects on Medicare spending (Bhattacharya et al., 2005). The National Academy of Sciences, Engineering, and Medicine used the framework in a report to measure how inequality in gains in life expectancy affects the distribution of benefits provided by programs like Medicare and Social Security in the United States (NASEM, 2015).

2.3.2

Evaluation

The primary strength of the microsimulation approach is the wide variety of outcomes and breadth of policies that it can be used to evaluate. As long as estimating an econometric relation between an outcome variable and the health state and demographic variables is possible, this framework can be used to forecast the outcome. This model does not necessarily need to be estimated using the same longitudinal dataset used to estimate the health transition or mortality functions. A second strength of this framework is that it addresses the problem of competing risks naturally within the structure of the model. This strength is essential when evaluating the effect of medical technological breakthroughs or population health improvements, both of which 27

Jay Bhattacharya

affect fiscal outcomes directly and indirectly by increasing longevity. The drawback of singledisease models is that they ignore the problem of competing risks. No matter how effective a new medical technology is, we will all die sometime from one condition or another. Suppose, for instance, that a magical inexpensive new medical treatment is developed that could cure all cancer with no side effects. The first and most obvious consequence of this treatment is that cancer mortality would drop to zero. Less obvious, however, is that the rate of mortality from all other causes would eventually increase. The cancer remedy cures cancer patients, but by doing so, keeps them alive long enough to die of a heart attack instead. In a certain sense, cancer and other diseases compete with each other to be the cause of mortality. As long as death is inevitable, reducing cancer mortality necessarily increases the summed mortality risk of all the other non-cancer causes. Healthcare expenditures might even rise overall because of the new cancer treatment, even if that treatment itself is costless. The FEM microsimulation will address this effect, while other frameworks ignore it and overestimate the impact of the technological breakthrough on health spending. The microsimulation approach has some important weaknesses as well. First, the data requirements for implementing the approach are significant. Without a high-quality, nationally representative longitudinal study that includes survey questions on chronic health, generating the health transition probabilities at the heart of the model is impossible. Even with such data available, important questions about the transition between health states in years t and t + 1 and the link between health states and the forecasted outcomes are hard to address without careful econometric work. For instance, in some datasets, a negative correlation between smoking and heart disease may be observed due to the higher selective mortality of smokers rather than the (nonexistent) protective effect of smoking. Sorting through these issues takes a great deal of care and medical expertise. In these studies, how study authors implement these considerations or what impact alternate specifications would have on results is not always clear. Second, the microsimulation modeling frameworks implemented to date make only partial equilibrium forecasts of items like future government spending on healthcare and pensions. The models are not set up to enforce budget constraints, so unbounded forecasts of government spending are permitted within the modeling infrastructure, with no mechanism to account for countervailing forces that would make this impossible in general equilibrium. This is a problem the accounting approach also shares. How important this criticism is in actual practice is difficult to assess without further research.

2.4

Overlapping Generations/Microeconomic Approach

The overlapping generations approach has a long history in economics, going back to at least Samuelson (1958), and applying it to the problem of population ageing is natural. The overarching idea is to map anticipated demographic changes in the population onto the model through population growth and ageing parameters. The model permits analysts to forecast the effects of demographic change on many key economic outcome variables, including GDP growth, labor force participation, interest rates, savings rates, and government obligations for pensions and healthcare spending.

2.4.1

Description

The starting point of the overlapping generations approach is the specification of a representative family’s utility function, which includes the preferences of both older and younger people. The approach adopts the standard neoclassical assumption that individuals care about consumption 28

Modeling the Impact of Population Ageing on Future Fiscal Obligations

goods and leisure. People derive utility from the well-being of other people: the young care about the old, and the old care about the young, at least in some versions of this approach. Younger people work, while older family members retire. The household intertemporal budget constraint balances return on investment in capital and income from wages versus spending on consumption goods and savings. The modeling infrastructure typically includes a representative firm that employs capital, labor, and technology to produce consumption goods that enter each person’s utility function. A government sector taxes work and returns on capital and uses the money to finance health spending for the elderly. All this setup borrows heavily from the literature on overlapping generations models. What distinguishes this approach is that assumptions about the longevity of individuals are calibrated to real-world demographic data, including data on fertility and longevity. The classic Bommier and Lee (2003) paper develops a continuous-time overlapping generations model of an economy that can be calibrated to realistic demographic data on population mortality and growth. However, much of the empirical literature has adopted a discrete-time approach with a simpler demographic structure than the flexible Bommier and Lee framework. Otsu and Shibayama (2016) build an overlapping generations model with two age bands—young and old—to measure the effect of population ageing on future pension and healthcare spending and economic growth in East Asia. I use their model to illustrate the application of overlapping generations models to this problem. There are five key modules and an equilibrium concept: •

Household utility functions. Otsu and Shibayama adopt a model in which households consist of two types of individuals—young and old—with fractions that vary over time based on changing demography. Utility functions are a weighted sum of the utility of the old and young. Individual utility functions include consumption goods and leisure as inputs. And families maximize a sum of discounted period utility functions, with a discount rate of β ∈ [0, 1]. U=

T X

β t [η(t)Uy (cyt , ht ) + (1 − η(t))Uo (cot )]

(10)

t=1

•

Here, η(t) is the fraction of youth in the population; Uy (cyt , hyt ) is utility experienced in youth, which is a function of consumption by the young in year t, cyt , and hours of work, ht ; and Uo (cot ), is utility experienced in old age, which is a function only of consumption in old age, cot because the old do not work. Budget constraints. The resources available to the family each period include wages, wt , earnings from capital holdings, kt , at the interest rate, rt , and government transfers, gt . They spend it on consumption goods and the single investment good, it . The government taxes labor at rate τt , but it does not tax capital. η(t)cyt + (1 − η(t))cot + it = (1 − τt )wt ht η(t) + rt kt + gt

•

(11)

Investment/capital market. While the budget constraint must be satisfied each year, household capital stock links the budget constraints over time. The modeling of the capital market is motivated by the literature on exogenous economic growth. In that literature, population growth mechanically dilutes the amount of capital held by each household, which in turn lowers the growth rate of physical capital per capita. Otsu and Shibayama incorporate this idea into their model by slowing the rate of per capita capital accumulation 29

Jay Bhattacharya

by a factor of (1 + nt ), where nt is the population growth rate. Capital stock depreciates at a fixed rate, δ, each period. So per capita capital grows according to (1 + nt )kt+1 = (1 − δ)kt + it

•

In the model, households choose consumption and investment levels by maximizing lifetime utility subject to the constraints. Particular simple, functional forms are assumed about utility to make the model tractable, but such simplifying assumptions could, in principle, be relaxed. Demographic and economic parameters, such as population growth, age structure, and interest rates, are drawn from standard sources. To complete the model requires assumptions about firms and the government sector. Firms. The model assumes a representative profit-maximizing firm that transforms labor and capital inputs into output, Yt —the model’s representation of GDP. Solving the profit maximization problem yields labor and capital demand functions that relate the young’s demographic inputs (labor supply), capital investments, and production technology to wages and interest rates. Under a set of simplifying assumptions about the firm’s production technology (e.g., Cobb-Douglass production functions), one can write these relationships as analytic functions where wages and the interest rate are functions of output, labor inputs, capital inputs, and a factor that captures exogenous improvements in the technology of production over time, At . wt = w(Yt , η(t), ht , kt ; At ) (13) rt = r(Yt , η(t), ht , kt ; At )

•

•

(12)

(14)

Government. A government sector taxes the labor of the young and transfers the proceeds to the household. Otsu and Shibayama choose a simple model of government spending as a fixed fraction of GDP. They assume that the government rebates all excess revenue to the household through a lump-sum transfer. Other authors, such as Sudo and Takizuka (2018), adopt a different assumption. They assume that the size of the government sector in each period grows with the proportion of elderly in the population as a simple way to model the higher healthcare demand of the elderly. In any case, this modeling approach requires a theory of how governments respond to population ageing. It also requires that the government finances its programs by taxes. Equilibrium. The equilibrium in this model depends on the endogenous variables—labor and capital levels and consumption and investment—being set such that simultaneously households maximize utility, firms maximize profits, and none of the budget constraints are violated. The prices in the model—wages and interest levels—move such that these conditions are met. In most empirical implementations of the overlapping generations framework, simplifying assumptions guarantee that a single equilibrium exists and can be calculated analytically. Data from standard sources can provide numerical estimates for most exogenous variables in the model. Answering counterfactual questions, such as what would happen if the population were to age, involves recalculating the equilibrium for different values of the exogenous variables, such as η(t).

Using these moving parts, Otsu and Shibayama (2016) calculate equilibrium levels of per capita GDP under alternative assumptions about population growth and ageing, growth in the size of the government sector, and income taxes. They forecast that population ageing will decrease economic growth through two mechanisms: (1) decreasing the size of the workforce available to produce goods and (2) reducing the marginal return to capital by reducing savings. Population 30

Modeling the Impact of Population Ageing on Future Fiscal Obligations

ageing induces increased health spending, which the government finances. A larger fraction of the population being old thereby increases government expenditures on health and tax rates, which directly reduces growth rates by reducing savings and investment. But the increase in tax rates increases labor force supply through a negative income effect, increasing the number of hours worked by the young. Overall, the authors conclude that “an increase in the share of the population over 64 years of age will significantly lower output growth through decreased labor participation. Population ageing can also reduce economic growth through increased labor income taxes and dampened productivity growth.”

2.4.2

Evaluation

Because this approach sits at the cross-section of several great streams of micro- and macroeconomic theory, economists are automatically attracted to it. One of its great features is that it enables authors to make explicit assumptions about how they anticipate various behaviors will change due to people living longer and healthier and the economic effects of an older population age structure, which will feature fewer workers per older person. These models endogenize these behaviors so that, for instance, people alter their work, fertility, and retirement decisions in coherent ways over their lifetimes to accommodate alterations in interest rates and other economic variables. Another advantage of the approach is that natural welfare implications arise from the model, so it can address questions about optimal policies that have welfare maximation as their goal, rather than just optimization of government finances. Unlike the other two approaches surveyed here, this approach integrates the idea that the outcomes forecasts—health spending, pension spending, labor market outcomes, and GDP growth, for example—are generated by an equilibrium process, with budget constraints that limit possible outcomes. The models also naturally permit the incorporation of assumptions about technological change and economic growth in the tradition of the Solow growth model. And finally, the data needed for calibrating models using this approach can be surprisingly modest, and the output of demographic projections about fertility and longevity readily accommodated. This approach also has weaknesses. One problem is that tractable implementations of the ideas require strong assumptions about utility function parameters, well-functioning capital markets, and production technology. These are the standard problems of nearly any empirical implementation of neoclassical microeconomic models. A more specific complaint is that models like this typically do not include any detailed modeling of health or the productivity of health technology developments that come naturally in the microsimulation framework. This is not a problem in principle because detailed health outcomes could be directly inserted into utility functions and a health capital variable of the sort made popular by Schultz (1962) and Grossman (1972) could be introduced, along with firms focused on health production. However, to my knowledge, this has not been done in the context of a forecasting model aimed at assessing the effects of demographic change.

2.5

Selected Review of Papers and Results

While this chapter focuses on methodological issues related to forecasting the effect of population ageing on future fiscal obligations by governments, readers will undoubtedly be interested in the substantive conclusions from this literature. Keeping with the approach I have adopted throughout the chapter, I provide a review based on some representative articles chosen based on the impression they have made on my thinking rather than a comprehensive meta-analytic review of this literature. Such a meta-analysis would be a real contribution to this voluminous 31

Jay Bhattacharya

literature, but to my knowledge has not been conducted. My goal instead is to provide my informal sense—based on decades of following this literature—of conclusions that are firmer and conclusions that remain contentious. Table 2.1 lists the eight papers that I have reviewed and provides a brief synopsis of the methods and aims of each paper and a quotation from the conclusions that describes the major substantive finding. Two papers (Elmendorf and Sheiner, 2000; National Research Council, 2012) follow some version of the accounting/macroeconomic approach. I also included Lee (2014) as a reference because it provides an excellent short exposition of National Research Council (2012). The primary aim of Elmendorf and Sheiner is to explore whether population ageing warrants an increased saving rate to finance the expected increase in Social Security and Medicare obligations in the coming years. They take standard forecasts of the old-age dependency ratio and labor force growth rates as inputs to generate consumption forecasts for the U.S. economy. Using a Solow growth model, they specify a macroeconomic identity setting consumption per person equal to a function of the old-age dependency ratio times consumption per worker. While their model finds that population ageing will increase future spending by the U.S. government on the elderly, they argue that future cohorts should bear that burden rather than the current generation through decreased consumption and increased saving. Table 2.1 Key contributions Paper/Approach

Brief Summary

Key Finding

Lee and Edwards (2002) FEM

The paper inputs stochastic forecasts of population age structure into a Markov model of U.S. Social Security and Medicare spending.

Elmendorf and Sheiner (2000) Macroeconomic approach

The paper’s forecasts build on an elaboration of a Solow growth model built on top of a macroeconomic identity setting consumption per person equal to a function of the old-age dependency ratio times consumption per worker. It inputs existing forecasts of the dependency ratio and labor force growth (both based on population forecasts) to generate consumption forecasts.

“Under current program structures, population ageing would be virtually certain to increase the costliness of Federal programs as a share of GDP by 35 percent (±2 percent) by the 2030s, and by 60 percent (±15 percent) in the second half of the century. . . We expect that governments will respond to these ageing-induced cost changes by altering program structures, so that these conditional projections will not be realized.” “[P]rojected population ageing in the United States still does not provide a rationale for large increases or decreases in current saving. Our best estimate is that because of the ageing of the population, present consumption should rise slightly, although other plausible specifications imply that consumption should decrease a little. This argument does not deny that ageing imposes an economic burden on the nation and that the projected increase in government spending on Social Security and Medicare is an important part of that burden. It simply argues that the optimal response to the ageing of the US population is to allow future cohorts to bear much or all of that burden.”

32

Modeling the Impact of Population Ageing on Future Fiscal Obligations Paper/Approach

Brief Summary

Key Finding

Goldman et al. (2010) FEM

Application of FEM framework comparing the effect of changes in health trends (e.g., obesity) and introduction of medical technologies on future federal health and Social Security spending.

Lee (2014), National Research Council (2012) Macroeconomic accounting framework

A forecast of the effect of population ageing in the United States on Social Security, Medicare, and long-term care obligations using an accounting framework.

B¨orsch-Supan (1995) Neoclassical/general equilibrium approach

Calibration of a Ramsey-Solow optimal growth model with demographic projections for Organisation for Economic Co-operation and Development countries. The setting is a general equilibrium analysis (including international trade flows) of the effect of population ageing on savings and investment.

“There is a popular perception that government programs benefit from worsened health. This perception is typically formed from the case of smoking, in which longevity reductions do in fact benefit a number of government programs. However, it does not hold universally. For example, lowering obesity would save Medicare and Medicaid more than it would cost in terms of public annuity obligations. Public health interventions in these areas, therefore, could produce very large gross returns for the public purse. Uncertainty due to the residual component of mortality improvements appears to have important fiscal consequences. Perhaps more than health trends, the pace of medical and pharmaceutical innovation is likely to be one of the main drivers of future expenditures and revenues.” “[T]the weighted support ratio is projected to decline by 12 percent between 2010 and 2050. However, that understates the budgetary pressures of population ageing on specific government programs for the elderly, such as Social Security, Medicare, or the institutional portion of Medicaid that pays long-term care costs for some elderly. Government programs finance a sizable proportion of consumption by the elderly, so population ageing will stress the budgets of these programs and contribute to the overall government deficit. Macroeconomic adjustment to population ageing will require some combination of lower consumption (through increased saving or higher taxes) and increased labor supply, perhaps through later retirement.” “As the public pay-as-you-go pension systems of the ageing industrialized countries are likely to become seriously strained under the growing dependency burden, the question arises whether a society should rely on private savings to finance old-age consumption. This is an empirical question about the magnitude and the flexibility of saving rates. The paper takes the German case as an example. (Continued)

33

Jay Bhattacharya Paper/Approach

Faruqee and M¨uhleisen (2003) Neoclassical life-cycle growth model

Brief Summary

Key Finding

Application of MULTIMOD, an international neoclassical lifecycle general equilibrium model of work, investment, consumption, and savings, to population ageing in Japan.

34

A simple computation of the pension gap shows that saving rates must increase in an unprecedented fashion in order to compensate for the dependency effect. However, the analysis of German age and cohort patterns shows that this is unlikely. First, the life cycle structure of German saving rates will not generate a lot of ‘excess’ savings. If one believes the (weak) evidence generated by the separation of age and cohort effects, the opposite is the case. Second, if cohort effects are governed by the rationale underlying a neoclassical growth model, i.e., if an aggregate saving rate emerges that maximizes the long-run welfare of a country, then saving will strongly decline, only somewhat moderated in the case of heavy investment in less ageing countries. In an economy with a shrinking labor force, funding a pension system cannot work as an escape route from a rising burden of dependency. Faltering rates of return prohibit this mechanism. . . . [U]nder moderately realistic assumptions, foreign direct investment helps—it actually helps considerably in terms of consumption possibilities—but the magnitude of the problem is too large to be offset. Because all industrialized countries are ageing, a full offset would require a very large emerging market for investment.” “As the [Japanese] workforce contracts [due to population ageing], output growth will be slow: absent any significant acceleration in total factor productivity, annual growth rates would be lower by about 0.5 percent over the next half century. In per capita terms, GDP per person could also decline (relative to the case of no demographic changes) since the rising share of the elderly suggests an even larger decline in effective labor supply than implied by the shrinking numbers of the workforce. . . . Somewhat surprisingly, saving rates and the current account ratio need not decline significantly with an ageing population. Though a higher share of older persons in the population would tend to reduce saving rates, other aspects of demographic changes in Japan, such as fewer young adults (and young dependents) and increased longevity,

Modeling the Impact of Population Ageing on Future Fiscal Obligations Paper/Approach

Bloom et al. (2015) Narrative review of population-ageing literature

Brief Summary

Key Finding

Review article that emphasizes behavioral and social adaptions to population ageing that may offset some of the anticipated negative economic consequences over the coming decades. The review relies on a wide set of studies that adopt the approaches covered in this chapter.

could counterbalance the negative effects of population ageing on saving rates. . . . Japan will need to implement a long-term fiscal strategy to make public finances sustainable. This paper suggests that public investment cuts, base broadening measures for income taxes, some increase in the consumption tax, and reductions in social security benefits, are likely to be the key building blocks of the longer term solution.” “Population ageing raises concerns about the economic security of older people, health spending, labour supply, tax receipts, savings, and growth of income per person. Assertions about the strong negative effects of population ageing on macroeconomic performance are overblown because countervailing behavioral changes and policy responses are possible. An important and continuing behavioral change—fertility decline—is mitigating the effects of population ageing because it is associated with rising labor force participation by women, increased human capital investment in children, and a decrease in youth dependency. The increasing rate of longevity is sparking another behavioral change—higher rates of savings in expectation of long periods of retirement. Public and private policy responses include a change in retirement policies, emphasis on disease prevention and early detection rather than treatment, a focus on noncommunicable diseases, and making better use of technology.”

National Research Council (2012) findings agree with Elmendorf and Sheiner that population ageing over the coming decades will sharply increase future U.S. government obligations on spending on the elderly (Social Security, Medicare, and nursing homes). They conclude, as do Elmendorf and Sheiner, that financing these obligations will require some “combination of lower consumption (through increased saving or higher taxes) and increased labor supply, perhaps through later retirement.” However, unlike Elmendorf and Sheiner, National Research Council does not specify whether the current generation, future generations, or both should bear these adjustments. Two papers (Faruqee and M¨uhleisen, 2003; B¨orsch-Supan, 1995) employ intricate neoclassical/life-cycle models to calculate how work, investment, consumption, and saving may change in response to population ageing. Faruqee and M¨uhleisen (2003) apply an international model called MULTIMOD to calculate these projections for the Japanese economy. Their 35

Jay Bhattacharya

model implies a sharp (0.5 percentage point) reduction in GDP growth and hence per capita consumption in the coming 50 years in Japan due to its ageing population. One attractive feature of MULTIMOD and similar equilibrium models is that they permit actors in the economies they simulate to adapt to changing economic circumstances that arise due to population ageing to soften the reductions in the number of working-age people in coming decades. These accommodations result in a forecast that the saving rate need not decline. They reason that “Though a higher share of older persons in the population would tend to reduce saving rates, other aspects of demographic changes in Japan, such as fewer young adults (and young dependents) and increased longevity, could counterbalance the negative effects of population ageing on saving rates.” Nevertheless, they forecast a substantial impact of ageing on the long-term fiscal health of the Japanese government because of higher spending on the elderly and suggest increases in taxes and reductions in benefits to address these trends. B¨orsch-Supan (1995) employs an intricate general equilibrium model of the German economy linked to an international investment sector to project the effect of population ageing. Like Faruqee and M¨uhleisen, his model permits substantial accommodations by private actors to ageing, including specifically to saving rates. However, unlike Faruqee and M¨uhleisen, B¨orschSupan’s conclusions are more pessimistic. With his model, he asks what would happen to the welfare of the population if the forecasted increase in retirement and healthcare spending were to be financed entirely by increases in private savings. He concludes that such a high level of private financing of future retirement obligations is unprecedented and that the accommodations predicted by a neoclassical growth model would tend to decrease optimal saving levels rather than increase them. He writes, “if cohort effects are governed by the rationale underlying a neoclassical growth model, i.e., if an aggregate saving rate emerges that maximizes the long-run welfare of a country, then saving will strongly decline, only somewhat moderated in the case of heavy investment in less ageing countries.” Two papers employ a Markov framework of the sort featured in the future elderly model framework. Lee and Edwards (2002) focus on measuring how the uncertainty inherent in population age forecasting affects uncertainty about the future fiscal impacts of ageing. The paper’s formal incorporation of uncertainty distinguishes it from many other papers in this literature, which often model uncertainty by forecasting two “extreme” scenarios and a middle scenario. Lee and Edwards (2002) incorporate formal uncertainty measures into a Markov model of future U.S. Social Security and Medicare spending. Their projects are pessimistic: “Under current program structures, population ageing would be virtually certain to increase the costliness of Federal programs as a share of GDP by 35 percent (±2 percent) by the 2030s, and by 60 percent (±15 percent) in the second half of the century.” However, they acknowledge that such sharp increases are unrealistic and that the U.S. government is likely to respond to such fiscal pressure by altering the structure of benefits provided to the elderly. This prediction about the political economy of the U.S. government does not come directly from Lee and Edwards’ (2002) modeling framework but seems like a sensible prediction nonetheless. Goldman et al. (2010) employ the FEM to ask whether increases in longevity—rather than just population ageing—are likely to increase or decrease fiscal obligations. Their goal is to assess the idea that shortened life expectancy actually decreases government fiscal obligations because, for instance, Social Security benefits do not continue after death. Their main conclusion from the model is that the effect of longevity on future fiscal obligations will depend on the form of medical or other technology that produces those changes in longevity. If, for instance, increased life expectancy arises from inexpensive improvements in the health of the population (perhaps better incentives to keep obesity under control), then increased longevity may reduce government fiscal obligations, as Medicare spending is higher for the unhealthy. However, expensive 36

Modeling the Impact of Population Ageing on Future Fiscal Obligations

new healthcare technologies financed by the U.S. Medicare system may increase longevity while also increasing government obligations. They conclude “perhaps more than health trends, the pace of medical and pharmaceutical innovation is likely to be one of the main drivers of future expenditures and revenues.” Ending this section with an excellent narrative review article of the population ageing literature, Bloom et al. (2015), is appropriate. This article emphasizes the behavioral and social adaptations to population ageing that may offset some of the anticipated negative economic consequences over the coming decades. Their conclusion is optimistic. While it is true, they write, that population ageing will place enormous pressure on future government fiscal obligations, behavioral and technological adaptations will help our societies adjust to the presence of older populations. These changes may include increased optimal private savings because of increased longevity, a shift in healthcare technology and public health toward preventing expensive chronic disease, and declines in fertility, making it more profitable for women to enter the labor force and for people to continue work in healthy old age.

2.6

Suggestions for Future Work

There is no best way to model the effect of population ageing on future outcomes. One goal of this survey is to highlight that regardless of which modeling framework an analyst chooses, there will be trade-offs. Most obviously, a sharp trade-off exists between the level of real-world institutional and policy details included in the model and the data requirements to populate the model’s parameters. Modelers who adopt each approach typically do so because they have a particular outcome that interests them. Hence, the goal is to find a good match between the approach chosen and the strengths of the approach in addressing the purpose of the forecast. That said, all three modeling approaches present difficulties in answering important policy questions. While the microsimulation approach provides the most detailed modeling of the healthcare sector, its focus is entirely on the demand side of the ledger, with an accounting of particular health conditions of the simulated individuals in the model. The accounting and overlapping generation frameworks have a more primitive model of the health status of individuals. None of the three frameworks yield much insight into the supply side of healthcare markets. This is an important gap in the literature because many policy questions require understanding how healthcare providers might respond to demographic shifts. Policies within health economics where suppliers play a crucial role include the imposition of controls on the adoption of medical technologies, physician and hospital payment reforms, pharmaceutical price controls, changes to the financing of health insurance systems, healthcare market structure reforms, and many others. Population ageing is likely to affect all these areas, as some diseases become more common or more expensive to treat. The supply side of healthcare markets needs to be taken more seriously in these models if reliable forecasts are to be made of the effects of these policies. The supply side is also important because changes in population structure are likely to impact the price of care for the elderly directly. As the labor force shrinks, the supply of trained healthcare workers (typically not elderly) who can care for the elderly in care homes will shrink. At the same time, more households will consist of only elderly people, with no family members among the younger generation available to care for them. This will increase the demand for institutional services to care for the elderly. Together, these supply and demand forces will put upward pressure on the equilibrium price of long-term care for the elderly. None of the three modeling frameworks currently incorporate modules that permit this sort of mechanism, which impinges directly on how expensive it will be to provide services for the future elderly. 37

Jay Bhattacharya

Finally, the impacts of the COVID-19 pandemic on population ageing, longevity, and the demands on healthcare systems have been enormous. Older people face the most significant risk of mortality from SARS-CoV-2 infection. According to one meta-analysis of seroprevalence studies, people over 70 face more than 2,000 times higher risk of death upon infection than people under 20 (Axfors and Ioannidis, 2021). A substantial fraction of COVID-19 associated deaths (80+ percent) have been among the elderly worldwide, and life expectancy decreased in developed countries between 2020 and 2021 (OECD, 2021). At the same time, lockdowns and other COVID-19 control measures have impinged on younger populations’ human capital investments and labor force participation. All these developments will undoubtedly have important effects on future population demographics and impacts on fiscal and economic outcomes that analysts in this literature will need to address.

Note 1 I discuss Clements et al. (2018) here at some length as a representative of this literature because of the clear and simple exposition of their methods. Papers that take a similar approach abound, some published long before Clements et al. (2018).

References ACEMOGLU, D., AND JOHNSON, S. (2007): “Disease and development: The effect of life expectancy on economic growth,” Journal of Political Economy, 115(6): 925–985. Available at https://doi.org/10.1086/529000. AXFORS, C., AND IOANNIDIS, J. (2021): “Infection fatality rate of COVID-19 in community-dwelling populations with emphasis on the elderly: An overview,” medRxiv, July 13, 2021. BHATTACHARYA, J., SHANG, B., SU, C. K., AND GOLDMAN, D. (2005): “Technological advance in cancer and the future of medical care expenditures by the elderly,” Health Affairs [Web Exclusive 10.1377/hlthaff.w5.r5-r17], September 26, 2005. BLOOM, D., CANNING, D., FINK, G., AND FINLAY, J. E. (2009): “Fertility, female labor force participation, and the demographic dividend,” Journal of Economic Growth, 14: 79–101. Available at https://doi.org/10.1007/s10887-009-9039-9. BLOOM, D. E., CHATTERJI, S., KOWAL, P., LLOYD-SHERLOCK, P., MCKEE, M., RECHEL, B., ROSENBERG, L., AND S MITH , J. P. (2015): “Macroeconomic implications of population ageing and selected policy responses,” The Lancet, 385(9968): 649–657. Available at https://doi.org/10.1016/S01406736(14)61464-1. BOMMIER, A., AND LEE, R. (2003): “Overlapping generations models with realistic demography,” Journal of Population Economics, 16: 135–160. Available at https://doi.org/10.1007/s001480100102. ¨ BORSCH -SUPAN, A. (1995): “The impact of population ageing on savings, investment and growth in the OECD area,” Institut f¨ur Volkswirtschaftslehre und Statistik. University of Mannheim working paper. Available at https://madoc.bib.uni-mannheim.de/1062/1/512.pdf. CLEMENTS, B., DYBCZAK, K., GASPAR, V., ET AL. (2018): “The fiscal consequences of shrinking and ageing populations,” Ageing International, 43: 391–414. Available at https://doi.org/10.1007/s12126017-9306-6. ELMENDORF, D. W., AND SHEINER, L. M. (2000): “Should America save for its old age? Fiscal policy, population ageing, and national saving,” Journal of Economic Perspectives, 14(3): 57–74. ¨ FARUQEE, H., AND MUHLEISEN , M. (2003): “Population ageing in Japan: Demographic shock and fiscal sustainability,” Japan and the World Economy, 15(2): 185–210. Available at https://doi.org/10.1016/S0922-1425(02)00017-8. GOLDMAN, D. P., HURD, M., SHEKELLE, P. G., NEWBERRY, S. J., PANIS, C. W. A., SHANG, B., BHATTACHARYA, J., JOYCE, G. F., AND LAKDAWALLA, D. (2004): Health Status and Medical Treatment of the Future Elderly: Final Report, TR-169-CMS, Santa Monica, CA: RAND. GOLDMAN, D. P., LEAF, D. E., AND TYSINGER, B. (2016): “The future Americans model: Technical documentation,” CEHD Working Paper, University of Chicago. Available at https://cehd.uchicago.edu/wp-content/uploads/2019/12/fam techdoc.pdf.

38

Modeling the Impact of Population Ageing on Future Fiscal Obligations

GOLDMAN, D. P., MICHAUD, P. C., LAKDAWALLA, D., ZHENG, Y., GAILEY, A., AND VAYNMAN, I. (2010): “The fiscal consequences of trends in population health,” National Tax Journal, 63(2): 307–330. GOLDMAN, D. P., SHANG, B., BHATTACHARYA, J., GARBER, A. M., HURD, M., JOYCE, G. F., LAKDAWALLA , D. N., PANIS, C., AND S HEKELLE , P. G. (2005): “Consequences of health trends and medical innovation for the future elderly,” Health Affairs (Millwood) 24(Suppl 2): W5R5–W5R17. Available at https://doi.org/10.1377/hlthaff.w5.r5. PMID: 16186147; PMCID: PMC6205231. GROSSMAN, M. (1972): “On the concept of health capital and the demand for health,” Journal of Political Economy, 80(2): 223–255. Available at http://www.jstor.org/stable/1830580. JOYCE, G. F., KEELER, E. B., SHANG, B., AND GOLDMAN, D. P. (2005): “The lifetime burden of chronic disease among the elderly,” Health Aff (Millwood) 24 (Suppl 2): W5R18–W5R29. Available at https://doi.org/10.1377/hlthaff.w5.r18. LEE, R. D. (2014): “Macroeconomic consequences of population ageing in the United States: Overview of a National Academy report,” American Economic Review, 104(5): 234–239. LEE, R., AND EDWARDS, R. (2002): “The fiscal effects of population ageing in the US: Assessing the uncertainties,” Tax Policy and the Economy, 16(2002): 141–180. LESTHAEGHE, R. (2014): “The second demographic transition: A concise overview of its development,” Proceedings of the National Academy of Sciences 111(51): 18112–18115. Available at https://doi.org/10.1073/pnas.1420441111. NASEM (NATIONAL ACADEMIES OF SCIENCES, ENGINEERING, AND MEDICINE). (2015): The Growing Gap in Life Expectancy by Income: Implications for Federal Programs and Policy Responses. Washington, DC: The National Academies Press. Available at https://doi.org/10.17226/19015. NATIONAL RESEARCH COUNCIL. (2012): Ageing and the Macroeconomy: Long-Term Implications of an Older Population. Washington, DC: The National Academies Press. Available at https://doi.org/10.17226/13465. OECD (ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT). (2021): “Health at a glance 2021: COVID-19 pandemic underlines need to strengthen resilience of health systems,” November 2021. Available at https://www.oecd.org/health/health-at-a-glance/. OTSU, K., AND SHIBAYAMA, K. (2016): “Population ageing and potential growth in Asia,” Asian Development Review, 33(2): 56–73. ROSER, M. (2019): “Future population growth,” Available at https://ourworldindata.org/futurepopulation-growth. Original version 2013. SAMUELSON, P. A. (1958): “An exact consumption-loan model of interest with or without the social contrivance of money,” Journal of Political Economy, 66(6): 467–482. SCHULTZ, T. W. (1962): Investment in Human Beings. Chicago: University of Chicago Press. SUDO, N., AND TAKIZUKA, Y. (2018): “Population ageing and the real interest rate in the last and next 50 years—A tale told by an overlapping generations model,” Bank of Japan Working Paper Series. No. 18-E-1. Tokyo, Japan.

39

3 MEDICAL INNOVATIONS AND AGEING: A HEALTH ECONOMICS PERSPECTIVE Volker Grossmann1

Abstract This chapter discusses the relationship between medical innovations and ageing from a health economics perspective and surveys empirical evidence on medical research and development incentives, research and development costs of pharmaceuticals, and the cost-effectiveness of health innovations. Particular focus is on the endogeneity of medical technological progress to expected market size and on the conceptualization of ageing as an accumulation of health deficits. This chapter also discusses the role of medical progress for longevity and health inequality and presents a framework to assess the effect of increased longevity on the value of life.

3.1

Introduction

Starting with 19th-century smallpox vaccination programs in most advanced countries, medical technological progress has had a major impact on the evolution of human well-being by reducing morbidity at any age and increasing life expectancy (Cutler et al., 2006; Skinner and Staiger, 2015). The extent of these effects is inevitably linked to the access of the population to health innovations. The well-known conjecture of Newhouse’s (1992) holds that medical innovations induce health spending to grow faster than income. Three channels are conceivable. First, diagnosing or treating any given health condition could become more expensive as medical technology advances. Examples are computed tomography (CT) and magnetic resonance imaging (MRI), which have greatly improved medical imaging compared with standard radiography (X-ray) but are also more costly. Similarly, personalized cancer medicine—based on an analysis of human gene mutations that cause cancer—is typically more expensive than standard chemotherapy (Tannock and Hickman, 2016). With respect to communicable diseases, human immunodeficiency virus (HIV) or hepatitis C virus (HCV) infections have only become treatable quite recently due to modern antiviral pharmaceuticals. Second, broad utilization of new health goods induces demographic change toward a larger fraction of the elderly in the population. This is likely to raise costs for medical treatment despite health status improvements over 40

DOI: 10.4324/9781003150398-4

Medical Innovations and Ageing: A Health Economics Perspective

time for a given age (e.g., Zweifel et al., 2005; Bech et al., 2011; Breyer et al., 2015). Third, better medical technology is likely to increase the demand for health insurance and treatments, thus raising health expenditure for given income (Weisbrod, 1991; Chandra and Skinner, 2012). In turn, medical technological progress itself is endogenous to healthcare utilization, as expected market size faced by potential innovators affects research and development (R&D) investments (e.g., Romer, 1990; Weisbrod, 1991). Thus, limiting healthcare access may have severe long-term consequences for morbidity and longevity advancements by disincentivizing medical R&D effort (Okunade and Murthy, 2002; B¨ohm et al., 2021). This chapter discusses the relationship among medical innovations, demographic change, health expenditure, longevity, and health inequality in the light of empirical evidence. Specifically, it explains how the dynamic interaction between medical technological progress and healthcare spending drives biological ageing.2 The chapter also develops a framework to evaluate the effect of increased longevity on the value of life, examines policy implications, and identifies open research questions. The remainder of the chapter is structured as follows. The following section presents evidence on the evolution of healthcare expenditure, morbidity, and life expectancy. Section 3 reviews the literature on the effectiveness of medical technological progress for improving longevity while section 4 discusses evidence on medical R&D costs and patent values. Section 5 develops a life-cycle model with stochastic survival and discusses how health outcomes affect the value of a statistical life (VSL). Section 6 presents possible conceptualizations to capture the dynamic relationship between medical technological progress and the ageing process in life-cycle models. Section 7 discusses the impact of health innovations on health inequality, and Section 8 suggests avenues for future research. The last section concludes.

3.2

Trends in Health Expenditure, Morbidity, and Life Expectancy

Okunade and Murthy (2002) analyze to what extent per capita real income and technological change (proxied by total R&D and health R&D spending) have driven real healthcare expenditure per capita in the United States during the 1960–1997 period. In support of Newhouse’s (1992), they find a stable, statistically significant, and positive long-run relationship among income, productivity, and health spending. Underlining the importance of health innovations in determining the fraction of total income devoted to healthcare, Acemoglu et al. (2013) and Baltagi et al. (2017) argue that rising income over time (exogenously driven by rising total productivity) cannot explain rising health expenditure shares. In fact, they estimate an income elasticity of health expenditure below unity. Also institutional changes like healthcare reforms and other only occasionally changing variables are inconsistent with the continuous rise of health expenditure shares over time (Chernew and Newhouse, 2011). Figure 3.1 shows the evolution of total health expenditure relative to the gross domestic product (GDP) of selected Organisation for Economic Co-operation and Development (OECD) countries. We see a secular increase in all countries. In the United States, the health expenditure share increased from 6.5 percent in 1970 to 17 percent in 2019, the highest level among OECD countries. Other advanced countries like Germany had similar health expenditure shares as the United States in 1970 but ended up with considerably lower ones (12 percent or less). Transition countries, represented in Figure 3.1 by the Czech Republic, started out at lower levels in 1990 but show a similar upward trend. About 12–14 percent of total health spending was on nondurable medical goods like pharmaceuticals in France, Germany, Switzerland, the United Kingdom, and the United States and about 18 percent in Italy and Japan (OECD, 2020). 41

Volker Grossmann

Figure 3.1 Evolution of total health expenditure as percentage of GDP, 1970–2018. Source: Data from OECD (2020).

Remarkably, according to Table 3.1, doctor visits per capita in the United States (2.5 in 2019) are considerably fewer than, say, in Germany and Japan (9.9 and 12.6 in 2019, respectively), despite considerably higher per capita health spending in the United States. Also the number of physicians and hospital beds relative to population size in the United States is at the lower end among advanced countries while there is a more extensive use of technology-intensive examinations (particularly CT exams) than in other countries. Switzerland stands out with respect to the number of nurses relative to its population size, whereas Japan has the highest number of hospital beds.

Table 3.1 Healthcare utilization and healthcare resources (selected indicators), 2018 or nearest year

Note: a. 2017, b. 2016, c. 2014, d. 2013, e. 2011, f. 2009, g. Exams in hospitals only. Source: Data from OECD (2020).

42

Medical Innovations and Ageing: A Health Economics Perspective

Figure 3.2 documents, for both males and females, considerable and gradual increases of remaining life expectancy at age 65 in most countries since 1960, based on contemporaneous mortality rates (period life expectancy). The increase was particularly high in Japan, with about 8

Figure 3.2 Evolution of remaining period life expectancy at age 65 (in years), 1960–2018. Source: Data from OECD (2020).

43

Volker Grossmann

years for males and 10 years for females. Japan also has the highest remaining life expectancy for females in 2018 (24.4 years) and is in the top group for males jointly with Switzerland, France, Italy, and Australia. The remaining life expectancy of males in the United States, Germany, and the United Kingdom is about 18 years—about 2 years less than in the leading countries. In these countries, females can expect to live about 21 more years when reaching age 65, which is about 3 years less than in Japan and France. Paralleling rising life expectancy over time is a slowdown in the biological ageing process as represented by the evolution of the so-called health deficit (or frailty) index, which is defined as the fraction of bodily impairments present in a person out of a sufficiently large list of potential health deficits.3 Empirically, health deficits correlate quasi-exponentially with age (e.g., Mitnitski et al., 2002a; Harttgen et al., 2013) and are a highly relevant determinant of the probability of death (e.g., Mitnitski et al., 2002a,b, 2005, 2006, 2007). Abeliansky and Strulik (2019) compute a health deficit index for a panel of 14 European countries and six waves of the Survey of Health, Ageing, and Retirement in Europe (SHARE). They document that later born cohorts, at the same age, are healthier than earlier born cohorts. For each year of later birth, health deficits decline by 1.4–1.5 percent on average. Differences between men and women, among countries, and over time are insignificant. The level of health deficits experienced at age 65 by individuals born in 1920 resembles that experienced at age 85 by individuals born in 1945. In a similar vein, Abeliansky et al. (2020) focus on the 50–90 age group to study the evolution of the health deficit index of the U.S. Americans born from 1904 to 1966. Using 13 waves of the Health and Retirement Study (compiled by the RAND Center of the Study of Ageing) from 1992 to 2016, they find that the average elderly American develops 5 percent more health deficits per year. For each year of later birth, health deficits decline by about 1 percent on average, documenting steady improvements in the health status over time. Figure 3.3 illustrates

Figure 3.3 Year-of-birth effects in the relationship between the log of health deficits and age in the United States, birth cohorts 1910–1959. Note: Displays coefficients of year-of-birth (yob) dummies in a regression of the type log(di ) = α + β · P agei + t γt · yobit + ϵi , where di denotes the health deficit index of individual i, t is the year of birth, and ϵi is the error term. See Abeliansky et al. (2020) for details. The lines show linear time trends of gender-specific coefficients γt . Source: Data from Abeliansky et al. (2020).

44

Medical Innovations and Ageing: A Health Economics Perspective

the health trend by plotting the estimated coefficients of year-of-birth dummy variables for birth cohorts 1910–1959 in a regression with the log of the individual health deficit index as dependent variable. The main explanatory variable is individual age, capturing that health deficits grow exponentially over the life-cycle (e.g., Mitnitski et al., 2002a; Harttgen et al., 2013).4 We see a slightly slower decline in health deficits over time for men than for women.

3.3

Health Effects and Cost-Effectiveness of Medical Innovations

Mounting evidence highlights the importance of medical innovations in health outcomes and the cost-effectiveness of health treatments. For instance, antibiotics like penicillin and sulfa drugs, which were invented in the 1930s and 1940s for treating bacterial infections that cause pneumonia, tuberculosis, syphilis, dysentery, and bacterial meningitis, substantially reduced mortality rates in the United States in the mid20th century (Cutler et al., 2006). Incremental advances to overcome resistance to the first- and second-generation antibiotics followed. Lichtenberg (2007) shows that pharmaceutical innovation significantly improves health outcomes for 92 potentially lethal diseases. Using prescription drug use data from 1996 to 2003, he finds that a higher percentage of prescriptions of later vintages has led to larger declines in mortality rates and hospitalization in the United States. More recently, Lichtenberg (2020) shows that increases in the approvals of new cancer drugs in the United States in the period 2000– 2014 have been associated with larger declines in premature mortality and hospitalization. Zhuo et al. (2020) developed a microsimulation model for Japan to show that directly-acting antivirals (DAAs), a class of HCV infection treatment available since 2014, cost less than US$10,000 per quality-adjusted life year (QALY) gained by the treatment.5 At the mean age of those infected with HCV (age 60), life expectancy rose by about 3 years. Drugs suppressing HIV may be less cost-effective. Borre et al. (2017) estimated that implementing the U.S. National HIV/AIDS Strategy would on average cost US$68,900 per QALY gained. However, that figure is still below the US$100,000–150,000 cost-effectiveness threshold suggested by more recent literature (e.g., Neumann et al., 2014). Critics point out that not all pharmaceutical R&D effort is targeted to improving health effects of treatments. So-called “me-too” drugs are a prime example. These have similar chemical structures as an original drug and are used for the same therapeutic purposes as “first-inclass” drugs (Gagne and Choudhry, 2011; Aronson and Green, 2020). Examples are the many tricyclic antidepressants, beta-blockers, and statins. Some “me-too” drugs are the outcome of parallel development in different companies, whereas others come from R&D targeted at the purpose of obtaining a new patent with comparably little effort (R´egnier, 2013). The latter is problematic when it raises costs for consumers vis-a-vis generics without extra value. However, some “me-too” drugs have fewer side effects, less drug-drug interactions, and show greater efficacy at least for some patient groups (Lakdawalla, 2018). Cutler et al. (2006) argue that, despite large morbidity effects, vaccination has played a minor role in short-run mortality reduction, except for the successful eradication of poliomyelitis in many countries. That said, morbidity caused by diseases like measles, hepatitis B, yellow fever, and tetanus may significantly raise mortality risk in the longer run. In other words, vaccination against these diseases could significantly improve longevity. The same is potentially true for COVID-19 vaccines that not only reduce the number of immediate deaths but also prevent longer-term bodily impairments after convalescence (Long COVID) that could lead to the development of further health deficits. 45

Volker Grossmann

An example of nonpharmaceutical medical technological progress involves treating stenosis, a coronary artery disease that narrows coronary arteries in a life-threatening way. Instead of placing bare-metal stents (BMS) via balloon dilation during a percutaneous coronary intervention, it is now possible to use coronary drug-eluting stents (DES). These release antiproliferative and anti-inflammatory substances to avoid the frequent recurrence of stenosis associated with BMS (Baschet et al., 2016). Second-generation DES have been introduced to prevent thrombosis that may be triggered by placing stents. In a meta-analysis, Baschet et al. (2016) show that DES leads to fewer complications than BMS and is generally cost-effective. Ford et al. (2007) argue that about half of the decrease in the age-adjusted death rate for coronary heart disease can be attributed to treatments and the other half to changes in risk factors (e.g., lower cholesterol levels and lower systolic blood pressure).

3.4

R&D Costs and Patent Values

An extensive literature addresses the R&D costs of new drug discoveries. Such information is important to discuss whether price-setting power is stronger than needed to provide R&D incentives. The potential to lower prices of pharmaceuticals without compromising medical progress may indirectly improve longevity by extending drug access in healthcare systems. In a meta-study based on 13 articles published from 1980 to 2009, Morgan et al. (2011) report estimates of the average non capitalized cost of drug development in the wide range of US$92 million to US$883.6 million in 2009 dollars, where estimates for later periods are larger.6 DiMasi et al. (2016) estimate the R&D costs of 106 randomly selected new drugs from 10 pharmaceutical firms that were first tested in humans between 1995 and 2007. They account for abandoned compounds during clinical trials by linking them to the costs of compounds that obtained market approval; that is, they correct for R&D failure risk.7 Their evidence suggests an average non capitalized R&D cost per new drug prior to approval of US$1,395 million in 2013 dollars (and, on average, additional R&D costs of US$566 million after initial approval). Average capitalized costs based on an annual real interest rate of 10.5 percent amount to US$2,558 million. DiMasi et al. (2016) also provide a literature review suggesting that R&D costs have risen substantially in the last decades while success rates have decreased. Estimating R&D costs from publicly available sources is preferable to self-reported costs.8 Wouters et al. (2020) provide high-quality estimates for 63 (out of 355) new therapeutic drugs and biologic agents approved by the U.S. Food and Drug Administration (FDA) between 2009 and 2018. Dividing R&D expenditures by clinical phase–specific success rates to correct for failed trials, the estimated median capitalized R&D cost (based on a real interest rate of 10.5 percent) for a single drug was US $985 million in 2018 dollars (mean costs were US$1,335.9 million). Prasad and Mailankody (2017) estimated the R&D costs for 10 cancer drugs approved by the FDA between 2005 and 2015, reporting noncapitalized median and mean cost of development of a single drug of US $648 million and US$719.8 million in 2017 dollars, respectively. Capitalized with an interest rate of 9 percent, the median cost was US$793.6 million and the mean cost was US$969.4 million. R&D costs for novel drugs were higher than for next-in-class drugs. Prasad and Mailankody (2017) report a mean sales revenue of US$6,699.1 million. This suggests that R&D costs were covered by a wide margin also when correcting for R&D failure risk. In a similar vein, Tay-Teo et al. (2019) estimate for 99 cancer drugs approved by the FDA from 1989 to 2017 that the median sales revenue was about 14.5 times higher than the median R&D costs estimated by Prasad and Mailankody (2017). They also found that the drugs continued to generate high revenues after patent expiry. 46

Medical Innovations and Ageing: A Health Economics Perspective

Revenue is not profit, however. Estimating profits also requires data on production and marketing costs. Songane and Grossmann (2021) estimate the ratio of R&D costs to the global present-discounted value (PDV) of annual profits until patent expiry of the leading human papillomavirus (HPV) vaccine Gardasil by Merck. They arrive at an estimate of 2.5–6.8 percent, depending on the assumed discount rate. This is considerably lower than the success rates in clinical trials for vaccines reported in the literature, suggesting stronger market power than needed to incentivize R&D.9 Songane and Grossmann (2021) also estimate that marketing costs for Gardasil are at least as high as R&D costs. This parallels Lakdawalla (2018, p. 438), who states: “The pharmaceutical industry spends approximately the same amount of money on marketing as it does on innovation investments.”

3.5

Life-Cycle Considerations and the Value of Life

This section presents a life-cycle model with stochastic survival. The framework enables us to discuss how health improvements affect welfare and to analyze the endogenous interaction between medical R&D investment and longevity.

3.5.1

Lifetime Utility, Budget Constraint, and Optimization

Consider an age-structured population in discrete time with a homogenous group of individuals of size Nv,t from cohort v in period t. Each period a new cohort is born. Accounting for survival probabilities and discounting, expected remaining lifetime utility at age s ≥ 0 of a representative agent from cohort v is given by10 Uvs (i) =

v+T−1 X

ρ t−v Nv,t Sv,t u(cv,t , ℓv,t ; dv,t ),

(1)

t=v+s

where Sv,t is the unconditional probability to survive to age t − v, cv,t denotes the per capita consumption level of a single final good (with price normalized to unity) in period t, ℓv,t the hours worked, dv,t the individual health deficits, ρ ∈ (0, 1] the discount rate, and T > 0 the maximum length of life. u(c, ℓ; d) is the instantaneous utility function, which is quasi-concave as a function of choice variables (c, ℓ) and has derivatives uc > 0, uℓ < 0, ud < 0. Moreover, suppose cross-derivative ucd < 0 holds, in line with the evidence that lower health status (more health deficits) is associated with a lower marginal utility of consumption (Finkelstein et al., 2013).11 For t ≥ v, financial wealth of cohort v accumulates according to Av,t+1 = (1 + rv,t )Av,t + Nv,t (yv,t − cv,t ),

(2)

where 1+rv,t is the cohort-specific interest factor in a perfect annuity market between date t and t + 1 and yv,t is (net) non-wealth income. Initial asset holding Av,v+s ≥ 0 is given and terminal condition Av,v+T ≥ 0 must hold. Denote by wv,t the cohort-specific wage rate. Non-wealth income consists of net earnings and other life-contingent income (like pension benefits), Iv,t , that, for simplicity, is considered exogenous: yv,t = Fv,t (wv,t ℓv,t ) + Iv,t ,

(3)

where function Fv,t transforms gross earnings, wv,t ℓv,t , into net earnings (after taxes, social security contributions, and health insurance contributions).12 47

Volker Grossmann

The mortality rate is the probability mv,t of a member from cohort v dying between period t and t + 1, conditional on having reached age t − v ≥ 0. By definition, survival rates, Sv,t , and S −Sv,t mortality rates are related by mv,t = − v,t+1 . For simplicity, consider a small open economy Sv,t with exogenous market interest rate, ¯r . In a perfect annuity market with fair insurance within a cohort,13 1 + ¯r . (4) 1 + rv,t = 1 − mv,t−1 Q Using (3), (4), mv,v+s−1 = 0, Sv,t = Sv,v+s t−1 u=v+s (1 − mv,u ), and the fact that terminal condition, Av,v+T ≥ 0, must be binding (as not holding wealth after certain death is optimal), equation of motion (2) for asset accumulation from the perspective of age s implies that ˜ v,v+s + 0=A

v+T−1 X 

Nv,t Sv,t

t=v+s

F(wv,t ℓv,t ) + Iv,t − cv,t (1 + ¯r )t−v−s



,

(5)

˜ v,v+s ≡ (1 + ¯r )Av,v+s ≥ 0. The utility maximization problem from the perspective of where A age s then reads as max

{cv,t ,ℓv,t }

v+T−1 X

ρ t−v Nv,t Sv,t u(cv,t , ℓv,t ; dv,t ) subject to (5).

(6)

t=v+s

The first-order conditions result in the standard Euler equation that governs the motion of consumption. When time path {Iv,t } is exogenous,14 the well-known labor supply condition that equates the marginal rate of substitution (MRS) between consumption and working hours with the derivative of net earnings (3) with respect to labor supply also holds.

3.5.2

Value of a Statistical Life

The VSL is defined as the willingness to pay for an additional person of age s in terms of wealth (Hall and Jones, 2007).15 Combining the first-order conditions of the household optimization problem (6) with (5), one can derive in analogy to Murphy and Topel (2006) the VSL for a member of cohort v at age s, VSLvs , as16 VSLvs =

v+T−1 X t=v+s

˜ v,v+s Sv,t u(cv,t , ℓv,t ; dv,t ) A . − (1 + ¯r )t−v uc (cv,t , ℓv,t ; dv,t ) Nv,v+s

(7)

It is given by the PDV of the expected stream of the life-year utility values, u/uc , minus initial per capita wealth. Adjustment for the marginal utility of consumption transforms utils into real dollars. Interestingly, for given survival rates and a given age, the VSL is not necessarily higher for those with better health. According to (7), this would require u(c, ℓ, d) to decrease faster in health deficits d than uc (c, ℓ, d). Murphy and Topel (2006) gauge the impact of exogenous longevity improvements on the VSL in a calibrated life-cycle model with stochastic survival.17 The quality of life (health status) entering instantaneous utility u is not linked to mortality and u/uc is independent of the quality of life. Notably, these assumptions imply that the VSL is independent of the quality of life, too. According to their analysis, the representative U.S. individual gained about 1.2 million US$ over the 20th century and that the gains to society between 1970 and 2000 are worth about half of GDP. This suggests that health innovations that improve longevity have substantial effects on the value of life—a result that is best understood from the feature that the marginal 48

Medical Innovations and Ageing: A Health Economics Perspective

intertemporal utility from increasing survival rates is non decreasing, according to (1) and (7). Lakdawalla et al. (2017) argue that Murphy and Topel even underestimate the benefits from improvements in the quality of life because they neglect the insurance value from lowering health risk that comes from health innovations. Jones and Klenow (2016) measure welfare gains from increased longevity by consumption equivalents rather than by estimating the effects on the value of life. They argue that Western Europe is closer to the United States in terms of welfare than in terms of GDP per capita thanks to higher life expectancy. In contrast, welfare differences between developing countries and the United States are greater than GDP per capita differences.

3.6

Morbidity, Healthcare Demand, and Medical R&D

This section links healthcare demand to incentives for health innovations under alternative conceptualizations of the relationship between health status and mortality.

3.6.1

Health Deficit Approach

Assuming that the quality of life enters instantaneous utility u independently of mortality, Murphy and Topel (2006) note that: “For example, technologies that improve mental health or reduce the effects of arthritis may increase instantaneous utility without affecting longevity” (p. 876). This view starkly contrasts gerontology research on the health deficit index, as surveyed by Strulik (2023). Mitnitski et al. (2002a,b, 2005, 2006) and Mitnitski et al. (2007) show that the presence of (many) health deficits is conducive to the development of further health deficits and that a lower health deficit index is associated with lower mortality rates in the elderly population. Their list of potential health deficits includes, for instance, the physical difficulty to move, which may contribute to developing cardiovascular disease. This blurs the border between life-threatening diseases and bodily impairments. Formally, the argument suggests that the (average) mortality rate within cohort v in period t can be written as mv,t = M(dv,t ), M ′ > 0, where dv,t follows a first-order difference equation. Consequently, cohort-specific mortality rates become path-dependent (i.e., technically, they are state variables). Hence, even if the quality of life as measured by health deficits did not affect u/uc in the VSL expression in (7), it could still affect the VSL by affecting mortality. Dalgaard and Strulik (2014) incorporate the notion of human ageing as an (approximately exponential) health deficit accumulation process in a life-cycle model.18 For our purposes, suppose health deficits of a member from cohort v evolve according to dv,t+1 − dv,t = αv · dv,t − g(hv,t ),

(8)

where hv,t is a measure of health inputs (preventive and curative healthcare) and g(·) is an increasing function. From the perspective of age s, dv,v+s > 0 is given. Parameter αv > 0 is the growth rate of the health deficit index absent health interventions. It may be cohort-specific because of environmental and cultural factors. To capture the higher need for health treatment of more impaired individuals, suppose the health input is an increasing function of both health deficits and the average quality (q) of the latest vintages in a wide range of health goods (and services) available in the market. B¨ohm et al. (2021) motivate the form g(hv,t ) = hv,t = κt · dv,t · qt , where κt may be interpreted as a policy parameter that captures the access to health treatment in the health insurance system. Combined with (8), the growth rate of (average) health deficits in cohort v is given by αv −κt ·qt . It depends on the interaction between access and quality of health goods, which is consistent with Murray (2017). Medical progress raising the quality of health goods thus reduces the growth rate of 49

Volker Grossmann

health deficits over time, consistent with the observation of Abeliansky and Strulik (2019) and Abeliansky et al. (2020) that later born cohorts display, for a given age, fewer health deficits (Figure 3.3). The effect is enlarged when access to modern health goods improves. Analogously to models of endogenous technological change in growth economics (e.g., Romer, 1990; Grossman and Helpman, 1991; Aghion and Howitt, 2005), the evolution of the average quality of health goods may be conceptualized as first-order difference equation qt+1 − qt = Q(Lt , qt ),

(9)

where Q is an increasing function of medical R&D (labor) input, L, and may depend on health good quality, q.19 For instance, B¨ohm et al. (2021) arrive at process (9) in a framework where competitive R&D firms invest in risky medical R&D. R&D incentives are determined by the expected PDV of the profit stream resulting from a successful innovation. Consequently, medical R&D investments critically depend on the size of the market for a new health good (vintage), i.e., on health deficits among the current and future age-structured population and on healthcare coverage (parameter κt ). The latter feature echoes Weisbrod (1991) who conjectures that the expansion of U.S. healthcare insurance has incentivized medical R&D. The baseline calibrated model of B¨ohm et al. (2021) suggests substantial future increases in cohort life expectancy.20 Cohort life expectancy at birth for those born in the 21st century is predicted to exceed 100 years, and considerably lower health deficits are expected among the elderly for later cohorts. This is caused by endogenous medical progress that raises qt over time. The associated change in the demographic structure is predicted to increase the health expenditure share moderately, which is in line with empirical evidence (e.g., Zweifel et al., 2005; Bech et al., 2011; Breyer et al., 2015). Pervasively limiting access to healthcare in advanced economies (leading to a decrease in κt over time) to prevent health expenditure shares from increasing is predicted to considerably lower life expectancy for future generations compared with the baseline model. Calibrating the VSL to US$6 million, this would lead to a welfare loss (measured by consumption equivalents) of about one-fifth for those born in the beginning of the 21st century. The welfare loss is even higher for later cohorts. The main reason for the considerable size of the detrimental effects of healthcare rationing is the associated reduction in market size, which disincentivizes health innovations. This echoes Chandra and Skinner (2012) who point to the economic and political resistance in the United States to finance possibly rising health expenditure, with potentially adverse effects on medical technological progress.

3.6.2

Health Capital Approach

A key feature of the health deficit approach that makes it particularly suitable for analyzing healthcare rationing effects is the path dependence of health deficits and mortality rates. The widely used health capital approach of Grossman (1972) also captures path dependence of health status. The Grossman model treats individual health status as a latent state variable (health capital) that individuals can invest in and views ageing as depreciation of the health capital stock rather than as an accumulation of health deficits. Formally, this is like capturing human capital accumulation (e.g., Lucas, 1988). It is fair to say, however, that it lacks foundation in medical science and has some undesirable implications. For instance, the health capital model assumes that an individual with good health experiences a greater decline in health than an individual with poor health (via health capital depreciation), contrary to the empirically established quasi-exponential growth of health deficits over the life cycle. It also implies that health investments decline in old age and near death (Strulik, 2015), contrary to the evidence suggesting the opposite. The most interesting application for the purpose of this chapter is the contribution by Fonseca et al. (2021), who analyze an elaborate stochastic life-cycle model, where agents choose 50

Medical Innovations and Ageing: A Health Economics Perspective

consumption, medical expenditure, and labor market participation (labor supply at the extensive margin). According to their analysis, medical technological progress is responsible for half of the increase in U.S. life expectancy over the period 1965–2005 but does not significantly contribute to observed healthcare spending growth. Rather, their evidence suggests that health insurance extension and general income growth have driven health expenditure growth. In line with this result, Finkelstein (2007) finds that the introduction of Medicare in 1965 led to a large increase in U.S. hospital spending. Her analysis suggests that the overall spread of health insurance since 1950 could be responsible for about half of the increase in real per capita health spending.

3.6.3

Non-Path-Dependent Mortality

Hall and Jones (2007) and Jones (2016) make eye-opening contributions to the evolution of socially optimal investments in health in a growing economy. They consider a representative individual of a single cohort, assuming that health spending contemporaneously affects mortality rates. In the context of an age-structured population, this means that the mortality rate of a member of cohort v at time t is given by mv,t = m(h ˜ v,t ) with m ˜ ′ < 0, where hv,t denotes health spending. Thus, unlike in the health deficit approach where mv,t = M(dv,t ), the mortality rate is a flow variable rather than a state variable. Both papers point to a major role of the elasticity of consumption utility, γ ≡ −cucc /uc . According to Hall and Jones (2007), if the marginal consumption utility is falling rapidly (i.e., if γ is significantly above unity), the health expenditure share in the economy should increase more than 30 percent until 2050. The result can be understood by the fact that, unlike consumption utility, increased survival linearly raises lifetime utility. As a result of rising health spending, life expectancy at birth would also increase significantly. In fact, γ ≫ 1 is in line with most empirical estimates of the intertemporal elasticity of substitution (e.g., Havr´anek, 2015).21 In Hall and Jones (2007), medical technological progress is exogenous and reduces mortality rates over time for given health spending. In contrast, Jones (2016) endogenizes health technology. R&D labor can be directed to improving the effectiveness of health spending on mortality reduction or the quantity of the material consumption flow. This puts the optimal direction of technological change at the center of the analysis. Assuming m(h) ˜ = h−β , Jones (2016) shows that in his favored case where γ > 1 + β it is optimal that the fraction of scientists into medical R&D and the fraction of non-R&D labor for producing the health input both converge to unity in the long run. For γ < 1 + β, the opposite result holds, i.e., the fraction of labor allocated to the life-saving sector should approach zero in the long run. Frankovic and Kuhn (2018) propose a multi period overlapping generations model with endogenous health innovations that interact with the demand for healthcare. Mortality risk depends on health spending and the state of medical knowledge. Medical R&D investments are paid for by profits made in the healthcare sector, which features decreasing returns. The analysis shows that the expansion of health insurance was more important for U.S. health expenditure growth in the period 1960–2010 than (exogenous) income growth. The model can also explain a significant part of the observed U.S. life expectancy gains. As in B¨ohm et al. (2021), better access to healthcare fosters medical technological progress. Both contributions shed light on the interrelation between health expenditure and health innovations that are driven by the demand for healthcare.

3.6.4

Market Size Effects: Empirical Evidence

Does the empirical evidence support the transmission channel from changes in market size to changes in medical technology, as hypothesized by Weisbrod (1991)? Empirical estimates differ 51

Volker Grossmann

in both the measure of market size and the outcome measure (medical R&D investments or patents). Acemoglu and Linn (2004) and Dubois et al. (2015) are interested in the determinants of the number of prescription drug approvals. Acemoglu and Linn (2004) exploit that market demand for each therapeutic class changes over time as the age structure changes because the risk of specific illnesses varies with age. Specifically, they construct for different drug categories potential drug market size by summing up the products of age-specific total income and agespecific drug expenditure shares over age groups, using U.S. data for the period 1970–2000. Their analysis suggests a major role of current market size for nongeneric drug approvals. Moreover, they find that both current and future market size considerably affect the approval of new molecular entities and generics. Anticipated increases in market size with a 10–20 year lead time significantly raises R&D investments. This is plausible given the considerable time span between the start of R&D investments and drug approvals. By contrast, past market size does not play a significant role. Dubois et al. (2015) exploit data for 14 countries (10 advanced countries plus Brazil, China, Mexico, and Turkey) and focus on new chemical entities. Market size for a therapeutic category is measured by the PDV of global sales revenue over 20 years (partly imputed). To address potential reverse causality problems, sales revenue for each country is instrumented by its GDP per capita, population size by gender and age (total and elderly), and mortality by disease targeted by a therapeutic category. The estimated equations include fixed effects for therapeutic categories and time. They find that a 1 percent increase in market size raises the number of innovations by 0.23 percent. Split by therapeutic categories, the market size elasticity is lowest for the cardiovascular system and blood-forming organs; it is highest for the nervous system and sensory organs. Finkelstein (2004) and Blume-Kohout and Sood (2013) exploit policy reforms in the United States that affect market size and look at their effect on clinical trials. Finkelstein (2004) estimates the effect on the number of new vaccine clinical trials of (1) the adoption of the 1991 Centers for Disease Control and Prevention (CDC) recommendation that all infants be vaccinated against hepatitis B; (2) the 1993 decision that Medicare (U.S. government health insurance for the elderly) covers all costs of influenza vaccinations for beneficiaries; and (3) the introduction of the Vaccine Injury Compensation Fund in 1986, which liberated manufacturers from liability in case of adverse reactions to vaccination against polio, diphtheria-tetanus, measles-mumpsrubella, and pertussis. On average, these policies increased the number of clinical trials for new vaccines by a factor of 2.5. Blume-Kohout and Sood (2013) examine the effect of introducing coverage of prescription drugs by Medicare Part D in 2006 on the number of pre-clinical and clinical trials. They find that a higher market share of prescriptions filled by Medicare beneficiaries is associated with larger increases of pharmaceutical R&D effort after implementation of Medicare Part D. Combined with the evidence that Medicare Part D has substantially increased the demand for pharmaceuticals among the elderly (e.g., Lichtenberg and Sun, 2007; Duggan et al., 2008; Schneeweiss et al., 2009), this is compelling evidence that market size matters for R&D incentives. Looking specifically at developing countries is also interesting. Using patent data for the period 1993–2009, Zhang and Nie (2021) show that the implementation of a public health insurance program for rural residents in China considerably affected the number of pharmaceutical patents targeting diseases that are prevalent in rural areas. This does not necessarily mean, however, that stronger patent protection in developing countries spurs pharmaceutical innovation of foreign-based multinationals. Kyle and McGahan (2012) estimate how market size—defined as the number of deaths from a certain disease—contemporaneously affects the number of new clinical trials targeting the disease, and how it interacts with the adoption of the 52

Medical Innovations and Ageing: A Health Economics Perspective

Trade-Related Aspects of Intellectual Property Rights (TRIPS) agreement.22 They find that TRIPS has not significantly raised R&D effort targeted to the diseases that are most prevalent in developing countries and explain the result with low drug affordability in those markets.

3.7

Effect of Health Innovations on Health Inequality

A pronounced socioeconomic health gradient exists. For instance, OECD (2019, Figure 3.5) indicates that the gap in life expectancy at age 30 between the tertiary educated and those with less than secondary schooling is, on average for 25 OECD countries, 6.9 years for men and 4 years for women. The gap among OECD members is highest in Eastern European countries. The literature offers various explanations for a socioeconomic health gradient related to differences in education, financial means, and the job environment. Ehrlich and Chuma (1990) show within the health capital framework of Grossman (1972) that optimal health expenditure and longevity that is endogenous to health spending depends on initial wealth. Becker (2007) discusses the education-health gradient in a two-period framework with endogenous investments in skill and health. Strulik (2018a,b, 2019) endogenizes unhealthy consumption within the health deficit framework. Strulik (2018a) shows that a higher return to education is not only linked to more education but also to a healthier lifestyle. Strulik (2019) explains unhealthy consumption by limited self-control. Strulik (2018b) and Galama and van Kippersluis (2019) present life-cycle models with unhealthy behavior suggesting that declining marginal consumption utility is key for understanding the socioeconomic health gradient. Galama et al. (2018) survey the literature on causal effects of education (and education policy) vis-a-vis noncognitive skills on health (and unhealthy behavior) and report mixed evidence. Conti et al. (2010) point to the important role of cognitive, noncognitive, and health endowments during childhood in health disparities among adults. How do medical innovations affect the socioeconomic health gradient? To address this question, Glied and Lleras-Muney (2008) employ data on disease-specific mortality rates for 1980 and 1990 and cancer registry data for 1973–1993. Using state-level and cohort-level variation in compulsory schooling years, they find that a higher number of active ingredients (new molecular entities) approved by the FDA for treating a disease increasingly reduces mortality risk from that disease as compulsory schooling levels rise. Chang and Lauderdale (2009) provide further evidence, finding that for the period 1976–2004 higher-income earners had higher cholesterol levels before and lower ones after the introduction of statins. Frankovic and Kuhn (2019) develop an overlapping-generations framework, where individual income is positively associated with healthcare utilization and education affects the effectiveness of medical progress on individual mortality reduction. Goldman and Lakdawalla (2005) show that the effects of health innovations on health inequality depend on the complexity of treatment regimens. If these are complex, as in the case of HIV medicine, better educated individuals benefit more. The opposite holds if treatments become easier to adopt. For instance, new drugs to treat hypertension have led to a decline in cardiovascular disparities. The evidence for the United States suggests rising survival inequalities over time (e.g., Chetty et al., 2016, as reviewed by Bor et al., 2017). For Europe though, Abeliansky and Strulik (2019) find that health deficits of individuals with low socioeconomic status decline at about the same rate as for individuals with high socioeconomic status. Even though socioeconomic health disparities may not be increasing, their evidence suggests long-run persistence of health inequality. Grossmann and Strulik (2019) also point to the possibility that, with path-dependent health status, medical technological progress benefits particularly those initially in good health, thus potentially raising health inequality. 53

Volker Grossmann

3.8

Avenues for Future Research

A particularly fruitful area for future research would be to examine in more detail the mechanisms that govern the interaction between the socioeconomic health gradient and medical technological progress. Allowing for out-of-pocket health spending or private health insurance in a heterogenous agent version of the presented health deficit framework with endogenous medical R&D would be interesting. Out-of-pocket spending would enable more affluent individuals to supplement publicly provided healthcare. Healthcare rationing would then potentially raise health inequality. A quantitative assessment would be highly desirable to understand the resulting distributional conflict. Also, more evidence is needed to understand potentially differential effects of health innovations on ageing in heterogeneous populations stratified by income, education, wealth, or occupation. For instance, knowing which policies provide information on the availability, safety, and effective use of state-of-the-art diagnosis, treatment, and vaccines could improve healthcare utilization of less educated individuals. We also need to know more about the role of differences in the practice patterns of medical professionals for healthcare access (e.g., Chandra and Skinner, 2004). Regarding the effect of health improvements on the VSL, incorporating a feedback effect of higher life expectancy on investments in human capital and entrepreneurship in the analysis of stochastic life-cycle models would be interesting. Particularly in a development context, an increase in longevity means that the returns from such investments can be spread over longer time horizons, in turn fostering economic activity and potentially raising welfare (e.g., Cervellati and Sunde, 2005, 2011, 2013; Strulik and Werner, 2016).23 Also importantly, longevity gains are not entirely exogenous to the individual. Future research should thus incorporate health investment choices and their interaction with public health systems and medical R&D incentives in VSL estimates. Accounting for unhealthy consumption choices would also be interesting. Relatedly, there is yet no conclusive evidence on the contribution of medical technological progress on longevity vis-a-vis other factors like behavioral changes and public health efforts. Finally, studying policies that foster the diffusion of health innovations among healthcare providers (e.g., hospitals) would be important. In fact, empirical evidence suggests that technology diffusion is slow, producing a significant time gap between early and late adoption (Skinner and Staiger, 2015). This may imply a sizable welfare loss (Frankovic et al., 2020).

3.9

Conclusion

In the medical literature, explaining and predicting life expectancy trends is typically based on estimating statistical time trends (e.g., Kontis et al., 2017). However, medical technological progress is largely affected by R&D incentives that, in turn, depend on the evolution of demand for health goods and services. Market size is critically determined by healthcare access, which is endogenous to public health policies. The key insights of this chapter may be summarized as follows. First, medical technological progress has potentially large effects on the evolution of life expectancy and the value of life. Second, new health treatments that improve health status are typically cost-effective, but likely to raise health expenditure as a fraction of income. Third, rising health expenditure shares over time can be socially optimal in a growing economy. Fourth, expected market size is an important determinant of medical R&D expenditure. Demand for health innovations in advanced countries is therefore the driving force of medical technological progress. Scope may still exist to reduce the market power of pharmaceutical companies, however. Typically, revenues and profits considerably exceed R&D costs even when accounting for failure risk. Fifth, the gerontologically founded health deficit model that displays positive path dependence of both 54

Medical Innovations and Ageing: A Health Economics Perspective

individual health status and mortality rates is particularly suited to address the future of human longevity in interaction with healthcare spending. It suggests that healthcare rationing could have severe negative consequences on the future of health and longevity not only for given medical technology but also by suppressing R&D incentives. Finally, medical technological progress that increases longevity could also raise health inequality. It may asymmetrically benefit more educated individuals, those with higher income, and those with better initial health status.

Notes 1 I am grateful to Johannes Sch¨unemann, Holger Strulik, and an anonymous referee for valuable comments and suggestions. I am also indebted to Lisa-Marie Wittler for excellent research assistance and Holger Strulik for providing me the data for Figure 3.3. Financial support of the Swiss National Fund (SNF) for the project “The Socioeconomic Health Gradient and Rising Old-Age Inequality” (grant no. 100018L 15009) is gratefully acknowledged. 2 The focus of this chapter is thus different from static contexts addressing, for instance, the role of co-payment rates in health insurance contracts and price-setting power of pharmaceutical companies for medical innovations (Garber et al., 2006; Lakdawalla and Sood, 2009, 2013; Grossmann, 2013). 3 See, e.g., Searle et al. (2008). Strulik (2023) provides an excellent introduction to the conceptualization of the frailty index as a measure of ageing. 4 Harttgen et al. (2013) document the pattern for 14 European countries and for China, Ghana, India, Mexico, the Russian Federation, and South Africa. 5 Nevertheless, given the comparably high treatment costs in absolute terms, severe rationing measures to limit coverage of such drugs had been in place in many advanced countries (WHO, 2016). 6 Noncapitalized cost does not account for opportunity costs of the use of capital that arises from the time lag between clinical trials and market introduction. 7 DiMasi et al. (2016) find that drugs entering clinical trials have a 12 percent probability of success (POS). Lo et al. (2020) evaluate the POS of clinical trials for 2,544 vaccine and 6,829 nonvaccine programs targeting infectious diseases. They arrive at average POS estimates of 39.6 percent for industry-sponsored vaccine programs and 16.3 percent for industry-sponsored anti-infective therapeutics. 8 Cost information provided by pharmaceutical companies is often upward biased for strategic reasons (e.g., for price negotiations with healthcare providers). Light et al. (2009) discuss the issue for the case of rotavirus vaccines. 9 Profits and mark-ups have been particularly high in the United States and China. The non capitalized Gardasil R&D costs in clinical trials were around US$1.1 billion in 2018 dollars. 10 See, e.g., Arthur (1981), Rosen (1988), Murphy and Topel (2006), and Hall and Jones (2007) for lifetime utility formulations in the context of stochastic survival. 11 We could also assume uℓd < 0 to capture that lower health status raises the disutility from effort provision (Grossmann et al., 2021). 12 Function F may differ before and after statutory retirement age and may depend on time because of social security reforms. We abstract from out-of-pocket health expenditure. This may be restrictive when health insurance does not cover innovative treatments that significantly impact life expectancy. 13 That is, zero-profit insurance companies pay a rate of return above the market interest rate, ¯r , and keep the wealth of the deceased. 14 When Iv,t is pension income, however, it is endogenous to labor supply—a fact that rational households consider (e.g., Grossmann et al., 2021). ˜ v,v+s , using the Lagrangian for the optimization 15 That is, the VSL is the MRS between Nv,v+s and A problem (6). In the presented framework, the definition of the VSL is identical to the MRS between wealth and mortality risk (Murphy and Topel, 2006). 16 Murphy and Topel (2006) derive the VSL expression in continuous time for a single agent. 17 Calibration typically uses estimates of the risk premium for the likelihood of fatal injury in the labor market. Viscusi and Aldy (2003) provide a meta-analysis of the related literature and find a median estimated VSL at birth of US$7 million in the United States (in 2000 US$). 18 Dalgaard and Strulik (2014) focus on a single cohort in continuous time to explain the Preston curve by deliberate health spending patterns. Their approach has been applied in numerous non-innovation

55

Volker Grossmann

19

20 21 22 23

contexts (e.g., Dalgaard and Strulik, 2017; Sch¨unemann et al., 2017a,b; Sch¨unemann et al., 2020; Strulik, 2018a,b; Strulik and Werner, 2021). That quality index q enters Q(L, q) may capture externalities from R&D investment (e.g., Jones and Williams, 2000) or quality depreciation that may follow from mutations of viruses and bacteria (B¨ohm et al., 2021). Quality depreciation may imply that Q(L, q) becomes zero in finite time, avoiding the “end of ageing” (De Grey and Rae, 2007) scenario, where αv ≤ κt · qt . Cohort life expectancy accounts for predicted future declines in mortality rates while period life expectancy employs contemporaneously observed mortality rates. The intertemporal elasticity of substitution measures the optimal substitution of present consumption for future consumption if the interest factor (1 + ¯r ) increases. It is given by 1/γ . TRIPS basically requires World Trade Organization members to secure intellectual property rights of multinational firms. For overviews on the effects of health improvements on economic growth, see Weil (2014) and Bloom et al. (2019, 2021).

References ABELIANSKY, A. L., EREL, D., AND STRULIK, H. (2020): “Ageing in the USA: Similarities and disparities across time and space,” Scientific Reports, 14, 14309. ABELIANSKY, A. L., AND STRULIK, H. (2019): “Long-run improvements in human health: Steady but unequal,” Journal of the Economics of Ageing, 14, 100189. ACEMOGLU, D., FINKELSTEIN, A., AND NOTOWIDIGDO, M. J. (2013): “Income and health spending: Evidence from oil price shocks,” Review of Economics and Statistics, 95(4), 1079–1095. ACEMOGLU, D., AND LINN, J. (2004): “Market size in innovation: Theory and evidence from the pharmaceutical industry,” Quarterly Journal of Economics, 119(3), 1049–1090. AGHION, P., AND HOWITT, P (2005): “Growth with quality-improving innovations: An integrated framework.” In Aghion, P., and Durlauf, S. (eds.): Handbook of Economic Growth (Volume 1A). Amsterdam: North-Holland, 67–110. ARONSON, J. K., AND GREEN, A. R. (2020): “Me-too pharmaceutical products: History, definitions, examples, and relevance to drug shortages and essential medicines lists,” British Journal of Clinical Pharmacology, 86(11), 2114–2122. ARTHUR, W. B. (1981): “The economics of risk to life,” American Economic Review, 71(1), 54–64. BALTAGI, B. H., LAGRAVINESE, R., MOSCONE, F., AND TOSETTI, E. (2017): “Healthcare expenditure and income: A global perspective,” Health Economics, 26(7), 863–874. BASCHET, L., BOURGUIGNON, S., MARQUE, S., DURAND-ZALESKI, I., TEIGER, E., WILQUIN, F., AND LEVESQUE, K. (2016): “Cost-effectiveness of drug-eluting stents versus bare-metal stents in patients undergoing percutaneous coronary intervention,” Open Heart, 3, e000445. BECH, M., CHRISTIANSEN, T., KHOMAN, E., LAURIDSEN, J., AND WEALE, M. (2011): “Ageing and healthcare expenditure in EU-15,” European Journal of Health Economics, 12(5), 469–478. BECKER, G. S. (2007): “Health as human capital: Synthesis and extensions,” Oxford Economic Papers, 59(3), 379–410. BLOOM, D. E., KUHN, M., AND PRETTNER, K. (2019): “Health and economic growth.” In Hamilton, J.H., Dixit, A., Edwards, S., and Judd, K. (eds.): Oxford Encyclopedia of Economics and Finance. Oxford: Oxford University Press, 1–25. BLOOM, D. E., KUHN, M., AND PRETTNER, K. (2021): “Modern infectious diseases: Macroeconomic impacts and policy responses,” Journal of Economic Literature, forthcoming. BLUME-KOHOUT, M. E., AND SOOD, N. (2013): “Market size and innovation: Effects of Medicare Part D on pharmaceutical research and development,” Journal of Public Economics, 97, 327–336. ¨ BOHM , S., GROSSMANN, V., AND STRULIK, H. (2021): “R&D-driven medical progress, healthcare costs, and the future of human longevity,” Journal of the Economics of Ageing, 18, 100286. BOR, J., COHEN, G. H., AND GALEA, S. (2017): “Population health in an era of rising income inequality: USA, 1980–2015,” The Lancet, 389(10077), 1475–1490. BORRE, E. D., HYLE, E. P., PALTIEL, A. D., NEILAN, A. M., SAX, P. E., FREEDBERG, K. A., WEINSTEIN, M. C., AND WALENSKY, R. P. (2017): “The clinical and economic impact of attaining national HIV/AIDS strategy treatment targets in the United States,” Journal of Infectious Diseases, 216(7), 798–807. BREYER, F., LORENZ, N., AND NIEBEL, T. (2015): “Healthcare expenditures and longevity: Is there a Eubie Blake effect?” European Journal of Health Economics, 16(1), 95–112.

56

Medical Innovations and Ageing: A Health Economics Perspective

CERVELLATI, M., AND SUNDE, U. (2005): “Human capital formation, life expectancy, and the process of development,” American Economic Review, 95(5), 1653–1672. CERVELLATI, M., AND SUNDE, U. (2011): “Life expectancy and economic growth: The role of the demographic transition,” Journal of Economic Growth, 16(2), 99–133. CERVELLATI, M., AND SUNDE, U. (2013): “Life expectancy, schooling, and lifetime labor supply: Theory and evidence revisited,” Econometrica, 81(5), 2055–2086. CHANDRA, A., AND SKINNER, J. (2004): “Geography and racial health disparities.” In Anderson, N. B., Bulatao, R. A., and Cohen, B. (eds.): National Research Council (US) Panel on Race, Ethnicity, and Health in Later Life. Washington, DC: National Academies Press, 604–639. CHANDRA, A., AND SKINNER, J. (2012): “Technology growth and expenditure growth in healthcare,” Journal of Economic Literature, 50(3), 645–680. CHANG, V. W., AND LAUDERDALE, D. S. (2009): “Fundamental cause theory, technological innovation, and health disparities: The case of cholesterol in the era of statins,” Journal of Health and Social Behavior, 50(3), 245–260. CHERNEW, M. E., AND NEWHOUSE, J. P. (2011): “Healthcare spending growth.” In Pauly, M. V., Mcguire, T. G., and Barros, P. P. (eds.): Handbook of Health Economics, (Volume 2). Amsterdam: Elsevier, 1–43. CHETTY, R., STEPNER, M., ABRAHAM, S., ET AL. (2016): “The association between income and life expectancy in the United States, 2001–2014,” JAMA, 315(16), 1750–1766. CONTI, G., HECKMAN, J., AND URZUA, S. (2010): “The education-health gradient,” American Economic Review, 100(2), 234–238. CUTLER, D. M., DEATON, A., AND LLERAS-MUNEY, A. (2006): “The determinants of mortality,” Journal of Economic Perspectives, 20(3), 97–120. DALGAARD, C.-J., AND STRULIK, H. (2014): “Optimal ageing and death: Understanding the Preston curve,” Journal of the European Economic Association, 12(3), 672–701. DALGAARD, C.-J., AND STRULIK, H. (2017): “The genesis of the golden age: Accounting for the rise in health and leisure,” Review of Economic Dynamics, 24, 132–151. DE GREY, A., AND RAE, M. (2007): “Ending ageing: The rejuvenation breakthroughs that could reverse human ageing in our lifetime,” New York: St. Martin’s Griffin. DIMASI, J. A., GRABOWSKI, H. G., AND HANSEN, R. W. (2016): “Innovation in the pharmaceutical industry: New estimates of R&D costs,” Journal of Health Economics, 47, 20–33. DUBOIS, P., DE MOUZON, O., SCOTT-MORTON, F., AND SEABRIGHT, P. (2015): “Market size and pharmaceutical innovation,” RAND Journal of Economics, 46(4), 844–871. DUGGAN, M., HEALY, P., AND SCOTT MORTON, F. (2008): “Providing prescription drug coverage to the elderly: America’s experiment with Medicare Part D,” Journal of Economic Perspectives, 22(4), 69–92. EHRLICH, I., AND CHUMA, H. (1990): “A model of the demand for longevity and the value of life extension,” Journal of Political Economy, 98(4), 761–782. FINKELSTEIN, A. (2004): “Static and dynamic effects of health policy: Evidence from the vaccine industry,” Quarterly Journal of Economics, 119(2), 527–564. FINKELSTEIN, A. (2007): “The aggregate effects of health insurance: Evidence from the introduction of Medicare,” Quarterly Journal of Economics, 122(1), 1–37. FINKELSTEIN, A., LUTTMER, E. F. P, AND NOTOWIDIGDO, M. J. (2013): “What good is wealth without health? The effect of health on the marginal utility of consumption,” Journal of the European Economic Association, 11(S1), 221–258. FONSECA, R., MICHAUD, P.-C., GALAMA, T., AND KAPTEYN, A. (2021): “Accounting for the rise of health spending and longevity,” Journal of the European Economic Association, 19(1), 536–579. FORD, E. S., AJANI, U. A., CROFT, J. B., CRITCHLEY, J. A., LABARTHE, D. R., KOTTKE, T. E., GILES, W. H., AND CAPEWELL, S. (2007): “Explaining the decrease in U.S. deaths from coronary disease, 1980–2000,” New England Journal of Medicine, 356(23), 2388–2398. FRANKOVIC, I., AND KUHN, M. (2018): “Health insurance, endogenous medical progress, and health expenditure growth,” ECON WPS, No. 01/2018, Vienna University of Technology, Institute of Mathematical Methods in Economics, Vienna. FRANKOVIC, I., AND KUHN, M. (2019): “Access to healthcare, medical progress and the emergence of the longevity gap: A general equilibrium analysis,” Journal of the Economics of Ageing, 14, 100188. FRANKOVIC, I., KUHN, M., AND WRZACZEK, S. (2020): “Medical innovation and its diffusion: Implications for economic performance and welfare,” Journal of Macroeconomics, 66, 103262.

57

Volker Grossmann

GAGNE, J. J., AND CHOUDHRY, N. K. (2011): “How many ‘me-too’ drugs is too many?” JAMA, 305(7), 711–712. GALAMA, T., LLERAS-MUNEY, A., AND VAN KIPPERSLUIS, H. (2018): “The effect of education on health and mortality: A review of experimental and quasi-experimental evidence.” In Jones, A. M. (ed.): Oxford Encyclopedia of Health Economics. Oxford: Oxford University Press, https://doi.org/10.1093/acrefore/9780190625979.013.7. GALAMA, T., AND VAN KIPPERSLUIS, H. (2019): “A theory of socio-economic disparities in health over the life cycle,” Economic Journal, 129(617), 338–374. GARBER, A. M., JONES, C. I., AND ROMER, P. (2006): “Insurance and incentives for medical innovation,” Forum for Health Economics & Policy, Article 4. GLIED, S., AND LLERAS-MUNEY, A. (2008): “Technological innovation and inequality in health,” Demography, 45(3), 741–761. GOLDMAN, D. P., AND LAKDAWALLA, D. N. (2005): “A theory of health disparities and medical technology,” Contributions to Economic Analysis & Policy, 4(1), Article 8. GROSSMAN, G. M., AND HELPMAN, E. (1991): “Quality ladders in the theory of growth,” Review of Economic Studies, 58(1), 43–61. GROSSMAN, M. (1972): “On the concept of health capital and the demand for health,” Journal of Political Economy, 80(2), 223–255. GROSSMANN, V. (2013): “Do cost-sharing and entry deregulation curb pharmaceutical innovation?” Journal of Health Economics, 32(5), 881–894. ¨ GROSSMANN, V., SCH UNEMANN , J., AND STRULIK, H. (2021): “Fair pension policies with occupationspecific ageing.” CESifo Working Paper No. 9180. GROSSMANN, V., AND STRULIK, H. (2019): “Optimal social insurance and health inequality,” German Economic Review, 20, e913–e948. HALL, R. E., AND JONES, C. I. (2007): “The value of life and the rise in health spending,” Quarterly Journal of Economics, 122(1), 39–72. HARTTGEN, K., KOWAL, P., STRULIK, H., CHATTERJI, S., AND VOLLMER, S. (2013): “Patterns of frailty in older adults,” PLOS ONE, 8(10), e75847. ´ HAVR ANEK , T. (2015): “Measuring intertemporal substitution: The importance of method choices and selective reporting,” Journal of the European Economic Association, 13(6), 1180–1204. JONES, C. I. (2016): “Life and growth,” Journal of Political Economy, 124(2), 539–578. JONES, C. I., AND KLENOW, P. J. (2016): “Beyond GDP? Welfare across countries and time,” American Economic Review, 106(9), 2426–2457. JONES, C. I., AND WILLIAMS, J. C. (2000): “Too much of a good thing? The economics of investment in R&D,” Journal of Economic Growth, 5(1), 65–85. KONTIS, V., BENNETT, J. E., MATHERS, C. D., LI, G., FOREMAN, K., AND EZZATI, M. (2017): “Future life expectancy in 35 industrialised countries: Projections with a Bayesian model ensemble,” The Lancet, 389(10076), 1323–1335. KYLE, M. K., AND MCGAHAN, A. M. (2012): “Investments in pharmaceuticals before and after TRIPS,” Review of Economics and Statistics, 94(4), 1157–1172. LAKDAWALLA, D. N. (2018): “Economics of the pharmaceutical industry,” Journal of Economic Literature, 56(2), 397–449. LAKDAWALLA, D. N., MALANI, A., AND REIF, J. (2017): “The insurance value of medical innovation,” Journal of Public Economics, 145, 94–102. LAKDAWALLA, D. N., AND SOOD, N. (2009): “Innovation and the welfare effects of public drug insurance,” Journal of Public Economics, 93(3-4), 541–548. LAKDAWALLA, D. N., AND SOOD, N. (2013): “Health insurance as a two part pricing contract,” Journal of Public Economics, 102, 1–12. LICHTENBERG, F. R. (2007): “The impact of new drugs on US longevity and medical expenditure, 1990–2003: Evidence from longitudinal, disease-level data,” American Economic Review, 97(2), 438–443. LICHTENBERG, F. R. (2020): “How cost-effective are new cancer drugs in the U.S.?,” Expert Review of Pharmacoeconomics & Outcomes Research, 20(1), 39–55. LICHTENBERG, F. R., AND SUN, S. X. (2007): “The impact of Medicare Part D on prescription drug use by the elderly,” Health Affairs, 26(6), 1735–1744. LIGHT, D. W., ANDRUS, J. K., AND WARBURTON, R. N. (2009): “Estimated research and development costs of rotavirus vaccines,” Vaccine, 27(47), 6627–6633.

58

Medical Innovations and Ageing: A Health Economics Perspective

LO, A. W., SIAH, K. W., AND WONG, C. H. (2020): “Estimating probabilities of success of vaccine and other anti-infective therapeutic development programs,” Harvard Data Science Review, https://doi.org/10.1162/99608f92.e0c150e8. LUCAS, R. E. (1988): “On the mechanics of economic development,” Journal of Monetary Economics, 22(1), 3–42. MITNITSKI, A. B., BAO, L., AND ROCKWOOD, K. (2006): “Going from bad to worse: A stochastic model of transitions in deficit accumulation, in relation to mortality,” Mechanisms of Ageing and Development, 127(5), 490–493. MITNITSKI, A. B., BAO, L., SKOOG, I., AND ROCKWOOD, K. (2007): “A cross-national study of transitions in deficit counts in two birth cohorts: Implications for modeling ageing,” Experimental Gerontology, 42(3), 241–246. MITNITSKI, A. B., MOGILNER, A. J., MACKNIGHT, C., AND ROCKWOOD, K. (2002a): “The accumulation of deficits with age and possible invariants of ageing,” Scientific World Journal, 2, 1816–1822. MITNITSKI, A. B., MOGILNER, A. J., MACKNIGHT, C., AND ROCKWOOD, K. (2002b): “The mortality rate as a function of accumulated deficits in a frailty index,” Mechanisms of Ageing and Development, 123(11), 1457–1460. MITNITSKI, A. B., SONG, X., SKOOG, I., BROE, G. A., COX, J. L., GRUNFELD, E., AND ROCKWOOD, K. (2005): “Relative fitness and frailty of elderly men and women in developed countries and their relationship with mortality,” Journal of the American Geriatrics Society, 53(12), 2184–2189. MORGAN, S., GROOTENDORST, P., LEXCHIN, J., CUNNINGHAM, C., AND GREYSON, D. (2011): “The cost of drug development: A systematic review,” Health Policy, 100(1), 4–17. MURPHY, K. M., AND TOPEL, R. H. (2006): “The value of health and longevity,” Journal of Political Economy, 114(5), 871–904. MURRAY, C. J. L. (2017): “Healthcare access and quality index based on mortality from causes amenable to personal healthcare in 195 countries and territories, 1990–2015: A novel analysis from the Global Burden of Disease Study 2015,” The Lancet, 390(10091), 231–266. NEUMANN, P., COHEN, J. T., AND WEINSTEIN, M. C. (2014): “Updating cost-effectiveness—The curious resilience of the $50,000-per-QALY threshold,” New England Journal of Medicine, 371(9), 796– 797. NEWHOUSE, J. P. (1992): “Medical care costs: How much welfare loss?” Journal of Economic Perspectives, 6(3), 3–21. OECD (ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT). (2019): Health at a Glance. Paris: OECD. OECD (ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT). (2020): “OECD health statistics 2020.” Available at www.oecd.org/els/health-systems/health-data.htm [Accessed November 11, 2020]. OKUNADE, A. A., AND MURTHY, V. N. (2002): “Technology as a major driver of healthcare costs: A cointegration analysis of the Newhouse conjecture,” Journal of Health Economics, 21(1), 147–159. PRASAD, V., AND MAILANKODY, S. (2017): “Research and development spending to bring a single cancer drug to market and revenues after approval,” AMA Internal Medicine, 177(11), 1569–1575. RE´ GNIER, S. (2013): “What is the value of ‘me-too’ drugs?,” Healthcare Management Science, 16, 300–313. ROMER, P. M. (1990): “Endogenous technological change,” Journal of Political Economy, 98(5), S71—S102. ROSEN, S. (1988): “The value of changes in life expectancy,” Journal of Risk and Uncertainty, 1(3), 285–304. SCHNEEWEISS, S., PATRICK, A. R., PEDAN, A., VARASTEH, L., LEVIN, R., LIU, N., AND SHRANK, W. H. (2009): “The effect of Medicare Part D coverage on drug use and cost sharing among seniors without prior drug benefits,” Health Affairs, 28(S1), w305–w316. ¨ SCH UNEMANN , J., STRULIK, H., AND TRIMBORN, T. (2017a): “The gender gap in mortality: How much is explained by behavior?,” Journal of Health Economics, 54, 79–90. ¨ SCH UNEMANN , J., STRULIK, H., AND TRIMBORN, T. (2017b): “Going from bad to worse: Adaptation to poor health, health spending, longevity, and the value of life,” Journal of Economic Behavior & Organization, 140, 130–146. ¨ SCH UNEMANN , J., STRULIK, H., AND TRIMBORN, T. (2020): “The marriage gap: Optimal ageing and death in partnerships,” Review of Economic Dynamics, 36, 158–176. SEARLE, S. D., MITNITSKI, A., GAHBAUER, E. A., GILL, T. M., AND ROCKWOOD, K. (2008): “A standard procedure for creating a frailty index,” BMC Geriatrics, 8(1): 1–10.

59

Volker Grossmann

SKINNER, J., AND STAIGER, D. (2015): “Technology diffusion and productivity growth in healthcare,” Review of Economics and Statistics, 97(5), 951–964. SONGANE, M., AND GROSSMANN, V. (2021): “The patent buyout price for human papilloma virus (HPV) vaccine and the ratio of R&D costs to the patent value,” PLOS ONE, 16(1), e0244722. STRULIK, H. (2015): “A closed-form solution for the health capital model,” Journal of Demographic Economics, 81(3), 301–316. STRULIK, H. (2018a): “The return to education in terms of wealth and health,” Journal of the Economics of Ageing, 12: 1–14. STRULIK, H. (2018b): “Smoking kills: An economic theory of addiction, health deficit accumulation, and longevity,” Journal of Health Economics, 62, 1–12. STRULIK, H (2019): “I shouldn’t eat this donut: Self-control, body weight, and health in a life cycle model,” Journal of the Economics of Ageing, 14, 100175. STRULIK, H. (2023): “Measuring ageing.” In Bloom, D. E., Sousa-Poza, A., and Sunde, U. (eds.): Handbook on the Economics of Ageing. London: Routledge, 455–473. STRULIK, H., AND WERNER, K. (2016): “50 is the new 30—Long-run trends of schooling and retirement explained by human ageing,” Journal of Economic Growth, 21(2), 165–187. STRULIK, H., AND WERNER, K. (2021): “Time-inconsistent health behavior and its impact on ageing and longevity,” Journal of Health Economics, 76, 102440. TANNOCK, I. F., AND HICKMAN, J. A. (2016): “Limits to personalized cancer medicine,” New England Journal of Medicine, 375(13), 1289–1294. TAY-TEO, K., ILBAWI, A., AND HILL, S. R. (2019): “Comparison of sales income and research and development costs for FDA-approved cancer drugs sold by originator drug companies,” JAMA Network Open, 2(1), e186875. VISCUSI, W. K., AND ALDY, J. E. (2003): “The value of a statistical life: A critical review of market estimates throughout the world,” Journal of Risk and Uncertainty, 27(1), 5–76. WEIL, D. N. (2014): “Health and economic growth.” In Aghion, P., and Durlauf, S. N. (eds.). Handbook of Economic Growth (Volume 2). Amsterdam: Elsevier, 623–682. WEISBROD, B. A. (1991): “The healthcare quadrilemma: An essay on technological change, insurance, quality of care, and cost containment,” Journal of Economic Literature, 29(2), 523–552. WHO (World Health Organization). (2016): Global Report on Access to Hepatitis C Treatment: Focus on Overcoming Barriers. Geneva: World Health Organization. WOUTERS, O. J., MCKEE, M., AND LUYTEN, J. (2020): “Estimated research and development investment needed to bring a new medicine to market, 2009–2018,” JAMA, 323(9), 844–853. ZHANG, X., AND NIE, H. (2021): “Public health insurance and pharmaceutical innovation: Evidence from China,” Journal of Development Economics, 148, 102578. ZHUO, Y., HAYASHI, T., CHEN, Q., AGGARWAL, R., HUTIN, Y., AND CHHATWAL, J. (2020): “Estimating the price at which hepatitis C treatment with direct-acting antivirals would be cost-saving in Japan,” Scientific Reports, 10, 4089. ZWEIFEL, P., STEINMANN, L., AND EUGSTER, P. (2005): “The Sisyphus syndrome in health revisited,” International Journal of Healthcare Finance and Economics, 5(2), 127–145.

60

4 MEDICAL PROGRESS, AGEING, AND SUSTAINABILITY OF HEALTHCARE FINANCE Michael Kuhn

Abstract This chapter explores the relationship among medical progress, ageing, and healthcare spending and summarizes what is known about the joint dynamics and their drivers and about the consequences for the sustainability of healthcare finance, economic performance, and ultimately welfare. Evidence from diverse strands of empirical research and theoretical analysis reveals strong complementarity and joint causation among medical progress, ageing, and healthcare spending and a strong influence of income growth and institutional structure on the dynamic process. The findings support a view that, despite manifold inefficiency involved in the process, the simultaneous growth of longevity and healthcare spending can be viewed as a welfare improvement. Further work is needed to explore the role of medical progress in exacerbating inequality in health outcomes, the role of healthcare institutions as important mediators, a potentially stabilizing role of income on healthcare spending growth, and what new forms of medical advances based on digitalization imply for healthcare spending and the ageing process.

4.1

Introduction

The political and academic discussion on the sustainability of healthcare spending has evolved around a triad of potential drivers: ageing, medical progress, and the inefficiency in healthcare spending. In this chapter, I focus on the first two drivers. With respect to inefficiency, I refer readers to an extant literature on inefficiency in providing and financing healthcare (e.g., Baicker et al., 2012; Cutler, 2018; Chandra and Staiger, 2020). The focus on ageing and medical progress is justified as the two forces are systematically and dynamically linked, whereas many concerns about inefficiency relate to standalone aspects of healthcare provision. Nonetheless, the chapter points out issues with efficiency where they are relevant for the relationship among ageing, medical progress, and healthcare spending.

DOI: 10.4324/9781003150398-5

61

Michael Kuhn

Even a casual inspection of time series data shows that for practically all high-income countries and many mid-income countries, life expectancy (and other measures of health) and healthcare spending per capita or as a share of a country’s gross domestic product (GDP) have increased in tandem over much of the second half of the 20th century and over the first decade of the 21st century (see, e.g., the data on life expectancy at birth and on total healthcare expenditures per unit GDP in OECD, 2019).1 Measures of medical progress, such as the medical patent share (as documented, e.g., in Jones, 2016) or the number of new U.S. Food and Drug Administration (FDA) approvals to the pharmaceutical market (as documented, e.g., in Acemoglu and Linn, 2004) show a similar upward trend over time. This naturally begs questions about causality, and, indeed, much of the empirical literature is motivated by identifying causal effects and pathways within what I will subsequently refer to as the ageing–medical progress–healthcare spending (AMH) nexus.2 Strikingly, much of the research is organized along the sides of the AMH nexus (or triangle), regressing healthcare expenditure, longevity or medical progress on a set of explanatory variables that typically includes the two other AMH components. Although much of this literature recognizes and seeks to address the issue of joint causality, the three strands of research on A, M, and H have, to a large extent, been developed in isolation with only recent studies probing into the joint causality in a deeper way. To some extent this is surprising, as a straightforward narrative would suggest codeterminedness of AMH. Medical advances—not all perhaps, but at least some—increase the effectiveness of healthcare treatments. With medical advances thus embodied in treatments (physically as in pharmaceuticals or medical devices, but also in terms of medical skills and knowledge and in organizational structures), individuals or medical practitioners on their behalf will be more inclined to order such treatments. While this does not necessarily imply an increase in healthcare spending per se, such a result is easily conceivable and well documented (see Section 4.2). The consumption of more and more effective healthcare is then prone to increase longevity—not as an exclusive or even primary driver but as one that has been shown to be relevant (see Section 4.3). With individuals consuming more healthcare in old age, this in turn tends to further boost spending (see the next section). Finally, to the extent that research and development (R&D) into medical advances is profit-driven—and again the evidence supports this—increases in healthcare spending spur innovations (see Section 4.5). This closes the loop and suggests that, while understanding and quantifying the pathways within the nexus is important, establishing hard and fast patterns of causality per se may well be a futile exercise. Indeed, this survey suggests as much. The chapter proceeds as follows. It sets out by focusing on partial approaches aimed at assessing the determinants of healthcare spending, longevity increases, and medical progress in isolation before turning toward more integrated approaches. Thus, the next section summarizes evidence on the determinants of healthcare spending (growth), with ageing and medical progress as key explanatory variables besides income growth and welfare state institutions. The following section reverses the roles and discusses analyses that treat longevity increases as an endogenous variable. Section 4.4 surveys approaches toward assessing the value of medical advances before the following section deals with the determinants of medical progress itself. Then Section 4.6 presents a set of macroeconomic models that integrate components of the AMH nexus to arrive at a deeper understanding of the underlying mechanics and incentives, to assess the welfare implications, and to derive policy recommendations. Again, we follow the AMH nexus by starting out with models addressing the relationship between healthcare spending and ageing, then considering the role of exogenous medical progress before turning to models that determine endogenously the full set of AMH relationships. The conclusion summarizes key insights and identifies future research needs. 62

Medical Progress, Ageing, and Sustainability of Healthcare Finance

4.2

Ageing and Medical Progress as Drivers of Health Expenditure Growth

In the following, I review some of the empirical evidence on the impact of ageing and medical progress on healthcare spending growth. In so doing, I include income growth and welfare state structure, in particular health insurance and social security, as two additional contributors. As we will see, these play an important background and translational role.

4.2.1

Ageing: A Mixed Bag

The public debate on the consequences of population ageing commonly includes a concern that an increasing population share of the elderly, coupled with high healthcare spending in old age, may cause healthcare expenditure to grow to unsustainable levels. As it turns out, the evidence supports this concern to a limited extent only.3 Building on individual-level data on healthcare spending over the life cycle up to and including the point of death, an extant literature shows that, after controlling for the time to death, age does not contribute significantly to the explanation of healthcare spending. It should, thus, be considered a “red herring,” i.e., a misleading clue (see, e.g., Zweifel et al., 1999; Seshamani and Gray, 2004; Breyer and Felder, 2006; Werblow et al., 2007 for early contributions; and Breyer et al., 2010, and Breyer and Lorenz, 2020, for surveys). The argument goes that if high healthcare expenditures in old age are primarily associated with the cost of dying (i.e., the high cost of intensive life-saving efforts), population ageing should either be neutral to aggregate healthcare spending (as people are dying anyway) or even result in lower period spending levels when deaths and the associated cost continue to be deferred into the future. This is because an ongoing increase in longevity would imply that fewer people are dying within each age group as time progresses, which should translate into reduced healthcare spending. This argument has come under scrutiny because ageing effects tend to re-emerge at the aggregate population level (e.g., Breyer and Felder, 2006; Dormont et al., 2006; Shang and Goldman, 2008; de Meijer et al., 2013; Breyer et al., 2015; European Commission, 2018) even if these are weaker than they would be without controlling for the time to death. Breyer and Lorenz (2020) provide an insightful discussion into why the results of cross-sectional individuallevel data studies cannot readily be translated into aggregate expenditure dynamics.4 One key reason is that, even if large at the level of individual decedents, the costs of dying constitute a share of only 8.5–11.2 percent of all healthcare expenditures and are dominated by chronic care costs (French et al., 2017). This is in line with findings that time to death serves as a proxy for morbidity and loses significance when the latter also is directly controlled for (Dormont et al., 2006; Howdon and Rice, 2018),5 which implies that the age structure of chronic care costs and their development is what really matters (Goldman et al., 2005; de Meijer et al., 2013; Wouterse et al., 2015; European Commission, 2018). Projections vary considerably between unhealthy ageing (e.g., Lakdawalla et al., 2004; Olshansky et al., 2005; Reither et al., 2011; Pandya et al., 2013), with considerable impact on public budgets (Goldman et al., 2010), and healthy ageing (Manton et al., 2006; Cutler et al., 2014). The literature also points at large heterogeneity within cohorts (Lowsky et al., 2014) and across countries (Michaud et al., 2011). Furthermore, much of the analysis takes treatment patterns as given. However, Breyer et al. (2015) show that both time to death and remaining life expectancy have a significant positive effect on healthcare expenditure. They provide the plausible interpretation of the latter effect that more intensive treatment is provided to individuals with a higher remaining life expectancy.6 Fang et al. (2007) provide similar evidence that individual health investments increase with 63

Michael Kuhn

life expectancy.7 Such behavior is entirely consistent with the dependency of the value of life and health, more generally, on remaining life expectancy (Murphy and Topel, 2006; Aldy and Viscusi, 2008). The notion that increases in life expectancy drive an increase in health expenditure is consistent with the finding by Dormont et al. (2006) that, while the direct effect of ageing on healthcare spending growth in France is moderate, “change in medical practice” can explain much of it. While they cannot test the role of ageing explicitly, the data reveal that expenditure increases in secondary care and for pharmaceuticals arise particularly among the higher age groups. These observations are corroborated by evidence that a trend toward more healthcare utilization in the Netherlands is particularly pronounced for ages 65+ and that, in case of secondary care, advances in medical technology reinforce this trend (Wong et al., 2012). Roham et al. (2014) provide similar evidence of spending growth in Ontario occurring particularly among the higher age groups and for technology-intensive care. This hints at an important role for the interaction of ageing and medical technology (de Meijer et al., 2013).

4.2.2

Medical Progress, Income Growth, and Welfare State Institutions: Will the Culprit behind Spending Growth Please Stand Up?

Even before ageing came into focus, a line of inquiry sought to identify from time series and later from panel regressions the drivers of healthcare spending growth (see Chernew and Newhouse, 2011, for a survey). Particular focus lies on medical progress, income (growth), and institutional features, especially health insurance and social security.

Medical progress: Since Newhouse’s (1992) contribution, which identified the impact of medical progress by the residual of a time series regression of various observable factors on medical expenditure, the literature has advanced to include explicit measures of medical technology (Okunade and Murthy, 2002; Smith et al., 2009; Murthy and Ketenci, 2017; You and Okunade, 2017). The results affirm the original finding that medical progress contributes a large share to spending growth. Smith et al. (2009), for instance, find that “medical technology explains about 27–48% of spending growth since 1960.” This is in line with the conjectures by Dormont et al. (2006) and Breyer and Felder (2006) that medical progress dominates “pure” ageing as a driver of healthcare expenditure. Income growth: Research in health economics has long sought to identify the income elasticity of healthcare spending. In a survey, Getzen (2000) finds a notable discrepancy between the spending elasticities estimated from individual-level data, which typically range between 0 and 0.7, and macroeconomic elasticities, which range from 0.5 to 1.5, leading him to conclude that healthcare is an individual “necessity” but a national “luxury.” More recent analysis confirms the earlier findings (Acemoglu et al., 2013; Baltagi et al., 2017; Fonseca et al., 2021). The striking difference between micro and macro elasticities is largely due to the macro elasticities typically including (unobserved) adjustments and price changes within the healthcare system.8 Recent analysis by Dranove et al. (2014) corroborates the important role of income as a driver by showing that the slowdown in U.S. healthcare spending growth since 2007 is largely explained by the economic downturn following the financial crisis. Health insurance and social security: Following the estimation of a low insurance elasticity of healthcare spending (−0.2) during the RAND health insurance experiment (Manning et al., 1987), an earlier view on a substantive role of health insurance in explaining healthcare 64

Medical Progress, Ageing, and Sustainability of Healthcare Finance

spending (Feldstein, 1977) has given way to a widespread understanding that health insurance is not a major driver of healthcare cost expansion. Recent findings by Fonseca et al. (2021) have now challenged this view. Estimating the contributors to U.S. healthcare spending growth between 1965 and 2000 from a structural model, they find a sizeable insurance elasticity between −0.61 and −0.74. Social security has received little attention as a driver of healthcare. Zhao (2014) points out, however, that social security schemes typically shift income to the older population, which has a stronger propensity to spend on healthcare. His structural model assigns a third of U.S. spending growth to the expansion of social security. The literature discussed so far has three shortcomings: (1) It neglects the fact that medical progress is likely to be a driver of ageing. (2) By broadly focusing on expenditure and, thus, on the social cost of ageing and medical progress, the literature fails to address the offsetting benefits. (3) The fact that healthcare spending growth and population ageing may drive medical progress itself is disregarded. We address these issues in the following subsections.

4.3

Healthcare and Medical Progress as Drivers of Longevity

So far, the discussion has been mostly about the impact of ageing and medical progress on the cost of healthcare. This begs the question as to whether sufficient benefits from health improvements and resulting increases in longevity offset these costs. Surprisingly little evidence is available on the extent to which more effective healthcare has translated into health improvements at the population level. Many studies seek to establish the impact of healthcare spending on mortality or survival, both adult and infant, or on life expectancy (e.g., Cr´emieux et al., 1999; Filmer and Pritchett, 1999; Lichtenberg, 2004; Zweifel et al., 2005; Bokhari et al., 2007; Hall and Jones, 2007; Baltagi et al., 2012; Moreno-Serra and Smith, 2015; Nakamura et al., 2020, the latter also discusses econometric pitfalls and the need for instrumental variable approaches) and find positive but modest effects. Increases in these effects over time should capture medical progress. Absent any direct evidence on this, Gallet and Doucouliagos (2017) find from meta-regressions that the elasticity of mortality with respect to health spending becomes stronger over time, which suggests medical progress. Noting that aggregate healthcare spending covers many conditions and treatments that involve morbidity but not mortality, the modest elasticities found for aggregate spending on aggregate measures of survival do not indicate poor effectiveness.9 For this reason, studying medical advances against specific diseases seems more appropriate. Cutler et al. (2006) and Catillon et al. (2018) point out that improvements in the treatment of infectious diseases and of cardiac disease had significant impacts on survival during the first and second halves, respectively, of the 20th century. More recently, treatment of cancer entered the stage (Cutler, 2008). We will deal with these developments in turn. Acemoglu and Johnson (2007) analyze extensively how the international epidemiological transition—strictly speaking its third stage (Omran, 1998)—around the middle of the 20th century bore on life expectancy and ultimately on economic growth.10 They argue that this stage was characterized by the invention and widespread diffusion of antibiotics as treatments against infectious diseases, such as tuberculosis, pneumonia, and cholera, but also by organizational innovation, such as the foundation of the World Health Organization, and innovation in controlling the epidemiological environment, namely through the use of DDT to control the vector-borne transmission of malaria. Using data on the arrival time of diseasespecific interventions, Acemoglu and Johnson (2007) estimate the effect of the availability of these interventions on disease-specific mortality and find a significant negative effect. Subsequently, they estimate the impact of predicted mortality on life expectancy. They find that 65

Michael Kuhn

almost two-thirds of the average 17-year increase in life expectancy between 1940 and 1980 across a set of globally selected countries (excluding Africa, Eastern Europe, and Russia) are explained by predicted mortality and, thus, by technological and organizational advances in medicine. Turning to the cardiac revolution, OECD (2018) data on the causes of death show that around 65 percent of the overall mortality decline in the United States over the time span 1980–2000 can be attributed to lower mortality from coronary heart disease. Considering that Ford et al. (2007) attribute around 50 percent of the U.S. mortality decline in coronary heart disease over that time span to changes in medical treatment, this suggests that cardiac treatments contributed more than 30 percent to the 1980–2000 decline in overall mortality. In a simulation calibrated to these data, Frankovic et al. (2020a) estimate that medical progress, as measured by an increase in the healthcare spending elasticity on mortality, explains about 23 percent of the increase in life expectancy (at birth) from around 75 years to around 79 years between 1980 and 2005, with the bulk of this increase accruing to the population aged 65+. Recent studies show that similar advances against mortality can be expected in the context of cancer treatment. Cutler (2008) estimates that 35 percent of the reduction in cancer mortality resulted from improved screening (especially colonoscopies) and 20 percent from improved therapies, in particular pharmaceuticals. More recent evidence corroborates this, which also shows, however, considerable variation depending on the form of cancer (Stevens et al., 2015; Li et al., 2020; MacEwan et al., 2020). Further support for the role of medical progress comes from evidence that focuses on treatments. Here, Lichtenberg (2007, 2009, 2013, 2018, 2019) and Cerda (2007) document health improvements that arise from pharmaceutical and therapeutic innovation spanning various country settings and time periods around the turn of the millennium. Considering the overall contribution of medical progress to the increase in life expectancy, Cutler et al. (2006b) attribute about half of the increase in survival in the United States between 1960 and 2000 to medical progress with the rest being down to lifestyle changes, in particular a reduction in smoking. This is well in line with the structural estimates by Fonseca et al. (2021) who find that medical technology improvements explain about 47 percent of the increase in U.S. life expectancy between 1965 and 2005. Recent analysis by Buxbaum et al. (2020) shows that the contribution of medicine continues to be strong, with pharmaceuticals explaining 35 percent of the increase and other medical care explaining 13 percent of the 3.3-year increase in life expectancy between 1990 and 2015. I conclude this section on two notes of caution. First, despite the evidence that medical care and advances in treatment play a role in increasing life expectancy, it is less clear whether the life years gained are healthy (or active). Goldman et al. (2005) consider the impact of a set of innovative life-expanding technologies, including some hypothetical, and illustrate by way of microsimulation the large extent to which the impact on healthcare spending, including survival-induced follow-up spending, depends on whether or not these technologies contribute to reducing disability. Based on a similar notion, Goldman et al. (2013) illustrate the advantage of innovations that delay ageing and lead to gains in healthy life years in excess of those afforded by isolated improvements against cardiovascular or cancer mortality. Second, evidence is mounting that the process of steady increases in longevity has stalled in many countries over the decade 2010–2020 and even reversed, in some countries and contexts (OECD, 2019). The reasons for this remain subject to investigation, but they appear multifaceted and related to a blend of deteriorating socioeconomic circumstances for certain societal groups and slowing medical advances (Raleigh, 2019).11 In particular, slackening progress against cardiovascular disease (Mehta et al., 2020) suggests declining returns to medical progress either 66

Medical Progress, Ageing, and Sustainability of Healthcare Finance

in terms of innovation per se or in terms of its diffusion to disadvantaged groups to an extent that is sufficient to offset deteriorating circumstances or health behaviors (see also the discussion on inequality in access to medical progress in Section 4.6.2).

4.4

What Is It Worth? Value of Medical Progress

The (unsurprising) finding that medical progress generates both costs (higher healthcare spending) and benefits (increase in longevity) begs the more important question of whether the latter outweigh the former. To arrive at an assessment, researchers typically draw on the value of a statistical life (year) concept (see Kniesner and Viscusi, 2019, for a recent overview), from which they can derive a value of medical care or a value of medical innovations for that matter. Analyzing the innovation toward revascularization as a treatment for heart attacks, Cutler and McClellan (2001), Cutler and Huckman (2003), and Cutler (2007) find that its value outweighs the cost by a large margin even when accounting for follow-up costs of (unhealthy) survival. Applying similar analysis to a range of innovations in other fields of medicine (treatment of low birth weight, cataract, depression, breast cancer), Cutler and McClellan (2001) find that treatment innovations typically create benefits that exceed their cost, with a zero net effect only for breast cancer. Skinner et al. (2006) and Chandra and Skinner (2012) suggest a more cautionary take on assessing the value of medical innovation. Their analysis shows that whether the adoption of medical innovation creates a net benefit depends both on the type of technology and on how it is deployed, which in turn depends on the organization of the healthcare system that adopts it. Here, Chandra and Skinner (2012) show many treatment “innovations” to be ultimately wasteful. Furthermore, a case exists for viewing innovation as a process involving different domains. Thus, while revascularization per se may have generated a net benefit in the treatment of heart attacks, the adequate application of antihypertensive drugs and beta blockers strongly leverages these benefits and may be much more cost-effective (Cutler et al., 2007; Rosen et al., 2007; Skinner and Staiger, 2015). While the previous literature assigns (exogenous) monetary values to the life years gained, Murphy and Topel (2006) calculate the willingness to pay for medical innovations based on a rigorous derivation of the value of a statistical life. Specifically, they estimate that the cumulative gain in life expectancy in 2007 was worth US$3.2 trillion per year, while a 1 percent future reduction in cancer mortality would be worth US$500 billion. Murphy and Topel (2006) also demonstrate the importance of complementarity across different dimensions of medical innovation. By enhancing future survival prospects, an improvement against cardiac mortality, for instance, raises the willingness to pay for innovations toward lowering cancer mortality as a competing risk. The argument extends to innovations against diseases that reduce the quality of life, e.g., Alzheimer’s disease: While improvements in the quality of life during old age raise the value of surviving to these ages, mortality reductions raise the value to pay for a better quality of life. Consequently, the sum of the values of a set of disease-specific medical innovations is likely to underestimate the value of the same innovations when considered across the board. In a similar vein, Goldman (2016) considers the value of delayed ageing, in the sense of progress against multiple mortality risks, and calculates a potential value of US$7.1 trillion of treatments that delay ageing. Several recent contributions assess the welfare impact of medical progress within a dynamic macroeconomic context. Based on a structural estimation of life-cycle preferences and medical technology, Fonseca et al. (2021) compare the life-cycle utility the 1940 cohort would 67

Michael Kuhn

receive if it was born into an economy with the factual medical progress of 1965–2005 with the utility the same cohort would receive in a counterfactual setting without medical progress. They find that individuals would be willing to forego around 8 percent of their consumption to benefit from medical progress. Modeling the development of the U.S. economy and healthcare sector over the period 1980–2005, Frankovic et al. (2020a) calculate the compensating variation, i.e., the increase in consumption that a birth cohort would require to be moved from a (counterfactual) economy without medical progress to the benchmark economy. Considering birth cohorts between 1910 and 1980 they find the compensating variation to increase with the birth year and to lie around 4 percent, 10 percent, and just below 40 percent for the 1930, 1940, and 1970 cohorts, respectively. In a similar analysis, but projecting lifesaving medical progress into the future, B¨ohm et al. (2021) confirm an unambiguously positive welfare impact.

4.5

Endogenous Medical Progress: Empirical Evidence

Turning to the final leg of the triangular relationship among health expenditure, ageing, and medical progress, let us consider healthcare expenditure and ageing as drivers of medical progress. Weisbrod (1991) points out that in an economy where private corporations carry out much of medical R&D, insurance-induced increases in healthcare spending should trigger incentives toward developing new treatments that, in turn, lead to further spending growth. A considerable body of empirical research has developed around Weisbrod (1991)’s conjecture. Following Acemoglu and Linn (2004), much of the literature studies the impact of market size on medical R&D, with some approaches employing the introduction of Medicare and other variations of insurance coverage as an instrument for market size and others considering age-structure effects. Finkelstein (2007) finds that the 1965 introduction of Medicare explains a spending increase more than six times larger than that suggested by the RAND health insurance experiment once macroeconomic responses, such as induced entry into the hospital market, are accounted for. She also provides evidence suggesting that new (cardiac) technologies were adopted following the introduction of Medicare. Clemens and Olsen (2021) estimate that the introduction of Medicare has spurred a 20–30 percent increase in medical patenting across U.S. states. Considering the impact of the introduction of Medicare on pharmaceutical innovation, Acemoglu et al. (2006) find little impact.12 This is no contradiction, however, as Medicaid only covered prescription drugs to a very limited extent before the introduction of Medicare Part D in 2006. Indeed, Blume-Kohout and Sood (2013) show that the 2006 reform boosted pharmaceutical R&D. Acemoglu and Linn (2004), Cerda (2007), and Dubois et al. (2015) employ demographic change as an instrument for changes in the size of (demographically delineated) markets.13 While all studies find significant market size effects, Acemoglu and Linn (2004) also demonstrate how pharmaceutical innovation in the United States, as measured by FDA approvals of new drugs, followed the ageing of the baby boom cohorts over the time frame 1970–2000. Strikingly, the finding that R&D activity tends to follow age-driven changes in disease prevalence extends to publicly funded scientific research (Bhattacharya and Packalen, 2011).

4.6

Putting Things Together: Integrated Macroeconomic Modeling of Healthcare

The empirical literature reviewed so far typically focuses on subparts of the AMH nexus. While insightful, such analyses allow only for a partial understanding of the mechanisms underlying 68

Medical Progress, Ageing, and Sustainability of Healthcare Finance

the nexus of dynamics and feedbacks and of the resulting economic, demographic, and ultimately welfare outcomes. These limitations may stand in the way of identifying inefficiency and assessing the scope for policymaking. Overcoming them requires an integrated macroeconomic analysis, efforts at which have been undertaken only recently. From early analyses studying the macroeconomic causes and consequences of increased healthcare spending (e.g., Van Zon and Muysken, 2001; Sanso and A´ısa, 2006; Hall and Jones, 2007), a literature has emerged that addresses in some detail the impact of ageing and medical progress on the macro-economy, focusing on healthcare spending and the role of the healthcare sector.14 In discussing it, we again follow the AMH triangle by first considering macroeconomic models that address the joint relationship between healthcare spending and ageing, then considering the role of exogenous medical progress within such models, and finally considering models in which all three factors—healthcare spending, ageing, and medical progress—are determined endogenously.

4.6.1

Ageing in Macroeconomic Models with a Healthcare Sector

Kuhn and Prettner (2016), Schneider and Winkler (2021), and Jung et al. (2017) explore the relationship between healthcare spending and ageing together with its macroeconomic impacts. In so doing, they take different perspectives. In Kuhn and Prettner (2016) and Schneider and Winkler (2021), healthcare spending lowers the mortality rate and thus acts, at least partially, as a driver of ageing. In contrast, Jung et al. (2017) explore the impact of exogenous changes in survival on the incentives to spend on healthcare with a view to remaining in good health over the prolonged life course. Despite their different starting points, the models share several important features, most notably the finding that the increase in the healthcare spending share that comes with an ageing population does not usually compromise GDP growth, a feature that is shared by the models with medical progress. This finding is striking, as it typically holds even when ageing comes with an increase in the old-age dependency ratio. Two channels are important here and represented in Kuhn and Prettner (2016), Schneider and Winkler (2021), and Jung et al. (2017). First, to the extent that health improvements raise productivity or allow the increase of labor participation at the extensive or intensive margins, this mitigates the tendency toward a higher dependency ratio due to increases in longevity.15 Second, an increase in old-age savings (e.g., Bloom et al., 2007) that is leveraged to the extent that individuals need to cover high healthcare expenditures in old age (De Nardi et al., 2010; Kopecky and Koreshkova, 2014) leads to a strong expansion in the capital stock. A reduction in the interest rate mirrors this effect, as Aksoy et al. (2019) show. While the expansion of the capital stock “levels up” GDP in Jung et al. (2017),16 the reduction in the interest rate raises the profitability of a (nonmedical) Romer (1990)–style R&D sector and thereby tends to raise the balanced growth rate in Kuhn and Prettner (2016). Kuhn and Prettner (2016) find that an excessive expansion of the healthcare sector will ultimately absorb labor from R&D activities, implying a growth-maximizing volume of healthcare.17 However, the willingness to pay for survival and its increase over time along with income growth (Hall and Jones, 2007) implies that an extension of healthcare beyond its growthmaximizing level constitutes a Pareto optimum. Indeed, building on Hall and Jones (2007), Chen et al. (2021) show that the healthcare spending shares in GDP in the years 2010 and 2015 were too low from a welfare perspective for a range of high- and middle-income countries, with the notable exception of the United States. To the extent that population ageing—embracing longevity expansions and fertility decline—depresses the support ratio, it has fiscal implications for the funding of healthcare 69

Michael Kuhn

that parallel those for the funding of pensions. In particular, the sustainability of pay-as-you-go funding that dominates public healthcare systems would then require raising contribution rates or cutting benefits. However, the policy problem is more complex in as far as health insurance primarily seeks to pool cross-sectional health risks but is then confounded by the longitudinal longevity risk. One suggested set of solutions builds on introducing tax-exempted health savings accounts that, by relying on the build-up of an individual capital stock, mimic funded pensions (Pauly et al., 1995). Such schemes have gained some prominence in several Eastern Asian countries and in the United States, and they lie at the heart of introducing ageing provisions within private health insurance contracts (e.g., Meier, 2005; Arentz et al., 2012; Atal et al., 2019) and public health insurance schemes (e.g., Hurley et al., 2008; Atal et al., 2019, 2020). According to some evidence, current and projected schemes induce individual savings among the more wealthy (Minicozzi, 2006) and reduce public expenditure (Hurley et al., 2008), but they have been criticized for inadequately pooling the highly skewed cross-cohort health (expenditure) risks and thus generating insufficient coverage and negative distributional consequences (Deber et al., 2004; Hurley et al., 2008; Wouters et al., 2016).18 Macroeconomic modeling by Jung and Tran (2016) also suggests that while healthcare spending can be contained by medical savings accounts, the tax breaks required for their implementation may lead to large fiscal losses.

4.6.2

Medical Progress in Macroeconomic Models

The assumptions of an exogenous increase in survival rates, as in Jung et al. (2017), or an increase driven by healthcare spending, as in Kuhn and Prettner (2016), leave no role for medical progress so far. Suen (2005) models medical progress as a decline in the price for medical care, which is equivalent to Kelly’s (2017) notion of medical progress as an increase in the total factor productivity within a distinct healthcare sector. While these models generate plausible healthcare expenditure growth and longevity increase, the assumptions of a declining unit price of medical care or, equivalently, extensive productivity growth in the healthcare sector sit uneasily against the observable medical price inflation19 and the absence of productivity growth in the data (F¨are et al., 1997; Okunade and Osmani, 2018).20 Importantly, a distinction can be made between process innovation, allowing the production of treatments of a given effectiveness at lower cost, and product innovation, allowing the production of treatments of greater effectiveness at a given cost. Frankovic et al. (2020b) consider a two-sector general equilibrium model with overlapping generations of households, in which a healthcare sector produces treatments, the effectiveness of which in lowering mortality depends on the state of medicine. They show that only product innovation can consistently explain the simultaneous increase in the volume-related price of healthcare (i.e., medical price inflation) and decline in the quality-adjusted price of healthcare observed in the data (e.g., Cutler et al., 1998; Lakdawalla et al., 2015; Hult et al., 2018).21 22 Calibrating the same model to reflect the development of the U.S. economy and demography over the time span 1980–2005, Frankovic et al. (2020a) trace the role of medical progress in the ageing process over time.23 They find that while medical progress explains about 23 percent of the increase in life expectancy, it explains 35 percent of the increase in per capita healthcare spending. Most of the spending increase is due to an expansion of demand, which is in line both with micro evidence (Cutler and Huckman, 2003; Wong et al., 2012; Roham et al., 2014) and macro evidence (Baker et al., 2003; Bundorf et al., 2009). With both the gains in life expectancy and the increase in per capita spending clustering in the elderly population, this leads to a considerable boost to Medicare spending and an according boost to the (implicit) Medicare tax rate of which 41 percent is due 70

Medical Progress, Ageing, and Sustainability of Healthcare Finance

to medical progress. This finding mirrors the strong fiscal implications of ageing found in Jung et al. (2017). Fonseca et al. (2021) model medical progress as a reduction in the marginal cost of improving health, reflecting greater effectiveness of healthcare. Estimating and simulating a partial equilibrium structural life-cycle model, they compare healthcare expenditure and life expectancy between 1965 and 2005, identifying the individual and collective contributions of income growth, medical progress, and insurance expansion by simulating counterfactual allocations in which relevant factors are held constant at the 1965 level. They attribute 47 percent of the increase in life expectancy (at age 25) to medical progress, with insurance and income explaining only 19 percent and 33 percent, respectively. Conversely, 37 percent of the spending growth is explained by the expansion of health insurance, 29 percent by income growth, and only 9 percent by medical progress. The residual gap of 25 percent arises from complementarity across the three factors, implying their mutual reinforcement. In Fonseca et al. (2021), complementarity is particularly pronounced between income and insurance. Frankovic and Kuhn (2018) identify strong additional complementarity between income and medical progress. This difference can be attributed to the way medical progress is modeled, which in Fonseca et al. (2021) leave little room for the demand expansion that is an important feature of the model in Frankovic and Kuhn (2018) and acts as a lever complementarity. Kelly and Kuhn (2022) consider the macroeconomic impact of an increase in medical effectiveness within a public healthcare system that exhibits congestion and a waiting time mechanism for the control of demand. In such a setting medical progress leads to a substantial increase in longevity if and only if healthcare capacity is increased simultaneously. Otherwise, the generation of additional demand for more effective treatments is merely boosting congestion, which lowers the quality, intensity, or timeliness of treatments and thereby offsets most of the gains from medical progress. Section 4.4 has already outlined that medical progress is found to improve welfare in the analyses by Fonseca et al. (2021) and Frankovic et al. (2020a), featuring exogenous medical progress, and by Frankovic and Kuhn (2018) and B¨ohm et al. (2021), featuring endogenous medical progress. One important driver is income growth, which raises the willingness to pay for longevity (Hall and Jones, 2007) and thus leverages the willingness to spend on increasingly more effective healthcare as time progresses. This sustains medical progress even if it is subject to decreasing returns. Worth stressing, however, is that the population does not equally share in the gains to medical progress. Relating evidence on a growing income gap in life expectancy (e.g., Chetty et al., 2016) to evidence that the socially and economically disadvantaged have poorer access to stateof-the-art healthcare (Glied and Lleras-Muney, 2008; Bago d’Uva et al., 2011; Vallejo-Torres and Morris, 2013; Fiva et al., 2014) or find it more difficult to utilize it productively (Avitabile et al., 2011; Lange, 2011; Hernandez et al., 2018), Frankovic and Kuhn (2019) consider the differential impact of medical progress in an overlapping generations economy with heterogeneous households. Specifically, they consider several pathways of socioeconomic disadvantage: a widening skill-driven earnings gap, a skill bias in the effective utilization of healthcare, and differential health insurance coverage. Replicating the 2.1-year increase in the gap in life expectancy between the high-skilled/income and low-skilled/income groups between 1960 and 2015, the model attributes 19 percent of this increase to skill-biased earnings growth, 57 percent to skill bias in the effective utilization of healthcare, and 5 percent to differential health insurance coverage. They also show that initial earnings-related differences in utilization are strongly leveraged by medical progress, which highlights once more the role of complementarity.24 Overall, their findings suggest that, while yielding large benefits on average, medical progress is also a strong contributor to inequality in life expectancy and acts as a strong lever to differential ageing. 71

Michael Kuhn

4.6.3

Macroeconomic Models with Endogenous Medical Progress

Mirroring the empirical research, treating medical progress as “manna from heaven” runs the risk of misattributing causality and obfuscating important channels of transmission and institutional influence. Frankovic and Kuhn (2023) and B¨ohm et al. (2021) address these concerns by including a medical R&D sector into general equilibrium economies with overlapping generations. In so doing, the two studies focus on different aspects of endogenous medical progress. Frankovic and Kuhn (2023) pick up on the empirical discussion concerning the extent to which health insurance expansion has led to greater rates of medical innovation (see Section 4.5). Treating insurance as a subsidy on the private demand for healthcare, they study the extent to which an insurance-driven expansion in demand stimulates (additional) medical innovation. Calibrating their model to reflect the expansion in private and public health insurance coverage in the United States between 1965, the year in which Medicaid was introduced, and 2005, they find that insurance expansion has raised the rate of medical progress by about 39 percent between 1965 and 1990, which is well in line with the estimates in Clemens and Olsen (2021). Their analysis documents a highly ambivalent welfare impact of the health insurance expansion, which explains about 63 percent of the increase in healthcare spending since 1965, mostly due to expanding utilization. For a fixed level of technology, the insurance-induced expansion of demand would have added only about 0.3 years to life expectancy, which is well in line with evidence that a given technology healthcare spending is subject to strongly diminishing returns (Skinner and Staiger, 2015) and that the introduction of Medicare per se has not led to significant improvements against mortality (Finkelstein and McKnight, 2008). Thus, the insurance-induced expansion of demand per se is mostly wasteful and, for a given level of medical technology, would indeed yield a welfare loss.25 This outcome is reversed when accounting for induced medical innovation, which adds a further 0.9 years to the increase in life expectancy over the time span 1965–2005. The expansion of health insurance then turns out to be Pareto optimal, benefiting all birth cohorts between 1910 and 1970, a result that holds even in the absence of income growth. Underlying these findings are two offsetting intergenerational externalities related to health insurance, particularly Medicare, that insures the high expenditures of the older population. On the one hand, excessive spending on the part of the elderly induces a negative externality on young working-age cohorts that, everything else equal, would lower their welfare. On the other hand, the excessive spending of the current elderly induces medical progress that predominantly benefits the future elderly, i.e., the current working-age cohorts.26 As it turns out, the latter externality more than offsets the former on a cohort-by-cohort basis (i.e., accounting for the full life cycle). This renders the expansion of health insurance a second-best intergenerational policy arrangement akin to unfunded social security that compensates older generations for educating younger cohorts (Boldrin and Montes, 2005; Andersen and Bhattacharya, 2017).27 B¨ohm et al. (2021) consider the impact of endogenous increases in the effectiveness of containing a Dalgaard and Strulik (2014)–type process of deficit accumulation. Given the forwardlooking simulation, their modeling, to some extent, reflects the scope for future advances of anti-ageing medicine. Their study also differs from Frankovic and Kuhn (2023) in that healthcare utilization is exogenously determined by a health authority, representing, e.g., the UK National Health Service to which the model is calibrated. Their findings are qualitatively like those in Frankovic and Kuhn (2023): the process of endogenous medical progress is strongly cumulative and generates increasing welfare gains over time. 72

Medical Progress, Ageing, and Sustainability of Healthcare Finance

While the modeling in Frankovic and Kuhn (2023) and B¨ohm et al. (2021) recognizes that the health and the medical R&D sectors absorb resources, their models assume that this does not bear on productivity growth in the rest of the economy. This may yield an overly optimistic view, especially of medical progress, given that the medical and nonmedical R&D sectors compete directly for skilled labor. Jones (2016) models such a rivalry, showing that if the marginal utility from consumption falls sufficiently fast relative to the marginal benefit from life expansion (which it seems to do for plausible parametrizations), it is optimal for medical R&D to crowd out conventional R&D over time. The increase in the willingness to pay for survival over time is then so large that individuals prefer life expansion if income growth is non-negative. Jones (2016) derives this result for a social planner setting, with work on a decentral setting in progress (Fernandez-Villaverde et al., 2017).

4.7

Conclusions

In conclusion, I summarize key insights and highlight directions for future research.

4.7.1

The Ageing–Medical Progress–Healthcare Spending Nexus

Ageing and medical progress jointly drive health expenditure, in particular among the elderly. For a given treatment style and state of medical technology, the impact of ageing hinges on the extent to which life years gained are spent in good health or with chronic disease. However, increases in longevity tend to induce additional treatment efforts and R&D effort into medical innovation. Furthermore, healthcare spending itself feeds back into R&D effort through market size effects. The joint endogeneity of all variables and their complementarity within the AMH nexus makes picking out causality a vain exercise. While the concept of the AMH nexus and its main macroeconomic dynamics are reasonably well understood by now, further research is needed with respect to quantifying the magnitudes of influences through various channels. While most empirical studies agree on the mostly positive correlations within the nexus, estimates still vary widely in terms of magnitude. This is due to differences in the conceptualization of medical progress and its impact on health, an understanding of which requires a multidisciplinary approach informed by medicine. Similarly, with the focus on mortality, survival, and longevity in many of the models, the role of morbidity and disability as key mediating factors remains insufficiently understood.

4.7.2

Role of Economic Development

Ageing and medical progress interact with income growth, in the presence of which increases in life expectancy brought about by additional health spending and induced medical innovation are optimal. Evidence that a decline in income growth tends to slow healthcare spending growth suggests the presence of automatic stabilization: Should healthcare spending growth compromise income growth by too much, the additional demand for health and healthcare tends to fall, which relieves the pressure on the economy. Altogether this suggests that concerns about unsustainable expansions of healthcare spending may be exaggerated. Nonetheless, the idea of “automatic stabilization” is only a conjecture based on very recent insight into the relationship between income growth and healthcare spending growth. More research is needed about the potential for stabilization that considers that the ageing process follows a lagged and to some extent independent dynamic. This research should also account for the question about the extent to which stagnant or declining healthcare spending leads to a deterioration of health and longevity. Recent work has started to probe into this, but reliable results are not yet available (e.g., McCartney et al., 2020). 73

Michael Kuhn

4.7.3

Role of Institutions

Institutions strongly shape the dynamics within the AMH nexus and their efficiency. Specifically, the generosity of health insurance acts as a further multiplier. While recent results suggest that an increase in health insurance coverage mitigates underinvestment in medical progress due to an intergenerational externality, more efficient means of stimulating medical innovation may exist. Furthermore, there are issues about how health insurance systems should be structured to improve inefficient spending dynamics. More generally, the role of institutions as mediators in the AMH dynamics has only begun to receive attention and predominantly with a focus on health insurance in the United States. However, relatively little is known about (1) the AMH nexus within other healthcare systems, including European style systems with their much larger role for public involvement and the healthcare systems of emerging economies; (2) the role of imperfect competition or government regulation in this process; and (3) which policies are conducive to improving the dynamic allocation, in particular in light of the fiscal pressures imposed by population ageing.

4.7.4

Type and Implementation of Medical Innovation

Further issues about efficiency arise with respect to the direction and implementation of medical innovation, where evidence reveals (1) a large proportion of innovations is poorly cost-effective (Chandra and Skinner, 2012); (2) some highly effective innovations are more of a gradual and organizational type and in some cases are rather slow and random in their diffusion (Skinner and Staiger, 2015); and (3) the effective and efficient implementation of medical innovations depends strongly on the institutional environment (Skinner et al., 2006). Despite first insights into the processes of medical innovation and its diffusion, a lack of (quantitative) understanding of the implications for the AMH nexus persists. In particular, how a large variety of treatments and treatment styles interact in determining healthcare spending and health outcomes at the macroeconomic level remains poorly understood, as does how competition and collaboration across providers and among medical practitioners shape the process of diffusion.

4.7.5

Distributional Concerns

The gains from medical progress are unevenly distributed across the population. Besides an increasing earnings and wealth gap, medical progress can be understood to be a key driver behind the gap in life expectancy. Indeed, medical progress is likely to lend strong leverage to the education and earnings gap as drivers of unequal health outcomes. Again, first research findings are only indicative of the strong distributional repercussions of medical progress, with much of this research again focusing on the U.S. healthcare system, which is known to be unequitable to begin with. Thus, more work is needed against the backdrop of healthcare systems of other economies. Finally, little is known about the interplay of medical advances and (differential) individual health behaviors, with this feedback being crucial for the extent to which medical innovations translate into (differential) health gains.

4.7.6

Future Development

The dynamics within the AMH nexus will in many ways turn on three factors. (1) The development of productivity growth both in general terms and within the healthcare sector in particular: It is important, however, that while the scope for future improvements to health and longevity depend on medical innovations, it does not depend on the extent of consumption 74

Medical Progress, Ageing, and Sustainability of Healthcare Finance

or even income growth per se as long as it does not turn negative (Jones, 2016). This suggests some scope for a decoupling of GDP growth and improvements in health and longevity. (2) The nature of medical innovations in their own right: To the extent that medical innovations expand healthy life time and to the extent they turn out to substitute rather than complement the utilization of healthcare (e.g., through low-intensity precision medicine or big-data-based personalized medicine), they decouple the ageing process from increases in healthcare spending. Furthermore, developments in the digitalization of healthcare and its automation based on robotics and artificial intelligence constitute process innovations that open significant scope for productivity increases and cost reductions. (3) The extent to which medical innovations are sufficient to offset adverse health trends due to deterioration in socioeconomic and environmental circumstances and in health behaviors: One crucial aspect of this question involves the diffusion of medical innovations to those social groups that are particularly disadvantaged. Overall, this loops back into the role of institutions and policymaking in fostering innovation and guaranteeing universal access. A full understanding of these processes requires more research but will in some ways have to wait for the frontier of medical technology to advance.

Notes 1 This is not without noting that broadly since the 2008 financial and economic crisis, both health expenditure growth and improvements in life expectancy have slowed down in many countries. This slowdown is discussed in Section 4.2.2 (with respect to healthcare spending), Section 4.3 (with respect to longevity expansion), and in the conclusion. 2 To the best of my knowledge this terminology has not been systematically used in the literature so far. 3 Population ageing, in the sense of a shift in the age distribution of a population toward higher age classes, is the result of two forces: increasing longevity and falling fertility. While the discussion about health expenditure growth is cast predominantly in terms of increasing longevity, fertility decline tends to raise health expenditure per capita by lowering the share of the young population with low expenditure levels while prospectively lowering the support ratio. As Breyer et al. (2015) and Breyer and Lorenz (2020) point out, this has important implications on the funding side. For a further exploration, see Section 4.6.1. 4 One aspect Breyer and Lorenz (2020) do not raise is that the neutrality of the cost of dying on healthcare spending relies on a continued deferral of mortality into higher age groups. Should advances in life expectancy slow down, the “tidal wave” of (deferred) deaths will ultimately break, leading to a peak in healthcare spending. 5 This ties in with the literature on health deficit accumulation (e.g., Rockwood and Mitnitski, 2007; Dalgaard and Strulik, 2014; Abeliansky and Strulik, 2018). A debate has emerged as to whether a slow-down in deficit accumulation has occurred for the European elderly (Abeliansky and Strulik, 2019; B¨orsch-Supan et al., 2021). 6 Zweifel et al. (2005) relate a similar finding to the ageing of the median voter and a consequent shift of preference expressions toward higher healthcare expenditure. 7 Fang et al. (2007) refer to this effect as the “Mickey Mantle effect,” following an alleged quote by a baseball star who, when mortally ill in advanced life, regretted his earlier abuse of alcohol: “If I’d known I was going to live this long, I would have taken better care of myself.” Breyer et al. (2015) argue that the quote should rather be attributed to the jazz musician Eubie Blake on the occasion of his 100th birthday. 8 In a dynamic context, this embraces the Baumol (1967) effect, according to which productivity growth in the conventional sectors of the economy triggers a shift of labor to the healthcare sector that typically exhibits low(er) productivity growth. In turn, this leads to medical price inflation as an additional driver of spending growth (Hartwig, 2008; Bates and Santerre, 2013). 9 For instance, Martin et al. (2008) consider specific National Health Service treatment programs for cancer and circulatory disease and report much higher spending elasticities (with respect to mortality) than those found in studies working with aggregate data. 10 Acemoglu and Johnson (2007) employ the epidemiological transition as a convenient instrument to address their original research question as to how health improvements bear on economic growth. This

75

Michael Kuhn

11

12

13 14 15

16 17

18

19 20

21

22

23 24

25

26 27

approach has been followed by a host of similar studies on the issue (e.g., Cervellati and Sunde, 2011; Bloom et al., 2014; Hansen and Lonstrup, 2015; Klasing and Milionis, 2020). Hansen and Strulik (2017) and Kotschy (2021) pursue this approach when estimating the impact of health improvements on education and productivity, based on data from the cardiovascular revolution. For this chapter, we restrict attention to the impact of the epidemiological transition and the cardiovascular revolution on longevity. The interpretation of a decline in period measures of life expectancy needs to be subject to caution. Luy et al. (2020) point out that fluctuations in period life expectancy may well be driven by cohort, tempo, and compositional effects rather than by period impacts. Finkelstein (2004) shows that policies aimed at stimulating the uptake of vaccination have triggered innovations, some of them being marginally wasteful but some yielding dynamic returns well above the static welfare gains. Cerda (2007) models the joint dynamics between induced drug innovation and induced ageing. This literature ties into a much broader literature on the macroeconomic impact of ageing that lies beyond the scope of this survey. See Bloom et al. (2019) for further detail. Kuhn and Prettner (2016) consider endogenous health-dependent participation in the labor market. In this setting health improvements may trigger a reduction in dependency if morbidity reductions lead to a deferral in the retirement age that overcompensates the increase in life expectancy. These channels are also present in Frankovic et al. (2020a,b) and Frankovic and Kuhn (2023). In Schneider and Winkler (2021), building on Romer’s (1986) model of capital-related technological spillovers, a similar mechanism is in place where the capital accumulation associated with old-age saving raises growth but the absorption of labor into the healthcare sector depresses it. Considering a decentral economy in which individuals invest in annuities, Schneider and Winkler (2021) identify a tendency toward excessive investments due to longevity moral hazard (Davies and Kuhn, 1992; Philipson and Becker, 1998; Kuhn et al., 2015). This notwithstanding, the introduction of long-term health insurance contracts (as, e.g., in Germany) may generate significant welfare gains compared with systems relying on serial short-term contracts (as, e.g., in the United States) (Atal et al., 2019, 2020). According to data from the Bureau of Economic Analysis, medical prices in the United States have risen 1.8 times faster than the overall consumption price index over the time span 1980–2005. It is important, here, to be clear about the definition of “output” and its “price.” In the present context, output refers to a pure volume measure of healthcare, e.g., hospital days or the size of a treatment bundle. This contrasts a setting where medical output and/or prices are adjusted for the quality of care, as measured, e.g., by survival gains. When using quality-adjusted measures of price and volume, productivity growth (Romley et al., 2015) and falling prices are found (e.g., Cutler et al., 1998; Lakdawalla et al., 2015; Hult et al., 2018). The observed inflation of volume-related prices arises when productivity growth in the healthcare sector falls short of productivity growth in the rest of the economy. This is tantamount to the so-called “Baumol” effect (Baumol, 1967; Acemoglu and Guerrieri, 2008; Frankovic et al., 2020a, b). Schneider and Winkler (2021) also model medical progress as an increase in medical effectiveness, i.e., a product innovation. They find it raises welfare if it does not lead to an excessive shift of the labor force from final goods production to healthcare and thereby stifles growth. Notably, and in contrast to most other studies, Frankovic and Kuhn (2018) and Frankovic et al. (2020b) provide full representations of the dynamics rather than steady-state comparisons. Dickman et al. (2016) compare healthcare spending by income group for the United States and document that after adjustments for age and health status the highest earners tend to spend most on health. To keep the analysis clean, Frankovic and Kuhn (2023) consider a risk-free setting and thereby abstract from any direct utility of insurance for risk-averse individuals. Thus, their welfare analysis presents a worst case for insurance expansion. Bhattacharya and Packalen (2012) consider a similar double externality in a static setting. Following empirical evidence that illustrates the relevance of slow diffusion of healthcare innovations (Skinner and Staiger, 2015), Frankovic et al. (2020a) consider the macroeconomic impacts of market size-dependent diffusion. Calibrating their diffusion process to represent some of the evidence from Skinner and Staiger (2015), they find that diffusion lags create a considerable drag on welfare, which justifies the subsidization of innovative healthcare.

76

Medical Progress, Ageing, and Sustainability of Healthcare Finance

References ABELIANSKY, A. L., AND STRULIK, H. (2018): “How we fall apart: Similarities of human ageing in 10 European countries,” Demography, 55(1): 341–359. ABELIANSKY, A. L., AND STRULIK, H. (2019): “Long-run improvements in human health: Steady but unequal,” The Journal of the Economics of Ageing, 14: 100189. ACEMOGLU, D., CUTLER, D., FINKELSTEIN, A., AND LINN, J. (2006): “Did Medicare induce pharmaceutical innovation?,” American Economic Review, 96(2): 103–107. ACEMOGLU, D., FINKELSTEIN, A., AND NOTOWIDIGDO, M. J. (2013): “Income and health spending: Evidence from oil price shocks,” Review of Economics and Statistics, 95(4): 1079–1095. ACEMOGLU, D., AND GUERRIERI, V. (2008): “Capital deepening and nonbalanced economic growth,” Journal of Political Economy, 116(3): 467–498. ACEMOGLU, D., AND JOHNSON, S. (2007): “Disease and development: The effect of life expectancy on economic growth,” Journal of Political Economy, 115(6): 925–985. ACEMOGLU, D., AND LINN, J. (2004): “Market size in innovation: Theory and evidence from the pharmaceutical industry,” The Quarterly Journal of Economics, 119(3): 1049–1090. AKSOY, Y., BASSO, H. S., SMITH, R. P., AND GRASL, T. (2019): “Demographic structure and macroeconomic trends,” American Economic Journal: Macroeconomics, 11(1): 193–222. ALDY, J. E., AND VISCUSI, W. K. (2008): “Adjusting the value of a statistical life for age and cohort effects,” The Review of Economics and Statistics, 90(3): 573–581. ANDERSEN, T. M., AND BHATTACHARYA, J. (2017): “The intergenerational welfare state and the rise and fall of pay-as-you-go pensions,” The Economic Journal, 127(602): 896–923. ¨ ARENTZ, C., EEKHOFF, J., AND KOCHSK AMPER , S. (2012): “Private health insurance: A role model for European health systems,” The European Journal of Health Economics, 13(5): 615–621. ATAL, J. P., FANG, H., KARLSSON, M., AND ZIEBARTH, N. R. (2019): “Exit, voice, or loyalty? An investigation into mandated portability of front-loaded private health plans,” Journal of Risk and Insurance, 86(3): 697–727. ATAL, J. P., FANG, H., KARLSSON, M., AND ZIEBARTH, N. R. (2020): “Long-term health insurance: Theory meets evidence.” Working Paper No. w26870. National Bureau of Economic Research, Cambridge, MA. AVITABILE, C., JAPPELLI, T., AND PADULA, M. (2011): “Cognitive abilities, healthcare and screening tests,” Journal of Population Ageing, Ageing 4(4): 251–269. BAGO D’UVA, T., LINDEBOOM, M., O’DONNELL, O., AND VAN DOORSLAER, E. (2011): “Educationrelated inequity in healthcare with heterogeneous reporting of health,” Journal of the Royal Statistical Society: Series A (Statistics in Society), 174(3): 639–664. BAICKER, K., CHANDRA, A., AND SKINNER, J. S. (2012): “Saving money or just saving lives? Improving the productivity of US healthcare spending,” Annual Review of Economics, 4(1): 33–56. BAKER, L., BIRNBAUM, H., GEPPERT, J., MISHOL, D., AND MOYNEUR, E. (2003): “The relationship between technology availability and healthcare spending: Attempts to address technology availability and rising costs could end up badly misguided if implications for quality are not considered,” Health Affairs, 22(Suppl 1): W3–537. BALTAGI, B. H., LAGRAVINESE, R., MOSCONE, F., AND TOSETTI, E (2017): “Healthcare expenditure and income: A global perspective,” Health Economics, 26(7): 863–874. BALTAGI, B. H., MOSCONE, F., AND TOSETTI, E. (2012): “Medical technology and the production of healthcare,” Empirical Economics, 42(2): 395–411. BATES, L. J., AND SANTERRE, R. E. (2013): “Does the US healthcare sector suffer from Baumol’s cost disease? Evidence from the 50 states,” Journal of Health Economics, 32(2): 386–391. BAUMOL, W. J. (1967): “Macroeconomics of unbalanced growth: The anatomy of urban crisis,” The American Economic Review, 57(3): 415–426. BHATTACHARYA, J., AND PACKALEN, M. (2011): “Opportunities and benefits as determinants of the direction of scientific research,” Journal of Health Economics, 30(4): 603–615. BHATTACHARYA, J., AND PACKALEN, M. (2012): “The other ex ante moral hazard in health,” Journal of Health Economics, 31(1): 135–146. BLOOM, D. E., CANNING, D., AND FINK, G. (2014): “Disease and development revisited,” Journal of Political Economy, 122(6): 1355–1366. BLOOM, D. E., CANNING, D., MANSFIELD, R. K., AND MOORE, M. (2007): “Demographic change, social security systems, and savings,” Journal of Monetary Economics, 54(1): 92–114.

77

Michael Kuhn

BLOOM, D. E., KUHN, M., AND PRETTNER, K. (2019): “Health and economic growth.” In: Hamilton, J. H., Dixit, A., Edwards, S., and Judd, K. (eds.), Oxford Research Encyclopedia of Economics and Finance. Oxford: Oxford University Press. doi: 10.1093/acrefore/9780190625979.013.36. BLUME-KOHOUT, M. E., AND SOOD, N. (2013): “Market size and innovation: Effects of Medicare Part D on pharmaceutical research and development,” Journal of Public Economics, 97: 327–336. ¨ BOHM , S., GROSSMANN, V., AND STRULIK, H. (2021): “R&D-driven medical progress, healthcare costs, and the future of human longevity,” The Journal of the Economics of Ageing, 18: 100286. BOKHARI, F. A., GAI, Y., AND GOTTRET, P. (2007): “Government health expenditures and health outcomes,” Health Economics, 16(3): 257–273. BOLDRIN, M., AND MONTES, A. (2005): “The intergenerational state education and pensions,” The Review of Economic Studies, 72(3): 651–664. ¨ BORSCH -SUPAN, A., FERRARI, I., AND SALERNO, L. (2021): “Long-run health trends in Europe,” The Journal of the Economics of Ageing, 18: 100303. BREYER, F., COSTA-FONT, J., AND FELDER, S. (2010): “Ageing, health, and healthcare,” Oxford Review of Economic Policy, 26(4): 674–690. BREYER, F., AND FELDER, S. (2006): “Life expectancy and healthcare expenditures: A new calculation for Germany using the costs of dying,” Health Policy, 75(2): 178–186. BREYER, F., AND LORENZ, N. (2020): “The ‘red herring’ after 20 years: Ageing and healthcare expenditures,” The European Journal of Health Economics, forthcoming. BREYER, F., LORENZ, N., AND NIEBEL, T. (2015): “Healthcare expenditures and longevity: Is there a Eubie Blake effect?,” The European Journal of Health Economics, 16(1): 95–112. BUNDORF, M. K., ROYALTY, A., AND BAKER, L. C. (2009): “Healthcare cost growth among the privately insured,” Health Affairs, 28(5): 1294–1304. BUXBAUM, J. D., CHERNEW, M. E., FENDRICK, A. M., AND CUTLER, D. M. (2020): “Contributions of public health, pharmaceuticals, and other medical care to US life expectancy changes, 1990–2015,” Health Affairs, 39(9): 1546–1556. CATILLON, M., CUTLER, D., AND GETZEN, T. (2018): “Two hundred years of health and medical care: The importance of medical care for life expectancy gains.” Working Paper No. w25330. National Bureau of Economic Research, Cambridge, MA. CERDA, R. A. (2007): “Endogenous innovations in the pharmaceutical industry,” Journal of Evolutionary Economics, 17(4): 473–515. CERVELLATI, M., AND SUNDE, U. (2011): “Life expectancy and economic growth: The role of the demographic transition,” Journal of Economic Growth, 16(2): 99–133. CHANDRA, A., AND SKINNER, J. (2012): “Technology growth and expenditure growth in healthcare,” Journal of Economic Literature, 50(3): 645–680. CHANDRA, A., AND STAIGER, D. O. (2020): “Identifying sources of inefficiency in healthcare,” The Quarterly Journal of Economics, 135(2): 785–843. CHERNEW, M. E., AND NEWHOUSE, J. P. (2011): “Healthcare spending growth.” In: Pauly, M. V., McGuire, T. G., and Barros, P. P. (eds.), Handbook of Health Economics, Vol. 2. Elsevier: Amsterdam et al., pp. 1–43. CHEN, S., KUHN, M., PRETTNER, K., BLOOM, D. E., AND WANG, C. (2021): “Macro-level efficiency of health expenditure: Estimates for 15 major economies,” Social Science & Medicine, 287: 114270. CHETTY, R., STEPNER, M., ABRAHAM, S., LIN, S., SCUDERI, B., TURNER, N., ... AND CUTLER, D. (2016): “The association between income and life expectancy in the United States, 2001–2014,” JAMA, 315(16): 1750–1766. CLEMENS, J. P., AND OLSEN, M. (2021): “Medicare and the rise of American medical patenting: The economics of user-driven innovation,” CESifo Working Paper 9008, CESifo, Munich. CR E´ MIEUX, P. Y., OUELLETTE, P., AND PILON, C. (1999): “Healthcare spending as determinants of health outcomes,” Health Economics, 8(7): 627–639. CUTLER, D. M. (2007): “The lifetime costs and benefits of medical technology,” Journal of Health Economics, 26(6): 1081–1100. CUTLER, D. M. (2008): “Are we finally winning the war on cancer?,” Journal of Economic Perspectives, 22(4): 3–26. CUTLER, D. M. (2018): “What is the US health spending problem?,” Health Affairs, 37(3): 493–497. CUTLER, D., DEATON, A., AND D LLERAS-MUNEY, A. (2006): “The determinants of mortality,” Journal of Economic Perspectives, 20(3): 97–120.

78

Medical Progress, Ageing, and Sustainability of Healthcare Finance

CUTLER, D. M., GHOSH, K., AND LANDRUM, M. B. (2014): “Evidence for significant compression of morbidity in the elderly US population.” In: Wise, D. A. (ed.), Discoveries in the Economics of Ageing, University of Chicago Press: Chicago and London, pp. 21–51. CUTLER, D. M., AND HUCKMAN, R. S. (2003): “Technological development and medical productivity: The diffusion of angioplasty in New York State,” Journal of Health Economics, 22(2): 187–217. CUTLER, D. M., LONG, G., BERNDT, E. R., ROYER, J., FOURNIER, A. A., SASSER, A., AND CREMIEUX, P. (2007): “The value of antihypertensive drugs: A perspective on medical innovation,” Health Affairs, 26(1): 97–110. CUTLER, D. M., AND MCCLELLAN, M. (2001): “Is technological change in medicine worth it?,” Health Affairs, 20(5): 11–29. CUTLER, D. M., MCCLELLAN, M., NEWHOUSE, J. P., AND REMLER, D. (1998): “Are medical prices declining? Evidence from heart attack treatments,” The Quarterly Journal of Economics, 113(4): 991–1024. DALGAARD, C. J., AND STRULIK, H. (2014): “Optimal ageing and death: Understanding the Preston curve,” Journal of the European Economic Association, 12(3): 672–701. DAVIES, J. B., AND KUHN, P. (1992): “Social security, longevity, and moral hazard,” Journal of Public Economics, 49(1): 91–106. DEBER, R. B., FORGET, E. L., AND ROOS, L. L. (2004): “Medical savings accounts in a universal system: Wishful thinking meets evidence,” Health Policy, 70(1): 49–66. DE M EIJER , C., WOUTERSE , B., P OLDER , J., AND KOOPMANSCHAP, M. (2013): “The effect of population ageing on health expenditure growth: A critical review,” European Journal of Ageing, 10(4): 353–361. DE NARDI, M., FRENCH, E., AND JONES, J. B. (2010): “Why do the elderly save? The role of medical expenses,” Journal of Political Economy, 118(1): 39–75. DICKMAN, S. L., WOOLHANDLER, S., BOR, J., MCCORMICK, D., BOR, D. H., AND HIMMELSTEIN, D. U. (2016): “Health spending for low-, middle-, and high-income Americans, 1963–2012,” Health Affairs, 35(7): 1189–1196. DORMONT, B., GRIGNON, M., AND HUBER, H. (2006): “Health expenditure growth: Reassessing the threat of ageing,” Health Economics, 15(9): 947–963. DRANOVE, D., GARTHWAITE, C., AND ODY, C. (2014): “Health spending slowdown is mostly due to economic factors, not structural change in the healthcare sector,” Health Affairs, 33(8): 1399–1406. DUBOIS, P., DE MOUZON, O., SCOTT-MORTON, F., AND SEABRIGHT, P. (2015): “Market size and pharmaceutical innovation,” The RAND Journal of Economics, 46(4): 844–871. EUROPEAN COMMISSION. (2018): “The 2018 ageing report: Economic and budgetary projections for the 28 EU member states (2016–2070).” European Economy Institutional Paper 079, Publications Office of the European Union: Luxembourg. FANG, H., KEANE, M., KHWAJA, A., SALM, M., AND SILVERMAN, D. (2007): “Testing the mechanisms of structural models: The case of the Mickey Mantle effect,” American Economic Review, 97(2): 53–59. ¨ , R., GROSSKOPF, S., LINDGREN, B., AND POULLIER, J. P. (1997): “Productivity growth in FARE healthcare delivery,” Medical Care, 35(4): 354–366. FELDSTEIN, M. (1977): “Quality change and the demand for hospital care,” Econometrica, 45(7): 1681–1702. ¨ , M. (2017): “An endogenous FERNANDEZ-VILLAVERDE, J., KRUEGER, D., LUDWIG, A., AND SCH ON growth model with a health sector,” Society for Economic Dynamics 2017 Meeting Paper No. 767. Society for Economic Dynamics. FILMER, D., AND PRITCHETT, L. (1999): “The impact of public spending on health: Does money matter?,” Social Science & Medicine, 49(10): 1309–1323. FINKELSTEIN, A. (2004): “Static and dynamic effects of health policy: Evidence from the vaccine industry,” The Quarterly Journal of Economics, 119(2): 527–564. FINKELSTEIN, A. (2007): “The aggregate effects of health insurance: Evidence from the introduction of Medicare,” The Quarterly Journal of Economics, 122(1): 1–37. FINKELSTEIN, A., AND MCKNIGHT, R. (2008): “What did Medicare do? The initial impact of Medicare on mortality and out of pocket medical spending,” Journal of Public Economics, 92(7): 1644–1668. FIVA, J. H., HÆGELAND, T., RØNNING, M., AND SYSE, A. (2014): “Access to treatment and educational inequalities in cancer survival,” Journal of Health Economics, 36: 98–111. FONSECA, R., MICHAUD, P. C., GALAMA, T., AND KAPTEYN, A. (2021): “Accounting for the rise of health spending and longevity,” Journal of the European Economic Association, 19(1): 536–579.

79

Michael Kuhn

FORD, E. S., AJANI, U. A., CROFT, J. B., CRITCHLEY, J. A., LABARTHE, D. R., KOTTKE, T. E., ... CAPEWELL, S. (2007): “Explaining the decrease in US deaths from coronary disease, 1980–2000,” New England Journal of Medicine, 356(23): 2388–2398. FRANKOVIC, I., AND KUHN, M. (2019): “Access to healthcare, medical progress and the emergence of the longevity gap: A general equilibrium analysis,” The Journal of the Economics of Ageing, 14: 100188. FRANKOVIC, I., AND KUHN, M. (2023): “Health insurance, endogenous medical progress, health expenditure growth, and welfare,” Journal of Health Economics, 102717. FRANKOVIC, I., KUHN, M., AND WRZACZEK, S. (2020a): “Medical innovation and its diffusion: Implications for economic performance and welfare,” Journal of Macroeconomics, 66: 103262. FRANKOVIC, I., KUHN, M., AND WRZACZEK, S. (2020b): “On the anatomy of medical progress within an overlapping generations economy,” De Economist, 168(2): 215–257. FRENCH, E. B., MCCAULEY, J., ARAGON, M., BAKX, P., CHALKLEY, M., CHEN, S. H., ... KELLY, E. (2017): “End-of-life medical spending in last twelve months of life is lower than previously reported,” Health Affairs, 36(7): 1211–1217. GALLET, C. A., AND DOUCOULIAGOS, H. (2017): “The impact of healthcare spending on health outcomes: A meta-regression analysis,” Social Science & Medicine, 179: 9–17. GETZEN, T. E. (2000): “Healthcare is an individual necessity and a national luxury: Applying multilevel decision models to the analysis of healthcare expenditures,” Journal of Health Economics, 19(2): 259–270. GLIED, S., AND LLERAS-MUNEY, A. (2008): “Technological innovation and inequality in health,” Demography, 45(3): 741–761. GOLDMAN, D. (2016): “The economic promise of delayed ageing,” Cold Spring Harbor Perspectives in Medicine, 6(2): a025072. GOLDMAN, D. P., CUTLER, D., ROWE, J. W., MICHAUD, P. C., SULLIVAN, J., PENEVA, D., AND OLSHANSKY, S. J. (2013): “Substantial health and economic returns from delayed ageing may warrant a new focus for medical research,” Health Affairs, 32(10): 1698–1705. GOLDMAN, D., MICHAUD, P. C., LAKDAWALLA, D., ZHENG, Y., GAILEY, A., AND VAYNMAN, I. (2010): “The fiscal consequences of trends in population health,” National Tax Journal, 63(2): 307. GOLDMAN, D. P., SHANG, B., BHATTACHARYA, J., GARBER, A. M., HURD, M., JOYCE, G. F., ... SHEKELLE, P. G. (2005): “Consequences of health trends and medical innovation for the future elderly: When demographic trends temper the optimism of biomedical advances, how will tomorrow’s elderly fare?,” Health Affairs, 24(Suppl2): W5-R5. HALL, R. E., AND JONES, C. I. (2007): “The value of life and the rise in health spending,” The Quarterly Journal of Economics, 122(1): 39–72. HANSEN, C. W., AND LONSTRUP, L. (2015): “The rise in life expectancy and economic growth in the 20th century,” The Economic Journal, 125(584): 838–852. HANSEN, C. W., AND STRULIK, H. (2017): “Life expectancy and education: Evidence from the cardiovascular revolution,” Journal of Economic Growth, 22(4): 421–450. HARTWIG, J. (2008): “What drives healthcare expenditure?—Baumol’s model of ‘unbalanced growth’ revisited,” Journal of Health Economics, 27(3): 603–623. HERNANDEZ, E. M., MARGOLIS, R., AND HUMMER, R. A. (2018): “Educational and gender differences in health behavior changes after a gateway diagnosis,” Journal of Ageing and Health, 30(3): 342–364. HOWDON, D., AND RICE, N. (2018): “Healthcare expenditures, age, proximity to death and morbidity: Implications for an ageing population,” Journal of Health Economics, 57: 60–74. HULT, K. J., JAFFE, S., AND D PHILIPSON, T. J. (2018): “How does technological change affect quality-adjusted prices in healthcare? Systematic evidence from thousands of innovations,” American Journal of Health Economics, 4(4): 433–453. HURLEY, J., GUINDON, G. E., RYNARD, V., AND MORGAN, S. (2008): “Publicly funded medical savings accounts: Expenditure and distributional impacts in Ontario, Canada,” Health Economics, 17(10): 1129–1151. JONES, C. I. (2016): “Life and growth,” Journal of Political Economy, 124(2): 539–578. JUNG, J., AND TRAN, C. (2016): “The macroeconomics of health savings accounts,” Towson University, Dept of Economics Working Paper No. 2010-12. Towson University: Towson, MD. JUNG, J., TRAN, C., AND CHAMBERS, M. (2017): “Ageing and health financing in the US: A general equilibrium analysis,” European Economic Review, 100: 428–462. KELLY, M. (2017): “Health capital accumulation, health insurance, and aggregate outcomes: A neoclassical approach,” Journal of Macroeconomics, 52: 1–22.

80

Medical Progress, Ageing, and Sustainability of Healthcare Finance

KELLY, M. (2017): “Congestion in a public health service: A macro approach,” Journal of Macroeconomics, 74: 103451. KLASING, M. J., AND MILIONIS, P. (2020): “The international epidemiological transition and the education gender gap,” Journal of Economic Growth, 25(1): 37–86. KNIESNER, T. J., AND VISCUSI, W. K. (2019): “The value of a statistical life.” In: Hamilton, J. H., Dixit, A., Edwards, S., and Judd, K. (eds.),Oxford Research Encyclopedia of Economics and Finance, Oxford: Oxford University Press. KOPECKY, K. A., AND KORESHKOVA, T. (2014): “The impact of medical and nursing home expenses on savings,” American Economic Journal: Macroeconomics, 6(3): 29–72. KOTSCHY, R. (2021): “Health dynamics shape life-cycle incomes,” Journal of Health Economics, 75: 102398. KUHN, M., AND PRETTNER, K. (2016): “Growth and welfare effects of healthcare in knowledge-based economies,” Journal of Health Economics, 46: 100–119. KUHN, M., WRZACZEK, S., PRSKAWETZ, A., AND FEICHTINGER, G. (2015): “Optimal choice of health and retirement in a life-cycle model,” Journal of Economic Theory, 158 (Part A): 186–212. LAKDAWALLA, D. N., BHATTACHARYA, J., AND GOLDMAN, D. P. (2004): “Are the young becoming more disabled?,” Health Affairs, 23(1): 168–176. LAKDAWALLA, D., SHAFRIN, J., LUCARELLI, C., NICHOLSON, S., KHAN, Z. M., AND PHILIPSON, T. J. (2015): “Quality-adjusted cost of care: A meaningful way to measure growth in innovation cost versus the value of health gains,” Health Affairs, 34(4): 555–561. LANGE, F. (2011): “The role of education in complex health decisions: Evidence from cancer screening,” Journal of Health Economics, 30(1): 43–54. LI, M., LAKDAWALLA, D. N., AND GOLDMAN, D. P. (2020): “Association between spending and outcomes for patients with cancer,” Journal of Clinical Oncology, 38(4): 323. LICHTENBERG, F. R. (2004): “Sources of US longevity increase, 1960–2001,” The Quarterly Review of Economics and Finance, 44(3): 369–389. LICHTENBERG, F. R. (2007): “The impact of new drugs on US longevity and medical expenditure, 1990–2003: Evidence from longitudinal, disease-level data,” American Economic Review, 97(2): 438–443. LICHTENBERG, F. R. (2009): “Have newer cardiovascular drugs reduced hospitalization? Evidence from longitudinal country-level data on 20 OECD countries, 1995–2003,” Health Economics, 18(5): 519–534. LICHTENBERG, F. R. (2013): “The impact of therapeutic procedure innovation on hospital patient longevity: Evidence from Western Australia, 2000–2007,” Social Science & Medicine, 77: 50–59. LICHTENBERG, F. R. (2018): “The impact of new drug launches on life-years lost in 2015 from 19 types of cancer in 36 countries,” Journal of Demographic Economics, 84(3): 309–354. LICHTENBERG, F. R. (2019): “The impact of pharmaceutical innovation on the burden of disease in Canada, 2000–2016,” SSM-Population Health, 8: 100457. LOWSKY, D. J., OLSHANSKY, S. J., BHATTACHARYA, J., AND GOLDMAN, D. P. (2014): “Heterogeneity in healthy ageing,” Journals of Gerontology Series A: Biomedical Sciences and Medical Sciences, 69(6): 640–649. LUY, M., DI GIULIO, P., DI LEGO, V., LAZAREVI Cˇ , P., AND SAUERBERG, M. (2020): “Life expectancy: Frequently used, but hardly understood,” Gerontology, 66(1): 95–104. MACEWAN, J. P., DENNEN, S., KEE, R., ALI, F., SHAFRIN, J., AND BATT, K. (2020): “Changes in mortality associated with cancer drug approvals in the United States from 2000 to 2016,” Journal of Medical Economics, 23(12): 1558–1569. MANNING, W. G., NEWHOUSE, J. P., DUAN, N., KEELER, E. B., AND LEIBOWITZ, A. (1987): “Health insurance and the demand for medical care: Evidence from a randomized experiment,” The American Economic Review, 77(3): 251–277. MANTON, K. G., GU, X., AND LAMB, V. L. (2006): “Long-term trends in life expectancy and active life expectancy in the United States,” Population and Development Review, 32(1): 81–105. MARTIN, S., RICE, N., AND SMITH, P. C. (2008): “Does healthcare spending improve health outcomes? Evidence from English programme budgeting data,” Journal of Health Economics, 27(4): 826–842. MCCARTNEY, G., FENTON, L., MINTON, J., FISCHBACHER, C., TAULBUT, M., LITTLE, K., ... MCMASTER, R. (2020): “Is austerity responsible for the recent change in mortality trends across high-income nations? A protocol for an observational study,” BMJ Open, 10(1): e034832. MEHTA, N. K., ABRAMS, L. R., AND MYRSKYL A¨ , M. (2020): “US life expectancy stalls due to cardiovascular disease, not drug deaths,” Proceedings of the National Academy of Sciences, 117(13): 6998–7000. MEIER, V. (2005): “Efficient transfer of ageing provisions in private health insurance,” Journal of Economics, 84(3): 249–275.

81

Michael Kuhn

MICHAUD, P. C., GOLDMAN, D., LAKDAWALLA, D., GAILEY, A., AND ZHENG, Y. (2011): “Differences in health between Americans and Western Europeans: Effects on longevity and public finance,” Social Science & Medicine, 73(2): 254–263. MINICOZZI, A. (2006): “Medical savings accounts: What story do the data tell?,” Health Affairs, 25(1): 256–267. MORENO-SERRA, R., AND SMITH, P. C. (2015): “Broader health coverage is good for the nation’s health: Evidence from country level panel data,” Journal of the Royal Statistical Society. Series A (Statistics in Society), 178(1): 101. MURPHY, K. M., AND TOPEL, R. H. (2006): “The value of health and longevity,” Journal of Political Economy, 114(5): 871–904. MURTHY, V. N., AND KETENCI, N. (2017): “Is technology still a major driver of health expenditure in the United States? Evidence from cointegration analysis with multiple structural breaks,” International Journal of Health Economics and Management, 17(1): 29–50. NAKAMURA, R., LOMAS, J., CLAXTON, K., BOKHARI, F., MORENO-SERRA, R., AND SUHRCKE, M. (2020): “Assessing the impact of healthcare expenditures on mortality using cross-country data.” In: Revill, P., Suhrcke, M., Moreno-Serra, R., and Sculpher, M. (eds.), Global Health Economics: Shaping Health Policy in Low-and Middle-Income Countries, Singapore: World Scientific, pp. 3–49. NEWHOUSE, J. P. (1992): “Medical care costs: How much welfare loss?,” Journal of Economic Perspectives, 6(3): 3–21. OECD (ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT). (2018): Health Statistics: Causes of Mortality. OECD library. Paris: OECD [accessed on June 25, 2018]. OECD (ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT). (2019): Health at a Glance 2019: OECD Indicators. Paris: OECD Publishing. OKUNADE, A. A., AND MURTHY, V. N. (2002): “Technology as a ‘major driver’ of healthcare costs: A cointegration analysis of the Newhouse conjecture,” Journal of Health Economics, 21(1): 147–159. OKUNADE, A. A., AND OSMANI, A. R. (2018): “Technology, productivity, and costs in healthcare.” In: Hamilton, J. H., Dixit, A., Edwards, S., and Judd, K. (eds.), Oxford Research Encyclopedia of Economics and Finance. Oxford: Oxford University Press, pp. 1–21. doi: 10.1093/acrefore/9780190625979.013.64. OLSHANSKY, S. J., PASSARO, D. J., HERSHOW, R. C., LAYDEN, J., CARNES, B. A., BRODY, J., ... LUDWIG, D. S. (2005): “A potential decline in life expectancy in the United States in the 21st century,” New England Journal of Medicine, 352(11): 1138–1145. OMRAN, A. R. (1998): “The epidemiologic transition theory revisited thirty years later,” World Health Statistics Quarterly, 53(2, 3, 4): 99–119. PANDYA, A., GAZIANO, T. A., WEINSTEIN, M. C., AND CUTLER, D. (2013): “More Americans living longer with cardiovascular disease will increase costs while lowering quality of life,” Health Affairs, 32(10): 1706–1714. PAULY, M. V., GOODMAN, J. C., FEDER, J., LEVITT, L., BUTLER, S. M., CUTLER, D. M., AND WILENSKY, G. R. (1995): “Tax credits for health insurance and medical savings accounts,” Health Affairs, 14(1): 125–139. PHILIPSON, T. J., AND BECKER, G. S. (1998): “Old-age longevity and mortality-contingent claims,” Journal of Political Economy, 106(3): 551–573. RALEIGH, V. S. (2019): “Trends in life expectancy in EU and other OECD countries: Why are improvements slowing?,” OECD Health Working Papers 108. OECD Publishing, Paris. REITHER, E. N., OLSHANSKY, S. J., AND YANG, Y. (2011): “New forecasting methodology indicates more disease and earlier mortality ahead for today’s younger Americans,” Health Affairs, 30(8): 1562–1568. ROCKWOOD, K., AND MITNITSKI, A. (2007): “Frailty in relation to the accumulation of deficits,” The Journals of Gerontology Series A: Biological Sciences and Medical Sciences, 62(7): 722–727. ROHAM, M., GABRIELYAN, A. R., ARCHER, N. P., GRIGNON, M. L., AND SPENCER, B. G. (2014): “The impact of technological intensity of service provision on physician expenditures: An exploratory investigation,” Health Economics, 23(10): 1224–1241. ROMER, P. M. (1986): “Increasing returns and long-run growth,” Journal of Political Economy, 94(5): 1002–1037. ROMER, P. M. (1990): “Endogenous technological change,” Journal of Political Economy, 98(5, Part 2): S71–S102. ROMLEY, J. A., GOLDMAN, D. P., AND SOOD, N. (2015): “US hospitals experienced substantial productivity growth during 2002–11,” Health Affairs, 34(3): 511–518.

