Employment, Retirement and Lifestyle in Aging East Asia (Social Policy and Development Studies in East Asia) 981160553X, 9789811605536

Citation preview

SOCIAL POLICY AND DEVELOPMENT STUDIES IN EAST ASIA

Employment, Retirement and Lifestyle in Aging East Asia Edited by Xinxin Ma

Social Policy and Development Studies in East Asia

Series Editors Joshua Mok, Lingnan University, Hong Kong Jiwei Qian, National University of Singapore, Singapore, Singapore

“Social Policy and Development Studies in East Asia” aims to provide a platform for academics, researchers and policy analysts to contribute their reflections and analysis of how rapid social, economic, cultural, political and even political economy changes would have affected the formulation, implementation and evaluation of social policy responses in handling/managing rapid changes and risk management issues confronting East Asian governments and societies.

More information about this series at http://www.palgrave.com/gp/series/15726

Xinxin Ma Editor

Employment, Retirement and Lifestyle in Aging East Asia

Editor Xinxin Ma Faculty of Economics Hosei University Tokyo, Japan

Social Policy and Development Studies in East Asia ISBN 978-981-16-0553-6 ISBN 978-981-16-0554-3 (eBook) https://doi.org/10.1007/978-981-16-0554-3 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Palgrave Macmillan imprint is published by the registered company Springer Nature Singapore Pte Ltd. The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore

Acknowledgments

The studies in the book were conducted with support from several research grants offered to the editor (Xinxin Ma) as the project leader as follows: Grant-in-Aid for Scientific Research (B) from the Japan Society for the Promotion of Science (JSPS) (Grand Number: 20H01520 from 2020 to 2022, “Economics Analysis on Social Security Policies in China: Empirical Studies Based on Survey Data”); Grant-in-Aid for Scientific Research (C) from the JSPS (Grand Number: 16K03611 from 2016 to 2018, “The Impact of Minimum Wage on the Wage Gaps between Local Urban Residents and Migrants in China”; and 25380297 from 2013 to 2015 “The Research on the Wage Gaps Between Public Sector and Private Sector in China”); Grand from the Joint Usage and Research Center, Institute of Economic Research, Kyoto University in 2020 (“Comparison Study on Corporate Governance Systems Between China and Eastern Europe”). It was also supported from several research grants offered to the editor as a project member as follows: Grant-in-Aid for International Joint Research Program from the JSPS (“Pension Reform in the PRC: Searching for a New Framework Based on Japanese Experiences” from 2017 to 2019); Grant-in-Aid for Scientific Research (B) from the Japan Society for the Promotion of Science (JSPS) (Grand Number: 20H01489 from 2020 to 2024, “Comparative Institutional Analysis of the Corporate Governance in Eastern Europe and China”); Grand from the Joint Usage and Research Center, Institute of Economic Research, Hitotsubashi University in 2015 (“Informal Sector and Income v

vi

ACKNOWLEDGMENTS

Inequality in Urban China”) and 2016 (“The Determinants of Elderly Labor Participation: A Comparison between China and Japan”). I would like to express my gratitude to my colleagues—Professor Kazuhiro KUMO (Hitotsubashi University), Professor Ichiro IWASAKI (Hitotsubashi University), and Professor Takashi OSHIO (Hitotsubashi University) for their helpful suggestions and encouragement for my research and for their warm friendships. I am grateful to the colleagues at the Institute of Economic Research (IER), Hitotsubashi University, for giving me such an excellent research environment and supports from 2015 to 2019. This book was written and completed when I was a Visiting Scholar at the Institute of Economic Research, Hitotsubashi University, from April 2020 to March 2021. I would like to express my deep gratitude to Hosei University for providing me a warm and excellent research environment, and so many supports to progress these research projects ongoing. I particularly acknowledge all contributors of this book for their hard work, efforts, cooperation, and so many helps. I greatly appreciate the Mr. Jacob Merge, Mr. Arun Kumar, and staff at Palgrave Macmillan for their interest in our research works, encouragement for the publishing, and hard editing works. Finally, I deeply acknowledge my family for their so many warm and strong support for my life and research work. Tokyo, Japan May 2021

Xinxin Ma

Contents

1

Introduction Xinxin Ma

Part I 2

3

4

5

1

Employment, Retirement, and Income Inequality of the Elderly in China

Health and Employment of the Younger, Middle-Aged, and Older Adults in China Xinxin Ma and Jingwen Zhang

19

Work Skills Gap and the Wage Differentials Between the Young, Middle-Aged, and Older Workers in China Xinxin Ma and Xiaobo Qu

41

The Impact of Social Insurance Contributions on Chinese Firms’ Employment and Wages Xinxin Ma and Jie Cheng

71

Pension Benefit Inequality of the Elderly in China Peng Zhan and Hanrui Jia

107

vii

viii

CONTENTS

Part II Employment, Retirement, and Lifestyles of the Elderly in Japan 6

7

8

9

Health Capacity to Work and Its Long-Term Trend Among the Japanese Elderly Takashi Oshio

133

Willingness to Continue Volunteering Among the Middle-Aged and Older Adults in Japan Xinxin Ma

161

Seniority Wage and Employment of the Older Workers in Japanese Firms Atsushi Seike and Xinxin Ma

187

Living Arrangement and Well-Being of the Middle-Aged and Older Adults in Japan Tsukasa Matsuura and Xinxin Ma

213

Part III 10

11

12

Population Aging, Work, and Lifestyles of the Elderly in Other East Asian Regions

Population Aging, Labor Force Participation, and Family Structure in the Republic of Korea Toru Suzuki

241

Retirement Timing and Post-retirement Employment in Taiwan Ruoh-Rong Yu and Ming-Chang Tsai

257

Population Aging, Low Fertility, and Social Security in Russia Kazuhiro Kumo

283

Index

301

Editor and Contributors

About the Editor Xinxin Ma is a Professor at Faculty of Economics, Hosei University. She was an executive board member of Japanese Association for Chinese Economy and Management (JACEM) from 2018 to 2020. She was the editorial board member of academic journals (e.g., Journal of Chinese Economy Studies, Asian Studies, Japanese Journal of Comparative Economics ) and a referee of various academic journals. She was an author of numerous books (Economic Transition and Labor Market Reform in China, Palgrave Macmillan, 2018, etc.), and numerous academic papers were published in peer-reviewed journals such as China Economic Review, Journal of Economics and Business, Journal of Happiness Studies, Emerging Markets Finance and Trade, Journal of Chinese Economic and Business Studies, BMC Public Health. Her current research project focuses on the empirical studies on social security policy reform and inequality in China and Japan.

Contributors Jie Cheng Institute of Population and Labor Economics, Chinese Academy of Social Sciences, Beijing, China Hanrui Jia Business School, Beijing Normal University, Beijing, China

ix

x

EDITOR AND CONTRIBUTORS

Kazuhiro Kumo Institute of Economic Research, Hitotsubashi University, Tokyo, Japan Xinxin Ma Faculty of Economics, Hosei University, Tokyo, Japan Tsukasa Matsuura Department of Economics, Chuo University, Tokyo, Japan Takashi Oshio Institution of Economics Research, Hitotsubashi University, Tokyo, Japan Xiaobo Qu Institute of Population and Labor Economics, Chinese Academy of Social Sciences, Beijing, China Atsushi Seike Faculty of Business and Commerce, Keio University, Tokyo, Japan Toru Suzuki Graduate School of Public Health, Seoul National University, Seoul, Korea Ming-Chang Tsai Research Center for Humanities and Social Sciences, Academia Sinica, Taipei, Taiwan Ruoh-Rong Yu Research Center for Humanities and Social Sciences, Academia Sinica, Taipei, Taiwan Peng Zhan School of Public Affairs, Zhejiang University, Hangzhou, China Jingwen Zhang School of Social Sciences, University of Manchester, Manchester, UK

List of Figures

Fig. 1.1

Fig. 1.2

Fig. 3.1 Fig. 6.1

Fig. 8.1

Fig. 8.2

International comparison of proportion of population aged 65+ to total population (Source Author’s creation based on data from World Population Prospects: The 2019 Revision published by United Nations International comparison of labor force participation rate of population aged 65+ (Note JPN: Japan; HK: Hongkong; SGP: Singapore; USA: United States; CAN: Canada; UK: United Kingdom; DEU: Germany; FRA: France; ITA: Italy. Source Author’s creation based on data from World Population Prospects: The 2019 Revision published by United Nations) Wage gap by task skill types and age group in China (Source Author’s creation based on the 2016 CLUS) Estimated potential and additional capacities to work in 2016, based on Model 1 results (Note The numbers in [ ] indicate the potential capacity to work. Source Author’s creation) Age-wage profile (seniority wage) and mandatory retirement age (Source Author’s creation based on Lazear [1979]) Age-wage profile by education in Japan (Source Author’s creation based on Elderly Employment and Recruitment Survey)

2

4 52

144

189

195

xi

xii

LIST OF FIGURES

Fig. 8.3

Fig. 8.4

Fig. 9.1

Fig. 9.2 Fig. 9.3 Fig. 10.1 Fig. 10.2 Fig. 12.1 Fig. 12.2

Fig. 12.3

Fig. 12.4

Fig. 12.5

Age-wage profile by firm size in Japan (Source Author’s creation based on Elderly Employment and Recruitment Survey) Age-wage profile by industrial sectors in Japan (Source Author’s creation based on Elderly Employment and Recruitment Survey) Proportion of older adults: comparison with Japan and other countries (Note The proportions of population aged 65 and older to total population are shown in the figure. Source Author’s creation based on data from World Population Prospects: The 2017 Revision published by United Nations. Japan: 1950–2015 Census data, Ministry of Internal Affairs and Communications, Japan: 2020–2060 data calculated by National Institute of Population and Social Security Research) Life satisfaction by living arrangement type for Japanese women (Source Author’s creation) Life satisfaction by living arrangement type for Japanese men (Source Author’s creation) International comparison of total fertility rate (TFR) (Source Author’s creation) Male labor force participation: 2015 (Korea and Japan) (Source Author’s creation) Number of people in poverty in Russia (Source Author’s creation based on data from Milanovic [1997]) Russia’s poverty headcount and per-capita GDP, 1989–2018 (Source Author’s creation based on data from Rosstat, Sotsial’noe polozhenie i uroven zhisni naseleniya Rossii, various years; Rosstat, Regiony Rossii, various years; World Bank, World Development Indicators ) Russia’s income disparities and per-capita GDP, 1980–2013 (Source Author’s creation based on data from Braithwaite [1995]; Rosstat, Sotsial’noe polozhenie i uroven zhisni naseleniya Rossii, various years; Rosstat, Regiony Rossii, various years) Average male life expectancy at birth (Source Author’s creation based on data from World Bank, World Development Indicators; Rosstat, Demograficheskii ezhegodnik Rossii, various years) Causes of death among Russian males, 1965–2018 (Source Author’s creation based on data from Rosstat, Demograficheskii ezhegodnik Rossii, various years)

195

196

214 221 222 243 252 285

286

287

288

289

LIST OF FIGURES

Fig. 12.6

Fig. 12.7

Fig. 12.8

Fig. 12.9

Crude birth rate and crude death rate in Russia, 1960–2018 (Source Author’s creation based on data from Rosstat, Demograficheskii ezhegodnik Rossii, various years) Proportion of population by age group/age composition indexes, 1989–2019 (Source Author’s creation based on data from Rosstat, Demograficheskii ezhegodnik Rossii, various years) Average pension benefit relative to basic cost of living and pensioner index, 1994–2018 (Source Author’s creation based on data from Rosstat, Sotsial’noe polozhenie i uroven zhisni naseleniya Rossii, various years) Russia’s total fertility rate and per-capita GDP, 1989–2018 (Source Author’s creation based on data from Rosstat, Demograficheskii ezhegodnik Rossii, various years; Rosstat, Regiony Rossii, various years)

xiii

291

292

293

295

List of Tables

Table 1.1 Table 1.2 Table 2.1 Table 2.2 Table 2.3 Table 2.4 Table Table Table Table

2.5 3.1 3.2 3.3

Table 3.4 Table 3.5 Table 3.6 Table 3.7 Table 3.8

International comparison of proportion of populations aged 15–64 to the total population International comparison of population turning point and per capita GDP Basic estimates of the influence of health on labor force participation in China Estimates of the influence of health on labor force participation by gender in China Estimates of the influence of health on labor force participation by age groups in China Estimates of the influence of health on labor force participation of individuals aged 16–50 in China Descriptive statistics of variables Task categories and measurement items in the CULS Task skill gap by different age groups in China Wage gap between age groups and the influence of task skill on wages in China The influence of task on wages by different age groups in China Decomposition results of wage gap between workers aged 16–49 and aged 50–59 in China Decomposition results of wage gap between workers aged 50–59 and aged 60+ in China Descriptive statistics of variables Results of wage function in China

