Digitalization and the Welfare State 9780192848369, 0192848364

This volume explores how digitalization--in different forms--affects the welfare state. Digitalization is likely to have

127 52 8MB

English Pages 433 Year 2022

Report DMCA / Copyright

DOWNLOAD PDF FILE

Recommend Papers

Digitalization and the Welfare State
 9780192848369, 0192848364

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Digitalization and the Welfare State

Digitalization and the Welfare State Edited by

MAR I U S R . BU SE M EY E R AC HI M K E M M E R L ING PAU L M A R X KE E S VAN K E R SBE RG EN

1

3 Great Clarendon Street, Oxford, OX2 6DP, United Kingdom Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries © Oxford University Press 2022 The moral rights of the authors have been asserted Impression: 1 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this work in any other form and you must impose this same condition on any acquirer Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America British Library Cataloguing in Publication Data Data available Library of Congress Control Number: 2021944086 ISBN 978–0–19–284836–9 DOI: 10.1093/oso/9780192848369.001.0001 Printed and bound by CPI Group (UK) Ltd, Croydon, CR0 4YY RYGER/Shutterstock.com. Links to third party websites are provided by Oxford in good faith and for information only. Oxford disclaims any responsibility for the materials contained in any third party website referenced in this work.

Preface This edited volume is the final outcome of a working process that started in early 2019. Since then, the world experienced a global pandemic of historic proportions. Our topic of digitalization and the challenges it poses for established welfare states and labor market institutions—even though temporarily less present in the media than before—has regained a new degree of urgency in the wake of the pandemic. This volume brings together internationally renowned scholars and experts on the welfare state to reflect on the implications of digitalization for social policymaking. We as editors have encouraged and pushed contributors to “think outside the box” in order to fully utilize the somewhat larger degrees of freedom that the publishing of an edited volume allows compared to standard journal articles. We are proud to say that the final contributions to this volume not only met our already high expectations, but clearly surpassed them. Hence, our first big “thank you” goes to the contributors of this volume who had to deal with our continuous requests to get things done within tight deadlines. The chapters were presented and discussed at two workshops that fortunately took place before the COVID-19 pandemic hit with full force. The first one took place in January 2019 in Bremen, followed by a second one in January 2020 in Konstanz. We are grateful to Frank Nullmeier and the University of Bremen for financial and organizational support in organizing the first workshop. We would also like to gratefully acknowledge funding support from the Excellence Cluster “The Politics of Inequality” (EXC 2035) for the second workshop as well as support for publication and language editing costs. Speaking of which, we would like to thank Carla Welch for excellent and highly professional language editing, provided in high quality and always on time. Special thanks also to Rohan Khan for excellent support as research assistant during the final (but crucial) stages of preparing the manuscript for submission to the publisher. Finally, we thank Dominic Byatt and Olivia Wells at Oxford University Press for their support during the whole project. Marius R. Busemeyer, Achim Kemmerling, Paul Marx, and Kees van Kersbergen Konstanz, Erfurt, Duisburg, and Aarhus, June 2021

Contents List of Figures List of Tables List of Contributors

1. Digitalization and the Welfare State: Introduction Marius R. Busemeyer, Achim Kemmerling, Paul Marx, and Kees van Kersbergen 2. Digitalization, Automation, and the Welfare State: What Do We (Not Yet) Know? Marius R. Busemeyer

ix xi xii

1

21

PART I . BROA DE R T R E N D S : I S T HI S TI ME DI FFE R E N T OR NOT ? 3. Digitalization and the Transition to Services Anne Wren 4. Welfare States, Labor Markets, Social Investment, and the Digital Transformation Werner Eichhorst, Anton Hemerijck, and Gemma Scalise 5. The Value and Future of Work in the Digital Economy Marius R. Busemeyer and Ulrich Glassmann 6. The Data Revolution and the Transformation of Social Protection Torben Iversen and Philipp Rehm 7. Social Solidarity in the Age of the Internet Paul Marx

41

64 83

99 118

PART II. T H E N EW NEW POL I T IC S OF TH E W ELFARE STATE IN T H E DIG I TA L AGE 8. Automation Risk, Social Policy Preferences, and Political Participation Thomas Kurer and Silja Ha¨usermann

139

contents

9. Gender, Technological Risk, and Political Preferences Jane Gingrich and Alexander Kuo

vii

157

10. The Case for a Basic Income in the Emergent Digitalized Economy Joe Chrisp and Luke Martinelli

174

11. Technology, Risk, and Support for Social Safety Nets. An Empirical Exploration Based on Italy Dario Guarascio and Stefano Sacchi

192

12. Tracing Fears About Digitalization and Automation in Social and Labor Market Policy Debates Achim Kemmerling and Stephanie Gast Zepeda

214

PA RT III. POLICIE S A N D P OL I T IC S : ADAPTAT ION, R E SI LIE NCE , V U L N E R ABI LI T I ES 13. Political and Institutional Limits to the Rise of Platform Work Georg Picot 14. Internet and Platform Work in Europe Jan Drahokoupil and Agnieszka Piasna

237 255

15. Digitalization and Automation: The Challenges for European Pension Policies David Natali and Michele Raitano

274

16. The Structural Adaptability of Bismarckian Social Insurance Systems in the Digital Age Frank Nullmeier

290

17. Transformation of Education Policy and Governance in the Digital Era Sigrid Hartong, Nelli Piattoeva, Antti Saari, and Glenn Savage

304

18. Digitalization and the Politics of Health Risks in Advanced Democracies Carsten Jensen and Kees van Kersbergen

319

19. Digitalization and the Effects of Internal and External Modernization in Health Care Systems Daniel Buhr and Rolf Frankenberger

336

viii

contents

20. The Politics of Tax Policy in the Digital Age Margarita Gelepithis

355

PART I V. C ONC LU SION S 21. Digitalization and the Future of the Democratic Welfare State Marius R. Busemeyer, Achim Kemmerling, Paul Marx, and Kees van Kersbergen Index

373

393

List of Figures 3.1. Sectoral composition of value added (gross) over time: UK, Germany, and Sweden (1970–2007)

48

3.2. Sectoral distribution of hours worked, in percent: UK, Germany, and Sweden (1970–2007)

50

3.3. Cross-sectoral distribution of employment, in percent, by skill level: UK (1970–2005)

52

3.4. Average sectoral compensation compared with economy-wide mean: UK (1970–2005)

54

3.5. Cross-sectoral distribution of employment, in percent, by skill level: Sweden (1970–2005)

56

3.6. Average sectoral compensation compared with economy-wide mean: Sweden (1970–2005)

57

3.7. Cross-sectoral distribution of employment by skill level: Germany (1970–2005)

58

3.8. Average sectoral compensation compared with economy-wide mean: Germany (1970–2005)

59

4.1. Comparative estimates of job automation risk in percent, 2013

65

4.2. Employment rate, equality, and welfare spending, 2016

68

5.1. Digitalization and the revalorization of self-determined forms of work

87

6.1. Predicted life insurance penetration

110

8.1. Distribution of automation risk

143

8.2. Share of the population that feels threatened by automation

144

8.3. Individual-level determinants of perceived automation risk

146

8.4. Individual-level determinants of perceived automation risk

147

8.5. Automation risk and social policy preferences

149

8.6. Subjective automation risk and policy preferences (position), by education

151

8.7. Subjective automation risk and political participation

152

9.1. Exposure across men and women

166

9.2. Redistributive preferences: Different measures of automation

168

9.3. Voting for right-wing populist parties: Different measures of automation

169

10.1. Probability of basic income support, by graduate status and RII

184

10.2. Probability of basic income support, by union membership status and RII

185

10.3. Probability of basic income support, by income and RII

186

x

list of figures

11.1. Objective (RTI) and subjective risk by occupation (1-digit, ISCO-08)

204

11.2. Support for social policy measures, by ESS Round

205

12.1. Aggregated word count for digitalization and automation in parliamentary debates

218

12.2. Phylogenetic trees showing the topics clustered hierarchically

220

12.3. Results of human coding

224

12.4. Partisan ideology and the “End of Work” idea

226

13.1. Platform work organized as genuine self-employment

240

13.2. Platform work organized as bogus self-employment

240

14.1. Conceptualization of Internet and platform work

257

14.2. Experience with platform work: ever performed

261

14.3. Frequent platform work: monthly and at least 50% of income

262

14.4. Internet work

263

14.5. Internet work, by type of activity

264

14.6. Annual income, by type of activity

265

14.7. Age of platform workers compared to other respondents

266

14.8. Gender composition of platform workers, by frequency of platform work

266

14.9. Gender composition of occasional Internet workers

267

14.10. Labor market status in the past 12 months

268

14.11. Main source of income, by type of employment contract in the past 12 months

269

19.1. e-Health in the EU

346

19.2. Internal and external modernization

350

20.1. Total tax revenue (2010–2018)

365

20.2. Composition of tax revenues, EU-27 plus UK

365

List of Tables 3.1. Contribution of ICT technology to the growth of value added: EU-15ex, annual average (1996–2005)

46

4.1. ICT usage indicators, 2019

69

4.2. Dualized labor markets and reform activity in Germany

74

5.1. Sources of employment growth in the digital(ized) economy

89

6.1. Information and social insurance

102

6.2. Life insurance penetration, information, and partisanship (ECM)

115

10.1. Probit regression results: main independent effects (Models 1 and 2) and interaction effects (Models 3–5)

183

11.1. ESS questions for UBI and GMI

197

11.2. Summary statistics, total sample of those in employment

203

11.3. Support for UBI, OLS estimation (clustered std errors at the 1-digit ISCO level) 206 11.4. Support for GMI, OLS estimation (clustered std errors at the 1-digit ISCO level)

207

12.1. Topic terms for selected topics in the Bundestag

229

12.2. Topic terms for selected topics in the House of Commons

230

12.3. Logit regressions for the “End of Work” idea

231

15.1. Potential challenges to pension systems

281

15.2. Possible public pension reform paths in the context of technological change

283

17.1. Contexts of datafication in policy in governance

307

18.1. Differences between labor market and health risks

321

19.1. Comparison of typologies of health care systems

341

List of Contributors Daniel Buhr is Head of the Steinbeis Transfercenter for Social and Technological Innovation and adjunct Professor for Policy Analysis and Political Economy at the Institute of Political Science, University of Tu¨bingen. His main research interests are economic and innovation policy, health, and care policy. Marius R. Busemeyer is Professor of Political Science and Speaker of the Excellence Cluster “The Politics of Inequality” at the University of Konstanz as well as Senior Research Fellow at the WSI (Du¨sseldorf). He recently published A Loud, But Noisy Signal? Public Opinion and Education Reform in Western Europe (Cambridge University Press) on the relative influence of public opinion on the politics of education. His work focuses on comparative political economy and welfare state research, theories of institutional change, and, more recently, analyses of public attitudes towards the welfare state. Joe Chrisp is a postdoctoral researcher at the Department of Comparative Politics at the University of Bergen, Norway. His research interests span various themes in public policy, comparative politics, and welfare states, and he is currently working on a project examining the growth in public consultations across EU member states. Joe completed his PhD on the political economy of basic income in Europe at the University of Bath, having studied at the University of Oxford and the University of Bristol. Jan Drahokoupil is a senior researcher at the European Trade Union Institute (ETUI) in Brussels where he coordinates research on digitalization and the future of work. His work in this context focuses in particular on internet and platform work. Werner Eichhorst is a Research Team Leader at the IZA Institute of Labor Economics (IZA), Bonn, and Honorary Professor of International and Comparative Labor Market Policy at Bremen University. His work focuses on comparative labor market research with a strong emphasis on institutions and the political economic of reforms. Rolf Frankenberger is a senior lecturer and researcher at the Institute of Political Science, University of Tu¨bingen. His research focuses on authoritarian regimes, welfare and capitalism, and political culture. Stephanie Gast Zepeda is a PhD candidate and research associate to the Gerhard Haniel Professor for Public Policy and International Development, Prof. Achim Kemmerling. She holds a BA in International Cultural and Business Studies from the University of Passau and an MA in International Political Economy from King’s College London. She is interested in political communication and computational social science. Margarita Gelepithis is a University Lecturer in Public Policy at the University of Cambridge. She has research expertise in the politics of economic and social policy, with a particular interest in policy responses to technological change and their distributional consequences.

list of contributors

xiii

Jane Gingrich is a Professor of Comparative Political Economy in the Department of Politics and International Relations (DPIR) and Tutorial Fellow in Politics at Magdalen College, University of Oxford. She is a fellow of the Canadian Institute for Advanced Research (CIFAR) Innovation Equity and the Future of Prosperity Program. Her interests lie in comparative political economy, the welfare state, and education. Ulrich Glassmann is professor of Comparative Institutional Analysis at the EuropaUniversity of Flensburg, Germany. After graduating he continued as a doctoral student at the Max Planck Institute for the Study of Societies, assistant professor at the University of Cologne and professor pro tempore at the University of Constance. His research is concerned with institutional regimes governing employment, education and innovation in regional economies. In particular, he focuses on the nexus between family networks and regional economic performance in Mediterranean countries. He has published his work in several edited volumes by Oxford University Press and journals such as Regional Studies, European Planning Studies, and Rationality & Society. Dario Guarascio is Assistant Professor in Economic Policy at Sapienza University of Rome. His main research interests are economics of innovation, labor economics, European economy, and industrial policy. Sigrid Hartong is professor for sociology at the Helmut-Schmidt-University in Hamburg. Her research focus is on the transformation of governance in education and society, with a particular emphasis on rising digitalization and datafication. She is founder of the initiative Unblack the Box (www.unblackthebox.org). Silja Ha¨usermann is Professor of Political Science at the University of Zurich. Her current research specializes in the fields of comparative welfare state research and comparative electoral research. She has been a Fellow at the Wissenschaftskolleg zu Berlin in 2018/2019 and directs the ERC-funded grant “welfarepriorities” (www.welfarepriorities.eu), which studies the transformation of distributive conflict in relation with the transformation of European mass politics. At the University of Zurich, she is the co-director of the University Research Priority Programme “Equality of Opportunities”. Homepage: www.siljahaeusermann.org. Anton Hemerijck (1958) is Professor of Political Science and Sociology at the European University Institute. Trained as an economist at Tilburg University in the Netherlands, he took his doctorate from Oxford University in 1992. Between 2009 and 2013, he was Dean of the Faculty of the Social Sciences at the Vrije Universiteit of Amsterdam. From 2013 and 2017, he was Centennial Professor of Social Policy at the London School of Economics and Political Science. He also directed the Scientific Council for Government Policy (WRR), the principal think tank in the Netherlands, while holding a professorship in Comparative European Social Policy at the Erasmus University Rotterdam from 2001 until 2009. Over the past decade he frequently served as an advisor on social policy, social investment and the welfare state for the European Commission. Key publications include Changing Welfare States (2013) and The Uses of Social Investment (2017), both published with Oxford University Press. In 2020, Anton Hemerijck won an ERC Advanced Grant for the research project WellSIRe (Wellbeing Returns on Social Investment Recalibration).

xiv

list of contributors

Torben Iversen is Harold Hitchings Burbank Professor of Political Economy in the Government Department at Harvard University. His research and teaching interests include comparative political economy, electoral politics, and applied formal theory. His most recent book with David Soskice is entitled Democracy and Prosperity: Reinventing Capitalism through a Turbulent Century (Princeton University Press 2019). He is currently completing a book with Philipp Rehm entitled Big Data and the Welfare State. How the Information Revolution Threatens Solidarity (Cambridge University Press, forthcoming). Carsten Jensen is Professor at Aarhus University, Denmark. His research interests lie within the broad field of comparative politics and political economy. His work has been published in the American Journal of Political Science, British Journal of Political Science, Comparative Political Studies, among others. He is the co-author (with Kees van Kersbergen) of The Politics of Inequality (Palgrave, 2017). Achim Kemmerling is Gerhard Haniel Professor for International Development and Public Policy and director of the Willy Brandt School of Public Policy, University of Erfurt. He holds a PhD from Freie Universita¨t in Berlin and has previously worked at Central European University Budapest, Jacobs University Bremen, and the Social Science Research Centre in Berlin. He works on the political economy of social, labour and tax policies, straddling both advanced economies and lower- and middle-income countries. Kees van Kersbergen is Professor of Comparative Politics at Aarhus University, Denmark. His research covers a wide range of topics and issues in comparative politics and political economy. He is the co-author (with Barbara Vis) of Comparative Welfare State Politics: Development, Opportunities, and Reform (Cambridge University Press, 2014), and (with Carsten Jensen) of The Politics of Inequality (Palgrave, 2017). Alexander Kuo is an Associate Professor in the Department of Politics and International Relations (DPIR) and Tutorial Fellow in Politics at Christ Church, University of Oxford. His research interests include comparative political economy and political behavior. Thomas Kurer is a Research Group Leader at the Cluster of Excellence “The Politics of Inequality” at the University of Konstanz. His research studies the politics of labor market inequality and occupational change. He is also part of the NORFACE Network project TECHNO on the political consequences of technological change. Homepage: www. thomaskurer.net. Luke Martinelli is a research fellow at the Institute for Policy Research (University of Bath). His research has involved questions around basic income and the political economy of the welfare state, including tax/benefit microsimulation and quantitative analysis of political survey data. Luke joined the IPR in May 2016 having recently completed a doctorate at the University of Bath’s Department of Social and Policy Science. He is a Fellow of the Higher Education Academy. Paul Marx is Professor of Socio-Economics and Political Science at the Institute for SocioEconomics, University of Duisburg-Essen. His research interests relate to comparative labor market analysis and political inequality. His work has been published in, among others, the British Journal of Sociology, the European Journal of Political Research, the European Sociological Review, Politics & Society, and the Journal of Politics.

list of contributors

xv

David Natali is Professor in EU and Comparative Politics at the S. Anna School of Advanced Studies in Pisa. His work focuses on the comparative politics of pension reform in OECD countries, employment policy and EU-level social policy. Frank Nullmeier is Professor of Political Science at the University of Bremen, Germany. He is head of the Department “Theoretical and Normative Foundations” at the SOCIUM Research Centre on Inequality and Social Policy and member of the Collaborative Research Centre “Global Dynamics of Social Policy”. Agnieszka Piasna is a Senior Researcher at the European Trade Union Institute (ETUI) in Brussels. Her research interests lie in areas of job quality, working time, gender equality, and labour market regulation. Her current work focuses on digital labour, and she coordinates research activities in the framework of the ETUI survey on Internet and Platform Work. She has a PhD in Sociology from the University of Cambridge. Nelli Piattoeva is an Associate Professor of education at Tampere University, Finland. Her research focuses on the role of quantification and knowledge-production in general in the governance of education systems in Russia and the post-Soviet space. Georg Picot is professor in comparative politics at the University of Bergen. His research interests lie in the fields of comparative welfare state research and comparative political economy. He is especially interested in precarious employment and regulation of wages. Michele Raitano is Professor of Economic Policy at the Department of Economics and Law of “Sapienza” University of Rome. His research interests include welfare state, pension systems, labor market, inequality and social mobility. As an expert of welfare policies, he is member of the independent experts’ network ESPN-European social policy network, financed by the European Commission. Philipp Rehm is Associate Professor of Political Science at Ohio State University. His work is located at the intersection of political economy and political behavior. In particular, he is interested in the causes and consequences of income dynamics (such as income loss, income volatility, and risk exposure). At the micro-level, his research explores how income dynamics shape individual preferences for redistribution, social policies, and parties. At the macro-level, his work analyzes the impact of labor market and income dynamics on polarization, electoral majorities, and coalitions underpinning social policy. Antti Saari is an associate professor in Tampere University Faculty of Education and Culture. His main research interests include history and philosophy of educational research and curriculum studies. Stefano Sacchi is Professor of Political Science at the Polytechnic University of Turin. His current research focuses on the political economy of welfare and labor policy, and on the socioeconomic and political impact of technological change. Glenn C. Savage is a policy sociologist with expertise in education reform, federalism, intergovernmental relations and global policy mobilities. He is an Associate Professor at the University of Western Australia in the School of Social Sciences and the Graduate School of Education

xvi

list of contributors

Gemma Scalise is assistant professor of Economic Sociology at the University of MilanBicocca. Her research interests include labour market and welfare regulation, comparative political economy and multi-level governance. Among her publications are The Political Economy of Policy Ideas and The European Strategy of Active Inclusion in Context (Palgrave Macmillan, 2020). Anne Wren is a political economist who has researched extensively on the political and distributional impact of de-industrialization, and the political economy of services-based societies. She obtained her PhD in political economy at Harvard University and has held positions at Stanford University and Trinity College Dublin. At Trinity College she directed a European Commission Murie Curie Excellence Team, researching “Political Responses to Economic Change: De-industrialization, Globalization, and Service Sector Development.” Her publications on this topic include The Political Economy of the Service Transition (ed., Oxford University Press 2013); and ‘Employment, Equality, and Budgetary Restraint: The Trilemma of the Service Economy’, World Politics, 50/4 (1998).

1 Digitalization and the Welfare State Introduction Marius R. Busemeyer, Achim Kemmerling, Paul Marx, and Kees van Kersbergen

“May you live in interesting times.” This, almost certainly apocryphal, Chinese malediction ironically expresses the wish for misfortune and hardship to befall us. However we measure and evaluate the challenges that rapid technological change poses to the labor market and the welfare state, it certainly isn’t far off the mark to say that there is broad consensus that we do, indeed, live in very interesting times. Arguably, the most important structural change of our times is that in today’s advanced economies, technological change in the fields of computing, communication technologies, artificial intelligence, and robotics is not just proceeding apace; it is advancing at an increasingly rapid rate. This volume starts from the conviction that such exponential change unsettles, and maybe even disrupts, the existing labor market and social policy arrangements that were developed in response to earlier socioeconomic transformations with the aim of protecting people against social risks and guaranteeing that various social needs are met. Rapid technological change (“digitalization”; for conceptual clarifications, see the following sections) not only reinforces some of the well-documented ongoing trends (e.g., labor market dualization, underemployment, precarious work), but also creates a new set of social risks and needs, for which the existing social policy architecture appears ill-suited. Digitalization is likely to have a lasting impact on work, welfare, education, and income distribution, radically transforming social risks and needs. Will the changing risks and needs also radically transform today’s labor market and social protection provisions and lead to a new welfare state paradigm? Of course, it is still impossible to say. Indeed, the purpose of this volume is to bring together a group of scholars to demonstrate various potential answers to this question. That said, what we do know from several decades of welfare state research is that technological change per se is not an automatic trigger of welfare state adjustment. There may be a functional need to adapt the labor market and social policies, but there is no guarantee that adjustment will happen. And if it does, it is likely to be shaped by prevailing political and institutional contexts (Hope and

Marius R. Busemeyer, et al., Digitalization and the Welfare State. In: Digitalization and the Welfare State. Edited by Marius R. Busemeyer, Achim Kemmerling, Paul Marx, and Kees van Kersbergen, Oxford University Press. © Oxford University Press (2022). DOI: 10.1093/oso/9780192848369.003.0001

2

marius r. busemeyer, et al.

Martelli 2019; Thelen 2019). In all probability, the prevailing and impending political struggles about the distribution and redistribution of the payoffs of the digital transformation of the world of work will have significant long-term consequences when it comes to the winners and losers (Frey 2019). There are several steps in the translation of socioeconomic problem pressures into policy output, each of which is conditioned and influenced by political processes. First, social needs and demands may or may not (easily) translate into collective societal demands. In fact, the social and political actors that once articulated such demands, such as unions for example, are themselves weakened by the impact of technological change, because the platform economy blurs the very categories (workers, employees, self-employed, contracted workers) on which collective organization has traditionally been based (Eichhorst et al. 2017), or because technological change fuels labor market polarization (Michaels et al. 2014). Further, even defining what counts as a societal “problem” in need of a political solution depends on the distribution of political power between the actors involved. Second, even if demands can be constructed socially and transformed into collective action, the political articulation of such demands remains a formidable hurdle, primarily but not exclusively, because political parties need to realign interests, reconstitute viable electoral and government coalitions in support of their new policy supply, and solve the distributional conflicts stemming from this. Lastly, even if political parties succeed in this, adequate social policy responses are not a given, because policy legacies and vested as well as new or emerging interests can slow down, impede, or circumvent policy reform and implementation. Budgetary or other economic constraints may further limit the scope for governmental responses. Therefore, the political and policy responses to rapid technological change are likely to be multifaceted, complex, and characterized by a number of political, fiscal, and economic trade-offs.

Research Questions, Goal, and Approach The overarching research question of this volume is therefore: to what extent and how will welfare states respond to the new challenges of digitalization? Analytically and in line with the previous considerations, this question has two constituent sub-questions. First, to what extent and under what conditions will digitalization trigger a (paradigmatic) reconfiguration of the welfare state’s policy space? This question addresses issues of changing individual policy preferences, policy positions (of interest groups, parties, and new actors), and the direction of actual and likely policy developments, including existing programs (e.g., social insurance, social investment, taxation) and new instruments (e.g., robot tax, lifelong learning provisions, revamped public employment schemes, universal basic income). Second, to what extent and under what conditions will digitalization generate

digitalization and the welfare state

3

a reconfiguration of the welfare state’s political space? This question pertains to changes in existing (voter) cleavage structures, the development of new cleavages, (party) political realignments and coalition potential, and the emergence of new and powerful actors, such as global tech companies with significant platform power (Culpepper and Thelen 2020). This volume has several goals: to identify the technological drivers of change and to describe the type of socioeconomic changes and conflicts of interests they generate; to map the extent to which and how such changes and conflicts challenge existing labor market and welfare state arrangements (protection and redistribution); to explain the (potential and real) political and policy responses to these challenges; and to understand the contours of future developments, as well as to reflect on whether these changes might lead to the emergence of new welfare state paradigms. The way we phrase our questions, i.e., in the future tense, reveals the forwardlooking, imaginative yet empirical perspective on the impact of digitalization on the welfare state we adopt. We do this, not because we think we can predict future developments (we obviously cannot), but rather because we believe that the rapid and accelerating pace of technological change has potentially radical ramifications for the welfare state, which demands such a new and forward-looking approach. Adopting a retrospective perspective to study the impact of technological change (i.e., by using data on past labor market transformations) is still important, but only provides a partial picture of the implications of fast-paced technological change. If we rely exclusively on the analysis of past data as most of the existing work has done, we could seriously underestimate the transformative impact of the digital revolution on the world of work and welfare. Thus, we think it is important to develop a theoretically informed empirical basis for social science and public debates about the long-term implications of the digital revolution for the welfare state. With our forward-looking empirical perspective, we propose to trace empirical implications and emerging contours of future policies and politics of the welfare state in the present. Concrete examples of empirical traces of future policies and politics are to be found, for instance, in citizens’ attitudinal patterns regarding the impact and desired policy consequences of technological change, collective actors’ positions regarding different reform options, and those actors’ interpretations of already ongoing trends in the labor market and the welfare state.

Optimists versus Pessimists: Is it Different this Time? The ongoing public and academic debates about the impact of technological change on work, welfare, and inequality center on two crucial questions: first, is this time really different or is the current wave of technological change just the same as previous waves? Second, are the long-term consequences of technological

4

marius r. busemeyer, et al.

change, on balance, positive for work, well-being, and welfare, or rather negative? To some extent, different contributions to the debate as well as the attitudes and positions of citizens and political actors themselves can be grouped into different camps representing different responses to these questions. For one, digitalization optimists see the opportunities rather than the problems. So far, history has falsified previous and similarly calamitous expectations about the impact of rapid technological, demographic, and social change on the welfare state (post-industrialization and globalization, for instance) and shown that such worries were conceivably mostly due to the limitations of our imagination. This time is no different, say the optimists. Digitalization, for instance the emergence of online platforms, helps to create new and uniquely flexible opportunities for work, especially in markets where individuals can monetize private assets (rooms, cars) with the help of digital technology and provide services at low prices (OECD 2017). The optimists believe that the digital transformation will lead to a more egalitarian labor market, because the platform economy breaks open “insider-dominated” markets and creates new ones (Eichhorst and Spermann 2016). The welfare state recalibration that this might require (e.g., in the fields of labor regulation, social security, interest representation, and mediation, and so on) is merely a temporary challenge. Governments are learning—some more quickly than others (OECD 2017)—how to adapt to digitalization and, more importantly, how to use the new abundance of information and social knowledge that is created and distributed by new technologies to the benefit of effective social policymaking. Digitalization may cause adaptive problems in the short term, but it creates the conditions for a better welfare state. Digitalization may very well upgrade the effectiveness of social service production and delivery and, by means of big data, it may provide governments with much-needed extended and improved social knowledge to serve as a rational basis for social policies (e.g., Baldwin-Philippi 2015; Gainous and Wagner 2014; Mayer-Scho¨nberger and Cukier 2013; McGinnis 2013). According to this optimistic perspective, it may take a few decades, but there is no reason, in principle, to expect the welfare state to be incapable of adapting to the new societal and political context of the digital age. Pessimists, in contrast, hold that this time is different and that the welfare state edifice (with some variation according to type) will have a hard time adapting to the challenges of digitalization. The major structural transformations of the past (e.g., deindustrialization) have already put established welfare states under a lot of pressure, but the latest technological changes pose an even more gargantuan task. To some extent, whether pessimism or optimism prevails depends on the time frame being referred to. Pessimists might admit that the long-term effects of technological change might turn out to be beneficial on average, but it is likely to be associated with a non-trivial period of adjustment and transition, in which the benefits and costs of digitalization are unequally distributed. Further, the digital revolution may make more information available more quickly, but

digitalization and the welfare state

5

more information does not imply the improved social knowledge that is useful for policymaking, according to the skeptics. In their view, digitalization is eroding the foundations of the welfare state, bypassing existing labor market regulation and industrial relations, undermining risk solidarity and social protection, and narrowing the tax base for the welfare state’s financial resources. “Platform workers,” for example, challenge traditional social insurance categories, tend to be unprotected, and can unfairly compete on price, both nationally and globally, with other types of workers with standard contracts (European Parliament 2017). Such platform workers will be more dependent on tax-financed social security, while the platforms themselves evade taxation because of their “reliance on intangible assets, massive use of data as a production factor, new business models, and the difficulty of determining the jurisdiction in which value creation occurs” (Eichhorst and Rinne 2017: 3). In addition, advancing technological change is expected to result in further labor market polarization and rising inequality (Michaels et al. 2014). The most pessimistic scenarios of automation and artificial intelligence advances envision the “End of Work” and unprecedented mass unemployment, while, at the same time, declaring solutions such as a universal basic income as economically unfeasible and politically unrealistic. Few, if any solutions can be seen on the pessimists’ horizon. From this perspective, there are no grounds to be confident that social policy change and innovation can expedite and catch up with digitalization. The position we take in this volume is that even if the optimists are right and we can look forward to the problem-solving capacity of current social policies being restored and its effectiveness and efficiency increased, this might take several decades during which time the unsettling forces of digitalization will be felt, potentially disrupting the welfare state’s core functions (e.g., social protection and social investment). For this reason, it is necessary to side with the pessimists and conduct novel empirical research on the (potential) adverse impact of digitalization on the welfare state and its prospective remedies. However, to come up with such remedies and solutions for the empirically identified problems, it makes perfect sense to listen carefully to the optimists who are pointing out the various options and strategies that are either already available or can be creatively envisaged as solutions to the welfare state’s current problems. We do not think that robots will completely take over existing jobs or that the rapidly developing platform economy will entirely obliterate the category of “worker” or “employee,” hence nullifying the whole edifice of social protection and services that is currently attached to labor market status and performance. However, the sheer pace and especially the accelerating rate of technological change are unprecedented and can be expected to cause extremely intricate and potentially explosive policy and political adjustment problems. In this sense, this time is most certainly different. What is more, this time may also be different because of certain new characteristics of the kind of technological change we are observing. For one,

6

marius r. busemeyer, et al.

whereas previous waves of technological change primarily promoted the automation of manual work, current and future advances in artificial intelligence and big data analysis will allow the automation of non-manual (white-collar) occupations and tasks that were previously relatively well protected. When automation anxiety spreads from the lower to the middle classes, this will likely (and partly already does) affect the political conflicts surrounding technological change (Kurer and Palier 2019). Moreover, as digitalization contributes to a further contraction of the employment share of the traditional manufacturing sector, new employment opportunities will emerge in the knowledge-driven service economy, in particular in occupations that require interpersonal, communicative, management, but also creative skills. The extent to which new kinds of personal and social services as well as new creative jobs in the digital economy (e.g., influencer, video game performer, or blogger) are in fact related to traditional notions of academic knowledge, is an issue for debate. Hence, the service part of the knowledge economy might—in the long term—be more important than the knowledge part, leading to a qualitative transformation of what is to be considered “gainful employment.” In yet another specific sense, however, this time may not be so different after all, because, in a way, history seems to be repeating itself. The Industrial Revolution dramatically increased prosperity for all eventually. But—as Frey et al. (2018) demonstrate—eventually is the operative word here, because it took no less than 40 years before the benefits of mechanization started to “trickle down” to ordinary workers. Similarly, it took 20 years before the benefits of automation, an ongoing process since 1980, began to have a positive impact on labor income in the American economy. The current phase of automation may perhaps give us, “good reasons to be optimistic about the long run,” writes Frey (2019: 365), because of the productivity and prosperity growth it promises to bring, but at the same time “such optimism is only possible if we successfully manage the short-term dynamics.”