82

Medical Progress, Ageing, and Sustainability of Healthcare Finance

ROSEN, A. B., CUTLER, D. M., NORTON, D. M., HU, H. M., AND VIJAN, S. (2007): “The value of coronary heart disease care for the elderly: 1987–2002,” Health Affairs, 26(1): 111–123. SANSO, M., AND A´I SA, R. M. (2006): “Endogenous longevity, biological deterioration and economic growth,” Journal of Health Economics, 25(3): 555–578. SCHNEIDER, M. T., AND WINKLER, R. (2021): “Growth and welfare under endogenous lifetimes,” The Scandinavian Journal of Economics, 123(4), 1339-1384. SESHAMANI, M., AND GRAY, A. (2004): “Ageing and health-care expenditure: The red herring argument revisited,” Health Economics, 13(4): 303–314. SHANG, B., AND GOLDMAN, D. (2008): “Does age or life expectancy better predict healthcare expenditures?,” Health Economics, 17(4): 487–501. SKINNER, J., AND STAIGER, D. (2015): “Technology diffusion and productivity growth in healthcare,” Review of Economics and Statistics, 97(5): 951–964. SKINNER, J. S., STAIGER, D. O., AND FISHER, E. S. (2006): “Is technological change in medicine always worth it? The case of acute myocardial infarction: Waste and inefficiency are not inevitable by-products of technological growth,” Health Affairs, 25(Suppl1): W34–W47. SMITH, S., NEWHOUSE, J. P., AND FREELAND, M. S. (2009): “Income, insurance, and technology: Why does health spending outpace economic growth?,” Health Affairs, 28(5): 1276–1284. STEVENS, W., PHILIPSON, T. J., KHAN, Z. M., MACEWAN, J. P., LINTHICUM, M. T., AND GOLDMAN, D. P. (2015): “Cancer mortality reductions were greatest among countries where cancer care spending rose the most, 1995–2007,” Health Affairs, 34(4): 562–570. SUEN, R. M. (2005): “Technological advance and the growth in healthcare spending.” Economie d’Avant Garde Research Reports 13, Economie d’Avant Garde. VALLEJO-TORRES, L., AND MORRIS, S. (2013): “Income-related inequity in healthcare utilisation among individuals with cardiovascular disease in England—Accounting for vertical inequity,” Health Economics, 22(5): 533–553. VAN ZON, A., AND MUYSKEN, J. (2001): “Health and endogenous growth,” Journal of Health Economics, 20(2): 169–185. WEISBROD, B. A (1991): “The healthcare quadrilemma: An essay on technological change, insurance, quality of care, and cost containment,” Journal of Economic Literature, 29(2): 523–552. WERBLOW, A., FELDER, S., AND ZWEIFEL, P. (2007): “Population ageing and healthcare expenditure: A school of ‘red herrings’?,” Health Economics, 16(10): 1109–1126. WONG, A., WOUTERSE, B., SLOBBE, L. C., BOSHUIZEN, H. C., AND POLDER, J. J. (2012): “Medical innovation and age-specific trends in healthcare utilization: Findings and implications,” Social Science & Medicine, 74(2): 263–272. WOUTERS, O. J., CYLUS, J., YANG, W., THOMSON, S., AND MCKEE, M. (2016): “Medical savings accounts: Assessing their impact on efficiency, equity and financial protection in healthcare,” Health Economics, Policy & Law, 11(3): 321–335. WOUTERSE, B., HUISMAN, M., MEIJBOOM, B. R., DEEG, D. J., AND POLDER, J. J. (2015): “The effect of trends in health and longevity on health services use by older adults,” BMC Health Services Research, 15(1): 1–14. YOU, X., AND OKUNADE, A. A. (2017): “Income and technology as drivers of Australian healthcare expenditures,” Health Economics, 26(7): 853–862. ZHAO, K. (2014): “Social security and the rise in health spending,” Journal of Monetary Economics, 64: 21–37. ZWEIFEL, P., FELDER, S., AND MEIER, M. (1999): “Ageing of population and healthcare expenditure: A red herring?,” Health Economics, 8(6): 485–496. ZWEIFEL, P., STEINMANN, L., AND EUGSTER, P. (2005): “The Sisyphus syndrome in health revisited,” International Journal of Healthcare Finance and Economics, 5(2): 127–145.

83

5 TECHNOLOGIES TO MITIGATE COGNITIVE AGEING Neil Charness1

Abstract I examine the role of technologies in mitigating cognitive ageing for ageing populations and their potential for increasing work productivity and independent living. Because older adult cohorts lag in technology adoption compared with their younger counterparts, overcoming the age-related digital divide is an important challenge to meet to support technology interventions. Normative cognitive declines occur in the decade of the 20s for fluid abilities like abstract problem solving, whereas crystallized abilities, knowledge acquired from one’s culture, tend to rise into the 50s or 60s and decline modestly. Information-processing capabilities such as working memory capacity and speed of processing decline between 20 and 100 percent from young to older adulthood. Attempts to rehabilitate cognitive decline with technology-assisted brain training have improved abilities but rarely show efficacy for outcomes such as work productivity or independence. Technologies that augment cognitive capabilities are more promising, particularly in the domain of advanced driver assistance systems for safe driving. For cognitive declines seen in dementia, technologies that substitute for lost functions may be feasible. Promising technologies that may benefit cognition-dependent work productivity include technologyassisted training and human–computer collaborations to support artificial intelligence expert systems. Systems that may promote greater independence at advanced ages by supporting activities of daily living, instrumental activities of daily living, and enhanced activities of daily living include mobile technology to support prospective memory, social media and videoconferencing technology to support social connectedness, and extended reality systems to support leisure activities.

5.1

Introduction

Technology advances have been essential in improving the human condition worldwide. Technologies such as vaccinations for children; the green revolution, which provided a reliable food supply for most of the world; engineering systems that provide access to clean water; and improvements in medical treatments have resulted in a longevity dividend of an additional 30 years in life expectancy from birth in the United States since 1900 and 25 years in the world 84

DOI: 10.4324/9781003150398-6

Technologies to Mitigate Cognitive Ageing

since 1950. Such longevity gains coupled with general fertility declines have led to ageing populations in many nations, with strong growth expected for those in older-age groups as shown in Figure 5.1. Adult development and ageing are associated with normative changes in informationprocessing abilities such as perceiving, attending to external and internal sources of information, learning, problem solving, and making decisions. Some changes are positive, such as increases in knowledge across the life span, while many others are negative, impairing ability to work or threatening the ability to live independently in the community. Age is also associated with nonnormative changes to brain function that are direct threats to independence, such as development of cognitive impairments from dementia, traumatic brain injuries (e.g., from falls), and strokes (from cardiovascular disease). Such age-related changes have implications for the well-being of both individuals and nation states. For instance, analyses for the United States (Maestas et al., 2016) and the European Union (Aiyar et al., 2016) projected a 1 percent decline in U.S. gross domestic product and a 2 percent decline in EU total factor productivity over the next decade, which they attributed to workforce ageing. Such projections should be viewed cautiously because the business cycle is likely to be a more powerful influence on decadal productivity than population ageing. Perturbations, such as a recession, or even a once-in-a-lifetime pandemic, can exert significant influence. In this chapter, I examine the prospects for technology to mitigate normative agerelated changes in cognition, prolonging independence for people and productivity for the labor force.

Figure 5.1 World population projections (millions) by older adult age groups age 65+ and age 85+ (millions). Source: https://population.un.org/wpp/Download/Standard/Population/.

85

Neil Charness

5.2

Age-Associated Changes in Cognition

Cross-sectional and longitudinal investigations of cognition across the adult life span show that some cognitive abilities are relatively age sensitive and others are not (Salthouse, 2010). For instance, knowledge-related capabilities tend to peak in the 50s and then stabilize into the 60s and 70s, often undergoing significant decline in some abilities only near end of life, the so-called “terminal drop” (White and Cunningham, 1988). Knowledge is measured with standardized tests of information (“what is the distance between New York and Paris”) and vocabulary (e.g., asking people to recognize the definitions of words), that is, information learned by people through continuous exposure to their culture. The types of cognitive abilities that tend to be well maintained with age have been called “crystallized abilities.” The types that show normative decline, often from the 20s, have been termed “fluid abilities” (e.g., ones that affect planning and abstract problem solving). The ability declines, measured in standard deviation units, are typically 1.5–2 SDs from the 20s to the 60s (Salthouse, 2010). Information-processing theorists (e.g., Newell and Simon, 1972) derived parameters for processing speed, memory capacity, and learning rate using an idealized model human processor (e.g., Card et al., 1983), and that work was extended to accommodate ageing adults (e.g., Jastrzembski and Charness, 2007). Table 5.1 shows that information-processing rates for cognitive activities such as perception, thinking, and motor performance undergo significant age-related slowing, leading to 20–100 percent increases in task completion duration between the decades of the 20s and the 60s. Working memory capacity (ability to store and manipulate information) declines by about 20 percent. See also Verhaeghen (2014). However, such declines are typically measured using novel tasks that minimize the ability to draw on knowledge and acquired skills. As mentioned earlier, knowledge tends to increase with age into midlife and sometimes can compensate for declines in basic abilities for job-related tasks, such as learning word processor software (Charness et al., 2001). Nonetheless, in a world that increasingly incorporates digital technology into everyday tasks, full participation in societal activities requires new learning. Seniors planning vacation trips Table 5.1 Information-processing parameters for young and old adults

Source: Jastrzembski and Charness (2007).

86

Technologies to Mitigate Cognitive Ageing

are forced to pay additional fees if they attempt to book flights using airline employees, rather than booking fares on company or aggregator websites. Hyper-competitive industries depend on rapid adoption of new technology to improve the quality of their human capital, thereby increasing worker productivity. So, slowing in information-processing rate, particularly in learning rate, can affect the bottom line of companies and potentially lead to declines in well-being of seniors navigating modern life.

5.3

The Role of Technology in Mitigation

Fortunately, we are on the cusp of development of intelligent technology systems that can support people for a wide variety of tasks. However, before technology interventions can take place, considering the Prevent-Rehabilitate-Augment-Substitute framework for such interventions is helpful (Charness, 2020; Figure 5.2). Prevention of impairments is an important principle for ensuring that much of the population reaches old age in the best possible shape. Public education and public health policy can play important roles in prevention activities (e.g., supporting pregnant women, vaccinating infants, encouraging regular exercise and good dietary practices across the life span, and encouraging best practices for health and safety at work). However, normative age-related changes in adulthood, even if slowed by preventive measures, are to be expected, so an important goal is finding techniques to compensate for the type of cognitive impairments outlined in Table 5.1. Compensatory strategies fall into three categories: rehabilitation, augmentation, and substitution.

Figure 5.2 The Prevent-Rehabilitate-Augment-Substitute framework (Charness, 2020) outlining most (bottom) to least preferred technology interventions.

87

Neil Charness

5.3.1

Rehabilitation

Epidemiological/association study evidence shows that older adults with more education (L¨ovd´en et al., 2020), those engaged in more challenging work (Andel et al., 2007), or those who seek out more demanding leisure activities (Verghese et al., 2003) show better cognitive performance in adulthood. Such relationships have led theorists to propose that such activities can build up (bank) “cognitive reserve”(Stern, 2012) that enables people to resist the development of cognitive impairment, such as in Alzheimer’s disease. Software companies, drawing on promising results from experimental studies capable of showing cause-and-effect relationships, such as the ACTIVE clinical trial (Ball et al., 2002), have developed “brain training” programs to improve cognition and counter normative changes in cognition, such as decline in working memory capacity and slowing in processing speed. Perhaps the earliest example was Nintendo’s Brain Age released in 2005 for Nintendo’s DS platform, though little evidence was provided to demonstrate cognitive benefits. Other companies quickly followed this pathway, with some like Lumosity making unwarranted claims about efficacy that resulted in fines from the U.S. Federal Trade Commission (https://www.ftc.gov/news-events/press-releases/2016/01/lumosity-pay-2-million-settleftc-deceptive-advertising-charges). A careful review of the studies cited as evidence for efficacy by the brain-training industry (Simons et al., 2016) concluded that such rehabilitative cognitive exercises were effective for improving performance on the cognitive abilities that were trained. Disappointingly, little evidence indicated that such training showed near transfer, i.e., transfer to related abilities, and no evidence indicated far transfer, i.e., transfer to activities in daily life (e.g., supporting independence or work performance). However, a recent meta-analysis indicated positive benefits for cognitive training from randomized controlled trials (Basak et al., 2020), so conclusions about the cognitive training field remain in flux. However, acquired human skills appear to be very narrow and generally show little evidence of far transfer to unrelated tasks (Sala and Gobet, 2017). As that review indicates, there is little expectation of much transfer between unrelated tasks such as using chess-in-the-school programs to improve a child’s ability to read. A major methodological problem in many intervention studies is the failure to provide an adequate control group (Boot et al., 2013b), complicating the interpretation of any effects obtained. Fortunately, more recent randomized controlled trials are using active control groups that control for participant expectations about outcomes. Another method for improving cognition broadly, which has shown promise in animal models, is physical exercise, through changes in brain chemistry that support new learning. An aerobic exercise intervention with sedentary older adults that used brisk walking (Colcombe and Kramer, 2003) has shown broad transfer to some cognitive abilities (e.g., executive function, which is responsible for planning and problem solving), though the size of the effect from early literature in this area (Voss et al., 2011) has been relatively modest, about a quarter to a half of a standard deviation change in the context of 1.5–2 SD declines in cognitive abilities across adulthood. The final sections of the chapter discuss more recent evaluations of physical activity interventions. In summary, attempts to rehabilitate cognition through technology-delivered cognitive exercises have shown low value in terms of potential outcomes sought by consumers (e.g., prevention of cognitive impairment, maintenance of independence), although they have benefited the bottom lines of the companies in this growing billion dollar industry (https://www.ft.com/content/c6028b80-3385-11e4-ba62-00144feabdc0). Given that rehabilitation of cognition through brain training has shown limited promise for reversing cognitive 88

Technologies to Mitigate Cognitive Ageing

decline, perhaps only slowing it (Rebok et al., 2014), can technology be harnessed to support or replace failing cognitive functioning?

5.3.2

Augmentation

The human species is renowned for its tool use (“homo habilis”) and for the superior knowledge that large brains and tools enabled (“homo sapiens”). Despite a slow start for computer-based technology in terms of improving the human condition (“You can see the computer age everywhere but in the productivity statistics.”—Robert Solow, New York Times Book Review, July 12, 1987, p. 36), microchip-based technology has become ubiquitous. It has also been essential for propping up the economy during the COVID-19 pandemic, allowing large swaths of the labor force to work from home via Internet connectivity and enabling tech-savvy, socially isolated older adults to visit family and friends virtually through videoconferencing. This section discusses the role of technology in buttressing cognitive performance in ageing adults, first noting the challenges posed by the age-related digital divide.

5.3.2.1

Digital Divide

Figure 5.3 shows that a pronounced age-related gap in technology adoption accompanies population ageing. The figure shows data from representative sampling of the U.S. population on Internet use, tracked by Pew Research from 2000 to 2021 (with 2020 missing, perhaps because of the disruptive effects of the pandemic). These data were gathered from annual cross-sectional studies, meaning that they are not longitudinal data following a particular age cohort over time. The three younger age/cohort groups have converged as of 2021 to virtually 100 percent of

Figure 5.3 Percent Internet use by age group (18–29 years, 30–49 years, 50–64 years, 65+ years) and year (2000–2021). Source: https://www.pewresearch.org/internet/fact-sheet/internet-broadband/?menuItem=9a15d0d33bff-4e9e-a329-6e328bc7bcce

89

Neil Charness

respondents indicating that they have used the Internet in the past year. Despite the obvious relevance of the Internet for dealing with the pandemic-related constraints, the rise in use, from 73 percent to 75 percent (probably within the margin of error for a survey) from 2019 to 2021, is modest for those age 65+ year. That is, the age gap for Internet use has been remarkably persistent for senior citizens. Similar cross-sectional gaps can be seen in the United States for smartphone ownership, where although 92 percent of those age 65+ years reported owning a mobile phone, only 61 percent reported owning a smartphone (Pew Research Center, 2021). For younger cohorts, smartphone ownership in 2021 was considerably higher (96 percent, 95 percent, and 83 percent for ages 18–29, 30–49, and 50–64, respectively). This digital divide is seen in many other countries and for a wide range of digital technologies (e.g., tablets, desktop computers, and laptop computers). Although nonuse probably has many reasons, such as expense, poor design, and inadequate instructional support (Czaja et al., 2019), cognitive ageing that affects both fluid and crystallized ability levels may be a contributing factor (Czaja et al., 2006). A Pew Research survey (Anderson and Perrin, 2017) asked respondents to indicate their degree of agreement with the statement “When I get a new electronic device, I usually need someone else to set it up or show me how to use it.” Those age 65+ either strongly agreed or agreed with this statement 72 percent of the time, compared with 17 percent of the time for the 18–29-year-old cohort. As mentioned previously, strong age-related declines occur in fluid ability, which underlies novel problem solving. What is striking in Figure 5.3 is that the 50–64-year-old cohort that lagged younger cohorts considerably in 2019 and reported needing help with setting up new technology about 62 percent of the time in the Anderson and Perrin (2017) report, suddenly adopted the Internet by 2021. Most of those individuals were working when the pandemic hit, and they probably had access to technology support through work, or through younger family members living at home, whereas senior citizens living alone were much less likely to have those resources available. Nonetheless, even very old adults can successfully adopt digital technology, if properly designed and provided with adequate instructional support (Czaja et al., 2017). Most theories of technology adoption and use (e.g., Venkatesh et al., 2012), and ones aimed specifically at older adult adoption (Chen and Lou, 2020) suggest that potential users weigh costs, such as product price, perceived ease of use, and privacy concerns, against benefits, such as perceived usefulness (utility), in their decision-making process. These same models, particularly UTAUT2 (Venkatesh et al., 2012), and others based primarily on qualitative (small sample interview) data, point to a range of person characteristics and social factors that influence technology adoption. Examples include subjective health, social influence, and alternative ways to achieve goals (family members who can help), particularly in terms of early considerations about adoption, the pre-implementation stage (e.g., Peek et al., 2014). In terms of general decision-making, in a life span sample of those using smartphones (not representative of older adult cohorts), older adults were more risk averse than younger adults (Rutledge et al., 2016). Other studies have attempted to disentangle the age-ability negative relationship, showing that cognitive ability seems to be an independent predictor (beyond age) for some types of risky choice (Dohmen et al., 2010, 2018). From both lines of research, we would expect that the barrier to technology entry (accepting the risk of changing the way to carry out tasks) would be somewhat higher for older adults than for younger cohorts.

5.3.2.2

Augmentation Candidates: Speed of Processing, Working Memory, and Prospective Memory

Table 5.1 shows the largest age gaps for performance occurring in processing speed, particularly for movement, and a reduction of 20 percent in working memory capacity. Working memory 90

Technologies to Mitigate Cognitive Ageing

capacity is critical for complex tasks that require holding and manipulating information. A good example is the task of driving a vehicle. Older adult crashes tend to result from missing critical features in the stream of perceptual information or from failing to act quickly and appropriately even when those features are noticed, that is, from attention and speed of processing weaknesses. Examples would be left turn crashes in drive-on-the-right countries, failure to predict closing speed from opposing lane traffic or to notice the presence of pedestrians in a crosswalk during a turn, merging into lanes, and collisions involving missed traffic signs and signals (Boot et al., 2013c). So providing technology that can support memory and bolster perceptual capabilities could support safer driving, such as advanced driver assistance systems (ADAS). Some evidence from simulator studies indicates that ADAS systems with forward collision warnings can improve some aspects of the driving process, such as inducing greater headway to a followed vehicle in case it stops suddenly (e.g., Souders et al., 2020). Another important aspect of memory that undergoes negative age-related change is the ability to remember to do actions in the future, or prospective memory functioning (Park et al., 1997). Prospective memory involves time-based cues such as remembering to go to a healthcare appointment in the future, or event-based cues such as remembering to feed a pet that greets you when you return home from work. Many people have learned to use calendars to create physical reminders for time-based events (though this assumes that the calendar is consulted in a timely way), and more recently some have learned to employ mobile devices to set alarms for time-based appointments, with the alarm serving as an event-related cue to trigger actions to take. In work environments, setting alarms within calendaring software is a typical method for remembering appointments such as attending meetings while working on existing tasks. However, older adults who have retired from work and who never adopted electronic assistants while working may not avail themselves of such support in their home environment. Mobile technology, particularly wearables such as smartwatches, has the potential to support prospective memory by providing vibration cues about upcoming events that could prompt people to look at the watch face for more information, if older adults learn how to use and maintain those devices. For those in home environments, smart speakers (digital assistants) also have reminder setting capabilities, though in small-scale studies with older adults less than 10 percent of older owners knew about those reminder capabilities (Koon et al., 2020). That is, the potential for memory support is there, but education and training are a necessary component for utilization. Given how costly technology support services are to implement, particularly those that allow people to query live humans about problems, few manufacturers of those devices are likely to support consumers directly in learning about such features (askan-expert services). Better artificial intelligence support systems could eventually help close this gap.

5.3.3

Substitution

Substitution is likely the least preferred option for maintaining productivity and supporting independence as it is both a visible indicator of disability (as are augmenting devices such as canes and walkers) and a signal that virtually no chance exists of being made whole again through rehabilitation. Another reason why substitution is likely to be the least attractive option is that it may assign older adults to a lesser class, much like the stigma associated with disability. There are parallels here to the concerns some economists express about the role of artificial intelligence (AI) and robotics in the workplace. Although the predictions are for long-term societal gains at the cost of short-term job losses and growing inequality (Berg et al., 2018), the 91

Neil Charness

substituted workers may disproportionately bear those costs. See also Abeliansky and Prettner (this volume). Nonetheless, much as robots can substitute for humans in manufacturing operations, digital technology might also substitute for failed cognitive functioning in ageing workers and consumers. Slower information-processing rates and working memory shrinkage make drivers more prone to judgment or execution errors, as mentioned earlier. While ADAS might keep ageing drivers above a “safe driver” threshold initially, further cognitive declines, particularly those due to progressive dementia, may make driving safely impossible, even with augmentation technology. At that point, ageing people would need to substitute another human to manage transportation needs (order a taxi or other ride-hailing service). Another option would be to substitute an autonomous vehicle (AV) for an ADAS-equipped one, assuming that the person with dementia could still direct that vehicle to and from the proper location by themselves or be aided by a system that supports the last 100 meters for wayfinding (e.g., also carries a robot assistant that can walk the person to the location). Although almost a decade has passed since the initial optimistic predictions of the arrival of fully autonomous vehicles in the next 5 to 10 years (https://www.driverless-future.com/?p=323; https://www. computerworld.com/article/2492744/autonomous-cars-will-arrive-within-10-years–intel-ctosays.html), progress is being made. If we judge by prior 10-year predictions, such as Herbert A. Simon’s optimistic prediction in 1957 that a computer would become world chess champion within the next 10 years (https://www.aaas.org/birth-modern-computing), we might expect that within the next 30 years we will see AVs that perform at the level of the best human drivers. (IBM’s Deep Blue chess-playing program beat the world chess champion, Gary Kasparov, during their second match in 1998.)

5.4

Promising Technologies and Domains for Intervention

Human functioning is often categorized into a hierarchy encompassing basic to advanced capabilities: activities of daily living (ADLs) and instrumental activities of daily living (IADLs) (see Lawton and Brody, 1969). More recently, researchers have argued for the importance of enhanced activities of daily living (EADLs) for maintaining an active, healthy lifestyle (Rogers et al., 2020). Basic activities (ADLs) include self-management of tasks such as eating, dressing, grooming, toileting, transferring (e.g., into or out of a bed), and walking. Such tasks are necessary for survival and/or maintaining dignity. More advanced tasks (IADLs) include shopping, cooking, managing medications, using a phone, doing housework, driving, and managing finances. The latter are necessary to remain independent in the community. More advanced tasks (EADLs) were discussed in the context of understanding leisure activities and include socializing, pursuing hobbies, and engaging in new learning. I add work and volunteering to that category for the purposes of this discussion, given the trend (until the recent pandemic) toward later retirement by ageing workers, particularly in the United States, that is resulting in projected increases in later-life labor force participation (Toossi, 2016). Also, the transition to retirement has shifted from a complete exit from paid work to leisure (which may include volunteer work), to a situation in which retirees may participate in part-time paid work (Coile, 2018). Further, as Coile notes, despite the general trend toward later retirement, a trend for disability-driven retirement is increasing. Technology can potentially support work and leisure participation. I propose examples within a few categories of functioning (ADLs, IADLs, and EADLs) and for domains such as work and leisure. 92

Technologies to Mitigate Cognitive Ageing

5.4.1

Prevention: Boosting Cognitive Health to Support ADLs and IADLs

Cognition is critical even for basic ADLs and becomes essential for success with instrumental activities (e.g., managing finances). Failure at ADLs is often a function of physical impairments brought on by conditions such as arthritis, heart disease, and stroke, but also by dementia, a catastrophic failure of cognition. First, note that technology for basic ADLs need not be high tech. Simple additions to home environments such as grab bars and shower seats could aid an estimated 5 million older Americans with unmet needs (Lam et al., 2021). The barrier to acquisition is mostly cognitive: lacking knowledge about the tools. I focus here on prevention of impairments that lead to disability, emphasizing cognitive ageing. Preventing cognitive decline and dementia that can affect even basic ADLs has traditionally involved identifying risk factors for cognitive decline, with the most prominent one being education, and targeting such risk factors (e.g., encouraging at-risk children to graduate from high school rather than drop out). Recent decline in dementia prevalence in the United States and elsewhere has been attributed at least in part to growing levels of education in the general population (Langa et al., 2017). However, risk factor/association studies cannot assess causal roles. Most theories about why factors such as education level, job type (intellectually challenging jobs yielding less cognitive decline than routine ones), and leisure activity type (more complex activities being protective for cognition) hypothesize that such factors improve cognition by increasing brain reserve (e.g., Stern, 2012). With association data, the directionality of the relationships is unclear, for instance, whether better cognitive health leads to adoption of more challenging work and leisure activities, or whether having a better brain initially leads to greater academic achievement (e.g., attaining a university degree). Longitudinal studies have found evidence for reciprocal relationships for work complexity (Schooler and Mulatu, 2001) and for education (L¨ovd´en et al., 2020), where the initially rich tend to get access to richer learning environments and hence develop more ageing-resistant brains. However, being cognitively rich does not seem to lead to slower cognitive decline; rather, such people require a greater accumulation of age-related or disease-related brain damage to fall below a threshold where they meet the diagnostic criteria for dementia. As mentioned at the outset, training to preserve cognition, often termed brain training, has become a central focus of intervention science. However, the evidence favoring such approaches is weak, and some recent investigations, both theoretical (L¨ovd´en et al., 2010) and empirical, suggest that only long-term training extending to a year or more (e.g., Hampshire et al., 2019) might be expected to yield any benefits. Older adults do report that they are willing to engage in cognitive training (Harrell et al., 2019). However, few adults tend to adhere to long-term programs of exercise or for sound nutrition, so reason exists to be skeptical that older adults will follow through on self-reported intentions. In fact, several studies where ageing adults were paid to participate have shown significant failures in adherence over training intervals lasting from 6 weeks (Souders et al., 2016) to 3 months (Boot et al., 2013a) to 2 years for a multi-domain Finnish intervention (Turunen et al., 2019). The largest-scale investigation of videogame training (Owen et al., 2010) enrolled 52,617 registrants, age 18–60, for online training. Owen et al. (2010) found that only 11,430 completed benchmarking assessments and at least two full training sessions for a requested 6 weeks of training that required three sessions a week play for an hour (10 minutes per task for six tasks). So even adopting minimal adherence standards of completion of assessments and two full training sessions (of the 18 sessions requested) yielded almost 80 percent attrition. Such results suggest that unless cognitive training can be made highly addictive, and possibly extended to periods of a year or longer, little population-level benefit can be expected from a brain-training approach 93

Neil Charness

to mitigating cognitive decline. Note, however, that sometimes benefits of cognitive training have a long latency, as in the ACTIVE randomized clinical trial, where benefits for self-reported IADL proficiency that was not apparent immediately post training (Ball et al., 2002) surfaced 10 years later (Rebok et al., 2014). Another approach to mitigating cognitive decline has been through short-term (e.g., 6week) physical exercise interventions, usually with sedentary older adults (for a brief review, see Voss et al., 2011). An early meta-analysis (Colcombe and Kramer, 2003) indicated that the benefits of exercise over a control condition averaged about 0.5 SD units. A later study by Barnes et al. (2013) failed to find effects on cognition for either physical exercise, mental exercise, or the two combined, relative to better control conditions than those used in earlier studies. Another study investigating a combined cognitive training and exercise intervention (Roig-Coll et al., 2020) found an exercise benefit, though with no added benefit for cognitive training. However, a recent Cochrane Library meta-analytic study that estimated effect sizes for interventions by pooling across studies deemed to be high quality concluded that while aerobic training has significant effects on fitness measures for older adults, none were evident for cognitive measures (Young et al., 2015). Thus, both cognitive training (Simons et al., 2016) and exercise training (Young et al., 2015) literatures reached similar pessimistic conclusions about improving cognition with interventions. Prevention of cognitive decline appears to be difficult to achieve. Hence, I examine prospects for compensating for cognitive decline by using technology interventions to augment capabilities.

5.4.2

Rehabilitation and Augmentation for Work

A major threat to productivity at work is obsolescence of knowledge. An early study of copy machine repair workers (Sparrow and Davies, 1988) found that worker age was much less critical to performance, measured as time to the next repair for a machine, than time since worker repair training. In today’s highly competitive work environment, worker knowledge is a prized commodity, but one that can become outdated quickly in fast-moving industries unless knowledge is updated frequently. Although larger enterprises have human resource departments that can serve as hubs for worker training and retraining, smaller ones, even those in the information technology industry, lack the resources to provide training on site and cannot easily send workers to formal training programs, so they rely on workers to train themselves (Charness and Fox, 2010). More concerning, older workers receive training less frequently, often because managers do not see them as appropriate candidates (Sharit et al., 2009) irrespective of their real value (e.g., Brooke, 2003). Early research on older worker training also found that older workers seek out training less often than their younger counterparts (e.g., Belbin and Belbin, 1972) and seem to be less successful in training (Kubeck et al., 1996). Fortunately, many educational and training programs are now available online, particularly for software applications, and although older workers can be expected to learn less quickly than their younger counterparts, those with knowledge in the domain can learn efficiently (Charness et al., 2001). Technology can help mitigate knowledge obsolescence when providing just-intime learning opportunities to employees. The recent pandemic provided a strong test of this proposition with respect to work-at-home adaptations and rapid adoption of videoconferencing software (e.g., Zoom). Providing workers with effective tutorials by using intelligent tutoring systems (e.g., Anderson et al., 1985), should help counteract knowledge obsolescence. Even if one can boost crystallized abilities (knowledge) in older workers through targeted training programs, strong age-related decline in fluid abilities remains a challenge in industries when new solutions are sought. As mentioned earlier, fluid abilities tend to decline from the 20s. 94

Technologies to Mitigate Cognitive Ageing

Creative productivity, a function of both forms of cognitive ability, on average has an inverted j-shaped function, rising rapidly in the 20s, peaking in the 30s and 40s, and declining slowly thereafter (Simonton, 1997), though the decade for the peak varies by discipline. Some disciplines like mathematics and physics (perhaps relying more on abstract problem-solving ability) show peaks in the 20s, and others, like history (requiring a large knowledge store: crystallized ability), tend to peak in the 50s. However, as Simonton notes, career age seems to be a better predictor of achievement peak than chronological age, though the two variables tend to be strongly correlated. Hence, changing fields may potentially reset the clock, perhaps providing a rationale for cross-training in industry. Cognitive training can boost abilities, including fluid ones, as reviewed previously. Whether that ability increase translates to work performance improvement is unknown. A more promising approach may stem from creating opportunities for human-computer symbiosis, forming collaborations between older workers and AI programs. Much of current AI research, particularly machine learning/deep learning approaches, focuses on creating expert systems based on pattern classification. Such systems are already being used for drug discovery and for identifying ageing biomarkers (Zhavoronkov et al., 2019). For some domains with welldefined outcomes (win, loss, draw), such as Go or chess, a program can learn from competing against itself to build up pattern-based knowledge from millions of games and achieve superhuman performance (Silver et al., 2018). However, few job-related patterns are likely to be so well defined. Particularly among knowledge workers, the type of tacit knowledge necessary to steer successfully through corporate environments rarely has clear-cut definitions (Wagner and Sternberg, 1991). However, hybrid systems of humans paired with AI systems, such as in the case of ADAS in driving discussed previously, may offer the best of both worlds. Such collaborations can capitalize on the ageing worker’s relatively intact knowledge base to label category instances that are difficult to classify by machine with unsupervised learning algorithms. With a hybrid human-computer system, human knowledge can be used to label cases for classification within a semi-supervised learning algorithm. So, achieving expert performance in a broad array of domains of interest to firms may be possible.

5.4.3

Augmentation and EADL Support

Among the EADLs mentioned earlier, socialization and pursuit of hobbies are potential targets for technology interventions. Social isolation (an objective measure of interpersonal connection) and loneliness (a subjective feeling of being alone) have only recently been identified as significant threats to human health (Steptoe et al., 2013) that are roughly equivalent to smoking (Holt-Lunstad et al., 2015). Loneliness is also associated with dementia (Holm´en et al., 2000). Although older adults are not necessarily the demographic at the highest risk for loneliness (young adults usually are), their tendency to be in single-person households (e.g., widowers, widows) makes them more at risk for social isolation. Such isolation may partly be due to normative events such as loss of friends and family members in very old age (the author’s father, who died at age 99, had outlived all his former age-mate friends) or to sensory, cognitive, and mobility impairments that make leaving a dwelling difficult. As the current COVID-19 pandemic made clear, technology has an important role to play in social connection and, potentially, in providing shut-ins with engaging leisure activities (EADLs). Telephony has been a means for remote communication and socialization since the early 1900s when wired telephones moved from businesses to homes. Modern videoconferencing software adds visual information (essential for processing nonverbal communication cues) to the sound provided by telephony, also permitting those with hearing deficits to lip read, 95

Neil Charness

and providing an option for automated captioning for those with more pronounced auditory deficits. Texting capability was an early feature of mobile phones, although those with arthritic fingers (about half of the older adult population has arthritis) may struggle with it, though smartphones can provide speech recognition that allows people to dictate messages. A caveat is that older-age cohorts are less likely to be able to set up and use new technology products on their own (Anderson and Perrin, 2017), so providing adequate technology support services will be necessary for successful adoption. Randomized clinical trials have examined how providing training and access to computer technology and the Internet can support social interaction (and cognition); however, the results suggest that even with carefully designed software/hardware platforms, positive social connection and well-being benefits fade within a year (Czaja et al., 2017) or are not observed (Slegers et al., 2008). Several promising technologies to support engagement in hobbies can enhance EADLs, including social media platforms and extended reality systems. A long-standing feature of Internet access is that it provides browser-based sources that allow for participation in social games such as bridge, mahjongg, and chess online. About a quarter of older Americans are still not online, however, and those who are may be unfamiliar with or not know how to join such online activities. Considerable evidence indicates that older adults who adopt such technology lag their younger counterparts in proficiency with these tools (Boot et al., 2015; Roque and Boot, 2016, 2020). Extended reality (XR) systems, a term that encompasses virtual reality, augmented reality, and mixed reality systems, can immerse people in either solitary or social environments as a form of entertainment using virtual reality headsets that provide three-dimensional information. Such technology is still moderately expensive, particularly for seniors on fixed incomes, but is dropping in price, making it affordable to a wider set of income levels. XR represents a way for those who are housebound to “tour” virtual sites such as museums or historical sites.

5.5

Conclusions

This chapter has described some of the challenges for ageing adults and the communities that they live in, for instance, managing productivity at work and well-being at home because of age-associated cognitive declines in various abilities: some declines (fluid) starting in the 20s and others (crystallized) in the 60s or 70s. Information-processing capabilities, particularly working memory capacity and processing speed, show pronounced differences between younger and older adults (Table 5.1). The chapter has also reviewed various ways in which technologies might mitigate such cognitive declines, introducing a framework of preferring interventions that first prevent decline, second rehabilitate normative decline, or third augment declining cognition. In more extreme circumstances such as dementia, technology may substitute for failed cognitive functions. A critical challenge for deriving benefits from technology interventions is how to mitigate the digital lag in adoption that handicaps older adult cohorts. Some protective factors have been identified (that might build brain reserve capacity). Those with more educational attainment, who are employed in complex jobs or engaged in complex leisure activities, exhibit higher cognitive functioning in later adulthood than those with less education, less challenging jobs, and passive leisure activities. Richer social networks are also protective of cognition. However, the rate of decline with age appears parallel for those with such advantages and those without them, except perhaps for the case of dementia-induced decline where the advantaged people may decline more quickly. Prevention of decline using technology to train cognitive functions directly seems to help boost the specific abilities trained, though with uncertain or negligible benefit for more distant 96

Technologies to Mitigate Cognitive Ageing

outcomes such as work productivity or independent living. Augmenting cognition with technology seems like a promising approach for improving worker productivity. Technology-assisted job training could combat knowledge obsolescence, and computer-human collaborative systems may blend the best of AI and human knowledge to provide greater productivity. Finally, technology that mitigates cognitive decline can play an important role in maintaining independence in the community. A salient example is the use of ADAS to augment older driver safety, having the car warn the driver about vehicles in their blind spot, indicate when a driver drifts from their lane, or warn and brake when a collision is imminent. Portable software systems, such as smartwatches and smartphones, exhibit promise for supporting failing prospective memory functioning that might result in missed appointments or deadlines at work or remind ageing adults about meetings with healthcare providers. Portable devices could also enhance adherence to technology-based interventions by providing timely prompts for physical and mental exercise routines. Aside from supporting ADLs and IADLs, technology products such as XR systems have a role to play in EADLs, particularly in supporting hobbies and socialization. Contrary to the human genome that governs ageing processes at the cellular level, technology systems can evolve within months and years rather than centuries and millennia. When designed with attention to usability, through careful product design, creation of appropriate instructional materials, and with adequate technical support, technology systems hold promise for supporting longer, more productive, and more enjoyable lives.

Note 1 This work was supported in part by a grant from the National Institute on Ageing, under the auspices of the Center for Research and Education on Ageing and Technology Enhancement (CREATE), 4 P01 AG 17211, and by a grant from the National Institute on Disability, Independent Living, and Rehabilitation Research, under the auspices of the Enhancing Neurocognitive Health, Abilities, Networks, & Community Engagement (ENHANCE) Center, grant 90REGE0012-01-00.

References ABELIANSKY, A. L., AND PRETTNER, K. (2022): “Automation and Ageing.” In Bloom, D. E., SouzaPosa, A., and Sunde, U. (eds.): The Routledge Handbook on the Economics of Ageing. New York: Routledge, 513–524. AIYAR, S., EBEKE, C., AND SHAO, X. (2016): “The impact of workforce ageing on European productivity.” IMF Working paper, WP/16/238. https://www.imf.org/media/Websites/IMF/ imported-full-text-pdf/external/pubs/ft/wp/2016/ wp16238.ashx [Accessed on April 4, 2021]. ˚ ANDEL, R., KAREHOLT , I., PARKER, M. G., THORSLUND, M., AND GATZ, M. (2007): “Complexity of primary lifetime occupation and cognition in advanced old age,” Journal of Ageing and Health, 19(3): 397–415. DOI: 10.1177/0898264307300171. ANDERSON, J. R., BOYLE, C. F., AND REISER, B. J. (1985): “ntelligent tutoring systems.,” Science, 228(4698): 456-462. DOI: 10.1126/science.228.4698.456. ANDERSON, M., AND PERRIN, A. (2017): “Tech adoption climbs among older adults.” Pew Research Center, May 2017, http://assets.pewresearch.org/wp-content/uploads/sites/ 14/2017/05/16170850/PI 2017.05.17 Older-Americans-Tech FINAL.pdf [Accessed September 20, 2017]. BALL, K., BERCH, D. B., HELMERS, K. F., JOBE, J. B., LEVECK, M. D., MARSISKE, M., MORRIS, J. N., REBOK, G. W., SMITH, D. M., TENNSTEDT, S. L., UNVERZAGT, F. W., AND WILLIS, S. L. (2002): “Effects of cognitive training interventions with older adults: A randomized control trial,” Journal of the American Medical Association, 288(18): 2271–2281. DOI:10.1001/jama.288.18.2271. BARNES, D. E., SANTOS-MODESITT, W., POELKE, G., KRAMER, A. F., CASTRO, C., MIDDLETON, L. E., AND YAFFE, K. (2013): “The mental activity and eXercise (MAX) trial: A randomized controlled trial to enhance cognitive function in older adults,” JAMA Internal Medicine, 173(9): 797–804. DOI:10.1001/jamainternmed.2013.189.

97

Neil Charness

BASAK, C., QIN, S., AND O’CONNELL, M. A. (2020): “Differential effects of cognitive training modules in healthy ageing and mild cognitive impairment: A comprehensive meta-analysis of randomized controlled trials,” Psychology and Ageing, 35(2): 220–249. DOI: 10.1037/pag0000442. BELBIN, E., AND BELBIN, R. M. (1972): Problems in Adult Retraining. London: Heineman. BERG, A., BUFFIE, E. F., AND ZANNA, L.-F. (2018): “Should we fear the robot revolution? (The correct answer is yes),” Journal of Monetary Economics, 97(8): 117–148. DOI: 10.1016/j.jmoneco.2018.05.014. BOOT, W. R., CHAMPION, M., BLAKELY, D. P., WRIGHT, T., SOUDERS, D. J., AND CHARNESS, N. (2013a): “Video games as a means to reduce age-related cognitive decline: Attitudes, compliance, and effectiveness,” Frontiers in Psychology, 4: 31. DOI: 10.3389/fpsyg.2013.00031. BOOT, W. R., CHARNESS, N., CZAJA, S. J., SHARIT, J., ROGERS, W. A., FISK, A. D., MITZNER, T., LEE, C. C., AND NAIR, S. (2015): “Computer proficiency questionnaire: Assessing low and high computer proficient seniors,” The Gerontologist, 55(3): 404–411. DOI: 10.1093/geront/gnt117. BOOT, W. R., SIMONS, D. J., STOTHART, C., AND STUTTS, C. (2013b): “The pervasive problem with placebos in psychology: Why active control groups are not sufficient to rule out placebo effects,” Perspectives on Psychological Science, 8(4): 445–454. DOI: 10.1177/1745691613491271. BOOT, W. R., STOTHART, C., AND CHARNESS, N. (2013c): “Improving the safety of ageing road users—A mini-review,” Gerontology, 60(1): 90–96. DOI: 10.1159/000354212. BROOKE, L. (2003): “Human resource costs and benefits of maintaining a mature-age workforce,” International Journal of Manpower, 24(3): 260–283. http://www.emeraldinsight.com/0143-7720.htm. CARD, S. K., MORAN, T. P., AND NEWELL, A. (1983): The Psychology of Human-Computer Interaction. Hillsdale, NJ: Lawrence Erlbaum Associates. CHARNESS, N. (2020): “A framework for choosing technology interventions to promote successful longevity: Prevent, rehabilitate, augment, substitute (PRAS),” Gerontology, 66(2): 169–175. DOI: 10.1159/000502141. CHARNESS, N., AND FOX, M. C. (2010): Formal training, older workers, and the IT industry. In McMullin, J., and Marshall, V. W. (eds.): Ageing and Working in the New Economy: Changing Career Structures in Small IT Firms. Williston, VT: Edward Elgar, pp. 143–162. CHARNESS, N., KELLEY, C. L., BOSMAN, E. A., AND MOTTRAM, M. (2001): “Word processing training and retraining: Effects of adult age, experience, and interface,” Psychology and Ageing, 16(1): 110–127. DOI: 10.1037/0882-7974.16.1.110. CHEN, K., AND LOU, V. W. Q. (2020): “Measuring senior technology acceptance: Development of a brief, 14-item scale,” Innovation in Ageing, 4(3): 1–12. DOI:10.1093/geroni/igaa016. COILE, C. (2018): “The demography of retirement.” In Hayward, M. D., and Mamjundar, M. K., (eds.): National Academies of Sciences, Engineering, and Medicine Future Directions for the Demography of Ageing: Proceedings of a Workshop, Washington, DC: The National Academies Press, pp. 217–246. DOI: 10.17226/25064. COLCOMBE, S., AND KRAMER, A. F. (2003): “Fitness effects on the cognitive function of older adults: A meta-analytic study,” Psychological Science, 14(2): 125–130. DOI: 10.1111/1467-9280.t01-1-01430. CZAJA, S. J., BOOT, W. R., CHARNESS, N., AND ROGERS, W. A. (2019): Designing for Older Adults: Principles and Creative Human Factors Approaches (3rd Edition). Boca Raton, FL: CRC Press. CZAJA, S. J., BOOT, W. R., CHARNESS, N., ROGERS, W. A., AND SHARIT, J. (2017): “Improving social support for older adults through technology: Findings from the PRISM randomized control trial,” The Gerontologist, 58(3): 467–477. DOI: 10.1093/geront/gnw249. CZAJA, S. J., CHARNESS, N., FISK, A. D., HERTZOG, C., NAIR, S. N., ROGERS, W. A., AND SHARIT, J. (2006): “Factors predicting the use of technology: Findings from the Center for Research and Education on Ageing and Technology Enhancement (CREATE),” Psychology and Ageing, 21(2): 333–352. https://doi-org.proxy.lib.fsu.edu/10.1037/0882-7974.21.2.333. DOHMEN, T., FALK, A., HUFFMAN, D., AND SUNDE, U. (2010): “Are risk aversion and impatience related to cognitive ability?,” American Economic Review, 100(3): 1238–1260. DOI: 10.1257/aer.100.3.1238. DOHMEN, T., FALK, A., HUFFMAN, D., AND SUNDE, U. (2018): “On the relationship between cognitive ability and risk preference,” Journal of Economic Perspectives, 32(2): 115–134. DOI: 10.1257/jep.32.2.115. HAMPSHIRE, A., SANDRONE, S., AND HELLYER, P. J. (2019): “A large-scale cross-sectional investigation into the efficacy of brain training,” Frontiers in Human Neuroscience, 13: 221. DOI: 10.3389/fnhum.2019.00221. HARRELL, E. R., KMETZ, B., AND BOOT, W. R. (2019): “Is cognitive training worth it? Exploring individuals’ willingness to engage in cognitive training,” Journal of Cognitive Enhancement, 3(4): 405–415. DOI: 10.1007/s41465-019-00129-4.

98

Technologies to Mitigate Cognitive Ageing HOLM E´ N, K., ERICSSON, K., AND WINBLAD, B. (2000): “Social and emotional loneliness among non-demented and demented elderly people,” Geriatrics, 31(3): 177–192. DOI: 10.1016/S01674943(00)00070-4. HOLT-LUNSTAD, J., SMITH, T. B., BAKER, M., HARRIS, T., AND STEPHENSON, D. (2015): “Loneliness and social isolation as risk factors for mortality: A meta-analytic review,” Perspectives on Psychological Science, 10(2): 227–237. DOI: 10.1177/1745691614568352. JASTRZEMBSKI, T. S., AND CHARNESS, N. (2007): “The model human processor and the older adult: Parameter estimation and validation within a mobile phone task,” Journal of Experimental Psychology: Applied, 13(4): 224–248. DOI: 10.1037/1076-898X.13.4.224. KOON, L. M., MCGLYNN, S. A., BLOCKER, K. A., AND ROGERS, W. A. (2020): “Perceptions of digital assistants from early adopters aged 55+,” Ergonomics in Design, 28(1): 16–23. DOI: 10.1177/1064804619842501. KUBECK, J. E., DELP, N. D., HASLETT, T. K., AND MCDANIEL, M. A. (1996): “Does job-related training performance decline with age?,” Psychology and Ageing, 11(1): 92–107. DOI: 10.1037/08827974.11.1.92. LAM, K., SHI, Y., BOSCARDIN, J., AND COVINSKY, K. E. (2021): “Unmet need for equipment to help with bathing and toileting among older US adults,” JAMA Internal Medicine, 181(5): 662–670. DOI: 10.1001/jamainternmed.2021.0204. LANGA, K. M., LARSON, E. B., CRIMMINS, E. M., FAUL, J. D., LEVINE, D. A., KABETO, M. U., AND WEIR, D. R. (2017): “A comparison of the prevalence of dementia in the United States in 2000 and 2012,” JAMA Internal Medicine, 177(1): 51–58. DOI: 10.1001/jamainternmed.2016.6807. LAWTON, M. P., AND BRODY, E. M. (1969): “Assessment of older people: Self-maintaining and instrumental activities of daily living,” Gerontologist, 9(3): 179–186. DOI: 10.1093/geront/ 9.3 Part 1.179. ¨ E´ N, M., BACKMAN ¨ LOVD , L., LINDENBERGER, U., SCHAEFER, S., AND SCHMIEDEK, F. (2010): “A theoretical framework for the study of adult cognitive plasticity,” Psychological Bulletin, 136(4): 659–676. DOI: 10.1037/a0020080. ¨ E´ N, M., FRATIGLIONI, L., GLYMOUR, M. M., LINDENBERGER, U., AND TUCKER-DROB, E. LOVD M. (2020): “Education and cognitive functioning across the life span,” Psychological Science in the Public Interest, 21(1): 6–41. DOI: 10.1177/1529100620920576. MAESTAS, N., MULLEN, K. J., AND POWELL, D. (2016): “The effect of population ageing on economic growth, the labor force and productivity,” National Bureau of Economic Research, Working Paper 22452, https://www.nber.org/papers/w22452.pdf [Accessed April 4, 2021]. NEWELL, A., AND SIMON, H. A. (1972): Human Problem Solving. Englewood Cliffs, NJ: Prentice-Hall. OWEN, A. M., HAMPSHIRE, A., GRAHN, J. A., STENTON, R., DAJANI, S., BURNS, A. S., HOWARD, R. J., AND BALLARD, C. G. (2010): “Putting brain training to the test,” Nature, 465(7299): 775–779. DOI: 10.1038/nature09042. PARK, D. C., HERTZOG, C., KIDDER, D. P., MORRELL, R. W., AND MAYHORN C. B. (1997): “Effect of age on event-based and time-based prospective memory,” Psychology and Ageing, 12(2): 314–327. DOI: 10.1037//0882-7974.12.2.314. PEEK, S. T., ET AL. (2014): “Factors influencing acceptance of technology for ageing in place: A systematic review,” International Journal of Medical Informatics, 83(4), 235–248. DOI: org/10.1016/ J.Ijmedinf.2014.01.004. PEW RESEARCH CENTER. (2021): Mobile fact sheet, https://www.pewresearch.org/internet/factsheet/mobile/ [Accessed April 11, 2021]. REBOK, G. W., BALL, K., GUEY, L. T., JONES, R. N, KIM, H.-Y., . . . ET AL. (2014): “Ten-year effects of the advanced cognitive training for independent and vital elderly cognitive training trial on cognition and everyday functioning in older adults,” Journal of the American Geriatrics Society, 62(1): 16–24. DOI: 10.1111/jgs.12607. ROGERS, W. A., MITZNER, T. L., AND BIXTER, M. T. (2020): “Understanding the potential of technology to support enhanced activities of daily living (EADLs),” Gerontechnology, 19(2): 125–137. DOI: 10.4017/gt.2020.19.2.005.00. ROIG-COLL, F., CASTELLS-SANCHEZ, A., LAMONJA-VICENTE, N., TORAN-MONSERRAT, P., PERA, G., . . . ET AL. (2020): “Effects of aerobic exercise, cognitive and combined training on cognition in physically inactive healthy late-middle-aged adults: The Projecte Moviment randomized controlled trial,” Frontiers in Ageing Neuroscience, 12: 590168 DOI: 10.3389/ fnagi.2020.590168.

99

Neil Charness

ROQUE, N. A., AND BOOT, W. R. (2016): “A new tool for assessing mobile device proficiency in older adults: The mobile device proficiency questionnaire,” Journal of Applied Gerontology, 37(2): 131–156. DOI: 10.1177/0733464816642582. ROQUE, N. A., AND BOOT, W. R. (2020): “A new tool for assessing older adults’ wireless network proficiency: The wireless network proficiency questionnaire,” Journal of Applied Gerontology, 40(5):541–546. DOI: 10.1177/0733464820935000. RUTLEDGE, R. B., SMITTENAAR, P., ZEIDMAN, P., . . . LINDENBERGER, U., DAYAN, P., AND DOLAN, R. J. (2016): “Risk taking for potential reward decreases across the lifespan,” Current Biology, 26(12): P1634–P1639. DOI: 10.1016/j.cub.2016.05.017. SALA, G., AND GOBET, F. (2017): “Does far transfer exist? Negative evidence from chess, music, and working memory training,” Current Directions in Psychological Science, 26(6): 515–520. DOI: 10.1177%2F0963721417712760. SALTHOUSE, T. A. (2010): Major Issues in Cognitive Ageing. Oxford: Oxford University Press. SCHOOLER, C., AND MULATU, M. S. (2001): “The reciprocal effects of leisure time activities and intellectual functioning in older people: A longitudinal analysis,” Psychology and Ageing, 16(3): 466–482. DOI: 10.1037/0882-7974.16.3.466. SHARIT, J., CZAJA, S. J., HERNANDEZ, M. A., AND NAIR, S. N. (2009): “The employability of older workers as teleworkers: An appraisal of issues and an empirical study,” Human Factors and Ergonomics in Manufacturing, 19(5): 457–477. SILVER, D., HUBERT, T., SCHRITTWEISER, J., ANTONOGLOU, I., LAI, M., . . . ET AL. (2018): “A general reinforcement learning algorithm that masters chess, shogi, and Go through self-play,” Science, 362(6419): 1140–1144. DOI: 10.1126/science.aar6404. SIMONS, D. J., BOOT, W. R., CHARNESS, N., GATHERCOLE, S. E., CHABRIS, C. F., HAMBRICK, D. Z., AND STINE-MORROW, E. A. L. (2016): “Do “brain training” programs work?,” Psychological Science in the Public Interest, 17(3): 108–191. DOI: 10.1177/1529100616661983. SIMONTON, D. K. (1997): “Creative productivity: A predictive and explanatory model of career trajectories and landmarks,” Psychological Review, 104(1): 66–89. DOI: 10.1037/0033-295X.104.1.66. SLEGERS, K., VAN BOXTEL, M. P. J., AND JOLLES, J. (2008): “Effects of computer training and Internet usage on the well-being and quality of life of older adults: A randomized, controlled study,” Journal of Gerontology: Psychological Sciences, 63b(3): P176–P184. DOI: 10.1093/geronb/63.3.p176. SOUDERS, D. J., BOOT, W. R., CHARNESS, N., AND MOXLEY, J. H. (2016): “Older adult video game preferences in practice: Investigating the effects of competing or cooperating,” Games and Culture, 11(1–2): 170–200. DOI: 10.1177/1555412015603538. SOUDERS, D. J., CHARNESS, N., ROQUE, N. A., AND PHAM H. (2020): “Ageing: Older adults’ driving behavior using longitudinal and lateral warning systems,” Human Factors, 62(2), 229–248. DOI: 10.1177/0018720819864510. SPARROW, P. R., AND DAVIES, D. R. (1988): “Effects of age, tenure, training, and job complexity on technical performance,” Psychology and Ageing, 3, 307–314. DOI: 10.1037/0882-7974.3.3.307. STEPTOE, A., SHANKAR, A., DEMAKAKOS, P., AND WARDLE, J. (2013): “Social isolation, loneliness, and all-cause mortality in older men and women,” PNAS, 110(15): 5797–5801. DOI: 10.1073/pnas.1219686110. STERN, Y. (2012): “Cognitive reserve in ageing and Alzheimer’s disease,” Lancet Neurology, 11(11): 1006–1012. DOI: 10.1016/S1474-4422(12)70191-6. TOOSSI, M. (2016): “Labor force projections to 2024: The labor force is growing, but slowly.” Monthly Labor Review, U.S. Bureau of Labor Statistics, December 2015, DOI: 10.21916/mlr.2015.48 accessed 11/8/2021. ¨ ¨ TURUNEN, M., HOKKANEN, L., BACKMAN , L., STIGSDOTTER-NEELY, A., HANNINEN , T., PAAJANEN, T., ET AL . (2019): “Computer-based cognitive training for older adults: Determinants of adherence,” PLoS ONE, 14(7): e0219541. DOI: 10.1371/journal.pone.0219541. VERGHESE, J., LIPTON, R. B., KATZ, M. J., HALL, C. B., DERBY, C. A., KUSLANSKY, G., AMBROSE, A. F., SLIWINSKI, M., AND BUSCHKE, H. (2003): “Leisure activities and the risk of dementia in the elderly,” New England Journal of Medicine, 348(25): 2508–2516. DOI: 10.1056/NEJMoa022252. VERHAEGHEN, P. (2014): The Elements of Cognitive Ageing: Meta-Analyses of Age-Related Differences in Processing Speed and Their Consequences. New York: Oxford University Press. VENKATESH, V., THONG, J. Y. L., AND XU, X. (2012): “Consumer acceptance and use of information technology: Extending the unified theory of acceptance and use of technology,” MIS Quarterly, 36(1): 157–178.

100

Technologies to Mitigate Cognitive Ageing

VOSS, M. W., NAGAMATSU, L. S., LIU-AMBROSE, T., AND KRAMER, A. F. (2011): “Exercise, brain, and cognition across the life span,” Journal of Applied Physiology, 111(5): 1505–1513. DOI: 10.1152/japplphysiol.00210.2011. WAGNER, R. K., AND STERNBERG, R. J. (1991): “Tacit knowledge: Its uses in identifying, assessing, and developing managerial talent.” In: Jones, J., Steffy, B., and Bray, D. (eds.): Applying Psychology in Business: The Manager’s Handbook. New York: Human Sciences Press, pp. 333–344. WHITE, N., AND CUNNINGHAM, W. R. (1988): “Is terminal drop pervasive or specific?,” Journal of Gerontology, 43(6): P141–P144. DOI: 10.1093/geronj/43.6.P141. YOUNG, J., ANGEVAREN, M., RUSTED, J., AND TABET, N. (2015): “Aerobic exercise to improve cognitive function in older people without known cognitive impairment,” Cochrane Library Report, 22(4): CD00538. DOI: 10.1002/14651858.CD005381.pub4. ZHAVORONKOV, A., MAMOSHINA, P., VANHAELEN, Q., SHEIBYE-KNUDSEN, M., MOSKALEV, A., AND A LIPER , A. (2019): “Artificial intelligence for ageing and longevity research: Recent advances and perspectives,” Ageing Research Reviews, 49: 49–66. DOI: 10.1016/j.arr.2018.11.003.

101

6 GENDER, AGEING, AND HEALTH Xiaoyan Lei and Yuqi Ta

Abstract This chapter provides an overview of the gender differences in health, measured in multiple ways, of the elderly and discusses various explanations for these gender health gaps. A consistent pattern emerges in developing and developed countries of female advantage in life expectancy and survival, but disadvantage in functional health. For other health indicators, such as self-rated health, depression, and cognition, relatively consistent evidence of female disadvantage is found in less developed regions, while mixed results are found in more developed regions, with the direction and size of the gaps depending on the age group, location, and time scale under study. To date, epidemiological, biological, social, contextual, and methodological explanations have been provided, but none alone can fully explain the complexities of the gender health gaps among the elderly. Nevertheless, most studies confirm the role of socioeconomic status.

6.1

Introduction

Population ageing has been a major worldwide demographic phenomenon over the past decades, with the global share of the population aged 65 years and older increasing from 5.1 percent in 1950 to 9.3 percent in 2020 and projected to reach 16 percent by 2050 (United Nations, 2019a). Considering that the pressure of population ageing depends on the health status of older people, attention should be paid not only to the length of life but also to the quality of life, that is, whether the prolonged life of older people means a healthier life. As women universally outlive men (Barford et al., 2006), ageing wears mainly a woman’s face. Globally, the percentages of those aged 65 years and older are projected to reach 17.3 percent of women and 14.5 percent of men by 2050 (United Nations, 2019a). Although women have longer life expectancy than men, many studies show that women are less healthy, which is the male-female health survival paradox (Oksuzyan et al., 2008). However, others have criticized these “taken-for-granted” assumptions (Lahelma et al., 2001; Schmitz and Lazareviˇc, 2020) because the size and the direction of the gender health gap may change over time, vary across countries, differ across birth cohorts, and depend on the health indicators under study. Meanwhile, whether the gender differences in health persist in later life is unclear (Girgus et al., 2017). Investigating gender differences in health among older adults matters for several reasons. To start, understanding the gender health gap in old age can help us better handle the pressure of population ageing. Given that older women have longer life expectancy but fewer women are DOI: 10.4324/9781003150398-7

102

Gender, Ageing, and Health

healthy in the remaining life span (Santosa et al., 2016), a huge care cost and medical burden for families and societies will arise if we ignore gender differences in health among older adults. In addition, exploring gender differences in health can help explain and reduce social inequality in multiple aspects, including gender, race, and social groups. Therefore, the gender health gap provides an important perspective for comprehensively understanding social inequalities, which may further provide insights into lessening social inequalities in health. Finally, increasing our understanding of the mechanisms underlying the gender health gap among older adults can not only help policymakers select proper interventions and adjust welfare provisions to correspond to the needs of the elderly but also ensure that policies do not unintendedly aggravate or contribute to gender health disparities. This chapter examines whether the gender health gap persists into old age and, if so, what factors contribute to the gender difference among the elderly. A review of recent relevant studies finds that the direction and magnitude of the gender health gap among older adults depend on the location, time scale, age group, birth cohort, and health indicator under study. Meanwhile, researchers have tried to explain the gender health gap through biological factors, social factors, or contextual factors, or by reporting heterogeneity, none of which solely explains the complexities of the gender health gap documented in empirical studies. The rest of the chapter is organized as follows. The following section describes the gender health gap in old age, covering a wide range of health indicators. The next section investigates explanations for gender differences in health among older adults. In both sections, the analyses are based on reviews and discussion of the relevant literature. The final section summarizes the main findings of this study and provides implications for future research.

6.2

Gender Differences in Health among Older Adults

According to the World Health Organization (WHO, 1946), health is a state of complete physical, mental, and social well-being and not merely the absence of disease or infirmity. Thus, a comprehensive evaluation of health should be based on multiple dimensions, including life span (e.g., life expectancy), physical health (e.g., physical functioning and the prevalence of diseases), mental health (e.g., depression), cognitive health, and subjective well-being [e.g., self-rated health (SRH)]. In addition, when talking about gender health gaps, demographic transition, especially fertility transition, which plays a significant role in women’s lives, should be considered. Previous studies show that fertility rates may be correlated with women’s socioeconomic status, such as education (Jain, 1981; McCrary and Royer, 2011) and labor force participation (Ahn and Mira, 2002), which can exert an influence on female health (Subbarao and Raney, 1995). Because low (high) fertility rates usually characterize developed (developing) regions (United Nations, 2019a), for each health dimension, we separately summarize these gender differences in more developed regions and less developed regions to examine whether different patterns of gender health gaps in old age exist in regions at different stages of development.1

6.2.1

Life Spans

Women live longer than men, on average. This evidence in the literature is consistent with two major indicators: the sex ratio and life expectancy at certain ages. The sex ratio, the number of men per 100 women, reflects whether the number of males and females are balanced in a country or region. To some extent, the sex ratio in old age depicts gender differences in survival status among the elderly. As Figure 6.1a shows, the sex ratios in more developed regions are smaller than those in less developed regions, with a U-shaped trend for more developed regions and an increasing trend for less developed regions from 1950 to 103

Xiaoyan Lei and Yuqi Ta

Figure 6.1 Sex ratios in more and less developed regions, males per 100 females. Source: Author’s compilation of data from United Nations (2019a).

2020. For example, the sex ratio for age 60 and older in more developed regions first declined from 75.2 in 1950 to 64.6 in 1965, then it rose to 77.9 in 2020; in less developed regions, it increased from 87.9 in 1950 to 89.2 in 2020. Considering the sex ratio across age groups, more developed and less developed regions undergo a declining trend with ageing (Figure 6.1b). For example, in more developed regions, the sex ratio for ages 0–14 is 105.3, lower for ages 60 and older (77.9), and lowest for ages 90 and older (39.4) in 2020. Life expectancy in old age, which is the average number of years from a certain age an older person expects to live, indicates the elderly’s survival status. In more and less developed regions, between 1950 and 2020, the trend in the life expectancy of older women and older men was increasing, with larger growth scales for older women and in less developed regions. For instance, at age 60, the life expectancy of women in less developed regions rose from 13.00 years in 1950 to 20.82 years in 2020, that is, the growth scale was about 60.20 percent, which was larger than that for men in less developed regions (58.53 percent) and women in more developed regions (40.21 percent) (Figure 6.2). Older women are expected to live longer than older men (the gender gap in life expectancy is positive) in both regions, but the gap in more developed regions is larger than that in less developed regions, with an inverted U-shaped trend for more developed regions and a growing trend for less developed regions. Taking the gender gap in life expectancy at age 60 as an 104

Gender, Ageing, and Health

Figure 6.2 Trends in life expectancy at age 60 by gender in more and less developed regions, years. Source: Author’s compilation of data from United Nations (2019a).

example, the gap in more developed regions first rose from 2.52 years in 1950 to 4.30 years in 1995 and then declined to 3.86 years in 2020, whereas the gap in less developed regions increased from 1.53 years in 1950 to 2.64 years in 2020 (Figure 6.3). Comparing the gender gap across age groups, the gap is small in older age groups in more and less developed regions. For instance, in 2020, the gender gap in life expectancy at age 80 in more developed regions is 1.52 years, which is 2.34 years less than that at age 60 (Figure 6.3).

6.2.2

Physical Health

Despite their longer life span, older women are paradoxically disadvantaged in physical functioning in more and less developed regions. Many evaluations find that older women on average live more years with functional limitations, and the gap may increase with age in developed countries (Arber and Cooper, 1999; Gorman and Read, 2006; Carmel, 2019; see the review in Leveille et al., 2000). For example, Crimmins et al. (2011) examine gender differences in functioning problems and disability among people aged 50 and older in 11 European countries, England, and the United States in 2004. They find that women in all these countries are more likely than men to have difficulties with instrumental activities of daily living and functioning problems. Using data from the Survey of Health, Ageing and Retirement in Europe (SHARE) 105

Xiaoyan Lei and Yuqi Ta

Figure 6.3 Trends in gender differences in life expectancy at age 60 and at age 80 in more and less developed regions, years. Note: Gaps are female life expectancy minus male life expectancy. Source: Author’s compilation of data from United Nations (2019a).

in 2015 (wave 6), Schmitz and Lazareviˇc (2020) also find that for ages 50 and older, the prevalence of activity limitations is greater in women in every age group (50–64, 65–79, and 80+) compared with men, and the gender gap increases with age at the expense of women. Similarly, for less developed regions, various studies show that older women are more likely to struggle with physical functioning, more likely to have functional dependency, and more likely to have a higher proportion of life expectancy with disability than their male peers (Huang et al., 2013; Jiao et al., 2021; Kaneda et al., 2009; Rotarou and Sakellariou, 2019; Roy and Chaudhuri, 2008; Zhang et al., 2015), although the sizes of these gender gaps vary across countries (Chirinda and Chen, 2017; Payne et al., 2017) and differ by birth cohort. For example, using 13-year nationwide survey data, the Chinese Longitudinal Healthy Longevity Survey (CLHLS), with respondents born between 1900 and 1945 in China, Jiao et al. (2021) find that the level of disabled life expectancy and its proportion in life expectancy for older women are higher than for older men at all ages, and the gender differences diminish with ageing. They also find that compared with earlier birth cohorts, later birth cohorts experience smaller gender gaps in proportion of disabled life expectancy in life expectancy when they reached the same age. A cross-national comparative study using data from wave 1 of the WHO’s Study on Global Ageing and Adult Health (SAGE) documents that older women have a higher proportion of 106

Gender, Ageing, and Health

life expectancy with moderate or severe disability and a lower proportion of life expectancy without disability than older men across six countries (China, Ghana, India, Mexico, Russia, and South Africa). These gender gaps may be more apparent in some countries than in others; for example, India has the largest gender difference in the proportion of life expectancy with severe disability at age 50 years (21.74 percentage points), while China has the lowest gap (3.46 percentage points) (Chirinda and Chen, 2017). Gender differences in the prevalence rates of chronic conditions among older adults are not so straightforward as life expectancy and functioning limitations. For chronic conditions, the direction and size of the gaps may vary across specific diseases and differ across the countries under study. In more developed regions, one strand of literature shows mixed results. For example, using data from the National Health Interview Survey (NHIS) from 1986 to 2001 in the United States, Case and Paxson (2005) find that older women suffer more from headaches, arthritis, bronchitis, asthma, lung problems, hypertension, vision problems, and reproductive cancers. Older men are more likely to report diabetes, cardiovascular disease, respiratory cancer, emphysema, and hearing loss. However, Schmitz and Lazareviˇc (2020) show that the gender gaps in diabetes and heart attacks are small and not statistically significant in most European countries, and a reverse gap disadvantages older women in diabetes in the oldest group in some European countries. Gorman and Read (2006) document that older men in the United States from 1997 to 2001 reported more life-threatening chronic medical conditions than older women, such as heart disease, emphysema, stroke, diabetes, and cancer, but the size of the difference was modest. Crimmins et al. (2011) observe that gender differences in stroke and diabetes were inconsistent in the United States, England, and 11 European countries in 2004. Ntritsos et al. (2018) conducted a systematic review and meta-analysis of the gender-specific prevalence of chronic obstructive pulmonary disease (COPD) globally. They find that although the prevalence of COPD is higher in men than their female peers across various regions of the world and different age groups (15–39, 40–69, 70+), the gaps are often not significant in developed and high-income countries. Turning to less developed regions, research observes that older women are more likely to suffer from some chronic diseases than older men, such as hypertension (e.g., Huang et al., 2013; Teh et al., 2014; Wandera et al., 2015; Zhang et al., 2015), heart disease (e.g., Wandera et al., 2015; Zhang et al., 2015), arthritis (e.g., Liu et al., 2017; Teh et al., 2014; Zhang et al., 2015), angina, and vision problems (e.g., Zhang et al., 2015). Older men have higher prevalence of other chronic illnesses, like hearing loss (e.g., Zhang et al., 2015), asthma (e.g., Teh et al., 2014), stroke, and lung disease (e.g., Zhang et al., 2015; Liu et al., 2017). For diabetes, no significant gender gaps are found in some developing countries, such as Malaysia (Teh et al., 2014), Uganda (Wandera et al., 2015), and China (Liu et al., 2017).

6.2.3

Mental Health

Depression symptoms, an indicator reflecting an individual’s mental health status, are commonly measured using the Center for Epidemiologic Studies Depression scale (CES-D), or the Geriatric Depression Scale (GSD). In more developed regions, mixed results emerge for gender gaps in depression. Some studies find that elderly women have significantly worse mental health than elderly men (e.g., Acciai and Hardy, 2017; Angelini et al., 2019; Crimmins et al., 2011; Imai et al., 2015; Schmitz and Lazareviˇc, 2020); some studies show the opposite, that is, elderly men suffer more from depression (e.g., Djukanovi´c et al., 2015; Lim et al., 2014); and other studies find no gender difference in depression among the elderly (e.g., Canoui-Poitrine et al., 2016; Forlani et al., 107

Xiaoyan Lei and Yuqi Ta

2014; Russell and Taylor, 2009). Lim et al. (2014) and Forlani et al. (2014) conduct analysis based on small samples (about 300 observations), so we should be cautious about their results. In less developed regions, where authors who do not find significant disadvantage among women in mental health are the exception (e.g., Zhang and Li, 2011), most studies show consistent evidence that older women suffer from higher levels of depressive symptoms than older men. A strand of research on China finds that the prevalence of depressive symptoms is significantly greater in older women, based on different data sets and the CES-D or GSD measure (Chan et al., 2012; Guo et al., 2017; Huang et al., 2013; Lei et al., 2014b; Li et al., 2014; Li et al., 2015; Chen and Fang, 2020). For instance, using data from the China Health and Retirement Longitudinal Study (CHARLS) with respondents ages 45 and older, Lei et al. (2014b) document that Chinese women had higher CES-D scores than their male counterparts in 2011– 2012, and the positive age/cohort gradient in depression symptoms was steeper for women than men. Studies on other developing countries also find female disadvantage in mental health (e.g., Brinda et al., 2016; Guerra et al., 2016; Sinha et al., 2021), although the magnitudes of the gaps might differ across countries. For instance, Brinda et al. (2016) find that older women have significantly higher prevalence of geriatric depression than older men in China, Ghana, India, Mexico, Russia, and South Africa. Another cross-national study also observes higher prevalence of depression in older women in China, Cuba, India, Mexico, Nigeria, Peru, Puerto Rico, and the Dominican Republic (Guerra et al., 2016).

6.2.4

Cognitive Health

Cognitive ability is an important factor for decision making, including for decisions related to health outcomes and finance. It is usually measured by multiple indicators, such as episodic recall, intact mental status, verbal ability, visuospatial ability, and Mini Mental State Exam (MMSE) scores. For more developed regions, the direction of gender gaps in cognition among the elderly depends on the indicators under study. Various studies find that older women perform better than men on episodic recall, facial and verbal recognition, and semantic fluency (e.g., De Frias et al., 2006; Maller et al., 2007; Munro et al., 2012), whereas older men outperform older women on visuospatial tasks (e.g., De Frias et al., 2006; Munro et al., 2012). A few studies on developed countries observe no gender difference in dementia incidence (e.g., Ruitenberg et al., 2001). As for less developed regions, the evidence of gender gaps in cognitive health in old age consistently appears in the literature. A series of studies on Chinese elderly has documented women’s disadvantages in cognitive ability over time by using multiple data sets. For example, Zhang (2006) uses two waves (1998 and 2000) of the CLHLS to estimate models of cognitive impairment (i.e., MMSE score lower than 18) for a nationwide sample of individuals ages 80 and older. The study finds that older women in China are at higher risk of cognitive impairment than their male counterparts. Using data from another nationwide representative survey in China, CHARLS, Lei et al. (2012) and Lei et al. (2014a) find that Chinese women ages 45+ scored much lower in episodic memory and intact mental status than their male peers in 2008 and 2011–2012, and such gender gaps grew among older cohorts. However, they document that gender disparity in cognitive abilities no longer exists among younger cohorts in China as the economy of China is growing rapidly. Gender differences in cognitive health can also be observed in other developing countries, such as India (Kumar et al., 2020; Lee et al., 2014; Oksuzyan et al., 2018; Weir et al., 2014; Xu et al., 2019), Indonesia (e.g., Strauss et al., 2018), South Africa (e.g., Harling et al., 2020), and Colombia (e.g., Mejia-Arango et al., 2021). For instance, Xu et al. (2019) use data from SAGE 108

Gender, Ageing, and Health

that includes older adults ages 50 years and older from China and India. They confirm that the disadvantage among women in cognitive function is consistent in these two countries and across different migration groups, measured by digit span tests, immediate and delayed recall tests, and a verbal fluency test.

6.2.5

Subjective Well-Being

SRH is a measure of subjective well-being. Although it is seemingly subjective, SRH may capture more information that other people (doctors, researchers, and so forth) do not know but the individual himself/herself recognizes and thus may be more complete. The gender gap in this dimension among elderly people also differs across countries and changes over time (Carmel, 2019). In more developed regions, a strand of research on the gender gaps in SRH shows mixed results. Crimmins et al. (2011) find that older women consistently reported worse SRH than older men across 13 countries (11 European countries, England, and the United States) in 2004. However, although Dahlin and H¨ark¨onen (2013) also document that elderly women report significantly worse SRH in Eastern and Southern Europe, differences were few or nonexistent in Estonia, Finland, and Great Britain. Likewise, Adjei et al. (2017) find significant gender gaps in SRH in Germany, Italy, and Spain, but no gender gap in the United States and the United Kingdom. A recent study on 16 European countries observes that Southern Europe stands out with a large gap to the disadvantage of older women for poor SRH, and Northern Europe had a rather small gender gap in 2015 (Schmitz and Lazareviˇc, 2020). Happiness and life satisfaction are also indicators that measure subjective well-being. On the gender gap, Stevenson and Wolfers (2009) observe that in the United States, women were more likely than men to report being “very happy” in the 1970s, but by 2006, women’s subjective well-being became lower than their male counterparts though the lives of women have improved over the past years by many objective indicators. Moreover, by implementing a crossnational study, Tesch-R¨omer et al. (2008) indicate that the sizes of the gender differences in life satisfaction and subjective health vary with the extent of gender inequality in different countries. Specifically, a negative correlation exists between the gender gap in subjective well-being and the extent of gender inequality for countries where gender inequality on the labor market is widely accepted, while a positive relationship emerges for countries where gender inequality on the labor market is widely rejected. Turning to less developed regions, most studies find that older women report a lower average level of SRH than older men (e.g., Boerma et al., 2016; Bora and Saikia, 2015; Huang et al., 2013; Roy and Chaudhuri, 2008; Zhang et al., 2015) and observe female disadvantage in quality of life (e.g., Lee et al., 2020). For instance, compared with the United States, where women’s disadvantage in SRH disappears at older ages in the early 60s (Case and Paxson, 2005), this gender gap persists into very old age in China (Zhang et al., 2015). Boerma et al. (2016) also document that women consistently reported poorer health than their male counterparts in other regions, such as Sub-Saharan Africa, Latin America, and South Asia. Using data from WHO’s SAGE wave 1, Lee et al. (2020) confirm that older men are more likely to have better quality of life than older women in China, Ghana, India, Russia, and South Africa.

6.2.6

Health Impact of Birth Control

Another important but understudied health aspect is gender differences in birth control. Its importance stems from the social and economic benefits of allowing women to decide whether and when to have children. Studies that examine the effects of contraception find that reliable 109

Xiaoyan Lei and Yuqi Ta

access to contraception can improve women’s educational attainment (Ananat and Hungerman, 2012) and labor market outcomes (Bailey et al., 2012), reduce the probability of woman living in poverty (Browne and LaLumia, 2014), and have a positive effect on the next generation (Bailey et al., 2014, 2019; Bernstein and Jones, 2019), while unplanned births may decrease women’s reported levels of happiness and be connected to depression and anxiety among women (Sonfield et al., 2013). Research also points out that women currently still bear most of the healthrelated risks, financial burden, and inconveniences of contraception (Campo-Engelstein, 2012): Female methods (barrier methods, hormonal methods, and long-acting reversible contraception) are often more expensive and have more side effects than male methods (condoms and vasectomy), and women have to spend time and effort to visit doctors to obtain contraception, acquire knowledge about the effectiveness and safety of contraception methods, and feel stress about the possibility of unplanned pregnancy. Consequently, gender differences in birth control may lead to female disadvantage in physical and mental health, which might have an enduring impact on female health in old age. How to improve the efficacy and safety of the existing female methods and make males share the responsibility for contraception should be emphasized. In summary, agreement exists that women have longer life spans than men, with female disadvantage in functioning. But inconsistency exists in the direction and magnitude of the gender gaps in other aspects, including the prevalence of chronic diseases, mental health, cognitive health, and subjective well-being, which differ by indicators, vary across countries, and change over time. Contraceptive injustice is also a potential channel of female disadvantage in health in old age. Most of the cited literature uses data from nationally representative surveys with large sample sizes and scientific designs, such as SHARE, HRS, NHIS, WHO’s SAGE, CHARLS, and CLHLS; therefore, they provide relatively reliable estimations.

6.3

Explanations for Gender Health Gaps among Older Adults

As a multifaceted phenomenon, health is not only an individual’s responsibility, but also affected by decisions made at multiple levels, such as family, work, community, and government (Bird and Rieker, 2008). Researchers have tried to explain gender health gaps by referring to various biological, social, and contextual factors that differentiate the lives of women and men (Read and Gorman, 2010). These contributors might be intertwined and framed in terms of differential “vulnerability” (women and men have varying reactions to similar risk factors) and differential “exposure” (women and men have different levels of contact with risk factors) (Kaneda et al., 2009; see the discussion in Denton et al., 2004). Among the explanations that have been put forth, relatively consistent evidence arises in socioeconomic status (SES), and contextual factors may explain why some gender health gaps differ across countries to some extent, whereas mixed results often appear for other factors, like biological factors, behavioral factors, and reporting heterogeneity. The complexity of gender differences in health among older adults means that neither epidemiological and biological explanations nor social and contextual factors alone can provide complete insights. By contrast, a combination of these explanations may better enhance our understanding of the underlying mechanisms behind gender health gaps.

6.3.1

Epidemiological Explanations

To explain why women have longer life expectancy but are more likely to be poor in other dimensions of health than men, a strand of literature has looked at epidemiological reasons (e.g., Case and Paxson, 2005; Verbrugge, 1989; Zhang et al., 2015). These explanations tend to 110

Gender, Ageing, and Health

explain the so-called male-female health survival paradox by gender differences in the distribution of chronic conditions; that is, women tend to suffer from nonfatal chronic conditions, while men tend to suffer from life-threatening chronic diseases. For instance, using an Oaxaca-like decomposition, Case and Paxson (2005) examine the relationship between chronic conditions and SRH of people ages 18–84 in the United States. They observe that female disadvantage in SRH can be entirely explained by differences in prevalence rates of chronic diseases between women and men. Similarly, another study on China finds that differences in the distribution of chronic conditions and health functioning among older women and men explain about 69 percent of the gender gap in SRH among older adults (Zhang et al., 2015). Research on some countries confirms that epidemiological explanations account for the gender differences in health to some extent, especially in SRH. However, this explanation does not address why gender differences exist in the distribution of health conditions, and it fails to explain why gender health gaps change over time or vary across countries. To explore the deeper causes of gender health gaps, more scholars have turned to biological, social, and contextual factors.

6.3.2

Biological Factors

Biological explanations argue that gender differences in physiological systems and genetic structure can explain gender health gaps. These explanations focus on the roles of hormones, the immune system, sex chromosomes, and the cerebral cortex. For example, estrogen protects women from the risk of heart disease by lowering the circulation of harmful cholesterol and provides a protective effect on verbal episodic memory in women (Verghese et al., 2000). Testosterone makes men at higher risk of life-threatening conditions by causing immunosuppression (Owens, 2002). In addition, women have a more robust immune system, which makes them less likely to die from infectious and parasitic diseases but puts them at greater risk of autoimmune diseases. Women have two X chromosomes, which may be associated with lower mortality and longer life spans than men, who only have one (Austad, 2006; Oksuzyan et al., 2008). On the cerebral cortex, research has associated gender gaps in cognitive abilities with the gender difference in the composition of gray matter (associated with mathematic skills) and white matter (associated with verbal skills), with men having about 6.5 times the amount of gray matter activated than women, and women having about 10 times the amount of white matter activated than men in IQ tests (Haier et al., 2005). Higher prevalence of dementia related to cognitive impairment is observed in older women, which results from greater prevalence of white matter disease in women than in men with ageing (De Leeuw et al., 2001). Although biological explanations may explain gender differences in health to some extent, especially for the female advantage in life expectancy, they fail to explain why some gender health gaps persist into old age and might differ across countries (Schmitz and Lazareviˇc, 2020). For instance, older women have lower levels of estrogen, which has a positive impact on women’s verbal episodic memory. Thus, we would expect smaller gender gaps in verbal recognition in menopausal age groups than in younger age groups. However, the results from De Frias et al. (2006) do not support this. Similarly, considering that men’s relatively lower volume of white matter activated in IQ tests and that air pollution mainly reduces the density of white matter (Wilker et al., 2015), we would expect air pollution exposure to have a more negative effect on men in verbal test scores. A study using nationally representative longitudinal survey data with respondents ages 25 and older in China confirms this hypothesis (Zhang et al., 2018). However, a study from the Republic of Korea finds that older women have higher risk than older men for decreased cognition with increased exposure to air pollution (Kim et al., 2019). Another study 111

Xiaoyan Lei and Yuqi Ta

from Mexico observes insignificant differences by gender among older adults in the relationship between air pollution and cognitive function (Salinas-Rodr´ıguez et al., 2018).

6.3.3

Social Explanations

Social explanations are the most frequently discussed factors when explaining gender health gaps. These explanations indicate that health disparities between men and women accrue from gendered access to health-related resources and differential exposure to health-related risks (Schmitz and Lazareviˇc, 2020). Hence, such explanations mainly focus on two aspects: socioeconomic factors and health behaviors.

6.3.3.1

Socioeconomic Factors

Of all the socioeconomic factors, SES, which is normally measured by education, occupation, or income, is the most widely documented. First, higher levels of education, prestigious occupations, or higher income may give an individual better access to the resources necessary for preventing or curing diseases, resulting in better health. In many developing countries, women are more poorly educated (e.g., Lei et al., 2012); less likely to be promoted and hold key positions (e.g., Hospido et al., 2020); more likely to work part time, with the double burden of unpaid household work and paid work (Backhans et al., 2007); and receive lower wages (Blau and Kahn, 2017; Read and Gorman, 2010) than their male counterparts. These female disadvantages in SES might lead to female disadvantage in health. Second, SES might indirectly affect an individual’s health through psychosocial characteristics, because low SES may be related to increased stress, low levels of self-esteem, and low perceived control (Denton et al., 2004). This may partly explain the female disadvantage in some health indicators. Various studies confirm the important role of SES in explaining gender health gaps among older adults. Some studies find that education is associated with gender gaps in SRH (e.g., Adjei et al., 2017; Ross et al., 2012; Zhang et al., 2015), depression (e.g., Li et al., 2014; Schmitz and Lazareviˇc, 2020), and cognitive health (e.g., Lei et al., 2012; Lei et al., 2014a; Xu et al., 2019; Casanova and Aguila, 2020). For instance, Zhang et al. (2015) observe that sociodemographic factors explain about 31 percent of gender differences in SRH in China among the elderly, and about 15 percent of the gap can be explained by education. For developed countries, Adjei et al. (2017) document that educational attainment is a large contributor to gender differences in SRH in Germany, Italy, and Spain, with contributions ranging from about 10 percent to 18 percent. Likewise, Casanova and Aguila (2020) find that among Mexican immigrants residing in the United States, after accounting for socioeconomic factors, the cognitive function of male Mexican immigrants is not significantly different from that of male non-Hispanic whites, whereas the cognitive scores of female Mexican immigrants remain significantly lower than those of female non-Hispanic whites. Other studies argue that marital quality (e.g., Umberson and Williams, 2005); childhood SES, like the number of rooms, books, and facilities in the home and the occupation of the breadwinner (e.g., Angelini et al., 2019); and other socioeconomic factors, like social cohesion and household burden (e.g., Stevenson and Wolfers, 2009) also partly explain older women’s disadvantage in health. Apart from SES, some studies refer to social support and living arrangements when discussing predictors of gender gaps in health among older adults, but the evidence on this point is inconclusive. For instance, given that social support is often seen as an important factor that protects people against depression (George et al., 1989), and most of the literature observes female disadvantage in mental health in old age, we would assume that older women have less social support 112

Gender, Ageing, and Health

than older men. However, many studies document that compared with older men, older women have more social contact and participation and more extensive social networks (e.g., Choi and Ha, 2011; Park et al., 2013). Meanwhile, a few studies find that older men perceived similar or higher availability of social support compared with their female counterparts (e.g., Antonucci et al., 2002; Glei et al., 2013). These contradictory results may accrue from differentials in the definitions of social support, which include an emotional form and an instrumental form (Girgus et al., 2017). On living arrangements, because women tend to live longer than men, older women are more likely to become widowed or live alone (Zunzunegui et al., 2007), which might result in higher prevalence of depression. Nevertheless, some studies observe that older men who live alone are more likely to have depression than older women who also live alone (Oh et al., 2015; Russell and Taylor, 2009).

6.3.3.2

Behavioral Factors

Behavioral explanations offer reasons for gender health gaps based on differentials in women’s and men’s lifestyle activities, which may protect or harm their health. For example, some studies document that men tend to have risky and harmful health behaviors, like heavy smoking and hazardous drinking, while women seek healthcare services more frequently and have healthier eating habits, such as consuming less salt and meat and more fruits and vegetables (Sch¨unemann et al., 2017). With a few exceptions, a strand of the literature has found evidence that gender differences in behaviors can account for gender health gaps in later life to a certain extent (e.g., Adjei et al., 2017; Carmel, 2019; Denton et al., 2004; Lei and Liu, 2018; Santosa et al., 2016; Sch¨unemann et al., 2017; Whitson et al., 2010). For instance, some studies in developed countries observe that the increased percentage of women adopting unhealthy lifestyles, like drinking and smoking, may explain the shrinking gaps in life expectancy (Carmel, 2019). And the higher prevalence of the obesity epidemic among older women than their male peers may partly explain the gender gap in functioning (Whitson et al., 2010; Crimmins et al., 2011). In addition, Adjei et al. (2017) document that the gender differential in the allocation of time between leisure and work accounts for much of the gender gap in SRH among the elderly. Turning to developing countries, some studies find that older women are more likely to have sedentary behavior than their male counterparts in low- and middle-income countries. Sedentary behavior leads to lower muscle strength and bone density and higher levels of body fat, thus causing relatively more women to develop more disability (Santosa et al., 2016). Moreover, Lei and Liu (2018) observe that differential behavioral changes at retirement can partly explain gender differences in cognitive abilities, with the beneficial effects of retirement on cognition being stronger for male blue-collar workers, who have a more active lifestyle in retirement. Yet, a study on the gender gap in SRH across 59 countries for people ages 18 years and older finds no effect of behavioral factors, such as smoking and alcohol use, between men and women (Boerma et al., 2016).

6.3.3.3

Gender Data Gap

The gender data gap arises from lacking female samples or underrepresentation of women in medical randomized clinical trials (RCTs), especially in RCTs related to cardiovascular diseases (WHO, 2016). The main reasons for underrepresentation or exclusion of female samples are controlling for variation in female hormonal cycles (Vitale et al., 2017) and lower willingness of women to be enrolled due to misperception of risk of diseases or difficulties in terms of support for the follow-up visits or transportation (Stramba-Badiale, 2010). A similar data gap 113

Xiaoyan Lei and Yuqi Ta

is evident for the elderly, resulting from difficulties in screening and retaining elderly patients in RCTs (Vitale et al., 2017). The failure of RCTs to enroll sufficient female (elderly) samples might lead to misdiagnosis, inadequate treatment, or adverse drug effects for women, particularly for elderly women, and thus can be regarded as an explanation for the gender health gap in older age. Take cardiovascular diseases as an example. Cardiovascular diseases are often seen as a male problem, despite the fact that they have been one of the major causes of mortality for women (WHO, 2020). Considering the fluctuations in sex hormones and difficulties in management, women (elderly) are usually excluded from or underrepresented in RCTs, particularly in studies of ischemic heart disease, heart failure, and cholesterol-lowering therapy (Stramba-Badiale, 2010). However, the variation of sex hormones is exactly what drives the differential response of women’s physiological system to cardiovascular drugs and treatments, and the ageing process is what distinguishes the elderly from young adults, which imply that extrapolating the safety and effectiveness of drugs and treatments to women and the elderly may not be possible. As a result, these data gaps lead to physicians offering suboptimal treatment and drugs to underrepresented patients, or even exposing them, especially elderly women, to unintended risks (Vitale et al., 2017). For instance, more adverse effects of procedures or drugs have been found in women than men (Stramba-Badiale, 2010), and female gender has become an independent risk factor for heart surgery survival (Eifert et al., 2014).

6.3.4

Contextual Factors

Apart from individual-level factors, such as biological processes, SES, and health behaviors, contextual factors might also shape gender differences in health. This is because individuals are trapped in broader cultural and political contexts, and these contextual influences (e.g., community/neighborhood actions and social policy) may have effects that vary by gender through different levels of exposure or/and vulnerability (Bird and Rieker, 2008; Read and Gorman, 2010). Research so far has had mixed results on the differential significance of residential contexts for the health of women and men (Stafford et al., 2005). However, some studies focusing on the elderly find that residential contexts are associated with gender gaps in health. For instance, Dahlin and H¨ark¨onen (2013) examine whether gender gaps in health are associated with levels of gender equality (measured by the gender inequality index) and overall social development (measured by the human development index) in each European country. They observe larger gender gaps in SRH in countries with more gender inequality or lower levels of social development, but they find insignificant results for longstanding illness. Turning to developing countries, a series of studies confirms the role of contextual factors in gender health gaps among the elderly. For example, using the CHARLS Pilot Survey data (2008), Lei et al. (2012) find that gender differences in cognition in China are concentrated within and related to poorer communities (measured by per capita expenditures), where strong economic incentives to favor boys at the expense of girls are observed in their educational outcomes and nutrition. Further, Lei et al. (2014a) use the first national wave of CHARLS data (2011–2012) to investigate how the historical, geographical, and cultural characteristics of communities affect the cognitive ability of the Chinese elderly. They document that the improvement in infrastructure (e.g., availability of electricity) and growth of the green coverage ratio are conducive to reducing women’s disadvantage in cognition. In addition, Lei and Liu (2018) find that the gender difference in mandatory retirement ages in China—that is, Chinese women are forced to retire at a younger age than their male 114

Gender, Ageing, and Health

counterparts—is a possible mechanism underlying the gender differential in the effects of retirement on cognition, with a negative effect for older women and a positive effect for their male counterparts.

6.3.5

Methodological Explanations

Because many studies on gender differences mainly use self-reported survey data, differentials in health reporting may be another reason for the observed gender health gaps, which is called a methodological explanation (Schmitz and Lazareviˇc, 2020). To be specific, such explanations assume that women tend to be more health conscious, more likely to seek medical services (Acciai and Hardy, 2017), and more likely to admit vulnerability and seek help (Courtenay, 2000). Thus, women might have different cut-points for moving between health categories when reporting health status compared with men, forming reporting heterogeneity (Lindeboom and Van Doorslaer, 2004). Empirical studies support or cast doubt on methodological explanations (e.g., Acciai and Hardy, 2017; Bago d’Uva et al., 2008a,b; Case and Paxson, 2005; Grol-Prokopczyk et al., 2011; Oksuzyan et al., 2019; Zhang et al., 2015). For example, Grol-Prokopczyk et al. (2011) find that older women in the United States tend to report better SRH than older men, and the gender difference in SRH disappears when adjusting women’s health optimism by using anchoring vignettes. Oksuzyan et al. (2019) use wave 1 of SHARE to examine whether differential reporting can explain gender differences in general health among older adults in 11 European countries. They find no clear and significant gender-specific patterns in reporting poor or good health. Acciai and Hardy (2017) also find that gender differences in depression in European countries are not an artifact of a gender-specific reporting style. For developing countries, mixed results also emerge. Using anchoring vignettes, a study documents that among six health domains (mobility, cognition, pain, self-care, difficulty in work or household activities, and mental health), there are gender-specific reporting patterns in two of them in China (mobility and pain), three in Indonesia (pain, self-care, and mental health), and all six in India (Bago d’Uva et al., 2008b). However, using anchoring vignettes and the Chinese sample ages 50+ from wave 1 of SAGE, Zhang et al. (2015) observe that gender-specific reporting does not exist in many aspects of health in China. Thus, they conclude that reporting heterogeneity is unlikely to explain the female disadvantage in health in China. In this section, we review studies on explanations for gender health gaps in old age, including epidemiological explanations, biological factors, social factors, contextual factors, and methodological explanations. Most of the cited studies explore associations between factors of concern and gender health gaps, but few can provide causality due to cross-sectional data structure or mortality selection effect. Some studies have sought causal explanations by using experimental methods (e.g., Austad, 2006; Grol-Prokopczyk et al., 2011), or empirical identification strategies such as controlling for confounding factors and fixed effects (e.g., Zhang et al., 2018) and instrumental variable methods (e.g., Lei and Liu, 2018).

6.4

Conclusion

This chapter reviews gender differences in health among older adults, including a range of diverse health indicators, and discusses several explanations for these gender health gaps from a worldwide perspective. Three major findings emerge. First, compared with older men, older women show consistent advantage in life expectancy and survival, but disadvantage in functional health. The literature on more and less developed regions confirms that although older women 115

Xiaoyan Lei and Yuqi Ta

are expected to live longer than their male peers, older women are more likely to live with functional limitations or disabilities. Second, the magnitude and direction of gender health gaps in old age may differ across countries, change over time, and vary across age groups or birth cohorts. The sex ratios (males to females) are smaller in older age groups and more developed regions, with a U-shaped time trend for more developed countries and an increasing time trend for less developed regions for 1950–2020. For life expectancy, a larger gap exists in more developed regions and in older age groups, with an inverted U-shaped time trend for more developed regions and a rising time trend for less developed regions for 1950–2020. For morbidity profiles, the direction of the gaps in chronic conditions may depend on the diseases under study. For other health indicators, like SRH, depression, and cognition, relatively consistent evidence of female disadvantage is found for less developed regions, while mixed results are found for more developed regions, with the direction and size of gaps depending on the age group, location, and time scale under study. One aspect in understanding the different patterns of gender health gaps in old age is different stages of demographic transition—less developed regions have higher fertility rates than more developed regions. Another perspective related to these heterogeneities is female-male ratios. Compared with more developed regions, countries in less developed regions, such as China and India, witness low female-male ratios that are associated with cultural norms and societal factors like son preference (Sen, 2003). Such gender-distorted populations are often related to higher female infant mortality and lower socioeconomic status of women and thus might be associated with elderly female disadvantage in health. Therefore, to some extent, lower female-male ratios to begin with might explain why gender health gaps are larger and more consistent in some less developed countries. Third, research to date has explored several explanations for gender health gaps, including epidemiological explanations, biological explanations, social explanations, contextual factors, and methodological explanations. But no explanation alone can fully explain the complexities of the gender health gaps among the elderly. Of all these explanations, the role of SES has been consistently confirmed in most research. Nevertheless, SES might not be solely responsible for all gender health gaps in later life. For example, biological factors may provide insights into female advantage in life span, and contextual factors can explain heterogeneities in gender health gaps across different countries to a certain extent. The findings here have some implications for future research. First, the complexities of gender differences in health among older adults imply that the gaps found in a single country, in one period, in one age group, or in a single birth cohort are not necessarily generalizable to other countries, other time scales, other age groups, or other cohorts. Thus, conducting more global assessments that include a wide range of unified health indicators is necessary to re-examine gender health gaps periodically. Meanwhile, combining cross-sectional studies with longitudinal studies can help in monitoring health disparities between older women and older men, without being confounded by selective mortality effects. Second, obtaining complete insight into the reasons for gender health gaps among the elderly calls for interdisciplinary research. Because various biological, social, and contextual factors may explain these gender health gaps to some extent, enhancing cooperation among relevant disciplines and applying a comprehensive framework are crucial to obtain a fuller understanding of the determinants of gender health gaps among the elderly. Third, due to omitted variable bias or reverse causality, most studies on gender health gaps in later life cannot present a causal relationship between the examined factors and health by gender. Therefore, room exists for future studies to isolate the causal impacts, which may help policymakers to figure out potential interventions to reduce health disparities by gender. 116

Gender, Ageing, and Health

Finally, the cooperation of governments and societies can make a difference in reducing gender health gaps in old age, which will not only benefit women but the whole society. Current developments in national policies should be intensified to increase women’s socioeconomic status by removing barriers that impede women’s access to resources and providing pathways for personal development, such as education, labor participation, job promotion, and increased pay. The policies should encourage both genders and all age groups to change risky lifestyles, such as smoking, excess alcohol consumption, unhealthy eating habits, and sedentary behaviors, especially among the elderly and retired. The policies should also promote efforts to ease the current contraceptive burden on women by promoting more equal shared responsibility between men and women. Moreover, to guarantee the safety and effectiveness of drugs and treatments for women and the elderly, more female and elderly samples should be included in RCTs.

Note 1 According to the United Nations definitions of development groups, more developed regions comprise Europe, North America, Australia/New Zealand, and Japan, while less developed regions or developing regions comprise Africa, Asia (except Japan), Latin America and the Caribbean, plus Melanesia, Micronesia, and Polynesia (United Nations, 2019b).

References ACCIAI, F., AND HARDY, M. (2017): “Depression in later life: A closer look at the gender gap,” Social Science Research, 68, 163–175. ADJEI, N. K., BRAND, T., AND ZEEB, H. (2017): “Gender inequality in self-reported health among the elderly in contemporary welfare countries: A cross-country analysis of time use activities, socioeconomic positions and family characteristics,” PLOS ONE, 12(9), e0184676. AHN, N., AND MIRA, P. (2002): “A note on the changing relationship between fertility and female employment rates in developed countries,” Journal of Population Economics, 15(4), 667–682. ANANAT, E. O., AND HUNGERMAN, D. M. (2012): “The power of the pill for the next generation: Oral contraception’s effects on fertility, abortion, and maternal and child characteristics,” Review of Economics and Statistics, 94(1), 37–51. ANGELINI, V., HOWDON, D. D., AND MIERAU, J. O. (2019): “Childhood socioeconomic status and late-adulthood mental health: Results from the survey on health, ageing and retirement in Europe,” The Journals of Gerontology: Series B, 74(1), 95–104. ANTONUCCI, T. C., LANSFORD, J. E., AKIYAMA, H., SMITH, J., BALTES, M. M., TAKAHASHI, K., E´ AND DARTIGUES, J. F. (2002): “Differences between men and women in social relations, resource deficits, and depressive symptomatology during later life in four nations,” Journal of Social Issues, 58(4), 767–783. ARBER, S., AND COOPER, H. (1999): “Gender differences in health in later life: The new paradox?,” Social Science & Medicine, 48(1), 61–76. AUSTAD, S. N. (2006): “Why women live longer than men: Sex differences in longevity,” Gender Medicine, 3(2), 79–92. ˚ BACKHANS, M. C., LUNDBERG, M., AND MANSDOTTER , A. (2007): “Does increased gender equality lead to a convergence of health outcomes for men and women? A study of Swedish municipalities,” Social Science & Medicine, 64(9), 1892–1903. BAGO D’UVA, T., O’DONNELL, O., AND VAN DOORSLAER, E (2008a): “Differential health reporting by education level and its impact on the measurement of health inequalities among older Europeans,” International Journal of Epidemiology, 37(6), 1375–1383. BAGO D’UVA, T., VAN DOORSLAER, E., LINDEBOOM, M., AND O’DONNELL, O. (2008b): “Does reporting heterogeneity bias the measurement of health disparities?,” Health Economics, 17(3), 351–375. BAILEY, M. J., HERSHBEIN, B., AND MILLER, A. R. (2012): “The opt-in revolution? Contraception and the gender gap in wages,” American Economic Journal: Applied Economics, 4(3), 225–254. BAILEY, M. J., MALKOVA, O., AND MCLAREN, Z. M. (2019): “Does access to family planning increase children’s opportunities? Evidence from the war on poverty and the early years of Title X,” Journal of Human Resources, 54(4), 825–856.

117

Xiaoyan Lei and Yuqi Ta

BAILEY, M. J., MALKOVA, O., AND NORLING, J. (2014): “Do family planning programs decrease poverty? Evidence from public census data,” CESifo Economic Studies, 60(2), 312–337. BARFORD, A., DORLING, D., SMITH, G. D., AND SHAW, M. (2006): “Life expectancy: Women now on top everywhere,” BMJ, 332, 808. BERNSTEIN, A., AND JONES, K. M. (2019): “The economic effects of contraceptive access: A review of the evidence,” Institute for Women’s Policy Research. Retrieved from https://iwpr.org/iwprissues/reproductive-health/the-economic-effects-of-contraceptive-access-a-review-of-the-evidence/. BIRD, C. E., AND RIEKER, P. P. (2008): Gender and Health: The Effects of Constrained Choices and Social Policies. Cambridge: Cambridge University Press. BLAU, F. D., AND KAHN, L. M. (2017): “The gender wage gap: Extent, trends, and explanations,” Journal of Economic Literature, 55(3), 789–865. BOERMA, T., HOSSEINPOOR, A. R., VERDES, E., AND CHATTERJI, S. (2016): “A global assessment of the gender gap in self-reported health with survey data from 59 countries,” BMC Public Health, 16(1), 1–9. BORA, J. K., AND SAIKIA, N. (2015): “Gender differentials in self-rated health and self-reported disability among adults in India,” PLOS ONE, 10(11), e0141953. BRINDA, E. M., RAJKUMAR, A. P., ATTERMANN, J., GERDTHAM, U. G., ENEMARK, U., AND JACOB, K. S. (2016): “Health, social, and economic variables associated with depression among older people in low and middle income countries: World Health Organization study on global ageing and adult health,” The American Journal of Geriatric Psychiatry, 24(12), 1196–1208. BROWNE, S. P., AND LALUMIA, S. (2014): “The effects of contraception on female poverty,” Journal of Policy Analysis and Management, 33(3), 602–622. CAMPO-ENGELSTEIN, L. (2012): “Contraceptive justice: Why we need a male pill,” AMA Journal of Ethics, 14(2), 146–151. CANOUI-POITRINE, F., REINALD, N., LAURENT, M., GUERY, E., CAILLET, P., DAVID, J. P., ... ELCAPA STUDY GROUP (2016): “Geriatric assessment findings independently associated with clinical depression in 1092 older patients with cancer: The ELCAPA cohort study,” Psycho-Oncology, 25(1), 104–111. CARMEL, S. (2019): “Health and well-being in late life: Gender differences worldwide,” Frontiers in Medicine, 6(218), 1–11. CASANOVA, M., AND AGUILA, E. (2020): “Gender differences in cognitive function among older Mexican immigrants,” The Journal of the Economics of Ageing, 16(100226), 1–14. CASE, A., AND PAXSON, C. (2005): “Sex differences in morbidity and mortality,” Demography, 42(2), 189–214. CHAN, D., KWOK, A., LEUNG, J., YUEN, K., CHOY, D., AND LEUNG, P. C. (2012): “Association between life events and change in depressive symptoms in Hong Kong Chinese elderly,” Journal of Affective Disorders, 136(3), 963–970. CHEN, Y., AND FANG, H. (2020): “The state of mental health among the elderly Chinese.” NBER Working Paper No. w26690. National Bureau of Economic Research, Cambridge, MA. CHIRINDA, W., AND CHEN, H. (2017): “Comparative study of disability-free life expectancy across six low-and middle-income countries,” Geriatrics & Gerontology International, 17(4), 637–644. CHOI, N. G., AND HA, J. H. (2011): “Relationship between spouse/partner support and depressive symptoms in older adults: Gender difference,” Ageing & Mental Health, 15(3), 307–317. COURTENAY, W. H. (2000): “Constructions of masculinity and their influence on men’s well-being: A theory of gender and health,” Social Science & Medicine, 50(10), 1385–1401. CRIMMINS, E. M., KIM, J. K., AND SOL E´ -AUR O´ , A. (2011): “Gender differences in health: Results from SHARE, ELSA and HRS,” European Journal of Public Health, 21(1), 81–91. ¨ ONEN ¨ DAHLIN, J., AND HARK , J. (2013): “Cross-national differences in the gender gap in subjective health in Europe: Does country-level gender equality matter?” Social Science & Medicine, 98, 24–28. DE FRIAS, C. M., NILSSON, L. G., AND HERLITZ, A. (2006): “Sex differences in cognition are stable over a 10-year period in adulthood and old age,” Ageing, Neuropsychology, and Cognition, 13(3–4), 574–587. DE LEEUW, F. E., DE GROOT, J. C., ACHTEN, E., OUDKERK, M., RAMOS, L. M. P., HEIJBOER, R., ... BRETELER, M. M. B. (2001): “Prevalence of cerebral white matter lesions in elderly people: A population based magnetic resonance imaging study. The Rotterdam Scan Study,” Journal of Neurology, Neurosurgery & Psychiatry, 70(1), 9–14.

118

Gender, Ageing, and Health

DENTON, M., PRUS, S., AND WALTERS, V. (2004): “Gender differences in health: A Canadian study of the psychosocial, structural and behavioural determinants of health,” Social Science & Medicine, 58(12), 2585–2600. DJUKANOVI C´ , I., SORJONEN, K., AND PETERSON, U. (2015): “Association between depressive symptoms and age, sex, loneliness and treatment among older people in Sweden,” Ageing & Mental Health, 19(6), 560–568. EIFERT, S., GUETHOFF, S., KACZMAREK, I., BEIRAS-FERNANDEZ, A., SEELAND, U., GULBINS, H., ... GANSERA, B. (2014): “Applying the gender lens to risk factors and outcome after adult cardiac surgery,” Visceral Medicine, 30(2), 99–106. FORLANI, C., MORRI, M., FERRARI, B., DALMONTE, E., MENCHETTI, M., DE RONCHI, D., AND ATTI, A. R. (2014): “Prevalence and gender differences in late-life depression: A population-based study,” The American Journal of Geriatric Psychiatry, 22(4), 370–380. GEORGE, L. K., BLAZER, D. G., HUGHES, D. C., AND FOWLER, N. (1989): “Social support and the outcome of major depression,” The British Journal of Psychiatry, 154(4), 478–485. GIRGUS, J. S., YANG, K., AND FERRI, C. V. (2017): “The gender difference in depression: Are elderly women at greater risk for depression than elderly men?,” Geriatrics, 2(4), 35. GLEI, D. A., GOLDMAN, N., LIU, I. W., AND WEINSTEIN, M. (2013): “Sex differences in trajectories of depressive symptoms among older Taiwanese: The contribution of selected stressors and social factors,” Ageing & Mental Health, 17(6), 773–783. GORMAN, B. K., AND READ, J. N. G. (2006): “Gender disparities in adult health: An examination of three measures of morbidity,” Journal of Health and Social Behavior, 47(2), 95–110. GROL-PROKOPCZYK, H., FREESE, J., AND HAUSER, R. M. (2011): “Using anchoring vignettes to assess group differences in general self-rated health,” Journal of Health and Social Behavior, 52(2), 246–261. GUERRA, M., PRINA, A. M., FERRI, C. P., ACOSTA, D., GALLARDO, S., HUANG, Y., ... PRINCE, M. (2016): “A comparative cross-cultural study of the prevalence of late life depression in low and middle income countries,” Journal of Affective Disorders, 190, 362–368. GUO, M., CHI, I., AND SILVERSTEIN, M. (2017): “Intergenerational support and depression among Chinese older adults: Do gender and widowhood make a difference?,” Ageing & Society , 37(4), 695–724. HAIER, R. J., JUNG, R. E., YEO, R. A., HEAD, K., AND ALKIRE, M. T. (2005): “The neuroanatomy of general intelligence: Sex matters,” NeuroImage, 25(1), 320–327. HARLING, G., KOBAYASHI, L. C., FARRELL, M. T., WAGNER, R. G., TOLLMAN, S., AND BERKMAN, L. (2020): “Social contact, social support, and cognitive health in a population-based study of middleaged and older men and women in rural South Africa,” Social Science & Medicine, 260(113167), 1–10. HOSPIDO, L., LAEVEN, L., AND LAMO, A. (2020): “The gender promotion gap: Evidence from central banking,” The Review of Economics and Statistics, 1–45. https://direct.mit.edu/rest/articleabstract/doi/10.1162/rest a 00988/97741/The-Gender-Promotion-Gap-Evidence-fromCentral?searchresult=1. HUANG, W., LEI, X., RIDDER, G., STRAUSS, J., AND ZHAO, Y. (2013): “Health, height, height shrinkage, and SES at older ages: Evidence from China,” American Economic Journal: Applied Economics, 5(2), 86–121. IMAI, H., OKUMIYA, K., FUKUTOMI, E., WADA, T., ISHIMOTO, Y., KIMURA, Y., ... MATSUBAYASHI, K. (2015): “Association between risk perception, subjective knowledge, and depression in communitydwelling elderly people in Japan,” Psychiatry Research, 227(1), 27–31. JAIN, A. K. (1981): “The effect of female education on fertility: A simple explanation,” Demography, 18(4), 577–595. JIAO, K., LIU, M., AND XU, M. (2021): “Age and cohort trajectories of gender inequality in health among elderly people in China,” Journal of Women & Ageing, 33(3), 247–267. KANEDA, T., ZIMMER, Z., FANG, X., AND TANG, Z. (2009): “Gender differences in functional health and mortality among the Chinese elderly: Testing an exposure versus vulnerability hypothesis,” Research on Ageing, 31(3), 361–388. KIM, H., NOH, J., NOH, Y., OH, S. S., KOH, S. B., AND KIM, C. (2019): “Gender difference in the effects of outdoor air pollution on cognitive function among elderly in Korea,” Frontiers in Public Health, 7, 1–10. KUMAR, H., AROKIASAMY, P., AND SELVAMANI, Y. (2020): “Socioeconomic disadvantage, chronic diseases and their association with cognitive functioning of adults in India: A multilevel analysis,” Journal of Population Ageing, 13(3), 285–303.

119

Xiaoyan Lei and Yuqi Ta

LAHELMA, E., ARBER, S., MARTIKAINEN, P., RAHKONEN, O., AND SILVENTOINEN, K. (2001): “The myth of gender differences in health: Social structural determinants across adult ages in Britain and Finland,” Current Sociology, 49(3), 31–54. LEE, J., SHIH, R., FEENEY, K., AND LANGA, K. M. (2014): “Gender disparity in late-life cognitive functioning in India: Findings from the longitudinal ageing study in India,” Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 69(4), 603–611. LEE, K. H., XU, H., AND WU, B. (2020): “Gender differences in quality of life among communitydwelling older adults in low-and middle-income countries: Results from the Study on Global Ageing and Adult Health (SAGE),” BMC Public Health, 20(1), 114. LEI, X., HU, Y., MCARDLE, J. J., SMITH, J. P., AND ZHAO, Y. (2012): “Gender differences in cognition among older adults in China,” Journal of Human Resources, 47(4), 951–971. LEI, X., AND LIU, H. (2018): “Gender difference in the impact of retirement on cognitive abilities: Evidence from urban China,” Journal of Comparative Economics, 46(4), 1425–1446. LEI, X., SMITH, J. P., SUN, X., AND ZHAO, Y. (2014a): “Gender differences in cognition in China and reasons for change over time: Evidence from CHARLS,” The Journal of the Economics of Ageing, 4, 46–55. LEI, X., SUN, X., STRAUSS, J., ZHANG, P., AND ZHAO, Y. (2014b): “Depressive symptoms and SES among the mid-aged and elderly in China: Evidence from the China Health and Retirement Longitudinal Study national baseline,” Social Science & Medicine, 120, 224–232. LEVEILLE, S. G., RESNICK, H. E., AND BALFOUR, J. (2000): “Gender differences in disability: Evidence and underlying reasons,” Ageing Clinical and Experimental Research, 12(2), 106–112. LI, D., ZHANG, D. J., SHAO, J. J., QI, X. D., AND TIAN, L. (2014): “A meta-analysis of the prevalence of depressive symptoms in Chinese older adults,” Archives of Gerontology and Geriatrics, 58(1), 1–9. LI, L. W., LIU, J., ZHANG, Z., AND XU, H. (2015): “Late-life depression in rural China: Do village infrastructure and availability of community resources matter?,” International Journal of Geriatric Psychiatry, 30(7), 729–736. LIM, E. J. (2014): “Gender differences in the relationship between physical functioning and depressive symptoms in low-income older adults living alone,” Nursing & Health Sciences, 16(3), 381–386. LINDEBOOM, M., AND VAN DOORSLAER, E. (2004): “Cut-point shift and index shift in self-reported health,” Journal of Health Economics, 23(6), 1083–1099. LIU, Y., LIU, G., WU, H., JIAN, W., WILD, S. H., AND GASEVIC, D. (2017): “Sex differences in non-communicable disease prevalence in China: A cross-sectional analysis of the China Health and Retirement Longitudinal Study in 2011,” BMJ Open, 7(12), e017450. MALLER, J. J., ANSTEY, K. J., RE´ GLADE-MESLIN, C., CHRISTENSEN, H., WEN, W., AND SACHDEV, P. (2007): “Hippocampus and amygdala volumes in a random community-based sample of 60–64 year olds and their relationship to cognition,” Psychiatry Research: Neuroimaging, 156(3), 185–197. MCCRARY, J., AND ROYER, H. (2011): “The effect of female education on fertility and infant health: Evidence from school entry policies using exact date of birth,” American Economic Review, 101(1), 158– 195. MEJIA-ARANGO, S., GARCIA-CIFUENTES, E., SAMPER-TERNENT, R., BORDA, M. G., AND CANOGUTIERREZ, C. A. (2021): “Socioeconomic disparities and gender inequalities in dementia: A community-dwelling population study from a middle-income country,” Journal of Cross-Cultural Gerontology, 36(1), 105–118. ˜ , B., ... MUNRO, C. A., WINICKI, J. M., SCHRETLEN, D. J., GOWER, E. W., TURANO, K. A., MU NOZ WEST, S. K. (2012): “Sex differences in cognition in healthy elderly individuals,” Ageing, Neuropsychology, and Cognition, 19(6), 759–768. NTRITSOS, G., FRANEK, J., BELBASIS, L., CHRISTOU, M. A., MARKOZANNES, G., ALTMAN, P., ... EVANGELOU, E. (2018): “Gender-specific estimates of COPD prevalence: A systematic review and meta-analysis,” International Journal of Chronic Obstructive Pulmonary Disease, 13, 1507–1514. OH, D. H., PARK, J. H., LEE, H. Y., KIM, S. A., CHOI, B. Y., AND NAM, J. H. (2015): “Association between living arrangements and depressive symptoms among older women and men in South Korea,” Social Psychiatry and Psychiatric Epidemiology, 50(1), 133–141. ´ , M. J., CAPUTO, J., JASILIONIS, D., AND SHKOLNIKOV, V. M. (2019): “Is the OKSUZYAN, A., DA NKO story about sensitive women and stoical men true? Gender differences in health after adjustment for reporting behavior,” Social Science & Medicine, 228, 41–50. OKSUZYAN, A., JUEL, K., VAUPEL, J. W., AND CHRISTENSEN, K. (2008): “Men: Good health and high mortality. Sex differences in health and ageing,” Ageing Clinical and Experimental Research, 20(2), 91–102.

120

Gender, Ageing, and Health

OKSUZYAN, A., SINGH, P. K., CHRISTENSEN, K., AND JASILIONIS, D. (2018): “A cross-national study of the gender gap in health among older adults in India and China: Similarities and disparities,” The Gerontologist, 58(6), 1156–1165. OWENS, I. P. (2002): “Sex differences in mortality rate,” Science, 297(5589), 2008–2009. PARK, N. S., JANG, Y., LEE, B. S., HALEY, W. E., AND CHIRIBOGA, D. A. (2013): “The mediating role of loneliness in the relation between social engagement and depressive symptoms among older Korean Americans: Do men and women differ?,” Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 68(2), 193–201. ´ PAYNE, C. F., GOMEZ -OLIV E´ , F. X., KAHN, K., AND BERKMAN, L. (2017): “Physical function in an ageing population in rural South Africa: Findings from HAALSI and cross-national comparisons with HRS sister studies,” The Journals of Gerontology: Series B, 72(4), 665–679. READ, J. N. G., AND GORMAN, B. K. (2010): “Gender and health inequality. Annual Review of Sociology,” Annual Review of Sociology, 36, 371–386. ROSS, C. E., MASTERS, R. K., AND HUMMER, R. A. (2012): “Education and the gender gaps in health and mortality,” Demography, 49(4), 1157–1183. ROTAROU, E. S., AND SAKELLARIOU, D. (2019): “Structural disadvantage and (un) successful ageing: Gender differences in activities of daily living for older people in Chile,” Critical Public Health, 29(5), 534–546. ROY, K., AND CHAUDHURI, A. (2008): “Influence of socioeconomic status, wealth and financial empowerment on gender differences in health and healthcare utilization in later life: Evidence from India,” Social Science & Medicine, 66(9), 1951–1962. RUITENBERG, A., OTT, A., VAN SWIETEN, J. C., HOFMAN, A., AND BRETELER, M. M. (2001): “Incidence of dementia: Does gender make a difference?,” Neurobiology of Ageing, 22(4), 575–580. RUSSELL, D., AND TAYLOR, J. (2009): “Living alone and depressive symptoms: The influence of gender, physical disability, and social support among Hispanic and non-Hispanic older adults,” Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 64(1), 95–104. ´ ˜ , J. A., MANRIQUE-ESPINOZA, B., MORENOSALINAS-RODR´I GUEZ, A., FERN ANDEZ -NI NO BANDA, G. L., SOSA-ORTIZ, A. L., QIAN, Z. M., AND LIN, H. (2018): “Exposure to ambient PM2.5 concentrations and cognitive function among older Mexican adults,” Environment International, 117, 1–9. ¨ SANTOSA, A., SCHR ODERS , J., VAEZGHASEMI, M., AND NG, N. (2016): “Inequality in disability-free life expectancies among older men and women in six countries with developing economies,” Journal of Epidemiology and Community Health, 70(9), 855–861. SCHMITZ, A., AND LAZAREVI Cˇ , P. (2020): “The gender health gap in Europe’s ageing societies: Universal findings across countries and age groups?,” European Journal of Ageing, 17(4), 509–520. ¨ SCH UNEMANN , J., STRULIK, H., AND TRIMBORN, T. (2017): “The gender gap in mortality: How much is explained by behavior?,” Journal of Health Economics, 54, 79–90. SEN, A. (2003): “Missing women—revisited,” BMJ, 327, 1297–1298. SINHA, P., HUSSAIN, T., BOORA, N. K., RAO, G. N., VARGHESE, M., GURURAJ, G., ... GROUP, N. I. C. (2021): “Prevalence of common mental disorders in older adults: Results from the National Mental Health Survey of India,” Asian Journal of Psychiatry, 55(102463), 1–8. SONFIELD, A., HASSTEDT, K., KAVANAUGH, M. L., AND ANDERSON, R. (2013): “The social and economic benefits of women’s ability to determine whether and when to have children.” Guttmacher Institute, New York. STAFFORD, M., CUMMINS, S., MACINTYRE, S., ELLAWAY, A., AND MARMOT, M. (2005): “Gender differences in the associations between health and neighbourhood environment,” Social Science & Medicine, 60(8), 1681–1692. STEVENSON, B., AND WOLFERS, J. (2009): “The paradox of declining female happiness,” American Economic Journal: Economic Policy, 1(2), 190–225. STRAMBA-BADIALE, M. (2010): “Women and research on cardiovascular diseases in Europe: A report from the European Heart Health Strategy (EuroHeart) project,” European Heart Journal, 31(14), 1677–1681. STRAUSS, J., WITOELAR, F., MENG, Q., CHEN, X., ZHAO, Y., SIKOKI, B., AND WANG, Y. (2018): “Cognition and SES relationships among the mid-aged and elderly: A comparison of China and Indonesia.” NBER Working Paper No. w24583. National Bureau of Economic Research, Cambridge, MA. SUBBARAO, K., AND RANEY, L. (1995): “Social gains from female education: A cross-national study,” Economic Development and Cultural Change, 44(1), 105–128.

121

Xiaoyan Lei and Yuqi Ta

TEH, J. K., TEY, N. P., AND NG, S. T. (2014): “Ethnic and gender differentials in non-communicable diseases and self-rated health in Malaysia,” PLOS ONE, 9(3), e91328. ¨ TESCH-ROMER , C., MOTEL-KLINGEBIEL, A., AND TOMASIK, M. J. (2008): “Gender differences in subjective well-being: Comparing societies with respect to gender equality,” Social Indicators Research, 85(2), 329–349. UMBERSON, D., AND WILLIAMS, K. (2005): “Marital quality, health, and ageing: Gender equity?,” The Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 60(Special Issue 2), S109–S113. UNITED NATIONS, DEPARTMENT OF ECONOMIC AND SOCIAL AFFAIRS, POPULATION DIVISION. (2019a): “World population prospects 2019, online edition. Rev. 1.,” Available at https://population.un.org/wpp/. UNITED NATIONS, DEPARTMENT OF ECONOMIC AND SOCIAL AFFAIRS, POPULATION DIVISION. (2019b): “World population prospects 2019, definition of regions.” Available at https://population.un.org/wpp/DefinitionOfRegions/. VERBRUGGE, L. M. (1989): “The twain meet: Empirical explanations of sex differences in health and mortality,” Journal of Health and Social Behavior, 30(3), 282–304. VERGHESE, J., KUSLANSKY, G., KATZ, M. J., SLIWINSKI, M., CRYSTAL, H. A., BUSCHKE, H., AND L IPTON, R. B. (2000): “Cognitive performance in surgically menopausal women on estrogen,” Neurology, 55(6), 872–874. VITALE, C., FINI, M., SPOLETINI, I., LAINSCAK, M., SEFEROVIC, P., AND ROSANO, G. M. (2017): “Under-representation of elderly and women in clinical trials,” International Journal of Cardiology, 232, 216–221. WANDERA, S. O., KWAGALA, B., AND NTOZI, J. (2015): “Prevalence and risk factors for self-reported non-communicable diseases among older Ugandans: A cross-sectional study,” Global Health Action, 8(1), 27923. WEIR, D., LAY, M., AND LANGA, K. (2014): “Economic development and gender inequality in cognition: A comparison of China and India, and of SAGE and the HRS sister studies,” The Journal of the Economics of Ageing, 4, 114–125. WHITSON, H. E., LANDERMAN, L. R., NEWMAN, A. B., FRIED, L. P., PIEPER, C. F., AND COHEN, H. J. (2010): “Chronic medical conditions and the sex-based disparity in disability: The Cardiovascular Health Study,” Journals of Gerontology Series A: Biomedical Sciences and Medical Sciences, 65(12), 1325–1331. WHO (WORLD HEALTH ORGANIZATION). (1946): “Summary report on proceedings minutes and final acts of the international health conference,” Official records of the World Health Organization No. 2. Available at https://apps.who.int/iris/bitstream/handle/10665/85573/ Official record2 eng.pdf?sequence=1. WHO (WORLD HEALTH ORGANIZATION). (2016): “Women’s health and well-being in Europe: Beyond the mortality advantage,” Regional Office for Europe, Copenhagen. Available at https://apps.who.int/iris/handle/10665/332324. WHO (WORLD HEALTH ORGANIZATION). (2020): “Global health estimates 2019: Deaths by cause, age, sex, by country and by region, 2000–2019,” Geneva. Available at https://www.who.int/ data/gho/data/themes/mortality-and-global-health-estimates/ghe-leading-causes-of-death. WILKER, E. H., PREIS, S. R., BEISER, A. S., WOLF, P. A., AU, R., KLOOG, I., ... MITTLEMAN, M. A. (2015): “Long-term exposure to fine particulate matter, residential proximity to major roads and measures of brain structure,” Stroke, 46(5), 1161–1166. XU, H., VORDERSTRASSE, A. A., DUPRE, M. E., MCCONNELL, E. S., ØSTBYE, T., AND WU, B. (2019): “Gender differences in the association between migration and cognitive function among older adults in China and India,” Archives of Gerontology and Geriatrics, 8, 31–38. ZHANG, B., AND LI, J. (2011): “Gender and marital status differences in depressive symptoms among elderly adults: The roles of family support and friend support,” Ageing & Mental Health, 15(7), 844–854. ZHANG, H., D’UVA, T. B., AND VAN DOORSLAER, E. (2015): “The gender health gap in China: A decomposition analysis,” Economics & Human Biology, 18, 13–26. ZHANG, X., CHEN, X., AND ZHANG, X. (2018): “The impact of exposure to air pollution on cognitive performance,” Proceedings of the National Academy of Sciences, 115(37), 9193–9197. ZHANG, Z. (2006): “Gender differentials in cognitive impairment and decline of the oldest old in China,” The Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 61(2), S107–S115. ZUNZUNEGUI, M. V., MINICUCI, N., BLUMSTEIN, T., NOALE, M., DEEG, D., JYLH A¨ , M., ... CLESA WORKING GROUP. (2007): “Gender differences in depressive symptoms among older adults: A cross-national comparison,” Social Psychiatry and Psychiatric Epidemiology, 42(3), 198–207.

122

7 ECONOMICS OF DISEASE PREVENTION IN THE ELDERLY JP Sevilla1

Abstract The intrinsic value of elderly disease prevention (EDP) lies in promoting elderly health and wellbeing. But the broad socioeconomic challenges of population ageing make instrumental values increasingly important: facilitating paid and unpaid work and active leisure, reducing health sector and fiscal burdens associated with population ageing, and reducing care and support burdens on households. Instrumental values are especially associated with interventions that reduce morbidity or increase functional ability. Elderly individuals will not choose socially optimal levels of investment in prevention because of externalities; the significant informational and cognitive burdens of rational choice in this area; problems associated with self-control, biases, and heuristics; and equity. Thus, socially optimal investments in EDP require a strong government role. Among likely high value-for-money investments are those addressing basic needs; taxes and regulations, especially regarding cigarettes, alcohol, and food; universal healthcare; vaccination; and various preventive clinical services. Standard economic evaluations of EDP interventions adopt a health payer perspective and cost-utility analysis. These focus only on interventions’ health gains and payer budget implications, ignoring the socioeconomic consequences of improved health and risking underinvestment. Remedying such risk requires greater use of the societal perspective and cost-benefit analysis in economic evaluation. Better valuation methods can facilitate socially optimal investments in EDP and send better market signals for research and development into future generations of preventive technologies. Difficult policy questions remain regarding how to balance dynamic efficiency on the one hand and static efficiency and equity on the other.

7.1

Introduction and Summary

The core benefit of disease prevention in the elderly (elderly disease prevention, or EDP) is the promotion of their health and well-being. However, given the broad socioeconomic challenges of population ageing, the instrumental benefits of such prevention in terms of addressing such challenges are increasingly important. These value elements include facilitating paid and unpaid work and active leisure, reducing health sector and fiscal burdens associated with population ageing, and reducing the burdens on households of providing care and support. These instrumental values are especially associated with interventions that reduce morbidity or increase functional 123

DOI: 10.4324/9781003150398-8

JP Sevilla

ability. Elderly individuals are unlikely to choose socially optimal levels of investment in prevention on their own because of externalities; the significant informational and cognitive burdens of rational choice in this area; problems associated with self-control, biases, and heuristics; and equity. Thus, socially optimal investments in EDP require a strong government role. Among likely high value-for-money health promotion and disease prevention interventions are those attending to the elderly’s basic needs; taxes and regulations, especially regarding cigarettes, alcohol, and food; universal healthcare; vaccination; and various preventive clinical services. Standard economic evaluations of disease prevention interventions adopt a health payer perspective and cost-utility analysis, which focuses only on interventions’ health gains and implications for the payer’s budget. This approach ignores the various socioeconomic consequences of improved health and, in particular, ignores the instrumental value of EDP in terms of delaying retirement, allowing part-time work in retirement, and reducing the fiscal burdens and informal and formal support costs associated with elderly disability. Such an approach therefore risks underinvestment in EDP. Remedying such underinvestment requires greater use of a societal perspective and cost-benefit analysis in economic evaluation. Better valuation methods can facilitate socially optimal investments in EDP and can send better market signals for research and development (R&D) into future generations of preventive technologies. Difficult policy questions remain regarding how to balance dynamic efficiency on the one hand and static efficiency and equity on the other.

7.2

The Economic Challenges of Ageing Populations and the Role of Disease Prevention in Addressing Those Challenges

Population ageing brings well-known socioeconomic challenges (Bloom, 2019). Central among these are risks of (1) slowdown in economic growth resulting from the lower labor force participation and productivity of the elderly relative to the working-age population (Gordon, 2016; cited in Acemoglu and Restrepo, 2017); (2) the rise in the informal and formal costs that workers must bear to support nonworkers (support costs) (Lee and Mason, 2017); and (3) the increased fiscal burdens from reduced earnings tax revenues and greater expenditures on pensions, healthcare, and long-term care. Traditional economic evaluation of health technologies typically focuses narrowly on two main elements of value: the health gains that result from their use (typically measured in qualityor disability-adjusted life years, or QALYs and DALYs) and any savings in expenditures from the health payer’s budget (Bloom et al., 2018). Recent work on economic evaluation has focused on broadening the scope of such valuation to include, among other things, the broader socioeconomic benefits of improved health (Bloom et al., 2018). This broader perspective suggests we expand our understanding of the value of disease prevention in the elderly (EDP) to include its potential contribution to raising economy-wide labor force participation and productivity and reducing support costs and fiscal burdens. This requires emphasizing the investment character of EDP, revising economic evaluation approaches accordingly, and using these approaches to facilitate expenditure on EDP commensurate not only to its intrinsic value to the elderly but also to its instrumental value in addressing the socioeconomic challenges of ageing.

7.3

Burden of Disease in the Elderly

The burden of disease borne by those aged 60 and older is about a quarter of the burden across all ages globally, about half the burden in developed nations, and about a fifth of it in developing nations (Prince et al., 2014). Infectious disease constitutes less than 10 percent of this burden, and noncommunicable diseases (NCDs) about 85 percent (Prince et al., 2014). 124

Economics of Disease Prevention in the Elderly

The largest contributors to this burden are cardiovascular diseases (30.3 percent), malignant neoplasms (cancers, 15.1 percent), chronic respiratory diseases (9.5 percent), musculoskeletal disease (7.5 percent), and neurological and mental disorders (6.6 percent) (Prince et al., 2014). The central risk factors for NCDs are smoking, alcohol abuse, physical inactivity, poor diet, and air pollution (Mikkelsen et al., 2019). Burdens of disease are not uniformly distributed across the elderly. Rather, a well-known socioeconomic gradient in health means that risks of mortality and morbidity are often much higher among those of low socioeconomic status as measured by education, wealth, social and occupational status, and neighborhood characteristics (like walkability, recreational opportunities, and accessibility of healthy foods) (Steptoe and Zaninotto, 2020). Some argue that only a relatively small part of these health inequalities can be attributed to differential healthcare use and that much of it results from a socioeconomic gradient in behavioral risk factors (like smoking and inactivity) (Steptoe and Zaninotto, 2020) and exposures to social and environmental risk factors (like occupational status, control over one’s environment, and stress) (Braveman and Gottlieb, 2014; Sassi et al., 2015).

7.4

Scope Setting

The two goals of health promotion and disease prevention are related but distinct (Tengland, 2010). The former is a positive goal, while the latter is the prevention of a negative one. All disease prevention is health promotion (they make health greater than it would be otherwise), but some health promotion is not disease prevention (two individuals may be equally disease free, but one could be made healthier in the sense of having more functional ability) (Tengland, 2010). Disease prevention is a historically medicalized term focused on the condition of specific organs. Health promotion is more holistically focused on the condition of the whole individual and, because less medicalized, is more likely to consider the socioeconomic gradient of health (Tengland, 2010). Exercise, healthy eating, and fall prevention, for example, seem more usefully thought of as health promotion than aimed at preventing specific diseases. In practice, health promotion and disease prevention activities often overlap (Tengland, 2010). However, as a conceptual matter, health promotion is a superset of disease prevention, and we stand a better chance of addressing social goals and addressing the challenges of ageing by expanding our policy repertoire. So, often, the appropriate conceptual focus is health promotion. However, “disease prevention” is an ingrained term in the literature so I shall often use that term, even in the context of making points that apply to health promotion more generally. I discuss the prevention of both infectious and noncommunicable diseases, for though NCDs constitute the bulk of the burden of disease, significant value for money (VfM) may arise from preventing infectious diseases such as through elderly vaccination. I shall focus on public sector interventions rather than those of private sector actors like employers or nongovernment organizations. Among public sector interventions, I will not limit myself to health technologies (i.e., interventions deployed within the health sector), but will discuss other policy instruments as well, including taxation, regulation, agricultural policy, etc., allowing for “whole-of-government” approaches (Sassi et al., 2015). I discuss both developed and developing country settings.

7.5

The Intrinsic and Instrumental Value of Elderly Health

Health is a form of human capital (Becker, 2007): a stock or asset embodied in our physical and mental states that yields a future stream of benefits in terms of time spent alive and in good health. It is a stock whose monetary value is the expected present discounted monetary 125

JP Sevilla

value of each moment within that stream. Like any asset, health can be run down through neglect, disease, poor living conditions or habits, and ageing. But it can also be maintained and augmented through investments in the form of health promotion or disease prevention activities that increase subsequent time spent alive and healthy. The value of each moment spent alive and in a particular state of health can be decomposed into its intrinsic and instrumental value. In utility-based formulations of this idea, an individual’s utility directly depends on their survival prospects, health-related quality of life (HRQoL), consumption, and nonmarket time. Consumption includes that of regular goods and services and health-related ones like healthcare. Nonmarket time is time spent on unpaid work—which consists of time spent on such things as housework, caregiving, and volunteering—and leisure. Survival has the intrinsic value of being a prerequisite to enjoying a positive level of utility from good health, consumption, and nonmarket time.2 Having good HRQoL in turn has the intrinsic value of being a natural complement to consumption and nonmarket time: the higher (lower) the level of HRQoL, the larger (smaller) the marginal utility of consumption and nonmarket time. (Natural complementarity is symmetric—think of right shoes and left shoes—so it also implies that consumption and nonmarket time enhance the value of health). The instrumental value of an individual’s survival prospects and HRQoL derives from their contribution to the non-health direct determinants of utility: consumption and nonmarket time. However, from a societal perspective, what matters is not only the consumption and nonmarket time of the individual whose health is at issue but also those of other individuals in society. A standard assumption is that individual utility or well-being (I use these words interchangeably, and I use them to refer to “overall” utility, that is, utility spanning all aspects of life including health and economic aspects) is a concave function of both consumption and nonmarket time. This means that smooth, stable, and certain lifetime trajectories are preferable to those that, though they have equal expected values, are cyclical, volatile, or risky. Thus, individuals desire not only higher levels of consumption and nonmarket time but smooth and certain trajectories in those quantities. Good health has financial risk protection value because it insulates smooth consumption from the potentially catastrophic and impoverishing out-of-pocket costs of illness. The impact of health on various aspects of paid work (or equivalently, market production) is relevant to health’s instrumental value because such paid work finances and therefore constrains consumption. Such constraint works at the level of individuals: I pay for this month’s groceries out of this month’s paycheck, or out of savings from past paychecks. But it also works at the level of households and societies. Individuals’ earnings can be transferred, informally within households or through formal tax-and-transfer programs run by government, to support the consumption of others. Such informal and formal transfer programs are critical elements of support costs, that is, the costs of supporting the consumption and health expenditures of the elderly who are often retired or, if not, do significantly less paid work than the working aged. Three such formal transfer programs are noncontributory pensions, social or statutory health insurance, and long-term care benefits. The human capital approach acknowledges that not just paid work or market activities matter. Nonmarket time in the form of unpaid work and leisure also matter, because these too are sources of utility for both the person who spends this time and others who may benefit from this person’s unpaid work. Unpaid work produces valuable outcomes such as clean and wellorganized homes, cared-for children and disabled family members, and contributions to the activities of nonprofit organizations and to community life. Loss in the ability to do unpaid work would often require hiring paid surrogates to perform such work. Leisure itself, that is, the time we spend being with friends, family, and loved ones; on recreational and entertainment 126

Economics of Disease Prevention in the Elderly

activities; in pursuit of personal life goals, commitments, or hobbies—colloquially, the time spent having a life, or living a life outside of work—is central to well-being. An elderly person’s health and freedom from disability will have an instrumental value related to their own or others’ market and nonmarket time. This is because elderly disability will impair their ability to do paid and unpaid work. This in turn can compel friends and family to pick up the slack, with potentially adverse effects on their own paid and unpaid work. On the one hand, such friends and family may have to do more paid work than they would otherwise to maintain household consumption. On the other, they may have to forego paid work or their own leisure to do unpaid work required to care for elderly disabled. These adverse effects of elderly disability on others’ time spent on market and nonmarket activities are another central element of support costs of ageing populations. The elderly’s longevity and HRQoL both have positive intrinsic value to their own lives (and indeed to the lives of others who love them). But this longevity and HRQoL can have asymmetric instrumental values because of their effects on others. As just discussed, improvements in elderly HRQoL can have strong positive instrumental value both for themselves and for others: it allows the elderly to do more paid and unpaid work, which in turn reduces the informal and formal support costs others must bear. However, improvements in longevity can by themselves increase the support costs borne by households and formal transfer systems and thereby create elements of negative instrumental value for society. Because of this dynamic, the total value to society of promoting elderly health depends on the extent of morbidity compression: the relative pace of improvements in life expectancy and disability-free life expectancy. If the former grows faster than the latter, then the elderly will live longer but also spend much of that longer life disabled and dependent on the productive contributions of others, thereby raising support costs. But if the latter grows faster, then the elderly will have growing abilities for self-sufficiency during their longer lives, reducing support costs. Thus, the value of elderly health promotion depends on its relative impact on adding years to life and adding life to years. Though life extension without morbidity compression can have negative instrumental value, that does not mean its total value is negative, because its intrinsic value would still in many cases be positive and larger. In sum, the value of investments in elderly health has intrinsic and instrumental elements. The instrumental value of such health to the elderly themselves involves health’s ability to facilitate their own paid and unpaid work and leisure. The instrumental value of the elderly’s health arises from reducing informal and formal support costs. These support costs can increase in response to longevity improvements but can decline in response to morbidity compression. In these areas lie the investment case for health promotion and disease prevention in the elderly: realizing the intrinsic and instrumental value of elderly health and thereby addressing some of the socioeconomic challenges of population ageing such as reduced market production and increased support costs.

7.6

Rationale for the Government’s Role in Optimal Investment in Elderly Health Promotion

Investments in health promotion yield the aforementioned benefits for the elderly and the rest of society but can also be costly. Socially optimal investments are those whose total social benefits exceed their costs. Ideally, each elderly individual, of their own volition, would undertake all health promotion activities that satisfy this socially optimal rule. However, multiple circumstances prevent such optimal investment decisions by the elderly, which rationalizes the government playing a role in realizing such optimal investments. 127

JP Sevilla

One important circumstance is that of positive externalities from prevention. Individuals are typically self-interested when making decisions, weighing up the costs and benefits to themselves and to their loved ones of alternative courses of action, ignoring or weighing less heavily those borne by others with whom they have no personal connections. Thus, for example, in societies with collectively financed healthcare systems, affordable means of preventing illness could significantly reduce financial burdens to those systems (in other words, an ounce of prevention could be worth a pound of cure), but individuals’ incentives to prevent that illness may diminish because those burdens have been insured away (i.e., the sick individual does not pay for the pound of cure because we collectively do). Another externality is that formal, collectively financed transfer programs often bear the support costs of elderly disability. So, although investments to prevent such disability can effectively reduce those support costs, the elderly themselves may be indifferent to such fiscal savings. Yet another important positive externality in the field of infectious disease prevention is that of herd effects from vaccination, say from influenza or COVID-19. When a person gets vaccinated, they protect not just themselves but also others around them, because these others are now surrounded by one less person from whom they can contract an infection. Yet many such beneficiaries from herd protection are strangers whose interests individuals may fail to fully consider when making vaccination decisions. Indeed, as a general matter, to the extent that health promotion investments in the elderly can help society cope with the socioeconomic burdens of ageing—raising society-wide productivity and lowering publicly financed health expenditures and support costs—the elderly themselves may not take such macrosocial values into consideration when making health-promotion-related decisions. Another circumstance is that making rational personal decisions about health is informationally and cognitively demanding. Such decision-making, for example, requires knowing the probabilities of various diseases, the range and likelihoods of various health and economic outcomes associated with that disease, the range of actions available to either prevent or treat those diseases, and the relative costs and effectiveness of those various actions. Collecting high-quality information on these various aspects and identifying best courses of action would take a significant amount of time, effort, and thought. Most if not all of us, left to our own devices, would be incapable of coping with those demands and be paralyzed into inaction. A third set of circumstances involves systematic shortfalls from rationality that behavioral economists and psychologists have studied in recent years. These shortfalls can be found in three areas: biases in the perception or understanding of information, the use of heuristics (or rules of thumb or cognitive shortcuts), and self-control problems (Sassi et al., 2015, p. 10). Selfcontrol problems arise when we acknowledge the principle of “no pain no gain” and indeed are willing to incur pain (say) tomorrow in return for gains thereafter, but are incapable of acting according to that principle when the pain falls today. Optimism bias leads us to overestimate the probability of good outcomes (like not getting sick) and underestimate the probability of bad ones (like getting sick) (Sharot, 2011). The availability heuristic leads us to judge events as being more likely the more easily examples of the event come to mind. This can lead us to overestimate the risk of highly publicized and sensational outcomes like plane crashes and nuclear accidents yet underestimate the risks of silent killers like chronic diseases (Sunstein, 2002). Many other such biases and heuristics can adversely affect the elderly’s adopting optimal health promotion behaviors.3 Finally, there is the issue of equity. I had earlier discussed that health risks are distributed unequally. Such distribution often reflects, interacts with, and can reinforce underlying socioeconomic inequalities. We presume that individuals’ health-related decisions are influenced by self-interest and that the most important redistributive mechanisms are those within the capacity 128

Economics of Disease Prevention in the Elderly

of governments, so we expect government rather than individual action to be the primary mechanism for achieving equity goals. Controversies remain regarding equity that I cannot address in much detail, including whether the goal is equity in health (Braveman and Gruskin, 2003) or well-being (see e.g., Adler, 2019), whether the goal should be equality (a goal that can be facilitated by reducing the well-being of the very well off) or priority for the worse off (a goal that can only be facilitated by giving larger weight to benefits received by the badly off and that can never be facilitated by reducing the well-being of the very well off) (Parfit, 2002), and whether equity should be seen from a lifetime perspective (on the one hand, given that one of the most important determinants of inequality in lifetime well-being is inequality in longevity, the elderly are the well off from a lifetime perspective, and so all else equal should receive lower priority than younger populations; on the other hand, the young may have higher cohort life expectancies than the old, which all else equal shifts priority in the direction of the elderly) or in cross-section (in which if the elderly have HRQoL lower than working-age adults, then they are worse off).4 For these reasons, government plays a central role in ensuring adequate investment in health promotion and disease prevention in the elderly.

7.7

Evidence on Value for Money of Health Promotion and Disease Prevention Policies

We now discuss which specific health promotion and disease prevention activities of government might represent good value for money. In discussing the evidence, having a taxonomy of such policies is helpful. A traditional distinction exists among primary prevention, which aims at preventing disease from occurring; secondary prevention, which seeks to prevent clinical manifestation of occurrent disease; and tertiary prevention, which attempts to reduce the severity of occurrent clinical manifestations (Duplaga et al., 2016). We can also distinguish between clinical prevention, which is prevention that takes place in a healthcare setting, and population-based prevention, which takes place outside such settings. Following are some preliminary notes. First, some of the most important risk factors for disease and disability are smoking, alcohol and drug abuse, physical inactivity, poor diet, air pollution, injury and violence, and poor mental health (itself a disease and disability) (Merkur et al., 2019). Much of health promotion and disease prevention therefore aims to address these risk factors. Second, consider the famous saying that “an ounce of prevention is worth a pound of cure.” This saying has some critical aspects worth disentangling. One, prevention can help maintain existing levels of health, while treatments, even when successful, may never be able to fully restore pre-disease levels of health. An example of this is HIV prevention versus treatment. Thus, the health gains from prevention can be attractive relative to those of treatment. Two, prevention has cost-saving potential: spending a little today averts a larger expense tomorrow. Such potential would be important given concerns that population ageing will drive healthcare costs upward worldwide (WHO et al., 2022). This potential has been investigated empirically. Results suggest that cost-saving preventive strategies do exist, but many strategies are not cost saving (Goldman et al., 2006; Roehrig, 2013; Cohen et al., 2008). Rather, they raise costs and so must be assessed in terms of VfM, i.e., whether they produce sufficient health and socioeconomic value for individuals and societies per extra dollar spent. A third interpretation, blending the first two, is that prevention tends to be greater VfM than treatment. While there is not a significant empirical literature that has tested this hypothesis, some preliminary evidence suggests that such a hypothesis should be handled with care. Very high VfM treatments (e.g., treatment of acute myocardial infarction with aspirin, or of asthma 129

JP Sevilla

with low dose inhaled beclometasone; WHO, 2017) and low VfM prevention strategies (some preventive technologies like vaccines are patented and so could have high prices and therefore low VfM) both exist. One suggestive study found that the distribution of incremental costeffectiveness ratios (ICERs) among preventive clinical services resembled that of treatments (Cohen et al., 2008). Thus, what follows shall not presume that prevention is generally cost saving or better VfM than treatment. Whether these are the case are empirical matters to be demonstrated on a case-to-case basis. Finally work on health promotion and disease prevention increasingly emphasizes the importance of a life-course approach to these goals (Merkur et al., 2019). Such an approach recognizes that the scope for keeping an individual healthy exists as early as before and during the mother’s pregnancy (e.g., by promoting women’s health and prenatal care), during infancy (e.g., through exclusive breastfeeding and vaccinations), childhood (e.g., through quality education), adolescence (e.g., through support for sexual and reproductive health and restriction of exposure to alcohol and tobacco), and working-age adulthood (see interventions that follow). Perhaps the most important place to start is population-based primary prevention, which emphasizes that prevention begins before individuals interact with the health system. Indeed, a core theme in the literature is that prevention begins outside the health system and should include attending to individuals’ basic needs (U.S. National Prevention Council, 2011). These include clean air and water; safe and affordable housing; personal, food, and financial security; convenient and supportive transport and infrastructure; employment opportunities even for the elderly; and educational and training opportunities for individuals of all ages (see, e.g., U.S. National Prevention Council, 2011; Schroeder, 2007). Other population-based primary prevention efforts explicitly target the risk factors mentioned previously. These include regulation and taxation of cigarettes, alcohol, junk food, and sugar, some evidence of which suggests may be cost saving at least in terms of public sector spending (Merkur et al., 2019; WHO, 2017; Van der Vliet et al., 2020; Merkur et al., 2013). Relevant regulations of these include certain bans on cigarette- and alcoholrelated advertising, promotion, and sponsorships and restrictions on where cigarettes and alcohol can be sold or consumed (e.g., no smoking indoors or on public transport). Other efforts include health warnings and labeling on tobacco, alcohol, and high-salt, high-saturatedfat, and high-sugar foods; mass media campaigns regarding smoking, alcohol, physical activity, and diet; reformulations and regulations of food products to reduce salt, saturated fats, and sugar; fall prevention programs (WHO, 2017; Laxminaran et al., 2006); clean air legislation (Merkur et al., 2013, p. 50); development of bike paths and trails (Merkur et al., 2013, p. 52); drunk driving legislation and enforcement (Merkur et al., 2013, p. 61); speed bumps, enforcement of traffic laws, and use of helmets and seat belts (Laxminaran et al., 2006, p. 1206); and national toll-free phone lines for advice on quitting smoking (WHO, 2017). A fundamental step toward linking the elderly with clinical preventive services is making sure that they have affordable access to clinical services in the first place. This is most likely to be achieved through mandatory universal statutory or social health insurance, whether financed by taxes or premiums. Such systems are already in place in many developed countries. In 2015, many low-and-middle-income countries committed to move toward universal health coverage (UHC) as part of their commitments to achieving the Sustainable Development Goals (WHO, 2021). UHC aims to provide individuals with “the full spectrum of essential, quality health services, from health promotion to prevention, treatment, rehabilitation, and palliative care across the life course” without having to incur “financial hardship” from such coverage.

130

Economics of Disease Prevention in the Elderly

Though much of the burden of disease in the elderly comes from NCDs, infectious disease risks and burdens exist throughout life, in both developed and developing countries. These infectious disease risks can also rise with age because of (1) the waning of the effects of childhood vaccinations; (2) biological and behavioral factors in adolescence and motherhood (see the following); (3) immunosenescence, the natural weakening of the immune system with age; and (4) the weakening of the immune system in response to chronic disease, which itself grows with age. The resulting health burdens include death and severe long-term disabilities. Vaccination is one of the most important public health innovations of all time and cost-effectively saves millions of lives each year worldwide (WHO et al., 2009). But it is still too often thought of as a childcentric policy. A life-course approach to vaccination promotes vaccination through all stages of life, including during adolescence, motherhood, the working years, and retirement. Such an approach promotes vaccination both in the general population and in specific subpopulations, such as those with chronic disease and the immunocompromised. Through all stages of life from adolescence onward, we benefit from periodic booster doses of vaccines against tetanus, diphtheria, and perhaps pertussis.5 The incidence and severity of seasonal influenza, invasive pneumococcal disease, pneumonia, and herpes zoster rise in the elderly because of immunosenescence and chronic disease. Chronic disease and compromised immune systems raise the risk and severity of infectious disease, and infection in turn raises the risk of complications from the former. This makes vaccination against seasonal influenza and pneumococcal disease an especially high priority for the elderly; for adults with heart disease, lung-disease-like asthma and chronic obstructive pulmonary disease, and diabetes (U.S. Centers for Disease Control, 2017); and for people with immune systems compromised by either disease (like HIV) or treatment (like cancer therapy or stem cell or solid organ transplantation) (Ljungman, 2012). The hepatitis B vaccine can benefit those with diabetes, HIV, and liver disease (U.S. Centers for Disease Control, 2016). High-value clinical prevention services include single- and multi-pill drug regimens (including aspirin, statins, beta blockers, diuretics, ACE inhibitors, and others) for lowering cardiovascular disease risks among those with high risks and for secondary prevention of stroke (Laxminaran et al., 2006; Prince et al., 2014); psychosocial interventions for reducing smoking and alcohol consumption (WHO, 2017; U.S. National Prevention Council, 2011, p. 18); nicotine replacement therapy (Laxminaran et al., 2006, p. 1207; Merkur et al., 2013); physical activity and diet counseling (WHO, 2017, p. 9; Merkur et al., 2013, p. 51); preventive foot care, diabetic retinopathy screening, glycemic control, and home glucose monitoring of insulin treatments among diabetics (WHO, 2017, p. 9); lifestyle interventions to prevent diabetes (Zhou et al., 2020) and diabetes screening among high risks (Chatterjee et al., 2013); colonoscopies starting at age 50 and mammograms for women below 70 (WHO, 2017, p. 13); and vision screening (Woolf, 2010).

7.8

Does Government Invest Enough in Elderly Disease Prevention?

We have seen the reasons we cannot expect individuals to invest socially optimal amounts in health promotion and disease prevention. We have also seen the role that governments can and must play in making such investments. We now discuss the prospects for such optimal investments. There are three layers of governmental decision-making to consider. First is the Ministry of Health (MoH), which often receives a fixed budget from the Ministry of Finance (MoF) and must allocate that health budget across different health sector activities, including public

131

JP Sevilla

health activities and various clinical services being offered as part of a (say) UHC benefits package. Given the fixed budget, spending more on one class of activities or services necessarily implies spending less on others, and the MoH must therefore decide on the relative priority of preventive services, screening activities, treatments, and so on. The second layer involves the MoF, which often receives a more or less fixed budget that the tax-and-transfer policies determined by a legislature have set. The MoF must then allocate that public sector budget across different public sector activities and line ministries. The MoF has the power to expand the MoH budget, thereby relaxing the trade-offs the MoH must perform between preventive and nonpreventive health services. However, given the fixed public sector budget, the more the MoF allocates to the MoH, the less it will have to allocate to non-healthrelated activities and line ministries, e.g., education, the environment, labor and employment, social protection and assistance, infrastructure, etc. The third layer involves the legislature, which determines through its tax policies what share of the national income is reallocated to the public sector and what stays in private hands. It can raise taxes, thereby easing the trade-offs an MoF must face, but at the expense of reducing households’ after-tax income. (MoFs can propose tax-and-transfer policies to legislatures and can raise funds by issuing debt, thus exercising some control over the level of public expenditure, but legislatures constrain such control because they have ultimate authority to decide on taxes, transfers, and borrowing limits.) Each of these decision makers can directly or indirectly facilitate or hinder optimal spending on health promotion and disease prevention in the elderly. The MoH can spend more on such activities while foregoing other health spending. The MoF can expand the MoH budget, allowing for more such spending on the elderly without having to forego other health spending, but at the cost of foregoing public spending outside the health sector. And legislatures can expand the MoF budget, relaxing the various health sector and public sector trade-offs, but at the cost of reducing household after-tax income. Simplifying somewhat, economic principles suggest that to promote social welfare, each decision maker should compute the full social costs and benefits of its options and then prioritize those options that yield the largest net social benefit per dollar spent from the relevant budget. Thus, to achieve optimal spending on disease prevention in the elderly, we need to measure the net social benefit of such prevention per dollar spent out of the MoH budget (or equivalently, the social rate of return). The MoH should compare this social rate of return with that of other health expenditures, the MoF should compare it with that of public expenditures outside the health sector, and the legislature should compare it with that of an extra dollar of household post-tax income. Thus, from a broader societal perspective (discussed and defended subsequently), optimal government investment at all three levels depends on having quantitative estimates of the full social benefits and costs per MoH dollar spent on health promotion and disease prevention in the elderly. Such estimates must include all elements of the intrinsic and instrumental values described earlier, whether such values fall on the elderly, their friends and family and households, or the informal or formal bearers of elderly support costs. Such optimal investment also requires government decision makers to commit to allowing such quantitative estimates of social value to drive their decisions. There are unfortunately challenges in these areas. Many MoHs, when deciding whether to reimburse a health technology (e.g., a preventive clinical service) as part of a UHC benefits package, will perform an economic evaluation of that technology. Performing such evaluation requires making two specification choices. First is the choice of perspective, and the most important options here are the health payer’s perspective 132

Economics of Disease Prevention in the Elderly

and the societal perspective. Second is the choice of analysis, and the most important options here are cost-utility analysis (CUA) and cost-benefit analysis (CBA). The health payer’s perspective values only a technology’s health gains (typically denominated in QALYs or DALYs) and consequences for the MoH’s budget. The fundamental goal underlying this perspective is to allocate the MoH budget to maximize the health produced from that budget. It therefore reimburses a candidate preventive service only if its ICER (or the ratio of incremental costs falling on the MoH budget to the incremental QALY/DALY produced by the candidate service relative to a counterfactual policy) falls below the ICER of marginal health technology (where the marginal health technology is the program that is likely to be displaced to accommodate the candidate technology given the fixed MoH budget). This perspective assumes or does not question the optimality of the MoH budget and so eschews informing MoF and legislative decisions. The societal perspective, in contrast, in principle considers the full health and socioeconomic impacts of a health technology and indeed any technology or policy under consideration for public funding whether health-related or not. CUA and CBA differ in that CUA assumes that every QALY or DALY has equal value, while CBA assumes that every dollar has equal value. These can differ when health gains have heterogeneous socioeconomic implications. For example, competing policies may all raise quality-adjusted population life expectancy by 1 year, but have different macroeconomic or fiscal impacts. Perhaps one such policy tends to raise elderly longevity without raising their functional ability, thus contributing to added fiscal burdens in terms of increased noncontributory pensions, disability benefits, or long-term care. And perhaps the other policy tends to raise longevity less and instead raises functional ability more. Such a policy could have more attractive fiscal effects by reducing (or leading to smaller increases in) support costs. In such case, CBA would be better able than CUA to reflect the superior fiscal impacts of the latter policy. However, many MoHs worldwide, in both developed and developing country settings, adopt a health payer perspective CUA as the fundamental baseline specification for evaluating health technologies. This may result in underinvestment in health promotion and disease prevention in the elderly. The health payer’s perspective, by focusing only on health gains and health payer budget impacts, ignores important value elements of elderly health promotion. More specifically, it ignores any positive effects on the ability of the elderly to delay retirement, or to keep doing part-time paid work while retired, or the fiscal benefits in terms of increased earnings taxes. It ignores any positive effects on the ability to do unpaid work and any consequent reduction in informal and formal support costs (to the extent these formal costs are borne by the public sector but not by the health payer’s budget). Because of such omission, a risk exists that given the size of the health payer’s budget, spending out of that budget on elderly health promotion may be suboptimally low. (The magnitude of this risk depends, among other things, on the relative size of these instrumental productivity and fiscal benefits in interventions that disproportionately promote elderly health to interventions that disproportionately promote non-elderly health. If the latter set of interventions produces larger such instrumental benefits, then consideration of such instrumental benefits in economic evaluation could result in less spending on the elderly out of the health payer’s fixed budget. This issue should be resolved empirically.) A second aspect of the health payer’s perspective is that it takes the size of the health payer’s budget as either given or optimal, neither questioning nor scrutinizing it relative to the socioeconomic costs and benefits of such spending. To discuss in detail the reasons behind this assumption would take me too far afield, but a simplified justification is that MoFs typically have the responsibility within governments to make judgments about the relative value of competing priorities like health and education and therefore their judgments should not be questioned but rather deferred to (Brouwer et al., 2008). This is, of course, a logical non sequitur. Individuals 133

JP Sevilla

and groups within societies can defer to MoF decisions in the sense of obeying them, while offering up critiques of such decisions with a view to their future reform. But more worryingly, this unquestioning approach to the payer’s budget creates considerable risk that this budget will be too small. Assume, to be conservative, that among the benefits of a health program, we value only its health consequences as measured by QALYs. Assume, however, that we value such QALYs in terms of what I earlier called their intrinsic value and that we can use individual willingness to pay (WTP) per QALY as a proxy for such intrinsic value. Prominent if not uncontroversial estimates of such individual WTP put that value at between one to three times per capita gross domestic product per QALY (Commission on Macroeconomics and Health, 2001; Robinson et al., 2017; Bertram et al., 2016). Some estimates indicate that in many countries around the world, the marginal cost to the health system of producing a QALY is about half of per capita gross domestic product (Revill et al., 2015; Claxton et al., 2015). Thus, the ratio of marginal benefit to marginal cost of spending out of the health payer’s budget is at least two and may be as high as six according to these benchmark estimates. This suggests that simply taking the health payer’s budget as given and not questioning its optimality probably results in vast underspending on health relative to its socially optimal level (i.e., the level that would equalize marginal benefits and costs). And such calculation only considers the intrinsic value of health, ignoring any instrumental values such as the contribution, say, of elderly morbidity reduction to higher levels of own paid and unpaid work and reduced support costs borne by others. To facilitate setting the health payer’s budget at a socially optimal level and thereby facilitate optimal spending out of that budget on elderly health promotion, a shift from a health payer perspective to a societal one is critical. More generally, a shift away from health payer perspective CUAs toward societal perspective CBAs that incorporate the full social benefits and costs of elderly health promotion and that can flexibly incorporate the productivity and fiscal benefits of such promotion, unconstrained by simplifying assumptions like equal values per QALY, will facilitate optimal spending and decision-making by ministries of health and finance and by legislatures. One problem with CBA, however, is that its equal value per dollar assumption makes it give excessive priority to the needs of wealthy individuals with greater ability to pay. This regressive implication plays an important role in many preferring the health payer perspective and CUA. Recent work in the welfarist tradition remedies this shortfall of CBA by using social welfare functions (SWFs) that effectively scale individuals’ WTP by their marginal utility of income (to neutralize differences in ability to pay) and distributional weights (to allow for priority to the worse off) (Adler, 2019). More widespread use of SWF-based extensions of CBA will allow for a robust incorporation of equity considerations into economic evaluation.

7.9

R&D

In many cases, for example vaccines for the elderly, ensuring optimal health promotion globally and over time requires dynamic efficiency, that is, optimal levels of technological innovation and R&D into new vaccines and other preventive technologies, e.g., a universal influenza vaccine, or vaccines for gonorrhea or syphilis. The financing of such R&D has three important sources: basic research funded by governments (like the U.S. National Institutes of Health), R&D performed by for-profits like pharmaceutical companies, and nonprofits like the Bill & Melinda Gates Foundation. For-profits, of course, invest in R&D only if they can make sufficient risk-adjusted profits on such investment. The underspending by national governments and health systems discussed in the previous section therefore risks sending adverse market signals to for-profits and could 134

Economics of Disease Prevention in the Elderly

therefore produce suboptimally low levels of R&D expenditure into vaccines and other preventives. Thus, the broadening of economic evaluation discussed earlier involving greater use of societal perspective CBAs and SWF can, by stimulating greater government expenditures on elderly health promotion, give stronger incentives for R&D into future technologies for achieving the same goal. Such broader economic evaluation can also improve the R&D decisions of governments and nongovernmental nonprofit R&D funders by providing more comprehensive estimates of the social value that would accrue from future innovations in disease prevention. (Indeed, governmental R&D priority setting can be considered a fourth level of decision-making—in addition to the three discussed in the previous section—that adoption of societal perspective CBAs could improve). The problem of providing optimal R&D incentives falls at the nexus, however, of a trade-off between dynamic efficiency on the one hand and static efficiency and equity on the other. In a patent-driven system for financing R&D such as ours, allowing for-profits to earn sufficient riskadjusted profits on their R&D spending is socially optimal. This in turn requires adequate sales volumes and, importantly, adequate profit margins or markups per unit sales. However, if prices are too high, then profits may exceed what would be required to incentivize R&D, in which case such profits become rents: pure transfers from global consumers of preventive services (individual taxpayers and premium payers who are the ultimate financiers of national health systems) to for-profits. Such rents are problematic: they unnecessarily reduce global access to preventive services, create deadweight losses, and exacerbate global inequality. Thus, reconciling dynamic efficiency with static efficiency and equity in elderly health promotion and disease prevention requires balancing for-profits’ need for adequate risk-adjusted profits and goals of facilitating global access and equity and reducing deadweight losses. This has stimulated research into a wide range of responses, including value-based pricing (Danzon et al., 2015), fair pricing (Moon et al., 2020), innovative financing mechanisms like advanced market commitments (Gavi, 2021; Kremer et al., 2020), and more radical proposals like eliminating patents (Boldrin and Levine, 2013).

7.10

Conclusion

Disease prevention in the elderly has intrinsic value to the elderly but also instrumental value in helping society cope with the socioeconomic challenges of population ageing. This is especially so to the extent such disease prevention can reduce morbidity risk and increase functional ability, allowing the elderly to retire later, do part-time paid work after retirement, and do unpaid work, thus reducing the fiscal burdens and support costs of elderly dependency. Governments are central to mobilizing socially optimal spending on EDP, but key to facilitating such optimal spending is adopting economic evaluation methods that use societal perspective costbenefit analyses to quantify the broad socioeconomic benefits of such investment accurately. Such broader valuation and the increased expenditure will facilitate more socially optimal levels of R&D into future preventive technologies. But challenges remain in addressing trade-offs between dynamic efficiency on the one hand and static efficiency and equity on the other.

Notes 1 I thank the Harvard Program on the Global Demography of Ageing and the Bill & Melinda Gates Foundation for financial support. I also thank a reviewer, Mark Jit, for very valuable comments. All remaining mistakes are mine alone. 2 See, e.g., Murphy and Topel (2006). 3 See, e.g., Luz et al. (2020), Bonner et al. (2018), and Davenport (2020). 4 See discussion and references in Adler and Holtug (2018).

135

JP Sevilla

5 See Adult Vaccination: A Key Component of Healthy Ageing. The Benefits of Life-Course Immunization in Europe. Report commissioned by the Supporting Active Ageing Through Immunization (SAATI) Partnership. (2013). https://www.ifa-fiv.org/wp-content/uploads/2015/09/SAATIReport-2013.pdf.

References ACEMOGLU, D., AND RESTREPO, P. (2017): “Secular stagnation? The effect of ageing on economic growth in the age of automation,” American Economic Review: Papers & Proceedings, 107(5): 174–179. https://doi.org/10.1257/aer.p20171101. ADLER, M. (2019): “Measuring social welfare: Introduction.” New York: Oxford University Press ADLER, M., AND HOLTUG, N. (2018): “Prioritarianism: A response to critics,” Duke Law School Public Law & Legal Theory Series No. 2018-32. Available at https://papers.ssrn.com/ sol3/papers.cfm?abstract id=3154777. BECKER, G. (2007): “Health as human capital: Synthesis and extensions,” Oxford Economic Papers, 59: 379–410. DOI: 10.1093/oep/gpm020. BERTRAM ET AL. (2016): “Cost-effectiveness thresholds: Pros and cons,” Bulletin of the WHO, 94: 925–930. DOI: 10.2471/BLT.15.164418. BLOOM, D. (2019): “The what, so what, and now what of population ageing,” In: Bloom, D. (ed.), Live Long and Prosper: The Economics of Ageing Populations (pp. 8–14). London: CEPR Press. BLOOM, D., ET AL. (2018): “The broad socioeconomic benefits of vaccination,” Science Translational Medicine, DOI: 10.1126/scitranslmed.aaj2345. BOLDRIN, M., AND LEVINE, D. (2013): “The case against patents,” Journal of Economic Perspectives, 27(1): 3–22. DOI: 10.1257/jep.27.1.3. BONNER, C., ET AL. (2018): “Heuristics and biases in cardiovascular disease prevention: How can we improve communication about risk, benefits and harms?,” Patient Education and Counseling, 101(5): 843–853. DOI: 10.1016/j.pec.2017.12.003. BRAVEMAN, P., AND GOTTLIEB, L. (2014): “The social determinants of health: It’s time to consider the causes of the causes,” Public Health Reports. DOI: 10.1177%2F00333549141291S206. BRAVEMAN, P., AND GRUSKIN, S. (2003): “Defining equity in health,” Journal of Epidemiology & Community Health, 57: 254–258. https://jech.bmj.com/content/57/4/254. BROUWER, W., ET AL. (2008): “Welfarism vs. extra-welfarism,” Journal of Health Economics, 27(2): 325–338. CHATTERJEE, R., ET AL. (2013): “Screening for diabetes and prediabetes should be cost-saving in patients at high risk,” Diabetes Care, 36(7): 1981–1987. DOI: 10.2337/dc12-1752/. CLAXTON, K., ET AL. (2015): “Methods for the estimation of the National Institute for Health and Care Excellence cost-effectiveness threshold,” Health Technology Assessment, 19(14): 1366–5278. DOI: 10.3310/hta19140. COHEN, J., ET AL. (2008): “Does preventive care save money? Health economics and the presidential candidates,” New England Journal of Medicine, 358(7): 661–663. COMMISSION ON MACROECONOMICS AND HEALTH. (2001): “Macroeconomics and health: Investing in health for development.” Geneva: World Health Organization. Available at https://apps.who.int/iris/handle/10665/42435. DANZON, P., ET AL. (2015): “Value-based differential pricing: Efficient prices for drugs in a global context,” Health Economics, 24: 294–301. DAVENPORT, T. (2020): “How to make better decisions about coronavirus,” MIT Sloan Management Review. Available at https://sloanreview.mit.edu/article/how-to-make-better-decisions-aboutcoronavirus/. DUPLAGA, M., ET AL. (2016): “Scoping review of health promotion and disease prevention interventions addressed to elderly people,” BMC Health Services Research, 16(Suppl 5): 278. DOI: 10.1186/s12913016-1521-4. GAVI. (2021): Innovative financing. Available at https://www.gavi.org/investing-gavi/innovative-financing. GOLDMAN, D., ET AL. (2006): “The value of elderly disease prevention,” Forum Health Econ Policy, 9(2). DOI: 10.2202/1558-9544.1004. GORDON, R. (2016): The Rise and Fall of American Growth: The U.S. Standard of Living since the Civil War, Princeton: Princeton University Press.

136

Economics of Disease Prevention in the Elderly

KREMER, M., ET AL. (2020): “Advanced market commitments: Insights from theory and experience,” AEA Papers and Proceedings, 110: 269–273. DOI: 10.1257/pandp.20201017. LAXMINARAN, R., ET AL. (2006): “Advancement of global health: Key messages from the Disease Control Priorities Project,” The Lancet, 367: 1193–208. DOI: 10.1016/S0140-6736(06)68440-7. LEE, R., AND MASON, A. (2017): “Cost of ageing,” Finance and Development, 54(1). Available at https:// www.imf.org/external/pubs/ft/fandd/2017/03/lee.htm#: :text=An%20aging%20population%20and %20slower,retirement%20programs%20for%20old%20people. LJUNGMAN, P. (2012): “Vaccination of immunocompromised patients,” Clinical Microbiology and Infection, 18(5): 93–99. DOI: 10.1111/j.1469-0691.2012.03971.x. LUZ, P., ET AL. (2020): “How heuristics and cognitive biases affect vaccination decisions,” Cad Saude Publica, 36(Suppl 2). DOI: 10.1590/0102-311X00136620. MERKUR, S., ET AL. (2013): “Promoting health, preventing disease: Is there an economic case?,” Policy Summary, 6. Available at https://www.euro.who.int/ data/assets/pdf file/0004/235966/e96956.pdf. MIKKELSEN, B., ET AL. (2019): “Life course approach to prevention and control of non-communicable diseases,” BMJ, 365(Suppl 1). Available at https://www.bmj.com/content/bmj/364/bmj.l257.full.pdf. MOON, S., ET AL. (2020): “Defining the concept of fair pricing for medicines,” BMJ, 368. DOI: 10.1136/bmj.l4726. MURPHY, K., AND TOPEL, R. (2006): “The value of health and longevity,” Journal of Political Economy, 114(5): 871–904. PARFIT, D. (2002): “Equality or priority?” In: Clayton, M., and Williams, A. (eds.), The Ideal of Equality. New York: Palgrave Macmillan, (pp. 81–125). PRINCE, M., ET AL. (2014): “The burden of disease in older people and implications for health policy and practice,” The Lancet, 385(9967). DOI: 10.1016/S0140-6736(14)61347-7. REVILL, P., ET AL. (2015): “Cost-effectiveness thresholds: Guiding healthcare spending for population health improvement.” In: Revill, P., Suhrck, M., Moreno-Serra, R., and Sculpher, M. (eds.), Global Health Economics: Shaping Healthcare Policy in Low- and Middle-Income Countries. Singapore: World Scientific Publishing Co. Pte. Ltd., (pp. 75–97). ROBINSON ET AL. (2017): “Understanding and improving the one and three times GDP per capita costeffectiveness thresholds,” Health Policy and Planning, 32: 141–145. DOI: 10.1093/heapol/czw096. Epub July 24, 2016. ROEHRIG, C. (2013): “The complex economics of disease prevention and longevity,” Health Affairs Blog, Available at https://www.healthaffairs.org/do/10.1377/hblog20130122.027393/full. SASSI, F., ET AL. (2015): “Introduction.” In: McDaid, D., et al. (eds.), Promoting Health, Preventing Disease: The Economic Case. Buckingham: Open University Press - McGraw-Hill, (pp. 3–18). SCHROEDER, S. (2007): “We can do better: Improving the health of the American people,” New England Journal of Medicine, 357: 1221–1228. DOI: 10.1056/NEJMsa073350. SHAROT, T. (2011): “The optimism bias,” Current Biology, 21(23): R941–R945. DOI: 10.1016/j.cub.2011. 10.030. STEPTOE, A., AND ZANINOTTO, P. (2020): “Lower socio-economic status and the acceleration of ageing: An outcome-wide analysis,” PNAS, 117(26): 14911–14917. DOI: 10.1073/pnas.1915741117. SUNSTEIN, C. (2002): “Hazardous heuristics.” University of Chicago Public Law & Legal Theory Working Paper 33. Available at https://chicagounbound.uchicago.edu/cgi/viewcontent.cgi?referer= &httpsredir=1&article=1137&context=public law and legal theory. TENGLAND, P. (2010): “Health promotion and disease prevention: Logically different conceptions?,” Healthcare Analysis, 18: 323–341. DOI: 10.1007/s10728-009-0125-0. U.S. CENTERS FOR DISEASE CONTROL. (2016): “Vaccine information for adults.” Available at https://www.cdc.gov/vaccines/adults/rec-vac/index.html. U.S. CENTERS FOR DISEASE CONTROL. (2017): “Adults with chronic conditions: Get vaccinated.” Available at https://www.cdc.gov/features/vaccineschronicconditions/index.html. U.S. NATIONAL PREVENTION COUNCIL. (2011): “National prevention strategy: America’s plan for better health and wellness.” Available at https://www.hhs.gov/sites/default/files/disease-preventionwellness-report.pdf. VAN DER VLIET, N., ET AL. (2020): “Ranking preventive interventions from different policy domains,” International Journal of Environmental Research and Public Health, 17(6). DOI: 10.3390/ijerph17062160. WHO (WORLD HEALTH ORGANIZATION). (2017): “‘Best buys’ and other recommended interventions for the prevention and control of noncommunicable diseases.” Available at https://www.who.int/ncds/management/WHO Appendix BestBuys.pdf.