3 4 27 28 30 33 34 49 51 54 57 60 61 63 66

xv

xvi

LIST OF TABLES

Table 4.1 Table 4.2 Table 4.3 Table 4.4 Table 4.5

Table 4.6 Table 4.7

Table 4.8 Table 4.9 Table 4.10 Table 5.1 Table 5.2 Table 5.3 Table 5.4 Table Table Table Table

5.5 5.6 5.7 5.8

Table 5.9

Table 5.10

Table 5.11 Table 5.12 Table 5.13

Four combinations of wages and employment Results for the impact of social insurance (CCR) on Chinese firms’ wages and employment in China Results for the impact of social insurance (CE) on Chinese firms’ wage and employment in China Results for the impact of social insurance (FAR) on Chinese firms’ wage and employment in China Summary of results for the impact of social insurance on wages and employment according to ownership types in China Summary of results for the impact of social insurance on wages and employment by firm size in China Summary of results for the impact of social insurance on wages and employment for capital-intensive firms and labor-intensive firms in China Summary of results for the impact of social insurance on wage and employment by region in China Descriptive statistics Results of wage and employment functions in China Proportion of the elderly who did not receive pension benefit in 2013 Inequality and decomposition of pension benefit Pension benefit based on different enterprise ownership and occupation type before retirement Regional gap and urban–rural gap of pension benefit (Unit: Yuan) Educational gap of pension benefit (Unit: Yuan) Gender gap in pension benefit Age gap of pension benefit (Unit: Yuan) Theil decomposition of pension benefit in 2013 for China Decomposition results of contribution of each variable to pension benefit inequality based on the regression model Decomposition results of contribution of each variable to pension benefit inequality based on the regression model, specific pension type Theil decomposition of pension benefit for 2013 in rural China Theil decomposition of pension benefit for 2013 in urban China OLS model results of pension benefit equations for 2013

83 85 86 87

89 90

91 92 97 99 112 114 115 116 117 117 118 120

122

123 125 126 128

LIST OF TABLES

Table 6.1 Table 6.2 Table 6.3

Table 6.4 Table 6.5 Table 6.6 Table 6.A1 Table 6.A2

Table 7.1 Table 7.2 Table Table Table Table

7.3 7.4 7.5 8.1

Table 8.2 Table 8.3 Table 8.4 Table 8.5 Table 8.6 Table 8.7 Table 8.8 Table 8.9 Table 9.1

Table 9.2

Summary statistics of working and health statuses in 1986, 2001, and 2016 Estimated potential and additional capacities to work in 2016 (% of total respondents in each age group) Estimated capacity to full- and part-time work in 2016: multinomial logistic models (% of total respondents in each age group) Estimated capacity to work in 1986, 2001, and 2016 Decomposition of the change in the additional capacity, based on Models 2 results Distribution of the estimated probabilities of work, based on Models 1 results (%) Regression results of Models 1 and 2 for those aged 50–59 years in 2016 (dependent variable = no work) Regression results of multinomial logistic models for individuals aged 50–59 years in 2016 (Base outcome = Work) Results of willingness to continue volunteering in Japan Results of probabilities of willingness to continue volunteering by age in Japan Summary of results Descriptive statistics of variables Results of volunteering reward function in Japan Probability of performing the mandatory retirement system in Japanese firms Results of mandatory retirement age function in Japan Results of retirement age function by firm size in Japan Probability of performing the elderly reemployment system in Japanese firms Probability of setting the oldest age in the elderly reemployment system in Japanese firms Results of reemployment age function in Japan Results of reemployment age function by firm size in Japan Descriptive statistics of variables Wage function by age in Japan Proportions of each types of family structure of household head aged 65 and older and aged 75 and older in Japan (Unit: %) Descriptive statistics of variables

xvii

142 143

145 146 148 149 153

156 169 171 173 175 180 198 199 200 202 204 206 207 208 209

215 219

xviii

LIST OF TABLES

Table 9.3

Table 9.4

Table 9.5

Table 9.6

Table 9.7

Table 9.8

Table 9.9

Table 9.10

Table Table Table Table Table

10.1 10.2 10.3 10.4 10.5

Table 11.1 Table 11.2

Results of living alone and well-being of Japanese individuals aged 45 and older using ordered probit regression model (Dependent variable: Happiness in the past one year) Results of living alone and well-being of Japanese individuals aged 45 and older using random effects probit model (Dependent variable: Happiness in the past one year) Results of living alone and well-being of Japanese individuals aged 60 and older using ordered probit regression model (Dependent variable: Happiness in the past one year) Results of living alone and well-being of Japanese individuals aged 60 and older using random effects probit model (Dependent variable: Happiness in the past one year) Results of living alone and well-being of Japanese individuals aged 45 and older using ordered probit regression model (Dependent variable: Lifetime happiness) Results of living alone and well-being of Japanese individuals aged 45 and older using random effects probit model (Dependent variable: Lifetime happiness) Results of living alone and well-being of Japanese individuals aged 60 and older using ordered probit regression model (Dependent variable: Lifetime happiness) Results of living alone and well-being of Japanese individuals aged 60 and older using random effects probit model (Dependent variable: Lifetime happiness) Projected percentages of elderly population aged 65+ Demographic indicators in 2065 Elderly poverty rate and suicide rates Percentage of elderly aged 65+ living alone Percentage of single based on 2015 Census in Korea and Japan Means of variables for analyses on retirement timing: by gender Means of variables for analyses on post-retirement employment: by gender

224

225

226

227

232

233

234

235 246 246 248 250 253 266 268

LIST OF TABLES

Table 11.3 Table 11.4

Cox model on retirement transition: by gender and retirement definition Random-effects multinomial logit model on post-retirement employment: by gender

xix

269 272

CHAPTER 1

Introduction Xinxin Ma

1.1

Background and Aim of This Book

The developed countries in East Asia, such as Japan and Republic of Korea (hereafter, abbreviated as Korea), have experienced population aging. As a developing country, China has also experienced population aging since 2010. The main feature of population aging in East Asia can be summarized in three points. First, the speed of population aging is faster for East Asia (e.g., Japan‚ China, Korea) than for European countries (i.e., France, United Kingdom (UK), etc.) and the United States (US). For China, based on the data of the National Bureau of Statistics (NBS, 2019), the ratio of the population aged 65 and over (65+) has increased from 4.2% in 1982 to 7.0% in 2000, and 10.5% in 2015. (Figure 1.1 shows the international comparison of population aging.) In 2015, ratio of the aged 65+ population in comparison with the total population was 9.3% for China and 26.0% for

X. Ma (B) Faculty of Economics, Hosei University, Tokyo, Japan e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 X. Ma (ed.), Employment, Retirement and Lifestyle in Aging East Asia, Social Policy and Development Studies in East Asia, https://doi.org/10.1007/978-981-16-0554-3_1

1

2

X. MA

Japan. Estimates for 2060 suggest that these ratios will become 29.8% for China and 38.3% for Japan. As population aging progresses, the labor force decreases. Table 1.1 reports the international comparison between the labor force (the individual aged 15–64) and the total population. The labor force decreased more starkly for countries in East Asia, as compared to those in Europe and North America. Concretely, from 2000 to 2050, the proportion is projected to decrease from 68.5% to 59.7% for China, from 68.2% to 51.1% for Japan, and from 72.2% to 53.2% for Korea. The ratio decreases are higher for countries in East Asia than for European countries (i.e., the UK, Germany, France, Sweden, etc.) and North American countries (i.e., the US, Canada). The labor force shortage is expected to become a severe problem in East Asia. Addressing the labor force shortage in aging societies and devising policies to increase the labor force participation of older adults is a challenge for governments. Figure 1.2 presents the labor force participation rate of the 65+ population. It is clear that they are different by country. It % 45.0 40.0 35.0 30.0 25.0 20.0 15.0 10.0 5.0 0.0 China

Japan

Republic of Korea

France

UK

US

Fig. 1.1 International comparison of proportion of population aged 65+ to total population (Source Author’s creation based on data from World Population Prospects: The 2019 Revision published by United Nations

1

INTRODUCTION

3

Table 1.1 International comparison of proportion of populations aged 15–64 to the total population

CHN JPN KOR IND VNM USA CAN UK DEU FRA ITA SWE RUS

1980

2000

2010

2015

2020

2030

2050

59.6 67.5 62.0 57.1 53.8 65.7 67.8 64.0 65.8 64.0 64.7 64.1 68.2

68.5 68.2 72.2 60.9 61.9 66.0 68.3 65.1 67.8 65.2 67.5 64.3 69.3

73.8 64.1 73.2 64.0 69.8 66.9 69.3 65.9 65.9 64.8 65.5 65.3 72.0

72.6 61.0 73.1 65.7 70.2 66.1 67.9 64.3 65.8 62.8 63.9 63.1 69.7

70.4 59.1 71.1 66.9 69.0 64.8 65.6 63.1 64.7 61.5 62.8 61.9 66.3

67.6 57.5 63.0 68.0 67.0 61.4 61.2 61.0 59.6 59.3 59.1 60.3 63.4

59.7 51.1 53.2 67.7 61.6 60.7 59.1 58.3 56.4 56.7 52.3 58.9 60.5

Note CHN: China; JPN: Japan; KOR: Republic of Korea; IND: India; VNM: Vietnam; USA: United States; CAN: Canada; UK: United Kingdom; DEU: Germany; FRA: France; ITA: Italy; SWE: Sweden; RUS: Russia Source Author’s creation based on data from World Population Prospects: The 2017 Revision published by United Nations. The data after 2020 are imputed values based on the median of birthrate, mortality

increased in Japan (37.0% in 1985; 33.9% in 2018) and Singapore (25.9% in 1985; 38.2% in 2018), and lowered in France (5.3% in 1985; 4.0% in 2018) and Italy (8.4% in 1985; 7.7% in 2018). Why does labor force participation of older adults differ by country? It may be related to labor supply side factors (e.g., willingness to work, income or wealth of the elderly, personal preference for work, human capital of older workers, etc.), labor demand side factors (i.e., hiring of older workers by firms, etc.), and institutional factors (e.g., public pension, public medical insurance, or disability insurance). Therefore, the empirical studies on the determinants (or mechanism) of employment of older adults are needed to promote their labor force participation. Second, in contrast to the developed countries in East Asia (e.g., Japan and Korea), China is a developing country with an aging population. Table 1.2 summarizes the population turning point (the year of labor force decrease) and per capita GDP at the population turning point year. It indicates that in the turning point year of labor force decrease, the per capita GDP was 23,504 dollars (1990) for Japan and 27,724 dollars

4

X. MA

% 45.0 40.0 35.0

38.2

37.0 33.9

30.0

26.7

25.9

1985

24.0

25.0 18.6

20.0

2018

15.8

15.0

18.1

14.0 11.4

10.3

8.5

10.0

5.1

5.0

8.4 7.7 5.3 4.0

0.0

JPN

HK

SGP

USA

CAN

UK

DEU

FRA

ITA

Fig. 1.2 International comparison of labor force participation rate of population aged 65+ (Note JPN: Japan; HK: Hongkong; SGP: Singapore; USA: United States; CAN: Canada; UK: United Kingdom; DEU: Germany; FRA: France; ITA: Italy. Source Author’s creation based on data from World Population Prospects: The 2019 Revision published by United Nations) Table 1.2 International comparison of population turning point and per capita GDP

China Japan Republic of Korea Singapore Thailand Malaysia Indonesia India Vietnam Philippines

Turning point year

Per capita GDP

2015 1990 2017 2010 2010 2020 2030 2035 2020 2040

9,722 23,504 27,724 30,391 8,740 15,571 6,207 7,758 4,762 12,289

Note Per capita GDP is adjusted by PPP (2000 standard) Source Author’s creation based on data from World Bank (WDI)

1

INTRODUCTION

5

(2017) for Korea; in China, it was 9,722 dollars (2010). Therefore, the financial burden of establishing a social security system to address the aging population is more severe for China. The Chinese government may learn from experiences in Japan and elsewhere to address the problems of its aging society. China is experiencing the unique economy transition from a planned economy to a market-oriented economy. A set of social and economic policies and institutions have been replaced. New policies were established and implemented. For example, the Chinese government reformed the social security systems and established new organizations, such as the New Rural Cooperative Medical Scheme (NRCMS) in 2003 and the New Rural Social Pension Scheme (NRSP) in 2009. By 2014, the public pension and medical insurances have legally covered the whole population. However, because society and institutions are segmented by the population registration system (“Hukou” in Chinese), workplace sectors (e.g., government organization, state-owned enterprises, privately-owned enterprises, foreign-owned enterprises, etc.), and the social security systems are also segmented. Therefore, establishing universal social security systems, such as those in Japan, Korea, and Taiwan, is an important issue for the Chinese government. Third, in comparison with younger and middle-aged generations, the elder generation can easily become impoverished. Reducing poverty and income inequality among that generation should be reviewed. Additionally, in East Asia, Confucianism’s filial piety influences the lifestyle (e.g., co-resident with older parents, income transfer from adult children to their parents, parental care, etc.) of middle-aged and older individuals. These lifestyles and their financial burden will affect the well-being of younger individuals. The balanced association between work and life at advanced ages also should be noted. Regarding the features of population aging in East Asia, the aims and scope of this book are as follows: First, this book focuses on the employment and lifestyle of middle-aged and older populations in East Asia, including China, Japan, and Korea. It also offers comparisons with Taiwan and Russia to enrich studies of these issues. This book is the first to challenge these issues with a comparison study between East Asia and other regions. Based on the principle of “evidence-based policymaking,”1 we employed the empirical studies based on the latest national individual and household survey data, firm survey data, and official data in each country

6

X. MA

or region. We investigated the determinants and mechanisms of the influence of social security policies on behaviors of individuals, households, and firms. The results of empirical studies can provide new information and evidence for academic research and policy decisions. Third, this book bases its empirical studies on theories from labor economics, family economics, population economics, demography, and sociology. The different methodologies used in this book can provide rich evidence from multiple perspectives.