Laying the Groundwork: Definitions and Research Questions To sum up the previous section, this Introduction puts forward two core theses: (1) This time is different in the sense that we expect the current wave of rapid technological change to have lasting and significant consequences for the future development of the welfare state, partly because of the scale and specific nature of the digital transformation, partly simply because each wave of technological change is different from the last. (2) Current day and near future distributive and redistributive politics (continue to) matter. Welfare state scholars have long been aware that politics matters with regard to welfare state policies and outcomes but given

digitalization and the welfare state

7

the rapid and potentially ground-breaking nature of technological and socioeconomic developments, we posit that politics matters now even more than it did in the past. As mentioned previously, we endorse a forward-looking empiricism that aims at identifying the traces of future developments in current day policies and politics. Different contributions to this volume adopt different approaches here, and certainly not all of the authors fully agree with our two core theses. The purpose of this volume, however, is not to settle the debate about the implications of digitalization for the welfare state, but rather to serve as a starting point for this debate, which we can expect to continue for some time to come. In the following section and the remainder of the Introduction, we lay the conceptual groundwork for the volume. First, we present and discuss working definitions of the different aspects and facets of technological change, which are often subsumed under the term “digitalization.” Second, we briefly review previous work on the impact of technology on work and welfare—a broader and more detailed review of the state of the art is presented in Chapter 2 of this volume. Third, we discuss implications of technological change for the political space and, fourth, its impact on the range of options available in welfare state policymaking. Fifth, we briefly address the role of the global pandemic in how digitalization will shape the welfare state. Lastly, we introduce the various contributions to this volume.

Key Dimensions of Technological Change There is no shortage of novel, some would say new-fangled terms to characterize the rapid transformation of the contemporary economy. We now have the digital economy (and digitalization), the gig economy, the platform economy (and platformization), the information economy, big data society (and datafication), the second Machine Age, the third Industrial Revolution, Industry 4.0, the robot economy (and robotization), surveillance capitalism, the attention economy, zero marginal cost society, post-capitalism, e-commerce, e-business, and the collaborative or sharing or circular economy, and this, in addition to—now familiar but not so long ago still new—terms, such as the knowledge economy, the network society, the service economy, and postindustrial society. Let us just say that this abundance of neologisms can be taken as a token of how interesting our times really are. Yet, interesting times can also be confusing, and we need conceptual clarification and specification of the relevant concepts that are used throughout this volume. Let us start with digitization, which is the process of converting analog data into digital form. The crucial quality is that digitized data can be easily, cheaply, and accurately stored, processed, transmitted, analyzed, manipulated,

8

marius r. busemeyer, et al.

interacted with, and replicated. The digital economy is the economic model where firms employ computer technology and digital infrastructures (say the Internet) to produce and exploit digitized data as a strategic resource, and profit from the positive network externalities and (next to) zero marginal costs that this offers. Digitalization refers to the dissemination of digital technologies, services, goods, techniques, and skills across economies; it is “the way in which many domains of social life are restructured around digital communication and media infrastructures” (Brennen and Kreiss 2014). The contemporary usage of the concept of digitalization also highlights that “the digital economy is becoming increasingly inseparable from the functioning of the economy as a whole” (UNCTAD 2019: 4). Eichhorst and Rinne (2017: 1) usefully distinguish two aspects of digitalization: automation, which is “the increasing use of robots, machines and algorithms in value chains, which is moreover no longer restricted to simple routine tasks”; and platform economy, which is “an entirely new business model that includes new real and virtual services and, importantly, online outsourcing.” The gig economy is a more encompassing term that refers to an economy in which contracted or independent work has come to play an increasingly important role. Its development is greatly stimulated by the platform economy, because it makes “gigs” much more easily accessible to a growing number of people. The final concept that we define is platform work, which is “an employment form in which organisations or individuals use an online platform to access other organisations or individuals to solve specific problems or to provide specific services in exchange for payment.” Platform work has the following features: “Paid work is organised through an online platform; Three parties are involved: the online platform, the client and the worker; The aim is to carry out specific tasks or solve specific problems; The work is outsourced or contracted out; Jobs are broken down into tasks; Services are provided on demand” (Eurofound 2018: 9).

Looking Back: The Impact of Technological Change on Work and Welfare So Far The latest round of technological change has many dimensions and its (potential) socioeconomic impact is complex. In the following, we can merely highlight some of the most important predictions currently found in the literature. Chapter 2 of this volume provides a more detailed review of the literature. The future of work is a recurring topic in human history (Mokyr et al. 2015), but technological innovation in both robots and algorithms has created new machines that look and act much more like human beings. This image has spawned both hopes and fears about employment in the public discourse and in the academic debate. Some studies highlight the risk of massive job loss as a result of automation (Frey and Osborne 2017). Others predict much less job destruction or even job

digitalization and the welfare state

9

growth (Acemoglu and Restrepo 2019; Arntz et al. 2018; Nedelkoska and Quintini 2018). The differences in these studies lie in how they operationalize automatable tasks, nested in jobs, occupations, and sectors (see Busemeyer (Chapter 2) in this volume). Here, in the Introduction, we simply seek to highlight the remarkable scientific uncertainty still shrouding political debates. There is greater consensus when it comes to the asymmetric effects of recent technological change. Recent scholarship confirms that technological change is one of the most important drivers of changes in labor market participation and—ultimately—inequality (e.g., Kristal and Cohen 2015). The technological race between education and technology (Goldin and Katz 2008) increases the demand for highly educated workers over time, while the forecasts for other workers are less optimistic. Famously, Autor et al. (2015) argue that technological change could lead to a hollowing out of the middle classes, because it will primarily affect routine tasks as opposed to non-routine tasks (Autor et al. 2003). The recent wave of technological change has allowed routine cognitive tasks to be automated, e.g., clerical tasks in accounting, legal, and administrative support services, affecting the employment prospects of typical middle-class occupations (see also Bu¨hrer and Hagist 2017: 115; OECD 2019). Eventually this would imply increasing (wage) inequality (but see Mishel et al. 2013). Much of this literature focusses on quantitative changes. However, there is an equally, if not even more fundamental qualitative transformation on the way: the death and birth of new types of work, occupations, and employment (Neufeind et al. 2018). From a regulatory perspective, this is a hard nut to crack: how to deal with mini-jobs and gig work is just one example. Another is the question of how the expansion of social services and creative occupations might fundamentally change perceptions of what is considered “gainful employment” (see Chapter 5 by Busemeyer and Glassmann in this volume). The impact of technological change on welfare is even harder to project, but a few trends are becoming increasingly obvious. Technological change will not only amplify (income or wealth) inequality through the labor market channel, but also through a reorganization of production. The idea of zero marginal costs means that once a company gets past the huge start-up costs (to become a predominant platform for social exchange or selling services, for instance), further growth comes at little additional cost. Facebook, Amazon, and Uber are all examples of this mechanism. Eventually, this implies much larger economies of scale in production and new levels of inequalities (Guellec and Paunov 2017; see also Kenny and Zysman in Neufeind et al. 2018, as well as Culepper and Thelen (2020) on platform power). Theoretically, one taxi company in Seattle could organize taxi services in all corners of the globe, which implies both domestic and global forms of inequality (Brynjolfsson and McAfee 2014). The platform economy makes such nonlinearities feasible. This would have far-reaching consequences for the organization of work and the distribution of welfare.

10

marius r. busemeyer, et al.

A more positive prediction is that lower production costs could lead to enormous productivity gains, improving services and making service delivery cheaper (see Wren (Chapter 3) in this volume). In medicine, doctors can use algorithms for diagnosis, and tax and welfare bureaucracies could see reductions in costs. Yet, it is also clear that technological change does not always and immediately translate into higher productivity and economic growth. A pertinent example is the introduction of personal computers in the 1980s (Manyika 2017), which sparked exaggerated hopes of short-run productivity gains. In fact, it took much longer to realize the potential of such innovations (Gordon 2016). In winner-take-all markets, technological change could also mean creating rents instead of productivity gains. In this sense, platforms such as Facebook merely outcompete other firms, rather than creating something new (Komlos 2016). In some instances, effects are more immediately visible. Nowadays, for instance, it is much easier to store big data on health and make it accessible to everyone. This also creates new challenges for regulating access to information and protecting privacy (see Iversen and Rehm (Chapter 6) in this volume). Technological change has made regulation more complicated, in part because the business model has changed. A large share of the productivity gains is not (directly) monetized, if, for instance, Google allows doctors or educators to use its search engine free of charge (Brynjolfsson and McAfee 2014). In the long term, however, the apparently free offer of services from Google and other leading tech companies might establish dependency relations in the sense that those who have got used to benefiting from digital services become dependent on their continued provision via centralized platforms (Culpepper and Thelen 2020), which might lead to significant power (and economic) imbalances in the long term. Further, digital platforms and algorithms capture information that Google, and others monetize by selling ads or information. This makes it hard to respect and ensure privacy and the functioning of insurance and other markets.

The Future Impact of Technological Change on the Political Space Rapid technological change is likely to reshape the political space in welfare states in the form of a shifting balance of power between established actors, the emergence and empowerment of new political actors and—potentially—the emergence and realignment of political cleavage structures. Again, the question in the background is whether “this time will be different.” We tend to believe that technological change will be rapid and significant. First of all, ex ante, it is an open question whether technological change leads to the mobilization of individual and collective actors or, in contrast, promotes political apathy and disengagement. Historically, of course, the negative side effects of the first Industrial Revolution triggered large-scale mobilization of the labor

digitalization and the welfare state

11

movement and thereby laid the groundwork for the modern welfare state. Today, we do not (yet) see similar movements opposing technological change, although this situation might change once the effects of automation and digitalization fully kick in on the labor market. In contrast, if technological change does indeed contribute to higher levels of inequality and labor market polarization, those on the losing end of this change might become more alienated from politics and disengage, as research on the depressing effect of inequality on participation has shown (Scha¨fer 2010; Solt 2010). In line with our forward-looking empirical perspective, we can already spot some traces of future development in current day politics. For instance, new research argues that the emergence of right-wing populist parties across many Organization for Economic Co-operation and Development (OECD) countries is related to real and perceived declines in status among the lower middle classes whose prospects of upward mobility are strongly conditioned by the forces of technological change (Anelli et al. 2019; Frey et al. 2018; Im et al. 2019; Kurer and Palier 2019). Empirically, it is challenging to disentangle the effect of technological change on support for right-wing populism from other confounders, in particular globalization. However, it may well turn out to be the case that the culprit for the rise of right-wing populism is not solely migration or economic globalization, but also technological change, and the way these processes reinforce each other. A related question is whether technological change triggers the emergence of a new political conflict line. Undoubtedly, the distributional patterns of the costs and benefits of globalization may not be that different from those observed in the case of technological change. Nevertheless there are some notable differences. For instance, manufacturing workers in the exposed, export-oriented sectors of the economy are among the winners of globalization, but still might fall victim to future rounds of automation. Vice versa, sociocultural professionals in the protected public sector might be under increasing pressure because of the constraints on public budgets related to globalization, but they are likely to be winners of technological change since their occupations are unlikely to become automated. Besides the threat of automation, the digital transformation is likely to create or at least reinforce the gap between those with easy access to digital technology and those who lack such access opportunities. This “digital divide” might be more pronounced outside the OECD world, but it certainly also matters in advanced economies. Further, as Culpepper and Thelen (2020) and Thelen (2018) have argued, a new conflict might emerge between the consumers and the producers of services offered on digital platforms. Consumers of services (Airbnb users, Uber customers, etc.) have a strong interest in keeping prices low and service quality high, whereas the “producers” of these services face precarious employment conditions and potentially ruinous price competition. Taken together, these examples demonstrate the potential for new political conflict lines that pit the losers and winners of technological change against each other.

12

marius r. busemeyer, et al.

For now, it is not yet clear whether any of these potential conflicts will become fully-fledged political cleavages, triggering the formation of new parties or new collective actors. The short-term rise and fall of the “Pirate Party” across many European countries might be seen as an indicator that this could indeed happen at some point in the future if established parties do not manage to own issues related to the digital transformation. Another indication of the mobilization potential of these issues were the large-scale protests against the enactment of the EU’s copyright reform bill which took place across Europe in the spring of 2019. Whether or not new political conflict lines emerge, consolidate, and become politically activated, however, is an open question and one certainly worthy of further empirical investigation. Lastly, and related to shifting power balances at the macro level, technological change is associated with the emergence of new actors in political arenas as well as the further weakening of traditional established collective actors, in particular trade unions. First and foremost, these new actors are the leading global tech companies (Apple, Facebook, Google/Alphabet, Microsoft, and Amazon). Due to their monopoly or platform power in the digital economy (Culpepper and Thelen 2020; Zuboff 2019), these actors occupy a historically unique position in the global economy. Even though monopolies also emerged during previous industrial revolutions, until now they have all been reined in by public regulation at some point. Whether the same thing will happen to the new tech giants is an open empirical and political question. In this sense, the digital revolution could indeed lead to a significant qualitative change in the relationship between the state and business. Further, the ability of global tech companies to establish monopoly-like positions on particular markets is intrinsically related to technology itself: the platform economy has an inherent “winnertake-all” dynamic as the value of the platform increases exponentially with its size. Second, technological change poses a significant challenge for the collective mobilization of workers’ interests, as the failed attempt to unionize an Amazon warehouse in Alabama, USA in April 2021 showed. Admittedly, new technology, in particular social media, might support mobilization efforts to some extent, but the new sectors of the digital economy and their new types of employment have fragmented employment relations and made it harder for trade unions to reach out to new workers. The ambiguities of platform workers’ employment status threaten to undermine existing labor market regulation and social insurance schemes. A further weakening of collective worker representation might result from the abovementioned emergence of the crosscutting cleavage between consumers and producers of services in the digital economy, in the worst case pitting different vulnerable groups against each other and thereby undermining the potential for unions to mobilize.

digitalization and the welfare state

13

Reshaping the Policy Space How will the digital transformation of work and welfare affect and reshape the policy space, i.e., the range of policy options and instruments that policymakers have at their disposal to respond to technological change? Will it simply reinforce the usage of existing welfare state programs and policies; or will it lead to the creation of entirely new instruments, potentially constituting a new welfare state paradigm? We believe that the answer is: both. Undoubtedly, some issues related to technological change, e.g., increasing job market polarization or large-scale layoffs due to automation, can and will be tackled by means of the traditional welfare state arrangements. At the same time, however, new kinds of social risks are likely to emerge, e.g., social exclusion due to digital divides or precariousness caused by new employment types. Traditionally, the welfare state has served three main functions: to insure against particular kinds of social risk (social insurance), to equalize and mitigate inequalities between the rich and the poor (redistribution), and to promote the formation and mobilization of human capital for the labor market (social investment). Of course, the weight a particular welfare state gives to each function, how it balances them against each other, and how exactly they are implemented in a particular setting are a matter of political contestation and depend on historical context. By and large, however, the following is a plausible working hypothesis. In the short run, new types of employment and job loss due to automation will undermine the basic logic of insurance-based social security systems, but at the same time will reinforce the demand for adequate social protection against the financial consequences of under- and unemployment. In the medium term, the digital transformation will increase the need for more social investment, while the redistributive pillar will become more contested due to the changing political space outlined in the previous section. These pressures on the three pillars of the welfare state will potentially boost the interest in and public support for a universalist safety net that is independent of employment status and income. Because such a scheme would, in reality, be strongly redistributive, it resembles the redistribution function of the welfare state. Whether the introduction of a universal basic income is politically feasible and/or desirable is a hotly debated topic (see Chrisp and Martinelli (Chapter 10) and Guarascio and Sacchi (Chapter 11) in this volume). Politically, it has become intertwined with debates about the potential negative side effects of the digital transformation, as even leading figures of the tech world (most prominently Mark Zuckerberg from Facebook) have expressed their support. If implemented, the large-scale introduction of universal basic income schemes would indeed represent a paradigmatic change of what the welfare state is about and what it is supposed to do. For the moment, there is no evidence of the political momentum

14

marius r. busemeyer, et al.

that would be necessary for such a change (e.g., Vlandas 2020). Hence, the most likely development trend in the medium term is that digitalization will reinforce and provide support for the transformation of welfare states toward universalist arrangements, loosening the connection between employment status and status in the social insurance system.

Excursus: COVID-19, Digitalization, and the Welfare State During the process of editing this book, we were (academically and privately) confronted with a major shock that brought the link between digitalization and social policy into the spotlight: the COVID-19 pandemic. At the time of writing, the crisis is ongoing, and it is impossible to tell what its outcomes will be. That said, we believe that it has the potential to accelerate some of the effects of digitalization on the welfare state. In other words, it will, by and large, enhance and accelerate some of the dimensions we have discussed so far. Because the pandemic has intensified pre-existing problems and pressures and increased public attention, it also is likely to serve as an impetus for policy innovation. Already with the global financial crisis, governments broke with the (neo)liberal zeitgeist of the 1980s and 1990s and became more interventionist. COVID-19 has certainly pushed the envelope for further interventions, especially in the OECD world: nation-states have dramatically expanded the scope of fiscal policy and regulatory competences, often at the expense of international regimes and supranational organizations. Governments have felt compelled to react to several immediate consequences of the crisis: the enormous economic downturn due to a shock of production systems; the concomitant loss of employment opportunities; as well as the huge asymmetries the shock has created. For instance, while some economic sectors such as physical retail or tourism have taken enormous hits, parts of the health care and pharma industries as well as online sales have seen tremendous growth rates as the pandemic unfolds. For the political space, this implies that digital divides between voters will matter even more, but they will also become more complex. The COVID-19 crisis has pooled occupations and sectors in three categories. First, there are those who can easily work from home without much loss of productivity or employment opportunity (Dingel and Neimann 2020). These are often the higher skilled and better paid. Those occupations and sectors that depend on physical contact fall into two further subcategories. The first is sectors in which physical contact is either relatively minimal (often manufacturing) or essential (e.g., retail in food and health care)—these will remain relatively unaffected. However, there is a second subcategory: those occupations for which physical contact is important, but the services offered are deemed non-essential (e.g., entertainment, gastronomy, and so on). Economic activity in this category has plummeted, especially where governments resorted to lockdown measures.

digitalization and the welfare state

15

COVID-19 will also accelerate ongoing technological change for the policy space. It is quite evident, for instance, that governments have ramped up their use of various aspects of employment policy, from furlough measures to cash benefits for the self-employed (Eichhorst et al. 2020). The pandemic has also thrown into sharp relief the ongoing technological disparities between OECD countries. For instance, the crisis so far has revealed enormous differences between health care systems that have already undergone digital modernization and those that mainly react to digitalization (see Buhr and Frankenberger (Chapter 19) in this volume). Similar observations could be made about how and the degree to which education systems have used digital tools in the crisis (see Hartong et al. (Chapter 17) in this volume). For these reasons, we believe that, if anything, the pandemic serves to highlight and accelerate certain aspects of the fundamental diagnosis we have sketched out in this introduction and which we will further corroborate in the following chapters.

Preview of the Book’s Structure and Chapters In the following, we provide a short preview of the chapters included in this volume. The structure of the volume largely follows the outline set out in this introductory chapter. The first part of the volume discusses general trends in the development of labor markets and welfare states. Chapter 2 by Busemeyer comprises a detailed reflection on the state of the art in research on the association between technological change and the welfare state. The chapter discusses the most important scholarly contributions in this area, but also highlights the research gaps that this volume seeks to address. Chapter 3 by Wren provides a long-term historical perspective on how technological change has affected labor markets and the productivity of the services sector. Chapter 4 by Eichhorst, Hemerijck, and Scalise focuses on the contemporary period, reflecting on the rise of the social investment model as a potentially new paradigm of welfare state policymaking and how this has helped countries engaging with technological change. In Chapter 5, Busemeyer and Glassmann adopt a qualitative and normative perspective by discussing how digitalization has affected and will influence the nature of employment and work, emphasizing that taking a purely quantitative perspective on how many jobs are destroyed or created may impede a deeper reflection on this issue. Iversen and Rehm (Chapter 6) adopt a politico-economic perspective on risk in insurance markets. They analyze how new technologies such as trackers can undermine political coalitions for public insurance, e.g., in health services by allowing individualized risk calculations. In a similar vein, Marx (Chapter 7) discusses how digitalization, in particular the rise of social media, changes the social fabric of societies by promoting status competition rather than social solidarity, a development which has potentially detrimental effects on the fundamental social pillars supporting the welfare state.

16

marius r. busemeyer, et al.

The second part of the volume centers on the impact of technological change on the political space. Kurer and Ha¨usermann (Chapter 8) explore the level of individual attitudes and preferences on social policy, showing that individual automation risk is not associated with increased demand for social investment policies, but rather with demand for redistribution, even though the former is often recommended by experts and pundits. Chapter 9 by Gingrich and Kuo adds a gender perspective to the discussion, arguing that men tend to be more affected by increasing automation risks than women, mostly because female employment tends to already be more precarious. This has important implications for the politics of welfare state adjustment, since the losers of digitalization are more likely to be men than women. Moving from political participation to policy preferences, two chapters by Chrisp and Martinelli (Chapter 10) and Guarascio and Sacchi (Chapter 11) discuss in detail the politics and policy logic of universal basic income, combining empirical analysis of attitudinal data with normative reflections on the potential of a universal basic income to serve as a solution to the challenges of the digital knowledge economy. While Martinelli and Chrisp conduct a comparative analysis of several OECD countries, Guarascio and Sacchi focus on the case study of the recent mobilization for a minimum income scheme in Italy. Chapter 12 by Kemmerling and Gast Zepeda moves from individual social policy attitudes to political discourses at the elite level. Analyzing parliamentary plenary debates, the authors show that countries differ significantly in when and how parliamentarians discuss digitalization. They also show that fears about job loss are more relevant at the two extremes of the ideological spectrum. The third part of the volume focuses on the implications of digitalization across a range of policy areas. To start, two chapters discuss the implications of the platform economy. Picot (Chapter 13) offers a conceptual and theoretical analysis of the inherent growth limitations of the platform economy, providing an important countervailing perspective to potentially exaggerated claims about the destructive potential of this sector. Complementing this theoretical perspective, Drahokoupil and Piasna (Chapter 14) provide an empirical overview of the size and economic importance of the platform economy across a range of European countries, as well as a critical discussion of the problems and challenges of measuring the size and impact of this economic sector. The two subsequent chapters focus on the consequences of technological change for pension policy—often considered the linchpin of the social insurance state. Natali and Raitano (Chapter 15) adopt a broadly comparative perspective, arguing that digitalization has different effects on Bismarckian systems than on Beveridgean systems, with Bismarckian systems being likely to face greater pressure to adapt. Nullmeier’s chapter (Chapter 16) provides a complementary perspective, by honing in on Bismarckian welfare states, i.e. Germany and Austria in particular. He, too, sees great challenges ahead, but argues that Bismarckian systems have at their disposal the institutional

digitalization and the welfare state

17

prerequisites to adjust to the digitalized knowledge economy. The chapter by Hartong, Piattoeva, Saari, and Savage on education policy (Chapter 17) argues that this domain has also undergone significant transformation. Global tech companies as well as EdTech start-ups have become more involved in providing the digital infrastructure that education systems rely on, e.g., to measure student and teacher performance and to provide accountability to stakeholders. This policy change fundamentally transforms power relations between actors, for instance between public school administration and private firms. The next two chapters tackle health policies. The chapter by Jensen and van Kersbergen (Chapter 18) emphasizes how technological change promotes significant advances in the provision of care, but also contributes enormously to cost pressure, which, in turn, may crowd out investments in other areas of the welfare state. Buhr and Frankenberger (Chapter 19) provide a complementary perspective, showing how health care systems fall into two groups: those who successfully modernize internally by using digitalization to improve health services, and those for who digitalization is an external source of pressure. Last but not least, Gelipithis (Chapter 20) discusses how the advent of the digital knowledge economy is affecting the realm of tax policy. Issues in this domain include the question of how and where global tech companies (in particular platform companies) are to be taxed and whether new types of taxes should be levied on the winners of the technological transformation (e.g., a robot tax). This chapter gives valuable insights into how the changing political-economic environment will affect the welfare state’s overall capacity to perform its insurance, investment, and redistribution functions. In the final (Chapter 21), we bring together and discuss the variegated and multifaceted responses to the overarching research question of this volume: to what extent and how will welfare states respond to the new challenges of digitalization? To provide just a short preview of the conclusions of this volume, we see significant support for the two core theses mentioned previously. Even though many of the dramatic predictions found in the airport bestsellers are most certainly overblown, we contend that the current wave of technological change is likely to fuel transformative changes in the welfare states of advanced democracies, in terms of both the political space and the policy space. In the short term, technological change may add tinder to the fire of populism, in the mid to long term, it is likely to rebalance the power relationships between established actors and might lead to new cleavages developing between the consumers and producers of digital services. Further, taking another angle, the future politics of the welfare state is likely to be a continuation of the past: redistributive struggles about the payoffs of societal and economic transformations are likely to remain central in political debates, and it is the outcomes of these struggles that will determine who the winners and losers of technological change are and just how big the differences in their life chances will be.

18

marius r. busemeyer, et al.

References Acemoglu, Daron, and Pascual Restrepo. 2019. “Automation and new tasks: How technology displaces and reinstates labor.” Journal of Economic Perspectives 33(2):3–30. Anelli, Massimo, Italo Colantone, and Piero Stanig. 2019. “We Were the Robots: Automation and Voting Behavior in Western Europe.” IZA. Institute for Labor Economics Discussion Paper 12485. Arntz, Melanie, Terry Gregory, and Ulrich Zierahn. 2018. Digitalisierung und die Zukunft der Arbeit: Makroo¨konomische Auswirkungen auf Bescha¨ftigung, Arbeitslosigkeit und Lo¨hne von Morgen. Mannheim: Zentrum fu¨r Europa¨ische Wirtschaftsforschung. Autor, David H., David Dorn, and Gordon H. Hanson. 2015. “Untangling trade and technology: Evidence from local labour markets.” The Economic Journal 125(584):621–46. Autor, David H., Frank Levy, and Richard J. Murnane. 2003. “The skill content of recent technological change: An empirical exploration.” The Quarterly Journal of Economics 118(4):1279–333. Baldwin-Philippi, Jessica. 2015. Using Technology, Building Democracy. Digital Campaigning and the Construction of Citizenship. Oxford: Oxford University Press. Brennen, Scott, and Daniel Kreiss. 2014. “Digitalization and digitization.” Culture Digitally. Retrieved April 15, 2021 (https://culturedigitally.org/2014/09/digitalizationand-digitization/). Brynjolfsson, Erik, and Andrew McAfee. 2014. The Second Machine Age: Work, Progress and Prosperity in a Time of Brilliant Technologies. New York: W.W. Norton and Company. Bu¨hrer, Christian, and Christian Hagist. 2017. “The effect of digitalization on the labor market.” Pp. 115–137 in The Palgrave Handbook of Managing Continuous Business Transformation, edited by H. Ellermann, P. Kreutter, and W. Messner. London: Palgrave Macmillan. Culpepper, Pepper, and Kathleen Thelen. 2020. “Are we all Amazon Primed? Consumers and the politics of platform power.” Comparative Political Studies 53(2):288– 318. Dingel, Jonathan I., and Brent Neiman. 2020. “How many jobs can be done at home?” Journal of Public Economics 189:104235. Eichhorst, Werner, Holger Hinte, Ulf Rinne, and Verena Tobsch. 2017. “How big is the gig? Assessing the preliminary evidence on the effects of digitalization on the labor market.” management revue 28(3):298–318. Eichhorst, Werner, and Ulf Rinne. 2017. “Digital Challenges for the Welfare State.” IZA. Institute for Labor Economics Policy Paper 134. Eichhorst, Werner, Ulf Rinne, Paul Marx, René Bo¨heim, Thomas Leoni, Pierre Cahuc, Tommaso Colussi, Egbert L. W. Jongen, Paul Verstraten, Priscila Ferreira, João Cerejeira, Miguel Portela, Raul Ramos, Martin Kahanec, Monika Martiskova, Lena Hensvik, Oskar Nordstro¨m Skans, Patrick Arni, Rui Costa, Stephen Machin, and Susan N. Houseman. 2020. “Short-Run Labor Market Impacts of COVID-19: Initial Policy Measures and Beyond.” IZA. Institute for Labor Economics Research Report 98. Eichhorst, Werner, and Alexander Spermann. 2016. “Sharing Economy: Mehr Chancen als Risiken?” Wirtschaftsdienst 96(6):433–9.

digitalization and the welfare state

19

Eurofound. 2018. Employment and Working Conditions of Selected Types of Platform Work. Luxembourg: Publications Office of the European Union. European Parliament. 2017. The Social Protection of Workers in the Platform Economy. Brussels: European Union. Frey, Carl Benedikt. 2019. The Technology Trap: Capital, Labor, and Power in the Age of Automation. Princeton: Princeton University Press Frey, Carl Benedikt, Thor Berger, and Chinchih Chen. 2018. “Political machinery: Did robots swing the 2016 US Presidential Election?” Oxford Review of Economic Policy 34(3):418–42. Frey, Carl Benedikt, and Michael A. Osborne. 2017. “The future of employment: How susceptible are jobs to computerisation?” Technological forecasting and social change 114(January):254–80. Gainous, Jason, and Kevin M. Wagner. 2014. Tweeting to Power. The Social Media Revolution in American Politics. Oxford: Oxford University Press. Goldin, Claudia, and Lawrence F. Katz. 2008. The Race between Education and Technology. Cambridge: Belknap Press. Gordon, Robert J. 2016. The Rise and Fall of American Growth: The U.S. Standard of Living since the Civil War. Princeton: Princeton University Press. Guellec, Dominique, and Caroline Paunov. 2017. “Digital Innovation and the Distribution of Income.” National Bureau of Economic Research (NBER) Working Paper 23987. Hope, David, and Angelo Martelli. 2019. “The transition to the knowledge economy, labor market institutions, and income inequality in advanced democracies.” World Politics 71(2):236–88. Im, Zhen J., Nonna Mayer, Bruno Palier, and Jan Rovny. 2019. “The ‘losers of automation’: A reservoir of votes for the radical right?” Research & Politics 6(1): 1–7. Komlos, John. 2016. “Has creative destruction become more destructive?” The B.E. Journal of Economic Analysis & Policy 16(4):1–12. Kristal, Tali, and Yinon Cohen. 2015. “What do computers really do? Computerization, fading pay-setting institutions and rising wage inequality.” Research in Social Stratification and Mobility 42:33–47. Kurer, Thomas, and Bruno Palier. 2019. “Shrinking and shouting: The political revolt of the declining middle in times of employment polarization.” Research and Politics 6(1):1–6. Manyika, James. 2017. “Technology, jobs, and the future of work.” McKinsey. Retrieved April 15, 2021 (https://www.mckinsey.com/featured-insights/employmentand-growth/technology-jobs-and-the-future-of-work). Mayer-Scho¨nberger, Viktor, and Kenneth Cukier. 2013. Big Data. A Revolution that Will Transform How We Live, Work, and Think. London: John Murray. McGinnis, John O. 2013. Accelerating Democracy: Transforming Governance through Technology. Princeton: Princeton University Press. Michaels, Guy, Ashwini Natraj, and John Van Reenen. 2014. “Has ICT polarized skill demand? Evidence from eleven countries over twenty-five years.” Review of Economics and Statistics 96(1):60–77. Mishel, Lawrence, John Schmitt, and Heidi Shierholz. 2013. “Assessing the Job Polarization Explanation of Growing Wage Inequality.” EPI. Economic Policy Institute Working Paper 295.

20

marius r. busemeyer, et al.

Mokyr, Joel, Chris Vickers, and Nicolas L. Ziebarth. 2015. “The history of technological anxiety and the future of economic growth: Is this time different?” Journal of Economic Perspectives 29(3):31–50. Nedelkoska, Ljubica and Glenda Quintini. 2018. “Automation, Skills Use and Training.” OECD Social, Employment and Migration Working Papers 202. Neufeind, Max, Jacqueline O’Reily, and Florian Ranft. (eds.) 2018. Work in The Digital Age. Challenges of the Fourth Industrial Revolution. Lanham: Rowman & Littlefield. OECD. 2017. OECD Digital Economy Outlook 2017. Paris: OECD Publishing. OECD. 2019. Under Pressure: The Squeezed Middle Class. Paris: OECD Publishing. Scha¨fer, Armin. 2010. “Die Folgen sozialer Ungleichheit fu¨r die Demokratie in Westeuropa.” Zeitschrift fu¨r vergleichende Politikwissenschaft 4(1):131–56. Solt, Frederick. 2010. “Does economic inequality depress electoral participation? Testing the Schattschneider hypothesis.” Political Behavior 32:285–301. Thelen, Kathleen. 2018. “Regulating Uber: The politics of the platform economy in Europe and the United States.” Perspectives on Politics 16(4):938–53. Thelen, Kathleen. 2019. “Transitions to the knowledge economy in Germany, Sweden and the Netherlands.” Comparative Politics 51(2):295–315. UNCTAD. 2019. Digital Economy Report 2019. Value Creation and Capture: Implications for Developing Countries. Geneva: United Nations. Vlandas, Tim. 2020. “The Political Economy of Individual-level Support for the Basic Income in Europe.” Journal of European Social Policy 31(1):62–77. Zuboff, Shoshana. 2019. The Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power. London: Profile Books.

2 Digitalization, Automation, and the Welfare State What Do We (Not Yet) Know? Marius R. Busemeyer

Introduction Technological change has been a driver of welfare state development (and inequality) for a long time. Nevertheless, as discussed in the Introduction to this volume, there may be good reasons for why the recent wave of technological change could signal a far-reaching transformation of labor markets and, ultimately, welfare state institutions in the coming years and decades. One of these reasons is that technological progress is advancing much faster than before and on several different dimensions at the same time, triggering synergies and complementarities, which in turn, promote further technological progress. Even scholars and experts underestimate the pace of change. For instance, in a seminal contribution to the literature on the labor market impact of technological change, Autor et al. (2003: 1283) refer to the navigation of a car through city traffic as an example of a task that would be very unlikely to fall prey to the forces of automation because it is too complex and difficult to routinize. A mere six years later, in 2009, Google launched a project to develop self-driving cars (now an independent company called “Waymo”) and by January 2020, Waymo had announced that its driverless cars had clocked 20 million miles on public roads.1 These quite spectacular technological advances have inspired a number of scholars and pundits to claim the advent of a new era in capitalism (Brynjolfsson and McAfee 2014; Mason 2015; Rifkin 2014; Schwab 2016). The findings of the more empirical and scholarly literature on the implications of these technological changes for labor markets and the welfare state are mixed, as I will show in greater detail in this short literature review. Optimists expect that technological change will promote growth and new employment opportunities (Acemoglu and 1 https://arstechnica.com/cars/2020/01/waymo-is-way-way-ahead-on-testing-miles-that-mightnot-be-a-good-thing/ (retrieved May 14, 2021).