137

JP Sevilla

WHO (WORLD HEALTH ORGANIZATION). (2021): “What is universal health coverage?” Available at https://www.who.int/news-room/fact-sheets/detail/universal-health-coverage-(uhc). WHO (WORLD HEALTH ORGANIZATION), UNICEF (UNITED NATIONS CHILDREN’S FUND), AND WORLD BANK . (2009): “State of the world’s vaccines and immunization.” Available at http://apps.who.int/iris/bitstream/handle/10665/44169/9789241563864 eng.pdf?sequence=1. WHO (WORLD HEALTH ORGANIZATION), U.S. NATIONAL INSTITUTE ON AGEING, U.S. NATIONAL INSTITUTES OF HEALTH, AND U.S. DEPARTMENT OF HEALTH AND HUMAN SERVICES. (2022): “Global health and ageing. Assessing the costs of ageing and healthcare.” Available at https://www.who.int/ageing/publications/global health.pdf [[accessed on January 11, 2022]]. WOOLF, S. (2010): “The price paid for not preventing diseases.” In: Institute of Medicine (ed.), The Healthcare Imperative: Lowering Costs and Improving Outcomes: Workshop Series Summary. Available at https://www.ncbi.nlm.nih.gov/books/NBK53920/pdf/Bookshelf NBK53920.pdf. ZHOU, X., ET AL. (2020): “Cost-effectiveness of diabetes prevention interventions targeting high-risk individuals and whole populations: A systematic review,” Diabetes Care, 43(7): 1593–1616. Available at https://doi.org/10.2337/dci20-0018.

138

8 THE ECONOMICS OF LONG-TERM CARE David N. F. Bell and Elizabeth Lemmon

Abstract Policy issues associated with the economics of long-term care (LTC) are increasingly important due to population ageing and consequent increased demand for care. Academic research on the economics of LTC spans many publications. These contributions are mainly, but not exclusively, found within health economics. We relate the market for LTC to the market for healthcare and argue that important differences exist between these research areas. We then use a Scopus search to identify publications relating to the economics of LTC. We use the results to identify the main issues that this body of research addresses and then list the most cited articles within the field. Among the most popular topics are the demand for LTC, the related demand for LTC insurance, and LTC financing. Unpaid care, child–parent interactions, the labor market for LTC workers, and projections of future spending also feature among the most cited pieces of work. We summarize some of the arguments relating to these issues. We conclude by arguing that to provide a clearer overview for researchers and policymakers, a case exists for bringing the literature on the economics of LTC together, either in the form of an entirely new journal, or as a satellite to an existing related outlet, and for researching several additional topics that the existing literature has paid scant attention to.

8.1

Introduction

Policymakers have had a long-running interest in the market for long-term care (LTC). This is understandable: Government or compulsory insurance spending on LTC accounted for an average 1.7 percent of gross domestic product (GDP) across Organisation for Economic Cooperation and Development (OECD) countries in 2017 (OECD, 2019). This share is set to increase due to population ageing. The provision of LTC also poses a range of policy questions relating to equity, efficiency, market structure, generational equity, and regulation. The pandemic has had severe consequences for the recipients of LTC, largely because the prevalence of COVID-19 increased steeply with age and because many LTC facilities were susceptible to the transmission of the disease. This crisis has focused unprecedented attention on the sector—including its economic sustainability. However, the economics of LTC is not delineated as a distinct area of study within economics. Instead, it is typically approached as a satellite of health economics. Many publications relating to LTC appear in health economics journals. DOI: 10.4324/9781003150398-9

139

David N. F. Bell and Elizabeth Lemmon

No category for LTC exists within the Journal of Economic Literature classification, though there are eight covering health economics. Yet, although it shares many issues with health economics, the economics of LTC and health economics balance these issues somewhat differently. In addition, although health economics contains the largest reservoir of articles on LTC, other contributions can be found across various economics and non-economics journals. Health economics emerged in the 1940s as an independent field of inquiry. This came about when researchers recognized that healthcare markets have distinctive features that require extensions to standard market models. The economics of LTC has not been similarly identified, despite sharing some features with healthcare markets and having some distinctive features of its own. In this chapter, we seek to identify the current scope of the “economics of LTC” by conducting a systematic search of existing literature to identify economic issues relevant to LTC. We then outline key debates relating to some of these issues. Constraints on space mean that our list is not exhaustive. Nevertheless, we cover what seems to us to be the most important debates. But first, we define what we mean by LTC and outline how the market for LTC is organized. We then explore the characteristics of the LTC market and, in relation to the market for healthcare, establish what we believe are its unique characteristics. We then discuss key LTC economics questions that economists have addressed and report main results. Our concluding comments argue that publications relating to the economics of LTC are too widely dispersed and that providing a more focused research platform and additional research on some underrepresented topics in the LTC literature would provide clear advantages.

8.2

The Market for Long-Term Care

LTC refers to services that assist individuals with personal care tasks and services that enable them to live independently. Loss of independence may be associated with chronic disease and/or increased frailty. Because this volume concerns ageing, we concentrate on loss of function associated with growing older. A continuum of LTC services operates to meet care client needs. At one end are services aimed at helping those living at home maintain their independence. These might include help with laundry, shopping, and managing finances—these are examples of instrumental activities of daily living (ADLs). At the other end of the spectrum is help with personal care tasks such as with washing, eating, and dressing, or ADLs. These can be provided at home or in an institutional setting. If an individual can no longer live independently at home even with support, they will also require accommodation services. These come in various types, with an important distinction between nursing homes, which also provide healthcare support, and residential homes, which do not have specialist healthcare facilities. The organization of LTC services differs across the world, but in general, LTC is provided either through paid (formal) or through unpaid (informal) care. Unpaid LTC refers to care that is provided by spouses/partners, family members, friends, or neighbors who, in most cases, have an existing social relationship with the person being cared for. These tend to be female or male spouses/partners or adult daughters/daughters-in-law. The gender balance of carers becomes more equal with increasing age, with a greater proportion of men taking a caring role. Paid LTC services refer to services that professional care workers who are employed by an organization provide. These services can be provided in an institutional setting, such as an LTC facility like a nursing home or residential care home, or in a community setting, for example in a person’s own home or in day-case or respite centers. 140

The Economics of Long-Term Care

In 2017, around 10.8 percent of those aged 65+ received paid LTC in OECD countries (OECD, 2019). This share varied substantially among countries. In Poland, for example, only 1 percent of those aged 65+ received paid LTC. In contrast, 22 percent of this age group received paid LTC in Switzerland (OECD, 2019). This contrast partly reflects differences in the composition of LTC services between formal and informal provision and in sharing caregiving between the state and the individual receiving care. The makeup of formal care provision also differs between community and institutional settings. For example, in 2017, 32 percent of those who were aged 65 and older and receiving LTC in Portugal received care at home, compared with 92 percent of those in Israel (OECD, 2019). The division between community care and institutional care is sometimes known as the “balance of care” and varies from country to country. Where individuals lack the means to purchase care and would pass eligibility criteria for state-provided care but do not receive such care, an “unmet need” exists. Countries also differ in the latitude provided to care recipients: on the one hand, to freely select care services using a state-provided budget, and on the other hand to having to accept the services the state chooses to provide. Furthermore, the resources devoted to formal LTC contrast widely internationally (Figure 8.1). In this figure, LTC spending encompasses LTC health spending and LTC social spending. The former comprises spending on nursing and personal care services (or ADLs), while the latter relates to spending on services helping with instrumental ADLs. On average, government and compulsory insurance scheme spending on these two LTC expenditure components accounted for 1.7 percent of GDP in the OECD17 countries in 2017 (OECD, 2019).1 Compared with government and compulsory insurance scheme expenditure on healthcare, spending on LTC is relatively low (OECD, 2019). In 2018, healthcare expenditure from government and compulsory insurance schemes comprised 6.6 percent of GDP (OECD, 2019). The demand for formal LTC services increases substantially with age. Across OECD countries, around half of those who receive LTC are aged 80 and older (OECD, 2019). While some

Figure 8.1 Long-term care expenditure (health and social components) by government and compulsory insurance schemes, as a share of GDP, 2017 (or nearest year). Source: OECD (2019).

141

David N. F. Bell and Elizabeth Lemmon

younger people, particularly people with disabilities, do receive LTC services, this chapter will focus on older people aged 65+. Generally, those who receive services within the community are younger than those receiving care in an institutional setting. Many of those receiving services, particularly in institutional settings, have dementia, the prevalence of which increases steeply with age, from 2 percent among those aged 65–69 to more than 40 percent for those aged 90+ (OECD, 2018). The OECD estimates that among its member states, the number of people with dementia will rise from 19 million in 2017 to 40.9 million by 2050 (OECD, 2018). Over the last few decades, the share of the population aged 65 and older in OECD countries has almost doubled (OECD, 2019). The fastest growth has been in the share of the population aged 80+. This demographic shift is a result of increases in life expectancy (and healthy life expectancy) and a sharp fall in fertility rates since the 1960s, resulting in older people making up an increasing share of the population—this is known as the “demographic transition.” This trend is expected to continue in the coming decades. In 2017, those aged 65+ made up 17.4 percent of the population in OECD countries and those aged 80+ made up 4.6 percent of the population. By 2050, those figures are expected to be 27.1 percent and 10.1 percent respectively (OECD, 2019). Therefore, take-up of LTC services will likely grow. To counter these demographic trends, societies are seeking to promote healthy ageing and thereby extend healthy life expectancy and reduce the age-specific demand for LTC services. The ageing process will inevitably lead to an increased volume of care needs for older people. An important economic challenge for all countries will be the financing of this increase in demand. Countries with younger populations may postpone adjustment, though not indefinitely because declining birth rates are almost universal. There are also labor supply issues to consider. LTC services are labor intensive (predominantly female labor) and poor working conditions in the sector, other societal changes such as increased labor market participation of females, increases in divorce rates, and reduction in co-residence threaten the supply of caregivers, both paid and unpaid. The changes in supply may affect the composition of demand through implicit price signals. Thus, the observed balance between paid and unpaid care, between institutional and community settings, is responsive to supply conditions in the LTC market.

8.3

The Contrast between Health Economics and the Economics of Long-Term Care

Healthcare and LTC differ both in process and outcome. People do not consume healthcare because they derive utility from it. The objective of healthcare is to raise the patient’s level of health, through a process that is frequently unpleasant. The purpose of LTC is not to improve health. Its purpose is to maintain, preserve, and slow its deterioration as a person ages. The process is not unpleasant in the way that healthcare treatments can be. Healthcare may be seen as a derived demand for health, while the demand for LTC is a derived demand for maintaining health and function. Arrow (1963) asserts that the existence of uncertainty warrants the “special place for medical care in economic analysis” and that the special features of the market for healthcare stem from this uncertainty (Arrow, 1963, p. 948). Uncertainty arises because people cannot predict when they will become ill or how much healthcare they will require if they do. In the market for LTC, the eventual need for care is less uncertain, but considerable uncertainty exists about the length of time over which care will be needed and the intensity of that care. As discussed, people generally demand LTC resources as they age and their physical health deteriorates. Therefore, individuals can plan to meet the costs of their potential dependence on LTC. However, what 142

The Economics of Long-Term Care

we see in practice is an LTC “insurance puzzle,” where demand for insuring against this risk is low. We discuss this phenomenon later in this chapter. Both markets also involve asymmetric information, but quality of treatment in the healthcare market is assured through the credentials of suppliers: quality in the LTC market is much more difficult to assure because widely accepted signals of quality are not generally available. Why this form of market failure is more prevalent with LTC than healthcare is not obvious. An element of path dependency may exist: whereas healthcare as an activity has been around for millennia, the LTC sector is relatively new, a product of the advances in healthcare that have led to rapid increases in longevity. The industry has therefore had much less time to establish quality assurance mechanisms. The interaction between the economics of LTC and the economics of the household has also hampered this process. When life expectancy was shorter, LTC was principally supplied within the household. In this setting, information flows are less problematic and the reputational and financial incentives relating to care provision are perhaps clearer. One further difference between health economics and LTC is the lack of contagion externalities in the provision of LTC. Many healthcare interventions are intended not just to restore the health of the patient but also to prevent others from catching disease. Contagion is not an issue with LTC per se, though the COVID-19 pandemic did expose the risks of disease contagion within LTC facilities.

8.4

Key Issues for the Economics of Long-Term Care

In general, economics is concerned with how individuals, firms, and governments make choices in the presence of scarcity. LTC resources are scarce, and decisions must be made about who should get LTC, who should provide it, and who should pay for it. Those with sufficient wealth or insurance cover can purchase care, irrespective of need. The less wealthy must be assessed as exceeding some threshold level of frailty before qualifying for state-subsidized care. Those falling below the threshold have little option but to rely on unpaid care, while those who exceed the threshold may also need to make co-payments from current income or from insurance. The design of insurance policies and co-payment mechanisms strongly influences the division of LTC costs between the individual and the state. Much of the literature relating to the economics of LTC addresses these or related issues. Several authors summarize the economics of LTC. Nortons (2000) chapter on LTC in the Handbook of Health Economics provides a thorough overview of the LTC sector, focusing on the key supply and demand side issues specifically related to nursing home care. Fern´andez et al. (2011) also discuss the LTC market, focusing on how care is delivered and organized and how changing demographics pressure public finances. Siciliani (2014) provides an overview of the main issues around financing LTC and projecting future expenditures on LTC. He also explores the role of government in ensuring quality of care within care home settings. Finally, Norton (2016) reviews issues relating to both the demand and supply side of LTC, including an extensive discussion of the provision of unpaid care. Thus, although review papers have brought together the literature on LTC heuristically, no previous attempt to systematically categorize the literature relating to the economics of LTC has occurred. This is what we now endeavor to do.

8.5

Systematic Search for Articles Relating to the Economics of Long-Term Care

In this section, we explore the interest in LTC economics from English language economics journals. To explore the distribution of relevant articles, we conducted a Scopus search for 143

David N. F. Bell and Elizabeth Lemmon

the phrase “long term care” in either the abstract, keywords, or title, within the “economics, econometrics, and finance” subject area.2 Given that several key economics journals were not included within this subject area of the Scopus source lists, we also conducted the same search in economics journals within the “health policy” subject area—specifically, Journal of Health Economics; Pharmaco Economics; Health Economics; Health Economics, Policy and Law; and the Journal of Medical Economics.3 Potentially we may not have captured several further outlets for studies relating to the economics of LTC. However, we believe that our approach captures the main areas of research within the field. This search generated 917 results. After removing irrelevant and duplicate entries, 679 documents remained. Most results were articles (82 percent), with the remainder comprising books, book chapters, conference proceedings, editorials, etc. The search yielded several interesting findings: 1. Publications relating to LTC have increased substantially, particularly since the beginning of this century. Figure 8.2 shows a rapid increase, averaging around 50 articles per year since the middle of the last decade. This broadly coincides with a growing policy focus on LTC as demand has grown following the demographic transition. Nevertheless, given that LTC accounts for around one-quarter of the contribution that healthcare makes to GDP, journals continue to disproportionately focus on health economics relative to LTC. 2. Documents were published across 207 unique sources. Figure 8.3 shows the most common. Interestingly, among the most common outlets for research on the economics of LTC are the health economics journals. 3. Most of the literature is from the United States (US) and the United Kingdom (UK). Specifically, about one-third of documents were from the US and 17 percent from the UK.4 This may reflect English language bias. The dominance of the US is reflected in the focus on insurance markets and unpaid care among the search results.

Figure 8.2 Total number of documents per year (excludes 2021).

144

The Economics of Long-Term Care

Figure 8.3 Number of documents published in the top 11 sources. Key: HE = Health Economics; JHE = Journal of Health Economics; GPRI = Geneva Papers on Risk and Insurance: Issues and Practice; EJHE = European Journal of Health Economics; JEA = Journal of the Economics of Ageing; JRI = Journal of Risk and Insurance; FLTCE = Financing Long-Term Care in Europe: Institutions, Markets and Models; RLTCQ = Regulating Long-Term Care Quality: An International Comparison; JFEI = Journal of Family and Economic Issues; BEJEAP = B.E. Journal of Economic Analysis and Policy.

8.6

Issues Identified by Systematic Search

Table 8.1 categorizes the documents revealed by the Scopus search into several topic groupings. These are not exclusive, and some articles could fit into more than one grouping. Nevertheless, they reflect the key themes emerging from the economics of LTC literature. The most popular area comprises the organization of care, including demand, financing, and policy. Table 8.2 lists the 20 most cited articles from the search. We now outline some of the research areas that emerge from Table 8.1. We cannot cover all of these within the constraints of this chapter. Instead, we focus on the main debates, referencing some of the highly cited articles listed in Table 8.2, where appropriate. We begin with unpaid care.

8.6.1

Unpaid Care

As previously mentioned, one of the main characteristics that differentiates the market for LTC from the market for healthcare is the existence of a large, unpaid workforce. Estimates from the US and the UK suggest that around 17 percent of the adult population provide unpaid care to other adults (Carers UK, 2020; NAC, 2015). The number of publications in this area reflect the importance of this issue, seven of which feature in the 20 most cited articles in Table 8.2. The publications categorized under the “unpaid care interactions” rubric include empirical research on the relationship between paid care (including both LTC services and 145

David N. F. Bell and Elizabeth Lemmon

health services) and unpaid LTC. Much attention has focused on whether unpaid and paid care are substitutes or complements. If unpaid care acts as a substitute for formal care, governments faced with growing pressure on public finances could adopt policies that encourage unpaid care in the hope that they may reduce reliance on publicly funded services. It may also be an inferior substitute where public authorities ration paid care so that an “unmet need” exists within the community. However, if unpaid care facilitates access to paid care, by for example providing advocacy services, government support for unpaid carers could have negative effects on public finances. Van Houtven and Norton’s 2004 research on this topic is the most cited piece of work identified in our search. In their paper, they seek to understand how unpaid care influences older people’s use of formal health and care services including nursing home use, inpatient care, outpatient surgery, and physician visits. Using data on a representative sample of Americans aged 70+, their two-part models show that unpaid care reduces older persons’ use of nursing home and home healthcare, indicating that unpaid and paid care are indeed substitutes. Other key papers in this area include Charles and Sevak (2005), whose findings on unpaid care and nursing home care corroborate the substitution effect demonstrated by Van Houtven and Norton.

Table 8.1 Main groupings of articles relating to the economics of long-term care Category

Description

No. of documents

Percentage of documents (%)

Demand for LTC

Research in this category focuses mainly on the LTC insurance market, the demand for LTC more generally, and the demand implications of population ageing. It has a particular focus on the “LTC insurance puzzle,” that is, the very low uptake of private insurance for LTC even though the need for LTC is a potentially catastrophic financial risk. The literature provides descriptive overviews of LTC financing systems within and among countries. It also often reflects concerns with the financial sustainability of LTC. Papers in this category often overlap with those relating to LTC policy. The literature comprises theoretical models of insurance markets in the presence of asymmetric information, adverse selection, and variations in individuals’ attitudes to risk. The research analyzes government LTC policies and their evolution set against the backdrop of an ageing population. The research focuses on savings and investment behavior of individuals planning for retirement, extending to the role of pension funds and housing equity in LTC financing. There is some overlap with the LTC financing category.

102

15

84

12.4

66

9.7

69

10.2

46

6.8

Financing of LTC

Insuring against LTC risk LTC policy

Retirement planning

146

The Economics of Long-Term Care Category

Description

No. of documents

Percentage of documents (%)

Cost of care

The literature aims to provide cost estimates for different types of LTC. An important example is the estimation of the cost of unpaid care through the application of various techniques to value caregivers’ time. The literature focuses on who provides care and interactions among state, local, and private provision. Several papers explore issues relating to inequalities in care provision and the identification of factors influencing care provision, for example, children’s education and their financial circumstances, changes in LTC policy, etc. The research explores the quality of care in the care home setting. This work includes theoretical models of price and quality competition and the role of regulatory frameworks in ensuring quality of care. This category is linked to the LTC policy category. The literature examines the interaction between paid and unpaid care. Are these types of care substitutes or complements? This includes the interaction between healthcare and unpaid care. Several papers investigate the effects of unpaid care on labor supply, particularly female labor supply. Papers in this category frequently project countryspecific future expenditures on LTC, often alongside expenditure on healthcare. Several articles focus on the role of out-of-pocket expenditures in LTC finance. The expenditure theme is also relevant to the papers relating to the “red herring hypothesis,” the idea that health and care expenditures rise with time to death rather than age. Works in this category are mostly theoretical, strategic bargaining models between parents and children that provide explanations of care provision within the family. Models invoke parent’s wish to purchase care elsewhere, children’s bequests motives, and altruism. This category involves models that forecast future rates of disability, morbidity, and mortality, and the implications for future demand for LTC services. A key issue for these projections is whether morbidity will be compressed or expanded. Some articles within this category also provide a broad overview of the potential impact of ageing in several areas, including LTC.

41

6.0

38

5.6

29

4.3

46

6.8

32

4.7

26

3.8

22

3.2

Provision of LTC

Quality of care

Unpaid care interactions

Expenditure on LTC

Parent-child interactions

Ageing

(Continued)

147

David N. F. Bell and Elizabeth Lemmon Category

Description

No. of documents

Percentage of documents (%)

Labor market

This grouping focuses on paid rather than unpaid care workers. Important issues include wages, training, hours of work, professionalization, turnover, and remittances abroad where care workers have migrated from a lower-income country. Papers in this category examine whether the distribution of ownership (for example, among private for-profit, private not-for-profit, and public) in the nursing and care home sector affects outcomes. This includes comparisons of market share, wage differentials, and quality of care. Some related work considers the home care and hospice settings. Studies in this category include cost-effectiveness analyses of specific interventions, for example, differing home care programs or treatments specific to Alzheimer’s disease. It also contains methodological papers concerned with the inclusion of indirect costs and estimating quality of life. This category covers other areas such as productivity, efficiency, and technological change; the interaction between LTC and healthcare services; inequalities in access to services; economic evaluations of LTC, often for specific types of disease or conditions; housing and living arrangements of older people; overprescribing in nursing homes; and well-being and happiness of older adults using LTC.

14

2.1

11

1.6

10

1.5

43

6.3

679

100

Ownership

Economic evaluation

Miscellaneous

TOTAL

The complex nature of the relationship between paid and unpaid care makes determining causality difficult. Both forms of care provision can occur simultaneously, and much of the literature has grappled with attempts to disentangle the effects. These statistical problems may partly explain contrary findings. For example, Bonsang’s (2009) analysis of the impact of unpaid care on home care and nursing home care finds evidence of substitution for nursing care, but complementarity for home care. These results might suggest that the mechanism depends on the needs of the cared for and thus the type of care being provided. Other work also supports this finding (Jim´enez-Mart´ın and Prieto, 2012; Balia and Brau, 2013) and some find no significant effect of unpaid care at all (Mellor, 2001; Weaver and Weaver, 2014). More fundamentally, underlying the relationship between unpaid and formal services is the caregiver’s motivation for providing care in the first place. Understanding those motivations is crucial for implementing policy to support both caregivers and the cared for. Some of the wellcited economic literature in this area has focused on game theoretic models of the relationship between parents and adult children (and among siblings). These models incorporate altruism by allowing parents’ health status to jointly enter both parent and child’s utility functions (Pezzin and Schone, 1999). 148

149

Charles and Sevak Brown and Finkelstein

Hirth

Werblow, Felder, and Zweifel Brown and Finkelstein Van Houtven, Coe, and Skira Pezzin and Schone

Brown and Finkelstein

Why is the market for LTC insurance so small? The effect of informal care on work and wages Intergenerational household formation, female labor supply, and informal caregiving: A bargaining approach Consumer information and competition between nonprofit and for-profit nursing homes Can family caregiving substitute for nursing home care? The private market for LTC insurance in the United States: A review of the evidence

Informal care and healthcare use of older adults Multiple dimensions of private information: Evidence from the LTC insurance market Does informal care from children to their elderly parents substitute for formal care in Europe? The interaction of public and private insurance: Medicaid and the LTC insurance market Population ageing and healthcare expenditure: A school of “red herrings”?

Van Houtven and Norton Finkelstein and McGarry

Bonsang

Title

Author

Journal of Human Resources

1999

Journal of Health Economics Journal of Risk and Insurance

2005 2009

Journal of Health Economics

Journal of Health Economics

2013

1999

Journal of Public Economics

Health Economics

American Economic Review

2007

2007

2008

Journal of Health Economics

American Economic Review

2006

2009

Journal of Health Economics

Source Title

2004

Year

Table 8.2 Articles from search results ordered by number of citations

104

123

129

134

150

150

163

166

246

276

303

Cited by

Unpaid care interactions Demand

Ownership

Unpaid care interactions Parent-child interactions

Insurance

Expenditure

Insurance

Unpaid care interactions

Unpaid care interactions Insurance

Group

(Continued)

Article

Article

Article

Article

Article

Article

Article

Article

Article

Article

Article

Document Type

The Economics of Long-Term Care

150

Chapter 17, Long term care

Informal care and Medicare expenditures: Testing for heterogeneous treatment effects

Van Houtven and Norton

2008

Handbook of Health Economics Journal of Health Economics

Journal of Health Economics

2010

2000

Journal of Human Resources

American Economic Review

Journal of Health Economics

Journal of Finance

Health Economics, Policy and Law Journal of Risk and Uncertainty

Source Title

2002

2008

Preference heterogeneity and insurance markets: Explaining a puzzle of insurance

Shared caregiving responsibilities of adult siblings with elderly parents Who will care? Employment participation and willingness to supply informal care

2011

2011

Determinants of LTC spending: Age, time to death or disability?

Ameriks, Caplin, Laufer, and Van Nieuwerburgh De Meijer, Koopmanschap, d’Uva, and van Doorslaer Cutler, Finkelstein, and McGarry Checkovich and Stern Carmichael, Charles, and Hulme Norton

1997

2010

The 10 characteristics of the highperforming chronic care system Adverse selection, bequests, crowding out, and private demand for insurance: Evidence from the LTC insurance market The joy of giving or assisted living? Using strategic surveys to separate public care aversion from bequest motives

Ham

Sloan and Norton

Year

Title

Author

75

82

83

87

87

92

92

94

99

Cited by

LTC economics Expenditure

Parent-child interactions Unpaid care interactions

Insurance

Expenditure

Demand

Demand

Quality of care

Group

Article

Review

Article

Article

Conference Paper

Article

Article

Article

Article

Document Type

David N. F. Bell and Elizabeth Lemmon

The Economics of Long-Term Care

Checkovich and Stern (2002) set out to understand if children’s caregiving decisions are dependent across time and among siblings. Their theoretical model assumes that a child’s caregiving supply is a function of their own characteristics, their parent’s characteristics, and the total supply of care from their siblings. Using data from the National Long Term Care Survey in the US, they estimate structural models and find that children who live farther from their parents and who work and are married will provide less care to parents. As expected, their results also show that female children provide more care than men (Checkovich and Stern, 2002). They also find that a child’s decision to provide care depends on their expectations about the care provided by their siblings. Some of these models make simplifying assumptions that appear quite different from the reality of children’s decisions to provide care for a parent in need of care. For example, in some, children can make a simple decision to opt out of care provision (Engers and Stern, 2008), while in others, children compete with one another for future bequests (Bernheim et al., 1985). Nevertheless, these models provide the building blocks to understand a complex relationship between parents and their children in relation to care giving. One key determinant and implication of care provision is caregiver labor supply. Most evidence suggests that unpaid carers spend less time in the labor market than their noncaring counterparts. Once again, the endogenous nature of labor supply in the unpaid care provision decision makes establishing causality difficult. Are children supplying less labor because of demands on their time to provide unpaid care, or are children who supply less labor more likely to become unpaid carers? Understanding the labor market consequences of care provision has received significant attention in the economic literature. Van Houtven et al. (2013), who provide an overview, find that female unpaid caregivers who remain working reduce their time spent in the labor market by between 3 and 10 hours, and they also face wages that are around 3 percent lower than those of non-caregivers. Potential causes for the reduction in wages may be due to carers choosing to switch to lower paid and less demanding jobs, as they anticipate the need to increase care provision in the future (Van Houtven et al., 2013). However, research into the effect of labor market participation on caregiving suggests that labor market participation reduces the probability of unpaid care provision (Carmichael et al., 2010; He and McHenry, 2016). Moreover, individuals with higher wages are less likely to provide care (Carmichael et al., 2010), supporting the hypothesis that individuals with higher wages face higher opportunity costs in their decision to provide care (Nortons, 2000). To conclude, the evidence relating to unpaid care points to the need for policymakers to carefully balance both labor market and LTC policies to ensure that caregivers can manage employment and caregiving responsibilities.

8.6.2

Demand/Insurance

LTC insurance may be provided within the family. Kotlifkoff and Spivak (1981) argue that within families many of the problems of insurance markets, such as moral hazard and adverse selection, are lessened due to the level of trust and information sharing that exists within families. Arguing that consumption and bequest-sharing arrangements within marriage offer better forms of risk management than public markets, they argue for increased family formation. Family care remains the norm in many developing countries (Hirschfeld, 2003) but is increasingly under pressure due to population ageing, reductions in fertility, and growing pressure on women’s time. A long-standing puzzle has been the failure of the private insurance market to adequately support the LTC sector. LTC insurance supports individuals with physical and/or cognitive 151

David N. F. Bell and Elizabeth Lemmon

impairments to deal with ADLs either at home or in an institution. But the insurance market for LTC is weak, covering only 6.2 percent of LTC expenditures in the US, while government programs (Medicare and Medicaid) cover 67 percent of costs (Hagen, 2013). Brown and Finkelstein (2008) review the literature on private LTC insurance and argue that it demonstrates that, although standard theory suggests individuals are willing to insure against the potentially catastrophic financial risk posed by a need for LTC, take-up is limited. This is caused by actuarially unfair premiums, which in turn result from imperfect competition in the market, asymmetric information, and issues associated with dynamic contracting—problems that arise with insurance contracts where the payoff may be some time in the future and parties may have an incentive to renege on the insurance agreement. For example, as time elapses, individuals may decide that they are less likely to need LTC based on their current health status and decide to drop out of their policy, thus creating an adverse selection problem for insurance providers. Dynamic contracting problems can be avoided using a “point of need” policy based on housing equity. Point of need policies are responses to an established demand for long-term care. A fair premium should equate to the costs of the expected duration of stay in residential care, which in most cases equates to remaining life expectancy. Housing wealth provides the insurance against the need for LTC. In the UK, this approach is mandated to offset public sector care costs. Where individual assets exceed a (low) wealth threshold, public authorities can offset care costs by sequestering housing equity after their death. Pauly (1990) argues that government support for LTC undermines private sector insurance. He argues that Medicaid crowds out the private sector due to the “implicit tax” that it applies to the private product. This is due to the services that Medicaid provides for free, overlapping with those covered by private insurance policies, providing a disincentive to opting for private provision. For homeowners with a living partner, the use of rental income from the original dwelling to pay for accommodation costs in the new care setting will likely be ruled out. Consumption of housing by the remaining partner will rise, though the marginal improvement in utility is likely to be small. Even if there is no partner, market imperfections and/or legacy considerations may reduce rental opportunities.

8.6.3

Expenditure

Economists are consistently interested in how much is spent on LTC and how much is likely to be spent as the demographic transition progresses. A key debate is whether LTC spending increases with age or with time to death (Zweifel et al., 1999). Resolving what is known as the “red herring” hypothesis—i.e., that population ageing does not lead to increases in healthcare expenditure, rather, expenditures are high toward the end of life and as age increases more individuals are in their final year of life—has important implications for projections of LTC expenditure. These will rise with population ageing if age is the main determinant of LTC spending. Note that Stearns and Norton (2004) find that health expenditure would grow around 9–15 percent more slowly if time to death is the main determinant. However, the timeto-death argument in respect of LTC expenditure is unresolved. Norton (2016) summarizes relevant literature, much of which is concerned with disentangling the joint determination of LTC expenditures and time of death. He argues that empirical work favors the time-to-death argument, but that further work is required to determine how it varies by type of health or care spending and by relevant institutional and cultural contexts. Another perspective on LTC expenditure is its distribution across recipients. As with healthcare, spending on LTC appears to be highly skewed. Bakx et al. (2016) show that for those 152

The Economics of Long-Term Care

receiving LTC in the Netherlands, the top 1 percent account for more than half of total spending. This result is sensitive to how care needs are assessed because eligibility criteria define the left truncation point of the formal care costs distribution. Our search did not encompass all articles on LTC spending. Important omissions were the Hurd et al. (2013, 2015) articles on costs associated with dementia. This disease has an important bearing on trends in LTC costs because its prevalence increases sharply with age, treatments are limited, and most of the associated costs are care rather than health costs. These articles use survey information to inform projections of the costs of care in the US based on assumptions about future prevalence, out-of-pocket costs, claims on Medicare, and imputed costs of unpaid care. Cost projections are sensitive to the valuation of unpaid care. Relative to 2002 and across a plausible range of prevalence, costs are projected to increase by around 400 percent by 2040 if unpaid care is valued using foregone wages and by more than 500 percent where it is valued at replacement cost.

8.6.4

Labor Market

Our Scopus search clearly indicated interest in the labor market for care workers, which is an unusual labor market in several respects. A recent OECD report on the LTC workforce indicates that around 90 percent of care workers are women, who are mostly middle aged, and 20 percent are foreign born (OECD, 2020). More than 70 percent of LTC workers are personal carers with low qualification requirements, 56 percent are employed in institutions, and the rest work in domestic residences. Working conditions are often poor and wages low. Median wages for LTC workers in European countries were EUR 9 per hour, while hospital workers in broadly similar occupations earned EUR 14 per hour (OECD, 2020). Job-related physical and mental stresses are high, and contractual arrangements often provide little security. Qualifications are low and training limited. This combination of job characteristics leads to high turnover. This has become a major policy issue worldwide given that the rate of growth of the population aged 80+ now exceeds the rate of growth of care workers (OECD, 2020). Given these characteristics, one focus of the literature on the labor market for care workers has been the causes and effects of low wages. Machin et al. (2003) investigate the effect of the introduction of a minimum wage on the previously unregulated market for care workers. They find that the minimum wage raised the wages of the lowest paid care workers, but also compressed differentials, reducing the incentive to take additional responsibility. Matsudaira (2014) tests for the existence of monopsony—where the value of the wage falls below the value of the marginal product and another possible cause of low wages—in the market for LTC nurse aides in the US. He fails to establish its existence, arguing instead that “employers forced to increase their staffing levels are able to recruit as many new workers as they require at the market wage” (Matsudaira, 2014, p. 101). Whether this outcome holds in the future must be subject to some doubt, with the OECD projecting, for example, that based on present demographic and productivity trends, OECD countries jointly need to increase their LTC workforce by 60 percent between 2016 and 2040. Currently there are five LTC workers for every 100 people aged 65+. To hold that ratio through to 2040 would require an increase of 13.5 million care workers (OECD, 2020).

8.7

Conclusion and Discussion

This chapter aims to identify research relating to the economics of LTC across economics journals using a systematic search. It then summarizes some of the key issues that emerge from the most cited articles. Our analysis of the key research issues in the economics of LTC clearly 153

David N. F. Bell and Elizabeth Lemmon

shows that unpicking causal relationships is problematic, rendering the task of establishing effective policy interventions extremely difficult. What has become evident during this endeavor is that research on this important topic lacks focus, making it difficult to develop an overview of the main issues and to explore gaps. Although the OECD covers key issues in its “Ageing and Long-term Care” hub, research on the economics of LTC is spread across mainstream journals and health economics and medical journals. Another difficulty facing researchers is the paucity of good-quality data on LTC. This may deter researchers from engaging with LTC issues. The lack of quality data became obvious during the pandemic, as countries sought to understand the severity of the threat facing care homes due to COVID-19. In attempting to catalogue care-home deaths in the UK nations, Bell et al. (2020) first estimate the number of care homes in the UK, a previously unpublished statistic. If such rudimentary data are not available, more sophisticated analysis inevitably faces serious obstacles. The coverage of LTC data by international organizations, such as the OECD and the European Union, is also problematic. Tables are often incomplete because countries do not provide relevant data, and topics selected for inclusion in these tables only give a very partial view of the sector. Notable omissions from international comparisons include unpaid care, the organization of care at home, unmet need, care client scope to vary service provision, labor market conditions for care workers, and importantly the outcomes of care interventions. Whereas economic evaluation of healthcare treatments has been common practice in health economics for decades, collection of data on the outcomes of LTC provision, or specific interventions relating to LTC, has been limited. Our review identifies just 10 papers relating to economic evaluation. Although the application of economic evaluation within LTC is increasing, progress remains slow and concerns over methodological consistency have been raised (Tinelli et al., 2020; Weatherly et al., 2017). Our analysis shows that important research contributions have been made in areas such as the determination of costs, insurance, unpaid care, the labor market for care workers, etc. However, the lack of a key reference source for the economics of LTC handicaps researchers and policymakers seeking to build a comprehensive picture. The importance of such clarity seems now to justify the establishment of a dedicated journal, or at least a satellite journal around one of the main health economics or ageing journals, devoted to the economics of LTC.

Notes 1 Czech Republic, Denmark, Finland, France, Germany, Israel, Latvia, Lithuania, Luxembourg, Netherlands, Norway, Portugal, Slovenia, Spain, Sweden, Switzerland, and the United Kingdom. 2 Search strategy: SUBJAREA (econ) TITLE-ABS-KEY (“long term care”). Please see https://www. scopus.com/sources.uri for the full list of sources included in this subject area. 3 Search strategy: SUBJAREA (medi) TITLE-ABS-KEY (“long term care”) SUBJTERMS (2719) AND (LIMIT-TO (EXACTSRCTITLE, “Journal Of Medical Economics”) OR LIMIT-TO (EXACTSRCTITLE, “Pharmacoeconomics”) OR LIMIT-TO (EXACTSRCTITLE, “Health Economics United Kingdom”) OR LIMIT-TO (EXACTSRCTITLE, “Journal Of Health Economics”) OR LIMIT-TO (EXACTSRCTITLE, “Health Economics Policy And Law”) OR LIMIT-TO (EXACTSRCTITLE, “Health Economics”)). 4 These figures are based on Scopus results prior to removing duplicate and irrelevant documents.

References AMERIKS, J., CAPLIN, A., LAUFER, S., AND VAN NIEUWERBURGH S. (2011): “The joy of giving or assisted living? Using strategic surveys to separate public care aversion from bequest motives,” Journal of Finance, 66(2): 519–561. https://doi.org/10.1111/j.1540-6261.2010.01641.x.

154

The Economics of Long-Term Care

ARROW, K. J. (1963): “Uncertainty and the welfare economics of medical care,” American Economic Review, 53(5): 941–973. https://doi.org/10.1016/B978-0-12-214850-7.50028-0. BAKX, P., O’DONNELL, O., AND VAN DOORSLAER, E. (2016): “Spending on healthcare in the Netherlands: Not going so Dutch,” Fiscal Studies, 37(3–4): 593–625. https://doi.org/10.1111/j.14755890.2016.12114. BALIA, S., AND BRAU, R. (2013): “A country for old men? Long-term home care utilisation in Europe,” Health Economics, 23(10): 1185–1212. https://doi.org/10.1002/hec.2977. BELL, D. N. F., COMAS-HERRERA, A., HENDERSON, D., JONES, S., LEMMON, E., MORO, M., MURPHY, S., O’R EILLY, D., AND PATRIGNANI , P. (2020): “COVID-19 mortality and long-term care: A UK comparison.” Available at https://ltccovid.org/2020/08/28/covid-19-mortality-and-long-termcare-a-uk-comparison/. BERNHEIM, B. D, SHLEIFER, A., AND SUMMERS, L. H. (1985): “The strategic bequest motive,” Journal of Political Economy, 93(6): 1045–1076. BONSANG, E. (2009): “Does informal care from children to their elderly parents substitute for formal care in Europe?,” Journal of Health Economics, 28(1): 143–154. https://doi.org/10.1016/ j.jhealeco.2008.09.002. BROWN, J. R., AND FINKELSTEIN, A. (2007): “Why is the market for LTC insurance so small?,” Journal of Public Economics, 91(10): 1967–1991. https://doi.org/10.1016/j.jpubeco.2007.02.010. BROWN, J. R., AND FINKELSTEIN A. (2008): “The interaction of public and private insurance: Medicaid and the LTC insurance market,” American Economic Review, 98(3): 1083–1102. https://doi.org/10.1257/aer.98.3.1083. BROWN, J. R., AND FINKELSTEIN, A. (2009): “The private market for LTC insurance in the United States: A review of the evidence,” Journal of Risk and Insurance, 76(1): 5–29. https://doi.org/ 10.1111/j.1539-6975.2009.01286.x. CARERS UK. (2020): “Carers week 2020 research report.” Available at CW 2020 Research Report WEB.pdf (carersuk.org). CARMICHAEL, F., CHARLES, S., AND HULME, C. (2010): “Who will care? Employment participation and willingness to supply informal care,” Journal of Health Economics, 29(1): 182–190. https://doi.org/10.1016/j.jhealeco.2009.11.003. CHARLES, K. K., AND SEVAK, P. (2005): “Can family caregiving substitute for nursing home care?,” Journal of Health Economics, 24(6): 1174–1190. https://doi.org/10.1016/j.jhealeco.2005.05.001. CHECKOVICH, T. J., AND STERN, S. (2002): “Shared caregiving responsibilities of adult siblings with elderly parents,” Journal of Human Resources, 37(3): 441–478. https://doi.org/10.2307/3069678. CUTLER, D. M., FINKELSTEIN, A., AND MCGARRY, K. (2008): “Preference heterogeneity and insurance markets: Explaining a puzzle of insurance,” American Economic Review: Papers and Proceedings, 98(2): 157–162. https://doi.org/10.1257/aer.98.2.157. DE MEIJER, C., KOOPMANSCHAP, M., D’UVA, T. B., AND VAN DOORSLAER, E. (2011): “Determinants of LTC spending: Age, time to death or disability?,” Journal of Health Economics, 30(2): 425–438. https://doi.org/10.1016/j.jhealeco.2010.12.010. ENGERS, M, AND STERN, S. (2008): “Long-term care and family bargaining,” International Economic Review, 43(1): 73–114. https://doi.org/10.1111/1468-2354.t01-1-00004. ´ FERN ANDEZ , J.-L., FORDER, J. E., AND KNAPP, M. R. J. (2011): “Long-term care.” In: Smith, P., and Glied, S. (eds.), The Oxford Handbook of Health Economics. Oxford: Oxford University Press, pp. 578–601. FINKELSTEIN, A., AND MCGARRY, K. (2006): “Multiple dimensions of private information: Evidence from the LTC insurance market,” American Economic Review, 96(4): 938–958. https://doi.org/10.1257/aer.96.4.938. HAGEN, S. (2013): “Rising demand for long-term services and supports for elderly people.” Congress of the United States Congressional Budget Office. Available at https://www.cbo.gov/sites/ default/files/cbofiles/attachments/44363-LTC.pdf. HAM, C. (2010): “The ten characteristics of the high-performing chronic care system,” Health Economics, Policy and Law, 5(10): 71–90. https://doi.org/10.1017/S1744133109990120. HE, D., AND MCHENRY, P. (2016): “Does formal employment reduce informal caregiving?,” Health Economics, 25(7): 829– 843. https://doi.org/10.1002/hec.3185. HIRSCHFELD, M. (2003): Long-Term Care in Developing Countries: Ten Case-Studies. World Health Organization, Geneva, and JDC-Brookdale Institute, The WHO Collaborating Centre for Health and Long-Term Care Policy and Research.

155

David N. F. Bell and Elizabeth Lemmon

HIRTH, R. A. (1999): “Consumer information and competition between nonprofit and for-profit nursing homes,” Journal of Health Economics, 18(2): 219–240. https://doi.org/10.1016/S0167-6296(98)00035-6. HURD, M. D., MARTORELL, P., DELAVANDE, A., MULLEN, K. J., AND LANGA, K. M. (2013): “Monetary costs of dementia in the United States,” New England Journal of Medicine, 368(14): 1326–1334. https://doi.org/10.1056/NEJMsa1204629. HURD, M. D., MARTORELL, P., AND LANGA, K. (2015): “Future monetary costs of dementia in the United States under alternative dementia prevalence scenarios,” Journal of Population Ageing, 8(1–2): 101. 10.1007/s12062-015-9112-4. JIM E´ NEZ-MART´I N, S., AND PRIETO, C. V. (2012): “The trade-off between formal and informal care in Spain,” European Journal of Health Economics, 13(4): 461–490. https://doi.org/10.1007/s10198-0110317-z. KOTLIFKOFF, L., AND SPIVAK, A. (1981): “The family as an incomplete annuities market,” Journal of Political Economy, 89(2): 372–391. https://doi.org/10.1086/260970. MACHIN, S., MANNING, A., AND RAHMAN, L. (2003): “Where the minimum wage bites hard: Introduction of minimum wages to a low wage sector,” Journal of the European Economic Association, 1(1): 154–180. https://doi.org/10.1162/154247603322256792. MATSUDAIRA, J. D. (2014): “Monopsony in the low-wage labor market? Evidence from minimum nurse staffing regulations,” Review of Economics and Statistics, 96(1): 92–102. https://doi.org/10.1162/REST a 00361. MELLOR, J. M. (2001): “Long-term care and nursing home coverage: Are adult children substitutes for insurance policies?,” Journal of Health Economics, 20(4): 527–547. https://doi.org/10.1016/S01676296(01)00078-9. NAC (NATIONAL ALLIANCE FOR CAREGIVING). (2015): “Caregiving in the US.” Available at https://www.aarp.org/content/dam/aarp/ppi/2015/caregiving-in-the-united-states-2015-reportrevised.pdf#: :text=Caregiving%20in%20the%20U.S.%202015%20should%20be%20considered,for%20 examining%20changes%20to%20caregiving%20in%20the%20future. NORTON, E. (2016): “Health and LTC.” In: The Handbook of the Economics of Population Ageing, Volume 1. Chapter 16, pp. 951–989. http://dx.doi.org/10.1016/bs.hespa.2016.06.001. NORTON, E. C. (2000): “Long term care.” In: Anthony, J. C., and Newhouse, J. P. (eds.), Handbook of Health Economics, Volume 1, Part B. Chapter 17, Amsterdam: Elsevier, pp. 955–994. OECD (ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT). (2018): “Care needed: Improving the lives of people with dementia,” OECD Health Policy Studies, OECD Publishing, Paris. http://dx.doi.org/10.1787/9789264085107-en. OECD (ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT). (2019): “Health at a glance.” Available at https://www.oecd-ilibrary.org/sites/4dd50c09-en/1/3/11/7/ index.html?itemId=/content/publication/4dd50c09-en& csp =82587932df7c06a6a3f9dab95304095d& itemIGO=oecd&itemContentType=book. OECD (ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT). (2020): “Who cares? Attracting and retaining care workers for the elderly.” Available at https://www. oecd.org/health/who-cares-attracting-and-retaining-elderly-care-workers-92c0ef68-en.htm. PAULY, M. V. (1990): “The rational nonpurchase of long-term care insurance,” Journal of Political Economy, 98(1): 153–168. https://doi.org/10.1086/261673. PEZZIN, L. E., AND SCHONE, B. S. (1999): “Intergenerational household formation, female labor supply and informal caregiving: A bargaining approach,” Journal of Human Resources, 43(3): 475–503. https://doi.org/10.2307/146377. SICILIANI, L. (2014): “The economics of long-term care,” The B. E. Journal of Economic Analysis and Policy, 14(2): 343–375. https://doi.org/10.1515/bejeap-2012-0063. SLOAN, F. A., AND NORTON, E. C. (1997): “Adverse selection, bequests, crowding out, and private demand for insurance: Evidence from the LTC insurance market,” Journal of Risk and Uncertainty, 15: 201–219. https://doi.org/10.1023/A:1007749008635. STEARNS, S. C., AND NORTON, E. C. (2004): “Time to include time to death? The future of healthcare expenditure predictions,” Health Economics, 13(4): 315–327. https://doi.org/10.1002/hec.831. TINELLI, M., BAUER, A., KNAPP, M., KETTUNEN, A., AND GUY, D. (2020): “Making the economic case for adult social care: The EconomicS of Social carE CompEndium (ESSENCE) Project,” Journal of Long-Term Care, 174–184. Online continuous volume and issue. https://doi.org/10.31389/jltc.3. VAN HOUTVEN, C. H., COE, N. B., AND SKIRA, M. M. (2013): “The effect of informal care on work and wages,” Journal of Health Economics, 32(1): 240–252. https:// doi.org/10.1016/j.jhealeco.2012.10.006.

156

The Economics of Long-Term Care

VAN HOUTVEN, C. H., AND NORTON, E. C. (2004): “Informal care and healthcare use of older adults,” Journal of Health Economics, 23(6): 1159–1180. https://doi.org/10.1016/j.jhealeco.2004.04.008. VAN HOUTVEN, C. H., AND NORTON, E. C. (2008): “Informal care and Medicare expenditures: Testing for heterogeneous treatment effects,” Journal of Health Economics, 27(1): 134–156. https://doi.org/10.1016/j.jhealeco.2007.03.002. WEATHERLY, H., NEVES DE FARIA, R., VAN DEN BERG, B., ET AL. (2017): “Scoping review on social care economic evaluation methods.” Discussion Paper. CHE Research Paper. Centre for Health Economics, York, UK. Available at http://eprints.whiterose.ac.uk/135405/. WEAVER, F. M., AND WEAVER, B. A. (2014): “Does availability of informal care within the household impact hospitalisation,” Health Economics Policy and Law, 9(1): 71–94. https://doi.org/10.1017/1S744133113000169. WERBLOW, A., FELDER, S., AND ZWEIFEL, P. (2007): “Population ageing and healthcare expenditure: A school of ‘red herrings’?,” Health Economics, 16(10): 1109–1126. https://doi.org/10.1002/hec.1213. ZWEIFEL, P., FELDER, S., AND MEIERS, M. (1999): “Ageing of population and healthcare expenditure: A red herring?,” Health Economics, 8(6): 485–496. https://doi.org/10.1002/(SICI)10991050(199909)8:6¡485::AID-HEC461¿3.0.CO;2-4.

157

9 IN GOOD AND BAD TIMES—ASSOCIATIONS BETWEEN SPOUSAL HEALTH AND ASSORTATIVE MATCHING ON EARLY-LIFE FACTORS IN EUROPE1 Iris Kesternich, Bettina Siflinger, and James P. Smith

Abstract In this chapter, we analyze spousal associations in late-life spousal health, both physical and mental, in a sample of elderly Europeans from the Survey of Health and Retirement in Europe. We document that correlations in couples’ health are strong and exhibit strong regional differences. Health associations in couples are strongest in the South, followed by Central Europe, and they are lowest in Northern Europe. We investigate the role of assortative matching on early-life factors to explain both associations in late-life health and their regional patterns. We estimate a matching model that allows for multiple continuous attributes. Assortative mating on earlylife factors is strong, and it follows a similar regional pattern as associations in spousal health. By linking our matching estimates to spousal health correlations, we find that matching on early-life factors explains only little of the variation in late-life spousal health associations. This is in line with research showing that matching on early-life factors matters more for health correlations in the early stage of marriage. We conduct a counterfactual analysis by imposing matching preferences from one region on couples in another region, showing that regional differences in spousal associations are indeed at least in part caused by differences in matching on early-life factors. Our findings open up new possibilities for future research combining insights from health economics with the matching literature.

9.1

Introduction

In this chapter, we study the association in spousal health across regions in Europe. The literature discusses four factors that may cause positive correlations in spousal health: First, assortative matching on health and socioeconomic status in early life may lead to positive associations in spousal health later in life. Second, spouses share a joint environment. For example, they face 158

DOI: 10.4324/9781003150398-10

Spousal Health and Assortative Matching

the same prices and are exposed to the same pollutants. Third, spousal health behaviors may become more similar over time, or they may have direct spillover effects on partners’ health. Fourth, the health of one spouse may have direct effects on the other’s health, for example, when one spouse infects the other with an infectious disease or when the burden of providing care affects spousal health. In past research, we have documented surprising regional differences in spousal lung function, a continuous measure of health, across European regions (Banks et al., 2021). In this chapter, we take a broader look at whether and why associations in spousal health across different regions in Europe are different. In particular, we want to gain a better understanding of the role of assortative matching on pre-marriage characteristics, i.e., childhood health, childhood socioeconomic status (SES), and education. Thus, this chapter focuses on the first factor that causes positive associations in spousal health. Our analysis is based on the theory of marriage markets as first proposed by Becker (1973, 1974). If two individuals decide to form a couple, they can combine time and market goods to produce household commodities, for example, health. If a couple is more productive in producing these commodities than two single households, there are gains from marriage. Individuals possess certain pre-marriage attributes, for example, education. These attributes can augment joint productivity. If they are complementary in producing gains, we would observe positive assortative matching at the time of marriage. Concerning health, we can assume that pre-marriage attributes can then help to produce late-life health in the spirit of Grossman’s (1972) health production function approach. Associations in spousal health are also interesting to study in their own right. Positive correlations in spousal health may exacerbate existing inequalities in income or socioeconomic status. Care needs may be larger if spousal health is positively correlated. As the recent COVID-19 crisis shows, healthcare systems may want to take a more systematic approach to testing and curing disease at the family level instead of at the individual level. Spillover effects of vaccination policies in the family were discussed already before the COVID-19 epidemic (Bouckaert et al., 2020, for influenza vaccinations). Targeting health policies at families rather than individuals has also been suggested for reducing obesity rates (Brown et al., 2014). Studying differences in spousal health associations among regions helps to shed light on the long-run determinants of health. Differences across regions allow us to observe more variation in explanatory factors than studying one region alone. In addition, much of the variation across regions is exogenous from an individual perspective. For our empirical analysis, we use the Survey of Health and Retirement in Europe (SHARE).2 One advantage of SHARE is that it collects health and sociodemographic information using the same measures across Europe, thus allowing for cross-national comparisons. Moreover, SHARE collects information on both spouses. And because SHARE uses retrospective modules to collect data on early-life health and socio demographics, we can link current associations in health to partners’ pre-marriage characteristics. We concentrate our analysis on a random sample of older couples from 11 European countries. We define “couple” here as a man and a woman who cohabit. To avoid having to deal with many different countries with small sample sizes, we split Europe into three regions (South, Central, and North) so that grouped countries are similar in terms of economic development, climate, and culture. The first step in our analysis is documenting patterns of associations in different health measures across regions in Europe. As outcomes, we choose two measures of physical health that are frequently used in the literature (major and minor conditions) and one measure of mental health (depression). To have a relatively homogeneous sample and to avoid having to deal with the age-cohort problem, we take the age of 70 as our focal age and take for each respondent the 159

Iris Kesternich et al.

SHARE wave where the woman is closest to 70 years old. We take women as the starting point for our analysis, so we select female age and use female measures as outcomes.3 Across Europe, all spousal health measures show strong and significantly positive associations. Health associations are strongest in the South of Europe, followed by Central Europe, and they are weakest in the North. For instance, an increase in the spouse’s depression score by one standard deviation (SD) increases a woman’s by 0.42 SD in the South, by 0.30 SD in Central Europe, but only by 0.19 SD in the North. Correlations in physical health show a similar pattern. Patterns of associations in health behaviors (smoking, drinking, and exercising) mainly show the opposite regional pattern than those of physical and mental health measures. Smoking and drinking behaviors are much more strongly related in the North of Europe than they are in Central Europe, and they are least strongly related in the South. Thus, while health behaviors may well contribute to differences in health levels, they do not seem to be drivers of differences in spousal health associations across regions. By contrast, associations in childhood health and SES follow the same pattern as health associations, i.e., stronger associations in the South than in Central and Northern Europe. We then investigate assortative mating based on early-life factors and their role in health associations later in life more closely. To this end, we estimate a static matching model with transferable utility and multiple continuous matching attributes (Dupuy and Galichon, 2014; Chiappori et al., 2020a,b). This matching model provides us with structural estimates of matching preferences, thus allowing us to investigate which and to what extent pre-marital factors matter for assortative matching in the three European regions. With these structural estimates at hand, we conduct a counterfactual exercise to predict how similar spousal adult health would have been in the South when sorting patterns had been those of the North and vice versa. Our results show that assortative matching is strong and significant in all three dimensions of early life. Assortative matching on early-life factors is much stronger in the South of Europe than in Central Europe and is weakest in the North. In all three regions, matching is stronger on education and childhood SES than on childhood health. In the North, education is the most important matching attribute, while in Southern Europe childhood SES is the dominant matching factor. In Central Europe, education and childhood SES are equally important. Our model predicts that assortative matching on these early-life factors explains only a small share of associations in adult health ranging from 0.5–10 percent depending on the health measure and the region. By contrast, the stark regional gradient in associations in late-life health can be partially explained by differences in assortative matching among the regions. This is in line with findings by Guner et al. (2018) showing that selection into marriage on health-related factors is more important for young couples than for old couples. However, we believe that more research is needed to investigate the role that early-life factors play in spousal health associations over the life course. Combining insights from the literature on assortative matching with the literature on health economics may be a fruitful way forward. Another avenue for future research that we want to suggest with this chapter is to learn more about and from regional differences in spousal health associations. Using cross-country surveys and life-history data may prove a good combination for doing this (Banks et al., 2020).

9.2

Literature

One strand of the health economics literature investigates the effect of the discrete state of being single, married, divorced, or widowed on health (e.g., Siflinger, 2017). It is well documented that being married or cohabiting tends to increase individuals’ health, but causal effects are more difficult to establish as unobservables may determine both health and marital decisions (Kohn 160

Spousal Health and Assortative Matching

and Averett, 2014). Another strand goes one step further and investigates what we can learn from spousal associations in health, combining Becker’s (1973, 1974) theory of sorting in marriage markets with Grossman’s (1972) model of health production. The seminal contribution on the model side is Wilson (2002); a more recent model of cumulative health exposures was developed in Davillas and Pudney (2017). The main difference to the Grossman model is that not only own attributes (e.g., education) are used to produce health but also the partner’s attributes. Spousal associations in health help to clarify which factors impact health over the life course. One important question is whether partners influence each other in their health behaviors or whether they were similar in the first place. If spousal health is mainly triggered by the partner’s behavior, this implies that health problems might be better addressed at the family level instead of the individual level. Several empirical studies are investigating the causes and consequences of spousal associations in health and health behaviors. These studies examine spousal associations in body mass index (BMI) to address the obesity epidemic (Brown et al., 2014), or correlations in health behaviors such as smoking (Clark and Etile, 2006), exercising (Farrell and Shields, 2002), and drinking behavior (Leonard and Mudar, 2003). Associations in spousal health are also documented in the literature on assortative matching. The main goal of this literature is to understand the sorting of the marriage market and its implications for fertility, investments in education, labor market participation, as well as inequality and intergenerational mobility (Becker, 1973; Fernandez and Rogerson, 2001; Fernandez et al., 2005). Showing that partners match on health is normally just a side-product of documenting patterns of assortative matching. For example, while Dupuy and Galichon (2014) are mainly interested in the role of personality traits on the marriage market, they also control for height and BMI showing that assortative matching on both factors is strong. Chiappori et al. (2012) document that there exist trade-offs between socioeconomic status and BMI on the marriage market. Chiappori et al. (2020a) analyzed multidimensional sorting, and they show that partners are similar regarding anthropometric measures and health behaviors. We also contribute to the literature on childhood health and SES on later life health (Almond and Currie, 2011; Goodman et al., 2011). While literature has documented that childhood is crucial in determining late-life health, it concentrates on individual outcomes rather than on the couple level. If it turns out that selection at the time of marriage matters most for spousal health later in life, this would suggest that factors determined early in life play a more important role than later life behaviors and environment. An implication would then be that policies should address health inequalities early in life rather than targeting later-life behavior. Table 9.1 gives a (by no means complete) overview of correlations of spousal health found in the literature for different health outcomes and countries. Most studies take a single-country perspective (exceptions are Banks and Smith, 2012; Banks et al., 2021), and only a few account for childhood conditions. With this chapter, we thus would like to bridge the literature in health economics and the matching literature by using a structural model on assortative matching on early-life factors and linking it to patterns of spousal health associations across European regions.

9.3

Data

This study uses data from SHARE. The SHARE baseline wave (2004/2005) includes nationally representative samples in 11 European countries (Austria, Belgium, Denmark, France, Germany, Greece, Italy, Netherlands, Spain, Sweden, and Switzerland) drawn from population registries,or using multistage sampling (http://www.share-project.org/). SHARE collects information on sociodemographics, health (self-reported health, health conditions, height, BMI, biomarkers), and health behaviors (smoking, drinking, exercise) of the respondents. Importantly, SHARE 161

Iris Kesternich et al. Table 9.1 Literature overview Authors (year)

Data

Health Measure

Spousal Correlation

Banks et al. (2014)

United Kingdom: Longitudinal Survey of Ageing (ELSA), United States: Health and Retirement Study (HRS) The Netherlands: cross-sectional health interview survey 1997–2008 British Household Panel Survey (BHPS)

Correlations in adult health measures and smoking

Strong and significant correlations (stronger in the United Kingdom than in the United States) when controlling for both partners’ ages

Spillover in partner’s vaccination rate

0.037, not significant at 10% level

Correlation in BMI

United States: Panel Study of Income Dynamics (PSID) Framingham Heart Study 1971–2003

Correlations in BMI

Men: spousal BMI 0.125** Women: spousal BMI 0.174** Controls: own and spousal characteristics Raw spousal correlation: 0.094**

Clark and Etile (2006)

BHPS

Davillas and Pudney (2017)

United Kingdom Household Longitudinal Survey, BHPS

Conditional probability of smoking if partner smokes Age-adjusted measure of homogamy based on self- and nurse-administered biomarker measures

Di Castelnuovo et al. (2009)

Seventy-one papers from MEDLINE, PubMed, and EMBASE databases

Many major coronary risk factors

Dupuy and Galichon (2014)

The Netherlands: De Nederlandsche Bank Household Survey, 1993–2002 1997 Health Survey of England

BMI associations

Bouckaert et al. (2020)

Brown et al. (2014)

Chiappori et al. (2012)

Christakis and Fowler (2007)

Farrell and Shields (2002)

Spousal risk for obesity

Sports participation

162

Men: 44%** increase if wife becomes obese Women: 37* increase if husband becomes obese Men: 52.8%*** Women: 51.2%***

Self-assessed health: 0.172–0.266*** Self-reported cardiovascular condition: 0.077*** Obesity: 0.382–0.398*** Blood pressure + heart rate: 0.181–0.284*** (not: diastolic blood pressure) Biomarkers: cholesterol 0.232*** Strongest spousal correlations: Smoking: 0.23**, BMI: 0.15** Significant positive correlations for diastolic blood pressure, triglycerides, total and low density lipoprotein cholesterol, weight, and waist/hip ratio (0.06–0.11**) Affinity matrix estimates: BMI: 0.21**, height: 0.18**, selfrated health: 0.14** Error correlation (random effect probit) for participation in sports: 0.357***

Spousal Health and Assortative Matching Authors (year)

Data

Health Measure

Spousal Correlation

Guner et al. (2018)

United States: PSID Medical Expenditure Panel Survey

Self-reported health: 0.17–0.29*** Innate permanent health: 0.28–0.37***

Howe et al. (2019)

United Kingdom Biobank United States: PSID

(1) Self-reported health (2) Innate permanent health (from group fixed effects estimator) Weekly alcohol consumption Association with spousal BMI

Oreffice and QuintanaDomeque (2010)

Palali and Van Ours (2017)

The Netherlands: DNB Household Survey

Decision to quit smoking

Saarela et al. (2019)

Finland: Register data, individuals aged 40–65, 1987–2011

Silventoinen et al. (2003)

Finnish Twin Cohort Study

Receipt of sickness allowance (SA) Receipt of disability pension (DP) Height and BMI: phenotypic assortment (PA) versus social homogamy (SH)

Wilson (2002)

United States: 1992 Health and Retirement Study

Self-assessed general health status (SAGHS), disability index (IFLAR), chronic disease index (WCDI)

Spousal alcohol consumption: 0.26*** units White couples (marital duration ≤ 3 years > 3 years): Women: 0.394***/0.261*** Men: 0.351***/0.181*** Controls: own and partner’s age, number of children, education, earnings, health. Same pattern but lower associations for black couples. No significant effect of spousal decision after controlling for common unobserved heterogeneity and common shocks. Risk of SA 50%** higher in first year of partner’s receipt of SA Risk of DP 100%** higher in first year of partner’s receipt of DP Height: PA: 0.27** men, 0.28** women SH: 0.24** men, 0.29** women BMI: PA: 0.13** men, 0.13** women SH: 0.31** men, 0.28** women Age 51–55: SAGHS: 0.249***, IFLAR: 0.227***, WCDI: 0.179*** Age 56-61: SAGHS: 0.260***, IFLAR: 0.245***, WCDI: 0.221**

Note: *** p < 0.01, ** p < 0.05, * p < 0.1 indicates statistical significance of estimates in studies.

conducts interviews with both partners in the household. Waves 3 and 7 of SHARE (SHARELIFE) collect retrospective histories covering the period in respondents’ lives before the first baseline interview. This information was collected using autobiographical life history calendar methods. Information in SHARELIFE includes a history of marriage over the full life course of respondents and several measures of early-life health and socioeconomic status. We also use information on current and previous relationships of our respondents. We use waves 1–7, covering the period until 2017. We group the 11 European countries into three regions. Southern Europe consists of Greece, Italy, and Spain; Denmark and Sweden represent Northern 163

Iris Kesternich et al.

Europe; and the remaining countries (Austria, Belgium, France, Germany, Netherlands, and Switzerland) form Central Europe. Table 9.2 shows how we constructed our estimation sample. We go from the full sample of women aged 65–74 [Column (1)] to our estimation sample of cohabiting individuals in their first relationship with nonmissing key variables for both partners [Column (5)]. We use only respondents who are in their first relationship because we do not observe information about previous partners. One concern may be that the institution of marriage means something different in the three regions, with marriage and cohabitation being much less common in Northern European countries, and couples being more likely to separate in the North than in the South of Europe. While this may certainly be true today, it is less of a concern for the generation of respondents in our sample. Column (2) of Table 9.2 shows that the share of women who are married or cohabiting is only somewhat lower in Northern Europe (90 percent) than in Central (92 percent) and Southern Europe (97 percent). Regions differ in whether individuals live in their first relationship: 70 percent in the North, 81 percent in Central Europe, and 96 percent in Southern Europe [see Column (4)]. However, within our sample of first-relationship cohabitations, the share of married individuals (99–100 percent) and the average relationship duration are very similar across regions (between 45.6 and 46.5 years).4 Table 9.3 presents the definitions of the variables used in our analysis. Most variables we use throughout the empirical analysis are standardized by region and gender. Most variable definitions are standard. An exception is our measure of childhood SES, which is an index constructed from a principal component analysis based on variables capturing different dimensions of early-life SES.5 Such early-life health and SES measures have been shown to pass several tests of internal and external validity when compared with other data on the same cohorts (Smith, 2009; Havari and Mazzonna, 2015) and have now been used in several applied studies in labor and health economics (see, for instance, Kesternich et al., 2014, 2015; van den Berg et al., 2016). Table 9.4 presents descriptive statistics for all variables used in the empirical analysis by gender and region. Overall, most variables follow a regional pattern. Regardless of gender, the number of minor conditions and the depression score are highest in Southern Europe and lowest in the North with Central Europe being in between. Moreover, spouses in the South have lower childhood SES than in other regions, and most individuals living in Southern European countries

Table 9.2 Sample selection: women of age 65–74 years (1)

(2)

(3)

(4)

(5)

(6)

(7)

European regions

Age 65–74

Married/ Cohabiting

Matched couples SHARE

First relationship

Nonmissing variables

Married

Duration relationship, years

Northern Europe

1,497

Southern Europe

2,615

767 (69.98%) 2,492 (80.91%) 2,074 (95.84%)

613 (99.03%) 1,940 (99.59%) 1,697 (99.24%)

46.48

3,954

1,096 (73.21%) 3,080∗ (77.99%) 2,164 (82.75%)

619

Central Europe

1,351 (90.25%) 4,657 (92.49%) 2,525 (96.56%)

Total

8,066

7,533

6,340

5,333

4,277

4,250

46.12

Note: ∗ Five same-sex couples removed.

164

1,948 1,710

46.50 45.60

Spousal Health and Assortative Matching

have achieved only primary education (59–65 percent). In the North, about one-third of men and women have obtained a tertiary degree. The gender pattern in health measures is consistent across regions. Women tend to have a higher depression score while men have a higher number of major and minor health conditions. Men and women seem to be similar in childhood SES, but women report a somewhat lower health status during childhood. In Southern and Central Europe, a higher share of men than women holds a tertiary educational degree, while educational achievement in couples is rather balanced in the North.

Table 9.3 Variable definitions Adult health outcomes Nr major conditions Nr minor conditions Depression

Cancer, heart attack, chronic lung disease, stroke, hip/femoral fracture, Parkinson’s disease Cataracts, diabetes/high blood sugar, high blood cholesterol, high blood pressure/hypertension, stomach/duodenal/peptic ulcer The EURO-D depression scale from 0 (Not depressed) to 12 (Very depressed). Following Prince et al. (1999), the EURO-D score is a composite index computed from a 12-item question battery: (1) being sad or depressed in the last month, (2) personal hopes for future, (3) suicidal feelings in last month, (4) tendency toward self-blame or guilt, (5) recent sleeping trouble, (6) having and keeping up with interest in things in last month, (7) being irritable recently, (8) increasing or decreasing appetite and eating, (9) too little energy to do things, (10) difficulty concentrating, (11) recent enjoyment, (12) crying in last month.

Adult behaviors Smokes now

Dummy indicating 1 if smokes at the present time and 0 otherwise

Ever smoked

Dummy indicating 1 if ever smoked daily and 0 otherwise

Drinks a lot (3 months)

Dummy indicating 1 if drinks at least 5 days per week in the past 3 months and 0 otherwise

Drinks a lot (6 months)

Dummy indicating 1 if drinks at least 5 days per week in the past 6 months and 0 otherwise

Exercising

Dummy indicating 1 if exercises weekly (once or more a week) and 0 otherwise

Childhood health and socioeconomic status Childhood health

Self-rated childhood health on a scale from 1 to 5 according to poor, fair, good to very good, and excellent

Childhood SES

We follow Mazzonna (2014) and run a principal component analysis based on the following variables measuring childhood SES at the age of 10: rooms per capita in individual’s dwelling, number of books in household, occupation of main breadwinner (white- or blue-collar worker), presence of various facilities and amenities in house (fixed bath, cold running water supply, inside toilet, central heating). The index for childhood SES corresponds to the score of first factor that is standardized by region and gender. (Continued)

165

Iris Kesternich et al. Adult health outcomes Regions North

Dummy indicating 1 if from Northern Europe (Sweden, Denmark) and 0 otherwise

Central

Dummy indicating 1 if from Central Europe (Austria, Belgium, France, Germany, Netherlands, Switzerland) and 0 otherwise

South

Dummy indicating 1 if from Southern Europe (Greece, Italy, Spain) and 0 otherwise

Education Education

Highest level of education attained according to ISCED-97, ranging from ISCED 0 (pre-primary) to ISCED 6 (second stage of tertiary). These are regrouped into three levels: up to primary (ISCED 0–1), secondary (ISCED 2–4), tertiary (ISCED 5–6). For details see https://ec.europa.eu/eurostat/ cache/metadata/Annexes/educ uoe h esms an2.html.

Table 9.4 Descriptive statistics by region and gender South Male Adult health outcomes Nr major conditions

Central Female

Male

Female

North Male

Female

0.38 (0.63) 1.10 (0.99) 2.00 (2.10)

0.21 (0.48) 1.04 (0.98) 2.87 (2.51)

0.38 (0.63) 0.94 (0.95) 1.63 (1.71)

0.22 (0.50) 0.90 (0.93) 2.31 (1.98)

0.37 (0.61) 0.90 (0.95) 1.32 (1.48)

0.21 (0.47) 0.84 (0.92) 1.67 (1.62)

72.80 (4.31)

69.38 (2.50)

71.61 (3.99)

69.45 (2.55)

71.91 (3.82)

69.66 (2.34)

Childhood health and SES Childhood SES -0.47 (0.91) Childhood health 4.12 (0.92)

-0.46 (0.93) 4.00 (0.95)

0.23 (0.94) 3.77 (1.02)

0.22 (0.94) 3.69 (1.01)

0.57 (0.89) 4.17 (0.96)

0.54 (0.87) 4.09 (0.99)

Education Up to primary Secondary Tertiary

0.65 0.29 0.06

0.14 0.56 0.30

0.18 0.63 0.19

0.24 0.45 0.32

0.20 0.45 0.35

Nr minor conditions Depression

Age

Observations

0.59 0.32 0.10

1,710

1,948

Note: Standard deviations are in parentheses.

166

619

Spousal Health and Assortative Matching

9.4

Associations in Spousal Adult Health and Its Determinants across Regions

In this section, we estimate associations between partners’ physical and mental health for the pooled sample and separately for each region. We begin by regressing a woman’s health (major or minor conditions for physical and depression for mental health) on her partner’s health. To control for any association that is due to partners being of similar age, we include an age quadratic of both partners in our models. Our model is specified as follows. healthf = α + βhealthm + γ agepolm + δagepolf + uf , where f denotes female and m denotes male. healthf is a woman’s health, healthm is her partner’s health, and agepolf and agepolm denote age and age squared of a woman and her partner, respectively. α is the intercept, and uf is an iid error term. The spousal health correlations are captured by the parameter β. The first column of Panel A in Table 9.5 reports the estimated spousal health ˆ pooling all three regions. We find strong and significant associations in spousal correlations β, health for all three health measures. Associations are weakest for major conditions, which are relatively rare. An increase in male number of major conditions by one SD, increases female major conditions by 0.06 SDs. An increase in male minor conditions by 1 SD increases a woman’s number of minor conditions by 0.15 of an SD. The increase is largest for depression, with an estimated association of 0.33 SDs. All estimated coefficients are significant at the 1 percent level. Table 9.5 Regression coefficients of own health outcomes on partner’s health outcomes, controlling for both partners’ ages and ages squared Panel A: Female health outcomes on partner’s health outcomes, female sample

Nr major conditions, std Nr minor conditions, std Depression, std Observations

All regions

South

Central

North

0.06∗∗∗

0.08∗∗∗

0.06∗∗

(0.02) 0.15∗∗∗ (0.02) 0.33∗∗∗ (0.02)

(0.03) 0.19∗∗∗ (0.02) 0.42∗∗∗ (0.02)

(0.03) 0.14∗∗∗ (0.02) 0.30∗∗∗ (0.02)

0.04 (0.04) 0.08∗ (0.04) 0.19∗∗∗ (0.04)

4,277

1,710

1,948

619

Panel B: Male health outcomes on partner’s health outcomes, male sample

Nr major conditions, std Nr minor conditions, std Depression, std Observations

All regions

South

Central

North

0.06∗∗∗ (0.02) 0.13∗∗∗ (0.02) 0.34∗∗∗ (0.02)

0.09∗∗∗ (0.03) 0.14∗∗∗ (0.02) 0.40∗∗∗ (0.03)

0.05∗∗ (0.02) 0.14∗∗∗ (0.02) 0.33∗∗∗ (0.02)

-0.01 (0.04) 0.06 (0.04) 0.18∗∗∗ (0.04)

4,863

2,064

2,141

658

Note: Robust standard errors in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1; ordinary least squares regression of women’s health measures on men’s health measures. Variables are standardized (std) by region. 167

Iris Kesternich et al.

In Columns (2)–(4), we report these estimated coefficients separately for each region. Our estimation results show a strong South-North gradient. Associations of both physical and mental health within couples are larger in the South of Europe than in Central Europe, and they are smallest in the North. Differences are largest for depression, both in absolute and in relative terms. In the South, a 1 SD increase in the male depression score is associated with an increase in the female depression score by 0.42 SD compared with an increase of 0.30 SD in Central Europe and of only 0.19 SD in the North. Panel B of Table 9.5 shows the estimated correlations based on a sample selected on men’s age and using male health measures as outcomes. The results are qualitatively similar. Table 9.6 presents associations in possible determinants of the health associations. Panel A presents the estimated associations in health behaviors, i.e., smoking, drinking, and exercising.

Table 9.6 Regression coefficients of own attributes on partner’s attributes, controlling for both partners’ ages and ages squared, female sample Panel A: Health behaviors

Smokes now Observations Ever smoked Observations Drinks a lot (3 months) Observations Drinks a lot (6 months) Observations Exercising Observations

All regions

South

Central

North

0.13∗∗∗ (0.02) 3,085

0.09∗∗∗ (0.03) 1,174

0.14∗∗∗ (0.03) 1,487

0.25∗∗∗ (0.08) 424

0.13∗∗∗ (0.01) 4,262

0.08∗∗∗ (0.02) 1,705

0.16∗∗∗ (0.02) 1,941

0.23∗∗∗ (0.04) 616

0.42∗∗∗ (0.02) 2,095

0.29∗∗∗ (0.03) 728

0.39∗∗∗ (0.03) 1,061

0.65∗∗∗ (0.05) 306

0.48∗∗∗ (0.04) 366

0.39∗∗∗ (0.07) 127

0.47∗∗∗ (0.06) 190

0.70∗∗∗ (0.09) 49

0.29∗∗∗ (0.01) 4,273

0.26∗∗∗ (0.02) 1,710

0.29∗∗∗ (0.02) 1,945

0.22∗∗∗ (0.04) 618

All regions

South

Central

North

0.52∗∗∗ (0.01) 0.29∗∗∗ (0.02) 0.51∗∗∗ (0.01) 4,277

0.60∗∗∗ (0.02) 0.41∗∗∗ (0.02) 0.57∗∗∗ (0.02) 1,710

0.49∗∗∗ (0.02) 0.22∗∗∗ (0.02) 0.46∗∗∗ (0.02) 1,948

0.36∗∗∗ (0.04) 0.17∗∗∗ (0.04) 0.50∗∗∗ (0.03) 619

Panel B: Childhood SES and education

Childhood SES, std Childhood health, std Education, std Observations

Note: Robust standard errors in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1; ordinary least squares regression of women’s attributes on men’s attributes. Variables on health behaviors are dummies. Variables on childhood SES, childhood health, and education are standardized (std) by region. 168

Spousal Health and Assortative Matching

Health behaviors are binary variables that take the value 1 if an individual exhibits this type of behavior and is zero otherwise (see Table 9.3 for their definitions). For drinking behavior, we use two different measures because the time frame for reporting drinking behavior in SHARE was changed several times. Health behaviors show strong and positive associations—more so for drinking than for exercising and smoking. However, health behavior is an unlikely explanation for the South-North pattern in health associations. Unlike for late-life health measures, where we observe the strongest associations in the South and the weakest in the North, smoking and drinking behaviors are very strongly related in the North (0.65–0.70 for drinking, 0.23–0.25 for smoking), while these associations are much weaker in Central Europe (0.39–0.47 for drinking, 0.14–0.16 for smoking) and the South (0.29–0.39 for drinking, 0.08–0.09 for smoking). By contrast, associations in exercising show a moderate North-South gradient: with an estimated coefficient of 0.29, the association is strongest in Central Europe, followed by the South (0.26), and is weakest in the North (0.22). Panel B of Table 9.6 shows estimated spousal associations in early-life factors, i.e., childhood SES, childhood health, and education. These three early-life measures are standardized by region and gender to ensure comparability across regions. In contrast to health behaviors, all early-life factors show the same South-North pattern as associations in late-life health. In the South, spousal correlations are between 0.41 of an SD for childhood health and 0.60 for childhood SES, while they are only between 0.17 (childhood health) and 0.50 (education) of an SD in the North. Because our measure of self-rated childhood health may suffer from social desirability or recall bias, as a robustness check we use an alternative measure of childhood health. Respondents were asked about illnesses and health conditions they had during childhood (up to and including age 15). They could select from a list of 19 childhood health conditions. We sum the number of health conditions by respondent and standardize by region. Using this alternative measure of child health, our estimation results show a strong South-North gradient and are qualitatively the same as for self-rated childhood health.

9.5

Assortative Matching on Childhood Background and Adult Health

In the last section, we documented strong associations in early-life factors with a similar SouthNorth pattern as associations in adult health measures. We now estimate a structural model to investigate the importance of assortative mating on early-life factors across regions, and we explore how different patterns of matching translate into differences in spousal adult health.

9.5.1

The Matching Model

Our structural matching model is based on Becker’s (1973) seminal analysis of matching in marriage markets. He proposed a two-sided assignment model with transferable utility, allowing the joint utility generated by a couple to be shared between both partners. The key idea of Becker’s model is that many goods (e.g., health) are produced within the household, using time and market goods. Men and women are characterized by several attributes, for example, education, that can increase household productivity. Utility maximization becomes equivalent to maximizing household production. When forming marriages, both spouses are assumed to match with the opposite sex to maximize their own payoffs. With one-dimensional matching characteristics, Becker (1973) shows that positive complementarities in household production imply positive assortative matching. Empirically, we follow Choo and Siow (2006), assuming that spouses match on multiple attributes where some are observed and some are unobserved by the researcher. 169

Iris Kesternich et al.

Consider a man i with multiple attributes y ∈ Y ⊆ Rdy and a woman j with multiple attributes x ∈ X ⊆ Rdx . The separable extreme value (SEV) approach assumes that the joint surplus from matching, 8(x, y) + ϵij (x, y), is the sum of a deterministic joint utility and a stochastic term that reflects unobserved heterogeneity in matching. This unobserved heterogeneity term is assumed to be additive separable in individual-specific terms, depending only on the other spouse’s matching attribute, i.e., ϵij (x, y) = αi (y) + βj (x). Under the assumption that αi (y) and βj (x) are extreme value type I distributed, Choo and Siow (2006) show that the joint utility is the sum of deterministic components of individual utilities, 8(x, y) = U(x, y) + V (x, y), such that at stable matching the respective utilities of matched men and women are given by U(x, y) + αi (y)

and

V (x, y) + βj (x).

Recent work by Dupuy and Galichon (2014) generalizes the SEV approach to continuous matching attributes and thus to a continuous choice setting. A key idea is to control the dimensionality of the estimation problem by assuming a parametric (quadratic) form for the deterministic component of joint utility, 8(x, y) = x Ay = ′

dy dx X X

Akl xl yk ,

l=1 k=1

where the affinity matrix A is a dx ×dy matrix, and k, l are specific attributes in X and Y . For two attributes yk and xl , the parameter Akl measures the change in the marginal gain in joint utility from an increase in a man’s kth attribute as a woman’s lth attribute increases. For Akl > 0, yk and xl are complements and reflect positive assortative matching between a man’s kth attribute and a woman’s lth attribute. If Akl < 0, the attributes k and l of a man and a woman, respectively, are substitutes that indicate negative assortative matching between yk and xl . Thus, the affinity matrix provides a measure of the intensity or strength of positive or negative assortativeness (mutual attractiveness) between all pairs of attributes X and Y , identifying the most important determinants for matching among several correlated attributes. A is estimated by maximum likelihood (for the estimation algorithm, see Chiappori et al., 2020a).

9.5.2

Results: Matching Attributes and Health Outcomes

We apply the matching model with continuous matching attributes to the three European regions.6 Our matching attributes for men and women are childhood SES, childhood health, and the highest educational degree obtained. For all health measures, we use residuals net of quadratic age and we standardize them for each region. Table 9.7 presents the estimated affinity matrix. The main diagonal reflects the intensity in complementarity/substitutability between the same type of attributes of men and women. The interactions in different attributes are presented by the off-diagonal elements. While we find strong and positive matching of early childhood characteristics in all three regions, there are important regional differences. Assortative matching is weakest in Northern Europe and strongest in Southern Europe, in particular with respect to childhood SES and childhood health. Increasing childhood SES of both partners by 1 SD increases the joint utility 170

Spousal Health and Assortative Matching Table 9.7 Estimates of the affinity matrix for three European regions, quadratic specification A. Southern Europe Men/Women

Childhood SES

Childhood health

Education

Childhood SES

0.82 (0.05) -0.08 (0.04) 0.04 (0.04)

-0.01 (0.04) 0.46 (0.03) -0.00 (0.04)

-0.03 (0.04) 0.03 (0.04) 0.70 (0.04)

Childhood SES

Childhood health

Education

0.54 (0.04) 0.06 (0.03) 0.13 (0.03)

-0.00 (0.03) 0.23 (0.02) 0.03 (0.03)

0.12 (0.03) -0.05 (0.03) 0.55 (0.04)

Childhood health Education

B. Central Europe Men/Women Childhood SES Childhood health Education

C. Northern Europe Men/Women Childhood SES Childhood health Education

Childhood SES

Childhood health

Education

0.27 (0.05) 0.06 (0.05) 0.21 (0.06)

0.07 (0.05) 0.16 (0.04) -0.01 (0.05)

0.18 (0.06) 0.02 (0.05) 0.58 (0.06)

Note: Southern European countries are Greece, Italy, and Spain. N = 1,710 couples with women aged 65–74 during the observation period. Central European countries are Austria, Belgium, France, Germany, the Netherlands, and Switzerland. N = 1,948 couples with women aged 65–74 during the observation period. Northern European countries are Denmark and Sweden. N = 619 couples with women aged 65–74 during the observation period. The samples contain one couple observation between 2003 and 2017 in SHARE. For couples observed several times during the observation period, we choose that observation in which a woman is closest to age 70. Positive values on main diagonal reflect positive assortative matching; bold-faced coefficients are statistically significant at the 5% level.

from matching by 0.27 units in Northern Europe, 0.54 units in Central Europe, and 0.82 units in Southern Europe. Increasing childhood health of both partners by 1 SD increases the joint utility by 0.16 units in the North, 0.23 units in Central Europe, and 0.46 units in the South. Education tends to be a very important matching attribute in all three regions. Increasing education of both partners by 1 SD increases the joint matching utility by 0.58 units in the North, 0.55 units in Central Europe, and 0.70 units in the South. This implies that the relative importance of attributes differs across regions. In the North, education is by far the most important matching attribute, while in Central Europe, childhood SES and education are equally important. In the South, childhood SES is the most important matching attribute. For all regions and all three 171

Iris Kesternich et al.

attributes, all estimates on the main diagonal (thus matching the same type of attribute between men and women) are statistically significant at the 5 percent level. We next link the estimated matching patterns to spousal health correlations. Because the largest differences in matching patterns occur between the South and the North, we focus on these two regions. For each of these two regions, we use the estimated preference parameters, i.e., the estimated affinity matrix, to compute the equilibrium probability distribution for all potential matches between men and women, and then calculate spousal correlations in both matching characteristics and health outcomes. These are the correlations from matching in the equilibrium. In equilibrium, every individual from the same region with the opposite gender is assigned a positive probability to be a potential match, not only the own spouse. Column (1) in Table 9.8 shows the actual sample correlations between spouses’ matching characteristics and health outcomes in Southern Europe. As expected, correlations are positive and highly significant. Column (2) shows the corresponding correlations obtained from equilibrium matching. Matched spouses are very similar to each other in their childhood characteristics. Columns (4) and (5) of Table 9.8 show the corresponding actual and equilibrium correlations for Northern Europe. Spousal correlations on matching characteristics are high and follow the expected pattern with the highest similarity in education. In both regions, actual correlations and equilibrium correlations are essentially identical, which demonstrates that our model has done a good job in fitting the equilibrium early-life characteristics to the actual ones. The lower panel of Table 9.8 shows that in such an optimal matching situation, spousal health correlations are surprisingly low. We compare observed correlations to those predicted in equilibrium in the South [Columns (1) and (2)]. We find that correlations in equilibrium are 5.1 percent of the actual correlations for major conditions, 1.1 percent for minor conditions, and 2.9 percent for depression. In the North, correlations in equilibrium are 5.4 percent of the actual correlations for major conditions, 10.4 percent for minor conditions, and only 0.5 percent for depression [Columns (4) and (5)]. Overall, this suggests that only a minor fraction of observed spousal health correlations can be explained by assortative matching on early-life

Table 9.8 Correlations in matching characteristics and health outcomes with actual preferences and with counterfactual preferences Southern Europe Sample correlation

Northern Europe

Preferences

Sample

Preferences

South

North

correlation

North

South

A. Matching attributes Childhood SES 0.609∗∗∗ Childhood health 0.415∗∗∗ Education 0.623∗∗∗

0.609∗∗∗ 0.415∗∗∗ 0.623∗∗∗

0.387∗∗∗ 0.163∗∗∗ 0.600∗∗∗

0.361∗∗∗ 0.178∗∗∗ 0.542∗∗∗

0.361∗∗∗ 0.178∗∗∗ 0.542∗∗∗

0.583∗∗∗ 0.431∗∗∗ 0.568∗∗∗

B. Health measures Nr major conditions Nr minor conditions Depression

0.004∗∗ 0.002 0.012∗∗∗

0.003∗∗ 0.001 0.010∗∗∗

0.037 0.077∗ 0.188∗∗∗

0.002 0.008∗∗ 0.001

0.002 0.011∗∗ 0.001

0.079∗∗∗ 0.191∗∗∗ 0.414∗∗∗

Note: *** p < 0.01, ** p < 0.05, * p < 0.1; p-values for spousal health and early-life correlations in equilibrium and in the counterfactual are obtained from 1,000 bootstrap replications.

172

Spousal Health and Assortative Matching

characteristics. In Southern and Northern Europe, about 5 percent of the observed correlations in spouse’s number of major conditions can be explained by assortative matching on earlylife characteristics. For minor conditions, the explained share is even lower in the South, but matching explains about 10 percent in the North. For depression, the pattern differs significantly between the South and the North. In the South, about 3 percent of the correlation of late-life depression is explained by matching on early childhood characteristics, while almost none of the correlation in the North is explained by assortative matching on early-life characteristics. Why are these correlations so low compared with health correlations of actual couples? In the observed sample, the matching probability with the observed spouse is equal to one. With optimal matching, a large portion of the matching probability is shifted to other potential partners in the sample. As a consequence, the strong observed health correlations between spouses reduce considerably if potential partners are less similar in their health outcomes than actual partners. If we allow for optimal mating, the predicted spousal health correlations are very small. This implies that assortative matching purely based on childhood characteristics contributes only little to the large and positive spousal health correlations observed in our sample. We conclude that even though assortative matching on childhood characteristics is strong and exhibits a similar regional pattern as late-life associations in health, it is not the main driver of adult health associations. In our sample of elderly Europeans, other factors, for example, a shared environment, must be responsible for associations in physical and mental health conditions later in life. How do our results compare with findings from the literature based on different samples, health measures, and models? Our findings are well in line with Guner et al. (2018) who use a grouped fixed effects estimator to investigate the relation between marriage and health. They find based on U.S. data that selection based on health-related characteristics is important for young couples, while it plays less of a role for older couples, such as those in our sample. Using UK data, Davillas and Pudney (2017) estimate a model of cumulative health exposures to show that shared lifestyle factors and assortative matching are about equally important. The average marriage duration in their sample is 22 years, so about half of that in our sample. Thus, the importance of early-life factors depends heavily on the type of health measure used, but even more important might be the length of marriage. One hypothesis that future research may address is whether the importance of early-life factors indeed declines with relationship duration. We finally assess how the correlation in spousal matching characteristics and health would change if the matching preferences in one region were assigned to another region. This informs us about how spousal correlations would change in a new, counterfactual matching equilibrium in each region.7 We perform this counterfactual analysis for the South and the North. In the ˆN first scenario, we use the estimated matching preferences from Northern Europe (e.g., A opt ) and S S the observed matching characteristics and health outcomes (X , Y ) from Southern Europe to compute the new optimal probability distribution of matches in the South [Column (3) in Table 9.8]. In the second scenario, we impose the Southern preferences on the North [Column (6)]. The upper panel shows that, as expected, Northern preferences would strongly reduce the correlations in childhood SES and childhood health in the South, while Southern preferences would lead to much stronger correlations in all early-life factors in the North. What does this mean for health correlations? The most important message is that both regions would become more similar regarding health outcomes. Take depression as an example. Actual spousal associations in depression in the South are a factor of 2.2 larger than in the North [see Columns (1) and (4) in Table 9.8]. Taking the numbers in the matching equilibrium [Columns (2) and (5)], we would even expect associations to be 12 times stronger in the South than in the North. When imposing Northern preferences on the South and Southern preferences on 173

Iris Kesternich et al.

the North, the North-South difference reduces to a factor of 10 [Columns (3) and (6)]. We see a similar pattern for correlations in the number of major health conditions: Correlations are larger in the South than in the North by a factor of 2, regardless of whether we use the actual correlations or the correlations predicted in equilibrium. When we swap preferences, the difference in the correlations reduces to a factor of 1.5. With minor conditions, the predicted pattern in the counterfactual would reverse. While actual correlations are larger in the South than in the North by a factor of 2, correlations in equilibrium matching are about four times stronger in the North than in the South and would increase by a factor of 10 if Southern preferences were imposed on couples living in Northern Europe. By and large, we observe that in the counterfactual scenario spousal health correlations would become smaller in Southern Europe while they would be (slightly) stronger in the North. Thus, we conclude that regional differences in spousal associations are indeed linked to differences in early-life factors. Note that we only use childhood SES, childhood health, and highest education for matching. Certainly, other attributes may also matter for assortative matching, which are omitted in our empirical analysis. Some of these attributes may contribute to a similar health development of couples. If we had information on such (omitted) attributes we may obtain stronger correlations in health between spouses. However, given that in equilibrium assortative matching on early-life characteristics is so strong and the associations in late-life health are so small, these omitted factors would have to change health outcomes by orders of magnitude to change our conclusion that assortative matching on early-life factors is not a major predictor of spousal heath associations later in life. Another reason for our findings is that couples are heterogeneous in their health development after matching has taken place. Even though different couples may be similar with respect to their matching attributes, they may face different health developments during their relationships. For instance, they live in different environments and are exposed to different couple-specific shocks. Thus, couples who are similar in pre-marital characteristics may be quite different in health later in life. To understand this mechanism in more detail, one would have to come up with couples’ health measures, e.g., after marriage, and follow them into late life. Thus, one potential for future research could be to formulate a dynamic structural model that accounts for endogenous choices during marriage and for differences in environments and shocks and incorporates assortative matching in equilibrium.

9.6

Conclusion

In this chapter, we investigate associations of spousal health, their regional patterns, and the role of assortative matching on early-life factors for determining these associations. We document that associations in spousal health are strong and that there are strong regional differences in these associations. Spouses are more similar regarding their late-life health in the South than in Central Europe and least similar in the North. We also show that assortative matching on early-life factors is strong and follows the same pattern across the regions. However, assortative matching on childhood characteristics explains only little variation in late-life spousal health variations. The magnitude of the share of late-life spousal health associations explained by assortative matching on early-life factors ranges from a very small fraction of associations in depression (0.5 percent) to a bit more than 10 percent of the variation in minor conditions in Northern Europe. In the South, between 1.1 and 5.1 percent of the variation in late-life health outcomes is explained. We then conduct a counterfactual analysis in which we impose Northern preferences for earlylife matching in the South and Southern preferences on the North. Here, we indeed observe 174

Spousal Health and Assortative Matching

that regions become much more similar. Thus, regional differences in spousal associations are indeed at least in part caused by differences in matching early-life factors. Understanding regional differences in spousal health associations may become a fruitful strand of future research. An interesting extension of our research would be to investigate how matching on early-life factors differs between younger and older couples. Also, it would be interesting to learn more about how differences in exposure to different environments contribute to spousal health associations, something that is less explored in the literature. Regional differences in assortative matching on early-life characteristics are also an intriguing topic for future research. Are they driven by culture, institutions, or inequality?

Notes 1 We thank Edoardo Ciscato for providing us with his estimation code and with very helpful comments. We thank Zelie Dresse and Han Pham for valuable research assistance. 2 This paper uses data from SHARELIFE release 7.1.0, as of June 26, 2020, and SHARE release 7.1.0, as of June 26, 2020. SHARE data collection was primarily funded by the European Commission through the 5th framework and 6th framework program. SHARELIFE was supported through the 7th framework program. 3 Alternatively, we select male age and use male measures as outcomes. We show that our results also hold if we use men instead. 4 We also re-estimate our correlations for individuals who are not in their first relationships, and our results remain qualitatively very similar. 5 We perform the principal-component analysis separately for region and gender. The index measuring childhood SES corresponds to the score of the first factor. Across regions and gender, the first factor explains between 41 and 48 percent of the total variance. 6 Edoardo Ciscato kindly shared his R code to estimate the matching model, see Chiappori et al. (2020a). 7 Our counterfactual analysis is similar to the one in Ciscato and Weber (2020) who use investigate how income inequality would change if marital preferences from earlier years were used to match in later years.

References ALMOND, D., AND CURRIE, J. (2011): “Killing me softly: The fetal origins hypothesis,” Journal of Economic Perspectives, 25(3): 153–172. BANKS, J., BRUGIAVINI, A., AND PASINI, G. (2020): “The powerful combination of cross-country comparisons and life history data,” The Journal of the Economics of Ageing, 16 [online]. Available at: 10.1016/j.jeoa.2019.100206 [Accessed: 27th April 2021]. BANKS, J., KELLY, E., AND SMITH, J. P. (2014): “Spousal health effects: The role of selection,” In: Wise, D. (ed.), Discoveries in the Economics of Ageing. Chicago: University of Chicago Press, pp. 255–283. BANKS, J., KESTERNICH, I., AND SMITH, J. P. (2021): “International differences in interspousal health correlations,” Health Economics. DOI: 10.1002/hec.4253. BANKS, J., AND SMITH, J. P. (2012): “International comparisons in health economics: Evidence from ageing studies,” Annual Reviews in Economics, 4(1): 57–81. BECKER, G. S. (1973): “A theory of marriage: Part I,” Journal of Political Economy, 81(4): 813–846. BECKER, G. S. (1974): “A theory of marriage: Part I,” Journal of Political Economy, 82(2): S11–S26. BOUCKAERT, N., GIELEN, A. C., AND VAN OURTI, T. (2020): “It runs in the family—Influenza vaccination and spillover effects,” Journal of Health Economics, 74: 102386. BROWN, H., HOLE, A. R., AND ROBERTS, J. (2014): “Going the same ‘weigh’: Spousal correlations in obesity in the United Kingdom,” Applied Economics, 46(2): 153–166. CHIAPPORI, P. A., CISCATO, E., AND GUERRIERO, C. (2020a): “Analyzing matching patterns in marriage: Theory and application to Italian data,” HCEO Working Paper Series, (No. 2020–080). CHIAPPORI, P. A., COSTA-DIAS, M., CROSSMAN, S., AND MEGHIR, C. (2020b): “Changes in assortative matching and inequality in income: Evidence for the UK,” Fiscal Studies, 41(1): 39–63.

175

Iris Kesternich et al.

CHIAPPORI, P. A., OREFFICE, S., AND QUINTANA-DOMEQUE, C. (2012): “Fatter attraction: Anthropometric and socioeconomic matching on the marriage market,” Journal of Political Economy, 120(4): 659–695. CHOO, E., AND SIOW, A. (2006): “Who marries whom and why,” Journal of Political Economy, 114(1): 175–201. CHRISTAKIS, N. A., AND FOWLER, J. H. (2007): “The spread of obesity in a large social network over 32 years,” New England Journal of Medicine, 357(4): 370–379. CISCATO, E., AND WEBER, S. (2020): “The role of evolving marital preferences in growing income inequality,” Journal of Population Economics, 33(1): 307–347. CLARK, A. E., AND ETIL E´ , F. (2006): “Don’t give up on me baby: Spousal correlation in smoking behaviour,” Journal of Health Economics, 25(5): 958–978. DAVILLAS, A., AND PUDNEY, S. (2017): “Concordance of health states in couples: Analysis of selfreported, nurse-administered and blood-based biomarker data in the UK Understanding Sociey Panel,” Journal of Health Economics, 56: 87–102. DI CASTELNUOVO, A., QUACQUARUCCIO, G., DONATI, M.B., DE GAETANO, G., AND IACOVIELLO, L. (2009): “Spousal concordance for major coronary risk factors: A systematic review and meta-analysis,” American Journal of Epidemiology, 169(1): 1–8. DUPUY, A., AND GALICHON, A. (2014): “Personality traits and the marriage market,” Journal of Political Economy, 122(6): 1271–1319. FARRELL, L., AND SHIELDS, M. A. (2002): “Investigating the economic and demographic determinants of sporting participation in England,” Journal of the Royal Statistical Society: Series A (Statistics in Society), 165(2): 335–348. FERNANDEZ, R., GUNER, N., AND KNOWLES, J. (2005): “Love and money: A theoretical and empirical analysis of household sorting and inequality,” The Quarterly Journal of Economics, 120(1): 273–344. FERNANDEZ, R., AND ROGERSON, R. (2001): “Sorting and long-run inequality,” The Quarterly Journal of Economics, 116(4): 1305–1341. GOODMAN, A., JOYCE, R., AND SMITH, J. P. (2011): “The long shadow cast by childhood physical and mental problems on adult life,” Proceedings of the National Academy of Sciences, 108(15): 6032–6037. GROSSMAN, M. (1972): “On the concept of health capital and the demand for health,” Journal of Political Economy, 80(2): 223–255. GUNER, N., KULIKOVA, Y., AND LLULL, J. (2018): “Marriage and health: Selection, protection, and assortative mating,” European Economic Review, 109: 162–190. HAVARI, E., AND MAZZONNA, F. (2015): “Can we trust older people’s statements on their childhood circumstances?,” European Journal of Population Health, 31(3): 237–257. HOWE, L. J., ET AL. (2019): “Genetic evidence for assortative mating on alcohol consumption in the UK Biobank,” Nature Communications, 10(1): 1–10. KESTERNICH, I., SIFLINGER, B., SMITH, J. P., AND WINTER, J. K. (2014): “The effects of World War II on economic and health outcomes across Europe,” The Review of Economics and Statistics, 96(1): 103–118. KESTERNICH, I., SIFLINGER, B., SMITH, J. P., AND WINTER, J. K. (2015): “Individual behavior as pathway between early life shocks and adult health: Evidence from hunger episodes in post-war Germany,” The Review of Economics and Statistics, 125(November): F372–F393. KOHN, J. L., AND AVERETT, S. L. (2014): “The effect of relationship status on health with dynamic health and persistent relationships,” Journal of Health Economics, 36: 69–83. LEONARD, K. E., AND MUDAR, P. (2003): “Peer and partner drinking and the transition to marriage: A longitudinal examination of selection and influence processes,” Psychology of Addictive Behaviors, 17(2): 115. MAZZONNA, F. (2014): “The long-lasting effects of family background: A European comparison,” Economics of Education Review, 40: 25–42. OREFFICE, S., AND QUINTANA-DOMEQUE, C. (2010): “Anthropometry and socioeconomics among couples: Evidence in the United States,” Economics & Human Biology, 8(3): 373–384. PALALI, A., AND VAN OURS, J. C. (2017): “Love conquers all but nicotine: Spousal peer effects on the decision to quit smoking,” Health Economics, 26(12): 1710–1727. PRINCE, M. J., ET AL. (1999): “Development of the EURO-D Scale: A European Union initiative to compare symptoms of depression in 14 European countries,” British Journal of Psychiatry, 174(4): 330–338.

176

Spousal Health and Assortative Matching

SAARELA, J., STANFORS, M., AND ROSTILA, M. (2019): “In sickness or in health? Register-based evidence on partners’ mutual receipt of sickness allowance and disability pension,” Social Science & Medicine, 240: 112576. SIFLINGER, B. (2017): “The effect of widowhood on mental health—An analysis of anticipation patterns surrounding the death of a spouse,” Health Economics, 26(12): 1505–1523. SILVENTOINEN, K., KAPRIO, J., LAHELMA, E., VIKEN, R. J., AND ROSE, R. J. (2003): “Assortative mating by body height and BMI: Finnish twins and their spouses,” American Journal of Human Biology, 15(5): 620–627. SMITH, J. P. (2009): “Re-constructing childhood health histories,” Health Economics, 26(12): 387–403. VAN DEN B ERG, G. J., P INGER , P.R., AND S CHOCH , J. (2016): “Instrumental variable estimation of the causal effect of hunger early in life on health later in life,” Economic Journal, 126(591): 465–506. WILSON, S. E. (2002): “The health capital of families: An investigation of the inter-spousal correlation in health status,” Social Science & Medicine, 55(7): 1157–1172.

177

10 MENTAL HEALTH AND ILLNESS IN AGEING Sherry Glied, Carolyn D. Gorman, and Richard Frank

Abstract This chapter discusses the economics of mental health and illness in the context of ageing. We describe how key characteristics of mental illnesses may affect social and economic outcomes across the life course. We provide perspectives on the epidemiology of these illnesses in a life course context. We then discuss the social welfare challenges generated by the interplay of mental illness and ageing, including effects on the health system.

10.1

Introduction

Mental illnesses impose significant burdens on older adults. Many global burden of disease studies point to mental illnesses (and substance use disorders) as the leading cause of disability across the entire population (Whiteford et al., 2016). As the global population ages, and as more people with mental illness survive to older ages, the share of older adults with mental illness in the overall population is likely to rise. Mental illnesses, however, manifest themselves differently in older adults; for example, the suicide rate for men over the age of 85 is more than four-fold that of the general population (Conwell et al., 2011). The category “mental illness” comprises a broad range of conditions, many quite common, that account for substantial impairment and disability worldwide (Ormel et al., 1994). Though these conditions differ in specifics, they share characteristics that generate distinctive disruptions to economic circumstances and affect human welfare in ageing. They interfere with cognitive and social functioning and are stigmatized in many cultures (Krendl and Pescosolido, 2020). Most mental illnesses have a relatively early age of onset and subsequently reoccur (or become exacerbated) episodically over the life course (Kessler et al., 2007). Their manifestation and exacerbation are sensitive to individuals’ physical, economic, personal, and social circumstances (Regier et al., 1993). In this chapter, we first describe how key characteristics of mental illnesses may affect social and economic outcomes across the life course. We then describe the epidemiology of these illnesses in a life course context. In the third section, we discuss the social welfare challenges generated by the interplay of mental illness and ageing, including their effects on the functioning of the overall healthcare system. 178

DOI: 10.4324/9781003150398-11

Mental Health and Illness in Ageing

10.2

Common Characteristics of Mental Illness Relevant to Ageing

Mental illnesses are common. Though estimates vary cross-nationally, between one-fifth and one-half of the population will experience symptoms consistent with a mental illness in their lifespans (Kessler et al., 2007). According to the National Alliance for Mental Illness, the largest U.S. advocacy group for people with mental illness: “A mental illness [or a mental health condition] is a condition that affects a person’s thinking, feeling, behavior, or mood. These conditions deeply impact day-to-day living and may also affect the ability to relate to others” (NAMI, 2021). While cognitive dysfunction is not an inevitable consequence of all mental illnesses, many of these conditions affect attention, motivation, working memory, executive function, speed of processing information, and social cognition (Millan et al., 2012). People with mental illness are more likely to be socially isolated and disengaged and/or disconnected from the labor market. For example, analyses of the General Social Survey in the United States show that people who experienced eight or more days of poor mental health in the prior month were about 10 percentage points less likely to vote in presidential elections than were those with no days of poor mental health (authors’ analysis of the General Social Survey 1972—2014 datafile; 2008 election). People in poor mental health, across the full range of conditions, are two to four times as likely as those without any conditions to report themselves as not in the labor force, not a student, not a homemaker, and not retired. Mental illnesses are associated with narrower social networks (Macdonald et al., 2000). They affect individuals’ ability to marry or stay married and reduce family size (in regards to number of children) (Bartel and Taubman, 1986). The importance of social connections in later life has been documented: in a large sample from the U.S. Health and Retirement Study (12,030 participants), loneliness is associated with a 40 percent increased risk of dementia when controlling for social isolation, clinical behavioral, and genetic risk factors (Sutin et al., 2020). Among individuals with Alzheimer’s disease, friendships as part of social networks can be significantly related to cognition (Balouch et al., 2019). The cognitive and interpersonal consequences of mental illnesses can intensify the challenges of ageing. The previously noted functional impairments (attention, working memory, executive function, speed of processing information, and social cognition) are often confused with the normal ageing process and frequently go unaddressed, limiting the opportunity for disease management. These functional impairments also mean that individuals with mental illness may make decisions that do not reflect the choices they would make if they were healthy, affecting both physical and economic well-being. Some research by the U.S. Federal Reserve Board suggests, for example, that household financial health itself may decline among those whose primary financial decisionmaker’s cognitive capacity declines (Hsu and Willis, 2013).

10.2.1

Compounded Consequences of Mental Illness for Social and Financial Well-Being due to Early Onset, Chronic Persistence, and Episodic Recurrence

Even though mental illnesses can occur at any point over the life course, the most common age of initial onset for some of the most severe illnesses is adolescence and early adulthood (Jones, 2013). Common mental disorders that are present in teenage years are strongly associated with poor mental health in middle age. Some of the most severe illnesses like schizophrenia and bipolar disorder, which begin early in life, persist throughout the life course, though the severity of episodes may vary over time (Haro et al., 2018). Of course, many people experience transient symptoms or an episode of mental illness—anxiety or depression, for example—in adolescence, early adulthood, and later ages, and 179

Sherry Glied et al. Table 10.1 Ages at which 25, 50, 75, and 99 percent of respondents are first affected by select mental disorders

Notes: SE = standard error. Projected lifetime risk indicates the proportion of the entire population that will have experienced the listed disorder by age 75. Data for standardized age-at-onset distributions of selected DSM-IV diagnoses derive from the World Health Organization Composite International Diagnostic Interview with projected lifetime risk at age 75. Surveys collecting this cross-national data were completed between 2002 and 2005. Source: Table adapted from Kessler et al. (2007); greater detail on specific diagnoses and sample available there.

the condition resolves, either through treatment or on its own, and does not recur (Bock et al., 2009). However, about half the time common mental illnesses, such as depression, recur episodically (Burcusa and Iacono, 2007). While symptoms that develop indeterminately or in a transitory fashion make determining the true earliest instances of onset difficult, several longitudinal studies following broad population samples defined by birth cohort or period contribute to the understanding of the prevalence of mental disorders by examining features in retrospect (Wadsworth, 1991; Colman et al., 2007; Poulton et al., 2015). In addition, cross-sectional surveys, such as the National Comorbidity Survey Replication (NCS-R) provide age of onset estimates for some diagnoses. Table 10.1 indicates the standardized ages at which select mental disorders first affect NCS-R respondents and the proportion of the population that will have experienced a mental disorder by age 75 (their projected lifetime risk). The early onset of severe and persistent mental illnesses means that people with these conditions often cannot accumulate human or financial capital and must rely on public income support programs (such as Supplemental Security Income in the United States; Patel et al., 2018). While early treatment may attenuate this trajectory, existing treatments typically cannot reverse the full functional effects of these conditions (Millan et al., 2015). A small number of individuals with severe and persistent mental illnesses become and may remain unable to live independently, and many depend on social programs or family members for assistance navigating daily life (Tj¨ornstrand et al., 2020). They are persistently overrepresented among the most disadvantaged groups—those who are homeless, incarcerated, or live in extreme poverty—in most countries (Fazel et al., 2008; Hirschtritt and Binder, 2017; Banks et al., 2017). Individuals with severe and persistent mental illnesses constitute about 2–5 percent of the population, depending on the definition used (Henderson et al., 2012; Frank and Glied, 2006). Effective treatments—both pharmacological and psychosocial—exist for most mental illnesses, though these treatments do not always work for everyone (Conley and Buchanan, 1997). Even when they do, the appropriate treatment regime can be difficult to determine (Schill and Olsson, 2016; Stroup and Marder, 2021). Nonetheless, recurrent episodes of many mental illnesses can have a cumulative effect and exact a significant economic and social toll. People who 180

Mental Health and Illness in Ageing

experience mental illnesses are more likely to be unemployed and are less productive at work (Frank et al., 2019). Those with mental illnesses who are working are much more likely than those without such illnesses to be employed in occupations with low nonroutine analytical or interpersonal skill demands and high routine manual or routine cognitive skill requirements (Frank et al., 2019). Typically, earnings in these occupations are relatively low. The effects of reduced earnings associated with even a single episode of mental illness in some cases may persist over time through their effects on human capital accumulation (Bartel and Taubman, 1986). These illness patterns can generate poor financial outcomes. A recent study reports the negative impact of mental illnesses on wealth accumulation: in a representative panel of U.S. households, psychological distress among heads of households is associated with an increased risk of entering into and staying in debt; this effect of psychological distress remains strong for several years (Balloch et al., 2020). Mentally ill individuals are also more likely to retire early and/or receive disability income (Conti et al., 2006). A cross-sectional analysis of Australians aged 45–64 shows the important impact of leaving the labor force due to mental illness. That study finds that individuals who retired early from the labor force because of depression, or another mental illness, saw wealth accumulation between 78 percent and 93 percent lower, respectively, compared with others with the same age, gender, and education who remained in the labor force with no chronic health conditions (Shrestha et al., 2011). The symptoms of, and the stigma associated with, mental illness may also make creating and maintaining the types of strong social capital that can provide a financial buffer in times of economic distress difficult. Assistance from friends and family can reduce the likelihood of hardship related to paying for bills, food, and healthcare—and this assistance is particularly valuable protection for lower-income individuals who cannot self-insure against material hardship as individuals with higher household incomes may be able to do (Campbell and Pearlman, 2019).

10.2.2

Responsiveness to Economic, Social, and Personal Circumstances

A growing literature documents the social determinants of many illnesses. Mental illnesses can be the product of individual personal and social circumstances (Compton and Shim, 2015). Researchers have concluded that both genes and environmental factors play roles in the development of mental illnesses, with varying contributions across conditions, and genetic predispositions and environmental circumstances also interact to raise risk (Ellis and Every-Palmer, 2017; Stilo and Murray, 2019). Life events—bereavement, job loss, physical injury—can trigger the development or exacerbation of a mental illness. Older adults are especially prone to suffer bereavements, transitions in their social roles through retirement, and loss of independent living. Losing a spouse, in particular, is an important risk factor for mental illness and is often associated with dramatic declines in life satisfaction (Lucas, 2007). For example, around half of individuals widowed may develop depressive symptoms in the year after a spouse dies and public U.S. survey data finds that newly widowed individuals above age 70 face a risk for syndromal depression that is nearly nine times that of married couples (Komaroff and Harvard Medical School, 1999; Siflinger, 2017; Turvey et al., 1999). In 2014, among adults 75 years or older in the United States who had ever been married, 58 percent of women and 28 percent of men had been widowed at least once (Mayol-Garcia et al., 2021). Chronic medical conditions and chronic pain are associated with stress and physiological changes that can lead to depression (CMHA, 2008). Depression earlier in life has separately been associated with a much higher risk of dementia in later life (Byers and Yaffe, 2011). Individuals with mental illness are more likely to smoke and to have relatively low levels of 181

Sherry Glied et al.

physical activity; these behavior patterns can affect long-term physical health (Ohrnberger et al., 2017). Poverty and mental illness have a bidirectionally causal relationship (Ridley et al., 2020). For example, inadequate housing and sleeping environments due to poverty can lead to mental illness. Conversely, mental illness can worsen economic outcomes and lead to poverty; treatment has been associated with an increase in days worked. In many studies, associations are found between depression and indicators of socioeconomic status, and for older adults specifically (Beekman et al., 1999; Compton and Shim, 2015; Ellis and Every-Palmer, 2017).

10.3

Epidemiology of Mental Illnesses in a Life Course Context

Mental illnesses are diagnosed based on self-reported or practitioner-observed symptoms, not laboratory tests or imaging studies (Manderscheid et al., 2010). This creates the impression of illnesses that are more subjective in their definition, leading to suspicion regarding the degree to which people with ordinary problems of living are experiencing an illness (Webber and Bjelland, 2015). The challenges of diagnosing mental illnesses have important implications for policy. Without clear markers of illness, its severity, and potential for recovery, rationing care based on a particular mental condition or apparent need is more difficult. The managed behavioral healthcare sector in the United States arose specifically to manage the nature of care provided to people with mental illness under health insurance contracts (Frank and Garfield, 2007). Conditioning receipt of other social supports based on a particular mental condition or apparent need is similarly difficult. The assessment of work-limiting disability related to mental illness is costly and errorprone; sometimes people with less serious illnesses qualify and people with more serious illnesses are denied benefits (Webber and Bjelland, 2015; Garc´ıa-G´omez and Gielen, 2018). Psychiatric epidemiologic studies, historically, often exclude elderly subjects, making it difficult to assess the prevalence of illness among older adults (Jeste et al., 1999). These conditions often go undiagnosed in older adults in part because primary care or generalist physicians are the main points of contact with the healthcare system for the older population but often have little specialized training in recognizing or treating mental illness (Bor, 2015). However, evidence suggests that the share of the global population over age 65 with mental illness is increasing rapidly (Jeste et al., 1999; SAMHSA, 2019). Part of this growth is due to general population ageing. In addition, cases of mental illness have increased significantly among younger cohorts, suggesting that higher rates among older adults should be expected as they age, based on increasing baseline rates accompanied by high rates of recurrence. Some evidence also suggests that the oldest old (those 85 or older) have higher rates of some disorders than do the younger old (Koenig et al., 1994). The rapid growth rate of the oldest old population group may further contribute to increased rates of mental illness in the population aged 65 and older. In the past, shorter life expectancy among those with mental illness contributed to lower prevalence of mental disorders in older age groups as compared with younger age groups. Today, people are living longer due to advancements in medicine, improved treatment of diseases, and overall higher standards of living. Individuals with chronic mental illnesses benefit from these developments, and more will reach old age than in the past. Longer overall life expectancy also means that more individuals without previous psychiatric diagnoses may develop late-onset mental disorders. Although direct information on the prevalence of mental illnesses is not available for all countries, cross-national collaboration and recent data collection efforts have allowed for a more global perspective than in the past. The Institute for Health Metrics and Evaluation facilitates data collection and analysis across a large consortium of researchers to populate the Global 182

Mental Health and Illness in Ageing

Burden of Disease study, a cross-national effort estimating health loss from hundreds of diseases, including many mental illnesses (Ritchie and Roser, 2018). Table 10.2 shows prevalence estimates of select mental disorders among all adults and older adults in 27 countries. In those countries, the share of the population with depression ranges between around 2 and almost 6 percent, with rates that remain stable or rise into old age. The prevalence of anxiety disorders overall appears highest among those aged 50–69, declining in later ages. Prevalence of both bipolar disorder and schizophrenia decline with age in almost all countries. This could be a function of differential mortality, as these disorders have higher mortality rates than both anxiety and depression. Finally, rates across countries appear comparable, though countries differ widely in characteristics. Although direct studies are limited, the data in Table 10.2 and other evidence suggest that the prevalence of some mental disorders and symptoms increases with age (McCombe et al., 2018; Volkert et al., 2013). Misattribution and confusion regarding the normal ageing process may lead to a substantial underreporting of psychiatric disorders in the older population. Research conducted 20 years ago suggested that reported prevalence rates at that time might have been 25 percent lower than real prevalence rates (Jeste et al., 1999). While the extent of underreporting is likely to have diminished, it remains a concern. The most common neuropsychiatric disorder in older individuals is dementia, including Alzheimer’s disease. Dementia refers to a range of diseases that cause progressive and severe cognitive decline. Symptoms of dementia often manifest earliest as memory loss, followed by disorientation, mood swings, confusion, greater memory loss, behavioral changes, and difficulties speaking and swallowing (Winblad et al., 2016). The prevalence of dementia roughly doubles every 5 years after age 65 and, worldwide, has been increasing with the ageing population, though decreasing in high-income countries (Prince et al., 2013; Wu et al., 2017; Matthews et al., 2013; Lilford and Hughes, 2020).

10.3.1

Life Satisfaction

The sparse literature on the epidemiology of mental illness across the life course can be juxtaposed with two related literatures. First, research using large surveys to measure subjective happiness, well-being, and life satisfaction has been considerable. While the mapping between mental illness and poor well-being is not precise, the concepts are clearly related (Charara et al., 2016). They also share some measurement approaches such as the General Health Questionnaire. Most of the research using cross-sectional data finds a U-shaped pattern of self-reported happiness or life satisfaction relative to age (Blanchflower and Oswald, 2007; Blanchflower, 2020; Deaton, 2008). On average, life satisfaction is higher at younger ages after the teen years and continuously decreases to reach a low point around the ages of 40–50, after which life satisfaction continuously increases into later life. This pattern has been shown to be consistent across many datasets cross-nationally, though the low point of life satisfaction varies by country (largely based on national income). However, analyses using cross-sectional data may be biased by nonrandom attrition, especially because mental illness (and perhaps low life satisfaction) is associated with early mortality. Analyses using longitudinal data from the Health and Retirement Study in the United States report evidence of stable life satisfaction between ages 65 and 75, followed by a rate of decline that accelerates at older ages (Hudomiet et al., 2020).

10.3.2

Suicide

A persistent, cross-national increase in suicide exists among older adults. Older men and women show the highest rates of suicide across age groups in almost all countries (Shah et al., 2007). 183

Sherry Glied et al.

Table 10.2 Prevalence of select mental disorders by country and age groups (including both males and females), 2017 (%)

Notes: Data for prevalence rates come from Our World in Data, based on the Institute of Health Metrics and Evaluation Global Burden of Disease estimates, which are generated from a combination of sources including medical and national records, epidemiological data, and survey data (Ritchie and Roser, 2018). Greater detail on specific diagnoses and sample is available there.

184

Mental Health and Illness in Ageing

Suicide completion is not homogeneous across older adults; for example, men 75 and older have suicide rates that are 51 percent higher than males aged 65–74 (Hedegaard et al., 2021). Likewise, men aged 65 and older have suicide rates that are significantly higher than those of women in the same age bracket—men over 75 completed suicide at rates that were 35.6 percentage points higher than females over 75. Psychiatric illness is a clear factor in completed suicides, particularly major depression. Physical conditions, dementia, and substance use disorders appear to play a role as well (Conejero et al., 2018). Many studies suggest severe pain is associated with suicide and self-harm among older men (Li and Conwell, 2010). For older adults in the community aged 55 and older, some research finds that perception of burdensomeness accounts for significant variance in suicidal ideation—more so than other commonly found risk factors such as depressive symptoms or hopelessness (Cukrowicz et al., 2011). However, the extent to which suicidal ideation or suicidal attempts are aligned with suicide completion is unclear (Lukaschek et al., 2015). Poverty and isolation are important in understanding contributing factors to suicide completion. A longitudinal study using claims data in South Korea examined the role of poverty in suicide, finding that the risk of suicide was higher for those in poverty—and significantly higher for males in poverty—compared with those with high income in the same age group (Choi et al., 2019). Social isolation, which almost one-quarter of all noninstitutionalized adults over 65 in the United States experience, is not only an important contributor to suicide but also to premature mortality from all causes (NASEM, 2020). Finally, substance use, access to firearms, and a history of involvement with the criminal justice system are all important predictors of suicide (Lankford, 2016; Webb et al., 2011). Substance use among the elderly who complete suicide varies across countries, as rates of problematic drinking differ between Western and Eastern countries (Conwell et al., 2011). While risk factors for suicide may differ in some respects in different countries, and while overall rates among countries are variable, the outcome is the same in most places: suicide among older age groups is higher than among the younger (Shah et al., 2016). These high rates of suicide, and measures of declining life satisfaction, offer additional evidence that rates of mental illness in older adults may be underreported.

10.3.3

Mortality and Morbidity

Mental illness consistently connects to excess mortality. The relative risk of mortality among individuals with mental illness is significantly higher than for comparison populations; a systematic review and meta-analysis of cohort studies from 29 countries in six continents finds a median of 10 years of potential life lost (Walker et al., 2015). Life expectancy for schizophrenia may be 20 years less than that of the general population—and the mortality gap has not improved since the 1970s, suggesting that these individuals have not enjoyed any greater benefits than average of the improved health outcomes experienced among the general population (Laursen et al., 2014; Saha et al., 2007). Increased risk of mortality also exists in major depression and in subclinical forms of depression (Cuijpers and Smit, 2002). Some evidence suggests that depressive symptoms in younger individuals who do not meet criteria for major depression are associated with negative outcomes in older age, such as increased risk for death (Gallo et al., 1997). Excess mortality among individuals with severe and persistent mental illness is mainly due to physical illness (Hert et al., 2011). Older adults with mental illness typically have comorbid physical health problems (Ohrnberger et al., 2017; Vogeli et al., 2007). In part, this is because the challenges associated with mental illness earlier in life, such as impaired decision-making 185

Sherry Glied et al.

processes, hinder efforts to improve physical health or prevent physical conditions. Some individuals with serious mental illness are less likely to seek physical healthcare or mental health treatment voluntarily or face barriers to access (Lehrer and Lorenz, 2014; Bastiampillai et al., 2016). Disparities in physical treatment provision are documented: many patients with serious and persistent mental illness do not receive tests to assess even relatively simple metabolic risk factors such as obesity and high blood pressure (Hert et al., 2011). Being overweight or obese for men, and being physically inactive for women, is strongly associated with mental health in European adults over the age of 50 (Linardakis et al., 2015). The relationship between mental illness and mortality also exists within categories of illness; for example, depression among cancer patients is associated with higher mortality (Satin et al., 2009). A bidirectional relationship exists between heart disease and depression: depression can worsen the prognosis for older patients with heart disease and patients with heart disease have an increased risk of developing depression (NHLBI, 2017). Hopelessness may be an important risk factor for early mortality as it is associated with serious mental illness, increased alcohol consumption, poor physical health such as hypertension or high blood pressure, congestive heart failure, and chronic obstructive pulmonary disorder (Sokol et al., 2021). Quality of life of persons with or recovering from chronic mental illness in the geriatric population has largely been a neglected area of research. A study of 305 older persons (aged 60 and older) with and without chronic mental illness in India found a high level of disability among those with mental illness and poor quality of life compared with those without mental illness. Deficits in the domains of moving around, getting along with people, engaging in life activities, and participating in society contributed most to disability (Ramaprasad et al., 2015). In a study drawing on a cross-sectional sample of individuals from 14 countries, disability levels in all countries increased substantially among persons with definite psychiatric disorders, controlling for physical disease severity, as compared with those without; major depression had the strongest relationship with disability including daily functioning (Ormel et al., 1994). Depression is also associated with increased use of healthcare services, increased morbidity, and increased mortality. For those living in the community over age 70, an International Classification of Diseases, Tenth Revision (ICD-10) diagnosis of depressive disorder or dysthymia (depressive patients) predicted an increased use of psychiatric services—including significantly more psychiatric hospital days, outpatient visits, medication use, and general practitioner services—and more psychiatric diagnoses and increased mortality (Djernes et al., 2011). While morbidity and mortality may be higher among individuals with mental illnesses, the significant role of medical health problems points to modifiable risk factors that could be improved through treatment and through the promotion of behavioral changes to improve physical health. High-quality healthcare services that address both physical and mental health issues among the mentally ill are likely necessary to reduce this excess morbidity. Depression, for example, is a treatable disorder and, as such, could serve as a potential target area for real improvement (Lilford and Hughes, 2020).

10.4

Policy Challenges Posed by Mental Illness in the Context of Ageing

We now turn to some salient economic and policy consequences stemming from mental illness among older adults. Mental illness and mental health do not receive substantial attention in policy discussions about health and ageing (Knight and Sayegh, 2011). Very little research has 186

Mental Health and Illness in Ageing

been conducted on the economics of mental illness and ageing. The characteristics of mental illness, combined with the prevalence of these conditions and the morbidity they generate, intensify several social welfare challenges posed by ageing.

10.4.1

Financial Security, Savings, and Investments

Much of the literature on the economics of ageing focuses on systems to ensure financial security for people who have retired from work. Problems related to a lack of retirement security that may be present in the general population are likely to be even more serious among the population ageing with mental illness because they often have had lower wages, less labor market participation, and earlier than average retirement. People with mental illness are likely to enter retirement with fewer assets than those in better mental health. Psychological distress has been linked to a large and significant reduction in savings for retirement with a large and significant effect (Bogan and Fertig, 2018). Mental health has also been shown to affect individuals’ financial market participation and outcomes through investment behavior (Bogan and Fertig, 2012). Cognitive impacts of depression may make it harder for individuals to make choices that promote financial security; in a study of individuals in Philadelphia who received housing foreclosure counseling, more than a third had major depressive symptoms (Pollack and Lynch, 2009). To the extent that retirement security systems are progressive in income and assets, they should automatically respond to these mental illness-related differences. However, programs that focus only on incomes may fail to account for the effects of a lifetime of episodic disability on asset accumulation. A recent review identifies five studies that examine the effects of increased generosity in retirement benefits on the mental health of older adults and the near-elderly (Simpson et al., 2021). Two U.S. studies and a study in Chile found some improvements in mental illness with increased generosity, while two studies in South Korea did not find changes in life satisfaction. The two U.S. studies point to both improved mental health status and a decrease in suicide among older adults associated with retirement benefit receipt. First, exogenous increases in Social Security (retirement) benefits are associated with lower depressive symptoms, and these findings are statistically significant and economically and clinically meaningful (Golberstein, 2015). These effects are stronger for women than for men. Second, in the United States, reaching the age at which individuals may claim early Social Security retirement benefits (age 62) is associated with a discontinuous drop in suicides, concentrated among men. The magnitude of the drop has increased considerably since the mid-2000s. By comparing receipt of income to the timing of retirement itself, the author concludes that this decline in suicides is much more likely to be reflecting the impact of Social Security receipt than the impact of retirement itself (DeSimone, 2018). In non-U.S. contexts, a longitudinal study on individuals in Chile who were not yet retired suggests that means-tested pension reforms slightly improved measures in subjective well-being among males (Lopez Garcia and Otero, 2017). However, an evaluation of the efficacy of 2014 social pension reform in South Korea did not find strong evidence of improvements to overall quality of life, nor did an evaluation of two other retirement policy interventions implemented around the same time (Lee and Wolf, 2014). Research suggests that expansion of health insurance benefits improves mental health: Medicare Part D prescription drug benefits in the United States, for example, both significantly reduced depressive symptoms among older adults and decreased average out-of-pocket spending on prescription drugs (by approximately US$113 annually) (Ayyagari and Shane, 2015). 187

Sherry Glied et al.

The likelihood of an individual incurring an out-of-pocket expense greater than US$2,000 also declined after the introduction of Medicare Part D. Thus, the new coverage somewhat improved financial security. Housing may also play an important role in promoting psychological well-being and in reducing stress that can lead to poor mental health (Rohe and Lindblad, 2013). Research suggests that homeowners have lower levels of distress than renters (Cairney and Boyle, 2004). Among older homeless adults, obtaining housing reduced depressive symptoms and the incidence of acute care use (Brown et al., 2015). Some research has suggested that the positive psychological impacts of homeownership arise from self-perceptions related to self-esteem, perceived control, and financial security (Rohe and Basolo, 1997; Rohe and Lindblad, 2013; Rohe and Stegman, 1994; Rohe et al., 2002). Consistent with a relationship between housing assets and psychological health, research following the 2008–2009 financial crisis and housing market crash found negative psychological impacts of homeownership, finding that mortgage delinquency was associated with reports of many negative emotions including anxiety, stress, fear, hopelessness, and depression (Fields et al., 2010). Those with mental illness may face greater housing challenges than any other group, given the substantial portion of mentally ill in homeless populations. Estimates vary, but an international review suggests that between 60 and 93 percent of homeless persons have a mental illness (Padgett et al., 2011; Schreiter et al., 2017). U.S. studies found that a substantial portion of the homeless populations in New York City, Boston, and Los Angeles are older than age 50 (Culhane et al., 2019). As the general population ages, this proportion will rise. The nature of some serious and persistent mental illnesses may make housing stability more difficult—even among those participating in a program that provided a housing unit to homeless mentally ill individuals, fewer than half spent every night in their unit and, notably, the level of individuals’ impairment related to psychiatric symptoms and substance abuse after a year of the program did not improve significantly (Pearson et al., 2009).

10.4.2

Social Capital, Social Networks, and Caregiver Burden

Informal care through family members and friends is a significant part of the care provision of older individuals with mental illnesses (Diminic et al., 2021). Ageing often makes living independently and managing activities of daily living, such as bathing, dressing, or eating, more challenging. In many cases, the symptoms of mental illnesses exacerbate these challenges, further leaving individuals unable to care for themselves without significant assistance. Functional impairment and reduced energy and motivation, for example, can be disabling symptoms of late-life depression (Verma and Silverman, 2006). To offset the costs of formal healthcare services, in many cases of severe illness, family members provide care to relatives with mental illness at younger ages. Caregiving can lead to burnout and produce caregiver burdens including financial strain, depression, physical morbidity, and even excess mortality, among other negative consequences (Bolin et al., 2008). When family members can no longer provide such care over time, social networks may shrink even further (Perese and Wolf, 2005). Family members who have supported people with mental illness throughout their lives may be at greater risk of burnout and inability to provide such care as people age (Hubbell and Hubbell, 2002). Some research indicates that caregivers of mentally ill older adults, who had been providing care for an extended time and had low social support, were at high risk for psychological distress or depression (Baillie et al., 1988). The effects of mental illness on narrowing social networks are likely to exacerbate these caregiver effects. This means that people with mental illness are less likely to be able to continue living independently 188

Mental Health and Illness in Ageing

without assistance as they get older, increasing the need of this group for structured home care or long-term care services. A majority of caregivers cross-nationally are older than age 50 themselves (Shahly et al., 2013). Studies show that the burden of caregiving and anticipation of widowhood for older caregivers providing for a spouse are associated with poor mental health (Christakis and Allison, 2006). The impact on caregiver health of having a spouse hospitalized with a serious disease, such as dementia or psychiatric condition, is almost as negative as it is for that spouse to die (Verma and Silverman, 2006). Mental illnesses pose a significant risk to the psychological wellbeing of family members, in that the presence of a family member with a mental illness increases the likelihood of experiencing psychological distress by a factor between one-third and one-half (Holmes and Deb, 2003). Given the high indirect costs of mental illnesses in older age, including from caregiver burden, treating these disorders has high indirect benefits (Martire et al., 2010). Treating late-life depression, for example, may help mitigate negative mental health outcomes both for mentally ill individuals and for their caregivers. Perceived social support for caregivers is also a major factor correlated with burnout; providing adequate support to family members is an appropriate policy goal (Zivin et al., 2013).

10.4.3

Implications for Health Systems

Predictable heterogeneity in the cost of care generates challenges for health insurance markets and for capitated payment systems (including case rates or bundled payments); because many mental illnesses are costly and have high rates of recurrence, insurers and capitated providers have an incentive to structure their offerings and services to avoid attracting people with these conditions (McGuire et al., 2014; Keller et al., 2020). Similar challenges arise when payments (or other rewards) are adjusted based on patient assessments of the quality of care because the presence of mental illness can affect both the quality of care patients receive and their perceptions of the quality of care (Hermann et al., 1998; Travers et al., 2021). Unregulated insurers can avoid customers with mental illness directly by either underwriting mental illness or excluding these conditions from coverage. Regulated insurers can avoid customers with mental illness indirectly by discouraging enrollment, taking advantage of the distinct services that mental illness often requires for treatment (specialized providers, pharmacotherapies, and utilization management techniques), and designing insurance products and provider networks accordingly (Keller et al., 2020; Donohue et al., 2009). Roughly 50 percent of the increased cost associated with individuals with a mental illness is attributable to medical care costs that are not directly related to a mental illness. A high rate of comorbidity exists between mental illnesses and other physical illnesses, as noted. The prevalence of physical health conditions rises with age, making the incentive effects of comorbidities an increasing problem over the life course. Integrated treatment can help reduce the costs of such comorbidities but implementing such treatments has faced challenges (Kearns et al., 2017). As the cost of care for comorbid medical conditions rises with age, the incentive to avoid people with comorbid mental health conditions rises with it. Many competitive insurance systems around the world and capitated payment systems use some form of risk adjustment to diminish selection incentives (Ettner et al., 1998). However, designing risk adjustment systems that appropriately address comorbidities can be complicated (Shrestha et al., 2011). The difficulty in measuring the presence and severity of mental illnesses compounds this problem. In programs where risk adjustment increases payments for patients with recorded mental health conditions, providers may have an incentive to overdiagnose such 189

Sherry Glied et al.

conditions. This appears to have been the case in the U.S. Medicare Advantage program; beneficiaries enrolled in risk-adjusted Medicare plans were substantially more likely to have a diagnosis of major depression, bipolar disorder, or paranoid disorder recorded over time than were those who remained enrolled in traditional, non-risk-adjusted plans (Kronick and Welch, 2014). The result is two side-by-side sets of incentives for inefficient conduct. Selection incentives remain because mental health risk adjustors are crude. In addition, most risk adjustment systems create incentives to “upcode” and collect extra payments. The subjective features of mental health diagnosis make auditing and enforcing appropriate coding conventions challenging.

10.5

Concluding Observations

Because mental illnesses frequently have early-life origins, they have profound impacts on an individual’s age-earnings profile, level of attachment to the labor market, and connections to social structures of life such as marriage, friendships, and family that provide support and happiness throughout a lifetime. The implication is that many mental illnesses trigger a dynamic of powerful impediments to successful ageing. Increasing longevity means that the population at risk for dementia is growing rapidly as are medical conditions associated with the old-old population segment. These factors contribute to the growing prevalence of mental illness among older adults. The disruption to human capital accumulation and the weakened attachment to the labor force caused by mental disorders and the consequent flattening of the age-earnings profile for people affected by mental illnesses early in life mean that a substantial segment of the population arrives at old age with little savings and few employment-related benefits. Moreover, a segment of this population will have fewer friends and family attachments, thereby weakening other sources of financial and emotional support. All this means that people arriving at old age with a history of mental illnesses face a greater likelihood of economic, health, and emotional stress than the rest of the population. These additional stressors aggravate existing mental health conditions. Evidence suggests that the social safety net can attenuate the economic and social deficits created by mental illness. Our review shows that income support and stable housing can improve the mental health and general well-being of older adults. Yet social welfare systems have not kept up with the ageing of the mentally ill populations in many countries. In the United States, major housing programs for low-income older adults have not expanded in more than a decade. The combined effects of early onset of mental illnesses, lower work force attachment, longevity, and a tattered safety net is a set of undesirable outcomes like elevated suicide rates among old-old males, higher rates of poverty, and increased homelessness among older adults. Deepening our understanding of the mechanisms involved in these complex dynamics could improve the development of targeted programs aimed at ameliorating these circumstances.

References AYYAGARI, P., AND SHANE, D. M. (2015): “Does prescription drug coverage improve mental health? Evidence from Medicare Part D,” J Health Econ, 2015(41): 46–58. BAILLIE, V., NORBECK, J. S., AND BARNES, L. E. A. (1988): “Stress, social support, and psychological distress of family caregivers of the elderly,” Nurs Res, 37(4): 217–222. BALLOCH, A., ENGELS, C., AND PHILIP, D. (2020): “When it rains it drains: Psychological distress and household net worth.” June 15, 2021. Available at SSRN: https://ssrn.com/abstract=3521323. BALOUCH, S., RIFAAT, E., CHEN, H.L., ET AL. (2019): “Social networks and loneliness in people with Alzheimer’s dementia,” Int J Geriatr Psychiatry, 34(5): 666–673. BANKS, L., KUPER. H, AND POLACK S. (2017): “Poverty and disability in low- and middle-income countries: A systematic review,” PLOS ONE, 12(12): e0189996.

190

Mental Health and Illness in Ageing

BARTEL, A., AND TAUBMAN, P. (1986): “Some economic and demographic consequences of mental illness,” J Labor Econ, 4(2): 243–256. BASTIAMPILLAI, T., SHARFSTEIN, S. S., AND ALLISON, S. (2016): “Increase in US suicide rates and the critical decline in psychiatric beds,” JAMA, 316(24): 2591–2592. BEEKMAN, A. T., COPELAND, J. R., AND PRINCE, M. J. (1999): “Review of community prevalence of depression in later life,” Br J Psychiatry, 174(4): 307–311. BLANCHFLOWER, D. G. (2020): “Unhappiness and age,” J Econ Behav Organ, 2020(176): 461–488. https://doi.org/10.1016/j.jebo.2020.04.022. BLANCHFLOWER, D. G., AND OSWALD, A. (2007): “Is well-being U-shaped over the life cycle?,” NBER Working Paper Series No. 12935, National Bureau of Economic Research, Cambridge, MA. BOCK, C., BUKH, J. D., VINBERG M., ET AL. (2009): “Validity of the diagnosis of a single depressive episode in a case register,” Clin Pract Epidemiol Ment Health, 5(1): 4. BOGAN, V. L., AND FERTIG, A. R. (2012): “Portfolio choice and mental health,” Rev Financ, 17(3): 955–992. BOGAN, V. L., AND FERTIG, A. R. (2018): “Mental health and retirement savings: Confounding issues with compounding interest,” Health Econ, 27(2): 404–425. BOLIN, K., LINDGREN, B., AND LUNDBORG, P. (2008): “Your next of kin or your own career?: Caring and working among the 50+ of Europe,” J Health Econ, 27(3): 718–738. BOR, J. S. (2015): “Among the elderly, many mental illnesses go undiagnosed,” Health Aff (Millwood) 34(5): 727–731. BROWN, R. T., MIAO, Y., MITCHELL, S. L., ET AL. (2015): “Health outcomes of obtaining housing among older homeless adults,” Am J Public Health, 105(7): 1482-1488. BURCUSA, S. L., AND IACONO, W. G. (2007): “Risk for recurrence in depression,” Clin Psychol Rev, 27(8): 959–985. BYERS, A. L., AND YAFFE, K. (2011): “Depression and risk of developing dementia,” Nat Rev Neurol, 7(6): 323–331. CAIRNEY, J., AND BOYLE, M. H. (2004): “Home ownership, mortgages and psychological distress,” Hous Stud, 19(2): 161–174. CAMPBELL, C., AND PEARLMAN, J. (2019): “Access to social network support and material hardship,” Soc Curr, 6(3): 284–304. CHARARA, R., BCHERAOUI C. E., KRAVITZ H., ET AL. (2016): “Mental distress and functional health in the United States,” Prev Med, 89: 292–300. CHOI, J. W., KIM T. H., SHIN J., ET AL. (2019): “Poverty and suicide risk in older adults: A retrospective longitudinal cohort study,” Int J Geriatr Psychiatry, 34(11): 1565–1571. CHRISTAKIS, N. A., AND ALLISON, P. D. (2006): “Mortality after the hospitalization of a spouse,” N Engl J Med, 354(7): 719–730. CMHA (CANADIAN MENTAL HEALTH ASSOCIATION). (2008): “The relationship between mental health, mental illness and chronic physical conditions.” Available at https://ontario.cmha.ca/ documents/the-relationship-between-mental-health-mental-illness-and-chronic-physical-conditions/ [accessed on April 30, 2021]. COLMAN, I., WADSWORTH M. E. J., CROUDACE T. J., ET AL. (2007): “Forty-year psychiatric outcomes following assessment for internalizing disorder in adolescence,” Am J Psychiatry, 164(1): 126–133. COMPTON, M. T., AND SHIM, R. S. (2015): “The social determinants of mental health,” FOCUS, 13(4): 419–425. CONEJERO, I., OLIE E., COURTET P., ET AL. (2018): “Suicide in older adults: Current perspectives,” Clin Interv Ageing, 2018(13): 691–699. https://doi.org/10.2147/CIA.S130670. CONLEY, R. R., AND BUCHANAN, R. W. (1997): “Evaluation of treatment-resistant schizophrenia,” Schizophr Bull, 23(4): 663–674. CONTI, R. M., BERNDT, E. R., AND FRANK, R. G. (2006): “Early retirement and public disability insurance applications: Exploring the impact of depression,” NBER Working Paper Series No. 12237, National Bureau of Economic Research, Cambridge, MA. CONWELL, Y., VAN ORDEN, K., AND CAINE, E. D. (2011): “Suicide in older adults,” Psychiatr Clin North Am, 34(2): 451–468, ix. CUIJPERS, P., AND SMIT, F. (2002): “Excess mortality in depression: a meta-analysis of community studies,” J Affect Disord, 72(3): 227–236. CUKROWICZ, K. C., CHEAVENS, J. S., VAN ORDEN, K. A., ET AL. (2011): “Perceived burdensomeness and suicide ideation in older adults,” Psychol Ageing, 26(2): 331–338.

191

Sherry Glied et al.

CULHANE, D., DORAN, K., SCHRETZMAN, M., ET AL. (2019): “The emerging crisis of aged homelessness: Could cost avoidance in healthcare fundhousing solutions?,” Int J Popul Data Sci, 4(3), 24. DEATON, A. (2008): “Income, health, and well-being around the world: Evidence from the Gallup world poll,” J Econ Perspect, 22(2): 53–72. DESIMONE, J. (2018): “Suicide and the social security early retirement age,” Contemp Econ Policy, 36(3): 435–450. DIMINIC, S., LEE Y. Y., HIELSCHER E., ET AL. (2021): “Quantifying the size of the informal care sector for Australian adults with mental illness: Caring hours and replacement cost,” Soc Psychiatry Psychiatr Epidemiol, 56(3): 387–400. DJERNES, J. K., GULMANN, N. C., FOLDAGER, L., ET AL. (2011): “13 year follow up of morbidity, mortality and use of health services among elderly depressed patients and general elderly populations,” Aust N Z J Psychiatry, 45(8): 654–662. DONOHUE, J. M., HUSKAMP, H. A., AND ZUVEKAS, S. H. (2009): “Dual eligibles with mental disorders and Medicare part D: How are they faring?,” Health Aff, (Millwood) 28(3): 746–759. ELLIS, P., AND EVERY-PALMER, S. (2017): “What causes mental illness?,” In Bloch, S., Green, A., Janca, A., Mitchell, P. B., and Robertson, M., (eds.), Foundations of Clinical Psychiatry Fourth Edition, Melbourne: Melbourne University Publishing, Chapter 4, pp. 87-126. ETTNER, S. L., FRANK, R. G., MCGUIRE, T. G., ET AL. (1998): “Risk adjustment of mental health and substance abuse payments,” Inquiry, 35(2): 223–239. FAZEL, S., KHOSLA V., DOLL H., ET AL. (2008): “The prevalence of mental disorders among the homeless in western countries: Systematic review and meta-regression analysis,” PLOS Med, 5(12): e225. FIELDS, D., LIBMAN, K., AND SAEGERT, S. (2010): “Turning everywhere, getting nowhere: Experiences of seeking help for mortgage delinquency and their implications for foreclosure prevention,” Hous Policy Debate, 20(4): 647–686. FRANK, R. G., AND GARFIELD, R. L. (2007): “Managed behavioral healthcare carve-outs: Past performance and future prospects,” Annu Rev Public Health, 28: 303–320. FRANK, R. G., AND GLIED, S. (2006): Better but Not Well: Mental Health Policy in the United States since 1950, Baltimore, MD: Johns Hopkins University Press. FRANK, R. G., GLIED, S. A., MARPLE, K. B., ET AL. (2019): “Mental illness and the evolving labor market: Impacts on work & disability,” NBER Center Paper No. NB19-05. https://www.nber. org/programs-projects/projects-and-centers/retirement-and-disability-research-center/centerpapers/nb19-05-0, Accessed October 23, 2021. GALLO, J. J., RABINS P. V., LYKETSOS C. G., ET AL. (1997): “Depression without sadness: Functional outcomes of nondysphoric depression in later life,” J Am Geriatr Soc, 45(5): 570–578. ´ GARC´I A-GOMEZ , P., AND GIELEN, A. C. (2018): “Mortality effects of containing moral hazard: Evidence from disability insurance reform,” Health Econ, 27(3): 606–621. GOLBERSTEIN, E. (2015): “The effects of income on mental health: Evidence from the social security notch,” J Ment Health Policy Econ, 18(1): 27–37. HARO, J. M., ALTAMURA C., CORRAL R., ET AL. (2018): “Understanding the course of persistent symptoms in schizophrenia: Longitudinal findings from the pattern study,” Psychiatry Res, 2018(267): 56–62. https://doi.org/10.1016/j.psychres.2018.04.005. HEDEGAARD, H., CURTIN, S. C., AND WARNER, M. (2021): “Suicide mortality in the United States, 1999–2019.” NCHS Data Brief no. 398. Hyattsville, MD: National Center for Health Statistics. http://dx.doi.org/10.15620/cdc:101761. HENDERSON, A., ROBINSON, W., AND FINEGOLD, K. (2012): “The Affordable Care Act and Latinos.” The Assistant Secretary for Planning and Evaluation. Accessed October 23, 2021. Available here: https://aspe.hhs.gov/reports/affordable-care-act-latinos-0. HERMANN, R. C., ETTNER, S. L., AND DORWART, R. A. (1998): “The influence of psychiatric disorders on patients’ ratings of satisfaction with healthcare,” Medical Care, 36(5): 720–727. HERT, M. D. E., CORNELL C. U., BOBES J., ET AL. (2011): “Physical illness in patients with severe mental disorders. I. Prevalence, impact of medications and disparities in healthcare,” World Psychiatry, 10(1): 52–77. HIRSCHTRITT, M. E., AND BINDER, R. L. (2017): “Interrupting the mental illness–incarceration– recidivism cycle,” JAMA, 317(7): 695–696. HOLMES, A. M., AND DEB, P. (2003): “The effect of chronic illness on the psychological health of family members,” J Ment Health Policy Econ, 6(1): 13–22.

192

Mental Health and Illness in Ageing

HSU, J. W., AND WILLIS, R. (2013): “Dementia risk and financial decision making by older households: The impact of information,” Finance and Economics Discussion Series, J Hum Cap, 45. doi: 10.2139/ssrn.2339225. HUBBELL, L., AND HUBBELL, K. (2002): “The burnout risk for male caregivers in providing care to spouses afflicted with Alzheimer’s disease,” J Health Hum Serv Adm, 25(1): 115–132. HUDOMIET, P., HURD, M. D., AND ROHWEDDER, S. (2020): “The age profile of life-satisfaction after age 65 in the U.S.,” NBER Working Paper No. 28037, National Bureau of Economic Research, Cambridge, MA, USA. JESTE, D. V., ALEXOPOULOUS G. S., BARTELS, S. J., ET AL. (1999): “Consensus statement on the upcoming crisis in geriatric mental health: Research agenda for the next 2 decades,” Arch Gen Psychiatry, 56(9): 848–853. JONES, P. B. (2013): “Adult mental health disorders and their age at onset,” Br J Psychiatry Suppl, 202(s54): s5–s10. KEARNS, B., RAFIA, R., LEAVISS, J., ET AL. (2017): “The cost-effectiveness of changes to the care pathway used to identify depression and provide treatment amongst people with diabetes in England: A model-based economic evaluation,” BMC Health Serv Res, 17(1): 78. KELLER, M. S., XU, H., AZOCAR, F., ET AL. (2020): “The role of behavioral health diagnoses in adverse selection,” Psychiatr Serv, 71(9): 920–927. KESSLER, R. C., ANGERMEYER M., ANTHONY, J. C., ET AL. (2007): “Lifetime prevalence and age-ofonset distributions of mental disorders in the World Health Organization’s World Mental Health Survey Initiative,” World Psychiatry, 6(3): 168–176. KNIGHT, B. G., AND SAYEGH, P. (2011): “Mental health and ageing in the 21st century,” J Ageing Soc Policy, 23(3): 228–243. KOENIG, H. G., GEORGE, L. K., AND SCHNEIDER, R. (1994): “Mental healthcare for older adults in the year 2020: A dangerous and avoided topic,” Gerontologist, 34(5): 674–679. KOMAROFF, A. L., AND HARVARD MEDICAL SCHOOL. (1999): The Harvard Medical School Family Health Guide, New York: Simon & Schuster. KRENDL, A. C., AND PESCOSOLIDO, B. A. (2020): “Countries and cultural differences in the stigma of mental illness: The east–west divide,” J Cross Cult Psychol, 51(2): 149–167. KRONICK, R., AND WELCH, P. W. (2014): “Measuring coding intensity in the Medicare advantage program,” Medicare Medicaid Res Rev, 4(2). http://dx.doi.org/10.5600/mmrr.004.02.a06. LANKFORD, A. (2016): “Public mass shooters and firearms: A cross-national study of 171 countries,” Violence Vict, 31(2): 187–199. LAURSEN, T. M., NORDENTOFT, M., AND MORTENSEN, P. B. (2014): “Excess early mortality in schizophrenia,” Annu Rev Clin Psychol, 2014(10): 425–448. http://dx.doi.org/10.1146/annurevclinpsy-032813-153657. LEE, H. S., AND WOLF, D. A. (2014): “An evaluation of recent old-age policy innovations in South Korea,” Res Ageing, 36(6): 707–730. LEHRER, D. S., AND LORENZ, J. (2014): “Anosognosia in schizophrenia: Hidden in plain sight,” Innov Clin Neurosci, 11(5–6): 10–7. LI, L. W., AND CONWELL, Y. (2010): “Pain and self-injury ideation in elderly men and women receiving home care,” J Am Geriatr Soc, 58(11): 2160–2165. LILFORD, P., AND HUGHES, J. C. (2020): “Epidemiology and mental illness in old age,” BJPsych Adv, 26(2): 92–103. LINARDAKIS, M., PAPDAKI, A., SMPOKOS EM., ET AL. (2015): “Association of behavioral risk factors for chronic diseases with physical and mental health in European adults aged 50 years or older, 2004–2005,” Prev Chronic Dis, 2015(12): E149. http://dx.doi.org/10.5888/pcd12.150134. LOPEZ GARCIA, I., AND OTERO, A. (2017): “The effects of means-tested non-contributory pensions on poverty and well-being: Evidence from the Chilean pension reforms.” Michigan Retirement Research Center at University of Michigan. Research Paper No. 2017-358. Ann Arbor, MI. Available at SSRN: https://ssrn.com/abstract=3195288 or http://dx.doi.org/10.2139/ssrn.3195288. LUCAS, R. E. (2007): “Adaptation and the set-point model of subjective well-being: Does happiness change after major life events?,” Curr Dir Psychol Sci, 16(2): 75–79. LUKASCHEK, K., ENGELHARDT, H., BAUMERT, J., ET AL. (2015): “No correlation between rates of suicidal ideation and completed suicides in Europe: Analysis of 49,008 participants (55+ years) based on the Survey of Health, Ageing and Retirement in Europe (SHARE),” Eur Psychiatry, 30(7): 874–879.

193

Sherry Glied et al.

MACDONALD, E., HAYES, R. L., AND BAGLIONI, A. J. (2000): “The quantity and quality of the social networks of young people with early psychosis compared with closely matched controls,” Schizophrenia Res, 46(1): 25–30. MANDERSCHEID, R. W., RYFF, C. D., FREEMAN, E. J., ET AL. (2010): “Evolving definitions of mental illness and wellness,” Prev Chronic Dis, 7(1): A19. MARTIRE, L. M., SCHULZ, R., REYNOLDS, C. F., ET AL. (2010): “Treatment of late-life depression alleviates caregiver burden,” J Am Geriatr Soc, 58(1): 23–29. MATTHEWS, F. E., ARTHUR A., BARNES, L. E., ET AL. (2013): “A two-decade comparison of prevalence of dementia in individuals aged 65 years and older from three geographical areas of England: Results of the Cognitive Function and Ageing Study I and II,” The Lancet, 382(9902): 1405–1412. MAYOL-GARCIA, Y., GURRENTZ, B., AND KREIDER, R. M. (2021): Number, timing and duration of marriages and divorces: 2016. In: Current Population Reports, Report Number P70-167, United States Census Bureau. Accessed October 24, 2021. Available from: https://www.census.gov/content/dam/Census/library/publications/2021/demo/p70-167.pdf. MCCOMBE, G., FOGARTY, F., SWAN, D., ET AL. (2018): “Identified mental disorders in older adults in primary care: A cross-sectional database study,” Eur J Gen Pract, 24(1): 84–91. MCGUIRE, T. G., NEWHOUSE, J. P., NORMAND, S., ET AL. (2014): “Assessing incentives for servicelevel selection in private health insurance exchanges,” J Health Econ, 35(2014): 47–63. ¨ MILLAN, M. J., AGID Y., BR UNE M., ET AL. (2012): “Cognitive dysfunction in psychiatric disorders: Characteristics, causes and the quest for improved therapy,” Nat Rev Drug Discov, 11(2): 141–168. MILLAN, M. J., GOODWIN, G. M., MEYER-LINDENBERG, A., ET AL. (2015): “Learning from the past and looking to the future: Emerging perspectives for improving the treatment of psychiatric disorders,” Eur Neuropsychopharmacol, 25(5): 599–656. NAMI (NATIONAL ALLIANCE ON MENTAL ILLNESS). (2021): “Mental health conditions.” Available here: https://nami.org/About-Mental-Illness/Mental-Health-Conditions [Accessed October 24, 2021]. NASEM (NATIONAL ACADEMIES OF SCIENCES, ENGINEERING, AND MEDICINE). (2020): Social Isolation and Loneliness in Older Adults: Opportunities for the Healthcare System. Washington, DC: The National Academies Press, p. 316. NHLBI (NATIONAL HEART, LUNG, AND BLOOD INSTITUTE). (2017): “Heart disease and depression: A two-way relationship.” Available at https://www.nhlbi.nih.gov/news/2017/heart-disease-anddepression-two-way-relationship. OHRNBERGER, J., FICHERA, E., AND SUTTON, M. (2017): “The relationship between physical and mental health: A mediation analysis,” Soc Sci Med, 195(2017): 42–49. ORMEL, J., VONKORFF, M., USTUN, T. B., ET AL. (1994): “Common mental disorders and disability across cultures. Results from the WHO Collaborative Study on Psychological Problems in General Healthcare,” JAMA, 272(22): 1741–1748. PADGETT, D. K., STANHOPE, V., HENWOOD, B. F., ET AL. (2011): “Substance use outcomes among homeless clients with serious mental illness: Comparing housing first with treatment first programs,” Community Ment Health J, 47(2): 227–232. PATEL, V., BURNS, J. K., DHINGRA, M., ET AL. (2018): “Income inequality and depression: A systematic review and meta-analysis of the association and a scoping review of mechanisms,” World Psychiatry, 17(1): 76–89. PEARSON, C., MONTGOMERY, A. E., AND LOCKE, G. (2009): “Housing stability among homeless individuals with serious mental illness participating in housing first programs,” J Community Psychol, 37(3): 404–417. PERESE, E. F., AND WOLF, M. (2005): “Combating loneliness among persons with severe mental illness: Social network interventions’ characteristics, effectiveness, and applicability,” Issues Ment Health Nurs, 26(6): 591–609. POLLACK, C. E., AND LYNCH, J. (2009): “Health status of people undergoing foreclosure in the Philadelphia region,” Am J Public Health, 99(10): 1833–1839. POULTON, R., MOFFITT, T. E., AND SILVA, P. A. (2015): “The Dunedin multidisciplinary health and development study: Overview of the first 40 years, with an eye to the future,” Soc Psychiatry Psychiatr Epidemiol, 50(5): 679–693. PRINCE, M., BRYCE, R., ALBANESE, E., ET AL. (2013): “The global prevalence of dementia: A systematic review and metaanalysis,” Alzheimers Dement, 9(1): 63–75 e2. RAMAPRASAD, D., RAO, N. S., AND KALYANASUNDARAM, S. (2015): “Disability and quality of life among elderly persons with mental illness,” Asian J Psychiatry, 18(2015): 31–36.

194

Mental Health and Illness in Ageing

REGIER, D. A., FARMER, M. E., RAE, D. S., ET AL. (1993): “One-month prevalence of mental disorders in the United States and sociodemographic characteristics: The epidemiologic catchment area study,” Acta Psychiatr Scand, 88(1): 35–47. RIDLEY, M., RAO, G., SCHILBACH, F., ET AL. (2020): “Poverty, depression, and anxiety: Causal evidence and mechanisms,” Science, 370(6522): p.eaay0214. RITCHIE, H., AND ROSER, M. (2018): “Mental health.” Published online at OurWorldInData.org. ROBISON, J., SCHENSUL, J. J., COMAN, E., ET AL. (2009): “Mental health in senior housing: Racial/ethnic patterns and correlates of major depressive disorder,” Ageing Ment Health, 13(5): 659–673. ROHE, W. M., AND BASOLO, V. (1997): “Long-term effects of homeownership on the self-perceptions and social interaction of low-income persons,” Environ Behav, 29(6): 793–819. ROHE, W. M., AND LINDBLAD, M. (2013): “Reexamining the social benefits of homeownership after the housing crisis.” In: Belsky, E. S., Herbert, C. E., Molinsky, J. H. (eds.), Homeownership Built to Last: Balancing Access, Affordability, and Risk After the Housing Crisis, New York: Brookings Inst, pp. 99–140. ROHE, W. M., AND STEGMAN, M. A. (1994): “The impact of home ownership on the social and political involvement of low-income people,” Urban Aff Q, 30(1): 152–172. ROHE, W. M., VAN ZANDT, S., AND MCCARTHY, G. (2002): “The social benefits and costs of homeownership.” In: Retsinas, N. P. and Belsky, E. S. (eds.), Low-Income Housing: Examining the Unexamined Goal, Brookings Institution Press, Chapter 13, pp. 281–406. http://www.jstor.org/ stable/10.7864/j.ctt1280j2. SAHA, S., CHANT, D., AND MCGRATH, J. (2007): “A systematic review of mortality in schizophrenia: Is the differential mortality gap worsening over time?,” Arch Gen Psychiatry, 64(10): 1123–1131. SAMHSA (SUBSTANCE ABUSE AND MENTAL HEALTH SERVICES ADMINISTRATION). (2019): “Key substance use and mental health indicators in the United States: Results from the 2018 National Survey on Drug Use and Health.” HHS Publication No. PEP19-5068, NSDUH Series H-54. Rockville, MD: Center for Behavioral Health Statistics and Quality, Substance Abuse and Mental Health Services Administration. Retrieved from https://www.samhsa.gov/data/. SATIN, J. R., LINDEN, W., AND PHILLIPS, M. J. (2009): “Depression as a predictor of disease progression and mortality in cancer patients: A meta-analysis,” Cancer, 115(22): 5349–5361. SCHILL, J., AND OLSSON, H. (2016): “Do doctors agree on doses of antipsychotic medications?,” Ther Adv Psychopharmacol, 6(6): 349–354. SCHOFIELD, D. J., SHRESTHA, R. N., PERCIVAL, R., ET AL. (2011): “Quantifying the effect of early retirement on the wealth of individuals with depression or other mental illness,” Br J Psychiatry, 198(2): 123–128. SCHREITER, S., BERMPHOL, F., KRAUSZ, M., ET AL. (2017): “The prevalence of mental illness in homeless people in Germany: A systematic review and meta-analysis,” Deutsches Aerzteblatt Int, 114(40): 665. SHAH, A., BHAT, R., MCKENZIE, S., ET AL. (2007): “Elderly suicide rates: Cross-national comparisons and association with sex and elderly age-bands,” Med Sci Law, 47(3): 244–252. SHAH, A., BHAT, R., ZARATE-ESCUDERO, S., ET AL. (2016): “Suicide rates in five-year age-bands after the age of 60 years: The international landscape,” Ageing Ment Health, 20(2): 131–138. SHAHLY, V., CHATTERJI, S., GRUBER, M. J., ET AL. (2013): “Cross-national differences in the prevalence and correlates of burden among older family caregivers in the World Health Organization World Mental Health (WMH) Surveys,” Psychol Med, 43(4): 865–879. SIFLINGER, B. (2017): “The effect of widowhood on mental health—An analysis of anticipation patterns surrounding the death of a spouse,” Health Econ, 26(12): 1505–1523. SIMPSON, J., ALBANI, V., BELL, Z., ET AL. (2021): “Effects of social security policy reforms on mental health and inequalities: A systematic review of observational studies in high-income countries,” Soc Sci Med, 272(2021): 113717. SOKOL, Y., RIDLEY, J., GOODMAN, M., ET AL. (2021): “Continuous identity cognitive therapy: Feasibility and acceptability of a novel intervention for suicidal symptoms,” J Cogn Psychother, 35(1): 64-80. STILO, S. A., AND MURRAY, R. M. (2019): “Non-genetic factors in schizophrenia,” Curr Psychiatry Rep, 21(10): 1–10. STROUP, T. S., AND MARDER, S. (2021): “Schizophrenia in adults: Maintenance therapy and side effect management.” In: Stein, M. B. (ed.), UpToDate. Wolters Kluwer. Available from: https://www.uptodate.com/contents/schizophrenia-in-adults-maintenance-therapy-and-side-effectmanagement [Accessed October 24, 2021].

195

Sherry Glied et al.

SUTIN, A. R., STEPHAN Y., LUCHETTI M., ET AL. (2020): “Loneliness and risk of dementia,” J Gerontol B Psychol Sci Soc Sci, 75(7): 1414–1422. TAKAHASHI, M., TANAKA, K., AND MIYAOKA, H. (2005): “Depression and associated factors of informal caregivers versus professional caregivers of demented patients,” Psychiatry Clin Neurosci, 59(4): 473–480. ¨ TJ ORNSTRAND , C., EKLUND, M., BEJERHOLM, U., ET AL. (2020): “A day in the life of people with severe mental illness living in supported housing,” BMC Psychiatry, 20(1): 508. TRAVERS, J. L., LE, C., DESAI, M. M., ET AL. (2021): “Factors associated with dissatisfaction in medical care quality among older medicare beneficiaries suffering from mental illness,” J Ageing Soc Policy, 33(1): 15–66. TURVEY, C. L., CARNEY, C., ARNDT, S., ET AL. (1999): “Conjugal loss and syndromal depression in a sample of elders aged 70 years or older,” Am J Psychiatry, 156(10): 1596–1601. VERMA, K., AND SILVERMAN, B. C. (2008): “Economic burden of late-life depression: Cost of illness, cost of treatment, and cost control policies.” In: Ellison, J. E., Etter, P. C., Kyomen, H. H., and Verma, S. (eds.), Mood Disorders in Later Life, 2nd Edition, Boca Raton: CRS Press, pp. 110–122. https://doi.org/10.3109/9781420053302. VOGELI, C., SHIELDS, A. E., LEE, T. A., ET AL. (2007): “Multiple chronic conditions: Prevalence, health consequences, and implications for quality, care management, and costs,” J Gen Intern Med, 22(Suppl 3): 391–395. VOLKERT, J., SCHULZ, H., HARTER, M., ET AL. (2013): “The prevalence of mental disorders in older people in Western countries—A meta-analysis,” Ageing Res Rev, 12(1): 339–353. WADSWORTH, M. E. J. (1991): The Imprint of Time: Childhood, History, and Adult Life, Oxford: Clarendon Press, Oxford University Press, xiv. WALKER, E. R., MCGEE, R. E., AND DRUSS, B. G. (2015): “Mortality in mental disorders and global disease burden implications: A systematic review and meta-analysis,” JAMA Psychiatry, 72(4): 334–341. WEBB, R. T., QIN, P., STEVENS, H., ET AL. (2011): “National study of suicide in all people with a criminal justice history,” Arch Gen Psychiatry, 68(6): 591–599. WEBBER, D. A., AND BJELLAND, M. J. (2015): “The impact of work-limiting disability on labor force participation,” Health Econ, 24(3): 333–352. WHITEFORD, H., FERRARI, A., AND DEGENHARDT, L. (2016): “Global burden of disease studies: Implications for mental and substance use disorders,” Health Aff, 35(6): 1114–1120. WINBLAD, B., AMOUYEL, P., ANDRIEU, S., ET AL. (2016): “Defeating Alzheimer’s disease and other dementias: A priority for European science and society,” Lancet Neurol, 15(5): 455–532. WU, Y. T., BEISER, A. S., BRETELER, M. M. B., ET AL. (2017): “The changing prevalence and incidence of dementia over time—Current evidence,” Nat Rev Neurol, 13(6): 327–339. ZIVIN, K., WHARTON, T., AND ROSTANT, O. (2013): “The economic, public health, and caregiver burden of late-life depression,” Psychiatr Clin North Am, 36(4): 631–649.

196

PART II

Pensions and Social Security

11 SOCIAL SECURITY REFORMS IN HETEROGENEOUS AGEING POPULATIONS Miguel S´anchez-Romero and Alexia Prskawetz

Abstract As populations age, government programs that redistribute resources from the working-age population to the dependent elderly are increasingly under fiscal pressure, requiring policy adjustments. At the same time, many of these government programs are also challenged because mortality differentials by socioeconomic status increase. If ex ante mortality differences are ignored in the contribution and benefit structure of government programs, a redistribution from shortto long-lived individuals will be the result and actuarial fairness may be violated. In this chapter, we first review evidence of the increasing mortality differential by socioeconomic status (SES). Out of all elderly government programs, we focus on public pensions. We proceed by discussing the role of differential mortality on the internal rate of return (IRR) of the pension system and discuss how contributions and benefits would need to adjust to restore equality of the IRR among different socioeconomic groups. We continue discussing several potential future research topics related to the policy implementation of such reforms. First, an investigation of which SES variable to use in pension reforms to better account for mortality differentials would be required. Second, empirical studies on the variation of the IRR of pensions across SES groups are important for the new design of the parametric components of pension plans. Because individuals will react to public policy reforms, we also propose studying these policy reforms in behavioral models to account for general equilibrium effects, the transition costs of such reforms, and the possibility of using a multi-pillar approach.

11.1

Introduction

Persistent low fertility and increasing longevity imply that most industrialized countries will face pronounced population ageing in the coming decades. To sustain the current system of reallocation of resources across ages that is based on contributions of a shrinking workforce and benefits that accrue to a growing share of elderly dependent people, reforms of social security systems are inevitable. Such reforms need to consider not only population ageing at the aggregate level but also the fact that individual ageing is heterogeneous across socioeconomic groups. Ignoring these heterogeneities might jeopardize any reform as we ignore inequalities in the length of life and income within generations for the sake of increasing sustainability of public finances. As several recent studies have shown (Ayuso et al., 2017; Chetty et al., 2016; DOI: 10.4324/9781003150398-13

199

Miguel S´anchez-Romero and Alexia Prskawetz

Haan et al., 2020; Lee and S´anchez-Romero, 2020), people with lower socioeconomic status and with lower life expectancy will receive fewer benefits compared with those who survive longer. A redistribution from short-lived to long-lived individuals may result in transforming a progressive pension system into a regressive one. As Alvarez et al. (2021) show, pension reforms such as the increase in normal retirement age or indexing benefits to average life expectancy will disadvantage short-lived individuals of lower socioeconomic status. Reducing regressive distributions of pension systems caused by inequality in the length of life may indeed be a prerequisite to implement necessary pension reforms as Richter and Werding (2020) recently highlight. A striking fact observed during the last decades is that these differences in life expectancy by socioeconomic status (SES), such as occupation, education, income, and even wealth, have grown over time. For instance, Waldron (2007) obtains that in the United States the difference in life expectancy at age 65 between those in the top half of the income distribution and those in the bottom half has increased from less than 1 year to more than 5 years for the cohorts born between 1912 and 1941. Chetty et al. (2016) consider trends in life expectancy by income groups in the United States between 2001 and 2014 and find a gap between the richest and poorest 1 percent of the individuals up to 14.6 years for men and 10.1 years for women, and this inequality increased over time. These differences in life expectancy imply that in the top 1 percent of the income distribution individuals expect Social Security and Medicare for about 11.8 and 8.3 more years compared with individuals in the bottom 1 percent of the income distribution. For Germany, Haan et al. (2020) document earnings-related heterogeneity in life expectancy between the top and bottom decile that increased from 4 years for cohorts born in 1926–1928 to 7 years for cohorts born in 1947–1949. These inequalities imply that the German pension system is becoming regressive and may increase inequality in retirement. For a selection of European Union (EU) countries, Corsini (2010) finds that differences in mortality by educational level are more pronounced among men (often more than 10 years) than among women (with a gap of about 5 years). Using more recent mortality data that include EU and non-EU countries, Murtin et al. (2021) find similar result as Corsini (2010) for 18 Organisation for Economic Co-operation and Development countries. Caselli et al. (2015) suggest several causes for the observed differences in mortality by SES: (1) early life conditions, (2) health-related behavior, (3) access to healthcare, and (4) environmental changes, among others. Early life conditions are important determinants of the life expectancy of individuals, among which endogenous factors such as educational attainment have been found to explain approximately 30 percent of the total difference in mortality by socioeconomic status in the United States (Hummer and Hernandez, 2013), whereas exogenous factors such as genetic background only explain 25–30 percent (Christensen et al., 2006). After controlling for other characteristics, health-related behaviors such as smoking, obesity, and physical activity are estimated to explain between 20 and 25 percent of the mortality gradient (Contoyannis and Jones, 2004; Fenelon and Preston, 2012). Other factors are also important. For instance, because individuals with high income levels have early access to new medical technology, a positive correlation between income and life expectancy is observed (Goldman and Lakdawalla, 2005; Rogers et al., 2013). Similarly, environmental changes also explain differences in mortality by socioeconomic status, because individuals with a higher income level can afford better houses and locations that protect them from environmental stress (Healy, 2003; Wen and Gu, 2012). As a result, finding that mortality differences by SES are even larger than mortality differences by gender and across countries is not surprising (Luy et al., 2011). As many studies (Preston and Elo, 1995; Doblhammer et al., 2005; Shkolnikov et al., 2006; Manchester and Topoleski, 2008; Klotz, 2010; Luy et al., 2011; Olshansky et al., 2012) argue, differences of longevity 200

Social Security Reforms in Heterogeneous Ageing Populations

among educational groups may be explained by education-specific individual behavior. Understanding the underlying mechanisms of the longevity gap is important because the length of life will induce different life-course decisions, including educational investment and labor market behavior at the intensive and extensive margin. As these life-course decisions will be linked to contributions and benefits of social security systems, the heterogeneity of individuals cannot be ignored when considering reforms of the social security system. A comprehensive research agenda on the economic effects of social security reforms in heterogeneous ageing populations entails the simultaneous analysis of these reforms on life-cycle decisions across different socioeconomic groups and on the redistributive nature of these reforms in retirement. Within such a framework, researchers can study which policy reforms are required to guarantee the sustainability of the welfare state and maximize the well-being of the population (including health, education, etc.), without inducing an increase in inequality by income within generations. The impact of the increasing gap in life expectancy on lifetime pension benefits will depend on the specific pension system being implemented. The first classification of pension systems differentiates between pay as you go and a fully funded set up. In a pay-as-you-go system, the currently working population pays for the current inactive elderly population. In contrast, in a fully funded system, retirement benefits are paid in relation to own contributions. A second classification distinguishes between defined contribution (DC) and defined benefit (DB) systems. In DC systems, contributions are set and benefits will be adjusted to guarantee the sustainability of pension systems. For DB systems, the benefit formula is defined a priori and contributions are adjusted to guarantee sustainability of the system. A third classification differentiates between progressive and nonprogressive/flat pension systems. A progressive system redistributes ex ante from high- to low-income earners. In a nonprogressive/flat pension system, contributions paid during working age are related one to one to the benefits received in old age (i.e., a constant replacement rate is assumed). In a recent contribution Lee and S´anchez-Romero (2020) and S´anchez-Romero et al. (2020) apply a behavioral model (that also accounts for the feedback of pension systems on individual life-cycle decisions) for the United States. They show that a DC system that transforms the pension wealth acquired at the end of the working life into an annuity for the pension period applying a cohort-specific life table for each income quintile may be regarded as the benchmark against which other pension systems can be assessed. Not accounting for differences in mortality, a DC system or a DB system will redistribute from short- to long-lived individuals. This redistribution is higher if the DB system applies a flat rather than a progressive replacement rate and if the mortality differences across income quintiles are increasing. Because public pension systems may also distort individual life-cycle decisions (labor, education, and consumption decisions) the resulting inequality may be even higher. A comparison of the implicit tax rate on work across income quantiles indicates that the distortionary effect of pension systems is highest for a progressive DB system, but it is most unequal across income quintiles for the DB system with flat replacement rates and DC systems that ignore the inequality in the length of life. Overall, Lee and S´anchez-Romero (2020) and S´anchez-Romero et al. (2020) clearly indicate that pension systems that do not account for the increasing mortality differential will become highly regressive ex post. To compare different pension reform approaches that account for differential mortality across socioeconomic groups, we propose a general framework that defines the internal rate of return (IRR) of any pension system as a function of pension parameters and the mortality differential. Within this framework, we discuss various reform options that aim to equalize the IRR across socioeconomic groups. We continue our review with a selection of future research topics that include the importance of empirical studies to understand the differences of the IRR across SES, 201

Miguel S´anchez-Romero and Alexia Prskawetz

the role of behavioral models that account for distortions and incentives that pension reforms imply, the costs of transitions to a new pension system, and finally the role of diversification of pension systems (multi-pillar pension systems) to better account for demographic and financial risk together with differential mortality.