1.2

Main Contents of This Book

This book consists of three parts—Part I Employment, Retirement, and Income Inequality of the Elderly in China (Chapters 2–5), Part II Employment, Retirement, and Lifestyle of the Elderly in Japan (Chapters 6–9), and Part III Population Aging, Work, and Lifestyle of the Elderly in Other Regions (Chapters 10–11). The main contents of each chapter are as follows. Health status is a main factor influencing individuals’ work participation. Considering the rapid process of population aging and the rising burden of public financial security (e.g., public medical expenditures faced by many governments), it is important for policymakers to develop a comprehensive understanding of the impact of health on work participation. Chapter 2 uses nationwide longitudinal survey data from the China Health and Nutrition Survey (CHNS) from 1997 to 2006 to investigate the influence of health on work participation in China among young, middle-aged, and older adults. The lagged variable probit regression model is used to address the reverse causality problem. The fixed-effects model is used to address the heterogeneity problem, and the Heckman two-step model is used to address the sample selection problem in the working hours function. The main findings are as follows: First, generally, health positively affects work participation. Concretely, as compared with the poor health group, the probability of working is 10.4–12.0% points higher for the healthy group. However, the influences of health on the probability of becoming a regular worker and working hours are not statistically significant. Second, the influences of health on work participation differ among groups. Concretely, the probability of working is 15.5 ~ 20.7% points higher for healthy men than for those in poor health; the probability of becoming a regular worker is 42.1% points higher for the healthy younger generation. Third, robustness checks using different

1

INTRODUCTION

7

samples (e.g., the group aged 16–50) confirm the main findings for the probability of becoming a regular worker and for longer working hours. It is expected that the public medical insurance policy may positively affect the health status of the population and promote the desired increase in work participation. Therefore, results indicate that the public medical insurance system should be supported by the Chinese government to improve the health status of individuals in China and, possibly, to increase the long-term labor supply in each age group. Since the 1980s and 1990s, the wage gap between highly skilled and low-skill workers has grown. The skill-biased technological change hypothesis advocates that the adoption of computer-based technologies and the increased employment of higher-educated labor within industries, firms, and workplaces causes increased employment and wages in more highly skilled jobs. It is thought that learning ability and human capital may decrease with age. Therefore, the wage gaps among different age groups may be caused by skill gaps among the age groups. However, empirical studies are scarce because there is no suitable data. Chapter 3 uses a unique survey—China Urban Labour Survey (CULS), conducted in 2016, used a task skill approach to investigate how the work skill gap affects the wage gap between the middle-aged and older worker groups in Chinese firms. Based on the CULS2016 questionnaire, we constructed four types of work skills: Task I (non-routine cognitive analytical), Task II (non-routine cognitive personal), Task III (routine cognitive), and Task IV (routine manual). The Heckman two-step model and Blinder-Oaxaca model were used. Four conclusions can be summarized as follows: First, the skill level of younger workers is higher than that of older workers in each task. This suggests that task skills decline with age. Second, the influence of task skills on wages differs by age group. For the younger worker group and the middle-aged worker group, each task positively affects wage levels, but for the older worker group, the influences of task skills are not statistically significant for each task. Third, the decomposition results of the wage gap between the younger worker group and the middle-aged worker group indicate that the explained part (53.7%) is greater than that of unexplained part (46.7%). Regarding the contributions of tasks to the wage gap, the differences in task skill levels (38.4%) contribute to expanding the wage gap, while differences in the returns on task skills (−18.0%) may reduce the wage gap. For the wage gap between the middle-aged worker group and the older worker group, it is shown that, generally, the influence of the unexplained part (66.3%) is

8

X. MA

greater than that of the explained part (33.7%). The differences in task skill levels (3.7%) expand the wage gap, while differences in the return on task skills (−35.7%) may reduce the wage gap. This suggests that the policy of work skill training during a lifelong working career, particularly for older workers, may reduce the wage gap among different age groups in China. Since 1978, the Chinese government has gradually implemented an economy transition and reformed the social security systems. The total social insurance contribution ratio paid by firms, excluding the housing fund, was around 43% in 2015. China has become one of the countries with higher social insurance premiums in the world. According to the rules of perfect market competition, to gain maximum profit, a firm may transfer the burden of increased social insurance contributions onto its employees by reducing wages and the number of employees. Yet, empirical study results are scarce and inconclusive on the issue for China. Chapter 4 helps to fill this gap in research knowledge. Using Chinese Employer-Employee Matching Survey (CEES) data from 2011 to 2014, and imputed values based on wage and employment functions, this chapter uses three indices of social insurance enforcement (the city contribution rate, the rate of city enforcement of social insurance, and the firms’ actual contribution rate) and four types of combinations of wage and employment (LWLE: a firm with low wages and less employment; HWME: a firm with high wages and more employment; LWME: a firm with low wages and more employment; HWLE: a firm with high wages and less employment) to investigate the influence of social insurance contributions on wages and employment in Chinese firms. The main findings are as follows: First, a bipolarization phenomenon is found in the relationship between firms’ actual social insurance contributions and the adjustments of wages and employment. It also found that the impact of the actual contribution rate for social insurance appears to be greater on levels of employment than on wages. Second, the impact of social insurance contribution on firms’ level of wages and employment differs for each group. In response to an increased social insurance contributions, privately-owned enterprises are likely to reduce employment. The negative impact of an increased social insurance contributions on wages and employment is greater for small and labor-intensive firms. Chapter 5 investigates the inequality of pension benefits among older people and the structural differences in the pension benefit composition between the high-income group and the low-income group using 2013

1

INTRODUCTION

9

China Household Income Project survey (CHIPs). Since family support is important for the income security of the elderly, this chapter also accounts for the impact of income-sharing within a household. There is a significant inequality in the pension benefits for the elderly in China, and its Gini coefficient is 0.59. The Gini coefficient of pension benefits is 0.89 in rural areas, which is greater than that of urban areas. This suggests that the inequality of pension benefits is greater in rural areas than in urban areas. The contribution rate of pension benefits to income inequality of the elderly in China is as high as 81%. The contribution rate of pension benefit inequality is much higher for urban areas than that in rural areas. Older people in the high-income group receive higher pensions, which means that a disparity in the social security system remains among different age groups, which may increase China’s income inequality. The results indicate that public transfers (public pensions) are more important for income redistribution than private transfers (family support). This suggests that, to reduce income inequality among the elderly, pensions play a greater role than income-sharing within a household. In addition, the results of the decomposition analysis based on the regression model indicate that the pension benefit inequality is mainly influenced by the disparities in the pension systems (Civil Servant Pension, Urban Employee Basic Pension, Urban Resident Basic Pension, or New Rural Social Pension Scheme), regional disparities (urban areas versus rural areas), and different levels of educational attainment. Among them, disparities in pension systems (differences in types of pension systems) are also manifested as a factor causing regional disparities. It is also found that differences in pension benefits by different pension systems are greatly affected by situations of an individual’s pre-retirement job. These results suggest that to reduce income inequality among the elderly in China, a policy for reducing the pension benefit inequality between rural areas and urban areas is necessary. Also, the Chinese government should consider reforming the defined benefit pension system based on individuals’ employment careers. Chapter 6 examines the health-based capacity of the elderly to work in Japan—that is, how much longer the elderly can work judging by their health. This chapter relies on the long-term trend microdata from 1986 to 2016 obtained from the nation-wide, population-based survey—Comprehensive Survey of the Living Conditions, which was conducted and released by the Ministry of Health, Labour and Welfare of the Japanese Government. Based on the estimated relationship between health and work status among individuals in their 50s, this study simulated their

10

X. MA

capacity to continue working in their 60s and early 70s. The simulation results revealed a large additional work capacity among the elderly, as well as the possibility of some shifting from part-time to full-time jobs among elderly males. This chapter further observed that this additional work capacity has increased over the past 30 years as health has improved, although health conditions still prevent some individuals from working. These results underscore the need for policy measures that will allow for the utilization of the unexploited work capacity of older adults. From a labor supply perspective, because older adults can receive pensions, it is thought that, unlike young and middle-aged groups, older adults are more likely to exit the labor market and participate in social activities. From a broader perspective, to produce an age-free society, it is important to build a social environment that enables older people to participate in various social activities, from volunteering to employment. Using the 2014 Activities and Working Status Survey of individuals and NPO organizations conducted by the Japan Institute for Labor Policy and Training (JILPT), Chapter 7 constructs an employee-employer matched dataset that includes individuals and NPO organizations to investigate the influence of four kinds of factors on individuals’ willingness to continue to volunteering in NPO organizations. The four factors are (1) human capital; (2) household income; (3) activity motivation; and (4) rewards. The main findings are as follows: First, generally, the human capital factor and the motivation factor significantly affect willingness to continue volunteering; however, the influences of household income and rewards from the NPO are not significant. Second, the influences of four kinds of factors on willingness to continue volunteering differ among groups. Specifically, the influences of human capital and income factors are confirmed for the group aged 50–59, and the influences of human capital and reward factors are significant for the group aged 60–64. The influences of income and motivation factors are confirmed for the group aged 65 and above. This suggests that the probability of participating in social activities is greater for middle-income and high-income groups than for the low-income group. To establish an active society to address the aging of the population, for the low-income group, implementing income redistribution policies, such as minimum income protection policy, is necessary; it will also be necessary to consider policies that support and promote participation in social activities. It is expected that social participation (volunteering) can improve health status and increase the social

1

INTRODUCTION

11

capital of older adults, which may improve the well-being of the older people in a society with an aging population. In Japan, to address the labor force shortage and to reduce the burden of public pensions funded by the government, the Japanese government has, since the 1980s, promoted a set of policies intended to raise the pension eligibility age and the legal retirement age. Particularly, in accordance with the 2007 Elderly Employment Security Act, firms are obliged to continue recruiting workers aged 60 and over. From the labor demand side, wage and employment institutions may have a significant impact on the employment of older workers. In most Japanese firms, particularly in large firms, seniority wages, long-term employment, and mandatory reteirment systems are known as the “Japanese style management.” However, there are few empirical studies on the influence of wage structures for firms’ employment of older workers. Using the Elderly Employment and Recruitment Survey (EERS) conducted by the Japan Institute for Labor Policy and Training (JILPT), Chapter 8 investigates seniority wage on the setting of the mandatory retirement age and on reemployment age after retirement. Maddala’s model and the 2SLS model are used to address the sample selection bias and endogeneity problems. New findings emerge. First, the steepness of the age-wage profile influences the employment of older workers. Concretely, for a firm with a steeper seniority wage profile, the probability of implementing the retirement system will increase, while the probability of implementing the elderly reemployment system will decrease. When the average annual wage per employee increases by 10,000 Japanese yen, the mandatory retirement age reduces by four years, and the oldest reemployment age will reduce by two years. Second, the influences of the seniority wage profile differ by firm size. Specifically, (1) when the average annual wage per employee increases by 10,000 Japanese yen, the mandatory retirement age will reduce five years for small firms, three years for medium-sized firms, and three years for large firms. (2) When the average annual wage per employee increases by 10,000 Japanese yen, the oldest reemployment age will be reduced by five years for large firms, but a negative effect on reemployment age is not found for small and medium-sized firms. The results indicate that the seniority wage negatively affects the employment of older workers, and the negative effect is greater for large firms than for medium-sized and small firms. To address the labor force shortage problem of population aging in Japan, building an “age-free” career society has become important. Reforming the wage system from