Marius R. Busemeyer, Digitalization, Automation, and the Welfare State. In: Digitalization and the Welfare State. Edited by Marius R. Busemeyer, Achim Kemmerling, Paul Marx, and Kees van Kersbergen, Oxford University Press. © Oxford University Press (2022). DOI: 10.1093/oso/9780192848369.003.0002

22

marius r. busemeyer

Restrepo 2019; Brynjolfsson and McAfee 2014). Other voices, however, raise concerns about the negative side effects of rapid technological change, such as mass unemployment (Ford 2016), itself associated with collapsing economic demand and a fundamental crisis of democratic capitalism (Collins 2013). Mokyr et al. (2015) as well as Gordon (2016) highlight the historical dimension by pointing out that previous waves of technological change did not lead to mass unemployment, despite similar initial fears. Historical experience also suggests that societies respond differently to technological change, either actively promoting technological innovation or deliberately slowing it down. Or put differently, the pace of technological innovation is not a completely exogenous factor, but also very much depends on the presence or absence of political forces supporting or opposing it (Frey 2019). The purpose of the following literature review is to provide some background as well as a common reference point for the remaining chapters in this volume. I do not claim to cover each and every piece of scholarship that has been published on the subject, which would be nigh on impossible in this highly dynamic and growing field of research. I will, however, point out the main areas of academic research as well as, more importantly, the blind spots that we hope to address with this edited volume. Very simply put, quite a lot of research has been done on the labor market implications of past and future waves of technological change, but so far, there is very little scholarship on the impact of digitalization and automation on welfare state policymaking or the politics underpinning welfare state change. In the following sections, I will first review existing research on the labor market implications of technological change (mostly from the field of labor market economics), before introducing in greater detail the limited number of contributions in the field of comparative welfare state research. I close by discussing how the contributions to this edited volume address the existing blind spots in the research.

Technological Change and Labor Markets Much of the research on the labor market implications of the recent wave of rapid technological change builds on and is inspired by previous work on “skill-biased technological change (SBTC)” (Acemoglu 2002). According to this argument, overall, technological change in recent decades has tended to privilege highly educated workers as demand for their services increased in line with developing technology, whereas low-skilled work is being replaced by machines or offshored to other countries. In line with this idea, technological change has been identified as an important driving force of increasing inequality and labor market polarization (Kristal and Cohen 2016). A crucial factor influencing the association between technological change and inequality is whether the education system continues to churn out a sufficient number of highly educated graduates to be able

digitalization, automation, and the welfare state

23

to counteract the unequalizing impact of technological change (Goldin and Katz 2008), which may become increasingly difficult to achieve as levels of educational attainment approach an upper limit. An important amendment and further extension of this basic argument has been made by Autor et al. (2003) who point out that, in fact, technological change is biased toward eliminating routine tasks rather than low-skilled tasks per se. For non-routine tasks that involve high cognitive and analytical as well as social (interactive) skills, the probability of these being taken over by robots or advanced software algorithms is lower than for routine tasks. From the perspective of routine-biased technological change (RBTC), the recent wave of technological change might also signal a fundamental transformation of labor markets: whereas previous waves of technological change mostly eliminated manual routine tasks, the more recent wave promises (or threatens?) to automate routine cognitive (i.e., white collar) tasks, e.g., clerical tasks in accounting, legal services, and administration, with important implications for employment prospects in middle-class occupations (Bu¨hrer and Hagist 2017: 115). Thus, the empirical implications of the RBTC argument differ slightly from the more traditional SBTC view. According to the latter, technological change simply leads to more inequality between the high- and low-skilled. In contrast, the RBTC view tends to expect a pattern of labor market polarization because the risk of automation is not evenly spread across the skills and income scales. Occupations involving complex analytical and interactive tasks, such as consulting, teaching, research, and management jobs, are concentrated at the high end of the skills/income scale. They are likely to be resilient to technological change or even grow proportionally in the knowledge-driven economy of the future. However, some occupations at the low end of the skills/income scale are equally resilient to the forces of technological change because they require the mastering of interactive skills or manual skills that are easy to perform for a human, but (still) difficult for robots (e.g., service sector jobs such as cleaning, waitressing, and hairdressing) (Goos and Manning 2007). There might also be societal resistance against the automation of certain service sector jobs for the foreseeable future, even though such automation would be technologically feasible, e.g., in the social care sector. Such resistance might only be felt as long as wages for these jobs stay relatively low, however. Given this polarization of employment prospects, middle-class jobs involving cognitive routine tasks are under particular pressure (OECD 2019). Some of these occupations will be upgraded to include more complex tasks, forcing workers to adjust and upgrade their qualifications accordingly. Others will be downgraded or eliminated, potentially resulting in lower wages, underqualified employment, or even unemployment. As will be discussed subsequently in greater detail, the fading job prospects of the (lower) middle class seem to be an important driving

24

marius r. busemeyer

force behind the rise of populism across many Organization for Economic Cooperation and Development (OECD) countries (Anelli et al. 2019; Frey et al. 2018; Kurer 2020; Kurer and Palier 2019). Many studies have confirmed the increasing polarization of employment opportunities—a “hollowing out” of the middle—which can eventually contribute to higher levels of wage inequality (Acemoglu et al. 2021; Autor 2015; Autor et al. 2015; Autor and Dorn 2013; Breemersch et al. 2017; Goos et al. 2014; Goos and Manning 2007; Graetz and Michaels 2018; Michaels et al. 2014). Others have pointed out important cross-national differences in the extent to which the transition toward the knowledge economy has fueled inequality in relation to national institutions, such as collective wage bargaining, strong unions, and employment protection legislation (Hope and Martelli 2019; Iversen and Soskice 2019; Thelen 2018, 2019). One drawback of these studies is that, since they work with historical labor market data, they can only provide limited insights on the impact of future waves of technological change. It may well be the case that the future implications of the digital transformation are similar to those of past waves (Mokyr et al. 2015); however, if the digital transformation of labor markets is, as some would argue, qualitatively different from previous waves, then the implications of this kind of research would be more limited. In light of this, a different set of studies adopts a “forward-looking” perspective, and we take the same approach in this volume (see Introduction). The pioneering study by Frey and Osborne (2017), for instance, derived estimates on the future automation potential of different occupations based on experts’ assessments regarding the most likely future development of certain technologies that could replace manual labor across a broad range of tasks. Most strikingly, the study posits that 47 percent of total employment in the US is in occupations at high risk of automation (probability of over 70 percent) in the coming years. Frey and Osborne’s approach has been reproduced and applied to other countries (see, e.g., Bonin et al. (2015) for Germany, Berriman and Hawksworth (2017) for the UK and Japan, Nedelkoska and Quintini (2018) and Arntz et al. (2016) for OECD countries). Most of these studies produce lower estimates regarding the automation risk for employment. For instance, Nedelkoska and Quintini (2018) posit that 14 percent of employment across OECD countries is at very high risk of automation (more than 70 percent) in the coming years and 32 percent of employment is at high risk of being automated (between 50 and 70 percent automation probability). Without a doubt, these figures signal potentially dramatic changes on the labor markets of advanced democracies, but some caveats need to be added. First of all, the loss of jobs due to digitalization and automation might be compensated by the creation of new employment opportunities, which are less prone to automation. Hence, overall employment levels in the economy could actually increase in the wake of the digital transformation if the number of newly created jobs is larger than the number of jobs that are eliminated, in particular if new technology is

digitalization, automation, and the welfare state

25

designed to complement human labor rather than to simply replace it (Acemoglu and Restrepo 2019, 2020). Empirical evidence for the case of Germany suggests that the latter might actually be the case (Arntz et al. 2018). Second, as already mentioned at previously, the pace of technological change is not a purely exogenous force. Technological change can be fostered by government policies promoting investments in research and development (R&D) and human capital (Buhr et al. 2016). Conversely, technological change can also be blocked by political or societal actions against the introduction of new technologies in the workplace. The latter is more likely if the nature of technology is such that it displaces manual labor rather than complementing it (Frey 2019). Third, Frey and Osborne (2017) assess the automation potential on the level of occupations rather than individual tasks (even though they claim to follow the taskbased perspective pioneered by Autor et al. (2003)). However, the occupationbased approach neglects to take into account that occupations or jobs always contain a mix of tasks, only some of which are at high risk of being automated, whereas others are not. More fine-grained, task-based analyses by Arntz et al. (2016) report much lower shares of employment at high risk of automation: from about six percent of employment in Korea to 12 percent in Austria (with an average of nine percent of jobs at high risk of automation across 21 OECD countries) (see also Bonin et al. 2015; Dengler and Matthes 2015). That said, even though a fully task-based perspective may lead to lower estimated levels of overall automation risk, the shift from “routinizable” to “non-routinizable” tasks within occupations is likely to be accompanied by a significant degree of occupational and skills upgrading, which could pose a challenge for those not able to keep up with the fast pace of transformative occupational changes. To sum up, studies in empirical labor market economics and sociology diverge somewhat regarding the potential labor market implications of the recent wave of technological change. In studies adopting a backward-looking perspective (i.e., working with observational labor market data), there is a growing consensus that technological change is one of the main drivers of labor market inequalities, with influence similar to that of globalization (Autor et al. 2015). At the same time, studies also suggest that the overall impact of technological change is limited, in the sense that it will not lead to the large-scale revolutionary developments that many airport bestsellers are announcing. Not entirely surprisingly, studies pursuing a more forward-looking approach tend to diverge regarding their predictions of overall levels of automation risk. They largely agree, however, that we can expect an increase in labor market polarization in terms of employment opportunities and wages, in particular in the short to medium term, while labor markets are transitioning to the knowledge economy. Compared to the backwardlooking empirical studies discussed previously, this line of scholarship produces

26

marius r. busemeyer

more dramatic assessments regarding the extent of the transformation on labor markets. Something that has received less attention in academic research to date, however, is to what extent and by which means the welfare state reacts to the digital transformation of the world of work. In the following section, I explore the small number of contributions in this area in greater detail.

Implications for Welfare State Politics and Policymaking As outlined in the Introduction, our guiding research questions are to what extent and how the digital transformation of employment affects and transforms both the process of welfare state policymaking and the underlying political process (politics). In general, as already noted, the extent of research on these issues is limited; nevertheless, there are some pioneering studies that help map the field. In the following, I will briefly review the existing scholarship, while also pointing out gaps in knowledge and avenues for future research.

The Politics of the Welfare State in the Digital Era As mentioned, previously, there is a long tradition of studying the connection between technological change and welfare state development. To highlight just a few seminal contributions in this tradition: Iversen and Wren (1998), for instance, pointed out emerging policy dilemmas in the service economy a good number of years ago; Armingeon and Bonoli (2006) explored the role of new social risks in the postindustrial welfare state (see also Rehm (2016)), whereas others studied the changing politics of the “new” welfare state (Pierson 2001) in the (now) past era of fiscal austerity of the 1980s and 1990s. Furthermore, a different line of scholarship has explored citizens’ attitudes toward the welfare state (Svallfors 2012) and how public policies shape these attitudes through policy feedback effects (Kumlin and Stadelmann-Steffen 2014). The following review builds on these two strands in the welfare state literature by exploring, first, how the digital transformation of work might affect welfare state attitudes and, second, how it could affect political processes both on the individual and the collective level. Regarding attitudes and preferences (see Chapter 8 by Kurer and Ha¨usermann; Chapter 11 by Guarascio and Sacchi, and Chapter 10 by Chrisp and Martinelli in this volume), a well-established finding in the literature is that the position of individuals on the labor market in terms of income, socioeconomic class, occupation, and labor market risk is an important driver of welfare state preferences (Kitschelt and Rehm 2014; Rehm 2009, 2016). A related strand in the literature has explored the link between economic globalization and welfare state attitudes (Busemeyer

digitalization, automation, and the welfare state

27

and Garritzmann 2019; Walter 2017), finding that exposure to globalizationrelated risks boosts support for the welfare state. Economic globalization and technological change are interrelated forces shaping patterns of inequality and labor market participation (Autor et al. 2015). For instance, digitalization allows some services to be offshored, increasing the pressure on white collar occupations. Moreover, competitive pressures related to globalization increase the incentives for domestic employers in wealthy democracies to substitute labor for capital (i.e., robots and software). Hence, the essential argument of the literature previously cited is that increasing labor market risk (or subjective perceptions of increased risk) due to technological change, globalization, or other factors is associated with an increase in demand for redistribution, compensation, or insurance via the welfare state. If technological change triggers large-scale transformations of the labor market, for which, as discussed, there is some evidence, it seems plausible to expect that it will also lead to citizens/workers placing new kinds of demands on the welfare state. The crucial questions to consider here, however, are: first, to what extent does the labor market, risk related to the digital transformation of work, actually constitute a new source of risk that is orthogonal to existing cleavages? And second, to what extent does technology-related risk trigger new kinds of demand on the welfare state rather than simply reinforcing existing demands? The pioneering study by Thewissen and Rueda (2019) goes some way toward answering these questions. Using a number of measures of automation risk from the labor market economics literature introduced previously, Thewissen and Rueda study the association between the individual’s risk of falling victim to automation in the near future and their support for governmental redistribution. Controlling for a range of potential confounding variables, they find robust evidence of a positive association between individual automation risk and support for more governmental redistribution. Hence, their study provides evidence that the ongoing and future automation of work does indeed constitute a new and different cleavage in preferences, since their findings show a robust effect of automation risk on support for redistribution. However, since they only focus on redistribution preferences, their study does not really give us any meaningful insights into whether automation might lead to new kinds of policy demands (the second question mentioned previously) from workers most affected by automation that go beyond existing welfare state institutions and policies. To date, there has been very little research on whether technology-related automation risk is associated with the demand for other kinds of welfare state policies, in particular social insurance and social investment policies. Chapter 8 by Kurer and Ha¨usermann in this volume provides some initial evidence that automation risk might be positively associated with support for redistribution, but not necessarily with support for social investment policies. Using data from the

28

marius r. busemeyer

European Social Survey (ESS), Busemeyer and Sahm (2021) report a similar finding. In contrast, studies by Jeffrey (2021), based on data from a survey experiment in the UK, and Zhang (2019), using similar data for the US, do not find evidence that (simulated) automation shocks increase support for redistribution or other redistributive policies. Lastly, Im (2021) argues that automation risk is positively associated with support for active labor market policies. In short, these findings are quite evidently mixed.2 A small number of papers have explored the driving forces of support for a universal basic income, since this policy is often discussed as a potential solution to the negative side effects of digitalization and automation (see following points). Again, making use of the 2016 wave of the ESS, Vlandas (2019) and Roosma and van Oorschot (2019) find that support for basic income tends to be more prevalent among left-leaning, but also low-income individuals. These papers did not explicitly study the association between technological change and support for basic income, but the study by Dermont and Weisstanner (2020) does, and it finds no association. Chapter 10 by Chrisp and Martinelli as well as Chapter 11 by Guarascio and Sacchi in this volume continue along these lines, finding mixed evidence of an association between automation risk and support for basic income. A further line of research could explore the reverse association, namely whether the existence of a generous welfare state affects support for technological change. A similar argument has been made in the globalization literature by Hays et al. (2005) who find that generous social insurance provision reduces opposition toward free trade in those exposed to globalization. So far, this kind of policy feedback effect has not yet been investigated in detail for digitalization and automation. Dekker et al. (2017) started to explore attitudes toward technology (robots in particular), confirming that these are mostly determined by economic self-interest (and therefore potentially amenable to buffering effects of policies). Moving from attitudes and preferences to political participation, some studies have started to explore the association between rapid technological change and political behavior. The various contributions to a special issue of the journal Research & Politics edited by Thomas Kurer and Bruno Palier deserve special mention here. One of the main findings of this issue is that the shrinkage of the middle class, largely related to and caused by rapid technological change, fuels support for socially conservative and right-wing populist parties across a range of OECD countries (Im et al. 2019; Kurer and Palier 2019; see also Kurer 2020). In line with the abovementioned idea regarding the buffering effect of the welfare state, Gingrich (2019) finds that generous welfare state policies do indeed have a certain

2 A likely explanation for why the findings are mixed is the fact that the different studies rely on different data sources (either comparative surveys such as the ESS or self-collected data for individual countries) or use question wordings and items that are ill-suited to studying technology-related social policy preferences.

digitalization, automation, and the welfare state

29

mitigating effect on political behavior, but this effect is limited and does not prevent groups of workers affected by rapid technological change from turning toward right-wing populist parties. Other research in this area confirms that economic shocks due to technological change play a role in the rise of nationalism in the West (Anelli et al. 2019; Colantone and Stanig 2019) and may have contributed to the election of Trump to the US presidency (Frey et al. 2018). It is interesting to note in this context that the economic risks associated with rapid technological change apparently have a mobilizing effect, although not to the benefit of the traditional defenders of the interests of the working class (i.e., social democratic parties and unions). Nevertheless, this finding stands in contrast to previous research that found a demobilizing effect of economic inequality on political turnout (Solt 2010).

Welfare State Policymaking in the Digital Era Moving beyond the microlevel of attitudes, preferences, and political behavior, I now discuss research that studies the transformation of welfare state policymaking at the macro level, focusing on collective actors such as unions, employers’ associations, and political parties, as well as the changing development trajectories of welfare state institutions and policies. As Kemmerling and Gast Zepeda mention in their contributions to this volume (Chapter 12) we know even less about how elites systematically perceive the threats and opportunities of digitalization than we do about how mass opinion is affected. An important area in this domain of research is the study of the repercussions of the platform economy, both for the changing balance of power between collective actors and for welfare state policymaking per se (see also the chapters by Picot (Chapter 13) as well as Nullmeier (Chapter 16) in this volume). As these chapters show, the rise of the so-called platform economy presents a challenge, both as regards empirically measuring the extent of the “sharing” and “gig” economy and when it comes to conceptually defining what these terms actually stand for. A prominent issue in the emerging literature on this topic is the question of how the rise of the platform economy might pose a challenge for labor market regulation and collective wage bargaining, as well as the role of established collective actors, especially unions, in these domains (Eichhorst et al. 2017; Eichhorst and Rinne 2017). The rise of digital platforms presents challenges for the traditional welfare state as it results in the boundary between regular employment and freelance self-employment becoming increasingly blurred (Eichhorst and Rinne 2017: 2–3), ultimately fueling “unfair competition with traditional firms employing dependent employees, parallel labor markets, and an erosion of labor law” (ibid.: 3). Furthermore, as argued more recently by Culpepper and Thelen (2020) as well as Rahman and Thelen (2019), a further source of power of the

30

marius r. busemeyer

digital platforms stems from the “winner-take-all” logic of the digital economy: the more customers/sellers/buyers a platform brings together, the more attractive it becomes for further customers/sellers/buyers, leading to the emergence of digital monopolies—or at least oligarchies—in the digital economy, whose economic power is increasingly mirrored in commensurate political influence (see Chapter 20 on the implications of this development for tax policy by Gelipithis in this volume). Viewed from a different perspective with more emphasis on the positive, employment-enhancing, and empowering effects of the digital economy, digital platforms might be perceived as opening up protected markets for new service suppliers, challenging “entrenched interests working to preserve their positions” (McAfee and Brynjolfsson 2016: 141). For instance, Uber’s ride-hailing activities could be regarded as opening up the protected market of taxi driving, allowing new types of workers to enter the market, who would struggle to do so in the traditional model (due to the high costs of buying a license, for instance) (Hall and Krueger 2018). So far, however, it remains an open empirical question whether the promises of empowerment made by the digital platforms of the sharing economy will lead to a more egalitarian capitalism or rather a “nightmarish form of neoliberal capitalism” (Martin 2016). Currently, the platform economy is still small, even in advanced postindustrial economies, but its future growth could result in more precarious employment. The rise of the platform economy and the digital transformation of economies also challenges the entrenched positions of collective actors. Ilsøe (2017) conducted expert interviews with representatives from trade unions and employers’ associations in Denmark, Sweden, and Germany on their perceptions of the expansion of the platform economy, showing that collective actors are still pondering on appropriate policy responses, most likely because a potentially new cleavage line between the winners and losers of the digital revolution cuts through their traditional constituencies. Thelen (2018) analyzes political reactions among policymakers and collective actors to the advance of Uber into local transportation markets in a sample of European countries and the US. In a related paper, Thelen (2019) studies how different countries (Germany, Sweden, and the Netherlands) have reacted to the rise of knowledge economy. Rooted in scholarship on varieties of capitalism (Hall and Soskice 2001; Thelen 2014), Thelen’s main argument is that political economies react very differently to the common challenge of “Uberization” and the rise of the knowledge economy. National responses are mediated by political and socioeconomic institutions as well as the prevailing balance of power between organized interests (see also Hope and Martelli (2019) as well as Iversen and Soskice (2019) for a similar argument, based on a quantitative analysis of aggregate data). Collier et al. (2018) complement these cross-national perspectives with a study on regional variation in the politics of regulation related to Uber

digitalization, automation, and the welfare state

31

within the U S, showing that in this case, the local context is certainly also of relevance. Apart from considerations related to the platform economy, which primarily concern the area of labor market regulation and collective wage bargaining, the digital transformation of the world of work has important implications for other policy areas as well. The individual chapters in this volume go into much more detail on the challenges and potential of the digital revolution for different policy areas, such as education, pensions, labor market policy, tax policy, and health care. It is obviously impossible for this short literature review to cover all aspects related to the digital transformation of welfare state policymaking in these different policy domains. Rather than repeating what the individual chapters of this volume have to say on these issues, I focus on how the emerging literature on policy responses to digitalization fits into broader discussions in comparative welfare state research, in particular the rise of the so-called “social investment state” (see also Chapter 4 by Eichhorst et al. in this volume). Broadly speaking, one of the few things that scholars and pundits discussing the impact of digitalization and automation on the labor market and the welfare state agree on is that it requires additional investment in various different types of education, including schools and vocational training, university, and lifelong learning. According to Buhr et al. (2016: 25), in the era of digitalization, education “becomes one of the crucial fields of future welfare state action” (see also Colin and Palier 2015: 29). This line of reasoning is consistent with the growing attention that scholars in comparative politics and welfare state research are paying to education (Gift and Wibbels 2014; Moe and Wiborg 2017), as well as the rapidly expanding scholarship on the social investment welfare state (Bonoli 2013; Garritzmann et al. 2017; Hemerijck 2013, 2017; Kvist 2015; Morel et al. 2012). Hence, the guiding question for research on the policy responses of the welfare state to the challenge of rapid technological change is whether the latter will simply (1) reinforce the ongoing trend toward a more employment-centered, activating, and human capital-oriented social investment state, (2) fuel political support for a more transfer-oriented, compensatory model of the welfare state, or (3) result in entirely new policy instruments (see Kurer and Ha¨usermann (Chapter 8) in this volume). For now, the dominant trend in welfare state policymaking is the gradual expansion of the social investment pillar—albeit at different speeds in different countries and on the European level (Hemerijck 2013). Undoubtedly, promoting educational opportunities at different stages in the life course and throughout the entire employment career holds significant potential for facilitating the adjustment of labor markets to rapid technological change, e.g., by fostering opportunities for lifelong learning, establishing statutory rights for further training, setting up levy-grant schemes among employers in order to finance further training, or by expanding the supply of training courses at public institutions of higher tertiary and

32

marius r. busemeyer

other forms of post-secondary education. Furthermore, policymakers can promote policies that prevent the emergence of barriers to access to higher levels of education, for instance by increasing the permeability between different segments of the education system (vocational and academic), by expanding student subsidy schemes, and by lowering tuition fees, where they exist. Even though, due to its emphasis on skill formation, the social investment model can certainly be regarded as an effective policy response to technological change, it also has downsides and blind spots, in particular if the expansion of social investment policies comes at the cost of downsizing compensatory transfer programs. Placing a strong emphasis on employability and skill formation in welfare state policymaking could then reinforce the preexisting cleavage lines in the knowledge economy, i.e., between dynamic and growing metropolitan areas and declining rural areas, as well as between high- and low-skilled workers (Iversen and Soskice 2019). Furthermore, proposals to promote educational opportunities in initial education and lifelong learning rest on the implicit assumption that changes to the supply side of the labor market (i.e., the education and training system) will result in commensurate changes in labor market participation. To some extent, this may be a reasonable and empirically valid assumption (Nelson and Stephens 2012). However, the ability of economies to absorb highly qualified employees is not without limits. Hence, even a dedicated strategy of upskilling may reach inherent limits related to the natural distribution of skills and talents in the population. In the worst case, skill mismatches following a disruptive technology shift could be so severe that even upskilling and retraining for prime age workers is barely feasible. Thus, pursuing a strategy of expanding educational opportunities in initial and further education may be an important and necessary part of policy responses to digitalization, but, on its own, it is not a panacea for the likely negative side effects of digitalization. So far, more radical proposals such as a universal basic income (UBI) (Pulkka 2017; Van Parijs and Vanderborgt 2017) have not yet won the support of mainstream political actors of the center-left (or right), even though they are gaining traction when it comes to support from citizens (Roosma and van Oorschot 2019; Vlandas 2019), albeit not necessarily from those most affected by digitalization and automation (Dermont and Weisstanner 2020). Although basic income schemes differ significantly in the details (see Chapter 10 by Chrisp and Martinelli in this volume), a fundamental difference between UBI and the social investment approach is that the former is geared toward reducing the dependence of the individual on the labor market, whereas the latter aims at enhancing the individual’s “employability,” i.e., their ability to perform well on the labor market. For the short to medium term, it seems more likely that policymakers will pursue the gradual expansion and adaptation of existing policy instruments according to the social investment logic, rather than opting for more radical solutions.

digitalization, automation, and the welfare state

33

Conclusions and Discussion This short and necessarily incomplete literature review has provided an introductory overview of existing research on the impact of rapid technological change on the welfare state. To sum up, a first takeaway from this exercise is that, to date, the number of studies on the implications of digitalization and automation for welfare state politics and policymaking is still very limited, in particular in comparison to the growing field of labor market research on the effects of technological change on labor market outcomes. The purpose of this edited volume is to narrow this significant gap in the research. Second, research on the association between the digital transformation of work and welfare state change can easily tie in with and build on existing lines of research in comparative welfare state scholarship, e.g., studies on labor market risk and welfare state attitudes, on employment position and political behavior, on industrial relations and collective wage bargaining, on the rise of the social investment model as a newly emerging paradigm of welfare state policymaking, as well as work on the political economy of the knowledge economy. Hence, the contours of an exciting and highly relevant research agenda are clearly discernible. Third, from a political perspective, it is far from clear that technological change will more or less automatically contribute to and fuel progressive welfare state reform. Although scholars and pundits regularly point to the benefits of placing more emphasis on investment in education throughout the life course, workers directly affected by the fallout of technological change might instead demand some form of immediate compensation, at least in the short term, or simply oppose further technological progress, in particular if the latter leads to the development of labor-displacing rather than labor-enhancing technology (Frey 2019). In the worst case, growing anxiety about the negative side effects of technological change and globalization might lead to a backlash and rising support for political parties that advocate a turn toward a more nationalist, regulatory, and transfer-heavy welfare state model (Anelli et al. 2019). This poses the risk of creating a vicious cycle of supporting welfare state policies that could limit the growth potential of economies in the long term, making it ever harder to establish policy instruments that compensate the losers of digitalization.

References Acemoglu, Daron. 2002. “Technical change, inequality and the labor market.” Journal of Economic Literature 40(1):7–72. Acemoglu, Daron, David Autor, Jonathan Hazell, and Pascual Restrepo. 2021. “AI and Jobs: Evidence from Online Vacancies.” NBER Working Paper 28257. Acemoglu, Daron, and Pascual Restrepo. 2019. “Automation and new tasks: How technology displaces and reinstates labor.” Journal of Economic Perspectives 33(2):3–30. Acemoglu, Daron, and Pascual Restrepo. 2020. “Robots and Jobs: Evidence from US Labor Markets.” Journal of Political Economy 128(6):2188–244.

34

marius r. busemeyer

Anelli, Massimo, Italo Colantone, and Piero Stanig. 2019. “We Were the Robots: Automation and Voting Behavior in Western Europe.” IZA. Institute for Labor Economics Discussion Paper 12485. Armingeon, Klaus, and Giuliano Bonoli. eds. 2006. The Politics of Post-Industrial Welfare States: Adapting Post-War Social Policies to New Social Risks. London: Routledge. Arntz, Melanie, Terry Gregory, and Ulrich Zierahn. 2016. “The Risk of Automation for Jobs in OECD Countries: A Comparative Analysis.” OECD Social, Employment and Migration Working Papers 189. Arntz, Melanie, Terry Gregory, and Ulrich Zierahn. 2018. Digitalisierung und die Zukunft der Arbeit: Makroo¨konomische Auswirkungen auf Bescha¨ftigung, Arbeitslosigkeit und Lo¨hne von Morgen. Mannheim: Zentrum fu¨r Europa¨ische Wirtschaftsforschung. Autor, David H. 2015. “Why are there still so many jobs? The history and future of workplace automation.” Journal of Economic Perspectives 29(3):3–30. Autor, David H., and David Dorn. 2013. “The growth of low-skill service jobs and the polarization of the US labor market.” The American Economic Review 103(5):1553– 97. Autor, David H., David Dorn, and Gordon H. Hanson. 2015. “Untangling trade and technology: Evidence from local labour markets.” The Economic Journal 125(584):621–46. Autor, David H., Frank Levy, and Richard J. Murnane. 2003. “The skill content of recent technological change: An empirical exploration.” The Quarterly Journal of Economics 118(4):1279–333. Berriman, Richard, and John Hawksworth. 2017. “Will robots steal our jobs? The potential impact of automation on the UK and other major economies.” UK Economic Outlook 2017(March):30–47. Bonin, Holger, Terry Gregory, and Ulrich Zierahn. 2015. Übertragung der Studie von Frey/Osborne (2013) auf Deutschland: Kurzexpertise Nr. 57. Mannheim: Zentrum fu¨r Europa¨ische Wirtschaftsforschung. Bonoli, Giuliano. 2013. The Origins of Active Social Policy: Labour Market and Childcare Policies in a Comparative Perspective. Oxford: Oxford University Press. Breemersch, Koen, Jozˇe P. Damijan, and Jozef Konings. 2017. “Labour Market Polarization in Advanced Countries: Impact of Global Value Chains, Technology, Import Competition from China and Labour Market Institutions.” OECD Social, Employment and Migration Working Papers 197. Brynjolfsson, Erik, and Andrew McAfee. 2014. The Second Machine Age: Work, Progress and Prosperity in a Time of Brilliant Technologies. New York: W.W. Norton and Company. Buhr, Daniel, Claudia Christ, Rolf Frankenberger, Marie-Christine Fregin, and Markus Tra¨mer. 2016. On the Way to Welfare 4.0? Digitalisation of the Welfare State in Labour Market, Health Care and Innovation Policy: A European Comparison. Berlin: Friedrich-Ebert-Stiftung. Bu¨hrer, Christian, and Christian Hagist. 2017. “The effect of digitalization on the labor market.” Pp. 115–37 in The Palgrave Handbook of Managing Continuous Business Transformation, edited by H. Ellermann, P. Kreutter, and W. Messner. London: Palgrave Macmillan.

digitalization, automation, and the welfare state

35

Busemeyer, Marius R., and Julian L. Garritzmann. 2019. “Compensation or social investment? Revisiting the link between globalisation and popular demand for the welfare state.” Journal of Social Policy 48(3):427–48. Busemeyer, Marius R., and Alexander Sahm. 2021. “Social Investment, Redistribution or Basic Income? Exploring the Association Between Automation Risk and Welfare State Attitudes in Europe.” Journal of Social Policy, FirstView online, DOI: https:// doi.org/10.1017/S0047279421000519. Colantone, Italo, and Piero Stanig. 2019. “The surge of economic nationalism in Western Europe.” Journal of Economic Perspectives 33(4):128–51. Colin, Nicolas, and Bruno Palier. 2015. “The next safety net: Social policy for a digital age.” Foreign Affairs 2015(July/August):29–33. Collier, Ruth Berins, V.B. Dubal, and Christopher L. Carter. 2018. “Disrupting regulation, regulating disruption: The politics of Uber in the United States.” Perspectives on Politics 16(4):919–37. Collins, Randall. 2013. “The end of middle-class work: No more escapes.” Pp. 37–70 in Does Capitalism Have a Future? edited by I. Wallerstein, R. Collins, M. Mann, G. Derluguian, and C. Calhoun. Oxford: Oxford University Press. Culpepper, Pepper, and Kathleen Thelen. 2020. “Are we all Amazon Primed? Consumers and the politics of platform power.” Comparative Political Studies 53(2): 288–318. Dekker, Fabian, Aanna Salomons, and Jeroen van der Waal. 2017. “Fear of Robots at Work: The Role of Economic Self-Interest.” Socio-Economic Review 15(3): 539–62. Dengler, Katharina, and Britta Matthes. 2015. “Folgen der Digitalisierung fu¨r die Arbeitswelt: Substituierbarkeitspotenziale von Berufen in Deutschland.” IAB Forschungsbericht 11. Dermont, Clau, and David Weisstanner. 2020. “Automation and the future of the welfare state: Basic income as a response to technological change?” Political Research Exchange 2(1):1–11. Eichhorst, Werner, Holger Hinte, Ulf Rinne, and Verena Tobsch. 2017. “How big is the gig? Assessing the preliminary evidence on the effects of digitalization on the labor market.” Management Revue 28(3):298–318. Eichhorst, Werner, and Ulf Rinne. 2017. “Digital Challenges for the Welfare State.” IZA. Institute for Labor Economics Policy Paper 134. Ford, Martin. 2016. The Rise of the Robots: Technology and the Threat of Mass Unemployment. London: Oneworld Publications. Frey, Carl B. 2019. The Technology Tap: Capital, Labor, and Power in the Age of Automation. Princeton: Princeton University Press. Frey, Carl B., Thor Berger, and Chinchih Chen. 2018. “Political machinery: Did robots swing the 2016 US Presidential Election?” Oxford Review of Economic Policy 34(3):418–42. Frey, Carl B., and Michael A. Osborne. 2017. “The future of employment: How susceptible are jobs to computerisation?” Technological Forecasting and Social Change 114(January):254–80. Garritzmann, Julian L., Silja Ha¨usermann, Bruno Palier, and Christine Zollinger. 2017. “WOPSI: The World Politics of Social Investment.” LIEPP Working Paper 64. Gift, Thomas, and Erik Wibbels. 2014. “Reading, writing, and the regrettable status of education research in comparative politics.” Annual Review of Political Science 17:291–312.