11.2

Redistribution in Pension Plans: The Internal Rate of Return

The most convenient measure for analyzing how pension plans redistribute income across heterogeneous pension participants is the IRR. Compared with other measures such as social security wealth (SSW) and lifetime pension benefits, the IRR is independent of the size of social contributions paid (see the proof of the independence of the IRR to different income levels in the Appendix).1 Because high-income earners pay more contributions than low-income earners, their social security wealth and lifetime pension benefits are by default higher. Other alternative measures for analyzing the redistributive properties of pension plans across heterogeneous individuals are the implicit tax/subsidy of the pension plan (see Ayuso et al., 2017) or the ratio between lifetime benefits and lifetime contributions. The IRR of a pension plan, denoted by i, is the discount factor for which the stream of present value pension benefits received from the pension plan equals the stream of present value social contributions paid to the pension plan. Or, equivalently, the IRR is the discount factor for which social security wealth equals zero. Thus, taking all contributions and benefits received by an individual, the IRR is equivalent to finding the value of i that satisfies Z ω Z R SSW (i) := e−ia S(a)b(a)da − e−ia S(a)t(a)da = 0 (1) 0

R

where R is the retirement age, ω is the maximum age, S(a) is the age-specific survival probability, b(a) is the pension benefit received, and t(a) is the social contribution paid. Pension benefits received at each age, b(a), are calculated as a fraction ψ(a) of a stock K˜ that is accumulated until the age of retirement. The stock accumulated until the age of retirement can be real (as in funded pension plans) or virtual (as in pay-as-you-go pension plans). The stock K˜ is the result of capitalizing a stream of flows during working ages that are a function of labor income, ˜ The survival function S˜ is taken according to a capitalization index r, and a survival function, S. from a life table, which generally corresponds to that of the average participant in the pension plan. The term ψ(a) = ϕexp{ρ(a − R)} is a conversion factor at each age a that transforms K˜ into pension benefits at age a. The definition of the fraction ϕ as we suggest also includes the potential disincentives/incentives for early/late retirement, while ρ is the rate of increase (in real terms) of pension benefits. For convenience, we will refer to ϕ as the replacement rate in what follows. Thus, the pension benefits at age a are given by2 ˜ b(a) = ψ(a)K.

(2)

The contributions paid at each age, denoted by t(a), are proportional to the labor income earned, t(a) = τ y(a), where τ is the social contribution rate and y(a) is the labor income. By substituting the benefits received, (2), and the social contributions paid at each age in (1), it can be shown that the IRR is a function of the replacement rate ϕ; the remaining years of life at retirement, denoted by LE; the capitalization index r; the social contribution rate τ ; ˜ and the capital that and the ratio between the capital actually accumulated until retirement (K) could have been accumulated until retirement if the life table that corresponds to the individual were used (K).3 The Appendix illustrates how these five components will determine the IRR. ˜ For notational simplicity, let us denote the last ratio by P = K/K. The capital stock in the 202

Social Security Reforms in Heterogeneous Ageing Populations

˜ does not depend on the survival of the individual (but on the average survival), numerator, K, while the capital stock in the denominator, K, depends on the individual survival. Thus, if individuals face a mortality risk lower (higher) than the average mortality risk, their value of P will be higher (lower) than one, which implies that the pension system has accumulated more (less) capital than what would have corresponded to their mortality risk. Specifying these five components is very convenient because it allows us to mathematically express the IRR as i := f (r, τ, P, ϕ, LE). Taking the total derivative of the IRR gives4 di = fr dr + fτ dτ + fP dP + fϕ dϕ + fLE dLE.

(3)

The first derivative on the right-hand side of (3) is the impact of changes in the capitalization index on the IRR. Because a higher capitalization index increases the capital stock accumulated until retirement, a higher value of r implies a higher pension benefit, ceteris paribus all other variables, i.e., fr > 0. For simplicity, we assume r is constant (i.e., dr = 0) and is equal to the growth rate of the total wage bill (or, in national accounts terminology, the growth rate of the total compensation of employees). The second term is the impact of the social contribution rate on the IRR. Because a higher social contribution rate reduces social security wealth, ceteris paribus all other variables, the social contribution rate has a negative effect on the IRR, i.e., fτ < 0. The third term represents the impact of the ratio of the capital stocks accumulated on the IRR. Because an increase in the value of P implies that the mortality risk applied by the system increases relative to the one that corresponds to the individual, which raises pension benefits, a higher value of P has a positive effect on the IRR, i.e., fP > 0. The fourth and fifth terms represent the impact of the replacement rate and the remaining years of life at retirement on the IRR. These two terms have a positive effect on the IRR, fϕ > 0, fLE > 0, because both factors increase social security wealth. Note that the second and third terms are associated with the contribution period, while the fourth and fifth terms are associated with the retirement period. Also important to note is that the ratio of the capital stocks accumulated until retirement, P, and the remaining years of life, LE, are a function of the individual-specific survival probability.

11.3

Mortality Gradient by SES and the Impact on the IRR

The negative correlation between mortality rates and higher SES implies that pension plans become regressive. In other words, for a constant replacement rate and social contribution rate, pension plans provide a higher IRR to individuals with higher SES. This is because individuals with higher SES live longer and receive benefits for more years than individuals with low SES. To estimate the life expectancy by SES, education (e) and the relative position in the labor income distribution (y) have frequently been used as a measure of SES. Both measures account equally well for differences in life expectancy, and neither one alone fully captures all the covariation of life expectancy by SES (Bosworth et al., 2016). So, many of the proposed pension reforms correcting for the increasing regressivity of the pension system caused by the mortality gradient use one or both measures of SES. To understand how SES affects the IRR, let us first take the total derivative of P and LE, because both depend on the survival function S, with respect to income y and education e: dP = PS (Se de + Sy dy),

(4)

dLE = LES (Se de + Sy dy),

(5)

where PS > 0 and LES > 0 are the partial derivatives of P and LE with respect to the survival function S and Se , Sy > 0 are the partial derivatives of the survival function with respect to 203

Miguel S´anchez-Romero and Alexia Prskawetz

education and relative position in the labor income distribution, respectively. Hence, (4) and (5) imply that an increase in education or in the position along the earnings distribution leads to an increase in survival and hence an increase in P and LE. Assuming the social contribution rate and the replacement rate remain constant (i.e., dτ = dϕ = 0), substituting (4) and (5) in (3) gives di = (fP PS + fLE LES )Se de + (fP PS + fLE LES )Sy dy > 0. (6) The positive sign of (6) implies that for constant parametric components of the pension system (τ and ϕ) by SES, the IRR is higher for individuals with high SES, either by educational attainment or by labor income, than for individuals with low SES. Hence, (6) confirms that whenever a positive correlation exists between lower mortality and higher SES, pension plans give a higher return to individuals with higher SES than to those with low SES, ceteris paribus the social contribution rate and the replacement rate.

11.4

Correction of Pension Plans for the Mortality Gradient by SES

The basic idea of old-age pension plans is to secure individuals against the risk of outliving their resources in old age. To do so, pension plans transfer income from short-lived individuals to long-lived individuals. In addition, pension plans can be used as a fiscal instrument for transferring resources from high to low SES groups. Hence, pension plans can also be classified as progressive or nonprogressive. Progressive pension plans [e.g., the U.S. Old-Age, Survivors, and Disability Insurance (OASDI) program] transfer income from high-income earners to lowincome earners, which implies that the IRR should decrease with higher SES. In nonprogressive pension plans, no transfer takes place across SES groups, and therefore the IRR should be the same for all SES groups. However, (6) shows that the IRR is higher for higher SES groups, which implies that pension plans become regressive due to the mortality gradient by SES. For expositional clarity, we consider only nonprogressive pension plans. To counterbalance the regressivity induced by the mortality gradient, (3) shows that the parametric components of the pension system (τ ,ϕ) can be modified to guarantee that the IRR is the same across SES groups, i.e., di = 0. This reform can be introduced either during the contribution or accumulation period, through τ , or during the retirement period, through ϕ.

11.4.1

Contributions

If we assume that the replacement rate does not change across SES groups, dϕ = 0, and the capitalization index is the same across SES groups, dr = 0, then substituting (4) and (5) in (3) yields that the social contribution rate that guarantees the same IRR across SES (i.e., di = 0) will satisfy fLE LES + fP PS (Se de + Sy dy) > 0. (7) dτ = − fτ According to (7) the social contribution rate should be higher for individuals with higher survival probability S, either because they have a higher educational attainment or because they have a higher position in the earnings distribution. Nonetheless, accounting for education in (7) may discourage high educational attainment and therefore using the labor income distribution as the adjustment factor (i.e., de = 0) is more appealing. Holzmann et al. (2020) propose and analyze five alternative designs of reforms of the social contribution rate by SES for a nonfinancial defined contribution pension plan. These five proposals combine a social contribution rate that is specific to each SES group based on its life table, as in (7), with a flat social contribution in a setting with and without a maximum contribution cap. The maximum contribution cap 204

Social Security Reforms in Heterogeneous Ageing Populations

in public pension plans is generally introduced to increase private savings among workers at the top of the income distribution. In addition, the maximum contribution cap can be used as a redistributive tool when the maximum benefit is not aligned with the maximum contribution. However, the maximum contribution cap will limit the efficacy of this reform because it limits the amount of contributions transferred from high SES to low SES individuals.

11.4.2

Benefits

Another alternative is to assume that the social contribution rate is the same across SES groups (dτ = 0) and introduce the reform during the benefit period, by modifying the pension replacement rate (ϕ) and/or the rate of change in the pension benefit throughout retirement. For instance, if we assume that the capitalization index is the same across SES groups, dr = 0, then substituting (4) and (5) in (3) yields that the pension replacement rate by SES should satisfy dϕ = −

fLE LES + fP PS (Se de + Sy dy) < 0. fϕ

(8)

Equation (8) implies that individuals with higher life expectancy, either because of higher education or because of higher income, should have a lower replacement rate than individuals with lower education or lower income. Like with the contribution rate, the introduction of education as an explanatory factor of the mortality gradient may discourage individuals who are indifferent between two educational groups to attain the highest educational level (S´anchezRomero and Prskawetz, 2020). For this reason, most proposals only consider the level of income or lifetime income. Several reforms have been proposed for modifying the replacement rate formula. For instance, Ayuso et al. (2017) suggest adjusting the replacement rate only by the difference in the life expectancy at retirement (i.e., PS = 0 and de = 0). This reform has the advantage that it is simple to explain and implement, because only a life table for each SES group is necessary. However, the mortality gradient not only affects survival during the retirement period but also during the working period. For this reason, S´anchez-Romero and Prskawetz (2020) suggest a reform in which the replacement rate accounts for the impact of the mortality gradient by SES on both the contribution period and the retirement period, as in (8) but with de = 0. S´anchez-Romero et al. (2021) present a comparison between Ayuso et al. (2017) and S´anchez-Romero and Prskawetz (2020). A second alternative policy is to modify the disincentives/incentives for early/late retirement, which are also captured by the parameter ϕ. If pension plans provide disincentives/incentives for early/late retirement that are actuarially fair toward the average member of a cohort, a delay in the retirement (e.g., early retirement age) will reduce the IRR for low SES groups and increase the IRR for high SES groups. As a result, delaying the early retirement age will exacerbate the regressivity of the pension system and widen the difference in the IRR across SES groups—see Simulation 1 in National Academies of Sciences, Engineering, and Medicine (2015) and Alvarez et al. (2021). That the disincentives for early retirement (i.e., before the normal retirement age) are actuarially fair toward the low SES group, while the incentives for late retirement (i.e., after the normal retirement age) are actuarially fair toward the high SES group, as suggested by Breyer (2013) and Lee and S´anchez-Romero (2020), seems more appropriate.5 However, this policy alone will not reduce the regressivity of the system and will need the simultaneous implementation of (8). A third alternative is to introduce more flexible benefit schemes and allow individuals to choose whether their pension benefits will decrease (i.e., be front-loaded), remain constant, or increase (i.e., be back-loaded) during the retirement period. If we assume all parameters remain 205

Miguel S´anchez-Romero and Alexia Prskawetz

unchanged except for the rate of increase of pension benefits (ρ), using the total derivative of LE with respect to S and ρ [i.e., dLE = LES (Se de + Sy dy) + LEρ dρ] and (4) in (3) yields that to provide the same IRR across SES groups the rate of increase of pension benefits should satisfy dρ = −

fP PS + fLE LES (Se de + Sy dy) < 0. fLE LEρ

(9)

Equation (9) implies that individuals with higher life expectancy should have a lower rate of increase of their pension benefits than those with lower life expectancy. Indeed, according to Richter and Werding (2020), individuals with a below-average life expectancy will find having a higher pension benefit at the beginning of the retirement period than at the end (i.e., frontloaded) more profitable, while individuals with an above-average life expectancy will benefit from having a lower pension benefit at the beginning of the retirement period than at the end (i.e., back-loaded).

11.5

Directions of Future Research

The previous section shows how the main parametric components of pension plans can be modified to guarantee the same IRR across SES groups. However, (7) and (8) only give the basic properties that are necessary for these reforms and are not the final formulas to be implemented. In what follows, we discuss potential lines of research for estimating the components of these two equations.

11.5.1

Estimation of the Mortality Gradient by SES

An incorrect estimation of the mortality gradient by SES can potentially cause an unintended imbalance in the benefits received between those who live longer and those with shorter lives. If the mortality gradient is overestimated, it will imply on average a transfer from those with longer lives to those with shorter lives. In contrast, if the mortality gradient is underestimated, it will imply on average a transfer from those with shorter lives to those with longer lives. Therefore, the first step is to choose the appropriate SES measure to use by accounting for the pros and cons of each SES measure. Standard measures of SES are educational attainment, labor earnings, occupation, income, and wealth. Educational attainment is frequently used as an SES measure for the mortality gradient because mortality later in life does not affect it and it is highly correlated with life expectancy and income. However, this measure suffers from two potential problems. First, the relationship between the mortality gradient and educational attainment can be overestimated due to an adverse selection problem (Dowd and Hamoudi, 2014; Goldring et al., 2016). Indeed, with the expansion of secondary and tertiary education, individuals with low education are becoming more negatively selected. A recent paper by Luy et al. (2019) provides empirical evidence for the role of the compositional effect of the educational structure in explaining increasing life expectancy. For Denmark, Italy, and the United States between 1990 and 2010 between 15 percent (for men in the United States) and 40 percent (for women in Denmark) of the gains in life expectancy are attributable to shifts in the educational structure. To control for the upward bias that the educational expansion creates in the mortality gradient, Hendi et al. (2021) suggest regressing the logarithm of relative mortality on the relative position in the educational distribution, because they find that this relationship is more stable across cohorts. Second, modifying the parametric components of the pension system in relation to educational attainment seems inappropriate, because it might create disincentives for acquiring higher education. Thus, using 206

Social Security Reforms in Heterogeneous Ageing Populations

education to control for the mortality gradient may, ultimately, imply lower educational attainments, which will reduce economic growth and increase the cost of the pension system for all SES groups. Lifetime labor income is also used as an SES measure for the mortality gradient. Important reports that use this variable as a measure of SES are Bosworth and Burke (2014), Bosworth et al. (2015, 2016), National Academies of Sciences, Engineering, and Medicine (2015), and Waldron (2007). Lifetime labor income has the advantage that it better represents the average income history of the pension participant and avoids being negatively influenced by short-term unemployment spells. Nonetheless, because lifetime labor income can also be prone to reverse causation and changing group sizes over time, many of the aforementioned articles stratify the income level mid-career earnings, or lifetime labor incomes, by income quintiles. Another potential SES candidate to control for the mortality gradient in pension plans is occupation. Like education and income, occupation has the advantage that it is observable and reported in many databases. However, individuals can have several occupations over their working lives and therefore it is not a stable SES measure. In addition, using occupation can produce adverse selection effects in which long-lived and well-paid individuals with risky jobs end up being better off than short-lived individuals in safe jobs (Pestieau and Racionero, 2016). SES variables are not limited to education, lifetime labor income, and occupation. For instance, one could think about using as SES variables for the mortality gradient gender; marital status; region; and health indicators, such as smoking and drinking habits, grip strength, and genomes, among many others. However, despite the sizable list of potential SES variables, the choice of SES variable for the mortality gradient in pension plans should be based on three criteria. First, the SES variable(s) should largely capture the variance in life expectancy and its evolution over time. As Baurin (2020) showed recently, a huge variance of longevity exists within socioeconomic status groups and the longevity dispersion is higher for lower socioeconomic status categories. Similarly, Alvarez et al. (2021) argue that lower socioeconomic groups not only exhibit lower life expectancy, but that their higher lifespan inequality will also disadvantage those groups if the retirement age is only adjusted to the increase of the average life expectancy in a population. Second, the SES variable(s) should have a predictable relationship to mortality (Bravo et al., 2021). This criterion is necessary for reducing uncertainty in the estimation of new life tables. When the uncertainty surrounding future benefits is too high, individuals cannot link their contributions to their pension benefits and therefore the contributions paid will be considered effectively as taxes. Third, the SES variable(s) should induce minimum behavioral reactions.

11.5.2

Empirical Evidence on the Variation of the IRR across SES Groups

An important step toward correcting for the mortality gradient by SES is to investigate empirically how the IRR varies across SES groups. Unfortunately, studies about the IRR of the pension system from an intragenerational perspective are scarce. Most quantitative analyses focus on comparing the IRR across generations. The main reason for the lack of intragenerational studies is the necessity of using social security biographies of pension participants together with their pension entitlements. Moreover, for the calculation of the IRR applying the correct SESspecific life table is necessary, which is subject to the problems stated in the previous subsection. Nonetheless, there are important studies in this regard. For instance, in the United States, Aaron (1977), Hurd and Shoven (1985), Duggan et al. (1993), Gustman and Steinmeier (2001), and Liebman (2001) show that the IRR that high-income earners receive is roughly the same as that of low-income earners, even though the U.S. pension system is progressive (i.e., the replacement 207

Miguel S´anchez-Romero and Alexia Prskawetz

rate of the old-age system declines as lifetime labor earnings increases). Indeed, they find that the mortality gradient across race, education, and marital status significantly, or completely, offset the progressivity of the U.S. pension system. In addition, late entry into the pension system, due to higher educational attainment, and becoming a widow(er) reinforces the effect of the mortality gradient. Given that most pension systems do not use the total amount contributed but the best or the last years contributed, paying contributions to the pension system early in life reduces the IRR because these contributions are not generally used in the pension formula. Moreover, the results reflect the fact that in the case of becoming widow(er), the positive effect of education on mortality increases the likelihood of receiving benefits over a longer period and hence of having a higher IRR. Similar results have been obtained in the United Kingdom (Atkinson, 1970) and in Germany (Schr¨oder, 2012; Haan et al., 2020).

11.5.3

Impact of Pension Reforms on Inequality (Behavioral Reactions)

All reforms of pension plans need to evaluate the potential behavioral effects and all potential general equilibrium effects. Pension plans, especially if they are pay as you go, induce behavioral effects because they change individuals’ budget constraints and may also change the relative factor prices (i.e., wage rates and interest rates). The most convenient approach for this analysis is to use dynamic growth models with overlapping generations a` la Auerbach and Kotlikoff (1987). These theoretical models can analyze practical policy issues with very complex institutional settings and can accommodate realistic demography. The standard behavioral reactions induced by pension plans, which are studied with dynamic growth models of overlapping generations, are changes in savings, labor supply, and retirement. For instance, using a dynamic general equilibrium–overlapping generations model Cubeddu (2000) and S´anchez-Romero and Prskawetz (2017) show how pension systems with a flat replacement rate partly explain the variation in lifetime welfare within cohorts. This variation is explained because pension plans with a flat replacement rate (or a strong link between benefits and contributions) give a higher IRR to individuals with above-average life expectancy than to those with below-average life expectancy—see Equation (6). Because a higher IRR of a pension plan raises the return to labor, individuals with above-average life expectancy supply more labor, which increases the return to education and ultimately the accumulation of savings. As a result, pension plans with a flat replacement rate increase inequality. These authors also show that pension plans with a progressive replacement rate can mitigate the inequality in lifetime income caused by nonprogressive pension plans. Nevertheless, given that a progressive replacement rate reduces the link between contributions and benefits for high-income workers, these plans produce labor distortions, which reduce labor supply and economic growth when markets are complete. When markets are incomplete (e.g., liquidity constraints), however, Fehr et al. (2013) show for the German economy that implementing a progressive pension plan with a degree of progressivity in the replacement rate of 30 percent is more efficient than the status quo. Thus, conducting/performing similar analyses as that of Fehr et al. (2013) for other countries with different pension plans would be important. Pension plans also affect individuals’ retirement decisions. This is because the present value of pension benefits received is a direct function of the retirement age. Thus, even when pension plans are on average actuarially fair, people whose life expectancy is above average have an incentive to retire later, while individuals whose life expectancy is below average have an incentive to retire earlier. This behavioral reaction raises inequality across individuals who differ by life expectancy because an additional year of work increases individuals’ wealth. Indeed, actuarially fair pension plans induce individuals with above-average life expectancy to accumulate more 208

Social Security Reforms in Heterogeneous Ageing Populations

wealth and those with below-average life expectancy to reduce their wealth. Therefore, pension reforms need to evaluate these incentives in models in which the population is heterogeneous by life expectancy. One possible solution suggested by Breyer (2013) and Lee and S´anchezRomero (2020) is to introduce different disincentives/incentives before and after the normal retirement age.6 In particular, early retirement disincentives (before the normal retirement age) should be more actuarially fair toward individuals whose life expectancy is below average, while incentives for delaying retirement (after the normal retirement age) should be more actuarially fair toward individuals whose life expectancy is above average. Thus, this policy will make postponing their retirement decision more attractive for individuals with a below-average life expectancy. Another alternative proposed by Richter and Werding (2020) is to allow retirees to choose between front-loaded and back-loaded pension benefits. Such a policy is preferable for short-lived individuals because it allows them to anticipate their retirement benefits by choosing a front-loaded pension benefit. However, it may induce individuals to retire earlier and increase inequality at very old ages between those who choose a front-loaded pension benefit versus those who choose a back-loaded pension benefit. These models also show in a closed economy without altruistic linkages across cohorts and complete markets that the introduction of a pay-as-you-go plan crowds out private capital, causing interest rates to rise and wages to fall. Given that higher interest rates and lower wages are more beneficial for high-income earners with savings than for low-income earners, inequality could increase. However, when individuals have an altruistic behavior toward their children, or private markets are not complete, private capital may not necessarily be reduced.

11.5.4

Transition Costs

Analyzing the transition costs from one pension plan to a new pension plan within and across generations is important. In this regard, the dynamic feature of growth models with overlapping generations is key, because these models can provide information about which generations are better off, and which ones are worse off, after the implementation of a pension reform. This is done using a three-step procedure. First, the model is solved for the initial pension system. This simulation represents the status quo and gives a specific transition path and the welfare of all individuals. Second, the new policy is implemented in the model to obtain the new transition path and the new welfare level of all individuals. Because a policy change will cause welfare losses for some individuals and gains for others, those who become better off need to compensate those who become worse off. This policy is implemented in the third step by introducing the Lump-Sum Redistribution Authority (LSRA), as proposed by Auerbach and Kotlikoff (1987). The LSRA is an additional government agency that uses lump sum taxes and transfers to keep those individuals who are born before the implementation of the policy at their status quo level of welfare. For those individuals born after the introduction of the policy, the LSRA levies a lump sum tax (or transfer) at the beginning of their economic lives that is the same across all future cohorts. When intra-cohort heterogeneity exists, the lump sum tax (or transfer) should also differ across SES groups (Kotlikoff et al., 1999). Thus, the LSRA is used in the literature to assess efficiency and redistributive effects of policy reforms. Fehr (2016) provides an excellent summary of the literature on the efficiency gains from the privatization of the social security pension system under perfect and imperfect markets.

11.5.5

Multi-Pillar Approach

The implementation of a progressive pension system across SES groups can be done not only through the introduction of a progressive replacement rate formula, as in Equation (8), but also 209

Miguel S´anchez-Romero and Alexia Prskawetz

through the introduction of a multi-pillar pension system (also referred to as a pension system with multiple tiers). The first pillar usually takes care of poverty, and it is of the Beveridgean type. This pillar guarantees a minimum benefit regardless of the income level of the pension participant. The second pillar is a mandatory pension plan of a Bismarckian type, in which a tight link exists between contributions and benefits. The third pillar comprises voluntary occupational and personal retirement saving plans. Thus, in a multi-pillar pension system, the total pension benefit received, b, is calculated P as the weighted sum of pension benefits earned ˜ across the different pension pillars, i.e., b = i αi ψi Ki , where i denotes the pillar type and P αi (with i αi = 1) is the weight of pillar i in the total pension benefit earned. Huggett and Ventura (1999) and Fehr and Uhde (2013, 2014) provide examples of the optimal design of pension systems, where in a two-tier pension system (first and second pillars) they derive the weight α and the generosity of each pension system ψi that maximize welfare and efficiency.7

11.5.6

Dealing with the Source of the Inequality Problem

In this chapter, we have focused on the pension reforms necessary for tackling the regressivity problem caused by the increasing mortality gradient by SES. Although the mortality gradient by SES has existed for decades, if not centuries, the importance of these proposals can be minimized if the mortality gradient by SES is reduced. Thus, studying policies that reduce heterogeneity/inequality in those dimensions that are prone to increase inequality would also be necessary. An important step for future research should be the integration of the educational system, the healthcare system, and labor market policies with the pension system into dynamic general equilibrium–overlapping generation models. Fehr et al. (2013), Laun and Wallenius (2016), and Laun et al. (2019) already present such models, which also consider disability, labor income, and mortality risks that differ by SES. More recently, S´anchez-Romero et al. (2021) endogenize the education decision in a dynamic general equilibrium–overlapping generations model in which individuals differ by labor income and mortality risk. However, to our knowledge no model fully integrates all four components (health, labor market, education, and pension systems).

11.6

Conclusion

In this chapter we first analyzed, using a general setting, the redistributive properties of pension systems when a mortality gradient by SES exists. Second, we reviewed the literature on the potential reforms necessary to cope with the mortality gradient by SES. We have not covered the extensive literature on pension systems but instead focused on the specific literature that focuses on the mortality gradient by SES as a core component. Specifically, using the IRR as a measure of redistribution of a pension system we have shown how wealthier individuals who are ex ante long-lived may receive a higher IRR than poor individuals who are ex ante short-lived. This result implies that the pension system redistributes income from short-lived, poor individuals to long-lived, richer individuals. To correct for the regressivity caused by the mortality gradient by SES, we have shown the properties that the parametric components of the pension system (e.g., contributions and the benefit formula) need to satisfy. There are several potential extensions and future lines of research in addition to those we could discuss in this short overview. Our analysis is restricted to old-age pension systems and does not consider widowhood pensions, disability pensions, survivors’ pensions, etc. Moreover, our framework ignores the fact that many behavioral effects of pension systems should be considered in the framework of a household decision model. A natural extension of the models analyzing the efficiency and welfare of the alternative pension reforms would include children 210

Social Security Reforms in Heterogeneous Ageing Populations

and spouses and hence account for the insurance role of families (Fuster et al., 2007; Hong and R´ıos-Rull, 2007; Ortigueira and Siassi, 2013; Fehr et al., 2017). In addition, individuals with worse health status and hence with potentially higher mortality risks can also claim disability benefits, which can increase the IRR received from the social security system. Finally, we have abstracted from how these reforms on accounting for differential mortality should be combined with other important reforms needed to guarantee the long-run sustainability of social security systems (S´anchez-Romero and Prskawetz, 2019).

Acknowledgment This project has received funding from the Austrian National Bank (OeNB) under Grant no. 17647 and Grant no. 18744.

Notes 1 Social security wealth and lifetime pension benefits are less appropriate measures for analyzing heterogeneous individuals because they are affected by the size of the contributions paid, which is positively related to the SES. 2 For expositional simplicity we have not explicitly considered the age of entrance into the labor market, which is influenced by educational attainment, and we have further assumed that the pension plan only insures old-age benefits. Notice that when survival benefits (i.e., childhood and widowhood) are ˜ In the section “Directions considered, the capital stock K˜ should not be capitalized according to S. of Future Research,” we give further details about how the age of entrance into the labor market and survival benefits affect the IRR. 3 For notational convenience, the remaining years of life, LE, is assumed to include the parameter ρ. 4 For notational simplicity we denote the partial derivative of variable Z with respect to X, or ∂Z/∂X, as ZX . 5 The normal retirement age is the first age at which pension claimers are not penalized in their replacement rate for retiring. 6 The normal retirement age is the age at which individuals can receive full retirement benefits. 7 An additional, but necessary, complexity of the multi-pillar pension analysis is studying whether the different pillars should be integrated or not. If pension pillars are not integrated, pension participants receive pension benefits from all pension pillars. If pension pillars are integrated, then receiving pension benefits from one pillar may partially or fully exclude receiving benefits from another pillar. The integration is in general implemented through means-testing.

References AARON, H. J. (1977): “Demographic effects of social security benefits.” In: Feldstein, M. S., and Inman, R. P. (eds.), The Economics of Public Services. New York: Macmillan, pp. 151–173. ALVAREZ, J. A., KALLESTRUP-LAMB, M., AND KJAERGAARD, S. (2021): “Linking retirement age to life expectancy does not lessen the demographic implications of unequal lifespans. Insurance,” Mathematics and Economics, 99: 363–375. ATKINSON, A. B. (1970): “National superannuation: Redistribution and value for money 1,” Bulletin of the Oxford University Institute of Economics & Statistics, 32(3): 171–185. AUERBACH, A. J., AND KOTLIKOFF, L. J. (1987): Dynamic Fiscal Policy, Cambridg: Cambridge University Press. AYUSO, M., BRAVO, J., AND HOLZMANN, R. (2017): “Addressing longevity heterogeneity in pension scheme design,” Journal of Finance and Economics, 6(1): 1–24. BAURIN, A. (2020): “The limited power of socioeconomics status to predict longevity: Implications for pension policy,” LIDAM Discussion Papers IRES 2020/019. Institut de recherche e´conomiques et sociales, UC Louvain. BOSWORTH, B., AND BURKE, K. (2014): “Differential mortality and retirement benefits in the Health and Retirement Study,” Working paper 2014-4. Center for Retirement Research at Boston College, Boston, MA.

211

Miguel S´anchez-Romero and Alexia Prskawetz

BOSWORTH, B., BURTLESS, G., AND ZHANG, K. (2015): “Sources of increasing differential mortality among the aged by socioeconomic status,” Economic Studies at Brookings, Working Paper 2015-10. Center for Retirement Research at Boston College, Boston, MA. BOSWORTH, B., BURTLESS, G., AND ZHANG, K. (2016): “Later retirement, inequality in old age, and the growing gap in longevity between rich and poor,” Economic Studies at Brookings, 87. Brooking Institution, Washington, DC. BRAVO, J. M., AYUSO, M., HOLZMANN, R., AND PALMER, E. (2021): “Addressing the life expectancy gap in pension policy,” Insurance: Mathematics and Economics, 99: 200–221. BREYER, F (2013): “Gesetzliche Rente: Bedeutet Teilhabe¨aquivalenz Verteilungsneutralit¨at?—Ein weiterer Irrtum,” Wirtschaftsdienst, 93: 117–119. https://doi.org/10.1007/s10273-013-1494-4. CASELLI, G., DREFAHL, S., WEGNER-SIEGMUNDT, C., AND LUY, M. (2015): “Future mortality in low mortality countries.” In: Lutz, W., Butz, W. P., and Samir, K. C. (eds.), World Population and Human Capital in the Twenty-First Century, Oxford: Oxford University Press, pp. 226–272. CHETTY, R., STEPNER, M., ABRAHAM, S., LIN, S., SCUDERI, B., TURNER, N., BERGERON, A., AND C UTLER , D. (2016): “The association between income and life expectancy in the United States, 2001–2014,” JAMA, 315(16): 1750–1766. CHRISTENSEN, K., JOHNSON, T. E., AND VAUPEL, J. W. (2006): “The quest for genetic determinants of human longevity: challenges and insights,” Nature Reviews Genetics, 7(6): 436–448. CONTOYANNIS, P., AND JONES, A. M. (2004): “Socio-economic status, health and lifestyle,” Journal of Health Economics, 23(5): 965–995. CORSINI, V. (2010): “Highly educated men and women likely to live longer. Life expectancy by educational attainment,” Statistics in Focus 24/2010: Population and Social Conditions, Eurostat. CUBEDDU, L. (2000): “Intragenerational redistribution in unfunded pension systems,” MF Staff Papers, 47(1): 90–115. DOBLHAMMER, G., RAU, R., AND KYTIR, J. (2005): “Trends in educational and occupational differentials in all-cause mortality in Austria between 1981/82 and 1991/92,” Wiener Klinische Wochenschrift, 117(13–14): 468–479. DOWD, J. B., AND HAMOUDI, A. (2014): “Is life expectancy really falling for groups of low socioeconomic status? Lagged selection bias and artefactual trends in mortality,” International Journal of Epidemiology, 43(4): 983–988. DUGGAN, J. E., GILLINGHAM, R., AND GREENLEES, J. S. (1993): “Returns paid to early social security cohorts,” Contemporary Policy Issues, 11(4): 1–13. FEHR, H. (2016): “CGE modeling social security reforms,” Journal of Policy Modeling, 38(3): 475–494. FEHR, H., KALLWEIT, M., AND KINDERMANN, F. (2013): “Should pensions be progressive?,’, European Economic Review, 63: 94–116. FEHR, H., KALLWEIT, M., AND KINDERMANN, F. (2017): “Families and social security,” European Economic Review, 91: 30–56. FEHR, H., AND UHDE, J. (2013): “On the optimal design of pension systems,” Empirica, 40(3): 457–482. FEHR, H., AND UHDE, J. (2014): “Means-testing and economic efficiency in pension design,” Economic Modelling, 44(S1): S57–S67. FENELON, A., AND PRESTON, S. H. (2012): “Estimating smoking-attributable mortality in the United States,” Demography, 49(3): 797–818. ˘ , A., AND ˙IMROHORO GLU ˘ , S. (2007): “Elimination of social security in a FUSTER, L., ˙IMROHORO GLU dynastic framework,” The Review of Economic Studies, 74(1): 113–145. GOLDMAN, D. P., AND LAKDAWALLA, D. N. (2005): “A theory of health disparities and medical technology,” Contributions in Economic Analysis and Policy, 4(1): article 8. GOLDRING, T., LANGE, F., AND RICHARDS-SHUBIK, S. (2016): “Testing for changes in the SESmortality gradient when the distribution of education changes too,” Journal of Health Economics, 46: 120–130. GUSTMAN, A. L., AND STEINMEIER, T. L. (2001): “How effective is redistribution under the social security benefit formula?,” Journal of Public Economics, 82(1): 1–28. ¨ HAAN, P., KEMPTNER, D., AND LUTHEN , H. (2020): “The rising longevity gap by lifetime earnings— Distributional implications for the pension system,” The Journal of the Economics of Ageing, 17: 100199. HEALY, J. D. (2003): “Excess winter mortality in Europe: A cross country analysis identifying key risk factors,” Journal of Epidemiology and Community Health, 57(10): 784–789. HENDI, A. S., ELO, I. T., AND MARTIKAINEN, P. (2021): “The implications of changing education distributions for life expectancy gradients,” Social Science & Medicine, 272: 113712.

212

Social Security Reforms in Heterogeneous Ageing Populations HOLZMANN, R., ALONSO-GARC´I A, J., LABIT-HARDY, H., AND VILLEGAS, A. M. (2020): “NDC schemes and heterogeneity in longevity: Proposals for redesign,” In: Holzmann, R., Palmer, E., Robert Palacios, R., and Sacchi, S. (eds.), Progress and Challenges of Nonfinancial Defined Pension Schemes, 1, Washington, DC: World Bank, pp. 307–332. HONG, J. H., AND R´I OS-RULL, J. V. (2007): “Social security, life insurance and annuities for families,” Journal of Monetary Economics, 54(1): 118–140. HUGGETT, M., AND VENTURA, G. (1999): “On the distributional effects of social security reform,” Review of Economic Dynamics, 2(3): 498–531. HUMMER, R. A., AND HERNANDEZ, E. M. (2013): “The effect of educational attainment on adult mortality in the United States,” Population Bulletin, 68(1): 1–16. HURD, M. D., AND SHOVEN, J. B. (1985): “The distributional impact of social security.” In: Wise, D. (ed.), Pensions, Labor, and Individual Choice, Chicago: University of Chicago Press for NBER, pp. 193–222. KLOTZ, J. (2010): “Convergence or divergence of educational disparities in mortality and morbidity? The evolution of life expectancy and health expectancy by educational attainment in Austria in 1981–2006,” Vienna Yearbook of Population Research, 8: 139–174. KOTLIKOFF, L. J., SMETTERS, K., AND WALLISER, J (1999): “Privatizing social security in the United States—Comparing the options,” Review of Economic Dynamics, 2(3): 532–574. LAUN, T., MARKUSSEN, S., VIGTEL, T. C., AND WALLENIUS, J. (2019): “Health, longevity and retirement reform,” Journal of Economic Dynamics and Control, 103: 123–157. LAUN, T., AND WALLENIUS, J. (2016): “Social insurance and retirement: A cross-country perspective,” Review of Economic Dynamics, 22: 72–92. ´ LEE, R. D., AND SANCHEZ -ROMERO, M. (2020): “Overview of the relationship of heterogeneity in life expectancy to pension outcomes and lifetime income.” In: Holzmann, R., Palmer, E., Robert Palacios, R., and Sacchi, S. (eds.), Progress and Challenges of Nonfinancial Defined Pension Schemes, 1, Washington, DC: World Bank, pp. 261–279. LIEBMAN, J. B. (2001): “Redistribution in the current U.S. Social Security System,” NBER Working Paper Series 8625. National Bureau of Economic Research, Cambridge, MA. LUY, M., DI GIULIO, P., AND CASELLI, G. (2011): “Differences in life expectancy by education and occupation in Italy, 1980–94: Indirect estimates from maternal and paternal orphanhood,” Population Studies, 65(2): 137–155. LUY, M., ZANNELLA, M., WEGNER-SIEGMUNDT, C., MINAGAWA, Y., LUTZ, W., AND CASELLI, G. (2019): “The impact of increasing education levels on rising life expectancy: A decomposition analysis for Italy, Denmark, and the USA,” Genus, 75(1): 11. MANCHESTER, J., AND TOPOLESKI, J. (2008): “Growing disparities in life expectancy,” Economic and budget issue brief, Congressional Budget Office, Washington, DC. MURTIN, F., MACKENBACH, J. P., JASILIONIS, D., AND D’ERCOLE, M. M. (2021): “Educational inequalities in longevity in 18 OECD countries,” Journal of Demographic Economics, 1–29. https://doi.org/10.1017/dem.2021.22. NATIONAL ACADEMIES OF SCIENCES, ENGINEERING, AND MEDICINE. (2015): The Growing Gap in Life Expectancy by Income: Implications for Federal Programs and Policy Responses, Washington, DC: The National Academies Press. OLSHANSKY, S. J., ANTONUCCI, T., BERKMAN, L., BINSTOCK, R. H., BOERSCH-SUPAN, A., CACIOPPO, J. T., CARNES, B. A., CARSTENSEN, L. L., FRIED, L. P., GOLDMAN, D. P., JACKSON, J., KOHLI , M., ROTHER , J., Z HENG, Y., AND ROWE , J. (2012): “Differences in life expectancy due to race and educational differences are widening, and many may not catch up,” Health Affairs, 31(8): 1803–1813. ORTIGUEIRA, S., AND SIASSI, N. (2013): “How important is intra-household risk sharing for savings and labor supply?,” Journal of Monetary Economics, 60(6): 650–666. PESTIEAU, P., AND RACIONERO, M. (2016): “Harsh occupations, life expectancy and social security,” Economic Modelling, 58: 194–202. PRESTON, S. H., AND ELO, I. T. (1995): “Are educational differentials in adult mortality increasing in the United States?,” Journal of Ageing and Health, 7(4): 476–496. RICHTER, W. F., AND WERDING, M. (2020): “Unterschiedliche Lebenserwartungen und Rentenanpassung: Ein Beitrag zur L¨osung eines vernachl¨assigten Verteilungskonflikts,” Perspektiven der Wirtschaftspolitik, 21(4): 389–402.

213

Miguel S´anchez-Romero and Alexia Prskawetz

ROGERS, R. G., HUMMER, R. A., AND EVERETT, B. G. (2013): “Educational differentials in US adult mortality: An examination of mediating factors,” Social Science Research, 42(2): 465–481. ´ SANCHEZ -ROMERO, M., LEE, R. D., AND PRSKAWETZ, A. (2020): “Redistributive effects of different pension systems when longevity varies by socioeconomic status,” The Journal of the Economics of Ageing, 17: 100259. ´ SANCHEZ -ROMERO, M., AND PRSKAWETZ, A. (2017): “Redistributive effects of the US pension system among individuals with different life expectancy,” The Journal of the Economics of Ageing, 10: 51–74. ´ SANCHEZ -ROMERO, M., AND PRSKAWETZ, A. (2019): “Social security: Long-term financing and reform.” In: Gu, D., and Dupre, M. (eds), Encyclopedia of Gerontology and Population Ageing, Cham: Springer. https://doi.org/10.1007/978-3-319-69892-2 535-1. ´ SANCHEZ -ROMERO, M., AND PRSKAWETZ, A. (2020): “The impact of reducing the pension generosity on inequality and schooling,” De Economist, 168(2): 279–304. ´ SANCHEZ -ROMERO, M., SCHUSTER, P., AND PRSKAWETZ, A. (2021): “Redistributive effects of pension reforms: Who are the winners and losers?” ECON WPS - Working Papers in Economic Theory and Policy No. 06/2021. TU Wien, Institute of Statistics and Mathematical Methods in Economics, Research Unit in Economics, Vienna. ¨ SCHR ODER , C. (2012): “Profitability of pension contributions—Evidence from real-life employment biographies,” Journal of Pension Economics and Finance, 11(3): 311. SHKOLNIKOV, V. M., ANDREEV, E. M., JASILIONIS, D., LEINSALU, M., ANTONOVA, O. I., AND MCKEE, M. (2006): “The changing relation between education and life expectancy in central and eastern Europe in the 1990s,” Journal of Epidemiology and Community Health, 60(10): 875–881. WALDRON, H. (2007): “Trends in mortality differentials and life expectancy for male Social Securitycovered workers, by socioeconomic status,” Social Security Bulletin, 67(3): 1–28. WEN, M., AND GU, D. (2012): “Air pollution shortens life expectancy and health expectancy for older adults: The case of China,” The Journals of Gerontology Series A: Biological Sciences and Medical Sciences, 67(11): 1219–1229.

11.7 11.7.1

Appendix

Determinants of the IRR

In this section we demonstrate that the IRR i can be written as a function of the five main components discussed in the main text—i.e., the replacement rate φ; the remaining years of life at retirement, denoted by LE; the capitalization index r; the social contribution rate τ ; and the e) and the capital that could ratio between the capital actually accumulated until retirement (K have been accumulated until retirement if the life table that corresponds to the individual were used (K). Our analysis proceeds in several steps. First, we derive a general form of the pension benefit. Second, we substitute pension benefits and contributions in (1) that implicitly define the IRR. Third, we rearrange the terms in the implicit equation of the IRR to see its dependency on the five components. Finally, we show that the IRR does not depend on the income level of the individual and hence that the IRR is an appropriate measure for analyzing the redistributive properties of pension plans. e that Pension benefits received at each age, b(a), are calculated as a fraction ψ(a) of a stock K e is the result of capitalizing a stream of is accumulated until the age of retirement. The stock K flows during working ages that is a function of labor income, according to a capitalization index e are given by r and a mortality hazard rate, e µ. Thus, the dynamics of K e ∂K e + φ (y(a)) = (r + e µ(a)) K ∂a

(10)

where e µ is the mortality hazard rate at age a associated with the survival function e S used by the pension plan. The function φ (y(a)) captures how the pension plan uses the labor income at age a to compute the pension benefits. The functional form of φ(y(a)) = φ0 y(a), with φ0 214

Social Security Reforms in Heterogeneous Ageing Populations

being a positive scalar, differs by the pension system analyzed. For instance, in a DC system, φ (y(a)) is equal to the contributions paid τ y(a), where τ is the social contribution rate. In a DB system, φ (y(a)) = βy(a), where β is the accrual rate. In the German pension point system, φ (y(a)) = y(a)/¯y(a), where y¯ (a) is the economy-wide average labor income when the individual is of age a. Thus, capitalizing (10) until the retirement age R, the pension benefits at age a are given by Z R e S(x) e φ (y(x)) dx, (11) b(a) = ψ(a)K (R) = ψ(a) er(R−x) e S (R) 0 with ψ(a) = ϕeρ(a−R) where ϕ is the pension replacement rate and ρ is the rate of increase (in real terms) of pension benefits. The contributions paid at each age, denoted by t(a), are given by t(a) = τ y(a).

(12)

Substituting the benefits received, (11), and the social contributions paid, (12), in (1), we obtain Z R Z ω e(R)da − e−ia S(a)ϕeρ(a−R) K e−ia S(a)τ y(a)da = 0, (13) 0

R

where i is the IRR and S(a) is the probability of surviving to age a of the individual. e(R) outside of the first integral, and Dividing both sides of (13) by eiR /S(R), leaving ϕ K moving the second integral to the right-hand side of (13) gives Z ω Z R S(a) (ρ−i)(a−R) S(a) e ϕ K (R) e da = τ e−i(a−R) y(a)da. (14) S(R) S(R) R 0 Now, let us denote by K the hypothetical capital that the individual would have accumulated in case of using the individual-specific survival probability S. The total capital at the age of retirement R would be Z R S(a) K(R) = e−r(a−R) φ (y(a)) da. (15) S(R) 0 Also, let us denote by P the ratio between the capital stock accumulated until retirement e(R) and the hypothetical capital stock that would have been accumulated in case of using K e(R)/K(R). the individual-specific survival probability K(R), i.e., P = K Multiplying and dividing the right-hand side of (14) by K(R), dividing both sides of e(R), and using the definition of P, gives Equation (14) by K R R −i(a−R) S(a) Z ω τ 0 e S(R) y(a)da (ρ−i)(a−R) S(a) da = R R . (16) ϕ e S(a) −r(a−R) S(R) P R 0 e S(R) φ (y(a)) da Now using the fact that φ (y(a)) is a constant fraction, φ0 , of y(a), we have R R −i(a−R) S(a) Z ω τ/φ0 0 e S(R) y(a)da (ρ−i)(a−R) S(a) . ϕ e da = RR −r(a−R) S(a) y(a)da S(R) P R 0 e S(R)

(17)

From Equation (17) we can show that the IRR, i, is a function of the conversion factor a weighted measure of the remaining years of life at retirement, denoted by LE = R ω ϕ, (ρ−i)(a−R) S(a)/S(R); the social contribution rate τ ; the ratio P; and the capitalization index e R r. Note that we have not included φ0 because its inverse (i.e., 1/φ0 ) has the same influence on i as the social contribution rate τ . So, for simplicity, we have skipped this term. In sum, we have shown that i can be expressed in terms of these five components, i.e., i := f (r, τ, P, φ, LE). 215

Miguel S´anchez-Romero and Alexia Prskawetz

11.7.2

Independence of the IRR to Different Income Levels

Let there be two SES groups {1, 2} whose members only differ with respect to their labor income level. Let y1 (a) = ξ1 y(a) and y2 (a) = ξ2 y(a) be the labor income at age a of individuals belonging to SES groups 1 and 2, respectively, with ξ1 > ξ2 > 0. Using (14) and dividing both sides of the equation by the hypothetical capital stock Kj (R) that would have been accumulated if age-specific survival probabilities were applied, with j ∈ {1, 2}, gives ej (R) Z ω (ρ−i)(a−R) S(a) K ϕ e da = τ Kj (R) R S(R)

RR 0

S(a) yj (a)da e−i(a−R) S(R)

Kj (R)

,

(18)


Given that φ yj (a) = φ0 yj (a) = φ0 ξj y(a) we obtain RR R R −r(a−R) eS(a) e(a) φ0 ξj 0 e−r(a−R) eSS(R) y(a)da y(a)da ej (R) K e 0 e S(R) = P= = , RR R R −r(a−R) S(a) S(a) Kj (R) φ0 ξj 0 e−r(a−R) S(R) y(a)da 0 e S(R) y(a)da

(19)

which implies that P is not affected by the income level ξj , because the terms in the numerator and the denominator cancel each other, as shown on the right-hand side of (19). Similarly, we have that the ratio on the right-hand side of (18) is ξj

RR

φ0 ξj

0

RR 0

RR

S(a) e−i(a−R) S(R) y(a)da S(a) e−r(a−R) S(R) y(a)da

=

0

φ0

S(a) e−i(a−R) S(R) y(a)da

RR 0

S(a) e−r(a−R) S(R) y(a)da

,

(20)

which also does not depend on the income level ξj . Thus, plugging (19) and (20) into (18) and dividing both sides by P gives (17), which shows that the IRR is not affected by the income level.

216

12 ECONOMIC PREPARATION FOR RETIREMENT1 Michael D. Hurd and Susann Rohwedder

Abstract We review three main approaches to assessing economic preparation for retirement: the income replacement rate, a life-cycle model optimization approach, and a consumption-based approach. The income replacement rate is widely used by financial advisors and in research on economic preparation for retirement. However, it is ill-equipped to capture some key features of the financial planning for retirement problem. The other two approaches address the main shortcomings of the income replacement rate. The low correlation between the income replacement rate and the preparation measures from the other two approaches suggests that the income replacement rate can be misleading for many households and that it is unsuitable as a rule of thumb to guide household planning for retirement. The consumption-based approach, informed by life-cycle model considerations, is simple enough that it can be adapted for individuals to use if those individuals are provided with information on some inputs such as survival risk and spending trajectories for persons with similar demographics. According to the consumption-based approach, 51 percent of single persons near retirement in the United States and 81 percent of married persons are adequately prepared, that is, they have a small probability (5 percent or less) to run out of wealth at advanced age, accounting for differential survival and out-of-pocket medical expenditure risk. Unlike the income replacement method, the consumption-based measure conveys the expected financial advantage of retirement-age couples over single people. We highlight several promising avenues for future research.

12.1

Introduction

Economic preparation for retirement addresses whether, on entering retirement, a household has access to economic resources sufficient to permit an adequate level of spending with high probability until death, where in the case of a couple, it would be the death of the surviving spouse. The purpose of this chapter is to discuss several approaches to assessing economic preparation for retirement. The most widely used method is the income replacement rate. However, it has several drawbacks that we spend some time elucidating. We contrast it with two other methods: The dynamic programming (DP) approach, which we only briefly outline because it is impractical for widespread use, and a consumption-based method that we deem relatively easy to implement and that addresses many of the shortcomings of the income replacement rate. DOI: 10.4324/9781003150398-14

217

Michael D. Hurd and Susann Rohwedder

A main issue is what constitutes an “adequate level of spending.” One might think of two definitions: (1) A minimal level of spending such as the poverty level or (2) a level of spending that is consistent with the lifetime economic resources of the household or perhaps the level of accustomed spending close to retirement. The first has received no support in advanced economies as evidenced by the linkage of public and private pensions to lifetime earnings (so that the level of pensions varies with lifetime earnings) and the emphasis on the need for private wealth accumulation. Measure (2) assesses adequacy comparatively within a household, that is, relative to each household’s economic resources measured at a moment in time or across multiple periods. We will only discuss measure (2). With extensive data, one could imagine an ex-post measure found by following a cohort over its entire lifetime, observing and accumulating its access to resources, mainly earnings but also other sources such as inheritances and transfers and then counting how many households maintained adequate levels of spending until the death of the surviving spouse. Data requirements make this unfeasible, and defining “adequate” would still be necessary. Further, we want an ex-ante evaluation: What fraction of households approaching retirement is adequately prepared? The institutional context is important. At one extreme, retirement resources are composed of defined benefit (DB) pensions only (public or private). In this case, no accumulation of bequeathable wealth occurs, and adequacy would be gauged by whether the level and trajectory of pension payments are correctly set by public and industry policy. One would find whether expected discounted marginal utilities of consumption are equalized intertemporally. At the other institutional extreme, the only retirement resources are bequeathable wealth. Then the research question is whether assets at retirement are at the right level to equate expected, discounted marginal utilities in retirement to average achieved discounted marginal utilities prior to retirement; that is, have households saved optimally. The economies of developed countries have a mix of both where public and private DB pensions and bequeathable wealth, including defined contribution (DC) pensions, are important. In this context, a life-cycle DP model is a natural vehicle for evaluation. Conditional on parameters, earnings paths, and specification of the stochastic environment, it solves for optimal consumption and wealth paths. (For simplicity we ignore choice of retirement age in this part of the discussion.) Within the model one could ask: At retirement what is the optimal level of wealth for a person with some set of characteristics? Those with less than optimal wealth are underprepared; those with more are overprepared. An extension would allow some band around the optimal level of wealth to account for stochasticity (e.g., +/− 15 percent). Ideally, the model should account for the shape and stochasticity of earnings paths, which would include health shocks affecting earnings and unemployment; rates of return on asset classes; demographics including marriage, divorce, widowing, and children; and out-of-pocket spending for healthcare. Following retirement, the model should account for mortality, especially differential mortality, widowing and the accompanying change in spending requirements, spending for healthcare, rates of return on assets including inflation, taxes (especially tax treatment of tax-advantaged savings), and possibly a change in the utility production function that uses health as an input. One weakness of using a DP model is that the data requirements are very substantial; another is that individual beliefs about the distributions of stochastic events, which help determine optimal choices, are not usually observed. More fundamentally, the DP model assumes optimality, and it is often used to choose utility function parameters that match central tendencies (conditional on observables) of patterns in the data, assuming these reflect optimizing behavior. Thus the ratio of overprepared households to underprepared households will be about unity absent macro shocks that unexpectedly affect most people. To the extent 218

Economic Preparation for Retirement

that macro shocks occur, such as the Great Recession or a stock market boom, majorities could be either underprepared or overprepared. Because of the data requirements and the restrictions imposed by economic theory, which may not reflect actual behavior, at least two other approaches are in use that we will discuss: an income replacement rate method and a consumption-based simulation method. Much of the research on adequacy of retirement resources is based on the Health and Retirement Study (HRS). The HRS is a general-purpose survey of the U.S. population over age 50. It is longitudinal, interviewing the same age-eligible respondents and their spouses biennially since 1992. Refresher cohorts are added every 6 years. It elicits comprehensive information on the economic situation of the household that includes detailed measures of household sources of income, total assets and their components, and employment. It collects information across multiple domains to support cross-disciplinary research. The HRS expends considerable effort tracking the vital status of study participants, which is important for studying survival. The purpose of this chapter is first to discuss the income replacement rate, which is most widely used because of its relative simplicity, and to highlight several key features of the retirement planning problem that the income replacement rate does not easily incorporate. We then discuss two other main approaches to assessing the adequacy of retirement resources that are conceptually more flexible than the income replacement rate approach to incorporate the complexities of the retirement planning problem: (1) a life-cycle DP model and (2) a consumptionbased simulation model. For each of these, we compare how the results relate to the income replacement rate approach. The conclusions summarize the implications for informative ex-ante assessments of households’ economic preparation for retirement.

12.2

The Income Replacement Rate

The income replacement rate (income immediately following retirement divided by income preceding retirement) became widely used as a measure of economic preparation for retirement by financial advisors and in the popular financial literature. For example, various online financial calculators suggest that retirement income replacement rates should exceed 70 percent. The target income replacement rate is meant to serve as a rule of thumb to provide a simple starting point for individuals and households in their financial planning for retirement. Prior research has investigated various conceptual issues concerning the implementation of the income replacement rate and its extensions to deal with the shift from DB retirement resources and to account for various types of financial and real assets.2 Recommended target replacement rates tend to range between 65 and 95 percent, depending on the source considered. They are less than 100 percent because for most retired households taxes are lower, work-related expenses are eliminated, and there is less need to save. The large range reflects in part differences in definitions of the replacement rate used (Biggs and Springstead, 2008), but also the fact that the appropriate target replacement rate may differ across households, depending on marital status, number of earners, and income group (Munnell et al., 2006) and on whether the household has children (Scholz and Seshadri, 2009).3 Despite these complexities, the income replacement rate continues to occupy an important role in retirement planning and assessments of adequacy of retirement resources. For example, the National Retirement Risk Index produced by the Retirement Research Center at Boston College (Munnell et al., 2006) computes projected retirement income replacement rates and compares them with target replacement rates. By this measure economic preparation for retirement has been and continues to be inadequate for many approaching retirement: Since 2010 219

Michael D. Hurd and Susann Rohwedder

the National Retirement Risk Index has hovered around 50 percent, meaning that only about half of working households are on track to be economically prepared for retirement (Munnell et al., 2021).

12.2.1

Measurement of Replacement Rates

The income replacement rate is the ratio of income following retirement to income prior to retirement. But the measurement of either is not straightforward. What should go into the denominator? Earnings can fluctuate, so one may not want to use earnings just prior to retirement. If both spouses work but retire at different times, when should we measure earnings? Some papers use “permanent” earnings, which are proportional to lifetime earnings when the retirement age is fixed. Others use earnings when the earners are in their late 50s. Regardless of the denominator, the numerator needs to account both for annuity income (Social Security and DB pensions) and bequeathable wealth. A common recommendation is to draw down some fraction of bequeathable wealth, say 4 percent per year, and add that amount to annuity income. A second is to annuitize bequeathable wealth and add it to annuity income. However, neither of these methods addresses the role of children in the allocation of lifetime consumption across the life cycle. Holding constant lifetime earnings, households with (grown) children should consume less in retirement than households without children, and households with more children should consume less than households with fewer children. For example, a couple with three children will have allocated more of its lifetime earnings to the pre-retirement years than a couple with two children and so would have less to consume after retirement. Thus, a lower ratio of total annuity income (annuitized wealth plus annuities) to earnings by the first couple should not necessarily be taken to indicate worse economic preparation, yet it would be under the income replacement rate method. More generally and to a first order, a household will allocate resources according to person-years lived at various ages: More children will shift consumption toward pre-retirement, long-lived parents toward post-retirement, and so forth. An important and unresolved issue is the treatment of owner-occupied housing. It produces a large implicit flow of both income and consumption before and after retirement that, when added to both numerator and denominator, increases the replacement rate. A further issue is whether to annuitize housing wealth in measuring the replacement rate.

12.2.2

Shortcomings of the Income Replacement Rate

No doubt the simplicity and transparency of the income replacement concept have contributed to its use. Yet the concept inadequately captures several issues relevant to economic preparation for retirement. We list some here, followed by a discussion and examples of several. 1. Just about one-third of full-time workers follow traditional retirement paths, transitioning from working full time to being fully retired. For the other two-thirds, their late-in-life work trajectories involve continued full-time or part-time work, unretirement, unemployment, and disability (Maestas, 2010; Hudomiet et al., 2017). For them, it is not obvious when to stop measuring income as pre-retirement and when to start measuring it as post-retirement. The situation becomes substantially more complicated when assessing retirement preparation for couples with two earners. 2. People can finance consumption out of savings; most of their savings are not annuitized, and so the financing of consumption is not recorded as income but as a drawdown of capital. 3. The time horizon or survival curve of the household means resources do not have to last indefinitely; due to differential mortality, the time horizon varies across households. 4. Spending is reduced following widowing. 220

Economic Preparation for Retirement

5. The demographic structure of the household, especially children, will help determine the fraction of lifetime resources to be consumed during each phase of life and hence the amount that should be carried into retirement. 6. The consumption path is not flat, but as observed empirically declines with age. 7. Taxes differ from pre- to post-retirement, and this difference varies by household according to the level of income, but also according to how much of their wealth has been sheltered from taxes during the accumulation phase and hence will be taxed when withdrawn from a tax-advantaged account. 8. Households vary in their economic preferences and constraints, with implications for needed resources.

12.2.3

Financing of Consumption Out of Savings

Increased diversification of financial resources in retirement in recent decades highlights the deficiencies of a simple replacement rate as an indicator of economic preparation for retirement. When private savings were small, the sum of post-retirement Social Security benefits and DB pensions could be compared with pre-retirement earnings to yield a replacement rate. This comparison depicted the impact of retirement on income and thus consumption. With the growth in savings and in DC pension plans, this simple comparison is inaccurate. Replacement rate calculations often account for bequeathable wealth by converting it to an income flow via annuitization: At retirement what level of annuity would the stock of bequeathable wealth (including DC pension accounts) be able to purchase? That income flow would be added to Social Security and DB benefits to yield an income equivalent measure of the total economic resources of the household. An objection to this practice is that households very rarely annuitize any bequeathable wealth, and the longevity insurance provided by annuities is not actually in effect under this synthetic annuitization. Assuming complete annuitization therefore misrepresents feasible consumption in case of survival to advanced ages. Nonetheless, adding in annuitized wealth is certainly better than ignoring it completely: Even if not formally annuitized, bequeathable wealth can be drawn down over time to support consumption needs.

12.2.4

Differential Mortality

The previous example assumes equal-length lifetimes, but life expectancy varies substantially by observable characteristics, resulting in sizeable differences in individuals’ time horizon both within couples and especially across households. Figure 12.1 shows survival curves for females estimated from HRS data. Single females lacking a high school degree have about a 50 percent chance of surviving from age 65 to age 81, whereas married females with a college degree have a 50 percent chance of surviving to age 96. Exacerbating the comparison between married households and single households is that the end of the lifetime of a married household occurs at the death of the surviving spouse, which in expectation is substantially greater than the expected age at the death of either spouse. However, these increased life years following retirement are at least partially offset by reduced spending needs by the survivor compared with the couple.

12.2.5

The Role of Children

Broadly conceived, demographics importantly determine desired wealth at retirement and, hence, annuitized income from bequeathable wealth. To illustrate the role of children, consider a maximally simplified life-cycle model. Lifetime wealth is given as the present value of lifetime earnings, and the retirement age is fixed. The length of life is given, which is the same for everyone. There are no returns to scale in consumption, family utility is additive to the 221

Michael D. Hurd and Susann Rohwedder

Figure 12.1 Differential survival, females. Source: Authors’ calculations based on HRS waves 1992–2016: Logistic regression of survival from wave t to t+1 (on average 2-year survival), controlling for education and a spline for age band.

utility of members, and all family members have the same utility function. The family should maximize within-period utility by equal allocations of spending to each person, and it should maximize utility across periods by equal spending per person in each period. The implication is that spending each year should be proportional to the number of persons in the household that year. At retirement, the fraction of lifetime wealth already spent should be proportional to the fraction of household’s lifetime person-years already lived. Or equivalently, wealth accumulation should be proportional to the fraction of lifetime person-years remaining to be lived. For example, suppose that life begins at age 20, and people work to age 65 and die at age 80. A single person will have 60 person-years of lifetime in total and 45 person-years by age 65. Thus wealth at retirement should be one-quarter of lifetime wealth. A couple in which the spouses are the same age will have 120 person-years of lifetime with 90 person-years already lived by age 65. Wealth at retirement should also be one-quarter of lifetime wealth. But the same couple with two children who both lived at home for 20 years each will experience 160 person-years of lifetime, of which 130 person-years are lived by age 65. Wealth at retirement should be 3/16 of lifetime wealth, which is 75 percent of the wealth of a couple without children.

12.2.6

Life-Cycle Consumption Path Is Not Flat, Varies by Observables

A simple life-cycle model predicts that spending by single persons is not flat as individuals age but rather declines due to the risk of dying early and “wasting” economic resources (Yaari, 1965). A decline could also be due to a complementarity between some types of spending and health. For example, the marginal utility of spending on private transportation will likely decline as health worsens because of fear of accidents. Indeed the share of total spending on private transportation declines among single persons from about 12 percent of spending at ages 60–64 to about 5 percent at ages 85–89 (Hurd and Rohwedder, 2010). Survey results further support the importance of reduced enjoyment of various spending-related activities at older ages: When asked how enjoyment from leisure, travel, new clothes, cars, or appliances changed compared with 6 years ago, most older respondents reported declines, and more so at advanced 222

Economic Preparation for Retirement

Figure 12.2 Fitted life-cycle consumption paths, single females. Source: Authors’ calculations, Consumption and Activities Mail Survey 2001–2017, based on methodology employed in Hurd and Rohwedder (2012), using more recent data.

older ages (Rohwedder et al., 2022). Figure 12.2 shows the spending path by single females estimated from panel spending data in the HRS. It shows that the path is flatter among those with a college degree, which the Yaari model would predict because of greater survival. From the perspective of economic preparation for retirement, this decline in total spending reduces the amount of wealth needed at retirement and hence the income replacement rate (including annuitized bequeathable wealth).

12.3

Preparation for Retirement in the Context of a Structural Life-Cycle Model

Scholz et al. (2006) use a life-cycle model to find the fraction of households that reached age 65 with adequate wealth, which is the wealth their DP model solves for in a stochastic lifecycle context. Most importantly, their model accounts for earnings histories that were linked to respondents in the HRS. Thus they were able to estimate the trajectory and volatility of earnings, which are inputs into the choice of optimal consumption and achieved wealth. They also accounted for marital status, the number of children, and several other aspects previously listed. Rather than estimating preference parameters that would fit the observed data, they simulated the optimal wealth paths of households using parameter values taken from prior studies. To assess empirically whether households were adequately prepared for retirement, they compared the optimal level of bequeathable wealth with observed wealth for a subsample of HRS households in the 1992 wave. They found that about 80 percent of households reached or exceeded their optimal wealth. Perhaps surprisingly, they found that the likelihood of having undersaved was independent of the position of the household in the distribution of lifetime earnings once marital status is controlled. 223

Michael D. Hurd and Susann Rohwedder

Scholz and Seshadri (2009) use the life-cycle model to calculate what a replacement rate should be (the “optimal target” replacement rate). Thus their approach addresses many of the shortcomings of the typically calculated income replacement rate. They find that the average optimal replacement rate from their model was 70 percent, a rate commonly specified by planners, but that the specific target was widely dispersed, depending on the economic circumstances. Thus a target replacement rate of, say, 70 percent could be highly inaccurate for many individuals. If individuals strived to achieve the 70 percent target, some individuals would increase their saving rate to reach the 70 percent target, resulting in their oversaving, yet other individuals already at 70 percent ought to increase their saving. This heterogeneity makes the 70 percent target a flawed planning tool. We have used the 70 percent target as an example because it is often cited, but the statement would be true for other potential target replacement rates. An important point of Scholz and Seshadri is that it quantifies some weaknesses of the income replacement rate discussed previously.

12.4

Consumption-Based Measure of Economic Preparation for Retirement

This method begins from the observation that consumption, not income, produces utility. This is particularly relevant after retirement because consumption is not the same as income, especially because of the ability to use wealth to finance consumption.4 Broadly speaking, the measure finds whether a household has, with high probability, the resources to finance a trajectory of spending from shortly following retirement until death (in the case of a single person) or until the death of the surviving spouse (in the case of a couple). The initial level of spending (or consumption, which will be used interchangeably with spending) is determined from the observed level of spending at retirement and the life-cycle path from observed rates of change of spending in panel data. Framing the question in this way is appealing to the nonspecialist because it aims to answer the relevant question: Can the household maintain its accustomed level of material well-being? Someone is adequately economically prepared for retirement if, with high probability, he or she dies with positive wealth (Hurd and Rohwedder, 2012). The method accounts for uncertainty about the date of death, differential mortality, taxes, spending out of assets, marital status, and the consumption path (across time and persons). The method accounts for heterogeneity by age, sex, marital status, education, and initial economic conditions. An important difference from a full life-cycle DP approach, as taken by Scholz and Seshadri, is that it begins at about the typical retirement age (65–69), which reduces data requirements, and that it makes heavy use of longitudinal spending data. It does not impose utility maximization but rather uses observations on all the important data elements, that is, data on what people actually do. For example, rather than assuming utility function parameters and utility optimization to derive a consumption trajectory, we use panel observations on spending to find empirically the consumption trajectory actually taken. We think this approach has considerable merit because of its close ties to data and its relatively modest assumptions. However, as with the DP model, it does depend on assumptions. Most critical is the assumption that, conditional on the consumption level near retirement, the consumption trajectory will follow the path observed in past data, modified by demographics. When this approach is weighed against a DP approach, the choice of which to use will depend on the research question under consideration and on the trade-off between their differing assumptions and the empirical support for the assumptions. Our data are from the HRS. In addition to data on income and wealth, we use data from the Consumption and Activities Mail Survey (CAMS). CAMS is a panel survey supplement to 224

Economic Preparation for Retirement

the HRS that measures spending in some 36 categories to arrive at a measure of total spending for a random subset of households in the HRS. The CAMS data aggregate to population levels that closely match aggregates from the Consumer Expenditure Survey. We selected a sample of single and married persons observed at ages 66–69 in HRS data at some point in the years 2000–2016 and asked whether they have the financial resources needed to finance a consumption plan from retirement through the end of life. That plan begins at an observed starting consumption level for each household and follows a path whose shape is determined by observed consumption change with age in panel. Because the age at death is unknown and because wealth is not completely annuitized, someone who dies unexpectedly early may have been adequately prepared ex post, and someone who survives to extreme old age will not have been adequately prepared ex post. To account for this randomness, we used simulation to find the fraction of times ex post that each household was adequately prepared. We stratified by education, sex, and initial marital status because of differential mortality and variation in the shape of the consumption path. Economic resources are a combination of post-retirement income such as Social Security and pension benefits, housing wealth, post-tax non-housing wealth, and pre-tax retirement accounts.5 The estimations and simulations account for differential mortality risk and, in the case of couples, the lifetime of the couple and the subsequent loss of returns to scale in consumption at the death of the first spouse. The estimations recognize that consumption need not be constant with age nor by socioeconomic status, both for reasons of mortality risk and for reasons of a complementarity between health and consumption. In fact, among single persons, observed consumption declines with age and the decline is greater for less educated persons. For example, based on panel observations on change in spending from CAMS, females ages 70–74 with a high school education reduced spending by 1.4 percent per year, but by 3 percent per year at ages 75–79 (Hurd and Rohwedder, 2012).6 We incorporated the risk of large out-of-pocket spending on healthcare by assuming that the distribution of out-of-pocket spending used in the simulations was the same as the observed distribution. We accounted for taxes, which for some households substantially reduce resources available for consumption because of large holdings of tax-advantaged savings. For other households, however, taxes are nil because of the sheltering of Social Security benefits from taxation at low levels of income. We found via repeated simulation the fraction of times each person dies with positive wealth, and we defined adequate preparation as when that fraction is 0.95 or greater. Table 12.1 shows the percentage of persons ages 66–69 adequately prepared, updated from our earlier work by incorporating HRS data through 2016. Single persons are much less likely to be adequately prepared than married persons: 51 percent versus 81 percent. There is a sharp gradient by education, particularly at the low end of the education distribution. The very low level of preparation among single women who lack a high school degree is particularly notable: Just 25 percent are adequately prepared. Stratifying by race and ethnicity, we find the fraction adequately prepared is substantially lower among Blacks and Hispanics than among non-Hispanic Whites: about 30 percentage points lower among single persons and 15 percentage points among married persons. We conducted additional simulations to find how modest reductions in initial spending would affect preparation. The thought experiment is that perhaps reducing initial spending by, say, 10 percent would not be too difficult and that doing so would be worthwhile to avoid running out of money late in life. Such a reduction would increase economic preparation to 56 percent and 85 percent among singles and couples, respectively, but it would increase the level of preparation among single females lacking a high school degree to just 29 percent. 225

Michael D. Hurd and Susann Rohwedder Table 12.1 Percent adequately prepared, defined as 95–100 percent chance of dying with positive wealth

Note: ∗ The category “other” in race/ethnicity has few observations; so we do not comment on it in the text. Source: Authors’ calculations, HRS 2002–2016 and CAMS 2003–2017, based on methodology employed in Hurd and Rohwedder (2012), using more recent data.

12.5

Comparison with Income Replacement Rates

For a comparison of economic preparation based on our consumption model with preparation based on income replacement rates, we draw on results from Hurd and Rohwedder (2015). In that paper we use a sample from HRS waves 2002–2008 and implement the income replacement rate and the consumption-based approach on the same sample. We use data on individuals, both single and married, ages 66–69, who were included in CAMS (so that consumption was observed). We chose individuals who were observed in some prior wave (going back as far as 1992) to have had earnings when aged 59–61. Thus, we could calculate the simple income replacement rate with data on pre-retirement earnings and on post-retirement Social Security and DB pension benefits. Sample members may have had earnings in up to three waves before age 62. For each individual, we took an average across waves when earnings were positive to help compensate for any observation error and to account for any years not in the labor force. For couples estimates of the replacement rate are more complicated because of the need to account for two persons. One of the two people in a couple, for example, may not have earnings. One may reach retirement at a different age from the other. What “pre-retirement income” means in household terms is not even obvious if the difference in ages is substantial. Our approach was 226

Economic Preparation for Retirement

Figure 12.3 Cumulative distributions (percent) of replacement rates, single and married persons. Source: Adapted from Hurd and Rohwedder (2015).

to treat each spouse separately, that is, as if they were single persons. His or her pre-retirement income is what he or she earned if working at ages 59–61. Post-retirement income of each married person includes his or her DB pension and Social Security benefits and 50 percent of the income resulting from annuitizing the assets of the couple. Figure 12.3 shows the cumulative distributions of replacement rates for single and married persons. That is, any point on a line shows the percentage of persons (read off the y-axis) having the replacement rate (or less) shown on the x-axis. For example, 38 percent of married persons have a replacement rate of 0.5 or less, and 25 percent of single persons have a rate of 0.5 or less. Notice some very low replacement rates: 9 percent of singles and 14 percent of married persons have an income replacement rate of 0.3 or less. In contrast, 69 percent of singles and 76 percent of married persons have replacement rates of 1.0 or less, meaning that 31 percent and 24 percent have replacement rates of more than 1.0. Many financial advisors favor a 70 percent replacement rate. The cumulative distribution of replacement rates on the y-axis reaches 46 percent for a replacement rate of 0.7 on the x-axis. That is, 46 percent of older singles have an income replacement rate of 0.7 or less, implying that 54 percent have a replacement rate of more than 0.7. This suggests that most single persons are well prepared economically for retirement at least by this measure. In contrast, 59 percent of married persons have a replacement rate of 0.7 or less, implying that just 41 percent of married persons are economically prepared for retirement, substantially fewer than single persons. The results of these analyses are counter to conventional thinking, which anticipates higher post-retirement incomes for married persons than for single persons because couples have considerably more than twice the assets of single persons. But these higher-income replacement rates for single persons are actually due to their considerably lower pre-retirement earnings, not due to higher post-retirement incomes. The inability of the income replacement rates to distinguish between the economic preparation of singles and couples argues against their use more generally for comparatively assessing economic preparation for retirement. We measured replacement rates in other ways such as only including DB pensions and Social Security income and adding a 4 percent drawdown rate of assets. These different methods produced quite different measures of the replacement rate: For example, just 35 percent of 227

Michael D. Hurd and Susann Rohwedder

single persons achieved a replacement rate of more than 0.7 (rather than 54 percent) when postretirement income only included DB pension income and Social Security income. The large variation in replacement rates according to what is included in post-retirement income adds further weight against their use. Rather than treating married persons as single persons, in an alternative version, we treated them as a couple to measure the pre-retirement income and the post-retirement income of the couple. This change added persons who had no earnings at ages 59–61 but whose spouse did have earnings. We also pooled the post-retirement pension and annuity income of each spouse. However, because of a wide range of age differences between the spouses, defining the retirement of the couple became even more difficult, so we limited the age difference to 5 years. These changes resulted in an increase in the percentage of economically prepared married persons from 41 percent to 54 percent. Which treatment is to be preferred is not obvious to us. We now turn to a comparison of economic preparation for retirement as measured by the income replacement rate with our consumption-based measure.7 Hurd and Rohwedder (2015) report that under the consumption-based measure of adequate economic preparation for retirement the fraction of single persons adequately prepared is 59 percent, which is reasonably close to the percent adequately prepared under a 0.7 target replacement rate: 54 percent. Some 80 percent of married persons are adequately prepared according to our consumption-based measure but using a 0.7 target replacement rate just 41 percent are adequately prepared. At the population level, the consumption-based and the replacement rate-based give comparable results for singles, but grossly different results for couples. Furthermore, the advantageous position of singles relative to couples in the replacement rate measure defies comparisons of economic resources such as wealth between the two groups. A second comparison is at the individual level, specifically the correspondence between economic preparation according to the replacement rate and economic preparation according to the consumption measure. Figure 12.4 shows for each replacement rate the percentage of persons adequately prepared according to the consumption measure. For example, among single persons with an income replacement rate of 0.11–0.20, 38 percent are adequately prepared according to the consumption measure. As the figure shows, the consumption measure increases as the replacement rate increases but the relationship is weak, particularly among married persons. For example, 73 percent are deemed adequately prepared according to the consumption measure even though their replacement rates are in the 0.31–0.40 band. One might ask how this could happen, that is, how could households where about three-quarters are adequately prepared have such a low replacement rate. It is particularly puzzling because the replacement rate includes annuitized bequeathable wealth, which is an important component of the consumption-based evaluation. An important proximate answer is that the replacement rate does not reference spending, which is the mechanism used by the household to affect the likelihood of not exhausting wealth. For example, a household with the following combination of circumstances might end up in this situation: unusually high earnings just prior to retirement and relatively low levels of bequeathable wealth (say, the result of rapidly increasing earnings) resulting in a low replacement rate, combined with a low level of spending appropriate to the low level of wealth, resulting in adequate preparation for retirement. As with Scholz and Seshadri, our overall level of economic preparation for retirement based on the income replacement rate is similar for single persons to the average frequency of economic preparation for retirement based on our consumption approach, but at the individual level, there is little concordance. Among couples the levels are greatly different: 41 percent versus 80 percent. 228

Economic Preparation for Retirement

Figure 12.4 Percent adequately prepared according to consumption-based measure, single and married persons, by income replacement rate. Source: Adapted from Hurd and Rohwedder (2015). Horizontal axis shows 11 intervals covering the range of observed income replacement rates. For the group of households whose income replacement rate falls in an interval we compute the fraction adequately prepared according to the consumptionbased measure, which is shown on the vertical axis. For example, among households with an income replacement rate between 0.41 and 0.50, about 72 percent were adequately prepared among married persons and 53 percent among single persons according to the consumption-based measure.

Note that only the consumption-based measure conveys the expected retirement-age financial advantage of couples over single people. Mean reported post-retirement income is twice as high for couples as for singles. Mean reported financial wealth, including individual retirement accounts, is three times as much for couples as for singles. These multiples may understate the advantages accruing to couples. Because of returns to scale, couples do not need to spend twice as much as singles to be as well off. For example, the structure of spouse and survivor benefits under Social Security recognizes returns to scale. In the simplest one-earner case, the couple’s benefit would be 1.5 × Primary Insurance Amount and the survivor’s benefit would be 1.0 × Primary Insurance Amount. Under equal sharing of income, the structure implies that a married person needs just 75 percent of the income of a single person to be as well off. Furthermore, as married persons become widowed, and if we assume that the survivor inherits the wealth of the deceased spouse, the survivor will tend to have similar levels of wealth supporting lower consumption. The lack of a sensible or anticipated relationship between the replacement rate of single persons and those of couples argues against using these ratios to indicate retirement preparation adequacy. The consumption-based measure does differ in the expected way across singles and couples. Moreover, Figure 12.4 shows that little relationship exists between the income replacement measures and the consumption-based measures, implying that the income replacement rate is not suitable to be used as a rule of thumb because it would be highly misleading for many households.

12.6

Summary and Conclusions

Because of the substantial variation across persons in lifetime earnings, evaluating economic preparation for retirement requires normalization to adjust for that variation, that is, a comparison of economic resources at retirement with some measure of lifetime economic resources. In 229

Michael D. Hurd and Susann Rohwedder

the case of the income replacement rate, those measures are post-retirement income and preretirement income. In the case of a DP approach, the comparison is between actual wealth at retirement with optimal wealth, or the wealth that results from optimal choices given the economic environment including the earnings path. In the case of the consumption measures, the comparison is implicitly between the mix of economic resources that can support a consumption path from retirement to the end of life and the level of consumption that the individual chooses around the time of retirement. Because of its simplicity, the income replacement rate is widely used as a measure of economic preparation for retirement by financial advisors and in the popular financial literature. Yet, our review shows that the income replacement rate is conceptually ill-equipped to capture key features of financial planning for retirement. Even defining pre-retirement income is complicated. A household often has two earners with different retirement ages, either because the two differ substantially in age, or because one prefers to retire at a different age than the other. Furthermore, some workers now have part-time options and can move among various income-earning states. How to determine the timing of a couple’s retirement and quantify their pre- and post-retirement income is unclear. These obstacles can be addressed by defining pre-retirement income as “permanent income,” which is proportional to lifetime earnings, but at the cost of a loss of simplicity. Post-retirement income was relatively easy to define in an era where most income was from Social Security and DB pensions. However, DB plans have been largely replaced by defined contribution plans, which support spending by a drawdown of wealth rather than by an income flow. Merging DC pensions and bequeathable wealth with DB plans and Social Security can be addressed by annuitizing wealth so that all resources are put on an income or flow basis for comparison with pre-retirement income. But little bequeathable wealth is actually annuitized, so the procedure overstates “usable” income. And it results in the loss of simplicity. Even if pre- and post-retirement income were easy to define and measure, however, heterogeneity across persons according to observable characteristics means that any income replacement rate that applies to everyone will be wrong for many. If the lifetime economic environment facing an individual can be characterized and if preferences of the individual were known, DP models could find the level of wealth at retirement that would result from optimal choices. If the observed level of wealth is greater than the optimal level, that individual would be adequately prepared (or perhaps overprepared). A major problem is characterizing the lifetime economic environment that the individual perceived when making choices, which may be difficult based on historical data. But because of misperceptions by the individual, the perceived environment may not coincide with the objectively measured environment. A second problem is knowing the preferences of the individual. Both are required to find the optimal wealth to compare with actual wealth. As a practical matter, an evaluation based on such a DP model is unlikely to be in wide use. Life-cycle considerations inform the consumption-based approach, but it does not depend on past events that can no longer be changed. It asks the question, which perhaps is the most natural query by an individual approaching retirement, of whether the individual’s economic resources can support the current level of consumption with some adjustments for eventual widowing, for observed spending reductions at advanced old age, and for personalized life expectancy. The approach is simple enough to adapt for individual use, especially when providing people with some information on survival probabilities and spending trajectories of people like them. For individual households, assessing retirement preparedness in their 50s seems timely: Children are often out of the house (or the financial commitments associated with children are better 230

Economic Preparation for Retirement

known), the household has fairly good information about future earnings, and time remains to adjust saving and timing of retirement. A weakness of the consumption-based approach is that it permits someone with a very low level of consumption to be economically prepared. In the context of a life-cycle model, this is not necessarily a classification error: Some people will have had very low levels of earnings (or possibly no earnings, living off income support or transfers), and even a very low level of consumption would be in line with those resources. Nonetheless, most people, as support for needs-based social programs shows, would want an additional constraint, a consumption floor. Implementing a floor is beyond the scope of this paper, but it would be practical using the methods we outlined. While financial planners still widely use income replacement rates, some financial planning tools suggest that households first try to predict their lifestyle in retirement, acknowledging that this may vary a lot across households. For example, the Fidelity retirement planning tool now queries the individual about the household’s expected expenses in retirement.8 This moves the discussion toward determining expenses in retirement and comparing them with the household’s income and asset portfolio. The assessment is whether the household’s resources can support that lifestyle, which is conceptually much closer to the approach we put forward in the consumptionbased measure of retirement adequacy than an income replacement rate approach.

12.7

Future Research

In our view, the evaluation of economic resources at retirement is one of the more pressing issues facing families and public policy. How to fix the financing problems of Social Security and Medicare in the U.S. context, or the funding shortfalls of public pension systems in other developed economies, depends on the economic circumstances of the cohorts approaching retirement. While we think the consumption-based evaluation is a useful approach, there are many unanswered questions that future research should address. Here are a few among many. 1. How to integrate in the decumulation phase asset management with the desired spending path; 2. The treatment of housing given its dual roles as providing a flow of consumption services and as an asset that can be used to finance non-housing consumption following downsizing; 3. The use of reverse-annuity mortgages to tap housing equity; 4. The life-cycle evolution of non-healthcare consumption and how it interacts with changes in health and with changes in marital status at advanced age; 5. The substitution between desired bequests and buffer-stock saving; 6. Financing of long-term care and willingness to spend for a high-quality nursing home; and 7. Interactions between informal care and formal care and how they interact with demographics.

Notes 1 The material presented in this chapter draws on prior work supported by grants from the Social Security Administration through the Michigan Retirement and Disability Research Center and from the National Institute on Aging (P01 AG008291). The findings and conclusions expressed are solely those of the authors and do not represent the views of the Social Security Administration, any agency

231

Michael D. Hurd and Susann Rohwedder

2 3

4

5

6 7 8

of the Federal government, the Michigan Retirement and Disability Research Center, or the National Institute on Aging. Joanna Carroll provided excellent programming assistance. See for example, Boskin and Shoven (1984), Au et al. (2004), Munnell and Soto (2005), Munnell et al. (2007), Munnell et al. (2009), and Munnell et al. (2013). Furthermore, some planning advice suggests that the target replacement rate might have to be higher for households that aspire to a generous lifestyle in retirement or anticipate other large expenditures (such as health or long-term care expenditures for themselves or other loved ones). Examples of other studies that adopted a consumption-based approach include Brady (2010) or MacDonald et al. (2016). Although different in some important ways from our approach, they reach similar conclusions about the low correlation between their adequacy measure and the income replacement rate. In the simulations, consumption in each period is financed by drawing as needed from resources in the following order (informed by empirical observation): (1) after-tax income; (2) (after-tax) non-housing/non-individual retirement account assets; (3) pre-tax financial assets, such as individual retirement accounts; and (4) housing. These rates of change are calculated over persons who are not liquidity constrained so as not to mix freely chosen rates of change with constrained rates of change. Our 2015 paper used a sample from HRS 2002–2008 so the results on economic preparation are slightly different from Table 12.1 of this chapter, which used a sample from HRS 2002–2016. See https://www.fidelity.com/retirement-planning/envision-your-retirement.

References AU, A., MITCHELL, O. S., AND PHILLIPS, J. W. R. (2004): “Modeling lifetime earnings paths: Hypothetical versus actual workers.” Working Paper 2004-3. Boettner Center for Pensions and Retirement Research, University of Pennsylvania. BIGGS, A. G., AND SPRINGSTEAD, G. R. (2008): “Alternate measures of replacement rates for social security benefits and retirement income,” Social Security Bulletin, 68(2): 1–19. BOSKIN, M. J., AND SHOVEN, J. B. (1984): “Concepts and measures of earnings replacement during retirement.” NBER Working Paper Series No. 1360. National Bureau of Economic Research, Cambridge, MA. BRADY, P. J. (2010): “Measuring retirement resource adequacy,” Journal of Pension Economics and Finance, 9(2): 235–262. DOI: 10.1017/S1474747208003806. HUDOMIET, P., PARKER, A., AND ROHWEDDER, S. (2017): “Cognitive ability, personality and pathways to retirement: An exploratory study,” Work, Ageing and Retirement, 4(1): 52–66. DOI: 10.1093/workar/wax030. HURD, M. D., AND ROHWEDDER, S. (2010): “Spending patterns in the older population.” In: Drolet, A., Schwarz, N., and Yoon, C. (eds.), The Ageing Consumer: Perspectives from Psychology and Economics, New York: Routledge, pp. 25–50. HURD, M. D., AND ROHWEDDER, S. (2012): “Economic preparation for retirement.” In: Wise, D. A. (ed.), Investigations in the Economics of Ageing, Chicago: University of Chicago Press, pp. 77–113. HURD, M. D., AND ROHWEDDER, S. (2015): “Measuring economic preparation for retirement: Income versus consumption.” University of Michigan Retirement Research Center (MRRC) Working Paper, WP 2015-332. Ann Arbor, MI. MACDONALD, B., OSBERG, L., AND MOORE, K. (2016): “How accurately does 70% final employment earnings replacement measure retirement income (in)adequacy? Introducing the living standards replacement rate (LSRR),” ASTIN Bulletin, 46(3): 627–676. DOI: 10.1017/asb.2016.20. MAESTAS, N. (2010): “Back to work: Expectations and realizations of work after retirement,” Journal of Human Resources, 45: 718–748. DOI: 10.3368/jhr.45.3.718. MUNNELL, A., CHEN, A., AND SILICIANO, R. L. (2021): “The National Retirement Risk Index: An update from the 2019 SCF.” January, No. 21-2. Center for Retirement Research at Boston College. MUNNELL, A. H., GOLUB-SASS, F., AND WEBB, A. (2007): “What moves the National Retirement Risk Index? A look back and an update.” January. Issue Brief Number 7-1. Center for Retirement Research at Boston College. MUNNELL, A. H., AND SOTO, S. (2005): “What replacement rates do households actually experience in retirement?” Working Paper 2005-10. Center for Retirement Research at Boston College.

232

Economic Preparation for Retirement

MUNNELL, A. H., WEBB, A., AND DELORME, L. (2006): “A new National Retirement Risk Index.” Issue Brief No. 48. Center for Retirement Research at Boston College. MUNNELL, A. H., WEBB, A., AND FRAENKEL, R. C. (2013): “The impact of interest rates on the National Retirement Risk Index.” June. Issue Brief Number 13-9. Center for Retirement Research at Boston College. MUNNELL, A. H., WEBB, A., GOLUB-SASS, F., AND MULDOON, D. (2009): “Long-term care costs and the National Retirement Risk Index.” January. Issue Brief Number 9-7. Center for Retirement Research at Boston College. ROHWEDDER, S., HURD, M. D., AND HUDOMIET, P. (2022): “Explanations for the decline in spending at older ages.” Michigan Retirement and Disability Research Center, Working Paper 2022-440. Ann Arbor, MI. SCHOLZ, J. K., AND SESHADRI, A. (2009): “What replacement rates should households use?” Michigan Retirement Research Center Working Paper 2009-214. Ann Arbor, MI. SCHOLZ, J. K., SESHADRI, A., AND KHITATRAKUN, S. (2006): “Are Americans saving ‘optimally’ for retirement?,” Journal of Political Economy, 114(4): 607–643. YAARI, M. E. (1965): “Uncertain lifetime, life insurance, and the theory of the consumer,” The Review of Economic Studies, 32(2): 137–150.

233

13 PENSION POLICY IN EMERGING ASIAN ECONOMIES WITH POPULATION AGEING: WHAT DO WE KNOW, WHERE SHOULD WE GO? George Kudrna, Philip O’Keefe, and John Piggott

Abstract This chapter reviews the current state of knowledge about pension policy and pension policy formulation in emerging economies undergoing demographic transition, and with this background indicates directions for further policy development. The countries we consider are primarily located in East and Southeast Asia, a region that is home to more than 30 percent of the world’s population, and are characterized by increasing life expectancy, falling and/or low-fertility rates, relatively immature social protection policy structures, high rates of informal employment, and in many cases, high rates of co-residency. These features point to the relevance of strands of research that do not normally sit together in thinking about the evidence base for pension policy formulation and its impacts. They include fiscal implications, impacts on economic growth and intergenerational affordability, the relationship between alternative pension models and labor market (in)formality, the role of public benefits in the context of multigenerational households and intergenerational transfers, and issues related to pension administration for older people who have worked in the informal sector for most or all of their lives. The chapter documents what we know about these various aspects of the issue and identifies knowledge gaps. Based on the evidence we do have, we indicate policy reform directions, in particular regarding expansion—or for some countries—introduction of social pensions directed to older people who have worked in the informal sector.

13.1

Introduction

This century has been characterized as the Asian Century, but even more, it is the Ageing Century. Population ageing is a global phenomenon, exerting unprecedented pressures on long-established social norms and policy institutions around the world. The rapid demographic transition in many Asian countries presents a special challenge. These societies will 234

DOI: 10.4324/9781003150398-15

Pension Policy in Emerging Asian Economies with Population Ageing

get old before economic growth delivers affluence, complicating the further development of strong social (income, health, and long-term care) support structures. Typically, much of the labor force operates informally. Further, regional and within-country migration, leading to separation of families across generations, is combined with declining fertility to reduce effective intergenerational family support, especially family-based care of older people, which has for so long been a traditional part of the Asian way of life. While some countries within the Asian region have developed policy structures designed to address, at least in principle, some of these challenges, their scalability, efficacy, and sustainability are in most cases yet to be proven. This chapter focuses on retirement income policy in emerging economies of East Asia and Southeast Asia (EA and SEA), which face these shared challenges. We argue that an effective, affordable, and implementable strategy in developing retirement income policy in these circumstances is to establish, or further develop, social pensions. We define social pensions to be noncontributory income payments available to eligible residents who have reached some threshold age, which we call the access age. Social pensions are therefore to be distinguished from more general welfare payments available to those in extreme poverty. The social pension may be payable to all who have reached the access age, or may be means-tested, either through valuing resources available to potential recipients, or through the application of some other criterion, such as eligibility for a contributory pension. Some form of such programs already exist in several of the countries under consideration, but even in those countries, significant needs remain for further policy development related to coverage and benefit adequacy. This policy design, appropriately parameterized, efficiently reduces poverty among the elderly, many of whom will have exhausted their earning capacity. Because social pensions target older cohorts, benefits can affordably be higher than general welfare payments and have limited labor market disincentives. Simulations, both drawn from the literature and presented here, suggest that the costs of providing a social pension are manageable even for countries with low per capita incomes. We begin in the next section with documenting common economic and demographic forces at play in EA and SEA countries, reporting on their rapid population ageing and informality in the context of both the labor force and the overall population. The following section then provides an overview of retirement income policy initiatives, referring to the common three-pillar pension policy design and emphasizing the diversity across the region, issues of administration, and evidence on household-level impacts of pension policies. The section after that examines the main design issues that EA and SEA countries could address to enhance the welfare impact and sustainability of existing social pension programs or when introducing such programs. This includes a numerical example costing the introduction of social pensions calibrated to Indonesia. The last section summarizes our conclusions.

13.2

Ageing and Informality

In this section, we consider the interaction of population ageing and informality in the main emerging economies of EA and SEA, which include the world’s most populous country, China, and the fourth most populous country, Indonesia. We begin by discussing rapid population ageing in the two Asian subregions. We also provide comparative demographic data for developed regions (Europe, Northern America, Australia/New Zealand, and Japan). The demographic data are mainly based on United Nations (2019) population estimates and projections. We then report on employment informality and informality at older ages. Household data derive from 235

George Kudrna et al.

major reports on EA and SEA regions by the World Bank (e.g., World Bank, 2016) and other international institutions (e.g., ILO, 2018).

13.2.1

Demographic Change and Population Ageing

Population ageing in emerging EA and particularly many SEA economies is at an earlier demographic stage than ageing societies in the developed world. But the speed of demographic transition is much faster. Emerging EA and SEA economies are projected to face very pronounced population ageing over the next 80 years, driven by substantial declines in fertility rates and projected improvements in survival rates and longevity. This demographic change will result in rapidly growing elderly populations and large declines in population growth. We first document the past and projected changes in the key demographic drivers: fertility and life expectancy implied by mortality rates. We then present and discuss the demographic estimates and trajectories for the old-age dependency ratio and the annual growth rate of the total population.

13.2.1.1

Demographic Drivers

Table 13.1 shows the past, current, and future total fertility rate and implied life expectancy at birth (for both sexes) in the main emerging EA and SEA economies and compares these with developed regions. The projected rates are based on the medium fertility variant of the United Nations, representing the most used future scenario. Total fertility rates (TFRs) (i.e., live births per woman of reproductive age) have undergone pronounced declines in the selected emerging Asian countries in the past 60 years. For example, in China, the TFR has declined from 5.5 births per woman in 1955–1960 (with the peak of almost 6.3 births in 1965–1970) to currently about 1.7.1 Similar developments were seen in SEA. For instance, the TFR in Indonesia and Vietnam declined from 5.5 to 6.2 in 1955–1960 to current (2015–2020) TFRs of 2.3 and 2.1, respectively. Notice that these declines are much more pronounced than those observed in developed regions, where the average TFR was 2.8 births per woman in 1955–1960 and current TFRs are Table 13.1 Demographic drivers and population ageing in EA and SEA

Notes: a Five-year average ending in displayed year; b births per woman of reproductive age; c both sexes (in years); d ratio of population aged 65+ to population aged 20–64 (in %); e annual growth rate (in %); f only selected “emerging” economies are considered here; g Europe, Northern America, Australia/New Zealand, and Japan. Source: United Nations (2019).

236

Pension Policy in Emerging Asian Economies with Population Ageing

not that different from those in the displayed Asian economies. Furthermore, many emerging SEA economies are projected to undergo further declines in the TFR over the course of this century. For instance, the Indonesian TFR is projected to further decline to 1.8 by 2100. Table 13.1 shows that sustained improvements in mortality/survival rates have led to greater longevity. In China, the average life expectancy at birth (for both sexes) has increased by more than 30 years in the last 60 years, from less than 45 years of expected life in 1955–1960 to more than 75 years now.2 Compared with more developed regions, emerging Asian economies have lower life expectancies, but they have experienced more significant improvements in the past. Life expectancy at birth has increased by more than 25 years in Indonesia and by about 18 years in Vietnam in the last 60 years, compared with 11.5 years in developed regions over the same period. To a lesser extent, migration has impacted and will impact the demographic transition in the EA and SEA regions. While developed countries have experienced and are projected to experience positive net (international) immigration rates, the opposite holds for many emerging Asian economies facing net migration outflows. For example, in Vietnam, the average net migration rate, defined as immigrants less emigrants per 100,000 of population, was around -1 percent, indicating almost 1 million net emigrants (United Nations, 2019). Furthermore, as Chomik and Piggott (2013) report, EA and SEA countries are the source of 25 million migrants, with high levels of international labor migration that have grown 6 percent per year and with migration flows expected to intensify due to increased skilled migration.3

13.2.1.2

Population Ageing

To demonstrate the rapid ageing of EA and SEA populations, Table 13.1 reports estimates and projections for the old-age dependency ratio and the annual growth rate of the total population. We define the old-age dependency ratio here as the percentage of the population aged 65+ relative to the working-age population aged 20–64. Note that the inverse of the age-dependency ratio is the potential support ratio, which indicates how many people of working age there are to “potentially” support the elderly. In the emerging EA and SEA economies, most of the elderly population works and relies on own earnings at older age, with only limited public support in most countries and declining private transfers for older people. Table 13.1 shows that in China, the old-age dependency ratio has more than doubled over the last 20 years, increasing to the current 18.5 percent, and it is expected to further increase by a multiple of 3 to more than 64 percent by 2100. In fact, the old-age dependency ratio in China is projected to be higher in 2100 than the average ratio among developed regions. The displayed SEA economies have experienced much smaller increases in the old-age dependency ratio to date than China has, but they will face much larger increases over the course of this century. For instance, in Indonesia, the old-age dependency ratio increased only slightly in the last 60 years to slightly more than 10 percent currently, but the ratio is projected to increase by a multiple of 5 by 2100. These dramatic changes in the old-age dependency ratio are due mainly to projected increases in elderly populations but also due to declines in (growth rates of) working-age populations. Currently about 12 percent of the total Chinese population is aged 65+, but the elderly population share is expected to reach almost 32 percent in 2100 (United Nations, 2019). The increases in elderly population shares that are projected for Indonesia and Vietnam are even more pronounced, e.g., in Vietnam, increasing from around 8 percent now to almost 30 percent in 2100. 237

George Kudrna et al.

Declining population size will accompany population ageing, as changes in the population growth rates in Table 13.1 show. Although these populous regions have experienced much higher growth of their populations in the past (compared with developed regions) driven by very high TFRs, the future population growth rates will become negative in future years. The total population in China will start declining after 2030, reaching a negative growth rate of more than 0.5 percent per year by 2100. In Indonesia and Vietnam, populations are projected to start declining in the second half of this century. Such pronounced demographic changes over the course of this century will have vast economy-wide implications, directly affecting the lives of more than 30 percent of the world population residing in EA and SEA countries. In emerging Asian economies and particularly in SEA, these demographic changes are occurring simultaneously with other challenges. These include (1) high informality, with both large informal employment and informal social support via co-residency and private support; (2) large regional labor migration from the countryside to towns and cities; and (3) undeveloped formal social security and government support, particularly for the elderly operating outside the formal sector. In the following, we focus on informal employment and the elderly population, with public policy initiatives discussed in detail in the section “Pension Policy in Emerging EA and SEA Economies.”

13.2.2

Informality and the Elderly Population

We now document informality and regional migration in emerging EA and SEA economies. We begin with the labor force and working-age population, reporting on developments in labor force composition and labor incomes and accounting for rural-urban differences. We then focus on older people, especially their living arrangements and income sources.

13.2.2.1

Informal Employment

A common feature across emerging Asian economies is very high informal employment, defined by the International Labour Organization in terms of the employment relationship and protections associated with the job of the worker.4 Drawing on ILO (2018) data, Table 13.2 reports on the share of informal employment in total employment (including agriculture) in selected EA and SEA countries and provides the decomposition into informal workers operating in the informal sector, in the formal sector, and in households (i.e., domestic workers producing only for Table 13.2 Share in percentage of informal employment in total employment (including agriculture)

Notes: a Developed countries with annual per capita income of $12,236 or more. Source: ILO (2018).

238

Pension Policy in Emerging Asian Economies with Population Ageing

final use). As shown, total informal employment in all selected countries is more than 50 percent, and in some countries more than 90 percent. In China, informal employment is 54 percent, with only 5.1 percent of informal employment in the formal sector (i.e., workers not covered by any protection policies). In Indonesia, total informal employment is more than 85 percent, with 67.5 percent operating in the informal sector, 5.8 percent in the formal sector, and 12.2 percent in households. ILO (2018) also shows that informal employment in EA and SEA differ significantly by region (rural vs. urban), by education, and over the life cycle. In China, rural informal employment is much higher at 82 percent, compared with urban informal employment at 36 percent. In Indonesia, the difference is much smaller, with rural informal employment at 91.3 percent and high urban informal employment at 80.3 percent. As expected, informal employment is negatively correlated with educational attainment. The decline in informal employment for those with higher education is much more pronounced in SEA than in developed countries. The life-cycle profile of informal employment is U-shaped, being very high in young populations (15–24) and older populations (55+), exceeding 90 percent of total employment for those age groups in some Asian countries. Those workers classified as in informal employment also work significantly more hours in the emerging economies of Asia and the Pacific, with 40 percent reporting working 48+ hours per week, compared with 15 percent across developed countries. Although those in informal employment work more hours, their overall labor earnings are significantly smaller than those in formal employment. Using the Indonesian Family Life Survey, Kudrna et al. (2020) find that the average earnings gap between those in formal and informal employment is almost 135 percent and that the earnings gap widens with age, exceeding 200 percent for 55+ cohorts. Comparing Indonesian Family Life Survey 2007 and 2014 waves, Kudrna et al. (2020) show that the movers from the informal sector have higher earnings than the stayers in 2014, with the gap between earnings increasing initially but declining at older ages. In addition to being very high, informal employment in most of emerging EA and SEA has been remarkably persistent over time, with the partial exception of China, and this is especially so in SEA. Hence, high informal employment will likely remain a key feature of most Asian economies for some decades into the future, even as they grow and experience pronounced population ageing.

13.2.2.2

Older People

We now focus on older people and their income sources and living arrangements. Figure 13.1, from World Bank (2016), depicts the sources of income at older age in various EA and SEA countries. It reveals that the main source of income at older age is own labor income and that the reliance on labor income is higher in rural areas. In contrast, public transfers represent a minor share of income at older age in most Asian countries, although heterogeneity is significant in the reliance on public transfers/pensions. Only in urban China and Mongolia are public transfers the main source of elderly income. Figure 13.1 indicates that private transfers represent the second most important source of elderly income in rural areas in most Asian emerging economies. Furthermore, the relative importance of private transfers in the total income of older people increases with age. Kudrna et al. (2020) find that in Indonesia for people aged 50–59, private transfers account for less than 20 percent of total income, but those aged 70+ derive more than 50 percent of their income from private transfers. While most of the old-age population does not receive a public pension, more than 70 percent of those aged 70+ receive positive net money transfers from their children. 239

George Kudrna et al.

Figure 13.1 Income sources for people aged 60–85 in EA and SEA. Source: World Bank (2016).

Not only financial private transfers but also living arrangements with high elderly coresidence rates confirm the importance of family for old-age support in emerging EA and SEA economies. According to World Bank (2016), for people aged 60+, co-residence rates range from 25 percent to more than 80 percent across emerging Asian countries. They are particularly high in lower-income countries and for women at very old age. For example, Kudrna et al. (2020) show that in Indonesia most older people live with their children and that share (of those with no spouse but with a child or children) increases significantly with age, especially among women. Elderly co-residency rates have declined over time, most notably in China and to a lesser extent Thailand, but overall rates are high by global standards for middle-income regions (Palacios and Evans, 2015). While the starting point on elderly poverty relative to younger cohorts varies across EA/SEA countries, population ageing with fewer adult children to provide within-family financial support for older people and larger elderly populations living longer lives will challenge social cohesion in emerging Asian economies, many of which will experience increasing hardship and inequality without government policy interventions and public transfers, particularly among older people.

13.3

Pension Policy

In this section, we first introduce a simple schema to characterize and distill the various features of retirement policies. The focus of this section is then on pension policies in emerging EA and SEA countries, emphasizing both their common features and their diversity. 240

Pension Policy in Emerging Asian Economies with Population Ageing

Figure 13.2 Multi-pillar retirement income designs. Source: Authors’ compilation.

13.3.1

Multi-Pillar Pension Taxonomy

Each country has a distinctive retirement income system. They are best systematized in terms of the three-pillar system (Bateman et al., 2001; OECD, 2017), as depicted in Figure 13.2. First-pillar benefits are unrelated to an individual’s past earnings and are sometimes referred to as noncontributory or social pensions. The payment, financed out of general revenue, may be universal or targeted. These structures serve as a safety net for those without other savings and aim to reduce poverty among the elderly. From a social welfare perspective, social pensions are perhaps the most important of the pillars, although that they be seen as one of several complementary pillars is important. If, for example, we evaluate retirement income design by appealing to the United Nations Sustainable Development Goals (SDGs) to be achieved by 2030, we find they envision social protection with substantial coverage of the poor and vulnerable that insures against economic shocks and addresses inequality (SDG1 and SDG10)5 (United Nations, 2015). As directed toward older cohorts, this is the key role of a social pension. By contrast, the aim of the second pillar is to provide income replacement, partly replacing income enjoyed pre-retirement. The benefits are typically related to a person’s salary, based either on salary profile or on contributions into a pension fund whose value is geared to salary, usually up to some ceiling. Second-pillar benefits can be unfunded, where the individual has a claim on the future revenues of employers, insurers, or governments, or prefunded with an underlying accumulation converted into a lump sum or income stream. Such schemes can be administered through public social security programs or through private pension funds, at arm’s length from government. The reason for mandating saving in a second pillar is twofold. First, without it, some people with the capacity to save will nevertheless choose to rely (or “free-ride”) on the first pillar. Second, mandatory saving acts as a commitment device. Many people may want and can afford to save a portion of their income for retirement but do not get around to it, displaying myopic behavior (Mitchell and Piggott, 2016). There is also often a third pillar, involving incentives for further voluntary saving, for people who want retirement incomes beyond the mandated level, though this is marginal currently in emerging EA and SEA. 241

George Kudrna et al.

13.3.2

Pension Policy in Emerging EA and SEA Economies

Pension systems in emerging EA and SEA countries combine considerable diversity in design and instrument mix with several common features and challenges. Table 13.3 shows the key elements of the pension systems of EA and SEA countries. Most countries in the region rely on contributory pension schemes. However, progress in expanding coverage of contributory pension schemes has been slow and often stalled at low levels, primarily due to high labor market informality and demand-side issues among workers and employers (Chomik and Piggott, 2015). At the risk of oversimplification, the pattern in contributory schemes is one where unfunded defined benefit (DB) plans exhibit limited coverage and frequently fiscal sustainability challenges, while prefunded defined contribution (DC) plans offer very low benefits.6 This has in turn led to diversified approaches to providing old-age financial protection, with a proliferation of noncontributory social pension schemes in EA and some SEA countries and, in a subset of countries, matching defined contribution (MDC) programs targeting informal workers. Nonetheless, major challenges of financial protection remain for those outside mandated contributory schemes, in the case of social pensions due to low benefit levels (and in some cases, tight targeting or absence of a program in some SEA countries), and in the case of MDCs due to limited take-up. This section examines past developments in and the current state of play with pension systems in emerging EA and SEA economies in terms of coverage, adequacy, and sustainability.7 We look first at noncontributory social pensions, then at matching schemes for informal workers, and finally at contributory schemes for the formal sector.

Table 13.3 Pension policy in emerging EA and SEA countries

Notes: a PAYG stands for pay as you go; b Malaysia PAYG defined benefit scheme only for civil servants; c Thailand has two separate matching defined contribution schemes, and a defined contribution pillar mandated only for public sector; d Vietnam social pension, targeted 65–79 and pension-tested from 80. Source: World Bank (2016); OECD (2018).

242

Pension Policy in Emerging Asian Economies with Population Ageing

13.3.2.1

Social Pension Programs

Social pension programs have proliferated in the past two decades, and they are now part of the pension systems in much of emerging EA and parts of SEA (Palacios and Knox-Vydmanov, 2014). But a significant unfinished expansion agenda remains even for those with programs in place. For all EA/SEA countries with social pension programs, adequacy remains a major concern, and for a significant subset an additional challenge concerns covering all elderly without formal sector pensions. Table 13.4 provides key features of the schemes, revealing variations and commonalities in design.8 The most significant variation is coverage, which ranges from 82 percent of the relevant age cohort in Thailand, through around 70 percent for those 65+ in Korea and 50 percent of those 60+ in Philippines, to 13.5 percent of those 60+ in Vietnam and even lower in Myanmar. Cambodia and Lao PDR have no elderly social pension at all, while Indonesia covers only 0.1 percent of the 60+ population with its social pension and a further 4 percent with an elderly window of a targeted cash transfer program (though the elderly must live with school-age children to qualify for the latter program). The approach to eligibility and targeting of the social pension primarily drive coverage differences. First, the programs exhibit variation in coverage due to differences in eligibility age thresholds. While most EA/SEA economies have an eligibility age of 60, it is as high as 85 in Myanmar and 80 for most beneficiaries in Vietnam. Apart from variations in eligibility age for social pensions, coverage differences are also due to the strictness of means-based targeting, which also exhibits variation (targeting methods for social pensions will be discussed further in the following). Despite the variation in the coverage of social pensions across EA/SEA, in all countries, the generosity of benefits and public spending on the schemes are very modest. No country has an elderly social pension benefit valued at more than 10 percent of per capita gross domestic product (GDP), and the benefit relative to overall living standards is much lower in China, Philippines, and Thailand. All are well below levels seen in most Organisation for Economic Cooperation and Development (OECD) countries and other developing countries such as Brazil, South Africa, and most Eastern European countries.9 In general, no explicit indexation rules are present in the social pension programs of emerging EA and SEA countries, with adjustments in benefit levels largely ad hoc and discretionary. Even in EA/SEA countries with higher coverage of social pensions, public spending remains modest, with Thailand easily the highest at 0.4 percent of GDP. All social pension programs in EA/SEA countries are general revenue funded. Not surprisingly, given the low levels of benefit, the limited available studies for developing Asia on the impact of social pensions on elderly poverty find very modest reductions in poverty (Badiani-Magnusson, 2016; Giles and Huang, 2016, for Thailand; Zhang et al., 2014, for China; HelpAge International, 2018, for Philippines). Thailand provides an interesting case with respect to social pensions.10 A tightly means-tested social pension introduced in 1993 covered around 400,000 elderly in the first half of the 2000s. Subsequently, targeting was decentralized and loosened in 2005 to include around 1.8 million elderly persons before the program became quasi-universal from 2009, subject only to a formal sector pensions test for those 60+ and currently covering more than 80 percent of the elderly cohort. Just as importantly, coverage of the bottom quintile increased more than three times from 2006 to 2010, from just under 30 percent in the initial year (Badiani-Magnusson, 2016). The basic flat benefit level roughly doubled in real terms from 2000 to the mid-2010s, and an increase in the benefit level for each decade of age was introduced (see Table 13.4).

243

George Kudrna et al. Table 13.4 Key indicators for social pensions in emerging EA and SEA countries

Notes: GDP = gross domestic product; IDR = Indonesian rupiah; KRW = Korean won; MMK = Myanmar kyat; MYR = Malaysian ringgit; PHP = Philippines peso; RMB = renminbi; THB = Thai baht; VND = Vietnamese dong. a China’s informal sector pension scheme is a hybrid design, with a basic flat benefit financed from general revenues and paid after age 60 as in standard social pensions, but also requires a period of 15 years of flat contributions to trigger the basic benefit entitlement. The accumulated contributions in the individual account are paid out on top of the basic benefit, but account for only around 15 percent of the total benefit stream after 60. Source: World Bank staff, government documents, and budgetary sources; HelpAge International (2018); OECD (2019).

244

Pension Policy in Emerging Asian Economies with Population Ageing

13.3.2.2

Implementation

A key factor in terms of potential coverage and impact of social pensions is the ability of countries to reliably identify, target, and pay eligible beneficiaries. While these continue to need improvement, advances in technology, public administration, and financial services across the region have meant that the primary barriers to social pension coverage expansion are increasingly less likely to be technical ones.11 First, with respect to identification of potential beneficiaries, progress in EA/SEA has been considerable in recent years (Gelb and Diofasi Metz, 2018; O‘Keefe et al., forthcoming). Historically, documentation to establish identity and age was scarce, as coverage of birth certificates or equivalent documentation was at best partial, especially for older people. That situation has changed sharply with the spread of national ID systems, which provide more robust proof of identity and age. Countries such as Indonesia and Thailand already have foundational biometric national IDs that uniquely identify people and basic details such as age and gender, and these facilitate application for benefits. Philippines is rolling out an open-source biometric ID presently, and Vietnam is accelerating roll-out of its national biometric ID and had covered around 58 million of an eligible population over 14 years old of 72 million by late 2021. As of 2018, the shares of population without some form of reliable identity document had fallen rapidly, with full coverage in Thailand and almost complete coverage in China (98 percent), Vietnam (96 percent), and Indonesia (92 percent), and high coverage in Philippines (85 percent) and Cambodia (84 percent) (Identitification for Development database, 2018, World Bank). Where no official document is available to verify age when registering older citizens for national IDs, countries rely on different combinations of less formal documents (e.g., baptismal certificates or even horoscopes) and/or some form of community validation of approximate age. With respect to benefit payment, bank accounts and even mobile-based payment have also become increasingly common in the region, especially for those receiving public transfers. Overall, the rate of bank account ownership was mixed across EA/SEA as of 2017, with high rates in countries such as China (80 percent), Malaysia (85 percent), and Thailand (82 percent), but very incomplete coverage in Indonesia (49 percent), Vietnam (31 percent), and Philippines (34 percent) (Global Findex Database, 2017, https://globalfindex.worldbank.org). Despite the mixed picture for the overall population, coverage of accounts among recipients of social transfers is increasingly high, with transfers acting as a key driver of financial inclusion. The overwhelming majority of cash transfer beneficiaries across the region now receive payments through bank accounts and in some cases mobile money accounts, with the notable exception of Vietnam (and the failure to pay all beneficiaries digitally there is not a technical barrier but a policy choice). Increasingly EA/SEA countries sign up unbanked transfer beneficiaries for bank accounts at the enrollment stage. Indonesia is a case in point, where poor households receiving targeted social transfers into bank accounts rose from only around one-fifth in mid-2001 to more than 90 percent by 2019, acting as a key driver of financial inclusion for the poor. The COVID-19 pandemic has accelerated these trends, with, for example, Thailand enrolling and paying emergency benefits to around 22 million informal workers through mobile-based apps and digital bank accounts. A key facilitator of this trend in most EA/SEA countries has been the spread of e-KYC (electronic know your customer), which has made signing up for bank accounts simpler. While challenges remain in remote areas with getting to bank branches, increased reliance on banking agents and expanded use of mobile money are helping address these last-mile problems.

245

George Kudrna et al.

The biggest implementation challenge remains targeting in cases where the program is not universal or has simple and inclusive targeting rules. Social pensions in emerging EA/SEA countries use different targeting approaches. Myanmar is fully universal past the eligibility age, reflecting limited capacity to target based on means. Thailand and Vietnam for the 80+ population use pension testing, whereby only those receiving formal sector pensions are excluded from social pension receipt. Countries such as Malaysia, Vietnam for those under 80, and Indonesia in its small program use tighter means-testing and leverage the general safety net targeting system. For targeted social assistance, proxy-means testing is used in targeted social assistance programs in Indonesia, Philippines, Vietnam, and Myanmar.12 While resulting in progressive targeting, significant challenges with exclusion errors of around 50 percent of the eligible poor population in Indonesia and Philippines remain (O‘Keefe et al., forthcoming). This suggests that a simpler targeting approach may be appropriate for social pensions, especially given the lesser concerns about work disincentives.

13.3.2.3

Part-Contributory and Matching Pension Programs

Distinct from formal sector contributory schemes and social pensions, some EA/SEA countries have introduced pension schemes that require a low (often flat) contribution from informal workers that is matched with a government subsidy. These MDC schemes are voluntary schemes not reliant on an employer relationship, which aim to incentivize participation of informal workers through a publicly financed contribution match that varies in generosity and structure. MDCs have been introduced in Korea, Vietnam, Thailand, and Malaysia.13 The impact on take-up of MDC schemes in EA and SEA has largely been modest or even marginal, adding only around 1.5 percent to membership in the Malaysian provident fund and around 1.5 percent of the working-age population in Vietnam. Thailand has achieved somewhat more success with two MDC schemes adding around 8 percent of the working-age population, though some double counting may occur across schemes. The notable exception is Korea, with its MDC scheme for “farmers and fishermen” more than doubling coverage between 1995 and 1999 with a generous match. The modest impact on participation of informal workers in most of EA/SEA is consistent with global experience in developing and developed countries, where MDCs have typically tended to have modest take-up and play at best a supplementary role in the broader pension architecture.14 The global experience suggests a potential supplementary role for MDCs with attractive matching rates, simple design, and accessible administrative arrangements. However, it is important not to expect too much from MDCs, which are unlikely to be the “silver bullet” that adequately addresses the coverage gaps for informal workers in contributory systems. The most innovative case of these programs is China, which from 2009 rolled out a unique hybrid pension scheme for rural and later urban informal workers, which requires a modest annual contribution to an individual account from any income source, and after 15 years of contributions triggers payment of a flat basic monthly pension at age 60 that is funded from general revenues, plus the proceeds of the individual account (see Dorfman et al., 2013). In principle, the local government level also provides a matching contribution to the individual account during working life at the rate of 30 percent of the individual contribution, though how consistently this occurs in practice is not well understood. Individual contributions are low and flat but tiered (as low as 100 RMB per year, though with considerable variation across provinces in terms of minimums and maximums). The basic benefit level is modest, with a national minimum benefit of 93 RMB per month past age 60 as of mid-2021, but with some provinces topping up to as much as 1,100 RMB monthly in Shanghai. The central authorities finance the basic benefit entirely in western and central provinces and half in coastal ones. The 246

Pension Policy in Emerging Asian Economies with Population Ageing

design results in the total flow of benefits to a typical participant being 80–85 percent public subsidy. The scheme is voluntary, but local authorities apply considerable suasion to participate, resulting in high coverage among informal workers. The contributory element of the scheme can perhaps more usefully be characterized as a membership fee of sorts during working life, with the policy motivation (in contrast to a pure social pension) of building a record-keeping base on informal workers that may facilitate migration to formal sector schemes at a later point.

13.3.2.4

Contributory Earnings-Related Pension Programs

Contributory pension programs continue to make up the major component of the pension system in much of emerging EA and SEA in expenditure terms, but considerable diversity exists in the policy framework of contributory pension programs in the region. A first source of diversity is between countries that rely primarily on DB schemes and those that rely primarily on DC schemes. Countries that were formerly under British rule tend to have DC/provident funds as the primary approach. In contrast, others tend to have DB schemes as the bedrock of their systems (e.g., Vietnam, Thailand, Philippines, Cambodia, and Lao PDR) or hybrid DB/DC (China). For both types of schemes, coverage beyond the formal sector is a continuing challenge, with adequacy a key additional challenge for DC schemes and sustainability for several DB schemes. The combined employer/employee contribution rates also range in both DB and DC systems, from as low as 6 and 8 percent in Thailand and Korea, respectively, to as high as the mid-20 percent range in China, Vietnam, and Malaysia (Social Security Administration, 2019).15 An additional source of diversity in contributory systems is between the bulk of Asian countries with parallel (and more generous) schemes for civil servants and private sector workers, and a small number with integrated (e.g., Mongolia, Singapore, and transition underway in China) or partly integrated (Vietnam) schemes. As already noted, across emerging Asia, voluntary occupational schemes are either largely absent in practice, or provide modest coverage (e.g., China).

13.3.2.5

Policy Assessment

Coverage of contributory schemes in much of emerging EA and SEA currently sits around or below average global coverage by country income level. However, coverage rates have seen only modest gains in recent decades, with limited expansion beyond the formal sector. In effect, demographics are winning the “race” against expansion of contributory schemes. A few standouts exist in emerging EA and SEA, with Malaysia and Mongolia notably above global coverage averages by income level, and a group of countries that cluster around the global average, including China, Vietnam, Korea, and Philippines. In contrast, in lower-income Association of Southeast Asian Nations countries, participation is in single digits. With few exceptions, a strong relationship exists between higher informal and rural worker shares and lower coverage. From a gender perspective, countries vary, with female pension receipt rates low relative to those of men in Indonesia, Korea, and Malaysia, but closer in China, Vietnam, and Myanmar, mirroring the share of women in wage employment relative to men (Chlon-Dominczak, 2015; Majoka and Palacios, 2018). With respect to adequacy of benefits, for DB schemes, considerable variation exists in target replacement rates across (and within) emerging EA and SEA countries, ranging from around 40 percent in Indonesia, Philippines, and Mongolia to as high as 75 percent in Vietnam for private sector workers and 80–90 percent of civil servants in countries such as China and Philippines.16 Overall, DB schemes in developing Asia have reasonable replacement rates for those with full 247

George Kudrna et al.

work histories, but adequacy challenges remain in immature schemes for the initial generation of workers with partial contribution histories (e.g., Thailand, Korea, and Vietnam). The substantial gap between the generosity of civil service and private sector pensions seems hard to justify in terms of equity, fiscal implications, and administrative and labor market efficiency and is increasingly out of line with global trends.17 The small number of EA/SEA countries with DC schemes face adequacy challenges due to factors including generous early withdrawal rules, low contribution density (especially for women), and low withdrawal ages in some cases. In Asia, countries with DC schemes avoid explicit government return guarantees over the long term. As for eligibility, the retirement age is a key factor. For some emerging EA and SEA countries, retirement or eligibility ages are low relative to life expectancy of older workers. This includes China, Vietnam, Thailand, and Indonesia. Low official retirement ages are exacerbated in some by early retirement provisions that lower the median age at retirement in contributory schemes by 2–3 years. In socialist countries and Mongolia, there are also 5-year differences in female and male retirement ages. Several countries have implemented or initiated gradual increases in retirement ages, including Vietnam, Indonesia, and Malaysia, and China has indicated that retirement ages are likely to rise in the 14th Five-Year Plan period. A second set of countries, including Korea, Philippines, and Singapore, has retirement ages more in line with OECD averages. A final crucial factor to consider with pension systems is financial sustainability. The challenge for DB schemes differs sharply from DC schemes. The latter are inherently financially sustainable, absent a guaranteed rate of return or some form of minimum benefit, which does not happen in emerging EA/SEA. For DB schemes, in contrast, parameter design is crucial to sustainability. In this respect, diversity exists between and within national schemes. One consistent feature is that many of the standalone civil servant schemes are not internally sustainable at current levels. Private sector schemes vary considerably, ranging from schemes that are already unsustainable (such as that in Mongolia, which has an annual pension deficit of around 2.5 percent of GDP and rising) or will be soon if not reformed further (such as that in Vietnam, which enters a cash flow deficit this decade), to a second group with significant fiscal pressures likely to emerge over the longer run absent deeper reforms (including Thailand, China, and Philippines), to less mature schemes that do not present imminent fiscal pressures (such as the poorer Association of Southeast Asian Nations countries).

13.3.2.6

Welfare and Behavioral Impacts of Pensions

Before turning to policy implications for retirement income policies, examining the impact of receipt of distinct types of pensions on the welfare and work behavior of older people is useful. Here, the literature for emerging EA and SEA is limited but instructive. With respect to formal contributory schemes, the impact of pension receipt on the likelihood of stopping work is clear in countries for which multivariate analysis has been done. In China, especially urban China, the likelihood of withdrawal from the labor force when in receipt of a contributory pension is high (43 percent and 38 percent reductions in probability of working for urban men and women, respectively). This is also the case in Indonesia, with 19 percent and 25 percent reductions in probability of work for urban and rural men, respectively (O‘Keefe et al., 2021).18 In contrast, the impact of receipt of the hybrid social pension is minimal in China (Zhang et al., 2014). In Thailand, the (more generous) social pension receipt has some impact on work probability (Badiani-Magnusson, 2016; Huang and Giles, 2017). 248

Pension Policy in Emerging Asian Economies with Population Ageing

In terms of household poverty and consumption/income impacts, in China, Indonesia, Thailand, and Vietnam, contributory pension receipt is associated with significant poverty reduction in rural and urban areas, with reductions in income poverty greater in rural areas of all countries (Giles and Huang, 2016). For social pensions, various authors find statistically significant poverty reduction impacts for rural elderly from the Chinese contributory social pension (Zhang et al., 2014, 2020), and positive impacts on health and household income and food expenditure (Huang and Zhang, 2021), while for Thailand no significant poverty impacts have been found from the social pension (Huang and Giles, 2017; Badiani-Magnusson, 2016). The spillover impacts of noncontributory pension benefits are not typically considered in a developed country context but can be important in emerging economies.19 Intergenerational family transfers, facilitated for most EA/SEA countries by high co-residency across generations, mean that benefits paid to one generation may benefit others. This positive spillover has been documented in several developing countries. For EA/SEA, the literature is limited, but in China, for example, the association between social pension receipt and the health status of children up to 15 years of age is significant. The association is larger for children who are boys, “left behind” by migrant parents, and in poor health, with improved nutritional intake being the major channel underlying the impacts (Zheng et al., 2020). Other authors also find positive effects among children aged below 15, with the social pension leading to more pocket money received, more caring from grandparents, improved health, and higher schooling rates among children under 15 (Huang and Zhang, 2021). Analysis for Thailand finds positive impacts on educational choice and child labor from receipt of the social pension (Herrmann et al., 2021). These findings echo those in other developing countries, where receipt of elderly social pensions has been found to have positive effects on other household members, including improvements in child health and education in countries such as Kenya, Brazil, South Africa, and Uganda; reductions in child labor in Brazil; and positive effects on job search in South Africa (Ardlington et al., 2009; Moscona and Seck, 2021; De Carvalho Filho, 2012). A final important feature of the EA/SEA regional context to consider before looking at policy directions in the next section is the fiscal dimension, both in terms of overall government revenue performance and in terms of public expenditure on social protection. The East Asia and Pacific region has low public revenues on average relative to other developing regions, with average revenue below 20 percent of GDP in 2019 (IMF, 2019). There is also considerable variation, with China having total public revenues closer to 30 percent of GDP, but Indonesia, for example, collecting closer to 15 percent of GDP in public revenues. In addition, the average allocation to social assistance programs (including social pensions) is low by global standards, with all the countries considered in this discussion allocating less than 1 percent of GDP for total social assistance transfers in 2019 (O‘Keefe et al., forthcoming; World Bank, 2018). Public pension expenditure in most developing EA/SEA countries is also relatively low by global standards relative to their per capita income levels. For example, in Indonesia, public pension expenditure was less than 1 percent of GDP in 2015 (OECD, 2017).

13.4

Policy Directions—Expanding Social Pensions

As documented in the previous two sections, the emerging EA and SEA nations present a diverse range of economic and policy structures. At the same time, however, they share some critical salient characteristics: rapid growth, rapid demographic transition, persistent high informal labor force participation, and major rural-to-urban migration. In particular, their economywide retirement-oriented social protection structures are mostly immature, with low coverage and/or very low benefits. These combine to suggest an economic rationale for developing 249

George Kudrna et al.

social protection structures that will allow parametric adjustment, increasing sophistication, and improved coverage and adequacy with economic development. Perhaps most importantly, these structures need to be designed to be resilient in the face of future economic growth and shared prosperity. The various elements of retirement policy will need to operate in a coordinated and complementary fashion, encouraging rather than inhibiting development, while at the same time providing adequate financial relief to those older people who need it. We argue here that while contributory structures will continue to have their place, they will not address the most important social imperatives presented by persistent informality for older cohorts in these countries. Further, if they are used as the primary vehicle for retirement income policy, a genuine risk exists that the mistakes of the developed world in this space will be needlessly repeated in these countries. With demographic transition, such mistakes tend to become much more expensive over time. Instead, we join others (e.g., World Bank, 2019; Chomik et al., 2019; Kudrna and Piggott, 2019) in advocating for an expansion of noncontributory (social) pension entitlements. That is, we emphasize first-pillar structures that are not tied to employment history. Historically, these have had limited application in developing countries, but that situation has changed in more recent years; around 90 developing countries now have some elderly social pension as of 2018 (HelpAge International, 2018). Typically, however, benefits are still set at very low levels, and coverage is patchy. In what follows, we will assume an earnings-related retirement policy of some kind remains in place. This assumption admits a dynamic of economic growth that would see a contributory pension scheme designed to address income replacement goals gradually increase in importance and relieve fiscal pressure on the social pension. In an ideal world, this income replacement pillar would be prefunded, to avoid long-term sustainability issues that declining fertility and labor force may generate. But we will not consider the design of this pillar in any detail.

13.4.1

Policy Objectives and Economic Analysis

Policy assessment requires appeal to some set of criteria. In the present developing country context, the most natural include (1) equity, by which we mean (mainly) that the poorest, especially those whose earning capacity has been exhausted, are covered and adequately supported; (2) economic efficiency, or the best possible allocation of resources, including especially the supply of labor and education choices and saving; and (3) sustainability, by which we mean the capacity of the policy to be maintained as demographics and economic circumstances change. These criteria suggest that, whatever their detailed design, noncontributory pensions enjoy several advantages over their contributory counterparts, especially in highly informal economies. They can be effectively targeted to those retirees most in need, not just those who have made contributions. They can be financed not only by labor taxes (which may not be efficient) but by any feasible tax base, allowing public finance design to be more flexible. They do not depend on a formal employer-employee relationship, which is challenging in developing country labor markets with high informality and fluidity. Their payments also need not be related to earnings history, meaning that redistribution within retirement cohorts can be more effective. And complex record-keeping is unnecessary—all that is necessary is documentation regarding identity and age, and in the case of a targeted design, capacity to measure and rank individual or household resources (as pointed out in the section on social pension implementation). While this may not always be straightforward in the countries considered here, it is much less demanding than a record of labor payments and contributions spanning several decades. 250

Pension Policy in Emerging Asian Economies with Population Ageing

Finally, the relationship between the timing of the development or expansion of a social pension policy and the stage of demographic transition is worth noting. In the early stages of demographic transition, when fertility has fallen and life expectancy is just beginning to rise, there may be a period when lower demands on the public purse enable relatively easy initiation of new programs or further development of existing policies. Properly designed, the social pension is a strong candidate that will then be in place as the lives of individuals are extended. But to bring some rigor to this discussion requires a consistent macroeconomic framework. The state-of-the-art model for this purpose—long-term development with consuming units treated as moving through the life course—is the overlapping generations (OLG) model. This model was developed in the context of developed countries, with the seminal computational large-scale OLG model developed for U.S. tax and pension policy by Auerbach and Kotlikoff (1987). Among other advantages, this type of model can account for demographic change and the behavior of agents regarding saving and labor supply choices as they move through time. The model can also treat consistently the revenue collection and benefit disbursements of the pension and tax systems and feature general equilibrium with market-clearing conditions. Kumru and Piggott (2009, 2010) developed an early model to study the economy-wide effects of targeting noncontributory public pensions in the United Kingdom. More recently, Kudrna and his co-authors (Chomik et al., 2015; Kudrna, 2016; Kudrna et al., 2019, 2022) developed a more sophisticated model for Australia that features a noncontributory pension structure with targeted or means-tested pension benefits.20 Expanding this kind of model to countries with a large informal sector is a major objective of the research program of which this chapter forms a part. Nevertheless, this exercise provides major insights that can inform the present policy discussion. First, the Australian results indicate that a targeted public pension design dominates a universal pension, improving the welfare of older households across the income distribution. Multiple mechanisms contribute to this outcome. Targeting pension payments to poorer elderly households immediately redistributes resources to the poor. But these households tend to have lower life expectancies, which magnifies the reduction in the revenue requirement generated by denying the pension to the better off. Second, the price distortion facing higher-income households is confined to the revenue-collecting tax, because they receive no benefit. As these households are more able to accumulate assets over the life course, in part because the targeted pension system requires less revenue, more move into the resource range where benefits are not payable, thus further reducing the revenue requirement. That is, a targeted first-pillar pension is more sustainable: as second-pillar pensions and life-course resources build with economic growth, the cost of a targeted pension reduces relative to a universal pension. Overall costs, however, will depend on the changing longevity and size of the group receiving pensions. This chapter is an early output from a major research project being undertaken by the ARC Centre of Excellence in Population Ageing Research (CEPAR) with support from the Australian Research Council, which has as its goal the development and implementation of a large-scale stochastic OLG model that explicitly incorporates the informal sector and builds links between formal and informal activity. It aims to quantify the macroeconomic and welfare effects of retirement income policies discussed in this chapter, including our social pension proposal and its variations. In addition to formal and informal labor, the model includes different skill types of households based on household survey data and is applied to examine both the long-run and transition-path implications (including demographic transitions), accounting for behavioral life-cycle and general equilibrium effects. The model will be initially calibrated to Indonesia, but there are plans to make it available to policymakers in countries with broadly similar economic structures, such as Thailand and Vietnam. The aim is to provide policymakers 251

George Kudrna et al.

in emerging economies with policy tools to generate new insights and encourage more rigorous thinking around policy initiatives related to retirement income design, as currently happens in many developed countries, for example, the U.S. Congressional Budget Office OLG model (the so-called Life Cycle Growth Model).

13.4.2

Features of Social Pensions

The key issues regarding first-pillar pension designs relate to benefit level, coverage (universal versus targeted and to what degree if so), eligibility age, and financing source.

13.4.2.1

The Benefit Level and Its Adjustment

Choosing the benefit level is clearly value-laden, related to social views of poverty. For example, the beneficiaries’ needs and basic acceptable standard of living could be judged against some absolute value (e.g., a fixed basket of goods or poverty line) or against prevailing, economywide community standards (e.g., average earnings). Most developed countries adopt a community standard (OECD, 2018). Practice across the developing world is diverse, commonly using some absolute benchmark or proportion thereof, or reverse engineering benefit levels from the allocated fiscal envelope.21 For the benefit itself, the most common practice in emerging economy social pension programs is to have flat benefit levels for eligible recipients for simplicity of administration. The Thai example of increasing benefits with age provides an example that may also be worth considering to provide higher protection as capacity to earn falls and savings diminish. Reliance on flat benefits could be fine-tuned over time as capacity to measure means improves and offers the possibility of tapering benefits as a person’s wealth increases. Korea and Chile offer examples of such an approach in their social pension programs. Once a benefit level is set, it needs to be adjusted over time. If it is accepted that noncontributory pensions should become more prominent in the overall retirement policy structure, then consideration should be given to moving beyond subsistence levels of benefit and thinking instead about benefit levels set to remove poverty. Over time, poverty thresholds are generally viewed as moving with community standards. To strengthen first-pillar pensions, benefits should be indexed to wages (or preferably a wider measure of average labor earnings that better reflects informal sector trends), or perhaps a mixture of prices and wages, rather than prices alone (Whitehouse et al., 2009). Given current practice in emerging EA/SEA of no or very imprecise and discretionary indexation of social pensions, the need for rule-based indexation is pressing.

13.4.2.2

Targeting

Perhaps the most complex question in developing a noncontributory pension policy is whether to target it in some way. As shown in the section on pension policy, emerging EA/SEA countries have adopted both quasi-universal and targeted programs, and sometimes, as in the case of Vietnam, both designs apply depending on recipient age. As argued previously, however, available evidence points to the superiority of a targeted program, where the mode of targeting evolves over time to reflect the sources of income, administrative capacity, and broader developments in digitization and fintech. The key implementation challenge in emerging economies when trying to balance equity and efficiency considerations is administrative capacity to measure means where incomes are volatile and hard to observe precisely. Current poverty-targeted programs for general social assistance in several EA/SEA countries demonstrate significant exclusion errors in targeting, 252

Pension Policy in Emerging Asian Economies with Population Ageing

suggesting that a more inclusive form of targeting for elderly social pensions may be appropriate for the present. A straightforward initial step would be “pension testing” to determine eligibility for social pensions, targeting social pensions to those in informal employment with no or limited coverage by formal earnings-related pension programs. The introduction of population ageing reinforces the sustainability and equity of a meanstested pension design in two ways (see Kudrna et al., 2022). First, as people live longer, they save more for their retirement, which will reduce the payouts of a means-tested social pension program. Analysis for Asia suggests that this behavioral effect may outweigh the compositional effect of larger aged cohort shares (Kinugasa and Mason, 2007). Second, mortality differentials between poor and rich (documented by Waldron (2007); Cristia (2009); and Chetty et al. (2016), for the United States, and OECD (2016), for OECD countries) make a means-tested pension equity enhancing. That is, they provide a higher share of the noncontributory pension budget to lower-income, shorter-lived residents than would be the case with universal schemes (as shown by Kudrna et al., 2022). Available evidence for EA/SEA confirms such income-related mortality differentials (Banerjee and Duflo, 2007; OECD, 2016). A further point relates to resource allocation by households. The households with the greatest flexibility regarding human capital, labor supply, and saving tend to be the richest. A means test that excludes this group from benefits removes a price distortion that would otherwise adversely affect household choice. In addition, of course, consequent upon a lower revenue requirement because of means-testing, tax rates generally will be lower.

13.4.2.3

Choice of an Access Age

This parameter should be consistent with the policy’s purpose: to provide support to those whose earning capacity has been exhausted. As such, both fairness and sustainability are relevant to the choice of access age (Chomik and Whitehouse, 2010). The current range in eligibility ages for social pensions across EA/SEA is probably too low relative to rest-of-life expectancy and labor force behavior in countries such as China, Thailand, and Philippines, and too high to be an effective source of old-age income support for poorer people in Myanmar and in the quasi-universal program in Vietnam. Closer consideration of the current calibration of eligibility ages seems warranted (including the political economy of the relationship to official retirement ages in contributory systems). Also critical is that any access age be indexed to increasing life expectancy over time and that this calculation is taken from the age of access. In other words, access age is indexed to rest-of-life expectancy.22 This will make maintaining the sustainability of the program easier as health improves with economic development and life spans increase.

13.4.2.4

Eligibility

Most countries require residence and/or citizenship for eligibility for first-pillar pensions to avoid pension-based migration. In emerging economies, a citizenship requirement is most common, and that is the prevailing approach in emerging EA/SEA. While current residence suffices in some developed countries, requiring several years of residence after a certain age or within several years of claiming the pension is common.

13.4.2.5

Financing

In all EA/SEA countries, as is common across the developing world, social pensions are financed from general revenues. This has two advantages. First, it gives considerable flexibility to policymakers in choosing the revenue instrument to finance benefits rather than being tied to 253

George Kudrna et al.

formal sector labor incomes as in contributory schemes. It is also more intergenerationally neutral. Second, it does not impose a distortion in taxation across factors of production. Such distortions are quite pronounced in some EA/SEA tax systems (e.g., in China, Vietnam, and Malaysia), which tax property and capital lightly but formal sector labor quite significantly.

13.4.3

Fiscal Costs of Targeted Social Pensions—An Example

We do not yet have an OLG model that effectively represents a developing economy with a large informal sector. But we can nevertheless parameterize a simpler structure to gain some sense of the fiscal costs of social pensions. We now consider specific parameters for social pensions and project the social pension expenditure under various assumptions. The projections are carried out for Indonesia, a country that currently does not provide its elderly with noncontributory social pensions (World Bank, 2020). The pension expenditure in year t, SP(t), is calculated as follows: SP(t) = p(t) ∗ inf s(t) ∗ elderly s(t) ∗ Tpop(t),

(1)

where p(t) is the pension benefit, inf s(t) is the share of informal elderly, elderly s(t) is the share of age-eligible population, and Tpop(t) is the total population. Our projections for Indonesia start in 2019 (base year) and span to 2100. The demographic data [for variables elderly s(t) and Tpop(t)] derive from United Nations (2019). We use the share of informal labor force as the proxy for inf s(t), derived from the World Bank toolkit for informality scenario analysis documented by Loayza and Meza-Cuadra (2018). This toolkit projects the employment shares and other variables over the period 2016–2040. After 2040, we assume the same 2040 shares (and the GDP per capita growth rate). We also use the projected GDP per capita growth rate from the toolkit. Note that the per capita growth rates are used to index the pension benefit p(t) over our projection period 2020–2100. In 2019, the pension benefit [p(2019)] is based on World Bank (2021) data, with all the benefit measures related to the International poverty line (IPL) set at 10,282.4 IDR [or US$1.9 (2011 purchasing power parity or PPP)] per day, per capita, amounting to about 6.5 percent of GDP per capita in 2019. Table 13.5 reports our projections of the social pension expenditure as a percentage of GDP over 2020–2100 under the baseline and several alternative scenarios. We first discuss the expenditure under the baseline projection. That baseline scenario assumes (1) the medium (most likely) fertility variant population projection, (2) the social pension benefit at 50 percent of ILP (5,141.2 IDR per day, per capita), (3) age eligibility of the 65+ population, and (4) targeting all informal elderly (decreasing only slightly from about 82 percent in 2020 to 80.7 percent in 2040). As shown, the current expenditure (on social pensions targeting all informal elderly aged 65 and over) would be very modest, at 0.164 percent of GDP.23 Due to population ageing in Indonesia, that expenditure increases to 0.411 percent of GDP in 2050 and 0.696 percent of GDP in 2100. Note that this modest social pension benefit (at 3.2 percent of GDP per capita) with small fiscal costs to the government (as indicated in Table 13.5) would eliminate most old-age poverty in Indonesia. Table 13.5 also provides social pension costing under several alternative projections, assuming (1) different demographic projections (with low- and high-fertility variants), (2) different benefit levels (set at 17 percent and 100 percent of IPL), and (3) different age eligibilities (the population aged 60+ and the population aged 70+). Because the low-fertility projection implies more pronounced population ageing (further increasing the old-age dependency ratio over the projection period), the social pension expenditure increases to 0.965 percent of GDP in 2100, compared with the baseline projection at 0.696 percent of GDP in 2100. 254

Pension Policy in Emerging Asian Economies with Population Ageing Table 13.5 Projections of social pension expenditures for Indonesia (percentage of GDP)

Notes: a Assuming the benefit at 50 percent of IPL (5,141 IDR, or US$0.8, in 2011 PPP per day in 2019) to all informal 65+ under medium fertility projection; b this approximates the P1 measure (or poverty gap) - average per capita shortfall to IPL; c this implies no poverty of age-eligible population; d set at age 60 in 2020 and then indexed to target constant 18.3 years of age eligibility for pension over 2021-2100, increasing to age 68 in 2100. Source: Authors’ calculations using data from United Nations (2019); Loayza and Meza-Cuadra (2018); World Bank (2021).

Changing the benefit level alters the expenditure significantly over the entire projection period. As one alternative, we set the benefit to 17 percent of GDP per capita, which represents the poverty gap P1 measure (in 2019) (i.e., average per capita shortfall to IPL). Using that benefit level, the expenditure is significantly lower, at only 0.056 percent of GDP in 2020, increasing to 0.237 percent of GDP in 2100. The other benefit alternative sets the social pension benefit at IPL (or 6.5 percent of GDP per capita in 2019). Even here, the expenditure is still modest at 0.329 percent of GDP and increasing to 1.393 percent of GDP in 2100. This benefit level would imply no poverty at older ages (using the World Bank’s PPP IPL measure).24 The eligibility age for the pension payment also has significant impacts on the overall cost. However, even with the eligibility age set to 60, the social pension expenditure would be 0.264 percent of GDP now and 0.857 percent of GDP in 2100. As mentioned, an alternative policy in relation to the eligibility age would be to consider an increasing eligibility age corresponding to increasing life expectancy over the projection period. The results in the last row of Table 13.5 show the fiscal costs of an automatic pension eligibility indexation policy, based on the Dutch policy rules (see Ayuso et al. 2021, for details). Specifically, in our example for Indonesia, we set the access age for the proposed social pension payments to age 60 and then index this access age to target the constant (social pension payment) period of 18.3 years (the life expectancy for both sexes at age 60 in 2020) over the period 2020–2100. Because the life expectancy at age 60 is projected to increase from 18.3 years in 2015–2020 to 26.4 years in 2095–2100, the eligibility age (for social pensions) increases to 68 years in 2100 under this alternative projection. The proportion of the age-eligible population more than doubles by 2100, but it is significantly smaller than under the constant access age policies at age 60 and even age 65. Importantly, the increases are automatically linked to increasing life expectancy.25 Assuming this alternative, the fiscal costs of social pensions increase to just 0.6 percent of GDP in 2100. Overall, our projections for Indonesia show modest fiscal costs of introducing social pensions targeted to informal elderly. Importantly under our baseline setup, these social pensions would 255

George Kudrna et al.

eliminate most old-age poverty in Indonesia. Similar costing (and poverty applications) would apply to other emerging Asian economies with large informal sectors, which will undergo pronounced population ageing over this century.

13.5

Concluding Comments

This chapter brings empirical evidence and economic rigor to a discussion about a complex policy issue. We provide evidence that the issue is urgent because demographic transition is occurring very rapidly, family structures that have traditionally been supportive across generations are being stressed by large-scale rural-urban migration and falling fertility, informal employment is dominant and showing little sign of rapid formalization in most EA/SEA countries, and retirement-related income protection programs in emerging EA/SEA countries are underdeveloped. These social and economic forces combine to threaten the well-being of older cohorts in coming decades. At the forefront of concern is the welfare of informal sector workers who will live for longer over time, but who will not have the opportunity to accumulate resources to finance their increased life spans. This naturally leads to consideration of an expanding social, or nonemployment-related, pension to help reduce poverty within this cohort. Illustrative calculations suggest that such support, which is more generous than general subsistence welfare, can be offered at low revenue cost relative to GDP and is unlikely to generate major adverse behavioral impacts. New technology is enabling digital identification of previously undocumented residents, thus rendering feasible electronic transfer of retirement support with fewer possibilities for error, corruption, and fraud and more efficient implementation of social pensions to the informal sector (Gelb and Diofasi Metz, 2018). While several countries under consideration in this chapter already have a social pension structure in place, specified benefits are extremely low and coverage is patchy. If we specified minimum criteria for a “meaningful” social pension to be coverage of half the targeted population and a minimum benefit of US$2 a day, then at the national level, no social pension in the emerging economies of East or Southeast Asia would qualify, although in a few cases, narrower jurisdictions, such as Shanghai, may meet these thresholds. Even Singapore, a highly developed nation with one of the world’s highest GDPs per capita, has a very stringently means-tested pension, with a maximum monthly value of less than US$70.26 Available evidence suggests that a targeted pension, tested for affluence, whose benefit level is indexed to community standards and whose access age is related dynamically to matureage expected life span, offers the best design for social pensions under these circumstances. Our numerical illustrations for Indonesia suggest that an outlay as low as 0.16 percent of GDP currently and rising to around 0.7 percent of GDP by 2100 will provide modest support for older cohorts whose earnings capacity has eroded through age and deteriorating health status. But the design of such transfer instruments is not straightforward. They need to be resilient in the face of increasing longevity; they need to deliver support for those in need, in relation to community standards and in the face of economic growth; their design should account for disincentives toward formalization and be compatible with formal sector contributory pensions that will continue to play a role in old-age financial protection; they should be cognizant of implementation capacity constraints (particularly for fine-tuned targeting) but also leverage new technologies, and they should be designed to be fiscally sustainable in the face of the previously identified transitions. The social pensions already implemented in emerging EA/SEA are 256

Pension Policy in Emerging Asian Economies with Population Ageing

variably effective in balancing these multiple and sometimes competing objectives. Clearly all existing social pension programs in the region will need to enhance their adequacy over time to provide effective old-age financial protection. Several will also need to expand coverage among the elderly to address the coverage gaps of contributory schemes. But most can be fiscally sustainable even with such reasonable increases in adequacy and in some cases coverage. For countries in EA and SEA currently without elderly social pensions and with high informality, serious consideration should be given to introducing one. Given the low fiscal cost of a meaningful social pension and the technological capacity to identify and deliver benefits to the target population, one question is why their development has not been more widespread. To conclude on a speculative note, one reason may be that the potential recipients of such a policy initiative have little political leverage. And the transferrelated social protection budget in these countries is very low—typically less than 1 percent of GDP. Furthermore, a long tradition exists of thinking of contributory pensions as an aspirational goal. This has been the prototypical model for retirement income protection in the developed world; many policy officials received their training in countries with a pay-as-you-go system of this kind. Social pensions, however, are seen as akin to “welfare.” Yet, as we indicated previously, generous means-tested social pensions can work well, have many advantages in terms of economic incentives, and are consistent with strong economic development. A major goal of this chapter is to challenge this mindset.

Notes 1 Note that the United Nations population data for China do not include Hong Kong and Macao, Special Administrative Regions of China, and Taiwan Province of China, which all have a lower TFR. Several fertility studies on China using microdata found a lower TFR than the United Nations estimate. For example, Guo et al. (2019), using the Chinese 2015 1 percent sample census data, found the overall TFR to be 1.047 in 2015. 2 Significant mortality variations and life expectancy gaps also exist across regions. For example, in China the life expectancy gap (at birth for both sexes) between the richest and poorest regions exceeded 10 years in 2010 (for details, see Yang and Lu, 2019). 3 Many emerging Asian economies have also been undergoing large internal migration associated with economic development and rapid urbanization, with people largely migrating from rural to urban areas in search of jobs and higher wages. In China, internal migrants account for 20 percent of its population (Zheng et al., 2020). 4 Specifically, according to ILO (2018), all own-account employers and all contributing family members are classified as in informal employment and for employees to be considered as informal, the employment relationship should not be subject to national labor legislation, income taxation, social protection, or entitlement to certain employment benefits. 5 See https://sdgs.un.org/. 6 This pattern has some exceptions, such as Mongolia and China for DB schemes. 7 This section draws in part on work with Robert Palacios (World Bank) for a forthcoming World Bank report on social protection in Asia-Pacific and World Bank (2016). 8 Table 13.4 includes China, though the design of its pension scheme for informal workers is a globally unique hybrid of social pension and matching contributory pension given the combination of high public subsidy in benefit financing with a simplified MDC contribution structure during working life. The following section on MDC pensions discusses it in detail. However, given the scheme’s general revenues financing share and poverty alleviation objective, comparison of parameters with other EA/SEA social pensions is instructive. 9 See World Bank (2018), which provides global comparative figures on social pension generosity. 10 See Sakunphanit and Suwanrada (2011) for an overview of the evolution of the scheme. 11 For a global overview of the state of social protection delivery systems in developing countries, see Lindert et al. (2020), and for a more detailed review of delivery systems for emerging Asia Pacific, see O‘Keefe et al. (forthcoming).

257

George Kudrna et al.

12 Proxy-means testing uses household survey data to generate a formula for applicants that incorporates variables most strongly associated with household poverty. This in turn generates a weighted score for applicant households that is used to determine eligibility. 13 The Chinese informal sector scheme is also structured in part as an MDC, though closer to a social pension in financing terms, and hence is not discussed in this section. 14 See Hinz et al. (2013) for a review of experience in OECD, Latin America, Asia, and Africa. 15 Countries can also be divided into those that adopted a national pension scheme when the population was fairly young, often not long after World War II (China, Philippines, Malaysia, Mongolia, and Singapore), and a second set that adopted national schemes late in their demographic transition (Korea, Vietnam, Thailand, Indonesia, and Lao PDR). There is a further subset of very late adopters such as Cambodia and Myanmar, which are only now in the process of introducing a national contributory pension scheme. 16 Based on full contribution history from age 25 to official retirement age. While most contributory schemes in developing Asia are gender neutral (OECD, 2018), Vietnam, Philippines, and China have higher female replacement rates for equivalent work history, but in practice lower average female benefits due to lower wages and contribution densities. See also OECD (2018) for relative generosity across the income distribution. 17 Both China and Vietnam are gradually bringing the pensions of civil servants in line with those of private sector workers, though in both cases an extended transition period is anticipated. 18 Interestingly, no such impacts are seen in Korea and Japan (O‘Keefe et al., 2021). 19 Piggott et al. (2009) provide a brief overview. 20 The OLG model of the Australian economy features a detailed representation of the Australian fiscal and pension systems, including progressive taxation of total personal income, noncontributory and means-tested age pension, and mandatory DC private pension pillar. 21 The developing country’s global average social pension level in 2017 was very modest at 27 percent of the post-transfer household income of the bottom quintile (see World Bank, 2018). In OECD countries, it was around 17 percent of gross average wage in 2018 (OECD, 2019). 22 Note that OECD (2019) discusses and Ayuso et al. (2021), for example, model automatic pension access age indexation policies in OECD countries. 23 Lu et al. (2014) present a similar exercise estimating the current fiscal cost of introducing social pensions for China. Their central case results indicate that a social pension paid to all informal elderly aged 65 and over would cost less than 1 percent of GDP even in 2050. Nevertheless, their estimates are higher than our baseline case for Indonesia because of their higher benefit level (set equal to the national poverty line, indexed to 6.6 percent of per capita GDP growth) and the already more pronounced population ageing in China. 24 Using a more general benefit level at the lower-middle-income class poverty line (17,317 IDR per pay or annually at 10.8 percent of GDP per capita), would cost 0.55 percent of GDP in 2020 and 2.35 percent of GDP in 2100. Note that this benefit level is about three times smaller than the maximum rate for the Australian noncontributory and means-tested age pension. 25 Specifically, the pension access age is calculated as Page(t) = 60 + round [LE60(t) – LE60(2020), 0], where LE60 is life expectancy (for both sexes) at 60 for the projection period of t = 2020, . . . ., 2100. 26 See https://www.mom.gov.sg/employment-practices/silver-support-scheme.

References ARDLINGTON, C., CASE, A., AND HOSEGOOD, V. (2009): “Labor supply responses to large social transfers: Longitudinal evidence from South Africa,” American Economic Journal: Applied Economics, 1(1): 22–48. AUERBACH, A., AND KOTLIKOFF, L. (1987): Dynamic Fiscal Policy, Cambridge: Cambridge University Press. AYUSO, M., BRAVO, J., HOLZMANN, R., AND PALMER, E. (2021): “Automatic indexation of the pension age to life expectancy: When policy design matters,” Risks, 9(5): article 96. BADIANI-MAGNUSSON, R. (2016): Estimating the Impact of Thailand’s Elderly Social Pension on WellBeing. Mimeo, Washington, DC. BANERJEE, A., AND DUFLO, E. (2007): “Ageing and death under a dollar a day,” NBER Working paper 13683. National Bureau of Economic Research, Cambridge, MA.

258

Pension Policy in Emerging Asian Economies with Population Ageing

BATEMAN, H., KINGSTON, G., AND PIGGOTT, J. (2001): Forced Saving: Mandating Private Retirement Saving, Cambridge: Cambridge University Press. CHETTY, R., STEPNER, M., ABRAHAM, S., LIN, S., SCUDERI, B., TURNER, N., BERGERON, A., AND C UTLER , D. (2016): “The association between income and life expectancy in the United States, 2001–2014,” Journal of the American Medical Association, 315(16): 1750–1766. CHLON-DOMINCZAK, A. (2015): “Women’s work and pensions drawing lessons from Central and Eastern Europe.” In: Marin, B (ed.), The Future of Welfare in a Global Europe, London: Routledge, pp. 337–353. DOI: 10.4324/9781315239446. CHOMIK, R., AND PIGGOTT, J. (2013): “Asia in the ageing century: Part I—Population trends,” CEPAR research brief 2013/01, Sydney. CHOMIK, R., AND PIGGOTT, J. (2015): “Population ageing and social security in Asia,” Asian Economic Policy Review, 10(2): 199–222. CHOMIK, R., PIGGOTT, J., WOODLAND, A., KUDRNA, G., AND KUMRU, C. (2015): “Means testing social security: Modelling and policy analysis,” CEPAR working paper 2015/35, Sydney. CHOMIK, R., PIGGOTT, J., AND YAN, S. (2019): “Ageing, fiscal sustainability, and adequacy of social security systems.” In: Asian Development Bank Institute (ed.), Ageing Societies: Policies and Perspectives, Tokyo: Asian Development Bank Institute, pp. 52–64. CHOMIK, R., AND WHITEHOUSE, E. (2010): Trends in Pension. Eligibility Ages and Life Expectancy, 1950– 2050. Paris: OECD. CRISTIA, J. (2009): “Rising mortality and life expectancy differentials by lifetime earnings in the United States,” Journal of Health Economics, 28(5): 984–995. DE CARVALHO FILHO, I. (2012): “Household income as a determinant of child labor and school enrollment in Brazil: Evidence from a social security reform,” Economic Development and Cultural Change, 60(2): 399–435. DORFMAN, M., HOLZMANN, R., O’KEEFE, P., WANG, D., SIN, Y., AND HINZ, R. (2013): China’s Pension System: A Vision. Directions in Development: Human Development Series. (Directions in development—human development). Washington, DC: World Bank. GELB, A., AND DIOFASI METZ, A. (2018): Identification Revolution: Can Digital ID Be Harnessed for Development?, Centre for Global Development, Washington, DC: Brookings Institution Press. GILES, J., AND HUANG, Y. (2016): Are the Elderly Left Behind in a Time of Rapid Demographic and Economic Change? A Comparative Study of the Poverty and Well-Being of East Asia’s Elderly. Background paper for World Bank Regional Report on Ageing in East Asia, Washington, DC: World Bank. GUO, Z., GIETEL-BASTEN, S., AND GU, B. (2019): “The lowest fertility rates in the world? Evidence from the 2015 Chinese 1% sample census,” China Population and Development Studies, 2: 245–258. HELPAGE INTERNATIONAL. (2018): “Social pensions database.” Available at http: pensionwatch.net/social-pensions-database/social-pensions-database [accessed on April 28, 2021]. HERRMANN, T., LECKCIVILIZE, A., AND ZENKER, J. (2021): “The impact of cash transfers on child outcomes in rural Thailand: Evidence from a social pension reform,” Journal of the Economics of Ageing, 19. DOI: 10.1016/j.jeoa.2021.100311. HINZ, R., HOLZMANN, R., TUESTA, D., AND TAKAYAMA, N. (2013): Matching Contributions for Pensions: A Review of International Evidence, Washington, DC: World Bank. HUANG, Y., AND GILES, J. (2017): Universal Social Pension and Spousal Labor Supply in Rural Thailand. Mimeo, Washington, DC: World Bank. HUANG, W., AND ZHANG, C. (2021): “The power of social pensions: Evidence from China’s new rural pension scheme,” American Economic Journal: Applied Economics, 13(2): 179–205. ILO (INTERNATIONAL LABOUR ORGANIZATION). (2018): Women and Men in the Informal Economy: A Statistical Picture, Third Edition. Geneva: International Labour Office. IMF (INTERNATIONAL MONETARY FUND). (2019): Government Finance Statistics (GFS), IMF Data. Washington, DC: International Monetary Fund. KINUGASA, T., AND MASON, A. (2007): “Why countries become wealthy: The effects of adult longevity on saving.,” World Development, 35(1): 1–23. KUDRNA, G. (2016): “Economy-wide effects of means-tested pensions: The case of Australia,” The Journal of the Economics of Ageing, 7: 17–29. KUDRNA, G., LE, T., AND PIGGOTT, J. (2020): “Review report on demographics, labour force and older people in Indonesia.” CEPAR working paper 2020/27, Sydney. KUDRNA, G., AND PIGGOTT, J. (2019): “Means-tested public pensions: Designs and impact for an ageing demographic.” In: Bloom, D. (ed.), Live Long and Prosper? The Economics of Ageing Populations, London:

259

George Kudrna et al.

Centre for Economic Policy Research, Centre for Economic Policy Research (CEPR) Press, pp. 136143. KUDRNA, G., TRAN, C., AND WOODLAND, A. (2019): “Facing demographic challenges: Pension cuts or tax hikes?,” Macroeconomic Dynamics, 23(2): 625–673. KUDRNA, G., TRAN, C., AND WOODLAND, A. (2022): “Sustainable and equitable pensions with means testing in ageing economies,” European Economic Review, 141. DOI: 10.1016/j.euroecorev.2021.103947. KUMRU, C., AND PIGGOTT, J. (2009): “Should public retirement provision be means-tested?” Australian School of Business Research paper no. 2009 AIPAR 01, Sydney. KUMRU, C., AND PIGGOTT, J. (2010): “Social resilience, means-testing, and capital taxation—Reflections on economic paradigms.” In: Towards a More Resilient Society: Lessons from Economic Crises, Tokyo: The Japan Institute of Economic Affairs, pp. 45–56. LINDERT, K., GEORGE KARIPPACHERIL, T., RODRIGUEZ CAILLAVA, I., AND NISHIKAWA CHAVEZ, K. (2020): Sourcebook on the Foundations of Social Protection Delivery Systems, Washington, DC: World Bank. LOAYZA, N., AND MEZA-CUADRA, C. (2018): A Toolkit for Informality Scenario Analysis, Washington, DC: World Bank. LU, B., HE, W., AND PIGGOTT, J. (2014): “Should China introduce a social pension?,” The Journal of the Economics of Ageing, 4: 76–87. MAJOKA, Z., AND PALACIOS, R. (2018): Targeting versus Universality: Is There a Middle Ground? Washington, DC: World Bank. MITCHELL, O. AND PIGGOTT, J. (2016): “Workplace-linked pensions for an ageing demographic.” In: Piggott, J., and Woodland, A. (eds.), Handbook of the Economics of Population Ageing, Elsevier, pp. 865–904, DOI: 10.1016/bs.hespa.2016.08.001. MOSCONA, J., AND SECK, A. (2021): “Social structure and redistribution: Evidence from age set organization.” Working paper. Department of Economics, Massachusetts Institute of Technology, Cambridge, MA. OECD (ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT). (2016): “Fragmentation of retirement markets due to differences in life expectancy.” In: OECD Business and Finance Outlook 2016, Paris: OECD Publishing, pp. 177-205, DOI: .10.1787/9789264257573-11-en. OECD (ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT). (2017): Pensions at a Glance: OECD and G20 Indicators, Paris: OECD Publishing. OECD (ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT). (2018): . Pensions at a Glance Asia/Pacific 2018, Paris: OECD Publishing. OECD (ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT). (2019): Pensions at a Glance: OECD and G20 Indicators, Paris: OECD Publishing. O‘KEEFE, P., GILES, J. AND HUANG, Y. (2021): “Ageing, work and retirement in China, East and Southeast Asia,” China: An International Journal, 19(3): 53–78. O’KEEFE, P., VASUDEVA DUTTA, P., PALACIOS, R., AND MOROZ, H. (forthcoming): Diverse Paths: The Dynamic Evolution of Social Protection in Asia and the Pacific. Washington, DC: World Bank. PALACIOS, R, AND EVANS, B. (2015): An Examination of Elderly Co-residence in the Developing World. Social Protection and Labor Policy Note # 17. Washington, DC: World Bank. PALACIOS, R., AND KNOX-VYDMANOV, C. (2014): “The growing role of social pensions: History, taxonomy and key performance indicators,” Public Administration and Development, 34(4): 251–264. PIGGOTT, J., ROBALINO, D., AND JIM E´ NEZ-MART´I N, S. (2009): “Incentive effects of retirement income transfers.” In: Robalino, D., Holzmann, R., and Takayama, N. (eds.), Closing the Coverage Gap: The Role of Social Pensions and Other Retirement Income Transfers, Washington, DC: World Bank, pp. 125-148. SAKUNPHANIT, T., AND SUWANRADA, W. (2011): “The universal coverage scheme.” In: UNDP, SU/SSC, ILO (eds.), Sharing Innovative Experiences: Successful Social Protection Floor Experiences, New York: UNDP, pp. 385-400. SOCIAL SECURITY ADMINISTRATION. (2019): “Social security programs throughout the world: Asia and the Pacific, 2018.” SSA Publication No. 13-11802, Office of Retirement and Disability Policy, Washington, DC. UNITED NATIONS. (2019): “World population prospects: The 2019 revision.” Department of Economic and Social Affairs, Population Dynamics. New York. WALDRON, H. (2007): “Trends in mortality differentials and life expectancy for male social security covered workers by socioeconomic status,” Social Security Bulletin, 67(3): 1–28.

260

Pension Policy in Emerging Asian Economies with Population Ageing

WHITEHOUSE, E., D’ADDIO, A., CHOMIK, R., AND REILLY, A. (2009): “Two decades of pension reform: What has been achieved and what remains to be done?,” The Geneva Papers on Risk and Insurance—Issues and Practice, 34(4): 515–535. WORLD BANK. (2016): Live Long and Prosper: Ageing in East Asia and Pacific. World Bank East Asia and Pacific Regional Report, Washington, DC. WORLD BANK. (2018): The State of Social Safety Nets 2018, Washington, DC. WORLD BANK. (2019): Protecting All: Risk Sharing for a Diverse and Diversifying World of Work. World Bank Human Development Perspectives, Washington, DC. WORLD BANK. (2019): Identification for Development (ID4D) 2019 Annual Report (English), Washington, DC. WORLD BANK. (2020): Investing in People: Social Protection for Indonesia’s 2045 Vision, Jakarta. WORLD BANK. (2021): “World Bank open data.” https://data.worldbank.org/. YANG, M., AND LU, B. (2019): “An assessment of mortality and life expectancy for China’s provinces: Based on the 2010 provincial census data,” Population Research, 43(1): 70–80 (in Chinese). ZHANG, C., GILES, J., AND ZHAO, Y. (2014): “Policy evaluation of China’s new rural pension program: Income, poverty, expenditure and labor supply,” China Economic Quarterly, 14(1): 203–230 (in Chinese). ZHANG, Z., LUO, Y., AND ROBINSON, D. (2020): “Do social pensions help people living on the edge? Assessing determinants of vulnerability to food poverty among the rural elderly,” The European Journal of Development Research, 32(1): 198–219. ZHENG, Y., JI, Y., CHANG, C., AND LIVERANI, M. (2020): “The evolution of health policy in China and internal migrants: Continuity, change, and current implementation challenges,” Asia Pacific Policy Studies, 7(1): 81–94.

261

14 TRENDS IN PENSION REFORMS IN OECD COUNTRIES1 Herv´e Boulhol, Maciej Lis, and Monika Queisser

Abstract This chapter explores pension reform trends in Organisation for Economic Co-operation and Development (OECD) countries over the last decades and discusses how already legislated measures might affect the future of pensions. While policymakers increasingly recognize the substantial challenges driven by population ageing, progress in pension policies has been uneven. Reform options to tackle these challenges are well identified, but their implementation often raises thorny political issues. Policies to increase effective retirement ages tend to be particularly unpopular. A better understanding of how financial incentives and myopia affect the retirement decision is needed. Pension communication can help inform workers of the benefits of working longer, but adjusting regulations might be needed to influence savings behaviors substantially. In addition, more research on the distributional impact of pension reforms, including across genders and accounting for inequalities in life expectancy, would be useful. Even the wealthiest OECD countries still have some pension coverage gaps, notably for the self-employed, whose integration into pension systems raises specific challenges. With increasing longevity, flexible retirement has become an important topic for policymakers although some confusion exists about whether this should be considered an instrument primarily used to raise labor supply or to increase individual well-being. Persistently low interest rates might provide new inputs in the debate between pay-as-you-go versus funded pensions. Although the COVID-19 crisis is not over at the time of writing, this chapter includes a preliminary overview of its impact on pensions.

14.1

Introduction

Demographic trends in Organisation for Economic Co-operation and Development (OECD) countries clearly show that substantial reforms are needed to make pension systems fit for the future, providing adequate social protection while being affordable for ageing societies. Young people already face more economic and labor market challenges today than their parents’ generation did; the funding needs for pensions and healthcare will further increase the burden placed on them. It is not that policymakers do not recognize the need and urgency for pension reforms. Over the past 30 years, the combination of rising longevity and falling fertility rates, which have been on developed countries’ policy radar since the 1960s, has triggered reform debates in all OECD countries. However, policy action has often lagged or even gone in the opposite direction. 262

DOI: 10.4324/9781003150398-16

Trends in Pension Reforms in OECD Countries

The reason is that reforming pensions is one of the most difficult tasks for policymakers to tackle. Pension reform announcements have toppled governments around the world, and they often mobilize widespread protest, in particular against increases in retirement age. In the United States, the Social Security retirement system has often been called “the third rail of politics,” comparing it to the high-voltage line of railroads—touch it and you die. It is not uncommon for a government to enter its term with big plans to overhaul the pension system, but given public pushback to water down the reform or abandon it altogether. The pension policy debate has evolved over time, moving from a more structural discussion about the relative benefits of running pay-as-you-go (PAYG) versus funded, defined benefit (DB) versus defined contribution (DC) and public versus privately managed pension systems, to questions around retirement behavior; incentives to retire in pension systems; and the links among increasing longevity, retirement ages, and pension benefit levels. The pension reform discussion, including in the academic literature, has thus become more operational and focused on how best to design socially and financially sustainable pension systems that account for the diversity of people’s work experiences, health status, and retirement preferences while keeping cost manageable for future generations. In this chapter, we explore pension reform trends in OECD countries in more depth, including their impact on future retirement benefits. We look at how the emphasis of reforms has shifted over the past decades and what lessons can be drawn from political experiences with pension reform. We also examine several life-course inequalities to consider when devising future reform plans, such as inequality in increasing life expectancy and gender inequality. Given that increases of the retirement age have proven so unpopular much of the chapter is devoted to the discussion of options for raising retirement ages and to new avenues for more flexible retirement paths. Even the wealthiest OECD countries still have some pension coverage gaps, notably for the self-employed. The COVID-19 crisis has highlighted such gaps, apart from also having a broader impact on the financial balance of pension systems. The pandemic is not over at the time of writing, and further effects on pensions will become evident in the future; yet, given the scale of global health, social, and economic challenges, we also include a preliminary assessment of the situation. Finally, we close by taking another look at the structural debate on the relative merits of PAYG versus funded pension systems, which, given broader macroeconomic developments, is likely to gain more traction in the future.

14.2

Extent, Limit, and Impact of Pension Reforms Over the Last Decades

This section starts by assessing the extent of pension policy responses undertaken so far to deal with the challenges posed by population ageing and discusses the political economy of pension reforms. The need to assess the impact of pension reforms on old-age income inequality is then highlighted, before focusing on differences in pension benefits between men and women.

14.2.1

Greater Recognition of Challenges Ahead but Uneven Pension Policy Achievements

Pension reform activity in OECD countries has changed substantially over the past decades. In the late 1970s, a period marked by economic difficulties, pension policy measures mainly focused on facilitating early retirement with the aim of reducing unemployment rates and raising contribution rates (Whitehouse et al., 2009). Only since the 1990s have policy responses across countries started to deal more consistently with the impact of ageing.2 The World Bank (1994) took a leading role in highlighting pension challenges ahead in its report “Averting the Old-Age 263

Herv´e Boulhol et al.

Crisis,” which promoted multi-pillar pension systems. A large body of work, including Bl¨ondal and Scarpetta (1999) and Gruber and Wise (1999), highlighted the poor incentives built into pension systems to make people work after age 55.3 Between the mid-1990s and the mid-2010s, pension reform activity was particularly intense in Europe (Carone et al., 2016). Policy responses then typically involved reducing future benefits and encouraging people to retire later, as well as increasing coverage of voluntary pensions and combating old-age poverty. These reforms have been implemented gradually, sometimes too slowly, through a wide range of measures. While public finance pressure resulting from the financial crisis of 2007–2008 accelerated pension reforms, especially in southern European countries, their actual implementation during the recovery phase was unequal (OECD, 2017a). Another trend has been the harmonization of pension schemes for public-sector and privatesector workers (OECD, 2016). Some countries opted for systemic reforms to insulate pension finances from the impact of ageing, directly affecting pension benefit levels. Notional defined contribution (NDC) schemes were introduced in Sweden, Italy, Latvia, Poland, and Norway (Holzmann et al., 2020, for an anthology) to replace PAYG DB pensions; funded occupational pensions increasingly shifted from DB to DC; and mandatory funded DC plans were created as a top-up or substitute for public pension provision, but subsequently abolished in several countries (Ortiz et al., 2018). Many countries also introduced closer links between contributions and entitlements. These links improve transparency, but contrary to expectations from theoretical models, they have not been very successful in inducing individuals to move into formal markets and to make voluntary pension contributions (Schwarz and Arias, 2014). Adapting conceptual frameworks to account for this evidence and make the models more consistent with these limited incentives would contribute significantly to both research and to the policy debate. Between 1990 and 2020, public pension spending increased by about 2 points of gross domestic product (GDP), from about 6 percent to 8 percent, on average among OECD countries. Recent projections based on legislated measures suggest a further increase of about 1 point on average between 2020 and 2050 (EC, 2021; OECD, 2019a). This compares with 2.5 points based on projections made 10 years ago for the same period, suggesting that some progress has been made in reigning in spending. The average age of labor market exit, which reached its trough around 2000, has been moving upward again (Figure 14.1). Many authors, including Duval (2004), Blundell et al. (2016), and Coile et al. (2018), suggest that social security reforms were instrumental in increasing labor force participation at older ages. But more research is needed, along the lines of the analysis undertaken by Geppert et al. (2019), to disentangle the effects of pension policies from those of health improvements and higher education levels, among others, and to separate the effects of raising statutory ages, reducing early retirement options, tightening unemployment and disability pathways to retirement, and improving financial incentives to continue working. Quantifying these impacts also means distinguishing short-term and long-term effects and identifying complementary measures that improve the effectiveness of these reforms as the impacts are typically nonlinear.

14.2.2

Pension Reforms and the Political Process

Pensions are one of the most difficult policy areas to reform. This may explain why pressure from economic crises has often triggered important pension reforms or at least their timing (Beetsma et al., 2020). Key measures to deal with ageing-related challenges, such as raising retirement ages, lowering future benefit promises, or changing benefit indexation, tend to be unpopular, and the choice among various policy options is often controversial. Making pension systems 264

Trends in Pension Reforms in OECD Countries

Figure 14.1 Average effective age of labor market exit and normal retirement age, OECD average 1970–2018. Note: The average effective age of labor market exit is defined as the average age of exit from the labor force for workers aged 40 and older. The normal retirement age in a given country is defined by the OECD as the age of eligibility of all mandatory pension schemes combined, based on a full career from age 22 (OECD, 2019a). Source: OECD (2019a), Figure 6.10, average of men and women.

financially and socially sustainable requires long-term policy decisions that might be difficult when facing electoral pressure. Given health improvements associated with longevity gains, ample evidence indicates that measures can be taken to raise labor supply and largely attenuate the impact of ageing on pension systems (B¨orsch-Supan, 2000; Gruber and Wise, 2004). B¨orsch-Supan (2015) concludes that the primary problem is the political resistance against institutional changes, leading to procrastination. Moreover, going against the interest of the increasing number of retired or nearly retired persons who are more likely to vote may be increasingly costly politically (Sinn and ¨ Ubelmesser, 2003). The political economy literature reviewed by Casamatta and Batt´e (2016) highlights how ageing would lead to expanding PAYG pension systems by shifting the agerelated weights of the electorate. In these theoretical frameworks, ageing would then preclude reducing old-age benefits and would lead to expanding pension spending as a share of GDP, resulting in higher contribution rates as these models rule out financial imbalances. In these theoretical frameworks, raising retirement ages is often also among the outcomes of the electoral vote to limit the increase in contribution rates. Further research on the political economy of pension reforms should focus on several issues. First, although intergenerational redistribution and concerns about financial sustainability are central in the field, the models typically exclude financial imbalances. Yet, pension deficits are frequent in the real world, and the models should take this into account. Second, the emphasis on intergenerational conflicts seems to be exaggerated. While numerous surveys show that younger people have less interest in pensions, the opposition to reforms, especially to raising retirement ages, cuts across generations. Youth worry about the prospects of longer working lives while surveys show older people being more altruistic than assumed in the models (e.g., Angelini and Laferr`ere, 2013; Lockwood et al., 2021), countering the idea that “gerontocracy” 265

Herv´e Boulhol et al.

rules (Tepe and Vanhuysse, 2009). Theoretical frameworks should better account for individual preferences, which would improve the understanding of the obstacles to pension reform. Third, while pension reforms may trigger widespread public protest, a high degree of consensus exists among experts on the need for reform. More research is needed to explain this gap between public and expert opinions as suggested by B¨orsch-Supan (2013) on how to communicate scientific evidence better and demystify popular fallacies. More attention should also be paid to the institutional frameworks facilitating the design and implementation of pension reforms. Tompson (2009) highlights that having an electoral mandate for reform is crucial. Attempting to pass a reform “quietly” typically backfires. Successful pension reform needs strong government leadership based on a long-term vision (Holzmann and Hinz, 2005). Pension reform experiences in Central Europe highlight the importance of pension policy consistency over time (Schwarz and Arias, 2014). Failing to build consensus has often led to the reversal of pension reform by subsequent governments. In turn, frequent policy shifts hurt confidence and weaken public support. Good examples of consensus building are the governance of the systemic pension reform in Sweden (Palmer, 2000) and Norway (Hinrichs, 2021).4 By contrast, according to the latter author, the succession of significant pension reforms in Germany since 1989 has reduced confidence in the system. Barr and Diamond (2008) indicate that the pension system has changed too frequently and with too short a time horizon in the United Kingdom as well.