12

X. MA

the traditional seniority wage system to an ability-based wage system (or a pay for performance system) should be considered. When the wage-age profile becomes flatter, employees’ motivation may decrease. Considering the employment and living stability of older workers, the seniority wage should be transformed and implemented step by step. In Japan, as population aging progresses, elderly single-person households are predicted to increase. How do living arrangements (i.e., living alone) affect the well-being of the elderly in Japan? Chapter 9 employs an empirical study to answer this question. Using longitudinal data from the Japanese Household Panel Survey (JHPS) and the Keio Household Panel Survey (KHPS) from 2011 to 2018, this study investigates the influence of living alone on the well-being of middle-aged and older adults in Japan. The main findings are as follows: First, women living alone are likely to feel happy in both age groups—aged 45 and above, aged 60 and above. Second, living alone negatively affects the happiness of men aged 60 and above. However, the negative effect of living alone disappears when marital status is controlled. The results indicate that the positive effect of having a spouse on well-being is greater for Japanese men. In Japan, the influence of living alone on the well-being of the elderly differs by gender. Chapter 10 analyzes the situation of population aging, labor force participation, and lifestyle in the Republic of Korea. Together with Taiwan, the Korea has been suffering for more than a decade from the lowest levels of fertility in the world. This implies that the aging population in Korea will increase drastically, eventually overtaking Japan as the oldest country in the world. Korea’s compressed development from a poor agrarian society to an advanced economic power has brought about various socio-cultural challenges. Since the industrial structure and conditions of the labor market change so swiftly, acquired skills become outdated very quickly. Thus, labor productivity tends to be low in middle and older ages. Because of early retirements from large corporations, the percentage of self-employed workers has increased. The labor force participation of elderly men is also high due to the unsatisfactory social security system. In 2019, the value of the Total Fertility Rate was recorded as 0.92 in Korea. An interpretation of the causal relationship is that there is a significant discrepancy between the rapidly changing non-family system and the slowly changing family system. While gender equity in political, educational, and occupational sectors improved dramatically in Korea, gender relations in familial and kinship sectors have changed slowly.

1

INTRODUCTION

13

Strong parent–child ties resulted in the postponement of home-leaving, economic independence, marriage, and childbearing. Heated competition to enter high-ranked universities and corporations was reinforced by the Confucian value system that treated manual laborers poorly in the Korean context. Such cultural concerns are discussed by contrasting Korea with Japan and with Western countries. Chapter 11 aims to investigate the timing of retirement and postretirement employment of people aged 50 to 64 in Taiwan. In Taiwan, the retirement systems differ for private and public employees. The mandatory retirement age for private and public employees is 65 years. An employee is eligible for voluntary retirement if he or she meets certain age and/or service requirements. According to government statistics, in 2018, the average retirement age of private employees increased from 56.5 years in 2010 to 61.2 years, whereas that of public sector employees increased from 55.16 years in 2010 to 57.1 years. Despite these trends, people in Taiwan retired earlier than people at a similar income level in other countries. Taiwanese labor force participation rates of people aged 55–59, 60–64, and 65+ are 55.63%, 36.70%, and 8.43%, respectively. These figures are relatively low. Due to increases in life expectancy and low fertility rates, pension reform and labor force participation of the aged population have become critical issues for Taiwan. Chapter 11 uses data from the Taiwan Panel Study of Family Dynamics (PFSD) survey. This unique panel survey, started in 1999, contains rich information on current working status, retirement, job before retirement, income, personal demographic characteristics, family structure, and family interactions. The first research aim of this chapter is to analyze the retirement timing decisions for those who are eligible for voluntary retirement; it focuses on the effects of family support and economic resources. The Cox proportional hazards model is used to analyze the transition to retirement for those employed in the private and public sectors. The second research aim is to explore the decision to take a job after official retirement. In addition to describing the work status of those who had officially retired from their employment, this chapter also compares the traits of the preretirement job to those of the post-retirement job. The random-effects probit model is applied to analyze the decision to take a job after retirement; focus is on the influence of family support and economic resources on the likelihood of post-retirement employment. The basic statistics reveal that men’s probability of transition into retirement (or pension) is higher than that of women, probably because men are more likely to

14

X. MA

have a continuous work trajectory and thus hence a higher chance of voluntary retirement or pension receipt than women. The basic statistics for post-retirement working status reveal that, relative to women, men are more likely to engage in full-time or part-time employment after retirement. This might be attributed to the different gender roles of men and women. Chapter 12 uses government and survey data to investigate the aging population and social security in the old Soviet Union and the new Russian Federation. As is well known, social security policies are implemented because poverty arises from problems caused by illness, childbirth, and aging. The same is true in the old Soviet Union and the new Russian Federation. It is thought that the economic background and the social system were very different between these two periods. For example, there was no employment insurance system in the Soviet Union, which was said to have no unemployment. However, illness, childbirth, and aging are events that occur regardless of the economic system, and there is always a need to protect affected individuals. Chapter 12 first provides an overview of the underlying economic and social environment in Russia. This chapter does not go into detail about the Soviet era, but focuses on the socio-economic shocks Russia has experienced after the collapse of the Soviet Union. Thus, this chapter describes changes in the socio-economic environment that occurred after the collapse of the Soviet Union in late 1980s and changes since 2000, when sustainable economic growth was first seen. In addition, the ultimate aim of social security can be considered the reduction of poverty. This chapter also reviews increasing levels of poverty and economic inequality in Russia, since the collapse of the Soviet Union. It describes trends in causes of death and decreasing childbirth, which contribute to the population aging of modern Russia.

1.3

Significance and Contributions of the Book

First, this book uses national individual data, household long-term crosssectional survey data, and longitudinal survey data to investigate the mechanisms of individuals and firms, as related to the employment, retirement, and well-being (e.g., income inequality, living arrangement) of the older adults in East Asia (China in Chapters 2–5, Japan in Chapters 6– 9, and other regions including Republic of Korea, Taiwan, Russia in Chapters 10–12). The comparative empirical studies among countries and regions in East Asia can provide a picture of employment and lifestyle in

1

INTRODUCTION

15

aging East Asia. Furthermore, we can understand the myriad of challenges facing aging populations in the societies in each country. Second, from the labor demand perspective, empirical studies (Hutchens 1986; Heywood et al. 2010; Hirsch et al. 2000; Zwick 2012; Frimmel et al. 2018) on the influence of wage and employment institutions on employment of older adults for the developed countries found that the firms may affect the employment of older workers. Yet, evidence for East Asia is scarce. Using the data from the firm survey, this book also investigates Japanese firm behaviors, such as the association between seniority wage and the mandatory retirement system (Chapter 8), and using the Chinese employer-employee matched survey data, this book analyzes the influence of social insurance contributions on firms’ employment and wage in China (Chapter 4) which have been neglected in most studies of East Asia. Third, there are few empirical studies on the work skills gap among young, middle-aged, and older adults (Chapter 3), health-based work capacity of older adults (Chapter 6), social participation of older adults (Chapter 7), and the timing of retirement and post-retirement employment in East Asia (Chapter 11). This book fills those knowledge gaps. These issues are related to the arguments for the employability of post-retirement older adults (Sullivan and Ariss 2019). Fourth, we also focus on the family structure and its impact on employment and lifestyle in East Asia (Chapters 9 and 11). Finally, the latest national individual and household survey data can provide the newest information to understand the current work and lifestyles of older adults in East Asia. This text can enrich academic research and provide evidence for policymakers.

Note 1. EBPM is an approach that helps people make well-informed decisions about policies, programs, and projects. It puts the best available evidence from research at the heart of policy development and implementation (Davies et al. 2000; Davies 2004).

16

X. MA

References Davies, H. T. O. (2004). Is evidence-based government possible? Jerry Lee Lecture, presented at the 4th Annual Campbell Collaboration Colloquium, Washington, DC. Davies, H. T. O., Nutley, S. M., & Smith, P. C. (2000). What works? Evidencebased policy and practice in public services. London: P&E. Frimmel, W., Horvath, T., Schnalzenberger, M., & Winter-Ebmer, T. (2018). Seniority wages and the role of firms in retirement. Journal of Public Economics, 164, 19–32. Heywood, J., irjahn, U., & Tsertsvardze, G. (2010). Hiring older workers and employing older workers: German evidence. Journal of Population Economics, 23(2), 595–615. Hirsch, B., Macpherson, D., & Hardy, M. (2000). Occupational age structure and access for older workers. Industrial and Labor Relations Review, 51, 401– 418. Hutchens, R. (1986). Delayed payment contracts and a firm’s propensity to hire older workers. Journal of Labor Economics, 4(4), 439–457. National Bureau of Statistics (NBS). (2019). China Statistical Yearbook 2019. Beijing: China Statistics Press. Sullivan, S., & Ariss, A. A. (2019). Employment after retirement: A review and framework for future research. Journal of Management, 45(1), 262–284. Zwick, T. (2012). Consequences of seniority wages on the employment structure. Industrial and Labor Relations Review, 65(1), 108–125.

PART I

Employment, Retirement, and Income Inequality of the Elderly in China

CHAPTER 2

Health and Employment of the Younger, Middle-Aged, and Older Adults in China Xinxin Ma and Jingwen Zhang

2.1

Introduction

According to human capital theory (Becker 1964; Mincer 1974), poor health may lead to decreased labor force participation or lower performance at work, resulting in an inevitable loss of productivity for both individuals and firms. Health status can also be considered a main factor that influences the participation of older people. Considering the rapid progress of population aging and the rising burden of public medical expenditures faced by governments, it is important for policymakers to develop a comprehensive understanding of the impact of health on labor force participation, which may influence sustainable economic development in the long term.

X. Ma (B) Faculty of Economics, Hosei University, Tokyo, Japan e-mail: [email protected] J. Zhang School of Social Sciences, University of Manchester, Manchester, UK © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 X. Ma (ed.), Employment, Retirement and Lifestyle in Aging East Asia, Social Policy and Development Studies in East Asia, https://doi.org/10.1007/978-981-16-0554-3_2

19

20

X. MA AND J. ZHANG

There is abundant literature examining the relationship between health and the labor supply for developed countries. Previous studies indicate that there may be two-way causality. On the one hand, health may affect labor supply and/or labor productivity (Kessler and Frank 1997; Pelkowski and Berger 2004; Chatterji et al. 2007, 2011; Darr and Johns 2008; Johns 2010; Cai 2010; Tefft 2012; Halla and Zweimuller 2013; Lundbor et al. 2015; Dawson et al. 2015; Barnay 2016; Bubonya et al. 2017). On the other hand, many studies found that working conditions also influence the health status of workers (Otsuka et al. 2009; Ma 2009; Artazcoa et al. 2013, 2016; Bannai and Tamakoshi 2014; Song et al. 2014; Cho et al. 2015; Barnay 2016; Afonso et al. 2017; Cayuela et al. 2018; Cygan-Rehm and Wunder 2018). Furthermore, there may be unobservable factors, often called “third factors,” that could affect both health and participation in work. Therefore, the endogeneity problem should be addressed carefully when analyzing the influence of health on the labor supply. In contrast to the rich literature on developed countries, empirical studies on this issue for China are still scarce, and most previous studies on China did not address the endogeneity problem. This study investigates the impact of health on participation in work, employment status (becoming a regular or an non-regular worker), and the number of working hours in China and compares the health effect on labor supply between age groups. This study makes four useful contributions to this debate. First, previous studies generally focus on a segment of the Chinese population, such as the elderly or rural residents. The current study analyzes the population aged from 18 to 60, which includes younger, middle-aged, and older individuals. The results provide useful new evidence because the whole national population and full range of working ages are examined. We also compared younger and middle-aged groups. Second, we use a nationwide longitudinal survey to construct the lagged health status variables in the previous survey year to address the reverse causality problem and a fixed effects model to address the individual heterogeneity problem.1 Third, although previous studies usually use work participation and working hours as indicators of labor force participation, it is thought that health may also affect employment status (e.g., to becoming regular worker or non-regular worker). This study investigates the influence of health on employment status, particularly “the probability of becoming a regular worker.” To the best of our knowledge, this is the first study on this issue in China. Fourth, to consider the

2

HEALTH AND EMPLOYMENT OF THE YOUNGER, MIDDLE-AGED …

21

heterogeneity for different groups, estimations are used for gender and age groups. The remainder of this chapter is organized as follows. Section 2.2 provides and discusses the possible channels of the influences of health on labor force participation and summarizes the results of previous studies. Section 2.3 explains the methodology, including the models and data. Section 2.4 presents the estimation results, and the conclusions of this study are presented in Sect. 2.5.