36

marius r. busemeyer

Gingrich, Jane. 2019. “Did state responses to automation matter for voters?” Research & Politics 6(1):1–9. Goldin, Claudia, and Lawrence F. Katz. 2008. The Race between Education and Technology. Cambridge: Belknap Press. Goos, Maarten, and Alan Manning. 2007. “Lousy and lovely jobs: The rising polarization of work in Britain.” The Review of Economics and Statistics 89(1):118–33. Goos, Maarten, Alan Manning, and Anna Salomons. 2014. “Explaining job polarization: routine-based technological change and offshoring.” American Economic Review 104(8):2509–26. Gordon, Robert J. 2016. The Rise and Fall of American Growth: The U.S. Standard of Living since the Civil War. Princeton: Princeton University Press. Graetz, Georg, and Guy Michaels. 2018. “Robots at work.” Review of Economics and Statistics 100(5):753–68. Hall, Johnathan V., and Alan B. Krueger. 2018. “An analysis of the labor market for Uber’s driver-partners in the United States.” ILR Review 71(3):705–32. Hall, Peter A., and David Soskice. eds. 2001. Varieties of Capitalism: The Institutional Foundations of Comparative Advantage. Oxford: Oxford University Press. Hays, Jude C., Sean D. Ehrlich, and Clint Peinhardt. 2005. “Government spending and public support for trade in the OECD: An empirical test of the embedded liberalism thesis.” International Organization 59(Spring 2005):473–94. Hemerijck, Anton. 2013. Changing Welfare States. Oxford: Oxford University Press. Hemerijck, Anton. (ed.) 2017. The Uses of Social Investment. Oxford: Oxford University Press. Hope, David, and Angelo Martelli. 2019. “The transition to the knowledge economy, labor market institutions, and income inequality in advanced democracies.” World Politics 71(2):236–88. Ilsøe, Anna. 2017. “The digitalisation of service work – social partner responses in Denmark, Sweden and Germany.” Transfer 23(3):333–48. Im, Zhen J. 2021. “Automation risk and support for welfare policies: How does the threat of unemployment affect demanding active labor market policy support?” Journal of International and Comparative Social Policy 37(1):76–91. Im, Zhen J., Nonna Mayer, Bruno Palier, and Jan Rovny. 2019. “The ‘losers of automation’: A reservoir of votes for the radical right?” Research & Politics 6(1):1–7. Iversen, Torben, and David Soskice. 2019. Democracy and Prosperity: Reinventing Capitalism through a Turbulent Century. Princeton: Princeton University Press. Iversen, Torben, and Anne Wren. 1998. “Equality, employment, and budgetary restraint: the trilemma of the service economy.” World Politics 50(4):507–46. Jeffrey, Karen. 2021. “Automation and the Future of Work: How Rhetoric Shapes the Response in Policy Preferences.” QPE Working Paper, SSRN. Retrieved September 9, 2021 (https://ssrn.com/abstract=3493878). Kitschelt, Herbert, and Philipp Rehm. 2014. “Occupations as a site of political preference formation.” Comparative Political Studies 47(12):1670–706. Kristal, Tali, and Yinon Cohen. 2016. “The causes of rising wage inequality: The race between institutions and technology.” Socio-Economic Review 15(1):187–212. Kumlin, Staffan, and Isabelle Stadelmann-Steffen. 2014. “Citizens, policy feedback, and European welfare states.” Pp. 3–16 in How Welfare States Shape the Democratic Public: Policy Feedback, Participation, Voting, and Attitudes, edited by S. Kumlin and I. Stadelmann-Steffen. Cheltenham: Edward Elgar.

digitalization, automation, and the welfare state

37

Kurer, Thomas. 2020. “The declining middle: Occupational change, social status, and the populist right.” Comparative Political Studies 53(10–11):1798–835. Kurer, Thomas, and Bruno Palier. 2019. “Shrinking and shouting: The political revolt of the declining middle in times of employment polarization.” Research and Politics 6(1):1–6. Kvist, Jon. 2015. “A framework for social investment strategies: Integrating generational, life course and gender perspectives in the EU social investment strategy.” Comparative European Politics 13(1):131–49. Martin, Chris J. 2016. “The sharing economy: A pathway to sustainability or a nightmarish form of neoliberal capitalism?” Ecological Economics 121:149–59. Mason, Paul. 2015. Postcapitalism: A Guide to Our Future. London: Allen Lane. McAfee, Andrew, and Erik Brynjolfsson. 2016. “Human work in the robotic future: Policy for the age of automation.” Foreign Affairs 2016(July/August):139–50. Michaels, Guy, Ashwini Natraj, and John Van Reenen. 2014. “Has ICT polarized skill demand? Evidence from eleven countries over twenty-five years.” Review of Economics and Statistics 96(1):60–77. Moe, Terry M., and Susanne Wiborg. 2017. “Introduction.” Pp. 1–23 in The Comparative Politics of Education: Teachers Unions and Education Systems around the World, edited by T. M. Moe and S. Wiborg. Cambridge: Cambridge University Press. Mokyr, Joel, Chris Vickers, and Nicolas L. Ziebarth. 2015. “The history of technological anxiety and the future of economic growth: Is this time different?” Journal of Economic Perspectives 29(3):31–50. Morel, Nathalie, Bruno Palier, and Joakim Palme. 2012. “Beyond the welfare state as we knew it?” Pp. 1–30 in Towards a Social Investment Welfare State? Ideas, Policies and Challenges, edited by N. Morel, B. Palier, and J. Palme. Bristol: Policy Press. Nedelkoska, Ljubica, and Glenda Quintini. 2018. “Automation, Skills Use and Training.” OECD Social, Employment and Migration Working Papers 202. Nelson, Moira, and John D. Stephens. 2012. “Do social investment policies produce more and better jobs?” Pp. 205–34 in Towards a Social Investment Welfare State? Ideas, Policies and Challenges, edited by N. Morel, B. Palier and J. Palme. Bristol: Policy Press. OECD. 2019. Under Pressure: The Squeezed Middle Class. Paris: OECD Publishing. Pierson, Paul. ed. 2001. The New Politics of the Welfare State. Oxford: Oxford University Press. Pulkka, Ville-Veikko. 2017. “A Free lunch with robots – can a basic income stabilise the digital economy?” Transfer: European Review of Labour and Research 23(3):295–311. Rahman, K. Sabeel, and Kathleen Thelen. 2019. “The rise of the platform business model and the transformation of twenty-first-century capitalism.” Politics & Society 47(2):177–204. Rehm, Philipp. 2009. “Risks and redistribution. An individual-level analysis.” Comparative Political Studies 42(7):855–81. Rehm, Philipp. 2016. “Risk Inequality and Welfare States: Social Policy Preferences, Development, and Dynamics.” Cambridge: Cambridge University Press. Rifkin, Jeremy. 2014. The Zero Marginal Cost Society: The Internet of Things, the Collaborative Commons, and the Eclipse of Capitalism. New York: St. Martin’s Press. Roosma, Femke, and Wim van Oorschot. 2019. “Public opinion on basic income: Mapping European support for a radical alternative for welfare provision.” Journal of European Social Policy 30(2):190–205.

38

marius r. busemeyer

Schwab, Klaus. 2016. The Fourth Industrial Revolution. New York: Random House/Crown Business. Solt, Frederick. 2010. “Does economic inequality depress electoral participation? Testing the Schattschneider hypothesis.” Political Behavior 32:285–301. Svallfors, Stefan. 2012. “Welfare states and welfare attitudes.” Pp. 1–24 in Contested Welfare States: Welfare Attitudes in Europe and Beyond, edited by S. Svallfors. Stanford: Stanford University Press. Thelen, Kathleen. 2014. Varieties of Liberalization and the New Politics of Social Solidarity. Cambridge: Cambridge University Press. Thelen, Kathleen. 2018. “Regulating Uber: The politics of the platform economy in Europe and the United States.” Perspectives on Politics 16(4):938–53. Thelen, Kathleen. 2019. “Transitions to the knowledge economy in Germany, Sweden and the Netherlands.” Comparative Politics 51(2):295–315. Thewissen, Stefan, and David Rueda. 2019. “Automation and the welfare state: Technological change as a determinant of redistribution preferences.” Comparative Political Studies 52(2):171–208. Van Parijs, Philippe, and Yannick Vanderborght. 2017. Basic Income: A Radical Proposal for a Free Society and a Sane Economy. Cambridge, MA: Harvard University Press. Vlandas, Tim. 2019. “The politics of the basic income guarantee: Analysing individual support in Europe.” Basic Income Studies 14(1):1–17. Walter, Stefanie. 2017. “Globalization and the demand-side of politics: How globalization shapes labor market risk perceptions and policy preferences.” Political Science Research and Methods 5(1):55–80. Zhang, Baobao. 2019. “No Rage against the Machine: Threat of Automation Does Not Change Policy Preferences.” MIT Political Science Department Research Paper 2019–25.

PART I

BROADER TRENDS Is this Time Different or Not?

3 Digitalization and the Transition to Services Anne Wren

Introduction One revolutionary impact of digitalization has been to increase the capacity for productivity growth and international trade in many areas of service provision. In advanced economies faced with deindustrialization, an important effect of this has been to facilitate the development of alternative high-productivity growth models that are services based, with positive implications for overall growth in living standards (see Hassel and Palier 2021; Wren 2021). In discussions of more politically charged issues associated with the new technologies, such as the rise of the gig economy and workers’ rights, or whether robots will “steal” jobs currently occupied by human workers, this is a point which is sometimes overlooked. Its implications, however, are significant. As Baumol (1967) influentially argued, in many areas of service provision, the capacity for productivity growth is inherently constrained, and productivity growth rates have historically remained low in relative terms. In the absence of productivity-enhancing innovation in the services sector, therefore, deindustrialization poses a threat to economy-wide growth: as the proportion of production located in higher productivity manufacturing sectors declines, economic growth rates will depend increasingly on rates of productivity growth in low-productivity services, leading to economic stagnancy and declining overall economic welfare (Baumol et al. 1989). In tandem with this, it raises the prospect of governments facing unappealing sets of policy trade-offs around the distribution of a shrinking economic pie (Iversen and Wren 1998). In this context, the impact of digitalization in facilitating productivity growth in certain areas of services— primarily through the rapid accumulation, analysis, and global transmission of information—is profound. The diffusion of ICT increases the possibility that dynamic high-growth economies can still flourish in a context of deindustrialization and implies that the transition to services does not have to be associated with declining living standards. It affords countries the opportunity to transition to sustainable high-productivity services-based growth models, completing the radical

Anne Wren, Digitalization and the Transition to Services. In: Digitalization and the Welfare State. Edited by Marius R. Busemeyer, Achim Kemmerling, Paul Marx, and Kees van Kersbergen, Oxford University Press. © Oxford University Press (2022). DOI: 10.1093/oso/9780192848369.003.0003

42

anne wren

transformation from industrial to service societies that has been underway for over 50 years. In response to one of the core questions posed in this volume, therefore, I argue in this chapter that the impact of the current technological revolution is different, in the sense that it has facilitated a structural transformation of the economy of a scale that has not been seen since the Industrial Revolution. Further, adopting the “optimists”’ perspective (see Introduction to this volume), the technological revolution appears to offer an escape from the gloomy scenario of an end to growth and economic stagnancy associated with deindustrialization. Like the Industrial Revolution that preceded it, however, the transition toward a digitized services-based economy has had disruptive effects on labor markets and associated with this, also on distributional and political outcomes, as the evidence and analysis throughout this volume makes clear. In this chapter, I emphasize how the differential impact of the new technologies across different types of services sectors and on workers with different types of skills, aggravates the tendency toward labor market bifurcation associated with deindustrialization. It is well established at this point that the new ICTs complement the skills of those with tertiary education, leading to an increase in the relative demand for these types of workers, and strengthening their labor market position (see references in Busemeyer’s (Chapter 2) in this volume). The ICT-intensive service sectors that are playing an increasingly important role in generating economic growth (finance, communications, and other business services, for instance) are heavily dependent on highly skilled workers, and wages in these sectors tend to be relatively high cross-nationally. Conversely, at lower skill levels, the loss of manufacturing jobs has been compensated for by increases in employment in less ICT-intensive services where rates of productivity growth and relative wages are lower (such as catering, distribution, and community and personal services). In distributional terms, the big change at this end of the labor market is the loss of what were relatively well-paid jobs for medium- and less-skilled workers in manufacturing sectors, and their replacement by relatively low-paid employment in non-ICT-intensive service sectors. The extent to which these tendencies have translated into actual increases in inequality varies across countries. This is largely as a result of variation in rates of deindustrialization, in the expansion of market services, and in the extent to which the welfare state continues to act as an intervening mechanism to facilitate the creation of jobs for less-skilled workers at more generous rates of pay (see Wren 2013, 2021; see also Eichhorst et al. (Chapter 4)and Busemeyer and Glassmann (Chapter 5) in this volume). In all of the advanced economies, however, the same economic mechanisms are at play. The distributional tensions which this creates have had serious implications for politics, contributing to the new “digital divides” in political preferences and outcomes considered elsewhere in this volume (see, for example, Kurer and

digitalization and the transition to services

43

Ha¨usermann (Chapter 8) and Gingrich and Kuo (Chapter 9)). The loss of relatively well-paid manufacturing jobs, and the increased concentration of high-skilled workers and their less-skilled counterparts in very different types of economic sectors, means that the economic interests of high- and low-skilled workers are less aligned than previously. This increased divergence of economic interests places strain on historical political coalitions across skill groups, as the highly skilled reap a large share of the benefits of the ICT revolution, while those at lower skill levels are “left behind.” In this chapter, I develop these arguments as follows. I begin by outlining the significance of digitalization in facilitating the sectoral reorientation of the economy toward high-productivity services sectors in an era of deindustrialization. I then examine the impact of this transformation on labor markets, and the associated bifurcation of labor market opportunities for workers with different levels of skills. In my analysis, I make use of historical data from the UK, Sweden, and Germany to describe how this economic and distributional transformation has evolved over time. This historical approach is designed to complement the forward-looking analyses elsewhere in this volume, emphasizing the real and directly measurable socioeconomic changes which have already occurred in connection with the ICT revolution, and examining the pace of these changes across three different welfare production regimes.. The final section includes some concluding comments.

Deindustrialization, the Transition to Services, and the Impact of Digitalization Deindustrialization The economic effects of the process of deindustrialization have been far-reaching: the world’s most developed economies have experienced a rapid decline in the share of employment and output located in manufacturing sectors over the last half century. Automation has played a role in influencing these outcomes, as has increased competition in markets for labor-intensive manufactured goods from the developing world, but the root causes of deindustrialization lie in the longerrun dynamics of economic development (Lawrence and Edwards 2013; Iversen and Cusack 2000; Rowthorn and Ramaswamy 1999). On the demand side an important explanation for deindustrialization concerns the relationship between consumption patterns and income growth. As Engel’s law famously states, the proportion of incomes spent on agricultural products declines with increasing incomes once basic nutritional needs are met, leading to a decline in the economic significance of agricultural sectors as economies develop. As originally argued by Fisher (1935) and Clark (1940), similar mechanisms are at play in the shift away from manufacturing. The saturation of demand for a wide

44

anne wren

range of consumer durables, which were the basis of output and employment in the Fordist era, has resulted in a decline in the income and price elasticities of demand for these goods (Appelbaum and Schettkat 1995). As a result, the increases in productivity in manufacturing associated with automation, and the reductions in prices which accompany this, no longer generate sufficient expansion in manufacturing demand to prevent net manufacturing employment loss (in contrast with the productivity increases associated with Fordist technical innovations in the earlier period (Bessen 2019; Meidner 1974; Rehn 1985)). Demand for many services, on the other hand, remains price and income elastic (Appelbaum and Schettkat 1999; Fuchs 1968; Kalwij et al. 2007; Kongsrud and Wanner 2005), and as a result, services tend to occupy a greater share of consumption and employment as incomes rise (see also Lawrence and Edwards 2013). On the supply side, meanwhile, as Baumol (1967) pointed out, the shift from manufacturing to services in consumption and employment is partly an inevitable outcome of productivity differentials between the two sectors. In services sectors in which face-to-face human interaction is an important component of production (for example, nursing, caring for children), there are barriers to productivity growth that have yet to be resolved by technological innovation. Over time, therefore, relative prices in these sectors rise in comparison with manufacturing sectors, in which productivity growth is higher, leading them to occupy a greater share of consumption; and labor shedding in high-productivity manufacturing sectors leads to a greater concentration of employment in low-productivity services sectors. Hence, while much of the political discourse surrounding deindustrialization has emphasized the impact of automation and of increased competition in traded goods from the developing world, in reality, in the advanced economies, deindustrialization is deeply rooted in the longer-run dynamics of economic development.

Digitalization and the Transition to Services Although digitalization has not been the primary driver of the process of deindustrialization, where it has arguably had its most revolutionary impact is in facilitating productivity increases and trade in services. Baumol (1967) famously highlighted the obstacles to productivity growth stemming from the requirement for face-to-face interpersonal interaction in the provision of many types of services. Childcare or nursing are good examples here. There are limits to the extent to which the number of children supervised by one carer can be increased without reducing the quality of supervision—and hence the service output. Similarly, many of the services provided by nurses rely heavily on one-on-one (often physical) interaction between carer and patient, which are not particularly amenable to increased rates of output. Thus far, the capacity for international trade is also

digitalization and the transition to services

45

clearly limited for these types of services. Where face-to-face interpersonal interaction is a critical component of the service, providing that service from a distance is often not possible. In a context of deindustrialization, this characteristic of service provision creates two sets of problems. First, as manufacturing sectors shrink in size, overall economic growth rates are increasingly determined by rates of productivity and income growth in services sectors. Limits on the capacity for productivity growth in services, therefore, place limits on the capacity for economic growth, and raise the prospect of economic stagnancy in the long term, with negative implications for economic welfare (Baumol et al. 1985; Rowthorn and Ramaswamy 1999). Second, as described in Iversen and Wren (1998), without rising real incomes in highproductivity sectors to sustain demand, strategies for employment creation create new types of distributional tensions for governments. Low rates of productivity growth in services imply that, in the absence of rising real incomes elsewhere in the economy, expansions in demand and employment rely on keeping relative prices and wages low. For governments, this creates a trade-off between the potential policy goals of employment creation and equality, which did not exist in the so-called “Golden Age” of high-productivity manufacturing expansion. In some countries, governments have intervened to reduce the distributional impact of this tradeoff by expanding employment directly in public services sectors, at higher relative wage rates. Again, in the absence of growth elsewhere in the economy, however, these strategies also come at a cost, in terms either of higher taxation or public debt. Thus, in this scenario, a three-way trade-off (trilemma) exists between the policy goals of employment creation, equality, and low taxation or public debt, which is both economically and politically challenging. One truly radical impact of the digital revolution, therefore, has been to alter this somewhat gloomy postindustrial scenario by significantly enhancing the capacity for productivity growth and international trade in many areas of services. In sectors where physical interpersonal interaction is of lesser importance to service provision, the new technologies can significantly enhance labor productivity (Inklaar et al. 2008; Triplett and Bosworth 2004, 2006). Meanwhile, the ability to transmit data instantaneously around the globe, and to communicate long distance at low or no cost, fundamentally alters the nature of service trade (Freund and Weinhold 2002; Nath and Liu 2017). This transformation allows high-productivity service sectors to play an enhanced role as the drivers of economic growth, and at the same time to reduce the distributional tensions associated with the service transition. Since demand for many social and personal services is income elastic (especially in areas where home production is feasible—think, for example, of gardening services or restaurant meals), rising real incomes in high-productivity service sectors can facilitate the expansion of demand in these types of sectors without relying so heavily on keeping relative prices and wages low. In this way, the starkness of the trade-off between employment creation and equality in the

46

anne wren

service economy is reduced. Similarly, strong growth rates render the expansion of public service sector employment and welfare supports more sustainable (Wren 2013, 2021). Table 3.1 compares the contribution of ICT to value added across sectors in a range of EU countries between 1995 and 2005. The three outlying sectors in terms of the impact of the new technologies are all services sectors in which the face-to-face interpersonal component of provision is nonessential—financial intermediation, telecommunications, and other business services. In all other areas

Table 3.1 Contribution of ICT technology to the growth of value added: EU-15ex, annual average (1996–2005). TOTAL INDUSTRIES AGRICULTURE, HUNTING, FORESTRY, AND FISHING MINING AND QUARRYING TOTAL MANUFACTURING FOOD, BEVERAGES, AND TOBACCO TEXTILES, TEXTILE, LEATHER, AND FOOTWEAR WOOD AND OF WOOD AND CORK PULP, PAPER, PAPER, PRINTING, AND PUBLISHING CHEMICAL, RUBBER, PLASTICS, AND FUEL OTHER NON-METALLIC MINERAL BASIC METALS AND FABRICATED METAL MACHINERY, NEC ELECTRICAL AND OPTICAL EQUIPMENT TRANSPORT EQUIPMENT MANUFACTURING NEC; RECYCLING ELECTRICITY, GAS, AND WATER SUPPLY CONSTRUCTION WHOLESALE TRADE RETAIL TRADE HOTELS AND RESTAURANTS TRANSPORT AND STORAGE POST AND TELECOMMUNICATIONS FINANCIAL INTERMEDIATION REAL ESTATE OTHER BUSINESS & RENTING OF MACHINERY, EQUIPMENT PUBLIC ADMIN AND DEFENSE; COMPULSORY SOCIAL SECURITY EDUCATION HEALTH AND SOCIAL WORK OTHER COMMUNITY, SOCIAL, AND PERSONAL SERVICES

0.448 0.026 0.097 0.297 0.201 0.158 0.166 0.491 0.298 0.191 0.186 0.285 0.542 0.271 0.262 0.265 0.132 0.630 0.261 0.145 0.455 1.857 1.172 0.083 1.251 0.296 0.158 0.222 0.399

Note: EU-15ex represents the EU-15 Member States in the database for which growth accounting could be performed, namely: AUT, BEL, DNK, ESP, FIN, FRA, GER, ITA, NLD, and UK.

digitalization and the transition to services

47

of services, and also in manufacturing, however, the impact of ICT is thus far less significant.1 The increased economic significance of the leading ICT-intensive service sectors is illustrated in Figure 3.1,2 which shows how the composition of value added (gross) changed across three different welfare production regimes (the UK, Germany, and Sweden) between 1970 and the onset of the financial crisis in 2007.3 The sectoral classifications used here and throughout the chapter are as follows: ICTintensive services (financial intermediation, communications, and other business); non-ICT-intensive services (hotels and restaurants, wholesale and retail, and other community, social and personal services); welfare services (education, health and social work, public administration and defense), and manufacturing (see Table 3.1 and for a broader discussion, see Wren 2013, 2021). As is clear from the first panel in Figure 3.1, the UK has experienced the greatest economic transformation since the start of the 1970s. During this period, ICTintensive service sectors expanded to take the lead role in driving value added in the UK (doubling their contribution from just over 10 percent to 25 percent, with the majority of that expansion occurring from the mid to late 1980s on). At the same time, the UK manufacturing sector’s contribution shrank from more than one-third in 1970, to less than one-sixth in 2007. The contribution of other types of service sectors to value added also expanded somewhat in this period, but the increase was far less marked. In Germany and Sweden, meanwhile, manufacturing was still a considerably more significant contributor to overall value added at the end of the period than the UK, which in part reflects the continuing strength of the German and Swedish manufacturing sectors (Carlin and Soskice 2009; Hassel and Palier 2021; Thelen 2014, 2019), but also the slower rates of productivity growth in market services observed in many European countries, compared to the US and the UK, in the period since the mid-1990s (van Ark et al. 2008). Nonetheless, even in these two countries, ICT-intensive service sectors have expanded significantly in recent decades. At the macro level then, one transformative effect of the digital revolution has been to enhance the capacity of high-productivity service sectors to act as the 1 These aggregate figures will inevitably mask differences at the sub-sectoral level. For example, we would expect higher levels of ICT intensity to be observed in third-level education than among preschoolers; similarly the new technologies can clearly enhance the output of a radiographer more than a nurse’s aide. Meanwhile, as Busemeyer and Glassman (Chapter 5) correctly emphasize in this volume, the extent to which productivity may be enhanced in more areas of personal and creative services in the future—by allowing an online DIY trainer or an opera singer to provide a service simultaneously to a larger number of people, for example—remains to be seen. 2 Source for all figures: EU KLEMS database (Growth Accounting and Basic Tables, National Accounts), November 2009 release, March 2011 update. 3 The NACE classifications used in the EU KLEMS database changed after the 2011 release (from three to four), so the more recent (post-2007) data are incompatible with the historical data that go back to 1970. For the purposes of this chapter, therefore, the analysis is restricted to the pre-2006 period to allow for historical comparison.

48

anne wren

UK 40 35

20

ICT Intensive Service Sectors Non-ICT Intensive Service Sectors

15

Welfare Sectors

10

Manufacturing

30 25

5 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006

0 Germany 40 35 30

ICT Intensive Service Sectors

25 20

Non-ICT Intensive Service Sectors

15

Welfare Sectors

10

Manufacturing

0

1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006

5

Sweden 40 35 30 ICT Intensive Service Sectors Non-ICT Intensive Service Sectors

25 20

Welfare Sectors

15

Manufacturing

10 5

1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006

0

Fig. 3.1 Sectoral composition of value added (gross) over time: UK, Germany, and Sweden (1970–2007).

digitalization and the transition to services

49

engine for economic growth (see also Bosworth and Triplett 2007).⁴ As illustrated previously, countries do vary in the extent to which they have reoriented their economies away from manufacturing and toward hi-tech services (see also Hassel and Palier 2021; Wren 2013, 2021). However, the economic significance of ICTintensive service sectors is increasing across all of the advanced economies. In an era of deindustrialization, the implications of this for overall economic welfare are profound.

The Impact on Labor Markets Digitalization, Deindustrialization, and the Bifurcation of Labor Market Opportunities As with earlier technological revolutions, however, the advent of digitalization has had disruptive effects on labor markets, and associated with this, also on social and political outcomes. Even countries like Sweden and Germany, in which the relative contribution of manufacturing sectors to value added remains strong, have experienced significant labor shedding in manufacturing sectors as a result of the dynamics of demand and productivity growth described earlier. Moreover, to varying degrees across countries, deindustrialization and the transition toward services has been associated with an increased bifurcation of labor market opportunities along the lines of skill (and sometimes region), which is intimately connected with technical change. Figure 3.2 illustrates how the distribution of employment across economic sectors changed between the early 1970s and the mid-2000s in the UK, Sweden, and Germany. In all three countries, rapid labor shedding in manufacturing sectors since the start of the 1970s has been associated with significantly increased concentrations of employment in low-productivity services sectors. In the UK, non-ICT-intensive private services had replaced manufacturing as the dominant employers by the early 1980s; in Sweden, unsurprisingly, that role has been played by the welfare sectors; while Germany has seen roughly equal expansion in both. Since the start of the 1990s, however, the most rapid expansions in employment in all three countries have occurred in ICT-intensive services; and, in the UK at least, ICT-intensive services had outstripped manufacturing in terms of employment provision by the early 2000s. These changes have had effects on the allocation of economic welfare in society which are far from neutral. As economies have reoriented themselves in sectoral ⁴ Note that the argument here is not that the ICT revolution was the only source of increased productivity growth in services sectors in this period. Clearly deregulation and the liberalization of international services trade also played an important role. Nevertheless, the substantial contribution of technological change to services productivity growth in this period is well established at this point (Triplett and Bosworth 2006; van Ark et al. 2008).

anne wren UK 35 30 25

ICT-Intensive Services

20

Non ICT-Intensive Services

15

Welfare Services

10

Manufacturing

0

_1970 _1972 _1974 _1976 _1978 _1980 _1982 _1984 _1986 _1988 _1990 _1992 _1994 _1996 _1998 _2000 _2002 _2004 _2006

5

Germany 35 30 25

ICT-Intensive Services

20

Non ICT-Intensive Services

15 Welfare Services 10 Manufacturing

0

_1970 _1972 _1974 _1976 _1978 _1980 _1982 _1984 _1986 _1988 _1990 _1992 _1994 _1996 _1998 _2000 _2002 _2004 _2006

5

Sweden 35 30 25

ICT-Intensive Services

20

Non ICT-Intensive Services

15

Welfare Services

10

Manufacturing

5 0 _1970 _1972 _1974 _1976 _1978 _1980 _1982 _1984 _1986 _1988 _1990 _1992 _1994 _1996 _1998 _2000 _2002 _2004 _2006

50

Fig. 3.2 Sectoral distribution of hours worked, in percent: UK, Germany, and Sweden (1970–2007).

digitalization and the transition to services

51

terms, one of the most significant distributional effects of the digital revolution has been its impact on the relative demand for different types of labor market skills. In particular, it is well established at this point, that ICT tends to increase the relative demand for highly skilled (college-educated workers) (see the literature following Autor et al. 2003; Goldin and Katz 2008; see also references in Chapter 2 by Busemeyer in this volume). This is because the new technologies are not (as yet) particularly effective at performing non-routine cognitive tasks requiring what Hall and Soskice (2001) would describe as high-end general skills “flexibility, creativity, generalized problem solving, and complex communications” (Autor et al. 2003: 5). Instead, they serve to complement the skills of the (typically collegeeducated) workers who perform those tasks: faster access to more complete market information, for example, can improve managerial decision-making, but it cannot yet substitute for that decision-making. In many areas of services requiring lower skill levels, however, the role played by ICT as a complement to, or substitute for, human labor has thus far been less significant. A waitress or a hairdresser may benefit from the use of handheld technologies for taking orders or payments, for example, but there is, as yet, little evidence that productivity has been significantly enhanced in these kinds of sectors by the advent of ICT (Bosworth and Triplett 2007). At the same time, the new technologies cannot substitute for human labor in many of these areas (although this might change in the future). The importance of the impact of ICT on the demand for workers at the lowest skill levels is therefore less clear at this point. Rather, the expansion of low-skilled employment in non-ICT-intensive service sectors is driven by factors such as the growth of incomes elsewhere in the economy, and by relative prices (since the demand for many personal services is income and price elastic (Appelbaum and Schettkat 1995, 1999)). Some authors argue, meanwhile, that it is those with mid-level skills (with secondary, but not tertiary qualifications) whose jobs are most at risk from the advent of ICT, because these workers are most likely to be engaged in routine tasks which are amenable to automation (Autor et al. 2003).

Distributional Change in the UK Labor Market As can be seen from Figures 3.1 and 3.2, the UK, in particular, has experienced a radical transformation in terms of the sectoral composition of growth and the associated reorganization of labor markets since the early 1970s. The next two figures illustrate some of the distributional impacts of these changes. From Figure 3.3⁵ we ⁵ In all three countries shown in the following figures, according to EU KLEMS definitions, the “high-skilled category” is restricted to those with a university degree. In the UK and Germany, “lowskilled” workers are those with no formal qualifications, and the “medium-skilled” category includes all those with intermediate-level qualifications below those of a university degree. In Sweden, however, the “medium-skilled” category is explicitly restricted to those with higher and intermediate vocational training.