14.2.3

More Efforts Needed to Assess the Impact of Enacted Reforms on Old-Age Income Inequality

Many econometric studies estimate the impact of changes in pension policy on people’s behavior, especially on effective retirement ages. However, the consequences of reforms on pension benefits across generations are less studied. Two complementary approaches can be used to assess these effects. Pension benefit projections for typical agents with different earning levels, career profiles, sexes, etc., can illustrate the impact of reforms. They also allow for cross-country comparison, as undertaken regularly by the European Commission and the OECD in the Pension Adequacy Reports (EC, 2021) and Pensions at a Glance publications (OECD, 2019a). For example, based on 2018 legislation, future net pension replacement rates from mandatory schemes for averagewage full-career workers would be about 60 percent in the OECD on average, ranging from around 30 percent in Lithuania, Mexico, and the United Kingdom to around 90 percent in Austria, Denmark, Italy, Luxembourg, and Turkey. Figure 14.2 implicitly shows the impact of past reforms on pensions by comparing replacement rates after a full career until the retirement age for cohorts born in 1940, 1956, and 1996. At normal retirement ages, replacement rates are estimated to decrease in 21, increase in 10, and remain flat in 5 countries. Past reforms have reduced pensions of average-wage individuals born in 1996 by 10 percent relative to those born in 1940, on average in the OECD (OECD, 2019b), while life expectancy at age 65 is projected to increase by 6 years. Accounting for changes in normal retirement ages, the share of adult life spent in retirement would increase by almost 10 percent. Legislated increases in retirement ages are only about half of what would be needed to stabilize the share of the retirement period in adult life over the next five decades. The second approach to study the impact of reforms on benefits is to use microsimulation because typical cases are, by definition, not representative. Microsimulation can provide an aggregate picture of the impact of reforms, including on pension spending, and identify winners and losers. But it requires large datasets and numerous modeling assumptions, which can 266

Trends in Pension Reforms in OECD Countries pp 40 30 20 10 0 -10 -20 -30 -73

Me x Sw ico ed Po en lan d Ch ile Sw Sp itz ain er la Un ite Gr nd d K ee ing ce A dom Ne ustr the alia Lu rla xe nd mb s Ge ourg rm De any nm a OE rk No CD rw a Ja y p Fr an an Fin ce lan La d t Un Ice via ite lan dS d ta Be tes lgi Po um rtu Ca gal na Au da st Ne Slov ria Cz w Z eni ec ea a h R lan ep d ub lic Ita Tu ly r Sl ov Ir key ak ela Re nd pu Hu blic ng Es ary ton i Ko a re a Isr ae l

-40

Figure 14.2 Replacement rates will fall in most OECD countries. Change in theoretical replacement rate between the 1940 and 1996 birth cohorts, in percentage points, full-career average-wage workers. Source: OECD (2019b).

complicate the identification of driving factors, and estimates might not be very robust. With increasing computing power and data availability, microsimulation has become more important in social sciences (Dekkers and Keegan, 2014). Microsimulation techniques should be further exploited to analyze redistributive effects of pension reforms and to assess the future evolution of old-age inequality, as in the United States where this tool is, e.g., used to examine the impact of rising life expectancy inequality on the progressivity of Social Security (Goldman and Orszag, 2014). Public policy has become increasingly concerned with retirement income adequacy and prevention of poverty among retirees. OECD (2017b) estimates that about two-thirds of a persistent increase in income inequality during the working age is transmitted into old age. More research is needed to shed light on whether the current environment of ageing pressure combined with labor market developments is likely to lead to a more unequal and divided old age, with prosperity for some and deep poverty for others (Phillipson, 2015). Dekkers et al. (2018) study this for three European Union countries, estimating that old-age poverty risks will fall in Belgium, but increase in Italy and Sweden, potentially related to the close link between earnings and pensions in NDC schemes in the latter two countries and the lowering of minimum pension and old-age safety-net levels.

14.2.4

The Shrinking Pension Gender Gap

Inequalities in pensions stem largely from labor market inequalities (in employment, part-time work, and wages). The gender pension gap exceeds 30 percent in many OECD countries and is often much higher than the gender wage gap given employment differences between men and women (Lis and Bonthuis, 2019). Along with better pension protection of career breaks due to childcare, increasing female employment has contributed to narrowing the gender pension gap in many countries and projections suggest the gap will be almost eliminated in a few countries (Barslund et al., 2021). Yet, this will not be the case in countries with persistently high gender wage gaps. Since the late 1970s, countries have been making pension systems more gender neutral. The European Commission adopted a directive on the progressive implementation of equal treatment of men and women for social security matters in 1979, which forbids any sex discrimination in terms of access, contributions, and benefit calculation, but certain areas regarded as advantageous 267

Herv´e Boulhol et al.

to women, such as the retirement age, maternity-related pension rights, and survivor pensions, were excluded. In general, gender neutrality would include equal eligibility conditions on all pension benefits for men and women. In 2018, gender differences in the normal retirement age existed in almost one-third of OECD countries, but seven have legislated to equalize the ages in coming decades (OECD, 2019a). Much of the gender gap in employment is due to child-related career breaks, which are cushioned by childcare-related pension credits, often linked to maternal and parental leave. In some countries, e.g., Slovenia, pension credits also cushion the impact of reduced working hours due to childcare. However, these credits cannot compensate for lower wages related to career shifts or lost promotion opportunities of mothers (Berniell et al., 2021; CukrowskaTorzewska, 2020; Kleven et al., 2019). Some countries, including France, Germany, Italy, and Spain, increase the pensions of mothers independently of whether they take a childcare break (Lis and Bonthuis, 2019), effectively offsetting part of the gender wage gap. Survivor pensions used to cover only widows but are now often available to men and samesex couples (OECD, 2018a). Today survivor pensions mainly aim to reduce the income drop when the partner dies by complementing the survivor’s own pension. As women live longer, they are the primary beneficiaries of survivor pensions. Among pension rights derived from the partner’s entitlements, spousal supplements are granted only in Belgium, Japan, and the United States, although survivor pensions exist at different levels in almost all OECD countries. Further research could explore whether the existing rules for survivor benefits are still relevant given labor market developments and changes in family constellations and what other arrangements might be more appropriate. Population ageing is increasing the demand for long-term care, which women often provide within families, resulting in an additional gender employment gap that a specific pension instrument could address. At the same time, instruments that cushion the impact of labor market inequalities on pensions might perpetuate existing gender employment gaps by lowering incentives to work (Nishiyama, 2019; S´anchez-Marcos and Bethencourt, 2018); empirical evidence on this, however, is scarce.

14.3

New Avenues for Pension Policies

This section discusses new avenues for pension policies in order for them to better achieve their objectives, with focus on the following areas: effective retirement ages, flexible retirement, pensions for the self-employed, the impact of COVID-19 on pensions, inequality in life expectancy in the design of pensions, and new developments in the debate between PAYG and funded pensions.

14.3.1

Policy Options to Increase Effective Retirement Ages and Their Limitations

Given ageing pressure, many countries have tightened pension eligibility conditions to raise effective retirement ages, for example by increasing statutory and early retirement ages and minimum contribution requirements. These measures have contributed significantly to prolonging working lives (Geppert et al., 2019; Staubli and Zweim¨uller, 2013; Duval, 2004; Bl¨ondal and Scarpetta, 1999) or delaying benefit claims or both (Manoli and Weber, 2016). Effective retirement ages have also increased because younger cohorts are better educated and in better health while labor has become less physically demanding (Geppert et al., 2019; OECD, 2017b; French and Jones, 2017; Loichinger and Prskawetz, 2017). Yet, whether greater health and education levels translate into higher aggregate employment has been questioned because these explaining 268

Trends in Pension Reforms in OECD Countries

factors were constantly improving even when the effective retirement age was falling (Coile et al., 2018). Based on current legislation, more than half of OECD countries plan to increase retirement ages or have linked them to life expectancy. For every year of gains in life expectancy, Denmark, Estonia, Greece, and Italy will increase the retirement age by 1 year, while Finland, the Netherlands, and Portugal will increase it by 8 months (De Tavernier et al., 2021). For the second group of countries, this would allow spreading the overall gains in longevity between working life and retirement to more or less maintain the current splitting of adult life between work and retirement. Some countries, however, backtracked from previous commitments to increase the retirement age, including through not applying links to longevity. An extensive investigation of the reasons behind both the backtracking and the fierce opposition to raising retirement ages versus other alternative policy options might provide guidance for a more robust implementation of retirement-age increases where needed. Tightening pension eligibility conditions can be less effective when gatekeeping to unemployment or disability benefits fails, allowing workers to leave the labor market before retirement age (Wise, 2015; Bl¨ondal and Scarpetta, 1999). In Austria, increasing the retirement age led to higher uptake of disability and unemployment benefits (Staubli and Zweim¨uller, 2013), while in Germany it had no significant spillover effects to other schemes (Engels et al., 2017). To limit these side effects, tightening access to pensions needs to be accompanied by labor market policies to ensure that older workers have sufficient opportunities to prolong their careers and are helped in their efforts to search for and accept job offers. These include measures against age-based discrimination and limiting seniority-based wage setting and active labor market policies to enhance the skills and employability of older workers (OECD, 2020a, 2019c). Some groups of workers are exempt from the general eligibility rules, such as those in hazardous and arduous occupations who can retire earlier in some countries. Among others, Natali et al. (2016) and Zaidi and Whitehouse (2009) describe the rules of hazardous and arduous pension schemes. Offering the option to retire early to people who risk suffering from health problems and having lower life expectancy due to poor working conditions could be justified on equity grounds. However, creating or maintaining special rules for selected occupations raises serious concerns. First, the risk exists that the selection of eligible occupations is arbitrary rather than based on medical evidence. Second, the evidence on the differences in longevity and timing of retirement by occupation is not conclusive. Some jobs, especially manual work, have a negative impact on health (Qi et al., 2018). This might explain mortality differences by occupation (Brønnum-Hansen et al., 2019; Ervasti et al., 2018), even though the causal link between occupational risks and mortality is not clear as other factors related to socioeconomic disadvantage are likely to play an important role. Scharn et al. (2018) find in their review of the retirement timing literature weak support for differences in labor market exit ages by sector, while physically demanding job characteristics are not found to shorten working lives. Third, pension rules are not effective in mitigating health risks stemming from difficult working conditions during the career, which should be addressed by broader policies related to improvements in working conditions and disease preventive measures. More work is needed to better document the functioning of hazardous and arduous pension schemes in relation to their precise rationale and to provide causal evidence on occupational health risks. Given the political difficulties in increasing retirement ages, governments have attempted to increase effective retirement ages through modifying financial incentives, for example by introducing bonus-penalty schemes, which are less contentious. To encourage working longer, flexible options should include at least actuarially neutral adjustments of benefits (Queisser 269

Herv´e Boulhol et al.

and Whitehouse, 2006). Duval (2004) and Gruber and Wise (2004) indicate that labor market participation at older ages was low in many countries in part due to very low penalties for early retirement. Some countries raised these penalties (Whitehouse et al., 2009), resulting in higher effective retirement ages, e.g., in Germany (B¨orsch-Supan et al., 2020b; Engels et al., 2017). However, Lalive et al. (2020) show that, while tightening eligibility conditions both delayed the age of claiming pensions and increased exit ages from the labor market, only the former was true for reforms that increased penalties for early retirement in Switzerland. Recent studies show that people in poor health tend to use early retirement options more often (Scharn et al., 2018; Axelrad, 2018; Leijten et al., 2015). Moreover, those in good health and in stable jobs prolong working lives beyond statutory retirement age more frequently (Nivalainen et al., 2020; Le Duigou et al., 2020; Qi et al., 2018). More evidence is needed to address the question of whether people with short life expectancy use early retirement more often. If so, the flexibility of when to retire combined with actuarial adjustments might alleviate the concern that the regressive impact of a common retirement age for all harms workers with poor health, which is discussed in a subsequent section.5 One reason for tightening eligibility conditions to address ageing pressure even when benefits are adjusted in an actuarially neutral way is short-sightedness (myopia). This refers to “nonrational” or time-inconsistent decisions leading to short-term biases when balancing labor versus leisure over the lifetime and when trading current versus future consumption (B¨orsch-Supan et al., 2018b). As a result of cognitive limitations, underestimation of longevity, and low levels of financial literacy, many people tend to retire as early as possible even with low pensions (O’Dea and Sturrock, 2018; Hurd et al., 2004), save too little (OECD, 2018b; Benartzi and Thaler, 2007), and buy no annuities (Benartzi et al., 2011; Davidoff et al., 2005). On top of setting the retirement age high enough, improving financial literacy is often perceived as a way to address these issues (Fornero et al., 2019; Gustman et al., 2012; Clark et al., 2011). However, the literature shows that few people across countries are knowledgeable about even the basic concepts of interest compounding, inflation, and risk diversification, casting doubts on the effectiveness of short exposure to financial literacy training. Given the inability of many people to make informed financial decisions, the debate is still ongoing about whether regulation should be emphasized (more than financial education) by limiting or simplifying the choices that people face (Lusardi and Mitchell, 2014). Addressing low financial literacy and cognitive biases is not enough to prevent regrets when retired. B¨orsch-Supan et al. (2020a) show that those who regret having too little retirement savings have experienced substantial negative income shocks related to health, family circumstances, or unemployment, while they do not show a higher tendency to procrastinate. Furthermore, pension communication can help increase effective retirement ages by informing workers about how longer working lives would increase their pensions. Several studies suggest that proper communication can help to convince voters of the need to increase the retirement ages in response to population ageing (Fornero et al., 2019; Hagelund and Grødem, 2016; OECD, 2014; Boeri and Tabellini, 2010), but a rigorous empirical validation seems to be missing. Recently, the government of Sweden, where the system is already based on actuarial adjustments of benefits, introduced a target retirement age, which is supposed to influence the social norm about the “right” age to retire and to link this norm to improvements in longevity (OECD, 2019a). This new policy will also require further empirical investigation. Mandatory contributions, minimum retirement ages, and mandatory annuitization are key elements of mandatory pension schemes that constrain individual consumption choices to ensure consumption smoothing into retirement and prevent old-age poverty. On top of the previously discussed myopic behaviors, the rationale for limiting individual choices refers to broadly sharing 270

Trends in Pension Reforms in OECD Countries

the cost of providing minimum standards of living for the older population and to addressing the underdevelopment of annuity markets (Lindbeck and Persson, 2003). Recently, governments have attempted to fulfill the key objectives of pension systems while minimizing the constraints on individuals by introducing both flexible retirement options, discussed as follows, and automatic enrollment. Automatic enrollment involves signing people up automatically to a pension plan while giving them the chance to opt out. Even though these are private schemes, their design often includes tax incentives or contributions matched by the government. KiwiSaver in New Zealand and NEST in the United Kingdom are widely known examples that show that automatic enrollment can increase pension coverage, and more countries have experimented with such a design (OECD, 2019d; Rudolph, 2019).

14.3.2

The Question of Greater Sustainability of Pension Systems through Flexible Retirement

Flexible retirement has become an important topic for policymakers. Rigidly set retirement ages and a binary choice between being employed or retired might no longer reflect societal preferences and the diversity of older workers’ situations (Eurofound, 2012).6 Some are able and motivated to work longer for the income, the social interactions, or simply because they like their job; others want to stop working earlier because of health problems, to pursue other interests, or to care for elderly relatives or grandchildren. With higher pension ages, flexible retirement is often advocated to maintain labor market attachment while gradually moving to retirement. However, distinguishing between the need (given ageing) for longer working lives and more working hours throughout the career is important. Offering greater flexibility might encourage older workers to stay in work longer but it might also entice them to reduce working hours when they would otherwise have worked full time. Overall, recent flexibility reforms do not seem to have increased labor supply (B¨orschSupan et al., 2018a), and additional research might be needed to confirm this. According to the available evidence, flexible retirement should thus rather be thought of as a way to enhance well-being by expanding the retirement menu. As OECD (2017a) argues, total flexibility is inconsistent with the very idea of a mandatory pension system, which sets rules to ensure old-age income security. As people might underestimate their future income needs, they risk retiring too early, fully or partially, ending up with insufficient pensions. Thus, a trade-off exists between greater autonomy left to individuals and old-age income risks. The OECD recommends that flexibility should only be granted after reaching a sufficiently high minimum retirement age, with benefits adjusted to age. In addition, full flexibility to combine work and pensions should be given once the condition for a full pension is met. In addition, the rules to draw full or partial pensions should be linked to neither work status nor earnings levels. All OECD countries allow combining work and pensions after the official retirement age, but in some cases disincentives exist (OECD, 2017a). Australia, Denmark, Greece, Israel, Japan, the Republic of Korea, and Spain reduce pensions above a certain earnings amount. Finland, France, Italy, and Poland, for example, require termination of the initial work contract to claim a full pension. Moreover, in countries that offer gradual retirement schedules, pension providers have little flexibility in their provision and uptake is low. By contrast, surveys typically find large shares of workers wanting greater retirement flexibility, with substantial differences across countries (Aegon Center for Longevity and Retirement, 2015). In the European Union, almost two-thirds of citizens find combining a part-time job and partial pension more appealing than fully retiring (Eurofound, 2016). 271

Herv´e Boulhol et al.

In the United Kingdom, more than half would support a system of partial early pensions in return for a lower final pension (Berry, 2011). However, in the Netherlands, the majority of respondents still prefer full retirement at the retirement age over gradual retirement (Elsayed et al., 2018). Significant barriers to flexible retirement in the labor market and in the work culture of many countries remain. Few employers have programs to support a gradual exit from employment (Eurobarometer, 2012; TCRS, 2016). The fixed costs associated with employment, both for employers and employees, could also partially explain the limited use of part-time work for gradual retirement (Piggott and Woodland, 2016). Examining why such a gap exists between wishes and reality is a promising area for further research and could help devise workable solutions for flexible retirement.

14.3.3

Pension Issues Related to Self-Employment

Broad coverage is a prerequisite for delivering adequate pensions. Initially, old-age pensions were offered to employees of the public sector, or to workers of heavy industries, often as occupational pensions and based on the employer-employee relationship. The self-employed were often excluded from mandatory pension systems, apart from specific arrangements for liberal professions such as lawyers, doctors, or craftsmen. Today, the self-employed are an increasingly diverse group ranging from traditional liberal professions and craftsmen to information technology contractors and gig platform workers. In some OECD countries (e.g., Chile, Turkey), the self-employed are largely active in the informal sector, while in others they are formal workers but not covered by pension schemes, contrary to most employees. In most OECD countries, mandatory pensions now cover the self-employed, but they often contribute less and earn fewer entitlements compared with employees with similar earnings. When earning the average wage, they can expect to receive pensions about 20 percent less than employees, on average (Figure 14.3). The lower income of the self-employed might increase old-age poverty risks, but the evidence is still scarce (Fachinger and Frankus, 2017; Knoef et al., 2014).

120% 100% 80% 60% 40% 20%

Me xic Ne Ja o the pan rla nd S s De pain Sw nm itz ark er Ge land rm a Tu ny rke La y tv Po ia la Be nd lgi um Ch ile I Sw taly ed e OE n CD Is Sl rael ov en Cz ec Fr ia Sl h R anc ov ep e ak u Re blic pu Au blic str a No lia rw Ca ay n Lit ada hu an i Ko a r Un Es ea t ite on Ne d S ia w tate Ze s ala Un n ite Fin d d K lan ing d do Ire m lan Au d str Lu Gre ia xe ec mb e o Hu urg ng ar y

0%

Figure 14.3 Theoretical pensions of the self-employed are lower than those of employees. Theoretical pensions of a self-employed worker relative to an employee having both a taxable income (net income or net wage before taxes) equal to the average net wage before taxes, for individuals with a full career from age 22 in 2018 based on mandatory contributions. Source: OECD (2019a).

272

Trends in Pension Reforms in OECD Countries

Pension rules for employees need to be adjusted for the self-employed. Self-employment, for example, does not allow separation of the employer from the employee. Often, the profit from self-employment activity is not distinguishable from the wage paid to the owner, while in other cases, the self-employed can decide whether to receive their income as a wage, a profit, or a mix of both, irrespective of the economic distinction between labor and capital income. Upon retirement, the self-employed no longer receive any income from their business while the market value of the capital stock might only slightly add to retirement savings. Thus, for purposes of smoothing income upon retirement, good reasons exist to base contributions and pension entitlements of the self-employed on total income received from their self-employment activity. An additional problem is that the self-employed are often allowed to calculate their profits based on simplified and advantageous accounting rules. Given these complexities, some countries have forsaken the income-smoothing objective and excluded the self-employed from mandatory coverage in earnings-related schemes, or made them liable for only a lump-sum contribution regardless of their actual income. Several surveys of the selfemployed show that they tend to prefer to remain outside of the mandatory pension system (OECD, 2019a); in Chile, for example, auto-enrollment failed in achieving broad pension coverage of the self-employed (Rudolph, 2019). Thus, space exists for both theoretical and empirical research on ways to increase pension coverage of the self-employed. Further evidence is needed on whether unified solutions work well for all self-employed workers and to what extent rules should be adapted. Recent labor market developments strengthen the case for such further analyses. One challenge is bogus self-employment of workers whose work characteristics resemble those of regular employment but are instead registered as self-employed. Pension prospects of those forced into self-employment by the lack of better opportunities are particularly poor, as Hershey et al. (2017) document for the Netherlands. Nevertheless, self-employment is often preferred due to more favorable tax, including social security contributions, and thus higher take-home pay. Some new forms of work, such as gig work, can be considered self-employment though the borders are fleeting. Additional research is needed to assess the actual size of the gap in current and future pension protection for various categories and income groups of the self-employed.

14.3.4

Impact of COVID-19 on Pensions

The COVID-19 pandemic is affecting pension systems through several channels, but as of the first half of 2021 only preliminary evidence is available, e.g., COR (2020). Policies to contain the spread of the virus severely disrupted economic activity. Most countries introduced or extended work retention schemes to protect employment, and these schemes have generally preserved the accrual of pension entitlements in earnings-related schemes, at the cost of rising public debt (OECD, 2020b). The consequences for pension systems and entitlements will depend on the speed and strength of economic recovery and on the evolution of inflation, interest rates, and asset prices. Three labor market effects will require particular attention from pension policymakers: the long-term scarring effects for youth (Schwandt and von Wachter, 2019), early labor market exits of older workers (Feher and Bidegain, 2020), and labor market withdrawals of (mainly) women related to additional family responsibilities (OECD, 2021; Reichelt et al., 2020). All these effects will have an impact on workers’ future pension entitlements. Due to the pandemic mortality rates have risen worldwide. By mid-2021, more than 3.5 million people had died from COVID-19. From January 2020 to April 2021, deviations of mortality from the long-term trends show that total mortality increased by more than 12 percent on average across 29 European countries, leading to a decrease in the number of people aged over 65 of about 0.4 percent (EuroMOMO, 2021).7 Deaths have been concentrated among 273

Herv´e Boulhol et al.

older people and those with comorbidities, leading to an estimate of the average number of years lost per death at 16, based on 2020 data from 81 countries (Pifarr´e i Arolas et al., 2021). It is too early to say whether the mortality rate will fall back to previous levels (or even lower temporarily) or remain higher because of long-term health effects of the disease and reduced prevention and treatment of other health conditions during the pandemic. As more evidence on the long-term mortality impact of COVID-19 becomes available, analyzing the impact on pension financing in more detail will become possible.

14.3.5

Consideration of Inequality in Life Expectancy in the Design of Pension Policies

Ageing is not a uniform process. Mortality rates differ widely across socioeconomic groups in all age groups in most countries. A 65-year-old university-educated man can expect to live 3.5 years longer than his low-educated peer, on average across 22 OECD countries; for women the difference is lower at 2.4 years (Murtin et al., 2017). These life expectancy gaps are large in Belgium, Chile, Latvia, Poland, and Slovenia but comparatively low in Canada, Italy, and Mexico. On average, these gaps represent 18 percent of remaining life expectancy for men and 11 percent for women. The evidence on changes in socioeconomic inequality in longevity is mixed, varying across countries and measures, such as those based on education, income, or location. Banks et al. (2021) highlight that assessing these changes raises serious methodological issues. Using a wide range of analyses,8 inequality in longevity has consistently increased in Denmark, Finland, Lithuania, Norway, and the United States; consistently decreased in Estonia, Greece, Hungary, Italy, Poland, and Spain; and been stable in France and the Republic of Korea. In Canada, the Czech Republic, Japan, Portugal, Slovakia, Slovenia, Sweden, Switzerland, Turkey, and the United Kingdom the picture is unclear. To sharpen the analysis, measurement of changes in life-expectancy inequality needs to become more reliable. Income redistribution from those dying early to those dying late is the core insurance function of pension systems. If low earners have a shorter life expectancy and thus receive benefits over a shorter period, this reduces the progressivity of pension systems. Schemes that appear to be distribution-neutral, such as those delivering annuities from pure DC pensions, are in fact regressive, as annuities are typically computed from common mortality tables.9 Bommier et al. (2006) estimate that differential mortality offsets about one-third of the income redistribution built into the French PAYG pension system, while S´anchez-Romero et al. (2019) suggest it offsets redistribution fully in the United States. OECD (2017b) estimates that a 3-year gap in remaining life expectancy at retirement reduces total pensions received by low earners by 13 percent relative to those of high earners, on average across countries, on top of the effects from lower earnings. Addressing longevity inequality is a challenge for pension policies. Policymakers should account for this inequality when determining benefit levels for low-income workers, as large longevity gaps can justify increasing redistribution in pension systems, as Diamond and Orszag (2004), for example, argue for the U.S. Social Security benefit formula. By contrast, letting different groups retire at different ages would raise a host of other issues, such as how these groups would be defined and delineated, whether individual health status and behaviors should be considered, how retirement ages should be adapted to changing longevity in a group, etc. Many countries in the past allowed for different retirement ages according to occupational risks, and these were increasingly closed and replaced by disability pension schemes that grant benefits based on individual health status. 274

Trends in Pension Reforms in OECD Countries

The issue of accounting for life-expectancy inequality in a pension benefit formula is sometimes mixed up with the question of how retirement ages should respond to changes in life expectancy. Raising the retirement age with unchanged mortality tables means that the increase will shorten low earners’ average retirement period more due their lower life expectancy and thus be regressive, although this effect is quantitatively small (OECD, 2017b).10 However, this does not mean that introducing an automatic link that raises retirement ages in line with increasing life expectancy is regressive. The reason is the following: if nothing is done and pension ages remain at the same level despite longevity gains, those gains will, based on the same argument, benefit relatively more those with shorter expected lives, at least when longevity gains are broadly shared across socioeconomic groups.11 Therefore, implementing such a link to accompany health improvements will be neutral in terms of redistribution, i.e., neither progressive nor regressive. However, if life expectancy gaps between socioeconomic groups widen, linking the retirement age to life expectancy does raise equity concerns.

14.3.6

News in the Debate between PAYG and Funded Pensions

Starting in the early 1980s with the Chilean pension reform, demographic, economic, and political developments prompted especially Latin American and Central and Eastern European countries to move away from PAYG DB systems to funded systems with individual accounts (Mesa-Lago et al., 2020). While in early stages of demographic development it might be easier to accumulate assets to build a multi-pillar system, later shifts from PAYG to funded systems involve high transition costs and as such do not solve financial sustainability issues. During the transition, pension financing needs are actually higher as resources are required to both pay current PAYG pensions and accumulate new funded entitlements. In a context of strong public finance pressure, low returns of pension funds, and high administrative costs (Hinrichs, 2021), many countries decided to reverse these reforms over the last 15 years (Wang et al., 2016).12 One strong argument for combining funded and unfunded schemes is the diversification of risks that people face, such as financial versus political, social or purely labor market risks, or international versus domestic developments (Lindbeck and Persson, 2003; Shiller, 2003).13 Combining both types of financing pensions can help raise the risk-adjusted returns of contributions to the retirement system, at both the individual and aggregate level.14 Some countries, for example Canada, Finland, and Sweden, have accumulated assets to finance their mandatory public pensions and maintain a good mix between PAYG and funding to finance public pensions. The Republic of Korea and the United States also have a mixed source of financing their public pensions between PAYG and funding; however, financial assets might be depleted to finance the consequences of ageing. One central argument for funded schemes hinges on the assumption (r > g) that financial returns on pension assets (r) are larger than GDP growth (g), which is a good proxy for the internal rate of return of PAYG pensions.15 The r > g relation is generally referred to as dynamic efficiency (Diamond, 1965), implying that no generation can be made better off without making any other generation worse off. Under dynamic efficiency, funded schemes generate higher future pensions. An extensive literature since Breyer (1989) has highlighted that PAYG and funded systems are actuarially equivalent. This means that the accumulated losses for future retirees with PAYG if r > g are equal to the gift made to the first generation of PAYG pensioners who did not contribute fully.16 There is no free lunch, and someone must pay for this gift. Similar trade-offs arise when both pensions and contribution rates are low and must be raised, as in Chile recently.17 In short, replacing (part of) PAYG by funded systems generates losers, and the choice between the two systems is mainly a distributional matter, implying that issues 275

Herv´e Boulhol et al. Table 14.1 Ten-year government bond and nominal GDP growth rates, selected countries, average by periods

Source: Boulhol and L¨uske (2019), based on data from Economic Outlook No 104, November 2018; https://stats.oecd.org/index.aspx?queryid=51396.

of intergenerational equity inevitably arise (Breyer, 1989). If r is lower than g, the capital stock is too large, which is dynamically inefficient: consuming the surplus capital or expanding PAYG pensions is Pareto improving while debt is self-financing.18 Very low interest rates pose new questions in the debate between PAYG and funded pensions (Boulhol and L¨uske, 2019).19 First, which r should be used to assess dynamic efficiency is unclear. Blanchard (2019) shows that both the risk-free and risky rates matter due to uncertainty in financial markets. In the United States, the risky rate has exceeded output growth, contributing to dynamic efficiency. However, currently, the risk-free rate is lower than the growth rate, thus triggering dynamic-inefficiency effects, with an overall unclear effect. For G7 countries Australia and Sweden, Table 14.1 shows that in 1992–2005 there was dynamic efficiency, but this is less clear for 2006–2018, with the impact of COVID-19 on monetary policies exacerbating this effect. When interest rates are low, financing the transition cost might be cheaper but the benefit of the whole shift away from PAYG is not clear.20 No academic consensus exists on whether ageing lowers economic growth prospects (Crafts, 2019). As for interest rates, Carvalho et al. (2016) estimate that demographic trends substantially reduced them in the United States, while B¨orsch-Supan et al. (2016) find that the return on productive capital declines with ageing, but by less than the decrease in output growth. The current prolonged period of low interest rates raises the question whether the dynamic efficiency condition will hold in the foreseeable future. Important avenues for further research are to assess economic efficiency in greater depth, estimate the impacts of ageing on PAYG internal rates of return and on interest and financial market rates, and explore what this will mean for the design of pension reforms.

Notes 1 The opinions and arguments expressed herein are those of the authors and do not necessarily reflect the official views of the OECD or its member countries. We are grateful to a reviewer who made excellent comments on a preceding version. 2 In the United States, the 1983 reform ensured solvency for several decades, and no major action has been taken since. Given ageing trends, the U.S. Social Security trust funds are projected to be insolvent around the mid-2030s, leading to brutal adjustments if no corrective measures are adopted as the trust funds cannot borrow (Gannon et al., 2020). 3 Around this time, the European Commission and the OECD coordinated efforts to make long-term international projections of public pension spending driven by demographic changes (Economic Policy Committee of the European Union, 2000; OECD, 2001).

276

Trends in Pension Reforms in OECD Countries

4 Persuading voters and stakeholders across the political spectrum of the need for reform and the costs of lack of reform requires a solid communication strategy based on research and analysis. Especially research coming from a nonpartisan trusted research institution seems to be effective in giving reform proposals more gravitas. Having an expert body—the more independent the better—in charge of assessing pension schemes is also helpful in strengthening the diagnosis and the acceptance of reforms and to support sound management of the system (Fall and Bloch, 2014). Several countries, including Belgium, Canada, France, and Sweden, have such an independent body. However, distrust of experts is currently on the rise in many countries. Communication strategies that directly target the legitimate worries of voters and stakeholders might therefore play an important role. 5 Technically, actuarially neutral penalties are higher when remaining life expectancy is lower, hence a common penalty benefits those who expect to die early. 6 In its most common use, the term refers to the flexibility to combine work and pensions and in particular to “gradual” or “phased” retirement. A second dimension of flexibility refers to the moment of retirement. 7 The EuroMOMO project (EuroMOMO, 2021) monitors excess mortality in 29 European countries. More than half a million excess deaths were estimated between the start of 2020 until the end of the first quarter of 2021. This compares with the regular baseline number of deaths of around 4.7 million over the same period, which implies that the total mortality rate increased by 12 percent and that the number of people aged over 65 would decrease by about 0.4 percent. 8 These include Mackenbach et al. (2016), Eurostat (2020), and some country-specific studies: Auerbach et al. (2017), Baker et al. (2019), Blanpain (2020), Brønnum-Hansen and Baadsgaard (2012), Chetty et al. (2016), Khang et al. (2019), Insee (2016), Marshall-Catlin et al. (2019), and studies referenced in GAO (2016). 9 Chile, Indonesia, and Mexico use gender-specific tables that lower pensions for women, something that is not allowed in the European Union. 10 The issue is more serious in the United States given the very large increase in life-expectancy inequality (Auerbach et al., 2017), but the United States is clearly an outlier (Banks et al., 2021). This is based on new collected data that show estimated longevity gaps at age 65 between the highest and lowest education groups that are substantially larger in most countries than those found in previous studies, due to more precise information on mortality after the age of 75. 11 As discussed previously, the evidence is mixed about trends in life-expectancy inequality. 12 In turn, such huge and erratic policy shifts have undermined confidence in the pension systems. The retiring of the baby boom generations complicates further any attempt to shift to funded pensions (B¨orsch-Supan, 2015). 13 Other arguments in favor of shifting to funding based on the benefits of developing financial markets and increasing domestic savings have received mixed support (Barr and Diamond, 2008). The initial transformation of Poland’s public PAYG system into a multi-pillar DC approach helped Warsaw’s development as a financial center, while the introduction of funded DC pensions in Chile encouraged the growth of Chilean financial markets (OECD, 2018b). How these developments contribute to raising welfare is beyond the scope of this article. Also, PAYG pensions may generate an implicit tax on labor earnings—like most redistribution devices—thereby restricting labor supply (Sinn, 2000). This is formally the case only in the situation of dynamic efficiency. 14 Ideally, abstracting from transition costs, the exact weights of the different schemes thus depend on social preferences, which differ markedly across countries (Devesa and Dom´enech, 2020). 15 It should be clear from the outset that the variables considered here are net rates that account for fees in funded schemes and administrative costs in both funded and PAYGO schemes. The internal rate of return of PAYG pensions, i.e., the return that delivers pension promises in a financially sustainable way, corresponds to the growth rate of the contribution base (Samuelson, 1958), which is equal to the growth rate of the wage bill under a stable contribution rate and therefore to the GDP growth rate when assuming a constant labor share in GDP. 16 Beyond actuarial equivalence, accounting for some possible labor market distortions generated by PAYG pensions provides stronger support for funded schemes (Homburg, 1990). 17 Additional PAYG contributions create some fiscal space over a generation, which can be used in many ways, while additional funded contributions might generate higher pensions in the future. 18 Are economies dynamically efficient and are there too much capital and savings? Abel et al. (1989) conclude that the economies of the seven OECD countries under study were dynamically efficient. However, this is questioned by Geerolf (2018) who finds that Japan and the Republic of Korea have

277

Herv´e Boulhol et al.

accumulated too much capital, supporting the “savings glut” hypothesis. Kajitani et al. (2018) also question whether China is in a dynamic efficient equilibrium. Homburg (2014) theoretically argues that accounting for land, a nonreproducible factor, rules out dynamic inefficient options. 19 The following discussion goes beyond the challenges posed to pension funds and financial institutions offering life insurance policies that promise pre-crisis and fixed nominal returns (see, OECD, 2016; Munnell and Aubry, 2016). 20 Also, future PAYG promises have a higher present value given that discounting is expensive. Yet this era of low interest rates mandates a re-think of macroeconomic policies to better seize created opportunities (Furman and Summers, 2020).

References ABEL, A. B., MANKIW, N. G., SUMMERS, L. H., AND ZECKHAUSER, R. J. (1989): “Assessing dynamic efficiency: Theory and evidence,” Review of Economic Studies, 56(1): 1–19. AEGON CENTER FOR LONGEVITY AND RETIREMENT. (2015): The New Flexible Retirement. The Netherlands: Aegon, Hague. ANGELINI, V., AND LAFERR E` RE, A. (2013): “Parental altruism and nest leaving in Europe: Evidence from a retrospective survey,” Review of Economics of the Household, 11(3): 393–420. AUERBACH, A. J., ET AL. (2017): “How the growing gap in life expectancy may affect retirement benefits and reforms,” NBER Working Papers, No. 23329. National Bureau of Economic Research, Cambridge, MA. AXELRAD, H. (2018): “Early retirement and late retirement: Comparative analysis of 20 European countries,” International Journal of Sociology, 48(3): 231–250. BAKER, M., CURRIE, J., AND SCHWANDT, H. (2019): “Mortality inequality in Canada and the United States: Divergent or convergent trends?,” Journal of Labor Economics, 37(S2): 325–353. BANKS J., CURRIE, J., KRUTIKOVA, S., SALVANES, K. G., AND SCHWANDT, H. (2021): “The evolution of mortality inequality in 11 OECD countries: Introduction,” Fiscal Studies, 42(1): 9–23. BARR, N., AND DIAMOND, P. (2008): Reforming Pensions: Principles and Policy Choices. New York: Oxford University Press. BARSLUND, M., DEKKERS, G., AND VAN DEN BOSCH, K. (2021): Projections of the Gender Pension Gap: Comparative Report, Brussels: Belgian Federal Planning Bureau, http://www.migape.eu/ pubs/Comparative report WP3 v27 def.pdf. BEETSMA, R., KLAASSEN, F., ROMP, W., AND VAN MAURIK, R. (2020): “What drives pension reforms in the OECD?,” Economic Policy, 35(102): 357–367. BENARTZI, S., PREVITERO, A., AND THALER, R. (2011): “Annuitization puzzles,” Journal of Economic Perspectives, 25(4): 143–164. BENARTZI, S., AND THALER, R. (2007): “Heuristics and biases in retirement savings behavior,” Journal of Economic Perspectives, 21(3): 81–104. BERNIELL, I., BERNIELL, L., DE LA MATA, D., EDO, M., AND MARCHIONNI, M. (2021): “Gender gaps in labor informality: The motherhood effect,” Journal of Development Economics, 150(102599): 1–19. BERRY, C. (2011): Gradual Retirement and Pensions Policy. London: International Longevity Centre. BLANCHARD, O. (2019): “Public debt and low interest rates,” American Economic Review, 109(4): 1197–1229. ˆ Adulte pour les G´en´erations N´ees entre, 1941 et 1955. BLANPAIN, N. (2020): La Mortalit´e Stagne a` l’Age Paris: Insee. ¨ BL ONDAL , S., AND SCARPETTA, S. (1999): “The retirement decision in OECD countries,” OECD Economics Department Working Papers, No. 202, OECD Publishing, Paris. BLUNDELL, R., FRENCH, E., AND TETLOW, G. (2016): “Retirement incentives and labor supply,” In Piggot, J., and Woodland, A. (eds.): Handbook of the Economics of Population Ageing, Vol. 1, Chapter 8, North-Holland, Amsterdam: Elsevier. BOERI, T., AND TABELLINI, G. (2010): “Does information increase political support for pension reform?,” Public Choice, 150(1–2): 327–362. BOMMIER, A., MAGNAC, T., RAPOPORT, B., AND ROGER, M. 2006): “Droit a` la retraite et mortalit´e diff´erentielle.,” Economie et Pr´evision, 168(2): 1–16. ¨ BORSCH -SUPAN, A. (2000): “Incentive effects of social security on labor force participation: Evidence in Germany and across Europe,” Journal of Public Economics, 78(1–2): 25–49. ¨ BORSCH -SUPAN, A. (2013): “Myths, scientific evidence and economic policy in an ageing world,” The Journal of the Economics of Ageing, 1–2: 3–15.

278

Trends in Pension Reforms in OECD Countries ¨ BORSCH -SUPAN, A. (2015): “Challenges for European welfare state,” International Tax and Public Finance, 22: 534–548. ¨ BORSCH -SUPAN, A., BUCHER-KOENEN, T., KUTLU-KOC, V., AND GOLL, N. (2018a): “Dangerous flexibility—Retirement reforms reconsidered,” Economic Policy, 33(94): 315–355. ¨ ¨ BORSCH -SUPAN, A., HARTL , K., AND LEITE, D. N. (2016): “Social security and public insurance,” In Piggott J. and Woodland A. (eds.): Handbook of the Economics of Population Ageing, Vol. 1, Chapter 16: 781–863, North-Holland, Amsterdam: Elsevier. ¨ ¨ BORSCH -SUPAN, A., HARTL , K., AND LEITE, D. (2018b): “Who cares about the day after tomorrow? Pension issues when households are myopic or time inconsistent,” Review of Development Economics, 22(3): 953–989. ¨ BORSCH -SUPAN, A., HURD, M. D., AND ROHWEDDER, S. (2020a): “Saving regret: Self-assessed lifecycle saving behavior in the U.S. and Singapore,” University of Michigan Retirement and Disability Research Center Working Paper, MRDRC WP 2020-413, Ann Arbor, MI. ¨ BORSCH -SUPAN, A., RAUSCH, J., AND GOLL, N. (2020b): Social Security Reforms and the Changing Retirement Behavior in Germany. Cambridge, MA: National Bureau of Economic Research. ¨ BOULHOL, H., AND LUSKE , M. (2019): “What’s new in the debate about pay-as-you-go vs funded pensions?,” In Nazar´e, D., and Cunha Rodrigues, N. (eds.): The Future of Pension Plans in the EU Internal Market. Cham, Switzerland: Springer, 37–53. BREYER, F. (1989): “On the intergenerational Pareto efficiency of pay-as-you-go financed pension schemes,” Journal of Institutional and Theoretical Economics, 145(4): 643–658. BRØNNUM-HANSEN, H., AND BAADSGAARD, M. (2012): “Widening social inequality in life expectancy in Denmark. A register-based study on social composition and mortality trends for the Danish population,” BMC Public Health, 12(1): 1–7. BRØNNUM-HANSEN, H., FOVERSKOV, E., AND ANDERSEN, I. (2019): “Occupational inequality in health expectancy in Denmark,” Scandinavian Journal of Public Health, 48(3): 338–345. CARONE, G., ECKEFELDT, P., GIAMBONI, L., LAINE, V., AND PAMIES SUMNER, S. (2016): “Pension reforms in the EU since the early 2000’s: Achievements and challenges ahead,” European Economy Discussion Papers, No. 042, Publications Office of the European Union, Brussels. CARVALHO, C., FERRERO, A., AND NECHIO, F. (2016): “Demographics and real interest rates: Inspecting the mechanism,” European Economic Review, 88(September): 208–226. CASAMATTA, G., AND BATT E´ , L. (2016): “The political economy of population ageing,” In Piggott, J., and Woodland A. (eds.): Handbook of the Economics of Population Ageing, Vol. 1. Chapter 7, NorthHolland, Amsterdam: Elsevier. CHETTY, R., STEPNER, M., ABRAHAM, S., LIN, S., SCUDERI, B., TURNER, S., BERGERON, A., AND C UTLER , D. (2016): “The association between income and life expectancy in the United States, 2001–2014,” JAMA, 315(16): 1750–1766. CLARK, R., MORRIL, M., AND ALLEN, S. (2011): “The role of financial literacy in determining retirement plans,” Economic Inquiry, 50(4): 851–866. COILE, C., MILLIGAN, K., AND WISE, D. (2018): “Social security programs and retirement around the world: Working longer—Introduction and summary,” NBER Working Papers, No. 23329. National Bureau of Economic Research, Cambridge, MA. ´ COR (CONSEIL D’ORIENTATION DES RETRAITES). (2020): Evolutions et Perspectives des Retraites en France, Rapport annuel du COR. Conseil d’orientation des retraites, Paris. CRAFTS, N. (2019): The future growth path for Europe and Ireland. Third annual conference of the Irish Fiscal Advisory Council Path for the Public Finances, 2019: Long-Term Fiscal Sustainability: Winter is Coming, Dublin! CUKROWSKA-TORZEWSKA, E., AND MATYSIAK, A. (2020): “The motherhood wage penalty: A metaanalysis,” Social Science Research, 88–89(May–July): 1–19. DAVIDOFF, T., BROWN, J., AND DIAMOND, P. (2005): “Annuities and individual welfare,” American Economic Review, 95(5): 1573–1590. DE TAVERNIER, W., BOULHOL, H. AND FUJIKI, Y. (2021): “Automatic adjustment mechanisms in pension systems,” In Pensions at a Glance 2021: OECD and G20 Indicators, Chapter 2, Paris: OECD Publishing, 83–119. DEKKERS, G., CONTI, R., DESMET, R., SUNDBERG, O., AND VAN DEN BOSCH, K. (2018): What Are the Consequences of the AWG 2018 Projections and Hypotheses on Pension Adequacy? Simulations for Three EU Member States, Report Federal Planning Bureau 11732, June, Brussels. DEKKERS, G., AND KEEGAN, M. (2014): New Pathways in Microsimulation. New York: Routledge.

279

Herv´e Boulhol et al. DEVESA, E., AND DOM E´ NECH, R. (2020): “Pensions, economic growth and welfare in advanced economies,” Conference paper, Fiscal Policy and Ageing, Oesterrichische Nationalbank, Vienna. DIAMOND, P. (1965): “National debt in a neoclassical growth model,” American Economic Review, 55(1): 1126–1150. DIAMOND, P., AND ORSZAG, P. R. (2004): Saving Social Security: A Balanced Approach. Washington, DC: Brookings Institution Press. DUVAL, R. (2004): “Retirement behaviour in OECD countries: Impact of old-age pension schemes and other social transfer programmes,” OECD Economic Studies, 2003(2): 7–50. EC (EUROPEAN COMMISSION). (2021): The 2021 Pension Adequacy Report. Luxembourg: Publications Office of the European Union. ECONOMIC POLICY COMMITTEE OF THE EUROPEAN UNION. (2000): Report to the ECOFIN Council on the Impact of Ageing Populations on Public Pension Systems, November, Brussels. ELSAYED, A., DE GRIP, A., FOUARGE, D., AND MONTIZAAN, R. (2018): “Gradual retirement, financial incentives, and labour supply of older workers: Evidence from a stated preference analysis,” Journal of Economic Behavior and Organization, 150(C): 277–294. ENGELS, B., GEYER, J., AND HAAN, P. (2017): “Pension incentives and early retirement,” Labour Economics, 47(C): 216–231. ¨ ¨ ERVASTI, J., PIETIL AINEN , O., RAHKONEN, O., LAHELMA, E., LALLUKA, T., AND MANTY , M. (2018): “Long-term exposure to heavy physical work, disability pension due to musculoskeletal disorders and all-cause mortality: 20-year follow-up—introducing Helsinki Health Study job exposure matrix,” International Archives of Occupational and Environmental Health, 92(3): 337–345. EUROBAROMETER. (2012): “Special Eurobarometer 378 on Active Ageing,” European Commission, Brussels. EUROFOUND. (2012): “Income from work after retirement in the EU,” Publications Office of the European Union, Luxembourg. EUROFOUND. (2016): “Extending working lives through flexible retirement schemes: Partial retirement,” Publications Office of the European Union, Luxembourg. EUROMOMO. (2021): “European mortality monitoring project,” https://www.euromomo.eu/ [accessed on 20 April 2021]. EUROSTAT. (2020): “Eurostat database: Life expectancy by age, sex and educational attainment level (indicator “demo mlexpecedu”),” http://appsso.eurostat.ec.europa.eu/nui/show.do?dataset=demo mlexpecedu&lang=en. FACHINGER, U., AND FRANKUS, A. (2017): “Self-employed people and pension: Is old age poverty the inevitable dark side of an entrepreneurial society?” In Bonnet, J., Dejardin, M., Garcia Perez De Lema D (eds.): Exploring the Entrepreneurial Society: Institutions, Behaviors and Outcomes. Cheltenham: In Press E.E. Publishing. FALL, F., AND BLOCH, D. (2014): “Overcoming vulnerabilities of pension systems,” OECD Economics Department Working Papers, No. 1133. Paris: OECD Publishing. FEHER, C., AND BIDEGAIN, I. (2020): “Pension schemes in the COVID-19 crisis: Impacts and policy considerations,” IMF Special Series on COVID-19, Washington, DC. FORNERO, E., OGGERO, N., AND PUGLISI, R. (2019): “Information and financial literacy for socially sustainable NDC pension schemes,” In: Holzmann, R., Palmer, E., Palacios, R., and Sacchi, S. (eds.), Progress and Challenges of Nonfinancial Defined Contribution Pension Schemes: Volume 2. Addressing Gender, Administration, and Communication. Washington, DC: World Bank, 187–216. FRENCH, E., AND JONES, J. (2017): “Health, health insurance, and retirement: A survey,” Annual Review of Economics, 9(1): 383–409. FURMAN, J., AND SUMMERS, L. (2020): A Reconsideration of Fiscal Policy in the Era of Low Interest Rates. Washington, DC: Discussion draft, Brookings Institution. GANNON, F., LEGROS, F., AND TOUZ E´ , V. (2020): “Sustainability of pension schemes: Building a smooth automatic balance mechanism with an application to the US Social Security,” Revue de l’OFCE, 170(6): 377–401. GAO (U.S. GOVERNMENT ACCOUNTABILITY OFFICE). (2016): Retirement Security. Shorter Life Expectancy Reduces Projected Lifetime Benefits for Lower Earners, GAO-16-354. Washington, DC. GEEROLF, F. (2018): Reassessing Dynamic Efficiency. Los Angeles: Mimeo, UCLA. GEPPERT, C., GUILLEMETTE, Y., MORGAVI, H., AND TURNER, D. (2019): “Labour supply of older people in advanced economies: The impact of changes to statutory retirement ages,” OECD Economics Department Working Papers, No. 1554. Paris: OECD Publishing.

280

Trends in Pension Reforms in OECD Countries

GOLDMAN, D., AND ORSZAG, P. R. (2014): “The growing gap in life expectancy: Using the future elderly model to estimate implications for Social Security and Medicare,” American Economic Review, 104(5): 230–233. GRUBER, J., AND WISE, D. A., EDS. (1999): Social Security and Retirement around the World. Chicago: University of Chicago Press. GRUBER, J., AND WISE, D. A., EDS. (2004) Social Security Programs and Retirement around the World: Micro Estimation. Chicago: University of Chicago Press. GUSTMAN, A., STEINMEIER, T., AND TABATABAI, N. (2012): “Financial knowledge and financial literacy at the household level,” American Economic Review, 102(3): 309–313. HAGELUND, A., AND GRØDEM, A. (2016): “Build your own pension: Framing pension reform and choice in newspapers,” Journal of Ageing & Social Policy, 29(3): 218–234. HERSHEY, D., VAN DALEN, H., CONEN, W., AND HENKEN, K. (2017): “Are “voluntary” selfemployed better prepared for retirement than “forced” self-employed?,” Work, Ageing and Retirement, 3(3): 243–256. HINRICHS, K. (2021): “Recent pension reforms in Europe: More challenges, new directions. An overview,” Social Policy Administration, 55(3): 409–422. HOLZMANN, R., AND HINZ, R. (2005): Old Age Income Support in the 21st Century: An International Perspective on Pension Systems and Reform. Washington, DC: World Bank. HOLZMANN, R., PALMER, E., PALACIOS, R., AND SACCHI, S. (2020): Progress and Challenges of Nonfinancial Defined Contribution Pension Schemes, Volume 1: Addressing Marginalization, Polarization, and the Labor Market. Washington, DC: World Bank. HOMBURG, S. (1990): “The efficiency of unfunded pension schemes,” Journal of Institutional and Theoretical Economics, 146(4): 640–647. HOMBURG, S. (2014): “Overaccumulation, public debt and the importance of land,” German Economic Review, 15(4): 411–435. HURD, M., SMITH, J., AND ZISSIMOPOULOS, J. (2004): “The effects of subjective survival on retirement and Social Security claiming,” Journal of Applied Econometrics, 19(6): 761–775. INSEE. (2016): Les In´egalit´es Sociales Face a` la Mort. Paris: Insee. KAJITANI, K., KINUGASA, T., AND LUN, K. (2018): “Dynamic efficiency in world economy,” Discussion Paper No.1801, Graduate School of Economics, Kobe University. KHANG, Y.-H., BHAK, J., LIM, D., KANG, H.-Y., LIM, H.-K., KIM, Y.-Y., AND PARK, J. H. (2019): “Trends in inequality in life expectancy at birth between 2004 and 2017 and projections for 2030 in Korea: Multiyear cross-sectional differences by income from national health insurance data,” BMJ Open, 9(7): 1–8. KLEVEN, H., LANDAIS, C., AND SØGAARD, J. (2019): “Children and gender inequality: Evidence from Denmark,” American Economic Journal: Applied Economics, 11(4): 181–209. KNOEF, M., BEEN, J., ALESSIE, R., CAMINADA, K., GOUDSWAARD, K., AND KALWIJ, A. (2014): “Measuring retirement savings adequacy: Developing a multi-pillar approach in the Netherlands,” Journal of Pension Economics and Finance, 15(1): 55–89. LALIVE, R., MAGESAN, A., AND STAUBLI, S. (2020): The Impact of Social Security on Pension Claiming and Retirement: Active vs. Passive Decisions, Cambridge, MA: National Bureau of Economic Research. LE DUIGOU, S., LEGENDRE, B., AND SABATIER, M. (2020): “Financial incentives to postpone retirement. Estimating the impacts of the 2003 French pension reform on the employment and wages of older workers by using a lifetime job search model,” Revue de l’OFCE, Presses de Sciences-Po, 170(6): 321–344, https://ideas.repec.org/a/cai/reofsp/reof 170 0321.html. LEIJTEN, F., DE WIND, A., VAN DEN HEUVEL, S., YBEMA, J., VAN DER BEEK, A., ROBROEK, S., AND BURDORF, A. (2015): “The influence of chronic health problems and work-related factors on loss of paid employment among older workers,” Journal of Epidemiology and Community Health, 69(11): 1058–1065. LINDBECK, A., AND PERSSON, M. (2003): “The gains from pension reform,” Journal of Economic Literature, 41(1): 74–112. LIS, M., AND BONTHUIS, B. (2019): “Drivers of the gender gap in pensions: Evidence from EU-SILC and the OECD pension models,” In: Holzmann, R., Palmer, E., Palacios, R., and Sacchi, S. (eds.), Progress and Challenges of Nonfinancial Defined Contribution Pension Schemes: Volume 2. Addressing Gender, Administration, and Communication. Washington, DC: World Bank, 25–53. LOCKWOOD, P. L., ABDURAHMAN, A., GABAY, A. S., DREW, D., TAMM, M., HUSAIN, M., AND APPS, M. A. J. (2021): “Ageing increases prosocial motivation for effort,” Psychological Science, 32(5): 668–681.

281

Herv´e Boulhol et al.

LOICHINGER, E., AND PRSKAWETZ, A. (2017): “Changes in economic activity: The role of age and education,” Demographic Research, 36(40): 1185–1208. LUSARDI, A., AND MITCHELL, O. S. (2014): “The economic importance of financial literacy: Theory and evidence,” Journal of Economic Literature, 52(1): 5–44. MACKENBACH, J., ET AL. (2016): “Changes in mortality inequalities over two decades: Register based study of European countries,” BMJ, 353(1): 1–8. MANOLI, D., AND WEBER, A. (2016): The Effects of the Early Retirement Age on Retirement Decisions. Cambridge, MA: National Bureau of Economic Research. MARSHALL-CATLIN, E., BUSHNIK, T., AND TJEPKEMA, M. (2019): “Trends in mortality inequalities among the adult household population,” Health Reports, 30(12): 11–17. MESA-LAGO, C., AND VALERO, D. (2020): “The new wave of pension reforms in Latin America,” In ´ Peris-Ortiz M., Alvarez-Garc´ ıa J., Dom´ınguez-Fabi´an I., Devolder P. (eds.): Economic Challenges of Pension Systems, Chapter 4, 255–254, Cham: Springer. MUNNELL, A. H., AND AUBRY, J.-P. (2016): “An overview of the pension/OPEB landscape,” CRR Working Papers, No. 2016-11, Center for Retirement Research at Boston College, Chestnut Hill, MA. MURTIN, F., MACKENBACH, J., LASILIONIS, D., AND MIRA D’ERCOLE, M. (2017): “Inequalities in longevity by education in OECD countries: Insights from new OECD estimates,” OECD Statistics Working Papers, 2017/02. Paris: OECD Publishing. NATALI, D., SPASOVA, S., AND VANHERCKE, B. (2016): Retirement Regimes for Workers in Arduous or Hazardous jobs in Europe. A Study of National Policies, European Social Protection Network (ESPN) Brussels: European Commission, https://op.europa.eu/en/publication-detail/-/publication/e23e81809ac9-11e6-868c-01aa75ed71a1#. NISHIYAMA, S. (2019): “The joint labor supply decision of married couples and the U.S. Social Security pension system,” Review of Economic Dynamics, 31(1): 277–304. ¨ NIVALAINEN, S., TENHUNEN, S., AND JARNEFELT , N. (2020): “Carrots, sticks and old-age retirement. A review of the literature on the effects of the 2005 and 2017 pension reforms in Finland,” Nordisk V¨alf¨ardsforskning — Nordic Welfare Research, 5(02): 83–94. O’DEA, C., AND STURROCK, D. (2018): “Subjective expectations of survival and economic behaviour,” IFS Working Papers, W18/14, Institute for Fiscal Studies, London. OECD. (2001): “Fiscal implications of ageing: Projections of age-related spending,” In: OECD Economic Outlook, Vol. 1. Paris: OECD Publishing, Chapter 4, 145–167. OECD. (2014): “Pension communication: Pension statements and national campaigns,” In: OECD Pensions Outlook, 2014. Paris: OECD Publishing, 177–201. OECD. (2016): “Civil service pensions: Toward a unified system with the private sector,” In: OECD Pensions Outlook 2016. Paris: OECD Publishing, Chapter 6, 155–180. OECD. (2017a): Pensions at a Glance 2019: OECD and G20 Indicators. Paris: OECD Publishing. OECD. (2017b): Preventing Ageing Unequally. Paris: OECD Publishing. OECD. (2018a): “Are survivor pensions still needed?,” In: OECD Pensions Outlook 2018. Paris: OECD Publishing, 213–253. OECD. (2018b): “Improving retirement incomes considering behavioural biases and limited financial knowledge,” In: OECD Pensions Outlook 2018. Paris: OECD Publishing, 139–178. OECD. (2019a): Pensions at a Glance 2019: OECD and G20 Indicators. Paris: OECD Publishing. OECD. (2019b): “Will future pensioners work for longer and retire on less?,” OECD Policy Brief on Pensions. Paris: OECD Publishing. OECD. (2019c): Working Better with Age. Ageing and Employment Policies. Paris: OECD Publishing. OECD. (2019d): Financial Markets Insurance and Pensions: Inclusiveness and Finance. Paris: OECD Publishing. OECD. (2020a): Promoting an Age-Inclusive Workforce. Paris: OECD Publishing. OECD. (2020b): “Retirement savings and old-age pensions in the time of COVID-19,” In: OECD Pensions Outlook 2020. Paris: OECD Publishing, 15–50. OECD. (2021): “Towards gender-inclusive recovery,” OECD policy responses to coronavirus (COVID-19), https://www.oecd.org/coronavirus/policy-responses/towards-gender-inclusiverecovery-ab597807/. ORTIZ, I., DURAN-VALVERDE, F., URBAN, S., AND WODSAK, V. (2018): Reversing Pension Privatizations: Rebuilding Public Pension Systems in Eastern Europe and Latin America. Geneva: International Labour Office. PALMER, E. (2000): “The Swedish pension reform model: Framework and issues,” Social Protection Discussion Papers and Notes, No. 23086. Washington, DC: World Bank.

282

Trends in Pension Reforms in OECD Countries

PHILLIPSON C. (2015): “The Political Economy of Longevity: Developing New Forms of Solidarity for Later Life,” The Sociological Quarterly, 56(1): 80–100. ´ PIFARR E´ I AROLAS, H., ACOSTA, E., LOPEZ -CASASNOVAS, G., LO, A., NICODEMO, C., RIFFE, T., ¨ , M. (2021): “Years of life lost to COVID-19 in 81 countries,” Scientific Reports, 11(1): AND M YRSKYL A 3504. PIGGOTT, J., AND WOODLAND, A. (2016): Handbook of the Economics of Population Ageing. North-Holland, Amsterdam: Elsevier. QI, H., HELGERTZ, J., AND BENGTSSON, T. (2018): “Do notional defined contribution schemes prolong working life? Evidence from the 1994 Swedish pension reform,” The Journal of the Economics of Ageing, 12(C): 250–267. QUEISSER, M., AND WHITEHOUSE, E. (2006): “Neutral or fair?: Actuarial concepts and pension-system design,” OECD Social, Employment and Migration Working Papers, No. 40. Paris: OECD Publishing. REICHELT, M., MAKOVI, K., AND SARGSYAN, A. (2020): “The impact of COVID-19 on gender inequality in the labor market and gender-role attitudes,” European Societies, 23(sup1): S228–S245. RUDOLPH, H. (2019): “Pension funds with automatic enrollment schemes: Lessons for emerging economies,” Policy Research Working Paper, No. 8726. Washington, DC: World Bank. SAMUELSON, P. (1958): “An exact consumption-loan model of interest with or without the social contrivance of money,” Journal of Public Economics, 66(6): 467–482. ´ SANCHEZ -MARCOS, V., AND BETHENCOURT, C. (2018): “The effect of public pensions on women’s labor market participation over a full life cycle,” Quantitative Economics, 9(2): 707–733. ´ ¨ SANCHEZ -ROMERO, M., LEE, R. D., AND FURNKRANZ -PRSKAWETZ, A. (2019): “Redistributive effects of different pension systems when longevity varies by socioeconomic status,” NBER Working Papers, No. 25944. Cambridge, MA: National Bureau of Economic Research. SCHARN, M., SEWDAS, R., BOOT, C., HUISMAN, M., LINDEBOOM, M., AND VAN DER BEEK, A. (2018): “Domains and determinants of retirement timing: A systematic review of longitudinal studies,” BMC Public Health, 18(1): 1083. SCHWANDT, H., AND VON WACHTER, T. (2019): “Unlucky cohorts: Estimating the long-term effects of entering the labor market in a recession in large cross-sectional data sets,” Journal of Labor Economics, 37(S1): S161–S198. SCHWARZ, A. M., AND ARIAS, O. S. (2014): The Inverting Pyramid: Pension Systems Facing Demographic Challenges in Europe and Central Asia. Washington, DC: World Bank. SHILLER, R. (2003): “Social Security and individual accounts as elements of overall risk-sharing,” American Economic Review, 93(2): 343–353. SINN, H. W. (2000): “Why a funded pension system is useful and why it is not useful,” International Tax and Public Finance, 7 (4/5): 389-410. ¨ BELMESSER, S. (2003): “Pensions and the path to gerontocracy in Germany,” SINN, H.-W., AND U European Journal of Political Economy, 19(1): 153–158. ¨ STAUBLI, S., AND ZWEIM ULLER , J. (2013): “Does raising the early retirement age increase employment of older workers?,” Journal of Public Economics, 108(1): 17–32. TCRS (TRANSAMERICA CENTER FOR RETIREMENT STUDIES). (2016): “All about retirement: An employer survey,” 17th Annual Retirement Survey, Transamerica Center for Retirement Studies, Los Angeles. TEPE, M., AND VANHUYSSE, P. (2009): “Are ageing OECD welfare states on the path to gerontocracy? Evidence from 18 democracies, 1980–2002,” Journal of Public Policy, 29(1): 1–28. TOMPSON, W. (2009): The Political Economy of Reform: Lessons from Pensions, Product Markets and Labour Markets in Ten OECD Countries. Paris: OECD Publishing. WANG, X., WILLIAMSON, J. B., AND CANSOY, M. (2016): “Developing countries and systemic pension reforms: Reflections on some emerging problems,” International Social Security Review, 69(2): 85–106. WHITEHOUSE, E., D’ADDIO, A., CHOMIK, R., AND REILLY, A. (2009): “Two decades of pension reform: What has been achieved and what remains to be done?,” The Geneva Papers on Risk and Insurance—Issues and Practice, 34(4): 515–535. WISE, D. (2015): Social Security Programs and Retirement around the World. Chicago: University of Chicago Press. WORLD BANK. (1994): Averting the Old-Age Crisis: Policies to Protect the Old and Promote Growth. New York: Oxford University Press. ZAIDI, A., AND WHITEHOUSE, E. (2009): “Should pension systems recognise “hazardous and arduous work”?,” OECD Social, Employment and Migration Working Papers, No. 91. Paris: OECD Publishing.

283

PART III

Income and Economic Growth

15 ECONOMIC GROWTH, INTERGENERATIONAL TRANSFERS, AND POPULATION AGEING Ronald Lee

Abstract Population ageing affects the output and macroeconomy directly because labor force growth is slower and because capital per worker rises, raising productivity and leading to higher wages and lower interest rates. Population ageing also affects the public and private systems that redistribute output across age groups by altering the relative numbers of donors and recipients, typically requiring either that donors give more or that recipients get less. Consequently, population ageing has positive and negative effects on economic well-being at the individual level and alters market and nonmarket distributions of income. The net outcomes differ from country to country depending on demography, institutions, and policies as data from National Transfer Accounts can illuminate. Population ageing is also associated with changes in policy and individual behavior. The low fertility that causes population ageing is associated with higher investments in the human capital of children, raising their productivity. It also releases parental time for increased labor supply (partially offsetting slower growth in the working-age population), home production, or leisure. Longer life may lead to postponed retirement and/or increased saving. Possibly, an older labor force will be less flexible and less creative, leading to slower technological progress and productivity growth. Capital may flow through international markets from capitalrich ageing populations of higher-income countries to labor-rich and capital-poor countries with younger and more rapidly growing populations. In recent years, many new studies have examined these issues theoretically and empirically. Some have focused on Keynesian worries that ageing may bring secular stagnation: Low and declining real interest rates constrain policy options for central banks seeking to stimulate the economy. Slow population growth may reduce investment demand as investors anticipate less growth in demand for their products. The combined effect may mire ageing economies in slow growth and high unemployment.

15.1

Introduction

In the 21st century, falling fertility, slowing population growth, and ageing populations around the world are sparking economic concerns.1 Although these demographic trends are mostly in line with earlier projections, ageing populations are moving into uncharted waters. Japan has DOI: 10.4324/9781003150398-18

287

Ronald Lee

the oldest population of a major nation, but even so its old-age dependency ratio (population 65+/population 20–64) is projected to rise by an additional 60 percent from 2020 to 2060, from 0.52 to 0.83. For high-income countries as a group, the old-age dependency ratio is projected to nearly double by 2100, from 0.31 to 0.60 (United Nations, 2019). Most populations around the world are ageing, although some in earlier stages of the demographic transition are experiencing economically favorable changes in their age distributions that bring “demographic dividends” (Lee, 2003). Recent fertility declines in China and the United States, if not reversed, portend deeper ageing in coming decades than was previously expected, with declining rates of labor force growth or accelerated decline. The changing population age distribution has economic implications because economic needs and behavior vary with age across the life cycle: work, saving, asset ownership, consumption, health, capabilities, caregiving, and need for care. When the population age distribution changes then the relative numbers of workers and consumers, of caregivers and receivers, of savers and dissavers, and owners and users of assets change. These changes create imbalances, stresses, and changed relative prices and incentives throughout the economy and population. From an economic perspective, why should we care? In what sense is population ageing a problem? If individuals did not have economic interactions with one another then the age distributions of the populations in which they lived would be irrelevant. But individuals do interact in many ways, through markets and through social relationships, both public and private. Suppose that each individual accumulated assets by saving and investing during their working years to provide for consumption during retirement, using assets to purchase an annuity. In this case population ageing would lead to more capital relative to labor in the macroeconomy. In a market economy, wages and output per worker would rise while profit rates and interest rates would fall, affecting the economic well-being of workers and the elderly. Per capita income might go up or down. Gross domestic product (GDP) might be higher or lower than it would have been with the same size population but no population ageing. Would any of this be an economic problem? From a certain purist perspective, it would not, because in this hypothetical scenario all economic consequences of ageing arise here through the operation of markets. There are no “externalities,” no divergence between the prices individuals face in the market and the broader social costs and benefits of their actions. Individuals maximizing their utilities will bring society to a Pareto optimal outcome. From a slightly broader perspective, there might be a social concern about changing relative income distribution between workers and the elderly due to these market-mediated changes in wages and asset returns due to population ageing. Of course, the assumption here that individuals interact only through markets is artificial. Other consequences arise through nonmarket systems of intergenerational redistribution or transfers within the family or public sector. Private transfers include parents providing food, shelter, clothing, and so on for their children and the assistance flowing to and from grandparents. Public transfers include education, healthcare, pensions, long-term care, family assistance, and so on. Across regions of the world, 50–60 percent of GDP is transferred intergenerationally within families or through the public sector (Lee and Donehower, 2011, pp. 188–189). Other chapters of this volume discuss many transfer systems, dealing with such topics as healthcare, pensions, long-term care, education, labor supply, finances, and retirement. These interact with population ageing to create budgetary stresses and imbalances as the relative numbers of givers and receivers of transfers shifts. From the purist perspective, all decisions surrounding private/familial transfers are made by individuals who have full knowledge of the consequences, all of which are borne by them, so again no externalities exist. Parents may choose to have fewer children, thereby causing population ageing, in full knowledge of the trade-off between reduced costs of raising fewer children versus having fewer adult children to help them in old 288

Economic Growth, Intergenerational Transfers, and Population Ageing

age. Some may find this purist view difficult to reconcile with the real-world uncertainties of child rearing and survival; power differences within the family; and with the changing values, policies, information, and advice received from governments (e.g., China). Public transfers are a different matter. Parents do not typically consider the consequences of their fertility choices for public pension or healthcare systems, or for the costs of public education. These are external consequences of their actions. In this case, the combined actions of individuals, each seeking his or her own well-being, may not lead to a socially optimal outcome. Individuals may choose to have zero or one child, leading to rapid and deep population ageing and unsustainable pension systems. That would indeed be an economic problem and might merit government action to subsidize the costs of childrearing as many rich nations have done. Population ageing may affect economic growth and the macroeconomy through three main avenues (Cutler et al., 1990). One is the divergence of age profiles of consumption, on the one hand, and income sources from labor and assets, on the other, representing varying degrees of economic dependency across age. This divergence is possible due to public and private intergenerational transfers. Because of it, population ageing raises old-age dependency and reduces support ratios. A second avenue is changes in the relative supply of labor and capital, with population ageing raising the ratio of capital to labor and consequent changes in productivity, wages, and interest rates. For a formal treatment of the trade-off between these first two effects see Arthur and McNicoll (1987). A third arises through potential effects on rates of technological progress and the adoption of new technologies, with ageing perhaps both slowing technological progress and raising the rate of adoption of new technologies. I will consider each of these, emphasizing the first two.

15.2

The Recent Literature

A seminal paper by Cutler et al. (1990) addresses many of the issues that have occupied the literature ever since. The last three authors have continued to contribute important work on macroeconomic consequences, including Poterba (2014), Elmendorf and Sheiner (2000), Sheiner et al. (2006), Eggertsson et al. (2019a), and Rachel and Summers (2019). Other critical summaries of the literature are Weil (2008), National Research Council (2012), and Lee (2016). Recent years have seen a burst of new theoretical and empirical research on the macroeconomic consequences of ageing. These studies take various theoretical and empirical approaches, but their findings are largely but not completely in agreement on some key points. (1) Older populations are associated with higher per capita income, and increases in population ageing are associated with increased growth of per capita income (Acemoglu and Restrepo, 2017; Eggertsson et al., 2019a; Bloom et al., 2021, find that lower fertility and higher life expectancy raise growth). (2) Population ageing is associated with slower growth of output (Eggertsson et al., 2019a; Aksoy et al., 2019; Gagnon-Bartsch et al., 2021; Goodhart and Pradhan, 2020). (3) Population ageing is associated with an increase in capital per worker (Eggertsson et al., 2019a; Auclert et al., 2021). (4) Population ageing reduces real interest rates through both the demand and supply side of funds for investment (Aksoy et al., 2019; Rachel and Summers, 2019; Eggertsson et al., 2019a,b; Gagnon-Bartsch et al., 2021; Auclert et al., 2021). Goodhart and Pradhan (2020) argue that population ageing will raise inflation and thereby raise nominal interest rates, which is not necessarily inconsistent with falling real rates of interest. Acemoglu and Restrepo (2017) find that economies with population ageing adopt more labor-saving robotics, which they suggest could help explain the null or positive association of ageing with more rapid per capita income growth. Eggertsson et al. (2019a) argue that the 289

Ronald Lee

positive or neutral association of population ageing with high or rising per capita income holds only for economies with interest rates somewhat above zero. Looking separately at countries with interest rates near zero, they find that ageing reduces per capita income growth, consistent with the secular stagnation argument that such low interest rates deprive central banks of their most potent weapon for stimulating the economy due to a zero lower bound. In line with this theme of possible multiple equilibria, Fanti et al. (2013) endogenize fertility in a Solow growth model with detailed population age distribution and find five possible stable steady states, some with very low fertility and hence very old populations.

15.3

The Age-Distributed Economy

In what follows I will discuss in greater detail many of these topics and later illustrate them drawing on data from National Transfer Accounts. At the most aggregated level, in a closed economy where private and government variables are combined, we have the identity between income on the left and its allocation on the right, Yl + Ya = C + S,

(1)

which says that total labor income plus asset income equals consumption plus saving. At the level of individuals, though, this identity looks different because the incomes of some are augmented by receipt of transfers, while other individuals may transfer income to others, including through paying taxes. A transfer is an economic flow that involves no explicit quid pro quo. If we let lowercase letters represent individual level flows, with subscripts l, a, f , and g denoting labor, assets, family, and government, and τ denoting transfers with superscripts + and − denoting inflows (transfers received) and outflows, then at the individual level we have yl + ya + τf+ + τg+ = c + s + τf− + τg− .

(2)

Consumption includes both private expenditures and consumption of publicly provided goods and services. Labor income includes wages and salaries, fringe benefits, and labor’s share of self-employment income. Asset income can be positive or negative and includes interest paid or received, dividends, rents, and profits; borrowing, lending, and repaying; and the services received from an owned home. Saving by individuals and by the public sector may be positive or negative. Public saving or dissaving is allocated to individuals in proportion to their tax payments. For present purposes, public transfer outflows include taxes, social contributions, and grants paid to federal, state, and local governments making possible transfers including public and quasi-public goods. Private transfer outflows are made from current income (and thus exclude capital transfers) and occur mostly within a household of co-residing family members but also occur between households. Equation (2) applies to any individual, but we will use it here for averages of all individuals of the same age. For example, yl (x) is labor income at age x, averaged across all males and females in the population whether the individual labor income is zero or positive (Lee and Mason et al., 2011; United Nations, 2013). If we sum (2) across all individuals or all age groups we recover (1) because a dollar of transfer outflow for one individual must be a dollar of transfer inflow for some other individual recipient, so all transfers add up to zero. Public transfers sometimes result in public borrowing or surplus, and in open economies, international transfers may come into play (United Nations, 2013). The components of (2) are estimated for many countries around the world by the National Transfer Accounts project, or NTA, as will be discussed later. 290

Economic Growth, Intergenerational Transfers, and Population Ageing

15.4

Population Ageing, Economic Output, and Its Primary Distribution

Consider the left side of (2) and the effect of population ageing and slowing population growth on economic output and its primary distribution through markets. A useful question is “Should an ageing population save more to help fund the increasing consumption needs of ageing retirees?” Generally not, unless saving would be too low in the absence of ageing (Cutler et al., 1990). Output, the productivity of labor, and the level of per capita income all depend on the ratio of capital to labor in a basic economic model. With the slower population growth associated with population ageing, a constant saving rate would lead to a higher capital labor ratio, because a smaller share of savings would be needed to equip new workers (“capital widening”) and more could be used to equip existing workers (“capital deepening”) (Solow, 1956). A planner seeking to maximize per capita consumption (income less savings) would choose both a lower saving rate and a higher capital-labor ratio for an ageing population. In this sense population ageing and slower population growth bring per capita economic benefits: higher output per worker. At the same time, slower labor force growth would indeed bring slower GDP growth, even if output per worker rises. In the long run, the GDP growth rate varies one to one with the growth rate of both population and labor (Solow, 1956, unless we entertain the endogenous growth theory discussed later). If we are concerned with GDP growth and size rather than per capita amounts, then population ageing is a worry. One reason for such a concern could be geopolitical power if power derives from position in the international size ranking of GDP. Another reason could be that slower expected GDP growth might reduce investment demand and bring low interest rates and secular stagnation (Summers, 2015; Eggertsson et al., 2019a; Rachel and Summers, 2019), limiting the options of central banks. Now let us bring age distribution into the discussion. Labor supply varies strongly with age, in both quantity and quality. Quality variations reflect changes with age in experience, vigor, self-discipline, flexibility, and cognition. They also reflect differences in the amount of education different generations received when young. Quantity variations reflect the impact of fertility on female labor supply, the levels of educational enrollment at young adult ages, and retirement behavior, which is strongly affected by pension structures (Gruber and Wise, 1999). In each year cross-sectional labor income by age reflects all these factors and more. Asset holdings also vary strongly with age. In the simplest life-cycle saving story, workers save and accumulate assets throughout their working years and then dissave after retirement to fund their consumption. More realistically, individuals may receive inheritances or capital transfers from parents or grandparents throughout their lives, augmenting any saving from labor income. Relative prices of assets may rise or fall, sometimes dramatically. Individuals may save to leave bequests at death to their descendants. They may save as a buffer against unforeseen costs such as health shocks. In societies with public pension and/or healthcare programs or strong traditions of familial support of the elderly, individuals may reduce their saving rates in anticipation of old-age support from these other sources. For all these reasons, considering actual patterns of asset holding and savings across ages, rather than relying on simplistic assumptions about saving behavior, is important. The possibility that population ageing may contribute to the risk of secular stagnation has drawn increasing attention in recent years (Summers, 2015; Rachel and Summers, 2019; Eggertsson et al., 2019a). Eggertsson et al. report that the real rate of interest declined by 4 percentage points from 1970 to 2015. They develop and simulate an overlapping generations model under various counterfactual assumptions for the United States to decompose the sources of that decline. They conclude that the decline in fertility and mortality contributed far more 291

Ronald Lee

than any other factor (rising government debt, declining productivity growth, or several other factors they consider). This result is consistent with the hypothesis that population ageing has played an important role in secular stagnation.

15.5

Population Ageing and the Intergenerational Redistribution of Income

Having considered the effects of population ageing on output and its primary distribution, we now turn to the secondary redistribution of income through transfers, which may take the form of money or of goods and services. Government transfers include those targeted to individuals, such as public assistance, public education, publicly provided healthcare, long-term care, public pensions, and nontargeted in-kind transfers such as the use of roads, police protection, national defense, and so on. Private transfers occur mainly within the family, including the private costs of raising a child, such as food, housing, clothing, childcare, and private education. They also include help that grandparents may give to or receive from their adult children and grandchildren. Mostly these are intrahousehold transfers, but some occur between households. As populations age, the increasing number of elderly recipients of public or private net transfers relative to the number of workers who are the main source of these transfers, puts budgetary pressure on either the family (for private transfers, as in East Asia) or on government budgets, particularly for public pensions, healthcare, and long-term care. In many countries the elderly make net private transfers to younger people, in which case ageing will ease private budgetary pressures, as it does for public budgets in some developing countries where the elderly pay more in taxes than they receive in public benefits. At the same time, the costs of public and private transfers to children may fall with population ageing as the population share of children shrinks. I say “may fall” because as fertility declines (leading to population ageing) families and governments typically choose to increase investment in education and health per child.

15.6

Insights from National Transfer Accounts

As noted earlier, National Transfer Accounts (see Mason in this volume and the project website ntaccounts.org) measure the individual components of (2) and much more. In what follows we will look at more summary measures of the age profiles, but here Figure 15.1 shows the age profiles themselves for the United States in 2015. Sources of Total Income Received

(a)

Uses of Total Income

(b)

120000 110000 Private transfers received

80000

asses income

Public transfers received

60000 40000

Public transfers given

90000

Labor income

2011 US Dollars

2011 US Dollars

100000

Private transfers given

70000 50000

Consumption

30000

20000 10000 0

0

10

20

30

40

50

60

70

80

90

Age

-10000 0

Saving

10

20

30

40

50

60

70

80

90

Age

Figure 15.1 Total per capita income received by individuals by age and by source and uses made of it, United States (2015). Source: National Transfer Accounts data extracted from ntaccounts.org.

292

Economic Growth, Intergenerational Transfers, and Population Ageing

Panel A gives the primary per capita income distribution by age, the sum of labor income and asset income (ignoring for now the two tranches representing public and private transfers). At young adult ages, this is entirely labor income; in middle adult ages, it is a mixture with labor income predominant; and at older ages, it is a mixture with asset income predominant. Children contribute virtually nothing to national income, while the elderly contribute much more through asset income than a picture based solely on labor income would indicate. We also see the secondary income distribution in Panel A. Private and public transfers are received at every age. Private transfers received tend to be larger in childhood and smaller in old age. Public transfers jump for children when they reach kindergarten age at 5, are lower in middle years, and then rise dramatically between age 62 and 70 as people qualify for public healthcare at age 65 and become eligible for public pensions, which may be initiated starting at age 62. We see that income from transfers is very important at both younger and older ages. In Panel B we see how this income is allocated to uses at each age. Children’s income goes almost entirely for consumption, which is to be expected because they are children, but also because it is almost entirely received in kind rather than in cash. Young adults borrow (have negative savings), but after the mid-30s adults begin to have positive saving that peaks in the late 50s. After the mid-70s, individuals begin to dissave. Although it is a bit difficult to tell due to savings, consumption rises very strongly with age. More detailed data would show that much of the rise in consumption with age is due to increased consumption of healthcare, largely in the form of in-kind public transfers. During the middle years, public and private transfers given are similar. Familial transfers mainly reflect the costs of rearing children, while public transfers are mainly taxes and contributions. The elderly continue to make public transfers, mostly through taxes on non-labor income. Strikingly, they also make large private transfers mainly to their children and grandchildren. From year to year, these age profiles change because of macroeconomic fluctuations, rising productivity of labor, and changing rates of return on assets. Individuals may make different decisions about what to give to whom and how much to save or consume, and government policies and programs will change. Nonetheless, the shapes of these profiles remain quite stable, as comparisons of NTA show for different years for the same country and different countries in the same year, although these latter also reveal strong national and regional differences (Lee and Mason et al., 2011). The age profiles shown are average values for each age group, so if they are multiplied by the population at each age they give the corresponding aggregate value at that age. For example, in the United States in 2015 a total of 4,355,000 people were aged 28. On average, 28-year-olds had labor income of $41,672. Multiplying we find that as an age group they received $181.5 billion. The sum over all age groups is the total labor income of the United States in 2015. We can do the same thing for each item in the two graphs, finding (where uppercase letters indicate a total for the U.S. economy): Yl , Yk , Tf+ , Tg+ , S, C, Tf− , and Tg− . When we do this for 2015, the sums for private and public total transfer inflows and outflows should each be equal and cancel, and the identity Yl + Ya = C + S should hold. But if we now repeat the exercise with the population age distributions of 2016, 2017, . . . , 2070, these conditions will no longer be met, and the quantities that must be equal in the base year, 2015, may increasingly diverge. This divergence is informative. It locates and assesses the effects of the projected future population change. Such results tell us how large the necessary adjustments will be and pinpoint the sources of imbalances (when this is done with greater public program detail) and whether the projected population age distribution changes in themselves would lead to tightening or relaxing of budgets. The proportional change needed in total consumption 293

Ronald Lee

to achieve balance is a convenient metric for the total size of needed adjustments, but actual adjustments could come through any of the other variables, e.g., by raising Yl through later retirement. I will carry out these calculations for the aggregated variables, but for transfers, I will combine the inflows and outflows to get net familial and net government transfers at each age and then aggregate these. For some future year t the aggregated variables are referred to as “effective” values. For example, Yl,2023 is “effective labor” in 2023. I refer to this as labor rather than labor income because we can think of the labor income in 2015 as describing the efficiency-weighted quantity of labor a person at each age supplies. Adding time subscripts and letting t0 be the baseline year, for that year we have Yl,t0 + Ya,t0 = Ct0 + St0 , Tf ,t0 = Tg,t0 = 0.

15.7

(3)

Summary Measures of Imbalances Created by Population Ageing

We can also construct some simple indices of imbalance. The most widely used index, at least in NTA, is the weighted support ratio (SR) defined as effective workers per effective consumer (Cutler et al., 1990, p. 9; Mason et al., 2017), that is, SRt = Yl,t /Ct .

(4)

This measure is intuitive and useful, particularly its proportional change over time, reflecting the way that changing population age distribution affects the number of workers available to support the number of consumers, assuming the age schedules of labor and consumption remain unchanged. With population ageing, the SR typically declines. During the demographic transition, after the start of fertility decline, the SR typically rises for some decades as the proportion of the population in the working ages increases, giving rise to the so-called “demographic dividend.” Mason et al. (2017) present estimates and projections of SR for 186 countries based on NTA. However, asset income and labor income are both available to fund consumption. While younger and middle-aged adults mainly bring their labor to the economy, the elderly mainly bring their asset holdings (see Figure 15.1, Panel A). As populations age, assets and perhaps asset income should rise relative to labor and labor income as found in the empirical studies cited earlier. If we take the age profile of asset ownership as given at baseline levels and assume that the rate of return on assets is equal across age, then we can estimate the increase in assets relative to labor using the NTA asset income and labor income age profiles. The portion of asset income that is not saved is available to support consumption. This leads us to construct the general support ratio (GSR) as GSRt =

Yl,t + Ya,t − S . Ct

(5)

At time t0 , GSRt0 = 1.0. A companion measure is the “transfer load” or TL, which for private transfers is TLf ,t = Tf ,t /Ct (6) and similarly for government transfers. At t0 the transfer load is 0, because inflows and outflows cancel at baseline. Thereafter, however, population ageing leads to increases in the TL in some 294

Economic Growth, Intergenerational Transfers, and Population Ageing

countries and to decreases in others. The TL tells us the proportional change in consumption that would be needed to balance the transfer system in a future year. Once again, this is a convenient metric although actual adjustments could be made in other variables as well. The GSR and TL are related in each year by GSRt = 1 − TLt .

(7)

A rising transfer load implies a falling general support ratio, and variations of the transfer load about zero should inversely mirror variations in the general support ratio about unity. This symmetry arises because only through transfers can consumption by age diverge from labor income plus asset income not saved. The GSR indicates the changing degree to which effective primary income will support effective consumption for an open economy where wages and asset returns are set on the international market. In a closed economy, however, we would expect that increasing capital relative to labor would lead to rising wages and falling interest rates.2 Expecting the actual outcome to be something between these two would be realistic, but here we will stick with the open economy assumption for simplicity. The estimates that follow in the next section are not dynamic though they are plotted against time. Rather, they are a series of cross-sectional static calculations reflecting the changing population age distribution. For example, the asset income bears no relation to savings in earlier years. It depends only on the baseline age profile of asset income and the population age distribution. If we wanted to make a real forecast of future total labor income and other variables, we could assume some rate of productivity growth for labor, such as 1.5 percent per year, and shift the profiles upward by that amount each year. But our purpose here is to isolate the effect of population change alone.

15.8

Estimated Impact of Population Ageing on Economic Flows Over the Next 50 Years

With that introduction and those caveats, we can now look at estimates. We start with the component aggregate variables over time for a few selected countries and then turn to the summary indices. Figure 15.2 shows aggregates for the United States, Japan, Germany, Philippines, and Brazil, all adjusted to a base year of 2020 and projected to 2070, countries chosen to reflect both regional diversity and diversity in the strength of public transfer programs for the elderly.3 These five countries give some idea of the range of possibilities. The United States is projected to have relatively modest population ageing, while ageing in Japan will be much more severe and its population is already declining. The United States has a modest public pension system while Japan has a more generous one. In these figures, the heavy solid line is primary income, that is, labor income plus asset income, while the heavy dashed line is consumption plus savings. For the United States, the lines are virtually on top of one another and very flat, indicating that population ageing will have very little effect on either and that the costs of population ageing will be very low. The decline in labor income is offset by a rise in asset income. Savings remain near zero throughout. Public transfer costs will rise modestly while private transfer costs decline modestly, this last because the elderly on net give transfers to younger people. Little adjustment will be necessary, only 2 percent of consumption by 2050. In Japan, the situation is quite different. Labor income per capita declines strongly with only a small offset from rising asset income. Effective primary income will fall, while effective 295

Ronald Lee

consumption plus saving will rise, leading to an imbalance of 15 percent. Maintaining the identity would require an adjustment equivalent to reducing consumption by 15 percent (at every age) below what it would otherwise be (e.g., given trends in productivity). Savings are moderate but remain flat. Public transfers, however, will rise strongly due to the generous public programs for the elderly (pensions, healthcare, and long-term care) while net private transfers remain near zero. Germany is quite like Japan. Primary output drops rapidly while consumption drops only modestly, leading to a large imbalance of 16 percent. Labor income drops more sharply than in Japan, and asset income does not rise at all. As in Japan, savings are flat and public transfers increase very substantially. Brazil is a different story. Labor income declines only slightly while asset income rises strongly, leading to a strong increase in primary income by 27 percent. Consumption rises only modestly but savings rise strongly, resulting in a 31 percent increase in C+S. While public transfers increase strongly due to the generous pension program, private transfers decrease strongly because Brazilian elderly make large transfers to their children and grandchildren, so net transfers remain quite flat. Overall, an imbalance of about 4 percent must be addressed, much less than might be expected given their steep demographic ageing and their generous public transfers. The Philippines is again different. Both labor income and asset income rise strongly, leading to a 42 percent increase in primary income. At the same time, consumption and savings rise rather moderately, by 23 percent in sum. Consequently, there is a beneficial imbalance of 16

Figure 15.2 Projected changes in population age distributions interact with economic age profiles to generate differing budgetary stresses across nations. Lines represent aggregate amounts divided by population, 2020–2070. Notes: The variables are all per capita (aggregate totals divided by population size), as follows: C for consumption; YL for labor income; YA for asset income; S for saving; TF for private transfers; TG for public transfers; Yl+Ya for total income; and C+S for consumption plus saving. Because the age profiles are taken from earlier years for which NTA are available, Yl+Ya is not necessarily equal to C+S as it would be in the true base year. Source: Calculated from National Transfer Accounts data extracted from ntaccounts.org and United Nations (2019).

296

Economic Growth, Intergenerational Transfers, and Population Ageing

percent, a sizable demographic dividend. This occurs because both public and private transfers decline, because the elderly receive less in both public and private transfers than they give to those younger. Implicit in these results are increases in the ratio of asset income to labor income under the assumption of an open economy, with wages and asset returns determined on the international market. The ratio in 2070 in the United States is 1.21 times that in 2020. The corresponding figures for the other countries are Japan 1.22, Germany 1.27, Brazil 1.86, and Philippines 1.42. These increases help to offset the rising dependency ratio (falling support ratio) as the population ages. In a closed economy these increases would bring rises in wages and declines in profit rates and interest rates, but this case is not considered here.

15.9

Summary Indices: General Support Ratio and Transfer Load

Having looked in detail at how ageing affects the economies of those five countries, we will now turn to the summary indices discussed earlier, the GSR and the TL. These ways of viewing the same information clarify the role of transfer systems in mediating the economic impact of ageing and simplify comparison of a broader set of countries. We start with the private transfer load, expressing the change in effective aggregate net transfers as a ratio to effective aggregate consumption. This is identically equal to zero in 2020 but varies as time passes. Figure 15.3 plots data for 17 NTA countries selected to provide a range of transfer patterns and regional diversity. Figure 15.3A shows the private TL, which typically becomes negative as time passes and populations age, because in most cases the elderly make net transfers to their children and grandchildren. A relative increase in the elderly population means transfers given increase relative to those received and therefore the negative TL. However, in some NTA countries, mainly in East Asia, working-age adults give net support to their elderly parents. In this case population ageing raises transfers received more than those given, raising the private TL. In China the private TL rises most, by 23 percent up to 2050, perhaps reflecting the great difference in labor income between old and young. In the Republic of Korea and Japan it also rises, but by much less. By contrast, the private TL falls strongly in India, Indonesia, Mexico, Brazil, and the Philippines where the elderly make substantial net transfers to their descendants. Some of these countries, like Indonesia and the Philippines, lack robust public pension programs and the elderly consume and make net transfers out of their asset income. Some other countries, like Brazil, do have generous public pensions that are used to fund net transfers to their adult children. The public transfer load generally rises in countries due to pensions and healthcare for the elderly, but in some countries without such programs it falls because the elderly pay more in taxes than they receive in benefits. In Figure 15.3B it rises in all 17 countries shown except the Philippines, where it falls by 4 percent. In some other countries, there is very little change, such as India and Indonesia. By far the biggest increase is in Brazil at 27 percent, with Costa Rica, Spain, Hungary, and the Republic of Korea at 16–17 percent, followed by Chile, Germany, Japan, Sweden, and Slovenia all at around 11–13 percent. These are substantial increases, indicating the unsustainability of these programs as currently structured. For example, other things equal (which they will not be!), Brazil would have to reduce its consumption by 27 percent below what it would otherwise be in 2070 or make other adjustments of corresponding magnitude. The total transfer load is the sum of the public and private loads, shown in Figure 15.3C. Summing them assumes they are in some way fungible so that a government can raise taxes to capture the private savings arising from fewer children per household, for example. The 297

Ronald Lee

Figure 15.3 Projected changes in private, public, and total transfer load from baseline zero in 2020 to 2070 for selected countries, not all labeled. Notes: Changes are expressed as a proportion of total consumption in each year. Source: Calculated from National Transfer Accounts, ntaccounts.org, and United Nations (2019).

countries with the greatest rise in the total transfer load are China, Republic of Korea, and Brazil at 20–30 percent, followed by Spain, Hungary, Japan, and Slovenia at 15–18 percent, and Chile, Costa Rica, Germany, and Sweden at around 12 percent. India, Indonesia, Philippines, and Mexico all have falling ratios in the range of –8 percent to –13 percent. The United States and Uruguay have small increases. The general support ratio is 1 minus the total transfer load, so it is not shown. It gives results for the size of necessary adjustments that are identically equal to the transfer load. If fertility is high and population growth is rapid, then the population will be young, and the dependency costs of the elderly will be minimized. However, in this case a higher proportion of output will have to be invested in capital to provide for the growing population and labor force, which will be costly. If fertility is low, the population will be older, and elder dependency will be more costly while less output will have to be invested in capital. In between may be some level of fertility that would maximize consumption (Samuelson, 1975). Given the NTA age profiles for each country, the fertility level that maximizes consumption in steady state can be found, assuming a capital-output ratio of 3 is maintained, or assuming that the ratio is chosen to maximize consumption. For most countries it is a total fertility rate between 1.6 and 1.8, depending on the systems of transfers and the age shapes of consumption and labor income (Arthur and McNicoll, 1987; Lee and Mason et al., 2014) . This suggests that countries should not be concerned if fertility is modestly below replacement, even if it poses problems for the public sector. 298

Economic Growth, Intergenerational Transfers, and Population Ageing

15.10

Other Aspects of Population Ageing in Relation to Economic Growth

While the foregoing calculations capture in realistic detail many economic impacts of changing population age distributions, they are based on the assumption that baseline NTA age profiles remain unchanged in the future, other than a possible upward trend induced by productivity growth. There are reasons to expect important changes in economic behavior by age, some of them closely linked to population ageing, and a possibility also exists that technological progress is connected to population ageing. Likewise, the macroeconomic environment, including unused capacity in labor and capital, may be affected by population ageing as in secular stagnation. These possibilities complicate the question whether population ageing will accelerate or decelerate the growth of GDP or per capita GDP. 1. Slow population growth and technological progress: The rate of technological progress has been the dominating force behind economic growth. An influential body of research suggests that population growth is an important driver of technological progress because technology is a non-rival good, and with larger populations the investments by firms in research and development will be larger. For example, Jones (2022) concludes that in a declining population, income and technology would stagnate. However, assessing this theory empirically is extremely difficult. Acemoglu and Restrepo (2017) find that in ageing populations labor scarcity leads to more rapid adoption of labor-saving technologies such as robotics, but this finding relates to adoption and not creation of new technologies. 2. Population ageing and human capital intensification: The very low fertility that leads to slow population growth and population ageing is typically associated with increased investments in human capital per child (Galor and Weil, 2000). This could happen because some independent force, such as improved access to contraception, reduces fertility, making couples better able to afford to invest in the human capital of their reduced number of children. It could also happen because couples reduce their fertility to invest more in each child. Or because some third factor, such as rising incomes, leads them to choose both lower fertility and higher investments in each child, as in the quantity-quality theory of fertility (Becker and Lewis, 1973). Regardless of the reason for the association, population ageing goes with increased education of offspring, which in turn raises the productivity of labor. The higher quality of labor may then offset its reduced quantity (Lee and Mason, 2010). 3. Productivity of an ageing labor force: Will an older labor force be less flexible, mobile, innovative, and dynamic, and therefore lead to reduced productivity? Or will an older labor force be more experienced, disciplined, and reliable, and therefore more productive? Empirical findings are mixed on this topic. Simple calculations based on the age profile of labor income suggest that any effects will be minimal (National Research Council, 2012, p. 116). A critical review of the literature concludes that “there is likely to be a negligible effect of the age composition of the labor force on aggregate productivity over the next two decades” (National Research Council, 2012, p. 120) 4. Low fertility and parental labor supply: Parents, and particularly mothers, devote substantial time to caring for their young children. For this reason, we might expect that lower fertility would free up some parental time that might be used for increased market labor. This may seem obvious, but empirical work is complicated by the possibility that couples reduce fertility to work more rather than work more because their fertility is lower. A study by Bloom et al. (2009) uses changes over time in national abortion policies as an instrument and concludes that fertility decline causes an increase in female labor supply to the market and thereby boosts economic growth. This increase in participation rates will partially offset the reduction in the working-age population. 299

Ronald Lee

5. Pension incentive structures, retirement, and elder labor supply: A large body of research establishes that design of defined benefit pensions can create strong incentives to retire early and that reducing or eliminating these incentives leads to later retirement and increased labor supply by the elderly. An influential collection of studies by Gruber and Wise (1999) focuses on nine high-income countries, finding that the structure of defined benefit pension plans could be summarized by the implicit tax on continuing to work, which was found to be strongly correlated with labor supply of men at older ages. In part because of pension reforms and in part for other reasons such as rising life expectancy and often improving health and vigor at older ages, retirement ages have risen by an average of 2 years in Organisation for Economic Co-operation and Development countries over the past 25 years (Boissonneault et al., 2020). Increased labor supply at older ages will help to offset the reduction in the growth of the working-age population.

15.11

Research Directions

As Acemoglu and Restrepo (2017, p. 179) suggest, a need exists for “work that systematically investigates the relationship between demographic change and GDP growth as well as the channels via which this relationship works.” Earlier I briefly described some of the recent contributions to the literature on this point. This work is well founded in theory, and the empirical work generally seeks to explain changes over the past few decades within a single country or in an international panel. Further work building on these studies would be welcome. On the one hand, none of these studies is causal. They are all associational. Might some sort of causal analysis be feasible, for example along the lines of Bloom et al. (2009) using an instrument for fertility? On the other hand, when studies attempt to make realistic numerical estimates of effects, more careful attention to the underlying demography might be useful. For example, Eggertsson et al. (2019a) use an overlapping generations model with 56 ages and actual survival rates drawn from U.S. mortality, yet it assumes that no one survives beyond age 81. Recent period life tables for the United States indicate that more than 50 percent of births survive past age 81. Similarly, some of these studies assume that adulthood is reached in the mid-20s and that children in the household at earlier ages have no effect on consumption by their parents’ households. Such assumptions and other simplifications might or might not matter for qualitative results but would surely affect numerical estimates. Sanchez-Romero (2013) and S´anchez-Romero et al. (2018) provide examples of overlapping generations modeling with realistic demography for the macro effects of a demographic transition. Most high-income and upper-middle-income countries have pay-as-you-go public pensions, but coverage and generosity of benefits vary widely. Knowing how these features of unfunded public pensions affect saving behavior of workers and retirees, aggregate asset holdings, capital intensity of the economy, and perhaps international capital flows, would be useful. Research would be very welcome on whether population ageing influences the rate of technological progress, its rate of adoption, and any bias toward labor or capital. A start has been made on these questions but the first one—development of new technology—remains particularly difficult to investigate empirically. Because most countries have embarked on the demographic transition and have either arrived at the stage of population ageing or will arrive there soon, thinking about ageing in the context of international markets for capital, goods and services, and labor is important. Analyzing the economic consequences of population ageing within a single country, even a large one like the United States or China, requires careful consideration of the global context of ageing. 300

Economic Growth, Intergenerational Transfers, and Population Ageing

Acknowledgment Support was received from the Center for the Demography and Economics of Ageing (CEDA) through a grant to CEDA from the US National Institute of Ageing P30AG012839.

Notes 1 “Yet demographic stagnation could bring its own high costs, among them a steady reduction in dynamism, productivity and a slowdown in national and individual prosperity, even a diminishment of global power” Manjoo (2021). 2 We could model output using a Cobb-Douglas production function with constant returns to scale and inputs of labor and capital based on the effective labor and effective capital calculations for each year. We could then calculate changes in wages and the return to capital in the usual way. These could then be viewed as multipliers for the age profiles, each with an initial value of unity. 3 The actual NTA base years are earlier than 2020 for all countries. Consequently some of the accounting identities are not satisfied when applying the age profiles to the 2020 population age distributions.

References ACEMOGLU, D., AND RESTREPO, P. (2017): “Secular stagnation? The effect of ageing on economic growth in the age of automation,” American Economic Review: Papers & Proceedings, 107(5): 174–179. Available at https://doi.org/10.1257/aer.p20171101. AKSOY, Y., BASSO, S. H., SMITH, R. P., AND GRASL, T. (2019): “Demographic structure and macroeconomic trends,” American Economic Journal: Macroeconomics, 11(1): 193–222. Available at https://doi.org/10.1257/mac.20170114. ARTHUR, B. W., AND MCNICOLL, G. (1978): “Samuelson, population and intergenerational transfers,” International Economic Review, 19(1): 241–246 AUCLERT, A., MALMBERG, H., MARTENET, F., AND ROGNLIE, M. (2021): “Demographics, wealth, and global imbalances in the twenty-first century,” NBER Working Paper No. 29161. Available at http://www.nber.org/papers/w29161. BECKER, G., AND LEWIS, H. G. (1973): “On the interaction between the quantity and quality of children,” Journal of Political Economy, 81(2): S279–S288. BLOOM, D. E., CANNING, D., FINK, G., AND FINLAY, J.E. (2009): “Fertility, female labor force participation, and the demographic dividend,” Journal of Economic Growth, 14: 79–101. Available at https://doi.org/10.1007/s10887-009-9039-9. BLOOM, D. E., KHOURY, A., KUFENKO, V., AND PRETTNER, K. (2021): “Spurring economic growth through human development: Research results and guidance for policymakers,” Population and Development Review, 47(2): 377–409. BOISSONNEAULT, M., MULDERS, J. O., TUREK, K., AND CARRIERE, Y. (2020): “A systematic review of causes of recent increases in ages of labor market exit in OECD countries,” PLOS ONE, Available at https://doi.org/10.1371/journal.pone.0231897. CUTLER, D., POTERBA, J., SHEINER, L., AND SUMMERS, L. (1990): “An ageing society: Opportunity or challenge?,” Brookings Papers on Economic Activity, 1: 1–56, 71–73. EGGERTSSON, G. B., LANCASTRE, M., AND SUMMERS, L. H. (2019a): “Ageing, output per capita, and secular stagnation,” American Economic Review: Insights, 325–342. Available at https://doi.org/10.1257/aeri.20180383 325. EGGERTSSON, G. B., MEHROTRA, N. R., AND ROBBINS, J. A. (2019b): “A model of secular stagnation: Theory and quantitative evaluation,” American Economic Journal: Macroeconomics, 11(1): 1–48. Available at https://doi.org/10.1257/mac.20170367. ELMENDORF, D. W., AND SHEINER, L. M. (2000): “Should America save for its old age? Fiscal policy, population ageing, and national saving,” Journal of Economic Perspectives, 14(3): 57–74. FANTI, L., IANNELLI, M., AND MANFREDI, P. (2013): “Neoclassical growth with endogenous age distribution. Poverty vs low-fertility traps as steady states of demographic transitions,” Journal of Population Economics, 26(4): 1457–1484. Available at http://www.jstor.org/stable/43738955. ´ GAGNON, E., JOHANNSEN, B. K., AND LOPEZ -SALIDO, D. (2021): “Understanding the new normal: The role of demographics,” IMF Economic Review, 69(2): 357–390. Available at https://doi.org/10.1057/s41308-021-00138-4.

301

Ronald Lee

GALOR, O., AND WEIL, D. N. (2000): “Population, technology, and growth: From Malthusian stagnation to the demographic transition and beyond,” The American Economic Review, 90(4): 806–828. GOODHART, C., AND PRADHAN, M. (2020): The Great Demographic Reversal: Ageing Societies, Waning Inequality, and an Inflation Revival. Cham, Switzerland: Palgrave Macmillan. Available at https://doi.org/10.1007/978-3-030-42657-6. GRUBER, J., AND WISE, D. A. (1999): “Introduction and summary.” In: Gruber, J., and Wise, D. A. (eds.), Social Security and Retirement around the World, Chicago: University of Chicago Press, pp. 437–474. JONES, C. I. (2022): “The end of economic growth? Unintended consequences of a declining population,” The American Economic Review, 112(11): 3489–3527. LEE, R. (2003): “The demographic transition: Three centuries of fundamental change,” Journal of Economic Perspectives, 17(4): 167–190. LEE, R. (2016): “Macroeconomics, ageing and growth.” In: Piggott, J., and Woodland, A. (eds.), Handbook of the Economics of Population Ageing, Amsterdam, Netherlands and Oxford: Elsevier, chapter 2, pp. 59–118. LEE, R., AND DONEHOWER, G. (2011): “Private transfers in comparative perspective.” In: Lee, R., and Mason, A. (eds.), Population Ageing and the Generational Economy: A Global Perspective, Cheltenham and Northampton, MA: Edward Elgar, chapter 8. Available at http://www.idrc.ca/ EN/Resources/Publications/Pages/IDRCBookDetails.aspx?. LEE, R., AND MASON, A. (2010): “Fertility, human capital, and economic growth over the demographic transition,” European Journal of Population, 26(2): 159–182. LEE, R., MASON A., PRINCIPAL AUTHORS AND EDITORS, AND MEMBERS OF THE NTA NETWORK (2011): Population Ageing and the Generational Economy: A Global Perspective. Cheltenham: Edward Elgar. Available at http://www.idrc.ca/EN/Resources/Publications/Pages/ IDRCBookDetails.aspx?PublicationID=987. LEE, R., MASON, A., AND MEMBERS OF THE NTA NETWORK. (2014): “Is low fertility really a problem? Population ageing, dependency, and consumption,” Science, 346: 229–234. Available at https://doi.org/10.1126/science.1250542. MANJOO, F. (2021): “The world might be running low on Americans,” New York Times, May 20. Available at https://www.nytimes.com/2021/05/20/opinion/the-world-might-be-running-lowon-americans.html. MASON, A., LEE, R., ABRIGO, M., AND LEE, S.-H. (2017): “Support ratios and demographic dividends: Estimates for the world,” Technical Paper, Population Division of the Department of Economic and Social Affairs, United Nations, www.unpopulation.org. NATIONAL RESEARCH COUNCIL. (2012): Ageing and the Macroeconomy. Long-Term Implications of an Older Population. Committee on the Long-Run Macroeconomic Effects of the Ageing U.S. Population. Washington, DC: National Academies Press. POTERBA, J. M. (2014): “Retirement security in an ageing population,” American Economic Review, 104(5): 1–30. Available at https://doi.org/10.1257/aer.104.5.1. RACHEL, L., AND SUMMERS, L. H. (2019): “On secular stagnation in the industrialized world,” Brookings Papers on Economic Activity, Spring: 1–54. Available at https://doi.org/10.1353/eca.2019.0000. SAMUELSON, P. (1975): “The optimum growth rate for population,” International Economic Review, 16(3): 531–538. SANCHEZ-ROMERO, M. (2013): “The role of demography on per capita output growth and saving rates,” Journal of Population Economics, 26(4): 1347–1377. Available at https://doi.org/10.1007/s00148-0120447-3. ´ SANCHEZ -ROMERO, M., ABIO, G., PATXOT, C., AND SOUTO, G. (2018): “Contribution of demography to economic growth.” SERIEs, 9: 27–64. Available at https://doi.org/10.1007/s13209-017-0164-y SHEINER, L., SICHEL, D., AND SLIFMAN, L. (2006): “A primer on the macroeconomic implications of population ageing.” Staff working papers in the Finance and Economics Discussion Series Divisions of Research & Statistics and Monetary Affairs Federal Reserve Board, Washington, D.C (FEDS) 2007-01. SOLOW, R. M. (1956): “A contribution to the theory of economic growth,” Quarterly Journal of Economics, 70(1): 65–94. SUMMERS, L. H. (2015): “Demand side secular stagnation,” American Economic Review, Papers and Proceedings, 105(5): 60–65. UNITED NATIONS. (2013): National Transfer Accounts Manual: Measuring and Analysing the Generational Economy. New York: United Nations Population Division. UNITED NATIONS. (2019): “World population prospects 2019.” Available at https://population. un.org/wpp/Download/Standard/Population/ [accessed on June 14, 2021]. WEIL, D. N. (2008): “Population ageing,” In: The New Palgrave Dictionary of Economics, London: Palgrave Macmillan. doi: 10.1057/978-1-349-95121-5-2460-1.

302

16 CONSUMPTION, SAVING, AND WEALTH ACCUMULATION AT OLD AGE: COMPARING EVIDENCE FROM DEVELOPED AND DEVELOPING COUNTRIES1 Marco Angrisani, Jinkook Lee, and Giacomo Rebellato

Abstract In this chapter, we briefly review the literature about consumption, saving, and wealth accumulation decisions in developed and developing countries. Given the rapid ageing of the population worldwide and the health and longevity risks older households face, we focus on the behavior of older individuals and document wealth decumulation patterns in different cultural and economic contexts. We first discuss why the study of consumption and saving decisions in developed and developing countries may require different conceptual frameworks and empirical approaches. We then provide an overview of observed consumption and saving patterns at old age, highlighting key factors that might influence the shape of the age-wealth profile in different contexts. Finally, we hint at future research avenues opened by newly available microdata on consumption, saving, and wealth in developing countries.

16.1

Introduction

The theoretical paradigm for research on household consumption and savings decisions is the life-cycle model/permanent income hypothesis (Modigliani and Brumberg, 1954; Friedman, 1957). According to this framework, the main motivation for saving is to accumulate resources to finance consumption in later life and ensure a certain standard of living over time. Hence, savings should be positive while individuals work and negative during retirement, with wealth exhibiting a hump-shaped profile over the life cycle. The empirical validity of the life-cycle model/permanent income hypothesis predictions has been the subject of much debate and research, mostly in the context of developed economies. Studies focusing on developing countries are fewer, mainly due to lack of suitable data until recent years. Most importantly, analyzing consumption, saving, and wealth accumulation decisions in developing countries requires a conceptual framework that may depart from a standard life-cycle model. As first pointed out by Deaton (1989), households in developing countries tend DOI: 10.4324/9781003150398-19

303

Marco Angrisani et al.

to be large and intergenerational. Thus, the need to accumulate resources, either as a means of transferring income from high- to low-productivity phases or transferring wealth between generations, is less obvious than in developed economies. Furthermore, most households engage in agricultural activities, thereby facing high income uncertainty. In relatively poor contexts where borrowing constraints are binding, saving serves mainly as a buffer against short-term income fluctuations (Deaton, 1991; Carroll, 1997). Noting differences in wealth accumulation and decumulation patterns across income groups, Holzmann et al. (2019) propose a three-tier model that may explain some of the observed differences in behavior between developing and developed countries. At the lowest income level— which may well represent a developing country—little or no saving occurs and, consequently, little dissaving occurs after retirement. At the highest income level—which may correspond to the most developed economies—households accumulate resources until retirement and do not substantially dissave as predicted by a standard version of the life-cycle model, unless hit by major shocks. The hump-shaped age-wealth profile predicted by a standard life-cycle model is apparent in the mid-income tier, which may exemplify middle- and upper-middle income countries. A one-model-fits-all approach is inadequate when examining consumption and saving decisions across countries with different levels of economic development and sociocultural norms (Deaton, 2018). With more reliable and readily accessible household survey data in developing countries, new empirical evidence can help understand how to better model expenditure and saving behaviors in emerging economies. This chapter aims to review the current state of knowledge about consumption, saving, and wealth accumulation decisions across countries, especially at old ages, given the rapid ageing of the population worldwide and the health and longevity risks older households face. First, we briefly describe why the study of consumption and saving decisions in developed and developing countries may require separate conceptual frameworks and empirical approaches. Then, we provide an overview of the literature on consumption, saving, and wealth at old age, first for developed and then for developing economies. In both cases, we highlight key factors that might influence the shape of the age-wealth profile in different contexts. After that, we discuss future research avenues opened by newly available microdata on consumption, saving, and wealth in developing countries. In the final section, we conclude.

16.2

Consumption, Saving, and Wealth in Developed and Developing Countries

The literature about consumption, saving, and wealth over the life cycle in developed countries is vast (Attanasio and Weber, 2010). In contrast, these outcomes have been studied to a lesser extent in the context of developing economies, largely due to lack of suitable data. Deaton (1989) highlights that data inadequacies make it harder to reliably measure consumption, saving, and asset holdings in developing than in developed economies. The study of household consumption, saving, and wealth accumulation decisions in developing countries also presents unique challenges from a theoretical standpoint. Deaton (1989) provides compelling reasons for investigating saving behavior in developing countries separately from saving behavior in developed countries. At the microeconomic level, households in developing countries tend to be large and multigenerational. Most such households secure income through agricultural work, which is uncertain and dependent on exogenous factors (e.g., climate conditions and natural disasters). In view of limited access to formal credit markets, these two aspects have different implications for saving behavior. While higher income volatility should lead to higher saving rates to insure against negative income shocks, larger household structures 304

Consumption, Saving, and Wealth Accumulation at Old Age

reduce the need for precautionary savings because intra-household transfers can provide the necessary buffer against income fluctuations. The contributions of Deaton and Paxson (1992), using data from Cˆote d’Ivoire and Thailand for the period 1985–1986, and Rosenzweig (1988), using data from India for the period 1975–1983, provide empirical evidence that the latter mechanism dominates the former, resulting in lower saving rates in developing than in developed countries. Previous literature (see Mikesell and Zinser, 1973, for a review) already documented a discrepancy in saving rates between developing and developed countries and linked that to different levels of economic growth. As Morduch (1995) discusses, a further complication in analyzing consumption and saving dynamics in developing countries is that, when complete insurance markets do not exist, households can decide to smooth agricultural income (which most households rely on in underdeveloped and underindustrialized economies), through conservative production strategies aimed at minimizing the risk of cyclical and/or seasonal crop fluctuations, and/or consumption, through borrowing and saving. These two smoothing strategies can be adopted simultaneously, and assessing their relative importance is complex. From a macroeconomic point of view, a few developing countries have institutions permitting the implementation of fiscal policies aimed at stabilizing personal disposable income. Informal and underground sectors imply not only difficulties in detecting and measuring tax bases, which, in turn, leads to lower tax revenues, but also in effectively reaching the segments of the population most in need of social assistance. In the presence of severe borrowing constraints, saving in developing countries serves primarily as insurance against income shocks in the short term, rather than to accumulate resources for the long term. The latter need is typically met via intra-household/intergenerational insurance and welfare mechanisms. Consequently, the life-cycle profile of wealth does not exhibit a clear hump shape, with marked accumulation and decumulation phases, as predicted by a standard life-cycle model. Deaton (2018) reports age profiles of consumption and income in Cˆote d’Ivoire and Thailand in the mid-1980s. These profiles reveal little evidence of either hump saving among the young or dissaving among the elderly, with consumption tracking income very closely over the life cycle. Financial development also plays an important role in shaping financial outcomes over the life cycle and in determining differences between developed and emerging economies. In developing countries, not only households at the bottom of the income and wealth distributions experience difficulties in accessing credit and financial markets, but also households from wealthier backgrounds might suffer from higher costs of borrowing and financing, given troubled institutional and financial settings with limited competition. Household-level data on wealth composition have only become available recently. Combined with information about existing financial institutions and availability of saving and borrowing instruments, this has allowed researchers to shed light on the impacts of different institutional arrangements on household financial outcomes and behaviors and on the role that policymakers can play in fostering financial stability in developing contexts. Badarinza et al. (2019) provide a glimpse into the household finance landscape in developing countries using microdata on household balance sheets. The authors construct harmonized measures of household assets and liabilities across six emerging economies (Bangladesh, China, India, the Philippines, South Africa, and Thailand) to analyze wealth distributions and accumulation patterns among them. They also compare the composition of household balance sheets between developing and developed countries. Among the asset classes considered, two are especially relevant for our discussion: financial savings and retirement savings. In emerging economies, most households hold most of their wealth in tangible assets (e.g., real estate, livestock) and face significant barriers to financial savings. Both these phenomena are commonly 305

Marco Angrisani et al.

observed across countries and likely stem from two main factors. First, in developing countries, individuals receive relatively low levels of income and face high-income volatility, making saving flows very unpredictable. Second, financial savings are often associated with large transaction costs, which can be pecuniary—due to underdeveloped financial markets and limited availability/accessibility of financial institutions—and nonpecuniary—stemming from lack of trust in institutions, reluctance to receive financial advice, and behavioral biases. Retirement wealth represents another systematic difference between developed and developing countries. When developing countries have some sort of contributory pension schemes, the informality of labor markets and the underdevelopment of institutions complicate the collection of contributions and redistribution of resources. This undermines the efficiency and efficacy of public pension schemes in fostering long-term savings and consumption smoothing over the life cycle. Given the rapid ageing of the population in developing countries and changes in household demographic composition, sustainable public pension schemes may become essential to ensure financial security of the elderly, especially among most vulnerable groups. Bloom and McKinnon (2013) review options in public pension system design and implementation. They highlight that a one-system-fits-all approach is unlikely to succeed. In contrast, they advocate for a combination of policies to optimize coverage, benefits, and financing strategies in relation to countries’ demographic trends, history, household structure and support practices, political system, size of the informal labor market, and fiscal outlook.

16.3

Consumption, Saving, and Wealth at Older Ages

In this section, we zoom in on consumption, saving, and wealth accumulation behavior at older ages. More precisely, we refer to individuals over the age of 50, a group for which harmonized microeconomic data are currently available in 47 countries (Lee et al., 2019). We briefly review empirical evidence from developed and emerging economies and contrast it with theoretical predictions. We also emphasize the differences between developed and developing countries. The reference theoretical framework is the life-cycle/permanent income model originally developed by Modigliani and Brumberg (1954) and Friedman (1957). Over the years, the basic version of this model has been extended in various directions. While flexible versions allow the reconciliation of empirical patterns and theoretical predictions, some puzzles and inconsistencies remain unexplained (Attanasio and Weber, 2010). Unsurprisingly, divergences from the model exhibit different shapes depending on the economic and institutional contexts individuals live and operate in.

16.3.1

Developed Countries

In most developed countries, individuals tend to retain substantial amounts of wealth while approaching the end of their life. This empirical fact, which is largely at odds with a standard life-cycle model, has been the focus of a large body of research in recent years. Existing work has advanced several explanations for this phenomenon, ranging from survival uncertainty, bequest motives, and medical expenditures at old ages, to the generosity of public pension provisions, wealth allocation strategies, and family structure. Comparing empirical evidence across developed economies characterized by different institutions helps in gauging the merit of these explanations and the relative importance of the mechanisms underlying them. The early contribution of Shorrocks (1975) emphasizes the importance of dealing with both differential mortality and cohort effects when trying to estimate the trajectory of wealth over time, especially when the focus is on the second half of the life cycle. Cross-sectional analyses 306

Consumption, Saving, and Wealth Accumulation at Old Age

may reveal increasing wealth over the life cycle partly because survivors tend to be richer. Similarly, different generations may accumulate different amounts of wealth because they are subject to different institutional settings (e.g., generosity of public pension) or economic conditions (e.g., productivity). Failing to account for such cohort effects is likely to bias conclusions about the actual shape of the age-wealth profile. Because of that, research on this topic has grown significantly over time with the availability of longitudinal data, which more easily permit controlling for differential mortality and cohort effects. A pioneering example of this line of work is Bernheim (1987). He empirically rejects the theoretical predictions of a standard life-cycle model, finding lack of wealth decumulation in a panel of retirees in the United States. In Europe, B¨orsch-Supan and Stahl (1991) and B¨orsch-Supan (1992) document that German households decrease spending and increase asset holdings after age 70. These profiles do not appear to be confounded by cohort effects and mortality differences. Using longitudinal data from the Netherlands, Alessie et al. (1995) find weak evidence against the life-cycle hypothesis as the median household after age 65 no longer accumulates or decumulates significant amounts of wealth. Disney et al. (1998) study wealth accrual and decumulation in Britain, highlighting important differences in the trajectory of assets over time and across cohorts. They provide evidence of saving behavior in line with the life-cycle model, whereby wealth increases up to age 60 and declines monotonically thereafter. At the same time, they caution against this result, given that a large fraction of household wealth is held in the form of housing and an unprecedented fall in housing prices took place over the observation period used in their study. Exploiting data from the Survey of Health, Ageing and Retirement in Europe, Christelis et al. (2009) provide one of the first cross-country analyses of wealth holding and wealth management at older ages. They document that the net-worth-to-income ratio is higher in Southern European than in Northern European countries, mainly due to a larger fraction of homeowners in the former. In general, older individuals (age 76 or more) exhibit a lower net-worth-toincome ratio than their younger counterparts (age 66–75), suggesting a decreasing trajectory of wealth over the life cycle. These differences by age are less pronounced in Southern European countries, reflecting once more the relatively higher weight of housing value on household total wealth in those countries. Poterba et al. (2011) rely on longitudinal data from the Health and Retirement Study (HRS) in the United States over the period 1992–2008 to examine the drawdown of wealth in retirement. They document that most households reach retirement with relatively little financial wealth—a conclusion that contrasts the previous finding by Scholz et al. (2006) that less than 20 percent of households interviewed by HRS in 1992 have less wealth than the optimal target predicted by a life-cycle model. Among households in the upper half of the wealth distribution, which hold more substantial amounts of liquid wealth, little evidence exists of financial asset decumulation in the early years of retirement. These households appear to rely on pension and Social Security benefits and retirement account withdrawals as income sources and to deplete financial wealth only in the case of health shocks or spouse’s death. Love et al. (2009) also observe the lack of a downward wealth trajectory with age. These authors demonstrate that a life-cycle model with uncertain longevity, random medical expenses, and bequest motives (aspects we will return to in more detail) can better replicate this empirical pattern. Poterba et. al. (2015) compare assets of HRS respondents in the first and last year they were observed (FYO and LYO). Their results show that asset balances are quite persistent at old ages: across cohorts, assets remain roughly constant between FYO and LYO, especially in the lower half of the wealth distribution. For those who enter retirement with substantial levels of wealth, the main determinants of asset depletion are health shocks and changes in household composition (e.g., death of a spouse), two factors that will be examined in more 307

Marco Angrisani et al.

detail shortly. Van Ooijen et al. (2015) studied the saving behavior and portfolio choices of Dutch individuals after retirement using longitudinal administrative data over the period 2005– 2010. Like the previously described U.S. case, the authors find that the death of a spouse implies significant reductions in wealth. However, negative health events are associated with higher savings. This result may stem from a combination of factors, including the comprehensive health insurance system and generosity of public pensions in the Netherlands, which guarantee a certain standard of living, and a decline in the marginal utility of consumption after a health shock, which would lead to higher savings. Banks et al. (2010) explore differences in wealth trajectory by cognitive ability and numeracy using the first three waves of the English Longitudinal Study of Ageing. They find some tentative evidence that for better-skilled individuals, whose behavior could better approximate that of a rational, forward-looking representative agent, wealth is more hump-shaped than for their less-skilled counterparts. This, however, does not translate into higher income and consumption replacement rates (post-retirement income and consumption levels relative to pre-retirement) and into better subjective well-being in retirement, which are similar across cognitive ability and numeracy groups. Overall, the existing literature reveals a robust pattern across developed countries: individuals tend to keep relatively large amounts of wealth even at very old ages. The extent to which this happens depends on the institutional settings, sociocultural factors, and economic incentives faced by retirees. The empirical evidence from various developed economies points at the following main reasons for why the elderly may not dissave as much as a standard life-cycle model would predict: longevity risk, bequest motives, and medical expenditures.

16.3.2

Longevity Risk

Attanasio and Hoynes (2000) estimate the relationship between wealth and mortality and use it to correct wealth-age profiles. Specifically, they assume that survival probability at a given age depends on the relative position in the wealth distribution. They present wealth-age profiles of different cohorts with and without the correction for differential mortality. The uncorrected age-wealth profiles show no evidence of asset decumulation at old ages. While substantial, the adjustment induced by the correction for differential mortality does not generate a pronounced decline in wealth in the last part of the life cycle. Similarly, Love et al. (2009) find that differences in subjective survival expectations account for some of the cross-sectional variation in age profiles of annualized wealth among HRS respondents. However, even among the most pessimistic retirees, no evidence exists of a robust downward-sloping trajectory in wealth.

16.3.3

Bequest Motives

Bernheim (1991) documents the existence of strong bequest motives that prevent individuals from dissaving at the pace that a standard life-cycle model would suggest and from converting their assets into annuities, even if facing perfect insurance markets. Hurd (1987) reaches a different conclusion: in a 10-year panel, wealth decreases with age, a fact that would be at odds with the presence of bequest motives. Similarly, Hurd (1989) estimates the parameters of a life-cycle model with bequests and finds that the marginal utility of bequests is small. Consequently, actual bequests are not very large, on average, and mainly accidental, as they stem from uncertainty about longevity. More recently, Lockwood (2018) argues that the fact that the elderly accumulates assets and do not buy annuities or long-term care insurance constitutes evidence for bequest motives. Individuals without bequest motives would like to consume all their wealth and find setting aside resources to insure for late-life risks costly. In contrast, those with 308

Consumption, Saving, and Wealth Accumulation at Old Age

bequest motives have a lower opportunity cost of precautionary saving, which, at the same time, decreases the demand for annuities and long-term insurance and increases asset accumulation at old ages.

16.3.4

Medical Expenditures

With increasing availability of wealth, health, and medical expenditure data for the same individuals from longitudinal surveys like the HRS and the Panel Study of Income Dynamics, the number of studies focusing on the role of uncertain future medical expenses for asset decumulation at old ages has increased significantly in the last two decades. Palumbo (1999) is one of the earliest contributions in this body of literature. The author finds that, while uncertain outof-pocket medical expenses represent an important motive for precautionary saving among the elderly, it is still not enough to fully explain the reluctance of individuals in the United States to draw down their assets during retirement. De Nardi et al. (2010) show that medical expenses are much higher and more volatile than previously estimated. Importantly, they rise very fast with age and exhibit a steep income gradient. Incorporating these features into their model, they can replicate observed saving profiles in the data and explain why richer households dissave less at old ages than their poorer counterparts. Lee and Kim (2008) rely on the first four waves of the HRS to study the impact of health stocks on wealth. They document that adverse health events accelerate the rate of wealth depletion only if they occur later in life (from age 70 onward). In general, health shocks have significant negative impacts on wealth when they happen, but these effects tend to disappear over time. The existing literature suggests that uncertainty about medical costs in late life is an important determinant of the slow decumulation of retirement wealth in the United States. Gauging the role of medical expenses for savings decisions at old ages, however, requires further analyses and better measures of medical costs in the very last months of life, when large expenditures are often incurred. Preliminary evidence reveals that out-of-pocket medical expenses right before death can deplete assets entirely and represent a key reason for holding large amounts of wealth in old age (Hoover et al., 2002; French et al., 2006). The importance of medical costs in shaping the wealth trajectory over the life cycle is confirmed by comparing the United States with countries characterized by universal health systems and therefore lower levels and volatility of medical spending. An example of such an exercise is Banks et al. (2019), who document steeper declines in nondurable spending at old ages in the United Kingdom than in the United States. The main factor underlying this difference is the sharp increase with age of out-of-pocket medical costs in the United States, which the National Healthcare System mostly covers in the United Kingdom. When out-of-pocket medical expenditures are removed from the computation of nondurable spending, the difference between the consumption trajectories in the two countries shrinks by three-quarters. This aspect may, once more, justify why U.S. households decumulate retirement wealth relatively slowly if at all. Finally, wealth definition and composition can significantly affect the trajectory of wealth over the life cycle and the extent to which wealth is depleted at old ages. As mentioned previously, Christelis et al. (2009) perform one of the first cross-country comparisons of wealth, emphasizing the role of housing in shaping the age-wealth profile. Blundell et al. (2016) confirm the importance of the illiquid nature of housing in explaining wealth drawdown at old age. They examine the levels and trajectories of assets among individuals over the age of 70 in the United Kingdom and the United States over the period 2002–2011. They show that assets were drawn more slowly in the United Kingdom than in the United States, a fact largely driven by a significantly higher growth in house prices in the former country. A relevant lesson here is that the illiquid nature of 309

Marco Angrisani et al.

housing may prevent households from drawing on housing wealth during retirement and generate age-wealth profiles that are inconsistent with theoretical predictions. Another important aspect to consider when analyzing age-wealth profiles across developed countries is the generosity of public pension systems. Because pension benefits are effectively life annuities, they protect against longevity and other types of late-life risk and greatly affect how retirees manage their accumulated resources as they age (Mitchell, 2002).

16.3.5

Developing Countries

As noted earlier, the age-wealth profile in developing economies may differ from that observed in developed economies for many reasons. Comparing empirical evidence across developing countries, resource pooling within households clearly characterizes consumption and savings patterns; with economic development, the economic unit moves away from a household to an individual. Deaton and Paxson (1992) are among the first authors to examine consumption, saving, and wealth accumulation patterns at older ages in developing countries. They rely on householdlevel data from Cˆote d’Ivoire and Thailand that exhibited limited reliability and accuracy. Because of that, they view their results as preliminary empirical insights and indicative of future research questions. The authors stress that a traditional life-cycle model of saving and capital accumulation may not be the ideal reference framework for empirical analysis focusing on developing countries, where the role of the household in the production and distribution of resources differs significantly from that in developed economies. In developing settings, households tend to be large and multigenerational and provide old-age and other kinds of insurance internally. Deaton and Paxson (1992) find that, while individual labor market participation and earnings patterns show standard life-cycle hump shapes, savings do not. This, however, does not mean that consumption tracks income very closely, thereby exhibiting significant variability. Intra-household insurance mechanisms ensure that consumption is much less variable over the life cycle than individual income. In other words, households can sustain their members’ living standards at old age without an obvious need to accumulate and decumulate assets. When trying to estimate age-wealth profiles in developing countries, the measurement of wealth itself represents an important limitation. In contexts where ownership of financial assets is uncommon, appraisal of real estate is complicated, and evaluation of other assets can be very inaccurate, reliably measuring household wealth and following its evolution over the life cycle are hard. While more recent data collection efforts have partly addressed these measurement issues, these issues restricted the focus of early studies to income and consumption lifecycle patterns, inferring saving and wealth accumulation trajectories from the comparison of these two. Deaton and Paxson (1994) rely on a series of 15 household income and expenditure surveys in Taiwan and track cohorts over time to study the life-cycle dynamics of income, consumption, and saving. They document higher saving rates among younger cohorts likely driven by demographic changes taking place at the time of the study, such as falling fertility rates and rising life expectancy. The authors speculate that the young may save more in preparation for a “modern old age” where large households would no longer be the norm, intra-household transfers would be less common, and the young would have to provide their own resources to finance retirement. In line with the empirical evidence for developed countries described previously, old Taiwanese cohorts appear to slightly increase their savings as they age, and they certainly do not dissave. The authors highlight that these observed patterns are likely to reflect the process of household dissolution at old age. 310

Consumption, Saving, and Wealth Accumulation at Old Age

More than two decades later, Deaton (2018) concludes that, in developing countries, consumption and income life-cycle trajectories are more linked than what standard theory would predict, that savings mainly serve the purpose of protecting standards of living against short-term fluctuations in income, and that the accumulation and decumulation of wealth are only observed for a small minority of households. He emphasizes once again the key role that intra-household insurance mechanisms play. In his words, “The evidence that individuals or households in developing countries make effective provision for retirement by saving is weaker or nonexistent, nor is it even clear that it makes sense for them to try given traditional social and family arrangements.” Interesting case studies for this review are low-income countries in Asia and Southeast Asia, which have experienced significant and rapid changes in their economic and demographic structures in the last 20 years. Among them, China has attracted particular interest from researchers. Chamon and Prasad (2010) use household-level data over the period 1995–2005 to examine saving behavior in China. They estimate a saving-age profile with an atypical U-shape. Controlling for cohort effects, households headed by young and old persons exhibit the highest saving rates. This pattern clearly contrasts the traditional hump-shaped profile of savings predicted by a standard life-cycle model and is plausibly the result of the increasing cost of housing, education, and healthcare due to the liberalization and transition to a market economy. Rosenzweig and Zhang (2014) delve into the Chinese saving puzzle that young individuals save more than their middle-aged counterparts. They show that, in the context of increasing housing costs that has characterized China in the past decades, the option of intergenerational co-residence facilitates saving among the young. The main underlying mechanism is that, in the presence of high housing cost, young individuals may optimally decide to co-reside with their parents, save on housing cost, and provide in-kind assistance in exchange. In line with this argument, the paper finds that intergenerational co-residence correlates negatively with the income of the young and positively with the socioeconomic status of the parents. Moreover, while intergenerational co-residence is associated with higher savings for the young, it does not lead to higher savings for the old, who feel less the need to self-insure against the risk of healthcare costs. Household size appears to be a crucial determinant of saving behavior in China. Lugauer et al. (2017) consider a theoretical life-cycle model that includes finite lifetimes and saving for retirement and in which parents care about the consumption of their dependent children. The model predicts a negative relationship between the number of dependent children in the family and the saving rate of a household. This prediction is tested using data from the China Household Finance Survey and receives empirical support. Exploiting the differential enforcement of the one-child policy across counties as a source of exogenous variation in household composition, the authors find that families with fewer dependent children exhibit significantly higher saving rates. No study to date has examined the implications that such a pattern, mainly driven by structural demographic changes, may have for asset management and decumulation at older ages. Recent work has focused on consumption and saving behavior in African countries. Ting and Kollamparambil (2015) study income, consumption, and saving outcomes of South African households using data from the General Household Survey. The authors report age profiles of 12 cohorts based on the birth year of the household head and perform regression analyses to estimate the effect of age on the variables of interest. The cohort analysis shows a typical humpshaped income path, peaking in the early 40s (while the peak in other developed countries happens in the late 40s or early 50s). When including social transfers in the income computation, this hump shape is less apparent as income increases significantly in the post-retirement period. Both consumption and consumption-to-income ratio are relatively smooth over the life 311

Marco Angrisani et al.

cycle. This indicates that savings rates do not follow a hump-shaped pattern as predicted by the permanent income hypothesis. Zooming in on old ages, households appear to be able to maintain their standards of living in retirement through government grants. A significant part of these transfers, however, is saved, mainly to insure against future funeral expenses. Mshamu et al. (2020) conducted a survey of household heads in 68 rural villages surrounding Mtwara town in Tanzania. They estimate the relationship between household wealth, measured by housing quality and ownership of durable assets, and age of the household head. They find that wealth accumulation in participating villages increases monotonically from age 20 onward, peaks at age 50, and declines sharply thereafter. Larger households tend to live in better houses and are more likely to own durables. In line with the general evidence from developing countries, this suggests that larger households that include young adults are better positioned to share risks and can protect against old-age poverty via financial contributions and labor to maintain a basic wattle and daub home.

16.4

Advanced Research on Consumption, Savings, and Wealth Accumulation Decisions at Old Ages: New Data from Developing Countries

In the past two decades, data collection efforts in developing countries have burgeoned, offering unprecedented high-quality household data on consumption, savings, and wealth. In this section, we highlight these new sources of data and suggest important research questions to deepen our understanding of consumption, savings, and wealth accumulation decisions at old ages in developing and transition economies. To date, seven longitudinal studies survey middle-aged and older individuals in developing countries: the Mexican Health and Ageing Study, the Costa Rican Longevity and Healthy Ageing Study, the Brazilian Study of Ageing, the Indonesian Family Life Survey, the China Health and Retirement Longitudinal Survey, the Malaysian Ageing and Retirement Study, and the Longitudinal Ageing Study in India. The major strengths of these studies are their comprehensive questionnaires, which include detailed questions on consumption, income, and wealth; the representativeness of their samples; and ex-ante harmonization of survey instruments, facilitating comparative studies across countries. Lee et al. (2021a) briefly summarize the cohort characteristics.2 Modeled after the HRS, these studies collect information about household wealth composition, eliciting ownership and value of housing and other real assets; financial assets, sources, and amount of income for all household members; and household expenditures. Combined with rich sociodemographic and health data, such measures provide an excellent opportunity to study consumption, savings, and wealth accumulation decisions at old age. Moreover, analyses across countries with different institutional settings, levels of economic development, and sociocultural norms can help researchers identify the key factors that shape the age-wealth profile in developing countries, the sources of risk the elderly face in different emerging economies, and the insurance mechanisms available to households to ensure a certain standard of living for all its members. Another important area of research is financial inclusion. In the face of declining multigenerational co-residence rates and growing demand for more robust public social safety nets, financial inclusion in developing countries is an important driver of economic security (World Bank, 2018; Lee et al., 2021b). A key question to address is the extent to which increasing ownership of financial accounts in emerging economies can contribute to reducing disparities by gender and socioeconomic status groups. With continuing economic growth, the financial sector is rapidly evolving in developing countries, and the emerging institutional environment 312

Consumption, Saving, and Wealth Accumulation at Old Age

may play an important role in shaping financial outcomes. Newly available household-level data on wealth composition combined with information on changing financial sectors will enable researchers to shed light on how facilitating savings and financial planning among older generations may improve the economic security and overall well-being of the elderly. Finally, cash transfer programs have become the centerpiece of many developing countries’ social protection agenda (Bastagli et al., 2019) and an important pillar of pension systems in many developed economies (B¨orsch-Supan et al., 2021). Data on income from public old-age pension and transfers provide an opportunity to study the impact of these means-tested programs on consumption and saving behaviors. Empirical evidence on this matter is crucial to inform the debate on how to design and enhance such welfare programs.

16.5

Conclusion

Our review centers on the trajectories of consumption, saving, and wealth at old ages across countries with different levels of economic development. We adopt the life-cycle/permanent income model as our reference theoretical framework and underscore the divergence between observed empirical patterns and theoretical predictions. In developed countries, households appear to retain large amounts of wealth at old ages and exhibit little tendency to dissave unless they are hit by major health shocks. While this phenomenon contrasts the typical hump-shaped wealth profile predicted by a standard lifecycle model, it can be rationalized by the existence of bequest motives, longevity uncertainty, and risk of high medical expenditures at the end of life. The comparison of empirical evidence across countries characterized by different levels of public pension generosity, health insurance arrangements, and sociocultural norms helps in gauging the relative merit of these explanations. The consensus seems to be that a rich and fully fledged model that accounts for the existing institutional setting and economic incentives retirees face can largely replicate the observed saving and wealth patterns in developed countries. Relatively fewer studies focus on developing economies, mainly due to lack of comprehensive and reliable data. Yet, several important lessons can be learned. First, given significant differences in institutions and socioeconomic factors between developed and developing countries, studying consumption and saving in the latter requires distinct conceptual frameworks and empirical approaches. Second, as households tend to be large and intergenerational, the need to accumulate wealth to transfer resources from working to nonworking phases or between generations is less obvious. In fact, households in developing countries can sustain their members’ living standards at old age via intra-household transfers. As a result, savings mainly serve to insure against short-term fluctuations in income, which is predominately agricultural. Third, with the rapid changes in developing countries’ economic and demographic structures, the economic unit is moving away from a household to an individual. New tendencies in savings and wealth accumulation/decumulation at old ages may emerge. Recent advances in data collection in developing countries will allow researchers to study these patterns in more detail and with greater accuracy than was possible before. Further research in this area is crucial for the development of policies and programs to ensure old-age economic well-being.

Notes 1 Funding: this work has received funding by the National Institute on Aging (R01 AG030153). 2 Individual cohort profiles can be found in Wong et al. (2017); Rosero-Bixby et al. (2019); Lima-Costa et al. (2018); Serrato and Melnick (1995); Zhao et al. (2014); Mansor et al. (2019); IIPS et al. (2020).

313

Marco Angrisani et al.

References ALESSIE, R., LUSARDI, A., AND KAPTEYN, A. (1995): “Saving and wealth holdings of the elderly,” Ricerche Economiche, 49(3): 293–314. ATTANASIO, O. P., AND HOYNES, H. W. (2000): “Differential mortality and wealth accumulation,” Journal of Human Resources, 35(1): 1–29. ATTANASIO, O. P., AND WEBER, G. (2010): “Consumption and saving: Models of intertemporal allocation and their implications for public policy,” Journal of Economic Literature, 48(3): 693–751. BADARINZA, C., BALASUBRAMANIAM, V., AND RAMADORAI, T. (2019): “The household finance landscape in emerging economies,” Annual Review of Financial Economics, 11: 109–129. BANKS, J., BLUNDELL, R., LEVELL, P., AND SMITH, J. P. (2019): “Life-cycle consumption patterns at older ages in the United States and the United Kingdom: Can medical expenditures explain the difference?,” American Economic Journal: Economic Policy, 11(3): 27–54. BANKS, J., O’DEA, C., AND OLDFIELD, Z. (2010): “Cognitive function, numeracy and retirement saving trajectories,” The Economic Journal, 120(548): F381–F410. BASTAGLI, F., HAGEN-ZANKER, J., HARMAN, L., BARCA, V., STURGE, G., AND SCHMIDT, T. (2019): “The impact of cash transfers: A review of the evidence from low-and middle-income countries,” Journal of Social Policy, 48(3): 569–594. BERNHEIM, B.D. (1987): “Dissaving after retirement: Testing the pure life cycle hypothesis.” In: Z. Bodie, J. B. Shoven and D. A. Wise, eds., Issues in Pension Economics. Chicago: University of Chicago Press, pp. 237–280. BERNHEIM, B. D. (1991): “How strong are bequest motives? Evidence based on estimates of the demand for life insurance and annuities,” Journal of Political Economy, 99(5): 899–927. BLOOM, D. E., AND MCKINNON, R. (2013): “The design and implementation of public pension systems in developing countries: Issues and options,” IZA Policy Paper No. 59. BLUNDELL, R., CRAWFORD, R., FRENCH, E., AND TETLOW, G. (2016): “Comparing retirement wealth trajectories on both sides of the pond,” Fiscal Studies, 37(1): 105–130. ¨ BORSCH -SUPAN, A. (1992): “Saving and consumption patterns of the elderly,” Journal of Population Economics, 5(4): 289–303. ¨ BORSCH -SUPAN, A., BUCHER-KOENEN, T., GOLL, N., AND HANEMANN, F. (2021): “Targets missed: Three case studies exploiting linked SHARE-RV data,” Journal of Pension Economics and Finance, 1–121. doi:10.1017/S1474747220000359. ¨ BORSCH -SUPAN, A., AND STAHL, K. (1991): “Life cycle savings and consumption constraints,” Journal of Population Economics, 4(3): 233–255. CARROLL, C. D. (1997): “Buffer-stock saving and the life cycle/permanent income hypothesis,” The Quarterly Journal of Economics, 112(1): 1–55. CHAMON, M. D., AND PRASAD, E. S. (2010): “Why are saving rates of urban households in China rising?,” American Economic Journal: Macroeconomics, 2(1): 93–130. CHRISTELIS, D., JAPPELLI, T., PACCAGNELLA, O., AND WEBER, G. (2009): “Income, wealth and financial fragility in Europe,” Journal of European Social Policy, 19(4): 359–376. DEATON, A. (1989): “Saving in developing countries: Theory and review,” The World Bank Economic Review, 3(suppl 1): 61–96. DEATON, A. (1991): “Saving and liquidity constraints,” Econometrica, 59(5): 1221–1248. DEATON, A. (2018): “Saving and consumption smoothing.” In: Deaton, A. (ed.), The Analysis of Household Surveys (Reissue Edition with a New Preface): A Microeconometric Approach to Development Policy. Washington, DC: World Bank, pp. 335–400. DEATON, A. S., AND PAXSON, C. (1992): “Patterns of ageing in Thailand and Cˆote d’Ivoire.” In: Wise, D. A. (ed.), Topics in the Economics of Ageing. Chicago, IL: University of Chicago Press, pp. 163–206. DEATON, A. S., AND PAXSON, C. (1994): “Saving, growth and ageing in Taiwan.” In: Wise, D. A. (ed.), Studies in the Economics of Ageing, Chicago, IL: University of Chicago Press, pp. 331–362. DE NARDI, M., FRENCH, E., AND JONES, J. B. (2010): “Why do the elderly save? The role of medical expenses,” Journal of Political Economy, 118(1): 39–75. DISNEY, R., JOHNSON, P., AND STEARS, G. (1998): “Asset wealth and asset decumulation among households in the retirement survey,” Fiscal Studies, 19(2): 153–174. FRENCH, E., DE NARDI, M., JONES, J. B., BAKER, O., AND DOCTOR, P. (2006): “Right before the end: Asset decumulation at the end of life,” Economic Perspectives, 30(3): 2–13. FRIEDMAN, M. (1957): “The permanent income hypothesis.” In: Friedman, M. (ed.), A Theory of the Consumption Function. Princeton, NJ: Princeton University Press, pp. 20–37.

314

Consumption, Saving, and Wealth Accumulation at Old Age

HOLZMANN, R., AYUSO, M., ALAMINOS, E., AND BRAVO, J. (2019): “Life cycle saving and dissaving revisited across three-tiered income groups: Starting hypotheses, refinement through literature review, and ideas for empirical testing,” IZA Institute of Labor Economics Discussion Paper No. 12655, Bonn. HOOVER, D. R., CRYSTAL, S., KUMAR, R., SAMBAMOORTHI, U., AND CANTOR, J. C. (2002): “Medical expenditures during the last year of life: Findings from the 1992–1996 Medicare current beneficiary survey,” Health Services Research, 37(6): 1625–1642. HURD, M. D. (1987): “Savings of the elderly and desired bequests,” The American Economic Review, 77(3): 298–312. HURD, M. D. (1989): “Mortality risk and bequests,” Econometrica: Journal of the Econometric Society, 57(4): 779–813. IIPS, HSPH, AND USC (INTERNATIONAL INSTITUTE FOR POPULATION SCIENCES, HARVARD T.H. CHAN SCHOOL OF PUBLIC HEALTH, AND THE UNIVERSITY OF SOUTHERN CALIFORNIA ). (2020): “Longitudinal ageing study in India (LASI) wave 1, 2017–18, India report,” International Institute for Population Sciences, Mumbai. LEE, J., JAIN, U., GOVIL, D., LUBET, A., AND SEKHER, T. V. (2021a): “Ageing in the global south.” In: Ferraro, K. F., and Carr, D. (eds.), Handbook of Ageing and the Social Sciences, the 9th Edition. Cambridge: Academic Press, pp. 65–82. LEE, J., AND KIM, H. (2008): “A longitudinal analysis of the impact of health shocks on the wealth of elders,” Journal of Population Economics, 21(1): 217–230. LEE, J., PHILLIPS, D., AND WILKENS, J. (2019): “Gateway to global ageing data.” In: Gu, D., and Dupre, M. (eds.), Encyclopedia of Gerontology and Population Ageing. Cham: Springer, pp. 1–9. LEE, J., PHILLIPS, D., AND WILKENS, J. (2021b): “Gateway to global ageing data: Resources for crossnational comparisons of family, social environment, and healthy ageing,” The Journals of Gerontology: Series B, 76(Supplement 1): S5–S16. LIMA-COSTA, M. F., DE ANDRADE, F. B., SOUZA, P. R. B. D., NERI, A. L., DUARTE, Y. A. D. O., CASTRO-COSTA, E., AND DE OLIVEIRA, C. (2018): “The Brazilian longitudinal study of ageing (ELSI-Brazil): Objectives and design,” American Journal of Epidemiology, 187(7): 1345–1353. LOCKWOOD, L. M. (2018): “Incidental bequests and the choice to self-insure late-life risks,” American Economic Review, 108(9): 2513–2550. LOVE, D. A., PALUMBO, M. G., AND SMITH, P. A. (2009): “The trajectory of wealth in retirement,” Journal of Public Economics, 93(1–2): 191–208. LUGAUER, S., NI, J., AND YIN, Z. (2017): “Chinese household saving and dependent children: Theory and evidence,” China Economic Review 57: Article 101091. MANSOR, N., AWANG, H., AND AB RASHID, N. F. (2019): “Malaysia ageing and retirement survey.” In: Gu, D., and Dupre, M. (eds.), Encyclopedia of Gerontology and Population Ageing. Cham: Springer, pp. 1–5. MIKESELL, R. F., AND ZINSER, J. E (1973): “The nature of the savings function in developing countries: A survey of the theoretical and empirical literature,” Journal of Economic Literature, 11(1): 1–26. MITCHELL, O. S. (2002): “Developments in decumulation: The role of annuity products in financing retirement,” In: Auerbach, A.J., and Heinz, H. (eds.), An Ageing, Financial Markets and Monetary Policy. Berlin, Heidelberg: Springer, pp. 97–125. MODIGLIANI, F., AND BRUMBERG, R. (1954): “Utility analysis and the consumption function: An interpretation of cross-section data,” Franco Modigliani, 1(1): 388–436. MORDUCH, J. (1995): “Income smoothing and consumption smoothing,” Journal of Economic Perspectives, 9(3): 103–114. MSHAMU, S., PEERAWARANUN, P., KAHABUKA, C., DEEN, J., TUSTING, L., LINDSAY, S. W., KNUDSEN, J., M UKAKA , M., AND VON S EIDLEIN, L. (2020): “Old age is associated with decreased wealth in rural villages in Mtwara, Tanzania: Findings from a cross-sectional survey,” Tropical Medicine & International Health, 25(12): 1441–1449. PALUMBO, M. G. (1999): “Uncertain medical expenses and precautionary saving near the end of the life cycle,” The Review of Economic Studies, 66(2): 395–421. POTERBA, J., VENTI, S., AND WISE, D. (2011): “The composition and drawdown of wealth in retirement,” Journal of Economic Perspectives, 25(4): 95–118. POTERBA, J., VENTI, S., AND WISE, D. (2015): “What determines end-of-life assets? A retrospective view.” Working Paper No. 21682. Cambridge, MA: National Bureau of Economic Research.

315

Marco Angrisani et al.

ROSENZWEIG, M. R. (1988): “Risk, implicit contracts and the family in rural areas of low-income countries,” The Economic Journal, 98(393): 1148–1170. ROSENZWEIG, M., AND ZHANG, J. (2014): “Co-residence, life-cycle savings and inter-generational support in urban China.” Working Paper No. 20057. Cambridge, MA: National Bureau of Economic Research. ROSERO-BIXBY, L., DOW, W. H., AND BRENES, G. (2019): “Costa Rican longevity and healthy ageing study.” In: Gu, D., and Dupre, M. (eds.), Encyclopedia of Gerontology and Population Ageing. Cham: Springer, pp. 1–5. SCHOLZ, J. K., SESHADRI, A., AND KHITATRAKUN, S. (2006): “Are Americans saving ‘optimally’ for retirement?,” Journal of Political Economy, 114(4): 607–643. SERRATO, C. A., AND MELNICK, G (1995): The Indonesian Family Life Survey: Overview and Descriptive Analysis of the Population, Health and Education Data. Santa Monica, CA: RAND, Labor and Population Program. SHORROCKS, A. F. (1975): “The age-wealth relationship: A cross-section and cohort analysis,” The Review of Economics and Statistics, 57(2): 155–163. TING, L., AND KOLLAMPARAMBIL, U. (2015): “Nature and determinants of household retirement savings behaviour in South Africa,” Development Southern Africa, 32(6): 675–696. VAN OOIJEN, R., ALESSIE, R., AND KALWIJ, A. (2015): “Saving behavior and portfolio choice after retirement,” De Economist, 163(3): 353–404. WONG, R., MICHAELS-OBREGON, A., AND PALLONI, A. (2017): “Cohort profile: The Mexican health and ageing study (MHAS),” International Journal of Epidemiology, 46(2): 1–10. WORLD BANK. (2018): “Financial inclusion: Overview.” Available at https://www.worldbank.org/ en/topic/financialinclusion/overview [accessed on April 30, 2021]. ZHAO, Y., HU, Y., SMITH, J. P., STRAUSS, J., AND YANG, G (2014): “Cohort profile: The China health and retirement longitudinal study (CHARLS),” International Journal of Epidemiology, 43(1): 61–68.

316

17 AUTOMATION AND AGEING Ana L. Abeliansky and Klaus Prettner1

Abstract Most countries have experienced declining fertility, rising life expectancy, and, consequently, rapid population ageing over the last decades. This has raised concerns about labor shortages. Contemporaneously, a trend toward more automation has gained momentum. This, in turn, has raised concerns about rising technological unemployment. Yet, labor shortages and technological unemployment could hardly materialize at the same time. In this chapter, we discuss the extent to which automation in terms of industrial robots, three-dimensional printers, and algorithms based on machine learning, may be a response to the demographic challenges that we face. As such, population ageing and labor shortages may provide incentives to invent and install new labor-saving technologies in the first place. Theoretical considerations and first empirical evidence are indeed in line with this argument.

17.1

Introduction and Background

In high-income and low-income countries alike, fertility rates have fallen, life expectancy at birth has risen, and the mean age of the population has increased substantially over the last decades (Bloom et al., 2011). Table 17.1 shows data on total fertility rates (TFR), life expectancy at birth (LEXP), and the old-age dependency ratio (OADR) from 1960 to 2018 for the Group of Seven (G7) countries plus China and the Russian Federation (henceforth Russia) to illustrate this development (World Bank, 2020). These data show that the TFR has fallen far below the replacement rate of 2.1 children per woman that would lead to a stable population in the long run in all countries as of 2018. Over the same period, LEXP has increased by more than 10 years, except in Russia and the United States, where the increase was less pronounced. Mainly due to declining fertility and rising life expectancy, the OADR has more than doubled in many countries. These demographic developments have many economic consequences, such as reductions in the relative size of the labor force, which put upward pressure on wages and downward pressure on interest rates; a rise in savings, which reinforces the downward pressure on interest rates; and increased financial needs for social security systems and pension schemes, which put upward pressure on tax rates and social security contributions (Gruber and Wise, 1998). Declining fertility may also be a driving force for higher education investments via the quality-quantity trade-off effect (Becker and Lewis, 1973; Becker and Tomes, 1976; Becker, 1993; Galor and DOI: 10.4324/9781003150398-20

317

Ana L. Abeliansky and Klaus Prettner Table 17.1 Total fertility rate (TFR), life expectancy at birth (LEXP), and the old-age dependency ratio (OADR) of people older than 64 in relation to the working-age population in the G7 countries plus China and Russia in 1960 and 2018

Source: World Bank (2020).

Weil, 2000; Galor, 2005, 2011; Li and Zhang, 2007). Increasing life expectancy reinforces this effect via the Ben-Porath mechanism (Ben-Porath, 1967; Cervellati and Sunde, 2013; Strulik and Werner, 2016). Because other contributions in this volume cover these and related aspects extensively (B¨orsch-Supan, 2023; de Grip, 2023; Hurd and Rohwedder, 2023; Piggot et al., 2023; Lee, 2023; Prskawetz and Sanchez-Romero, 2023; Queisser et al., 2023; Scott, 2023; Skirbekk, 2023), the rest of this chapter focuses on the effects of demographic developments on automation. Against the described demographic backdrop, the ongoing trend of automation, which has accelerated after the global economic and financial crisis, may be a silver lining as it could mitigate labor shortages. Figure 17.1 displays the worldwide stock of operative industrial robots according to International Federation of Robotics data (International Federation of Robotics, 2016, 2017, 2018a). The stock of robots was negligible until the end of the 1980s, started to increase slowly afterward and took off after the global economic and financial crisis. Since then less than 1 million industrial robots in 2008 increased roughly threefold to almost 2.8 million in 2019. This is, of course, faster than economic growth and population growth in any country in the world, which implies an intensification of automation, i.e., an increase in robot density (the number of industrial robots per 1,000 manufacturing workers). Figure 17.2 follows Abeliansky and Prettner (2017, 2023) and displays a world map of robot density according to the International Federation of Robotics database (2019), while Figure 17.32 shows a world map of the OADR. In both figures, the countries with a darker color have either a higher robot density (Figure 17.2) or a higher OADR (Figure 17.3). Robot density tends to be higher in countries with a higher OADR, suggesting the silver lining on the horizon that we have alluded to. In addition, robot density is particularly high in Germany, Italy, and Japan, which are ageing rapidly. By contrast, robot density is lower in countries such as France and the United Kingdom, which are not ageing as quickly because they face comparatively high fertility and/or strong inward migration from abroad (cf. Prettner and Bloom, 2020). 318

Automation and Ageing

Figure 17.1 Worldwide stock of operative industrial robots in millions according to the International Federation of Robotics (2016, 2017, 2018a). Note: Values for years that are not reported by the International Federation of Robotics are interpolated.

[0,1.14] (1.14,15.4] (15.4,58.3] (58.3,600] No data

Figure 17.2 Average robot density, 2018. Note: Robots per 100,000 workers in 2018. Data source: International Federation of Robotics (2019) database. A darker color indicates a higher robot density. The cutoffs correspond to the 25th, 50th, 70th, and 100th percentiles.

17.2

The Role of Automation in Compensating for Ageing: A Simple Theoretical Illustration and the Current State of Empirical Research

As mentioned before, automation could help alleviate the burden of population ageing because robots replace workers, and therefore they substitute for low fertility when it comes to labor market consequences. We illustrate the main mechanisms in a more formal manner. To this end, consider a Cobb-Douglas production function of the form Yt = Ktα (Lt + Pt )1−α ,

(1)

where Yt denotes aggregate output, Kt is the stock of traditional physical capital (machines, assembly lines, production halls, etc.), Lt is employment, Pt refers to the stock of automation 319

Ana L. Abeliansky and Klaus Prettner

[0,5.9] (5.9,10] (10,21.4] (21.4,100] No data

Figure 17.3 Old-age dependency ratio, 2018. Note: Old-age dependency ratio: ratio of people older than 64 relative to the 15–64-year-old workingage population in 2018. Data source: World Bank (2020). A darker color indicates a higher old-age dependency. The cutoffs correspond to the 25th, 50th, 70th, and 100th percentiles.

capital (robots, three-dimensional or 3D printers, and other smart machines that operate on their own), and α ∈ (0, 1) refers to the elasticity of output with respect to traditional physical capital. In this specification, we follow Abeliansky and Prettner (2017, 2023) and Prettner (2019) in assuming that automation capital is a perfect substitute for labor. While this conforms to the very definition of automation, the main insights also hold true when labor and automation capital are imperfect (gross) substitutes as in Steigum (2011), Lankisch et al. (2019), Cords and Prettner (2022), and Gasteiger and Prettner (2022).3 We see immediately from this production function that increases in traditional physical capital Kt cannot easily compensate for a decline in labor Lt because operating traditional machines and staffing assembly lines require human workers in the absence of robots. Thus, compensating declining Lt with more Kt becomes increasingly difficult because of the diminishing marginal productivity of Kt . This problem does not occur, however, when we consider compensating a decline in Lt by investing in robots and 3D printers. In the case of perfect substitutability (the strict definition of automation), industrial robots can fully replace assembly line workers, and 3D printers can fully replace workers who produce customized parts and prototypes. Thus, an increase in Pt can perfectly compensate any decline in Lt in such a setting. To see this formally, assume that there is perfect competition in goods and factor markets, such that production factors are remunerated according to their marginal value product. In this case, the wage rate of workers (wt ) would equal the rate of return on automation capital (Rtautom ) and be given by  α Kt autom wt = Rt+1 = (1 − α) . (2) Lt + Pt By contrast, the rate of return on traditional physical capital would be given by   Lt + Pt 1−α trad Rt+1 = α . (3) Kt We observe that, ceteris paribus, a decline in Lt raises the wage rate, wt , and decreases the rate trad . Both of these developments foster automation of return on traditional physical capital, Rt+1 320

Automation and Ageing

investments because, (1) as labor becomes more expensive, the adoption of industrial robots becomes increasingly attractive and (2) as the rate of return on traditional capital decreases, this renders investing in the other type of capital (automation) more attractive. In an interior equilibrium with rational investors, both effects imply higher investment in automation and lower investment in traditional physical capital in the face of population ageing until the rates of return on both types of capital are equal again. Overall, we therefore expect to see more investment in automation in those countries that are ageing most rapidly.4 Two studies have estimated empirically the idea that population ageing leads to higher incentives to automate production. Abeliansky and Prettner (2017, 2023) propose a simple overlapping generations setting with automation as captured by the production function depicted in Equation (1). As is standard in this literature, population ageing is conceptualized as a decline in fertility, which is its most important driver. The authors show that a decrease in population growth raises the number of industrial robots per 1,000 manufacturing workers over and above the level that would be expected by the implied reduction in capital dilution. Controlling for income per capita, investment, and trade openness (besides an array of other control variables), Abeliansky and Prettner find that a decrease in population growth by 1 percent leads to a 2 percent increase in the growth rate of robot density. These results align with those of Acemoglu and Restrepo (2022), who analyze the effects of ageing on robot adoption by regressing robot density on the share of the population over the age of 56 and various important control variables. They show that a 10 percentage point increase in the share of the population over the age of 56 implies 1.6 additional industrial robots per 1,000 workers. Given the observed increase in their ageing measure and in the stock of industrial robots per 1,000 workers between 1993 and 2014, Acemoglu and Restrepo (2022) conclude that population ageing alone explains around 35 percent of the observed cross-country differences in robot adoption over this period. This strong effect shows the importance of demographic changes on the underlying incentives to invest in automation. Acknowledging that automation is, to a certain extent, an endogenous response to ageingdriven labor supply reductions could help to explain why we have not (yet?) observed labor shortages in the face of population ageing.5

17.3

Potential Displacement of Workers by Robots

To what extent could automation alleviate the burden of ageing by replacing workers? Most of the estimates of how many workers robots could replace are based on the technological feasibility of substituting workers with robots (cf. Frey and Osborne, 2017; Arntz et al., 2017; McKinsey Global Institute, 2017; Nedelkoska and Quintini, 2018). These estimates vary widely, depending on the country under consideration and the method used (e.g., whether the estimates are based on the replacement of occupations as in Frey and Osborne, 2017, or on the replacement of tasks within jobs as in Arntz et al., 2017). Most estimates suggest that robots could do 9–47 percent of all jobs until about 2030. While focusing on the technological feasibility of substituting workers with robots yields important benchmark estimates, doing so disregards economic considerations and general equilibrium effects that may greatly reduce the actual number of jobs that robots could (economically) do. The main reasons are that (1) replacing an inexpensive, long-lived worker without maintenance requirements with an expensive, often relatively short-lived and maintenance-intensive robot may not pay off; (2) producing robots requires time and resources such that imagining a situation in which enough robots could be built over the coming decade to replace billions of workers is difficult; and (3) the deployment of robots may reduce the prices 321

Ana L. Abeliansky and Klaus Prettner

of goods produced with robots and increase the income of capital owners. Rising (real) incomes imply an increasing demand for all goods, not only those produced by robots but also those produced by workers (e.g., personal services). Thus, employing robots instead of workers in some parts of the economy could be compensated by more (human) employment in other parts (Acemoglu and Restrepo, 2018; Prettner and Bloom, 2020). Accounting for economic considerations in replacing workers with robots and potential bottlenecks in robot production, Abeliansky et al. (2020a) calculate scenarios of the number of jobs that robots could substitute using information on current trends in robot production as inferred from International Federation of Robotics data (International Federation of Robotics, 2018b). In the baseline scenario of Abeliansky et al. (2020a), the number of industrial robots until 2030 grows according to the trend observed from 2010 to 2017. They also consider two alternative scenarios: a low robot adoption situation in which the growth rate from 2018 to 2030 is 50 percent lower than the trend growth rate from 2010 to 2017 and a high robot adoption scenario in which the growth rate from 2018 to 2030 is 50 percent higher than in the baseline scenario. In the next step, Abeliansky et al. (2020a) use the stock of robots that is predicted for the year 2030 and assess how many workers robots can replace based on the estimates from Acemoglu and Restrepo (2017b) for the United States and Dauth et al. (2017) for Germany. Because the former contribution estimates that each additional robot replaces a comparatively high number of 6.2 manufacturing workers, while the latter paper’s estimate is a substantially smaller number of 2 workers, these two studies span a plausible range of replacement scenarios.6 In the baseline scenario of medium robot adoption, Abeliansky et al. (2020a) find that robots could replace 12–38 million jobs worldwide until 2030 with the lower number referring to the low-replacement scenario and the higher number to the high-replacement scenario. In the high robot adoption and high displacement scenario, this number rises to 68 million. This estimate is a bit higher than the Bloom et al. (2019) estimate, which relies on International Federation of Robotics (2017) predictions as their baseline trend for robot adoption. Although these numbers look high, they do not even come close to the lower end of the estimates that are based on the possibility for technological substitution (Arntz et al., 2017). To summarize, economic considerations significantly lower the scope for automation. Thus, while a substantial part of the demographically induced decline in the labor force could be compensated by automation, fears of widespread “technological unemployment” due to robot adoption seem (as of yet) to be unfounded.

17.4

The Effects of Automation on Health

One dimension of the demography-automation nexus that the literature has so far neglected refers to the effects automation has on workers in terms of their health. While introducing industrial robots might improve the mental health of workers by alleviating their tasks or making them more interesting, the labor market effects of automation could have adverse effects.7 For example, the fear of “technological unemployment” might increase stress/dissatisfaction at work, thereby decreasing labor productivity. Evidence so far is scarce and mostly relates to countries with high robot adoption rates. Morikawa (2017) relies on a cross-sectional survey of Japanese individuals and finds that about 30 percent of the population fears losing their jobs to artificial intelligence and automation—especially the young (aged between 20 and 30 years old) and those working in manufacturing and in clerical occupations. By contrast, those with a higher education, commonly referred to as “high-skilled,” fear the least. Women seem to fear losing their jobs less, on average. Similar evidence from Schwabe and Castellacci (2020) focuses on the Norwegian case. They find that recent waves of automation have induced fear 322

Automation and Ageing

in about 40 percent of currently employed workers that a smart machine might do their work in the future. Further evidence suggests that those performing low-skilled jobs fear the most. Schwabe and Castellacci (2020)’s identification strategy relies on instrumenting the possibility that a worker’s job can be performed by a machine to later analyze the relation between this fear and job satisfaction. Using the German Socioeconomic Panel, Abeliansky and Beulmann (2019) find that in Germany increased robot adoption is associated with lower mental health of workers (measured as a composite index). Their heterogeneity analysis suggests that those in the middle-skill and middle-occupation groups are those most affected by increased robot adoption as are those who perform more routine tasks. The effect is driven by a particular subcomponent of the index, which considers whether individuals feel that they are achieving less than they would like at work and whether they are less thorough in their work, which the authors interpret as stress at work. The transmission channel is that individuals worry more about their economic situation as automation increases. The latter also affects their income.8 In terms of physical health, Gunadi and Ryu (2021) investigate whether increased robot adoption in U.S. Metropolitan Statistical Areas is associated with better or worse reported health. Evidence suggests that lower-skilled workers in areas with high robot adoption tend to improve their health status and have less work disability. Introducing industrial robots seems to benefit workers by making their jobs less physically demanding. In addition, further results suggest that workers improve their “bad behavior” because they reduce their smoking intensity (for everyday smokers).9 Along the same lines, Gihleb et al. (2022) find that increased robot adoption in the United States is associated with fewer work-related injuries and with fewer “high-burden” jobs in the manufacturing sector. However, further evidence from Gihleb et al. (2022) contradicts the results from Gunadi and Ryu (2021) in terms of “bad behaviors,” finding that increased robot exposure is associated with more deaths due to drug or alcohol abuse. In addition, Gihleb et al. (2022) find that higher robot adoption is associated with more “mentally unhealthy days.” Related to Germany, the same authors investigate the physical health of individuals and find that increased robot exposure is associated with less disability, work accidents, and physical intensity of jobs. Overall, the literature has documented some negative effects of increased robot exposure, particularly on the mental health of young workers and workers in low-skill jobs. Evidence is less clear in terms of physical health in the United States. These aspects are important in the context of ageing because reduced mental health and physical health could reduce life expectancy. Other indirect effects of automation on demographics include those reported by Anelli et al. (2021), who find that a more intense robot use leads to a decline in new marriages and an increase in divorces and cohabitations when analyzing data from the United States. The authors claim that this is partly due to rising economic uncertainty and a reduction in the relative marriage-market value of men. These effects could have direct implications on ageing not only via lower fertility but also because marriage is associated with higher life expectancy (Sch¨unemann et al., 2020).

17.5

Important Future Research Questions

The following are important areas for future research on the relation between ageing and automation. The effects of automation in diagnosing and treating ageing-related diseases on health outcomes and on medical costs. Medical apps and robots are already very proficient in predicting and diagnosing diseases, e.g., by helping radiologists to screen X-rays, by helping to monitor cardiovascular 323

Ana L. Abeliansky and Klaus Prettner

functions and summon help in emergencies via wearable sensors, or by assisting care assistants and nurses in their daily jobs with robotic exoskeletons when they need to lift patients. All this could imply considerable relief for medical and care professionals in dealing with ageing-related diseases that are bound to increase due to observed demographic changes (Bloom et al., 2011, 2014, 2020; Chen et al., 2018; Prettner and Bloom, 2020). 3D printing medical implants, prostheses, hearing aids, replacement teeth, and other medical products should help to reduce the price of these products, increase their quality, and reduce the associated waiting times. Because mainly older adults demand these products, such developments could improve the well-being of older people, and because ageing-related diseases are such an important contributor to the economic burden of population ageing, automation technologies could have considerable scope in containing the costs of age-related healthcare. Whether and to what extent these effects will materialize is a highly promising area for future research. In addition, a need exists for research on the design of policies and regulations in ensuring beneficial effects of these technologies for older adults. The impact that automation will have on the life of older adults. Robotic exoskeletons could help older adults to stay active for longer and to cope with the ageing-driven decline in physical strength. In general, automation could extend the working life of older adults by decreasing the physical burden of their jobs. This would help to ameliorate the demographic challenges of ageing countries. In addition, autonomous driving technology could help older adults to stay mobile for much longer and therapeutic robots could even alleviate loneliness in old age (see, for example Ford, 2015; The Economist, 2016, 2019; Financial Times, 2019; Abeliansky et al., 2020a). All this could imply tremendous shifts in the way that older adults live and in the way families organize the task of caring for their older members (Prettner and Bloom, 2020). Whether automation will increase or decrease international migration. While automation technologies reduce the upward pressure on wages in high-income countries that is associated with demographic change, it also leads to reshoring. Thus, it lessens the ability of low-income countries to develop via industrialization due to the cost advantage emanating from lower wages (Abeliansky et al., 2020b; Krenz et al., 2021). As a consequence, automation technologies reduce the wage growth prospects in low-income countries, which could contribute to increasing migration pressure on an international scale (Zhou et al., 2019). The ways that automation affects demographic change. Plausibly, automation could exert a reverse causal effect on demographic outcomes, in general, and on ageing, in particular (Prettner and Bloom, 2020). If routine workers are displaced by robots, their incomes shrink, which could have two opposing effects on fertility. On the one hand, lower incomes could imply that families would not be able to afford to have children anymore, thereby reducing fertility. In addition, parents may want to invest more in the education of their children to make them more competitive on labor markets that are characterized by a high degree of automation and technology use. This increase in education tends to imply a decrease in fertility because of the quality-quantity trade-off effect (cf. Becker, 1960; Galor and Weil, 2000; Galor, 2011; Strulik et al., 2013). On the other hand, a decrease in wages (or the direct job displacement and, thus, a decrease in the employment of parents) reduces the opportunity costs of children, which are largely due to parental time requirements (Becker, 1965). According to this channel, automation could increase fertility. On top of these, more conventional and opposing effects of automation on fertility, the potential future widespread availability of sex robots could affect partnership formation, adding another plausible channel by which automation could affect fertility negatively. The question as to whether automation will decrease or increase fertility is surely an important one for future research. 324

Automation and Ageing

17.6

Conclusions

A simple theoretical model of automation suggests that (1) automation can compensate for the labor supply effects of ageing and that (2) the incentives to invest in automation rise with demographic developments that lead to ageing. Empirical evidence is highly consistent with these theoretical findings and suggests that ageing is a strong force in driving robot adoption. In addition, projections of the number of jobs that robots could substitute suggest sizeable effects, although projections that consider economic factors (whether employing a robot instead of a human pays off) do not come close to the estimates that are solely based on technological feasibility of substitution. As far as the effects of automation on (individual) ageing are concerned, some studies suggest positive physical health effects because manual labor may become less physically demanding. However, automation can also have adverse effects, particularly on mental health, because many workers fear losing their jobs to robots. The mental health consequences could then also exert negative effects on physical health outcomes. The following areas are highly promising for further research: (1) the effects of automation in old-age-related healthcare on health outcomes and on treatment costs, (2) the effects of automation on the quality of life of older adults and on their labor-force participation, (3) the effects of automation and of automation-driven reshoring on international migration, and (4) the potential reverse causal effects of automation on fertility and thereby on ageing in the long run.

Notes 1 This article reflects the state of knowledge at the time of writing, which was December 2020. 2 The percentiles in Figure 17.3 are based on more countries than those of Figure 17.2 due to data availability. 3 For an overview, see Prettner and Bloom (2020). 4 For more thorough theoretical analyses of the relationship between ageing and technological change, in general, and between ageing and automation, in particular, see Acemoglu and Restrepo (2017a, 2022), Prettner and Trimborn (2017), Baldanzi et al. (2019), Gehringer and Prettner (2019), St¨ahler (2021), and Belyakov et al. (2021). 5 In terms of another demographic process, Faber et al. (2019) analyze the migration response of U.S. American workers to two shocks: import competition from China and the use of industrial robots. While both shocks steeply reduce manufacturing employment, robots cause a strong reduction in population size (while trade with China does not). 6 The estimates of the published version of Acemoglu and Restrepo (2017b, 2022) lie within this range. 7 Hessel et al. (2017) show that people with long-standing illnesses face a greater risk of losing their jobs due to automation than healthy workers do. 8 Gihleb et al. (2022) do not find evidence on high psychological intensity (compared with Abeliansky and Beulmann, 2019). The difference is probably due to the fact that Gihleb et al. (2022) use a timeinvariant index of psychological intensity (which also applies to the physical burden intensity) as the dependent variable that is associated with each type of job. The variation that they exploit in their analysis seems to be coming from people changing jobs in terms of intensity and not from higher robot exposure that workers themselves are experiencing and how this relates to their (changing) mental health. 9 The authors argue that increased automation could be considered as a shock (the authors compare it to recessions that are associated with healthier behaviors given the reduction in work-stress).

References ABELIANSKY, A. L., ALGUR, E., BLOOM, D. E., AND PRETTNER, K. (2020a): “The future of work: Meeting the global challenges of demographic change and automation,” International Labour Review, 159(3): 285–306.