2.2

Literature Review

There is a large body of literature on the impact of health on labor force participation from both the economics and psychological/medical perspectives (Ettner et al. 1997; Hamilton et al. 1997; Kessler and Frank 1997; Chatterji et al. 2007, 2011; Darr and Johns 2008; Johns 2010; Tefft 2012; Frijter et al. 2014; Dawson et al. 2015; Bubonya et al. 2017). The results of previous studies generally indicate that illness negatively affects work participation and working hours. For example, Chatterji et al. (2011) found that for males, psychiatric disorder is associated with a 9% points reduction in the likelihood of current labor force participation and 14% in the likelihood of future employment, and 19% and 13% points reductions in these outcomes among females, respectively. Tefft (2012) analyzed the impact of health-related quality of life (HRQOL) on employment using a constructed index related to seasonal affective disorder (SAD) to instrument HRQOL. The results reveal that each additional day of poor mental health per month increases the probability of unemployment by 0.76% points among women. Chatterji et al. (2011) indicated that the effects of health status are greater for men than women. To address the endogeneity of health, previous studies generally rely on an instrumental variable (IV) method and/or the use of longitudinal survey data. Parental psychological problems (e.g., Ettner et al. 1997; Chatterji et al. 2011), individual experiences of illness in the past (Ettner et al. 1997; Hamilton et al. 1997; Chatterji et al. 2007, 2011), degree of religiosity (Alexandre and French 2001; Chatterji et al. 2007), perceived social support (Hamilton et al. 1997; Alexandre and French 2001; Ojeda et al. 2010), participation in physical activity (Hamilton et al. 1997), and days of darkness (Tefft 2012) have been used as IVs in these previous studies.

22

X. MA AND J. ZHANG

In contrast to the literature for developed countries, most previous studies for China show serious methodological problems (Wei 2004; Liu 2008; Lu et al. 2009; Xie 2011; Pan et al. 2011; Liu and Li 2012; Qin et al. 2012, 2015; Zhang et al. 2013; Li et al. 2014; Sun and Feng 2015; Yang et al. 2015; Qin and Wang 2015; Lu 2017; Jiang et al. 2019). First, empirical studies that appropriately address the endogeneity resulting from reverse causality problems are still rare (Xie 2011; Qin et al. 2012; Jiang et al. 2019). Previous studies do not consider the simultaneous occurrence of the heterogeneity and reverse causality problems. Second, most of the objects analyzed in previous studies are often middle-aged and older adults; therefore, it is not clear how health status influences the whole population, including the younger generation in China. Lastly, most previous studies did not address the heterogeneity problem. Unlike previous research, the current study uses nationally representative longitudinal survey data that covers young, middle-aged, and older individuals and includes both urban and rural residents. Lagged variables are used to address the reverse causality problem, the fixed effects (FE) model is used to address the heterogeneity problem, and the Heckman two-step model is used to address sample selection bias. This study investigates the variation in the way health status influences different groups (gender and age).

2.3

Methodology and Data 2.3.1

Model

This study investigates the causal effects of health on three major aspects of labor force participation: (a) work participation, (b) employment status (becoming a regular worker), and (c) working hours. For individual i, the relationship between health status and labor force participation can be described by Eq. 2.1.1. Yi = a + β1H Hi + β1X X i + εi

(2.1.1)

where Yi is a variable that indicates the work status of individual i, that is, participation in work, employment status (becoming a regular worker), or working hours. H represents health status, and X is a set of individual and household characteristics that may affect the labor supply. Specifically, X stands for the control variables (i.e., educational attainment,

2

HEALTH AND EMPLOYMENT OF THE YOUNGER, MIDDLE-AGED …

23

age, ethnicity, family structure, children aged 6 and younger, household income, co-resident with parents, and regional dummies). β is the coefficient to be estimated and εi is an error item. To consider the sample selection bias problem, we also use the Heckman two-step model2 for the working hours estimations (see Eq. 2.1.2). In Eq. 2.1.2, the inverse Mills ratio (λi ) is calculated based on the labor force participation function shown in Eq. 2.1.1. Yi = a + β1H Hi + β1X X i + β1λi + εi

(2.1.2)

If H and X are uncorrelated with the error terms, a standard probit model and an ordinary least squares (OLS) model will produce unbiased estimates for β. Unfortunately, health status is likely to be affected by some factors included in the errors, such as unobservable individual characteristics that can affect both work participation and health, thus resulting in biased estimates. The following three methods are used to address the endogeneity problem. First, as there may be a reverse causality relation between health and work participation, the lagged variable (LV) of health status in the previous survey year is used. It is expressed by Eq. 2.2 Yit = a + β2H Hit−1 + β2X X it + u i

(2.2)

Second, panel data analysis methods (e.g., the fixed or random effects models) are used to address the heterogeneity problem, as shown in Eq. 2.3. Yit = a + β3H Hit−1 + β3X X it + εi + vit

(2.3)

In Eq. 2.2, u it is composed of an unobserved time-invariant individual effect (εi ) and true error term (vit ) because εi may remain in Eq. 2.2, which may cause an estimation bias. The FE model, or the random effects (RE) model expressed by Eq. 2.3, was used to address the heterogeneity problem. Finally, various analyses investigate the heterogeneous effects of health. Specifically, this study explores how these effects vary by gender as well as by age groups (the youth, middle age, and older age groups), educational attainment groups (low, middle, and high educational level groups), and Hukou type (rural residents and urban residents).

24

X. MA AND J. ZHANG

2.3.2

Data and Variables

The data for this study were obtained from the China Health and Nutrition Survey (CHNS) from 1997 to 2006.3 CHNS is a nationwide longitudinal survey conducted by the Carolina Population Center at the University of North Carolina and the National Institute for Nutrition and Health (NINH, former National Institute of Nutrition and Food Safety) at the Chinese Center for Disease Control and Prevention (CCDC). The survey uses a multistage, random cluster process to draw samples from 15 provinces and municipal cities (Beijing, Liaoning, Heilongjiang, Shanghai, Jiangsu, Shandong, Henan, Hubei, Hunan, Guangxi, Guizhou, and Chongqing) that vary substantially in geography, economic development, and government public resources. The CHNS provides extensive information about health as a basic social demographic and employment information, which enables the empirical analysis herein. The following three types of variables of work participation are used as the dependent variables: 1. A binary variable that is equal to 1 when an individual worked in the survey year, and 0 otherwise. 2. Employment status is a binary variable that is equal to 1 when the individual is a regular worker, and equal to 0 when they are an non-regular worker. In this study, irregular work was defined by the employment position of occupation in the CHNS questionnaire. Notably, non-regular workers comprise those self-employed workers with no employees (including farmers), temporary, paid family, and unpaid family workers. Regular workers comprise the self-employed who employ other workers, those who are employed in a workplace (i.e., firms, organizations) as a permanent employee, or those with a long-term labor contract. 3. Working hours are calculated from the information in the CHNS.4 The main indicator of health status is the SRH, which is a binary dummy variable that is equal to 1 when the answer is “excellent” or “good,” while it is equal to 0 when “fair” and “poor.” based on the questionnaire item “How do you feel about your health status?”. Five covariates are used to control for other influences on labor force participation. First, at the individual level, according to human capital

2

HEALTH AND EMPLOYMENT OF THE YOUNGER, MIDDLE-AGED …

25

theory (Becker 1964; Mincer 1974), an individual’s wage level is determined by human capital such as education or years of work experience. In addition, wage level is an important determinant of the labor supply. Therefore, this study uses the number of years of schooling, age, and age squared as the proxy variables of human capital. The decision about labor supply may vary by ethnic groups, rural, and urban residents. Thus, a dummy variable for ethnic groups (Han = 1, ethnic minority = 0) and household registration (Hukou) type (rural = 1, urban = 0) are controlled. Second, for the family factors, (1) Douglas (1934) and Arisawa (1956) suggest that women’s labor supply behaviors are influenced not only by the wage level in the labor market but also by family factors such as household income. Hence, annual household income (excluding the respondent’s own income) is included in the model. (2) Co-residence with parents may increase household income as well as the availability of childcare support, but it may also generate new consumption for the household and demand for elderly care. It has been found that living with parents influences women’s labor supply (Ogawa and Ermisch, 1996; Compton and Pollak, 2014; Shen et al. 2016). Additionally, there is a gender difference in terms of grandchild care support. In particular, women will be more involved in taking care of young children than men. Therefore, two dummy variables are used in our models: the coresidence-with-mother dummy and the co-residence-with-father dummy. (3) Having a child aged 6 and younger dummy variable is constructed to control for the influence of childcare on work participation. Third, based on the labor market segmentation hypothesis (Piore, 1970), work status may be influenced by the characteristics of the labor market. Therefore, a variety of ownership type dummy variables5 are used to control the influence of labor market segmentation on working hours. Fourth, the survey year dummy variables were constructed to control the government policy and economic environment influences. Finally, culture, lifestyle, and economic development level may differ in each region; hence, four region dummy variables (Eastern Region, Central Region, Western Region, and Northeast Regions) are used to capture the regional disparity. The youngest legal working age is 16 years, and the oldest eligible mandatory retirement age6 is 60 years, and those within this range are selected as the analyzed sample. The total number of observations was 35,729. The sample aged 16–50 were used to conduct robustness checks.

26

X. MA AND J. ZHANG

The descriptive statistics for the main variables used in the analysis are summarized in Table 2.5.

2.4 2.4.1

Results Baseline Results

The results of basic estimates of the influence of health on labor force participation are summarized in Table 2.1. First, the results based on the probit regression model, LV probit regression, and the FE model indicate that SRH positively affects work participation. Second, the results based on the probit regression model and LV model indicate that the SRH positively affects the probability of becoming a regular worker, but it is not statistically significant in the results of the FE model. Third, the results based on the LV model and Heckman two-step model indicate that healthy individuals work for more hours, but the influence of health is not statistically significant in the results of the FE model. The results of the F-test and Hausman test7 indicate that the FE model is more appropriate than the other models. Therefore, there may be a bias in the results for the effects of health on employment status (to become a regular worker) and working hours. The effects of subjective health status may be overestimated in the results based on the probit regression model or the LV probit regression model, which does not address the heterogeneity problem. When the heterogeneity problem is addressed, health positively affects work participation. Concretely, to compare with the poor health group, the probability of work participation is 10.4–12.0% points higher for the healthy group. However, the influences on the probability of becoming a regular worker and working hours are not statistically significant. 2.4.2

Results by Gender

Table 2.2 summarizes the results by gender. (1) The results for the LV model and the FE model show that health positively affects work participation for men, but for women, the effect of health on the labor supply is not statistically significant. Specifically, compared with the poor health male group, the probability of participation in work is 15.5–20.7% points higher for the healthy male group. There is a clear gender difference in the influence of health on work participation, and the influence of health

2

HEALTH AND EMPLOYMENT OF THE YOUNGER, MIDDLE-AGED …

Table 2.1 Basic estimates of the influence of health on labor force participation in China

27

(a) Participation in work (1) Probit SRH

(2) LV

(3) FE

0.108*** (0.017)

SRHt − 1 Observations Groups

0.104* (0.053)

35,729

0.120*** (0.021) 22,813

10,748

(1) Probit

(2) LV

(3) FE

(b) Regular worker

SRH

0.0805*** (0.024)

SRHt − 1 Observations Groups

0.122 (0.090) 0.115*** (0.032) 14,274

23,370

3,489

(c) Working hours (1) LV

athrho lnsigma Observations Groups

(3) Heckman

−0.158 (−0.470)

SRH SRHt − 1

(2) FE

1.120*** (0.405)

10,142

15,073 9,199

1.197*** (0.405) 0.008 0.138 18,088

Note 1. ***p < 0.01, **p < 0.05, *p < 0.10; Clustered standard errors are in parentheses 2. SRH: self-reported health; LV: lagged variable of SRH in the prior survey year is used in the probit regression model; FE: fixed effects model: Heckman: Heckman two-step estimation. The other variables including the gender, age, education, ethnic, urban, co-residence with father, co-residence with mother, youngest child aged 6 and younger, household income, region, and survey year are estimated in labor force participation and regular worker functions. In addition, the occupation dummy variables are added and estimated in the working hours functions. The results of these variables are not expressed in this table Source Author’s creation based on the data from CHNS 1997–2006

28

X. MA AND J. ZHANG

Table 2.2 Estimates of the influence of health on labor force participation by gender in China (a) Participation in work (1) LV Men

(2) FE Women

SRH SRHt − 1

0.207*** (0.031) 11,169

Observations Groups

0.041 (0.028) 11,644

Men

Women

0.155* (0.081)

0.071 (0.071)

4,548 1,468

6,200 1,954

(b) Regular worker (1) LV

(2) FE

Men

Women

SRH SRHt − 1

0.155*** −0.042 7,760

Observations Groups

0.054 (0.049) 6,514

Men

Women

0.116 (0.110)