52

anne wren High-Skilled 45 40 35 30 25 20 15 10 5 0

0 2 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 97 97 97 97 97 98 98 98 98 98 99 99 99 99 99 00 00 00 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _2 _2 _2 Medium-Skilled 45 40 35 30 25 20 15 10 5 0

0 2 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 97 197 197 197 197 198 198 198 198 198 199 199 199 199 199 200 200 200 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

_1

Low-Skilled 45 40 35 30 25 20 15 10 5 0

0 2 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 97 97 97 97 97 98 98 98 98 98 99 99 99 99 99 00 00 00 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _2 _2 _2 Manufacturing Non ICT-Intensive Services

ICT-Intensive Services Welfare Services

Fig. 3.3 Cross-sectoral distribution of employment, in percent, by skill level: UK (1970–2005).

digitalization and the transition to services

53

can see that the changes which have occurred in employment patterns have been very different at different skill levels. Thus, in 1970, a college-educated worker in the UK was, by some margin, most likely to be employed in one of the welfare service sectors. By 2007, however, he or she, was equally, if not slightly more likely, to be employed in ICT-intensive services. In contrast, the lowest-skilled workers, who were most likely to find employment in manufacturing in 1970, were heavily concentrated in low-productivity, non-ICT-intensive services by the end of the period, as were those with medium-level skills. The real distributional impact of these changes becomes clear when we consider how wages diverge across different types of economic sectors. Figure 3.4 describes the evolution of average wages at the sectoral level when compared with the economy-wide mean. The first important point to note here is that manufacturing has always been, and remains, a relatively well-paid sector. In the UK, in 1970, average manufacturing wages clearly outstripped the economy-wide mean and that differential increased throughout the late 1970s and 1980s, finally hovering around 20 percent in the early 2000s. The only other three sectors to record wages that were consistently well above the economy-wide average were two ICT-intensive services sectors—financial intermediation and post and telecommunication⁶—along with public administration and defense (see Panel (A)). In contrast, the non-ICTintensive service sectors—hotels and restaurants, wholesale and retail, and other community, social, and personal services—were among the lowest-paid sectors throughout the period (see Panel (B)). Taken together, the data in Figures 3.3 and 3.4 therefore illustrate the bifurcation of labor market opportunities associated with deindustrialization, digitalization, and the transition to a services-based economy in the UK. At one end of the labor market, the story is one of an increased concentration of highly skilled employment in consistently highly paid ICT-intensive services sectors from the late 1980s on. This partly reflects a shift away from manufacturing, but also a movement away from welfare sectors. In contrast, at lower skill levels, as deindustrialization has proceeded, there has been a marked movement away from relatively highly paid manufacturing sectors and toward non-ICT-intensive services sectors, which are among the worst paid in the UK economy. This expansion was facilitated by the removal of welfare state protection on the wages of low-paid workers, which occurred in the UK in the 1980s, in particular. Further, regional mapping of these data (not shown here) shows that, over the past 40 years, this ⁶ Relative wages in the post and telecommunications sector dip sharply in the mid-1980s, possibly as a result of the Thatcher government’s privatization initiatives, but have subsequently increased significantly, as we would expect in the wake of the revolution in ICT.

54

anne wren

(a) 1.8 MANUFACTURING

1.6

POST AND TELECOMMUNICATIONS

1.4 1.2

FINANCIAL INTERMEDIATION

1 OTHER BUSINESS, RENTING MACHINERY

0.6

PUBLIC ADMIN AND DEFENCE

_1970 _1972 _1974 _1976 _1978 _1980 _1982 _1984 _1986 _1988 _1990 _1992 _1994 _1996 _1998 _2000 _2002 _2004

0.8

(b) 1.2

WHOLESALE AND RETAIL TRADE

1 0.8

HOTELS AND RESTAURANTS

0.6

OTHER COMMUNITY, SOCIAL AND PERSONAL SERVICES

0.4

EDUCATION

0.2 0 _1970 _1972 _1974 _1976 _1978 _1980 _1982 _1984 _1986 _1988 _1990 _1992 _1994 _1996 _1998 _2000 _2002 _2004

HEALTH AND SOCIAL WORK

Fig. 3.4 Average sectoral compensation compared with economy-wide mean: UK (1970–2005).

skill-based cleavage has been accompanied by a growing regional divide: the impact of deindustrialization has been most strongly felt in the north and west of the UK—while the explosion in high-productivity services jobs has been centered in the southeast. As described in the emerging literature on postindustrial growth models, however, these effects have thus far been more muted in countries which, on the one hand, have not transitioned as fully toward a private services-led model, and, on the other, have used welfare state interventions to manage the transition with less severe effects (see, for example, the contributions in Hassel and Palier 2021).

digitalization and the transition to services

55

Germany and Sweden In Sweden, for example, there has been a high level of consistency in terms of the distribution of employment at different skill levels, across sectors, over time⁷ (see Figure 3.5). Throughout the period under investigation, the Swedish economy has displayed a continued ability to generate jobs for medium-skilled and low-skilled workers in manufacturing sectors. As in the UK, however, in Sweden, non-ICTintensive services sectors were equally, if not more important than manufacturing in terms of employment at the lowest skill levels by the time of the onset of the financial crisis in 2005. At medium skill and high skill levels, meanwhile welfare sectors continued to dominate in terms of employment. Nevertheless, the primary alternative to employment in public service sectors for those with college educations was in high-end, ICT-intensive services, and the importance of these sectors in generating employment for the highly skilled has increased in Sweden over time, as it has elsewhere. Further, it is wages in these sectors (finance and general business services), along with manufacturing, that have pulled away from the economy-wide average as the Swedish wage bargaining system decentralized (see Figure 3.6). In Germany, too, (see Figure 3.7) the manufacturing sector has continued to prove more successful at creating jobs at the medium-skill level than in the UK— reflecting the continued relative strength of the German manufacturing sector in general. Even in Germany, however, only one-fifth of the workforce with midlevel skills were employed in manufacturing by 2005: instead, workers at this skill level were marginally more likely to find employment either in welfare sectors or in non-ICT-intensive services. At the lowest skill levels, meanwhile, the evolution of labor market performance has been much closer to that of the UK, with the widespread loss of manufacturing jobs associated with a significant expansion in employment in non-ICT-intensive services. This expansion has been accompanied by a significant decline in the relative wage rates of often insecure and un-unionized workers in low-end services sectors, compared with the protected industrial core (Gernandt and Pfeiffer 2007; Hassel 2012; Palier and Thelen 2010; Wren 2021). At the other end of the skills distribution, German college-educated workers, like their Swedish counterparts, remain heavily concentrated in welfare sectors, although in Germany ICT-intensive service sectors have also considerably increased in importance as employers of graduates since the early 1990s (Wren 2021). Figure 3.8. Illustrates the increasing divergence in the welfare of workers in the “winning” and “losing” sectors as the structural transformation of the German ⁷ The EU KLEMS data on sectoral employment by skill level is available for a shorter historical period in Germany and Sweden than in the UK.

56

anne wren High-Skilled 60 50 40 30 20 10 0

0 2 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 97 97 97 97 97 98 98 98 98 98 99 99 99 99 99 00 00 00 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _2 _2 _2 Medium-Skilled 45 40 35 30 25 20 15 10 5 0

0 2 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 97 97 97 97 97 98 98 98 98 98 99 99 99 99 99 00 00 00 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _2 _2 _2 Low-Skilled 45 40 35 30 25 20 15 10 5 0

0 2 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 97 97 97 97 97 98 98 98 98 98 99 99 99 99 99 00 00 00 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _2 _2 _2 Manufacturing

ICT-Intensive Services

Non ICT-Intensive Services

Welfare Services

Fig. 3.5 Cross-sectoral distribution of employment, in percent, by skill level: Sweden (1970–2005).

digitalization and the transition to services

57

(a) 1.8

MANUFACTURING

1.6

WHOLESALE AND RETAIL TRADE

1.4

POST AND TELECOMMUNICATIONS

1.2

FINANCIAL INTERMEDIATION

1

OTHER BUSINESS, RENTING EQUIPMENT

0.6

_1970 _1972 _1974 _1976 _1978 _1980 _1982 _1984 _1986 _1988 _1990 _1992 _1994 _1996 _1998 _2000 _2002 _2004

0.8

PUBLIC ADMIN AND DEFENCE; COMPULSORY SOCIAL SECURITY

(b) 1.2 1 0.8

HOTELS AND RESTAURANTS EDUCATION

0.6

HEALTH AND SOCIAL WORK

0.4

OTHER COMMUNITY, SOCIAL AND PERSONAL SERVICES

0

_1970 _1972 _1974 _1976 _1978 _1980 _1982 _1984 _1986 _1988 _1990 _1992 _1994 _1996 _1998 _2000 _2002 _2004

0.2

Fig. 3.6 Average sectoral compensation compared with economy-wide mean: Sweden (1970–2005).

economy proceeds. While, over time, the average wages of the remaining manufacturing “insiders,” along with finance workers, have risen steadily compared with the rest of the economy (see Panel (A)), average wages in the hotel and restaurant sector have plummeted, while those in the other less ICT-intensive services sectors (personal services and wholesale and retail), in which most of the least-skilled German workers are concentrated, are well below the economy-wide average (see Panel (B)). The data do suggest, therefore, that many of the same tendencies exist in Sweden and Germany as in the UK. However, the effects are muted. Welfare state interventions have a role to play in explaining this divergence. In Sweden,

anne wren High-Skilled 45 40 35 30 25 20 15 10 5 0 0 2 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 97 97 97 97 97 98 98 98 98 98 99 99 99 99 99 00 00 00 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _2 _2 _2 Medium-Skilled 45 40 35 30 25 20 15 10 5 0 0 2 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 97 97 97 97 97 98 98 98 98 98 99 99 99 99 99 00 00 00 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _2 _2 _2 Low-Skilled 45 40 35 30 25 20 15 10 5 0 _1 97 _1 0 97 _1 2 97 _1 4 97 _1 6 97 _1 8 98 _1 0 98 _1 2 98 _1 4 98 _1 6 98 _1 8 99 _1 0 99 _1 2 99 _1 4 99 _1 6 99 _2 8 00 _2 0 00 _2 2 00 4

58

Manufacturing

ICT-Intensive Services

Non ICT-Intensive Services

Welfare Services

Fig. 3.7 Cross-sectoral distribution of employment by skill level: Germany (1970–2005).

digitalization and the transition to services

59

(a) 1.6 MANUFACTURING

1.4

POST AND TELECOMMUNICATIONS

1.2

FINANCIAL INTERMEDIATION

1

PUBLIC ADMIN AND DEFENCE; COMPULSORY SOCIAL SECURITY

0.8

EDUCATION _1970 _1972 _1974 _1976 _1978 _1980 _1982 _1984 _1986 _1988 _1990 _1992 _1994 _1996 _1998 _2000 _2002 _2004

0.6

(b) 1.2 1 WHOLESALE AND RETAIL TRADE

0.8

HOTELS AND RESTAURANTS

0.6

OTHER BUSINESS, RENTING EQUIPMENT HEALTH AND SOCIAL WORK

0.4

OTHER COMMUNITY, SOCIAL AND PERSONAL SERVICES

0.2

_1970 _1972 _1974 _1976 _1978 _1980 _1982 _1984 _1986 _1988 _1990 _1992 _1994 _1996 _1998 _2000 _2002 _2004

0

Fig. 3.8 Average sectoral compensation compared with economy-wide mean: Germany (1970–2005).

for example, public investment in school and college-based education, coupled with a strong vocational training regime, have underpinned strong economic performance across ICT-intensive sectors in both manufacturing and services. And this, in turn, has served both to facilitate continued financial support for high levels of public service employment, and to support demand for non-ICT-intensive services, with less reliance on dismantling protection on the relative wages of lowpaid workers (see Hassel and Palier 2021; Thelen 2019; Wren 2013). In Germany, the vocational training regime has also played a critical role in the country’s continued strong manufacturing performance (Carlin and Soskice 2009; Hassel and Palier 2021; Thelen 2019). In recent years, however, limits on the capacity of the German model to generate sufficient employment have led governments to reduce

60

anne wren

protection on the wages of low-paid workers, facilitating an increased concentration of less-skilled labor in non-ICT-intensive sectors at reduced wage rates, at a cost of increased inequality (Gernandt and Pfeiffer 2007; Hassel 2012; Palier and Thelen 2010; Wren 2021). A complete analysis of the institutional underpinnings of alternative postindustrial growth models is beyond the scope of this short chapter, but for a more detailed discussion, see Wren (2013, 2021) and the various contributions in Hassel and Palier (2021).

Conclusions The collective goal of the contributions in this volume is to initiate a discussion of the implications of digitalization for the politics of the welfare state and inequality. As the editors point out in the Introduction, the first question that we might ask on this issue is whether digitalization does in fact have any implications that are genuinely significant or new. How, if at all, do the implications of the current technological revolution differ from earlier waves of technological change? The second question that arises concerns the nature of these implications. Do we expect the impact of digitalization on economic welfare to be generally positive (the “optimists”’ view), or negative (as assumed by the “pessimists”)? For example, are the dominant effects likely to be an increase in aggregate economic welfare, or the emergence of a new “digital divide” in the labor market and in politics with the potential to undermine the welfare state? In answer to the first of these questions, I have advanced the argument in this chapter that digitalization does have implications for economic development, growth, and welfare, and these are both new and important. As argued by Baumol (1967), many areas of service provision have typically been subject to technical constraints on productivity growth. Given these constraints, deindustrialization, and the associated decline in the proportion of production located in high-productivity manufacturing sectors appeared, at one point, to raise the prospect of a future in which economic growth rates are low (since growth rates will increasingly rely on rates of productivity growth in low-productivity sectors), and distributional battles are increasingly intense (since they are centered on the division of an economic pie which is shrinking over time). The advent of the new technologies has therefore had a truly revolutionary impact, which significantly alters this scenario. By enhancing rates of productivity growth and international trade in certain areas of services, the ICT revolution (along with other factors such as deregulation and the liberalization of services trade) has facilitated the development of new high-productivity growth models that are services led. At the macro level, therefore, (and supporting the “optimists”’ view), digitalization significantly improves the prospects for economic growth and for aggregate economic welfare in an era of deindustrialization. Further, it contributes directly to the radical transformation of the advanced economies from manufacturing to services based.

digitalization and the transition to services

61

As with any radical technological change, however, there is little doubt that the ICT revolution has had disruptive effects on existing labor market norms and outcomes. Some of these effects are already visible, others remain within the realm of prediction or speculation, and many are discussed in this volume. In this chapter, I have focused on one of the most direct and established effects of digitalization on labor markets, which is closely associated with the transformation from an industrial to a services-based society. That is, that the “benefits” of digitalization have not been felt evenly across different skill groups and economic sectors. The ICT revolution has created an increased demand for college-educated workers in the high-productivity, ICT-intensive global services sectors at high rates of pay. In contrast, at lower skill levels, deindustrialization has been associated with a loss of millions of relatively well-paid manufacturing jobs and, in many countries, with an increased concentration of labor in low-productivity service sectors, where the impact of ICT is less significant and wage rates are low in relative terms. This bifurcation of labor market opportunities is clearly visible from the data on the UK—the country within Europe which has experienced perhaps the greatest shift toward services over the past half century. A “pessimistic” view on the impact of digitalization on economic welfare and on welfare politics, therefore, might emphasize that the downside of the new technology is a new digital divide between college-educated workers and “the rest.” This divide is measurable in economic terms by increased inequality. In political terms, it is manifest in growing skillsbased cleavages on issues such as immigration, globalization, and redistribution, which create challenges for any political party aiming to form coalitions across skills groups. The emergence of these cleavages has been highly visible in recent years in debates over Brexit in the UK, and over the election of Donald Trump in the US (the other country that has proceeded furthest along the road of transformation from manufacturing toward digitized services). However, these cleavages have also, of course, been closely associated with the growth in strength of radical right parties in many European countries (see Iversen and Soskice 2019, among others). Whether the “optimists” or the “pessimists” will ultimately be proved correct in their assessment of the overall impact of digitalization on economies and polities, therefore, depends on how the substantial aggregate welfare gains of technological change in facilitating the development of new high-productivity growth models in an era of deindustrialization are balanced against their distributional and political effects. As always, this depends, in part, on the institutional and political capacity of governments to manage the transition in ways that ensure a more equitable distribution of the welfare benefits of the new technologies. Clearly the impact of ICT on labor market outcomes in Sweden and Germany in the 20 years following the onset of the technological revolution was not as extreme as it was in the UK, and welfare state interventions have a key role to play in explaining this divergence. However, it is important to recognize that the cross-national divergence in labor market outcomes described here, also reflects the faster pace of transition

62

anne wren

toward high-productivity services in the UK. Thus, the question as to whether these divergent paths are ultimately sustainable cannot yet be answered. The economic transition toward digitalized services is a dynamic process, which is still incomplete. Its distributional and political implications will therefore be the subject of investigation for many years to come.

References Appelbaum, Eileen, and Ronald Schettkat. 1995. “Employment and productivity in industrialized countries.” International Labour Review 134(4–5):605–23. Appelbaum, Eileen, and Ronald Schettkat. 1999. “Are prices unimportant? The changing structure of the industrialized economies.” Journal of Post Keynesian Economics 21(3):387–98. Autor, David H., Frank Levy, and Richard J. Murnane. 2003. “The skill content of recent technological change: An empirical exploration.” The Quarterly Journal of Economics 118(4):1279–333. Baumol, William J. 1967. “Macroeconomics of unbalanced growth: The anatomy of urban crisis.” American Economic Review 57(3):415–26. Baumol, William J., Sue Anne Blackman, and Edward N. Wolff. 1985. “Unbalanced growth revisited: Asymptotic stagnancy and new evidence.” The American Economic Review 75(4):806–17. Baumol, William J., Sue Anne Blackman, and Edward N. Wolff. 1989. Productivity and American Leadership. Cambridge: The MIT Press. Bessen, James. 2019. “Automation and Jobs: When Technology Boosts Employment.” Economic Policy 34(100):589–626. Bosworth, Barry P., and Jack E. Triplett. 2007. “Services productivity in the United States: Griliches’s Services volume revisited.” Pp. 413–47 in Hard-To-Measure Goods and Services: Essays in Honor of Zvi Griliches, edited by E. R. Berndt and C. R. Hulten. Chicago: University of Chicago Press. Carlin, Wendy, and David Soskice. 2009. “German economic performance: Disentangling the role of supply-side reforms, macroeconomic policy and coordinated economy institutions.” Socio-Economic Review 7(1):67–99. Clark, Colin. [1940] 1983. The Conditions of Economic Progress. New York: Garland Pub. Fisher, George Bernard. 1935. The Clash of Progress and Security. London: Macmillan. Freund, Caroline, and Diana Weinhold. 2002. “The internet and international trade in services.” American Economic Review 92(2):236–40. Fuchs, Victor R. 1968. The Service Economy. Cambridge: National Bureau of Economic Research. Gernandt, Johannes, and Friedhelm Pfeiffer. 2007. “Rising wage inequality in Germany.” Jahrbu¨cher fu¨r Nationalo¨konomie und Statistik 277(4):358–80. Goldin, Claudia, and Lawrence F. Katz. 2008. The Race between Education and Technology: How America once Led and Can Win the Race for Tomorrow. Cambridge: Harvard University Press. Hall, Peter A., and David W. Soskice. 2001. Varieties of Capitalism: The Institutional Foundations of Comparative Advantage. Oxford: Oxford University Press. Hassel, Anke. 2012. “The paradox of liberalization—understanding dualism and the recovery of the German political economy.” British Journal of Industrial Relations 52(1):57–81.

digitalization and the transition to services

63

Hassel, Anke, and Bruno Palier. eds. 2021. Growth and Welfare in Advanced Capitalist Economies. How Have Growth Regimes Evolved? Oxford: Oxford University Press. Inklaar, Robert, Marcel P. Timmer, and Bart van Ark. 2008. “Market services productivity across Europe and the US.” Economic Policy 23(53):139–94. Iversen, Torben, and Thomas R. Cusack. 2000. “The causes of welfare state expansion: Deindustrialization or globalization?” World Politics 52(3):313–49. Iversen, Torben, and David Soskice. 2019. Democracy and Prosperity: Reinventing Capitalism through a Turbulent Century. Princeton: Princeton University Press. Iversen, Torben, and Anne Wren. 1998. “Equality, employment, and budgetary restraint: The trilemma of the service economy.” World Politics 50(4):507–46. Kalwij, Adriaan S., Stephen Machin, Laura Blow, Marijke van Deelen, Frangois Gardes, Maria-Jose Luengo-Prado, Javier Ruiz-Castillo, John Schmitt, and Christophe Starzec. 2007. “Comparative service consumption in six countries.” Pp. 109–40 in Services and Employment: Explaining the US-European Gap, edited by M. Gregory, W. Salverda, and R. Schettkat. Princeton: Princeton University Press. Kongsrud, Per Mathis, and Isabelle Wanner. 2005. “The Impact of Structural Policies on Trade-Related Adjustment and the Shift to Services.” OECD Economics Department Working Papers 427. Lawrence, Robert Z., and Lawrence Edwards. 2013. “US Employment Deindustrialization: Insights from History and the International Experience.” Policy Brief 13-27, Peterson Institute for International Economics (Washington DC) (October 2013). Meidner, Rudolf. 1974. Coordination and Solidarity: An Approach to Wages Policy. Stockholm: Bokfo¨rlaget Prisma. Nath, Hiranya, and Lirong Liu. 2017. “Information and communications technology (ICT) and services trade.” Information Economics and Policy 41:81–7. Palier, Bruno, and Kathleen Thelen. 2010. “Institutionalizing dualism: Complementarities and change in France and Germany.” Politics & Society 38(1):119–48. Rehn, Go¨sta. 1985. “Swedish active labor market policy: Retrospect and prospect.” Industrial Relations: A Journal of Economy and Society 24(1):62–89. Rowthorn, Robert, and Ramana Ramaswamy. 1999. “Growth, trade, and deindustrialization.” IMF Staff Papers 46(1):18–41. Thelen, Kathleen. 2014. Varieties of Liberalization and the New Politics of Social Solidarity. Cambridge: Cambridge University Press. Thelen, Kathleen. 2019. “Transitions to the knowledge economy in Germany, Sweden, and the Netherlands.” Comparative Politics 51(2):295–315. Triplett, Jack E., and Barry P. Bosworth. 2004. Productivity in the US Services Sector: New Sources of Economic Growth. Washington, DC: Brookings Institution Press. Triplett, Jack E., and Barry P. Bosworth. 2006. “Baumol’s Disease’ Has Been Cured: IT and multifactor productivity in U.S. services industries.” Pp. 34–71 in The New Economy and Beyond. Past, Present and Future, edited by D. W. Jansen. Cheltenham: Edward Elgar Publishing. Van Ark, Bart, Mary O’Mahony, and Marcel P. Timmer. 2008. “The productivity gap between Europe and the United States: Trends and causes.” Journal of Economic Perspectives 22(1):25–44. Wren, Anne. 2013. “Introduction: The political economy of post-industrial societies.” Pp. 1–72 in The Political Economy of the Service Transition, edited by A. Wren. Oxford: Oxford University Press. Wren, Anne. 2021. “Strategies for growth and employment creation in a services based economy: Skill formation, equality, and the welfare state.” Pp. 255–90 in Growth and Welfare in Advanced Capitalist Economies. How Have Growth Regimes Evolved?, edited by A. Hassel and B. Palier. Oxford: Oxford University Press.

4 Welfare States, Labor Markets, Social Investment, and the Digital Transformation Werner Eichhorst, Anton Hemerijck, and Gemma Scalise

Introduction Throughout history, technological change has been accompanied by both job destruction and employment creation. In hindsight, the net labor market effect of landmark industrial shifts has been positive, albeit with important differences across time and space. Although past conjectures of jobless growth have thus far proven to be off the mark, this time it could be different (see also the Introduction to this volume). Digitalization, artificial intelligence, and the platform economy will have profound consequences for the quality and diversity of future employment relations, if not on the number of jobs, by massively reducing transaction, coordination, and monitoring costs of employment relations (Weil 2014). Given that current welfare state policies, pension, health, and unemployment benefits were developed and drew on the standard (male breadwinner) model of employment relations of the postwar era, the digital transformation will have profound consequences for welfare provision. In Section 2, we discuss how technological change puts pressure on existing welfare state arrangements, emphasizing the key role of social investment reform—broadly defined—as a policy response to the challenge of digitalization. Section 3 focuses on three countries—the Netherlands, Germany, and Italy— which are taking very different approaches to the adjustments required by the digital and knowledge economy. The section examines how these three countries have pursued reforms, following our conception of the social investment, in the area of human capital “stock” development; labor market regulation to ease the gendered “flow”; and securing contemporary labor and family life-course transitions, and social protection “buffers” to mitigate income volatility. The three countries, all of which share a policy legacy of employment-based social insurance, are experiencing variegated reform trajectories, with the Netherlands following

Werner Eichhorst, Anton Hemerijck, and Gemma Scalise, Welfare States, Labor Markets, Social Investment, and the Digital Transformation. In: Digitalization and the Welfare State. Edited by Marius R. Busemeyer, Achim Kemmerling, Paul Marx, and Kees van Kersbergen, Oxford University Press. © Oxford University Press (2022). DOI: 10.1093/oso/9780192848369.003.0004

welfare states, labor markets, social investment

65

the Nordic example and jumping on the social investment bandwagon as early as the 1990s, with Germany as a latecomer not following suit until the early 2000s, and Italy lacking the endogenous impetus for social investment reform until very recently. The concluding section reflects on how countries that adopted the social investment agenda early on have been more successful in transforming themselves into knowledge economies and digital societies. Social investment reform has inadvertently prepared the way for more effective and legitimate welfare reform options to accommodate potentially more disruptive technological change.

The Changing Nature of Jobs in the Age of Digitalization and Social Investment Reforms

70 60 50 40 30 20 10 0

Norway Finland Sweden New Zealand United States Korea Denmark Netherlands United Kingdom Estonia Canada OECD Belgium Japan Italy Czech Republic Ireland France Turkey Austria Israel Germany Poland Lithuania Chile Spain Greece Slovenia Slovak Republic

European labor markets have constantly been undergoing transformation due to the dynamics of regulation and global integration, as well as permanent structural change. More recently, however, the digital transformation has started to affect job content, business models, and employment levels on a more fundamental level. As discussed in Chapter 2 of this volume, digitalization threatens jobs largely characterized by routine tasks and shifts task structures toward more non-routine tasks, both at high and low levels of skills. Recent research (Arntz et al. 2016; Nedelkoska and Quintini 2018: Figure 4.1) shows that, given the intra-occupational heterogeneity of jobs and the tasks actually performed, the expected job displacement risk might be smaller than originally expected, while job change might in fact be more important. Moves toward jobs in labor-intensive industries characterized by task content that is currently hard to automate imply observable changes, but also further

High risk of automation

Risk of significant change

Fig. 4.1 Comparative estimates of job automation risk in percent, 2013. Source: OECD calculations based on the Survey of Adult Skills (PIAAC) in Nedelkoska/Quintini (2018).

66

werner eichhorst, anton hemerijck, and gemma scalise

shifts within and between sectors and occupations (OECD 2019a). However, given cross-national differences in the industrial composition and the job/task structure within industries, the levels of estimated risks of substitution and job change as well as polarization vary considerably between countries (OECD 2019b). Technological change tends to put particular pressure on traditional medium-skilled jobs (with an above-average share of routine tasks). Consequently, there is a risk of ever deeper labor market polarization to the detriment of medium-skilled occupations deeply embedded in social protection and industrial relations systems, the core pillar of European welfare states. Some countries have already shifted quite rapidly toward more automation-proof jobs, while others still rely more heavily on routine-intensive (industrial) employment and therefore appear to be more vulnerable. The digital transition may also be associated with heavier reliance on both internally and externally flexible types of work, including temporary or freelance jobs and platform work, meaning that the exclusion or inclusion of social protection for the self-employed or hybrid workers becomes even more relevant (see Picot, (Chapter 13) in this volume). As the extent of actual technical change and its implications for employment depend on several parameters, such as institutional regulation patterns, relative prices of capital and labor, and consumer and societal preferences, global scenarios are of limited reliability. Moreover, the impact of technological change goes beyond exogenous forces that governments, workers, and the social partners need to “respond to.” Technological change should also be thought of as endogenous in the sense that technological applications are shaped by the institutional environment of existing employment relations and welfare arrangements. Managing the ongoing transformation toward an increasingly digital economy in an equitable and sustainable fashion touches on important functions of the welfare state, including income protection and social insurance, active labor market policies, education and training, and, more broadly, labor market institutions. Advanced European welfare states share a common legacy, dating back to the “Golden Age” of economic and welfare growth in the postwar decades, when systems were put in place for social protection programs whose aim was to provide industrial workers with ex-post income compensation in case of sickness, injury, unemployment, and for old age, but these welfare states have also long been under pressure to adapt and develop new tools to keep up with changing economies, societies, and labor markets. In fact, over the past two decades, practically all European welfare states have been recalibrating the basic policy mixes upon which they were built to address new social risks of demographic ageing, the feminization of the labor market, and the shift to the service and knowledge society. Since the turn of the century, the notion of social investment (SI) has gained purchase as a novel welfare concept to address these postindustrial economic and social changes in an integrated fashion (Hemerijck 2017). SI reform tilts the welfare balance from ex-post

welfare states, labor markets, social investment

67

compensation in times of economic or personal hardship to ex-ante risk prevention, enhancing people’s opportunity and ability to mitigate social risks before they materialize, while ensuring the high levels of (quality) employment (which are sustainable in the digital era) and income support necessary to sustain what John Myles (2002) has called the “carrying capacity” of popular welfare states. In fact, three complementary policy functions underpin the SI edifice: (1) investing in quality education and training to raise and maintain the “stock” of human capital and capabilities throughout the life course; (2) easing the “flow” of contemporary labor market and life course transitions; (3) providing inclusive social safety nets to serve as income protection and economic stabilization “buffers.” The complementarity of these three policy functions is key to reducing the adverse effects of the spread of the digital economy. Intrusive technological change underscores the importance of lifelong human capital “stock” and the continuous development of new skills over the life course. The emergence of new forms of employment, such as platform work, with growing numbers of de jure self-employed workers who are actually de facto dependent employees, sees policymakers confronted with the predicament of needing to update employment regulation to manage entirely new patterns of labor market “flows,” while at the same time exposing outdated social insurance “buffers” tied to stable breadwinner employment patterns. There is a real need to explore new policy mixes of stock (building human capital), flow (managing labor market transitions), and buffer (providing income protection in times of need) for an entirely new class of workers who are not adequately covered by existing “stock,” “flow,” and “buffer” policies, such as pensions, unemployment benefits, and paid sick leave (Eichhorst and Rinne 2017). There is ample evidence that SI reform is an effective tool for boosting employment, while mitigating inequality. Thanks to their relatively lean welfare states, the US and, to a lesser extent, the United Kingdom (UK) achieve relatively high employment levels at the cost of high inequality (see Figure 4.2; the size of the bubbles is proportional to welfare spending in the respective countries). By contrast, many welfare states in continental and northern Europe—countries where digitalization is progressing fast—have proven capable of reconciling the world’s highest levels of employment with comparatively low levels of inequality (upper-right half of Figure 4.2), and are potentially also better prepared for the future challenge of creating knowledge-intensive jobs, while minimizing polarization. Some big welfare spenders, such as France, do seemingly well in terms of redistribution but have failed to raise employment levels above the Lisbon employment target of 70 percent (the dashed line in Figure 4.2). More worryingly, southern European countries fall short of both objectives: they face both low employment and high levels of inequality, despite substantial welfare spending. For our in-depth case studies, we decided to focus our analysis on two social investment “bandwagon” countries and one “latecomer.” The Netherlands, a country at the forefront in terms of digitalization, adopted a more comprehensive strategic

68

werner eichhorst, anton hemerijck, and gemma scalise 80 Sweden Germany

75

Japan

Employment rate

Uk

70

Netherlands

Australia Estonia

US

Hungary

Ireland Portugal

Luxembourg

Slovenia Poland

France

60

Czech Republic

Austria

Finland

Latvia

65

Denmark

Spain

Belgium

Slovak Republic

Italy

55 Greece

50 45 –0,40

–0,35

–0,30

–0,25

Equality (reverse Gini index)

Fig. 4.2 Employment rate, equality, and welfare spending, 2016. Note: the size of the bubbles in the graph is proportional to welfare spending in each country, measured by government spending on education and social protection. The dashed line indicates the Lisbon employment target (increase employment to or above 70 percent). Source: own figure based on data from OECD.Stat.

approach to welfare restructuring and employment creation with the revitalization of corporatist agreements between the social partners and the government from the 1980s onwards. Germany, which scores above the EU average in the digital economy and society index, had moved toward SI by the mid-2000s; Italy—a laggard in the challenge of digitalization—with its strong traits of the familialist southern European model, has not (yet) moved away from the welfare-withoutwork policy conundrum. As one of the largest European welfare states in terms of spending, it retains a bias toward passive compensation and traditional labor market and social services. As the following case studies will show, countries that adopted the SI agenda early on have been more successful in moving toward digital economies and societies: national strategies for human capital growth, together with public and private investment in education and research and development (R&D), have strongly influenced the different paths of institutional change of the three countries (see Table 4.1).

The Netherlands In terms of labor market vulnerability to technological change, the Dutch labor market seemed quite resilient before the onslaught of the COVID-19 pandemic.

Table 4.1 ICT usage indicators, 2019.

EU-28 (2013–2020) Netherlands Germany Italy

Companies that employ ICT specialists*

Persons in jobs using computers and requiring access to Internet*

Persons with at least basic Internet skills (2020)

Scientists and engineers**

Persons employed in science and technology**

Persons with tertiary education (ISCED) and/or employed in science and technology**

20 26 19 16

55 69 59 50

20 27 24 15

5 7.4 5.4 2.6

22.1 29.9 28.1 17

35.4 42.9 39 24.4

Source: Eurostat, percent of active population; * not including financial sector; ** percent of total population.