325

Ana L. Abeliansky and Klaus Prettner

ABELIANSKY, A. L., AND BEULMANN, M. (2019): “Are they coming for us? Industrial robots and the mental health of workers,” cege Discussion Paper No. 379. Center for European, Governance and Economic Development Research (cege), University of G¨ottingen. ABELIANSKY A. L., MARTINEZ-ZARZOSO, I., AND PRETTNER K. (2020b): “3D printing, international trade, and FDI,” Economic Modelling, 85: 288–306. ABELIANSKY, A. L., AND PRETTNER, K. (2017): “Automation and demographic change.” Hohenheim Discussion Papers in Business, Economics, and Social Sciences 05-2017, University of Hohenheim. ABELIANSKY, A. L., AND PRETTNER, K. (2023): Automation and population growth: Theory and cross-country evidence. Journal of Economic Behavior & Organization 208: 345–358. ACEMOGLU, D., AND RESTREPO, P. (2017a): “Secular stagnation? The effect of ageing on economic growth in the age of automation,” American Economic Review, Papers & Proceedings, 107(5): 174–179. ACEMOGLU, D., AND RESTREPO, P. (2017b): “Robots and jobs: Evidence from US labor markets.” NBER Working Paper No. 23285, National Bureau of Economic Research, Cambridge, MA. ACEMOGLU, D., AND RESTREPO, P. (2022): “Demographics and automation,” The Review of Economic Studies, 89: 1–44. ACEMOGLU, D., AND RESTREPO, P. (2018): “The race between man and machine: Implications of technology for growth, factor shares and employment,” American Economic Review, 108(6): 1488–1542. ACEMOGLU, D., AND RESTREPO, P. (2020): “Robots and jobs: Evidence from US labor markets,” Journal of Political Economy, 128(6): 2188–2244. ANELLI, M., GIUNTELLA, O., AND STELLA, L. (2021): ”Robots, marriageable men, family, and fertility,” Journal of Human Resources, 1020–11223R1. ARNTZ, M., GREGORY, T., AND ZIERAHN, U. (2017): “Revisiting the risk of automation,” Economics Letters, 159(C): 157–160. BALDANZI, A., PRETTNER, K., AND TSCHEUSCHNER, P. (2019): “Longevity-induced vertical innovation and the trade-off between life and growth,” Journal of Population Economics, 32(4): 1293–1313. BECKER, G. S. (1960): “An economic analysis of fertility.” In: Roberts, G. B. (ed.): Demographic and Economic Change in Developed Countries, Cambridge, MA: National Bureau of Economic Research, Vol. 11, pp. 209–231. BECKER, G. S. (1965): “A theory of the allocation of time,” The Economic Journal, 75(299): 493–517. BECKER, G. S. (1993): A Treatise on the Family. Cambridge, MA: Harvard University Press. BECKER, G. S., AND LEWIS, H. G. (1973): “On the interaction between the quantity and quality of children,” Journal of Political Economy, 81(2): 279–288. BECKER, G. S., AND TOMES, N. (1976): “Child endowments and the quantity and quality of children,” The Journal of Political Economy, 84(4): 143–162. BELYAKOV, A. O., KURBATSKIY, A. N., AND PRETTNER, K. (2021): “The growth effects of anticipated versus unanticipated population ageing,” Journal of Mathematical Economics, 94: 1–11. BEN-PORATH, Y. (1967): “The production of human capital and the life cycle of earnings,” Journal of Political Economy, 75(4): 352–365. BLOOM, D. E., CAFIERO, E. T., JAN E´ -LLOPIS, E., ABRAHAMS-GESSEL, S., BLOOM, L. R., FATHIMA, S., FEIGL, A. B., GAZIANO, T., MOWAFI, M., PANDYA, A., PRETTNER, K., ROSENBERG, L., SELIGMAN, B., S TEIN, A., AND W EINSTEIN, C. (2011): The Global Economic Burden of Non-communicable Diseases. Geneva: World Economic Forum. BLOOM, D. E., CAFIERO-FONSECA, E. T., MCGOVERN, M. E., PRETTNER, K., STANCIOLE, A., WEISS, J., BAKKILA, S., AND ROSENBERG, L. (2014): “The macroeconomic impact of noncommunicable diseases in China and India: Estimates, projections, and comparisons,” The Journal of the Economics of Ageing, 4 (December 2014): 100–111. BLOOM, D. E., CHEN, S., KUHN, M., MCGOVERN, M., OXLEY, L., AND PRETTNER, K. (2020): “The economic burden of chronic diseases: Estimates and projections for China, Japan, and South Korea,” The Journal of the Economics of Ageing, 17: 1–13. BLOOM, D. E., MCKENNA, M., AND PRETTNER, K. (2019): “Global employment and decent jobs, 2010–2030: The forces of demography and automation,” International Social Security Review, 72(3): 43– 78. ¨ BORSCH -SUPAN, A. (2023): “Ageing and productivity.” In: Bloom, D. E., Sousa-Poza, A., and Sunde, U. (eds.), The Routledge Handbook of the Economics of Ageing (forthcoming, this volume). CERVELLATI, M., AND SUNDE, U. (2013): “Life expectancy, schooling, and lifetime labor supply: Theory and evidence revisited,” Econometrica, 81(5): 2055–2086.

326

Automation and Ageing

CHEN, S., KUHN, M., PRETTNER, K., AND BLOOM, D. (2018): “The macroeconomic burden of noncommunicable diseases in the United States: Estimates and projections,” PLOS ONE, 13(11): e0206702. CORDS, D., AND PRETTNER, K. (2022): “Technological unemployment revisited: Automation in a search and matching framework,” Oxford Economic Papers, 74(1): 115–135. DAUTH, W., FINDEISEN, S., SUEDEKUM, J., AND WOESSNER, N. (2017): “German robots—the impact of industrial robots on workers,” CEPR Discussion Paper 12306, Centre for Economic Policy Research, London. DE G RIP, A. (2023): “Education and ageing.” In: Bloom, D. E., Sousa-Poza, A., and Sunde, U. (eds.), The Routledge Handbook of the Economics of Ageing (forthcoming, this volume). FABER, M., SARTO, A., AND TABELLINI, M. (2019): “The impact of technology and trade on migration: Evidence from the US,” Harvard Business School BGIE Unit Working Paper (20-071), Cambridge, MA. FINANCIAL TIMES. (2019): “Robots/ageing Japan: I, carebot.” Available at https://www.ft.com/ content/314c65a8-8829-11e9-a028-86cea8523dc2 [accessed on November 24, 2019]. FORD, M. (2015): Rise of the Robots: Technology and the Threat of a Jobless Future. New York: Basic Books. FREY, C. B., AND OSBORNE, M. A. (2017): “The future of employment: How susceptible are jobs to computerisation?,” Technological Forecasting and Social Change, 114(C): 254–280. GALOR, O. (2005): “From stagnation to growth: Unified growth theory.” In: Aghion, P. and Durlauf, S. (eds.), Handbook of Economic Growth, Amsterdam: Elsevier, volume 1, chapter 4, pp. 171–293. GALOR, O. (2011): Unified Growth Theory. Princeton, NJ: Princeton University Press. GALOR, O., AND WEIL, D. (2000): “Population, technology, and growth: From Malthusian stagnation to the demographic transition and beyond,” The American Economic Review, 90(4): 806–828. GASTEIGER E., AND PRETTNER K. (2022): “Automation, stagnation, and the implications of a robot tax,” Macroeconomic Dynamics, 26(1): 218–249. GEHRINGER, A., AND PRETTNER, K. (2019): “Longevity and technological change,” Macroeconomic Dynamics, 23(4): 1471–1503. GIHLEB, R., GIUNTELLA, O., STELLA, L., AND WANG, T. (2022): “Industrial robots, workers’ safety, and health,” Labour Economics, 78: 102205. GRUBER, J., AND WISE, D. (1998): “Social security and retirement: An international comparison,” American Economic Review, 88(2): 158–163. GUNADI, C., AND RYU, H. (2021): “Does the rise of robotic technology make people healthier?” Health Economics, 30(9): 2047–2062. HESSEL, O., CHRISTIANSEN, S., AND SKIRBEKK, V. (2017): “Poor health as a potential risk factor for job loss due to automation: The case of Norway,” Occupational and Environmental Medicine, 75(3): oemed-2017-104349. HURD, M., AND ROHWEDDER, S. (2023): “Assessments of adequacy of retirement resources.” In: Bloom, D. E., Sousa-Poza, A., and Sunde, U. (eds.), The Routledge Handbook of the Economics of Ageing (forthcoming, this volume). INTERNATIONAL FEDERATION OF ROBOTICS. (2016): “Executive summary world robotics 2016 industrial robots.” Available at https://ifr.org/news/world-robotics-report-2016 [accessed on August 1, 2018]. INTERNATIONAL FEDERATION OF ROBOTICS. (2017): “Executive summary world robotics 2017 industrial robots.” Available at https://ifr.org/downloads/press/Executive Summary WR 2017 Industrial Robots.pdf [accessed on August 1, 2018]. INTERNATIONAL FEDERATION OF ROBOTICS. (2018a): “Executive summary world robotics 2018 industrial robots.” Available at https://ifr.org/downloads/press2018/Executive Summary WR 2018 Industrial Robots.pdf [accessed on February 10, 2019]. INTERNATIONAL FEDERATION OF ROBOTICS. (2018b): “World robotics industrial robots and service robots.” Available at https://ifr.org/worldrobotics/ [accessed on March 7, 2019]. INTERNATIONAL FEDERATION OF ROBOTICS. (2019): “World robotics industrial robots and service robots.” Available at https://ifr.org/worldrobotics/ [accessed on December 21, 2020]. KRENZ, A., PRETTNER, K., AND STRULIK, H. (2021): “Robots, reshoring, and the lot of low-skilled workers,” European Economic Review, 136: 103744. LANKISCH, C., PRETTNER, K., AND PRSKAWETZ, A. (2019): “How can robots affect wage inequality?,” Economic Modelling, 81(September 2019): 161–169.

327

Ana L. Abeliansky and Klaus Prettner

LEE, R. (2023): “Economic growth and ageing.” In: Bloom, D. E., Sousa-Poza, A., and Sunde, U. (eds.), The Routledge Handbook of the Economics of Ageing (forthcoming, this volume). LI, H., AND ZHANG, J. (2007): “Do high birth rates hamper economic growth?,” Review of Economics and Statistics, 89(1): 110–117. MCKINSEY GLOBAL INSTITUTE. (2017): “Jobs lost, jobs gained: Workforce transitions in a time of automation.” Available at https://www.mckinsey.com/ /media/mckinsey/featured%20 insights/future%20of%20organizations/what%20the%20future%20of%20work%20will%20mean%20for %20jobs%20skills%20and%20wages/mgi-jobs-lost-jobs-gained-report-december-6-2017.ashx [accessed on December 31, 2019]. MORIKAWA, M. (2017): “Who are afraid of losing their jobs to artificial intelligence and robots? Evidence from a survey.” SSPJ Discussion Paper Series No. DP 17-007. Available at http://hermes-ir.lib.hitu.ac.jp/hermes/ir/re/30381/DP17-007.pdf [accessed on August 11, 2021]. NEDELKOSKA, L., AND QUINTINI, G. (2018): “Automation, skill use and training.” OECD Social, Employment and Migration Working Paper No. 202, Paris. PIGGOT, J., KUDRNA, G., AND O’KEEFE, P. (2023): “Pension policy in emerging economies with population ageing: What do we know, where should we go?” In: Bloom, D. E., Sousa-Poza, A., and Sunde, U. (eds.), The Routledge Handbook of the Economics of Ageing (forthcoming, this volume). PRETTNER, K. (2019): “A note on the implications of automation for economic growth and the labor share,” Macroeconomic Dynamics, 23(3): 1294–1301. PRETTNER, K., AND BLOOM, D. E. (2020): Automation and Its Macroeconomic Consequences: Theory, Evidence, and Social Impacts. Cambridge, MA: Academic Press. PRETTNER, K., AND TRIMBORN, T. (2017): “Demographic change and R&D-based economic growth,” Economica, 84(336): 667–681. PRSKAWETZ, A., AND SANCHEZ-ROMERO, M. (2023): “Social security reforms in heterogeneous ageing populations.” In: Bloom, D. E., Sousa-Poza, A., and Sunde, U. (eds.), The Routledge Handbook of the Economics of Ageing (forthcoming, this volume). QUEISSER, M., BOULHOL, H., AND MACIEJ, L. (2023): “Trends in pension reforms in OECD countries.” In: Bloom, D. E., Sousa-Poza, A., and Sunde, U. (eds.), The Routledge Handbook of the Economics of Ageing (forthcoming, this volume). ¨ SCH UNEMANN , J., STRULIK, H., AND TRIMBORN, T. (2020): “The marriage gap: Optimal ageing and death in partnerships,” Review of Economic Dynamics, 36: 158–176. SCHWABE, H., AND CASTELLACCI, F. (2020): “Automation, workers’ skills and job satisfaction,” PLOS ONE, 15(11): 1–26. SCOTT, A. J. (2023): “Labor supply and ageing.” In: Bloom, D. E., Sousa-Poza, A., and Sunde, U. (eds.), The Routledge Handbook of the Economics of Ageing (forthcoming, this volume). SKIRBEKK, V. (2023): “Ageing and cognitive developments.” In: Bloom, D. E., Sousa-Poza, A., and Sunde, U. (eds.), The Routledge Handbook of the Economics of Ageing (forthcoming, this volume). ¨ ST AHLER , N. (2021): “The impact of ageing and automation on the macroeconomy and inequality,” Journal of Macroeconomics, 67: 103278. STEIGUM, E. (2011): “Robotics and growth.” In: Beladi, H., and Choi E. K. (eds.), Frontiers of Economics and Globalization: Economic Growth and Development, Bingley: Emerald Group, chapter 21, pp. 543–557. STRULIK, H., PRETTNER, K., AND PRSKAWETZ, A. (2013): “The past and future of knowledge-based growth,” Journal of Economic Growth, 18(4): 411–437. STRULIK, H., AND WERNER, K. (2016): “50 is the new 30—Long-run trends of schooling and retirement explained by human ageing,” Journal of Economic Growth, 21(2): 165–187. THE ECONOMIST. (2016): “A printed smile. 3D printing is coming of age as a manufacturing technique,” April 30, 2016. THE ECONOMIST. (2019): “Grandma’s little helper. An ageing world needs more resourceful robots,” February 14, 2019. WORLD BANK. (2020): World Development Indicators 1960–2020. Washington, DC: World Bank. ZHOU, Y., BLOOM, D. E., AND TYERS, R. (2019): “Implications of automation for global migration,” Paper presented at the 11th International Symposium on Human Capital and Labor Markets in Beijing, December 13–14, 2019.

328

18 WORKING LIFE—LABOR SUPPLY, AGEING, AND LONGEVITY1 Andrew J. Scott

Abstract The labor supply of older workers will be a key policy issue in the future. An ageing society is leading to an increasing number of older people and longevity to longer lives and changes in the life course. Both forces are increasing labor market participation at older ages. As retirement age extends, it becomes less of a discrete event, more variable in its timing, and characterized by considerable diversity across individuals. Therefore, understanding labor supply at older ages increasingly requires a focus on the intensive margin and hours worked and on the characteristics of work. As tax and benefit schemes change to reduce incentives to retire and as health at older ages improves, an issue emerges as to what makes older workers distinctive. The corollary is to either subsume older workers in more general theories of labor supply or to develop a richer understanding of ageing and its impact on preferences.

18.1

Introduction

Two trends—ageing and longevity—make the labor supply of older workers a key economic issue, now and in the future. In response to a demographic transition, an “ageing society” is emerging, which is reflected in a changing age structure of the population. By 2050, the United Nations (2020) estimates that one in four persons living in Europe or North America will be aged over 65 years. The number of persons globally over the age of 80 years is projected to triple, from 143 million in 2019 to 426 million in 2050. This shift leads to economic concerns of declining gross domestic product (GDP) growth due to reductions in the working-age population and worries regarding public finances in the face of rising pension and health costs (Aksoy et al., 2019; Cooley and Henriksen, 2018). From this perspective, understanding how to keep people working for longer, both pre- and postretirement are needed to prevent stalling economic growth (Maestas and Zissimopoulos, 2010). An ageing society reflects an aggregate phenomenon, but from an individual perspective, a striking feature of the 20th century was a marked increase in life expectancy (Oeppen and Vaupel, 2002). Eggleston and Fuchs (2012) emphasize that the demographic transition has become a longevity transition with most of the gains in life expectancy now occurring at later ages. Across high-income countries, in the decade up to 2019, most life expectancy gains arose from declines DOI: 10.4324/9781003150398-21

329

Andrew J. Scott

in mortality in the years after age 65. In 1960, a 65-year-old male in the United Kingdom could expect a further 12 years of life; by 1990, this had risen to 14 years and by 2018 to nearly 19 years. This increase in life expectancy raises broad issues as how to structure the life course to make best use of this additional time (Gratton and Scott, 2016). It also raises the possibility of a longevity dividend (Olshansky et al., 2006) as longer lives should be a positive for the economy if health and productivity can be maintained (Scott, 2021). Achieving the economic dimension of a longevity dividend requires supporting employment and productivity in later years. Determining the length of working life, how to support longer careers, what type of work to do and when to do it are key issues for individuals as they adapt to these longer lives. Considerable research and policy discussion around labor supply at older ages focus on retirement age. This is understandable given the 20th century created retirement as a mass phenomenon defined by a common age at which withdrawal from the labor market occurred.2 The emergence of this new stage of life had sizable practical and conceptual impacts: (1) Employment at ages above 65 years reduced dramatically. For example, between 1890 and 1990, U.S. male labor force participation for those aged over 65 fell from 74 percent to 16 percent and in France between 1890 and 2000 from 54 percent to 2.1 percent. (2) Substantial social experimentation and commercial innovation occurred to define the use of leisure post retirement.3 (3) A focus on the extensive margin of labor supply at older ages occurred, centering on binary outcomes of full-time work or retirement (Blundell et al., 2016). (4) A three-stage lifecycle model of education, work, and retirement was established, e.g., Modigliani and Brumberg (1954). (5) A cut-off age defined when individuals are assumed to be unproductive, leading to the definition of an old-age dependency ratio (OADR). (6) By emphasizing nonemployment and dependency at older ages, this new stage of life abstracted from various ways in which individuals make non-market-based contributions to their family, community, and society (Lee, 2010; Bloom et al., 2020). (7) This emergence also led to a lack of focus on economic behavior at older ages. The third stage of life in a standard life-cycle model is seen as empty of substantive economic decisions. Rather it is a time when plans made in the past unfold as assets decumulate, labor supply is zero, and consumption declines with age and health. What is striking is how the rising importance of labor supply among older adults requires focusing on the opposite issues. In particular, (1) In high-income countries, where longevity increases are most pronounced, proportions of people working beyond 65 years are rising. Since its 1990 low, U.S. male labor force participation has risen to reach 25 percent and continues to increase. (2) A growing popular focus centers on how to be active in later years rather than just consume leisure, either through longer careers (Conley, 2018) or through broader social contribution (Friedman, 2018). (3) The labor market experience of older workers is disproportionately involved in part-time and flexible work. The intensive margin of labor supply seems more important for this group as are qualitative issues about the type of work done. (4) As careers lengthen and the nature of work at older ages shifts, there is the nascent appearance of a multistaged working life (Gratton and Scott, 2016), bridge jobs spanning the gap between full-time work and retirement (Alcover, 2016), and “unretirement” (Maestas, 2010). (5) The definition of “old age” used to construct OADRs faces criticism (Sanderson and Scherbov, 2019). (6) Ageist assumptions that older people are unproductive or should be judged based on their productivity alone are being rejected (Applewhite, 2016). (7) Interest in the motivation of older workers and consumers is growing as they increase in number due to an ageing society and shift behavior due to a longevity agenda. To provide greater insight into these issues, this chapter is structured as follows. The following section documents cross-country facts about labor supply at older ages and considers longer-run 330

Working Life—Labor Supply, Ageing, and Longevity

trends among high-income countries. The next section considers the optimal retirement age and seeks to understand past stylized facts and consider implications for future trends. Greater longevity points toward longer working careers, which in turn raises a broad range of factors that go well beyond retirement age. The next section therefore considers the intensive margin of work and the role of health, education, and wider preferences in determining labor supply at older ages. The subsequent section develops these themes and considers whether a better understanding of the preferences of older individuals requires a more foundational theory of ageing. A final section concludes.

18.2

Cross-Country and Historical Trends

This section presents cross-country evidence about older age groups in the labor force and how these have changed over time.4

18.2.1

Fact 1. Labor Force Participation Shows an Inverted U-Shaped Relationship with Age

Figure 18.1 shows that labor force participation rates across the world initially rise with age as education is completed and individuals enter the workforce. However, from the mid to late 40s, people begin to withdraw from the labor market. The speed of withdrawal increases with age and is fastest in high-income countries. High-income countries have the lowest level of participation in the labor force among those aged over 65 years and low-income countries the highest.

18.2.2

Fact 2. Labor Force Participation of Men and Women Show Broadly Similar Variations with Age

Figure 18.1 shows that the pattern of entry and withdrawal from the labor market is broadly similar for men and women. However, women tend to withdraw from the labor market slightly before men and are less likely to be in the labor force at all ages, although the differences

Figure 18.1 Labor force participation rates (%) by age, 2019. Source: ILO, https://ilostat.ilo.org/topics/population-and-labour-force/.

331

Andrew J. Scott

reduce with age. Although not shown in Figure 18.1, women are also more likely to engage in part-time work across the life course.

18.2.3

Fact 3. Since the 1990s Labor Force Participation Rates in High-Income Countries Have Been Rising, Reversing a Century-Long Period of Decline

Although Figure 18.1 shows that high-income countries have low labor force participation rates for older workers, this has not always been the case (Figure 18.2). Starting in the second half of the 19th century and accelerating during the 20th century, as public and private pension schemes were extended, labor force participation rates declined from around three-quarters to four-fifths of men to reach less than one in five in the United States and even lower in Europe. However, starting in the mid-1990s, this trend reversed, and participation rates have been rising.

18.2.4

Fact 4. Increased Employment in Older Age Groups Has Become an Important Source of Overall GDP Growth

Table 18.1 shows that the rising participation of older workers combined with growing numbers in these age cohorts has contributed substantially to recent GDP growth. Across Organisation for Economic Co-operation and Development (OECD) countries, rising employment among workers aged over 50 accounts for more than 70 percent of total employment growth post 2010 and in the European Union (EU28) for more than 100 percent. In the OECD, a third of this increase in employment of older workers arose from an increased participation rate with the remainder accounted for by a larger cohort size. In the EU28, in this age group, the split was 60 percent to 40 percent between increased participation rate and cohort size.

18.2.5

Fact 5. While Employment at Older Ages Has Increased, It Is Highest and Increasing Fastest among Those with Higher Levels of Education

Although the rising participation of older workers is a common phenomenon, workers with higher levels of education are more likely to work at older ages. Figure 18.3 shows they have also experienced sharper increases in participation in many countries.

Figure 18.2 Labor force participation rates (%) for men aged over 65 years. Source: Costa (1998) and OECD, https://data.oecd.org/emp/labour-force.htm.

332

Working Life—Labor Supply, Ageing, and Longevity Table 18.1 Percentage of total employment growth explained by age group 50 years plus

Source: OECD, https://data.oecd.org/emp/labour-force.htm.

Table 18.2 Part-time employment as percentage of full-time employment by age, 2018/2019

Source: Eurostat (European data), Current Population Survey (U.S. Data).

18.2.6

Fact 6. Older Workers Are More Likely to Be Engaged in Part-Time Work and Be Self-Employed

The type of work older workers engage in also tends to be different. Table 18.2 shows that involvement in part-time work follows a U-shaped pattern as it is most common in early and later years. 333

Andrew J. Scott

Figure 18.3 Percentage of population employed at older ages by education (Europe 65–74 years, United States over 65 years). Source: Eurostat Labour Force Surveys (European data), Current Population Survey (U.S. data).

Figure 18.4 shows that full-time work remains the most common form of employment for older workers, but this age group engages in part-time work and a broader range of part-time work more frequently than other age groups. In addition to being more engaged in part-time work, older workers are also more likely to be self-employed (Abraham et al., 2020). Between ages 18 and 29, only 19 percent of U.S. workers are self-employed. Between 50 and 64 years of age, that number is around 25 percent, 334

Working Life—Labor Supply, Ageing, and Longevity

rises to 46 percent for those between 65 and 69, and reaches 68 percent for those aged between 75 and 79.

18.2.7

Fact 7. Older Age Groups Remain Active and Productive Outside of the Labor Market

Bloom et al. (2020) show that as involvement in the labor market declines from age 50, contributions in terms of caring (for children or adult members of the household) and volunteering increase until around ages 65–69 years and then diminish. UK evidence shows that those aged 65–74 are most likely to volunteer regularly for charitable activity.5 Eurostat time use surveys (Figure 18.5) suggest that, in general, older individuals spend more time caring for others than average.

18.3

The Age of Retirement

Given the centrality of retirement to discussions of labor supply at older ages, we focus first on shifts in the retirement age. Just as the introduction of mass pension schemes accelerated the

Figure 18.4 Distribution of average weekly hours worked by age group across the OECD. Source: OECD, https://stats.oecd.org/Index.aspx?DataSetCode=USLHRS D.

Figure 18.5 Proportion of time spent caring for others. Source: Eurostat, aggregated from Harmonised European Time Use Surveys.

335

Andrew J. Scott

Figure 18.6 Current and future retirement ages across the OECD. Source: OECD (2019).

emergence of retirement, so have changes in these schemes influenced recent increases in the labor force participation rate at older ages. In response to public finance concerns, governments have taken steps (OECD, 2019) to increase the age at which state pensions are payable, announced future increases (see Figure 18.6), and changed the tax system to provide fewer biases toward retirement.6 Rogerson and Wallenius (2021) made the case that some of these shifts were not connected to an ageing society but were instead unwinding policies enacted in the high unemployment decades of the 1980s. While changes in social security incentives are undoubtedly important in explaining shifts in labor force participation at older ages, Laun and Wallenius (2016) also noted other factors in explaining cross-country variations. OECD policies imply that the proportion of adult life spent in retirement is expected to increase slightly on average (but not in all countries). For the cohort entering the labor force today, the OECD predicts 33.6 percent of adult life spent in retirement compared with 32 percent for those just retiring. In the United Kingdom, the government has explicitly committed to changes in the state pension age such that “people should spend, on average ‘up to one third’ of their adult life over State Pension Age” (UK Government, 2013). In other words, in the face of growing life expectancy policy has been to preserve the structure of a three-stage life but change its key parameters: age at which state pension is payable, the level of the pension, and the contribution rate.

18.3.1

Optimal Retirement

These changes have been made with the aim of achieving fiscal sustainability and with an eye on what is politically feasible. However, there is a deeper issue about how individuals should respond to longer lives and how to allocate the additional time between work and leisure. A natural way to introduce retirement into an optimizing dynamic life-cycle model (Bloom et al., 2014) is to pin down the age of retirement (R) by a condition equating the marginal benefit from work to its marginal disutility, i.e., setting R equal to the a that satisfies the following condition: W (a)xMUC (a) = MDL (a) (1) The left-hand side of (1) gives the marginal benefit of work, where W (a) is the wage paid at age a and MUC (a) is the marginal utility of consumption at age a. This should be set equal 336

Working Life—Labor Supply, Ageing, and Longevity

to the marginal disutility of work at age a, MDL (a). An obvious assumption to make is that the marginal disutility of work increases with age (in Bloom et al., 2014, this is motivated by reference to declining health such that the relative value of leisure increases).7 Equation (1) can help shed light on the different time series and cross-sectional trends of the previous section. Take the case of an increase in life expectancy but no improvement in wages or the relationship between health and age. If retirement age does not change, then lifetime income is spread more thinly across each year, leading to a rise in the marginal utility of consumption. This requires accepting a higher disutility of work, which means a later retirement, which in turn helps finance higher lifetime consumption. If at the same time health is also improving at each age [leading to a lower MDL (a)], then the increase in retirement is even larger. Thus, longer life expectancy leads to an increase in the age of retirement and the more that health improves the bigger is this increase. The impact of higher wages on retirement is more complex, as wages play two roles in life-cycle models. The higher the wage, the more income an individual has and the more consumption and leisure they want to consume, which points toward retirement at earlier ages. However, higher wages also make leisure more expensive [see (1) leading to a substitution effect that increases retirement age]. Bloom et al. (2014) show that for the relative risk aversion case, the income effect dominates if the elasticity of intertemporal substitution is less than one. Under this assumption, higher wages lead to a decline in the age of retirement. This simple framework helps explain the trends of the previous section. Higher-income countries will have lower labor force participation rates at older ages if the income effect dominates. Similarly, rapid improvements in income offsetting increases in life expectancy explain the fall in labor force participation rates during the 20th century in high-income countries. However, as Figure 18.7 shows, more recently, several countries have seen income growth slow while life expectancy continues to rise and so find themselves in the top left quadrant of Figure 18.7. In terms of our model, this requires increases in the retirement age. For emerging markets undergoing a rapid demographic transition, the concern is that income growth will decline faster than improvements in life expectancy, leading to the phenomena of countries becoming old before they become rich.

18.3.2

The Link between Retirement and Longevity

While the link between retirement and life expectancy is positive, assuming no wage growth, the precise link between them depends on the years that are affected by declining mortality and their health status. In other words, the link between optimal retirement and life expectancy depends upon how we age. To analyze this link, consider the following model of Bloom et al. (2014). This is a dynamic intertemporal model where utility is assumed strongly separable in goods and leisure such that c(a)1−β − d∗ F(a) 1−β

(2)

where c(a) denotes consumption at age a, β is the coefficient of relative risk aversion, d∗ is a parameter reflecting the disutility of work, and F(a) denotes frailty (ill health) at age a (see Strulik, 2023, for an overview on modeling frailty and mortality). We assume for simplicity that the choice to work is discrete, i.e., an individual either works full time or does not work. We discuss later the importance of the hours worked decision for older workers. Let frailty, F(a), be defined by F ∗ eφ(a−T) so that ill health increases at the rate φ with age ′ [F (a) > 0] reaching a level F ∗ at age T. Assume that the mortality rate [µ(a)] also depends on 337

Andrew J. Scott

Figure 18.7 Dynamics of life expectancy increases and GDP growth. Source: Life expectancy data from Human Mortality Database; GDP growth from Penn World Table.

frailty8 such that µ(a) = ρ + M ∗ F(a)λ

(3)

Under this Gompertz-Makeham formulation, mortality depends on age-independent risks ρ but also rises with age (senescence) at the rate φλ, reaching the level ρ + M ∗ F ∗λ at age T. We can therefore think of T as a limit to both lifespan and ageing, and it denotes the age at which mortality and ill health reach a maximum. In this formulation, the frailty index F(a) effectively captures the ageing process (see Scott et al., 2021, for an application to valuing improvements in ageing). Increases in life expectancy arise from improvements in F(a), which also determines the relationship between health and age.9 Given our formulation improvements in ageing can be achieved in two ways. The first (elongation) is through raising T to T ∗ . Figure 18.8 shows that a change in T elongates the ageing process and reduces frailty at each age. If T ∗ = T + 20 then F(a, T) = F(a + 20, T ∗ ) and from a frailty point of view “70 is the new 50” (or in our case “a + 20 is the new a”). The other way of changing frailty is through changes in φ (compression), which influences the rapidity of ageing. In our model, we have assumed a compensating effect (Strehler and Mildvan, 1960) such that reductions in frailty before age T are matched by sharper increases in frailty at ages near to T. In other words, there is a compression of morbidity (Fries, 1980). In the limit, as (φ) gets larger, frailty falls to zero before rising sharply to F ∗ at T. 338

Working Life—Labor Supply, Ageing, and Longevity

Figure 18.8 Modeling improvements in frailty by elongation and compression. Note: Horizontal axis shows frailty index running from 0 to 1.

While both elongation and compression reflect improvements in health and life expectancy they represent different forms of ageing. Under elongation, life expectancy increases, all years benefit from improvements, and the probability of living beyond T increases. Under compression, life expectancy gains focus increasingly on years close to T and the probability of reaching age T rises but not of living beyond age T. Based on this frailty model, assuming no wage growth and no consumption growth (the interest rate is assumed equal to the discount rate), then Equation (1) becomes
W ∗ c −β = d ρ + M ∗ F ∗ eφλ(R−T) (4) where W ∗ is the wage and c is the constant age-independent level of consumption. Under the assumption that health is homogenous of degree zero in health and life expectancy, Bloom et al. (2014) show that retirement age increases with life expectancy but less than one for one. The homogeneity assumption implies that health at age a is pinned down by age as a proportion of life expectancy, e.g., if life expectancy rises to 100 from 80 then health at age 75 is the health previously associated with 60-year olds (75/100 = 60/80). Under elongation, health and life expectancy are not homogenous of degree zero, but as Figure 18.9a shows (based on full solutions to this dynamic model) this result still broadly holds. As life expectancy increases so too does retirement age with a gradient very close to 1. Under elongation, if life expectancy increases by 20 years, then health shifts by 20 years. The result is that retirement age increases by slightly less than 20 years. As per our previous discussion, the higher wages are the earlier retirement occurs, but in all cases, the level of wages (here with no wage growth) does not substantively affect the slope between retirement age and life expectancy. However, under compression (Figure 18.9b), the link between optimal retirement age and life expectancy is different. In this case, in the limit retirement age tends toward life expectancy. Further, at lower levels of life expectancy, increases in retirement age are less than one for one but as life expectancy increases retirement age rises much faster than one for one. The result is that under compression retirement age, as a proportion of life expectancy, initially declines but then increases, and in the limit working life extends to cover all adult life. The reason for this difference is the role of health in determining the disutility of work. Under compression, frailty decreases at older ages as life expectancy increases, and so the disutility of work falls sharply. Effectively healthy life expectancy is growing faster than life expectancy and retirement age rises sharply. Under elongation, frailty still accumulates but does so more 339

Andrew J. Scott

Figure 18.9 Optimal retirement age and life expectancy.

slowly, and healthy life expectancy and life expectancy are increasing broadly in step and so retirement rises in step too. As this discussion reveals, simple formulae linking retirement age to life expectancy are unlikely to be optimal—much depends on how future health and life expectancy gains accrue. The notion that a third of adult life should be spent in retirement is certainly unsupported. If healthy life expectancy and life expectancy maintain a broadly constant ratio then indexing retirement to life expectancy is sensible. If, however, a compression of morbidity is achieved then retirement should also be compressed. This analysis also shows that a larger economic longevity dividend will be generated through improving ageing via compression rather than elongation.

18.4

Beyond Retirement

A focus on retirement reflects the historical importance of a three-stage life for issues of labor supply at older ages. However, in the face of longer lives, a three-stage structure may no longer be optimal. As the previous section showed, in the case of a full compression of frailty, retirement could potentially disappear. Even in the case of elongation, working careers lengthen almost one for one with increases in life expectancy, pointing to lengthy and extended careers. Technological obsolescence, a desire for change, and the need to reskill all point toward these longer careers becoming “multi-staged” (Gratton and Scott, 2016) and accompanied by late life transitions (Johnson et al., 2009). Longer lives and more varied career paths with shifting employment status suggest a greater focus on labor supply issues at older ages before retirement is needed. The need to focus on pre-retirement also arises from the stylized facts outlined in the section on cross-country and historical trends. Withdrawal from the labor market starts around age 50 as does the rising importance of self-employment and flexible working. Realizing an economic longevity dividend will require boosting labor supply from the age of 50 onward rather than just focusing on extending retirement age.

18.4.1

Health

As the difference between elongation and compression showed, issues around working at older ages depend heavily on health. Not only do people age differently (Hosseini et al., 2021), but there are systematic socioeconomic influences on health and longevity (Chetty et al., 2016; Marmot and Allen, 2014). Growing inequality in health and life expectancy (Case and Deaton, 340

Working Life—Labor Supply, Ageing, and Longevity

2020) will therefore lead to increased dispersion in the optimal retirement age. Reflecting this, many governments have switched from a single state pension age to offering a range of ages over which a pension can be claimed with financial incentives for claiming later. Such policies recognize that retirement as a mass phenomenon is ever less useful for understanding labor supply at older ages. They also point to the importance of public health policies aimed at older individuals as a key plank for achieving an economic longevity dividend. While the focus so far has been on the health of the individual, another challenge for an ageing society is care costs. Kydland and Pretnar (2019) document that the need to care for older relatives is an important contributor to withdrawal from the labor market after the age of 50. We can introduce care into our analysis by modifying (1) in the following manner: W (a)xMUc (a) = MDL (a) + κS(a + X)F(a + X)

(5)

The term κ reflects the costs of caring, X is the age gap between generations,10 S(a + X) is the probability of someone aged a + X still being alive, and F(a + X) is frailty at that age. While κ may reflect the monetary costs of paying for care, it can also be interpreted more positively. Caring for elderly relatives may also be a source of lifetime satisfaction and thus also add to the disutility of working. A rising probability of being alive in years of poor health will create a larger care burden and reduce labor supply of those who must provide care. Life expectancy gains driven by compression, however, will increase S(a + X) but lower F(a + X) and so could reduce care costs. Elongation has the same but weaker effects for ages < T but increases S(a+X) for a > T and so does less to alleviate the care burden.11 This adds an additional indirect channel through which healthy ageing via compression is important for boosting labor supply of older workers.

18.4.2

Wages

As with all ages, the labor supply of older workers depends on the wage offered. French (2005) estimates that older workers actually have a higher elasticity of labor supply with respect to wages. In our optimal retirement calculations, we assume no increases in wages across a lifetime for analytical simplicity, but evidence suggests that wages fall at retirement and potentially with age, which would also lower labor supply (although see Blundell et al., 2016, for a summary suggesting that evidence for this decline is not firmly established). One reason wages may decline with age is that the productivity of older workers declines. This could relate to declining physical health or cognitive abilities or reduced innovation, all of which could potentially be offset by healthier ageing (and could also explain why more recent cohorts of older workers are more engaged in the labor market).

18.4.3

Hiring and Labor Demand

Even if productivity does not decline with age, firms may believe it does due to ageism (Allen, 2019). In this case, not only will older workers find themselves more likely to be made unemployed (an effect that is sharper in downturns; Dahl and Knepper, 2020) but also less likely to be hired and more likely to withdraw from the labor market due to age discrimination. A study for the advocacy group AARP found that 10 percent of men aged 50 years and older and 7.5 percent of women who had withdrawn from the labor market did so due to age discrimination. Age discrimination also helps explain why older workers are disproportionately represented among the self-employed (Fact 6). Carlsson and Eriksson (2019) and Neumark et al. (2019) found evidence of ageism in hiring decisions. However, discrimination is not uniform among all older workers: both studies 341

Andrew J. Scott

find older women are affected more than men. These findings point to three issues that repeatedly come up around older workers. The first is diversity: not all older workers face the same problems. The second is the need for policies specifically aimed at older workers, in this case tackling age discrimination. However, the third issue also shows that many of the issues around older workers are not unrelated to those found at all ages, in this case the prevalence of gender discrimination.

18.4.4

Education

Another solution to declining productivity at older ages is education. Hazan (2009) and Cervellati and Sunde (2013) emphasized that longer working lives require more years of education. The earlier analysis of optimal retirement age was based on length of career. If longer careers also require more education, then careers will start later or be interrupted more frequently. The result is that retirement age will likely rise more than one for one with life expectancy once allowance is made for additional years of education and even greater dispersion in retirement ages based on differences in educational achievement. While more years of education are required for longer careers, the issue is when education should occur. If skills and knowledge decline over time (either through decay or through obsolescence), then the obvious implication is to add some of the additional years of education into later life. This is even more important if the rate of technical obsolescence increases. Initiatives such as a “60-year curriculum” (Dede and Richards, 2020) and “long-life learning” (Conley and Rauth, 2020) are already beginning to emerge to support this.

18.4.5

The Nature of Work

Also important in determining labor supply is d∗—the parameter in (1) that captures the nonhealth aspects of the disutility of work. This parameter captures a broad range of features connected to work. For instance, Acemoglu and Restrepo (2021) provide evidence that more rapidly ageing countries are investing in robotics to support older workers in production for longer. Using robotics will help maintain productivity and thus wages, but to the extent their use also reduces the physical demands of a job it may also lower d∗ and so boost labor supply. Similarly, B¨orsch-Supan and Weiss (2016) described several changes BMW made in its Bavarian plant to make work more productive and less physically challenging for older workers. As older workers increase as a proportion of the workforce, more and more firms are likely to redesign work to accommodate their needs and preferences. An important aspect of such “age-friendly” policies appears to be the offer of flexible work. Growing evidence points to older workers being prepared to accept lower wages in return for greater flexibility (Maestas et al., 2018; Ameriks et al., 2020). Cook et al. (2019) also present evidence that older workers trade off wages against job attributes. They find that hourly earnings are lower for older Uber drivers because they drive at less busy times and in less congested areas. The nature of jobs also has an influence. For instance, Shavit and Carstensen (2019) find older workers prefer jobs that involve helping others. Johnson et al. (2009) document that career changes in the United States at later ages are common either through voluntary actions or through layoffs. They find that these career changes typically involve moving into lower-paying jobs with fewer benefits but also more flexible work arrangements, less stress, and fewer responsibilities. Coronado (2016) notes that many older workers feel forced to retire early because of a lack of firm flexibility in the positions on offer. 342

Working Life—Labor Supply, Ageing, and Longevity

This also helps explain why older workers are disproportionately part of the contingent workforce if firms are reluctant to offer flexible working hours (Hurd, 1996). Linked with this transition is the notion of “bridge” jobs (Alcover, 2016), which create a transition between traditional full-time career jobs and a shift into retirement. Together these forces point toward two research shifts. The first is a shift away from focusing on retirement and instead focusing more broadly on the intensive margin, e.g., how many hours to work at older ages. Gotbaum and Wolfe (2018) notes that “most people hope their retirement will be like a warm bath: you work your way in slowly and gradually,” while traditional retirement practice is “more like a cold shower.” In response to longer lives and the need to work for longer, retirement seems to be taking on the quality of the limit point in a continuous process rather than that of an abrupt change. The second consequence is that whereas in the 20th century much of the increase in life expectancy translated into leisure post-retirement, in response to future life expectancy gains a greater increase in leisure pre-retirement is likely. Understanding the preference for leisure at older ages is a key determinant of labor supply in these years.

18.5

Toward a Theory of Ageing

The previous section outlined several issues regarding the preferences of older workers and how these might influence labor supply. Viewed from the perspective of an ageing society, with its focus on changes in the population’s age structure, understanding these issues requires shining a research spotlight on a group whose needs and desires have previously received relatively little attention.

18.5.1

Different Concepts of Ageing

In contrast, the longevity agenda focuses on changes in how individuals are ageing due to longer lives and changing health dynamics. From this perspective understanding the behavior of older workers requires a deeper understanding of ageing. One such approach is to draw a distinction between chronological and biological age, where the latter reflects underlying health. Several studies point to improvements in the relationship between chronological and biological age (Cutler et al., 2014; Levine and Crimmins, 2018; Abeliansky et al., 2020). Shoven and Goda (2011) emphasize that chronological age is a nominal measure and biological age a real measure and so adjustment needs to be made for “age inflation.” Using this approach considering the labor supply of older workers as a separate issue is unnecessary. Once adjustment is made for age inflation older workers are no longer “old” (60 is the new 50, etc.) and can be subsumed into existing theories and policies of labor supply. However, the concept of ageing that implicitly defines older workers may be broader than just biological age. Consider once more the previous example where a full compression of morbidity is achieved, and retirement age increases to match life expectancy. A 200-year life expectancy would suggest a working career lasting from age 20 to 200. In this case, a compression of morbidity makes biological age constant regardless of chronological age. This means health does not change, and so economic decision-making is stationary and the same decisions regarding work and leisure are made every period. But key to the notion of ageing is a sense of nonstationarity, of decisions changing with the passage of time. Take the extreme case of immortality12 whereby an individual becomes infinitely lived. Would such an individual really behave like the textbook case of an infinitely lived consumer? Or would boredom or ennui become a challenge? 343

Andrew J. Scott

18.5.2

The Role of Time in Ageing

The standard in many economic models is to assume that individuals prefer variety in terms of the goods they consume. Over a long life would they have a similar preference for variety in terms of behavior over time? Heaton (1993) proposes a model of preferences where commodities can be complements or substitutes depending upon the passage of time. Would something similar be needed to properly understand the preferences of long-lived individuals around leisure and work? Effectively these questions point toward the need for a proper theory of ageing. A focus on biological age leads to conclusions that “60 is the new 50,” but this is true only if health is the only determinant of how we age. If the passage of time also affects our preferences, then 60 is nothing other than the new 60. Another view of ageing (Sanderson and Scherbov, 2019) is to focus on “prospective” or thanatological age—that is, how many years of life remain in expectation. This view of ageing has two important implications. The first is that it could provide a theory of retirement based on a “final” chapter of life, consistent with historical notions of retirement. In this case, even under full compression, retirement might occur before the expected end of life. Prospective age also helps explain why “60 is the new 60.” Increases in life expectancy mean a 60-year old today has more life ahead of them than past cohorts of 60-years olds. That should make them more interested in investing in their health, in their human capital, and in prolonging their working career. However, they still have 60 years of past experience and several years of declining human capital. Time is Janus-like looking both backward and forward, and both aspects seem important for properly understanding changing labor supply at older ages. Carstensen’s (2006) Socioemotional Selectivity Theory argues that as time horizons shorten individuals shift their activities and resources into emotionally more meaningful activities. This emphasis “on what matters” would also explain the role of older individuals in nonmarket activities aimed at caring and social contribution. Once again through increasing d∗, this would lead to a decline in labor supply in the formal economy and a shift into other areas. Such a theory of ageing would imply preference parameters that vary with age. For instance, evidence indicates that risk aversion increases with age as do preferences for flexible work. Better understanding of how deep primitives that define motivation change over time is needed to better understand labor supply decisions of older workers.

18.5.3

Are Older Workers Different?

If social security arrangements and health play a diminishing role in determining labor supply at older ages, then what, if anything, takes their place? This question creates something of a dilemma in ageing research. Just as in the famous physics conundrum regarding whether Schrodinger’s cat is dead or alive, so the issue is whether ageing is a real issue or not for older workers. The more that health pins down age, then the more health improves the less age matters and the more similar older individuals become to younger people. However, the more that ageing matters in shaping our preferences as life lengthens, the more economists need to understand the impact of ageing on the labor supply of older workers. In terms of policies, this implies a twofold response. To the extent to which living longer and healthier lives is about “60 becoming the new 50,” then many issues around older workers can be swept into current labor market institutions, policies, and analysis. However, to the extent to which preferences and needs evolve over time then new forms of corporate and government policies will be needed to boost the labor supply of older workers and achieve a 344

Working Life—Labor Supply, Ageing, and Longevity

longevity dividend. Further complicating the situation is the diversity in how people age, requiring a mixture of both approaches. The tendency is to assume that “older workers” represent a homogenous group, but as their numbers swell perhaps the greatest insight will be recognizing the within-group diversity of issues that impact labor supply.

18.6

Conclusions

Traditionally, the labor supply choices of older workers have been seen as determined by the tax and benefit system and declining health. With social security reforms making the former more neutral and evidence that recent cohorts of older workers are healthier, focus needs to shift to other issues to better understand labor supply at older ages. In the three-stage life that characterizes traditional life-cycle models, retirement age is a key labor supply issue. However, as working careers lengthen, the intensive margin of work becomes more important as leisure is distributed more equally across later years of life and retirement becomes more of a transition than an event. One possibility is that as people age better “60 becomes the new 50” and the analysis of older workers can be subsumed within more standard labor supply analysis. Another possibility is that understanding labor supply at older ages requires a deeper understanding of preferences of this age group, which in turn requires a theory of ageing. A nuanced theory of ageing would also provide insight into the diversity that characterizes older workers. As an ageing society leads to more older people and the malleability of age leads to greater diversity, the need will be to understand different preferences and constraints that affect older people in different ways. A better understanding of ageing would also help sharpen insights into whether older age groups differ in their preferences around flexible work or whether this is a response to constraints arising from corporate policies and ageism. Separating labor demand and labor supply issues for older workers is especially challenging, and a sharper model of older-age preferences would help to disentangle them. A combination of increased longevity and a rising older population is leading to more meaningful economic decisions occurring at later ages. That requires a greater research focus on better understanding what exactly, if anything, makes this age group different both from other age groups and from past cohorts. Age would appear to be anything but a number.

Notes 1 Funding from ESRC Research Grant T002204 is gratefully acknowledged as well as excellent research assistance from Jonathan Old. 2 See Thane (2006) for a broad historical perspective and Graebner (1988) and Costa (1998) and Macnicol (1998) for detailed U.S. and UK histories. 3 See Blechman (2009) for a history of Youngtown and the establishment of retirement communities for the over 50s. 4 The focus is on cross-country averages. Given the diversity that characterizes ageing, both within populations and across countries, these average patterns inevitably abstract from many key issues attached to ageing and longevity. 5 Source: https://data.ncvo.org.uk/volunteering/demographics/. 6 See Blundell et al. (2016) for a survey and analysis. 7 A theme of this chapter is the need to develop a theory of ageing based around an understanding of how individual preferences may change over time. In using the model of Bloom et al. (2014), we focus on changes brought about by changes in health and that affect the within-period preferences of consumption and leisure. Richer models could introduce intertemporal shifts in preferences (see Dohmen et al., 2023, for evidence that preferences around risk and patience shift with age) or

345

Andrew J. Scott

8

9 10

11

12

shifts in constraints (such as wages or employment opportunities due to ageism—see the subsequent discussion). The notion of a frailty index common to both health and mortality is consistent with biological models (see Gavrilov and Gavrilova, 1991; Strulik and Vollmer, 2013) and empirical data (Mitniski et al., 2002; Abeliansky and Strulik, 2018). By assuming mortality and health both depend on the same frailty index we assume that “frailty,” “age,” and “proximity to death” are all synonyms. We focus here on caring for older generations rather than children. The model could also be extended to the case where healthier grandparents are able to take on a greater caring role for children and so provide a boost to labor supply for families with children. Elongation will also increase S(a + 2X) and raise the prospect of four-generation households and an even larger caring burden. If steps are not taken to ensure that caring is spread proportionately across genders, then this rising burden could lead to major discrepancies between male and female labor supply at older ages. A key feature of our frailty and mortality functions is a positive derivative with respect to age. Biological research into ageing seeks to reduce the magnitude of that derivative. Immortality requires not just that mortality rates do not rise with age but that they are zero. For instance, the naked mole rat is an animal of great interest for longevity researchers as its mortality rate is independent of age but not zero (“Calvo ageing”). As a result, it is not immortal and has a laboratory lifespan of up to 30 years. Therefore, this example of immortality is intended to be more philosophical than practical.

References ABELIANSKY, A. L., EREL, D., AND STRULIK, H. (2020): “Ageing in the USA: Similarities and disparities across time and space,” Nature Scientific Reports, 10: 14309. ABELIANSKY, A., AND STRULIK, H. (2018): “How we fall apart: Similarities of human ageing in 10 European countries,” Demography, 55: 341–359. ABRAHAM, K. G., HERSHBEIN, B., AND HOUSEMAN, S. (2020): “Contract work at older ages,” NBER Working Paper No. 26612. National Bureau of Economic Research, Cambridge, MA. ACEMOGLU, D., AND RESTREPO, P. (2021): “Demographics and automation,” Review of Economic Studies, 1–44, https://doi.org/10.1093/restud/rdab031. AKSOY, Y., BASSON, H., SMITH, R., AND GRASI, T. (2019): “Demographic structure and macroeconomic trends,” American Economic Journal: Macroeconomics, 11(1): 193–222. ALCOVER, C. M. (2017): “Bridge employment: Transitions from career employment to retirement and beyond.” In: Parry, E., and McCarthy, J. (eds.), The Palgrave Handbook of Age Diversity and Work, London: Palgrave Macmillan, pp. 225–262. ALLEN, S. G. (2019): “Demand for older workers: What do economists think? What are firms doing?,” NBER Discussion Paper 26597. National Bureau of Economic Research, Cambridge, MA. AMERIKS, J., BRIGGS, J., CAPLIN, A., LEE, M., SHAPIRO, M. D., AND TONETTI, C. (2020): “Older Americans would work longer if jobs were flexible,” American Economic Journal: Macroeconomics, 12(1): 174–209. APPLEWHITE, A. (2016): This Chair Rocks: A Manifesto Against Ageism, New York: MacMillan Books. BLECHMAN, A. D. (2009): Leisureville: Adventures in a World Without Children, New York: Grove Press. BLOOM, D. E., CANNING, D., AND MOORE, M. (2014): “Optimal retirement with increasing longevity,” Scandinavian Journal of Economics, 116(3): 838–858. BLOOM, D. E., KHOURY, A., ALGUR, E., AND SEVILLA, J-P. (2020): “Valuing productive non-market activities of older adults in Europe and the US,” de Economist, 168(2): 153–181. BLUNDELL, R., FRENCH, E., AND TETLOW, G. (2016): “Retirement incentives and labor supply.” In: Piggot, J., and Woodland, A. (eds.), Handbook of Economics of Population Ageing, Volume 1B. Elsevier, North Holland, pp. 457–566. ¨ BORSCH -SUPAN, A., AND WEISS, M. (2016): “Productivity and age: Evidence from work teams at the assembly line,” Journal of the Economics of Ageing, 7: 30–42. CARLSSON, M., AND ERIKSSON, S. (2019): “Age discrimination in hiring decisions: Evidence from a field experiment in the labor market,” Labour Economics, 59: 173–183. CARSTENSEN, L. (2006): “The influence of a sense of time on human development,” Science, 312(5782): 1913–1915.

346

Working Life—Labor Supply, Ageing, and Longevity

CASE, A., AND DEATON, A. (2020): Deaths of Despair and the Future of Capitalism, Princeton: Princeton University Press. CERVELLATI, M., AND SUNDE, U. (2013): “Life expectancy, schooling and lifetime labor supply theory and evidence revisited,” IZA DP N0.7286. CHETTY, R., STEPNER, M., ABRAHAM, S., LIN, S., SCUDERI, B., TURNER, N., BERGERON, A., AND C UTLER , D. (2016): “The association between income and life expectancy in the United States, 2001–2014,” Journal of the American Medical Association, 315(16): 1750–1766. CONLEY, C. (2018): Wisdom@Work: The Making of a Modern Elder, London: Currency Portfolio Penguin. CONLEY, C., AND RAUTH, I. (2020): “The emergence of long life learning.” Modern Elder Academy White Paper. COOK, C., DIAMOND, R., AND OYER, P. (2019): “Older workers and the gig economy,” American Economic Review, Papers and Proceedings, 109: 372–376. COOLEY, T., AND HENRIKSEN, E. (2018): “The demographic deficit,” Journal of Monetary Economics, 93: 45–62. CORONADO, J. (2016): “The changing nature of retirement.” In: Mitchell, O., and Shea, R. C. (eds.), Reimagining Pensions: The Next Years, Oxford: Oxford University Press. COSTA, D. (1998): The Evolution of Retirement: An American Economic History, Chicago: University of Chicago Press. CUTLER, D. M., MEARA, E., POWELL, W. F., AND RICHARDS-SHUBIK, S. (2014): “Health and work capacity of older adults: Estimates and implications for social security policy.” Available at https://papers.ssrn.com/sol3/papers.cfm?abstract id=2577858. DAHL, G. B., AND KNEPPER, M. (2020): “Age discrimination across the business cycle.” NBER Working Paper 27581, National Bureau of Economic Research, Cambridge, MA. DEDE, C., AND RICHARDS, J. (2020): The 60 Year Curriculum: New Models for Lifelong Learning in the Digital Economy, Abingdon on Thames, UK: Routledge. DOHMEN, T., HUFFMAN, D., AND SUNDE, U. (2021): “Ageing and economic preferences.” In: Bloom, D. E., Sousa-Poza, A., and Sunde, U. (eds.), The Routledge Handbook of the Economic of Ageing, Oxon: Routledge. EGGLESTON, K., AND FUCHS, V. (2012): “The new demographic transition: Most gains in life expectancy now realised late in life,” Journal of Economic Perspectives, 26(3): 137–156. FRENCH, E. (2005): “The effects of health, wealth, and wages on labor supply and retirement behaviour,” Review of Economic Studies, 72(2): 395–427. FRIEDMAN, M. (2018): How to Live Forever: The Enduring Power of Connecting the Generations, New York: Public Affairs. FRIES, J. (1980): “Ageing, natural death, and the compression of morbidity,” New England Journal of Medicine, 303(3): 130–135. GAVRILOV, L. A., AND GAVRILOVA, N. S. (1991): The Biology of Human Lifespan: A Quantitative Approach, Reading, UK: Harwood Academic. GOTBAUM, J., AND WOLFE, B. (2018): “Help people work longer by phasing retirement.” Available at https://www.brookings.edu/opinions/help-people-work-longer-by-phasing-retirement/. GRAEBNER, W. (1980): A History of Retirement: The Meaning and Function of an American Institution 1885– 1978, New Haven, CT: Yale University Press. GRATTON, L., AND SCOTT, A. (2016): The 100 Year Life—Living and Working in An Age of Longevity, London: Bloomsbury Press. HAZAN, M. (2009): “Longevity and lifetime labor supply: Evidence and implications,” Econometrica, 77(6): 1829–1863. HEATON, J. (1993): “The interaction between time-nonseparable preferences and time aggregation,” Econometrica, 61(2): 353–385. HOSSEINI, R., KOPECKY, K., AND ZHAO, K. (2021): ”The evolution of health over the life cycle,” Review of Economic Dynamics, 45: 237–263. HURD, M. (1996): “The effect of labor market rigidities on the labor force behaviour of older workers.” In: Wise, D. (ed.), Advanced in the Economics of Ageing, Chicago: University of Chicago Press. JOHNSON, R. W., KAWACHI, J., AND LEWIS, E. K. (2009): Older Workers on the Move: Recareering in Later Life, Washington, DC: AARP Public Policy Institute. KYDLAND, F., AND PRETNAR, N. (2019): “Who will care for all the old people?.” In: Bloom, D. E. (ed.), Live Long and Prosper? The Economics of Ageing Populations, VoxEU.org Ebook.

347

Andrew J. Scott

LAUN, T., AND WALLENIUS, J. (2016): “Social insurance and retirement: A cross country perspective,” Review of Economic Dynamics, 22: 72–92. LEE, R. (2010): “Population ageing and the economic role of the elderly: Bonanza or burden? commentary on Coall and Hertwig, ‘Grandparental investment: Past, present, and future,’” Behavioral and Brain Sciences, 33(1): 31–32. LEVINE, M. E., AND CRIMMINS, E. M. (2018): “Is 60 the new 50?: Examining changes in biological age over the past two decades,” Demography, 55(2): 387–402. MACNICOL, J. (1998): The Politics of Retirement in Britain 1878–1948, Cambridge: Cambridge University Press. MAESTAS, N. (2010): “Expectations and realizations of work after retirement,” Journal of Human Resources, 45(3): 718–748. MAESTAS, N., MULLEN, K., POWELL, D., WACHTER, T. V., AND WENGER, J. B. (2018): “The value of working conditions in the United States and implications for the structure of wages,” NBER Working Paper 25204, National Bureau of Economic Research, Cambridge, MA. MAESTAS, N., AND ZISSIMOPOULOS, J. (2010): “How longer work lives ease the crunch of population ageing,” Journal of Economic Perspectives, 24(1): 139–160. MARMOT, M., AND ALLEN, J. J. (2014): “Social determinants of health equity,” American Journal of Public Health, 104(4): S517–S519. MITNITSKI, A., MOGILNER, A. J., MACKNIGHT, C., AND ROCKWOOD, K. (2002): “The accumulation of deficits with age and possible invariants of ageing,” Scientific World, 2: 1816–1822. MODIGLIANI, F., AND BRUMBERG, R. H. (1954): “Utility analysis and the consumption function: An interpretation of cross-section data.” In: Kurihara, K. K. (ed.), Post- Keynesian Economics, New Brunswick, NJ: Rutgers University Press, pp. 388–436. NEUMARK, D., BURN, I., AND BUTTON, P. (2019): “Is it harder for older workers to find jobs? New and improved evidence from a field experiment,” Journal of Political Economy, 127(2): 922–970. OECD. (2019): OECD pensions at a glance. Available at https://www.oecd.org/publications/oecdpensions-at-a-glance-19991363.htm. OEPPEN, J., AND VAUPEL, J. W. (2002): “Broken limits to life expectancy,” Science, 296(5570): 1029–1031. OLSHANSKY, S. J., PERRY, D., MILLER, R. A., AND BUTLER, R. N. (2006): “In: pursuit of the longevity dividend: What should we be doing to prepare for the unprecedented ageing of humanity?,” The Scientist, 20(3): 28–33. ROGERSON, R., AND WALLENIUS, J. (2021): “Shocks, institutions, and secular changes in employment of older individuals,” NBER Macroeconomics Annual, 177–216. SANDERSON, W., AND SCHERBOV, S. (2019): Prospective Longevity—A New Vision of Population Ageing, Cambridge, MA: Harvard University Press. SCOTT, A. J. (2021): “Achieving a three dimensional longevity dividend,” Nature Ageing, 1: 500–505. SCOTT, A. J., ELLISON, M., AND SINCLAIR, D. A. (2021): “The economic value of targeting ageing,” Nature Ageing, Available at https://doi.org/10.1038/s43587-021-00080-0. SHAVIT, Y. Z., AND CARSTENSEN, L. L. (2019): “Age and occupational time perspective are associated with preference for helping others at work,” Innovation in Ageing, 3: S127. SHOVEN, J., AND GODA, G. S. (2011): “Adjusting government policies for age inflation.” In: Shoven, J., (ed.), Demography and the Economy, Chicago: University of Chicago Press. STREHLER, B. L., AND MILDVAN, A. S. (1960): “General theory of mortality and ageing,” Science, 132(3418): 14–21. STRULIK, H. (2021): “Measuring ageing.” In: Bloom, D. E., Sousa-Poza, A., and Sunde, U. (eds.), The Routledge Handbook of the Economic of Ageing, Oxon: Routledge. STRULIK, H., AND VOLLMER, S. (2013): “Long run trends of human ageing and longevity,” Journal of Population Economics, 26: 1303–1323. THANE, P. (2006): “The history of retirement.” In: Clark, G. L., Munnell, A. H., and Orszag, J.M. (eds.), The Oxford Handbook of Pensions and Retirement Income, Oxford: Oxford University Press, pp. 33–51. U.K GOVERNMENT. (2013): “Autumn statement announcement on a core principle underpinning future state pension age rises.” https://assets.publishing.service.gov.uk/government/uploads/system/uploads/ attachment data/file/263660/spa-background-note-051213 tpf final.pdf. UNITED NATIONS. (2020): “2019 revision of world population prospects.” dataset https://population. un.org/wpp/.

348

19 EDUCATION AND AGEING: HUMAN CAPITAL INVESTMENTS AND AGEING Andries de Grip and Raymond Montizaan

Abstract Increasing life expectancy, falling birth rates, and reforms of pension systems have all contributed to a global ageing of the labor force over the past few decades. At the same time, rapid technological change leads to changing skills demands, implying that workers are increasingly challenged to keep their skills up to date at later ages. This chapter discusses the importance of human capital investments for maintaining older workers’ employability in this dynamic ageing environment. It provides an overview of how production of human capital evolves through the life cycle and which types of human capital investments are available to workers. It discusses the impact of skills obsolescence for human capital investments over the life cycle and examines in more detail the importance of lifelong learning and the returns to human capital investments later in life. We further discuss the mutual relation between human capital investments and retirement behavior. Finally, this chapters formulates a research agenda that stresses the importance of further random control trials to investigate the effectiveness of training policies aimed at older workers. We argue that more research is needed on the specific preferences of older workers with respect to the skills they would like to invest in and on which training modes are most effective in raising the returns to training for older workers. Along these lines we should find answers to the crucial question of how we can enable human capital investments to account for the changing needs and preferences over workers’ lengthening life cycle.

19.1

Introduction

Increasing life expectancy and falling birth rates have contributed to the graying of the labor force of Organisation for Economic Co-operation and Development (OECD) countries over the past few decades. In 1950, there were more than seven working-age people for every one of pension age. By 2047, there will be just two workers per pensioner (D’Addio et al., 2010). Consequently, pension reforms have been introduced in most industrialized countries to increase the retirement age (B¨orsch-Supan and Coile, 2018; Lindeboom and Montizaan, 2020). Mandatory retirement ages have subsequently either increased dramatically or been abolished in many countries, implying that both workers and employers recognize that workers will continue their 349

DOI: 10.4324/9781003150398-22

Andries de Grip and Raymond Montizaan

employment until a higher age. At the same time, rapid technological change leads to changing skills demands on the labor market (Weinberger, 2014; Deming, 2017; Autor and Salomons, 2018; Agrawal et al., 2019). This implies that workers are increasingly challenged to keep their skills up to date to later ages and must invest in new skills that are complementary to new technologies to remain employable. Human capital investments are crucial for maintaining older workers’ employability in this dynamic environment, but the question is how to optimize these investments. To answer this ambitious question, this chapter will give an overview of various studies on the complex relationship between human capital investments and ageing. We will provide this overview by first discussing how the production of human capital evolves throughout the life cycle. In this discussion, the optimal timing of human capital investments within the life cycle will be examined. In addition to this literature, we discuss how different human capital investments—general versus vocational—at an early age can affect human capital investments in later stages of the life cycle and discuss the literature on more informal modes of learning through the life cycle such as learning by doing and learning from peers in the workplace. Second, we will discuss various causes of skills obsolescence that affect the development of workers’ human capital through their life cycle and the need for further later-life human capital investments to remain employable. Third, we will examine in more detail the importance of lifelong learning for ageing workers, whether the returns to human capital investments in later stages of the life cycle outweigh the costs, and the barriers to continuing to invest in human capital. Finally, we discuss the importance of human capital investments for retirement behavior. In the concluding section, we present an outline of a future research agenda on human capital investments in an ageing workforce.

19.2

Production of Human Capital through the Life Cycle

In the initial human-capital model (Becker, 1962), it is already obvious that human capital investments at a young age are most beneficial as these investments have a longer period over which to increase a worker’s earnings. Ben-Porath (1967) relates the life cycle of earnings to the life cycle of investments in human capital. In his seminal paper, he develops a production function of human capital that underlies the optimal path of accumulation of human capital through the life cycle to maximize the present value of lifetime earnings. The costs and benefits of human capital investments determine this optimal path of human capital development. In his model of the life cycle of earnings, Ben-Porath distinguishes three phases in developing the fraction of human capital allocated to human capital production: In phase 1, the available stock of human capital is entirely allocated to the further production of human capital. This reflects the period of full-time initial education during childhood and adolescence. In phase 2, the available stock of human capital is allocated to the production of human capital and to income-generating productive services. The gross additions to the stock of human capital are positive until t = T, the date of retirement, i.e., phase 3, in which no production is undertaken, and one’s human capital loses its value. In phase 2, the production of human capital declines as t rises due to (1) the shorter period during which human capital investments at a later age can generate returns by using the skills acquired in the labor market and (2) the higher foregone earnings because of the higher stock of human capital. The latter, however, only holds if the greater productivity of time allocated to productive services is not fully matched by a greater productivity in human capital production, but this is usually the case as older workers generally face a decreasing capacity to learn (Rosen’s, 1975). 350

Education and Ageing: Human Capital Investments and Ageing

Apart from the flow of human capital produced (Q), Ben-Porath’s definition of human capital ˙ also includes a rate of depreciation (δ): development (K) K˙ t = Qt − δKt

(1)

Due to the depreciation of the stock of an individual’s human capital, the net development of the stock of human capital is likely to be negative in the years before retirement as this depreciation exceeds the new investments in human capital that are hampered because they can generate returns only for a short time. In his paper, Ben-Porath assumes that δ is a fixed rate but notices that it could depend on the allocation of human capital to different activities. In the following, we will further discuss the impact skills obsolescence could have on the human capital of older workers. In their paper on the technology of skill formation, Cunha and Heckman’s (2007) even more strongly emphasize the importance of human capital investments in early life. They show that investing in human capital in early life is particularly beneficial as both the self-productivity of skills and the dynamic complementarity boost the returns of human capital investments at a young age. The self-productivity of skills refers to the finding that skills produced at one stage augment the skills attained at later stages in life, whereas the dynamic complementarity of human capital investments refers to the finding that skills produced at one stage raise the returns on new human capital investments at subsequent stages in life. Heckman (2007) extends this model by broadening the framework of human capital investments that includes the dynamic complementarity among investments in cognitive skills, noncognitive skills, and health. Note that although Cunha and Heckman’s (2007) model strongly emphasizes the importance of human capital investments in early life, it also shows that later investments in human capital are needed to harvest the dynamic complementarity between early and later investments. The optimal ratio of early and later human capital investments, therefore, depends on the degree of complementarity between early and late investments, although a high interest rate and a high rate of skill depreciation will also lead to later investments.

19.2.1

Lifetime Effects of General versus Vocational Initial Education

The returns on early investments in human capital and participation in further education later in life also depend on the kind of skills youngsters acquire. Various studies focus on the different lifetime effects of vocational versus general initial education and show that life-cycle outcomes of youngsters who received vocational education differ from those who participated in general education, with respect to employment, wages, and participation in further education later in life (e.g., Hanushek et al., 2017; Golsteyn and Stenberg, 2017). These studies show that the initial gains of vocational education for the transition of school to work (as shown by e.g., Ryan, 2001) are offset by a lower adaptability and diminished employment later in life. This pattern appears to be most pronounced in the countries where apprenticeships dominate vocational initial education: Denmark, Germany, and Switzerland (Hanushek et al., 2017). Gould et al. (2001) show that technological progress leads to a higher depreciation of technology-specific vocational skills versus general skills. Moreover, those with general education more often invest in further training, which suggests that general education makes subsequent educational investments cheaper. Hanushek et al. (2017) show that in the aforementioned apprenticeship countries, those with a general education are indeed more likely to participate in training when they become older. However, they do not find a similar pattern in other “vocational countries.” Golsteyn and Stenberg (2017) find that particularly males with general initial education are more likely to participate in further education. Yet, Brunello and Rocco (2017) show that this primarily holds for the first part of their working careers. 351

Andries de Grip and Raymond Montizaan

19.2.2

Learning by Doing

Ben-Porath’s (1967) model of the life cycle of investments in human capital assumes that individuals either allocate their time and current stock of human capital to further investments in their human capital or to activities in the labor market. This neglects the possibility of joint production, which is quite common in the labor market as not only investments in further training but also learning by doing affect the human capital development of workers. In early human capital theory, Mincer (1962) develops the “earnings function,” which relates workers’ wages to their educational background (S) and their work experience (X) as a measure of the skills acquired by learning by doing. In this earnings function, Mincer not only includes a worker’s years of experience, but also the square of this number of years to account for the diminishing return on learning by doing in a worker’s later career. Mincer’s estimation results show that an additional year of experience gives the worker on average a return of 8 percent. However, the negative coefficient of the square term shows that the rate of return levels off as years of experience increase (see Table 19.1). This shows that, at some point in their career, workers are virtually no longer learning anything during their work. However, in their study on a large car manufacturing plant, B¨orsch-Supan and Weiss (2016) show that the average age-productivity profile of individual workers increases until the age of retirement at 65 years. When decomposing workers’ productivity profiles into the effects of job tenure and age, they find that experience keeps older workers’ productivity from falling. In a further decomposition of this productivity increase, they find that the older workers’ competence is their ability to avoid severe errors. While tenured workers occasionally make errors, they rarely make severe errors. This suggests that older workers are especially good at vital tasks.

19.2.3

Learning from Peers in the Workplace

The feedback that workers get from their more senior colleagues could also affect informal learning in the workplace. In their meta-analysis of 46 estimates from 34 field studies and laboratory experiments, Herbst and Mas (2015) find a substantial positive spillover effect of worker productivity on the productivity of co-workers of about 12 percent. This might be due to either increasing the efforts of initially less productive workers or to human capital spillovers that increase co-workers’ productivity. The findings of Cornelissen et al. (2017) suggest that the pressure for higher efforts operates predominantly in low-skilled occupations, while human capital spillovers are most pronounced in high-skilled occupations. De Grip et al. (2016) show that placing new workers in more experienced teams reduces the time until new hires become equally productive as an experienced worker by an average 36 percent, compared with being placed in less experienced teams. As the productivity of new hires is much lower than the productivity of tenured workers, clearly higher efforts cannot close the productivity gap. This suggests that older workers might also contribute to the production of the human capital of the less experienced workers in their team via knowledge spillovers in the workplace. Table 19.1 Mincer’s earnings function

Source: Mincer (1974), Table 5.1. 352

Education and Ageing: Human Capital Investments and Ageing

19.3

Skills Obsolescence

In Ben-Porath (1967), the given rate of skill depreciation is not further discussed. Rosen’s (1975) is the first paper to highlight that due to skills obsolescence, the net human capital development of older workers will be substantially lower or even negative. Older workers may suffer from both technical and economic skills obsolescence. Technical obsolescence refers to the wear of skills in the course of time, whereas economic obsolescence negatively affects the value of a worker’s skills in the labor market due to shifts in the skills demanded. Building on Rosen’s work, De Grip and Van Loo (2002) develop a more extensive typology of skills obsolescence (see Table 19.2). This typology distinguishes between two causes of technical skills obsolescence (wear and atrophy) and three causes of economic skills obsolescence due to job-, firm-, and sector-specific shifts in the demand for skills. Pfeiffer and Reuß (2008) include age-dependent skill depreciation in Cunha and Heckman’s (2007) model on the technology of skill formation. They assume that depreciation of skills is modest in childhood and accelerates with increasing age. This reflects Rosen’s (1975) notion that older workers suffer from a general deterioration of their human capital (“wear”) and a decreasing capacity to learn and adapt to new situations. Strictly speaking, the latter does not refer to skills obsolescence as such but to a lower return on new investments in human capital.

Table 19.2 Causes of skills obsolescence

Source: Adjusted from De Grip and Van Loo, (2002). 353

Andries de Grip and Raymond Montizaan

Rosen’s (1975, pp. 199–200) considers skills obsolescence due to technological change as a vintage effect: “Production innovations often render useless skills associated with prior methods.” Neuman and Weiss (1995) show that high-educated workers are generally more affected by skills obsolescence than low-educated workers and face a particularly high depreciation of their skills when they are employed in high-tech sectors, as firms in these sectors frequently innovate their production process. The rate of depreciation of human capital also differs by field of education. Building on an analysis of the relation between more recent versus older citations in the literature, McDowell (1982) shows that fields of expertise such as physics and chemistry suffer more from skills obsolescence than expertise in humanities. However, a high rate of depreciation of human capital may also drive regular investments in further education to keep one’s skills up to date. Bartel and Sicherman (1993) show that in this respect particularly the pattern of technological change matters. On the one hand, an unexpected technological shock increases the depreciation rate of workers’ human capital, which induces older workers to retire earlier. On the other hand, more gradual technological change will lead to flatter investment profiles in human capital as workers attempt to meet the continuously changing skill demands in their industry or occupational field. This induces workers to retire later because more years are needed to recoup the returns on the later investments in further education. Allen and De Grip (2012) show that (older) workers’ perception of skills obsolescence drives this process as changes in skill requirements and the learning of new skills keep each other roughly in balance. They find that perceived skills obsolescence decreases the probability that older workers lose their employment in the coming years because those who experience a depreciation of their skills appear to learn more on the job and participate more often in training. Economic skills obsolescence also occurs when workers lose their job because of shrinking employment in the occupation or sector of industry in which they were employed. Autor and Dorn (2009) show that a lower demand for routine tasks within local markets disproportionately raises the share of older workers employed in middle-skill, nonroutine jobs. While young workers reallocate downward and upward to different occupations, older workers only move downward. The latter suffer most from shrinking employment because their high skill specificity raises the costs of occupational mobility. The graying of low-skill, nonroutine jobs is aggravated because it deters young workers from entering occupations in which these tasks are more prevalent. Hence, shrinking occupations are “getting old.” Indeed, various studies confirm that displaced workers suffer from severe employment and income losses (e.g., Couch and Placzek, 2010) and that these losses are highest for older displaced blue-collar workers with vocational training (Hanushek et al., 2017). Jacobson et al. (2005), however, show that displaced workers can derive significant net benefits from retraining around the time of their job loss although the returns associated with different types of courses vary substantially. Moreover, for older displaced workers retraining will be less beneficial because of the shorter time they can generate returns on new human capital investments by using it in the labor market.

19.4

Lifelong Learning: Returns on Human Capital Investments of Older Workers

Based on the human capital theory prediction that older workers have less returns on their investments in training because of the shorter remaining time in the labor market to generate these returns, we would expect that older workers and their employers have fewer incentives to invest in their human capital. As such, lower training rates among older people would be consistent with human capital theory. Nevertheless, because older workers could suffer from 354

Education and Ageing: Human Capital Investments and Ageing

both technical and economic skills obsolescence, it is crucial that they continue their human capital investments to ensure that they remain employable until the end of their working life. Many studies have found that the training rate among older workers is indeed substantially lower than among younger and middle-aged workers. In their review of the existing evidence on workplace training in Europe in different data sources, the Continuing Vocational Training Survey (CVTS), OECD data, and the European Community Household Panel, Brunello et al. (2007) find that training in Europe increases with education and skill intensity of occupations, while it decreases with age. Age-training profiles are downward sloped in all countries, although the slope differs among countries. Using data from the European Union (EU) Labour Force Survey that allows making a distinction between different types of training, Carmichael and Ercolani (2014) examine the relationship between age and training in the 15 European Union countries before the EU enlargement in 2004. Their main finding is that across the EU15 countries those aged 50–64 are less likely to participate in both general and work-related training. One should be cautious, however, interpreting these negative age-training profiles as evidence confirming the human capital theory. First, the validity of the human capital theory crucially depends on the expected duration of employment, rather than age, for the decision to participate in training (Taylor and Urwin, 2001). Montizaan et al. (2010), however, find strong causal evidence suggesting that the expected duration of employment plays a crucial role in determining the training participation of older workers. Their study uses a natural experiment in which a pension reform reduced future pension benefits for workers who were born in 1950 or later, while those born in 1949 remained entitled to the old pension rights. They found that this exogenous shock to pension rights postponed the expected retirement date and increased participation in training courses among older workers, although exclusively for those employed in large organizations. Brunello and Comi (2015) have a similar result, analyzing the impact of an Italian pension reform: Training incidence increased 9 percent following a 1-year increase in the minimum retirement age. Second, the lower training rates of older people may be caused by having fewer opportunities for training at an older age or can be caused by institutional limitations. Third, whether the negative age-training profiles are caused by lower incentives from older workers, their employers, or both, is unclear. From the workers’ point of view, on-the-job training has a significant positive effect on their employability. Although training policies that provide monetary subsidies to stimulate training participation among older workers have been criticized, in particular because of the substantial deadweight losses involved and the small expected returns (see Heckman, 2000; Falch and Oosterbeek, 2011; Abramovsky et al., 2011), several studies have found substantial individual returns to training. As mentioned earlier, Cunha and Heckman’s (2007) show that later investments in human capital are needed to harvest the dynamic complementarity between early and later investments, where the optimal ratio of early and later human capital investments depends on the degree of complementarity between early and late investments. While the substantial deadweight losses are particularly found for lower-educated individuals without employment, for workers who regularly invest in their human capital the deadweight losses are lower and expected returns remain sizable. Picchio and van Ours (2013) show for the Netherlands that firm-provided training significantly increases workers’ future employment prospects. This also holds for older workers, suggesting that firm-provided training remains an important instrument to retain older workers at work. Dostie and L´eger (2014) further find, using longitudinal linked employer-employee data for Canada and accounting for endogenous training decisions, that the training wage premium only diminishes slightly with age. Hence, older workers should, on average, have positive incentives to continue their investments in training. 355

Andries de Grip and Raymond Montizaan

Carmichael and Ercolani (2015) performed Fairlie-Blinder-Oaxaca decompositions to disentangle the extent to which age-training gaps are due to direct negative age-training coefficient effects and differences in observable characteristics or employer behavior. Most interesting is that they find that for 15 European Union countries the negative age-training slope is only partly caused by the preferences and lack of motivation of older people; barriers created by employers are also an important factor. They further find that higher overall training rates and higher rates of employment participation for older people are important mitigating factors that can reduce the age-training gap. This hints at the importance of having access to training. Several other studies further find that employers play an important role in the lower training participation of older workers. Using a large, linked employer-employee data set, Zwick (2015) shows that the training of older employees is indeed less effective according to a self-assessment of the training participants. The main reason for the age difference in training effectiveness during the life cycle is that firms do not consider the differences in training motivation between younger and older workers. While older workers get higher returns from informal and directly relevant training and from training content that are more easily tackled by their higher crystallized abilities (Salthouse, 2004), these better matching training types are not more often provided to older employees. Zwick (2015) thus confirms that training access seems to be a vital problem, although the problem seems to be nuanced, i.e., even when older workers do have access to training courses, they do not have access to the training courses that match their preferences. To avoid deadweight losses and low effectiveness of training investments, training programs need to be better designed to match the preferences and current skills of older workers (Beier, 2008).

19.5

Training and Retirement of Older Workers

The relationship between older workers’ training participation and employability and their retirement age runs in both directions. A higher expected retirement age might increase the likelihood that workers and employers are willing to invest in older workers’ human capital, but training participation can also have a motivational impact to continue working. Several studies show that developmental opportunities for older workers can have motivation-enhancing effects (Polat et al., 2017), can reduce intentions to retire early (Armstrong-Stassen and Schlosser, 2008; Armstrong-Stassen and Ursel, 2009), and can delay actual retirement decisions (Herrbach et al., 2009). Yet, several other studies suggest that human resource practices other than training (such as job flexibility) are more influential in affecting older workers’ motivation to continue working (G¨obel and Zwick, 2013; Veth et al., 2015). More recent evidence, however, shows a more subtle view on the motivational impact of training. De Grip et al. (2020) examine the incremental effect of having training opportunities over and above workers’ actual training participation. Grounded in social exchange theory, they argue that the effect of training opportunities on the expected retirement age depends on employees’ positive reciprocity orientation. Using linked employer-employee data they show that training opportunities positively associate with workers’ expected retirement age over and above their actual training participation, but only for those with strong positive reciprocity beliefs. Hence, the offer of training itself is a credible gesture to facilitate employees’ future employability, stimulating exclusively the workers who are positively reciprocal to return this gift by raising their expected retirement age. However, employers might not be inclined to postpone the retirement age of all older employees, depending on the type of human capital these employees have acquired in their working careers. In the case of firm-specific training investments, life-cycle theory predicts that employers enter implicit contracts with newly hired employees to ensure rent sharing and to 356

Education and Ageing: Human Capital Investments and Ageing

decrease turnover after firm-specific training investments (Kennan, 1979; Blinder, 1982; Johnson, 1996; Fella, 2005). The starting point of this literature is the bilateral monopoly problem between workers and their employers after firm-specific training investments that may lead to undesirably high employee turnover. The two parties must agree on how the rents from firmspecific human capital investments will be shared after the investment. Implicit contracts that include upward-sloping earning profiles, mandatory retirement, and severance pay if workers are prematurely discharged are a solution to this problem. Because older workers with firm-specific skills due to the upward-sloping earning profiles earn more than their productivity at the end of their career, they have an incentive to retire later, while employers have an incentive to terminate the labor contract as soon as workers’ earnings exceed their productivity. Mandatory retirement is, therefore, a necessary feature of the implicit contract. Hence, workers with firm-specific skills and long careers at one firm might be confronted with barriers to retire later than initially planned. However, employers have no incentives to enforce mandatory retirement of workers with general skills. Montizaan et al. (2013) test this hypothesis using the U.S. National Longitudinal Survey of Older Men and find that workers who participated in firm-specific training in their early careers do indeed retire earlier than those with general skills. This suggests that older workers with firm-specific human capital must retire when they reach the common mandatory retirement age of 65. These results indicate that the effectiveness of institutional arrangements to postpone retirement in many industrialized countries will therefore also depend on the training policies of employers and the type of skills workers acquired in the past. Workers with general skills can expect to continue working to a later age, while those with firm-specific human capital will experience barriers to continued employment raised by their employers.

19.6

Research Agenda

From an early stage, research on human capital and ageing could build on the straightforward theoretical models of Becker (1962) and Ben-Porath (1967). These models clearly explain why human capital investments of older workers are less beneficial than investments earlier in life. However, both the depreciation of human capital due to the shifting skill demands induced by technological change (Neuman and Weiss, 1995) and the postponement globally of retirement eligibility (B¨orsch-Supan and Coile, 2018) increase the demand for human capital investments later in life. Although the literature shows that demands for new skills and the postponement of retirement increase the human capital investments of older workers, these investments are particularly hampered by shocks due to technological change (Bartel and Sicherman, 1993) and shocks in retirement eligibility (Montizaan et al., 2010). This, in turn, has increased resistance among older workers to pension reforms in many countries. Therefore, increasing our insights into ways to increase human capital investments of older workers that increase their employability is important. Insights into the effect of specific policy interventions in this field are best obtained by random control trials (RCTs) on various interventions that aim to increase training participation of older workers with high returns in the labor market. These RCTs have a strong internal validation that allows for finding causal effects. However, as they focus on specific interventions in a specific institutional context, the external validity of the effects found is often less evident. Therefore, our first suggestion for a future research agenda on human capital and ageing would be to focus on a broad series of RCTs on various policy interventions in countries with different institutional contexts, including innovative interventions and replication studies in different contexts. These RCTs should not only focus on the deadweight losses of training for unemployed, lower-educated individuals but also on the returns of smaller training courses aimed at workers who have regularly invested in their human capital during their life course. 357

Andries de Grip and Raymond Montizaan

Our second suggestion for the future research agenda follows from the studies that showed that firms do not sufficiently consider the differences in training motivation between younger and older workers and that older workers lack access to the training courses that match their preferences (Beier, 2008; Zwick, 2015). More research is needed on the specific preferences of older workers with respect to the skills they would like or need to invest in and on the training modes that raise the returns to training for older workers most effectively, thereby stimulating employers to keep investing in the human capital of their older employees. Is developing low-cost training courses that optimally use the crystallized intelligence of older workers as reflected in their current skills possible? And to what extent do on-the-job training, informal learning in the workplace, or hybrid training forms lead to higher returns for older workers than more formal training forms? Along these lines, we should find answers to the crucial question of how we can better customize human capital investments to account for the changing needs and preferences over workers’ lengthening life cycle. Answering these research questions requires developing new data sources combining surveys and register data and possibly big data on training platforms that can lead to much more detailed information on skills development and preferences. Our final suggestion stems from the speed of technological changes and their impacts on changing skills demands in the labor market. Because workers are increasingly challenged to keep their skills up to date and must continue investing until a later age than in the recent past, having better foresight into the skills that new technologies demand and their impact on the depreciation of the skills of older workers is important. Future technological changes such as automation, robotization, and artificial intelligence may displace various skills but can also increase the demand for skills complementary to these new technologies. To enable workers to make timely human capital investments in anticipation of these changes, regular forecasts and simulations of plausible scenarios are needed to assess the impact of future technological developments on various segments of the labor market.

References ABRAMOVSKY, L., BATTISTIN, E., FITZSIMONS, E., GOODMAN, A., AND SIMPSON, H. (2011): “Providing employers with incentives to train low-skilled workers: Evidence from the UK employer training pilots,” Journal of Labor Economics, 29(1): 153–193. AGRAWAL, A., GANS, J. S., AND GOLDFARB, A. (2019): “Artificial intelligence: The ambiguous labor market impact of automating prediction,” Journal of Economic Perspectives, 33(2): 31–50. ALLEN, J., AND DE GRIP, A. (2012): “Does skill obsolescence increase the risk of employment loss?,” Applied Economics, 44(25): 3237–3245. ARMSTRONG-STASSEN, M., AND SCHLOSSER, F. (2008): “Benefits of a supportive development climate for older workers,” Journal of Managerial Psychology, 34(4): 419-437. ARMSTRONG-STASSEN, M., AND URSEL, N. D. (2009): “Perceived organizational support, career satisfaction, and the retention of older workers,” Journal of Occupational and Organizational Psychology, 82(1): 201–220. AUTOR, D., AND DORN, D. (2009): “This job is ‘getting old’: Measuring changes in job opportunities using occupational age structure,” American Economic Review, P&P, 99(2): 45–51. AUTOR, D., AND SALOMONS, A. (2018): “Is automation labor-displacing? Productivity growth, employment, and the labor share,” Working Paper No. w24871. National Bureau of Economic Research, Cambridge, MA. BARTEL, A. P., AND SICHERMAN, N. (1993): “Technological change and retirement decisions of older workers,” Journal of Labor Economics, 11(Part 1): 162–183. BECKER, G. S. (1962): “Investment in human capital: A theoretical analysis,” Journal of Political Economy, 70(5, Part 2): 9–49. BEIER, M. E. (2008): “Age and learning in organizations.” In: Hodgkinson, G. P., and Ford, J. K. (eds.), International Review of Industrial and Organizational Psychology, Vol. 23. Chichester: John Wiley & Sons, pp. 83–106.

358

Education and Ageing: Human Capital Investments and Ageing

BEN-PORATH, Y. (1967): “The production of human capital and the life cycle of earnings,” Journal of Political Economy, 75(4, Part 1): 352–365. BLINDER, A. S. (1982): “Private pensions and public pensions: Theory and fact,” NBER Working Paper No. w0902. National Bureau of Economic Research, Cambridge, MA. ¨ BORSCH -SUPAN, A. H., AND COILE, C. (2018): “Social security programs and retirement around the world: Reforms and retirement incentives–Introduction and summary,” Working Paper No. w25280. National Bureau of Economic Research, Cambridge, MA. ¨ BORSCH -SUPAN, A., AND WEISS, M. (2016): “Productivity and age: Evidence from work teams at the assembly line,” The Journal of the Economics of Ageing, 7(April): 30–42. BRUNELLO, G., AND COMI, S. (2015): “The side effect of pension reforms on the training of older workers. Evidence from Italy,” The Journal of the Economics of Ageing, 6(December): 113–122. BRUNELLO, G., GARIBALDI, P., AND WASMER, E. (EDS.). (2007): Education and Training in Europe, Oxford, New York: Oxford University Press. BRUNELLO, G., AND ROCCO, L. (2017): “The labor market effects of academic and vocational education over the life cycle: Evidence based on a British cohort,” Journal of Human Capital, 11(1): 106–166. CARMICHAEL, F., AND ERCOLANI, M. G. (2014): “Age-training gaps in the European Union,” Ageing & Society, 34(1): 129–156. CARMICHAEL, F., AND ERCOLANI, M. G. (2015): “Age-training gaps across the European Union: How and why they vary across member states,” The Journal of the Economics of Ageing, 6(December): 163–175. CORNELISSEN, T., DUSTMANN, C., AND SCHOENBERG, U. (2017): “Peer effects in the workplace,” American Economic Review, 107(2): 425–456. COUCH, K., AND PLACZEK, D. (2010): “Earnings losses of displaced workers revisited,” American Economic Review, 100(1): 572–589. CUNHA, F., AND HECKMAN, J. (2007): “The technology of skill formation,” American Economic Review, 97(2): 31–47. D’ADDIO, A. C., KEESE, M., AND WHITEHOUSE, E. (2010): “Population ageing and labour markets,” Oxford Review of Economic Policy, 26(4): 613–635. DE GRIP, A., FOUARGE, D., MONTIZAAN, R., AND SCHREURS, B. (2020): “Train to retain: Training opportunities, positive reciprocity, and expected retirement age,” Journal of Vocational Behavior, 117(103332): 1–15. DE GRIP, A., SAUERMANN, J., AND SIEBEN, I. (2016): “The role of peers in estimating tenureperformance profiles: Evidence from personnel data,” Journal of Economic Behavior & Organization, 126(Part A): 39–54. DE GRIP, A., AND VAN LOO, J. (2002): “The economics of skills obsolescence: A review.” In: De Grip, A., Van Loo, J., and Mayhew, K. (eds.), The Economics of Skills Obsolescence, Research in Labor Economics, vol. 21. Amsterdam/Boston, MA: JAI Press, pp. 1–26. DEMING, D. J. (2017): “The growing importance of social skills in the labor market,” The Quarterly Journal of Economics, 132(4): 1593–1640. DOSTIE, B., AND LE´ GER, P. T. (2014): “Firm-sponsored classroom training: Is it worth it for older workers?,” Canadian Public Policy, 40(4): 377–390. FALCH, T., AND OOSTERBEEK, H. (2011): “Financing lifelong learning: Funding mechanisms in education and training,” EENEE Analytical Report n. 10. EENEE, Munich. FELLA, G. (2005): “Termination restrictions and investment in general training,” European Economic Review, 49(6): 1479–1499. ¨ GOBEL , C., AND ZWICK, T. (2013): “Are personnel measures effective in increasing productivity of old workers?,” Labour Economics, 22(June): 80–93. GOLSTEYN, B., AND STENBERG, A. (2017): “Earnings over the life course: General versus vocational education,” Journal of Human Capital, 11(2): 167–212. GOULD, E. D., MOAV, O., AND WEINBERG, B. A. (2001): “Precautionary demand for education, inequality, and technological progress,” Journal of Economic Growth, 6(4): 285–315. HANUSHEK, E. A., SCHWERDT, G., WOESSMANN, L., AND ZHANG, L. (2017): “General education, vocational education, and labor-market outcomes over the life-cycle,” Journal of Human Resources, 52(1): 48–87. HECKMAN, J. J. (2000): “Policies to foster human capital,” Research in Economics, 54(1): 3–56. HECKMAN, J. J. (2007): “The economics, technology, and neuroscience of human capability formation,” Proceedings of the National Academy of Sciences, 104(33): 13250–13255. HERBST, D., AND MAS, A. (2015): “Peer effects on worker output in the laboratory generalize to the field,” Science, 350(6260): 545–549.

359

Andries de Grip and Raymond Montizaan

HERRBACH, O., MIGNONAC, K., VANDENBERGHE, C., AND NEGRINI, A. (2009): “Perceived HRM practices, organizational commitment, and voluntary early retirement among late-career managers,” Human Resource Management, 48(6): 895–915. JACOBSON, L. S., LALONDE, R., AND SULLIVAN, D. (2005): “Estimating the returns to community college schooling for displaced workers,” Journal of Econometrics, 125(1–2): 271–304. JOHNSON, R. W. (1996): “The impact of human capital investments on pension benefits,” Journal of Labor Economics, 14(3): 520–554. KENNAN, J. (1979): “Bonding and the enforcement of labor contracts,” Economics Letters, 3(1): 61–66. LINDEBOOM, M., AND MONTIZAAN, R. (2020): “Disentangling retirement and savings responses,” Journal of Public Economics, 192: 104297. MCDOWELL, J. M. (1982): “Obsolescence of knowledge and career publication profiles: Some evidence of differences among fields in costs of interrupted careers,” American Economic Review, 72(4): 752–768. MINCER, J. (1962): “On-the-job training: Costs, returns, and some implications,” Journal of Political Economy, 70(5, Part 2): 50–79. MINCER, J. (1974): Schooling, Experience and Earnings, New York: National Bureau of Economic Research. ¨ MONTIZAAN, R., CORVERS , F., AND DE GRIP, A. (2010): “The effects of pension rights and retirement age on training participation: Evidence from a natural experiment,” Labour Economics, 17(1): 240–247. ¨ MONTIZAAN, R., CORVERS , F., AND DE GRIP, A. (2013): “Training patterns and early retirement,” Applied Economics, 45(15): 1991–1999. NEUMAN, S., AND WEISS, A. (1995): “On the effects of schooling vintage on experience-earnings profiles: Theory and evidence,” European Economic Review, 39(5): 943–955. PFEIFFER, F., AND REUSS, K. (2008): “Age-dependent skill formation and returns to education,” Labour Economics, 15(4): 631–646. PICCHIO, M., AND VAN OURS, J. C. (2013): “Retaining through training even for older workers,” Economics of Education Review, 32(1): 29–48. POLAT, T., BAL, P. M., AND JANSEN, P. G. (2017): “How do development HR practices contribute to employees’ motivation to continue working beyond retirement age?,” Work, Ageing and Retirement, 3(4): 366–378. ROSEN, S. (1975): “Measuring the obsolescence of knowledge.” In: Juster, F. T. (ed.), Education, Income and Human Behavior, New York: Carnegie Foundation, pp. 199–232. RYAN, P. (2001): “The school-to-work transition: A cross-national perspective,” Journal of Economic Literature, 39(1): 34–92. SALTHOUSE, T. A. (2004): “What and when of cognitive ageing,” Current Directions in Psychological Science, 13(4): 140–144. TAYLOR, P. E., AND URWIN, P. (2001): “Age and participation in vocational education and training,” Work, Employment and Society, 15(4): 763–779. VETH, K. N., EMANS, B. J., VAN DER HEIJDEN, B. I., KORZILIUS, H. P., AND DE LANGE, A. H. (2015): “Development (f) or maintenance? An empirical study on the use of and need for HR practices to retain older workers in healthcare organizations,” Human Resource Development Quarterly, 26(1): 53–80. WEINBERGER, C. (2014): “The increasing complementarity between cognitive and social skills,” Review of Economics and Statistics, 96(4): 849–861. ZWICK, T. (2015): “Training older employees: what is effective?,” International Journal of Manpower, 36(2): 136–150.

360

PART IV

Work and Employment

20 THE EMPLOYMENT OF OLDER WORKERS Hippolyte d’Albis

Abstract The participation of older workers in the labor market has increased in most high-income countries since the mid-1990s. This can be explained by changes in the characteristics of recent generations of older workers. The labor supply is healthier and better trained and includes more women. This can also be explained by reforms of retirement systems and of health and unemployment insurance systems, even if modifications may still be required to guarantee sustainability. In addition, recurring concerns about the productivity of older workers and their ability to adapt to technological change seem unfounded and likely to stem from prejudice against them. Finally, the promises associated with greater flexibility in working conditions have not, to date, led to a significant increase in the work of older individuals.

20.1

Introduction

The employment of older individuals is a key variable in the adaptation of the economies of high-income countries to the ageing population. Increasing their participation in the labor market can help both improve the financial sustainability of the social welfare system and secure individuals’ resources after retirement. Nevertheless, determining whether workers can work longer both in terms of their health and labor market prospects is crucial. For instance, raising the legal retirement age in a country where older people cannot find work is a difficult policy to implement. The employment of older individuals is often assessed through the participation rate of people aged 55 and older in the labor market. Two major stylized facts characterize this participation. First, strong heterogeneity exists among countries. The labor force participation rate of people aged 55–64 was on average 64.4 percent in 2019 in Organisation for Economic Co-operation and Development (OECD) countries, ranging widely from less than 37 percent in Turkey to more than 82 percent in Iceland. Second, over a long period, the trend has been U-shaped. At the end of the 1960s, participation was more than 62 percent (on average in OECD countries); it then steadily declined to reach a low of less than 49 percent in the mid-1990s and has since DOI: 10.4324/9781003150398-24

363

Hippolyte d’Albis

gone back up. Note that this evolution can be found in numerous countries, even if the date of the reversal varies. For the United States, for example, the low point was reached in 1986 with 54 percent compared with more than 65 percent in 2019. This chapter presents the recent literature on the main factors influencing the employment of older individuals. It is structured in three parts. The first part examines the determinants of the labor supply of older individuals, particularly the evolution of their health and education level and the composition effect due to the increase in the share of women in the workforce. The second part deals with the transformation of the demand for older workers and employers’ attitudes toward them. It addresses issues related to the productivity of older individuals, technological transformations, and working conditions. The third part explores the public institutions and policies that affect the work of older individuals. In particular, it analyzes the role of social welfare based on retirement, health insurance, and unemployment insurance systems.

20.2

Changes in the Labor Supply

The characteristics of older individuals have undergone three major changes in recent decades. The health of recent generations and therefore their ability to work have improved considerably. Their qualifications have also been profoundly modified, with a strong increase in the duration of their initial training. Finally, the proportion of women in labor supply has clearly increased. These three developments have affected the employment rate of older individuals.

20.2.1

Health and Mortality

The health of populations, and in particular of older individuals, has improved considerably. According to the United Nations (2019), in high-income countries, the life expectancy of 50year-old women is currently 35.2 years, whereas it was only 31.7 years 25 years ago and 27.9 years 50 years ago. This dramatic improvement can also be observed for men at the same age, although their levels remain lower: 30.8 years today compared with 26.5 years 25 years ago and 23.2 years 50 years ago. At first glance, the consequences of this change on the employment of older people appear quite intuitive and are often put forward in the public debate on the legal retirement age. The idea is based on a principle of proportionality for major life-cycle events. If individuals live longer, they are expected to enter the labor market later (because they train longer) and exit later, which should therefore increase the employment rate of older individuals. However, studies based on life-cycle theory show that the relationship may be more complex. An increase in life expectancy automatically leads to having to allocate resources over a longer period of life, which, in a theoretical framework in which a rational agent makes optimal choices, is an incentive to consume less in each period and to work more. Consumption and leisure are indeed correlated in these models. However, the increase in life expectancy may also lead to an increase in human wealth, calculated as the expected present value of future income. This “income effect” translates into the opposite effect to that generated by the planning period: It implies an increase in consumption and a decrease in the labor supply, i.e., retirement at a younger age. Various authors note this ambiguity, highlighting the importance of uncertainty over individuals’ lifetime (Chang, 1991; Kalemli-Ozcan and Weil, 2010), the modification of the preference for leisure generated by an increase in life expectancy (Bloom et al., 2007, 2014), the role of the epidemiological transition (d’Albis et al., 2020), and the effect of longevity on the duration of education (Hazan, 2009; Cervellati and Sunde, 2013; S´anchez-Romero et al., 2016). The consensus is that if mortality gains occur at the end of life, as is currently the case in high-income countries, then the planning effect dominates the income effect: The increase in life expectancy is then associated with a higher labor supply for older individuals. 364

The Employment of Older Workers

Milligan and Wise (2015) propos one method to assess the effect of improved longevity on the labor supply. The authors consider the death rate at a given age to be an indicator of the state of health at that age and, therefore, of the capacity to work. For a given country and for different dates, they analyze the relationship between the death rate and the employment rate at each age; then, they observe how the employment rate changes over time as a function of the death rate. Wise (2017) presents an application of this method to a large number of high-income countries. According to this study, for example, the employment rate was 56.2 percent for American men aged 60–64 in 2010. However, in 1977, people with the same death rate (who were therefore younger) had an employment rate of 83.2 percent. This means that the improvement in health did not translate into a proportional increase in the employment rate, and according to the terminology proposed by Milligan and Wise, this age group has a 27 percent “additional employment capacity.” This capacity has nevertheless fallen sharply since the 2000s. Countries also differ considerably, primarily due to differences in the level of employment: The additional employment capacity is high in the United Kingdom and France, whereas it is low in the Netherlands and Sweden (Coile at al., 2017). These initial studies could be improved by accounting for longevity differences between social categories (Waldron, 2007). Recent evidence indeed suggests that the gap in life expectancies is actually expanding. For instance, for U.S. men born in the 1930 cohort, the highest income quintile’s life expectancy at age 50 is 5.1 years longer than the lowest quintile’s, while for those born in the 1960 cohort, the projected gap widens to 12.7 years (National Academies, 2015). More recently, using statistical data from American counties, Currie and Schwandt (2016) show that since 1990 life expectancy inequalities have increased among those over 50, while they have decreased for the population as a whole. The health of the wealthiest older individuals has improved more than that of the poorest. Clearly, not only changes in longevity but also changes in health at older ages affect employment trends. The link between mortality and health is intuitive—a generation’s state of health at a given age is correlated with its life expectancy—but proving this in a systematic manner is not easy because health data are not necessarily comparable over time and across countries. Several studies have nevertheless made it possible to assess an active life expectancy (or healthy life expectancy), that is to say a lifetime without being incapacitated (Robine et al., 2003), whether through an international comparison (Salomon et al., 2012) or for specific countries. For example, Manton et al. (2006) estimate that 65-year-old American men experienced an increase in their active life expectancy from 10.9 years in 1965 to 13.9 years in 1999, while their life expectancy went from 15 years to 17.7 years over the same period. This leads to the conclusion that the proportion of life spent in good health has been increasing (from 72.7 percent to 78.5 percent), while the years spent in poor health have been decreasing (from 4.1 years to 3.8 years). Serious illnesses such as cancer and heart or lung disease (McClellan, 1998) affect the labor supply of older individuals, mainly at the extensive margin (French, 2005; Coile at al., 2017), but also at the intensive margin, although to a lesser extent (Blundell et al., 2016). As French and Jones (2017) point out, poor health can affect the labor supply because it increases the disutility of labor (Capatina, 2015), reduces productivity and thus labor compensation, and decreases the need for retirement savings.

20.2.2

Education and Qualifications

In all high-income countries, the increase in education levels has been dramatic. According to OECD (2021), 43.4 percent of Americans aged 55–64 had a tertiary education in 2019, compared with 28 percent in 1999. When taking the average for OECD countries, the share was 365

Hippolyte d’Albis

28.4 percent, compared with 13.7 percent, i.e., more than doubling in 20 years. The differences among countries are considerable, ranging, for example, from 12.8 percent in Italy to 49.8 percent in Canada. This upward trend can be found in all age groups and when looking at women only. For men, the situation has been improving on average for the OECD, but it has been stagnating in many countries. In the United States, for example, 42.6 percent of men aged 55–64 had a tertiary education in 2019, a figure that has remained unchanged for 10 years. This improvement in qualifications has changed the tasks older workers perform. In the United States, according to Belbase and Chen (2019), 38 percent of workers aged 55–64 perform routine occupations today, compared with 55.3 percent in 1979, and 85.6 percent use a computer at work, compared with 49 percent in 1998. The authors highlight that these proportions are practically the same for workers aged 25–54. This educational improvement has come with greater professional mobility, with an increase in job changes after age 50. However, Rutledge et al. (2017) note that since the 1990s, the range of professional options has grown for more educated individuals and narrowed for less educated men. The situation of older individuals on the labor market is closely linked to their education level. Labor market participation increases with education level at any age (in the OECD, the employment rate of 25–64-year-olds is 84.6 percent for people with tertiary education versus 57.7 percent for those with less than upper secondary education), and the desire to stay in employment after age 60 also increases with education level (Wahrendorf et al., 2013). At older ages, health inequalities, which correlate with educational inequalities, can explain part of these differences (Carr et al., 2018). This implies that the least educated older individuals are more likely to leave the labor market involuntarily, for health reasons or due to layoffs (M¨acken et al., 2021). In some countries, the rules of the retirement system can also explain the differences by education level. As the more qualified entered the labor market later, they may want to prolong their activity to receive a full pension. Finally, diplomas and qualifications are a protection against layoffs. However, while diplomas have become more widespread among the population, some authors underline that their role as employment protection is increasingly less effective. The case of “displacement,” that is a job loss associated with a change in the labor demand, is particularly telling. Zhivan et al. (2012) show that in the United States between 1984 and 2004, the probability of being displaced increased with age when controlling for tenure and that tertiary education protects less and less against displacement. While education is no longer sufficient to guarantee the security of older workers on the labor market, continuing education seems to be an interesting opportunity because it makes it easier for older workers to find a job in the event of unemployment. However, the incentive to train decreases as workers approach retirement because the expected gains are lower (Ben-Porath, 1967; Neumann and Weiss, 1995). Countries where the retirement system favors early departure are thus those where older workers train the least (Lau and Poutvaara, 2006; Fouarge and Schils, 2009). The overall change in education levels therefore certainly explains part of the improvement in the employment rate of older people observed since the early 2000s, but cannot fully explain the magnitude of the increase (Coile, 2018).

20.2.3

Women’s Labor Force Participation

For decades, high-income countries have been undergoing a revolution in gender relations, which has led to decreasing participation gaps both for paid activities (Goldin, 2006; Gershuny, 2000) and for domestic work (Pailh´e et al., 2021). This phenomenon has also affected older women workers. Thus, among people aged 55–64 living in OECD countries, women’s participation rate was 76 percent of men’s in 2019 compared with 61 percent in 2000. Older women are likely to continue to increase their participation in the labor market (Goldin and Katz, 366

The Employment of Older Workers

2018), but once again the differences among countries are considerable. In the United States, for example, the ratio is higher than the average for OECD countries—it reached 83.3 percent in 2019—but it has been declining slightly since 2010. Nonetheless, a gap between men and women remains and can still be explained for some countries by family taxation (Apps and Rees, 2009) and by an unequal distribution of work, particularly with regard to caring for dependent relatives. This is especially the case for dependent elderly parents (Johnson and Lo Sasso, 2006), the spouse when he is ill (Pozzoli and Ranzani, 2009), or grandchildren (Zamarro, 2020). The question of the effect of children on the work of mothers also arises for older women workers, as the average age at childbearing has increased. However, providing an answer is difficult because the literature mainly considers the case of young mothers: The negative effect of children on their labor supply has been demonstrated in many contexts (Del Boca et al., 2009). Among older mothers, recent evidence suggests instead that the departure of children is associated with a decrease in the labor supply (d’Albis et al., 2021). With the feminization of the labor supply, the proportion of couples with both spouses working has increased. Any institutional reform—in particular pension reforms—aimed at increasing one spouse’s work may therefore potentially affect the other spouse’s labor supply. The theoretical framework, which is that of collective choice models (Chiappori, 1988), identifies two contradictory effects. The increase in one spouse’s work first has an income effect that is likely to reduce the other spouse’s labor supply. Conversely, if the two spouses’ leisure time is complementary (i.e., they enjoy spending time together), then they are likely to coordinate their retirement and the increase in one spouse’s labor supply is likely to lead to an increase in the other one’s labor supply. Empirical studies in different countries have found a certain complementarity of leisure activities (Baker, 2002; Casanova, 2010; Michaud and Vermeulen, 2011). These studies have also revealed a certain asymmetry, whose direction varies among countries. For example, in the United States, men modify their labor supply following a reform affecting their wives, but the reverse is not true (Coile, 2004), while in Norway, women modify their labor supply following a reform affecting their husbands (Bratsberg and Stancanelli, 2018). The consequences of the complementarity of leisure activities within couples are essential to analyze the evolution of the work of older individuals. According to Schirle (2008), between 1995 and 2005, the increase in women’s participation could explain nearly 50 percent of the increase in older men’s participation rate in Canada. For the United States and the United Kingdom, this figure reaches 33 percent. Therefore, any policy promoting women’s employment is favorable to the employment of older individuals. In particular, policies promoting work-family balance, which encourage work in the years following childbirth, have a positive impact because they minimize the effects of career interruptions and make women’s employment likelier after age 55.

20.3

Technological and Structural Developments

The economic environment for older workers has also changed dramatically in recent decades. New technologies have disrupted production and the tasks of employees. Human resource management has also evolved, and new working conditions have been rolled out. These two major developments have changed the role of older workers and sometimes the attitude of firms toward them.

20.3.1

The Productivity of Older Workers

The public debate about older workers often revolves around the issue of their productivity. A very simplistic conception of firms consists of assuming that workers are paid at their marginal 367

Hippolyte d’Albis

productivity. If productivity decreases with age, it follows that it would be in the interest of firms to reduce the remuneration of their older workers, which is difficult especially if these workers have seniority; the default solution would therefore be to lay off older individuals as a priority or not to hire them. The decline in productivity with age could thus be a cause of the underemployment of older individuals. The issue is therefore empirical, but except for very specific professions (such as scientists, whose productivity is measured by the number and quality of publications), it is ultimately extremely difficult to settle (Allen, 2019). Clearly, certain physical abilities decrease with age, and psychological research on attention, perception, and working memory shows a sharp decrease in the latter with age (McDaniel et al., 2012). In addition, many studies report cognitive decline after age 50 (Anghel and Lacuesta, 2020). However, older workers benefit from experience, which in economics has long been known to be crucial in the production process or even in innovation (Arrow, 1962), and which is especially reflected in better work organization and planning (Romeu Gordo and Skirbekk, 2013). To determine which of the two effects prevails, empirical assessments of the relative productivity of older individuals are based either on the comparison of similar firms with different proportions of older workers or on firm-specific analyses. The former suffers from multiple biases and has conflicting results, while the latter are difficult to generalize (Allen, 2019). Researchers therefore have struggled to determine which effect prevails, suggesting that the drop in productivity, if it does exist, must not be very significant. This vision, which centers on the individual, is nevertheless somewhat outdated, and examining the age structure of the workforce in terms of complementarity seems more relevant. Younger workers benefit from more recent initial training and are therefore more in tune with technological developments, while older workers benefit from their experience. If the substitutability between workers is assumed to be imperfect, it follows (provided that the usual assumptions on the production functions are verified) that the marginal productivity of the younger workers increases with the number of older workers, and vice versa. The effect depends on the elasticity of substitution but also on the relative proportion of older individuals among workers. In this respect, if the parameters of the function are known, the optimal proportion of older workers can be determined. However, understanding that the increase in the productivity of young workers resulting from an increase in the number of older workers does not necessarily imply an increase in the demand for young people is important. If a company is “compelled” to keep its older workers, an adjustment may occur through a decrease in the number of young people. Empirically, this effect can be observed in studies that have, for example, evaluated the effect of an increase in the retirement age in Italy (Boeri et al., 2016; Bovini and Paradisi, 2019) or in Norway (Vestad, 2013). In the United States, the adjustment instead occurs through wages or the quality of the jobs reserved for young people (Mohnen, 2019). However, these microeconomic studies do not account for the general equilibrium effects generated by the increase in the employment rate of older individuals, particularly the increase in firms’ total production and the improved balance of public finances due to more taxes collected. Those effects are likely to improve labor demand for both younger and older workers. Thus, it is unsurprising that macroeconomic studies lead to a more positive view of increased employment of older individuals. Using a panel of 22 OECD countries over the period 1960– 2008, Kalwij et al. (2010) show that the employment of older individuals does not replace that of the young. The results are similar for the United States (Gruber and Milligan, 2010; Munnell and Wu, 2012, 2020).

368

The Employment of Older Workers

20.3.2

Sectorial Transformations

Since the First Industrial Revolution, technical progress has caused fears of downgrading and obsolescence among workers. The current context is nevertheless particularly worrisome due to the scale of activities and professions that the digital revolution may affect (Brynjolfsson and McAfee, 2014; Schwab, 2017). Traditionally, technical progress has competed with low-skilled and unskilled workers who perform routine occupations. Today, big data and artificial intelligence are likely to replace qualified workers in intellectual occupations. For example, according to some projections, computers could replace 47 percent of the American workforce (Frey and Osborn, 2017) and 55 percent of the Japanese workforce (David, 2017). These projections have been much discussed in recent literature that assesses the aggregate effects by accounting for the productivity gains associated with the automation of certain tasks (Acemoglu and Restrepo, 2020). For older workers, the problem seems particularly crucial. Indeed, science, technology, engineering, and mathematics (STEM) skills, which seem complementary to new technologies, are particularly scarce among older workers. According to OECD (2016), 55 percent of people aged 55–64 lack the basic skills to use information and communication technology, and only 10 percent can perform more complex tasks involving several steps, compared with 42 percent of 25–64-year-olds. The problem is not only due to differences across cohorts in initial training (Hudomiet and Willis, 2021) but also to the very rapid obsolescence of the STEM skills acquired in initial training. According to Deming and Noray (2018), computer science and engineering students have lost 50 percent of their earning premium after a 10-year career. Technical progress, which is well known to generate a skill bias (Berman et al., 1998), is therefore likely to also generate a bias in favor of the younger who are more educated and more capable with new technologies. Almost 30 years ago, Bartel and Sicherman (1993) tackled the issue by emphasizing the crucial role of lifelong learning and anticipation. They show that if technological changes are anticipated and workers trained, then technical progress tends to prolong the careers of older workers: those working in industries where technological growth is strong retire later. Friedberg (2003) corroborates this, showing that, among older workers, computer users retire later than others, and Haegeland et al. (2007) highlight that older workers leave companies that adopt significantly different technologies later. Several studies since then (Ahituv and Zeira, 2011; Messe et al., 2014; Burlon and Vilalta-Buf´ı, 2016) also underline the heterogeneity of reactions to technical progress. This literature therefore puts the damage that the digital revolution may cause to the employment of older individuals into perspective, all the more so as new generations of older workers are better trained. Technical progress is undoubtedly a driver of economic growth, but the activities in which it is concentrated are not necessarily the only ones that are useful to society. The care sector for the elderly—home care providers in particular—is booming, and the needs due to population ageing will be numerous, as the situation in Japan suggests (Martin, 2018). More generally, the entire service sector is likely to undergo changes due to the digital revolution. The case of the “gig economy” is a good illustration. A priori, the flexibility provided by its platforms in terms of having multiple jobs or combining a job and a pension should appeal to older individuals and allow a smoother transition between working life and retirement. Nevertheless, studies show that a few older individuals participate in the gig economy (Oyer, 2016; Katz and Krueger, 2019), even if their behavior differs from that of younger workers or even complements it. Using data on Uber drivers in Chicago, Cook et al. (2019) show, for example, that older drivers prefer less congested places and times. They are, therefore, paid less than younger Uber drivers, but also than taxi drivers of the same age. The sectorial composition of the economy is changing, and so is employment. Workers—young and old alike—are adapting to these changes. The debate 369

Hippolyte d’Albis

can also be viewed from the opposite angle: do older individuals contribute to technological change? The literature on the topic has received much attention (see, for example, Acemoglu and Restrepo, 2018) by going against received ideas that associate risk-taking with youth and conservatism with older ages. Unfortunately, prejudices against older individuals are omnipresent in the labor market.

20.3.3

The Transformations of Working Conditions

If the importance of a research topic is measured by the number of academic publications dealing with it, then discrimination against older individuals is clearly an important topic. This concern is also very present in public debates, despite a protective legislation. In the United States, a set of laws has protected older workers since 1967, and disputes are often taken to court. In the European Union, the Treaty of Amsterdam signed in 1997 has provided the legal basis for making age discrimination illegal. Like other forms of discrimination at work, age discrimination can affect hiring, firing, and promotions. Academic research mainly focuses on discrimination in hiring because it is clearly the most difficult to establish legally and must therefore be the most widespread. This research is based on responding to job advertisements by sending pairs of resumes that are identical, except for the person’s age. Lahey (2008) sent the resumes of women aged 35–62, for example, and found that younger women were about 45 percent more likely to get job interviews. However, this type of study suffers from two major biases. First, producing similar resumes for two people of different ages is difficult because this implies either that the younger one has a lot of experience or the older one has little. Second, the heterogeneity of career paths increases sharply with age, which means that the information provided by an older individual’s resume is less precise than that on a young individual’s resume. To minimize these biases, Neumark et al. (2019) conducted a large-scale experiment (based on 40,000 applications!) and showed clear discrimination against older women (backing up the work of Farber et al., 2015) and against people aged 64–66, who are therefore very close to retirement. Discrimination against younger men may occur, but the authors lack strong evidence of this. In addition to hiring discrimination, firms may also offer less favorable terms to the older workers they do hire. By analyzing the ads posted on a site geared to older workers, Munnell et al. (2020) show that wages are rather more advantageous than the average for the entire population, but that health and retirement benefits are rarely mentioned. Such discrimination stems from ageism (Butler, 1969), which is rampant in society and in the labor market in particular. One of its manifestations is the unwillingness of employers to offer alternative working conditions, which are popular with older workers. Rowe and Kahn (1998), in their book Successful Ageing, identify the unyieldingness of managers in the face of requests for more flexible hours, reduced working hours, and more diversified tasks as a powerful hurdle to remaining in the workforce. Numerous studies corroborate these intuitions (Cahill et al., 2015; Earl and Taylor, 2015; Damman, 2016; Van Yperen and W¨ortler, 2017), highlighting the positive correlation between various forms of workplace flexibility and older workers’ willingness to remain employed. Such arrangements also have positive consequences for the health and wellbeing of older individuals (Vanajan et al., 2020; Piszczek and Pimputkar, 2021). Such flexibility is particularly essential in allowing a smooth transition between a full-time working life and a full retirement. The development of bridge jobs among older individuals (Ruhm, 1990) and jobs during retirement (Maestas, 2010) thus crucially depends on flexibility. Work-life balance is currently a major objective of family and gender equality policies to allow young parents to thrive. Time is of the essence to apply this objective to the entire population, particularly to employed older individuals. The COVID-19 pandemic has caused many hardships, but it has 370

The Employment of Older Workers

also been a global experiment in the vastly increasing flexibility of working conditions, notably through an unprecedented generalization of remote work. Shortly after the start of the pandemic, Chen and Munnell (2020) showed that older workers were a priori no less capable of working remotely than younger workers. Working conditions in the post-COVID-19 world will likely be more flexible and therefore more favorable to older workers.

20.4

Social Security and Public Policies

Workers in high-income countries operate in a complex institutional system that covers their unemployment and illness risks and determines their retirement income. Its main objectives, which vary from one country to another, range from a balanced budget to the fight against poverty and for intergenerational equity. The employment of older individuals is not always one of these objectives, and yet these systems produce numerous incentives that are favorable—or unfavorable—to keeping the elderly employed.

20.4.1

Retirement Systems

Public retirement systems differ tremendously from one country to the next because they are based on principles that depend on the historical context in which they were created (d’Albis, 2020). Several studies suggest that differences in retirement systems explain a significant share of the differences in the employment of older individuals among countries (Erosa et al., 2012). However, all systems in high-income countries face an ageing population, which increases the number of retirees relative to the number of working individuals. To remain financially sustainable, they must therefore either reduce pensions or increase contributions on salaries. A hybrid measure (which amounts to both reducing the sum of pensions and increasing the sum of contributions over the life cycle) is to increase the retirement age. This can be achieved by increasing the legal retirement age, which is typically the minimum age below which a pension is not paid. This can also be achieved through a system of financial incentives to prolong work, which can take the form of higher pensions for those who leave later. At the end of the 1990s, the studies of Gruber and Wise (1999) and Bl¨ondal and Scarpetta (1999) convinced the political world that retirement systems reduced older individuals’ incentive to work, and these studies, therefore, had a considerable influence because they were at the root of many reforms. The effect of the increase in the legal retirement age on the employment of older individuals may seem obvious a priori: If the minimum retirement age goes from 60 to 62, for example, the employment rate of people aged 55–64 should increase, at least for those who are already employed. It may also increase the incentives of those who are looking for work (Seater, 1977; Hairault et al., 2010). However, indirect effects may go in the opposite direction, which could be induced in particular by the behavior of employers, who may hire fewer (or lay off more) older workers. Recent literature has analyzed different pension reforms in Austria (Staubli and Zweim¨uller, 2013), Australia (Atalay and Barrett, 2015), the United Kingdom (Cribb et al., 2016), Germany (Geyer and Welteke, 2021), France (Rabat´e, 2019; Rabat´e and Rochut, 2020), etc., to identify the causal effect of postponing the minimum retirement age. All these studies find that pension reforms increase labor supply, which depends, of course, on the share of people leaving exactly at the minimum retirement age. But adverse effects are also observed, for example, on the employment of more vulnerable workers (Vigtel, 2018) or on the incentive to work before retirement (d’Albis et al., 2020). In addition, such reforms are extremely unpopular in public opinion. Other reforms have therefore focused on modifying the “normal” retirement 371

Hippolyte d’Albis

age, which is used as the reference age for the various financial conditions associated with calculating pensions. These reforms are based on the net gains from an additional year of work and, more specifically, on whether the resulting increase in the discounted sum of future pensions is greater than the cost of not receiving a pension during the additional year of work. If the gains equal the costs, the system is termed actuarially neutral. Goda et al. (2009) show that for the American system the costs are sometimes higher than the gains, leading to what they call an implicit taxation of the work of older individuals. This taxation is reinforced by the presence of income thresholds that determine the calculation of pensions (Auerbach et al., 2016). This nonneutrality is also observed outside the United States (B¨orsch-Supan and Coile, 2020). However, various studies show that the responses of the employment of older individuals to the changes in the financial pension rules are modest, except for the change in the reference age to calculate pensions (Mastrobuoni, 2009; Behaghel and Blau, 2012; Brown, 2013; Lalive et al., 2017). This is probably due to the population’s lack of financial literacy, which an abundant literature highlights. In contrast, other more recent reforms have sought to give more flexibility to older workers as to how they may transition to retirement. The idea is to remove the rigidity of a system that abruptly shifts older individuals from full-time work to a life without work. The most emblematic case are the measures implemented in 2000 in Sweden where retirement is possible from the age of 61, one’s pension is calculated according to one’s retirement age, and retirement can be combined with a professional activity. Other European countries have also introduced flexibility measures. Their evaluation provides conflicting results. Based on data from nine countries that have implemented such reforms, B¨orsch-Supan et al. (2018) estimate that increased flexibility has a positive impact on the participation rate, but a negative one on the number of hours, and that the final effect on the labor supply of older individuals may be negative. National studies such as Hernæs et al. (2020) in Norway or Albanese et al. (2020) in Belgium corroborate these conclusions. While the effect on the work of older workers remains uncertain, these measures are nonetheless likely to increase the well-being of older workers and therefore potentially their productivity at work.

20.4.2

Health Insurance

As mentioned earlier, the health of older individuals has improved overall in high-income countries, entailing an increase in life expectancy. However, considerable differences exist within the population, and health problems can still prevent some older worker from working. The health and disability insurance system hence plays an important role in the employment of older workers. Disability insurance is one of the oldest components of social welfare, which was initially introduced in the 19th century in some companies or by workers’ mutual benefit societies. Most high-income countries include disability insurance in their social security. However, as Coile et al. (2016) note, the share of recipients varies greatly from country to country. In 2009, for example, 6 percent of men received it in Sweden compared with 16 percent in the United Kingdom. Based on the various reforms of disability insurance systems, these proportions seem to be closely linked to the rules of eligibility and the generosity of the systems. For instance, in some cases, disability pensioners are allowed to work, which mechanically affects the participation of older individuals in the labor market. As Yin (2015) points out, some countries have proposed the possibility of combining reduced disability benefits with retention in the labor market; the simulations performed suggest that such a reform would improve the financial situation of the U.S. Social Security Disability Insurance (SSDI) program and increase the well-being of people with disabilities. An important issue is that disability is not necessarily fully observable, which 372

The Employment of Older Workers

creates some moral hazard. According to Maestas et al. (2013), the SSDI program induces a significant disincentive to labor supply, especially for individuals with unobservably less severe impairments. Employers may play an important role by improving the accommodation of newly disabled workers (Hill et al., 2016) or by participating in the disability insurance system as, for instance, in the Netherlands (Kalwij et al., 2016). Moreover, much of the literature highlights the fact that the effects of the disability insurance system interact with those produced by other social security programs. Kitao (2014) shows, for example, that the presence of a disability insurance system magnifies the effects of a pension reform. Health insurance more directly affects the labor supply of older individuals. In countries where medical costs are high and employers supply health coverage, the incentive to extend one’s working life is very strong as long as one is ineligible for social security. This can be explained by an aversion to the risks associated with health (Palumbo, 1999). Once the eligibility age for health coverage has been reached (65 in the case of Medicare), this incentive to work disappears. Several studies show this effect empirically. French and Jones (2011) analyze the labor supply of people whose health coverage is linked to their jobs. They estimate that raising the Medicare eligibility age by 2 years could have twice the effect of raising the normal retirement age by 2 years. However, health coverage offered unconditionally to people with low incomes (or low wealth) can reduce the incentive to work. However, studies disagree on this effect: Pashchenko and Porapakkarm (2017) estimate that 23 percent of able-bodied, Medicare-eligible individuals would work if they could keep their public insurance, while Levy et al. (2015) and Heim and Lin (2017) find that local Medicare extensions have had little impact on retirement. In European countries where the health system is universal, the role of health insurance in the labor supply of older individuals is obviously less significant. The greater issue in these countries is their important labor market regulations.

20.4.3

Labor Market Regulations

Various labor market institutions and regulations also influence the labor supply of older individuals. The main regulation concerns unemployment insurance, which provides financial protection for employees who lose their jobs. As with health insurance, understanding the possible complementarities with the retirement system is important. Hutchens (1999) thus defends the idea that an overly generous retirement system may implicitly subsidize firms by leading them to encourage their older workers to retire rather than to apply for unemployment benefits. Firms are indeed the winners because the increase in pensions is not subject to experience rating. However, the argument also applies in the other direction: An overly generous unemployment insurance system may be used as subsidized early retirement by those who are ineligible for retirement. The question crucially depends on the duration of unemployment benefits. Employees who are entitled to decent unemployment benefits for N months after being laid off may be willing to be laid off N months before their retirement. Baguelin and Remillon (2014) show that a 2003 reform in France aimed at reducing the duration of benefits for the older unemployed had a positive effect on the oldest age at which people were laid off before retirement. This type of result has been reproduced in several countries. Inderbitzin et al. (2016) examine an Austrian reform between 1988 and 1993 that consisted of increasing the duration of unemployment benefits for older individuals in certain regions. They find that the reform considerably reduced the age of permanent exit from the labor market in the regions concerned. In Finland, Kyyr¨a and Pesola (2020) show that a higher threshold age for a longer benefit period increases the employment of older individuals without affecting their health. According to Dlugosz et al. (2014), the 2006 Hartz reforms in Germany, which substantially reduced 373

Hippolyte d’Albis

the duration of benefits for the older unemployed, also reduced the layoffs of older workers. They also observe that firms used the period leading up to the implementation of the reform to significantly increase their layoffs. This confirms that labor market rules influence the behavior of firms and employees and have repercussions for the employment of older individuals. To encourage the retention of older individuals in employment, in 1987 France introduced a tax on the layoffs of employees over the age of 50, but this led to less hiring of older workers (Behaghel et al., 2008). Another form of regulation is promoting nonstandard employment, such as part-time work or self-employment. These forms of employment are booming, especially among older individuals in high-income countries. In the United States, self-employment at older age is more common among the highly educated, accounting for much of the difference in employment rates across education groups (Abraham et al., 2021). This reflects the fact that some of them wish to adjust their working hours upward or downward (Bell and Rutherford, 2013; Ramnath et al., 2021), particularly in anticipation of the decrease in both full-time employment opportunities (Dorn and Sousa-Poza, 2010) and early retirement possibilities (Casey et al., 2003). Self-employment appears more as a way out of unemployment than as a smooth transition to retirement (Been and Knoef, 2017). Been and van Vliet (2017) analyze 13 European countries between 1995 and 2008 and find a positive link between the employment rate of older men and the share of older individuals working part time. The link with self-employment is weaker. These new forms of employment do not seem to be a robust solution to the underemployment of older individuals. What is more, the increased precariousness that they imply can be problematic, as was the case when restrictions were implemented in the context of the pandemic.

References ABRAHAM, K., HERSHBEIN, B., AND HOUSEMAN, S. (2021): “Contract work at older ages,” Journal of Pension Economics and Finance, 20(3): 426–447. ACEMOGLU, D., AND RESTREPO, P. (2018): “Demographics and automation,” NBER Working Paper No. 24421. National Bureau of Economic Research, Cambridge, MA. ACEMOGLU, D., AND RESTREPO, P. (2020): “Robots and jobs: Evidence from US labor markets,” Journal of Political Economy, 128(6): 2188–2244. AHITUV, A., AND ZEIRA, J. (2011): “Technical progress and early retirement,” Economic Journal, 121(551): 171–193. ALBANESE, A., COCKX, B., AND THUY, Y. (2020): “Working time reductions at the end of the career: Do they prolong the time spent in employment?,” Empirical Economics, 59(1): 99–141. ALLEN, S. G. (2019): “Demand for older workers: What do economists think? What are firms doing?,” NBER Working Papers No. 26597. National Bureau of Economic Research, Cambridge, MA. ANGHEL, B., AND LACUESTA, A. (2020): “Ageing, productivity and employment status,” Economic Bulletin, Banco de Espa˜na, issue 1/2020: 1–20. APPS, P., AND REES, R. (2009): Public Economics and the Household, Cambridge: Cambridge University Press. ARROW, K. (1962): “The economic implications of learning by doing,” The Review of Economic Studies, 29(3): 155–173. ATALAY, K., AND BARRETT, G. (2015): “The impact of age pension eligibility age on retirement and program dependence: Evidence from an Australian experiment,” Review of Economics and Statistics, 97(1): 71–87. AUERBACH, A. J., KOTLIKOFF, L. J., KOEHLER, D., AND YU, M. (2016): “Is Uncle Sam inducing the elderly to retire?” In: Moffitt, R. A. (ed.), Tax Policy and the Economy, Volume 31. Cambridge, MA: National Bureau of Economic Research, pp. 1–42. BAGUELIN, O., AND REMILLON, D. (2014): “Unemployment insurance and management of the older workforce in a dual labor market: Evidence from France,” Labour Economics, 30(C): 245–264.

374

The Employment of Older Workers

BAKER, M. (2002): “The retirement behavior of married couples: Evidence from the spouse’s allowance,” The Journal of Human Resources, 37(1): 1–34. BARTEL, A., AND SICHERMAN, N. (1993): “Technological change and retirement decisions of older workers,” Journal of Labor Economics, 11(1): 162–183. BEEN, J., AND KNOEF, M. (2017): “Job-search requirements for unemployed at the end of working life: Effects on unemployment dynamics and self-employment probabilities,” Journal of Human Resources, 52(2): 491–530. BEEN, J., AND VAN VLIET, O. (2017): “Early retirement across Europe. Does non-standard employment increase participation of older workers?,” Kyklos, 70(2): 163–188. BEHAGHEL, L., AND BLAU, D. M. (2012): “Framing social security reform: Behavioral responses to changes in the full retirement age,” American Economic Journal: Economic Policy, 4(4): 41–67. BEHAGHEL, L., CR E´ PON, B., AND SE´ DILLOT, B. (2008): “The perverse effects of partial employment protection reform: The case of French older workers,” Journal of Public Economics, 92(3–4): 696–721. BELBASE, A., AND CHEN, A. (2019): “How did computerization since the 1980s affect older workers?,” Issue in Brief 19-19. Center for Retirement Research at Boston College, Chestnut Hill, MA. BELL, D. N. F., AND RUTHERFORD, A. C. (2013): “Older workers and working time,” The Journal of the Economics of Ageing, 1–2: 28–34. BEN-PORATH, Y. (1967): “The production of human capital and the life cycle of earnings,” Journal of Political Economy, 75(4): 352–365. BERMAN, E., BOUND, J., AND MACHIN, S. (1998): “Implications of skill-biased technological change: International evidence,” The Quarterly Journal of Economics, 113(4): 1245–1279. ¨ BL ONDAL , S., AND SCARPETTA S. (1999): “The retirement decision in OECD countries,” Economics Department Working Paper 202, Organisation for Economic Co-operation and Development, Paris. BLOOM, D. E., CANNING, D., AND MOORE, M. (2007): “A theory of retirement,” NBER Working Paper No. 13630. National Bureau of Economic Research, Cambridge, MA. BLOOM, D. E., CANNING, D., AND MOORE, M. (2014): “Optimal retirement with increasing longevity,” The Scandinavian Journal of Economics, 116(3): 838–858. BLUNDELL, R., FRENCH, E., AND TETLOW, G. (2016): “Retirement incentives and labor supply.” In: Piggott, J., and Woodland, A. (eds.), Handbook of the Economics of Population Ageing, Volume 1, North Holland, Elsevier, pp. 457–566. BOERI, T., GARIBALDI, P., AND MOEN, E. R. (2016): “A clash of generations? Increase in retirement age and labor demand for youth,” CEPR Discussion Papers 11422, London. ¨ BORSCH -SUPAN, A., BUCHER-KOENEN, T., KUTLU-KOC, V., AND GOLL, N. (2018): “Dangerous flexibility—Retirement reforms reconsidered,” Economic Policy, 33(94): 315–355. ¨ BORSCH -SUPAN, A., AND COILE, C. (2020): Social Security Programs and Retirement around the World: Reforms and Retirement Incentives. Chicago: University of Chicago Press. BOVINI, G., AND PARADISI, M. (2019): “Labor substitutability and the impact of raising the retirement age.” Available at https://scholar.harvard.edu/files/paradisi/files/paradisi jmp.pdf. BRATSBERG, B., AND STANCANELLI, E. (2018): Partial Retirement and Spousal Labor Supply: Learning from a Norwegian Retirement Reform. Paris: Mimeo. BROWN, K. M. (2013): “The link between pensions and retirement timing: Lessons from California teachers,” Journal of Public Economics, 98: 1–14. BRYNJOLFSSON, E., AND MCAFEE, A. (2014): The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies. New York: Norton. BURLON, L., AND VILALTA-BUF´I, M. (2016): “A new look at technical progress and early retirement,” IZA Journal of Labor Policy, 5: 5. BUTLER, R. N. (1969): “Age-ism: Another form of bigotry,” The Gerontologist, 9(4): 243–246. CAHILL, K. E., JAMES, J. B., AND PITT-CATSOUPHES, M. (2015): “The impact of a randomly assigned time and place management initiative on work and retirement expectations,” Work, Ageing and Retirement, 1(4): 350–368. CAPATINA, E. (2015): “Life-cycle effects of health risk,” Journal of Monetary Economics, 74(C): 67–88. CARR, E., FLEISCHMANN, M., GOLDBERG, M., KUH, D., MURRAY, E. T., STAFFORD, M., STANSFELD, S., VAHTERA , J., X UE , B., Z ANINOTTO, P., Z INS, M., AND H EAD, J. (2018): “Occupational and educational inequalities in exit from employment at older ages: Evidence from seven prospective cohorts,” Occupational and Environmental Medicine, 75(5): 369–377. CASANOVA, M. (2010): Happy Together: A Structural Model of Couples’ Joint Retirement Choices. Mimeo, UCLA, Los Angeles, CA.

375

Hippolyte d’Albis

CASEY, B., OXLEY, H., WHITEHOUSE, E., ANTOLIN, P., DUVAL, R., AND LEIBFRITZ, W. (2003): “Policies for an ageing society: Recent measures and areas for further reform,” OECD Economics Department Working Paper, No. 369, Paris. CERVELLATI, M., AND SUNDE, U. (2013): “Life expectancy, schooling, and lifetime labor supply: Theory and evidence revisited,” Econometrica, 81(5): 2055–2086. CHANG, F.-R. (1991): “Uncertain lifetimes, retirement and economic welfare,” Economica, 58(230): 215–232. CHEN, A., AND MUNNELL, A. H. (2020): “Can older workers work from home?,” Issues in Brief 2020-9. Center for Retirement Research at Boston College, Chestnut Hill, MA. CHIAPPORI, P.-A. (1988): “Rational household labor supply,” Econometrica, 56(1): 63–90. COILE, C. (2004): “Retirement incentives and couples’ retirement decisions,” The B.E. Journal of Economic Analysis & Policy, 4(1): 1–30. COILE, C. (2018): “Working longer in the US: Trends and explanations,” NBER Working Paper No. 24576. National Bureau of Economic Research, Cambridge, MA. COILE, C., MILLIGAN, K. S., AND WISE, D. A. (2016): “Introduction and summary.” In: Wise, D. A. (ed.), Social Security Programs and Retirement around the World; Disability Insurance Programs and Retirement, Conference report, 1–14. Cambridge, MA: National Bureau of Economic Research. COILE, C., MILLIGAN, K. S., AND WISE, D. A. (2017): “Introduction and summary.” In: Wise, D. A. (ed.), Social Security Programs and Retirement around the World; The Capacity to Work at Older Ages, Conference report, 1–13. Cambridge, MA: National Bureau of Economic Research. COOK, C., DIAMOND, R., AND OYER, P. (2019): “Older workers and the gig economy,” AEA Papers and Proceedings, 109: 372–376. CRIBB, J., EMMERSON, C., AND TETLOW, G. (2016): “Signals matter? Large retirement responses to limited financial incentives,” Labour Economics, 42: 203–212. CURRIE, J., AND SCHWANDT, H. (2016): “Inequality in mortality between rich and poor U.S. counties decreased among the young while increasing for older adults, 1990–2010,” Science, 352(6286): 708–712. D’ALBIS, H. (2020): “Social security: History and operations.” In: Gu D., and Dupre, M. (eds.), Encyclopedia of Gerontology and Population Ageing, Cham: Springer, pp. 1–5. D’ALBIS, H., DOORLEY, K., AND STANCANELLI, E. (2021): “Older mothers’ employment and marriage stability when the nest is empty,” PSE working paper N° 2021–24, Paris. D’ALBIS, H., FOUG E` RE, D., AND GOU E¨ DARD, P. (2020): “Slow down before you stop: The effect of the 2010 French pension reform on older teachers’ sick leaves,” CEPR Discussion Paper 15142, London. ´ D’ALBIS, H., LAU, S. P., AND SANCHEZ -ROMERO, M. (2012): “Mortality transition and differential incentives for early retirement,” Journal of Economic Theory, 147(1): 261–283. DAMMAN, M. (2016): “Blended work and employment participation of older workers: A further discussion,” Work, Ageing and Retirement, 2(4): 384–389. DAVID, B. (2017): “Computer technology and probable job destructions in Japan: An evaluation,” Journal of the Japanese and International Economies, 43: 77–87. DEL BOCA, D., PASQUA, S., AND PRONZATO, C. (2009): “Motherhood and market work decisions in institutional context: A European perspective,” Oxford Economic Papers, 61(1): i147–i171. DEMING, D. J., AND NORAY, K. L. (2018): “STEM careers and the changing skill requirements of work,” NBER Working Paper No. 25065. National Bureau of Economic Research, Cambridge, MA. DLUGOSZ, S., STEPHAN, G., AND WILKE, R. A. (2014): “Fixing the leak: Unemployment incidence before and after a major reform of unemployment benefits in Germany,” German Economic Review, 15(3): 329–352. DORN, D., AND SOUSA-POZA, A. (2010): “ ‘Voluntary’ and ‘involuntary’ early retirement: An international analysis,” Applied Economics, 42(4): 427–438. EARL, C., AND TAYLOR, P. (2015): “Is workplace flexibility good policy? Evaluating the efficacy of age management strategies for older women workers,” Work, Ageing and Retirement, 1(2): 214–226. EROSA, A., FUSTER, L., AND KAMBOUROV, G. (2012): “Labor supply and government programs: A cross-country analysis,” Journal of Monetary Economics, 59(1): 84–107. FARBER, H. S., SILVERMAN, D., AND VON WACHTER, T. (2015): “Factors determining callbacks to job applications by the unemployed: An audit study,” NBER Working Paper No. 21689 National Bureau of Economic Research, Cambridge, MA. FOUARGE, D., AND SCHILS, T. (2009): “The effect of early retirement incentives on the training participation of older workers,” Labor, 23(s1): 85–109.

376

The Employment of Older Workers

FRENCH, E. (2005): “The effects of health, wealth, and wages on labour supply and retirement behaviour,” The Review of Economic Studies, 72(2): 395–427. FRENCH, E., AND JONES, J. B. (2011): “The effects of health insurance and self-insurance on retirement behavior,” Econometrica, 79(3): 693–732. FRENCH, E., AND JONES, J. B. (2017): “Health, health insurance, and retirement: A survey,” Annual Review of Economics, 9(1): 383–409. FREY, C. B., AND OSBORN, M. A. (2017): “The future of employment: How susceptible are jobs to computerisation?,” Technological Forecasting & Social Change, 114: 254–280. FRIEDBERG, L. (2003): “The impact of technological change on older workers: Evidence from data on computer use,” ILR Review, 56(3): 511–529. GERSHUNY, J. (2000): Changing Times: Work and Leisure in Postindustrial Society, Oxford: Oxford University Press. GEYER, J., AND WELTEKE, C. (2021): “Closing routes to retirement for women: How do they respond?,” Journal of Human Resources, 56(1): 311–341. GODA, G. S., SHOVEN, J. B., AND SLAVOV, S. N. (2009): “Removing the disincentives in social security for long careers.” In: Brown, J. R., Liebman, J. B., and Wise, D. A. (eds.), Social Security Policy in a Changing Environment, Cambridge, MA: National Bureau of Economic Research, pp. 21–38. GOLDIN, C. (2006): “The quiet revolution that transformed women’s employment, education, and family,” American Economic Review, 96(2): 1–21. GOLDIN, C., AND KATZ, L. F. (2018): “Women working longer: Facts and some explanations.” In: Women Working Longer: Increased Employment at Older Ages, Chicago: University of Chicago Press, pp. 11–54. GRUBER, J., AND MILLIGAN, K. (2010): “Do elderly workers substitute for younger workers in the United States?” In: Gruber, J., Wise, D. A., Goldin, C., and Katz, L. F. (eds.), Social Security Programs and Retirement around the World: The Relationship to Youth Employment, Cambridge, MA: National Bureau of Economic Research, pp. 345–360. GRUBER, J., AND WISE, D. A. (1999): Social Security and Retirement Around the World, Chicago: University of Chicago Press. HAEGELAND, T., RONNINGEN, D., AND SALVANES, K. (2007): “Adapt or withdraw? Evidence on technological changes and early retirement using matched worker-firm data,” Research Department Discussion Paper n°509, Statistics Norway, Chicago. HAIRAULT, J.-O., LANGOT, F., AND SOPRASEUTH, T. (2010): “Distance to retirement and older workers’ employment: The case for delaying the retirement age,” Journal of the European Economic Association, 8(5): 1034–1076. HAZAN, M. (2009): “Longevity and lifetime labor supply: Evidence and implications,” Econometrica, 77(6): 1829–1863. HEIM, B. T., AND LIN, L. (2017): “Does health reform lead to an increase in early retirement? Evidence from Massachusetts,” ILR Review, 70(3): 704–732. HERNÆS, E., JIA, Z., PIGGOTT, J., AND VIGTEL, T. C. (2020): “Work less but stay longer. Mature worker response to a flexibility reform,” Discussion Papers 937. Research Department, Statistics Norway. HILL, M. J., MAESTAS, N., AND MULLEN, K. J. (2016): “Employer accommodation and labor supply of disabled workers,” Labour Economics, 41: 291–303. HUDOMIET, P., AND WILLIS, R. J. (2021): “Computerization, obsolescence, and the length of working life,” NBER Working Paper No. 28701. National Bureau of Economic Research, Cambridge, MA. HUTCHENS, R. (1999): “Social security benefits and employer behavior: Evaluating social security early retirement benefits as a form of unemployment insurance,” International Economic Review, 40(3): 659–678. ¨ INDERBITZIN, L., STAUBLI, S., AND ZWEIM ULLER , J. (2016): “Extended unemployment benefits and early retirement: Program complementarity and program substitution,” American Economic Journal: Economic Policy, 8(1): 253–288. JOHNSON, R. W., AND LO SASSO, A. T. (2006): “The impact of elder care on women’s labor supply,” Inquiry, 43(3): 195–210. KALEMLI-OZCAN, S., AND WEIL, D. (2010): “Mortality change, the uncertainty effect, and retirement,” Journal of Economic Growth, 15: 65–91. KALWIJ, A., KAPTEYN, A., AND DE VOS, K. (2010): “Retirement of older workers and employment of the young,” De Economist, 158: 341–359. KALWIJ, A., KAPTEYN, A., AND DE VOS, K. (2016): “Health, disability insurance, and labor force exit of older workers in the Netherlands.” In: Wise, D. A. (ed.), Social Security Programs and Retirement Around the World: Disability Insurance Programs and Retirement, Chicago: University of Chicago Press, pp. 211–223.

377

Hippolyte d’Albis

KATZ, L. F., AND KRUEGER, A. B. (2019): “The rise and nature of alternative work arrangements in the United States, 1995–2015,” ILR Review, 72(2): 382–416. KITAO, S. (2014): “A life-cycle model of unemployment and disability insurance,” Journal of Monetary Economics, 68: 1–18. KYYR A¨ , T., AND PESOLA, H. (2020): “Long-term effects of extended unemployment benefits for older workers.,” Labour Economics, 62(C): 101777. LAHEY, J. N. (2008): “Age, women, and hiring: An experimental study,” Journal of Human Resources, 43(1): 30–56. LALIVE, R., MAGESAN, A., AND STAUBLI, S. (2017): “Raising the full retirement age: Defaults vs incentives,” NBER Retirement Research Center Paper No. NB 17-12. National Bureau of Economic Research, Cambridge, MA. LAU, M. I., AND POUTVAARA, P. (2006): “Social security incentives and human capital investment,” Finnish Economic Papers, 19(1): 16–24. LEVY, H., BUCHMUELLER, T. C., AND NIKPAY, S. (2015): “The effect of health reform on retirement,” Michigan Retirement Research Center Research Paper No. 2015-329, Ann Arbor, MI. ¨ ¨ , P., HESS, M., AND ELLWARDT, L. (2021): “Educational inequalities in labor market MACKEN , J., PR AG exit of older workers in 15 European countries,” Journal of Social Policy, 51(2), 435–459. MAESTAS, N. (2010): “Back to work: Expectations and realizations of work after retirement,” Journal of Human Resources, 45(3): 718–748. MAESTAS, N., MULLEN, K. J., AND STRAND, A. (2013): “Does disability insurance receipt discourage work? Using examiner assignment to estimate causal effects of SSDI receipt,” American Economic Review, 103(5): 1797–1829. MANTON, K. G., GU, X. L., AND LAMB, V. L. (2006): “Long-term trends in life expectancy and active life expectancy in the United States,” Population and Development Review, 32(1): 81–105. MARTIN, J. P. (2018): “Live longer, work longer: The changing nature of the labour market for older workers in OECD countries,” ZA DP No. 11510, Bonn. MASTROBUONI, G. (2009): “Labor supply effects of the recent social security benefit cuts: Empirical estimates using cohort discontinuities,” Journal of Public Economics, 93(11–12): 1224–1233. MCCLELLAN, M. B. (1998): “Health events, health insurance, and labor supply: Evidence from the health and retirement survey.” In: Wise D. A. (ed.), Frontiers in the Economics of Ageing, Chicago: University of Chicago Press, pp. 301–350. MCDANIEL, M., PESTA, B., AND BANKS, G. (2012): “Job performance and the ageing worker.” In: Hedge, J. W., and Borman, W. C. (eds.), Oxford Handbook of Work and Ageing, New York: Oxford University Press, pp. 281–297. MESSE, P.-J., MORENO-GALBIS, E., AND WOLFF, F.-C. (2014): “Retirement intentions in the presence of technological change: Theory and evidence from France,” IZA Journal of Labor Economics, 3(8): 1–28. MICHAUD, P.-C., AND VERMEULEN, F. (2011): “A collective labor supply model with complementarities in leisure: Identification and estimation by means of panel data,” Labour Economics, 18(2): 159–167. MILLIGAN, K., AND WISE, D. A. (2015): “Health and work at older ages: Using mortality to assess the capacity to work across countries,” Journal of Population Ageing, 8(1–2): 27–50. MOHNEN, P. (2019): “The impact of the retirement slowdown on the U.S. youth labor market,” Available at https://conference.nber.org/conf papers/f131435.pdf. MUNNELL, A. H., WETTSTEIN, G., AND WALTERS, A. N. (2020): “What jobs do employers want older workers to do?,” Working Papers 202011. Center for Retirement Research at Boston College, Chestnut Hill, MA. MUNNELL, A. H., AND WU, A. (2012): “Will delayed retirement by baby boomers lead to higher unemployment among younger workers?,” Center for Retirement Research at Boston College, Chestnut Hill, MA. MUNNELL, A. H., AND WU, A. (2020): “Do older workers squeeze out younger workers?,” SIEPR Discussion Paper No. 13-011, Stanford. NATIONAL ACADEMIES OF SCIENCES, ENGINEERING, AND MEDICINE, AND COMMITTEE ON POPULATION. (2015): “The growing gap in life expectancy by income: Implications for federal programs and policy responses,” National Academies Press, Washington. NEUMANN, S., AND WEISS, A. (1995): “On the effects of schooling vintage on experience-earnings profiles: Theory and evidence,” European Economic Review, 39(5): 943–955. NEUMARK, D., BURN, I., AND BUTTON, P. (2019): “Is it harder for older workers to find jobs? New and improved evidence from a field experiment,” Journal of Political Economy, 127(2): 922–970.

378

The Employment of Older Workers

OECD (ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT). (2016): “Skills matter: Further results from the survey of adult skills,” OECD Skills Studies. OECD Publishing, Paris. OECD (ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT). (2021): Population with tertiary education (indicator). DOI: 10.1787/0b8f90e9-en. OYER, P. (2016): “The independent workforce in America,” Upwork White Paper. PAILH E´ , A., SOLAZ, A., AND STANFORS, M. (2021): “The great convergence: Gender and unpaid work in Europe and the United States,” Population and Development Review, 47(1): 181–217. PALUMBO, M. G. (1999): “Uncertain medical expenses and precautionary saving near the end of the life cycle.,” Review of Economic Studies, 66(2): 395–421. PASHCHENKO, S., AND PORAPAKKARM, P. (2017): “Work incentives of Medicaid beneficiaries and the role of asset testing,” International Economic Review, 58(4): 1117–1154. PISZCZEK, M. M., AND PIMPUTKAR, A. S. (2021): “Flexible schedules across working lives: Age-specific effects on well-being and work,” Journal of Applied Psychology, 106(12): 1907–1920. POZZOLI, D., AND RANZANI, M. (2009): “Old European couples’ retirement decisions: The role of love and money,” Working Papers 09-2, Aarhus School of Business, Department of Economics, University of Aarhus. RABAT E´ , S. (2019): “Can I stay or should I go? Mandatory retirement and labor force participation of older workers,” Journal of Public Economics, 180: 104078. RABAT E´ , S., AND ROCHUT, J. (2020): “Employment and substitution effects of raising statutory retirement age in France.,” Journal of Pension Economics and Finance, 19: 293–308. RAMNATH, S., SHOVEN, J., AND SLAVOV, S. (2021): “Pathways to retirement through self-employment,” Journal of Pension Economics and Finance, 20(2): 232–251. ROBINE, J.-M., JAGGER, C., MATHERS, C. D., CRIMMINS, E. M., AND SUZMAN, R. M. (2003): Determining Health Expectancies, West Sussex: John Wiley & Sons. ROMEU GORDO, L., AND SKIRBEKK, V. (2013): “Skill demand and the comparative advantage of age: Job tasks and earnings from the 1980s to the 2000s in Germany,” Labour Economics, 22(C): 61–69. ROWE, J. W., AND KAHN, R. L. (1998): Successful Ageing, New York: Pantheon Books. RUHM, C. J. (1990): “Bridge jobs and partial retirement,” Journal of Labor Economics, 8(4): 482–501. RUTLEDGE, M. S., SASS, S. A., AND RAMOS-MERCADO, J. D. (2017): “How does occupational access for older workers differ by education?,” Journal of Labor Research, 38(3): 283–305. SALOMON, J. A., WANG, H., FREEMAN, M. K., VOS, T., FLAXMAN, A. D., LOPEZ, A. D., AND MURRAY, C. J. L. (2012): “Healthy life expectancy for 187 countries, 1990–2010: A systematic analysis for the Global Burden Disease Study 2010,” The Lancet, 380(9859): 2144–2162. ´ SANCHEZ -ROMERO, M., D’ALBIS, H., AND PRSKAWETZ, A. (2016): “Education, lifetime labor supply, and longevity improvements,” Journal of Economic Dynamics and Control, 73: 118–141. SCHIRLE, T. (2008): “Why have the labor force participation rates of older men increased since the mid-1990s?,” Journal of Labor Economics, 26(4): 549–594. SCHWAB, K. (2017): The Fourth Industrial Revolution, New York: Crown Publishing Group. SEATER, J. (1977): “A unified model of consumption, labor supply, and job search,” Journal of Economic Theory, 14(2): 349–372. ¨ STAUBLI, S., AND ZWEIM ULLER , J. (2013): “Does raising the early retirement age increase employment of older workers?,” Journal of Public Economics, 108: 17–32. UNITED NATIONS. (2019): World population prospects 2019, custom data acquired via website. Department of Economic and Social Affairs, Population Division, New York. ¨ VANAJAN, A., BULTMANN , U., AND HENKENS, K. (2020): “Health-related work limitations among older workers-the role of flexible work arrangements and organizational climate,” The Gerontologist, 60(3): 450–459. ¨ VAN YPEREN, N. W., AND WORTLER , B. (2017): “Blended working and the employability of older workers, retirement timing, and bridge employment,” Work, Ageing and Retirement, 3(1): 102–108. VESTAD, O. L. (2013): Early Retirement and Youth Employment in Norway, Oslo: Statistic Norway. VIGTEL, T. C. (2018): “The retirement age and the hiring of senior workers,” Labour Economics, 51(C): 247–270. WAHRENDORF, M., DRAGANO, N., AND SIEGRIST, J. (2013): “Social position, work stress, and retirement intentions. A study with older employees from 11 European countries,” European Sociological Review, 29(4): 792–802. WALDRON, H. (2007): “Trends in mortality differentials and life expectancy for male social security-covered workers, by socioeconomic status,” Social Security Bulletin, 67(3), 1–28.

379

Hippolyte d’Albis

WISE, D. A. (2017): “Social Security Programs and Retirement around the World; The Capacity to Work at Older Ages,” National Bureau of Economic Research Conference Report. Cambridge, MA: National Bureau of Economic Research. YIN, N. (2015): “Partial benefits in the Social Security Disability Insurance Program,” Journal of Risk and Insurance, 82(2): 463–504. ZAMARRO, G. (2020): “Family labor participation and child care decisions: The role of grannies,” SERIEs, 11: 287–312. ZHIVAN, N. A., SOTO, M., SASS, S. A., AND MUNNELL, A. H. (2012): “How the risk of displacement for older workers has changed,” Labor, 26(1): 90–107.

380

21 RETIREMENT AND HEALTH Jan C. van Ours

Abstract Retiring is a labor market transition that affects the personal life of individuals. Retiring may have short-term or long-term effects on mental health, physical health, or even mortality. Empirical studies show health outcomes varying from negative to positive with many studies finding no health effects at all. This range of outcomes is partly related to heterogeneity in terms of personal characteristics, type of job, institutional arrangements, and whether retiring was voluntary or mandatory. However, part of the variation in outcomes also relates to differences in how health is measured and the methods used to establish causality from retirement to health. The variation in outcomes makes it hard to advocate evidence-based retirement policies that account for health effects. In the policy debate on the increase of the statutory retirement age or the age of eligibility to early retirement, health effects cannot be an unimportant ingredient. Introducing more flexibility in the timing of retirement may be the only policy measure that is unambiguously beneficial for the health of all workers.

21.1

Introduction

There are two types of retirement. First is statutory retirement, i.e., retiring at the standard retirement age. Second is early retirement, i.e., retiring earlier than the statutory retirement age.1 Retiring usually means leaving the labor market to never return to work. Nevertheless, retirement is not necessarily an absorbing state. Workers may partially retire, i.e., they start collecting pension benefits but keep on working part time. Workers could also retire from the job they had for a long time but start a part-time (or full-time) job at another firm or become a self-employed worker who uses the retirement benefits as a basic income.2 Three types of health effects of retirement can be distinguished: two short-term effects, i.e., on mental health and physical health, and one long-run effect, i.e., on mortality. The effects on mental health and physical health may not be instantaneous, i.e., occurring shortly after retirement, but may gradually evolve and therefore duration in retirement needs to be considered. Retirement is often an expected transition from work to full-time leisure. If one stays employed until the standard retirement age, the time of the transition is known well in advance. The worker often does not have a choice about whether to make this transition. It is forced upon the worker, like it or not. Mandatory retirement may have negative health consequences because workers are forced to do something that they dislike. Thus, retirement may come with DOI: 10.4324/9781003150398-25

381

Jan C. van Ours

depression and anxiety. Nevertheless, while for some workers retirement is an unwanted transition, other workers look forward to using the option to retire early. This transition is not forced but voluntary, and these workers use it as soon as formal regulations allow. Because the transition is voluntary, positive health effects may be expected in terms of mental health. Whereas the mental health effects of retirement may depend on the voluntary or involuntary nature of the labor market transition, physical health may be affected similarly if, for example, the retiree spends more time on physical exercise than in pre-retirement. The exact health consequences of retirement are an empirical matter. This holds for immediate health effects but also for long-term health consequences, including mortality. The health effects of retirement may be interesting from a policy point of view because average retirement ages are going up due to institutional increases in standard retirement ages and disappearing early retirement programs (Boeri and Van Ours, 2021). Health is a multidimensional phenomenon that can be measured by objective characteristics, such a body temperature, blood pressure, physical, or mental fitness tests, but it can also be measured through self-assessment. The mental health of an individual can be established in various ways. It can be done through self-reporting, i.e., by answering several questions that can be combined into a summary measure like the Mental Health Inventory-5. However, it can also be single questions on mood, depression, and anxiety that may be used as indicator.3 For example, tests on speed of processing words, word reading, and recall can measure cognitive ability. Cognition is considered a mental health indicator. The theory of “mental retirement” or “use it or lose it” theory suggests that working is more mentally stimulating than retirement. For example, cognitive decline may accelerate after retirement because of fewer opportunities to communicate or collaborate. After retirement workers may also lose the need for self-discipline. Rohwedder and Willis (2010) refer to this as the “unengaged lifestyle” explanation for the cognition effects of retirement, also mentioning a mental retirement effect related to human capital theory. Because workers have little incentive to invest in their human capital toward the end of their career the reduction in mental exercise may start before actual retirement, i.e., there is an “on-the-job” mental retirement effect. To establish the causal effect of retirement on mental health, one must consider that joint unobserved characteristics and reverse causality may also cause the link between the two variables. An example of an unobservable characteristic is the preference for leisure time that may affect both mental health and the desire to retire. Reverse causality may play a role if the mental or physical condition of a worker affects the retirement decision of that worker. To deal with potential reverse causality, two methods are generally used, instrumental variables (IV) and a regression discontinuity design (RDD). Eligibility ages for early retirement or retirement-related social security benefits are popular instrumental variables. RDD analysis typically exploits the sudden increase in the retirement probability as soon as an individual attains the age for pension eligibility. Sometimes an increase in statutory retirement age is exploited to establish causality using a difference-in-differences approach.4 Panel datasets, some of which facilitate multi-country studies because of the harmonization of questionnaires, support the identification of treatment effects of retirement on health outcomes. Frequently used datasets are the U.S. Health and Retirement Study (HRS); the English Longitudinal Study of Ageing (ELSA); and the Survey of Health, Ageing and Retirement in Europe (SHARE). The European dataset has been extended to other countries across the world including Brazil, China, India, and Japan. These panel datasets allow researchers to account for association through joint time-invariant unobserved characteristics by introducing individual fixed effects. Finally, in terms of the nature of the data a distinction exists between studies focusing on data from one country and studies analyzing cross-country data. 382

Retirement and Health

The latter type of studies use data that are collected in a harmonized way in different countries. This chapter focuses on the causal effect of retirement on health and not the other way around.5 This implies that studies that do not go beyond the association between retirement and health are not discussed as they do not reveal which way causality goes and do not exclude the possibility that causality runs both ways. This chapter is organized as follows. A crosscountry comparison of retirement statistics including healthy years in retirement sets the stage. The comparison is limited to the extent that it only presents information from Organisation for Economic Co-operation and Development (OECD) countries because comparable statistics are often collected in a harmonized way for these countries. Nevertheless, within the OECD labor market institutions—including retirement programs—and labor market outcomes vary substantially. The overview of previous studies is also OECD-based and focuses on individual health effects for the retiree. First, studies on mostly short-term mental and physical health effects of retiring are discussed. After that, previous work on long-term effects of retirement, i.e., mortality, is discussed. A brief presentation of studies that consider cross-partner effects of retirement follows. Studies differ with respect to the identification strategy, IV, and RDD. The studies are discussed chronologically according to the publication year and, where relevant, IV studies are discussed separately from RDD studies. There are many overviews of work on retirement and health if only because every empirical study provides a description of results from previous studies. In addition there are overview studies that are themselves the contribution to the research literature. Here, the overview studies are presented separately from studies that contribute to the empirical literature on retirement and health. All studies included are published from 2010 onward. This chapter concludes with a summary of the main findings, thoughts about future research, and potential policy implications.

21.2

Setting the Stage—Cross-Country Comparison of Retirement-Related Statistics

Figure 21.1 shows international developments in life expectancy at age 65, effective retirement age, and life expectancy after retirement over the period 1970–2018. The figure shows averages for 11 OECD countries separately for females and males.6 The developments for males and females are very similar, though at different levels. Panel a shows that life expectancy at age 65 is about 3 years longer for females than for males. Life expectancy increased steadily with almost 10 years for females and more than 11 years for males. Panel b shows that the effective retirement age is always 1–2 years higher for males.7 The effective retirement age goes down until the early 21st century to increase from then on. Over the whole period of 50 years, the effective retirement age dropped about 3.5 years for females and about 3 years for males. Panel c shows approximate life expectancy after retirement.8 The difference between life expectancy at age 65 and effective retirement age increased up to the early 21st century from 10 to almost 20 years for males and from 14 to about 24 years for females. After that, the difference is approximately constant. Clearly, over the past decades the life perspective of retirement has changed a lot. On the one hand, the retirement period has expanded by about a decade for both females and males. On the other hand, despite the increase in the effective retirement age in the 21st century the remaining life expectancy after retirement is constant. 383

Jan C. van Ours

Figure 21.1 Life expectancy at age 65, effective retirement age and life expectancy after retirement, females and males; average over 11 OECD countries 1970–2018. Note: Average of 11 countries: Australia, Austria, Belgium, France, Japan, Netherlands, Norway, Portugal, Sweden, Switzerland, United States. Source: OECD statistics. Life expectancy at age 65 is the average number of years that a person of that age can be expected to live, assuming that age-specific mortality levels remain constant. Life expectancy after retirement is calculated as (life expectancy after age 65 + 65 – effective retirement age).

Cross-country differences in the retirement perspectives are substantial. Table 21.1 provides information about life expectancy at age 65 for some countries, distinguishing between healthy and non-healthy years.9 The bottom line of the table shows that on average for both women and men from age 65 onward about 9.5 years are healthy years. Women have about 12 additional non-healthy years, while men have about 8.5 additional non-healthy years beyond age 65. These numbers exhibit substantial cross-country differences. Life expectancy at age 65 ranges for women from 18.5 in Hungary to 24.2 in Japan and for men from 14.1 in Latvia to 20.2 in Switzerland. The range in life expectancies at age 65 of about 5–6 years is even larger when it comes to expected healthy life years with a range of 10–11 years: From less than 5 years in Latvia and Slovakia to more than 15 years in New Zealand and Sweden and for men in Iceland. Somewhat surprising are the differences in unhealthy years between men and women because the expected healthy years are not very gender specific. The last two columns of Table 21.1 show substantial differences in life expectancy by educational attainment.10 Highly educated individuals live substantially longer than less educated individuals. There is an educational gap in life expectancy of more than 10 years for men in many Eastern European countries. Because life expectancy in healthy years in these countries is on the low end of the distribution retiring in these countries is something completely different from retiring in Nordic countries with many healthy years after age 65. 384

Retirement and Health Table 21.1 Life expectancy at age 65 (2018) and educational gap in life expectancy at age 30 (2015) Country

Women life expectancy at 65 Total Healthy Other

Men life expectancy at 65 Total Healthy Other

Education Gap Women Men

Australia Austria Belgium Canada Chile Czech Republic Denmark Estonia Finland France Germany Greece Hungary Iceland Ireland Israel Italy Japan Korea, Republic Latvia Lithuania Luxembourg Mexico Netherlands New Zealand Norway Poland Portugal Slovak Republic Slovenia Spain Sweden Switzerland Turkey United Kingdom United States

22.6 21.6 21.9 22.1 21.7 19.8

19.9 18.5 18.6 19.4 18.3 16.2

Average

21.4

12.0 10.8

6.5 7.8

3.5 3.0 6.7 2.6

6.3 6.2 9.5 4.0

8.1

8.1

2.7

10.6

18.0 15.7 18.6 19.7 18.0 19.1 14.6 19.5 19.1 19.7 19.6 19.6 18.7

10.8 5.6 7.8 10.2 11.5 7.4 6.9 15.2 7.5

7.2 10.1 10.8 9.5 6.5 11.7 7.7 4.3 11.6

4.0 8.7 3.6 2.6

5.5 12.3 5.3 6.5

2.4 6.4

5.7 11.5

9.8

9.8

5.0 2.7

7.4 4.4

4.2 5.6 9.2

9.9 8.9 9.6

8.0

11.0

11.6 6.1

14.1 14.5 18.8 16.8 18.7 19.4

9.9 15.3

8.8 4.1

4.6 4.8

5.2 4.6

8.8 9.4 4.6

11.3 12.6 14.7

19.4 15.8 18.2 15.4

8.2 9.5 4.0

7.6 8.7 11.4

3.2 5.1 2.8 6.3

4.8 11.8 5.1 14.4

7.4 11.3 15.8 10.2

14.4 12.2 5.8 12.8

7.5 11.5 15.6 10.6

10.3 8.0 3.6 9.6

2.7

6.4

2.6

4.0

10.7

10.4

10.2

8.7

2.0 4.0

3.2 4.4

3.8

6.9

4.2

7.1

13.8 11.4

7.8 10.5

8.5

11.3

20.7 20.6 22.0 23.8 21.1 21.9 18.5 21.5 21.6 22.1 22.8 24.2 22.8

11.8 5.8 6.9 11.3 12.2 7.2 7.4 14.6 7.4

8.9 14.8 15.1 12.5 8.9 14.7 11.1 6.8 14.2

9.2

13.6

19.0 19.7 22.2 18.7 21.1 21.7

4.7 6.3 8.4

14.3 13.4 13.7

9.5 15.6

21.7 20.1 22.0 19.3 21.8 23.5 21.6 23.0 19.4 21.1 20.7

17.8 19.5 19.2 20.2 16.2 18.9 18.1

9.6

11.7

18.1

9.4

8.5

Note: Life expectancy at age 65 is the average number of years that a person of that age can expect to live, assuming age-specific mortality levels remain constant, 2018 (some countries 2017). Healthy life years are defined as the number of years spent free of long-term activity limitation (this is equivalent to disability-free life expectancy). Healthy life years are calculated annually by Eurostat based on life table data and age-specific prevalence data on long-term activity limitations. Educational gap: the gap in the expected years of life remaining at age 30 between adults with the highest level (“tertiary education”) and the lowest level (“below upper secondary education”) of education, mostly 2015, some countries earlier years. Source: OECD Statistics. 385

Jan C. van Ours

Table 21.1 presents cross-country differences in retirement-related characteristics. The first three columns show employment rates for the age groups 55–59, 60–64, and 65–69. The differences are substantial for the 60+ categories. In the age range 55–59, employment rates are above 80 percent in many countries, in the age category 60–64, the highest employment rate is 78 percent (in Iceland), but the second highest is below 70 percent. In the 65–69 category only in Iceland is the employment rate above 50 percent. Average employment rates across the OECD countries are 72.5 percent in the age category 55–59, 49.6 percent in the age category 60–64, and 22.3 percent in the age category 65–69. The normal retirement age for workers with an uninterrupted working career from age 22 for most countries and for both males and females is age 65. For Iceland, Ireland, Israel, and Norway, it is age 67. In some countries institutional gender-based discrimination favors women regarding the normal retirement age. In Austria and Israel the normal retirement age for women is 5 years earlier than for men. In Poland, Turkey, and the United Kingdom, the differences between men and women are smaller but still more than 2 years. In many countries but not all the effective labor market exit age is close to the normal retirement age. The cross-country distribution in effective labor market exit ages is not that wide, although exceptional cases exist, such as France with 60.8 years and the Republic of Korea with 72.3 years as the effective labor market age. The gender differences within countries are often limited. For men, the average effective labor market exit age is 65.4, for women this is 63.7. Expected years in retirement vary from a Korean low of 12.9 years (men) and 16.3 years (women) to a French high of 22.7 years (men) and 26.9 years (women). Considering that many of the expected years in retirement are not in good health, the perspective of a retiree in the Republic of Korea is very different from the perspective of a retiree in France. The last two columns of Table 21.1 show the healthy years in retirement.11 These confirm the large cross-country variation. On average, people spend 10–12 healthy years in retirement, but the range is from a low of 3–5 years in Latvia to a high of 14–15 years in Belgium, France, Spain, and Sweden.

21.3

How Retirement Affects Mental Health and Physical Health

There are three types of studies on the effects of retirement on mental health and physical health; cross-country studies, single-country studies, and overview studies.

21.3.1

Cross-Country Studies

Comparable data on retirement and health collected through HRS, ELSA, and SHARE facilitate cross-country studies. Which countries are included varies from study to study and is not important for this overview. What matters is that cross-country differences in retirement institutions can be exploited to identify causal effects even though cross-country heterogeneity often exists in the relevant relationships. All studies presented and discussed in this section used an IV approach for identification. Rohwedder and Willis (2010) find that early retirement had negative effects on cognitive skills of workers in their early 60s. Coe and Zamarro (2011) find no significant retirement effects on depression and cognitive functioning for men. However, they do find a significant positive retirement effect on overall health. Mazzonna and Peracchi (2012) conclude that retirement not only has a one-time negative effect on the cognition of workers, but retirement also increases the cognitive decline. Belloni et al. (2016) conclude that retirement has a positive effect on mental health of men while the mental health of women is unaffected. The positive mental health effect is stronger for blue-collar men in areas that were strongly hit by the Great Recession. Hessel (2016) studies the effect of retirement on self-reported

386

71.5 72.8 68.4 71.8 70.8 85.9 81.3 78.0 79.1 72.5 80.8 52.3 74.0 82.9 68.1 72.8 64.7 81.7 72.7 76.5 76.9 57.9 61.3 76.2 82.5 79.1 65.8

Australia Austria Belgium Canada Chile Czech Republic Denmark Estonia Finland France Germany Greece Hungary Iceland Ireland Israel Italy Japan Korea, Republic of Latvia Lithuania Luxembourg Mexico Netherlands New Zealand Norway Poland

55.2 30.8 30.0 53.1 59.6 46.4 60.3 58.8 51.7 30.8 60.3 30.1 38.2 78.0 51.8 61.3 41.1 68.8 59.4 53.0 58.2 19.0 47.7 58.1 72.7 64.4 33.7

Employment Rate by Age % %

Country

28.5 9.4 5.3 25.3 39.9 13.9 19.2 33.3 14.1 6.5 17.0 10.6 6.9 52.6 22.9 41.9 12.3 46.6 46.2 24.6 22.0 3.3 37.6 17.0 44.0 29.5 10.5

%

Table 21.2 Labor market at old age, OECD countries, 2018

65.0 65.0 65.0 65.0 65.0 63.2 65.0 63.3 65.0 63.3 65.5 62.0 63.5 67.0 66.0 67.0 67.0 65.0 61.0 62.8 63.6 62.0 65.0 65.8 65.0 67.0 65.0

65.0 60.0 65.0 65.0 65.0 62.7 65.0 63.3 65.0 63.3 65.5 62.0 62.0 67.0 66.0 62.0 66.6 64.0 61.0 62.8 61.9 62.0 65.0 65.8 65.0 67.0 60.8

Normal retirement age Men Women 65.3 63.5 61.6 65.5 70.0 63.2 65.1 65.5 64.3 60.8 64.0 61.7 63.4 68.1 65.6 69.4 63.3 70.8 72.3 65.7 64.3 60.5 71.3 65.2 69.8 66.1 62.8

64.3 60.8 60.5 64.0 66.7 61.3 62.5 65.7 63.4 60.8 63.6 60.0 60.0 65.9 64.1 66.0 61.5 69.1 72.3 64.7 63.0 61.3 66.5 62.5 66.4 64.1 60.6

Effective retirement age Men Women 19.8 19.3 21.1 18.9 14.7 17.7 17.8 15.4 19.1 22.7 19.1 21.8 15.9 17.0 18.7 16.2 20.7 15.5 12.9 13.9 15.1 22.3 12.8 18.6 15.6 18.0 17.8

23.3 25.0 25.5 23.0 19.7 22.8 22.7 20.1 23.5 26.9 22.5 26.4 22.9 20.7 22.2 20.7 25.7 21.0 16.3 19.7 22.1 25.0 17.5 23.4 20.5 22.5 24.3

Expected years in retirement Men Women 18.0 15.9

12.2 14.3 5.1 8.5 15.5 13.6 12.2 12.4 13.7 8.3 12.7

5.0 8.3 12.1 12.0 14.2 13.2 (Continued)

13.5 14.2

9.9 10.7 5.1 8.5 14.4 12.5 10.7 8.5 12.1 6.9 11.5

3.5 6.3 13.7 9.7 10.5 10.4

Healthy years in retirement Men Women

Retirement and Health

387

388

72.5

49.6

22.3

64.2

65.2 62.2 62.0 65.0 65.0 65.0 51.0 65.0 66.0 63.5

65.2 62.2 61.7 65.0 65.0 64.0 48.0 62.7 66.0

Normal retirement age Men Women

65.4

68.5 61.1 63.1 62.1 66.4 66.4 66.3 64.7 67.9 63.7

65.4 59.9 60.1 61.3 65.4 65.0 64.9 63.6 66.5

Effective retirement age Men Women

17.8

15.7 17.8 19.0 21.7 18.0 18.8 15.3 18.9 16.4 22.5

21.6 23.4 25.6 26.6 21.3 22.6 19.8 22.2 19.8

Expected years in retirement Men Women

12.0

12.1

10.5 10.2

9.0 9.7 12.3 15.0 15.4 10.2

6.0 7.9 9.4 14.4 14.2 9.2

Healthy years in retirement Men Women

Note: Employment rate is employment as a percentage of the population. Normal retirement age is for an individual retiring in 2018 after an uninterrupted career from age 22. The average effective age of labor force exit is calculated as a weighted average of (net) withdrawals from the labor market at different ages over a 5-year period for workers initially aged 40 and over. Expected years in retirement is a calculation of remaining life expectancy from the time of effective age of labor force exit for men and women. Estimates of the number of years of additional life are calculated based on the United Nations World Population Prospects. Healthy years in retirement are approximated assuming that these are equal to 65+ healthy life expectancy at age 65—effective labor market exit age. Source: OECD Statistics.