0.141 (0.158)

2,382 832

1,107 406

(c) Working hours (1) LV Men

(2) FE Women

SRH SRHt − 1 athrho lnsigma Observations Groups

0.993* (0.560)

1.199** (0.587)

5,623

4,519

Source and Note See Table 2.1

(3) Heckman

Men

Women

−0.116 (−0.626)

−0.153 (−0.712)

8,419 4,903

6,654 4,296

Men

Women

0.569 (0.639) −0.269 −4.608 8,682

1.309*** (0.561) 0.80 1.343 9,406

2

HEALTH AND EMPLOYMENT OF THE YOUNGER, MIDDLE-AGED …

29

status is greater for men than for women. These results are consistent with those of Chatterji et al. (2011) and Xie (2011). (2) Although the results for the LV model indicate that health may positively affect the probability of becoming a regular worker for men, the results for the FE model indicate that the influence of health status on the probability of becoming a regular worker is not statistically significant for both men and women. (3) The results for the LV model and Heckman two-step model show that both healthy men and healthy women may work for more hours. The results for the FE model indicate that the influence of health on work participation is not statistically significant for both men and women. These results indicate that when the heterogeneity problem is addressed, health only affects work participation for men, and the influence of health on the probability of becoming a regular worker and the number of working hours are not statistically significant for both men and women. 2.4.3

Results by Different Age Groups

Table 2.3 summarizes the results using three subsamples: the younger generation aged 30 and younger, the middle-aged generation aged 30– 50 years, and the older generation aged 50 and older. First, the results for the LV model show that health may positively affect work participation for the younger, middle-aged, and older generations. The results for the FE model indicate that the influence of health status on work participation is not statistically significant for the younger, middle-aged, and older generations. This suggests that the influence of heterogeneity on work participation is large for each generation. Second, the results for the LV model show that health may positively affect the probability of becoming a regular worker for the middle-aged and older generations, but the results for the FE model indicate that the positive influence of health on the probability of becoming a regular worker only exists for the younger generation at a statistically significant level of 10%. Specifically, compared with the poor health group, the probability of becoming a regular worker is 42.1% points higher for the healthy group aged 30 and younger. Third, the results for the LV model and Heckman two-step model reveal that the younger generation of healthy individuals may work for longer hours, while the results for the FE model indicate that

Observations Groups

SRHt − 1

SRH

(b) Regular worker

Observations Groups

SRHt − 1

SRH

0.106 (0.078) 2,573

0.0883** (0.043) 7,600

0.165** (0.065) 4,101

591 251

0.421* (0.233)

50

30 X. MA AND J. ZHANG

1,612

3.045*** (1.173)

5,595

0.509 (0.545)

2,935

1.180* (0.698)

2,991 2,308

1.196 (1.486)

(2) FE < 30

> 50

< 30

age 30–50

(1)LV

Sources and Note See Table 2.1

athrho lnsigma Observations Groups

SHSt-1

SHS

(c) Working hours

3,711 2,601

−0.0801 (−1.064)

−0.893 (−0.632)

8,371 5,249

> 50

age 30–50

3.215*** (1.096) −0.103 2.743*** 2,640

0.516 (0.526) −0.099 2.819*** 7,283

age 30–50

(3) Heckman < 30

1.444** (0.710) 0.404 2.902*** 8,165

> 50

2 HEALTH AND EMPLOYMENT OF THE YOUNGER, MIDDLE-AGED …

31

32

X. MA AND J. ZHANG

the influences of health are not statistically significant for the younger, middle-aged, and older generations. These results suggest that when the heterogeneity problem is addressed, subjective health status only affects the probability of becoming a regular worker for the younger generation, and there are only negligible differences in the influences of health on work participation and working hours between the younger, middle-aged, and older generations. 2.4.4

Robustness Checks Using Different Analyzed Samples

In China, based on the public pension policy and mandatory retirement system, the eligible mandatory retirement age in the public sector (i.e., government organizations, state-owned enterprises) is 50 for a female worker and 55 for a female cadre, and 60 for both male workers and male cadres. The subsample aged 16–50 years are used to reduce the influence of the mandatory retirement system on work participation. The results are summarized in Table 2.4. It is confirmed once more that although the results for the probit regression, the LV, and the Heckman two-step models show that health positively affects work participation, when the FE model is used to address the heterogeneity problem, the effects of health are not statistically significant. These results are consistent with most of the results reported in Table 2.1. It should be noted that the results for the FE model indicate that health may positively affect work participation at the 10% level for the subsample aged 16–60 years (Table 2.1), while the statistical significance disappears for individuals aged 16–50 years (Table 2.4). These results indicate that the positive influence of health on work participation may be greater for the older generation than for the younger and middle-aged generations.

2.5

Conclusions

This study investigates the influence of health on labor force participation in China. It uses nationwide longitudinal survey data from the CHNS from 1997 to 2006. The lagged variable probit regression model (LV) is used to address the reverse causality problem, the FE model is used to address the heterogeneity problem, and the Heckman two-step model is used to address the sample selection problem in the working hours function. The main findings are as follows.

2

HEALTH AND EMPLOYMENT OF THE YOUNGER, MIDDLE-AGED …

Table 2.4 Estimates of the influence of health on labor force participation of individuals aged 16–50 in China

33

(a) Participation in work (1) Probit SRH

(2) LV

0.068*** (0.022)

SRHt − 1 Observations Number of groups

(3) FE 0.005 (0.070)

24,185

0.130*** (0.028) 14,578

(1) Probit

(2) LV

5,788 3,422

(b) Regular worker

SRH

0.074*** (0.027)

SRHt − 1 Observations Number of groups

0.114 (0.098) 0.095*** (0.035) 11,179

18,987

(3) FE

3,004 1,082

(c) Working hours (1) LV

(2) FE

(3) Heckman

−0.281 (0.521)

SRH SRHt − 1

1.076** (0.464)

Athrho Lnsigma Observations Number of groups

12,390

7,988 9,199

0.978** (0.453) −0.159 2.807*** 10,939

Source and Note See Table 2.1

First, when the heterogeneity problem is addressed, health positively only affects work participation. Specifically, compared with the poor health group, the probability of participation in work is 10.4–12.0% points higher for the healthy group. However, the influences on the probability of becoming a regular worker and working hours are not statistically significant. Second, the influences of health on labor force participation differ by each group. (1) Health status positively affects the probability of work

34

X. MA AND J. ZHANG

participation solely for men. Concretely, to compare with the poor health male group, the probability of participation in work is 15.5–20.7% points higher for the healthy men group. (2) Health status positively affects the probability of becoming regular workers for the younger generation. The probability of becoming a regular worker is 42.1% points higher for the healthy younger generation. Third, the robustness checks using different samples (i.e., samples aged 16–50) confirm the main findings for the probability of becoming a regular worker and for longer working hours once more. This indicates that health status positively affects labor force participation even when the heterogeneity problem is addressed. China has an aging population; thus, maintaining and increasing labor force participation now presents an important challenge for the government. It can be expected that the public healthcare insurance policy may positively affect the health status of the population and the desired increase in labor force participation (Ma 2015). The public medical insurance system should be supported by the Chinese government to improve the health status of the whole population in China, which may increase the labor supply for the long term.

Appendix See Table 2.5. Table 2.5 Descriptive statistics of variables (a) Healthy group

(b) Poor health group

Variables

Means

S.D

Means

Labor participation Regular worker Working hours Female Age

0.707

0.455 0.558

0.350 25.232 0.488 41.253

0.477 24.654 0.500 15.264

0.257 16.509 0.567 52.002

D

S.D 0.497 0.437 22.853 0.496 15.969

t-test p-value

0.149

30.494***

0.093 14.816*** 8.723 30.324*** −0.079 −15.229*** −10.749 −66.599***

(continued)

2

HEALTH AND EMPLOYMENT OF THE YOUNGER, MIDDLE-AGED …

35

Table 2.5 (continued)

Variables Education Primary Junior high school Senior high school and higher Han Urban Household Income (10 thousand RMB) Co-residence with mother Co-residence with father Child aged 6 and younger Region Eastern Central Western Northeastern Survey year 1997 2000 2004 2006 Observations

(a) Healthy group

(b) Poor health group

D

Means

S.D

Means

0.359 0.384

0.003 0.554 0.003 0.268

0.004 0.004

0.257

0.003 0.178

0.003

0.079

0.877 0.407 0.586

0.328 0.873 0.491 0.413 0.805 0.561

0.333 0.492 0.795

0.004 −0.006 0.025

0.260

0.439 0.110

0.313

0.151

36.956***

0.295

0.456 0.137

0.344

0.157

36.386***

0.152

0.359 0.080

0.271

0.072

21.419***

0.242 0.338 0.226 0.194

0.003 0.003 0.002 0.002

0.191 0.347 0.293 0.169

0.003 0.004 0.004 0.003

0.051 12.148*** −0.009 −1.913* −0.067 −15.272*** 0.025 6.257***

0.300 0.244 0.234 0.222 28,210

0.003 0.003 0.003 0.002

0.202 0.236 0.283 0.279 14,833

0.003 0.003 0.004 0.004

0.098 22.042*** 0.008 1.827* −0.049 −11.119*** −0.057 −13.238***

S.D

Note ***p < 0.01, **p < 0.05, *p < 0.10; D = a − b Source Author’s creation based on the data from CHNS 1997–2006

t-test p-value

−0.195 −38.3749*** 0.116 23.516*** 18.031***

1.192 1.118 2.931**

36

X. MA AND J. ZHANG

Notes 1. As summarized in Sect. 2.2, in previous studies, a set of instrumental variables (IVs), such as parental psychological problems, individual experiences of mental illness in the past, degree of religiosity, perceived social support, participation in physical activity, and darkness days, are used as IVs. The information in the CHNS, participation in physical activity, is used as an IV, which is similar to Hamilton et al. (1997) and Zhang et al. (2013). However, the results of the F-test and Hausman test indicate that these instrumental variables lack validity. Faulty instrument variables may cause an enlarged bias; therefore, the results based on these instrumental variable methods are not reported herein. 2. This study uses the regional dummy variables (Eastern region, Central region, and Western region) and ownership sector dummy variables (stateowned enterprises, government organizations, privately-owned enterprises, others) as identification variables for the first-stage estimation (labor force participation function) and two-stage estimation (working hours function) in the Heckman two-step model. 3. The total CHNS survey data obtained were 10 waves, including CHNS 1989, 1991, 1993, 1997, 2000, 2004, 2006, 2009, 2011, and 2015. However, because we can only obtain information on self-rated health status from the date of CHNS from 1997 to 2006, only the CHNS data from 1997 to 2006 were used in this study. 4. Using the CHNS, the number of daily and weekly working hours can be determined, and the number of working hours per week was calculated. 5. The ownership type dummy variables include government organizations, state-owned enterprises (SOEs), privately-owned enterprises (POEs), and others. 6. In China, the eligible mandatory retirement age is 50 for a female worker, 55 for a female cadre, and 60 for both male and male cadres. 7. Based on the results of a Hausman test, the FE model was selected for this study.

References Alexandre, P. K., & French, M. T. (2001). Labor supply of poor residents in Metropolitan Miami, Florida: The role of depression and the co-morbid effects of substance use. The Journal of Mental Health Policy and Economics, 4, 161–173. Afonso, P., Fonseca, M., & Pires, J. F. (2017). Impact of working hours on sleep and mental health. Occupational Medicine, 67 (5), 377–382.