70

werner eichhorst, anton hemerijck, and gemma scalise

Its relatively good performance in employment, education, and skills, and inclusive social protection, however, is not simply a matter of policy virtu, but also of sectoral fortuna. The Netherlands is a trading, services-based economy with a relatively small, but highly competitive industrial base (Nedelkoska and Quintini 2018). The Dutch labor market is flexible, underpinned by a strong regulatory framework of gender-balanced flow, supported by comprehensive, if expensive, childcare provision. With respect to the buffer function of the social investment welfare state, the Dutch social security system is based on compulsory unemployment insurance and two universal provisions: mandatory health insurance for the entire population (Zorgverzekeringswet), paid for by each individual, and a basic pension scheme paid out of taxes. Over the years of the Great Recession, as might be expected, the stock function of the Dutch welfare state has been neglected, with Programme for International Student Assessment (PISA) scores falling below the EU average. Historically, the social partners, coming together in bi- and tripartite policy platforms, such as the Foundation of Labour and the Socio-Economic Council, have been strongly involved in the introduction of new technologies ever since the 1950s. The 1970s stagflation crisis, which hit the Netherlands pretty badly, brought industrial policy into disrepute. This was accompanied by a decline in political interest in technology. Over the course of the 1980s and 1990s, the social partners and the government aligned wage restraint and cuts in social benefits geared toward activation with the expansion of flexible, part-time service sector jobs, which boosted female employment (Visser and Hemerijck 1997). Dutch trade unions responded to the sustained erosion of standard employment relations by making sure, in consultation with employers and the government, that nonstandard employment relations were well regulated. With respect to part-time work, this strategy was a resounding success. In the Netherlands, in terms of labor market flow, many employees want to work part-time, especially women working in care, education, and the public sector. However, the normalization of part-time work did not make it cheap for employers. Despite the long-term success of the Dutch polder model, a new fault line thus emerged, in part as an unintended consequence of effective job protection and inclusive social security for part-time and full-time work. From 2004 to 2015, flexible contracts as a share of total labor market contracts rose from 15 percent to 22 percent (CBS and TNO 2016), while at the same time the number of self-employed, own-account workers also grew to over a million out of a working-age population of nine million, the fastest rise in Europe (OECD Gender Entrepreneurship Database). As a consequence, wage dispersion between those in regular employment, including part-timers, covered by the Dutch flexicurity regime, and the number of people in independent work (which was not covered) ballooned. Due to its large financial sector, the Dutch economy was hit especially hard by the Great Recession. Overnight, the Dutch state had to bail out four out of its

welfare states, labor markets, social investment

71

six largest financial corporations. As a consequence, the budget deficit went up from practically zero in 2007 to 5.4 percentage points of gross domestic product (GDP) in 2009, while public debt rose from 42 to 58 percent within a year. Fiscal dire straits made austerity reform, under the auspices of the Stability and Growth Pact (SGP), imperative. Austerity reform was supported by the social partners in different political coalitions. By 2009, the Balkenende IV Cabinet, a coalition of Christian Democratic Appeal (CDA), Partij Van De Arbeid (Labor Party, Netherlands) (PvdA), and ChristenUnie (CU), agreed to respect existing dismissal protection and unemployment benefit duration, in exchange for lower subsidies to childcare for high-income brackets, while trade unions agreed to restrain wages. In the fall of 2010, a short-lived minority coalition of the Volkspartij voor Vrijheid en Democratie (Dutch: People’s Party for Freedom and Democracy; Netherlands) (VVD) and the CDA, supported in parliament by the Party of Freedom, led by the Islamophobic Geert Wilders, came to office. In June 2011, the populist Partij Voor de Vrijheid (Party For Freedom; the Netherlands) (PVV) refused to support the pension deal negotiated with the social partners a year earlier, and the Rutte I government resigned. In the 2012 elections, the VVD and PvdA became the two largest parties and decided to form a new government. In the coalition agreement, the regressive mortgage interest rate tax subsidy, popular with VVD voters, was traded for a relaxation of dismissal protection, a typical PvdA stronghold. Moreover, the social democrats were bent on restoring relations with the social partners, especially the trade unions. After three months in office, on April 11, 2013, the Rutte II administration signed a Social Pact, negotiated in secret sessions between the leaders of the main employer organization Confederation of Netherlands Industry and Employers (VNO-NCW) and the Federation of Dutch Trade Unions (FNV). For the PvdA and the trade unions, a key impetus behind the 2013 social accord was to stem the tide of “excessive” flexibilization of the Dutch labor market. Yet, the VVD and PvdA continued to hold divergent views, especially on platform work. For the liberal VVD, in the digital age, platform work represented a novel entrepreneurial initiative. For the PvdA, own-account work would remain precarious if not covered by inclusive social protection. Although issues of digitalization and the rise of the platform economy were discussed at the level of the tripartite Social and Economic Council (SER), the Rutte II administration was unable to make progress. In June 2019, the Rutte III cabinet, made up of four political parties (VVD, D66, CDA, and CU), finally agreed to a pension pact with the social partners, largely based on the 2010 agreement discussed earlier and secured with a four-billion Euro government investment fund. The retirement age will rise to 67 in 2024, but on a gentler incline than agreed in 2009. In 2018, the Organization for Economic Co-operation and Development (OECD) identified the uncontrolled rise of nonstandard work as a fundamental change to the Dutch welfare state. Subsequently, two leading reports were published on the future of work and welfare in the Netherlands. In 2019, the

72

werner eichhorst, anton hemerijck, and gemma scalise

Scientific Council for Government (WRR) published Het Betere Werk. De Nieuwe Maatschappelijke Opdracht (The Better Job. The New Societal Imperative). On January 20, 2020, a high-level state commission published In Wat voor Land Willen we Werken? Naar een Nieuw Ontwerp voor de Regulering van Werk (What Country Do We Want to Work In? Toward a New Design for Labor Regulating) advocating mandatory social insurance for the self-employed (Borstlap 2020). Notably, both reports touched on digitalization and the need for permanent upskilling. The main conclusion of the Borstlap Commission was that that, over the past two decades, employers have increasingly opted for independent work subcontracting, as (semi-)permanent employment proved more and more costly. According to the commission, the growing share of independent employees in the workingage population is bound to become a drag on Dutch competitiveness. Without using the functional triad of social investment “stock,” “flow,” and “buffer” provision, the report intimates that, if uncorrected, current labor market conditions will result in less inclusive social security buffering, fragmentary and less flexible labor market transitions, and huge underinvestment in human capital. In the view of the Borstlap Commission, in terms of flow, sustainable (semi-)permanent employment relations should reemerge as the dominant norm in the labor market, with a stronger emphasis on improving internal flexibility in employment organizations. In terms of regulation, more transparency is called for across three distinct types of career path: (1) the norm of (semi-)permanent contracts; (2) part-time employment and temporary work, and (3) independent self-employment. The choice between (permanent) employment, temporary work, and entrepreneurship, should be based on substantive grounds, and not driven by tax or regulatory (dis)incentives. It is imperative, in terms of stock, that workers, whether in semipermanent employment relations or not, are provided with resources for lifelong human capital development. To improve overall social resilience, human capital development should be undergirded by a foundation of inclusive social security and income protection for all, independent of career modalities. This implies a further transition from selective “Bismarckian” social insurance principles toward “Beveridgean” public social security for unemployment, sickness and disability, and skill depletion, beyond the public social assistance and basic pension provisions that already exist. The new aspect is that independent entrepreneurs will have to pay into the Beveridgean funds for basic social security for disability and skill depletion. Another concrete recommendation is to contain external flexibility by making temporary agency work more expensive, based on a clear-cut delineation of the “temporary” nature of agency work, whereby the de facto employer should be the legal one. The aim here would be to disincentivize excessive sub-contracting. Before the COVID-19 pandemic, the vocal high-skill segment of the Dutch selfemployed strongly opposed integration into a social security regime for all. As many independent jobs were under immediate threat, the Dutch government has stepped up to soften the blow for freelancers and platform workers. The upshot

welfare states, labor markets, social investment

73

is that the Rubicon is crossed, and the self-employed will be covered by a hybrid Dutch Beveridgean-Bismarckian welfare state, as suggested by the Borstlap Commission. Because of the intrusive austerity drive launched in 2009, the Netherlands took a backseat on social investment reform. Beyond regulatory overhaul, a debate on social innovation in relation to digitalization, along the lines of the German Industry 4.0 initiative, to which we now turn, is sorely lacking.

Germany Comparative estimates of substitution risks caused by technology show that jobs in Germany are highly vulnerable to such a risk. In fact, Germany exhibits one of the highest substitution risks of all OECD countries (Nedelkoska and Quintini 2018). This particularly applies to the manufacturing sector, which continues to be the backbone of the German employment model and is still larger than in many other OECD countries. Hence, while the overall number of jobs is likely to remain more or less stable or even slightly increase in the digital era, profound changes within and between sectors, occupations, and jobs are expected. This raises doubts about the existing organization of work and the sectoral structure which contribute to the high exposure to automation. Furthermore, lifelong learning in Germany, considered a key priority for human capital stock, is institutionally fragmented and biased in favor of better skilled and younger people, as well as company-initiated training provided to core staff. Collective bargaining and firm-level worker participation (codetermination) might help facilitate change, but the scope of both of these mechanisms has been on the decline over the last decade. Large parts of the service sector are not covered by collective bargaining, with the same applying to many smaller firms, while the metal and chemical sectors continue to be strongholds of industrial relations. Finally, the buffering function of a Bismarckian welfare state might be affected by a potential erosion of social insurance funding, in particular if self-employment/platform work increases (although this type of work remains very limited so far). In response to these challenges, the early 2010s were dominated by a statesponsored research and industrial policy which aimed to look into innovative business processes (Industry 4.0), with a focus on the engineering core of the economy (see Buhr and Frankenberger (Chapter 19) in this volume). Only somewhat later, encouraged by trade unions becoming increasingly aware of the challenges to the manufacturing sector, did attention shift toward labor market and social policy issues. This triggered a government-initiated institutional dialogue between the Federal Ministry of Labour and Social Affairs, the social partners, academic experts, and the wider public. The main goal was to explore the need and possibility of modernizing the labor market and human resource and social policies in

74

werner eichhorst, anton hemerijck, and gemma scalise

light of the digital transformation. This was based on broad stakeholder participation, ultimately aimed at stimulating an iterative policymaking process, starting with a Green Paper put forward by the ministry in April 2015, defining a set of policy priorities, and concluding with a White Paper published in early 2017 (Federal Ministry of Labour and Social Affairs 2017). The stakeholders involved identified four main topics: (1) Lifelong learning was considered essential in order to continuously keep up with rapidly evolving technological developments. (2) Flexibility at work and new working time arrangements were discussed in order to increase business flexibility, as well as employee autonomy, while addressing the issue of a potential dissolution of the boundary between working time and leisure. (3) Social protection of the self-employed was perceived as a contentious issue as the lines between employment and selfemployed work are increasingly blurred, such that some actors argued that it was appropriate and reasonable to include self-employed individuals in the statutory pension insurance system alongside employees. (4) Industry 4.0 offers new opportunities to shape work and production processes and to relieve workers of routine activities, but this was seen as a potential opportunity that could only be used to the full by organizing work in new ways and adapting workers’ skills. However, a closer look at actual policy initiatives shows a certain tension between the priorities identified and the reforms implemented in practice over the last few years. Taking a broader and longer-term perspective, we can distinguish two main areas of policy action that continue to be relevant in the digital context: human capital formation, on the one hand, and regulatory as well as social protection issues, on the other. Furthermore, it is useful to consider the duality of the German labor market, divided between a core that is still governed by strong collective bargaining and the margin of the labor market where state policies are more important (see Table 4.2).

Table 4.2 Dualized labor markets and reform activity in Germany. Core labor market (with collective bargaining)

Margin of the labor market

Human capital formation

Employer-funded continuing vocational education and training, extended via collective agreements

Increasing role of public employment agency/ALMP in training for employed people

Regulatory issues

Collectively agreed or company-based arrangements for mobile working, flexible working times etc., reorganization of work

Statutory minimum wage, reregulation of nonstandard work, steps toward expanding coverage of social insurance

welfare states, labor markets, social investment

75

In terms of the core labor market, a publicly supported industrial policy aimed at increasing investment, productivity, and competitiveness through the development and application of digital technologies continues to be combined with company-sponsored training for skilled workers and increasingly widespread collective agreements with training components with a view to maintaining high levels of productivity. This is being complemented by new forms of internal and functional flexibility, such as a more flexible organization of working time, internal collaboration, and new forms of work, partly embedded in sectoral or companylevel agreements, or driven by firms directly. In this segment, only a very limited role is played by legislation or policy intervention, such as new legislation on temporary part-time work for parents or the potential, but still highly controversial, reform of working time legislation. As was the case during the Great Recession, the COVID-19 pandemic has resulted in well-established instruments, such as publicly sponsored short-time work, being used heavily, thus avoiding or at least postponing layoffs. Short-time work is also available to smaller firms and to the service sector, but implementation there is less able to rely on established procedures, and the link between short-time work and training remains quite weak, with the same applying to policies that could help ease the flow from declining sectors or firms to areas with more robust labor demand. As for the margins of the labor market, the last few years were characterized by reregulatory policy reforms correcting some of the deregulatory steps taken in the 2000s, such as the introduction of a statutory minimum wage in 2015 and stricter regulation of temporary agency work in 2017. This can be interpreted as a response to growing public concerns about inequality and “precarity” in the labor market (Marx and Starke 2017). There has been some debate, albeit so far without concrete outcomes, on the boundary between dependent and self-employed work as regards the redefinition of the dependent worker status and/or the inclusion of the self-employed in social insurance. Notably, as a direct response to COVID-19, ad hoc support for freelancers and small companies was made available to help them maintain liquidity, ultimately putting the issue of a contributory unemployment insurance for the self-employed or for those combining different types of income on the agenda. Lastly, in Germany, there is increasing intervention from public Active Labour Market Policy (ALMP) to promote training of employed people, in particular medium- and low-skilled workers in small and medium enterprises (SMEs). However, a stronger institutional base for a more universal regime of lifelong learning is still absent. This latter point illustrates the difficulties in creating a more egalitarian lifelong learning environment in a county with fragmented adult learning. While there has been a broader expansion of childcare and quality improvements in schooling (more or less in line with social investment) in Germany over the last two decades, the realm of lifelong learning is still characterized by fundamental divides between company-initiated training addressing core (skilled) staff, public

76

werner eichhorst, anton hemerijck, and gemma scalise

ALMP generally targeting the unemployed, and a structural neglect of those groups that might be most at risk of skill obsolescence, in particular if they are not employed in firms covered by collective agreements with training components. The first national adult learning strategy, adopted in mid-2019, was the result of a difficult and complex process. The strategy has the potential to lead to better coordination, higher transparency, and more universal access to adult learning, but in terms of concrete implications it remains rather limited. A preliminary assessment shows that the main issues debated in Germany in the Work 4.0 context have long been topics of labor market and social policy, but they have been reframed and given new urgency, motivated by current and imminent technological change and automation. An earlier focus on stimulating innovative production technology was linked to social innovation as actors from trade unions and the area of social policy entered the discourse. The German experience shows that a “flexible” tripartite approach at different levels seems feasible due to shared interest in productivity, innovation, and jobs, as well as the joint interests of both labor and business. But this does not preclude conflicts and stalemate in critical areas, such as the responsibilities for the design, delivery, and funding of continuous vocational training or the regulation of and pension coverage for self-employed work. In fact, while there have been longer-term policy trends toward reregulation of the labor market and more emphasis on education, direct social policy responses to digitalization are hard to find.

Italy As in Germany, in Italy too, the substitution risk due to technology is above the OECD average (see Figure 4.1) and particularly affects the manufacturing sector. Based on SMEs in typical “Made in Italy” sectors, manufacturing is mostly associated with low and medium technology activities and clustered in industrial districts which are characterized by a deep regional divide. The northern “Industrial Triangle” (Milan-Turin-Genoa) is oriented toward capital, high-tech, and knowledge industries, in the northeastern and central regions, family enterprises mostly specialize in low-skilled light manufacturing, while the south relies mainly on tourism, with high levels of informality, and youth and female unemployment. Although digitalization is characterized by sectoral specificities associated with the skills required for particular professions, the employment shares of highskilled workers are growing and a phase of reprofiling of conventional jobs is expected, further increasing job polarization and internal disparities (Cirillo et al. 2019). In recent years, a growing awareness of the need to ensure the regulation of the labor market is in step with the knowledge society and the digital era has driven significant legislative initiatives. However, there are many constraints that

welfare states, labor markets, social investment

77

have put a brake on change, including: weak state-sponsored industrial and innovation policies combined with low private investment in R&D; delegitimized social dialogue; a lack of policy complementarity and administrative capacity. Taken together, these features represent a weak institutional setting for the development of SI responses to digitalization. Such shortcomings have been exacerbated by the prolonged public underfunding of education and research, which contributed to making Italy one of the European countries with the lowest levels of schooling and human capital and with the highest school dropout rate and young people neither in Employment or Education or Training (NEET) (European Commission 2018). Low levels of cognitive skills are combined with skills mismatch and surplus, reflecting ineffective regulation and low demand for skills (OECD 2017). Yet, in spite of very difficult fiscal and economic circumstances in the wake of the 2008 crisis, since 2014 Italy has passed a range of reforms to realign its socioeconomic policy strategy. The government relaunched its consultation with stakeholders to improve the responsiveness and inclusiveness of the labor market and provide the country with the essential technological infrastructure and skills to allow innovation to progress. Recent reforms addressed four main policy domains: (1) Labor market (2014 Jobs Act); (2) Education (2015 Good School Act and 2015 National Plan for Digital Schools); (3) Industrial and innovation policy (2016 Industry 4.0; 2017 Enterprise 4.0; 2020 Transition 4.0, and Italy 2025); (4) Social protection (2019 Citizenship Income Scheme). A neo-voluntarist social dialogue has gone through various stages. Stakeholders have been consulted to finetune measures, but non-institutionalized industrial relations, in a context of political instability, do not allow unions and employers to develop stable institutions and contribute to policymaking, leaving governments acting alone. The 2014 Jobs Act has been particularly heavily criticized by unions due to its introduction of a new open-ended contract with dismissal costs increasing in proportion to seniority and removing the reinstatement provided for in the workers’ statute for dismissals without just cause in companies with more than 15 employees. An attempt to expand social security was also made through a new unemployment benefit scheme (NASpI) introduced to extend benefits coverage to workers with atypical contracts. In 2019, the law was partially reformed and those working in digital labor platforms were included in its scope. A shift toward activation measures was also enforced: the link between benefit conditionality and activation was strengthened and the scope and duration of wage supplement schemes for industrial crises was limited. The National Agency for Active Labour Market Policies was created to harmonize standards and practices. However, territorial and policy fragmentation, combined with weak administrative capacity, greatly reduce the effectiveness of such measures. One example of this is the lack of coordination between the National Institute for Social Security (which manages income support schemes) and regional employment services

78

werner eichhorst, anton hemerijck, and gemma scalise

(responsible for ALMPs), which invalidates the conditionality mechanism. At the same time, regional employment offices are poorly equipped to provide adequate support for job reintegration. The training that is offered is not targeted, being neither linked to job demand or coordinated with firms. The reconciliation of work and private life was also addressed in the Jobs Act: maternity leave was made more flexible and both parental and paternity leave were extended to all categories of workers. Despite these measures, important inequalities persist in terms of employment protection and the generosity and coverage of unemployment benefit. To address digital competences and a shortage of job-related skills, the 2015 Good School Act funded infrastructure interventions to develop ICT-based learning environments (i.e. technological equipment, administrative digitalization, staff professional development) and addressed the lack of cooperation between companies and vocational schools. Inspired by the German dual education system, the School-Work Alternation scheme was developed, making traineeships compulsory in the last three years of upper-secondary education. However, there have been very few concrete initiatives to foster the local implementation of this measure (i.e. support for schools to establish partnerships with firms and workbased learning). Only “virtuous” schools and companies benefited from these policies, while most were not affected, especially in regions where there are fewer companies able to provide high-quality work experience. To facilitate the transition to digital technologies among firms, the 2016 Industry 4.0 Plan set up a network of technological hubs. The aim was to engage a broad range of actors, including large private players, universities, research centers, SMEs, and start-ups to promote the adoption of technologies in key industrial sectors. In fall 2017, the second phase of the plan, entitled Enterprise 4.0, was launched and then expanded in 2020 with the Transition 4.0 program. Incentives were made available for training start-ups and innovative companies using tax credits, and funding for digitalization vouchers for SMEs was increased. Finally, the Italy 2025 strategy for structural transformation was developed to expand digital infrastructures and collaboration between the public and private sector in fostering innovation. These policy packages represent a major effort to stimulate the digitalization of the economy and lead technological change. However, Italy’s investment in R&D is still the second lowest of the EU-15 countries and policies are still primarily based on indirect subsidies and tax incentives to firms, rather than direct state funding, which have limited capacity to promote private investment in skills and innovation (Burroni et al. 2019). Although private R&D expenditure has been on the rise in recent years (in 2018 it reached 0.86 percent of GDP), it remains well below the EU average. Restricted access to credit, low foreign direct investment (FDI), and limited venture capital markets are unfavorable conditions for the growth of R&Dintensive companies. Moreover, the low share of people employed in R&D in both

welfare states, labor markets, social investment

79

the public and private sector and the weak cooperation between universities and businesses slow down the transfer of knowledge and the sharing of risks related to R&D activities (Ramella 2015). In February 2019, a new income scheme, entitled Citizens’ Income Benefit but more similar to a guaranteed minimum income, has been launched. This targets jobseekers and low earners who agree to sign an employment pact declaring that they are immediately available for work. Although beneficiaries are expected to retrain and get back into work, regional employment services complain that they do not have sufficient human and economic resources to offer retraining and effective job matching. Since March 2020, unprecedented economic efforts have been undertaken to guarantee social safety nets and employment-related measures in response to the COVID-19 crisis. The government developed expansionary measures to support the health care system, households, workers, and firms affected by the pandemic (i.e. expansion of the ordinary wage guarantee; income support for workers not covered by any social safety net; dismissal procedures suspended; new income allowances for autonomous and seasonal workers; new parental leave and childcare allowance). Tax payments were suspended, a debt moratorium on bank loans was approved, and public guarantees on new loans to firms were increased. It is too early to evaluate the effectiveness of these measures, but the implementation of these emergency policies alone confirmed the weakness of the administrative system, and two months after the beginning of lockdown, these measures had not been implemented due to institutional layering and a lack of coordination. Despite significant policy reformism, poor implementation capacity and institutional weakness reduce policy effectiveness and efficiency. This is crucial also when it comes to shaping the impact of digitalization. Up till now, innovation and skills policies have been marginal and lifelong learning and policies aimed at facilitating female employment have been neglected. Mario Draghi publicly addressed these weaknesses in his first speech in parliament after his new government had taken office in February 2021, during which he pledged to employ the planned EU recovery fund to speed up plans for the digitalization of the different economic sectors, skills expansion, and the ecological transition.

Conclusion and Outlook Arguably, digitalization and the welfare state jointly saved Europe from economic meltdown when the Covid-19 pandemic struck and continued to smolder in 2020. Thanks to digitalization, most working-age adults were able to shift to working online. In addition, the welfare state proved proficient in buffering the economic costs of extended lockdowns. Supported by accommodating monetary and fiscal policy, precious time was bought to develop effective vaccines. On a less salutary

80

werner eichhorst, anton hemerijck, and gemma scalise

note, the pandemic exposed many of the preexisting fault lines inherited from the Great Recession, ranging from excessive inequality, wage stagnation for frontline public and private service workers, exacerbating fault lines in the work-life balance, especially for working mothers, and diminishing employment opportunities for the millennials. Our overall argument is that the digital revolution has had more of an incremental than a disruptive effect on welfare regimes. In comparative research, welfare state resilience, measured in terms of social protection spending, has often been wrongly understood as policy immobilism tout court. We concur with a more dynamic notion of welfare state resilience, bringing together two critical elements. On the one hand, there is the shock absorption capacity of social security buffers in times of recession. On the other, there is the dimension of adaptive change to more slow-burning social and economic challenges, such as technological innovation and demographic ageing. As a whole, the welfare state serves as an incremental catalyst and facilitator of structural change. Beyond immediate crisis-contingencies and more medium-term functional pressures, welfare regimes, which channel close to 30 percent of GDP across the EU, have a life of their own, shaped by idiosyncratic interactions between political actors, including governments, political parties, interest groups, and individual policymakers, each with their own views on the merits and limitations of digitalization and welfare provision. Our case studies thus illustrate institutionally bounded trajectories of regime adaptation rather than punctuated change. This should come as no surprise, because digitalization is part and parcel of the longer-term trend of the growth of the knowledge economy in ageing European societies. Unquestionably, the COVID-19 pandemic has accelerated the digital transformation of the way we work and organize our lives. Equally, if not more importantly, the disruptive nature of the pandemic has brought home just how imperative competent welfare states and resilient health systems are. Against the backdrop of an existential recognition of human frailty, normative arguments about social fairness across risk groups have been rekindled. And since COVID-19 knows no boundaries, the pandemic has highlighted the need for more effective EU cooperation and fiscal solidarity, policy ingredients that was sorely lacking over the decade of the Great Recession. In summary, the Covid-19 predicament reinforced the need for: 1. inclusive buffer policies in order to minimize the social and economic costs (scarring effects, depreciation of human capital) of high and persistent unemployment in a downturn; 2. gender-balanced flow policies in order to maximize employment and facilitate effective homeworking, together with economic adjustment (e.g. quick and painless reallocation of workers from declining industries to growing ones);

welfare states, labor markets, social investment

81

3. lifelong stock policy commitment to support a high-skill/high-productivity/high-wage equilibrium; 4. an E(M)U governance regime that effectively supports strong, inclusive, and smart welfare states. As the in-depth analysis of our case studies reveal, Germany and the Netherlands have moved in the direction of SI, largely driven by proactive adaptation to the reality of expanding services and a more heterogeneous labor force. With its strong rebound in exports in the aftermath of the global financial crisis, Germany significantly improved its social investment, catching up with the Netherlands, which, after 2009, took something of a backseat on social investment driven by a crossparty commitment to fiscal austerity. Italian policymakers tried to contain the trend toward labor market dualization and improve coverage for those most at risk. However, these reforms were enacted against the background of the sovereign debt crisis, which left Italy little fiscal leeway to upgrade social investment. Here, domestic institutional weaknesses in combination with macroeconomic imbalances related to eurozone asymmetries, and especially the austerity drive proposed by EU institutions in the immediate aftermath of the Greek sovereign debt crisis, arguably weakened the resilience of health and welfare provision before the pandemic struck in northern Italy in January 2020.

References Arntz, Melanie, Terry Gregory, and Ulrich Zierahn. 2016. “The Risk of Automation for Jobs in OECD Countries: A Comparative Analysis.” OECD Social, Employment and Migration Working Papers 189. Borstlap, Hans. 2020. “Commission on the regulation of work. in what kind of a country do we want to work? Towards a new design for the regulation of work.” Retrieved March 11, 2021 (https://www.rijksoverheid.nl/actueel/nieuws/2020/01/23/ koolmees-evenwichtige-analyse-commissie-borstlap). Burroni, Luigi, Alberto Gherardini, and Gemma Scalise. 2019. “Policy failure in the triangle of growth: Labour market, human capital, and innovation in Spain and Italy.” South European Politics and Society 21(1):29–52. CBS/TNO. 2016. Nationale Enquête Arbeidsomstandigheden 2016. Methodologie en globale resultaten. Den Hagg: TNO, CBS, Ministerie van Sociale Zaken en Werkgelegenheid. Cirillo, Valeria, Rinaldo Evangelista, Dario Guarascio, and Meeto Sostero. 2019. “Digitalization, Routineness and Employment: An Exploration on Italian Task-based Data.” INAPP Working Paper 10. Eichhorst, Werner, and Ulf Rinne. 2017. “Der digitale Gestaltungsauftrag.” ifo Schnelldienst 70(7):16–18. European Commission. 2018. Education and Training Monitor 2018 – Italy. Luxemburg: Publications Office of the European Union.

82

werner eichhorst, anton hemerijck, and gemma scalise

Federal Ministry of Labour and Social Affairs. 2017. White Paper Work 4.0. March 2017. Berlin: Federal Ministry of Labour and Social Affairs. Hemerijck, Anton. 2017. The Uses of Social Investment. Oxford: Oxford University Press. Marx, Paul, and Peter Starke. 2017. “Dualization as destiny? The political economy of the German minimum wage reform.” Politics & Society 45(4):559–84. Myles, John. 2002. “A new social contract for the elderly?” Pp. 130–72 in Why We Need a New Welfare State, edited by G. Esping-Andersen, D. Gallie, and A. Hemerijck. Oxford: Oxford University Press. Nedelkoska, Ljubica, and Glenda Quintini. 2018. “Automation, Skills Use and Training.” OECD Social, Employment and Migration Working Papers 202. OECD. 2016. Entrepreneurship at a Glance 2016. Paris: OECD Publishing. OECD. 2017. Education at a Glance 2017: OECD Indicators. Paris: OECD Publishing. OECD. 2019a. OECD Employment Outlook 2019. Paris: OECD Publishing. OECD. 2019b. Under Pressure: The Squeezed Middle Class. Paris: OECD Publishing. Ramella, Francesco. 2015. Sociology of Economic Innovation. London: Routledge. Visser, Jelle, and Anton Hemerijck. 1997. The Dutch Miracle. Job Growth, Welfare Reform, and Corporatism in the Netherlands. Amsterdam: Amsterdam University Press. Weil, David. 2014. The Fissured Workplace. Why Work Became so Bad for so Many and What Can Be Done to Improve it. Cambridge, MA: Harvard University Press

5 The Value and Future of Work in the Digital Economy Marius R. Busemeyer and Ulrich Glassmann

Introduction As discussed in the introductory chapter to this volume, there is a growing empirical literature on the labor market consequences of digitalization and automation (e.g., Arntz et al. 2016; Autor 2015; Ford 2016; Frey and Osborne 2017; see also Busemeyer (Chapter 2) in this volume). Instead of contributing to this empirical debate, this chapter seeks to provide a different, more theoretical and conceptual perspective. The starting point of our analysis is the observation that much of the literature is concerned with the quantitative dimension of employment, i.e., how many jobs are lost and gained in particular sectors of the economy. In contrast, we are more interested in the changing nature of employment as well as the changing value of work, not in purely monetary terms, but rather in terms of the substance of occupations and tasks that individuals engage in to make a living. New technology has already led to the emergence of entirely new job categories (“occupations” such as blogging or influencing) and types of employment (“semidependent work,” a hybrid between dependent and independent work), and this trend is likely to continue in the future. In part, these new occupations and employment opportunities are related to the establishment and maintenance of the IT infrastructure of the digital economy (i.e., the “producers” and administrators of the digital economy, including software engineers, systems administrators, programmers, etc.), and these are often the jobs which feature centrally in debates about the employment potential of the digital economy. However, besides IT jobs, there are (at least) two other categories of jobs with considerable growth potential in the future: social/personal services and creative occupations related to the production of digital content (rather than developing and maintaining the infrastructure of the digital economy).

Marius R. Busemeyer and Ulrich Glassmann, The Value and Future of Work in the Digital Economy. In: Digitalization and the Welfare State. Edited by Marius R. Busemeyer, Achim Kemmerling, Paul Marx, and Kees van Kersbergen, Oxford University Press. © Oxford University Press (2022). DOI: 10.1093/oso/9780192848369.003.0005

84

marius r. busemeyer and ulrich glassmann

This chapter emphasizes the importance of the latter two sectors relative to the first. We posit that, while current debates often center around the employment potential of the digital infrastructure economy only, there is, in the long term, likely to be considerable growth potential in the social service economy, especially in social services provided and funded by the welfare state, although the growth potential of this sector ultimately depends on the political support for the expansion of the welfare state and the revalorization of social services. We further argue that there is also likely to be an increase in employment opportunities in the creative digital content economy. A crucial difference between work in the digital content economy and traditional types of work is that the market price of work (in today’s perspective equated with its value, see the following section) is not primarily determined by investment in education, but rather measured in terms of the attention that “performers” (i.e., producers) and consumers in the digital content economy can attract (Reckwitz 2017). Although the number of “performers” in the digital attention economy is currently still very small relative to overall levels of employment, we believe its eventual growth could trigger a feedback dynamic with regard to the meaning and value of work with significant implications for education systems and the welfare state. The second key argument of this chapter is that a purely “quantitative” perspective (i.e., how many employment opportunities are destroyed, how many are created) fails to acknowledge the “qualitative” dimension of how employment is likely to change in the future. Current and future labor market developments can certainly be conceptualized and described drawing on widely used economic terms and concepts, i.e., with reference to labor market supply and demand. But what are the deeper implications of a likely future scenario, in which a large and growing share of the workforce engages in social and creative tasks and occupations which just a few years ago were simply not considered to be gainful employment, such as playing videogames, commenting on shopping experiences, or providing coaching seminars for those affected by heightened stress levels created by the digital economy via online platforms? Digital technology does not just reduce labor demand in some job categories and increase it in others; it completely changes our conceptual understanding of what activities are considered valuable work by transforming our currently narrow conception of work as formal and dependent employment. This will also shift the priorities of contemporary welfare states toward new forms of risk protection (a strengthening of universalist elements in social insurance schemes or, more radically, the introduction of universal basic income schemes, see chapters by Martinelli and Chrisp (Chapter 10) and Guarascio and Sacchi (Chapter 11) in this volume) and requires education policies to meet the needs of this future scenario—not only in the sense of providing more science, technology, engineering, and mathematics (STEM) education, but also by promoting social and creative skills.

the value and future of work in the digital economy

85

In the following pages, we first provide a brief discussion about how the value of work has been conceptualized and measured throughout the ideological history of the political economy, as well as how recent technological change holds the potential to significantly transform the way we think about categories such as “work” and “gainful employment,” both now and in the future. Subsequently, we take up the discussion of employment growth potential in the three sectors outlined previously: the skill-based digital infrastructure economy, the nonprofit sector of social and personal services, and the attention-driven (creative) digital content economy. Finally, we reflect on how these developments affect the priorities of welfare state and education policymaking, as well as their relevance with regard to inequality.