Average

19.1 8.1 8.6 6.1 24.0 22.9 20.0 21.3 31.9

71.3 76.5 68.6 63.2 85.5 81.9 39.6 74.4 70.2

Portugal Slovak Republic Slovenia Spain Sweden Switzerland Turkey United Kingdom United States

46.0 32.5 24.9 39.2 70.2 61.3 30.1 54.2 55.5

%

Employment Rate by Age % %

Country

Jan C. van Ours

Retirement and Health

and physical health, finding an improvement. The positive health effects of retirement occurred for males and females at all educational levels. Heller-Sahlgren (2017) focuses on regular state pension ages to study the effects of retirement on mental health. These turn out to be absent in the short run but large and negative in the long run. The effects are not heterogeneous with respect to gender, educational attainment, and occupation. Celidoni et al. (2017) find that the effect of retirement on cognition depends on the nature of the retirement. With statutory retirement the effect is detrimental and getting worse over time, while early retirement has beneficial effects on cognition. Mazzonna and Peracchi (2017) suggest that workers in physically demanding jobs benefit from retirement in both mental and physical health, while for the rest of the workforce retirement has long-run negative effects on health and cognition. Bertoni et al. (2018) focus on the relationship between retirement and muscle strength loss. They find positive short-run effects but negative long-run consequences of early retirement, which accelerate the age-related decline in muscle strength. Kolodziej and Garc´ıa-G´omez (2019) investigate the heterogeneity of causal positive effects of retirement on mental health, finding that these are larger for those in poor mental health. Motegi et al. (2020) analyze the effects of retirement on health behavior (alcohol consumption, smoking, and physical activity), finding that the patterns vary a lot across countries while being heterogenous by age and gender.

21.3.2

Single-Country Studies

Studies on the relationship between retirement and health focusing on single countries mainly use two identification strategies to establish causality, i.e., IV and RDD.12 These are discussed separately starting with the IV studies. Coe et al. (2011) conclude from an analysis of U.S. data that the negative association between retirement and cognition disappears once selectivity in retirement status and retirement duration are considered. Instead, for blue-collar workers retirement is beneficial for cognition. For white-collar workers there seems to be no relationship. The authors argue that these findings are not inconsistent with the “use it or lose it” hypothesis, speculating that for blue-collar workers engaging in intellectually stimulating activities is easier in retirement but more difficult for white-collar workers. Bonsang et al. (2012) base their IV strategy on eligibility for U.S. social security, concluding differently, i.e., finding retirement to have a significant negative, though not instantaneous, effect on cognitive functioning. Behncke (2012) finds that retirement in England has negative effects on physical health, i.e., cardiovascular disease, problems in physical activities, cancer, and perceived health, while no significant effects on mental health problems were evident. Insler (2014) uses a general health index based on doctor-diagnosed health variables and self-reported health status collected in U.S. data. Following an IV strategy based on self-reported probabilities of working past ages 62 and 65, he finds a positive health effect of voluntary retirement and attributes this to a behavioral change of retirees, who, for example, are more likely to attempt to quit smoking. Based on U.S. data, Gorry et al. (2018) find that the impact of retirement on happiness is immediate, while health effects show up later. Retirement does not seem to affect healthcare utilization. Atalay et al. (2019) studied the effects of retirement of Australian workers on their cognitive skills, finding that the effects for women are moderate while for men retirement has a negative effect on reading performance. Kuusi et al. (2020) analyzed Finnish administrative data on the use of antidepressants as an indicator for mental health and hospital visits for cardiovascular and musculoskeletal diseases as indicator for physical health. They find beneficial health effects of retirement though these effects are relatively small and rather short-lived. 389

Jan C. van Ours

Recently, RDD studies have become more popular. Eibich (2015) used an RDD based on age-related financial incentives in the German pension system to explain changes in measures of self-reported and mental health. Because of financial incentives, discontinuities exist in the age-retirement profile at 60 and 65. The author finds positive effects on mental health, which he attributes to relief from work-related stress and strain to an increase in sleep duration and to a more active lifestyle. Less-educated workers benefit especially in physical health; for higher-educated workers mental health improves at retirement. F´e and Hollingsworth (2016) investigated the retirement effects on mental health, health problems, and healthcare utilization for UK males. Using an RDD for the short-run effects and a panel data model for the long-run effects, they find that retirement has neither short-run nor long-run health effects. Clouston and Denier (2017) analyzed U.S. data, finding that retirees have lower cognition before retiring and once retired they experience a more rapid decline in cognition, phrasing this as a cumulative “mental retirement” effect. Picchio and Van Ours (2020) studied the effects of retirement in the Netherlands, finding that these are heterogeneous by gender and marital status. Retirement of partnered men has positive effects on their self-perceived health and happiness. Single men retiring experience a drop in self-assessed health. Whether partnered or single, retirement of women hardly affects their health. Rose (2020) analyzed the relationship in England between retirement and various health indicators using a range of datasets. He concludes that retirement improves self-reported health, while little evidence exists of an effect of retirement on health behavior and mental health. Finally, Nielsen (2019) applies both an IV approach and an RDD strategy to Danish data. A reform of the statutory retirement age is used as an IV while a discontinuity at the earliest age of retirement is analyzed in an RDD. The analysis focuses on healthcare utilization and health in terms of comorbidities and mortality. Statutory retirement has no effects at all while early retirement only has a negative effect on healthcare utilization.

21.3.3

Overview Studies

All empirical studies discuss prior studies. This discussion ranges from a brief presentation of the main findings to systematic tabulated overviews.13 Some studies also focus on providing an overview of previous studies without making an empirical contribution themselves. Van der Heide et al. (2013) discuss longitudinal studies, concluding that the effects on general health and physical health are unclear, while effects on mental health seem to be beneficial. Nishimura et al. (2018) investigate the differences in retirement effects across various studies, concluding that the choice of estimation method is the key factor in explaining these differences. Redoing several earlier studies using a fixed effects instrumental variable analysis, the authors conclude that results are more stable, indicating positive health effects of retirement, though some cross-country heterogeneity remains. Kuhn (2018) summarizes various issues in the relationship between retirement and health, emphasizing that even credible empirical evidence is inconclusive. Therefore, deriving uniform policy conclusions is not possible. Van Mourik (2020) systemizes the results of about 60 studies that claim to have investigated the causal effects of retirement on health using a wide range of indicators, i.e., self-reported health, depression, cognition, grip strength, and healthcare use. He finds that half of the reported effects are not significant, while about 15 percent of the effects are negative and about 35 percent are positive.

21.4

Effects of Retirement on Mortality

To establish the effect of retirement on mortality difference-in-difference methods, IV methods, and RDD are used exploiting changes in (early) retirement ages. Hernaes et al. (2013) find that 390

Retirement and Health

a retirement reform in Norway indeed induced some workers to retire early, but the mortality of workers who retired early does not differ from those who did not retire early. Hallberg et al. (2015) analyze the effects of an early retirement offer to Swedish army personnel, finding that mortality among early retirees is lower. Fitzpatrick and Moore (2018) use a change in eligibility for social security retirement insurance in the United States to establish the effects of retirement, finding an increase in male mortality, which the authors attribute to retirementassociated changes in unhealthy behaviors. The increase is largest for unmarried males and males with low education levels. For females retiring, mortality does not increase significantly. Using a temporary change in the rules for early retirement of older civil servants in the Netherlands, Bloemen et al. (2017) find that early retirement reduces mortality. Hagen (2018) exploits a Swedish reform of retirement allowing local government workers to retire early, finding no effects on mortality or healthcare utilization. Kuhn et al. (2020) use Austrian administrative data, finding that retirement increases mortality for men but not for women. Grøtting and Lillebø (2020) study the effects of retirement on health in Norway using administrative and survey data in an RDD based on the statutory retirement age of 67. They find no effect on mortality. In addition to separate country-level studies, there are also some overview studies. Van Mourik (2020) concludes that for mortality close to 90 percent of the studies find no significant effect of retirement.

21.5

Cross-Partner Effects of Retirement

Partnered individuals may make retirement decisions at the household level. Furthermore, the retirement decision of one partner may affect the health of the other partner. This implies that cross-partner effects may exist in decision-making and in the effects of those decisions. For example, from an overview of the literature, Coile (2015) concludes that in about one-third of working couples’ partners retire within 1 year of each other. Bloemen et al. (2019) also find cross-partner retirement effects of retiring in the Netherlands. By contrast, Picchio and Van Ours (2020) find no indication of coordinated retirement decisions in the Netherlands. Cross-partner effects may be present not only in terms of retirement decisions but also in terms of health effects. Using Japanese data, Bertoni and Brunello (2017) investigate the so-called “Retired Husband Syndrome,” suggesting that wives of retiring men experience a negative mental health shock. M¨uller and Shaikh (2018) use data from various European countries to investigate the causal health effects of the retirement of a partner. Based on an RDD, they conclude that subjective health is negatively affected by the retirement of the partner and positively by own retirement. These effects are heterogeneous: male health is not affected by the retirement of his spouse, while female health is negatively affected by the retirement of her partner. Analyzing French labor force survey data, Messe and Wolff (2019) find no cross-partner spillover effects of retirement on health. Picchio and Van Ours (2020) study cross-partner effects of retirement on health in the Netherlands. Retirement of partnered men has positive effects on the mental health of their partner.

21.6

What Have We Learned?

The overview of the various studies on retirement and health do not easily yield firm conclusions. The empirical evidence on the effects of retirement on health is mixed. Some studies find a positive effect, while other studies conclude there is no effect or a negative effect. Retirement always comes with an increase in leisure time with health effects also depending on changes in health-related behavior such as physical exercise. The type of work prior to retirement seems to matter. With stressful work or work in unhealthy situations retirement may benefit health. 391

Jan C. van Ours

Nevertheless, having a job can be satisfying because the work is interesting and meaningful. If so, retiring may be harmful for health. Retiring often comes with a drop in income, which may be harmful to health because low income is often associated with an unhealthy lifestyle. Whether this also holds for a negative income shock in retirement is open for discussion. Some of the differences in the mental health effects of retirement may have to do with the nature of the retirement decision, i.e., whether retirement is voluntary or mandatory. Bassanini and Caroli (2015), for example, find that voluntary retirement often has a positive effect on mental health, whereas involuntary job loss has a negative effect on mental health. The difference in effects may relate to the sense of control that people have over their retirement decision. All in all, the effects of retirement on physical health, mental health, and mortality vary from study to study depending on analysis method and the country or countries involved. Abundant heterogeneity exists in the relationship between retirement and health. Heterogeneity exists in the type of retirement, which can be voluntary or statutory, and related to that heterogeneity is the timing of retirement, i.e., early or late. Furthermore, personal characteristics such as marital status, education, gender, and age are also heterogeneous. And heterogeneity exists in the type of job and related work, which may or may not be physically or mentally demanding.

21.7

What Can We Do?

Retiring is a complex phenomenon and yet a simple event. International differences in the phenomenon of retirement are substantial. Not only does the age of retirement vary significantly, but wide variation also exists in the quantity and quality of expected years of life after retirement. In recent decades, early retirement programs have slowly faded away while standard retirement ages have increased. On the one hand, retiring early has become more difficult for some workers, and they are forced to keep working up to a higher age. If retirement improves health for them, postponing retirement may not be welfare improving. On the other hand, other workers have been allowed to keep working up to a higher age according to their preferences. For them, the policy changes have been beneficial, and postponing retirement may have improved their health thus increasing welfare. What can be done from a policy perspective? Some economists advocate evidence-based policies. As a leading principle this is indeed worth recommending, but from a practical point of view is not easy to implement as the evidence is often unclear. This also holds for the relationship between retirement and health. Different studies point in different directions. This does not mean that drawing policy conclusions is impossible. Flexible retirement combines a one-size-fits-all policy with a differentiated approach. There are various ins and outs of flexible retirement. Van Vuuren (2014) argues that flexible retirement opportunities provide insurance to individual workers. Health shocks can be absorbed through early retirement while a loss in pension wealth can be balanced by later retirement. Induced by union-supported collective agreements, the statutory eligibility of retirement age is often effectively a mandatory retirement age.14 Eligibility for retirement has gradually changed from a right to an obligation. Mandatory retirement is a clear example of institutional discrimination. Discrimination in the labor market is a phenomenon that is present in many dimensions but hard to tackle in terms of magnitude and mechanisms and even more so in terms of effective policy measures. This also holds for age discrimination, which is to some extent institutional as many countries have mandatory retirement. Clearly, governments and unions are not averse to discrimination when it comes to age. In fact, they advocate age discrimination as a valid instrument of labor market policy. On the one hand, older workers are stimulated to keep working for a long time. On the other hand, they are forced to stop working if they reach a particular age irrespective of whether 392

Retirement and Health

they are physically or mentally fit and irrespective of whether they themselves would prefer to keep working.

21.8

Where Do We Go from Here?

With a further increase of the normal retirement ages, more information will become available about the health effects of these increases. This will provide researchers with the opportunity to explore heterogeneity of the health effects in more detail. Furthermore, the use of retrospectively collected life-history information may allow researchers to study how early life events affect health.15 With this retrospective information, a more detailed analysis and interpretation of the health effects of later life events such as retirement can be provided. Nevertheless, whether these new insights from research will stimulate the development of new policy ideas is doubtful. Such is life. A policymaker cannot only rely on research, and a researcher’s work is never done.

Notes 1 A third type of retirement is indirect retirement, i.e., retiring through an intermediate stage of unemployment, sickness, or disability. As the name suggests, indirect retirement only has an indirect relationship with retirement, and these intermediate transitions are ignored here. 2 Various ways exist to define/determine in a survey whether an individual is retired. This can be done by asking whether the individual is working, whether the individual receives retirement benefits, or whether the individual considers himself or herself to be retired. 3 Studies with a sole focus on life satisfaction, happiness, or well-being are ignored. These variables may relate to mental health and indeed some mental health summary measures include happiness. Nevertheless, well-being has a much wider meaning than health. 4 Studies using instrumental variables usually rely on age of eligibility for early retirement or statutory retirement. Sometimes this is accompanied by instrumental variables based on the years to or years since reaching the eligibility age. Cross-country studies can exploit cross-country variation in eligibility ages. Thus, even in a cross-section variation exists in the instrumental variables. Similarly, RDD studies mostly use discontinuities related to early or mandatory retirement ages. Because using eligibility ages for identification purposes is common, unless otherwise mentioned, when a reference is made to the use of instrumental variables or an RDD strategy these are based on eligibility ages. 5 French and Jones (2017) provide an overview of studies on how health affects retirement. They suggest that health is not the primary source of the decline in the retirement age in the last decades of the 20th century, while the decline of health with age only explains a small share of the decline in employment near the retirement age. Blundell et al. (2021) find for England and the United States that declines in health explain between 3 and 15 percent of the decline in employment between ages 50 and 70. The effect is much stronger in the United States than in England as for institutional reasons unhealthy Americans have a strong incentive not to work. 6 Although this is just a subset of all OECD countries, the averages over all OECD countries for the available years are very much comparable to these for the 11 countries. 7 According to OECD statistics, the effective retirement age is defined as “the average age of exit from the labor force for workers aged 40 and over. To abstract from compositional effects in the age structure of the population, labor force withdrawals are estimated using changes in labor force participation rates rather than labor force levels. These changes are calculated for each (synthetic) cohort divided into five-year age groups.” 8 Life expectancy after retirement is not a formal OECD statistic but calculated by the author as 65 plus the years of life expectancy after age 65 minus the effective retirement age. 9 Note that one must be careful with cross-country comparisons of healthy and non-healthy years as the information on this is based on micro data from EU-SILC (Statistics on Income and Living Conditions). Cross-country data comparability is limited because of cultural factors and different formulations of questions in EU-SILC. 10 There are several explanations for the differences in life expectation according to education. These include socioeconomic conditions like the type of job, (un)healthy lifestyles, and access to appropriate healthcare.

393

Jan C. van Ours

11 The number of healthy years in retirement is an approximation assuming that they are equal to 65+ healthy life expectancy at age 65—the effective labor market exit age. This calculation is an approximation because the healthy life expectancy at age 65 is a number for the full population while the effective labor market exit age relates to the labor force. Furthermore, all healthy years in retirement are assumed to come first. 12 Other identification strategies also exist. Using an increase in the full retirement age for men in Israel to perform a difference-in-difference analysis, Shai (2018) finds that employment at older ages has negative health effects as measured through self-assessed health doctor visits and hospitalization, in particular for low-educated workers. 13 Hagen (2018), Kuusi et al. (2020), and Picchio and Van Ours (2020) present such overview tables. 14 B¨orsch-Supan et al. (2018) argue that, for example, relaxing mandatory retirement may stimulate some workers to retire later but other workers will use the flexibility to retire earlier. To increase labor supply at older ages increasing the statutory retirement age is much more effective. According to them, there is no economic reason why claiming a pension benefit must imply leaving the labor force. Pension benefits should be actuarially fair while the age of exit from the labor force should depend on individual preference for work and leisure. 15 Such retrospective information on life histories is already collected in ELSA and SHARE (see Banks et al., 2020, for a discussion).

References ATALAY, K., BARRETT, G. F., AND STANEVA, A. (2019): “The effect of retirement on elderly cognitive functioning,” Journal of Health Economics, 66: 37–53. BANKS, J., BRUGIAVINI, A., AND PASINI, G. (2020): “The powerful combination of cross-country comparisons and life-history data,” Journal of the Economics of Ageing, 16: 100206. BASSANINI, A., AND CAROLI, E. (2015): “Is work bad for health? The role of constraint versus choice,” Annals of Economics and Statistics, 119/120: 13–37. BEHNCKE, S. (2012): “Does retirement trigger ill health?,” Health Economics, 21(3): 282–300. BELLONI, M., MESCHI, E., AND PASINI, G. (2016): “The effect on mental health of retiring during the economic crisis,” Health Economics, 25 (S2): 126–140. BERTONI, M., AND BRUNELLO, G. (2017): “Pappa ante portas: The effect of the husband’s retirement on the wife’s mental health in Japan,” Social Science & Medicine, 175: 135–142. BERTONI, M., MAGGI, S., AND WEBER, G. (2018): “Work, retirement, and muscle strength loss in old age,” Health Economics, 27(1): 115–128. BLOEMEN, H., HOCHGUERTEL, S., AND ZWEERINK, J. (2017): “The causal effect of retirement on mortality: Evidence from targeted incentives to retire early,” Health Economics, 26(12): e204–e218. BLOEMEN, H., HOCHGUERTEL, S., AND ZWEERINK, J. (2019): “The effect of incentive-induced retirement on spousal retirement rates: Evidence from a natural experiment,” Economic Inquiry, 57(12): 910–930. BLUNDELL, R., BRITTON, J., COSTA DIAS, M., AND FRENCH, E. (2021): “The impact of health of labor supply near retirement,” Journal of Human Resources, forthcoming. BOERI, T., AND VAN OURS, J. C. (2021): The Economics of Imperfect Labor Markets, third edition. Princeton, NJ: Princeton University Press. BONSANG, E., ADAM, S., AND PERELMAN, S. (2012): “Does retirement affect cognitive functioning?,” Journal of Health Economics, 31(3): 490–501. ¨ BORSCH -SUPAN, A., BUCHER-KOENEN, T., KUTLU-KOC, V., AND GOLL, N. (2018): “Dangerous flexibility—Retirement reforms reconsidered,” Economic Policy, 33(94): 315–355. CELIDONI, M., DAL BIANCO, C., AND WEBER, G. (2017): “Retirement and cognitive decline. A longitudinal analysis using SHARE data,” Journal of Health Economics, 56: 113–125. CLOUSTON, S. A. P., AND DENIER, N. (2017): “Mental retirement and health selection: Analyses from the U.S. Health and Retirement Study,” Social Science & Medicine, 178: 78–86. COE, N. B., VON GAUDECKER, H.-M., LINDEBOOM, M., AND MAURER, J. (2011): “The effect of retirement on cognitive functioning,” Health Economics, 21(8): 913–927. COE, N. B., AND ZAMARRO, G. (2011): “Retirement effects on health in Europe,” Journal of Health Economics, 30(1): 77–86. COILE, C. C. (2015): “Economic determinants of workers’ retirement decisions,” Journal of Economic Surveys, 29(4): 830–853.

394

Retirement and Health

EIBICH, P. (2015): “Understanding the effect of retirement on health: Mechanisms and heterogeneity,” Journal of Health Economics, 43: 1–12. FE´ , E., AND HOLLINGSWORTH, B. (2016): “Short- and long-run estimates of the local effects of retirement on health,” Journal of the Royal Statistical Society Series A, 179(4): 1051–1067. FITZPATRICK, M., AND MOORE, T. (2018): “The mortality effects of retirement: Evidence from social security eligibility at age 62,” Journal of Public Economics, 157: 121–137. FRENCH, E., AND JONES, J. B. (2017): “Health, health insurance, and retirement: A survey,” Annual Review of Economics, 9: 383–409. GORRY, A., GORRY, D., AND SLAVOV, S. N. (2018): “Does retirement improve health and life satisfaction?,” Health Economics, 27(12): 2067–2086. GRØTTING, M. W., AND LILLEBØ, O. S. (2020): “Health effects of retirement: Evidence from survey and register data,” Journal of Population Economics, 33(2): 671–704. HAGEN, J. (2018): “The effects of increasing the normal retirement age on healthcare utilization and mortality,” Journal of Population Economics, 31(1): 193–234. HALLBERG, D., JOHANSSON, P., AND JOSEPHSON, M. (2015): “Is an early retirement offer good for your health? Quasi-experimental evidence from the army,” Journal of Health Economics, 44: 274–285. HELLER-SAHLGREN, G. (2017): “Retirement blues,” Journal of Health Economics, 54: 66–78. HERNAES, E., MARKUSSEN, S., PIGGOTT, J., AND VESTAD, O. L. (2013): “Does retirement age impact mortality?,” Journal of Health Economics, 32(3): 586–598. HESSEL, P. (2016): “Does retirement (really) lead to worse health among European men and women across all educational levels?,” Social Science & Medicine, 151: 19–26. INSLER, M. (2014): “The health consequences of retirement,” Journal of Human Resources, 49(1): 195–233. ´ KOLODZIEJ, I. W., AND GARC´I A-GOMEZ , P. (2019): “Saved by retirement: Beyond the mean effect on mental health,” Social Science & Medicine, 225: 85–97. KUHN, A. (2018): “The complex effects of retirement on health,” Bonn: IZA; IZA World of Labor, 430. ¨ KUHN, A., STAUBLI, S., WUELLRICH, J.-P., AND ZWEIM ULLER , J. (2020): “Fatal attraction? Extended unemployment benefits, labor force exits, and mortality,” Journal of Public Economics, 191: 104087. KUUSI, T., MARTIKAINEN, P., AND VALKONEN, T. (2020): “The influence of old-age retirement on health: Causal evidence from the Finnish register data,” Journal of the Economics of Ageing, 17: 100257. MAZZONNA, F., AND PERACCHI, F. (2012): “Ageing, cognitive abilities and retirement,” European Economic Review, 56(4): 691–710. MAZZONNA, F., AND PERACCHI, F. (2017): “Unhealthy retirement,” Journal of Human Resources, 52(1): 128–151. MESSE, P.-J., AND WOLFF, F.-C. (2019): “The short-term effects of retirement on health within couples: Evidence from France,” Social Science & Medicine, 221: 27–39. MOTEGI, H., NISHIMURA, Y., AND OIKAWA, M. (2020): “Retirement and health investment behaviors: An international comparison,” Journal of the Economics of Ageing, 16: 100267. ¨ MULLER , T., AND SHAIKH, M. (2018): “Your retirement and my health behavior: Evidence on retirement externalities from a fuzzy regression discontinuity design,” Journal of Health Economics, 57: 45–59. NIELSEN, N. F. (2019): “Sick of retirement?,” Journal of Health Economics, 65: 133–152. NISHIMURA, Y., OIKAWA, M., AND MOTEGIE, H. (2018): “What explains the difference in the effect of retirement on health? Evidence from global ageing data,” Journal of Economic Surveys, 32(3): 792–847. PICCHIO, M., AND VAN OURS, J. C. (2020): “Mental health effects of retirement,” De Economist, 168(3): 319–352. ROHWEDDER, S., AND WILLIS, R. J. (2010): “Mental retirement,” Journal of Economic Perspectives, 24(1): 119–138. ROSE, L. (2020): “Retirement and health: Evidence from England,” Journal of Health Economics, 73: 102352. SHAI, O. (2018): “Is retirement good for men’s health? Evidence using a change in the retirement age in Israel,” Journal of Health Economics, 57: 15–30. VAN DER HEIDE, I., VAN RIJN, R., ROBROEK, S., BURDORF, A., AND PROPPER, K. (2013): “Is retirement good for your health? A systematic review of longitudinal studies,” BMC Public Health, 13(1): 1180. VAN MOURIK, C. (2020): “A meta-analysis into the causal effect of retirement on health,” Tilburg: Netspar Academic Series, 2020-001. VAN VUUREN, D. (2014): “Flexible retirement,” Journal of Economic Surveys, 28(3): 573–593.

395

22 THE RELEVANCE OF COGNITION IN THE CONTEXT OF POPULATION AGEING Bernt Bratsberg, Ole Røgeberg, and Vegard Skirbekk

Abstract Cognitive ability has well-established links to education, job performance, health, and fertility. Using Norwegian administrative data with population coverage, we assess how outcomes across these life domains vary with age. We examine how age profiles changed across the 1950–1969 birth cohorts—all of whom can be observed through age 50. Higher cognitive ability is associated with longer education, delayed entry into the labor market, and a steeper age-earnings profile. In the family domain, higher ability is associated with a greater likelihood of partnering and delayed but ultimately higher fertility. In the health domain, higher ability is associated with reduced risk of death and disability pension. While these patterns persist and remain substantial across all cohorts studied, different patterns of gradual change emerge across domains: convergence across ability groups for educational attainment and disability and divergence for childbearing.

22.1

Background

Cognitive abilities (CA) vary systematically over the life course (Liem et al., 2017; Liverman et al., 2015; Prull et al., 2000; Salthouse, 2012). We might expect a broad measure of cognitive ability to show a humped shape over the lifespan, perhaps peaking in mid adulthood and gradually declining with age. In practice, such a pattern depends crucially on how our CA measure weights different underlying components with markedly different trajectories (Hartshorne and Germine, 2015). Fluid intelligence, the “raw” processing power and ability to reason abstractly in novel contexts, is held to typically peak in early adulthood. In tests of speed, reasoning, and memory, for instance, the average performance of those in their early 20s is in the 75th percentile of the population ability distribution, whereas the average performance of those in their early 70s is near the 20th percentile (Salthouse, 2004). Crystallized intelligence—which captures learned skills, knowledge, and facts—may keep increasing throughout the working life. This is well established for vocabulary-related tasks. Vocabulary tests show age-related increase in performance that plateaus around 50, whereas the performance of experienced crossword solvers 396

DOI: 10.4324/9781003150398-26

The Relevance of Cognition in the Context of Population Ageing

was found to increase even beyond this age (Salthouse, 2004). Measures of fluid and crystallized intelligence, in turn, reflect a mix of more narrowly specified cognitive abilities (e.g., processing speed, memory, and attention span), each of which may have its own life course trajectory and peak at different ages (Hartshorne and Germine, 2015). Cognitive ability also differs across individuals, with the most common CA measure—the IQ score—being constructed to have a mean of 100 and a standard deviation of 15. These individual differences persist over time, with about half of the between-individual variation at age 70 explained by differences present already at age 11 (Deary, 2014). Surprisingly, some data even indicate that the overall variation between individuals is constant or possibly shrinking with age (Salthouse, 2004). Cognition is important in many life arenas, which suggests that variation in cognitive ability both across and within individuals may influence and help determine the age trajectory of other life outcomes as well. Clarifying the causal mechanisms involved, however, is a tall order. In some cases, the link between cognitive ability and other outcomes may be quite direct. Using data from performance in chess tournaments and adjusting for selective attrition, for instance, chess performance over the life course seems to track the age profile of fluid intelligence, declining steadily from the early 20s onwards (Bertoni et al., 2015). For earnings and occupational status, however, we may suspect that performance in young adulthood will cast a longer shadow by influencing educational attainment and early accumulation of human capital. As an example, we might suspect that fluid intelligence helps predict the rate at which human capital (a form of “crystallized” intelligence) will increase with exposure to learning opportunities and depreciate with the passage of time). Early fluid CA scores would then predict future trajectories of yet-to-be-learned competences that are valued in the labor market. Dynamic feedback effects also reflect exposure to environmental influences such as education that have causal effects on CA (Ritchie et al., 2015; Schneeweis et al., 2014). For example, those with high initial ability may systematically receive stronger “doses” of education, healthy and supportive upbringing, or cognitively stimulating jobs that may boost or sustain their CA relative to what it would otherwise have been. Early differences in CA can thus be amplified or dampened as individuals are sorted and self-select into environments that themselves influence cognitive trajectories (Dickens and Flynn, 2001). Correlations between cognitive and other outcome trajectories may also reflect common causal factors: the robust and strong correlation between early cognitive ability and health and mortality outcomes (Calvin et al., 2010) has been hypothesized to also reflect more general biological processes of development and maintenance (Whalley and Deary, 2001). Consistent with this, some evidence indicates a shared genetic etiology between CA and several diseases and life expectancy (Arden et al., 2016; Hagenaars et al., 2016). In this chapter, we assess how age-related trajectories of employment, family, and health outcomes differ for males with different CA scores. We do this using Norwegian administrative register data with full population coverage and a stanine (nine-category) CA score for males measured at military conscription around the age of 18. Linking registers at the individual level, we construct a longitudinal panel data set with low levels of attrition covering the 1950–1969 birth cohorts and outcomes measured up to 2019. This allows us to assess whether males with different levels of CA in young adulthood differ systematically in their age profile of employment, earnings, marriage, childbearing, medical disability, and mortality. Our main descriptive strategy is simple by design: Aggregating CA scores into three categories (low, medium, and high), birth cohorts into groups of five, and measuring outcomes at the integer age level, we calculate average outcomes within each CA-cohort-age cell within our observation period. For the earliest cohorts, this allows us to characterize their trajectories from age 20 until retirement age and for the latest cohorts from 20 to the early 50s. Separately 397

Bernt Bratsberg et al.

assessing age profiles for different sets of birth cohorts is done to assess the robustness and stability of patterns across subsamples of the data population. This also speaks to contextual variation, in that different cohorts faced different educational opportunities, labor market structures, and social norms surrounding family and marriage at any given age. To further assess any changes over time, we also—separately for each birth cohort—compare average outcomes at age 50 in the low and high CA score groups to the average outcomes of the medium CA score group.

22.2

Data

Data on cognitive ability come from a three-part test given at military conscription. The three-part test covers vocabulary, arithmetic, and abstract reasoning. While the test remained unchanged throughout our data period, secular increases in CA scores across birth cohorts (Flynn effects) led to a re-norming of the test that affected those born April 1961 and later and shifted the stanine scale by one point relative to earlier cohorts. This should not affect our analyses, which compare low scores (1–4 for those born before April 1961, 1–3 for those born after), medium (5–7 before April 1961 and 4–6 for after), and high (8–9 before April 1961 and 7–9 after). The medium category covers about 60 percent of all scored males, while the low and high categories cover about 20 percent each. A clear majority of each birth cohort is scored, with coverage increasing across the birth years covered by our analyses (Figure 22.1). Unscored males are a heterogeneous group whose composition has likely changed over time, as indicated by the less consistent cohort trends in educational attainment relative to that for scored males (Figure 22.1). In the earlier cohorts, many of the unscored individuals were young men at sea who were exempted from conscription testing, while others were not scored because medical or other issues made them unfit for military service. Analyses on birth cohorts from the 1960s and later indicate that those unscored

Figure 22.1 Fraction of men with valid IQ score and educational attainment in analysis population, by birth year. Note: Population consists of men born in Norway and present in the country on their 18th birthday. Educational attainment refers to normed years of schooling of the highest completed education by age 50. N = 633, 575.

398

The Relevance of Cognition in the Context of Population Ageing

in later cohorts were disproportionately drawn from the lower part of the cognitive ability distribution (Bratsberg and Rogeberg, 2018). Data on employment and earnings come from the register underlying the national pension system, established in 1967, and from the registers of the welfare and tax administrations. Employment is defined as annual earnings above 1G, where G denotes the base amount of the national pension scheme and the minimum earnings for pension eligibility under the rules that prevailed through 2011. Data on marriage, cohabitation, and children are drawn from the population register and the registers of Statistics Norway. Civil status covers the years 1991–2019. Throughout this period, we can consistently identify marriage status and cohabiting unmarried partners with common children. Linkable parental identifiers are present for most children born since the mid-1960s, with father links missing for only 1.6 percent of children born in Norway in 1965, declining to 0.8 percent for those born in 1995. Mortality data come from the population register, while a separate register identifying all recipients of disability benefits covers 1992–2018. Unfortunately, the conscription register (with CA scores) excludes most men who died before 1979, implying that we cannot link CA and young-age mortality for those born in the early 1950s. In sum, our analysis population consists of all males born in Norway between 1950 and 1969, conditional on being alive and not emigrated at conscription age (year of 18th birthday). Apart from mortality analyses, we additionally condition observations on those alive and present in Norway at the end of the observation year.

22.3

Education, Employment, and Earnings

The cognitive ability score from military conscription testing at age 18 strongly predicts later educational attainment, and the extended educational period is mirrored as a delayed labor market entry for high-scoring males (Figure 22.2). At ages above 30, employment shares increase with CA, with the labor market participation of the low-CA group showing a particularly strong decline with age. At age 61 (the year before eligibility for early retirement), there is a 23 percentage point gap (66 vs. 89 percent) in employment rates for males from the lowest and highest scoring groups of the earliest cohort depicted (1950–1954). Comparing the cohorts, the qualitative patterns remain similar. Educational attainment increased over cohorts in the medium- and especially the low-CA group, and employment rates declined slightly during the 30–50-year age span, but, when evaluated at age 51, remained relatively stable across the four cohort groups. Relative to the patterns of employment by CA group, the earnings patterns are more striking (Figure 22.3). Whether we consider average earnings over the group as a whole or average earnings conditional on employment, higher scores are associated with markedly stronger age trends and higher peak levels of earnings. Again, the main takeaway from the visual cross-cohort comparisons is one of apparent stability—though with some indication that the last cohort group saw average earnings in the high-scoring group level off at an earlier age. These patterns are in line with much of the existing research literature: Cognitive ability in young adulthood is a strong predictor of educational attainment and occupational status at age 30 (Hegelund et al., 2019), and these associations remain strong (though somewhat attenuated) in analyses comparing siblings (Hegelund et al., 2019). A general measure of cognitive ability has been claimed as the best predictor of occupational level and job performance (Schmidt and Hunter, 2004), though measures of more narrowly specified cognitive abilities may do better on 399

Bernt Bratsberg et al.

Figure 22.2 Life-cycle education and employment, by CA and birth cohort group. Note: Attainment is the normed years of schooling of the highest completed education at each age. Employment is defined as annual earnings from work exceeding 1G (base amount of national pension scheme). Both series conditional on resident in Norway end of year. Data cover 1970–2018. To ease the comparison of panels, the vertical lines indicate the highest comparable age across the four cohort groups (age 51) and the age before eligibility for the early retirement program, “AFP” (age 61).

some criteria (Lang and Kell, 2020). Finally, the “cognitive reward” in earnings increases with age (Lin et al., 2018). Linking these trajectories to the underlying trajectories of cognitive change is more complicated, as multiple mechanisms may relate to the differences in earning profiles. In the simplest labor market models, earnings can be taken to reflect productivity. Defining a job by the set of tasks involved, as in Autor and Handel (2013), these tasks will require some mix of specific

Figure 22.3 Life-cycle earnings by CA and cohort group, unconditional (top panels) and conditional (bottom panels) on employment. Note: Earnings are inflated to 2018 NOK, and are depicted in units of 100,000. See also note to Figure 22.2.

400

The Relevance of Cognition in the Context of Population Ageing

cognitive and physical skills and abilities. Within this baseline model, we might interpret the earnings profile by age as reflecting the age-related change in the specific mix of abilities and acquired competencies required for some specific occupation. If those with higher CA select into jobs where CA is relatively important, we would expect the earnings of high-scoring males to mirror the development and maintenance of cognitive skills and abilities. The earnings profile of lower-ability males, however, might show a stronger influence of analogous performance trajectories in more physical skills and abilities, as age is related to an increased prevalence of illness and poor health and more disability and mobility restrictions (Beard et al., 2016; Chang et al., 2019). A natural extension of this model is to introduce human capital investments. Job performance often depends on learned skills, trained abilities, and accumulated knowledge. In jobs with a strong cognitive component, those with higher CA may be faster learners and more capable of accumulating and using accumulated knowledge, and the potential productivity payoff may be larger for investments in cognitive skills than for investments in more physical skills. This points toward a framework where earnings match productivity profiles that in turn reflect the trajectories of skills and abilities causally related to performance in any specific occupation (Skirbekk, 2008). Finally, it is worth noting that the models discussed so far share the assumption that earnings reflect the contemporaneous productivity of a worker. Economists have questioned this assumption along different lines. In jobs where monitoring is hard, for instance, a rising wage profile may reflect an implicit contract that serves as a disciplining device (Lazear, 1981): workers are paid below their marginal product early in their career, and the firm holds the surplus and uses it to raise the workers’ earnings when they are more senior at the firm. Those who shirk risk being detected and losing their job, thus forfeiting the “delayed earnings.” From this perspective, differences in earnings profiles could relate to differences in the costs of monitoring worker effort in more “cognitive” relative to more “manual” jobs. Others have suggested that earnings profiles may deviate from productivity profiles as a result of social norms (earnings “should” increase with age), that senior workers are better at extracting rents from a firm, or that workers desire rising consumption profiles and choose jobs with increasing wage profiles as a forced saving mechanism (Neumark, 1995).

22.4

Marriage, Cohabitation, and Fertility

Marriage/cohabitation rates increase with IQ for all cohorts studied, with low-scoring males having substantially reduced probabilities of being observed with a spouse/partner (see Figure 22.4). For example, for the 1950–1954 cohort, marriage rates at age 51 were 64 percent for the low-CA group compared with 75 percent for the high-CA group. This difference has widened over time; marriage rates at age 51 for the 1965–1969 cohort stood at 56 and 70 percent for the low- and high-CA groups, respectively. There is strong evidence suggesting a trend toward fertility postponement, where later cohorts consistently delay fertility to higher ages. Within each cohort, however, a slight but consistent difference occurs in fertility timing by cognitive ability, with the low-scoring males having more children at earlier ages, while medium- and higher-ability males have more children at a later age and end up with a higher number overall. To further assess how cognitive ability scores relate to marriage/cohabitation, we can also assess how civil status changes over time for those who are either partnered or nonpartnered at age 41 (the youngest age at which civil status is available for all birth years of the analysis population). This shows the net result of “churn” in the marriage market and finds that males 401

Bernt Bratsberg et al.

Figure 22.4 Marriage and children over the life cycle, by CA and cohort group. Note: Marriage rates include cohabiting couples with common children. Data on civil status cover the period 1991–2019 and children the period 1964–2019. See also note to Figure 22.2.

partnered at age 41 are more likely to remain partnered at subsequent ages, while nonpartnered males are more likely to become partnered the higher their CA score (Figure 22.5). Comparing the cohorts, we see a larger net dissolution of partnerships on average in more recent cohorts and an increased difference among the CA groups evaluated at age 51. For the nonpartnered group, the qualitative picture is stable over time, though the high-scoring group sees reduced partnering rates and increasingly resembles the medium-scoring group. The strength and direction of the statistical association between cognitive ability and fertility is a controversial and politically fraught topic, with some claiming that a persistent inverse relationships is both a natural feature of a modern society and an ongoing and serious threat

Figure 22.5 Marriage rates by age for those married (top panels) and not married (bottom panels) at age 41, by CA and cohort group. Note: Marriage rates include cohabiting couples with common children. Data on civil status cover 1991–2019. See also note to Figure 22.2.

402

The Relevance of Cognition in the Context of Population Ageing

to long-term human welfare (Lynn and Harvey, 2008; Nyborg, 2012). While a recent metareview argues that the number of children declines with cognitive ability (Reeve et al., 2018), this is not the case for Norwegian males born 1950–1969: the high-scoring group has a slightly delayed fertility profile, but when evaluated at higher ages the low-ability males have substantially fewer children on average. This is in line with a recent paper using conscription data from neighboring Sweden covering the same historical period, which also reports a positive relationship between male cognitive ability and lifetime fertility (Kolk and Barclay, 2019). Turning to marriage/cohabitation, others have also reported a positive association between cognitive ability and marriage probability for men (Aspara et al., 2018; Taylor et al., 2005), while reported associations between cognitive ability and divorce risk seem to vary more (Dronkers, 2002; Taylor et al., 2005). Cognition-related factors such as income, schooling length, and schooling test score performance, however, tend to relate to a lower risk of divorce in line with our results (Britt and Huston, 2012; Charles and Stephens, 2004).

22.5

Health and Mortality

Throughout the age span studied, males in the low-IQ group have substantially higher risk of mortality and receiving a disability pension (Figure 22.6). At ages 50+, the share of deceased in the low-scoring group is typically about twice that of the deceased share in the high-scoring group, while the share with disability pensions among the survivors is roughly four times higher for the low-scoring than the high-scoring group. A statistically and substantively strong association between cognitive ability and health/mortality has been robustly documented. A meta-analysis from 2010 covering 16 studies found a 24 percent (95 percent confidence interval 23–25) reduction in risk of death with every standard deviation increase in CA score (Calvin et al., 2010). Family-level confounders

Figure 22.6 Mortality and disability by age, by CA and cohort group. Note: Disability is measured by receipt of disability pension. Data on disability cover 1992–2018 and mortality 1968–2019. Note that, because conscription data are not available for a majority of those who died before 1979, the mortality curves by CA are drawn from ages 28 and 23 in the first two panels. See also note to Figure 22.2.

403

Bernt Bratsberg et al.

such as childhood socioeconomic status do not seem to drive the association: similar associations are found using family fixed effect models (Bratsberg and Rogeberg, 2017; Iveson et al., 2018), and genetically informed analyses suggest that the association is largely genetic in origin (Arden et al., 2016). Strong relationships with cognition have been reported for chronic health conditions at age 50 (Wraw et al., 2015); for unintentional injury from poisoning, fire, road traffic accidents, medical complications, and falling; for hospital admission due to assaults (fights, stabbing, blunt instruments, and firearms) (Whitley et al., 2010); for deaths caused by accidents, heart disease, and suicide (Batty et al., 2009); and for infectious disease exposure (Berkman et al., 2002; Chang et al., 2007; Gale et al., 2016). Notably, however, a large Swedish study failed to find evidence of an association with cancer (Batty et al., 2007). Multiple mechanisms have been suggested as potential contributors to these associations (Whalley and Deary, 2001), including the following: • •

•

•

Cognitive ability may be an indicator of “bodily insults,” in that nutrition and illnesses may permanently weaken both health and CA, including exposure to infectious diseases. Cognitive ability may be an indicator of “system integrity,” reflecting individual variation in the precision or error control of biological processes of development and maintenance. In support of this, it has been argued that simpler measures of “information processing efficiency,” such as reaction time, explain a substantial part of the CA-mortality association (Deary and Der, 2005). Cognitive ability may be a predictor of healthy behaviors, for instance because higher CA causally affects the ability to perceive and mitigate/avoid risks of acute or chronic health problems or injury, or adhere to complicated medical treatment regimens (Gottfredson and Deary, 2004). Cognitive ability may be a predictor of entry to safer and more stimulating contexts, e.g., less risky jobs and more supportive social environments incentivizing development.

22.6

Cohort Trends

The visual comparison of age profiles across 5-year cohort groups by and large leaves the impression of stability of differences across CA groups over time. To assess change across birth cohorts more directly, we now compare the average outcomes of high and low scorers at age 50 with the average outcomes of the medium-score group (Figure 22.7, with slopes of linear trend lines given in Table 22.1). This analysis finds that low-scoring males are converging on the reference group in terms of educational attainment, while diverging on employment rates, unconditional earnings, marriage, and family size. The high-scoring males, however, are converging on the medium-CA reference group in terms of education, earnings, and disability rates, while diverging on marriage rates. This means that educational attainment and disability have clearly converged across CA groups, but family outcomes have clearly diverged. Across all outcomes, however, cognitive ability remains a strong predictor for all cohorts considered. The trend in relative earnings of the high- vs. medium-CA group is perhaps the most surprising result from this analysis, in that it seems to run counter to the “enduring paradox” that less-educated workers have seen real wages fall “over the last four decades in industrialized economies” (Autor, 2019) and that a “dramatic growth in the wage premium associated with higher education and cognitive ability” has occurred (Autor, 2014). Such trends are held to reflect skill-biased technological change and labor market polarization. In brief, routine manual and labor tasks have been automated or outsourced to low-cost countries, hollowing out employment shares of occupations in the middle of the wage distribution and creating job 404

The Relevance of Cognition in the Context of Population Ageing

Figure 22.7 Relative socioeconomic, demographic and health outcomes of high and low vs. medium CA groups at the age of 50 by birth cohort. Note: Scatter points give the percent difference in outcome at age 50 between high vs. medium and low vs. medium CA groups, respectively.

growth in occupations characterized by either low wages and nonroutine manual tasks or high wages and nonroutine cognitive tasks. This, in turn, has led to marked growth in the returns to education and cognitive ability. These trends do not show up as rising divergence of earnings across cognitive ability groups in the Norwegian data: Although the low-score group has seen its (unconditional) earnings decrease relative to the middle group (through a modest relative employment decline), the high-score group has seen its relative earnings decline on average by about three percentage points per decade.

Table 22.1 Average change across 1950–1969 birth cohorts in outcomes at age 50 of high- and low-CA groups relative to reference group (medium score)

Note: ∗ Statistically significant at the 5 percent level. Note: Table entries are coefficient estimates of trend lines depicted in Figure 22.7. 405

Bernt Bratsberg et al.

Resolving this paradoxical result lies outside the scope of this chapter. One possibility is that the underlying economic trends have been dampened in Norway. Stronger shifts toward labor market polarization seem to occur in times of economic crises (Hershbein and Kahn, 2018), while Norway has largely avoided serious macroeconomic downturns in the last few decades—with unemployment remaining below 4 percent between the early 1990s and the COVID-19 pandemic. In addition, the underlying technological and economic trends may play out differently in the Norwegian institutional context: Norway is characterized by a strong welfare state with free and universal education, a legal right to secondary school education regardless of cognitive ability, strong labor unions, and a long history of relatively compressed wage distributions. From this perspective, the stable earnings difference (conditional on employment) between the low- and medium-CA group might reflect a mix of skill upgrading through increased education across cohorts along with strengthened selection into employment showing up as reduced employment shares in the low-CA group.

22.7

Conclusion

Using population-covering data for 20 birth cohorts of males, we document persistent and substantial differences in the age profiles of employment, earnings, family variables, and health across groups with different cognitive ability scores around age 18. Across all domains and cohorts, higher CA scores are associated with “better” outcomes on average: higher educational attainment, steeper and higher earnings profiles, higher employment shares with increasing age, higher shares with families, more children, more stable marriages, and lower rates of disability and death. For employment, family outcomes, and mortality, the low-scoring males seem to differ substantially from the more similar medium- and high-scoring males. For education, earnings, and disability, the differences seem to be more pronounced along the entire ability distribution. How ongoing skill-biased technological change will affect labor market outcomes remains to be seen. While cognitive abilities are increasingly important for employment, disability, and earnings, the threshold for what is acceptable cognitive functioning may change over time. Age-related cognitive decline can imply a stronger productivity decline in professions where cognitive functioning is more important than occupations where cognition matters less, particularly if fluid intelligence is required to stay abreast of rapidly changing technologies and knowledge. Population ageing implies that the cognition-related differences in average income will increase, as these differences widen throughout the working life. This is also the case for several of the other associations between cognition and family and health-related life outcomes. Cognition predicts careers, partnering, reproduction, and health more strongly at older ages. As cognitive ability tends to be positively associated with multiple beneficial life outcomes, this also suggests that correlated outcomes may stack up across different life domains, further amplifying the variation in overall life outcomes associated with cognition.

References ARDEN, R., M. LUCIANO, I. J. DEARY, C. A. REYNOLDS, N. L. PEDERSEN, B. L. PLASSMAN, M. MCGUE, K. CHRISTENSEN AND P. M. VISSCHER (2016): “The Association between Intelligence and Lifespan Is Mostly Genetic,” International Journal of Epidemiology, 45(1):178–185. ASPARA, J., K. WITTKOWSKI AND X. LUO (2018): “Types of Intelligence Predict Likelihood to Get Married and Stay Married: Large-Scale Empirical Evidence for Evolutionary Theory,” Personality and Individual Differences, 122:1–6.

406

The Relevance of Cognition in the Context of Population Ageing

AUTOR, D. H. (2014): “Skills, Education, and the Rise of Earnings Inequality among the ‘Other 99 Percent’,” Science, 344(6186):843–851. AUTOR, D. H. (2019): “Work of the Past, Work of the Future,” NBER Working Paper 25588. (National Bureau of Economic Research, Cambridge, MA. AUTOR, D. H. AND M. J. HANDEL (2013): “Putting Tasks to the Test: Human Capital, Job Tasks, and Wages,” Journal of Labor Economics, 31(S1):S59–S96. BATTY, G. D., K. M. WENNERSTAD, G D. SMITH, D. GUNNELL, I. J. DEARY, P. TYNELIUS AND F. RASMUSSEN (2007): “IQ in Early Adulthood and Later Cancer Risk: Cohort Study of One Million Swedish Men,” Annals of Oncology, 18(1):21–28. BATTY, G. D., K. M. WENNERSTAD, G D. SMITH, D. GUNNELL, I. J. DEARY, P. TYNELIUS AND F. RASMUSSEN (2009): “IQ in Early Adulthood and Mortality by Middle Age: Cohort Study of 1 Million Swedish Men,” Epidemiology, 20(1):100–109. BEARD, J. R., A. M. OFFICER AND A. K. CASSELS (2016): “The World Report on Ageing and Health,” The Gerontologist, 56(Suppl 2):S163–S166. BERKMAN, D. S., A. G. LESCANO, R. H. GILMAN, S. L. LOPEZ AND M. M. BLACK (2002): “Effects of Stunting, Diarrhoeal Disease, and Parasitic Infection During Infancy on Cognition in Late Childhood: A Follow-up Study,” The Lancet, 359(9306):564–571. BERTONI, M., G. BRUNELLO AND L. ROCCO (2015): “Selection and the Age–Productivity Profile. Evidence from Chess Players,” Journal of Economic Behavior & Organization, 110:45–58. BRATSBERG, B. AND O. ROGEBERG (2017): “Childhood Socioeconomic Status Does Not Explain the IQ-Mortality Gradient,” Intelligence, 62:148–154. BRATSBERG, B. AND O. ROGEBERG (2018): “Flynn Effect and Its Reversal Are Both Environmentally Caused,” Proceedings of the National Academy of Sciences, 115(26):6674–6678. BRITT, S. L. AND S. J. HUSTON (2012): “The Role of Money Arguments in Marriage,” Journal of Family and Economic Issues, 33(4):464–476. doi: 10.1007/s10834-012-9304-5. CALVIN, C. M., I. J. DEARY, C. FENTON, B. A. ROBERTS, G. DER, N. LECKENBY AND G. D. BATTY (2010): “Intelligence in Youth and All-Cause-Mortality: Systematic Review with Meta-Analysis,” International Journal of Epidemiology, 40:626–644. doi: 10.1093/ije/dyq190. CHANG, A. Y., V. F. SKIRBEKK, S. TYROVOLAS, N. J. KASSEBAUM AND J. L. DIELEMAN (2019): “Measuring Population Ageing: An Analysis of the Global Burden of Disease Study 2017,” The Lancet Public Health, 4(3):e159–e167. CHANG, L. Y., L. M. HUANG, S. S. F. GAU, Y. Y. WU, S. H. HSIA, T. Y. FAN, K. L. LIN, Y. C. HUANG, C. Y. LU AND T. Y. LIN. (2007): “Neurodevelopment and Cognition in Children after Enterovirus 71 Infection,” New England Journal of Medicine, 356(12):1226–1234. CHARLES, K. AND STEPHENS, JR, M. (2004): “Job Displacement, Disability, and Divorce,” Journal of Labor Economics, 22(2):489–522. DEARY, I. J. (2014): “The Stability of Intelligence from Childhood to Old Age,” Current Directions in Psychological Science, 23(4):239–245. DEARY, I. J. AND G. DER (2005): “Reaction Time Explains IQ’s Association with Death,” Psychological Science, 16(1):64–69. DICKENS, W. T. AND J. R. FLYNN. (2001): “Heritability Estimates Versus Large Environmental Effects: The IQ Paradox Resolved,” Psychological Review, 108(2):346. DRONKERS, J. (2002): Is There a Relation between Divorce Risk and Intelligence? Evidence from the Netherlands, Italia: European University Institute. GALE, S. D., L. D. ERICKSON, A. BERRETT, B. L. BROWN AND D. W. HEDGES (2016): “Infectious Disease Burden and Cognitive Function in Young to Middle-Aged Adults,” Brain, Behavior, and Immunity, 52:161–168. GOTTFREDSON, L. S. AND I. J. DEARY (2004): “Intelligence Predicts Health and Longevity, but Why?,” Current Directions in Psychological Science, 13(1):1–4. HAGENAARS, S. P., S. E. HARRIS, G. DAVIES, W. D. HILL, D. CM LIEWALD, S. J. RITCHIE, R. E. MARIONI, C. FAWNS-RITCHIE, B. CULLEN AND R. MALIK (2016): “Shared Genetic Aetiology between Cognitive Functions and Physical and Mental Health in UK Biobank (N= 112 151) and 24 GWAS Consortia,” Molecular Psychiatry, 21(11):1624–1632. HARTSHORNE, J. K. AND L. T. GERMINE. (2015): “When Does Cognitive Functioning Peak? The Asynchronous Rise and Fall of Different Cognitive Abilities across the Life Span,” Psychological Science, 26(4):433–443.

407

Bernt Bratsberg et al.

HEGELUND, E. R., T. FLENSBORG-MADSEN, J. DAMMEYER AND E. L. MORTENSEN (2018): “Low IQ as a Predictor of Unsuccessful Educational and Occupational Achievement: A Register-Based Study of 1,098,742 Men in Denmark 1968–2016,” Intelligence, 71:46–53. HEGELUND, E. R., T. FLENSBORG-MADSEN, J. DAMMEYER, L. H. MORTENSEN AND E. L. MORTENSEN (2019): “The Influence of Familial Factors on the Association between IQ and Educational and Occupational Achievement: A Sibling Approach,” Personality and Individual Differences, 149:100–107. HERSHBEIN, B. AND L. B. KAHN. (2018): “Do Recessions Accelerate Routine-Biased Technological Change? Evidence from Vacancy Postings,” American Economic Review, 108(7):1737–1772. ˇ UKI C´ , G. DER, G. D. BATTY AND I. J. DEARY. (2018): “Intelligence and All-Cause IVESON, M. H., I. C Mortality in the 6-Day Sample of the Scottish Mental Survey 1947 and Their Siblings: Testing the Contribution of Family Background,” International Journal of Epidemiology, 47(1):89–96. KOLK, M. AND K. BARCLAY. (2019): “Cognitive Ability and Fertility among Swedish Men Born 1951–1967: Evidence from Military Conscription Registers,” Proceedings of the Royal Society B, 286(1902):20190359. LANG, J. WB. AND H. J. KELL. (2020): “General Mental Ability and Specific Abilities: Their Relative Importance for Extrinsic Career Success,” Journal of Applied Psychology, 105(9):1047. LAZEAR, EDWARD P. (1981): “Agency, Earnings Profiles, Productivity, and Hours Restrictions,” The American Economic Review, 71(4):606–620. LIEM, FRANZISKUS, G. VAROQUAUX, J. KYNAST, F. BEYER, S. K. MASOULEH, J. M. HUNTENBURG, L. LAMPE, M. RAHIM, A. ABRAHAM AND R. C. CRADDOCK (2017): “Predicting Brain-Age from Multimodal Imaging Data Captures Cognitive Impairment,” NeuroImage, 148:179–188. LIN, D., R. LUTTER AND C. J. RUHM (2018): “Cognitive Performance and Labour Market Outcomes,” Labour Economics, 51:121–135. LIVERMAN, C. T., K. YAFFE AND D. G. BLAZER (2015): ”Cognitive ageing: Progress in understanding and opportunities for action,” Military Medicine, 180(11):1111–1113. LYNN, R. AND J. HARVEY (2008): “The Decline of the World’s IQ,” Intelligence, 36(2):112–120. NEUMARK, D. (1995): “Are Rising Earnings Profiles a Forced-Saving Mechanism?,” The Economic Journal, 105(428):95–106. NYBORG, H. (2012): “The Decay of Western Civilization: Double Relaxed Darwinian Selection,” Personality and Individual Differences, 53(2):118–125. PRULL, M. W., J. D. E. GABRIELI AND S. A. BUNGE. (2000): “Age-Related Changes in Memory: A Cognitive Neuroscience Perspective.” In: Craik, F. I. M., and Salthouse, T. A. (eds.), The Handbook of Ageing and Cognition, Mahwah, NJ: Lawrence Erlbaum Associates, pp. 91–153. REEVE, C. L., M. D. HEENEY AND M. A. WOODLEY OF MENIE (2018): “A Systematic Review of the State of Literature Relating Parental General Cognitive Ability and Number of Offspring,” Personality and Individual Differences, 134:107–118. RITCHIE, STUART J, TIMOTHY C BATES AND IAN J DEARY (2015): “Is Education Associated with Improvements in General Cognitive Ability, or in Specific Skills?,” Developmental Psychology, 51(5):573. SALTHOUSE, T. A. (2004): “What and When of Cognitive Ageing,” Current Directions in Psychological Science, 13(4):140–144. SALTHOUSE, T. A. (2012): “Does the Level at Which Cognitive Change Occurs Change with Age?,” Psychological Science, 23(1):18–23. SCHMIDT, F. L. AND J. HUNTER (2004): “General Mental Ability in the World of Work: Occupational Attainment and Job Performance,” Journal of Personality and Social Psychology, 86:162–173. doi: 10.1037/0022-3514.86.1.162. SCHNEEWEIS, NICOLE, VEGARD SKIRBEKK AND RUDOLF WINTER-EBMER. (2014): “Does Education Improve Cognitive Performance Four Decades after School Completion?,” Demography, 51(2):619– 643. SKIRBEKK, V. (2008): “Age and Productivity Potential: A New Approach Based on Ability Levels and Industry-Wide Task Demand,” Population and Development Review, 34:191–207. doi: 10.2307/25434764. TAYLOR, M. D., C. L. HART, G. D. SMITH, L. J. WHALLEY, D. J. HOLE, V. W. AND I. J. DEARY. (2005): “Childhood IQ and Marriage by Mid-Life: The Scottish Mental Survey 1932 and the Midspan Studies,” Personality and Individual Differences, 38(7):1621–1630. WHALLEY, L. J. AND I. J. DEARY. (2001): “Longitudinal Cohort Study of Childhood IQ and Survival up to Age 76,” British Medical Journal, 322(7290). doi: 10.1136/bmj.322.7290.819.

408

The Relevance of Cognition in the Context of Population Ageing

WHITLEY, E., G. D. BATTY, C. R. GALE, I. J. DEARY, P. TYNELIUS AND F. RASMUSSEN (2010): “Intelligence in Early Adulthood and Subsequent Risk of Unintentional Injury over Two Decades: Cohort Study of 1 109 475 Swedish Men,” Journal of Epidemiology & Community Health, 64(5):419–425. WRAW, C., I. J. DEARY, C. R. GALE AND G. DER (2015): “Intelligence in Youth and Health at Age 50,” Intelligence, 53:23–32.

409

23 PRODUCTIVITY IN AN AGEING WORLD Axel B¨orsch-Supan and Matthias Weiss

Abstract There is a widespread impression that older workers are less productive than their younger colleagues. If this impression were true, population ageing would have negative effects on overall productivity as the share of older workers is increasing and would thus directly reduce economic growth. A lack of economic growth would amplify the economic strains on ageing societies already exerted by increasing pension, healthcare, and long-term care expenditures. This chapter discusses many methodological challenges that have plagued the measurement of the link between age and productivity. It reviews micro- and macroeconomic studies and juxtaposes them with the literature on the relationship between population ageing and aggregate productivity. The methodologically most convincing studies on individual productivity in standard jobs find an initial increase in productivity, probably a learning effect, but then productivity remains flat until retirement. These micro-level studies measure productivity in a given technological environment and therefore ignore the contribution to productivity from improving this environment. This is where macro studies come into play, which suggest a negative influence of population ageing on total factor productivity. While the microeconomic and the macroeconomic evidence appear to contradict each other, they do not because they measure different relationships that are not linked through simple aggregation. This result has important implications for methodology and public policy.

23.1

Introduction

The link between age and productivity is an important relationship on all economic levels. In macroeconomics, the development of total factor productivity in an ageing population is a key issue for the economics of ageing because productivity growth dominates economic growth in modern economies (Maddison, 2001). Productivity growth stands in the middle of a yet unresolved controversy about whether population ageing is a major cause of the decline of total factor productivity observed after the turn of the millennium. In microeconomics, the development of individual labor productivity over the life cycle is a key determinant of labor supply and demand, including the retirement decision, and hence indirectly also determines the life-cycle path of consumption and saving. This is reflected in 410

DOI: 10.4324/9781003150398-27

Productivity in an Ageing World

managerial economics, where employers are keen to optimize personnel allocation. How individual labor productivity relates to age and how team composition and process design can influence this relationship are therefore important for management decisions. In economic policy, population ageing makes reforms of the welfare state an urgent but controversial issue because there are many trade-offs. Arguably, the most important one is between ageing-related social expenditures (e.g., pensions, healthcare, and long-term care) and productivity-enhancing investments (e.g., education and infrastructure). Moreover, the relationship between age and productivity plays an important role in the design of pension reform and the related controversial debates, especially regarding an increase of the retirement age. The belief that older workers are less productive dominates public discourse. This impression is widespread and implicit in many discussions about ageing, even in the economics profession. Examples are popular textbooks such as Lazear (1995, p. 40, Figure 4.1) and the macroeconomic literature, where many authors rather casually assume an increasing and then decreasing profile with a peak somewhere between age 30 and 45 (e.g., the seminal work by Altig et al., 2001, and more recently Kotschy et al., 2020). The belief that productivity peaks at relatively young ages and is low already in the 55–64-age interval is sometimes accompanied by straightforward age discrimination or mistrust against older workers. For example, 76 percent of French managers state that “an age over 55 years plays against a job candidate” (Prouet and Rousselon, 2018, based on Eurobarometer data). The European Union (EU) average of this negative image is 61 percent, and only in five EU countries is it less than 50 percent. The impression of declining productivity by age also dominates personnel policies by employers and retirement choices made by employees. In many countries, the assertion that productivity declines with age is used as a motivation for early retirement policies. If the impression were true, population ageing would have negative effects on overall productivity as the share of older workers is increasing and would thus directly reduce economic growth. This is the foundation for a rejuvenation of Hansen’s 1939 specter of another “secular stagnation” (Gordon, 2015, 2016). A lack of economic growth would amplify the economic strains on ageing societies already exerted by increasing pension, healthcare, and long-term care expenditures. The impression that productivity declines by age has a physiological foundation. Occupational medicine, cognitive psychology, and gerontology have documented that muscle strength; sight; lung, kidney, and heart functioning; and many other biometric indicators deteriorate from early age onward (Schmidt et al., 2000). We know from corresponding studies in medicine, psychology, and gerontology that physical and cognitive abilities that can be precisely measured in laboratory situations decrease with increasing age (Avolio and Waldmann, 1994; Ilmarinen, 1999). Figure 23.1 shows that basic physiological parameters such as the speed by which nerves transport electric signals, how much oxygen the blood can take up, or how much force a single muscle can exert, all have early peaks before age 30 and then decline. Measures of cognition, such as standard IQ tests, start declining at similarly early stages in the life course (Lehr, 2000; Park and Bischof, 2013), see Figure 23.2. However, cognitive abilities not only include “fluid intelligence,” but also have “crystalline” components. Fluid intelligence includes the fluency of adjustment, agility, ability to combine, coordination of cognitive processes, accuracy, orientation in new situations, etc. Crystalline intelligence, however, comprises skills that require general knowledge, experience, vocabulary, and understanding of language. With age, crystalline intelligence remains stable, while fluid intelligence declines (Weinert, 1992; Maercker, 1992; Park and Bischof, 2013). Staudinger and Baltes show that no age-related decline in performance can be observed in experience-related tasks (Staudinger and Baltes, 1996; Staudinger, 1999). Figure 23.3 shows the difference between 411

Axel B¨orsch-Supan and Matthias Weiss

Figure 23.1 Basic physiological parameters. Source: Schmidt et al. (2000).

Figure 23.2 Three estimates of the age-IQ relationship. Source: Lehr (2000).

measures of fluid and crystalline intelligence in a cross-sectional setting; Nyberg et al. (2012) show that the slopes are much flatter in longitudinal data. Complex professional tasks also require experience, social skills, and organizational talent, which are usually not included in the isolated laboratory measurements of physiological indicators and cognitive performance. This is reflected in the applied psychology literature. For example, Ng and Feldman (2008) study 10 dimensions of job performance: core task performance, 412

Productivity in an Ageing World

Speed of Processing Digit Symbol Letter Comparison Pattern Comparison

1 Fluid intelligence

Working Memory Letter Rotation Line Span Computation Span Reading Span

Z-Scores

0.5

0

Long-Term Memory Benton Rey Cued Recall Free Recall

-0.5 Crystallized intelligence

World Knowledge Shipley Vocabulary Antonym Vocabulary Synonym Vocabulary

-1

20’s

30’s

40’s

50’s

60’s

70’s

80’s

Age (Y)

Figure 23.3 Crystallized versus fluid intelligence. Source: Park and Bischof (2013).

creativity, performance in training programs, organizational citizenship behaviors, safety performance, general counterproductive work behaviors, workplace aggression, on-the-job substance use, tardiness, and absenteeism. Experience, political skills, and the ability to create and maintain large networks are important competencies for leadership positions. These more-difficult-tomeasure skills tend to grow with age; they only lose their functionality when they turn into mental immobility (e.g., unwise persistence on traditional procedures) or when they can no longer be used due to very large deficits in cognitive performance (Farr et al., 1998; Maier, 1998). Factors relating to the relationship between labor productivity and age are complex and mutually influencing. They also depend on the social and spatial environment and on individual conditions, including education, socioeconomic status, and health. Figure 23.4 graphically illustrates various influencing relationships. The relationship between age and work productivity is therefore much more complex than Figures 23.1 and 23.2 suggest. Moreover, while experience and the ability to deal with human nature appear to increase with age, these characteristics are much harder to measure than physiological indicators and cognitive performance. Hence, an innate bias toward direct measures that decline early in life may have contributed to the aforementioned impression that older individuals are less productive than their younger colleagues.1 Figure 23.5 can therefore schematically illustrate our idea of age-related performance and its uncertainties. While physiological and cognitive performance (red line) is precisely measured and peaks at relatively young ages, experience and the ability to deal with human nature (green line) peak much later—or maybe not at all—but can only be measured with substantial error (indicated by the green area between the error bands). Labor productivity is the sum of both (blue line, with error bands bordering the blue area). Depending on how important experience is, labor productivity changes as people age. Knowing whether the age at which labor productivity stagnates and begins to decline is currently on average closer to 30 years or beyond 70 years is obviously essential. Reviewing answers to this question is the aim of this contribution. 413

Axel B¨orsch-Supan and Matthias Weiss

Figure 23.4 Interactions in the age-productivity relationship. Source: B¨orsch-Supan et al. (2008).

Figure 23.5 is not universal. People differ. Therefore, knowing how the peak age varies by individual characteristics within and across sectors of the economy and the nature of jobs and how it can be influenced by economic policy and business management are important. Easily and precisely measurable individual top performances (e.g., records in sports or outstanding scientific achievements) are achieved in many areas almost exclusively at a young age. However, they cannot be transferred directly to everyday working life. After all, large parts of the professional world are deliberately organized such that the productivity of the organization does not hinge on the top performance of individuals. For example, assembly lines run slowly so that most workers can work largely without errors, as the subsequent rectification of assembly errors 414

Productivity in an Ageing World

Figure 23.5 Age and productivity over the life cycle. Source: Author depiction.

is very expensive. The division of labor only works in everyday life if it does not depend on the top performance of every individual involved. Moreover, individual productivity has external effects. The contribution of an individual to the operating result is also due to the contributions of others inside and outside the company, in which environment individuals work, how well production processes are designed, and how innovative products are. Aggregate productivity may therefore be a function of the average age of an entire population even if the shop-floor productivity of typical workers is independent of age. This chapter therefore starts with a review of methodological challenges in measuring productivity. We then review microeconomic and microeconometric studies with respect to these methodological challenges, followed by the literature on the macroeconomic relationship between population ageing and aggregate productivity. The final section concludes and draws some implications for public policy and future research.

23.2

Methodological Challenges

There are many definitions and measurement problems that have plagued the literature on the link between age and productivity. They occur both on the individual and at the country level.

23.2.1

Age-Productivity Profile of Individuals

Studies on the link between individual age and labor productivity of individuals face many methodological challenges: confusing age and cohort; the measurement of productivity; the appropriate level of aggregation; and, arguably the largest challenge, selectivity and endogeneity created by reverse causality. In a single cross-section, age is indistinguishable from birth cohort. Hence, the large performance losses in older people that often show up in cross-sectional studies may result from very different cultural and school learning and educational conditions in childhood (Lehr, 415

Axel B¨orsch-Supan and Matthias Weiss

2000). Panel data are essential to properly identify age-productivity trajectories and to purge them of cohort effects. Additional assumptions are necessary to account for period effects such as temporary productivity declines/increases in times of business-cycle down- and upturns, respectively.2 Conceptually, the labor productivity of a worker appears easy to measure. Individual-level physical productivity is the number of widgets of a given quality that a worker produces in a given time, e.g., apples harvested, nails forged, lines of code written, web pages designed, or judgments spoken. In the simplest cases, quality is obvious, e.g., the fraction of crooked nails or the number of bugs in the computer code. However, more often output is heterogeneous, e.g., designing simple text web pages versus sophisticated graphics. In this case, sales prices are used to measure quality. This leads to productivity measures based on value added per work hour. While value-added figures are easy to obtain at various levels of aggregation—because they are a byproduct of accounting at the plant, company, or country level—they may introduce bias because prices and thus value added may be distorted by monopolistic price setting and, in international comparisons, by exchange rate fluctuations due to other factors than the quality of products. Biases arise at the company level if the average age of workers in a company is correlated with that company’s power to set prices. In some cases, prices may not exist or are tautological. For example, national accounts measure the output of judges and other civil servants simply as the sum of their salaries. In this case, value-added measures of productivity are meaningless. Finding the right level of aggregation is a second challenge. An individualistic view fails to recognize that workers often work in teams and thereby affect one another’s productivity. This creates externalities. Older workers may devote some of their working time to helping or teaching younger workers. In this case, an individualistic approach would underestimate older workers’ and overestimate younger workers’ productivity. If, in turn, older workers depend on the help of younger colleagues, e.g., when lifting heavy loads, productivity at older ages is overestimated relative to that at younger ages. Salient aspects are workers’ contributions to their team’s work climate and how teams deal with emergency situations (Mas and Moretti, 2009; Backes-Gellner et al., 2011; G¨obel and Zwick, 2013). The experience-based productivity of employees depends heavily on the professional environment. The art of the division of labor consists precisely in assigning people to those tasks that they can do best in interaction with other people, including other ages. The concept of individualistic productivity, as measured in laboratory experiments, does not fit into a society that is strongly based on the division of labor. However, a plant or company view of productivity may obscure job heterogeneity and its interaction with motivation and thus productivity. One would expect, e.g., that the productivity effect of older workers on the shop floor, whose careers have peaked, differs significantly from the productivity effect of equally old managers, who might still have ambitions for a position at the company’s top or a realistic chance to move to another company. Promotions and the change from shop floor to management create jumps in the age-productivity profile. Plant-view regressions that average over different nonlinear age-productivity profiles might therefore create statistical artifacts and misinterpretations. A compromise between the individualistic approach and the plant or company view is to measure productivity at the level of work groups, which create a homogenous joint product and where strong interactions exist. Management science tries to optimize productivity at this level, hoping that experiences and ideas that have not yet been worked out of the older generation meet an environment in which younger colleagues persistently implement these suggestions. A serious methodological challenge is the potential endogeneity of age composition through various selection processes. Being in the labor force is endogenous because employers are more 416

Productivity in an Ageing World

likely to retain productive than unproductive workers. Hence, plant closures and early retirement tend to create a positive selection of productive workers. A similar positive selection occurs when managers assign easier tasks to workers who become less productive. The same happens if workers self-select into easier jobs when they become less productive. If workers were to become less productive with age, these selectivity effects would upwardly bias the slope of the age-productivity profile, i.e., they would suggest a spurious positive correlation between age and productivity, i.e., a rising age-productivity profile when it is flat or a flat one when it is negative. A related endogeneity problem exists for age structure at the company level. Because more productive firms are usually more profitable, they expand and increase their workforce. This leads to a rejuvenation of their workforce because new hires are more likely to be young. Similarly, innovative and technology-intensive companies with high labor productivity tend to have younger workforces because older employees were already employed in other companies when the innovative new companies were founded. Relating productivity to the age of the workforce in these cases results in a spurious negative correlation between productivity and age. This holds especially in economy-wide studies of innovations, labor productivity, and workforce age (Hellerstein and Neumark, 1995; Hellerstein et al., 1999). An assessment of the literature on age and productivity at the micro level thus needs to answer the following questions: How well do the data identify employees’ productivity, including their contribution to colleagues? How well do the studies deal with potential nonrandom sample selection? Are the studies successful in eliminating the bias stemming from the potential endogeneity of the age composition of the workforce?

23.2.2

Methodological Challenges at the Country Level

Distinguishing carefully between micro- and macro-level studies of the link between age and productivity is important. First, age in micro-level studies usually refers to the age of a single individual while in country-level studies it refers to the average age or the dependency ratio of a heterogeneous population. Second, productivity in micro studies usually refers to labor productivity of a single individual, while in country-level studies it mainly refers to total factor productivity, sometimes aggregate labor productivity, measured by the national accounts. Third and most importantly, what holds at the micro level does not necessarily translate to the macro level because additional processes are working at the macro level that are typically controlled for at the micro level. Examples are changes in the underlying technology. At the country level, labor productivity is defined as gross domestic product (GDP) divided by the number of workers or, more precisely, divided by the number of work hours (“hours’ productivity”). Because a country’s output also depends on other factors of production, total factor productivity (TFP) is a key measure at the macroeconomic level, defined as GDP divided by an aggregate of all relevant factors of production that is derived from a production function F, e.g., TFP = GDP/F(L,K) where L measures labor input and K capital input. Measurement challenges arise as work hours in the national accounts are often relatively rough estimates and the measurement of capital rests on many assumptions on depreciation. Moreover, separating the contributions of capital and labor to TFP is not only difficult due to measurement issues but also due to a conceptual challenge (B¨orsch-Supan, 1998). Measuring cross-national differences in productivity presents many additional difficulties. First, GDP in national accounts is aggregated from sectoral value-added measures that depend on prices that may not be comparable across countries. To purge country comparisons from exchange rate effects, purchasing power parities are used, which have their own measurement problems (European Commission, 2012). Second, if countries have differentially large public 417

Axel B¨orsch-Supan and Matthias Weiss

sectors, productivity comparisons are biased due to the tautological definition of public sector output mentioned in the preceding subsection. All these are rather mundane issues. However, they have serious implications for determining whether the ageing of a population endangers economic growth. If prices of capital goods do not properly reflect quality changes of these capital goods, then time-series analyses, which relate changes in economic growth to changes in the age structure of the population, are biased. Similarly, if purchasing power parities of countries in different stages of the demographic transition do not completely account for the different baskets of goods and services, cross-national studies trying to measure the effect of the average age in a population on its aggregate productivity will be biased. These mundane issues particularly affect the measurement of TFP as a Solow residual because this residual absorbs all measurement errors. Finally, the endogeneity problem for the age structure at the company level is still present at the country level, arguably to a lesser extent. Demography itself, especially average age, is endogenous if high productivity in one country causes higher immigration from other countries. Moreover, reverse causality exists if labor input depends on productivity. This especially affects labor force by age category as used, e.g., in the seminal work by Feyrer (2007). Finally, technical progress, especially in medicine, causes higher life expectancy and represents another channel of reverse causality. This has been addressed by using instruments, typically lags (Feyrer, 2007; F¨ollmi et al., 2019; Poplawsky-Ribeiro, 2020). Whether lags are indeed uncorrelated with the outcome variables is questionable given the low-frequency nature of demographics.3 Feyrer (2007) summarizes his paper as follows: “The evidence in this paper is not sufficient to establish a causal link between demographic change and productivity growth. An alternative possibility is that there is some omitted factor that had an impact on fertility in the past, but which affects productivity with long lags.” From a methodological perspective, this also holds for the long string of macroeconomic papers that have followed Feyrer and will be reviewed in Section 23.4.

23.3

Age and Productivity at the Micro Level

Labor economists have long been interested in estimating age-productivity profiles. Mark (1957) analyzes output figures of factory workers in 22 footwear and clothing establishments. Kutscher and Walker (1960) analyze office workers in five government agencies and 21 private companies. Both studies focus on workers with piece-rate pay schemes in jobs/establishments that counted the output. Both settings, factory and office work, experience little variation in average output per hour between age groups up to the age of 64, and the individual variation in performance within the age groups is always larger. Recent surveys on the subject are Skirbekk (2004), Gelderblom (2006), Chapter 6 in National Research Council (2012), and the special issues edited by Prskawetz et al. (2008) and Bloom and Sousa-Poza (2013) with their introductory summaries. Studies on the age-productivity relationship can be broadly divided into four groups: evaluations by superiors, wage-based productivity studies, the relationship between plant or company productivity and the age structure of its employees, and studies measuring individual productivity in a setting where such productivity can be measured well. One method of determining productivity over the life course is to evaluate appraisals from superiors. This is a common practice in industrial science and has contributed significantly to conveying the impression that older workers are less productive than younger ones. Examples are Medoff and Abraham (1980), Hunter and Hunter (1984), McEvoy and Cascio (1989), Salthouse and Maurer (1996), and Schneider and Stein (2006). The weakness of this approach, however, lies in the fact that productivity is not measured but “assessed.” Such supervisors’ assessments 418

Productivity in an Ageing World

are problematic because they may reflect prejudices about age-productivity profiles. Common images of old age and existing prejudices, e.g., about a supposedly decreasing productivity of the elderly, may be perpetuated, so that these studies are likely to suffer from systematic bias (Huber, 2002; Tuomi and Ilmarinen, 1999). A second group of studies uses wage and salary data as measures of individual productivity (e.g., Kotlikoff and Wise, 1989; Kotlikoff and Gokhale, 1992; Laitner and Stolyarov, 2005; McEvoy and Cascio, 1989). While wages and salaries reflect productivity to some extent, they are also determined by union negotiation, company policy, and seniority rules. The widespread use of pay schemes, and especially seniority-based pay scales, means that wages often increase and, more importantly, rarely fall with increasing age regardless of the progression of productivity. Lazear (1979, 1981) explains the increasing age-earning profiles with incentive effects. Loewenstein and Sicherman (1991) and Frank and Hutchens (1993) show in experiments that workers prefer increasing wage profiles and explain this with loss aversion and problems of self-control. Hence, the relationship between age and productivity in these studies tends to be positively biased. While the prevalence of seniority wages has decreased in recent years, it has not gone away, although it is small in some countries (e.g., Germany and Sweden) while it is large in others (e.g., France), see Prouet and Rousselon (2018). Third, by far the largest group of studies relate plant-level productivity to the age of the plants’ employees. Examples are Haltiwanger et al. (1999, 2007), Hellerstein et al. (1999), and Hellerstein and Neumark (2007) for the United States; Hægeland and Klette (1999) for Norway; Aubert (2003), Cr´epon et al. (2003), and Aubert and Cr´epon (2007) for France; Hellerstein and Neumark (1995) for Israel; Grund and Westerg˚ard-Nielsen (2008) for Denmark; Ilmakunnas and Maliranta (2005, 2007) and Daveri and Maliranta (2007) for Finland; Malmberg et al. (2008) for Sweden; Dostie (2011) for Canada; Prskawetz et al. (2006) and Mahlberg et al. (2013) for Austria and Sweden; Lallemand and Ryckx (2009) for Belgium; van Ours (2009) for the Netherlands; and Schneider (2007) and G¨obel and Zwick (2009) for Germany. Plant-level value added and the size and age structure of the plants’ employees are relatively accurate and directly measurable, e.g., from balance sheets and personnel records. However, the first shortcoming of this approach is that plant-level figures aggregate over heterogeneous jobs and positions. Averaging over very different nonlinear age-productivity profiles within a plant is likely to create aggregation biases if individuals have different impacts on the overall output and their productivity peaks at different ages, e.g., of production workers on the shop floor earlier than that of managers who might still have ambitions for a top-level position. An even more serious shortcoming is that the age structure of a plant tends to be endogenous because it is partly a function of the plant’s productivity. More productive firms are usually more profitable and therefore often expand and increase their workforce. Because new hires are likely to be younger (Quimet and Zarutskie, 2013), the workforce of more productive firms rejuvenates relative to less productive firms. This rejuvenating effect of more productive firms is of less importance for studies that focus on the difference between age-productivity and age-wage profiles and in which the estimated age-productivity profile is not the goal. Examples are Cr´epon et al. (2003), Ilmakunnas and Maliranta (2005, 2007), Haltiwanger et al. (1999, 2007), Hellerstein et al. (1999), and Hellerstein and Neumark (1995, 2007). For this difference, selectivity is not a problem. However, the age-productivity profiles, which are a side product of these studies, suffer from selection bias. If the main reason for the endogeneity of the age structure in a plant with respect to the plant’s productivity is that more productive firms are growing and employ new workers who are usually young, then this can be fixed by using firm fixed effects. However, firm fixed effects do not purge the bias that results if in an economic downturn productivity (in terms of value 419

Axel B¨orsch-Supan and Matthias Weiss

added) declines and, at the same time, no new (young) employees are hired such that the workforce ages. The same situation occurs when in an economic upturn, productivity rises and new (young) workers are hired such that the workforce becomes younger. In this case, one needs to instrument the age structure of the firm. Most papers in the literature use lagged values of their age variables as instruments and estimate an Arellano-Bond (1991) model in first differences (Cr´epon et al., 2003; Aubert and Cr´epon, 2007; Daveri and Maliranta, 2007; Lallemand and Ryckx, 2009; Malmberg et al., 2008; Prskawetz et al., 2006; van Ours, 2009; Hellerstein and Neumark, 2007; Schneider, 2007). The underlying assumption is that lagged differences in the age composition are exogenous with respect to current differences in value added. G¨obel and Zwick (2009) employ a more sophisticated dynamic panel model that uses lags as instruments in a more efficient way based on the econometric work by Arellano and Bover (1995), Blundell and Bond (1998), and Bond (2002), while Hellerstein and Neumark (2007) and Dostie (2006) derive their instruments from structural assumptions about the underlying economic process. The studies by Aubert and Cr´epon (2007), G¨obel and Zwick (2009), B¨orsch-Supan and Weiss (2016), and B¨orsch-Supan et al. (2021) suggest that the more one controls for endogeneity (fixed effects variants, instrumental variables), the more favorable the age-productivity profile becomes for older workers. These studies find an initial increase in productivity, probably a learning effect, but then productivity remains flat until retirement. Aubert and Cr´epon (2007) employ fixed effects (FE) and generalized methods of moments (GMM) estimators for a French matched employer-employee data set. While the within FE estimator has a peak of age productivity at about 35 years, the more sophisticated GMM estimator shows a significant decline only after the mid-50s. However, this decline is so small that the productivity of older workers never falls below average productivity. G¨obel and Zwick (2009) show how the estimated age-productivity profile depends on the methodology used. Figure 23.6 depicts four econometric methodologies to estimate the relationship between a plant’s productivity and the average age of its workforce, based on a German matched employer-employee data set. Panel (1) shows the age-productivity profile that results if the endogeneity problem is ignored. From the observation that plants with a younger (30– 35-year-old) workforce are more productive than plants with a 40–60-year-old workforce, one may wrongly conclude that a younger workforce is the cause of a plant’s higher productivity. In a second step, Panel (2) shows the age-productivity profile that results when plant-specific effects are corrected. Plants in highly productive (growing and thus rejuvenating) sectors are no longer compared with plants from unproductive (shrinking) sectors. Instead, the changes within plants are considered. This eliminates much of the endogeneity problem described previously. However, even when looking at the same plant over time, the endogeneity problem is not eliminated because the variation in the age of the workforce over time depends on whether the company is expanding (and thus rejuvenating) or whether a hiring freeze has led to the workforce slowly ageing. Therefore, in their third step, G¨obel and Zwick (2009) use a method that maps these internal dynamics and thus can also eliminate this part of the distortion. Panels (3) and (4) show the estimated age-productivity profiles. They increase up to the age group of 50–55-year-olds and do not significantly decrease afterward. A comparison of the four panels in Figure 23.6 shows that the relative productivity of older employees is estimated to be higher, the more sophisticated the method used to eliminate the endogeneity bias. However, the more sophisticated methodology comes at the expense of the precision of the estimates. As the vertical bars in Figure 23.6 show, the confidence intervals of the profiles increase with the panel number. In a related study, G¨obel and Zwick (2012) show that this pattern holds across all sectors. 420

Productivity in an Ageing World

Figure 23.6 Age-productivity profiles in German plants by econometric methodology. The brackets indicate 95 percent confidence intervals. Source: G¨obel and Zwick (2009).

A fourth type of study is based on direct productivity measures, for example, the number of publications in academic research (Oster and Hamermesh, 1998; Jones and Weinberg, 2011; Weinberg and Galenson, 2005; van Ours, 2009), Nobel Prizes (Jones, 2010), the value of artists’ paintings (Galenson and Weinberg, 2000, 2001; Galenson, 2009; Bayer et al., 2009), performance in sports (Fair, 1994, 2005, 2007; van Ours, 2009; Castellucci et al., 2011; Strittmatter et al., 2020), or the number and quality of completed court cases (Backes-Gellner et al., 2011). These studies often show a very early peak in productivity that probably underlies the popular impression also suggested by Figures 23.1 and 23.2. Although these studies assess productivity quite accurately, they are limited in terms of the range of professions they can feasibly measure. Moreover, these studies typically focus on individual top performers who probably differ from the average worker and a normal work setting. Hence, their relevance is small because in everyday work life the workflow is customized to average rather than to top performance. Studies that focus on average performance in standard jobs often suffer from a precise measurement of productivity. Early exceptions are Lazear (2000) and Shaw and Lazear (2007) who analyze individual output in an auto glass company, measured by the number of windshields installed by each worker during a given day. They focus on learning effects during young ages and find very steep learning curves that then turn into slightly declining trajectories at longer tenures. B¨orsch-Supan and Weiss (2016) and B¨orsch-Supan et al. (2021) observe productivity in Taylorized production processes in the manufacturing and service industry. They try to avoid the conceptual and methodological shortcomings of the previous studies by employing an identification strategy that exploits the day-to-day variation in team composition, while removing 421

Axel B¨orsch-Supan and Matthias Weiss

differences between workers and differences between teams using fixed effects for individualteam pairs. This eliminates the variation responsible for the various selectivity biases that mar many previous estimates of the age-productivity relation. The disadvantage of this method is that the double fixed effects strategy absorbs much of the variation in the data such that the remaining productivity signal is very small and thus easily dominated by the noise in the data. They therefore use very large panel data sets with millions of observations, collected in two German companies. B¨orsch-Supan and Weiss (2016) study age and productivity in a truck assembly plant of an internationally operating automotive company. The authors use a physical productivity measure that pertains to normal workers in a work setting found in many companies, namely errors made in the production of a standardized product manufactured on an assembly line that dictates the time for the given tasks. Their units of observation are work teams, i.e., an aggregation level between the individual and an entire plant. They identify the contribution of a worker to the productivity of the worker’s team by the daily variation in team composition due to holidays and leaves to compensate for overtime. Figure 23.7 shows the number of errors per day and the severity of errors given that an error occurred. The regression equation includes age splines and control variables for worker characteristics (such as sex and nationality) and work environment (such as day of week and temperature). The frequency of errors (Panel 1) exhibits a clearly increasing profile: Older workers make significantly more errors. However, the severity of errors strongly decreases with age (Panel 2). Weighting errors by severity and converting this into a productivity measure yields Figure 23.8. Productivity increases slightly until age 65. It is measured rather precisely up to age 60. The increase becomes insignificant at ages between 60 and 65 years where too few observations exist for precise estimation. A main result of this study that emerges is the lack of evidence for a productivity decline in this assembly plant at least until age 60. B¨orsch-Supan and Weiss interpret these results as follows. Errors are rare. They usually happen in especially tense situations, typically when things go wrong and there is little time to fix them. In these improvisational situations, older, more experienced workers seem to know better which severe errors to avoid. This concentration on the vital tasks, potentially at the cost of some minor errors, gives older workers an advantage in terms of the overall productivity measure, the severity-weighted sum of errors. A detailed study of their regression coefficients reveals

Figure 23.7 Age profile for the number and severity of errors. Figures depict the conditional mean and the two-standard-deviation error bands. Source: B¨orsch-Supan and Weiss (2016).

422

Productivity in an Ageing World

Figure 23.8 Age and productivity in the assembly line of a German automotive plant. Source: B¨orsch-Supan and Weiss (2016).

that the severity of errors is mostly explained by the age variables while the control variables are mostly insignificant. Hence, external conditions do not seem to matter much. Experience is what prevents severe errors. For the number of errors, the opposite is true: Only three of the nine age splines are significant but almost all control variables are. Higher age leads to more errors, but other factors seem more important. The study by B¨orsch-Supan and Weiss confirms the conclusion of Dittmann-Kohli and van der Heijden (1996) that little or no correlation exists between age and productivity. However, while Dittmann-Kohli and van der Heijden’s conclusion was based on the notion that the cognitive and physiological degradation observed in laboratory tests does not apply in most occupational fields, because age-sensitive skills are not decisive for end performance, the conclusion of B¨orsch-Supan and Weiss, based on their team-based approach, is that the different skills of younger and older worker complement each other in a well-managed production environment, such that productivity is managed to be flat across ages. The main limitation of B¨orsch-Supan and Weiss (2016) is its focus on a single plant in auto manufacturing with a relatively homogenous set of tasks. Moreover, while manufacturing jobs may require more physical strength, dexterity, and agility, and therefore provide a setting that is most likely to exhibit a declining age-productivity profile, the manufacturing sector has become less relevant, especially in countries with the most severe demographic ageing. By contrast, the service industry sector is expanding not only in economic importance and share of the labor force, but also in terms of the diversity of tasks performed by different teams (Uppenberg and Strauss, 2010). This diversity of tasks performed is important because we can expect systematic differences in the age composition among different types of teams. Though most jobs in this sector are not physically demanding, technological change and new or changing work tasks may pose related challenges to older employees, potentially resulting in reduced productivity. B¨orsch-Supan et al. (2021) therefore used a similar econometric methodology as B¨orschSupan and Weiss (2016) and apply it to an internationally operating financial company with a wide range of tasks typical for the service industry. On average, they reproduce the finding for the automotive industry: no decline in productivity until the typical retirement age. However, across tasks, they find large heterogeneity and conclude that work content strongly influences the age-productivity relation (Figure 23.9). When dealing with more demanding tasks, productivity increases with age, probably due to experience and high motivation. In turn, frustration 423

Axel B¨orsch-Supan and Matthias Weiss

Figure 23.9 Age-specific productivity in the insurance industry. Solid line: mean, thin lines: 95 percent confidence interval. Source: B¨orsch-Supan et al. (2021).

and tiredness may explain the decreasing profile in those teams that deal with potentially boring routine work. The main lesson, however, is that on average and for most of employees, productivity does not decline until mandatory retirement at age 65 (upper right-hand panel). This result is important because it is representative for a large sector of modern economies. We conclude that microeconomic studies that address the methodological challenges, especially selectivity and endogeneity, find that, on average, age-productivity profiles are essentially flat in standard jobs, while productivity peaks at relatively young ages in top-performance jobs, especially in sports and science. The National Research Council (2012, p. 120), Chapter 6, is more cautious. Based on its survey of the literature available until 2012 and its own calculations, the National Research Council concludes that “there is likely to be a negligible effect of the age composition of the labor force on aggregate productivity over the next two decades,” while cautioning that the evidence is very fragile.

23.4

Ageing and Productivity at the Country Level

As pointed out at the beginning of the Section 23.2.2, micro- and macro-level studies of the link between age and productivity must be carefully distinguished because they use different concepts of age and productivity. Moreover, the micro-level studies take the technology and organization of labor as given (“control it out”), while one of the main aims of macro-level studies is to observe changes in technology and organization as a determinant of intertemporal or cross-national productivity changes. This makes comparing micro- and macro-level estimates, even qualitatively, difficult. We will come back to this point in the concluding section. The relationship between age and productivity plays a crucial role in the controversial discussion about the macroeconomic implications of population ageing (e.g., Bloom et al., 2003; National Research Council, 2012; Lee, 2014; Lee and Mason, 2010). In a Solow-type growth model with a simple production function, such as 424

Productivity in an Ageing World

GDP/N = TFP ∗ F(K/N, L/N),

(1)

GDP per capita is related to total factor productivity times capital input K and labor input L per population N. Population ageing affects GDP per capita first through a decline of the labor force per population, L/N. This direct “labor scarcity effect” (Acemoglu and Restrepo, 2017) may be amplified or attenuated if TFP depends on the age structure of the population. Furthermore, a decline of capital per population, K/N, e.g., due to dissaving of older individuals and declining interest rates, is a third mechanism that may cause lower economic growth and lower labor productivity (Poterba, 2001; B¨orsch-Supan, 2008; Ferrero et al., 2019; R¨ohe and St¨ahler, 2020). On the pessimistic side, Cowen (2011), Eichengreen (2015), Summers (2015), Gordon (2015, 2016), and others see population ageing as a major threat to economic growth and as a new episode of what Hansen (1939) 80 years ago called “secular stagnation.” However, this conclusion may be premature and one step too many because population ageing has so far resulted in an ageing of the workforce but not a shrinking. Hence, population ageing could have had an adverse effect on economic growth only if older workers were less productive than their younger peers. The microeconomic evidence collected in the previous section sheds some doubt on the validity of this necessary condition. Werding (2008) and Vandenbroucke (2020) built calibrated simulation models along these lines by simply assuming declining age-productivity profiles like Skirbekk (2008). Others are more optimistic and argue that productivity increases because of, and not in spite of, population ageing. Kluge et al. (2014), Mokyr (2014), and Glaeser (2014) claim that better education and the effects of new technologies (e.g., robots or artificial intelligence) will dwarf demographic effects. While the theoretical arguments are straightforward, with both sides stressing the role of productivity, empirical evidence on the balance of these arguments is scant. Figure 23.10 shows prima facie evidence. The Organisation for Economic Co-operation and Development (OECD) and EU exhibit similar patterns. In both economic regions, GDP per person employed, a rough measure of labor productivity, exhibits an acceleration of the secular decline after the turn of the millennium. TFP in Europe even turns negative. The demographic support ratio—defined as the share of people of working age in the population, here 15–64 years—slopes steeply downward after about 2005. Hence, when only looking at the time after about 1985, this pattern would be compatible with the explanation that productivity declined most strongly when the working-age population declined. However, this evidence is not causal, and it is not even suggestive, if we also consider the time between 1960 and 1985. Most attempts to establish a causal relationship between population ageing and productivity at the country level employ national accounting data of a panel of countries over time. The foundation for econometric estimation are equations of the following type: ln(GDP/L) = ln(TFP) + α/(1 − α) ∗ ln(K/Y ) + ln(H/L).

(2)

By inserting time and country fixed effects; splitting L and N by age groups; using refined measures of physical and human capital, K and H; and adding other explanatory variables, this approach is very flexible. Separate regressions on the components of the right-hand side that use the same regressors—such as population or labor force composition by age group, dependency ratio, or average age of population or labor force—will produce a set of coefficients that sum to the coefficients of the same regression performed on the left-hand side, i.e., output per worker. This can shed light on the various channels by which output and productivity depend on population ageing. As Section 23.2.2 details, concerns exist about endogeneity due to reverse 425

Axel B¨orsch-Supan and Matthias Weiss

7.0% 60.%

EU

5.0%

19 60 19 66 19 72 19 78 19 84 19 90 19 96 20 02 20 08 20 14 20 20

OECD

OECD

4.0% 3.0% 2.0% 1.0% 0.0%

19 6 19 0 6 19 5 7 19 0 7 19 5 8 19 0 8 19 5 9 19 0 9 20 5 0 20 0 0 20 5 1 20 0 15

EU

TFP growth 4.5% 4% 3.5% 3.0% 2.5% 2.0% 1.5% 1.0% 0.5% 0.0% -0.5%

EU OECD

19 6 19 0 6 19 5 7 19 0 7 19 5 8 19 0 8 19 5 9 19 0 9 20 5 0 20 0 0 20 5 1 20 0 15

Labor productivity growth

Demographic support ratio 68 67 66 65 64 63 62 61 60 59 57

Figure 23.10 Old-age dependency ratio, GDP per person employed, and TFP, 1960–2019. Labor productivity and TFP smoothed. Source: Author calculation based on Penn World Tables Version 10, Feenstra et al. (2015), and World Bank (2021), last updated March 19, 2021.

causality, in particular because labor input is likely to depend on productivity. The literature addresses these concerns by using instruments, typically 10-year lags. Acemoglu and Restrepo (2017) is one of the few papers that finds no negative relationship between GDP per capita and conventional measures of ageing. They argue that directed technical change has compensated for an ageing workforce because demographic change has provided an incentive to invest more in automation, supporting the arguments of Acemoglu at al. (2014), Kluge et al. (2014), Mokyr (2014), and Glaeser (2014). However, Lindh and Malmberg (1999), Feyrer (2007, 2008), Aiyar et al. (2016), Maestas et al. (2016), F¨ollmi et al. (2019), Aksoy et al. (2019), and Kotschy and Sunde (2018) are examples of studies that find evidence supporting the notion that the population age 50 and older contributes negatively to GDP per capita, labor productivity, and/or TFP. The estimation results by Werding (2008) show an inverse U-shape in the relation between TFP and the age composition of the workforce, which confirms K¨ogel (2005), who finds a negative impact of the youth dependency ratio on TFP. Ilmakunnas and Miyakoshi (2013) add an interesting facet by distinguishing low and high skills in their analysis of TFP growth. They find that among the low-skilled the ageing process is a negative driver of productivity, but among the high-skilled it is a positive driver. This fits the more nuanced role of age in productivity set out in the introduction of this review. Employing more information and communication technology (ICT) is generally a positive driver of TFP, but it interacts negatively with ageing in the high-skilled labor category and positively with ageing in the low-skilled labor category. Because ICT may have positive or negative impacts on the productivity contributions of different age/skill groups, the impact of the ageing process is affected by workforce composition and overall ICT investments. Poplawski-Ribeiro (2020) and Gr¨undler and Potrafke (2021) provide some more recent evidence that TFP is a declining function of the old-age dependency ratio, thus amplifying the negative impact of a shrinking labor force on GDP per capita. Keeping the many econometric challenges in mind, the preponderance of evidence points to a negative influence of population ageing on both TFP and aggregate labor productivity. The obvious next question is then: Which mechanism in older economies makes the work environment less productive? Obvious candidates are innovation and entrepreneurship as the main channels for technological change. Feyrer (2007) and Aiyar et al. (2016) relate TFP growth on measures of public and private research and development (R&D) spending, patents, and similar measures using formal 426

Productivity in an Ageing World

regressions and informal correlation analyses. Gr¨undler and Potrafke (2021) regress the number of patents and the number of scientific journal articles on the dependency ratio and find a negative correlation. Unfortunately, this evidence is suggestive at best due to reverse causality and related identification problems. Many papers refer to Jones (2010) and the related literature reviewed in Section 23.3, which finds solid evidence that scientists and other top performers are most productive in their 30s, concluding that population ageing with its lower share of young and mid-age citizens, who are likely to develop path-breaking innovations, leads to less innovation activities. However, as pointed out, how relevant these observations are is unclear. The automotive study by B¨orsch-Supan and Weiss (2016) has data on small-scale process innovations by age of employee. Their estimated age-innovation profile is flat and precisely measured. This finding is in line with a much older study by Farr et al. (1998), who used a quasi-experimental setting. Due to a sales crisis in the automotive industry in the 1980s, the workforce at the Ford works in Detroit was reduced, with mostly younger workers being laid off. The average age of the workforce rose from 37.2 to 44.5 years. According to this study, older employees were just as creative and decisive as younger workers, took part in the same qualification programs as younger workers, and had the same level of success. Long-run historical evidence also points in this direction. Using a panel of 21 OECD countries over the period 1870–2009, Ang and Madsen (2015) show that highly educated workers are more innovative and that their propensity to innovate increases (!) sharply with age. With a view on developing countries, this view is also echoed by Kotschy et al. (2020), who point out the importance of interactions among productivity, ageing, and education.4 Evidence on the role of entrepreneurship is similarly scant and suggestive. Gr¨undler and Potrafke (2021) provid an overview and show that the number of entrepreneurs and selfemployed persons in a country is negatively related to the country’s old-age dependency ratio. The general notion behind this finding is that older individuals are less likely to start a new business (Azoulay et al., 2020). One channel is that the occupational choice between becoming an entrepreneur and being employed depends on the degree of risk tolerance (Doepke and Zilibotti, 2014). Moreover, evidence largely from laboratory experiments indicates that older individuals are more risk averse than younger individuals (Dohmen et al., 2011; Falk et al., 2018). Hence, fewer individuals will choose to start a new business in ageing populations. A second channel works through the interest rate. B¨orsch-Supan et al. (2006, 2019) and R¨ohe and St¨ahler (2020) use macro models with a life-cycle structure to show that population ageing will lower interest rates, mainly based on the underlying life-cycle mechanism of dissaving in old age, which reduces capital supply. A lower interest rate then reduces competition in the product market in favor of established firms because low interest rates facilitate refinancing assets. Unproductive firms are thus less likely to close, and new firms with greater innovation potential than established firms do not enter markets (Coad et al., 2016). Both channels imply that population ageing reduces factor productivity by decreasing firms’ fluctuations in product markets. Hopenhayn et al. (2018) and Karahan et al. (2019) provide formal models for this mechanism. Another channel through which population ageing affects innovation and entrepreneurship and thus productivity works through government expenditures. Innovation and entrepreneurship are often subsidized or enjoy tax-preferred treatment due to their positive externalities. However, governments in ageing populations are typically forced to spend large parts of their budgets on pensions, healthcare, and long-term care, thereby potentially crowding out spending on R&D that supports innovation and entrepreneurship. More generally, a large literature suggests that expenditures on the younger generation, including education and family support, are smaller in ageing societies (e.g., Poterba, 1998; Boeri et al., 2001, 2002; Breyer and Stolte, 2001; Galasso and Profeta, 2002, 2007; Razin et al., 427

Axel B¨orsch-Supan and Matthias Weiss

2002; Shelton, 2008; Meier and Werding, 2010; Potrafke, 2011; B¨orsch-Supan, 2012; Gr¨undler and K¨ollner, 2017). This evidence is not watertight as the discussion by Tepe and Vanhuysse (2009) points out. This mechanism is also unlikely to explain the current productivity slowdown because current workers grew up while the dependency ratio was still increasing. Nevertheless, this public expenditure channel with its crowding-out mechanism is likely to depress future productivity.

23.5

Summary and Conclusions

The relationship between age and productivity is more complex than widespread prejudices suggest and all too lightly taken as an underlying assumption in the literature. It is complex because agility, speed, and fluid intelligence decline in human bodies, while experience, political skills, and the size and quality of networks increase with age and crystalline intelligence remains up to very old ages. The age at which this combination of human characteristics peaks in terms of productivity depends on the work environment and the nature of the job. Moreover, even in a given work environment and for a given range of tasks, the heterogeneity across individuals is large. Shedding light on this complex relationship is difficult due to the measurement challenges discussed in Section 23.2. Distinguishing the link between the age and labor productivity of single individuals from the link between population ageing and aggregate productivity at the country level is also crucial. We first take stock of the micro-level results. Studies based on superiors’ assessments and age-wage profiles tend to be biased. Studies that employ direct measures of individual productivity typically describe top-performance jobs, e.g., the number and quality of publications in academic research, Nobel prizes, the value of artists’ paintings, or performance in sports and chess. They find that productivity peaks early, often in the 30s. While these studies can measure productivity quite precisely, their relevance is small because in everyday work life the workflow is customized to average rather than to top performance. Average performance is better measured in plant-level studies, which need to account for the endogeneity of the age distribution of employees. The methodologically most convincing studies use sophisticated econometric methods and can overcome these methodological problems. These studies find that an initial increase in productivity, probably a learning effect, but then productivity remains flat until retirement. The limitation of micro-level studies is that they measure productivity over time in a given technological environment and therefore ignore the contribution to productivity from improving this environment. This is where macro studies come into play. In contrast to the micro-level studies, the preponderance of evidence provided by macro-level studies points to a negative influence of population ageing on both TFP and labor productivity. Unfortunately, this evidence suffers from reverse causality between TFP or aggregate labor productivity on the one hand and population ageing on the other hand. To make matters worse, these aggregates are slow moving and cannot be easily instrumented by lags. Most macroeconomic evidence is therefore more suggestive than causal. Nevertheless, the agreement across the many existing studies is large. The microeconomic and the macroeconomic evidence appear to contradict each other. This is not the case. The microeconomic studies do not directly translate to the aggregate level because they measure the productivity of older workers relative to younger workers in a technological and organizational environment that is fixed either by design or by controlling for it. On the one hand, this is necessary for the research question of the micro-level studies: Is it the age of the observed individual that makes this individual productive? On the other hand, these studies ignore the contribution to productivity by improving this environment, which is

428

Productivity in an Ageing World

exactly the focus of most macroeconomic studies. This difference makes comparing micro- and macro-level estimates, even only qualitatively, difficult. Technological progress is widely considered the main source of long-run economic growth (Galor and Tsiddon, 1997; Galor and Weil, 2000; Hsieh et al., 2011; Mokyr et al., 2015). Channels are innovation and entrepreneurship. On a deeper level of causality, these may be driven by a higher appetite for risk of younger individuals; by a higher capital-to-worker ratio in younger populations as older individuals are assumed to dissave and ageing economies tend to have lower interest rates; and by higher public spending on R&D in younger economies because such spending is not crowded out by spending on pensions, healthcare, and long-term care. All these mechanisms are plausible and can be found in many regression variants that are not causal but suggestive. Several policy conclusions can be drawn. First, contrary to the popular impression, there is no decline of individual productivity in many standardized jobs in the manufacturing and service sectors that would make an increase in the retirement age a self-defeating proposal. While little is known about productivity at ages well beyond the current normal retirement age due to the lack of observations, no evidence exists for a decline until then. However, the trade-off between agility and experience rests highly on the value of job-specific experience. Hence, keeping older workers is important because job changes at old age may lead to a productivity decline due to lost experience. Moreover, seniority wages have harmful side effects. Even with perfectly flat productivity profiles, seniority wages make hiring younger workers more attractive than keeping older ones. Second, while government spending patterns are unlikely to have resulted in the productivity decline since the turn of the millennium, a short-sighted spending policy that crowds out education in favor of pensions is a self-defeating policy and a prescription for long-term harm to economic growth through higher productivity, which in turn is needed to finance the welfare state. Third, pension reforms that attempt to save costs by indexing pension benefits to prices rather than wages, counting on large productivity growth rates in the future, are a dangerous gamble because the preponderance of arguably weak evidence speaks against it. While we have learned much about the age-productivity profile of top- and average-performing individuals at the micro level, resolving the endogeneity problems at the macro level seems to promise little success because instruments are hard to find. Hence, future research on the relationship between age and productivity should focus on the microeconomics and microeconometrics of innovation, entrepreneurship, and how innovations change the day-to-day work environment of average workers.

Notes 1 Skirbekk (2008) shows the measurement problem particularly well. Skirbekk describes the physical and cognitive capabilities of a person with great care using well and reliably measured indicators such as grades in math, finger dexterity, and eye-hand-foot coordination while he assumes a purely hypothetical stepwise declining function to describe the life-course trajectory of experience. His conclusion (“productivity peaks for the 35–44-year age group”) is thus not based on evidence. The seminal paper by Altig et al. (2001) similarly postulates rather than estimates an inverse U-shaped age-productivity profile. 2 A large literature exists on how to identify age, cohort, and period effects, especially how to break the linear dependency by various assumptions, with applications, e.g., on life-cycle saving patterns; cf. the review by B¨orsch-Supan (2001) and the approaches by Attanasio (1994), Deaton and Paxson (1994), Alessie et al. (2005), and Heo et al. (2017).

429

Axel B¨orsch-Supan and Matthias Weiss

3 Gr¨undler and Potrafke (2021) use as an alternative the ratio of retirement age to average age. However, most of the variance of this instrument comes from average age, thus it fails to be exogenous. 4 The paper cites G¨obel and Zwick (2013) and B¨orsch-Supan and Weiss (2016) as indicating that there is a “peak in life-cycle productivity around age 40 to 54.” This is not correct, see the section “Age and Productivity at the Micro Level.”

References ACEMOGLU, D., DORN, D., HANSON, G. H., PRICE, B., ET AL. (2014): “Return of the Solow paradox? IT, productivity, and employment in US manufacturing,” American Economic Review, 104(5): 394–399. ACEMOGLU, D., AND RESTREPO, P. (2017): “Secular stagnation? The effect of ageing on economic growth in the age of automation,” American Economic Review, 107(5): 174–179. AIYAR, S., EBEKE, C., AND SHAO, X. (2016): “The impact of workforce ageing on European productivity,” IMF Working Paper 16/238, Washington, DC. AKSOY, Y., BASSO, H. S., SMITH, R. P., AND GRASL, T. (2019): “Demographic structure and macroeconomic trends,” American Economic Journal: Macroeconomics, 11(1): 193–222. ALESSIE R. J. M., KAPTEYN, A., AND LUSARDI, A. (2005): “Explaining the wealth holdings of different cohorts: Productivity growth and social security,” European Economic Review, 49(2005): 1361–1391. ALTIG, D., AUERBACH, A. J., KOTLIKOFF, L. J., SMETTERS, K. A., AND WALLISER, J. (2001): “Simulating fundamental tax reform in the United States,” American Economic Review, 3: 574–595. ANG, J. B., AND MADSEN, J. B. (2015): “Imitation versus innovation in an ageing society: International evidence since 1870,” Journal of Population Economics, 28(2): 299–327. ARELLANO, M., AND BOND, S. (1991): “Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations,” Review of Economic Studies, 58(2): 277. ARELLANO, M., AND BOVER, O. (1995): “Another look at the instrumental variable estimation of errorcomponents models,” Journal of Econometrics, 68(1): 29–51. ATTANASIO, O. P. (1994): “Personal saving in the United States.” In: Poterba, J. (ed.), International Comparisons of Household Savings, Chicago, London: University of Chicago Press, pp. 57–124. AUBERT, P. (2003): “Productivity, wage and demand for elder workers; an examination on French matched employer-employee data,” Working Paper INSEE, Paris. AUBERT, P., AND CR E´ PON, B. (2007): “Are older workers less productive? Firm-level evidence on ageproductivity and age-wage profiles,” Mimeo, INSEE, Paris. AVOLIO, B., AND WALDMAN, D. (1994): “Variations in cognitive, perceptual, and psychomotor abilities across the working life span: Examining the effects of race, sex, experience, education, and occupational type,” Psychology and Ageing, 9(3): 430–442. AZOULAY, P., JONES, B., KIM, J. D., AND MIRANDA, J. (2020): “Age and high-growth entrepreneurship,” American Economic Review: Insights, 2(1): 65–81. BACKES-GELLNER, U., SCHNEIDER, M. R., AND VEEN, S. (2011): “Effect of workforce age on quantitative and qualitative organizational performance: Conceptual framework and case study evidence,” Organization Studies, 32(8): 1103–1121. BAYER, T., PAGE, J., RAVIV, Y., AND ROSETT, J. (2009): “Age, human capital and the quality of work: New evidence from old masters,” Robert Day School Working Paper No. 2009-04. Claremont, CA. BLOOM, D., CANNING, D., AND SEVILLA, J. (2003): “The demographic dividend: A new perspective on the economic consequences of population change,” RAND Corporation. Santa Monica, CA. BLOOM, D., AND SOUSA-POZA, A. (2013): “Ageing and productivity: Introduction,” Labour Economics, 22: 1–4. BLUNDELL, R., AND BOND, S. (1998): “Initial conditions and moment restrictions in dynamic panel data models,” Journal of Econometrics, 87(1): 115–143. ¨ BOERI, T., BORSCH -SUPAN, A., AND TABELLINI, G. (2001): “Would you like to shrink the welfare state? The opinions of European citizens,” Economic Policy, 32: 9–50. ¨ BOERI, T., BORSCH -SUPAN, A., AND TABELLINI, G. (2002): “Would you like to reform the pension system? The opinions of European citizens,” American Economic Review, 92: 396–401. BOND, S. R. (2002): “Dynamic panel data models: A guide to micro data methods and practice,” Portuguese Economic Journal, 1: 141–162. ¨ BORSCH -SUPAN, A. (1998): “Capital’s contribution to productivity and the nature of competition,” Brookings Papers on Economic Activity: Macroeconomics, 1998: 205–248.

430

Productivity in an Ageing World ¨ BORSCH -SUPAN, A. (2001): “International comparison of household savings behavior: A study of lifecycle savings in seven countries,” Research in Economics, 55(2): 133–134. ¨ BORSCH -SUPAN, A. (2008): “The impact of global ageing on labor, product and capital markets.” In: Prskawetz, A., Bloom, D. E., and Lutz, W. (eds.), Population Ageing, Human Capital Accumulation, and Productivity Growth Population and Development Review, 34 (Supp.): 52–77. ¨ BORSCH -SUPAN, A. (2012): “Entitlement reforms in Europe: Policy mixes in the current pension reform process,” NBER Working Paper No. 18009. Available at http://www.nber.org/papers/w18009. ¨ BORSCH -SUPAN, A., DUEZGUEN, I., AND WEISS, M. (2008): “Labor productivity in an ageing society.” In: Broeders, D., Eijiffinger, S., and Houben, A. (eds.), Frontiers in Pension Finance and Reform, Cheltenham: Edward Elgar, pp. 83–96. ¨ BORSCH -SUPAN, A., HUNKLER, C., AND WEISS, M. (2021): “Big data at work: Age and labor productivity in the service sector,” Journal of the Economics of Ageing, https://doi.org/10.1016/ j.jeoa.2021.100319. ¨ BORSCH -SUPAN, A., LEITE, S., NUNO, D., AND RAUSCH, J. (2019): “Demographic changes, migration and economic growth in the Euro area.” In: European Central Bank (ed.), 20 Years of European Economic and Monetary Union, Frankfurt: European Central Bank, pp. 193–230. ¨ BORSCH -SUPAN, A., LUDWIG, A., AND WINTER, J. (2006): “Ageing, pension reform, and capital flows: A multi-country simulation model,” Economica, 73: 625–658. ¨ BORSCH -SUPAN, A., AND WEISS, M. (2016): “Productivity and age: Evidence from work teams at the assembly line,” Journal of the Economics of Ageing, 7: 30–42. BREYER, F., AND STOLTE, K. (2001): “Demographic change, endogenous labor supply and the political feasibility of pension reform,” Journal of Population Economics, 14(3): 409–424. CASTELLUCCI, F., PADULA, M., AND PICA, G. (2011): “The age-productivity gradient: Evidence from a sample of F1 drivers,” Labour Economics, 18(4): 464–473. COAD, A., SEGARRA, A., AND TERUEL, M. (2016): “Innovation and firm growth: Does firm age play a role?,” Research Policy, 45(2): 387–400. COWEN, T. (2011): The Great Stagnation: How America Ate All the Low-Hanging Fruit of Modern History, Got Sick, and Will (Eventually) Feel Better: A Penguin eSpecial from Dutton. New York City: Penguin. CR E´ PON, B., DENIAU, N., AND PE´ REZ-DUARTE, S. (2003): “Wages, productivity, and workers characteristics: A French perspective,” CREST Document de Travail No. 2003-04. Centre de Recherche en ´ Economie et Statistique, Paris. DAVERI, F., AND MALIRANTA, M. (2007): “Age, seniority and labour costs: Lessons from the Finnish IT revolution,” Economic Policy, 22(49): 117–175. DEATON, A., AND PAXSON, C. (1994): “Saving, ageing and growth in Taiwan.” In: Wise, D. (ed.), Studies in the Economics of Ageing, Chicago: University of Chicago Press, pp. 331–364. DITTMANN-KOHLI, F., AND VAN DER HEIJDEN, B. (1996): “Leistungsf¨ahigkeit a¨ lterer Arbeitnehmer—interne und externe Faktoren,” Zeitschrift f¨ur Gerontologie, 29: 323–327. DOEPKE, M., AND ZILIBOTTI, F. (2014): “Culture, entrepreneurship, and growth.” In: Aghion, P., and Durlauf, S. N. (eds.), Handbook of Economic Growth, Volume 2. Amsterdam: Elsevier, pp. 1–48. DOHMEN, T., FALK, A., HUFFMAN, D., SUNDE, U., SCHUPP, J., AND WAGNER, G. G. (2011): “Individual risk attitudes: Measurement, determinants, and behavioral consequences,” Journal of the European Economic Association, 9(3): 522–550. DOSTIE, B. (2006): “Wages, productivity, and ageing,” IZA Discussion Paper No. 2496. Institut zur Zukunft der Arbeit, Bonn, Germany. DOSTIE, B. (2011): “Wages, productivity, and ageing,” De Economist, 159(2): 139–158. EICHENGREEN, B. (2015): “Secular stagnation: The long view,” American Economic Review, 105(5): 66–70. EUROPEAN COMMISSION. (2012): Eurostat-OECD Methodological Manual on Purchasing Power Parities, Luxembourg: Publications Office of the European Union. FAIR, R. (1994): “How fast do old men slow down?,” The Review of Economics and Statistics, 76(1): 103–118. FAIR, R. (2005): Estimated Age Effects in Baseball. Mimeo. Cowles Foundation, Yale University. Available at http://fairmodel.econ.yale.edu/. FAIR, R. (2007): “Estimated age effects in athletic events and chess,” Experimental Ageing Research, 33: 37–57. FALK, A., BECKER, A., DOHMEN, T., ENKE, B., HUFFMAN, D., AND SUNDE, U. (2018): “Global evidence on economic preferences,” Quarterly Journal of Economics, 133(4): 1645–1692. FARR, J. L., TESLUK, P. E., AND KLEIN, S. R. (1998): “Organizational structure of the workplace and the older worker.” In: Schaie, K. W., and Schooler, C. (eds.), Impact of Work on Older Adults, New York:

431

Axel B¨orsch-Supan and Matthias Weiss

Springer Publications, pp. 143–185. FEENSTRA, R. C., INKLAAR, R., AND TIMMER, M. P. (2015): “The next generation of the Penn World Table,” American Economic Review, 105(10): 3150–3182. FERRERO, G., GROSS, M., AND NERI, S. (2019): “On secular stagnation and low interest rates: Demography matters,” International Finance, 22(3): 262–278. FEYRER, J. (2007): “Demographics and productivity,” Review of Economics and Statistics, 89(1): 100–109. FEYRER, J. (2008): “Aggregate evidence on the link between age structure and productivity.” In: Prskawetz, A., Bloom, D. E., and Lutz, W. (eds.), Population Ageing, Human Capital Accumulation, and Productivity Growth Population and Development Review, 34(Supp.): 78–102. ¨ ¨ FOLLMI , R., SCHMIDT, T., AND JAGER , P. (2019): “Demographischer Wandel und dessen Auswirkungen auf die Produktivit¨at und das Wirtschaftswachstum,” Grundlagen f¨ur die Wirtschaftspolitik Nr. 1. Staatssekretariat f¨ur die Wirtschaft SECO, Bern. FRANK, R. H., AND HUTCHENS, R. M. (1993): “Wages, seniority, and the demand for rising consumption profiles,” Journal of Economic Behavior and Organization, 21(3): 251–276. GALASSO, V., AND PROFETA, P. (2002): “The political economy of social security: A survey,” European Journal of Political Economy, 18(1): 1–29. GALASSO, V., AND PROFETA, P. (2007): “How does ageing affect the welfare state?,” European Journal of Political Economy, 33: 554–563. GALENSON, D. W. (2009): “The greatest artists of the twentieth century.” In: Galenson, D. W. (ed.), Conceptual Revolutions in Twentieth-Century Art, Cambridge: Cambridge University Press, pp. 30–61. GALENSON, D. W., AND WEINBERG, B. A. (2000): “Age and the quality of work: The case of modern American painters,” Journal of Political Economy, 108(4): 671–777. GALENSON, D. W., AND WEINBERG, B. A. (2001): “Creating modern art: The changing careers of painters in France from impressionism to cubism,” The American Economic Review, 91(4): 1063–1071. GALOR, O., AND TSIDDON, D. (1997): “Technological progress, mobility, and economic growth,” American Economic Review, 87(3): 363–382. GALOR, O., AND WEIL, D. N. (2000): “Population, technology, and growth: From Malthusian stagnation to the demographic transition and beyond,” American Economic Review, 90(4): 806–828. GELDERBLOM, A. (2006): “The relationship of age with productivity and wages.” In: Ageing and Employment: Identification of Good Practice to Increase Job Opportunities and Maintain Older Workers in Employment, Final Report to the European Commission, DG Employment. Available at http://ec.europa.eu/social/BlobServlet?docId=2073&langId=en. GLAESER, E. L. (2014): “Secular joblessness.” In: Teulings, C., and Baldwin, R. (eds.), Secular Stagnation: Facts, Causes, and Cures, London: CEPR Press, pp. 69–80. ¨ GOBEL , C., AND ZWICK, T. (2009): “Age and productivity: Evidence from linked employer employee data,” De Economist, 160: 35–57. ¨ GOBEL , C., AND ZWICK, T. (2012): “Age and productivity: Sector differences,” De Economist, 160: 35–57. ¨ GOBEL , C., AND ZWICK, T. (2013): “Which personnel measures are effective in increasing productivity of old workers?,” Labour Economics, 22: 80–93. GORDON, R. J. (2015): “Secular stagnation: A supply-side view,” American Economic Review, 105(5): 54–59. GORDON, R. J. (2016): The Rise and Fall of American Growth: The U.S. Standard of Living since the Civil War, Princeton, NJ: Princeton University Press. ˚ -NIELSEN, N. (2008): “Age structure of the workforce and firm GRUND, C., AND WESTERG ARD performance,” International Journal of Manpower, 29(5): 410–422. ¨ ¨ GR UNDLER , K., AND KOLLNER , S. (2017): “Determinants of governmental redistribution: Income distribution, development levels, and the role of perceptions,” Journal of Comparative Economics, 45(4): 930–962. ¨ GR UNDLER , K., AND POTRAFKE, N. (2021): Ageing and Productivity: Causal Evidence and Transmission Mechanisms. Mimeo. CESifo, Munich. HÆGELAND, T., AND KLETTE, T. J. (1999): “Do higher wages reflect higher productivity? Education, gender and experience premiums in a matched plant-worker data set.” In: Haltiwanger, J., Lane, J., Spletzer, J. R., Theeuwes, J., and Troske, K. (eds.), The Creation and Analysis of Employer-Employee Matched Data, Amsterdam: North Holland, pp. 231–259. HALTIWANGER, J. C., LANE, J. I., AND SPLETZER, J. R. (1999): “Productivity differences across employers: The roles of employer size, age, and human capital,” The American Economic Review, 89(2): 94–98. HALTIWANGER, J. C., LANE, J. I., AND SPLETZER, J. R. (2007): “Wages, productivity, and the dynamic interaction of business and workers,” Labour Economics, 14: 575–602.

432

Productivity in an Ageing World

HANSEN, A. H. (1939): “Economic progress and declining population growth,” American Economic Review, 29(1): 1–15. HELLERSTEIN, J. K., AND NEUMARK, D. (1995): “Are earnings profiles steeper than productivity profiles? Evidence from Israeli firm-level data,” Journal of Human Resources, 30(1): 89–112. HELLERSTEIN, J. K., AND NEUMARK, D. (2007): “Production function and wage equation estimation with heterogeneous labor: Evidence from a new matched employer-employee data set.” In: Hulten, C., and Berndt, E. (eds.), Hard-to-Measure Goods and Services: Essays in Memory of Zvi Griliches, Chicago: University of Chicago Press, pp. 31–71. HELLERSTEIN, J. K., NEUMARK, D., AND TROSKE, K. R. (1999): “Wages, productivity, and worker characteristics: Evidence from plant level production function and wage equations,” Journal of Labor Economics, 17(3): 409–446. HEO, J., JEON, S.-Y., OH, C.-M., HWANG, J., OH, J., AND CHO, Y. (2017): “The unrealized potential: Cohort effects and age-period-cohort analysis,” Epidemiology and Health, 39, https://doi. org/10.4178/epih.e2017056. HOPENHAYN, H., NEIRA, J., AND SINGHANIA, R. (2018): “From population growth to firm demographics: Implications for concentration,” NBER Working Paper No. 25382. National Bureau of Economic Research, Cambridge, MA. HSIEH, C. T., AND KLENOW, P. J (2011): “Development accounting,” American Economic Journal: Macroeconomics, 2(1): 207–223. HUBER, A. (2002): Strategien zur alternsgerechten Gestaltung von Gruppenarbeit—Gesundheitsf¨orderung und Qualifizierung, Stuttgart: Frauenhofer IRB Verlag. HUNTER, J. E., AND HUNTER, R. F. (1984): “Validity and utility of alternative predictors of job performance,” Psychological Bulletin, 96: 72–98. ILMAKUNNAS, P., AND MALIRANTA, M. (2005): “Technology, labour characteristics and wageproductivity gaps,” Oxford Bulletin of Economics and Statistics, 67(5): 623–645. ILMAKUNNAS, P., AND MALIRANTA, M. (2007): “Ageing, labor turnover, and firm performance,” ETLA Discussion Paper No. 1092. Available at http://www.etla.fi/. ILMAKUNNAS, P., AND MIYAKOSHI, T. (2013): “What are the drivers of TFP in the ageing economy? Ageing labor and ICT capital,” Journal of Comparative Economics, 41(1): 201–211 ILMARINEN, J. (1999): Ageing Workers in the European Union—Status and Promotion of Work Ability, Employability and Employment. Helsinki: Finnish Institute of Occupational Health. JONES, B. (2010): “Age and great innovation,” Review of Economics and Statistics, 92(1): 1–14. JONES, B. F., AND WEINBERG, B. (2011): “Age dynamics in scientific creativity,” Proceedings of the National Academy of Sciences, 108(47): 18855–19096. KARAHAN, F., PUGSLEY, B., AND SAHIN, A. (2019): “Demographic origins of the startup deficit,” NBER Working Paper No. 25374. National Bureau of Economic Research, Cambridge, MA. KLUGE, F., ZAGHENI, E., LOICHINGER, E., AND VOGT, T. (2014): “The advantages of demographic change after the wave: Fewer and older, but healthier, greener, and more productive?,” PLOS ONE, 9(9): 1–11. ¨ KOGEL , T. (2005): “Youth dependency and total factor productivity,” Journal of Development Economics, 76(1): 147–173. KOTLIKOFF, L. J., AND GOKHALE, J. (1992): “Estimating a firm’s age-productivity profile using the present value of workers’ earnings,” Quarterly Journal of Economics, 107(4): 1215–1242. KOTLIKOFF, L. J., AND WISE, D. (1989): “Employee retirement and a firm’s pension plan.” In: Wise, D. (ed.), The Economics of Ageing, Chicago: University of Chicago Press, pp. 279–334. KOTSCHY, R., SUAREZ URTAZA, P., AND SUNDE, U. (2020): “The demographic dividend is more than an education dividend,” PNAS, 117(42): 25982–25984. KOTSCHY, R., AND SUNDE, U. (2018): “Can education compensate the effect of population ageing on macroeconomic performance?,” Economic Policy, 33(96): 587–634. KUTSCHER, R. E., AND WALKER, J. F. (1960): “Comparative job performance of office workers by age,” Monthly Labor Review, 83(1): 39–43. LAITNER, J., AND STOLYAROV, D. (2005): Technological Progress and Worker Productivity at Different Ages. Mimeo. University of Michigan. LALLEMAND, T., AND RYCKX, F. (2009): “Are older workers harmful for firm productivity?,” De Economist, 157(3): 273–292. LAZEAR, E. P. (1979): “Why is there mandatory retirement?,” Journal of Political Economy, 87(6): 1261– 1284.

433

Axel B¨orsch-Supan and Matthias Weiss

LAZEAR, E. P. (1981): “Agency, earnings profiles, productivity, and hours restrictions,” American Economic Review, 71(4): 606–620. LAZEAR, E. P. (1995): Personnel Economics, Cambridge, MA: MIT Press. LAZEAR, E. P. (2000): “Performance pay and productivity,” American Economic Review, 90(5): 1346–1361. LEE, R. D. (2014): “Macroeconomic consequences of population ageing in the United States: Overview of a national academy report,” American Economic Review, 104(5): 234–239. LEE, R. D., AND MASON, A. (2010): “Some macroeconomic aspects of global population ageing,” Demography, 47: 151–172. LEHR, U. (2000): Psychologie des Alterns. Wiebelsheim: Quelle & Mayer. LINDH, T., AND MALMBERG, B. (1999): “Age structure effects and growth in the OECD, 1950–1990,” Journal of Population Economics, 12: 431–449. LOEWENSTEIN, G., AND SICHERMAN, N. (1991): “Do workers prefer increasing wage profiles?,” Journal of Labor Economics, 9(1): 67–84. MADDISON, A. (2001):The World Economy: A Millennial Perspective, Paris: OECD Development Center Studies. MAERCKER, A. (1992): “Weisheit im Alter [Wisdom in Old Age],” M¨unch. Med. Wschr. 134(33): 518–522. MAESTAS, N., MULLEN, K., AND POWELL, D. (2016): “The effect of population ageing on economic growth, the labor force and productivity,” NBER Working Paper No. 22452. National Bureau of Economic Research, Cambridge, MA. MAHLBERG, B., FREUND, I., CRESPO CUARESMA, J., AND PRSKAWETZ, A. (2013): “Ageing, productivity and wages in Austria,” Labour Economics, 22: 5–15. MAIER, G. (1998): “Formen des Erlebens der Arbeitssituation: ein Beitrag zur Innovationsf¨ahigkeit a¨ lterer Arbeitnehmer,” Zeitschrift f¨ur Gerontologie, 31: 127–137. MALMBERG, B., LINDH, T., AND HALVARSSON, M. (2008): “Productivity consequences at the plant level of work-force ageing: Stagnation or a Horndal effect?,” Population and Development Review, 34: 238–256. MARK, J. A. (1957): “Comparative job performance by age,” Monthly Labor Review, 80: 1467–1471. MAS, A., AND MORETTI, E. (2009): “Peers at work,” American Economic Review, 99(1): 112–145. MCEVOY, G. M., AND CASCIO, W. F. (1989): “Cumulative evidence of the relationship between employee age and job performance,” Journal of Applied Psychology, 74(1): 11–17. MEDOFF, J. L., AND ABRAHAM, K. G. (1980): “Experience, performance, and earnings,” Quarterly Journal of Economics, 95(4): 703–736. MEIER, V., AND WERDING, M. (2010): “Ageing and the welfare state: Securing sustainability,” Oxford Review of Economic Policy, 26(4): 655–673. MOKYR, J. (2014): “Secular stagnation? Not in your life.” In: Teulings, C., and Baldwin, R. (eds.), Secular Stagnation: Facts, Causes, and Cures, London: CEPR Press, pp. 83–89. MOKYR, J., VICKERS, C., AND ZIEBARTH, N. L. (2015): “The history of technological anxiety and the future of economic growth: Is this time different?,” Journal of Economic Perspectives, 29(3): 31–50. NATIONAL RESEARCH COUNCIL. (2012): Ageing and the Macroeconomy: Long-Term Implications of an Older Population, Washington, DC: National Academies Press. NG, T. W. H., AND FELDMAN, D. C. (2008): “The relationship of age to ten dimensions of job performance,” Journal of Applied Psychology, 93(2): 392–423. ¨ ¨ NYBERG, L., LOVDEN , M., RIKLUND, K., LINDENBERGER, U., AND BACKMAN , L. (2012): “Memory ageing and brain maintenance,” Trends in Cognitive Sciences, 16(5): 292–305. OSTER, S. M., AND HAMERMESH, D. S. (1998): “Ageing and productivity among economists,” Review of Economics and Statistics, 80(1): 154–156. PARK, D., AND BISCHOF, G. N. (2013): “The ageing mind: Neuroplasticity in response to cognitive training,” Dialogues in Clinical Neuroscience, 15: 109–119. POPLAWSKY-RIBEIRO, M. (2020): “Labour force ageing and productivity growth,” Applied Economics Letters, 27(6): 498–502. POTERBA, J. M. (1998): “Demographic change, intergenerational linkages, and public education,” American Economic Review, 88(2): 315–320. POTERBA, J. M. (2001): “Demographic structure and asset returns,” Review of Economics and Statistics, 83(4): 565–584.

434

Productivity in an Ageing World

POTRAFKE, N. (2011): “Does government ideology influence budget composition? Empirical evidence from OECD countries,” Economics of Governance, 12: 101–134. PROUET, E., AND ROUSSELON, J. (2018): Senior Citizens, Employment and Retirement. Paris: France Strategie. PRSKAWETZ, A., BLOOM, D. E., AND LUTZ, W., EDS. (2008): “Population ageing, human capital accumulation, and productivity growth,” Population and Development Review, 34(Supp): 78–102. PRSKAWETZ, A., MAHLBERG, B., SKIRBEKK, V., FREUND, I., WINKLER-DWORAK, M., LINDH, T., ¨ , O. S., AND ANDERSSON, F. (2006): The Impact of MALMBERG, B., JANS, A.-C., NORDSTR OM Population Ageing on Innovation and Productivity Growth in Europe, Research Report 28. Wien, Austria: Vienna Institute of Demography, Austrian Academy of Sciences. QUIMET, P., AND ZARUTSKIE, R. (2013): “Who works for startups? The relation between firm age, employee age, and growth,” FEDS Working Paper 2013-75. Available at https://papers.ssrn.com/sol3/ papers.cfm?abstract id=2976876 [accessed on May 4, 2020]. RAZIN, A., SADKA, E., AND SWAGEL, P. (2002): “The ageing population and the size of the welfare state,” Journal of Political Economy, 110: 900–918. ¨ , O., AND ST AHLER ¨ ROHE , N. (2020): “Demographics and the decline in firm entry: Lessons from a life-cycle model,” Discussion Paper Nr. 15, Frankfurt: Deutsche Bundesbank. SALTHOUSE, T. A., AND MAURER, T. J. (1996): “Ageing, job performance, and career development.” In: Birren, J. E., and Schaie, K. W. (eds.), Handbook of the Psychology of Ageing, 4th Edition. New York: Kluwer Academic Publishers, pp. 353–364. SCHMIDT, R. F., THEWS, G., AND LANG, F. (2000): Physiologie des Menschen Human Physiology, BerlinHeidelberg: Springer. SCHNEIDER, H., AND STEIN, D. (2006): “Personalpolitische Strategien Deutscher Unternehmen zur Bew¨altigung demografisch bedingter Rekrutierungsengp¨asse bei F¨uhrungskr¨aften.” Research Report 6. IZA, Bonn, Germany. SCHNEIDER, L. (2007): “Mit 55 zum alten Eisen? Eine Analyse des Alterseinflusses auf die Produktivit¨at anhand des LIAB,” Zeitschrift f¨ur ArbeitsmarktForschung, 40(1): 77–97. SHAW, K. L., AND LAZEAR, E. P. (2007): “Tenure and output.” NBER Working Paper No. w13652. Available at SSRN: https://ssrn.com/abstract=1043342. SHELTON, C. (2008): “The ageing population and the size of the welfare state: Is there a puzzle?,” Journal of Public Economics, 92: 647–651. SKIRBEKK, V. (2004): “Age and individual productivity: A literature survey.” In: Feichtinger, G. (ed.), ¨ Vienna Yearbook of Population Research, Vienna: Verlag der Osterreichischen Akademie der Wissenschaften, pp. 133–159. SKIRBEKK, V. (2008): “Age and productivity capacity: Descriptions, causes and policy options,” Ageing Horiz, 8: 4–12. STAUDINGER, U. M. (1999): “Older and wiser? Integration results on the relationship between age and wisdom-related performance,” International Journal of Behavioral Development, 23: 641–664. STAUDINGER, U. M., AND BALTES, P. B. (1996): “Weisheit als Gegenstand psychologischer Forschung,” Psychologische Rundschau, 47: 57–77. STRITTMATTER, A., SUNDE, U., AND ZEGNERS, D. (2020): “Life cycle patterns of cognitive performance over the long run,” Proceedings of the National Academy of Sciences of the United States of America, 117(44): 27255–27261. SUMMERS, L. H. (2015): “Demand side secular stagnation,” American Economic Review, 105(5): 60–65. TEPE, M., AND VANHUYSSE, P. (2009): “Are ageing OECD welfare states on the path to gerontocracy? Evidence from 18 democracies, 1980–2002,” Journal of Public Policy, 29: 1–28. TUOMI, K., AND ILMARINEN, J. (1999): “Work, lifestyle, health and work ability among ageing municipal workers in 1981–1992.” In: Ilmarinen, J., and Louhevaara, V. (eds.), Finn-Age—Respect for the Ageing: Action Programme to Promote Health, Work Ability and Well-Being of Ageing Workers in 1990–96, Helsinki: Finnish Institute of Occupational Health, pp. 220–232. UPPENBERG, K., AND STRAUSS, H. (2010): “Innovation and productivity growth in the EU service sector,” EIB Working Paper 07/2010. Available at https://www.eib.org/attachments/efs/ efs innovation and productivity en.pdf [accessed on May 4, 2020]. VANDENBROUCKE, G. (2020): “The baby boomers and the productivity slowdown,” European Economic Review, 132: 103609.

435

Axel B¨orsch-Supan and Matthias Weiss

OURS, J. C. (2009): “Will you still need me—when I’m 64?,” IZA Discussion Paper No. 4246. Institut zur Zukunft der Arbeit, Bonn. WEINBERG, B. A., AND GALENSON, D. W. (2005): “Creative careers: The life cycles of Nobel laureates in economics,” NBER Working Paper No. w11799. National Bureau of Economic Research, Cambridge, MA. WEINERT, F. E. (1992): “Altern in psychologischer Perspektive.” In: Baltes, P. B., and Mittelstrass, J. (eds.), Zukunft des Alterns und gesellschaftliche Entwicklung, New York, Berlin: Springer, pp. 180–203. WERDING, M. (2008): “Ageing and productivity growth: Are there macro-level cohort effects of human capital?,” CESifo Working Paper No. 2207, M¨unchen, Germany. WORLD BANK. (2021): World Development Indicators. Washington, DC. VAN

436

24 POPULATION AGEING AND GENDER GAPS: LABOR MARKET, FAMILY RELATIONSHIPS, AND PUBLIC POLICY1 Paola Profeta

Abstract This chapter assesses the complex relationships between population ageing and gender gaps in the labor market, family relationships, and public policy in OECD countries. Population ageing and gender gaps are closely related. Recent evidence shows that, in countries where women participate more in the labor market, fertility rates are higher, thus suggesting that gender equality in the labor market may counterbalance the ageing process. The current trends of ageing and gender equality are changing the labor force composition by increasing the proportion of elderly workers and of women. Ageing and gender equality also interact in the private sphere, influencing family and intergenerational relationships: the ageing process challenges the role of grandmothers as childcare providers and imposes a reconsideration of elderly women as care receivers. All these interactions create new pressure on public policy. The pandemic is challenging the current relationship between ageing and gender equality: gender equality in the labor market is stalling, birth rates are declining, and family relationships—both intergenerational and intra-couple—are under pressure, due to the increased amount of childcare and housework together without the support of grandparents. Further research is needed to understand how public policies will deal with population ageing and the closure of gender gaps in this new context.

24.1

Introduction

The dramatic change in women’s identities and of their roles in families and societies, which led to a substantial increase in the share of women in the workforce, pervaded the past century and continues to shape the current one. Goldin (2006) uses the phrase “quiet revolution” to characterize the emergence of a new economic role for women in the United States. In Europe, while gender gaps in educational attainment have nearly disappeared in most countries, the revolution is still underway in the Mediterranean area, where labor market differences between men and women remain wide. Culture and attitudes crucially shape the ongoing process, while at the same time they are challenged by the increasing participation of women in the workforce DOI: 10.4324/9781003150398-28

437

Paola Profeta

and decision-making processes. Public policies will have to deal with these major changes and face the challenge of redesigning adequate and sustainable institutions. The COVID-19 pandemic poses several challenges to the gender revolution. So far, the pandemic has been quite gendered. While more men than women are dying of COVID-19, women have shown higher vulnerability than men on the labor market (Alon et al., 2020). They are employed in sectors such as the service sector, which was severely hit by the pandemic; they are dominant in sectors with high risks of contagion (e.g., teachers and nurses); and they are less present in occupations that are telecommutable or critical, i.e., they are not affected by stay-athome restrictions. In addition to the labor market aspect, the gendered impact of the pandemic depends on the increase of domestic work and childcare, which has not been equally shared between partners but has mainly fallen on women (Del Boca et al., 2020). Thus women in many places have lost jobs more than men, which is why COVID-19, unlike previous economic crises, has been defined as a “she-cession,” pointing to the possibility of a stalling gender equality process. The new challenges posed by the pandemic interact with the well-established sociodemographic process of population ageing. The world is rapidly ageing. In Europe, the proportion of the population aged 65 and older was 15.7 in 2000 and 17.5 in 2010, and it is expected to increase to 29.7 in 2050 (World Bank Population Estimates and Projections). People live longer than ever before, mostly in good health, and birth rates continue to decline. As a result, the age structure of the population is dramatically changing all over the world, although with some differences across countries. Gender gaps and ageing are closely connected: while in the past gender equality has contributed to a decline in fertility, and thus to an exacerbation of the ageing process, more recent evidence points to the emergence of a positive cross-country relationship between female employment rates and fertility rates. In favorable contexts, where men are involved in household tasks and where appropriate family policies are promoted, more women at work may imply an increase of fertility rates, which translates into a less serious ageing process. Similarly, the pandemic instead seems to be associated with both stalled achievements of gender equality and a reduction of birth rates. Population ageing will lead to an unequal distribution of work because the proportion of the working population will decline unless reforms are implemented to change age-specific patterns of work and keep older individuals at work longer. In this context, the gender revolution will play a crucial role in the evolution toward a new equilibrium. More educated women and their higher participation in the labor market will change the labor force composition by gender and age in the future. A cohort effect emerges for women: older employees and early retirees are mostly men, but this is expected to change. The consequences of this process will go beyond the labor market and affect the family and social sphere, intragenerational relations, and care provisions. While elderly women caregivers, namely grandmothers, have so far played an important role in supporting women’s participation in the labor force and the dual-career model, this role is expected to change with the increase of mothers’ age at first birth and of the geographical mobility of young couples. In later life, women will more likely become care receivers. The coronavirus pandemic has also substantially affected intergenerational relationships: because elderly people are more at risk of contagion, families had to reduce or avoid the use of grandparents as childcare providers. Together with the increase of the amount of childcare and household duties due to the closure of schools and nurseries, the absence of grandparents’ support has put substantial pressure on families and, in most countries, particularly on women (Del Boca et al., 2020). 438

Population Ageing and Gender Gaps: Labor Market, Family Relationships, and Public Policy

The role of public policy in dealing with the ageing process while promoting gender equality is fundamental. Pay-as-you-go (PAYG) pension systems are under financial pressure in ageing countries (see, among others Vaupel and Kistowski, 2008; Christensen et al., 2009; Galasso, 2008; Bovenberg et al., 2010). The higher participation of women in the workforce can help alleviate the pressures of ageing in the labor market and pension systems by reducing the dependency ratio. At the same time, the gender revolution pushes the emergence of new needs, which translate into demand for more generous and better designed family policies. Overall, gender equality has the potential to mitigate the male gerontocracy (“the power of the old”) of the policy agenda in developed democratic countries. This chapter assesses the relationship between population ageing and gender gaps in developed countries. The next section presents evidence on the relationship between female employment and fertility rates. The following two sections show how age and gender interact in the labor market and in care provisions, respectively. The section then highlights the role of public policy, namely pensions and family policy. The final section concludes.

24.2

The Relationship between Female Employment and Fertility

The relationship between female employment and fertility rates has been widely studied. Traditionally, gender equality has proved to be related to the decline in fertility rates observed in all developed countries over time. As the decline in fertility, together with the increase in life expectancy, translates into an overall ageing process, a clear challenge has emerged: how to continue the process toward gender equality without exacerbating population ageing. The question is, however, not so simple as we have observed many changes in the relationship between female employment and fertility over the last decades. According to the principle of specialization formalized by Becker (1981), even though men and women are intrinsically identical, they gain from a division of labor between market and household activities. As long as men have a comparative advantage in working in the market and women in performing domestic and care work, due to their biological characteristics, full specialization arises, which explains the traditional male breadwinner family model where women do not work and do care for children. Several changes have challenged this equilibrium. Economic development and the rise of the service sector have reduced the role of physical strength underpinning the comparative advantage of men in the labor market. Similarly, medical progress has reduced the risk of disease related to pregnancy and birth and the need for breastfeeding (Albanesi and Olivetti, 2016), which represents a comparative advantage of women in caring for children. Moreover, since the second half of the 20th century, women have taken many steps toward independence: they have increased their attention and time devoted to building their individual identities and economic independences before the formation of the family (Goldin, 2006). This new role of women has led to the postponement of fertility choices or even their rejection. Demographers have shown that the postponement of parenthood explains the decline in fertility (Kohler et al., 2002; Lestaeghe, 2010), i.e., a negative relationship between gender equality and fertility rates emerges. From an economic perspective because maternity is a penalty in the labor market (Kleven et al., 2019) we expect a trade-off for women between having children and working or having a career. Not only lower birth rates provide women with more time to fully participate in the formal labor market, but the rise of female wages increases women’s labor supply and, due to the higher opportunity cost, lowers fertility (Galor and Weil, 1996). However, this is not a unique or stable equilibrium. Several recent studies have shown that the context influences the trade-off: culture, family relationships, and public policy have 439

Paola Profeta

Figure 24.1 Female employment and fertility. Source: Author’s elaboration on data from OECD https://data.oecd.org/.

an important role (Del Boca and Locatelli, 2006), and they may even reverse the relationship between fertility and gender equality. In this direction, recent evidence has challenged the traditional negative relationship between gender equality and fertility rates (Engelhardt and Prskawetz, 2004). Figure 24.1 indeed shows a positive correlation between female employment rates and fertility rates across Organisation for Economic Co-operation and Development (OECD) countries today. Macroeconomists have also identified a positive cross-country relationship between women’s empowerment and fertility rates (Doepke and Tertilt, 2018). Working and earning her own income enlarge the set of possible choices for a woman, including the choice of having children. This strongly depends on a context favorable to the double role of women as workers and mothers. In particular, the availability of childcare services, a flexible labor market, and a gender-equal intra-household division of family chores (Oppenheimer, 1994; Fuwa, 2004) contribute to an equilibrium where both fertility rates and female employment are high. In contrast, countries where policies do not support the double role of women as mothers and workers and women bear the entire burden of family care are more likely to be trapped in an equilibrium that features low fertility and low female employment. The reversal of the relationship has attracted the attention of demographers. Myrskyla et al. (2009) show that for countries with a human development index lower than 0.85–0.9 the relationship between total fertility rate and human development index is negative, while for higher levels, it becomes positive. They explain that the reversal trend depends on the level of gender equality in the country vis-`a-vis the level of gender equality within the family. At the beginning of the process toward gender equality the improvement is based on the role of women in the society, while gender gaps still dominate the private sphere, i.e., the division of household tasks (McDonald, 2000). This imposes a trade-off for women between working or 440

Population Ageing and Gender Gaps: Labor Market, Family Relationships, and Public Policy

having children. As the process of gender equality continues, we enter the so-called “second-half of the gender revolution” (Hofferth and Goldscheider, 2010; Goldscheider et al., 2015) where men become involved in household and care activities (Esping-Andersen and Billari, 2015), with the consequence that the trade-off is relaxed and the new positive relationship between female employment and fertility can emerge (Matysiak and Vignoli, 2008). A recent study by Fanelli and Profeta (2021) proves the role of the involvement of men in the family on fertility rates using detailed repeated individual data from the Generations and Gender Survey. In economic terms, a more balanced division of home and market work between men and women within the family is expected to narrow the gender wage gaps, which in turn reduces the opportunity cost of having an additional child for working women. By affecting intra-household bargaining, an increase of women’s empowerment within the couple may itself matter for fertility choices and fertility outcomes (Doepke and Kindermann, 2017). To sum up, gender equality and fertility are not in opposition, rather they can go hand in hand. Thus, promoting gender equality may imply an increase in fertility rates that translates into a less serious ageing process. However, the COVID-19 pandemic poses a new challenge. It has so far been associated with a stalling of both gender equality (Alon et al., 2020) and fertility rates (Aassve et al., 2020). The positive relationship between the two seems to be confirmed, although in a direction that exacerbates both gender gaps and the ageing process. Future research will show how public policy will deal with this new challenge.

24.3

The Labor Market

Age and gender are crucial characteristics of the labor market. The current trends of ageing and gender equality are changing the labor force composition, by increasing the proportion of elderly workers and of women. Governments and policymakers across the OECD are promoting employment at an older age and raising the effective age at which people exit the labor market (OECD, 2019). The labor force participation of older people (aged 55–64) in the OECD has increased by 8 percentage points and reached 64 percent in 2018. Women are working longer than before, which is a persistent trend (Goldin and Katz, 2018). However, a quarter of all older OECD workers still leave the labor force before the age of 60, and the gender gap in participation rates for the 55–64 year group was 18 percentage points in 2018, only 3 percentage points less than 10 years before (OECD, 2019).

24.3.1

The Labor Supply of Elderly Workers

Absorbing elderly workers involves both supply and demand, and both aspects are more critical for women than for men. From the supply side, incentives to continue working at an older age and retire later have been significant over the past decade. Pension reforms have been a key policy to reverse workers’ incentive to retire early (Gruber and Wise, 2008; Bl¨ondal and Scarpetta, 1999). Several studies show the causal impact of pension reforms on the increase of the share of elderly workers. These studies prove that the increase in retirement age increased the labor supply of elderly workers for both men and women (Staubli and Zweim¨uller, 2013; Mastrobuoni, 2009). Specific aspects of the pension reforms matter, such as the changes of pension benefit formulas that reduce pension benefits at an earlier age, the lower replacement rates, and the large impact on benefits of delaying pensions (Geppert et al., 2019). Some of these reforms target women more than men, for example, the increase of retirement age (in nine out of 35 OECD countries retirement age is still lower for women than men) and the generosity of survival pensions (Giupponi, 2020), thus predicting gender differences in response to them (Gruber and Wise, 2008). Women’s labor supply is expected to respond differently than that of men due to 441

Paola Profeta

changes in the internal rates of return of pension systems (Blundell et al., 2002; Disney, 2004). More generally, women’s labor supply is expected to react more to pension reforms compared with that of men, because the income and labor supply elasticities of labor supply are higher for women than for men, due to their different preferences and their role as primary caregivers (as we will explain in a later section). In fact, women tend to retire as early as they can. However, because women earn less than men and have interrupted careers, they may face more constraints than men in their retirement choices, which reduces their effective labor supply reaction to pension reforms (Boeri and Brugiavini, 2008). Moreover, the effective reaction of women to pension reform may be limited by joint retirement because a major determinant of women’s choice of retirement is the retirement of their partners (Ruhm, 1996; Gustman and Steinmeier, 2000; Banks et al., 2010; Stancanelli and Van Soest, 2012; Hospido and Zamarro, 2014).

24.3.2

The Labor Demand of Elderly Workers

Recently, the demand side has emerged as the more critical one: many older workers continue to struggle to access good jobs and training, as employers have few incentives to hire and retain older workers in good jobs. Surveys show that workers’ early retirement may not be a voluntary decision, but rather their employers’ choice (Dorn and Sousa-Poza, 2010; Marmot et al., 2002). The employability of elderly workers depends not only on their work choice but also on the fulltime earnings ratio of 55–64 year olds to 25–54 year olds, the opportunities that workers have to work at an older age, and the practices firms use to retain and encourage old-age workers, which remain limited in many countries (OECD, 2019). Thus, while past research concentrated on the supply side, more recent research has focused on the role of the demand side. Firms are reluctant to hire and maintain elderly workers: a greater presence of older workers may hamper firms’ productivity (Vandenberghe, 2013) and reduce future growth if older workers are less innovative or less willing to take risks than younger ones (Engbom, 2019). The seminal paper by Lazear (1979) provides a theoretical background for firms’ incentives to dismiss elderly workers. Additional incentives arise if firms need to restructure to become more competitive in new or existing markets and in contexts of with strict employment protection legislations, high firing costs, or the steep seniority wages (Bello and Galasso, 2020). Despite these incentives, recent evidence shows that firms may be wrong about considering elderly workers as a “burden”: maintaining the old-age workforce has benefits for productivity, mainly related to the larger experience of the elderly and their skills and attributes that are difficult to replace (Jaeger and Heining, 2019), and elderly workers do not come at a cost of younger workers or at a cost of labor productivity (Carta et al., 2019). In this context, the intersection between age and gender has so far received recent little attention. The evidence remains scarce, and more research is needed. Age discrimination appears stronger for women than for men. Research based on experiments has found evidence of age discrimination in hiring older women (Neumark et al., 2019; Lahey, 2008; Carlsson and Eriksson, 2019). Employers’ stereotypes about the ability to learn new tasks play an important role in age discrimination and are stronger against women than men. I highlight two major drawbacks of existing studies: (1) they focus mainly on low-income jobs and cannot show the complete differential effect for men and women throughout the income spectrum, and (2) they do not focus on workers above 50 or 60 years old, i.e., close to retirement age, which is however an important age group to analyze (UNECE, 2019). While some detailed evidence on the labor demand of elderly men exists (Ahmed et al., 2012; Albert et al., 2011; Duncan and Loretto, 2004; Riach and Rich, 2010), we need more complete studies focused on women. Gender differences in the demand side of old-age work will be important for many reasons. Among them, 442

Population Ageing and Gender Gaps: Labor Market, Family Relationships, and Public Policy

retaining elderly women at work is more difficult than retaining men because of the incentives that women have to retire earlier. The employability of elderly workers requires specific training programs. A third of 55–64 year-old workers have no computer skills and low technological skills (OECD Survey of Adult Skills, PIAAC). This is even more true for women, who experience a digital divide even among younger generations (OECD, 2019). Upskilling and investing in lifelong learning is crucial to help older workers work under rapidly evolving technological progress, particularly for women. A vicious circle arises because firms expect old workers to be less productive because their skills are not updated and not innovative and thus invest less in training and other programs of the elderly, further reducing their employability. Future studies will develop policy evaluations of gender-specific age-management practices and training programs.

24.3.3

The “Double Burden” of Age and Gender

The gender gap in old-age work is also connected to forms of discrimination that are amplified at the age-gender intersection. Although legislation bans age discrimination, ageism is still perceived real in the labor market. Several surveys in Europe confirm implicit biases against hiring older workers and in the workplace (Eurobarometer, 2015). As culture against gender equality is a major driver of gender gaps and a reality all over the world (as shown by data from the World Value Survey analyzed by several authors, starting with Inglehart and Norris, 2003), elderly women face a double penalty in the labor market. The double penalty in the labor market due to being old and a woman has emerged in several studies across social sciences (Profeta, 2020; Price, 2007; Lef`ebvre, 2007). However, when addressing the intersection between ageing and gender gaps research findings remain scattered and tend to evade making causal inferences. That is, most studies are cross-sectional and thus are ineffective in addressing causality. Recent studies have addressed specific subgroups of the female cohort, such as childless women vis-`a-vis mothers (e.g., Quashie et al., 2021; Verdery et al., 2019; Hadley, 2019; M¨ohring, 2018; Albertini and Kohli, 2016), relatively rich vis-`a-vis relatively poor women (e.g., Vlachantoni, 2019; M¨ohring, 2018; B´arcena-Mart´ın and Moro-Egido, 2013; Lewis, 2007), and relatively healthy vis-`a-vis relatively unhealthy women (e.g., Schmitz and Lazareviˇc, 2020; Boerma et al., 2016). However, while subgroups have been considered, public policy analyses for the ageing female cohort have not been fully developed. The pandemic seems to have a strong negative effect on older female workers, thus reinforcing the need to better understand the intersection between the disadvantages of age and gender in the labor market.

24.3.4

Gender Gaps from the Labor Market to Pensions

Economists have studied how gender gaps in the labor market translate into gender gaps in pensions. According to OECD (2019), women get 25 percent lower pensions than men, mainly due to shorter careers, fewer hours of work, contribution gaps related to delayed or interrupted employment, and lower wages. Labor market differences are the most important driver of the observed gender gaps in pensions (Ginn, 2001). Pensions rules also matter: pension systems have historically been designed according to typical male patterns of full-time, continuous employment and without including unpaid tasks such as care provision, even if these often prevent women from participating in the labor force full time. Thus, the current pension systems appear problematic from a gender perspective and lead to an unfair distribution of resources for olderage women in particular (Vlachantoni, 2019; M¨ohring, 2018). How the pension system considers unpaid care periods matters (D’Addio, 2015). Even when pension systems do place value on 443

Paola Profeta

unpaid care work and include this into their formulas, this may not be sufficient to compensate women for the time they spent outside the labor force. The redistributive design of pension systems seems to play a crucial role: redistributive, flat-rate pension systems tend to favor women, who typically belong to the weaker group of workers, in contrast with earnings-related systems (Ginn, 2004). Interestingly, the literature has shown that earnings-related pension systems entail a disincentive to fertility (see Cigno et al., 2003), again linking ageing and gender gaps. Defined contribution pension schemes are also expected to be associated with larger pension gender gaps than defined benefit ones, and the shift from defined benefits to defined contributions tends to reinforce the gender pension gaps. Lis and Bonthuis (2019), however, show that European countries with notionally defined contribution pension schemes (Italy, Latvia, Norway, Poland, and Sweden) are like the others in terms of pension outcomes for women, thus suggesting that labor market differences are more important than the design of pensions. The authors also highlight that indexation of pensions in payment and the existence of survivors’ pension options strongly affect gender inequalities. The existence of gender pension gaps is particularly critical due to the longer life expectancy of women, who tend to outlive their male counterparts by an average of 5.5 years. Despite the increase of retirement age, the increase of life expectancy implies that people will still spend almost a quarter of their life in retirement (OECD, 2019). This is particularly true for women, because of their longer life expectancy. However, their lower level of pensions exposes them to high risk of old-age poverty. According to OECD (2019), the average old-age poverty rates for women and men in the OECD are 15.7 percent and 10.3 percent, respectively. Lower earnings-related pension income and a longer life expectancy are among the main drivers of higher poverty incidence among women than among men. Women as “care receivers” in later life have recently attracted the attention of researchers. Pension gender gaps create a vicious circle in which elderly women see themselves trapped in economic dependence at old age, therefore relying on informal care provision by younger family members. Considering these family members are likely to be women who will interrupt their careers to provide such care, then the cycle commences again. The need for care in later life is amplified in case of childlessness. Population ageing and low fertility rates might translate into more elderly in need of care but without kin to support them (UNECE, 2020; Scheil-Adlung, 2015). The nearly linear relationship between low fertility rates and childlessness suggests that this cohort will increasingly require particular policy attention (Verdery et al., 2019). At the same time, evidence suggesting that the childless are effective in developing strong ties with other family members of friends who can provide support is not so straightforward when intensive support is required in old age (Albertini and Kohli, 2016). This gap in the supply of care is particularly relevant when considering a global decreasing trend in formal care provision (Hadley, 2019) and that, at least in Europe, family members informally provide 80 percent of the care supply (Zigante, 2018; Cascella Carb´o and Garc´ıa-Orell´an, 2020). Although both have consequences, childlessness may affect men and women as care receivers differently. The low level of economic independence of women and low pensions implies that women are more likely to be in need in later life and, in the case of childlessness, the absence of supporting children may create more problems for women than for men. Thus, differences between men and women are expected to emerge (Dykstra and Hagestad, 2007; Dykstra and Wagner, 2007; Albertini and Mencarini, 2012), which may also depend on marital status (Wenger et al., 2000; Larsson and Silverstein, 2004). At the same time, however, childless women are expected to be less negatively impacted economically, as they do not face the decision of interrupting their careers to care for dependent 444

Population Ageing and Gender Gaps: Labor Market, Family Relationships, and Public Policy

children. Future research will need to clarify the gender-specific challenges of care for childless elderly people.

24.4

Family Relationships

In addition to being recipients of care, women are also the largest providers of care—both formal and informal. The role of elderly women as “caregivers” instead of “care receivers” has been strong in European countries where grandmothers have played an important role in supporting women’s participation in the labor force and the dual-career model. Informal care by grandparents, mainly grandmothers, acts as a substitute for scarce formal childcare services, the insufficient numbers of school hours, and inadequate access to part-time work (Del Boca et al., 2005; Keck and Saraceno, 2008). Heterogeneity across countries is huge and is driven by needs, which in turn depend on the family structure, financial difficulties, and the presence of services. According to Survey of Health, Ageing and Retirement in Europe data, the percentage of grandparents who care for grandchildren at least once per week is 45 percent in Italy, 30 percent in France, and 20 percent in Sweden. They are mainly grandmothers. However, the ageing process and the gender equality process challenge the role of elderly women as caregivers. The ageing process will lead to an increase of retirement age that reduces the availability of grandparents. The increase of a mother’s age at first birth makes continuing to provide childcare at a very old age problematic for grandparents. Research also shows that the informal use of grandparents for childcare may have contributed to women’s participation in the labor market (Glasser et al., 2013), but at the cost of fewer children (Arpino et al., 2014) because caring for more than one small child can be complicated for grandparents. The coronavirus epidemic further challenges these relationships, having suddenly reduced the role of grandparents as caregivers. While the impact of the pandemic on intra-couple family relationships has already been investigated (see Del Boca et al., 2020, for Italy; Farr´e et al., 2020, for Spain; Sevilla and Smith, 2020, for the United Kingdom), with the general result being that the increased family burden (due for example to school closures) has fallen more on women than on men, the consequences of the reduced role of grandparents as caregivers have not yet received specific attention. Identifying the causal impact of grandparents on women’s employment and on the so-called “she-cession” and distinguishing it from the increased family burden is not obvious and will need careful analysis. The ageing process in a more gender equal labor market may also contribute to the rise of the so-called “sandwich generation” of women, those who find themselves as caregivers for both underage children and ageing parents or in-laws, while also holding down a job (K¨unemund, 2006). This popular concept has however generated little empirical evidence to date. This is probably because the true effects of the ageing process will only appear in the future, when the baby boom generations become old and in need of care, due to a longer life expectancy, and because of their low fertility, care will be divided among fewer siblings. Currently, elderly women (grandmothers) are in fact more often caregivers than care receivers. Future research will assess the evolution of these trends.

24.5

Public Policy

Pension expenditures, which are expected to increase due to population ageing, dominate current welfare systems (Galasso and Profeta, 2004). In a political economy perspective, an older median voter prefers a larger size of the pension system (Galasso and Profeta, 2002). A higher participation of women in the workforce may counterbalance population ageing and alleviate the financial pressures of an ageing population on PAYG social security systems. At the same 445

Paola Profeta

time, the changing role of women is forcing the welfare state to reorganize in favor of more and broader family policies, such as childcare services, which may facilitate female labor force participation (Profeta, 2020). A new trade-off may emerge. Because population ageing is positively related to expenditures in pensions (Galasso and Profeta, 2004) and gender equality is positively related to expenditures in family policies (Profeta, 2020), population ageing and gender equality may push a “competition” for the allocation of a given amount of resources between two different components of public spending. Figure 24.2 shows a (slightly) negative relationship between public expenditure on family policies and on pensions (both as a percentage of gross domestic product or GDP) across OECD countries, which, although only suggestive, confirms the possible trade-off. The interaction between ageing and gender equality puts additional pressure on public policies. Population ageing requires reforms of sustainable PAYG pensions systems. As increasing contribution rate is difficult, and reducing the generosity of pensions is strongly opposed by a cohesive group of elderly (Mulligan and Sala-i-Martin, 2003), the increase of retirement age is the more politically feasible option (Galasso and Profeta, 2004). Many countries have in fact increased retirement age. This, however, puts pressure on the use of informal childcare provided by grandparents, thus requiring more public spending on childcare. The coronavirus pandemic has amplified this effect because grandparents are less available due to their higher risk of contagion. How the composition of the welfare state will evolve under the ageing process and the reduction of gender gaps and how the COVID-19 pandemic will affect this evolution is a new, promising area of research. Given all these interactions, recent research suggests that pension reforms dealing with the ageing process cannot be implemented in isolation, but should be designed together with welfare and family policy reforms (OECD, 2019). Coda Moscarola et al. (2016) provide an example of the undesirable outcomes of a lack of comprehensive policies. They studied the postponement of retirement age for Italian women forced by the 2011 pension reform, finding that working grandmothers increased sick leave, which enabled them to care for grandchildren, and that this

Figure 24.2 Expenditure on family policies and pensions (as percentage of GDP). Source: Author’s elaboration on OECD data.

446

Population Ageing and Gender Gaps: Labor Market, Family Relationships, and Public Policy

effect was stronger in regions with low childcare services. Existing studies also show that pension reforms in response to population ageing may affect men and women differently. In addition to gender differences in labor supply reactions and gender pension gaps related to gender gaps in the labor market, mentioned previously, new challenges emerge. First, the trend of promoting private pensions as a response to ageing is gendered because women tend to invest fewer resources and for a shorter amount of time into the second and third pillar, due to their interrupted careers, lower wages, and lower financial literacy (Fornero and Monticone, 2011). Second, because women live longer than men, serious concerns about adequate income for women at old age arise. Because pension systems are not meant to address and solve gender gaps at old age, a more comprehensive approach to pension, social, and family policy reforms is recommended. An additional element of the described picture is that women’s empowerment may influence individuals’ voting decisions on the public policy platform and ultimately change the policy agenda. Although causal evidence remains scarce and not fully conclusive, the literature shows that more women in decision-making positions tend to reorient the policy agenda toward family, social, and education items (see Hessami and da Fonseca, 2020, for a complete review). Thus, more gender balance in decision-making positions may potentially mitigate the “male gerontocracy” of the welfare state (see Baltrunaite et al., 2014) that is exacerbated by the ageing process and improve the sustainability of welfare systems. A feedback effect may arise, because a society, an economy, and a political system that is open to the young will guarantee greater female representation, as women are more represented among young workers and professionals.

24.6

Conclusions

As the ageing process and the reduction of gender gaps represent two major and continuous trends of our times, research on the interaction between ageing and gender will continue to grow. This chapter has identified some of the main aspects of the interaction that have already attracted the attention of scholars: the relationship between gender equality and fertility and its evolution over time, the role of age and gender in the labor market and during retirement, their impact on family and intergenerational relationships, and the role of pensions and family policies. The review in this chapter is of course not exhaustive of the vast literature across the many disciplines that have addressed this topic. Yet it shows the scarcity of causal studies and the prevalence of cross-country analysis or country-specific descriptive studies that, while informative about the complex mechanisms in place and of their interesting possible consequences, are far from providing a causal impact of age and gender on outcomes. More studies are needed in this direction. The chapter also suggests how the coronavirus pandemic is challenging the relationship between ageing and gender equality. Future studies will assess the impact of the pandemic in each of the dimensions described. The pandemic can be also useful to establish more carefully the existence of causal links. Other demographic changes will interact with ageing and gender gaps, including, for example, migration patterns. Interestingly, we may expect the emergence of opposite effects. On the one side, the increasing emigration of adult children for professional reasons leaving their parents without a source of informal care provision, creates demand for more public policy interventions for the elderly. On the other side, migration is also gendered, with a prevalence of women who immigrate to provide informal and formal care, often employed under irregular contracts, which implies precarious conditions and limited access to social protection. This creates pressure for more gender equality interventions. Future studies will address how public policy will deal with these changes, thus adding additional dimensions to the age and gender intersection. 447

Paola Profeta

Note 1 The author thanks Ximena Cal´o for excellent research assistance.

References AASSVE, A., CAVALLI, N., MENCARINI, L., PLACH, S., AND LIVI BACCI, M. (2020): “The COVID-19 pandemic and human fertility,” Science, 369(6502): 370–371. AHMED, A. M., ANDERSSON, L., AND HAMMERSTEDT, M. (2012): “Does age matter for employability? A field experiment on ageism in the Swedish labour market,” Applied Economic Letters, 19(4): 403–406. ALBANESI, S., AND OLIVETTI, C. (2016): “Gender roles and medical progress,” Journal of Political Economy, 124(3): 650–695. ´ ALBERT, R., ESCOT, L., AND FERN ANDEZ -CORNEJO, J. A. (2011): “A field experiment to study sex and age discrimination in the Madrid labour market,” The International Journal of Human Resource Management, 22(02): 351–375. ALBERTINI, M., AND KOHLI, M. (2016): “Childlessness and intergenerational transfers in later life.” In: Kreyenfeld, M., and Konietzka, D. (eds.), Childlessness in Europe: Contexts, Causes, and Consequences. Switzerland: Springer Nature, pp. 351–368. ALBERTINI, M., AND MENCARINI, L. (2012): “Childlessness and support networks in later life: New pressures on familistic welfare states?,” Journal of Family Issues, 35(3): 331–357. ALON, T. M., DOEPKE, M., OLMSTEAD-RUMSEY, J., AND TERTILT, M. (2020): “The impact of COVID-19 on gender equality,” NBER Working Paper No. 26947. National Bureau of Economic Research, Cambridge, MA. ARPINO, B., PRONZATO, C. D., AND TAVARES, L. P. (2014): “The effect of grandparental support on mothers’ labour market participation: An instrumental variable approach,” European Journal of Population, 30(4): 369–390. BALTRUNAITE, A., BELLO, P., CASARICO, A., AND PROFETA, P. (2014): “Gender quotas and the quality of politicians,” Journal of Public Economics, 118(C): 62–74. BANKS, J., BLUNDELL, R., AND CASANOVA RIVAS, M. (2010): The Dynamics of Retirement Behavior in Couples: Reduced-form Evidence from England and the US. London: UCL. ´ BARCENA -MART´I N, E., AND MORO-EGIDO, A. I. (2013): “Gender and poverty risk in Europe,” Feminist Economics, 19(2): 69–99. BECKER, G. (1981): A Treatise on the Family. Cambridge, MA: Harvard University Press. BELLO, P., AND GALASSO, V. (2020): “Old before their time: The role of employers in retirement decisions,” International Tax Public Finance, 27(5): 1198–1223. ¨ BL ONDAL , S., AND SCARPETTA, S. (1999): “The retirement decision in OECD countries,” OECD Working Paper AWP 1.4, Paris. BLUNDELL, R., COSTAS, M., AND SMITH, S. (2020): “Pension incentives and the pattern of early retirement,” The Economic Journal, 112(478): 153–170. BOERI, T., AND BRUGIAVINI, A. (2008): “Pension reforms and women retirement plans,” IZA Institute of Labor Economics Discussion Paper No. 3821. Bonn, Germany. BOERMA, T., REZA HOSSEINPOOR, A., VERDES, E., AND CHATTERJI, S (2016): “A global assessment of the gender gap in self-reported health with survey data from 59 countries,” BMC Public Health, 16(675): 1–9. BOVENBERG, L., VAN SOEST, H. O., AND ZAIDI, A. (2010): Ageing, Health and Pensions in Europe: An Economic and Social Policy Perspective. London: Palgrave Macmillan. CARLSSON, M., AND ERIKSSON, S. (2019): “Age discrimination in hiring decisions: Evidence from a field experiment in the labor market,” Labour Economics, 59(C): 173–183. CARTA, F., D’AMURI, F., AND VON WACHTER, T. M. (2019): “Workforce ageing, pension reforms, and firm outcomes,” NBER Working Paper No. 28407. National Bureau of Economic Research, Cambridge, MA. ´ , R. (2020): “Burden and gender inequalities around CASCELLA CARB O´ , G. F., AND GARC´I A-ORELL AN informal care,” Investigaci´on y Educaci´on en Enfermer´ıa, 38(1): 1–14. CHRISTENSEN, K., DOBLHAMMER, G., RAU, R., AND VAUPER, J. W. (2009): “Ageing populations: The challenges ahead,” The Lancet, 374(9696): 1196–1208. CIGNO, A., CASOLARO, L., AND ROSATI, F. C. (2003): “The impact of social security on saving and fertility in Germany,” FinanzArchiv, 59(2): 189–211.

448

Population Ageing and Gender Gaps: Labor Market, Family Relationships, and Public Policy

CODA MOSCAROLA, F., FORNERO, E., AND STRØM, S. (2016): “Absenteeism, childcare and the effectiveness of pension reforms,” ZA Journal of European Labour Studies, 5(1): 1–18. D’ADDIO, A. C. (2015): “Explaining the gender pension gap in OECD countries: Socio-economic determinants and pension rules that matter,” Unpublished manuscript, OECD, Paris. DEL BOCA, D., AND LOCATELLI, M. (2006): “The determinants of motherhood and work status: A survey,” IZA Institute of Labor Economics Discussion Paper No. 2414. Bonn, Germany. DEL BOCA, D., OGGERO, N., PROFETA, P., AND ROSSI, M. C. (2020): “Women’s work, housework and childcare, before and during COVID-19,” IZA Institute of Labor Economics Discussion Paper No. 13409. Bonn, Germany. DEL BOCA, D., PASQUA, S., AND PRONZATO, C. (2005): “Employment and fertility in Italy, France and the UK,” Labour, 19(1): 51–77. DISNEY, R. (2004): “Are contributions to public pension programmes a tax on employment?,” Economic Policy, 39(19): 267–311. DOEPKE, M., AND KINDERMANN, F. (2017): “Intrahousehold decision making and fertility.” In: Cervellati, M., and Sunde, U. (eds.), Demographic Change and Long-Run Development. Cambridge, MA: MIT Press, pp. 159–182. DOEPKE, M., AND TERTILT, M. (2018): “Women’s empowerment, the gender gap in desired fertility, and fertility outcomes in developing countries,” American Economic Review Papers and Proceedings, 108: 358–362. DORN, D., AND SOUSA-POZA, A (2010): “Voluntary and involuntary early retirement: An international analysis,” Applied Economics, 42(4): 427–438. DUNCAN, C., AND LORETTO, W. (2004): “Never the right age? Gender and age-based discrimination in employment,” Gender, Work and Organization, 11(1): 95–115. DYKSTRA, P. A., AND HAGESTAD, G. O. (2007): “Childlessness and parenthood in two centuries: Different roads different maps?,” Journal of Family Issues, 28(11): 1518–1532. DYKSTRA, P. A., AND WAGNER, M. (2007): “Pathways to childlessness and late-life outcomes,” Journal of Family Issues, 28(11): 1487–1517. ENGBOM, N. (2019): “Firm and worker dynamics in an ageing labor market,” Working Papers 756, Federal Reserve Bank of Minneapolis. ENGELHARDT, H., AND PRSKAWETZ, A. (2004): “On the changing correlation between fertility and female employment over space and time,” European Journal of Population, 20(1): 35–62. ESPING-ANDERSEN, G., AND BILLARI, F. C. (2015): “Re-theorizing family demographics,” Population and Development Review, 41(1): 1–31. EUROBAROMETER SURVEY ON DISCRIMINATION. (2015): “General perceptions, opinions on policy measures and awareness of rights,” https://europa.eu/eurobarometer/surveys/detail/2077. FANELLI, E., AND PROFETA, P. (2021): “Fathers’ involvement in the family, fertility and maternal employment: Evidence from Central and Eastern Europe,” Demography, 58(5): 1931–1954. ´ FARR E´ , L., FAWAZ, Y., GONZ ALEZ , L., AND GRAVES, J. (2020): “How the Covid-19 lockdown affected gender inequality in paid and unpaid work in Spain,” IZA Institute of Labor Economics Discussion Paper No. 13434. Bonn, Germany. FORNERO, E., AND MONTICONE, C. (2011): “Financial literacy and pension plan participation in Italy,” Journal of Pension Economics and Finance, 10(4): 547–564. FUWA, M. (2004): “Macro-level gender inequality and the division of household labor in 22 countries,” American Sociological Review, 69(6): 751–767. GALASSO, V. (2008): The Political Future of Social Security in Ageing Societies. Cambridge, MA: MIT Press. GALASSO, V., AND PROFETA, P. (2002): “The political economy of social security: A survey,” European Journal of Political Economy, 18(1): 1–29. GALASSO, V., AND PROFETA, P. (2004): “Lessons for an ageing society: The political sustainability of social security systems,” Economic Policy, 19(38): 63–115. GALOR, O., AND WEIL, D. N. (1996): “The gender gap, fertility, and growth,” American Economic Review, 86(3): 374–387. GEPPERT, C., GUILLEMETTE, Y., MORGAVI, H., AND TURNER, D. (2019): “Labour supply of older people in advanced economies: The impact of changes to statutory retirement ages,” OECD Economics Department Working Papers No. 1554, OECD Publishing, Paris. GINN, J. (2001): Women, Work and Pensions. Maidenhead: Open University Press. GINN, J. (2004): “European pension privatisation: Taking account of gender,” Social Policy & Society, 3(2): 123–134.