2

HEALTH AND EMPLOYMENT OF THE YOUNGER, MIDDLE-AGED …

37

Arisawa, H. (1956). Wage structure and economic structure: The significance and background of low wage. In I. Nagayama (Ed.), Wage basic survey (pp. 40– 57). Tokyo: Toyo Kaizei Press. (In Japanese). Artazcoa, L., Cortes, I., Bernavides, F. G., Escriba-Aguir, V., & Bartoll, X. (2013). Long working hours and health in Europe: Between-country differences. Scandinavian Journal of Work, Environment & Health, 39(4), 369–378. Artazcoa, L., Cortes, I., Bernavides, F. G., Escriba-Aguir, V., & Bartoll, X. (2016). Long working hours and health in Europe: Gender and welfare state differences in a context of economic crisis. Health & Place, 40, 161–168. Bannai, A., & Tamakoshi, A. (2014). The association between long working hours and health: A systematic review of epidemiological evidence. Scandinavian Journal of Work, Environment & Health, 40(1), 5–18. Barnay, T. (2016). Health, work and working conditions: A review of the European economic literature. The European Journal of Health Economics, 17, 693–709. Becker, G. S. (1964). Human capital: A theoretical and empirical analysis, with special reference to education. New York: Columbia University Press. Bubonya, M., Cobb-Clark, D. A., & Wooden, M. (2017). Mental health and productivity at work: Does what you do matter? Labour Economics, 46, 150– 165. Cai, L. (2010). The relationship between health and labor force participation: Evidence from a panel data simultaneous equation model. Labour Economics, 17, 77–90. Cayuela, A., Martinez, J. M., Ronda, E., Delclos, G. L., & Conway, S. (2018). Assessing the influence of working hours on general health by migrant status and family structure: The case of Ecuadorian-, Colombian-, and Spanish-born workers in Spain. Public Health, 163, 27–34. Chatterji, P., Alegria, M., Lu, M., & Takeuchi, D. (2007). Psychiatric disorders and labor market outcomes: Evidence from the National Latino and Asian American study. Health Economics, 16(10), 1069–1090. https://doi.org/10. 3386/w14404. Chatterji, P., Alegria, M., Lu, M., & Takeuchi, D. (2011). Psychiatric disorders and labor market outcomes: Evidence from the national comorbidity surveyreplication. Journal of Health Economics, 30(5), 858–868. Cho, S., Li, M., Kim, K., Ju, Y., Paek, D., & Lee, W. (2015). Working hours and self rated health over 7 years: Gender differences in a Korean longitudinal study. BMC Public Health, 15(1), 1–11. Compton, J., & Pollak, R. A. (2014). Family proximity, childcare, and women’s labor force attachment. Journal of Urban Economics, 79, 72–90.

38

X. MA AND J. ZHANG

Cygan-Rehm, K., & Wunder, C. W. (2018). Do working hours affect health? Evidence from statutory workweek regulations in Germany. Labour Economic, 53, 162–171. Darr, W., & Johns, G. (2008). Work strain, health, and absenteeism: A metaanalysis. Journal of Occupational Health Psychology, 13(4), 293–318. Dawson, C., Veliziotis, M., Pacheco, G., & Webber, D. J. (2015). Is temporary employment a cause ore consequence of poor mental health? A panel data analysis. Social Science & Medicine, 134, 50–58. Douglas, P. H. (1934). The theory of wages (pp. 279–294). New York: The MacMillan Company. Ettner, S. L., Frank, R. G., & Kessler, R. C. (1997). The impact of psychiatric disorders on labor market outcomes. ILR Review, 51(1), 64–81. Frijter, P., Johnston, D. W., & Shields, M. A. (2014). The effect of mental health on employment: Evidence from Australian panel data. Health Economics, 23, 1058–1071. Halla, M., & Zweimuller, M. (2013). The effect of health on earnings: Quasiexperimental evidence from commuting accidents. Labour Economics, 24, 23– 38. Hamilton, V. H., Merrigan, P., & Dufresne, E. (1997). Down and out: Estimating the relationship between mental health and unemployment. Health Economics, 6, 397–406. Jiang, J., Huang, W., Wang, Z., & Zhang, G. (2019). The effect of health on labor supply of rural elderly people in China: An empirical analysis using CHARLS data. International Journal of Environmental Research and Public Health, 16(1195), 1–15. Johns, G. (2010). Presenteeism in the workplace: A review and research agenda. Journal of Organizational Behavior, 31(4), 519–542. Kessler, R. C., & Frank, R. G. (1997). The impact of psychiatric disorders on work loss days. Psychological Medicine, 27 (4), 861–873. Li, Q., Lei, X., & Zhao, Y. (2014). The effect of health on the labor supply of mid-aged and older Chinese. China Economic Quarterly, 13(3), 917–938. (In Chinese). Liu, S. (2008). Impact of health on labor force participation of rural residents. Chinese Rural Economics, 8, 25–33. (In Chinese). Liu, S., & Li, J. (2012). Health, labor force participation and elderly poverty in rural China. Chinese Rural Economics, 1, 56–68. (In Chinese). Lu, C., Frank, R. G., Liu, Y., & Shen, J. (2009). The impact of mental health on labour market outcomes in China. Journal of Mental Health Policy and Economics, 12(3), 157–166. Lu, H. (2017). Impact of farmers’ health on off-farm labor supply. Journal of Agro-Forestry Economics and Management, 16(4), 454–461. (In Chinese).

2

HEALTH AND EMPLOYMENT OF THE YOUNGER, MIDDLE-AGED …

39

Lundbor, P., Nilsson, M., & Vikstrom, J. (2015). Heterogeneity in the impact of health shocks on labor outcomes: Evidence from Swedish workers. Oxford Economic Papers, 67 (3), 715–739. Ma, X. (2009). Does long working hours cause to the mental health problem? In Y. Higuchi, Y. Seiko, & H. Terayama (Eds.), The dynamic of household behavior in Japan. Tokyo: Keio University Press. (In Japanese). Ma, X. (2015). Public health care insurance system reform in China. Kyoto: Kyoto University Press. (In Japanese). Mincer, J. (1974). Schooling, experience and earning. New York: Columbia University Press. Ogawa, N., & Ermisch, J. F. (1996). Family structure, home time demands, and the employment patterns of Japanese married women. Journal of Labor Economics, 14(4), 677–702. Ojeda, V. D., Frank, R. G., McGuire, T. G., & Gilmer, T. P. (2010). Mental illness, nativity, gender and labor supply. Health Economics, 19(4), 396–421. Otsuka, Y., Sasaki, T., Iwasaki, K., & Mori, I. (2009). Working hours, coping skills, and psychological health in Japanese daytime workers. Industrial Health, 47 (1), 22–32. Pan, J., Qin, X., & Liu, G. (2011). Does body size matter in urban employment? Evidence from China. Nankai Economic Studies, 2, 68–85. (In Chinese). Pelkowski, J. M., & Berger, M. C. (2004). The impact of health on employment, wages, and hours worked over the life cycle. The Quarterly Review of Economics and Finance, 44(1), 102–121. Piore, M. J. (1970). Job and training. In S. H. Beer & R. Barringer (Eds.), The state and the poor. Cambridge, MA: Winthrop. Qin, L., Cheng, J., & Pan, J. (2015). Impact of health on labor supply of China’s floating workers. Statistics & Information Forum, 30(3), 103–108. (In Chinese). Qin, L., Qin, X., & Jiang, Z. (2012). Impact of health on off-farm working hours of migrants. Chinese Rural Economics, 8, 38–45. (In Chinese). Qin, L., & Wang, Z. (2015). Impact of health status and related factors on labor market positions of urban mature Chinese. Journal of Labor Research., 36(2), 224–231. (In Chinese). Shen, K., Yan, P., & Zeng, Y. (2016). Coresidence with elderly parents and female labor supply in China. Demographic Research, 35, 645–670. Song, J., Lee, G., Kwon, J., Park, J., Choi, H., & Lim, S. (2014). The association between long working hours and self-rated health. Annals of Occupational and Environmental Medicine, 26(2), 1–12. Sun, D., & Feng, Z. (2015). Impact of health on non-farm employment in rural China: Efficiency effect and allocation effect: A case from Guannan and Xinyi of Jiangsu province. Issues in Agricultural Economy, 8, 28–34. (In Chinese).

40

X. MA AND J. ZHANG

Tefft, N. (2012). Mental health and employment: The SAD story. Economics & Human Biology, 10(3), 242–255. Wei, Z. (2004). The role of health on off-farm employment and wage decision. Economic Research Journal, 2, 64–74. (In Chinese). Xie, E. (2011). Impact of health on retirement. World Economics Paper, 1, 109– 120. (In Chinese). Yang, Z., Maierdan, T., & Wang, Y. (2015). Impact of health shock on agricultural labor supply of middle-aged and older rural residents’ an empirical study based on CHARLS. China Rural Survey, 3, 24–37. (In Chinese). Zhang, Y., Li, C., Liu, C., & Peng, S. (2013). The impact of health on work in China: A study using pilot survey data. The Geneva Papers on Risk and Insurance Issues and Practice, 38(4), 857–870.

CHAPTER 3

Work Skills Gap and the Wage Differentials Between the Young, Middle-Aged, and Older Workers in China Xinxin Ma and Xiaobo Qu

3.1

Introduction

Since the 1980s, the wage gap (or income inequality) between highly skilled workers and low-skilled workers has increased in developed economies like those of the USA and Europe. Autor et al. (2003), Autor and Dorn (2009, 2013), Acemoglu and Autor (2011), and Autor (2013) argue that advances in information and communication technology (ICT) contribute to the expansion of the wage gap. Following the task-based approach,1 a number of studies investigate how task type influences wage inequality, and find that the changes of employment share and wage levels

X. Ma (B) Faculty of Economics, Hosei University, Tokyo, Japan e-mail: [email protected] X. Qu Institute of Population and Labor Economics, Chinese Academy of Social Sciences, Beijing, China e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 X. Ma (ed.), Employment, Retirement and Lifestyle in Aging East Asia, Social Policy and Development Studies in East Asia, https://doi.org/10.1007/978-981-16-0554-3_3

41

42

X. MA AND X. QU

differ by task type (Autor et al. 2003; Goos and Manning 2007; Autor and Dorn 2009; Antonczyk et al. 2010; Acemoglu and Autor 2011; Autor and Price 2013; Autor and Dorn 2013; Autor 2013; Goos et al. 2009, 2014; Arias et al. 2014; Dicarlo et al. 2016). Peng et al. (2017) examine the impact of ICT on the demand for older workers (aged 50 and over), and finds the evidence of a decrease in demand for older workers in the 1970s and 1980s in nine European countries. Autor and Dorn (2009) and Lewandowski et al. (2017) indicate that age structures differ by task-specific skills, for example, the proportion of young workers is higher for non-routine tasks, and on the contrary, the proportion of older workers is higher for routine task. It is thought that the depreciation of skills over the life cycle may contribute to the work skills gap between the young, middle-aged, and older worker groups. Therefore, it is expected that the differences of task-specific skills for different age groups may contribute to the wage gap between different age groups (Desjardins and Warnke 2012; Lewandowski et al. 2017). How does the work skills gap between different age groups influence wage gaps in China? This chapter uses the unique survey data from the China Urban Labor Survey (CULS) conducted in 2016, and decomposition models, to investigate the influences of work skill levels on the wage gaps between three groups: (i) the younger group aged 16–49; (ii) the middle-aged group aged 50–59, and (iii) the older group aged 60 and over from a task-specific skills approach. This is the first study to explore this issue for China. The results provide new evidence that enables us to explore the relationship between work skill and wage gaps. The remainder of this chapter is structured as follows. Section 3.2 summarizes economic theory about the wage gap, and previous empirical studies on the correlation between task-specific skills and the wage gap. Section 3.3 describes the methods of analysis, including an introduction to models and data. Section 3.4 gives the calculated results, and Sect. 3.5 states and interprets the econometric results. The last section presents the four main conclusions.

3

WORK SKILLS GAP AND THE WAGE DIFFERENTIALS BETWEEN …

3.2 3.2.1

43

Literature Review

General Economic Theories That Explain the Influence of Task-Specific Skills on Wage Gap Between Different Age Groups

What causes the wage gap between different age groups? Two economic theories go some way to explain it. First, human capital theory (Becker 1964; Mincer 1974) suggests that in a perfectly competitive labor market the individual wage level is determined by a worker’s labor productivity. Labor productivity is in a positive relation with a worker’s human capital (e.g., years of schooling, years of experience). Task-specific skills can be thought of as a part of human capital, hence it is expected that when there is task skill gap between young, middle-aged, and older workers, a wage gap may occur. Second, according to the discrimination hypothesis (Becker 1957), discrimination against older workers may be shown by employers, customers, and colleagues: this discrimination can cause a wage gap between different age groups. For example, even though older workers perform a task that is similar with a task performed by their colleagues who are middle-aged workers, discrimination against older workers may cause the wage level to be set lower for the older workers than for middle-aged workers. Therefore when the wage set for the task skill levels differs for these two groups there may be a wage gap between the middle-aged group and the older group. 3.2.2

Empirical Studies of Task Types and Their Influence on Wage and Employment

Published empirical studies on the relationship between task skills and wage or employment can be summarized as follows. Existing published research about the evolving relationship between task and employment finds that since the 1970s tasks have changed from manual tasks and routine cognitive work toward non-routine cognitive work. The proportion of highly skilled workers with non-routine task and low-skilled workers with routine manual task has grown while the proportion of middle-skilled workers performing routine cognitive task has declined. This resulted in a polarization of employment in the USA and other OECD countries (Autor et al. 2003; Goos and Manning 2007; Autor and Dorn 2009; Antonczyk et al. 2010; Acemoglu and Autor 2011; Autor and Price 2013; Autor and Dorn 2013; Autor 2013;