What Is the Value of Work? From the Industrial to the Digital Economy The value of work in today’s economy is tightly bound to the concept of gainful employment via the formal labor market (Cole 2007; Jahoda 1982). It is due to this very reason that the digital revolution stimulates so much rethinking of our basic understanding of the value and meaning of work. As some traditional occupations are automized and as digitalization transforms the nature of work, formal labor market arrangements, which may have appeared efficient in the past, may become dysfunctional in the digital age. This problem can be analyzed from two angles: first, we can hold on to the assumption that value creation is primarily secured by traditional labor market institutions, formal work contracts, and work relationships situated in hierarchical company structures. Such a perspective is backed by the conjecture from transaction cost economics (Coase 1937; Williamson 1975) that institutions reduce uncertainty and facilitate efficient exchange (North 1990). Moreover, it is in line with established social policy concepts promoting activation geared toward formal employment. Quantifying how many workplaces in manufacturing and services remain after digitalization has reshaped the labor markets of advanced capitalist economies seems a logical consequence of such thinking; and for the majority, the debate on the future of work is still wedded to this perspective. However, from our point of view, a second, alternative perspective on this problem appears more fruitful. We suggest distinguishing between intrinsic and extrinsic value of work to create this alternative perspective (Hackman and Oldham 1976; Jahoda 1982; Ryan and Deci 2000). Extrinsic value comes from successful commodification of labor. Work does not carry extrinsic value because people necessarily want to produce a certain product, but because that product is in high demand. It is therefore identical with what advocates of neoclassical theory (Marshall 2013 [1890]; Walras 2010 [1874])—the founders of today’s mainstream economic thinking—meant by exchange value.

86

marius r. busemeyer and ulrich glassmann

By framing demand according to a purely subjective notion of utility, neoclassical theory promoted a broad understanding of the concept of productive labor, incorporating different kinds of services, from the work of creative artists to social or financial services, into the concept of the formal labor market. The downside of such an approach is that it reduces the value of work to its extrinsic value only— work is valuable to the extent that it can be traded on formal labor markets for a reasonable price. Mazzucato (2018) has argued very convincingly that this understanding of the value of work has given us a flawed idea of processes of value creation because it suggests a false equivalence between pricing and value creation: “Defining everything that commands a price as valuable led to the marginalists’ conclusion that what you get is what you are worth” (Mazzucato 2018: 69). What such a conception of work does not take into account is whether people actually want to produce a product because they enjoy making it, identify with it, or consider it useful or generally valuable for society. This intrinsic value of work is defined by the individual worker’s understanding of the value of the product that they are producing. In Richard Sennett’s words, it is the desire to “do a job well for its own sake” (Sennett 2008: 9). It is of course possible that extrinsic and intrinsic value assessments overlap, but under formal market rules, self-actualization is not the main work incentive; instead, the contractual arrangement removes many of the preconditions for intrinsic rewards—most obviously work autonomy. Thus, one crucial indicator for the relative importance of intrinsic and extrinsic value of work from the perspective of the individual is whether work is executed due to external incentives and control leading to “separable outcomes” (extrinsic motivation) or performed of the individual’s own volition (intrinsic motivation; Ryan and Deci 2000). Typically, this distinction accompanies the differentiation between an activity that is regarded as “gainful employment” and one that is regarded as “leisure activity.” Interestingly, activities which today are typically labelled as “hobbies,” have many of the features which Sennett (2008) describes as being lost in modern work relations. Not only are hobbies pursued due to intrinsic motivation, they are also very often linked to repetitive actions allowing for “slow learning” and “habit,” which are preconditions for good craftsmanship (ibid: 265). We believe that the digital revolution, paradoxically, may promote a return to intrinsic motivation and rewards, and sometimes even the virtues of craftsmanship as conventional forms of work and formal labor market institutions erode. This does not imply a liberation of the workforce from status competition, but we posit that, in principle, digital technology has the potential to promote more self-actualization through work. Figure 5.1 is a stylized representation of the long-term historical process. It shows how first, several processes promoted a shift from self-determined work, either in small-scale skilled craft occupations, private households, or the “informal” shadow economy, to the dominance of dependent work in formal employment relationships, accompanied by the expansion of the welfare state. Second, the

the value and future of work in the digital economy

87

Craft work self-employment

r ia

liz

ati on

Household work

Digitalization

-in

du st

Digitalization

Informal work

Po st

Formal work

Industrialization

Self-determined forms of work

Shadow economy Formal labor market focus of pre-digital welfare Redistributional social policies Dependent employment

Fig. 5.1 Digitalization and the revalorization of self-determined forms of work.

figure also depicts how digitalization might contribute to a partial reversal of this process. Initially, industrialization led to a shift toward dependent employment as a model for value creation. The technological innovations of the early and later stages of industrialism and the division of labor in factory systems, regulating workflows, labor time, output per head etc. promoted a shift toward formalized and dependent work.1 The formalization of labor also contributed to the emergence of the modern welfare state (Esping-Andersen 1999). The political setup of welfare institutions and implementation of redistributive social policies stabilized the role of gainful employment in creating and defining “valuable” work. The socalled “standard worker contract” encouraged workers to trade the intrinsic value of their work for secure entitlement to social benefits. The postmaterialist movement (Inglehart 1971) has attempted to shift the focus from extrinsic motivation to the intrinsic value of work, in line with its emancipatory appeal to redefine gender equality, promote self-determination, and push for social justice in broad terms. However, so far, this process of shifting from extrinsic 1 This stylized depiction of the transformation of forms of work shares some similarities with the analysis by Tilly and Tilly (1998) who refer to one of the analytical dimensions as the “extent of timediscipline.” We label this dimension formalization because time discipline is just one of many rigidities resulting from contractual arrangements in capitalist economies.

88

marius r. busemeyer and ulrich glassmann

to intrinsic motivation for work has not led to a revival of the informal economy, but instead further broadened the scope of formal labor markets. Empirical studies show that this process of post-industrialization is negatively associated with intrinsic work values (Parboteeah et al. 2013). Similarly, earlier calls for gender equality have not eliminated the gender bias in the distribution of housework and care work, but rather fostered the formalization of social services, i.e., the transformation of unpaid housework into formal employment, often as part and parcel of the social investment strategy (Estevez-Abé 2006; Isailovic 2020). We argue that digitalization holds the potential to partially reverse the longterm trajectory of formalization and organizational dependence. For instance, digital technology has the potential to eliminate many work processes perceived as having extrinsic but no intrinsic value, meaning it may not only be considered a job-eliminating force. Furthermore, by allowing for new work arrangements that fall between office and homeworking, digitalization may contribute to the blurring of boundaries between actual housework and some other activity (i.e., “gainful employment”). Home decorating may for instance be taught on YouTube and in this way become marketized. However, the new aspect here is that the direction and extent of commodification is defined by the worker. Self-employment may become a more achievable endeavor for people with limited initial capital endowment if investment costs remain limited to a few digital devices which people use to share content on social media. If these activities become established forms of work, they will certainly revolutionize the way we perceive work, but they will also have consequences for the way we aspire to jobs, knowledge, and careers.

Growth Potential of Employment in the Digital(ized) Economy As we saw in the previous section, what is considered valuable employment has frequently changed over time and very much depends on the theoretical lens used to describe and make sense of economic activities. The purpose of the following section is to reflect more systematically on the implications of the changing nature of employment and work in the digital(ized) economy. In the next section, we will discuss three likely sources of employment growth in the digital(ized) economy and their implications for the welfare state (see Table 5.1).

Digital Infrastructure Economy We start with what we call the “digital infrastructure economy.” This sector encompasses economic activities that are related to the establishment or maintenance of digital technology in the narrow sense, i.e., occupations such as software engineers,

the value and future of work in the digital economy

89

Table 5.1 Sources of employment growth in the digital(ized) economy. Digital infrastructure economy

Social nonprofit economy

Creative digital content economy

(Re)valorization of work

low intrinsic potential

medium intrinsic potential

high intrinsic potential

Growth potential

Short-term

Medium-term

Long-term

Scarce resource

Knowledge/skills

Public resources/subsidies

Attention

Distributional pattern

Skill gradient

Mixed

Winner-takes-all, but not linked to skills

“Development entrepreneurs”

Tech companies

Nonprofits and the state

Consumers/users of platform economy

Type of employment

Mix of traditional and project work

Potentially increasingly formalized work

Market/attentiondriven entrepreneurship

systems administrators, programmers, IT consultants, and hardware producers. Public debates about the future of work in the digital economy are often centered on this sector. However, the future of work in the digital infrastructure economy (narrowly defined) is not the same as the future of work in the digitalized economy, i.e., an economy that has undergone a significant digital transformation. Undeniably, the growth potential of the digital infrastructure economy over time and across countries is significantly determined and affected by the availability and distribution of the scarce resource of knowledge. And, vice versa, as we set out in the following, the growth potential of non-tech related labor market segments in the digitalized economy may be determined by entirely different factors. To be more precise, in a fully digitalized economy, all kinds of employment are more likely to be affected by digital technology than in an industrial economy (Reckwitz 2017), but the extent to which and how they are affected varies—and this is a crucial point. Our argument draws inspiration from a recent paper by Benzell et al. (2015). This paper develops a complex formal model explaining the short- and longterm consequences of technological change on employment (see Collins (2014) for a related, but more critical and non-formal version of the core idea). Simply put, the idea is that technological advances that promise to replace human labor with robots and software algorithms create significant employment growth in the digital infrastructure economy in the short term. Since technology that replaces humans could lead to significant savings in terms of wage costs, the demand for these tools is likely to grow exponentially and tech firms are likely to

90

marius r. busemeyer and ulrich glassmann

meet this demand, operating at the cutting edge of the technology front (i.e., delivering what is technologically feasible). This, in turn, is likely to generate huge demand for high-skilled tech workers specializing in setting up the infrastructure of the growing digital economy. This is, more or less, the state of affairs that many OECD countries are currently experiencing. In the medium to long term, however, the growth potential of the digital infrastructure economy is likely to decrease. For one, the growth potential of economic sectors that are using (rather than constructing) the infrastructure of the digital economy is likely to grow, relatively speaking. To use a crude analogy: The advent of the automobile as a revolutionary technology to promote mobility first created jobs for people to construct the roads, but then economic value was produced by people using cars to facilitate the production of other goods and services, whereas the relative importance of the “road construction and maintenance industry” declined. However, there is an important and critical difference between the current and previous waves of technological change: This time the “tools” that are used and developed to construct the digital economy are increasingly able to (re)construct and improve themselves (as if the road engineers of the past had built machines that would take over the business of road construction in the future). This, of course, implies that today’s software engineers might write software that, in the long term, is able to further develop, update, and improve itself, limiting the input from humans in the process (Domingos 2015). Put simply, even though there is likely to be increasing demand for IT experts to set up the infrastructure in the short to medium term, the importance of the creative part of how these platforms and services are used is growing in importance (i.e., while Netflix certainly hired a lot of software engineers in the beginning, it is now more likely to hire writers, directors, and actors, see also Reckwitz (2017) for a similar argument). Our argument draws inspiration from this idea. The ultimate implications of the digital transformation of the world of work for the welfare state and society are likely to depend on how political struggles about the payoffs generated by the digital economy are resolved. Put differently, from a purely economic labor market perspective, different trajectories regarding the future of work and employment are possible—with entirely different distributions of employment opportunities, wages, and relative degrees of precariousness. The shorter-term policy implications of the growth of the digital infrastructure economy are indeed very much in line with the recommendations of leading pundits and scholars to invest in education, lifelong learning, reskilling, and so on (Colin and Palier 2015; McAfee and Brynjolfsson 2016), with a particular focus on supporting low-skilled individuals. It is quite likely that these measures will help buffer the negative side effects of the advance of the digital infrastructure economy by supporting job creation and

the value and future of work in the digital economy

91

growth. However, if our assessment is correct, these measures may not be sufficient to reverse the long-term dynamics that Benzell et al. (2015) and others refer to. For a more medium—to long-term response to the negative side effects of rapid technological change, it is necessary to reassess and reassert the role of the state in a more all-encompassing manner, which we do in the following.

Social Nonprofit Economy Besides the growth of the digital infrastructure economy, we believe that the social service sector holds considerable growth potential for the future as well. Certainly, the social service sector was around long before the onset of the current wave of rapid technological change (see our previous discussion). Also, our main point is not that technology, as such, will directly transform employment in this sector (although this is happening as well, e.g., with the advent of care robots). Our focus is rather on how digitalization and automation indirectly affect the social service sector in the medium term by promoting rationalization of employment in other parts of the economy, so that the social service sector simply becomes a much more important source of employment as “employer of last resort” when employment opportunities in other parts of the economy decline. A simple reason for this assessment is that research into the automation potential of various occupations has repeatedly shown that occupations that require interpersonal communication, management, and team building skills are less likely to fall victim to routinization and automation (Frey and Osborne 2017). Furthermore, even if certain tasks could be automated from a purely technological point of view (again, think of care robots), it is likely that consumers would be more willing to pay for services delivered by humans rather than robots under certain circumstances. This probably also holds when it comes to other services, e.g., customers in a fancy restaurant would be willing to pay extra for human chefs and waiters rather than robots, although this may not apply to a fast-food restaurant. However, the price to be paid for services is likely to be strongly affected by the prevailing regulatory context (e.g., compare the salary of waiters in Norway and the US, which diverge strongly in spite of the fact that this is essentially the same service in terms of actual tasks). Hence, the state—both as a regulatory actor and as a financier of public social services—plays a central role here. Similar to the previous waves of technological/welfare state innovation, the current wave is likely to contribute to a further expansion of the scope of the welfare state, as already suggested. At the same time, the “boundaries” between public and private will become progressively more blurred, as private actors are put in charge of delivering publicly funded services and as formerly private (voluntary) services increasingly fall under the remit of the welfare state. Hence, to some extent, this is a continuation of previous development trends, but in other ways, it could become a paradigmatic shift in what the welfare state is about. Whether or not the latter

92

marius r. busemeyer and ulrich glassmann

occurs is of course, as previously mentioned, highly conditional on the political context. In what way could the digital transformation trigger a paradigmatic shift in welfare state policies? For one, it could reinforce the ongoing trend toward a socialinvestment oriented welfare state model (Garritzmann et al. 2017; Hemerijck 2018; Morel et al. 2012; see also the chapter by Eichhorst et al. (Chapter 4) in this volume). However, given its focus on skills, the social investment model lends itself more to the promotion of employment growth in the first of our three employment categories—the skill-based digital infrastructure economy. In contrast, employment growth in the nonprofit social economy is less clearly associated with the skills divide. There are high-skilled as well as low-skilled social services, including some that might require more “digital” skills, with others relying more on social and communicative skills. On average, and compared to the skill-biased digital infrastructure economy, employment in personal and social services is less knowledge-intensive and therefore opens up employment opportunities for those who fall through the cracks of the digital infrastructure economy. In other words: the policy challenge is to develop instruments that cater to the potential losers of the digital transformation, but to do so in a way that prevents a large-scale “Luddite” backlash that might make the further adoption of technological advances politically difficult (Frey 2019; see Kemmerling and Gast Zepeda (Chapter 12) in this volume). This requires the creation of employment categories and “occupations” that are valuable from a societal point of view and include intrinsic motivation as a stimulus for career development, i.e., those involving meaningful tasks and activities that allow workers to develop professional identities and either earn a market income with that activity or receive some other form of income support to be able to perform it. Thus, we posit that managing the transition to the digitalized economy will entail the creation of “employment” opportunities for those failing to keep up with the increasing demands of the skills-based knowledge economy. Put crudely, the deeper question is whether individuals who may be regarded as “low-skilled” according to the standards of the knowledge economy, should be forced to accept low-paid and potentially demeaning employment (demeaning in the sense that the job could also be performed by a machine) rather than being supported to pursue other activities (or forms of “employment” in this non-conventional sense) that are currently outside of the domain of the formal welfare state, but may have positive effects on society in general. Examples of the latter type of work could be training the youth football team, helping out with the local fire brigade, giving music lessons to neighbors’ children, or teaching others how to surf, etc. Some of these activities are already offered as “services” on the market (music and surfing lessons, for instance), and in fact, these kinds of service sector jobs have grown significantly in recent years. A non-market way of organizing such social and neighborhood services is the concept of the so-called “timebank,” which is based

the value and future of work in the digital economy

93

on the idea that not money but time must be considered the most important resource for work. Timebanks exist in various communities around the globe in the form of communal associations which give different services a standard value (time), thus eliminating economic status differences. It is likely that the mid- to long-term consequences of further digitalization and automation will continue to drive growth in these sectors, also stimulating experimentation with alternative concepts to standard market arrangements. An important caveat to add here is that the future of employment in personal and social services strongly depends on the extent and type of public regulation— hence, public commitment to regulation and adequate public spending are the scarce resources in this case (in contrast to the skill-based digital infrastructure economy, where skills are the scarce resource). Thus, this issue is directly related to distributive and redistributive struggles about the payoffs of the digital transformation (see Chapter 20 on taxation by Gelipithis in this volume). Implementing the policy recommendations previously outlined in the section on the digital infrastructure economy (i.e., increasing investment in human capital formation) would be one part of the compensatory package that may be necessary to support the losers of digitalization. However, this may not be sufficient if a significant expansion of employment opportunities in the social service economy is required. Hence, some of the payoffs of the digital transformation could be taxed (e.g., via a “robot” or similar taxes) in order to finance the expansion of the public and nonprofit social service sector.

Creative Digital Content Economy Finally, we want to outline a third sector in which employment growth, in particular in the long term and motivated by intrinsic value, is likely to occur: the creative digital content economy. Similar to the discussion in the previous section, we posit that the growth potential of non-technological sectors might actually be larger in the digitalized economy of the future, for the simple reason that social, communicative, and creative tasks are more difficult to automate. And even when they are (there is software that can write music or draw pictures, for example), human beings are simply more likely to appreciate (and pay for) these kinds of activities when they are performed by other human beings rather than a piece of software. The establishment of the digital infrastructure economy, especially what is known as the platform economy, is an important precondition for the growth of the creative economy in the digitalized economy of the future (Reckwitz 2017). Of course, and again similar to the social service sector, the creative economy has been growing for decades as citizens have become more able and willing to pay for creative services. Digital technology, however, is fundamentally changing the economic logic of the creative economy by opening up access to formerly fairly strictly

94

marius r. busemeyer and ulrich glassmann

regulated and protected “winner-take-all” markets. Certainly, the digital platforms that creative workers use to market their activities (YouTube, Spotify, etc.) have a pronounced “winner-take-all” quality, since the value of the platform increases disproportionately with its number of users. However, by lowering barriers to access, these platforms have allowed newcomers to easily tap into a large pool of potential customers/consumers/fans. Again, there is of course still a “winner-take-all” dynamic in the sense that attention is concentrated on a small subgroup of “influencers,” YouTube stars, and artists. However, compared to the previous regime, in which large broadcasting companies and record labels controlled access to the core of the “winner-take-all” market, access is more deregulated and decentralized in the platform economy. Furthermore, the range of activities that are “sellable” in the new creative digital content economy is much broader than the traditional creative industries (Florida 2002), including, for instance, cooking classes, advice on make-up and clothes shopping, gardening, arts and crafts, playing video games for a living, or life counselling. Hence, the boundary between the “creative economy” and the social service sector is, as discussed, undeniably blurry, and the COVID-19 pandemic is likely to contribute to a significant expansion of these kinds of services and activities in the long term. From our point of view, the important change to the previous economic regime is that in the new platform-based digitalized economy, individuals can actually earn a living by offering the previously mentioned services or engaging in creative activities. The striking fact about the new platform-based creative economy is that it seems to be largely divorced from traditional concepts of skilled employment. Certainly, an individual needs a particular skill set to stand out among the millions of potential “influencers” on YouTube and be able to make a living, but this skill set is likely to be quite different from the skills required in the knowledge-based digital infrastructure economy. Hence, members of the younger generation making decisions about their future employment and educational careers face a different set of choices nowadays, namely whether to continue investing in a traditional education or whether to use the time to build up a following on YouTube or engage in some other form of self-determined work. Relatedly, the scarce resource in the platform-based creative digital content economy is not skills in the traditional sense, but attention. Potential suppliers of creative content compete with each other to attract the attention of potential followers; the way in which attention is measured is the number of followers that are accumulated on digital platforms, which can—to some extent—be converted into economic value by selling advertising space, etc. Thus, the value of work in this sector is defined less by the level of skills required for a particular task, but more by the degree of attention that a supplier of creative content

the value and future of work in the digital economy

95

captures. The work is therefore less formal and certainly also allows for more selfactualization, but while the dependence on firm hierarchies declines, the dependence on other consumers who may “like” or “dislike,” give or withdraw attention increases. Much like with opera singing or any social service, value creation in the creative economy does not depend on the societal relevance of the activity but on the preferences of consumers. Consumers do not always spend money, but they always spend their time when they devote attention to online content. This allows (or by the system’s logic: forces) content producers to marketize attention, e.g., by using the attention span of consumers for product placement. But even creating attention itself can be commodified, be it in the form of click farms or firms which allow you to order “crowds on demand” in real locations to make a new shop look trendier or make a political party look as though it is more widely supported. All these kinds of activities can be understood as work. Whether this work carries intrinsic value depends on the degree of self-determination it entails. At some point, professed self-fulfillment may turn into self-exploitation and status competition (see Marx (Chapter 7) in this volume), in particular in the context of the marketized digital attention economy. In the long run, the balance between self-fulfillment and self-exploitation may depend less on the extent to which platform work becomes formalized and regulated, and more on the amount of space the marketized logic of the digital attention economy leaves for the development of individual ideas and creativity. The implications of the growth of the creative digital content economy for the welfare state are twofold. First, assuming that much of the economic activity in this sector is provided in the form of freelance or self-employed work and traded and offered via platforms, the provision of a stable safety net for individuals poses particular challenges for welfare states that are traditionally organized around dependent work (see Chapter 15 by Nullmeier as well as Chapter 16 by Natali and Raitano for a similar argument). The provision of social insurance for platform workers would then either lead to further individualization and privatization of risk and insurance, or promote universalistic approaches that provide citizenship-based types of social insurance. The second implication, more relevant for the education sector, is that a significant expansion of the creative digital content economy could fundamentally alter the relationship between the education system and the labor market. Currently and traditionally, the acquisition of human capital through the formal education system is directly related to access to more or less prestigious and attractive employment opportunities in the labor market. In a radical version of the creative digital content economy, this link could be severely weakened, when the essence of creativity is not about being particularly talented in producing videos or music, but rather about devising new forms of self-marketing, even if this is based on mundane activities such as buying and discussing beauty products and filming

96

marius r. busemeyer and ulrich glassmann

oneself playing video games. On the positive side, this could be a liberating experience for those who are good at self-marketing; on the negative side, it might contribute to a further erosion of educational standards and aspirations, which in the long term is likely to affect the quality of employment in the formal economy. This process would also pose challenges for those not able or willing to engage in self-marketing.

Conclusions and Further Implications for Policymaking Our brief and necessarily incomplete analysis of current and likely future employment trends in the digital(ized) economy has emphasized the need to think about changing employment patterns in a broader sense, i.e., to reflect more on the implications of the changing quality of employment rather than the quantitative aspects related solely to supply and demand of jobs. From this perspective, what should the priorities of policymaking in the welfare state be and to what extent are current policies moving in the right direction? The first issue to highlight here is that the potential contribution of the social and nonprofit sector to employment growth strongly depends on the political will to commit resources to the development of this sector. In a number of ways, the current trend of transforming welfare states in the direction of the social investment model paves the way for the long-term growth of the social service sector. The social investment approach aims at supporting individuals to improve their employability in the skill-based digital infrastructure economy, as well as at expanding employment opportunities in social services, in particular by means of early childhood education and care policies. Whether employment in these growing social services is considered a high-quality alternative to private sector employment in the knowledge-based digital economy ultimately depends on the willingness of policymakers to devote the necessary resources to pay for decent salaries and ensure high-quality working conditions. Similarly, new policy instruments in the platform-based creative digital content economy could support the further expansion of the sector, while preventing the emergence of precarious employment conditions. For instance, as the boundaries between dependent employment and independent self-employment become increasingly blurred, the link between employment status and access to social insurance should become weaker, e.g., by transforming traditional social insurance schemes into citizenship-based insurance schemes (see also Chapter 16 by Nullmeier, and Chapter 15 by Natali and Raitano in this volume). The second issue is education policy. Currently, when discussing how education systems adapt to the challenges of the digital economy, public debates commonly center on questions such as how to improve the provision of digital skills, programming, hard science, and math in schools (see Chapter 17 on education, by

the value and future of work in the digital economy

97

Hartong et al. in this volume). Furthermore, it is believed that using digital media (tablets, laptops, etc.) in schools is a better way of helping students to learn these kinds of skills than using traditional tools and methods. If our argument is right, these kinds of skills would only meet the needs of one of the segments in which employment is likely to grow, namely the skills-based digital infrastructure economy. If employment in communicative, interpersonal, and creative occupations is also likely to grow, however, schools (and universities) need to do a better job of teaching social and creative skills. These skills are often regarded as less important compared to the traditional school subjects and therefore not necessarily at the top of the agenda for school reformers, but—as our argument implies—taking social and creative skills more seriously is likely to maximize the employment potential of social services and the creative economy in the medium- to long-term future. Hence, one priority of education policy should be to think about the digital transformation of education in a broader sense, moving away from a narrow focus on digital skills to include social and creative skills, as well. Pursuing this priority would also support and promote intrinsic rather than extrinsic motivations for work.

References Arntz, Melanie, Terry Gregory, and Ulrich Zierahn. 2016. “The Risk of Automation for Jobs in OECD Countries: A Comparative Analysis.” OECD Social, Employment and Migration Working Papers 189. Autor, David H. 2015. “Why are there still so many jobs? The history and future of workplace automation.” Journal of Economic Perspectives 29(3):3–30. Benzell, Seth G., Laurence J. Kotlikoff, Guillermo LaGarda, and Jeffrey D. Sachs. 2015. “Robots Are Us: Some Economics of Human Replacement.” NBER Working Paper 20941. Coase, Ronald. 1937. “The nature of the firm.” Economica 4:386–405. Cole, Matthew. 2007. “Re-thinking unemployment. A challenge to the legacy of Jahoda et al.” Sociology 41(6):1133–49. Colin, Nicolas, and Bruno Palier. 2015. “The next safety net: Social policy for a digital age.” Foreign Affairs 2015(July/August):29–33. Collins, Randall. 2014. “Das Ende der Mittelschichtarbeit: Keine weiteren Auswege.” pp. 49–88 in Stirbt der Kapitalismus? Fu¨nf Szenarien fu¨r das 21. Jahrhundert, edited by I. Wallerstein, R. Collins, M. Mann, G. Derluguian, and C. Calhoun. Frankfurt A.M.: Campus. Domingos, Pedro. 2015. The Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our World. New York: Basic Books. Esping-Andersen, Gøsta. 1999. Social Foundations of Postindustrial Economies. Oxford: Oxford University Press. Estevez-Abé, Margarita. 2006. “Gendering the varieties of capitalism. A study of occupational segregation by sex in advanced industrial societies.” World Politics 59(1):142–75. Florida, Richard. 2002. The Rise of the Creative Class: And How It’s Transforming Work, Leisure, Community And Everyday Life. New York: Basic Books.

98

marius r. busemeyer and ulrich glassmann

Ford, Martin. 2016. The Rise of the Robots: Technology and the Threat of Mass Unemployment. London: Oneworld Publications. Frey, Carl B. 2019. The Technology Trap: Capital, Labor, and Power in the Age of Automation. Princeton: Princeton University Press. Frey, Carl B., and Michael A. Osborne. 2017. “The future of employment: How susceptible are jobs to computerisation?” Technological Forecasting and Social Change 114(January):254–80. Garritzmann, Julian L., Silja Ha¨usermann, Bruno Palier, and Christine Zollinger. 2017. “Wopsi: The World Politics of Social Investment.” LIEPP Working Paper 64. Hackman, Richard J., and Greg R. Oldham. 1976. “Motivation through the design of work. Test of a theory.” Organizational Behavior and Human Performance 16(2): 250–79. Hemerijck, Anton. 2018. “Social investment as a policy paradigm.” Journal of European Public Policy 25(6):810–27. Inglehart, Roland. 1971. “The silent revolution in Europe. Intergenerational change in post-industrial societies.” American Political Science Review 65(4):991–1017. Isailovic, Ivana. 2020. “Gender equality as investment: EU work-life balance measures and the neoliberal shift.” Yale Journal of International Law, forthcoming. Jahoda, Marie. 1982. Employment and Unemployment. Cambridge: Cambridge University Press. Marshall, Alfred. [1890] 2013. Principles of Economics. London: Palgrave Macmillan. Mazzucato, Mariana. 2018. The Value of Everything: Making and Taking in the Global Economy. Harmondsworth: Penguin Books. McAfee, Andrew, and Erik Brynjolfsson. 2016. “Human work in the robotic future: Policy for the age of automation.” Foreign Affairs 2016(July/August):139–50. Morel, Nathalie, Bruno Palier, and Joakim Palme. 2012. “Beyond the welfare state as we knew it?” Pp. 1–30 in Towards a Social Investment Welfare State? Ideas, Policies and Challenges, edited by N. Morel, B. Palier, and J. Palme. Bristol, UK: Policy Press. North, Douglass C. 1990. Institutions, Institutional Change and Economic Performance. Cambridge: Cambridge University Press. Parboteeah, K. Praveen, John B. Cullen, and Yongsun Paik. 2013. “National differences in intrinsic and extrinsic work values: The effects of post-industrialization.” International Journal of Cross-Cultural Management 13(2):159–74. Reckwitz, Andreas. 2017. Die Gesellschaft der Singularita¨ten. Berlin: Suhrkamp. Ryan, Richard M., and Edward L. Deci. 2000. “Intrinsic and extrinsic motivations: Classic definitons and new directions.” Contemporaray Educational Psychology 25(1):54–67. Sennett, Richard. 2008. The Craftsman. New Haven: Yale University Press. Tilly, Chris, and Charles Tilly. 1998. Work Under Capitalism. Boulder, CO: Westview Press. Walras, Léon. [1874] 2010. Elements of Pure Economics. London: Routledge. Williamson, Oliver E. 1975. Markets and Hierarchies. Analysis and Antitrust Implications. New York. The Free Press.

6 The Data Revolution and the Transformation of Social Protection Torben Iversen and Philipp Rehm

Introduction There is a broad consensus in the political economy literature that a central function of the welfare state is to provide social insurance (Baldwin 1990; EspingAndersen 1990; Iversen and Soskice 2001; Moene and Wallerstein 2001; de Swaan 1988). Underlying most of these analyses is an assumption, often implicit but virtually universal, that social insurance cannot be provided effectively through the market, mainly due to incomplete and asymmetric information (Akerlof 1970; Barr 2012; Boadway and Keen 2000; Stiglitz 1982). But, while this assumption may have applied in the past, the data revolution is making it untenable today. This chapter asks what happens to the politics of social protection and to inequality when information about risks to health, life, employment, credit, and so on, becomes more widely available and shareable. A hint of what is to come can be gleaned from the life insurance market, where ICT is radically transforming the status quo. For example, John Hancock Life Insurance, a major player in the US American market, introduced a policy that calculates annual premiums partially based on data collected by an “activity tracker” that policyholders receive for free when they sign up. These types of devices can track and instantly share (via an app) things like steps and stairs taken, active minutes, calories burned, heart rate, sleep quality and blood pressure, among others. The company marketed this life insurance policy as “an innovative solution that rewards you for living a healthy life. In fact, the healthier you are, the more you can save.”1 And it is not just insurance companies getting in on the action. The leading technology companies—Apple, Alphabet, Amazon, Microsoft, etc.—are all

1 FAQs on the John Hancock Vitality Program (https://www.johnhancockinsurance.com/vitalityprogram/vitality-faq.html, retrieved March 11, 2021). President and General Manager of John Hancock Insurance, Michael Doughty, assures customers: “You do not have to send us any data you are not comfortable with,” though he points out: “The trade-off is you won’t get points for that.” (Siegel Bernard 2015: p. B1). Torben Iversen and Philipp Rehm, The Data Revolution and the Transformation of Social Protection. In: Digitalization and the Welfare State. Edited by Marius R. Busemeyer, Achim Kemmerling, Paul Marx, and Kees van Kersbergen, Oxford University Press. © Oxford University Press (2022). DOI: 10.1093/oso/9780192848369.003.0006

100

torben iversen and philipp rehm

committing huge resources to develop a new data-based health industry, where, inter alia, doctors can interact with AI-enabled databases, and individuals can easily share their information with insurance companies. Similar efforts are underway in credit markets, where detailed information about credit history is linked to a trove of data on income, occupation, residence, credit history, and so on. There is currently no integrated analytical framework to help us better understand the consequences of big data for social policy and inequality. Our chapter seeks to provide such a framework and apply it to the political economy of social protection, with an emphasis on the knowledge economy and due consideration to the role of national political and regulatory institutions.