449

Paola Profeta

GIUPPONI, G. (2020): “When income effects are large: Labor supply responses and the value of welfare transfers,” Center for Economic Performance (CEP) Discussion Paper 1651. London, UK. GLASSER, K., PRICE, D., DI GESSA, G., RIBE, E., STUCHBURY, R., AND TINKER, A. (2013): “Grandparenting in Europe: Family policy and grandparents’ role in providing childcare.” Grandparents Plus, United Kingdom GOLDIN, C. (2006): “The quiet revolution that transformed women’s employment, education, and family,” American Economic Review, 90(4): 1091–1119. GOLDIN, C., AND KATZ, L. (2018): “Women working longer: Increased employment at older age,” NBER Conference Report. National Bureau of Economic Research, Cambridge, MA. GOLDSCHEIDER, F., BERNHARDT, E., AND LAPPEGARD, T. (2015): “The gender revolution: A framework for understanding changing family and demographic behavior,” Population and Development Review, 41(2): 207–239. GRUBER, J., AND WISE, D. A. (2008): Social Security and Retirement around the World. Chicago: University of Chicago Press. GUSTMAN, A., AND STEINMEIER, T. (2000): “Retirement in dual-career families: A structural model,” Journal of Labor Economics, 18(3): 503–545. HADLEY, R. A (2019): “Ageing without children, gender and social justice.” In: Westwood, S. (ed.), Ageing Diversity and Equality: Social Justice Perspectives. Oxford and New York: Routledge, pp. 66–81. HESSAMI, Z., AND DA FONSECA, M. (2020): “Female political representation and substantive effects on policies: A literature review,” European Journal of Political Economy, 63: 101896. HOFFERTH, S. L., AND GOLDSCHEIDER, F. (2010): “Family structure and the transition to early parenthood,” Demography, 47(2): 415–437. HOSPIDO, L., AND ZAMARRO, G. (2014): “Retirement patterns of couples in Europe,” IZA Journal of Labor Studies, 3(12): 1–18. INGLEHART, R., AND NORRIS, P. (2003): Rising Tide: Gender Equality and Cultural Change around the World. Cambridge: Cambridge University Press. JAEGER, S., AND HEINING, J. (2019): “How substitutable are workers? Evidence from worker deaths,” MPRA Paper 109757, University Library of Munich, Germany. KECK, W., AND SARACENO, C. (2008): “Grandchildhood in Germany and Italy: An exploration,” Comparative Social Research, 25: 133–163. ¨ KLEVEN, H., LANDAIS, C., POSCH, J., STEINHAUER, A., AND ZWEIM ULLER , J. (2019): “Child penalties across countries: Evidence and explanations,” American Economic Review Papers and Proceedings, 109: 122–126. KOHLER, H. P., BILLARI, F. C., AND ORTEGA, J. A. (2002): “The emergence of lowest-low fertility in Europe during the 1990s,” Population and Development Review, 28(4): 641–680. ¨ KUNEMUND , H. (2006): “Changing welfare states and the ‘sandwich generation’: Increasing burden for the next generation?,” International Journal of Ageing and Later Life, 1(2): 11–29. LAHEY, J. (2008): “Age, women, and hiring: An experimental study,” The Journal of Human Resources, 43(1): 30–56. LARSSON, K., AND SILVERSTEIN, M. (2004): “The effect of marital and parental status on informal support and service utilization: A study of older Swedes living alone,” Journal of Ageing Studies, 18: 231–244. LAZEAR, E. P. (1979): “Why is there mandatory retirement?,” The Journal of Political Economy, 87(6): 1261–1284. LEF E` BVRE, M. (2007): “The redistributive effects of pension systems in Europe: A survey of evidence,” Luxembourg Income Study (LIS) Working Paper Series No. 457. Luxembourg. LESTAEGHE, R. (2010): “The unfolding story of the second demographic transition,” Population and Development Review, 36(2): 211–251. LEWIS, J. (2007): “Gender, ageing and the ‘new social settlement’: The importance of developing a holistic approach to care policies,” Current Sociology, 55(2): 271–286. LIS, M., AND BONTHUIS, B. (2019): “Drivers of the gender gap in pensions: Evidence from EU-SILC and the OECD pension model,” Social Protection and Jobs. World Bank Discussion Paper No. 1917. Washington, DC. MARMOT, M., BANKS, J., BLUNDELL, R., LESSOF, C., AND NAZROO, J. (2002): “Health, wealth and lifestyles of the older population in England,” Institute for Fiscal Studies United Kingdom. MASTROBUONI, G. (2009): “Labor supply effects of the recent social security benefit cuts: Empirical estimates using cohort discontinuities,” Journal of Public Economics, 93(11): 1224–1233.

450

Population Ageing and Gender Gaps: Labor Market, Family Relationships, and Public Policy

MATYSIAK, A., AND VIGNOLI, D. (2008): “Fertility and women’s employment: A meta-analysis,” European Journal of Population, 24(4): 363–384. MCDONALD, P. (2000): “Gender equality in theories of fertility transition,” Population and Development Review, 26(3): 427–439. ¨ MOHRING , K. (2018): “Is there a motherhood penalty in retirement income in Europe? The role of lifecourse and institutional characteristics,” Ageing and Society, 38: 2560–2589. MULLIGAN, C. B., AND SALA-I-MARTIN, X. (2003): “Social security, retirement, and the singlemindedness of the electorate,” NBER Working Paper No. 9691. National Bureau of Economic Research, Cambridge, MA. MYRSKYL A¨ , M., KOHLER, H. P., AND BILLARI, F. C. (2009): “Advances in development reverse fertility declines,” Nature, 460(7256): 741–743. NEUMARK, D., BURN, I., AND BUTTON, P. (2019): “Is it harder for older workers to find jobs? New and improved evidence from a field experiment,” Journal of Economy, 17(2): 922–970. OECD (ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT). (2019): Working Better with Age: Ageing and Employment Policies. Paris: OECD Publishing. OPPENHEIMER, V. K. (1994): “Women’s rising employment and the future of the family in industrial societies,” Population and Development Review, 20(2): 293–342. PRICE, D. (2007): “Closing the gender gap in retirement income: What difference will recent UK pension reforms make?,” Journal of Social Policy, 36(4): 561–583. PROFETA, P. (2020): Gender Equality and Public Policy: Measuring Progress in Europe. Cambridge: Cambridge University Press. QUASHIE, N. T., ARPINO, B., ANTCZAK, R., AND MAIR, C. A. (2021): “Childlessness and health among older adults: Variation across five outcomes and 20 countries,” Journals of Gerontology: Social Sciences, 76(2): 348–359. RIACH, P., AND RICH, J. (2010): “An experimental investigation of age discrimination in the English labor market,” Annals of Economics and Statistics, 99–100: 169–185. RUHM, C. J. (1996): “Do pensions increase the labor supply of older men?,” Journal of Public Economics, 59(2): 157–175. SCHEIL-ADLUNG, X. (2015): “Long-term care protection for older persons: A review of coverage deficits in 46 countries,” Extension Social Security (ESS) Working Paper No. 50. ILO, Geneva, Switzerland. SCHMITZ, A., AND LAZAREVI Cˇ , P. (2020): “The gender health gap in Europe’s ageing societies: Universal findings across countries and age groups?,” European Journal of Ageing, 17: 509–520. SEVILLA, A., AND SMITH, S. (2020): “Baby steps: The gender division of childcare after COVID19,” Centre for Economic Policy Research (CEPR) Discussion Paper No. 14804. London, United Kingdom. STANCANELLI, E., AND VAN SOEST, A. (2012): “Retirement and home production: A regression discontinuity approach,” American Economic Review, 102(3): 600–605. ¨ STAUBLI, S., AND ZWEIM ULLER , J. (2013): “Does raising the early retirement age increase employment of older workers?,” Journal of Public Economics, 108(C ): 17–32. UNECE (UNITED NATIONS ECONOMIC COMMISSION FOR EUROPE). (2019): “Combating ageism in the world of work,” Policy Brief on Ageing, 21: 1–16. UNECE (UNITED NATIONS ECONOMIC COMMISSION FOR EUROPE). (2020): “Gender equality in ageing societies,” Policy Brief on Ageing, 23: 1–28. VANDENBERGHE, V. (2013): “Are firms willing to employ a greying and feminizing workforce?,” Labor Economics, 22(C): 30–46. VAUPEL, J. W., AND KISTOWSKI, K. G. (2008): “Living longer in an ageing Europe: A challenge for individuals and societies,” European View, 7(2): 255–263. VERDERY, A. M., MARGOLIS, R., ZHOU, Z., CHAI, X., AND RITTIRONG, J. (2019): “Kinlessness around the world,” Journals of Gerontology: Social Sciences, 74(8): 1394–1405. VLACHANTONI, A. (2019): “Socio-economic inequalities in later life: The role of gender.” In: Westwood, S. (ed.), Ageing, Diversity and Equality: Social Justice Perspectives. Oxford and New York: Routledge, pp. 25–35. WENGER, C. G., SCOTT, A., AND PATTERSON, N. (2000): “How important is parenthood? Childlessness and support in old age in England,” Ageing & Society, 20(2): 161–182. WORLD BANK POPULATION ESTIMATES AND PROJECTIONS: available at https://databank. worldbank.org/source/population-estimates-and-projections. ZIGANTE, V. (2018): Informal Care in Europe: Exploring Formalisation, Availability and Quality. Luxembourg: Publications Office of the European Union.

451

PART V

Data and Measurement

25 MEASURING AGEING Holger Strulik

Abstract This chapter is written for (health) economists interested in the conceptualization and measurement of ageing in demography, biology, and gerontology. I review two alternative methods to measure ageing and a theory that explains the outcome of both of these measurements. I explain how ageing, conceptualized as increasing probability of death, can be accurately and conveniently expressed by the Gompertz-Makeham law. I discuss the stability of the estimated parameters, inferences about human lifespan, and similarities and differences of ageing across sexes and countries and over time. Alternatively, ageing can be measured as accumulation of health deficits. The chapter reviews the frailty index, designed to measure biological ageing, and demonstrates important similarities and differences between the force of mortality and the accumulation of health deficits. Regularities of health deficit accumulation across individuals, at the level of (sub) populations, and across the world are discussed. The chapter also introduces reliability theory and shows how increasing mortality and frailty can be explained as a stochastic process of loss of built-in redundancy of organisms. The chapter concludes with some suggestions for the modeling of health and mortality in economics.

25.1

Introduction

In chronological terms, all humans age by 1 year every year. In physiological terms, however, individuals age quite differently. A 70-year-old person can be as healthy as a 50-year-old one and vice versa. (Physiological) ageing is defined as the intrinsic, cumulative, progressive, and deleterious loss of function that eventually culminates in death (Masoro, 2005; Arking, 2006; Lopez-Otin et al., 2013). Here we focus on ageing of organisms (mostly humans), but every car owner recognizes that the definition of ageing applies to nonliving matter as well. Actually, living and inanimate systems age in a similar way, a phenomenon to which we return later. Based on the definition of ageing, I discuss two possibilities to measure it as (1) increasing probability of death and (2) accumulation of health deficits and functional limitations. Ageing, conceptualized as increasing probability of death and measured by the force of mortality, will be discussed in the first half of the chapter. I show that ageing can be accurately and conveniently expressed by the Gompertz-Makeham law. I discuss the stability of the estimated parameters, inferences about human lifespan, and similarities and differences of ageing across sexes and countries and over time. DOI: 10.4324/9781003150398-30

455

Holger Strulik

While the force of mortality can measure ageing in populations, an index of health deficits is furthermore capable of measuring ageing at the level of individuals. It will be discussed in the second half of the chapter. After the introduction of the frailty index as a measure of biological ageing, I discuss important regularities of health deficit accumulation across individuals, at the level of (sub) populations, and across the world and highlight similarities and differences between the force of mortality and the accumulation of health deficits. The two measurement parts are linked by a section that provides an introduction to reliability theory and shows how increasing mortality and frailty can be explained as a stochastic process of loss of redundancy in organisms. The chapter concludes with some suggestions for the modeling of health and mortality in economics.

25.2

The Force of Mortality

Demographers employ several indices of mortality. For example, the probability of surviving to age x, the life expectancy at age x, or the probability of dying in the age interval x + 1x. Of these, the probability of dying is the most suitable measure of ageing because it provides the agespecific impact on mortality independently from age-specific events in other age groups. To see this, note that, for example, the probability of dying in the age interval (0,1) clearly affects the survival probability to any age x > 1 and life expectancy at age 0, but not the other way around. Biologists and gerontologists emphasize that individual ageing must be understood as an event-dependent, not as a time-dependent process (Arking, 2006). No such thing as a “biological clock” exists; the probability of dying is thus a measure of how members of a population (a species, a population of a country) age on average. Another implication of this view is that the probability of dying is not useful to assess or explain individual survival prospects. Given a sample of a population and the observation (from a life table) that a number S(x) thereof survives to age x and a number d(x) = S(x)−S(x+1x) dies between age x and x+1x, the probability of dying in age interval x + 1x is given by q(x) = d(x)/S(x). The discrete time measure has the inconvenient side effect that the probability to die depends on the length of the age-interval x. In order to get rid of this problem we take the continuous limit µ(x) = lim = 1x→0

˙ S(x) S(x) − S(x + 1x) =− , 1xS(x) S(x)

(1)

which provides the force of mortality: the conditional probability of dying at age x given survival up to age x. For empirical applications and small 1x, the force of mortality can be approximated, for example, by µ(x) = log [(S(x − 1x)/S(x + 1x)] /(21x) (Sacher, 1956). The notion of the force of mortality should be familiar not only to engineers (as the failure rate) but also to economists as the hazard rate; the hazard here is to die at age x. The “perpetual youth” model built on Blanchard (1985) and Yaari (1965) investigates the special case µ(x) = λ for all x. The assumption that the probability of dying at age x is independent from x, constitutes, in fact, the definition of a non-ageing organism in biology (Arking, 2006). Thus, the most popular economic models that account for probabilistic death and the finiteness of human life are inherently unsuitable to display and discuss aspects of human ageing. A feature, that is acknowledged by coining the “perpetual youth” expression.1

25.3

The Gompertz-Makeham Formula

The force of mortality for humans increases with age in a particular way. Figure 25.1 shows this for U.S. American men in the period 2010–2019. A structurally similar figure can be drawn 456

Measuring Ageing

for other human subpopulations (Strulik and Vollmer, 2013). In the figure, dots represent the actual age-specific force of mortality. Human life can be roughly subdivided into two periods: an initial phase of development ranging from birth to puberty during which the probability of dying decreases, immediately followed by the phase of ageing during which the probability of dying increases. Remarkably, almost nowhere along the human life cycle is the probability of dying constant.2 The most striking feature of Figure 25.1 is the long period of life, ranging from about 40 to 90 years of age, for which age and the force of mortality are log-linearly related. This relationship is known as Gompertz law after actuary Benjamin Gompertz (1825) who first observed and stated it formally: log µ(x) = a + αx, or equivalently, µ(x) = R exp(αx).

(2)

For humans, the estimate of α is around 0.09 implying that the probability of dying doubles approximately every 8 years. The Gompertz formula was further improved by William Makeham (1860) who added a constant reflecting an age-unrelated force of mortality to yield the famous Gompertz-Makeham formula, µ(x) = A + R exp(αx).

(3)

While the Gompertz formula approximates human mortality reasonably well for ages between 40 and 90 years, the Makeham amendment yields a good approximation also for ages between 15 and 40 years (Carnes et al., 2006). Subsequently, many researchers from various disciplines have tried to further improve the formula to little avail. Taking both simplicity and precision into account, the Gompertz-Makeham formula is to the present day the most appropriate, concise, and widely used formal description of ageing (Olshansky and Carnes, 1997). Not only for humans are its parameters estimated with great precision with a coefficient of determination above 0.9, also species as different as yeast, fruitflies, rats, and horses have been shown to age according to the Gompertz-Makeham formula. The estimated coefficients, of course, differ greatly, reflecting the large variation in lifespan across species (Arking, 2006; Gavrilov and Gavrilova, 1991).

age-specific mortality rate

1 0.1 0.01 0.001 0.0001 0

20

40

60

80

100

age Figure 25.1 Age-specific mortality rate: U.S. American men 2010–2019. Note: Dots indicate data points; period data from HMD (2020). The straight line shows the Gompertz estimate, µ = Reα×age with R = 4 · 10−5 and α = 0.093. Semi-logarithmic scaling of axes.

457

Holger Strulik

Notice that µ(x) is finite for finite x; there exists no age x at which µ(x) converges toward a pole. Or more vividly put “no matter how old one is, the probability to die the next day is never unity.” With respect to economic modeling, the Gompertz-Makeham law refutes the frequently made assumption of a given “capital T” that marks the end of life. Figure 25.2 compares mortality according to Gompertz-Makeham with mortality according to the perpetual youth assumption. Solid lines reflect estimates for Swedish men born 1901– 1910. Dash-dotted lines show perpetual youth outcomes for λ = 0.025. In the panel on the left-hand side, the Gompertz part begins to dominate at ages around 50, from that point the force of mortality appears to be log-linearly increasing in age. The perpetual youth model clearly overestimates mortality of young adults and underestimates it for the old. The middle panel of Figure 25.2 shows the probability of surviving to age x . It is obtained as the solution ˙ of S(x) + µ(x)S(x) = 0 given that S(0) = 1, such that using (3),  
R exp(αx) − 1 (4) S(x) = exp −Ax − α

Prob. to survive to age x

force of mortality (μ)

1 0.1 0.01

0.001 20

40

60 age (x)

80

100

life expectancy at age x

and, trivially, for the perpetual youth agent S(x) = exp(−λx). According to GompertzMakeham, the probability of surviving decreases relatively little with age at young ages and strongly at old ages. The perpetual youth model predicts just the opposite, i.e., steeply falling survival prospects at young ages. The panel on the right-handR side of Figure 25.2 shows life expectancy (expected remaining ∞ years to live) at age x, e(x) = x S(a)da/S(x). For the perpetual youth agent life expectancy is independent of age (and equals 1/λ) while, actually, according to Gompertz-Makeham life expectancy is almost linearly falling with age in the range of 20–70 years, after which it levels off. At the upper age range, for the oldest old, the Gompertz-Makeham formula loses precision. In contrast to what one might have expected, at ages of above 90 the force of mortality increases less than log-linearly with age. Various amendments to the original formula have been suggested to take the oldest old into account (e.g., Perks, 1932). If this leveling off becomes so strong that the force of mortality stops increasing, the oldest old indeed become non-ageing. The possibility that—once a certain age has been reached—humans may turn out to be non-ageing has inspired research and debate in demography and gerontology (Wachter and Finch, 1997; Carnes and Olshansky, 2007; Barbi et al., 2018; Gavrilova and Gavrilov, 2020; Vaupel et al., 2021). Over the last century, human life expectancy at birth has increased by more than 20 years in most fully developed countries (Riley, 2001). It is interesting to investigate how these huge

0.8 0.6 0.4 0.2 0 20

40

60 80 age (x)

100

60 50 40 30 20 10 0 20

40

60 80 age (x)

100

Figure 25.2 Ageing according to Gompertz-Makeham vs. probabilistic death. Note: Solid lines: estimates from Swedish life tables for men born 1901–1910: A = 0.00552, R = 0.000033, α = 0.1013 (Gavrilov and Gavriolova, 1991). Dash-dotted lines as for solid lines, but A = 0.00048. Dashed lines: hypothetical perpetual youth scenario: λ = 0.025.

458

Measuring Ageing

improvements of human longevity have affected the Gompertz-Makeham law. For this purpose, the notion is helpful that the Makeham-parameter A reflects the age-unrelated forces of mortality. In particular, we expect prevention, eradication, or cure of age-unrelated diseases to be reflected in A. We expect progress with respect to the ageing process itself to be reflected in a change of Gompertz parameters R and/or α. While the Gompertz parameters R and α are relatively stable over time, the background risk parameter A steeply declined over the last century. For Sweden, for example, the parameter A fell by about one order of magnitude from 0.55 percent to 0.048 percent (Gavrilov and Gavrilova, 1991). This means that so far advances in longevity manifested themselves mostly in reduced background risk. Human ingenuity has not yet been able to manipulate much the biological mechanism behind the ageing process, captured by the age-related parameters R and α. Results for A = 0.0048 are shown by dash-dotted lines in Figure 25.2. Apparently, until the 1980s, technological, economic, and cultural progress had predominantly improved the survival probabilities for young people. Life expectancy at age 80 in the 1980s differed little from what it was at the beginning of the 20th century. The structural stability of the Gompertz parameters is also helpful in explaining the ideas of “rectangularization” and “compression of morbidity” (Fries, 1980). The middle panel of Figure 25.2 conveys the information that compared with the beginning of the century a higher share of Swedish men reach an old age of, say 70 years, and then expire more quickly during their last years before death. Visually, the survival curve becomes closer to rectangular over time. On average, individuals spend more of their life in a relatively healthy condition. Across individuals, however, ageing became a more salient phenomenon: less young and intrinsically healthy people are killed from exogenous forces, implying a rising share of the population dying from age-related diseases and chronical illnesses.

25.4

Human Life Span and the Strehler–Mildvan Correlation

Whereas life expectancy is a population-specific characteristic, lifespan is usually conceptualized as a species-specific characteristic (the lifespan of humans, mice, or elephants). It has already been shown that strictly speaking such a thing as maximum human lifespan does not exist. Even under time invariance of the Gompertz-Makeham parameters, the simple fact that the “sample size” of people who ever lived on earth increases continuously implies that the maximum ever-observed life length will rise as time proceeds (Finch and Pike, 1996). Alternatively, it has been suggested to use the Gompertz parameter to infer lifespan as the age at which individuals from different populations face the same force of mortality. Although the parameters α and R show little variation over time they differ across sexes and across countries. On average, across countries, women face a lower R and a higher α (Gavrilov and Gavrilova, 1991), indicating that women have an initial advantage of lower ageing, which men eventually catch up to with rising age. In fact, a strong inverse (log-linear) association exists between R and α known as the Strehler-Mildvan correlation or the compensation effect of mortality (Strehler and Mildvan, 1960; Gavrilov and Gavrilova, 1991). Figure 25.3 shows the association between these parameters for men and women from 26 developed countries over the period 1975–1999. The estimated Strehler-Mildvan correlation is given by log Ri = log M − L · αi

(5)

where for i = 1, . . . , n the Ri and αi are the country- and sex-specific realizations of the Gompertz parameters and M and L are the invariant coefficients of the Strehler-Mildvan 459

Holger Strulik Men 1975-1999

-11

-9 -10 -11 0.08

Women 1975-1999

-10

log (R)

log (R)

-8

-12 -13

0.085

0.09

0.095

0.1

-14 0.09

0.105

0.1

0.11

0.12

0.13

Figure 25.3 The compensation effect of mortality. Note: Data for estimates of α and R for 26 countries from Strulik and Vollmer (2013); raw data from HMD (2020). Regression lines: log R = −0.76−96.6·α for men (R2 = 0.97); log R = −0.86−96.4·α for women (R2 = 0.97).

correlation. Inserting the Strehler-Mildvan correlation into the Gompertz-Makeham law µi = Ai + Ri exp(αi x) provides µi − Ai = M exp [αi (x − L)] . (6) The modified Gompertz-Makeham law implies a fixed point. It predicts that—controlling for country-specific background risk—men and women across countries share a common force of mortality M at age L. For humans, the point estimate is close to L = 96 years (Gavrilov and Gavrilova, 1991; Strulik and Vollmer, 2013). The time invariance of the Gompertz law and the Strehler-Mildvan correlation suggests that, controlling for country-specific background risk, humans share a common mechanism of ageing, a common stochastic process according to which individual bodies lose function over time and bodily failures and health deficits accumulate. A general pattern emerging from the Strehler-Mildvan correlation is that economically more advanced countries are characterized by a lower R and higher α, i.e., by a lower initial mortality rate and faster speed of ageing. This implies that the mortality rate is for all ages below L lower in the economically advanced country. An important conclusion from this observation is that ageing rapidly (i.e., being endowed with a high α) is actually not a burden but an indication of superior fitness. This notion will be further substantiated in the next section on reliability. Finally, note that if R falls continuously and α rises such that lifespan L remains constant, life expectancy improves continuously. If L is constant, however, improvements of life expectancy have an asymptotic limit. Strulik and Vollmer (2013) study the Strehler-Mildvan correlation from 1900 to 1999 and confirm that for the first half of the 20th century (and presumably also earlier in human history) the data supports the notion of an invariant human lifespan of about 87–89 years. Then, in the 1950–1974 period, female lifespan increased about 8 years to 96 years and male lifespan followed in the 1975–1999 period. Observable improvements in life expectancy during the first half of the century originated from declining background mortality (sanitation, vaccination) and from a reduction of R in association with a movement along the mortality compensation line, indicating better (initial) physiological conditions (better nutrition, e.g., Fogel and Costa, 1997). The results suggest that humans, unlike any other species, managed to increase lifespan, which increased in sync with life expectancy during the later 20th century, confirming that limits to life expectancy seem indeed to be “broken” (Oeppen and Vaupel, 2002). This “manufactured life-time” (Carnes and Olshansky, 2007) likely originated from medical technological progress, in particular the “cardiovascular revolution” (Hansen and Strulik, 2017) of the 1970s, which allowed for an extension of human life by repairing and replacing failed organs. 460

Measuring Ageing

25.5

Reliability Theory

The fact that all humans appear to age according to a common formula, which we moreover share with fruitflies, rats, and other animals, motivates the search for a common process that drives ageing. The quest for such a process is challenging because, as emphasized already, no biological clock exists. Trying to explain ageing following a line of reasoning by stating that humans age because their organs (e.g., the cardiovascular system) age, and that organs age because the tissue they are made of ages, etc., will turn out to be utterly futile. At some point a micro-level will be reached that consists of non-ageing entities, for example, atoms. Eventually, we want to explain why a system ages that consists of non-ageing elements. In explaining ageing systems, we do not have to start from scratch. We can build upon a subdiscipline in engineering, reliability theory, which was developed to understand how complicated mechanical systems consisting of non-ageing elements (like cars) are increasingly losing function over time so that the failure rate, i.e., the probability of the expiry of the system, increases with age (Barlow and Proschan, 1975). There are several reliability theories of human ageing available (Gavrilov and Gavrilova, 1991; Novoseltsev, 2006; Finkelstein, 2008). Here, we consider the basic idea by describing two particularly straightforward models. The presentation follows Gavrilov and Gavrilova (1991, 2001), who were the first to integrate reliability research into the realm of biology. Consider an organism constructed of n non-ageing blocks. Non-ageing means that the failure rate λ is constant over time. Given age x the probability a block will fail is 1 − exp(−λx). Blocks are connected in parallel, and the organism lives as long as at least one block is in order. The probability that the organism expires before age x is given by F(x) = [1 − exp(−λx)]n and the probability of surviving to age x is S(x) = 1 − F(x). The unconditional probability of dying at age x is thus given by dS/dx = −λn exp(−λx)[1 − exp(−λx)]n−1 and the force of mortality is µ(x) =

λn exp(−λx)[1 − exp(−λx)]n−1 . 1 − [1 − exp(−λx)]n

Taking an approximation at young ages, when 1 − exp(−λx) ≈ λx, the expression simplifies to µ(x) ≈ nλn xn−1 and, for old ages, using L’Hospital’s rule, we obtain limx→∞ µ(x) = λ. The simple model is thus capable of explaining ageing: the force of mortality µ(x) increases with age x. Ageing is explained as a loss of redundancy over time. This notion of ageing as accelerated loss of organ reserve is in line with the mainstream view in the medical science. For example, initially, as a young adult, the functional capacity of human organs has been estimated to be tenfold higher than needed for survival (Fries, 1980). The meta study of Sehl and Yates (2001) computes for 13 human organ systems a mean loss of functional capacity of 0.65 percent per year. The model is also useful in explaining mortality of the oldest old. With increasing age the organism loses redundancy until survival depends, in the limit, just on the survival of the last functioning non-ageing block. Thus the organism converges toward the constant mortality rate λ of its non-ageing elements. This is a general result from reliability theory: the rate of ageing, i.e. the age-dependent component of the mortality rate (the failure rate), is increasing in the complexity (redundancy) of the system. We can expect that, for example, light bulbs and bacteria age at a much lower rate than cars or humans. Seen this way, ageing appears as a positive trait. It occurs because complex systems start out at a high level of redundancy and are thus less likely to expire at a young age. The problem with the simple model is that the derived µ(x) at young ages does not obey the Gompertz law. It follows—similar to the failure rate of mechanical systems—a Weibull distribution. To describe the ageing process of humans, the model has to be made “more human.” 461

Holger Strulik

Next we consider one of several possible extensions. Suppose an organism consists of m irreplaceable blocks, i.e., blocks connected in series such that the organism dies if one block fails. Each block consists of n elements, connected in parallel with age-independent failure rate λ. Following the previous computations we know that the failure rate of a block is nλn xn−1 for small x and approaches the constant λ for large x. Because blocks are connected in series (each of them being essential), the failure rate of the organism equals the sum of the failure rates of blocks i.e., m · nλn xn−1 for small x and mλ for large x. Next suppose that many elements are initially defective and that the probability of an initially functioning element is given by q. The failure rate of a block with i initially functioning elements is thus µB (i) = iλi xi−1 for small x and µB (i) = λ for large x. Blocks, ordered according to their number of initially functioning elements, are binomially distributed. We approximate the binomial with a Poisson distribution: ki P(i) = c · exp(−k) , i! where k ≡ nq is the mean number of initially functioning elements and c is a normalizing constant ensuring that the sum of probabilities equals one. The failure rate of the system, computed as the sum of the failure rate of blocks, is then obtained as µ(x) =

n X

mP(i)µB (i) = mc · exp(−k)

i=1

n X ki i=1

i!

µB (i).

Now consider a complex, redundant organism with a large number of elements. In the limit, for n → ∞, it can be shown that µ(x) can be approximated by ( Reα·x for small x µ(x) ≈ (7) c · m · λ for large x with α = kλ and R ≡ mcλk exp(−k) (Gavrilov and Gavrilova, 1991). The organism ages according to Gompertz law. For large ages (the oldest old) the force of mortality converges to a high plateau. Taking logs of R we get log R = log(cmkλ) − k and inserting α = kλ we arrive at log R = log M − αL with M = αmc and L = 1/λ, which is the Strehler-Mildvan correlation. The reliability model is capable of generating the most important regularities of human ageing. For an interpretation of parameters, first note that L is uniquely pinned down by λ, the age-independent failure rate of an element. If λ is a species-specific constant, then the model predicts a unique focal point (lifespan) L for the species. Across species, L depends inversely on the robustness of its non-ageing elements. If we reasonably assume that m, the number of irreplaceable blocks, is also a species-dependent constant, then all variation within a species results from variation of k, the mean number of initially functioning elements. The parameter k is a compound parameter k = nq. Consider first variation in n, the number of elements per block. For humans, these differences could exist across countries because their citizens are on average of different size. This could in principle come through country-specific diet and/or Darwinian selection because optimal body size depends on ambient temperature (Bergmann, 1847; Dalgaard and Strulik, 2015). The model thus predicts in line with the evidence a positive association between lean body size (height) and life expectancy (Waaler, 1984; Peck and Vagero, 1989; Koch, 2011). It rationalizes why health and life expectancy improve as individuals, on average, grow taller, a phenomenon associated with economic development during the 20th century (Dalgaard and Strulik, 2016; Dalgaard et al., 2022). The explanation 462

Measuring Ageing

offered by reliability theory is increasing redundancy. Larger humans have a larger organ reserve to live off. Next, hold n constant and consider variation of q, the probability for an element to be initially functioning. The idea here is that available nutrition (for mother and child) and disease exposure early in life shape q. Taking the historic improvement in nutrition and health into account, the model predicts that individuals age faster today than one hundred years ago because they start out better at young age, endowed with more functioning organ reserve. Consequently, survival prospects and life expectancy have improved at any age. The model predicts “fetal origins” (Barker, 1995; Almond and Currie, 2011; Dalgaard et al., 2021a), i.e., lasting consequences of early events in life on late-life mortality and morbidity. Because individuals are also predicted to be taller because of the generally improved conditions (Fogel, 1994) and because of less exposure to inflammation and infections (Crimmins and Finch, 2006), mortality improves due to the simultaneous and amplifying effect of increasing n and q on k = nq. While the model does a fairly good job of explaining basic mechanisms of human ageing, a better approximation of the actual ageing process can be achieved by adding more details. Extensions of the basic model include cascading effects, i.e., the phenomenon that failure of one element of the system entails failure of other elements (Gavrilov and Gavrilova, 1991), mechanisms of imperfect (cell-) maintenance and repair (Finkelstein, 2008); and mechanisms for redundancy expansion in early life (Milne, 2008). Allowing more complex interaction between elements leads to network theories of ageing at the molecular level (based on Kowald and Kirkwood, 1996) and at the level of functional limitations and frailties (Rutenberg et al., 2018). A common characteristic of all reliability-based models is that organisms are conceptualized as complex systems consisting of essential parts (e.g., organs, tissue) connected in series, which are in turn built of smaller entities (such as cells and molecules) connected in parallel. Parallel connectivity means that every reliability theory is built upon the idea of redundancy. Another common theme is a stochastic failure rate for the basic entities. The notion of ageing as driven by a “natural” stochastic process helps to explain the “unfair” nature of human fate, i.e., why we actually observe large differences of ageing on the individual level. Reliability theory can explain why individuals raised under equal conditions and/or built from the same genes (monozygotic twins) can age and eventually die in very different ways. At the same time, the models provide a toehold to explain how genetic endowments, population- (e.g., country-) specific characteristics, and the environment early in life have bearing on aggregate ageing behavior of populations. In short, it integrates “good luck” as a major third important driver of longevity besides “good genes” and “good behavior.” The view that ageing and death are best conceptualized as accidental stochastic shocks at the molecular level is firmly established in the natural sciences and has entered biology textbooks (Arking, 2006).

25.6

The Frailty Index

Although age is a powerful predictor of mortality at the aggregate level, it is a relatively poor predictor at the individual level, and biologists are tirelessly emphasizing that ageing should not be conceptualized as a time-dependent but as an event-dependent process. We thus turn next to the measurement of ageing as it is expressed in individuals and the question of how it relates to chronological age. Physiological ageing is complex and several proposals of complex measurements of the state of health have been made, including, for example, healthy (disability-free) life expectancy (Jagger and Robine, 2011); activities of daily living (Katz, 1983); the healthy ageing index (Sanders et al., 2014; O’Connell et al., 2019); the healthy ageing score; or allostatic load scores of biomarkers (Seeman et al., 2001); see Michel et al. (2019) 463

Holger Strulik

for an overview. Here, we focus on a particularly straightforward and useful methodology, the frailty index. The frailty index, developed by Mitnitski and Rockwood and several coauthors in a series of articles (based on Mitnitski et al., 2001), simply records the fraction of a large list of ageingrelated health conditions that is present in an individual. The list of potential deficits ranges from mild ones to near-lethal ones, and it has been shown that it does not matter which particular health deficits are included in the unweighted index as long as there are sufficiently many. The intuition for this remarkable feature is that health deficits are connected to other health deficits. For example, hypertension is associated with the risk of stroke, heart diseases, kidney diseases, dementia, and problems of walking fast and sleeping well. This means that if a particular health deficit is missing from the list, its effect (on, for example, probability of death) is taken up by a combination of other health deficits. The health deficit index has a microfoundation in reliability theory and in a network theory of human ageing (Rutenberg et al., 2018). As the index rises toward one, the individual is viewed as increasingly frail and in this sense physiologically older. The gradual loss of functional capacity of human organs is expressed as a gradual increase of the frailty index. The index thus captures in one number the state of health and the biological ageing process defined as the intrinsic, cumulative, progressive, and deleterious loss of function. The literature has outlined five criteria for health deficits to be included in the frailty index (Searle et al., 2008): (1) A deficit’s prevalence must generally increase with age, although some clearly age-related adverse conditions can decrease in prevalence at very advanced ages due to survivor effects. (2) The deficits need to be associated with health status. Graying hair, for example, would be inadmissible although it is obviously age-related. (3) The chosen deficits must not saturate too early. For example, as humans age, focusing on close objects becomes harder (presbyopia); by around age 55, the disease is nearly universal and thus less than ideal to include. (4) The deficits that make up a frailty index must cover a range of systems. If the index becomes too narrowly focused, say on cognitive deficits, it potentially no longer captures overall ageing but simply cognitive ageing. (5) To ensure that the index is independent from the specific deficits considered, a sufficiently large number of deficits—30–40—needs to be included. The quality of the frailty index has been demonstrated by its predictive power for death at the individual level and for mortality at the group level and for other adverse health outcomes such as the risk of institutionalization in nursing homes and becoming a disability insurance recipient (Rockwood and Mitnitski, 2007; Hosseini et al., 2022). Another reason for the popularity of the frailty index is that it is easily comparable across samples, datasets, and populations (Searle et al., 2008). On average, individuals accumulate health deficits exponentially with age. The frailty index increases by about 2–5 percent from one birthday to the next. This regularity, akin to the Gompertz law, has been shown first for Canadians (Mitnitski et al., 2002a,b; Mitnitski and Rockwood, 2016) and replicated for populations from Europe, the United States, and across the world (Harttgen et al., 2013; Abeliansky and Strulik, 2018a; Abeliansky et al., 2020; Dalgaard et al., 2022). The exponential accumulation of health deficits suggests that biological ageing is a self-productive process (Dragone and Vanin, 2022), in which the presence of many health deficits is conducive to the faster development of new deficits. On average, women, at a given age, display more health deficits than men and men develop new health deficits faster than women (Mitnitski et al., 2002a,b, 2005; Yang and Lee, 2010; Mitnitski and Rockwood, 2016; Gordon et al., 2017; Abeliansky and Strulik, 2018a). Because mortality is usually lower for women than for men and because the frailty index has been shown to be highly predictive of mortality, these observations contribute to the morbiditymortality paradox. Apparently, a given score of the frailty index exerts a stronger effect on 464

Measuring Ageing

mortality for men (Mitnitski et al., 2002a,b; Gu et al., 2009; Romero-Ortuno and Kenny, 2012; Mitnitski and Rockwood, 2016). Potential explanations of the paradox within the frailty-index literature include the features that women suffer more often from nonlethal health deficits and that women visit doctors more often and report more diagnoses of deficits. The rich literature on the morbidity-mortality paradox also discusses explanations outside the frailty-index paradigm, including genetic gender differences, immune system responses, hormones, disease patterns, and gender differences in health behavior as potential explanations (e.g., Bird and Rieker, 1999; Case and Paxson, 2005; Oksuzyan et al., 2008; Schuenemann et al., 2017). As an example, we next consider some of these regularities within Europe. Abeliansky and Strulik (2018a) use the Survey of Health, Ageing and Retirement in Europe (SHARE dataset release 5.0.0, B¨orsch-Supan et al. (2013) and consider four waves with health-related information that took place between 2004 and 2013. The study computed a frailty index consisting of 38 symptoms, signs, and disease classifications for individuals aged 50–90 from the 10 countries that participated in all four waves: Austria, Belgium, Denmark, France, Germany, Italy, Netherlands, Sweden, Spain, and Switzerland. In log-linear regressions akin to the Gompertz law and controlling for individual fixed effects, the point estimate of the age coefficient was 0.023 for women and 0.026 for men. Inspired by the Gompertz-Makeham law of mortality and its adaption to frailty analysis by Mitnitski et al. (2002a,b), nonlinear regression was used to estimate Dag = Ag + Rg · exp (αg · ageag ) + ϵag .

(8)

0

0.6

-0.5

0.5 0.4 0.3 0.2 0.1

0

-1 -1.5

Austria Germany Sweden Netherlands Spain Italy France Denmark Switzerland Belgium

-2 -2.5 -3 -3.5 -4

0 50

60

70

80

age

90

100

50

60

70

80

90

age

100

110

120

Log health deficits (men)

0.7

Log health deficits (women)

frailty index (D)

To get reliable estimates, individuals were binned in 1-year age groups. The index a refers to age group and g refers to gender (male, female). Results using data pooled over all waves and countries are shown in the panel on the left-hand side of Figure 25.4. Women are represented by filled circles and men by empty circles. The predictions from the regression are shown by solid and dashed lines. The quasi-exponential growth equation fits the data well (R2 = 0.99). It shows that men start out healthier and then accumulate health deficits more quickly. At about age 103 the regression lines intersect. We next consider results from regressions of (8) for each country separately. Interestingly, the country-specific parameters of R and α, denoted by Rc and αc , are negatively associated, akin to the Strehler-Mildvan correlation. Regressing log(Rc ) = β + Fαc provides the estimate F = −103.2 (±2.6), suggesting a common age of about 103 years at which European men and women display, on average, the same frailty index. The center and right panels of Figure 25.4

-1 -2 -3 -4 -5 -6 50

60

70

80

90

100

110

120

age

Figure 25.4 Ageing within and across 10 European countries. Note: Left panel: frailty index by age for European women (filled circles) and men (empty circles) and prediction from nonlinear regression (solid and dashed lines). Center and right panels: Health deficit accumulation across countries, Dc = exp(β) exp(αc (age + F)), in which αc stems from estimating (8) by country and β and F stem for the compensation law regression log(Rc ) = β + Fαc , logarithmic scale of Dc .

465

Holger Strulik

illustrate the estimation results. For better visibility, deficits are displayed in logs and results are shown for men and women separately. We observe a compensation effect of morbidity: in countries where individuals are healthier at initial age 50, they experience subsequently faster ageing, confirming the prediction from reliability theory that initially more robust systems age faster. The intersection is about at D = 0.7, a value that has been suggested as the maximum observable frailty index (Rockwood and Mitnitski, 2006). The frailty index has been developed to measure visible health deficits in elderly persons. Basic principles of reliability theory suggest that damage of body cells occurs at any age (Kirkwood, 2005) albeit at young ages these micro damages are mostly not yet visible in form of diagnosed health deficits and functional limitations.3 To assess ageing in young adults, Belsky et al. (2015) use biomarkers to measure the physiological deterioration of multiple organ systems (pulmonary, periodontal, cardiovascular, renal, hepatic, and immune function). They collected 18 biomarkers for a cohort of New Zealanders of which only 1 percent had been diagnosed with an age-related chronic disease. Biological age at chronological age 38 was computed from biomarker function and found to be normally distributed ranging from age 28 to 61. Because biomarkers were also collected at age 26 and 32, age patterns of biomarkers could be obtained, and biologically older individuals were found to age at a faster rate. On average, each year increase of biological age was associated with a 5 percent faster deterioration of biomarkers, implying that deficits in organ systems grow similar to health deficits in elderly persons. Levine (2013) discusses the use of biomarkers for the computation of biological age as a predictor of mortality. Blodgett et al. (2017) compute a frailty index from abnormal laboratory test results and show that it correlates with and displays similar characteristics to the conventional frailty index. The lab-based index is higher at young ages and increases less steeply with age than the conventional index, indicating that deficits arise first at the cellular level before they become visible as accumulated damage at the organ and tissue level.

25.7

Individual Ageing

Whereas health deficits appear to increase monotonically with age at the population level, there is scope for repair and recovery at the individual level. Moreover, individual health trajectories are highly idiosyncratic, reflecting individual-specific exposure to stressors and individualspecific recovery time. The self-productive feature of health deficit accumulation, however, is also visible at the individual level. Specifically, health deficits have been shown to be approximately Poisson distributed whereby the probability of an individual with n deficits to display k deficits next period is given by Pn (k) = An (ρn )k /k! (Mitnitski et al., 2006), in which ρn is the mean and the variance. Introducing death as an absorbing state assumed with probability Pn (d) and requiring that probabilities add up to one, the normalizing constant can be eliminated and the transition probability can be written as Pn (k) = e−ρn (1 − Pn (d))

(ρn )k . k!

(9)

Mitnitski et al. (2006) show that a simple linear approximation, ρn = a1 + b1 n and Pn (d) = a2 + b2 n fits the observed health transition of Canadians aged 65+ quite well (R2 = 0.98). Stochastic changes in individual health status are thus described by an age-independent Markov chain such that the mean and variance of health deficits next period increases with the number of currently present health deficits. These results are illustrated in Figure 25.5. While transiting back to a state with fewer health deficits is always possible, the general trend goes toward the accumulation of more deficits capturing ageing as the cumulative, progressive, and deleterious 466

Measuring Ageing

0.3 n=0

probability (P (k)) n

0.25 0.2

n=2

0.15

n=5

0.1 0.05 0 0

5

10

15

health deficits (k) Figure 25.5 Health deficit transitions. Note: Probability distribution of health deficits (k) for individuals with n health deficits 2 years ago. Estimation for Canadians aged 65+ (estimates from Mitnitski et al., 2006).

loss of function that eventually culminates in death. Hosseini et al. (2022) refine the stochastic process for frailty dynamics with age-dependent variance. Grossmann and Strulik (2019) discuss implications for health inequality and public policy.

25.8

Ageing of Populations

Finally, we extend the methodology, which was until recently applied exclusively to individuals, to the use of macro data by computing the frailty index of populations, following Dalgaard et al. (2022). Consider the definition of the frailty index of individual i from country c, Dic =

N 1 X 1ic (d) , N

(10)

d=1

where 1ic (d) is an indicator function that takes on the value 1 if individual i suffers from deficit d and N is the number ofPpotential deficits. The average frailty index in country c, Dc , is computed as Dc = (1/Pc ) Pi c Dic , where Pc is the size of the population in country c. Using (10), a simple rearrangement of the sum allows us to write the average frailty index as Dc =

N 1 X Pdc , N Pc

(11)

d=1

where Pdc is the number of people in country c that suffer from deficit d. The aggregate frailty index is then simply computed as the average of N prevalence rates, Pdc /Pc , in each country. P Analogously, the frailty index for an age group a in country c is computed as Dac = (1/N) N d=1 (Pdac /Pac ), where Pdac /Pac is the prevalence rate of d within age group a in country c. Frailty indices for men and women were constructed analogously. Dalgaard et al. (2022) obtained data on prevalence rates for men and women aged 20–94 for 5-year age groups from the Global Burden of Disease Study for the period 1990– 2019 (Vos et al., 2020). Applying the methodology of selecting items for the index as explained previously, 467

Holger Strulik

they construct a frailty index consisting of 32 health conditions for 201 countries. Controlling for country and year fixed effects, Dalgaard et al. (2022) find that the average woman in the world accumulates 2.7 percent more health deficits from one birthday to the next. The rate is 2.8 for men, who start out healthier and display fewer deficits until about age 80. These growth rates are astonishingly stable, varying between 2.6 and 2.9 percent when different subdivisions of the world were considered. Next, Dalgaard et al. (2022) explore the mortality-morbidity nexus at the population level, for a selected group of countries for which age-specific mortality rates were available from HMD (2020). In log-log regressions, they find an elasticity of mortality rate with respect to frailty of 3.0 for men and 2.8 for women. These numbers are very robust to the inclusion of country and year fixed effects. Combined with the previous results, they confirm the morbidity-mortality paradox at the population level. Women display, on average, greater frailty but are less likely to die at the same level of the frailty index. The numbers are also intuitive. Recalling Gompertz law of mortality log µ = α M age, with α M ≈ 0.09 and inserting health deficit accumulation log(D) = α D age with α D ≈ 0.03 suggests a mortality elasticity of α M /α D = 3.

25.9

Discussion and Conclusion

This chapter introduced two measures of human ageing: the force of mortality and the frailty index. Both measures grow exponentially in chronological age. Chronological age, however, does not cause physiological ageing. Instead, ageing can be understood as the deterioration of a complex and highly redundant system experiencing stochastic damage at the subcellular level. This view, formalized in reliability theory, can explain the exponential increase of health deficits and mortality, the compensation laws of mortality and morbidity, and other regularities of human ageing. The frailty index has been introduced in health economics by the health deficit model (Dalgaard and Strulik, 2014). The health deficit model computes the derivative of the frailty index ˙ = αD − A, which is taken into account as a state equation in eco(8) with respect to age, D nomic life-cycle modeling. In the deterministic version of the model, death occurs when an upper limit of health deficits has accumulated. In the stochastic version, survival probability is a negative function of the frailty index. The parameter A is conceptualized as the main gateway through which health behavior (such as, for example, smoking or exercising) affects ageing: it lets health deficit grow (somewhat) faster or slower than exponentially. The model predicts that health expenditure increases with age, but it is also consistent with the observation that age per se does not much affect health expenditure once time to death is controlled for (Zweifel et al., 1999; Schuenemann et al., 2022). Current and future medical progress promises to reduce the natural force of ageing α, perhaps eventually to a level at which health investments in prevention and repair allow humans to become non-ageing (De Grey, 2013; Lopez-Otin et al., 2013; Sinclair and LaPlante, 2019). For a first discussion of these ideas in an economic life-cycle model, see Dragone and Strulik (2020). The main feature of the health deficit model is that it implements the self-productive nature of health deficits: the presence of many health deficits is conducive to the faster development of new deficits. Considering two individuals of the same age, the one in worse health is predicted to develop more health deficits in the next period. The feature of self-productivity captures human ageing as intrinsic, cumulative, progressive, and deleterious loss of function that eventually culminates in death, and it explains the exponential (or, more generally, convex) association of morbidity and mortality with chronological age. 468

Measuring Ageing

Against this background, it is remarkable that the health capital model (based on Grossman, 1972), which for a long time served as the main paradigm in life-cycle health economics, assumes just the opposite. There, health capital H is assumed to represent the state of health, whereby more health capital identifies a healthier person, and that health capital depreciates at ˙ = −δH. The process of health capital some positive rate δ, such that the loss of health is H depreciation is self-depleting (Dragone and Vanin, 2022): the presence of much health capital is conducive to larger losses of health capital. Considering two individuals of the same age, the healthier one (endowed with more health capital) is predicted to lose more health capital (faster decline of health) in the next period. Notice that introducing an age-dependent depreciation rate cannot “repair” the prediction of faster physiological ageing of physiologically younger individuals. This counterfactual feature can be regarded as the origin of the many shortcomings and limitations of the health capital model highlighted in the literature (e.g., Wagstaff, 1986; Zweifel and Breyer, 1997; Case and Deaton, 2005; Almond and Currie, 2011; Strulik, 2015; Dalgaard et al., 2021a). From the empirical side, the greatest limitation of the health capital model is perhaps that its main object of investigation, health capital, is unobservable and alien to medical scientists. Health capital may be constructed from the absence of health deficits or approximated by selfassessed health status, which likely depends on the presence of health deficits. The frailty index, in contrast, provides an opportunity to formulate, calibrate, and test health economic life-cycle models with a straightforward metric that measures health deficits directly. The measure has an established methodology in gerontology where it has been applied in hundreds of studies. It allows economists and medical scientists to discuss human ageing using a common language and methodology.

Notes 1 Of course, this criticism also applies to the deterministic overlapping generation (OLG) model, of which the popular two-period OLG model (based on Diamond, 1965) is a special case. From an ageing perspective, it should be called the “blade runner model” (perfect health until T, then death). 2 This rough division of human life ignores that the age at the trough of the mortality rate declined over the last century and is now reached in fully developed countries around age 10, i.e., several years before sexual maturity (Milne, 2006). Moreover, ageing understood as the increasing loss of bodily function starts at least at birth if not earlier because infants are already subject to somatic mutations and telomere shortening (Frenck et al., 1998; Kirkwood, 2005). See Dalgaard et al. (2021a) for a model of human ageing from conception to death. 3 Exploring the influence of early-life and in utero circumstances on old-age frailty is possible, however. Abeliansky and Strulik (2018b) show that individuals exposed to hunger in childhood develop health deficits faster in old age, and Abeliansky and Strulik (2020) show how the season of birth is related to late-life health deficit accumulation.

References ABELIANSKY, A. L., EREL, D., AND STRULIK, H. (2020): “Ageing in the USA: Similarities and disparities across time and space,” Scientific Reports, 10(1): 14309. ABELIANSKY, A. L., AND STRULIK, H. (2018a): “How we fall apart: Similarities of human ageing in 10 European countries,” Demography, 55(1): 341–359. ABELIANSKY, A. L., AND STRULIK, H. (2018b): “Hungry children age faster,” Economics & Human Biology, 29: 211–220. ABELIANSKY, A. L., AND STRULIK, H. (2020): “Season of birth, health and ageing,” Economics & Human Biology, 36: 100812. ALMOND, D., AND CURRIE, J. (2011): “Killing me softly: The fetal origins hypothesis,” Journal of Economic Perspectives, 25(3): 153–72. ARKING, R. (2006): The Biology of Ageing: Observations and Principles. Oxford: Oxford University Press.

469

Holger Strulik

BARBI, E., LAGONA, F., MARSILI, M., VAUPEL, J. W., AND WACHTER, K. W. (2018): “The plateau of human mortality: Demography of longevity pioneers,” Science, 360(6396): 1459–1461. BARKER, D. J. P. (1995): “Fetal origins of coronary heart disease,” British Medical Journal, 311 (6998): 171–174. BARLOW, R. E., AND PROSCHAN, F. (1975): Statistical Theory of Reliability and Life Testing: Probability Models. New York: Holt, Rinehart, and Winston. BELSKY, D. W., CASPI, A., HOUTS, R., COHEN, H. J., CORCORAN, D. L., DANESE, A., ... SUGDEN, K. (2015): “Quantification of biological ageing in young adults,” Proceedings of the National Academy of Sciences, 112(30): E4104–E4110. BERGMANN, C. (1847): “Ueber die Verh¨altnisse der W¨arme¨okonomie der Thiere zu Ihrer Gr¨osse,” G¨ottinger Studien, 3, 595–708. BIRD, C. E., AND RIEKER, P. P. (1999): “Gender matters: An integrated model for understanding men’s and women’s health,” Social Science & Medicine, 48(6): 745–755. BLANCHARD, O. J. (1985): “Debt, deficits, and finite horizons,” Journal of Political Economy, 93(2): 223–247. BLODGETT, J. M., THEOU, O., HOWLETT, S. E., AND ROCKWOOD, K. (2017): “A frailty index from common clinical and laboratory tests predicts increased risk of death across the life course,” Geroscience, 39(4): 447–455. BLODGETT, J. M., THEOU, O., HOWLETT, S. E., WU, F. C., AND ROCKWOOD, K. (2016): “A frailty index based on laboratory deficits in community-dwelling men predicted their risk of adverse health outcomes,” Age and Ageing, 45(4): 463–468. BORKAN, G. A., AND NORRIS, A. H. (1980): “Assessment of biological age using a profile of physical parameters,” Journal of Gerontology, 35: 177–184. ¨ BORSCH -SUPAN, A., BRANDT, M., HUNKLER, C., KNEIP, T., KORBMACHER, J., MALTER, F., ... ZUBER, S. (2013): “Data resource profile: The Survey of Health, Ageing and Retirement in Europe (SHARE),” International Journal of Epidemiology, 42(4): 992–1001. CARNES, B. A., HOLDEN, L. R., OLSHANSKY, S. J., WITTEN, M. T., AND SIEGEL, J. S. (2006): “Mortality partitions and their relevance to research on senescence,” Biogerontology, 7(4): 183–198. CARNES, B. A., AND OLSHANSKY, S.J. (2007): “A realist view of ageing, mortality, and future longevity,” Population and Development Review, 33(2): 367–381. CASE, A., AND DEATON, A.S. (2005): “Broken down by work and sex: How our health declines.” In: Wise, D. A. (ed.), Analyses in the Economics of Ageing, Chicago: University of Chicago Press, pp. 185–212. CASE, A., AND PAXSON, C. (2005): “Sex differences in morbidity and mortality,” Demography, 42(2): 189–214. COMFORT. (1979): The Biology of Senescence, 3rd edition, New York: Elsevier. CRIMMINS, E. M. (2004): “Trends in the health of the elderly,” Annual Review of Public Health, 25: 79–98. CRIMMINS, E. M., AND FINCH, C. E. (2006): “Infection, inflammation, height, and longevity,” Proceedings of the National Academy of Sciences, 103(2): 498–503. DAHL, R., AND FOREMAN, M. (1987): The Complete Adventures of Charlie and Mr Willy Wonka. London: Puffin Books. DALGAARD, C. J., HANSEN, C. W., AND STRULIK, H. (2021a): “Fetal origins: A life cycle model of health and ageing from conception to death,” Health Economics, 30(6): 1276–1290. DALGAARD, C. J., MADSEN, J. B., AND STRULIK, H. (2021b): “Physiological constraints and the transition to growth: Implications for comparative development,” Journal of Economic Growth, 26(3): 241–289. DALGAARD, C. J., HANSEN, C. W., AND STRULIK, H. (2022): “Physiological ageing around the world,” PloS One, 17(6): e0268276. DALGAARD, C. J., AND STRULIK, H. (2014): “Optimal ageing and death: Understanding the Preston curve,” Journal of the European Economic Association, 12(3): 672–701. DALGAARD, C. J., AND STRULIK, H. (2015): “The physiological foundations of the wealth of nations,” Journal of Economic Growth, 20(1): 37–73. DALGAARD, C. J., AND STRULIK, H. (2016): “Physiology and development: Why the West is taller than the rest,” Economic Journal, 126(598): 2292–2323. DE GREY, A. D. N. J. (2013): “Strategies for engineered negligible senescence,” Gerontology, 59(2): 183–189. DIAMOND, P. A. (1965): “National debt in a neoclassical growth model,” American Economic Review, 55(5): 1126–1150. DRAGONE, D., AND VANIN, P. (2022): “Substitution effects in intertemporal problems,” American Economic Journal: Microeconomics, 14(3): 791–809.

470

Measuring Ageing

DRAGONE, D., AND STRULIK, H. (2020): “Negligible senescence: An economic life cycle model for the future,” Journal of Economic Behavior and Organization, 171: 264–285. FINCH, C. E., AND AUSTAD, S. N. (2001): “History and prospects: Symposium on organisms with slow ageing,” Experimental Gerontology, 36, 593–597. FINCH, C. E., AND CRIMMINS, E. M. (2004): “Inflammatory exposure and historical changes in human lifespans,” Science, 305, 1736–1739. FINCH, C. E., AND PIKE, M. C. (1996): “Maximum lifespan predictions from the Gompertz mortality model,” Journal of Gerontology, 51(3): 183–194. FINKELSTEIN, M. S. (2008): Failure Rate Modeling for Risk and Reliability, London: Springer. FOGEL, R. W. (1994): “Economic growth, population theory, and physiology: The bearing of long-term processes on the making of economic policy,” The American Economic Review, 84(3): 369–395. FOGEL, R. W., AND COSTA, D. L. (1997): “A theory of technophysio evolution, with some implications for forecasting population, healthcare costs, and pension costs,” Demography, 34(1): 49–66. FREEDMAN, V. A., MARTIN, L.G., AND SCHOENI, R. F. (2002): “Recent trends in disability and functioning among older adults in the United States: A systematic review,” JAMA, 288, 3137–3146. FRENCK, R. W., BLACKBURN, E. H., AND SHANNON, K. M. (1998): “The rate of telomere sequence loss in human leukocytes varies with age,” Proceedings of the National Academy of Sciences of the United States of America, 95(10): 5607–5610. FRIES, J. F. (1980): “Ageing, natural death, and the compression of morbidity,” New England Journal of Medicine, 303(3): 130–135. GAVRILOV, L. A., AND GAVRILOVA, N. S. (1991): The Biology of Human Life Span: A Quantitative Approach, London: Harwood Academic Publishers. GAVRILOV, L. A., AND GAVRILOVA, N. S. (2001): “The reliability theory of ageing and longevity,” Journal of Theoretical Biology, 213(4): 527–545. GAVRILOVA, N. S., AND GAVRILOV, L. A. (2020): “Are we approaching a biological limit to human longevity?,” The Journals of Gerontology: Series A, 75(6): 1061–1067. GOMPERTZ, B. (1825): “On the nature of the function expressive of the law of human mortality, and on a new mode of determining the value of life contingencies,” Philosophical Transactions of the Royal Society of London, 115: 513–583. GORDON, E. H., PEEL, N. M., SAMANTA, M., THEOU, O., HOWLETT, S. E., AND HUBBARD, R. E. (2017): “Sex differences in frailty: A systematic review and meta-analysis,” Experimental Gerontology, 89: 30–40. GROSSMAN, M. (1972): “On the concept of health capital and the demand for health,” Journal of Political Economy, 80(2): 223–255. GROSSMANN, V., AND STRULIK, H. (2019): “Optimal social insurance and health inequality,” German Economic Review, 20(4): e913–e948. GU, D., DUPRE, M. E., SAUTTER, J., ZHU, H., LIU, Y., AND YI, Z. (2009): “Frailty and mortality among Chinese at advanced ages,” Journals of Gerontology: Series B, 64(2): 279–289. GURVEN, M., AND KAPLAN, H., (2007): “Longevity among hunter-gatherers: A cross-cultural examination,” Population and Development Review, 33: 321–365. HANSEN, C. W., AND STRULIK, H. (2017): “Life expectancy and education: Evidence from the cardiovascular revolution,” Journal of Economic Growth, 22(4): 421–450. HARTTGEN, K., KOWAL, P., STRULIK, H., CHATTERJI, S., AND VOLLMER, S. (2013): “Patterns of frailty in older adults: Comparing results from higher and lower income countries using the Survey of Health, Ageing and Retirement in Europe (SHARE) and the Study on Global AGEing and Adult Health (SAGE),” PLOS One, 8(10), e75847. HMD (HUMAN MORTALITY DATABASE). (2020): Human Mortality Database. University of California, Berkeley (USA), and Max Planck Institute for Demographic Research (Germany). Available at https://www.mortality.org. HOSSEINI, R., KOPECKY, K. A., AND ZHAO, K. (2022): “The evolution of health over the life cycle,” Review of Economic Dynamics, 45: 237–263. JACKSON, S. H. D., WEALE, M. R., AND WEALE, R. A. (2003): “Biological age – what is it and can it be measured?,” Archives of Gerontology and Geriatrics, 36, 103–115. JAGGER, C., AND ROBINE, J. M. (2011): “Healthy life expectancy.” In: Rogers, R. G., and Crimmins, E. M. (eds.), International Handbook of Adult Mortality, Dordrecht: Springer, pp. 551–568. KATZ, S. (1983): “Assessing self-maintenance: Activities of daily living, mobility, and instrumental activities of daily living,” Journal of the American Geriatrics Society, 31(12): 721–727.

471

Holger Strulik

KIRKWOOD, T. B. L. (1990): “The disposable soma theory of ageing.” In: Harrison, D. E. (ed.), Genetic Effects on Ageing II, Caldwell, NJ: Telford Press, pp. 9–19. KIRKWOOD, T.B.L. (1999): Time of Our Lives: The Science of Human Ageing, Oxford: Oxford University Press. KIRKWOOD, T. B. L. (2005): “Understanding the odd science of ageing,” Cell, 120(4): 437–447. KOCH, D. (2011): “Waaler revisited: The anthropometrics of mortality,” Economics & Human Biology, 9(1): 106–117. KONTIS, V., BENNETT, J. E., MATHERS, C. D., LI, G., FOREMAN, K., AND EZZATI, M. (2017): “Future life expectancy in 35 industrialised countries: projections with a Bayesian model ensemble,” Lancet, 389: 1323–1335. KOWALD, A., AND KIRKWOOD, T. B. L. (1996): “A network theory of ageing: The interactions of defective mitochondria, aberrant proteins, free radicals and scavengers in the ageing process,” Mutation Research/DNAgeing, 316(5–6): 209–236. LEVINE, M. E. (2013): “Modeling the rate of senescence: Can estimated biological age predict mortality more accurately than chronological age?,” Journals of Gerontology Series A: Biomedical Sciences and Medical Sciences, 68(6): 667–674. LOPEZ-OTIN, C., BLASCO, M. A., PARTRIDGE, L., SERRANO, M., AND KROEMER, G. (2013): “The hallmarks of ageing,” Cell, 153(6): 1194–1217. MAKEHAM, W. M. (1860): “On the law of mortality and the construction of annuity tables,” Journal of the Institute of Actuaries, 8(6): 301–310. MASORO, E. J. (2005): “Are age-associated diseases and integral part of ageing?” In: Masoro, E. J. and Austad, S. N. (eds.), Handbook of the Biology of Ageing, Amsterdam: Elsevier, pp. 43–62. MEDAWAR, P.B. (1952): An Unsolved Problem of Biology, London: Lewis. MICHEL, J. P., GRAF, C., AND ECARNOT, F. (2019): “Individual healthy ageing indices, measurements and scores,” Ageing Clinical and Experimental Research, 31(12): 1719–1725. MILNE, E. M. G. (2006): “When does human ageing begin?,” Mechanisms of Ageing and Development, 127(3): 290–297. MILNE, E. M. G. (2008): “The natural distribution of survival,” Journal of Theoretical Biology, 255(2): 223–236. MITNITSKI, A. B., BAO, L., AND ROCKWOOD, K. (2006): “Going from bad to worse: A stochastic model of transitions in deficit accumulation, in relation to mortality,” Mechanisms of Ageing and Development, 127(5): 490–493. MITNITSKI, A. B., MOGILNER, A. J., MACKNIGHT, C., AND ROCKWOOD, K. (2002a): “The accumulation of deficits with age and possible invariants of ageing,” Scientific World, 2: 1816–1822. MITNITSKI, A. B., MOGILNER, A. J., MACKNIGHT, C., AND ROCKWOOD, K. (2002b): “The mortality rate as a function of accumulated deficits in a frailty index,” Mechanisms of Ageing and Development, 123(11): 1457–1460. MITNITSKI, A. B., MOGILNER, A. J., AND ROCKWOOD, K. (2001): “Accumulation of deficits as a proxy measure of ageing,” Scientific World, 1: 323–336. MITNITSKI, A. B., AND ROCKWOOD, K. (2016): “The rate of ageing: The rate of deficit accumulation does not change over the adult life span,” Biogerontology, 17(1): 199–204. MITNITSKI, A., SONG, X., SKOOG, I., BROE, G. A., COX, J. L., GRUNFELD, E., AND ROCKWOOD, K. (2005): “Relative fitness and frailty of elderly men and women in developed countries and their relationship with mortality,” Journal of the American Geriatrics Society, 53(12): 2184–2189. NOVOSELTSEV, V. N. (2006): “Mathematical modeling in biology: Systems capable to live and die,” Automation and Remote Control, 67(6): 835–855. O’CONNELL, M. D., MARRON, M. M., BOUDREAU, R. M., CANNEY, M., SANDERS, J. L., KENNY, R. A., ... NEWMAN, A. B. (2019): “Mortality in relation to changes in a healthy ageing index: The health, ageing, and body composition study,” The Journals of Gerontology: Series A, 74(5): 726–732. OEPPEN, J., AND VAUPEL, J. W. (2002): “Broken limits to life expectancy,” Science, 296(5570): 1029–1031. OKSUZYAN, A., JUEL, K., VAUPEL, J. W., AND CHRISTENSEN, K. (2008): “Men: Good health and high mortality. Sex differences in health and ageing,” Ageing Clinical and Experimental Research, 20(2): 91–102. OLSHANSKY, S. J., AND CARNES, B. A. (1997): “Ever since Gompertz.,” Demography, 34(1): 1–15. PECK, A. M., AND VAGERO, D. H. (1989): “Adult body height, self perceived health and mortality in the Swedish population,” Journal of Epidemiology & Community Health, 43(4): 380–384. PERKS, W. (1932): “On some experiments in the graduation of mortality,” Journal of the Institute of Actuaries, 63(1): 12–57.

472

Measuring Ageing

RILEY, J. C. (2001): Rising Life Expectancy: A Global History, Cambridge: Cambridge University Press. ROCKWOOD, K., AND MITNITSKI, A. B. (2006): “Limits to deficit accumulation in elderly people,” Mechanisms of Ageing and Development, 127(5): 494–496. ROCKWOOD, K., AND MITNITSKI, A. B. (2007): “Frailty in relation to the accumulation of deficits,” Journals of Gerontology Series A: Biological and Medical Sciences, 62(7): 722–727. ROCKWOOD, K., SONG, X., MACKNIGHT, C., BERGMAN, H., HOGAN, D. B., MCDOWELL, I., AND MITNITSKI, A. B. (2005): “A global clinical measure of fitness and frailty in elderly people,” Canadian Medical Association Journal, 173(5): 489–495. ROMERO-ORTUNO, R., AND KENNY, R. A. (2012): “The frailty index in Europeans: Association with age and mortality,” Age and Ageing, 41(5): 684–689. RUTENBERG, A. D., MITNITSKI, A. B., FARRELL, S. G., AND ROCKWOOD, K. (2018): “Unifying ageing and frailty through complex dynamical networks,” Experimental Gerontology, 107: 126–129. SACHER, G. A. (1956): “On the statistical nature of mortality, with especial reference to chronic radiation mortality,” Radiology, 67(2): 250–258. SANDERS, J. L., MINSTER, R. L., BARMADA, M. M., MATTEINI, A. M., BOUDREAU, R. M., CHRISTENSEN, K., ... NEWMAN, A. B. (2014): “Heritability of and mortality prediction with a longevity phenotype: The healthy ageing index,” Journals of Gerontology Series A: Biomedical Sciences and Medical Sciences, 69(4): 479–485. SCHUENEMANN, J., STRULIK, H., AND TRIMBORN, T. (2017): “The gender gap in mortality: How much is explained by behavior?,” Journal of Health Economics, 54: 79–90. SCHUENEMANN, J., STRULIK, H., AND TRIMBORN, T. (2022): “Medical and long-term care with endogenous health and longevity,” The Journal of the Economics of Ageing, 23: 100400. SEARLE, S. D., MITNITSKI, A., GAHBAUER, E. A., GILL, T. M., AND ROCKWOOD, K. (2008): “A standard procedure for creating a frailty index,” BMC Geriatrics, 8(1): 1–10. SEEMAN, T. E., MCEWEN, B. S., ROWE, J. W., AND SINGER, B. H. (2001): “Allostatic load as a marker of cumulative biological risk: MacArthur studies of successful ageing,” Proceedings of the National Academy of Sciences, 98(8): 4770–4775. SEHL, M. E., AND YATES, F. E. (2001): “Kinetics of human ageing: I. Rates of senescence between ages 30 and 70 years in healthy people,” The Journals of Gerontology Series A: Biological Sciences and Medical Sciences, 56(5): B198–B208. SINCLAIR, D. A., AND LAPLANTE, M. D. (2019): Lifespan: Why We Age – and Why We Don’t Have To, New York: Atria Books. STREHLER, B. L., AND MILDVAN, A. S. (1960): “General theory of mortality and ageing,” Science, 132(3418): 14–21. STRULIK, H. (2015): “A closed-form solution for the health capital model,” Journal of Demographic Economics, 81(3): 301–316. STRULIK, H., AND VOLLMER, S. (2013): “Long-run trends of human ageing and longevity,” Journal of Population Economics, 26(4): 1303–1323. VAUPEL, J. W., VILLAVICENCIO, F., AND BERGERON-BOUCHER, M. P. (2021): “Demographic perspectives on the rise of longevity,” Proceedings of the National Academy of Sciences, 118(9): e2019536118. VOS, T., LIM, S. S., ABBAFATI, C., ABBAS, K. M., ABBASI, M., ABBASIFARD, M., ... BHUTTA, Z. A. (2020): “Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: A systematic analysis for the Global Burden of Disease Study 2019,” The Lancet, 396(10258): 1204–1222. WAALER, H. T. (1984): “Height, weight, and mortality: The Norwegian experience,” Acta Medica Scandinavica, 215(S679): 1–56. WACHTER, K. W., AND FINCH, C. E. (1997): Between Zeus and the Salmon: The Biodemography of Ageing, Washington, DC: National Academy of Sciences. WAGSTAFF, A. (1986): “The demand for health: some new empirical evidence,” Journal of Health Economics, 5(3): 195–233. YAARI, M. E. (1965): “Uncertain lifetime, life insurance and the theory of the consumer,” Review of Economic Studies, 32(2): 137–150. YANG, Y., AND LEE, L.C. (2010): “Dynamics and heterogeneity in the process of human frailty and ageing: Evidence from the US older adult population,” Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 65(2): 246–255. ZWEIFEL, P., AND BREYER, F. (1997): Health Economics, Oxford: Oxford University Press. ZWEIFEL, P., FELDER, S., AND MEIERS, M. (1999): “Ageing of population and healthcare expenditure: A red herring?,” Health Economics, 8(6): 485–496.

473

26 THE HEALTH AND RETIREMENT STUDY John W. R. Phillips and David R. Weir

26.1

Introduction

The Health and Retirement Study (HRS) is now a well-known resource for the study of ageing and uniquely so for the economics of ageing.1 It has been emulated by similar harmonized longitudinal studies around the world, creating a network of mutual support and multiple sources of innovation. The HRS familiar to researchers today has evolved greatly from its founding design, while holding to and building on its core principles of population representation, longitudinality, multidisciplinary content, and public data sharing. Because of that commitment to data sharing, the evolution and impact of HRS can readily be seen in the record of publications using the study, as shown in Figure 26.1. Shown there are the number of peer-reviewed journal publications using HRS in each year, as recorded in the HRS online bibliography (https://hrs.isr.umich.edu/publications). Figure 26.1 divides the total into papers on the economics of ageing and all others. The number of economics papers has grown steadily over the life of HRS, and there are now more than 1,300 papers on the economics of ageing. The growth in papers on other domains has been even more rapid, especially since about 2007. In 2020, 400 papers were published using HRS, bringing the total to about 3,800. Public data are especially important for training new generations of researchers, and HRS has been used in more than 700 PhD dissertations. The history of HRS can usefully, if somewhat arbitrarily, be divided into four phases. The first, from initial conception through 1996, set the foundations. Economists played key roles in its early development, and it was designed to further the study of the economics of ageing (Willis, 1999, pp. 119–145). The second, from 1998 to 2004, created the now-familiar steadystate design and broadened the scope of the study through ancillary studies. The first international partner studies began at the same time. The third phase, from 2006 to 2014, enhanced the diversity of the HRS sample and enriched its scientific content, largely through the addition of in-person interviewing to the core study. The fourth phase, begun in 2016, brought new activities outside the core.

26.2

Phase 1: HRS Origins

Since 1992, the Health and Retirement Study has received support from the National Institute on Ageing (NIA) via a cooperative award to the University of Michigan 474

DOI: 10.4324/9781003150398-31

The Health and Retirement Study

Figure 26.1 Annual peer-reviewed journal publications, by topic.

with co-funding from the Social Security Administration (SSA). Congress established the NIA to lead a national scientific effort to understand the nature of ageing to promote the health and well-being of older adults. While the primary role of SSA is the administration of retirement, disability, and survivors benefits, they also support one of 13 federal statistical agencies supporting research related to their benefit programs.2 The size and growth of older populations in the United States and around the world represent both a great achievement and an unprecedented challenge. Population ageing translates into a larger fraction of the population facing a significant risk factor for many health conditions such as Alzheimer’s disease (ageing) and dependency ratios with a large fraction of the population drawing benefits from Social Security and Medicare relative to those working to cover them. Understanding the evolving behavioral and social factors driving the health and welfare of older people as they transition out of the workforce required a significant scientific (and financial) investment. NIA’s Strategic Directions has long encouraged investments in data to support innovative research to better understand the dynamics of the ageing process and build knowledge regarding the consequences of an ageing society on population health.3 Since the HRS first interviewed participants in 1992, the study has provided a rich matrix of detailed multidisciplinary data describing the health and welfare of older adults in America as they age, data that has been used in research supported by both NIA and SSA extramural programs and the broader ageing research community. Where did the concept of an HRS come from?4 The idea for a new study started to emerge in the 1980s. The primary source of longitudinal data used to study the retirement transition at the time was the Retirement History Survey (RHS), a 10-year study based on a national sample of 11,353 people aged 58–63 when first interviewed in 1969.5 Supported by the SSA, the RHS included data on employment history, health, wealth, and retirement plans. Though the data, which included linkage to program records on earnings and benefits from Social Security, had many strengths, it was becoming clear that many emerging trends and topics could not be effectively studied using RHS. Extramural staff at NIA, other experts from federal statistical agencies, and the science community concluded that a new longitudinal study of ageing in America was overdue. 475

John W. R. Phillips and David R. Weir

Though many deserve credit for the conceptual development and eventual initiation of the HRS, it is fair to say that the catalyst was Dr. Richard Suzman, the eventual director of the Division of Behavioral and Social Research at NIA.6 Leading a collection of multiple expert meetings over several years, Dr. Suzman’s efforts identified the need for a new study. The expert group that conveyed final recommendations to NIA regarding a new study in 1988, the Ad Hoc Advisory Panel to the NIA Extramural Program on Priorities for Data Collection in Health and Retirement Economics, provided a focused description of the needs to be addressed by the study. They noted limitations of existing data to study the emerging issues of an ageing population with significant changes in demographic and labor force circumstances and that “none of the available data sets monitor retirement decisions, along with their antecedents and consequences, in sufficient numbers of individuals to draw firm conclusions” (Juster and Suzman, 1995, pp. S7–S56). Importantly, the principles provided by the experts are still core to HRS after almost 30 years: longitudinal data are essential for understanding the dynamics of ageing; linkages to administrative records are essential to obtain information that could not be readily obtained from the survey’s respondents; access to data, subject to maintaining respondent privacy and confidentiality is paramount; health and living arrangement components should be strengthened. The panel also recommended an alternative funding mechanism to the standard research grant to support enhanced involvement by the government in the study. The new study would address the underrepresentation of women, Blacks, and Hispanics in the RHS. It would include more information on physical, mental, and cognitive function to better understand the retirement and disability process—areas of interest for both NIA and SSA. Developing a protocol to consent for linkage to program data from Social Security and Medicare would be critical. Moreover, it would need to be able to support hypotheses on the retirement process across disciplines such as economics, sociology, psychology, epidemiology, demography, and biomedical disciplines. Maintaining and further developing the multidisciplinary and innovative aspects would require a process to regularly engage relevant experts regarding study content, in effect carrying forward the lessons learned from the many expert discussions that led to the study’s origin. Ultimately, after extensive discussion and deliberation among academic and federal experts that considered content, design, and various options for implementation, the study received Congressional and NIA support. Following a competitive peer-reviewed round of applications, the initial award of the HRS occurred in 1991.7 The result was a project that was true to the recommended guidance: a nationally representative sample of adults aged 51–61 including partners of any age, with oversamples of Black and Hispanic populations, with consents for federal program data linkage that provided longitudinal multidisciplinary data to study the retirement transition. Collected data were de-identified and shared publicly, while administrative data use required additional agreements to prevent inappropriate disclosures. The study was managed through a cooperative agreement between the NIA and the Institute for Social Research (ISR) at the University of Michigan with co-funding support from SSA via an interagency agreement. An investigative team with expertise in economics, medicine, demography, psychology, public health, and survey methodology at ISR, led by Dr. F. Thomas Juster and drawing key experts from the other applicant teams, designed, administered, and conducted the HRS. The cooperative agreement supporting HRS provided for significant involvement by the government in the conduct of the study relative to government involvement in a normal grant, which is generally limited to management of study scope and progress. The significant involvement came in the form of a Data Monitoring Committee (DMC) with a multidisciplinary group of subject matter experts charged with providing recommendations to the NIA and SSA regarding the activities of the study.8 The DMC also continues the collaborative scientific approach of expert input that led to the study’s creation. The first wave of HRS in 1992 completed with 12,652 interviews, 476

The Health and Retirement Study

a response rate of 81.6 percent, and initiated what would become a landmark study of ageing in the United States.9 Innovation in economic measurement was a hallmark of the first phase of HRS. Whereas income was measured in many surveys such as the Current Population Survey, financial wealth and pension entitlements were of central importance to the study of retirement and preparation for retirement and were not addressed in many economic surveys, and there are no administrative records with broad coverage. Juster’s experience with the Survey of Consumer Finances, the Federal Reserve Board’s longstanding survey of wealth, led to a particular focus on methods to overcome item non-response in self-reports. Missing data in wealth surveys was taken by some as an indication that the topic was too sensitive, or people too poorly informed, to provide good answers. Allowing people to provide answers in ranges rather than exact values was a successful approach, typically by showing respondents cards with preprinted ranges identified by letters and asking them to give the letter corresponding to their answer. This approach could not be used in the longitudinal HRS because follow-up interviews were done by telephone. There were also methodological concerns that showing respondents a set of ranges framed their answers in ways that might bias results. These challenges led HRS to the use of unfolding brackets in which a respondent who failed to provide an exact value would be asked if it was more or less than a value X, and then typically up to two additional brackets to narrow the range. Brackets were accepted well by respondents and greatly improved the accuracy of imputations for missing data (Hill, 1999, pp. 64–68). Juster was a pioneer of the use of subjective probabilities. He had first proposed it in the context of automobile purchases: asking people the percent chance they would purchase a car rather than a yes/no question. In HRS, this approach was used for similar consumer choices like working past a certain age and extended to other concepts like receipt of inheritance and survival to a certain age over which the respondent had little control (Dominitz and Manski, 1999, pp. 15–33). Survival probabilities are particularly difficult (Elder, 2013; d’Uva et al., 2020, pp. 569–589). External linkages were a key part of the initial design of the HRS. Excluding the top 10 percent of households, Social Security accounts for half of the income of Americans over 65. Linked administrative records are invaluable for measuring accurately the amount of benefits for retirees and projecting the expected amount for future retirees (Mitchell et al., 2000, pp. 327– 359). Private pensions are the second biggest source of retirement income. Linkage to employer pension plan descriptions can be used to project future individual benefits and retirement wealth (Gustman et al., 1999, pp. 150–208). The HRS as launched in 1992 represented only persons born 1931–1941 and focused on following them through the retirement transition. The NIA had other data needs, particularly for the oldest-old population. In 1993, the Asset and Health Dynamics of the Oldest-Old (AHEAD) study, targeting persons 70 and older (born before 1924) was launched as a partner to the HRS, sharing some of the same content and scientific personnel. Its interviews were scheduled for the odd-numbered years between HRS interviews. It devoted greater attention to health and the receipt of care for health, particularly within the family.

26.3 26.3.1

Phase 2: The Steady State The Steady-State Sample Design

At the end of 1996, there were three waves of data on the original HRS cohort and two waves of data on the AHEAD cohort. The two studies had some overlap in content and scientific leadership but functioned largely independently of one another. The cohorts were ageing—by 477

John W. R. Phillips and David R. Weir

the next wave in 1998 the youngest age-eligible members of the HRS cohort would be 57, limiting its ability to address retirement preparation issues. Because of the gap between the HRS and AHEAD cohorts, there was no way to fully represent the population 65 and older on a crosssection basis. Addressing the need for additional birth cohorts through more separate projects would have been unnecessarily expensive and the inevitable dis-harmonization scientifically undesirable. The solution led by Robert Willis was the creation of the “steady-state” design inaugurated in the 1998 wave. It called for integrating the HRS and AHEAD survey instruments into a single computer-assisted questionnaire that could target questions appropriate to age and circumstances. That allowed for the integration of the fieldwork and training of a single group of interviewers on the common instrument. The gap between the HRS and AHEAD cohorts was filled by the addition of a birth cohort born 1924–1930 called the Children of the Depression Age (CODA) cohort. The ageing of the HRS cohort was addressed by the addition of a new younger birth cohort born in 1942–1947 called the War Baby (WB) cohort. Together these changes enabled the HRS to represent in cross-section the full population 51 and older in 1998. Crucial to the steady-state design was its dynamic element. By adding a new 6-year birth cohort every 6 years the study would continually refresh its coverage of the pre-retirement ages. This design allowed for longitudinal analysis as originally envisioned for the HRS and AHEAD cohorts, but also repeated cross-section analysis of the population 55 and older in any year, and eventually cross-cohort comparisons of longitudinal trajectories. Figure 26.2 depicts the evolution of the HRS sample design in simplified form. It is a Lexis diagram, showing calendar time on the horizontal axis and age on the vertical. Any individual or cohort moves northeast through the diagram as it ages over time. For convenience, Figure 26.2 combines the early AHEAD waves with HRS even though they were done in separate years. The other simplification is suppression of the detail about spouses. The original HRS and AHEAD cohorts sampled persons in their target birth years (1931–1941 for HRS and 1923 and earlier for AHEAD) and included their spouses of any age. In the steady state, that design had to be modified. Sampling is of financial units that can be two people of different ages. Each combination of spousal ages is to be sampled once and only once. The CODA cohort did not include couples with one member in either the AHEAD or HRS target birth years, and the WB cohort did not include couples with one member born before 1942. For all subsequent refresher cohorts couples are assigned to a cohort based on the birth year of the older of the two persons.

26.3.2

Ancillary Studies

The integrated all-cohort core interview that emerged in 1998 is a lengthy and demanding survey. Additions must be very limited to manage respondent burden. And yet the scientific interest in other content is high. To create new scientific opportunities without overburdening the core survey the HRS began to investigate ancillary studies. These proved very successful and have become a key part of the overall HRS design. Mail surveys were the first approach to be explored. They are inexpensive and better-suited to longer questions and longer answer categories than telephone surveys. The uncertainty was whether participants would respond and whether the request to do mail surveys would affect future response rates to the core interview. In 1999 an experiment was conducted, sending mail surveys to 2,998 randomly selected participants. The response rate of 84 percent exceeded 478

The Health and Retirement Study

HRS Longitudinal Cohort Sample Design 100 95 90 85 80 75 70 65 60 55 50 2

9 19

-4 -6 998 000 002 004 006 008 010 012 014 016 018 020 022 2 2 2 2 2 93 995 2 2 2 2 1 2 2 2 9 1 1 EGENX 1966-71 LBB 1960-65 MBB 1954-59 EBB 1948-53

WB 1942-47

HRS 1931-41

CODA 1924-30

AHEAD 0 and u′′ (y) < 0, while the period utility conditional on not surviving is zero. Individuals have identical preferences. The expected lifetime utility of a young person (who is in period 1) is represented by p1 u(y1 ) + p1 p2 u(y2 ). In contrast, the expected lifetime utility of an old person (who has survived the first period and is now in period 2) is u(y1 ) + p2 u(y2 ). Let us consider an intervention that affects the current-period mortality risk, i.e., it increases p1 for the young person and p2 for the old one. Using the definition of VSL as the marginal rate of substitution between wealth (or income in 2 u(y2 ) , our case) and survival probability, we get that the VSL of a young person is VSLy = u(y1p)+p ′ 1 u (y1 )

2) while the VSL of an old person is VSLo = p2u(y u′ (y2 ) . In both expressions, the numerator represents the gain in lifetime utility if the individual survives the period, as compared with lifetime utility if she does not survive, where such a gain depends both on remaining life expectancy and on the

569

Maddalena Ferranna et al.

quality of remaining life, proxied by income. In contrast, the denominator is the opportunity cost of spending money on risk reduction measures rather than on other goods or services. The opportunity cost increases in survival probability and decreases in income. 2 u(y2 ) 2 )) VSL decreases with age if and only if u(y1p)+p > (p(u(y . Suppose y1 = y2 . Then, the ′ ′ 1 u (y1 ) 2 u (y2 )) utility gain from surviving (the numerator) decreases with age because older people have a shorter remaining life expectancy. That is, the difference between remaining life expectancy conditional on surviving and on not surviving the current period is 1 for an old person and 1 + p2 for a young person. A decreasing life expectancy implies that VSL can decrease with age. Note that if y1 < y2 (i.e., income increases with age), the utility gain from surviving may increase with age, thereby implying a positive relationship between VSL and age. However, the opportunity cost of spending (the denominator) is likely to decrease with age because p1 > p2 (i.e., the risk of dying increases with age) and because y1 is likely to be smaller than y2 . Therefore, the negative correlation between opportunity cost and age implies that VSL can increase with age. The example shows that the effect of age on VSL is indeterminate because it is mediated by the life-cycle profile of consumption, which, in turn, depends on individuals’ earnings, access to financial markets, and preferences for reallocating income to consumption across time. In particular, if individuals have access to actuarially fair annuities (i.e., they can easily transfer income across time) and optimal consumption is constant over time (e.g., when the interest rate is equal to the rate of time preference), VSL decreases with age. In contrast, if individuals cannot borrow against future income, VSL has an inverted U-shape throughout the life cycle, tracking typical lifetime income and consumption patterns closely (Shepard and Zeckhauser, 1984; Johansson, 2002). Empirical evidence is also inconclusive about the relation between VSL and age. Although most studies find that VSL is smaller for older people, the size of the difference varies widely. For example, Alberini et al. (2004) and Alberini et al. (2006) find weak support for the notion that willingness to pay for risk reduction declines with age, Andersson (2007) shows that the willingness to pay for road safety does decline with age, and Johannesson and Johansson (1996) focus on individuals at advanced ages and find that the willingness to pay to increase survival probability increases with a person’s age. Using a revealed preference approach instead of a stated preference one, Aldy and Viscusi (2008) find that the VSL-age relation has an inverted U-shape. The difference in results depends, among other things, on the adopted methodology and the characteristics of the selected sample (Krupnick, 2007).

32.3

The Social Welfare Function Approach

The SWF approach presupposes that the value of an intervention depends on its impact on social welfare. Such a method first determines the health and non-health impacts of the intervention on individuals’ well-being and then aggregates those well-being gains and losses to yield an overall measure of how beneficial the intervention is. The aggregation is performed through an SWF that assigns a measure of social value to each distribution of individual well-being (Adler, 2019). The intervention shifts the distribution of individual well-being across the population. The value of the intervention is established by comparing the social welfare associated with the original distribution of well-being and the social welfare associated with the distribution of well-being resulting from the policy. In BCA, the use of individual-specific VSL implies that benefits to the well-off are larger than similar benefits accruing to less well-off individuals because money has relatively lower marginal value for the wealthy. This issue prompts evaluators to adopt a population-average 570

Age and the Value of Life

VSL (Robinson, 2007). This bias toward the interests of the well-off is not present in the SWF approach because the unit of analysis is well-being and not money. This constitutes the main advantage of SWF analysis compared with BCA. Note that the insensitivity of the SWF approach to differences in ability to pay does not imply that mortality gains to the wealthy are always valued equally as similar mortality gains to the less wealthy, especially if income is taken as a proxy for quality of life (Adler et al., 2014). The SWF approach (or at least some SWFs) might count mortality gains as larger for wealthy rather than for less wealthy individuals not because the former have more financial resources and thus express a higher willingness to pay, but because the well-off are expected to experience a better life if saved. Unlike BCA, the SWF methodology requires interpersonally comparable well-being numbers. Individual well-being is assumed to depend on bundles of attributes that matter to individuals, such as income, health, and leisure. Specifying a well-being measure and its dependence on the bundle of attributes can be done in several ways (Adler and Decancq, 2022). One possibility is to measure individual well-being using von Neumann-Morgenstern utility functions that represent individual preferences regarding alternative probability distributions of attributes over the lifetime (Adler, 2019). This is the approach this chapter takes because of its comparability to the BCA methodology. Alternative approaches to measure well-being include happiness and subjective satisfaction (Clark et al., 2018); equivalent income, i.e., income adjusted by the value of nonmarket attributes (e.g., health and longevity) based on population preferences for those attributes (Fleurbaey et al., 2013; Samson et al., 2018); and measures based on capabilities and individuals’ opportunities (Sen, 1999). These measures can lead to different well-being rankings of individuals (Decancq and Neumann, 2016), highlighting the importance of the choice of well-being measure for policy evaluation. Preference-based approaches (e.g., equivalent income and von NeumannMorgenstern utility functions) dominate the literature because they account not only for the position of individuals along different attributes but also for individuals’ preferences for those attributes. A complementary issue is whether the metric of interest should be lifetime well-being or sub-lifetime well-being (Adler, 2012, chapter 6). The predominant view, and the one taken here, is that the situation of the individual over the whole life matters for policy evaluation. This includes not only the current and future positions of the individual along the different wellbeing-relevant attributes (e.g., income and health), but also how the individual fared in the past. Adopting a lifetime perspective is compatible with the shared view that the evaluation of public policies should be sensitive to pre-existing inequities (e.g., in income or in access to healthcare). The most used SWF is the utilitarian one, according to which the social value assigned to a distribution of individual well-being equals the total sum of those well-being levels. Considering again the previous example, and assuming Ny young people and No old persons, the utilitarian SWF equals W U = Ny [p1 u(y1 ) + p1 p2 u(y2 )] + No [u(y1 ) + p2 u(y2 )]. A limitation of utilitarianism is its insensitivity to inequalities in well-being across the population. In other words, utilitarianism is indifferent to whether a well-being gain accrues to a well-off person or to a less well-off one. For example, the utilitarian SWF would be indifferent to structural inequities and to the potential impact of the intervention in addressing those inequities. This issue can be solved by adopting a prioritarian SWF, i.e., an SWF that is sensitive to the distribution of well-being and that gives extra weight to the well-being of the worse-off (Adler, 2012). The main consequence of using prioritarian SWFs is that policies producing well-being gains for those with low well-being are ranked higher than policies producing the same well-being gains for those who are better off. The prioritarian SWF ranks distributions of well-being according to the sum of a strictly increasing and strictly concave transformation of individual well-being. 571

Maddalena Ferranna et al.

Let g(·) denote some strictly increasing and strictly P concave function and wi the well-being of individual i. The prioritarian SWF takes the form i g(wi ); when the function g is linear, we get the utilitarian SWF. By taking a strictly increasing and strictly concave transformation of wellbeing levels, the prioritarian SWF gives greater weight to well-being changes affecting worse-off individuals. Note that an SWF that gives absolute priority to the worse-off (e.g., a Rawlsian SWF that looks only at the well-being of the worst-off) can be considered an extreme case of prioritarianism. When considering mortality risks, individuals’ well-being is uncertain (individuals may survive or not), and the SWF requires an uncertainty module, i.e., a procedure for applying the SWF to probability distributions of well-being (Adler, 2012, 2019). The two dominant approaches to applying the prioritarian rule under risk are ex-ante prioritarianism and ex-post prioritarianism. Ex-ante prioritarianism assumes that the metric of interest is expected individual well-being, while ex-post prioritarianism concerns realized individual wellbeing.6 Thus, in the ex-ante case the worst-off person is the one with the worst fate in expectation, while in the ex-post case the worst-off person is the one who has lived the worst life. In the example considered here, the ex-ante prioritarian SWF is given by W EAP = Ny g(p1 u(y1 ) + p1 p2 u(y2 )) + No g(u(y1 ) + p2 u(y2 )), while the ex-post prioritarian SWF is given by W EPP = (Ny p1 + No )[(1 − p2 )g(u(y1 )) + p2 g(u(y1 ) + u(y2 ))].7 Note that young people are among the worse-off from both the ex-ante and ex-post perspectives. Expected lifetime well-being increases with age as young individuals are uncertain whether they will survive to old age, while older individuals have at least lived through the young age, i.e., p1 u(y1 ) + p1 p2 u(y2 ) < u(y1 ) + p2 u(y2 ). Realized well-being also increases with age because those who die young are undoubtedly among the worst-off in the population. This implies that both forms of prioritarianism can accommodate the fair innings argument.

32.4

Age and the Social Value of Mortality Risk Reduction

Adler et al. (2014) introduce the concept of “social value of risk reduction” (SVRR), defined as the rate of increase in social welfare resulting from a small increase in the survival probability of an individual. Mathematically, the SVRR is the partial derivative of social welfare with respect to an individual’s survival probability. SVRR is comparable to VSL. In the standard BCA calculus, the benefits from a mortality risk-reducing intervention are (approximately) equal to the sum of individual-specific VSLs multiplied by the magnitude of the risk reduction for each individual. Thus, VSL can be interpreted as the rate of increase in benefits resulting from a small increase in the survival probability of an individual. SVRR represents how socially valuable it is to (marginally) reduce the mortality risk of an individual. For example, if there are two patients and only one drug (or other medical resource), and the treatment increases the chances of survival of the two individuals by the same amount, the drug should be given to the person with the largest SVRR. In addition, as in the standard BCA calculus, the social benefit of a health intervention is approximately equal to the sum of individual-specific SVRRs multiplied by the magnitude of the risk reduction for each individual. Thus, other things equal, a budget-constrained government should finance primarily interventions that confer larger risk reductions to individuals with larger SVRRs. Adler et al. (2014) study the sensitivity of utilitarian and prioritarian SVRRs to wealth and baseline mortality risk in a single-period model, and Adler et al. (2021) extend the analysis to a multiperiod model under the assumption of myopic consumption (i.e., no borrowing or saving).8 Such a multiperiod model allows study of the impact of age on SVRR. Both papers show that the choice of value framework is consequential for evaluating risk reduction policies and that 572

Age and the Value of Life

the sensitivity of SVRR to individual characteristics differs across the three benchmark SWFs (i.e., utilitarian, ex-ante prioritarian, and ex-post prioritarian) and BCA (with VSL interpretable as a type of SVRR in the BCA case). Here, we focus on the sensitivity of SVRR to age, and we refer to the aforementioned papers for the sensitivity of SVRR to baseline risk and wealth. Let us first consider the utilitarian SWF. In such a case, the SVRR for the young person is u(y1 ) + p2 u(y2 ), while the SVRR for the old person is u(y2 ). Thus, the utilitarian SWF values risk reduction policies based on the size of the expected utility gains from the intervention. The utilitarian SVRR decreases with age if and only if u(y1 ) + (p2 − 1)u(y2 ) > 0. Two effects emerge. On the one hand, the young person has a larger increase in remaining life expectancy if she survives the current period. As pointed out earlier, the gain for the young person is 1 + p2 , while the gain for the old person is 1. On the other hand, the utility gains from the intervention depend also on the quality of life, proxied here by income. If a higher quality of life is attached to older ages (i.e., y2 > y1 ), the utilitarian SVRR may increase with age because the young individual may not reach old age at all (i.e., p2 < 1), while the older person already has (provided she survives the period). For example, if both p2 and y1 are very small while y2 is very large, the young person faces a long life of misery while the old person faces a short, pleasant life. In this case, from a utilitarian perspective saving the life of the old person may be more valuable. The utilitarian SWF is compatible with the approach commonly used in the health sector (e.g., in the evaluation of vaccines), according to which the value of a health intervention depends on the number of avoided deaths and on the (quality-adjusted) life expectancy of the affected individuals, where each additional (quality-adjusted) life year has equal social value across individuals. Thus, u(y1 ) = u(y2 ), and the utilitarian SVRR depends only on remaining life expectancy, thereby implying that the social value of reducing mortality risk is lower for older people than for younger ones. Like BCA, the utilitarian SWF does not capture the idea of fair innings, i.e., the view that younger individuals should get priority with respect to life-saving measures on fairness grounds because they have not yet had a chance to live a full life. Indeed, the utilitarian SVRR is sensitive to the size of the utility gains from the intervention, but not to the distribution of those gains across the population, e.g., whether a given utility gain occurs to a young person or to an old one. If a young and an old person benefit equally in utility terms from a risk-reducing intervention, their utilitarian SVRRs would coincide. For example, the utilitarian SWF would be indifferent between allocating a scarce medical resource to the young or to the old person. In contrast, the fair innings argument requires that, in such a situation, the scarce resource be allocated to the young because she has lived fewer years. Like the utilitarian SVRR, both the ex-ante and the ex-post prioritarian SVRRs may increase or decrease with age. However, compared with the utilitarian case, the prioritarian SWFs attach higher value to reducing the mortality risk of the young than to similar reductions benefiting the older person. In particular, Adler et al. (2021) show that the ratio of the SVRR for the young individual to the SVRR for the old individual is always larger under prioritarianism than under utilitarianism. Consequently, if the young and the old experience the same utility gains from the intervention (the ratio of utilitarian SVRRs equals 1), the prioritarian SVRRs are larger for the young than for the old (the ratio of prioritarian SVRRs is larger than 1), i.e., the prioritarian SWFs recommend giving the scarce medical resource to the young. Consequently, the prioritarian SWFs can capture the fair innings idea. In the two-period example, the utilitarian ratio of the SVRR for the young to the SVRR 2 u(y2 ) 1) = u(y for the old is u(y1 )+p u(y2 ) u(y2 ) + p2 . In the ex-ante prioritarian case, the SVRR for the young person equals g′ (p1 u(y1 ) + p1 p2 u(y2 ))[u(y1 ) + p2 u(y2 )], while the SVRR for the old person equals g′ (u(y1 ) + p2 u(y2 ))[u(y2 )] . For each age group, the ex-ante prioritarian SVRR 573

Maddalena Ferranna et al.

equals the utilitarian SVRR (the second term) multiplied by a factor that depends on individuals’ expected lifetime utility, conditional on their age. Because the priority function g is concave, the priority weight is larger for individuals with smaller expected lifetime utility, i.e., it is larger for the young than for   the old. Indeed, the ex-ante prioritarian SVRR ratio equals g′ (p1 u(y1 )+p1 p2 u(y2 )) u(y1 ) u(y1 ) + p 2 , which is larger than the utilitarian SVRR ratio u(y2 ) + p2 u(y2 ) g′ (u(y1 )+p2 u(y2 )) because of the concavity of the priority function g. Hence, the ex-ante prioritarian SVRR ratio is always larger than the utilitarian one, implying a fair innings effect. In the ex-post prioritarian case, the SVRR for the young person is g(u(y1 )) + p2 [g(u(y1 ) + u(y2 )) − g(u(y1 ))], while the SVRR for the old person is g(u(y1 ) + u(y2 )) − g(u(y1 )). The expost prioritarian SVRR is like the utilitarian SVRR, except it depends on transformed realized lifetime utilities. The ex-post prioritarian ratio of the SVRR for the young to the SVRR for 1 )+0)−g(0) + p2 , where g(0) = 0. This ratio is larger than the utilitarian the old equals g(u(yg(u(y 1 )+u(y2 ))−g(u(y1 )) one because the concavity of the priority function g implies that the first term is larger than u(y1 ) u(y2 ) . Once again, compared with the utilitarian case, the ex-post prioritarian SWF attaches higher value to interventions that benefit primarily the young.

32.5

Discussion and Conclusion

This chapter reviews the literature on the impact of age on the value of reducing mortality risks, focusing on the comparison between two alternative evaluation approaches: benefit-cost analysis and SWF analysis. We argue that the latter is a more attractive methodology to evaluate health improvements because it is compatible with the idea that each unit of health may have different value across individuals, and, unlike BCA, it need not have the ethically objectionable result that benefits accruing to the well-off are more valuable than similar benefits accruing to the less well-off. This chapter focuses on two SWF frameworks: utilitarianism and prioritarianism. Utilitarianism is sensitive to the size of the well-being gains induced by the health intervention. Under simplifying assumptions, this implies that young individuals are attached a larger weight in policy evaluation because of their longer remaining life expectancy.9 Prioritarianism is sensitive to both the size of the well-being gains and their distribution across the population. Consequently, young individuals are prioritized in the evaluation of life-saving measures not only because they are expected to live longer but also because being young makes them among the worst-off in terms of lifetime utility (a fair innings argument). The SWF methodology is more data-intensive than standard methods such as BCA. It requires information about the heterogeneous distribution of impacts and burdens and the correlation of those impacts with background characteristics. If we want to adopt a distributionsensitive SWF (e.g., a prioritarian one), we also need information about the past situation of the individual. A topic for future research is how to collect and process the information required to construct measures of lifetime well-being and their correlation with policy impacts. In this chapter, we point out the significant role of the chosen well-being measure for the ranking of individuals’ well-being and thus the social value of a given intervention. In applications, assuming that individuals have identical preferences is standard. But preferences about what matters in life can change with age (e.g., Dohmen et al., 2017). Failing to account for that can bias our measure of well-being at advanced ages, in particular whether people are ageing successfully or not. More research is needed to properly measure preferences of older people and how to incorporate such preferences into the evaluation of health interventions. Similarly, if individuals have different preferences for well-being attributes, such as health and consumption levels, SWF analysis requires these to be measured. 574

Age and the Value of Life

Most of the literature reviewed in this chapter focuses on the value of mortality risk reductions, while neglecting the value of reducing nonfatal risks. Even though reducing mortality risks accounts for the largest share of benefits of health-related interventions, another topic for future research is how to extend the SWF approach to the evaluation of reducing morbidity risks and how that value changes with age (see Hammitt and Treich, 2022, for a discussion of the social value of morbidity risks). That matters for health conditions that entail significant pain and suffering, e.g., some types of cancer. Two other little-explored topics are worth mentioning. First, how to account for issues of responsibility and deservingness in the evaluation of health interventions. For example, should the evaluation give less priority to individuals who are in bad health because of choices they made in the past? This is an issue, for instance, in liver transplantation for alcohol-related liver disease and how past behavior and the risk of returning to drinking should affect the allocation choice. Apart from the ethical stance (e.g., whether personal responsibility should be accounted for at all), the issue here is how to extend the SWF framework to differentiate between responsibility and luck (e.g., Ferreira and Peragine, 2016; Brunori et al., 2022). Second, some interventions (e.g., sexual and reproductive health interventions) affect not only the health and longevity of current people but also the well-being of people not yet born and whose existence may depend on the intervention itself. While population ethics has an extensive body of work on how to compare the well-being of existing and possible people, its application to the evaluation of policies has been limited (Broome, 2004; Blackhorby et al., 2005). This would be a challenging and interesting topic for future research.

Notes 1 This is a simplified account of the main trade-offs highlighted by the COVID-19 pandemic because it neglects other high-risk categories (e.g., individuals with comorbidities or individuals living in crowded settings) and the negative impacts of nonpharmaceutical interventions on older adults (e.g., worsening of mental health due to isolation). 2 Note that the negative correlation between age and remaining life expectancy holds as long as survival probabilities do not increase with age (Adler et al., 2021). 3 As the standard view in economics, this chapter assumes a welfarist approach, according to which the value of an intervention depends on its impacts on the total sum and distribution of individuals’ lifetime well-being. In contrast, a non-welfarist approach might support the idea that everyone has an equal right to life and that the evaluation of health interventions should not attach different values to benefits accruing to different individuals, e.g., based on their age. 4 However, note that applying country-specific VSL measures depending on a country’s stage of development is also common on the grounds that the value of money changes across countries (Robinson et al., 2019). 5 In this chapter, we restrict our attention to the differences between benefit-cost analysis and social welfare function (SWF) analysis in evaluating mortality risk reductions. In particular, we do not discuss the relationship between age and the value of life within a cost-effectiveness approach. Cost-effectiveness analysis is the most common method for economic evaluation in the health sector. The value of an intervention is based on its incremental costs per additional QALY generated by the policy. Consequently, each QALY is assumed to have the same value. See Cookson et al. (2020) for a review of cost-effectiveness analysis applied to health interventions and for a discussion of how to incorporate distributional concerns in the analysis. 6 Note that the choice between these uncertainty modules does not matter for the utilitarian SWF; the ex-ante utilitarian and ex-post utilitarian SWFs are identical. 7 From an ex-post perspective, individuals either die young or they die old. In the example, (Ny p1 + No )(1 − p2 ) individuals die young and (Ny p1 + No )p2 individuals survive until the end of old age. Note that Ny (1 − p1 ) individuals do not survive to the end of the first period and are assigned a zero-well-being level, with g(0) = 0 by assumption.

575

Maddalena Ferranna et al.

8 The “myopic consumption” assumption can be interpreted either as the absence of perfect financial markets or as the fact that individuals do not access financial markets. Adler et al. (2019) show that the main results of the model hold also when there are actuarially fair annuities (i.e., the individual can save and borrow) as long as the risk reduction intervention comes as a surprise, i.e., the individual does not alter her consumption plan. 9 The simplifying assumptions include nonincreasing consumption with age (i.e., quality of life, as proxied by income/consumption, does not increase with age) and nonincreasing survival probability with age (i.e., only if survival probabilities do not increase with age is it assured that life expectancy remaining decreases with age).

References ADLER, M. D. (2012): Well-Being and Fair Distribution: Beyond Cost-Benefit Analysis. New York: Oxford University Press. ADLER, M. D. (2019): Measuring Social Welfare: An Introduction. Oxford, New York: Oxford University Press. ADLER, M. D., AND DECANCQ, K. (2022): “Well-being measurement.” In: Adler, M. D., and Norheim, O. (eds.), Prioritarianism in Practice, Cambridge: Cambridge University Press, chapter 3, pp. 128–171. ADLER, M. D., FERRANNA, M., HAMMITT, J. K., AND TREICH, N. (2019): “Fair innings? The utilitarian and prioritarian value of risk reduction over a whole lifetime,” Duke Law School Public Law and Legal Theory, Working Paper n. 2019-79. ADLER, M. D., FERRANNA, M., HAMMITT, J. K., AND TREICH, N. (2021): “Fair innings? The utilitarian and prioritarian value of risk reduction over a whole lifetime,” Journal of Health Economics, 75: 102412. ADLER, M. D., HAMMITT, J. K., AND TREICH, N. (2014): “The social value of mortality risk reduction: VSL versus the social welfare function approach,” Journal of Health Economics, 35: 82–93. ALBERINI, A., CROPPER, M., KRUPNICK, A., AND SIMON, N. B. (2004): “Does the value of a statistical life vary with the age and health status?: Evidence from the USA and Canada,” Journal of Environmental Economics and Management, 48(1): 769–792. ALBERINI, A., HUNT, A., AND MARKANDYA, A. (2006): “Willingness to pay to reduce mortality risks: Evidence from a three-country contingent valuation study,” Environmental and Resource Economics, 33(2): 251–264. ALDY, J. E., AND VISCUSI, W. K. (2008): “Adjusting the value of a statistical life for age and cohort effects,” Review of Economics and Statistics, 90(3): 573–581. ANDERSSON, H. (2007): “Willingness to pay for road safety and estimates of the risk of death: Evidence from a Swedish contingent valuation study,” Accident Analysis and Prevention, 39(4): 853–865. BLACKHORBY, C., BOSSERT, W., AND DONALDSON, D. (2005): Population Issues in Social Choice Theory, Welfare Economics and Ethics, Cambridge: Cambridge University Press. BOGNAR, G. (2008): “Age-weighting,” Economics and Philosophy, 24(2): 167–189. BOGNAR, G. (2015): “Fair innings,” Bioethics, 29(4): 251–261. BROOME, J. (2004): Weighing Lives, Oxford, New York: Oxford University Press. BRUNORI, P., FERREIRA, F. H. G., AND PERAGINE, V. (2022): “Prioritarianism and equality of opportunity.” In: Adler, M. D., and Norheim, O. (eds.), Prioritarianism in Practice, Cambridge: Cambridge University Press, chapter 11, pp. 518–571. CLARK, A., FL E` CHE, S., LAYARD, R., POWDTHAVEE, N., AND WARD, G. (2018): The Origins of Happiness: The Science of Wellbeing over the Life-Course, Princeton, NJ: Princeton University Press. COOKSON, R., GRIFFIN, S., NORHEIM, O. F., AND CULYER, A. C. (EDS.). (2020): Distributional Cost-Effectiveness Analysis. Quantifying Health Equity Impacts and Trade-Offs, Oxford, New York: Oxford University Press DANIELS, N. (1988): Am I My Parents’ Keeper? An Essay on Justice between the Young and the Old, New York: Oxford University Press. DECANCQ, K., AND NEUMANN, D. (2016): “Does the choice of well-being measure matter empirically?” In: Adler, M. D., and Fleurbaey, M. (eds.), The Oxford Handbook of Well-Being and Public Policy, Oxford: Oxford University Press, chapter 9. DOHMEN, T., FALK, A., GOLSTEYN, B. H. H., HUFFMAN, D., AND SUNDE, U. (2017): “Risk attitudes across the life course,” The Economic Journal, 127(605): F95–F116.

576

Age and the Value of Life

DOLAN, P., AND TSUCHIYA, A. (2012): “It is the lifetime that matters: Public preferences over maximising health and reducing inequalities in health,” Journal of Medical Ethics, 38(9): 571–573. EMANUEL, E. J., PERSAD, G., UPSHUR, R., THOME, B., PARKER, M., GLICKMAN, A., ZHANG, C., BOYLE, C., SMITH, M., AND PHILLIPS, J. P. (2020): “Fair allocation of scarce medical resources in the time of Covid-19,” New England Journal of Medicine, 382: 2049–2055. FERREIRA, F. H. G., AND PERAGINE, V. (2016): “Individual responsibility and equality of opportunity.” In: Adler, M. D., and Fleurbaey, M. (eds.), The Oxford Handbook of Well-Being and Public Policy, Oxford, New York: Oxford University Press, chapter 25, pp. 746–784. FLEURBAEY, M., LUCHINI, S., MULLER, C., AND SCHOKKAERT, E. (2013): “Equivalent income and fair evaluation of healthcare,” Health Economics, 22(6): 711–729. ¨ FUCHS-SCH UNDELN , N., KRUEGER, D., LUDWIG, A., AND POPOVA, I. (2020): “The long-term distributional and welfare effects of COVID-19 school closures,” National Bureau of Economic Research, Working Paper No. 27773. HAMMITT, J. K. (2007): “Valuing changes in mortality risk: Lives saved vs. life years saved,” Review of Environmental Economics and Policy, 1(2): 228–240. HAMMITT, J. K., AND TREICH, N. (2022): “Prioritarianism and fatality risk regulation.” In: Adler, M. D., and Norheim, O. (eds.), Prioritarianism in Practice, Cambridge: Cambridge University Press, chapter 7, pp. 317–359. HARRIS, J. (1985): The Value of Life, London: Routledge and Kegan Paul. JOHANNESSON, M., AND JOHANSSON, P.-O. (1996): “To be, or not to be, that is the question: An empirical study of the WTP for an increased life expectancy at an advanced age,” Journal of Risk and Uncertainty, 13(2): 163–174. JOHANSSON, P.-O. (2002): “On the definition and age-dependency of the value of a statistical life,” Journal of Risk and Uncertainty, 25(3): 251–263. KRUPNICK, A. (2007): “Mortality risk valuation and age: Stated preference evidence,” Review of Environmental Economics and Policy, 1(2): 261–282. LOCKWOOD, M. (1988): “Quality of life and resource allocation.” In: Bell, J. M., and Mendus, S. (eds.), Philosophy and Medical Welfare, Cambridge: Cambridge University Press, pp. 33–55. MONGEY, S., PILOSSOPH, L., AND WEINBER, A. (2020): “Which workers bear the burden of social distancing?,” NBER Working Paper No. 27085. National Bureau of Economic Research, Cambridge, MA. NORD, E. (2005): “Concerns for the worse off: Fair innings versus severity,” Social Science and Medicine, 60(2): 257–263. PRATT, J. W., AND ZECKHAUSER, R. (1996): “Willingness to pay and the distribution of risk and wealth,” Journal of Political Economy, 104(4): 747–763. ROBINSON, L. A. (2007): “How US government agencies value mortality risk reductions,” Review of Environmental Economics and Policy, 1(2): 283–299. ROBINSON, L. A., HAMMITT, J. K., AND O’KEEFFE, L. (2019): “Valuing mortality risk reductions in global benefit-cost analysis,” Journal of Benefit-Cost Analysis, 10(S1): 15–50. ROSENBAUM, L. (2020): “Facing COVID-19 in Italy—Ethics, logistics, and therapeutics on the epidemic’s front line,” The New England Journal of Medicine, 382: 1873–1875. SAMSON, A. L., SCHOKKAERT, E., TH E´ BAUT, C., DORMONT, B., FLEURBAEY M., LUCHINI, S., AND VAN DE VOORDE, C. (2018): “Fairness in cost-benefit analysis: A methodology for health technology assessment,” Health Economics, 27(1): 102–114. SEN, A. K. (1999): Commodities and Capabilities, Oxford, New York: Oxford University Press. SHEPARD, D. S., AND ZECKHAUSER, R. J. (1984): “Survival versus consumption,” Management Science, 30(4): 423–439. VISCUSI, W. K. (2009): “The devaluation of life,” Regulation and Governance, 3(2): 103–127. WILLIAMS, A. (1997): “Intergenerational equity: An exploration of the ‘fair innings’ argument,” Health Economics, 6(2): 117–132.

577

33 HAPPINESS AND AGEING IN THE UNITED STATES1 David G. Blanchflower and Carol Graham

Abstract The years since the Great Recession have brought increasing concern, in countries all over the world, of declines in mental health and well-being. Globally, chronic depression and suicide rates peak in midlife. In the United States, deaths of despair are most likely to occur in these years, and the patterns are robustly associated with unhappiness and stress. A less-known relationship also exists between well-being and longevity among the elderly, particularly for those over age 70. In this chapter, we analyze several data sets for the United States, provide extensive evidence on middle-age patterns and how they differ between the married and unmarried, and review new work on the elderly. The relationship between well-being and ageing has a robust association with trends that can ruin lives and shorten life spans. It applies to much of the world’s population and links to behaviors and outcomes that merit the attention of scholars and policymakers alike.

33.1

Introduction

The post Great Recession years from 2010 onward have brought increasing concern, in countries all over the world, of declines in mental health and well-being. Globally, chronic depression and suicide rates peak in midlife. In the United States, deaths of despair are most likely to occur in middle age, from 35 to 55, and the patterns are robustly associated with unhappiness and stress. Well-being is also a factor in differential mortality rates among the elderly, particularly those over age 70. Better understanding the relationship between well-being and ageing is not just an academic exercise. It has a robust association with trends that can ruin lives and shorten life spans. In this chapter, we analyze several data sets for the United States, provide extensive evidence on that middle-age patterns and how they differ between the married and unmarried, and review new work on the elderly. Evidence for the United States and elsewhere of midlife lows is widespread in well-being data, using happiness and life satisfaction data. Evidence also exists of a humpshaped relationship in age in unhappiness data such that depression, stress, anxiety, despair, and unhappiness peaks in midlife. This is consistent with data on deaths of despair due to suicide and to drug and alcohol poisoning, which disproportionately occur among the prime age. In the United Kingdom, the proportion of individuals reporting that they were depressed rose sixfold from 1997 to 2018 (Bell and Blanchflower, 2019). In the United States, the rise in pain and premature mortality especially among the prime-age less-educated—the so-called 578

DOI: 10.4324/9781003150398-39

Happiness and Ageing in the United States

deaths of despair—is notable (Blanchflower and Oswald, 2020; Case and Deaton, 2015, 2020). For example, Nahin (2015) reported that a remarkable 126 million or 56 percent of American adults experienced some type of pain in 2012. Of these, 20 percent experienced pain daily (i.e., chronic pain). In Gallup’s U.S. Daily Tracker Poll, the proportion of people saying they experienced physical pain yesterday rose from 23.5 percent in 2008 to 27.2 percent in 2017. Meanwhile, almost 1 million Americans died due to suicide, opioid and other drug overdoses, and alcohol-related diseases from 2005 to 2018, a trend that is driving up the overall mortality rate in the United States and contributed to reduced life expectancy between 2014 and 2017 (Kochanek et al., 2020). In 2019, 70,630 drug overdose deaths occurred in the United States for an age-adjusted rate of 21.6 per 100,000 standard population, up from 20.7 in 2018 (https://www.cdc.gov/drugoverdose/deaths/index.html). The COVID-19 global pandemic has further heightened concerns about well-being. According to the most recent Census Bureau’s Household Pulse Survey #33 for June 23–July 5, 2021,2 80 million Americans age 16+ reported feeling down, depressed, or hopeless over the prior seven-day period: 47 million “on several days,” 15 million on “more than half the days,” and 18 million said “nearly every day.”3 According to weighted data from the 2020 Gallup Covid Tracker, mean life satisfaction measured with the Cantril ladder variable was 6.62 (n =109,596). This compares with 7.06 from the (much larger) 2017 Daily Tracker. Well-being collapsed in the United Kingdom in March 2020 as the country went into lockdown. The Labour Force Surveys conducted by the UK’s Office of National Statistics (ONS) showed life satisfaction averaging 7.7 (on a 0–10 scale) from April 2017 to March 2020. Figure 33.1 plots life satisfaction approximately weekly between March 2020 and June 2021

Figure 33.1 Life satisfaction in the United Kingdom. Source: www.covidsocialstudy.org.

579

David G. Blanchflower and Carol Graham

based on a University College London survey that asks the same question: “Overall, how satisfied are you with your life nowadays, where nought is ‘not at all satisfied’ and 10 is ‘completely satisfied.’ ” We include the score of 7.7 in the first week of March 2020 in the chart and show the precipitous drop that followed. The average life satisfaction score dropped to 5.4 in the last week of March 2020 as lockdown started (https://www.covidsocialstudy.org/results). While some adaptation occurred in the following months through September 2020 as the score rose to 6.5, the average life satisfaction score fell again to a low of 5.6 at the end of January 2021 as new restrictions were implemented. The life satisfaction score slowly rose from there and stood at 6.8 in June 2021, well below its pre-pandemic levels.4 We have also recently seen a big fall in happiness as measured by the General Social Survey in the United States in 2021—the percent saying they were very happy, for example fell from 31 percent in 2014, 28 percent in 2016, and 30 percent in 2018 to 20 percent in 2021. The big declines in well-being during the COVID lockdown stand in stark contrast to the Great Recession, which saw major declines in output and increases in unemployment around the world. With the major exception of Greece and Spain, the Great Recession was not happiness reducing (Bell and Blanchflower, 2015, Table 10). Given that these drops in well-being all occurred in during the period that COVID-19 incidence and death required significant government response in many countries, the large drops in well-being are clearly associated with the pandemic and are unlikely to have been driven by unrelated political trends in individual countries. Independent of the declines in well-being in recent years and of those related specifically to the pandemic, several earlier economic studies found evidence of a significant and empirically large downturn in human well-being during the midlife years, or so-called “happiness curves” (Rauch, 2018).5 Early work was based on life satisfaction and happiness data; the research now extends to trends in unhappiness, stress, lack of sleep, depression, and even suicide (Daly et al., 2011) and across multiple data sets and 145 countries (Blanchflower, 2020a). Evidence also indicates that unhappiness reaches a peak in midlife (Graham and Ruiz-Pozuelo, 2017; Blanchflower, 2020b). Thus, when significant general declines in well-being occur, those already in low well-being, middle-aged years are likely to be particularly vulnerable. Evidence from longitudinal surveys that focus on changes in life satisfaction controlling for personal fixed effects suggests that life satisfaction in individuals forms a U-shape (Cheng et al., 2017). Controlling for cohort effects has little or no impact on the U-shape (Clark, 2019; Blanchflower, 2020b). A hill shape in antidepressant use reaches its apex in the mid-40s in European countries (Blanchflower and Oswald, 2016; Blanchflower and Bryson, 2021b). The midlife U-shape pattern even applies to apes (Weiss et al., 2012). The recent increases in U.S. deaths of despair have occurred precisely in the middle ages of 35–64 years (Case and Deaton, 2015, 2020). The patterns in these deaths have a robust association with the same ill-being markers—unhappiness and stress—that increase in midlife (Graham and Pinto, 2019). Two new statistical releases in October 2020 showed drug poisoning deaths peaking in midlife. In England and Wales, the two age groups with the highest rates are 30–39 and 40–49 (Figure 33.2a). The older of the two groups showed the biggest rise in deaths from drug use over these years, taking over from the 30–39 age group in 2016. In the United States, the Centers for Disease Control (CDC) published new data in October 2020 on poisoning deaths from cocaine. These rates were stable between 2009 and 2013 and then nearly tripled from 2013 to 2018 (Hedegaard et al., 2020). In 2018, the rate of drug overdose deaths involving cocaine was highest for adults aged 35–44 (8.6 per 100,000). Figure 580

Happiness and Ageing in the United States

Figure 33.2 Mortality from drug overdose. Notes: Figure (a) illustrates age-specific mortality rates for deaths related to drug misuse, in England and Wales, registered between 1993 and 2019. Figure (b) shows rates of drug overdose deaths by cocaine, United States, 2009–2018. Source: (a) Office for National Statistics, deaths related to drug poisoning in England and Wales, 2019 registrations, October 2020. (b) Hedegaard et al. (2020).

33.2b shows that the age group 35–44 overtook the previously highest age range of 45–54 in 2018. Notably the death rate for Black Americans from cocaine is more than double that of White Americans (9.0/100,000 vs. 4.6 in 2018), while the death rate from opioid overdose and suicides is much higher for White Americans (Graham and Pinto, 2019). The latest data for 2019 and 2020 reported by Hedegaard et al. (2021) (Table 33.1) on drug overdose death rates per for the United States shows they peak in midlife. The increase between 2019 and 2020 was especially marked for those ages 35-44. 581

David G. Blanchflower and Carol Graham Table 33.1 Drug overdose deaths by age Year

Ages 15–24

Ages 25–34

Ages 35–44

Ages 45–54

Ages 55–64

Ages 65+

2019 2020

11.2 16.7

35.6 47.3

40.5 53.9

36.9 46.9

30.4 37.3

8.3 9.4

Source: Hedegaard et al (2021).

Curtin et al. (2021) show that for both men and women suicide rates in the United States peak in midlife. Provisional 2020 suicide rates per 100,000 population for males were as follows: 15 − 24 = 22.3 (5.7); 25 − 34 = 29.3 (6.9); 35 − 44 = 27.2 (7.4); 45 − 54 = 27.6 (8.4); 55 − 64 = 26.9 (7.2); 65 − 74 = 24.6 (5.6) with female rates in parentheses. Blanchflower and Oswald (2020) find a rise in despair, distress, and misery in the United States, using questions in the Behavioral Risk Factor Surveillance System (BRFSS) where despair is measured based on whether respondents reported that all of the previous 30 days were bad mental health days. Blanchflower and Feir (2021) find that poor mental health is especially prevalent among Native Americans. Despair peaks in midlife, especially for the least educated. Case et al. (2020) show that pain peaks in midlife and the most for the least educated, and Blanchflower and Bryson (2021a) show that sleep duration has a U-shape in age. Chronic depression and suicide occur disproportionately at midlife in Europe also (Blanchflower, 2020b). An analysis by the Organisation for Economic Co-operation and Development (OECD) in How’s Life? 2020 shows that deaths from suicide, alcohol abuse, or drug overdoses are higher in 10 OECD countries—Slovenia, Lithuania, Latvia, the Republic of Korea, Denmark, Belgium, Hungary, Austria, Finland, and Poland—than they are in the United States.6 Next we look at evidence on the U-shape in well-being in the literature and then present evidence from several U.S. data files. We identify differences in the raw data from the married and the unmarried. We then examine the well-being of the elderly, which is impacted by the fact that those with low levels of well-being at around age 65 have higher mortality rates. In the final section, we draw some conclusions.

33.2

The Existence of a U-Shape in Happiness and Life Satisfaction

Despite the large body of economics research finding a dip in midlife well-being, some prominent papers dismiss the midlife downturn as an illusion. As the following examples show, most of these critiques, many from psychologists, suffer from sample sizes that are too small to be representative, from flaws in econometric analysis, and from mixing and matching studies with and without controls for confounding factors, such as health, employment, marriage status, and others. Results based on patterns in the raw data reflect the effects of ageing plus all confounding factors as individuals age. Results based on regressions that include controls for these factors reflect the pure effect of ageing on well-being. Neither approach is “correct,” rather they are asking different scientific questions. What is incorrect is failing to distinguish the approach used to generate the findings. Blanchflower and Graham (2021a,b) address these questions in more detail. Some examples from the literature illustrate the variety in findings. An earlier review by Ulloa et al. (2013) drew the conclusion that “extant studies . . . show either a U-shaped, inverted Ushaped or linear relation between ageing and subjective well-being.” Myers (2000, p. 58) argues that “although many people believe there are unhappy times of life—times of adolescent stress, 582

Happiness and Ageing in the United States

midlife crisis, or old age decline—repeated surveys across the industrialized world reveal that no time in life is notably happiest and most satisfying.” In contrast, Michael Argyle concludes that studies of life satisfaction found happiness increased with age (Argyle, 1999, 2001), while Palmore and Luikart (1972) argue that age has little or no relationship with life satisfaction. Even when U-shapes are found, they are frequently dismissed as largely irrelevant and the scale of the effects are frequently classified as trivial. For example, Cantril (1965) is often cited as finding no evidence of a U-shape in well-being. Yet his study did in fact show them. Cantril reports that when asked to indicate their thoughts about their current life 24.2 percent of those age 0; ∂τ > 0; ∂τ < R,t P,t R,t P,t R,t

t 0; ∂M ∂γt > 0. Therefore, our model predicts that migration from P to R is increasing in the size of the young cohort in P, in the state of the technology in R, in the tax rate in P, and in the international applicability of migrants’ human capital (captured by an increase in γt ).

744

Population Ageing and Migration

Correspondingly, migration from P to R is decreasing in the size of the young cohort in R, in the state of the technology in P, in the tax rate in R, and in time-equivalent migration costs (captured by a decrease in γt ). With a balanced government budget, public expenditures equal tax revenues: GR,t = τR,t αAR,t (NR,t + γt Mt )α

(3)

GP,t = τP,t αAP,t (NP,t − Mt )α . ∂G

(4) ∂G

Differentiating (3) and (4) with respect to Mt gives ∂MR,tt > 0, ∂MP,tt < 0. Therefore, immigration improves public finances in the destination country if public expenditures do not increase correspondingly. This effect would weaken if a considerable share of tax revenues were spent on redistribution among the working-age generation, or on publicly provided private goods that also immigrants consume. The negative effect on the origin country could be alleviated at least partly if migrants were to send remittances or invest in their origin country. The model’s predictions help to explain major recent changes along the dominant migration corridors from Africa and the Middle East to Europe and from Latin America to North America, in line with econometric estimates in Hanson and McIntosh (2016) and numerical predictions from Dao et al. (2021). A youth bulge in Africa and the Middle East, corresponding to ever larger young cohorts NP,t , are a major push factor that combines with the pull factor of population ageing in Europe, corresponding to decreasing NR,t . In Latin America, instead, the population growth rate has leveled off, which reduces push factors to North America, while birth rates in North America have remained at a higher level than in Europe. Increasing migration from countries like India, in turn, reflects both steady increases in the number of graduates and increasing international applicability of their human capital. The model can also be applied to migration within Europe: Even though population growth rates in Eastern Europe have plummeted, much higher standards of living in Western Europe have exerted a powerful gravitational pull, together with free mobility of labor that can be interpreted as a reduction in time-equivalent migration costs. Finally, if the model were extended to migration between different destinations, it could capture the effects of shared language (Adser`a and Pytlikov´a, 2015) or network effects of previous migrants, which could be modeled as an increase in γt .

41.4

Empirical Evidence

Figure 41.4 shows the demographic divergence among continents when analyzing different age groups separately, and Table 41.1 shows how the share of global population living in different regions has developed. While Asia’s share of share of world population aged 0–19 has remained relatively stable, Africa’s share has increased from 10.6 percent in 1950 to 26.2 percent in 2020, and Europe’s share has declined from 17.2 percent in 1950 to 6.1 percent in 2020. Europe’s share decreased and Africa’s increased sharply also in the 20–64 age group. Europe’s share of the elderly has declined less, and Africa’s increased only marginally. Taken together, especially Europe and East Asian economies face decreasing labor force and challenges in taking care of their ageing populations in the absence of migration. In Africa and the Middle East, instead, the challenge is the lack of jobs for the youth bulge. Together with vast gaps in the standard of living, these demographic forces are major pull and push factors on migration. How well or badly does the theoretical framework in the previous section predict recent immigration flows? To evaluate this, we focus on net immigration into OECD countries, as immigration restrictions in destination countries that were not part of the model strongly affect 745

Panu Poutvaara

Figure 41.4 Population by age group: Estimates and projections 1950–2100. Source: United Nations (2019).

Table 41.1 Continental population shares in percent by age groups 1950

2020

2100 (projection)

Panel A: Continental population shares in age group 0–19 Africa 10.6 26.2 Asia 58.7 55.7 Europe 17.2 6.1 Latin America and the Caribbean 7.7 8.1 North America 5.3 3.5 Oceania 0.4 0.5

48.7 36.9 4.8 5.0 4.0 0.7

Panel B: Continental population shares in age group 20–64 Africa 7.9 13.7 Asia 53.7 62.3 Europe 24.2 10.0 Latin America and the Caribbean 6.0 8.6 North America 7.7 4.9 Oceania 0.6 0.5

41.6 42.2 5.4 5.8 4.3 0.7

Panel C: Continental population shares in age group 65+ Africa 5.7 6.5 Asia 44.0 56.6 Europe 34.0 19.6 Latin America and the Caribbean 4.6 8.1 North America 11.0 8.5 Oceania 0.7 0.7

24.3 52.9 7.8 8.7 5.6 0.7

Source: United Nations (2019).

emigration from many poor countries. Figure 41.5 shows how net immigration into OECD countries from 2015 to 2020 related to their gross domestic product (GDP) per capita from 2015 to 2019. Countries with higher GDP per capita attract higher net immigration, in line with the theory. 746

Population Ageing and Migration

Figure 41.5 GDP per capita (2015–2019) and net migration rate (2015–2020). Source: a United Nations (2019); b OECD (2021), Gross domestic product (GDP): GDP per head, US$, current prices, current PPPs, https://stats.oecd.org/index.aspx?queryid=60706, accessed on April 29, 2021.

Figure 41.6 shows how net immigration to OECD countries from 2015 to 2020 responds to demographic development, corresponding to the size of the young cohort NR,t in the model. We instrument the size of the young cohort entering the labor force from 2015 to 2020 by the population of those aged 0–9 in 1995, as they would be aged 20–29 in 2015. We instrument the size of the older cohort gradually leaving the labor force by the population of those aged 35–44 in 1995 and 55–64 in 2015, ageing to 60–69 by 2020. If the ratio is larger (smaller) than one, then the underlying demographic structure would be expected to increase (decrease) labor supply from 2015 to 2020. Figure 41.6 shows that most OECD countries experience underlying demographics that push for decreasing working-age population and net immigration. However, OECD countries experiencing increasing labor force have also received net immigration, except for Mexico. Although refugees from Syria could drive increases in Turkey and refugees from Venezuela could drive increases in Chile and Columbia, the overall conclusion from Figure 41.6 is that cross-country correlation between net immigration and cohort sizes for OECD countries is close to zero in the studied period. The link between demographic fundamentals and immigration appears in global analysis over longer time periods (Hanson and McIntosh, 2016; Dao et al., 2021). Figure 41.7 shows more detailed age distributions of world population and world migrant population in 2020. While people tend to migrate as young adults, the global migrant population, defined as being born outside the country of current residence, is older than the world population on average. Forty-one percent of the world population but only 21.4 percent of the world migrant population was younger than 25 in 2020, and 9.3 percent of the world population but 12.2 percent of the migrant population was 65 years or older. 747

Panu Poutvaara

Figure 41.6 Relative cohort size (1995) and net migration rate (2015–2020). Source: United Nations (2019).

Figure 41.7 Population distribution by age groups: World population vs. migrant population 2020. Source: United Nations (2019).

Figure 41.8 presents the corresponding comparison for Europe where natural population growth rates are low or even negative and immigrants represent a large share of the total population. Even in Europe, at the stock level immigrants comprise a smaller fraction of children and teenagers than natives, if looking only at own migration experience. 748

Population Ageing and Migration

Figure 41.8 Population distribution by age groups in Europe: Total population vs. migrant population 2020. Source: United Nations (2019).

Figure 41.9 Age group distribution: European population vs. incoming immigrant population from nonEU28 countries in 2019. Source: Eurostat, Population on 1 January by age group and sex (DEMO PJANGROUP) and Immigration by age group, sex, and country of previous residence (MIGR IMM5PRV).

However, Figure 41.9 for European Union member states and associated countries for which age information is available shows that the flow of new immigrants is considerably younger than existing populations. 749

Panu Poutvaara

Figure 41.10 Age group distribution of native Germans without migration background compared with first- and second-generation immigrants 2019. Source: Mikrozensus, https://www.destatis.de/DE/Themen/Gesellschaft-Umwelt/Bevoelkerung/ Migration-Integration/Tabellen/migrationshintergrund-alter.html, accessed on April 29, 2021.

Even more important for long-term demographics in destination countries than immigrants’ age when they arrive is their fertility pattern after arrival. Figure 41.10 shows the age distribution of Germans without immigrant background, first-generation immigrants, and secondgeneration immigrants. It highlights that the picture in Figure 41.8 underestimates the contribution of immigration to alleviating demographic change, as in Figure 41.8 second-generation immigrants are counted together with natives. Immigrants being younger than natives is a necessary but insufficient condition for immigration to alleviate demographic challenges. An additional condition is that immigrants must be employed. The first column of Table 41.2 shows the share of the foreign-born population in the working-age population aged 25–54 in Australia, Canada, France, Germany, Italy, the United Kingdom, and the United States. The lower limit is set at 25 to account for a high share of students in younger age groups and the upper limit of 54 reflects the gradual onset of early retirement in older age groups. The share of the foreign-born population ranges from 15 percent in France to more than 30 percent in Australia and Canada, with Germany, the United Kingdom, and the United States in between 22–24 percent shares. The other columns show how employment shares differ between native-born and foreign-born males and females. Employment shares of foreign-born males are almost the same as employment shares of native-born males in Canada, Italy, the United Kingdom, and the United States, and about 7 percentage points lower in France and Germany. Employment shares of foreign-born women are considerably lower than the employment shares of native-born women in all countries, ranging from a 4-percentage-point gap in the United States to a 23-percentage-point gap in France. In terms of addressing challenges arising from population ageing, most countries have considerable potential in increasing female labor force participation rates, especially among immigrants. 750

Population Ageing and Migration Table 41.2 Foreign-born population in selected countries Foreign-Born Population Working-Age (25–54)

Employment-to-Population Ratio (Age Group 25–54)

Native-Born Foreign-Born Native-Born Foreign-Born Males % Males % Females % Females %

Country

Year

Share in %

Australia Canada France Germany Italy United Kingdom United States

2017 2018 2019 2019 2019 2019 2020

35.6 30.7 14.8 23.9 16.9 21.8 21.6

86.4 86.2 91.2 80.6 90.0 93.2

86.0 78.8 84.4 82.1 91.0 92.8

82.5 80.4 85.7 61.8 81.7 93.4

71.8 57.0 66.6 52.6 72.4 89.6

Source: International Labour Organization, 2021, ILOSTAT database, available from https://ilostat.ilo.org/data/.

41.5

Research Agenda for the Future

Most of the world has undergone a demographic transition from an equilibrium with high fertility and low life expectancy to one with low fertility and high life expectancy. This has also transformed migration patterns, with Europe changing from a primary source of global migration flows to an important destination. One of the most important questions for demographic and migration research in the 21st century is how the fertility rate will develop in Africa. Based on current trends, the United Nations population forecast predicts the African population will increase from 1.3 billion in 2020 to 4.3 billion in 2100, and the number of those aged 0–19 will exceed 1 billion by year 2050. So far, Africa’s share of global migration flows has been way below its population share. Even modest increases in emigration from Africa would generate major increases in immigration pressure in the rest of the world, mostly in Europe. If population growth slowed significantly, as Vollset et al. (2020) suggest, this would reduce emigration pressure. Previous research shows that education, development, and population growth are closely related. Educated women tend to have fewer children (Currie and Moretti, 2003; Osili and Long, 2008), meaning that changes in educational trajectories could also change population forecasts and migration pressure. Mountford and Rapoport (2016) highlight the quantity-quality trade-off in the number and education of children for Africa and show that European immigration policy that selects primarily high-skilled migrants could depress fertility in Africa, by encouraging parents to move resources from investing in more children to providing their children with more human capital. This way, high-skilled migration flows from Africa to Europe that appear as a brain drain in a static context with given human capital investments could translate into a boost to human capital formation and a boon in slowing down population growth in a dynamic context, once accounting for parental incentives. These insights highlight that better modeling of future demographic development, especially the interaction between education and fertility, is centrally important to forecasting future migration flows and their effects. To provide more reliable policy advice, researchers should build in uncertainty about scenarios they use and analyze how their conclusions would change under different scenarios. The other crucial question is how climate change affects migration flows. Berlemann and Steinhardt (2017) summarize recent research on climate change and migration, concluding that 751

Panu Poutvaara

rising temperatures tend to induce outmigration, especially in countries dependent on agriculture. Climate change is also associated with more civil conflicts, thereby increasing the number of refugees fleeing war or persecution (Hsiang et al., 2011). The projected population increase of 3 billion by the end of the century in Africa alone coincides with potentially devastating effects of climate change on agricultural productivity and a specter of environmental disasters from the joint pressures of climate change and rapid population increase. Climate change and rising oceans also threaten the viability of densely populated coastal areas and whole Pacific nations, creating an increasing population of climate refugees. Beine and Parsons (2017) find significant heterogeneity in the effects of natural disasters on emigration: Natural disasters increase emigration to neighboring countries and, for middle-income origins, to former colonial powers. Therefore, climate change can be expected to strengthen the push and pull factors arising from population ageing in former colonial powers in Europe and high fertility rates in Africa and the Middle East where most global conflicts occur. Drastic improvements in communication technology, especially mobile Internet access, make migration much easier and allow migrants to remain in contact with family members left behind. Recent research has extended gravity models of migration decisions by modeling the decision whether to acquire costly information (Bertoli et al., 2020). Migrants rationally invest less in acquiring information about countries that they expect not to be optimal destinations or that have high information costs. Could improvements in information technology result in tipping points in which international migration patterns, say, between Africa and Europe, could rapidly change from a high-information-cost and low-migration equilibrium to a low-information-cost and high-migration equilibrium? Adema et al. (2021) analyze data from 120 countries and 2,200 subnational regions and find that mobile Internet adoption increases desire and plans to emigrate. Another important research frontier concerns the interaction of demographic change, migration, and healthcare. With increasing life expectancy, healthcare and personal services targeting the elderly will gain importance. An increasing fraction of elderly households in rich countries could afford to hire immigrants from low-wage countries to provide care at home, as an alternative to a nursing home. With sufficiently flexible regulations and moderate employer costs, governments in ageing countries could co-opt such services to reduce pressure on nursing homes. However, strict regulations that make legal hiring prohibitively expensive could force such services to remain in the shadow economy, to the detriment of migrants and public finances. This illustrates a more general choice facing governments in destination countries. They can try to stem migration flows or manage them with an aim of realizing potential mutual gains from labor shortages in Europe, East Asia, and other destinations, and youth bulges in Africa and the Middle East.

Note 1 The author thanks Clara Albrecht and Simon Meemken for outstanding research assistance and Cevat Giray Aksoy and an anonymous reviewer for helpful comments.

References ABEL, G. J., AND SANDER, N. (2014): “Quantifying global international migration flows,” Science, 343(6178): 1520–1522. Available at https://doi.org/10.1126/science.1248676. ADEMA, J., AKSOY, C. G., AND POUTVAARA, P. (2021): Mobile Internet access and the desire to emigrate. ifo Working Paper No. 365. Munich, Germany. Available at https://www.ifo.de/ en/publikationen/2021/working-paper/mobile-internet-access-and-desire-emigrate. ADSER A` , A., AND PYTLIKOV A´ , M. (2015): “The role of language in shaping international migration,” The Economic Journal, 125(586): F49–F81. Available at https://doi.org/10.1111/ecoj.12231.

752

Population Ageing and Migration

AKSOY, C. G., AND POUTVAARA, P. (2021): “Refugees’ and irregular migrants’ self-selection into Europe,” Journal of Development Economics, 152: 102681. Available at https://doi.org/10.1016/ j.jdeveco.2021.102681. AKSOY, C. G., POUTVAARA, P., AND SCHIKORA, F. (2020): “First time around: Local conditions and multi-dimensional integration of refugees,” IZA Discussion Paper No. 13914. Bonn, Germany. Available at https://docs.iza.org/dp13914.pdf. ANGRIST, J. D., AND KUGLER, A. D. (2003): “Protective or counter-productive? Labour market institutions and the effect of immigration on EU natives,” The Economic Journal, 113(488): F302–F331. Available at https://doi.org/10.1111/1468-0297.00136. BARSBAI, T., RAPOPORT, H., STEINMAYR, A., AND TREBESCH, C. (2017): “The effect of labor migration on the diffusion of democracy: Evidence from a former Soviet republic,” American Economic Journal: Applied Economics, 9(3): 36–69. Available at https://doi.org/10.1257/app.20150517. BATTISTI, M., FELBERMAYR, G., PERI, G., AND POUTVAARA, P. (2018): “Immigration, search and redistribution: A quantitative assessment of native welfare,” Journal of the European Economic Association, 16(4): 1137–1188. Available at https://doi.org/10.1093/jeea/jvx035. ¨ ZDEN, C BEINE, M., DOCQUIER, F., AND O ¸ . (2011): “Diasporas,” Journal of Development Economics, 95(1): 30–41. Available at https://doi.org/10.1016/j.jdeveco.2009.11.004. BEINE, M., AND PARSONS, C. R. (2017): “Climatic factors as determinants of international migration: Redux,” CESifo Economic Studies, 63(4): 386–402. Available at https://doi.org/10.1093/cesifo/ifx017. BERLEMANN, M., AND STEINHARDT, M. F. (2017): “Climate change, natural disasters, and migration—A survey of the empirical evidence,” CESifo Economic Studies, 63(4): 353–385. Available at https://doi.org/10.1093/cesifo/ifx019. ´ BERTOLI, S., AND FERN ANDEZ -HUERTAS MORAGA, J. (2013): “Multilateral resistance to migration,” Journal of Development Economics, 102: 79–100. Available at https://doi.org/10.1016/ j.jdeveco.2012.12.001. ´ BERTOLI, S., FERN ANDEZ -HUERTAS MORAGA, J., AND GUICHARD, L. (2020): “Rational inattention and migration decisions,” Journal of International Economics, 126: 103364. Available at https://doi.org/10.1016/j.jinteco.2020.103364. BHAGWATI, J. N., AND HAMADA, K. (1974): “The brain drain, international integration of markets for professionals and unemployment: A theoretical analysis,” Journal of Development Economics, 1(1): 19–42. Available at https://doi.org/10.7916/D8XK8RCB. BORJAS, G. J. (1987): “Self-selection and the earnings of immigrants,” American Economic Review, 77(4): 531–553. Available at http://www.jstor.org/stable/1814529. BORJAS, G. J. (1995): “The economic benefits from immigration,” Journal of Economic Perspectives, 9(2): 3–22. Available at https://doi.org/10.1257/jep.9.2.3. BORJAS, G. J. (1999): “Immigration and welfare magnets,” Journal of Labor Economics, 17(4): 607–637. Available at https://doi.org/10.1086/209933. BORJAS, G. J. (2001): “Does immigration grease the wheels of the labor market?” Brookings Papers on Economic Activity, 32(1): 69–133. Available at https://doi.org/10.1353/eca.2001.0011. BORJAS, G. J. (2003): “The labor demand curve is downward sloping: Reexamining the impact of immigration on the labor market,” The Quarterly Journal of Economics, 118(4): 1335–1374. Available at https://doi.org/10.1162/003355303322552810. BORJAS, G. J., KAUPPINEN, I., AND POUTVAARA, P. (2019): “Self-selection of emigrants: Theory and evidence on stochastic dominance in observable and unobservable characteristics,” The Economic Journal, 129(617): 143–171. Available at https://doi.org/10.1111/ecoj.12585. BRELL, C., DUSTMANN, C., AND PRESTON, I. (2020): “The labor market integration of refugee migrants in high-income countries,” Journal of Economic Perspectives, 34(1): 94–121. Available at https://doi.org/10.1257/jep.34.1.94. CARD, D., DUSTMANN, C., AND PRESTON, I. (2012): “Immigration, wages, and compositional amenities,” Journal of the European Economic Association, 10(1): 78–119. Available at https://doi.org/10.1111/j.1542-4774.2011.01051.x. CATTANEO, C., AND PERI, G. (2016): “The migration response to increasing temperatures,” Journal of Development Economics, 122: 127–146. Available at https://doi.org/10.1016/j.jdeveco.2016.05.004. CERVELLATI, M., AND SUNDE, U. (2017): “The demographic transition and long-term development,” In Cervellati, M., and Sunde, U. (eds.): Demographic Change and Long-Run Development. Massachusetts/London: MIT Press, pp. 1–28. CHASSAMBOULLI, A., AND PALIVOS, T. (2014): “A search-equilibrium approach to the effects of immigration on labor market outcomes,” International Economic Review, 55(1): 111–129. Available at

753

Panu Poutvaara

https://doi.org/10.1111/iere.12043. CLEMENS, M. A. (2011): “Economics and emigration: Trillion-dollar bills on the sidewalk?” Journal of Economic Perspectives, 25(3): 83–106. Available at https://doi.org/10.1257/jep.25.3.83. CLEMENS, M. A. (2013): “Why do programmers earn more in Houston than Hyderabad? Evidence from randomized processing of US visas,” American Economic Review, 103(3): 198–202. Available at https://doi.org/10.1257/aer.103.3.198. CURRIE AND MORETTI (2003): “Mother’s education and the intergenerational transmission of human capital: Evidence from college openings,” The Quarterly Journal of Economics, 118(4): 1495–1532. Available at https://doi.org/10.1162/003355303322552856. DAO, T. H., DOCQUIER, F., MAUREL, M., AND SCHAUS, P. (2021): “Global migration in the twentieth and twenty-first centuries: The unstoppable force of demography,” Review of World Economics, 157(2): 417–449. Available at https://doi.org/10.1007/s10290-020-00402-1. DISNEY, R. (2004): “Are contributions to public pension programmes a tax on employment?” Economic Policy, 19(39): 268–311. Available at https://doi.org/10.1111/j.1468-0327.2004.00124.x. DOCQUIER, F., AND RAPOPORT, H. (2012): “Globalization, brain drain, and development,” Journal of Economic Literature, 50(3): 681–730. Available at https://doi.org/10.1257/jel.50.3.681. ¨ ZTUNC, J., AND POUTVAARA, P. (2019): “Immigration and electoral EDO, A., GIESING, Y., O support for the far-left and the far-right,” European Economic Review, 115: 99–143. Available at https://doi.org/10.1016/j.euroecorev.2019.03.001. FACKLER, T. A., GIESING, Y., AND LAURENTSYEVA, N. (2020): “Knowledge remittances: Does emigration foster innovation?” Research Policy, 49(9): 103863. Available at https://doi.org/ 10.1016/j.respol.2019.103863. FLAHAUX, M.-L., AND DE HAAS, H. (2016): “African migration: trends, patterns, drivers,” Comparative Migration Studies, 4(1). Available at https://doi.org/10.1186/s40878-015-0015-6. GOULD, D. M. (1994): “Immigrant links to the home country: Empirical implications for U.S. bilateral trade flows,” The Review of Economics and Statistics, 76(2): 302. Available at https://doi.org/10.2307/2109884. GROGGER, J., AND HANSON, G. H. (2011): “Income maximization and the selection and sorting of international migrants,” Journal of Development Economics, 95(1): 42–57. Available at https://doi.org/10.1016/j.jdeveco.2010.06.003. HAAS, H. DE, VEZZOLI, S., SZCZEPANIKOVA, A., AND VAN CRIEKINGE, T. (2018): European Migrations. Dynamics, Drivers, and the Role of Policies, Luxembourg: Publications Office of the European Union. ¨ HALLA, M., WAGNER, A. F., AND ZWEIM ULLER , J. (2017): “Immigration and voting for the far right,” Journal of the European Economic Association, 15(6): 1341–1385. Available at https://doi.org/10.1093/jeea/jvx003. HANSON, G., AND MCINTOSH, C. (2016): “Is the Mediterranean the new Rio Grande? US and EU immigration pressures in the long run,” Journal of Economic Perspectives, 30(4): 57–82. Available at https://doi.org/10.1257/jep.30.4.57. HATTON, T. J. (2020): “Asylum migration to the developed world: Persecution, incentives, and policy,” Journal of Economic Perspectives, 34(1): 75–93. Available at https://doi.org/10.1257/jep.34.1.75. HSIANG, S. M., MENG, K. C., AND CANE, M. A. (2011): “Civil conflicts are associated with the global climate,” Nature, 476: 438–441. Available at https://doi.org/10.1038/nature10311. KENNAN, J., AND WALKER, J. R. (2011): “The effect of expected income on individual migration decisions,” Econometrica, 79(1): 211–251. Available at https://doi.org/10.3982/ECTA4657. KERR, W. R. (2008): “Ethnic scientific communities and international technology diffusion,” The Review of Economics and Statistics, 90(3): 518–537. Available at https://doi.org/10.1162/rest.90.3.518. KLEIN GOLDEWIJK, K., BEUSEN, A., AND JANSSEN, P. (2010): “Long-term dynamic modeling of global population and built-up area in a spatially explicit way: HYDE 3.1.,” The Holocene, 20(4): 565–573. Available at https://doi.org/10.1177/0959683609356587. KOETHENBUERGER, M., POUTVAARA, P., AND PROFETA, P. (2008): “Why are more redistributive social security systems smaller? A median voter approach,” Oxford Economic Papers, 60(2): 275–292. Available at https://doi.org/10.1093/oep/gpm043. LUCAS, R. E. B. (1997): “Internal migration in developing countries.” In Rosenzweig, M. R., and Stark, O. (eds.): Handbook of Population and Family Economics, Volume 1, Part B, chapter 13, pp. 721–798. Available at https://doi.org/10.1016/S1574-003X(97)80005-0. LUCAS, R. E. B., AND STARK, O. (1985): “Motivations to remit: Evidence from Botswana,” Journal of Political Economy, 93(5): 901–918. Available at https://doi.org/10.1086/261341.

754

Population Ageing and Migration

MAYDA, A. M. (2010): “International migration: A panel data analysis of the determinants of bilateral flows,” Journal of Population Economics, 23(4): 1249–1274. Available at https://doi.org/10.1007/s00148009-0251-x. MENG, X. (2012): “Labor market outcomes and reforms in China,” Journal of Economic Perspectives, 26(4): 75–102. Available at https://www.jstor.org/stable/23290281. MOUNTFORD, A. (1997): “Can a brain drain be good for growth in the source economy?” Journal of Development Economics, 53(2): 287–303. Available at https://doi.org/10.1016/S0304-3878(97)00021-7. MOUNTFORD, A., AND RAPOPORT, H. (2016): “Migration policy, African population growth and global inequality,” The World Economy, 39(4): 543–556. Available at https://doi.org/ 10.1111/twec.12268. OSILI, U. O., AND LONG, B. T. (2008): “Does female schooling reduce fertility? Evidence from Nigeria,” Journal of Development Economics, 87(1): 57–75. Available at https://doi.org/10.1016/ j.jdeveco.2007.10.003. OTTAVIANO, G. I. P., AND PERI, G. (2012): “Rethinking the effect of immigration on wages,” Journal of the European Economic Association, 10(1): 152–197. Available at https://doi.org/10.1111/j.15424774.2011.01052.x. PARSONS, C., AND VE´ ZINA, P.-L. (2018): “Migrant networks and trade: The Vietnamese boat people as a natural experiment,” The Economic Journal, 128(612): F210–F234. Available at https://doi.org/10.1111/ecoj.12457. POUTVAARA, P. (2000): “Education, mobility of labour and tax competition,” International Tax and Public Finance, 7(6): 699–719. Available at https://doi.org/10.1023/A:1008733527427. POUTVAARA, P. (2004): “Educating Europe: Should public education be financed with graduate taxes or income-contingent loans?,” CESifo Economic Studies, 50(4): 663–684. Available at https://doi.org/10.1093/cesifo/50.4.663. POUTVAARA, P. (2007): “Social security incentives, human capital investment and mobility of labor,” Journal of Public Economics, 91(7): 1299–1325. Available at https://doi.org/10.1016/j.jpubeco.2007.03.001. POUTVAARA, P. (2008): “Public and private education in an integrated Europe: Studying to migrate and teaching to stay?” Scandinavian Journal of Economics, 110(3): 591–608. Available at https://doi.org/10.1111/j.1467-9442.2008.00552.x. POUTVAARA, P., AND KANNIAINEN, V. (2000): “Why invest in your neighbor? Social contract on educational investment,” International Tax and Public Finance, 7(4): 547–562. Available at https://doi.org/10.1023/A:1008745708377. RAPOPORT, H., AND DOCQUIER, F. (2006): “The economics of migrants’ remittances.” In Kolm, S.-C., and Mercier-Ythier, J. (eds.): Handbook of the Economics of Giving, Altruism and Reciprocity, North Holland, The Netherlands, pp. 1135–1198. Available at https://doi.org/10.1016/S1574-0714(06)02017-3. RAUCH, J. E., AND TRINDADE, V. (2002): “Ethnic Chinese networks in international trade,” Review of Economics and Statistics, 84(1): 116–130. Available at https://doi.org/10.1162/003465302317331955. RAZIN, A., AND SADKA, E. (2000): “Unskilled migration: A burden or a boon for the welfare state?” Scandinavian Journal of Economics, 102(3): 463–479. Available at https://doi.org/10.1111/1467-9442. 00210. REHER, D. S. (2004): “The demographic transition revisited as a global process,” Population, Space and Place, 10(1): 19–41. Available at https://doi.org/10.1002/psp.313. RILEY, J. C. (2005): “Estimates of regional and global life expectancy, 1800–2001,” Population and Development Review, 31(3): 537–543. Available at https://doi.org/10.1111/j.1728-4457.2005.00083.x. ROY, A. D. (1951): “Some thoughts on the distribution of earnings,” Oxford Economic Papers, 3(2): 135–146. SINN, H.-W. (1997): “The selection principle and market failure in systems competition,” Journal of Public Economics, 66(2): 247–274. Available at https://doi.org/10.1016/S0047-2727(97)00043-1. SJAASTAD, L. A. (1962): “The costs and returns of human migration,” Journal of Political Economy, 70(5, Part 2): 80–93. Available at https://doi.org/10.1086/258726. SMITH, A. (1776, reprinted 1976): An Inquiry into the Nature and Causes of the Wealth of Nations, Indianapolis: Liberty Classics. SPILIMBERGO, A. (2009): “Democracy and foreign education,” American Economic Review, 99(1): 528–543. Available at https://doi.org/10.1257/aer.99.1.528. ¨ STARK, O., HELMENSTEIN, C., AND FURNKRANZ -PRSKAWETZ, A. (1997): “A brain gain with a brain drain,” Economics Letters, 55(2): 227–234. Available at https://doi.org/10.1016/S0165-1765(97)00085-2. STORESLETTEN, K. (2000): “Sustaining fiscal policy through immigration,” Journal of Political Economy, 108(2): 300–323. Available at https://doi.org/10.1086/262120.

755

Panu Poutvaara

STORESLETTEN, K. (2003): “Fiscal implications of immigration—A net present value calculation,” Scandinavian Journal of Economics, 105(3): 487–506. Available at https://doi.org/10.1111/1467-9442.t01-200009. UNITED NATIONS. (2019): World Population Prospects 2019. Online edition. Rev. 1. 2019. Department of Economic and Social Affairs, Population Division, New York. VAN M OL , C., AND DE VALK , H. (2016): Migration and immigrants in Europe: A historical and demographic perspective. In Garc´es-Mascare˜nas, B., and Penninx, R. (eds.): Integration Processes and Policies in Europe. Contexts, Levels and Actors, Cham, Springer Open, pp. 31–55. VOLLSET, S. E., GOREN, E., YUAN, C. W., CAO, J., SMITH, A. E., HSIAO, T., BISIGNANO, C., ET AL. (2020): “Fertility, mortality, migration, and population scenarios for 195 countries and territories from 2017 to 2100: A forecasting analysis for the Global Burden of Disease Study,” The Lancet, 396(10258): 1285–1306. Available at https://doi.org/10.1016/S0140-6736(20)30677-2. WILDASIN, D. E. (1991): “Income redistribution in a common labor market,” American Economic Review, 81(4): 757–774. Available at http://www.jstor.org/stable/2006641. WILDASIN, D. E. (2000): “Labor-market integration, investment in risky human capital, and fiscal competition,” American Economic Review, 90(1): 73–95. Available at https://doi.org/10.1257/aer.90.1.73. YANG, D. (2008): “International migration, remittances and household investment: Evidence from Philippine migrants’ exchange rate shocks,” The Economic Journal, 118(528): 591–630. Available at https://doi.org/10.1111/j.1468-0297.2008.02134.x. ZIJDEMAN, R., AND RIBEIRA DA SILVA, F. (2015): Life expectancy at birth (total). Available at https://datasets.iisg.amsterdam/dataset.xhtml?persistentId=hdl:10622/LKYT53 [accessed on April 21, 2021].

756

INDEX

accounting/macroeconomic approach 22, 24, 32 adverse selection 146, 151, 206 age discrimination 341, 370, 392, 411, 442, 731 age gradient 543 age inflation 343 age patterns 306, 496, 532, 542 age profile 289, 404, 422, 491, 525, 540, 716 age profile of preferences 539 age reallocations 487 age structure 343, 368, 417, 418, 438, 494, 496, 609, 644, 714 age-dependency ratio 237 age-patience nexus 542 age-productivity 352, 415, 418 age-specific mortality rates 468, 516, 716, 736 age-training 355 age-wage profiles 419 age-wealth profile 303, 307, 310 ageing society 329, 343 algorithms 95, 317 artificial intelligence (AI), 75, 91, 322, 358, 369 assortative matching 158, 169 asymmetric information 143 attitudes to risk 146 augmentation 89 authoritarianism 611 automation 75, 317, 319, 358, 369 Ballod’s Belastungs Koeffizienten 507 behavioral biases 306 benefit level 252, 264 benefit-cost analysis (BCA), 567 bequeathable wealth 218, 220 bequest motives 306, 308 bias in perception 128 biased beliefs 621, 624, 629 birth control 109

bridge jobs 330, 370, 521 budget constraint 28, 29 capital deepening 291, 497, 500, 502 capital-labor ratio 291, 717 capitalization index 202 care crisis 648, 708 casualty sensitivity 604, 607 changing skills demands 349, 358 characteristics approach 516 child deficit 490, 493 childlessness 444 chronic conditions 107, 664 climate change 666, 751 clinical knowledge 619 co-payment 680, 682 co-residency 234, 707 cognition 86, 108, 382, 386, 397, 411, 482, 546, 696 cognitive ability 90, 108, 308, 396, 399, 403, 546, 623, 730 cognitive ageing 84, 464, 546 cognitive reserve 88 cognitive reward 400 complementarity 71, 170, 225, 351, 355, 367, 741 consumption-based simulation model 219 contagion externalities 143 contributor-pensioner ratio 691, 693 cost-benefit analysis (CBA), 133 cost-effectiveness 40, 45, 148, 575 cost-utility analysis (CUA), 133 COVID-19, 1–3, 21, 38, 45, 95, 128, 139, 159, 245, 273, 370, 406, 438, 482, 516, 533, 566, 579, 629, 648, 666, 731 creative productivity 95 crude birth rate 658 crude death rate 658

757

Index

crystallized intelligence 396 cues 91, 619 deadweight loss 135, 355 decision making 90, 108, 128, 391, 492, 544, 696 delayed earnings 401 dementia 92, 108, 142, 464, 480 demographic change 22, 236, 321, 324, 447, 605, 616, 643, 724, 739 demographic dividend 294, 486, 495, 714 demographic transition 142, 234, 288, 329, 418, 495, 658, 713, 736 depression 107, 164, 382, 578, 582, 699 design frictions 620 digital divide 89, 443 digital revolution 369 digital world 667 digitalization 61, 75, 657 disability insurance 204, 372, 464 disability-adjusted life year (DALY), 124, 513 disease prevention policies 129 displacement 321, 366 dynamic life-cycle model 336 early retirement 205, 248, 263, 373, 381, 386, 391, 399, 411, 442, 750 early-life factors 160, 169 economic dependency 289, 507, 724 economic preferences 221, 540 education 88, 160, 200, 225, 239, 322, 332, 342, 350, 365, 399, 699, 742 educational attainment 206 educational gap 384 effectiveness 62, 110, 128, 264, 356, 623 efficiency, dynamic 134, 275 efficiency, static 135 elasticity of mortality 65, 468 elasticity of substitution 51, 368 elderly disease prevention 123, 131 emerging economies 74, 234, 305 empowerment 447, 605, 611 ethno-cultural anxiety 611 extended reality 96 fair innings argument 567 family relationships 437, 445 family structures 667 family support 235, 427, 707 family transfers 249, 487, 494 fertility rate 103, 142, 439, 498, 606, 659, 717 financial education 726 financial incentives 269, 371, 390, 625, 627 financial inclusion 245, 312 financial literacy 270, 372, 447 financial security 130, 187, 306 financial well-being 179 flexible retirement 271, 392

fluid intelligence 396, 411, 482 flynn effects 398 force of mortality 456 forecast 22, 24 foreign policy preferences 604 forgetting 623 formal education 726 frailty index 55, 338, 455, 463 fund transfers 490 future elderly model (FEM), 25, 36 gender differences 103, 268, 386, 441, 629, 729 gender equality 114, 370, 437, 439 gender health gaps 102, 110 gender imbalance 674 gender inequality 109, 263, 727 gender mortality gap 714 gender norms 646 gender pension gap 267, 444 gender ratio 674, 675, 679 gender-related attitudes 639, 645 Gini index 726 Gompertz-Makeham law 455, 458 Great Recession 219, 386, 525, 578 gross domestic product (GDP), 23, 30 happiness 109, 389, 533, 571, 578, 582 happiness curves 580 Health and Retirement Study 25, 44, 219, 307, 382, 474, 525, 692 health behavior 67, 112, 159, 389, 465, 623, 700 health capital approach 50 health deficit index 44, 49, 464 health deficits 44, 455, 466 health expenditure 24, 41, 63, 126, 152 health inequality 53, 467 health insurance 25, 41, 51, 64, 70, 126, 189, 308, 372, 703 health risk regulation 567 health shocks 218, 291, 307, 392, 598 health spending 25 health system 130, 189, 309, 373, 533, 619, 648, 668, 720 health system barriers 620 health transition 26, 28 health, cognitive 93, 108, 515, 547 health, mental 49, 107, 159, 167, 179, 323, 386, 578, 664, 691 health, physical 105, 159, 167, 182, 323, 386, 547, 695 healthcare spending 23, 61, 63 heterogeneity in preferences 539 heuristics 128, 616 hiring discrimination 370 Hukou 695, 739 human capital accumulation 50, 181 human capital investments 349, 355, 357, 401, 479, 751

758

Index

human capital theory 354, 382 hybrid pension scheme 246 immigration policy 643, 751 immigration restrictions 745 immigration-related attitudes 645 inattention 623 income elasticity 41 income elasticity of healthcare spending 64 income replacement rate 219 income volatility 304 incremental cost-effectiveness ratio (ICER), 130 independent living 84, 181 individual ageing 199, 466, 643, 666 industrial robots 318, 322 inflation 218, 270, 289, 512 informal childcare 446 informal employment 238 informality 235, 238, 306, 724 informality rates 714 information and communication technology 369, 426 information costs 752 insurance elasticity 64 interest rate 46, 276, 288, 427, 500 intergenerational equity 276, 371, 510 intergenerational relationships 437, 447 intergenerational transfers 234, 289, 724 internal rate of return (IRR), 202, 275 international migration 324, 666, 735 international peace 605 internationalism 607 IQ score 397 isolationism 607 job performance 396, 401, 412 Kaldor-Hicks efficiency 568 knowledge gaps 624 labor demand 341, 368, 442 labor force participation 22, 92, 103, 124, 249, 264, 331, 363, 441, 496, 646, 725, 750 labor market 109, 151, 242, 306, 322, 330, 350, 366, 386, 437, 441, 516, 585, 648, 694, 726, 739 labor market competition 651 labor market flexibility 741 labor market participation 51, 151, 329, 399 labor market polarization 404 labor productivity 322, 410, 413, 415, 442, 500 labor scarcity effect 425 labor supply 48, 208, 271, 291, 329, 335, 340, 364, 410, 439, 641, 731 labor supply elasticities 442 Latin American paradox 727 learning by doing 352 leisure 29, 88, 126, 330, 367, 381, 522

life expectancy 22, 41, 63, 104, 127, 142, 182, 200, 221, 236, 266, 317, 329, 364, 383, 444, 458, 501, 515, 520, 548, 567, 657, 661, 692, 714 life expectancy at birth 50, 237, 317, 458, 500, 662, 692, 735 life satisfaction 109, 183, 533, 579 life spans 103, 253, 459, 578 life-cycle deficit 489, 499, 714 life-cycle dynamic programming model 219 life-cycle effect 641 life-cycle model 35, 41, 71, 222, 305, 306, 310, 330, 468, 498 life-cycle model, stochastic survival 47, 50 life-cycle model/permanent income hypothesis 303 life-cycle optimization behaviors 521 life-cycle theory 356, 364 lifelong learning 349, 354, 369, 443 lifestyle habits 619 lifetime budget constraint 499 lifetime labor income 207 lifetime well-being 129, 571 long-term care 126, 139, 140, 189, 231, 268, 292, 427, 610, 707 longevity 24, 45, 62, 127, 200, 237, 265, 308, 329, 365, 459, 544, 597 longevity dividend 84, 330, 340 loss of redundancy 456 machine learning 95, 317 macroeconomic/accounting approach 22, 24, 32 male gerontocracy 439, 447 male-female health survival paradox 111 marginal productivity 320, 368 marginal rate of substitution 48, 569 market size effects 51, 68 marriage 151, 159, 323, 401, 590 marriage market 159, 161, 401 marriage/cohabitation rates 401 matching model 169 medical innovation 40, 45, 67 medical price inflation 70 medical progress 46, 61, 64, 439, 468 medical progress, endogenous 72 medical progress, exogenous 69 medical progress, private/public 72 medical technologies 25 mental disorder 125, 179, 699 mental illness 178, 179 mental retirement 382, 390 microsimulation techniques 267 midlife crisis 583 migration 237, 324, 499, 641, 666, 729, 735 military 398, 605, 729 moral hazard 76, 151, 373 morbidity 41, 63, 125, 186, 338, 459, 575, 663 morbidity-mortality paradox 464

759

Index

mortality 26, 45, 65, 125, 185, 199, 221, 253, 269, 306, 337, 364, 390, 403, 455, 501, 582, 658, 663, 692, 713 mortality gradient 203 mortality risk 26, 51, 203, 225, 566, 572, 595, 666, 716 mortality selection 115, 595 motivation 148, 188, 330, 356, 411, 628, 706 multi-pillar pension system 210, 264 multi-pillar pension taxonomy 241 myopia 270 national transfer accounts 292, 486, 508, 716 nationalism 611 natives 643, 741, 750 Neumann-Morgenstern utility functions 571 non-path-dependent mortality 51 noncommunicable diseases 124, 616, 695 noneconomic dependency 512 nonfinancial incentives 625 notional defined contribution 264 nudging 620 old-age deficit 490, 494 old-age dependency ratio 23, 69, 237, 288, 317, 330, 426, 506, 515, 644, 716 old-age income inequality 266, 710 old-age poverty 256, 264, 312, 444 opportunity costs 55, 151, 309, 324, 439, 569 optimal retirement 336, 339, 527 optimism bias 128 outmigration 691, 708, 729, 752 overlapping generations 28, 51, 70, 208, 251, 321 oversampling 482 pareto optimal 72, 288 patience 542 patient interventions 625, 629 pay schemes 418 pension coverage gaps 262 pension eligibility 255, 268, 382, 399 pension expenditures 255, 445 pension plans 202, 506 pension plans, nonprogressive 204 pension plans, progressive 204 pension policy 234, 240, 263 pension system, defined benefit 201, 218, 242, 263, 300, 444 pension system, defined contribution 201, 204, 218, 242, 263, 444 pension system, private 263, 720 pension system, public 201, 263, 295, 306, 509, 720 pension tax rate 509 pension wealth 201, 392, 499 pension, funded 70, 202, 262, 510 pension, pay-as-you-go 202, 262, 510, 741 pension-based migration 253

pensions 23, 126, 218, 292, 336, 371, 403, 427, 441, 477, 521, 610, 640, 721 perpetual youth model 458 population ageing 21, 63, 102, 123, 152, 182, 235, 268, 287, 319, 396, 411, 437, 492, 506, 604, 639, 657, 691, 714, 735 population growth 238, 321, 500, 609, 658, 709, 751 population profiles 660 post-retirement income 227, 308 poverty 182, 235, 267, 371, 707, 726 power liabilities 610 pre-retirement income 226 preference-based approaches 571 preparation for retirement 217, 223, 477 present bias 542, 622 Prevent-Rehabilitate-Augment-Substitute 87 primary income 293 prioritarian evaluation 568 prioritarianism 572 productivity growth 31, 70, 292, 410, 499, 609 prospective old-age threshold 516 public pensions 199, 275, 308, 509 public spending 22 quality-adjusted life year (QALY), 45, 124, 533, 567 quiet revolution 437 rationality 128 reallocation system 487, 490 red herring hypothesis 147 rehabilitation 88, 130, 665 rejuvenation 417, 666 reliability theory 455, 461 replacement rate 201, 247, 266, 317, 509, 659 resident pension 693 retired husband syndrome 391 retirement 36, 92, 113, 187, 201, 217, 240, 288, 303, 329, 350, 442, 475, 498, 520, 694, 717 retirement age 205, 262, 268, 330, 356, 364, 381, 411, 429, 441, 526, 694 retirement behavior 263, 291, 349 retirement decisions 208, 262, 356, 391, 521 retirement rates 598, 694 retirement system 263, 363, 371, 694 retirement wealth 306, 309, 477 return on pension assets 512 return to education 53, 208 risk attitudes 541 risk diversification 270 robotics 75, 91, 289, 299, 342 salience 623 sandwich generation 445 saving behavior 291, 304, 307 saving rate 36, 291, 304, 310, 713, 726 savings 28, 70, 126, 187, 208, 221, 295, 304, 306, 521, 548, 718

760

Index

scarcity 143, 299, 744 sectorial transformation 369 secular stagnation 290, 411, 425 selection bias 419, 595 self-control problems 128, 623, 628 self-employment 272, 374 self-productivity of skills 351 self-selection 740 sex discrimination 267 sex ratio 103 skill bias 71, 369 skill-biased earnings growth 71 skill-biased technological change 404 skills obsolescence 349, 353 social accountability 628 social assistance 672, 684 social inequality 103 social isolation 95, 185 social networks 96, 113, 179, 188 social norms 401, 530, 620 social pensions 235, 241, 243 social programs 180, 231, 714 social protection systems 671, 672, 685 social security 27, 65, 221, 241, 264, 336, 371, 382, 608, 643, 693, 714 social security systems 199, 317, 445, 529 social value of risk reduction 572 social welfare function approach 570 socioemotional selectivity theory 344 Solow growth model 31, 290 spillover effects 159 spousal health 158, 167 statutory retirement 381, 389 steady-state design 478 Strehler–Mildvan correlation 459 stress 153, 322, 390, 578, 582 subjective well-being 109, 187, 308, 533, 582 subsidies 355, 693 substitution 91, 146, 231, 322, 337, 528, 741 suicide 183, 404, 578, 582 support costs 127, 133 support ratio 69, 237, 289, 495, 497, 506, 717 support ratio, general 294 support ratio, weighted 294 surplus wealth 608 survival probability 202, 308, 456, 569, 572 survival rates 48, 70, 236, 300, 502

survival risk 217 survivor pensions 268 sustainability 61, 70, 139, 199, 211, 235, 248, 271, 363, 447, 649, 722, 741 targeting 159, 252 technological change 31, 41, 350, 354, 369, 423, 426, 497, 744 technological unemployment 317, 322 theory of ageing 343 time allocation 521, 731 time preferences 541 time use categories 524 total factor productivity 70, 85, 410, 417, 609 total fertility rates 236, 298, 317, 440, 692, 735 training motivation 356 transaction costs 209, 275 transfer load 294 transfer, private 239 transfer, public 239 two-child policy 708 uncertainty 36, 142, 207, 224, 276, 304, 323, 523, 736 unhealthy behaviors 391, 708 unionized labor market 741 universal health coverage 130, 657, 664, 720 unpaid workforce 145 upskilling 443 utilitarianism 571 value of a statistical life 48 value of a statistical life (VSL), 67, 567 wealth accumulation 181, 218, 304, 608 wealth accumulation decisions 303, 312 wealth decumulation 303, 307 well-being 129, 179, 183, 201, 256, 271, 288, 324, 372, 475, 570, 580, 659 widowhood 596 willingness to pay 48, 67, 134, 533, 569 women’s empowerment 440 women’s labor force participation 366 work productivity 84, 413 work-life balance 370 working conditions 153, 269, 370 working-age generation 745

761