Theory Akerlof ’s (1970) seminal QJE article “On the Market for Lemons” presents a key reason for the breakdown of markets: asymmetric information and the associated problem of adverse selection. Rothschild and Stiglitz (1976) and Stiglitz (1982) were the first to model the logic for insurance markets, with more recent extensions summarized in Boadway and Keen (2000); Przeworski (2003); and Barr (2012). In these models, individuals know their risk types, but insurers do not. Although the technical details are complex, the logic of the model is as simple as it is compelling. If insurers do not have individual information about risks, they will only know the mean risk in the pool of the insured, which can be inferred from insurance payouts. Based on this mean, the insurer can set a flat premium that is just high enough to settle all (actual and expected) insurance claims. If people have information about their own risks, such an insurance plan will be attractive to everyone whose risks are above the mean; they pay in less than they expect to collect. But the opposite is true for those whose risks are below the mean, because they will end up subsidizing high-risk types. Depending on the degree of risk aversion, some individuals who are not too far below the mean will still buy insurance, but others at low risk will opt out entirely. When low-risk types opt out, the result is that the average risk in the insurance pool rises, which, in turn, will prompt others to opt out, etc. This is the adverse selection problem, which led Akerlof to conclude that there would only be a market for “lemons,” although the process may in fact stop short of complete market failure because people with high risk aversion and moderate risks may stay in the pool despite the fact that they are subsidizing bad risks. In any event, there is market failure in the sense that many who want insurance will end up without insurance or with less insurance than they would like. This underinsured group may include both low-risk types, who opt out, and high-risk types with low incomes who cannot afford the higher premiums as good risks leave.

the data revolution, transformation of social protection

101

The economic analysis usually ends here, and public provision is “explained” as a solution to a problem of private under-provision—i.e. market failure. But politically, efficiency cannot be what drives the introduction of a public system. Historically, public systems were constructed as single pools with a common premium paid through income taxes (or similar income-related contributions), and in such all-encompassing pools, there must be some low-risk types who pay “too much” for their insurance and would want to opt out, just as there is with private insurance. The difference between a public and a private system is that the state can compel people to pay into the system; opt-out is not an option. From this perspective, the welfare state emerged not for reasons of efficiency, but through coercion of the majority. The transition to a public system therefore comes about, assuming democratic decision-making prevails, only because a majority prefers public to private provision. This is not a high bar to cross because in the private market adverse selection usually ensures that those with moderate risk and income will be saddled with very high insurance rates so as to pay for themselves and for those with higher risks, whereas good risks will not pay into the system at all. In the public system, those with median risks will still subsidize those at higher risk, but they will also be subsidized themselves by those with lower risks (and typically higher income). So, asymmetric information creates a political path to the welfare state. This is summarized as scenario (2) in Table 6.1. This account implies that the rise of the welfare state is driven by the middle class, but it also suggests that it is a contentious process. In this sense, it is entirely consistent with “power resource theory” (Esping-Andersen 1985; Korpi 1983; Stephens 1979) because income and risk determine how much an individual contributes and benefits, and income and risk are, in turn, correlated with class. Once placed in a public system, the logic translates into fiscal preferences: those with high income and low risk want to spend less on public insurance than those with low income and high risk. Much of the literature on the political economy of public opinion is concerned with documenting these differences in preferences, and how political parties respond to them. An important qualification, however, is that the degree to which people are divided over social spending depends on how well-informed they are about their own risks. If people are uncertain about where they are in the risk distribution, their prior is close to the mean risk, and support for spending will converge to the mean level.2 As a consequence, the greater uncertainty, the less 2 If i receives signals about his or her true risk, pi , from a noisy environment in which the overall mean is p̄ then poi = α · psi + (1 − α) · p̄ , where psi is a signal drawn from a distribution that is centered on the individual’s true risk (pi ) and α is a measure of the “precision” of that signal, which in our model, equals the private information available to i (a formal proof can be found in Iversen and Soskice (2015), Appendix B.). In the extreme case of no information (α = 0), poi = p̄ , so the range is zero; at the other extreme of complete information, poi = pi , the range equals the difference between those with the lowest and those with the highest risk.

102

torben iversen and philipp rehm Table 6.1 Information and social insurance. Distribution of information Asymmetric Level of information

Low

(1) High uncertainty case (Consensus)

High

(2) Classic adverse selection case (Polarization)

Symmetric

(3) Efficient market case (for majority able to pay)

polarization in preferences. A very simple way of expressing this general insight is that class conflict increases with information. This echoes Rehm’s (2016) observation that homogeneity in the risk distribution reduces conflict over the welfare state, but here homogeneity is induced by lack of information—the distribution of actual risk does not change. This is illustrated by scenario (1) in Table 6.1. Of course, some will be well informed about their risks, but in the early phase of welfare state development it is reasonable to assume that the vast majority only had limited information. In such a world of uncertainty, as long as a public system forces at least some (known) low-risk types into the insurance pool, the welfare state is likely to enjoy widespread cross-class support. We may see this as the foundation for the “Golden Age” of welfare state expansion. Scenario (3) in Table 6.1 is one in which both insurers and private individuals have complete information. Akerlof and Stiglitz did not discuss this case, since they were interested in exploring the consequences of private information. However, the symmetric information case is important to our story. Even when privacy protection limits the ability of insurers to acquire individual information, as is the case with medical records, for example, people may choose to share that information. The example in the introduction of using monitoring devices to reduce insurance premiums provides the intuition: good risks pay lower premiums if they can share their type. This logic applies to the important area of private health data where the level and credibility of information has vastly improved over time. There are three related forces driving this trend. First, the general advance of medicine has increased the detail and reliability of diagnostics (Shojania et al. 2003). Second, the explosion in the number and variety of tests that can be done by certified laboratories has made it possible to share this information credibly; DNA diagnostics in particular promises to offer an order of magnitude more information about health risks than

the data revolution, transformation of social protection

103

in the past. Finally, computing power combined with AI has made it possible to classify individuals in risk groups much more accurately than in the past. The fact that individual information can be acquired by, and credibly shared with, would-be insurers mitigates the asymmetric information problem and opens up the possibility of insurance being provided efficiently through the market to those who are able to pay. For each group of individuals with identical risk profiles, there would be a separate insurance pool/plan with its own cost and replacement rate. Realistically, there is likely to be some modest risk heterogeneity within groups that is unknown to the insurer and therefore pooled. As long as insurers have enough information to distinguish members of different groups, we would get a series of distinct insurance pools and plans. In this brave new world of near-complete information, there would be an effectively functioning market for the “creampuffs”—people with low risks that insurance companies crave. In fact, anyone with risks below the mean would be better off in such a world, assuming that private provision is no less efficient than public provision.3 Again, this is because, in a public system, all those with belowaverage risk subsidize those with above-average risk. Another implication is that those with the highest risk, who also tend to have the lowest incomes, may be unable to afford private insurance. For example, low-income people at serious risk of diabetes may be unable to effectively insure against that risk if is known to insurers. In this sense, the market is obviously not efficient, but it may work efficiently for a majority. Of course, high-risk types would want to protect their privacy. Some individuals with lower risks may also want to do so on principle, but this will be costly because “refusers” are automatically placed in the same pool as high-risk types, driving up their premiums. Everyone in the pool with risks below the group average will therefore have a financial incentive to divulge their information to get a cheaper plan. But if they do share their information, the same will be true for those with risks below the mean in the remaining pool, and so on. This process will continue until all the “lemons” have been called out. This is Akerlof ’s logic in reverse because the result will now be a segmented private market for risk where all information is common knowledge, and with an uninsured group of high risks who cannot pay. It is clear from this analysis that privacy laws are not sufficient to remedy the problem of information that can be credibly shared. Would there be majority support for privatization? Based on our assumptions, the answer may be yes if the risk distribution is bottom-heavy. This is true in the case of health risks,⁴ and it is ordinarily also true in the case of unemployment 3 This is of course an issue of considerable debate. Suffice it to say here that the assumption helps to home in on the effect of information on the direction of change in distributive politics. ⁴ While there is no direct data on risk, health spending is highly concentrated. In 2009, about half of US health care spending went to just five percent of the population (National Institute for Health Care Management 2012). Of course, much of this spending is on the elderly, and everyone grows old,

104

torben iversen and philipp rehm

risks (Rehm 2016). In an “up or down” vote, self-interested voters may therefore support privatization. Again, this conclusion only holds in a world of complete and shared information, and there are many complicating factors, most notably the relative cost of private insurance compared to a single-payer system. Our main argument is that big data will make the public system increasingly contested, and it could give way to private alternatives or market-conforming reforms of public systems. Broadly speaking, our argument implies that the information revolution results in pressures toward more market-conforming provision. Pressure to marketize is not the same as privatization, however. First, markets can be preempted by allowing greater differentiation in the public sector. In this case, we can think of private markets as having an effect through “shadow prices”: the public sector mimics the private in terms of choice and prices. This would imply policies that allow more differentiation and choice in the public system, use of credits for supplementary private insurance, cuts in benefits for high-risk groups (such as refusal to cover procedures for obesity), and introducing high copayments (which are highly regressive). Second, pressures for marketization will be tempered by governments using measures such as “price non-discrimination” clauses to rule out private markets. These prevent insurers from offering better deals to those with lower risk profiles. Since parties on the left of the political spectrum tend to represent constituencies that are lower income and higher risk, left governments are more likely to promote such policies. Uncertainty about the cost and effectiveness of private alternatives can also deter voters from supporting privatization, and the exchange of certain benefits for uncertain gains is known to cause resistance. When markets are blocked for any reason, increased information instead causes polarization in preferences regarding the level of public provision and the distribution of the costs. This polarization can intensify in the presence of effective private options because such options make the feasibility of markets clear—even if it is hard for individuals to take advantage of these options because of the double payment problem (paying for private plans, while also being taxed for public ones). Our argument implies the following two hypotheses, which we will explore in the next section: H1: As information and credible sharing of information improves, private insurance markets will rely on such information, when it becomes available, for purposes of risk classification and premium calculation.

so people must worry about insurance as they age. We consider this issue below. During pandemics, the health risk distribution may be top-heavy.

the data revolution, transformation of social protection

105

H2a: Where private insurance markets exist, they expand as information and credible sharing of information improve. H2b: The expansion of private insurance markets will be more rapid under right than left governments.

Empirics In this section, we explore two implications of our theoretical framework. To start with, we illustrate the basic logic of our argument by reporting some trends in private insurance markets that are the result of increased information, and the emergence of the technology to credibly share it. We then test the hypothesis that more information will lead to an expansion of insurance, using the life insurance market as an example.

Individualization of Life and Health Insurance Car insurance is a good example of how information technologies can radically transform an insurance market. A key problem in this market is asymmetric information: driving behavior is a major factor in accident proneness and should therefore be used in premium calculations, but insurance companies have traditionally been unable to monitor it. Advances in information technology have alleviated this asymmetric information problem, which radically transforms the status quo. In particular, insurance companies can now use tracking devices that collect— and instantly transmit—data on driving behavior, such as distance driven, acceleration, braking events, cornering forces, speeds relative to speed limits, and so on. This enables insurance companies to offer individualized insurance premiums that are directly tied to observed driving behavior, and there are now more than a dozen companies in the US offering these “pay as/how you drive” policies. These plans are attractive for safe drivers which insurance companies hope to sign up. Similar developments can be observed in the life and health insurance markets. In these domains, tracking devices are one tool for credible information sharing, with genetic testing another. Information obtained through these channels allows companies to offer individually targeted insurance plans, based on detailed risk classification. In this business model, insurance companies worldwide team up with firms like Discovery Limited which develops wellness programs branded as “Vitality – a wellness solution that changes the way insurance works.” Vitality uses data on consumer behavior, which are collected by fitness trackers (such as Fitbit,

106

torben iversen and philipp rehm

Jawbone, Misfit, Apple Watch) and transmitted to the company or insurer. Additional information, such as purchasing data, is sometimes collected as well. This detailed, constant, and instant tracking of consumers is useful for health and life insurance companies alike, as explained on the company’s (now defunct) website: Insurers traditionally use risk rating factors to access and underwrite risk. These include age, gender, socio-economic status as well as smoker status and medical history. These risk factors are mostly static and offer a limited view of a person’s risk. A person’s health behavior, however, provides a more accurate risk indication. Vitality, with its 17 years of wellness experience, data and understanding of wellness behavior, adds an additional dynamic underwriting rating factor. It takes into account the impact of chronic diseases and lifestyle factors, such as smoking, level of exercise, diet, alcohol consumption, blood pressure and cholesterol on a person’s risk profile. By integrating Vitality with insurance products, we have developed a scientific and dynamic underwriting model that uses highquality data about a person’s health, wellness, credit card spending and driving behavior to assess their risk more accurately over time. This results in: better benefits, lower and more accurate risk pricing, better selection, lower laps rates, [and] better mortality and morbidity experience.⁵

Tracking devices (“wearables”) are increasingly common in the life insurance market. And health insurance coverage plans tied to these tracking devices— frequently in combination with workplace wellness programs, which often perform health risk assessments and biometric screenings—are gradually being rolled out as well. At one point, policyholders with Oscar Health Insurance in the US, for example, received a free step tracker and could earn up to US$1/day for taking a particular number of steps. In a similar vein, health insurance company UnitedHealth and chipmaker Qualcomm have teamed up to develop a wearable device tied to a coverage plan that incentivizes health behavior by paying up to US$4/day to a covered employee and their spouse if they reach certain targets.⁶ The new tracking systems have become most widespread on the life insurance market and include John Hancock Life Insurance (US), Prudential’s Vitality Health (UK), AIA Australia life insurance and MLC On Track (Australia), and Generali (Austria, France, Germany). But they are equally useful for health insurers, and currently more than 100,000 employees at an undisclosed number of employers in a dozen US states are using wearables through UnitedHealth plans. ⁵ https://web.archive.org/web/20,160,322,151,205/http://www.vitalitygameon.com/vitalitygameon/ (retrieved March 11, 2021). ⁶ Technological progress makes tracking devices ever more sophisticated. One example is the Kolibree toothbrush whose 3-axis accelerometer, 3-axis gyrometer, and 3-axis magnetometer can decipher detailed subtle movements in order to provide real-time feedback that gets transferred to the brusher’s smartphone via Bluetooth, and from there can be shared with a dentist. Of course, it could also be shared with a dental insurance company.

the data revolution, transformation of social protection

107

Yet, this is surely only the beginning. Big Tech is committing huge resources to the advancement of a new data-based health industry, using a variety of related strategies. Apple and Alphabet are developing new tracking technologies—for heart rhythms, sleep patterns, circulation, and even blood sugar levels—using wearables (including a smart contact lens), and microchips can now be implanted directly into the body to constantly monitor health and relay information. Alphabet has recently created a new research unit, called Verily Life Sciences, to develop these technologies using AI-based approaches to data analysis, and Microsoft’s Healthcare NeXT is focused on collecting huge amounts of individual data from a variety of sources and transferring it to cloud-based systems, including a virtual assistant that takes notes at patient-doctor meetings using speech recognition technologies (Singer 2017). As we will document below, the number of reliable tests that can be conducted by independent laboratories has also greatly increased over time, and these data can be combined with data from the new tracking algorithms to produce detailed profiles of individual health parameters with enormous predictive power. The promise of “personalized medicine” is based on such individual information, and former US President Obama’s Precision Medicine Initiative reads like an impassioned call for more data on people’s underlying health risks—“including their genome sequence, microbiome composition, health history, lifestyle, and diet.” As AI crunches these numbers, much greater risk differentiation in insurance policies becomes feasible, which in turn expands the reach of markets, which is ultimately likely to result in greater inequality in coverage and cost. An extreme scenario would be if advances in technology and medical knowledge were to enable companies to predict someone’s medical future with great accuracy, rendering bad risks uninsurable in private markets, while offering good risks a whole range of attractive options.

Information and Private Market Penetration: Life Insurance To explore the effect of information on private market penetration, we turn to information about health and the development of life insurance markets. Apart from modest programs for survivor’s (widow’s) pensions, the public system offers no life insurance. This is therefore an obvious area of potential private expansion as more medical information becomes available that can be credibly shared. The principle of life insurance is very simple: people pay a predetermined monthly premium as long as they live, and the insurance company pays a predetermined amount to survivors when the policyholder dies. If the policyholder dies before the cost of the payout (adjusting for interest) is covered, the insurer loses, while the family of the insured gains. For the insurer to calculate the insurance premium it is therefore essential to be able to calculate life expectancy of potential

108

torben iversen and philipp rehm

policyholders accurately. The adverse selection problem is obvious in this context because people will buy plans that assume they will live longer than they themselves expect to live. The first life insurance plans were restricted to pools where members shared so many traits that their life expectancy could be calculated with great accuracy. The Scottish Presbyterian Widows Fund, commonly credited as the world’s first modern life insurance scheme, was restricted to Scottish Presbyterian clergymen—a very homogenous group with high entry barriers and an average life expectancy that was easy to calculate from carefully kept church records. Modern life insurance schemes rely on information about individual health, complemented by demographic and related data. The expectation is that better information regarding health risks leads to larger life insurance markets. Our dependent variable is life insurance market penetration measured as a ratio of direct gross life insurance premiums to gross domestic product (GDP). This measure was developed by the OECD and “represents the relative importance of the [life] insurance industry in the domestic economy” (OECD 2015). We have data covering 22 advanced economies for the period around 1983–2013. (The Appendix provides more details on the data, as well as methods, variables, and results). The key independent variable is private information that can be credibly shared with insurers. We do not, of course, have direct access to private information, but such information is reflected to some extent in the availability of diagnostic tests. Accurate tests by independent laboratories are one element of what insurance companies need to distinguish risk groups, and such tests—based on blood, saliva, urine, tissue, and increasingly also genetic samples, as well as CT and MRI scanning—have become much more common, accurate, and affordable. A striking example is the cost of sequencing the human genome, which has dropped from about US$300 million in 2001, US$1,000 in 2014, to less than US$200 in 2018.⁷ Correspondingly, the number of personalized gene-based diagnostic tests and treatments available in the US has also risen from 13 in 2006 to 113 in 2014 (Personalized Medicine Coalition 2014). The number of standard tests which can be carried out on a simple penetration blood sample increased from 130 in 1992 to 319 in 2014 (Pagana and Pagana 1992; Pagana et al. 2014). We can also trace the development of diagnostic capabilities via an authoritative and widely used indexing system for diagnostic tests operated by the National Library of Medicine. It maintains a list of 27,000 or so “Medical Subject Headings [MeSH]” (Coletti and Bleich 2001) that are designed to map the entire biomedical field based on English-language academic journals. The MeSH classification includes a hierarchical tree structure where one sub-branch indexes terms related

⁷ According to The Economist (March 14–20, 2020, p. 5), “the first genome cost, by some estimates, $3 bn.” Moreover, the costs for sequencing a human genome have been falling faster than Moore’s law (The Economist, March 14–20, 2020, p. 8).

the data revolution, transformation of social protection

109

to “Diagnosis” (E01). In 1971, there were 277 index entries; there were 450 in 1981, 600 in 1991, 701 in 2001, 914 in 2011, and 1,067 in 2014. Tests can be used to predict life expectancy by disease, and the more tests are conducted, the greater the accuracy of these predictions. The World Health Organization (WHO) collects detailed data on mortality by cause and calculates the “Potential Years of Life Lost” (PYLL) for each major cause of death (cancer, cardiovascular deceases, acquired immune deficiency syndrome (AIDS), etc.). PYLL is the absolute difference between how long people diagnosed with a particular disease actually live and the average life expectancy (weighting deaths occurring at younger ages more heavily).⁸ PYLL has become more detailed and accurate over time, it varies by country and year, and it is available for a broad set of countries. The business of life insurance is to predict life expectancy, and PYLL is precisely the information they need to estimate expected payouts for people with particular conditions. For healthy buyers, insurers will have to rely on diagnostic information that is predictive of such conditions, and we do not have access to this information as researchers. Yet, PYLL carries useful indirect information about risks. This is because accurate and timely diagnosis is a necessary condition for effective treatment, and therefore for a lower PYLL. For example, hereditary amyloidosis is a condition that is caused by an inherited genetic mutation, which can be identified through DNA testing long before symptoms arise.⁹ Once symptoms appear, there are blood and tissue tests that can pinpoint the exact form of the disease, which in turn decides treatment. Most who are diagnosed with hereditary amyloidosis eventually die from heart or kidney failure, but early detection and treatment— ranging from a strict diet to drugs and even liver transplants—create a wide PYLL range. Needless to say, a late or inaccurate diagnosis increases PYLL. In general, better diagnosis should be negatively related to PYLL, and this is indeed what we find when we regress the MeSH data on the PYLL series, along with a set of control variables.1⁰ Better diagnostics leads to better treatment, which reduces premature death. Hence, PYLL is also a good indicator of underlying risks that are not directly observed as a disorder.11 Specifically, if countries where people die earlier from particular diseases have fewer diagnostic tests available then this will limit the scope for the life insurance industry to flourish because it depends entirely on an established infrastructure of laboratories, testing technology, ⁸ A low PYLL is, however, not a necessary condition for the availability of information because some diseases, especially soon after they are discovered, are not treatable even if they can be accurately diagnosed (AIDS was a case in point). ⁹ A low-cost DNA sequencing service such as 23&Me tests for a common variant. 1⁰ Controls are health insurance coverage (percent of population), total health expenditures (percent of GDP), and economic growth rate. 11 The assumption is that if good diagnostics is a necessary condition for treatment; ipso facto effective treatment (fewer years of life lost) is a sufficient condition for accurate diagnostics. We realize that this will be a noisy indicator since some diagnoses may not be followed by treatment, and some treatment may make more effective use of information. But as long as the variance is not systematically related to our dependent variable (market development) it will only bias our results toward zero.

110

torben iversen and philipp rehm

and expertise to estimate individual life expectancy and therefore limit adverse selection. Our assembled dataset contains 486 country-year observations, covering 22 countries over the period from the early 1980s to the 2010s. Year coverage varies by country, giving us an unbalanced cross-section time series dataset. To emphasize the dynamic nature of our argument and data, we estimate an error correction model (ECM) with panel-corrected standard errors and with an AR1 autocorrelation structure. The estimation results are illustrated in Figure 6.1 (detailed results can be found in the Appendix). We find that going from the lowest to the highest level of information (0 to 1) raises life insurance market penetration by an average of about 4.5 percent in the first year, and by about 8 percent in the long run. This substantive effect is indicated by the solid upward sloping line in the left panel of Figure 6.1. The left panel also includes separate estimates for countries with frequent left and right governments (H2b). The effect of information on life insurance penetration is much stronger in countries with frequent right governments (p95, top line), whereas it is muted in countries with frequent left governments (p5, bottom line). The right panel of the figure shows the difference between these lines (also based on Model (3)), which is substantive at high levels of information, and statistically significantly different from zero throughout. Left governments (rightly) worry about the expansion of life insurance markets largely because they could be a trojan horse for the expansion of private health insurance. While there are no readily available data on the regulation of the life insurance industry, a typical restriction

15 10 5 0 –5 Information 90% Cls Right party dominance

20 15 10 5 0 –5

0. 1 0. 2 0. 3 0. 4 0. 5 0. 6 0. 7 0. 8 0. 9 1. 0

Predicted life insurance penetration

(b) 20

0. 1 0. 2 0. 3 0. 4 0. 5 0. 6 0. 7 0. 8 0. 9 1. 0

Predicted life insurance penetration

(a)

Information No partisanship Left party dominance

90% Cls Difference right vs. left party dominance

Fig. 6.1 Predicted life insurance penetration. Note: simulations are based on Models (1) and (3) in Table 6.2 in the Appendix.

the data revolution, transformation of social protection

111

is that insurance companies cannot use genetic information to set premiums. This creates an adverse selection problem for insurers. The quantitative results are clearly only suggestive—we do not claim to have identified causal effects—but they lend credence to the proposition that the increased availability of diagnostic testing has facilitated life insurance markets, barring regulations designed to counter this trend.

Conclusion Are the developments in the private insurance market—fueled by the information revolution—a harbinger of what is to come in social insurance? Will the information revolution undermine majority support for public social policy programs, lead to widespread privatization, and end the welfare state as we know it? Our argument implies that as more information can be credibly shared with insurers, private markets in social insurance are becoming more feasible. However, as mentioned, pressure to marketize is not the same as privatization, and some of the following factors are likely to prevent private insurance, even in the presence of extensive and symmetric information. First, markets can be preempted by allowing greater differentiation in the public sector. We have not presented data to explore this hypothesis, but there is ample evidence that this is in fact the case, albeit to different degrees depending on government partisanship (Gingrich 2011; Hacker 2004). Second, pressures to marketize will be tempered by governments using measures such as “price nondiscrimination” clauses to rule out private markets. These can prevent insurers from offering better deals to those with lower-risk profiles. Left governments are more likely to pursue such policies because they represent constituencies that are lower income and higher risk. When markets are blocked for any reason, increased information will tend to polarize preferences regarding the level of public provision and the distribution of the costs. This polarization can intensify in the presence of effective private options because such options make the feasibility of markets clear—even if it is hard for individuals to take advantage of the options because of the double payment problem. Third, correlated risk can undermine private insurance markets because insurers cannot rely on average risks to set premiums that equal payouts in expectation. Consequently, in areas where correlated risks are typical—such as the domain of unemployment12—social insurance may continue to be the only feasible option. In 12 There are, however, unemployment systems that have features of private markets (Sweden is a good example).

112

torben iversen and philipp rehm

these cases, increasing information should lead to more polarized attitudes (and demand for segmentation within a public system), but not to privatization. Fourth, time inconsistency problems arise in some insurance domains, and the state is able to address these more credibly than private companies. In commercial insurance, younger, healthier, and more employable workers “vote with their feet” by simply leaving the insurance scheme. This is analogous to the aforementioned adverse selection problem, but the difference is that these workers would prefer to stay and would do so if the insurer could make credible commitments to future benefits. Private insurers are not generally in a strong position to be able to solve this problem because they lack coercive powers, and instead they focus on product markets where the problem does not arise in the first place. In homogenous risk pools, the time inconsistency problem is muted for risks that are fairly stable over time, and where people continuously buy insurance so that they are covered for the next insurance term. For example, home insurance covers risks that usually do not change much from one contract period to the next, and consequently there is no implied transfer (in expectation) to others. Time inconsistency becomes a serious problem when large transfers are required across generations because younger generations know that they are unlikely to need the insurance until they grow old. This is obviously true of PAYG old-age insurance, but it is also true of health insurance because poor health is more prevalent among the old. If insurance companies offered health plans that insured people indefinitely, they may well maximize the lifetime utility of would-be buyers, but young, healthy people would not buy into the insurance unless they could be certain that it would pay out when they grew old and sick, and unless insurers could assess long-term risks accurately there would, again, be a standard adverse selection problem. The time inconsistency problem suggests that some types of insurance may continue to only be provided through the state. Fifth, switching from social to private insurance may incur considerable costs, especially for those that have contributed to the public system for an extended period of time. While these individuals may be better off in a private system, they may still continue to support the public system because they are already vested in it. For this reason alone, the transition from social to commercial insurance is likely to be gradual, even in the presence of breathtaking technological advances regarding information collection and sharing. Finally, there is no private insurance against poverty, such as in the case of chronic illness. Many middle-class Americans deplete their savings after exhausting private insurance to pay for long-term care because of Alzheimer’s and other debilitating diseases. In such instances, the individuals concerned will eventually qualify for Medicaid. Even though it targets the poor, Medicaid enjoys broad support among the middle class who tends to be privately insured (Busemeyer and Iversen 2020). A similar argument applies to the state as a backstop for insurance bankruptcies.

the data revolution, transformation of social protection

113

Market failure is the starting point for almost all research on social insurance— which is often treated as synonymous with the welfare state—but if the source of market failure is incomplete information, then surely more information will change the politics of social insurance. Based on our theoretical framework, the information revolution will lead to increased polarization of welfare state politics. Moreover, at least in some areas of social insurance, it might also lead to the (gradual) privatization of currently socialized risk pools. Either way, solidaristic risk-sharing will come under political attack, and inequality in social protection and incomes is likely to increase.

Appendix: details on data Life insurance penetration data The data are available at: https://stats.oecd.org/Index.aspx?DataSetCode=INSIND (OECD insurance indicators). See also OECD (2015). Sample: AUS (1984–2013), AUT (1987–2013), BEL (1983–2013), CAN (1984–2013), CHE (1983–2013), DEU (1987–2013), DNK (1983–2013), ESP (1983–2013), FIN (1983–2013), FRA (1983–2013), GBR (1996–2013), GRC (1992–2013), IRL (1983–2013), ISL (1983– 2013), ITA (1983–2013), JPN (1983–2013), NLD (1995–2013), NOR (1983–2013), NZL (1989–2003), PRT (1983–2013), SWE (1983–2013), USA (1983–2013). Mortality by cause data Our measure of information is based on data about premature mortality, as provided by the OECD (https://stats.oecd.org/index.aspx?DataSetCode=HEALTH_STAT). We make use of the “Potential Years of Life Lost” (PYLL) variable, which is defined as follows: “This indicator is a summary measure of premature mortality, providing an explicit way of weighting deaths occurring at younger ages, which may be preventable. The calculation of Potential Years of Life Lost (PYLL) involves summing up deaths occurring at each age and multiplying this with the number of remaining years to live up to a selected age limit (age 70 is used in OECD Health Statistics). In order to assure cross-country and trend comparison, the PYLL are standardized, for each country and each year. The total OECD population in 2010 is taken as the reference population for age standardization. This indicator is presented as a total and per gender. It is measured in years lost per 100 000 inhabitants (men and women) aged 0–69.” [Source: OECD (2015), Potential Years of Life Lost (indicator). doi: 10.1787/193a2829-en (retrieved September 1, 2015)]. We calculate Potential Years of Life Lost (PYLL) due to the following diseases: – – – – – – –

Certain infectious and parasitic diseases Neoplasms Diseases of the blood and blood-forming organs Endocrine, nutritional, and metabolic diseases Mental and behavioral disorders Diseases of the nervous system Diseases of the circulatory system

114 – – -

torben iversen and philipp rehm Diseases of the respiratory system Diseases of the digestive system Diseases of the skin and subcutaneous tissue Diseases of the musculoskeletal system and connective tissue Diseases of the genitourinary system Certain conditions originating in the perinatal period Congenital malformations and chromosomal abnormalities

These diseases account for about 75 percent of PYLL—the remaining PYLL are largely due to “external causes of mortality” (traffic accidents, accidental poisoning, suicides, etc.). The WHO data rely on the International Statistical Classification of Diseases and Related Health Problems (ICD). Over time, the ICD has been updated. In the empirical analyses, we include an indicator variable for changes to the ICD classification. The potential breaks occur in the following country-years: Australia (AUS): 1968, 1979, 1998. Austria (AUT): 1969, 1980, 2002. Belgium (BEL): 1968, 1979, 1998. Canada (CAN): 1969, 1979, 2000. Denmark (DNK): 1969, 1994. Finland (FIN): 1969, 1987, 1996. France (FRA): 1968, 1979, 2000. Germany (DEU): 1998. Greece (GRC): 1968, 1979. Iceland (ISL): 1971, 1981, 1996. Ireland (IRL): 1968, 1979, 2007. Italy (ITA): 1968, 1979, 2003. Japan (JPN): 1968, 1979, 1995. Luxembourg (LUX): 1971, 1979, 1998. Netherlands (NLD): 1969, 1979, 1996. New Zealand (NZL): 1968, 1979, 2000. Norway (NOR): 1969, 1986, 1996. Portugal (PRT): 1971, 1980, 2002. Spain (ESP): 1968, 1980, 1999. Sweden (SWE): 1969, 1987, 1997. Switzerland (CHE): 1969, 1995. United Kingdom (GBR): 1968, 1979, 2001. United States (USA): 1968, 1979, 1999. Partisanship variable and controls: Since partisanship only has an effect through slowly changing regulatory measures, we use Huber and Stephens’ (2001) cumulative measure of seats in government held by left parties (divided by the number of years), starting in 1960. This measure does not change a great deal in our sample, and since we estimate fixed effects models, it is not clear that estimating direct effects of partisanship is meaningful. We report results for including left partisanship both as an independent variable and as an interaction with information, but our focus is on whether left partisanship slows the progression of markets in response to information, as hypothesized above. Finally, we include three control variables that may influence life insurance penetration: (i) the percentage of the population covered by public or primary private health insurance; (ii) total health expenditure (all financing agents) as a percentage of GDP; and (iii) the rate of economic growth. Sample: The following country-years are included in our sample, which was determined by data availability: AUS (1985–2011), AUT (1988–2013), BEL (1984–2012), CAN (1985–2011), CHE (1984–2012), DEU (1993–2013), DNK (1984–2012), ESP (2011–2011), FIN (1984– 2013), FRA (1991–2011), GBR (1997–2013), GRC (1993–2008), IRL (1984–2010), ISL (1984–2009), ITA (1989–2012), JPN (1984–2012), NLD (1996–2013), NOR (1984–2013), NZL (1990–2001), PRT (1984–2013), SWE (1986–2013), USA (1984–2010). Results: The estimation results are shown in Table 6.2. The only difference between the models is whether and how cumulative partisanship is entered as an explanatory variable. Model (1)

the data revolution, transformation of social protection

115

Table 6.2 Life insurance penetration, information, and partisanship (ECM). (1)

(2)

(3)

Dependent variable: Life insurance penetration [first difference] Life insurance penetration

−0.185**

−0.218**

−0.199**

[lag]

(0.066)

(0.068)

(0.067)

Information

1.357*

2.968**

2.560*

[lag]

(0.597)

(1.064)

(1.058)

Information

2.789+

7.729+

6.322

[first difference]

(1.578)

(4.426)

(4.830)

Left partisanship X Information

−0.036**

−0.029*

[lag]

(0.014)

(0.015)

Left partisanship X Information

−0.151

−0.100

[first difference]

(0.106)

(0.111)

Left partisanship

0.031** (0.009)

Health insurance coverage

−0.012

−0.018

−0.010

(percent of population)

(0.023)

(0.023)

(0.023)

Total health expenditures

−0.059

−0.034

−0.059

(percent of GDP)

(0.066)

(0.060)

(0.066)

Economic growth rate

0.036+

0.035+

0.038+

(0.020)

(0.020)

(0.021)

Constant

2.011

1.208

1.951

(2.299)

(2.166)

(2.304)

Dummy for breaks in PYLL series

Yes

Yes

Yes

Country dummies

Yes

Yes

Yes

N

500

500

500

N of countries

22

22

22

Adj. R2

0.127

0.148

0.137

Note: Coefficients above SEs. + p