44

X. MA AND X. QU

Goos et al. 2009, 2014; Arias et al. 2014; Dicarlo et al. 2016). The demands of particular occupations or tasks have changed accordingly (Autor et al. 2003; Spitz-Oener 2006). Similar patterns have been identified in middle-income or emerging market countries, although routine cognitive employment has either remained stable or even increased in the southern emerging economies like India and Brazil (Aedo et al. 2013), in Russia (Gimpelson and Kapeliushnikov 2016), and in the economies of Central and Eastern Europe (Hardy et al. 2017). Adopting the SBTC (skill biased technology change) approach to explore the relationship between task and wage structure, Gathmann and Schonberg (2010) find task-specific human capital is an important factor for individual wage growth, accounting for up to 52% of overall wage growth in the USA. Antonczyk et al. (2009) investigate the changes in the German wage structure for full-time working males from 1999 to 2006. Using the Blinder-Oaxaca type decomposition model, they decompose the changes in the entire wage distribution between 1999 and 2006 into the separate effects of personal characteristics and task assignments. They found a noticeable increase of wage gap between 1999 and 2006, and the decomposition results show that the changes in personal characteristics explain some of the increase in wage gap whereas the changes in task assignments tend to reduce wage gap. The coefficient effect for personal characteristics tends toward an increase in wage gap at the top end of the wage distribution, the coefficient effect for the task assignments on the contrary shows an inverted U-shaped pattern. Antonczyk et al. (2010) compare the trends in wage gap in the USA and Germany between 1979 and 2004. They find there is evidence in both the USA and Germany which is consistent with a technology-driven polarization of the labor market but the patterns of trends in wage gap differ so strongly between the USA and Germany that technology effects alone cannot explain these findings. Firpo et al. (2011) use Current Population Survey (CPS) data to investigate the influence of the change in the return to occupational tasks on the change in wage distribution over the last three decades in the USA. They find that the polarization of wages in the 1990s can be explained by the changes in wage setting between and within occupational tasks, which are captured by task measures linked to technological change and offshore activities. Two papers about the age structure of task types are most relevant to this study. Autor and Dorn (2009) investigate the aging of performers of middle-skill jobs and routine task intensive jobs from 1980 to 2015

3

WORK SKILLS GAP AND THE WAGE DIFFERENTIALS BETWEEN …

45

in the USA. They argue that workers develop occupation-specific human capital as they gain work experience and skill specificity makes the costs of occupational mobility higher for older than for younger workers. When an occupation declines, older workers will face an incentive not to exit the occupation while younger workers will face an incentive not to enter. Firms may react to changing demand for occupations by hiring young workers for growing occupations. These suppositions imply that occupations will “get old” as their recruitment declines—that is, the mean age of an occupation’s workforce will rise in a declining occupation. An empirical study based on the OLS regression model confirms this argument. Using the data of 12 European countries between 1998 and 2014, and the OLS model, Lewandowski et al. (2017) develop the work of Autor et al. (2003). In order to analyze the age dimension of changes in the task composition of jobs, Acemoglu and Autor (2011) constructs five kinds of task type: non-routine cognitive analytical; nonroutine cognitive interpersonal; routine cognitive; routine manual; and non-routine manual physical. They find that the shift away from routine work and toward non-routine work occurs much faster among workers born between 1970 and 1989 (the younger generation) than among workers born between 1950 and 1969 (the older generation). They find in the majority of countries the aging of the workforce occurred more quickly in occupations that were more routine-intensive because the proportion of young workers in these occupations was declining. These two papers argue the age structure differs for different task types and the employment share differs for task types and age groups but it is not clear how the work skill gaps influence the wage gaps between the younger, the middle-aged, and older workers. This study explores this neglected area.

3.3

Methodology and Data 3.3.1

Models

First, wage function is used to estimate the wage gap for different age groups and the impact of task on wage, which is expressed as Eqs. 3.1 and 3.2.2 lnW i = βa Agei + β X X i + u i

(3.1)

lnW i = βa Agei + βt Task i + β X X i + u i

(3.2)

46

X. MA AND X. QU

In Eqs. 3.1 and 3.2, i represents the individual, lnW i is the logarithm of the average wage, Age expresses the index of age including age, age squared, age dummy variables, X represents factors (e.g., years of schooling, years of experience, gender, industry sector, migrant, urban Hukou, region) which may affect wage levels, Task expresses the task type dummy variables including Task I (non-routine cognitive analytical), Task II (non-routine cognitive personal), Task III (routine cognitive), and Task IV (routine manual), u is a random error item. The coefficient βa expresses the wage gap by different age groups. We can compare the values of βa from Eqs. 3.1 and 3.2. According to human capital theory, because the task skill can be considered as a task-specific human capital which may affect the individual’ wage level, the values of βa may differ for Eqs. 3.1 and 3.2. We compare the values of βt to investigate the differences of task skill effect by various task types. To distinguish the influence of task on wage for different age groups, we use Eq. 3.3 to estimate the coefficients βt using various age group samples separately. lnW i j = βt j Task i j + βx j X i j + u i j j:

(3.3)

  age groups age 16+, age 16-49, age 50-59, age 60+

It is thought there may be a sample selection bias problem (a worker can choose to work or not) in the OLS model. The Heckman two-step model (Heckman 1979) is used to address the selection bias problem. Based on the estimated results of the distribution function and the density function from the probit regression model, the selection bias correction item (λ = φ(.)/(.)) are calculated. The probit regression model includes the identification variables. The corrected wage function expressed by Eq. 3.4 can be estimated using these correction items. lnW i j = βt j Task i j + βx j X i j + βλj λi j + u i j

(3.4)

Second, we use the Blinder-Oaxaca decomposition model (Blinder 1973; Oaxaca 1973) to investigate the task effect on wage gap. The model is expressed as Eqs. 3.5 and 3.6.3 lnW a1 − lnW a2 = βa1 (X a1 − X a2 ) + (βa1 − βa2 )X a2

(3.5)

lnW a1 − lnW a2 = βa2 (X a2 − X a1 ) + (βa2 − βa1 )X a1

(3.6)

3

WORK SKILLS GAP AND THE WAGE DIFFERENTIALS BETWEEN …

47

In Eqs. 3.5 and 3.6, X a1 and X a2 are variable means of two groups (e.g., middle-aged worker and older worker). βm and β f are estimated coefficients in wage functions. Based on the human capital theory (Becker 1964; Mincer 1974) and discrimination hypothesis (Becker 1957), the decomposition model decomposes the wage gap between two age groups into two parts: the human capital endowment (known as “explained part”) [βa1 (X a1 − X a2 ) or βa2 (X a2 − X a1 )] and the endowment return (known as the “unexplained part”) ([(βa1 − βa2 )X a2 or (β a2 − βa1 )X a1 ]. The explained part expresses the differentials of individual characteristics such as the differences in human capital endowments including taskspecific human capital. The unexplained part includes the differences in return to task, wage determination systems, discrimination, or capabilities not at present measurable. The larger the estimated explained part is, the greater is the influence of human capital differences between two age groups on wage gap, and vice versa. 3.3.2

Data

Data from the China Urban Labor Survey (CULS) is used in this study. The CULS has been conducted by the Institute of Population and Labor Economics of the Chinese Academy of Social Science (CASS) since 2010. We use the third wave of CULS which includes the “skills used at work” questionnaire. The third wave of CULS is a unique survey of the skills and skill use at work in China, which is based on the OECD’s Programme for the International Assessment of Adult Competencies (PIAAC) and the World Bank’s Skills Measurement Program (STEP), and the basis of Occupational Information Network (O*NET) data for the USA. It is the first survey on skills at work for China; therefore, the third wave of CULS is the most suitable data for this study. The third wave of CULS was conducted in 2016 in six large cities in China (Guangzhou, Shanghai, and Fuzhou on the coast, Shenyang in the northeast, Xian in the northwest, and Wuhan in central China). The 2016 CLUS contains far more observations (almost 15,500) and covers a more comprehensive area. The analytic objects are workers, and the unemployed are excluded from this calculation. The analytic objects are limited tolocal urban residents aged 16 and above. Four kinds of sample are used to consider the influence of the legal retirement age4 in the public sector: the group aged 16 and above, the group aged 16–49, the group aged 50–59, and the group aged 60 and above. Abnormal value samples,5 no answer samples,

48

X. MA AND X. QU

and missing value samples are excluded. The samples used in this study are 7,197 for total samples aged 16 and above, 6,114 for samples aged 16–49, and 1,000 for samples aged 50–59, and 83 for samples aged 60 and above. 3.3.3

Variables

The dependent variable is the logarithm value of the hourly wage in the wage function and decomposition model. The hourly wage is calculated from wage and work hours. First, for the independent variables (see Appendix Table 3.7), the main independent variable is the task. Based on the World Bank’s Skills Measurement Program (STEP),6 the 2016 CLUS designed the almost similar questionnaires about the task items. We classified the four types of tasks according to the contents of task items (Table 3.1). They are Task I (Non-routine cognitive analytical task) including nine kinds of items: (1) frequency of learning new knowledge [0–5]; (2) longest materials typically read [0–5]; (3) complexity of mathematics use at work [0–4]; (4) read bill [0–1]; (5) read newspaper [0–1]; (6) read reports [0–1]; (7) under advanced math [0–1]; (8) solve complex problems [1–5]; and (9) use programming language [1–6]. Task II (Non-routine cognitive interpersonal) including five kinds of items: (1) use phone [0–1]; (2) interaction with customers [0–10]; (3) supervise coworkers [0–1]; (4) make presentations [0–1]; and (5) collaboration with coworkers [1–5]. Task III (Routine cognitive) including five kinds of items (1) complexity of mathematics use at work [0–4]; (2) time spent on repetitive work [0– 4]; (3) freedom in changing assignment order [1–10]; (4) fill form [0–1]; and (5) read manual [0–1]. Task IV (Manual) including four kinds of items: (1) drive vehicle [0–1]; (2) repair electronics [0–1]; (3) operate on heavy machines [0–1]; and (4) physical demand [1–10]. We calculated the total values of skill score for each task. The higher the skill score the higher the skill level in each task. Therefore the higher skill core of task represents the higher task-specific human capital. We use the skill score for four types of task in this study. Second, three kinds of age variables are used for wage function: (1) age and age squared; (2) six kinds of age category dummy variables: aged 16–24, aged 25–29, aged 30–39, aged 40–49, aged 50–59, aged 60 and above; (3) an aged 50 and above dummy variable which is equal to 1 when the age is 50 and above.

3

WORK SKILLS GAP AND THE WAGE DIFFERENTIALS BETWEEN …

Table 3.1 Task categories and measurement items in the CULS

49

Task category

Measurement items in the CULS

Task I (non-routine cognitive analysis)

1. Frequency of learning new knowledge [0–5] 2. Longest materials typically read [0–5] 3. Complexity of mathematics use at work [0–4] 4. Read bill [0–1] 5. Read newspaper [0–1] 6. Read reports [0–1] 7. Under advanced math [0–1] 8. Solve complex problems [1–5] 9. Use programming language [1–6] 1. Use phone [0–1] 2. Interaction with customers [0–10] 3. Supervise coworkers [0–1] 4. Make presentations [0–1] 5. Collaboration with coworkers [1–5] 1. Complexity of mathematics use at work [0–4] 2. Time spent on repetitive work [0–4] 3. Freedom in changing assignment order [1–10] 4. Fill form [0–1] 5. Read manual [0–1] 1. Drive vehicle [0–1]

Task II (non-routine cognitive interpersonal)

Task III (routine cognitive)

Task IV (manual)

(continued)

50

X. MA AND X. QU

Table 3.1 (continued)

Task category

Measurement items in the CULS 2. Repair electronics [0–1] 3. Operate on heavy machines [0–1] 4. Physical demand [1–10]

Source Author’s creation based on the 2016 CLUS

Third, the other factors which may influence the wage levels are conducted as follows: 1. The female dummy variable is constructed to estimate the gender wage gap. 2. The years of schooling, years of experience,7 and squared variables are used as the index of general human capital. 3. The labor market in China is segmented into rural regions and urban regions under the Hukou system. There is a wage gap between migrants and local urban residents (Wang 2005; Ma 2018a). The migrant dummy variable and urban dummy variable are used. 4. Four kinds of firm size dummy variables are used: firms with less than 20 employees, firms with 20–99 employees, firms with 100– 499 employees, and firms with more than 500 employees. 5. It is thought the nature of the labor contract may influence the employment and wage, three kinds of employment contact dummy variables are conducted. They are permanent contact, temporary contact, non-labor contact, and missing (no answer) dummy variables. 6. In China there is a wage gap between the public sector and private sector (Zhang and Xue 2008; Ye et al. 2011; Demurger et al. 2012; Ma 2018b, c) and five kinds of ownership type dummy variables are conducted to control the influence of ownership types on wage. They are state-owned enterprise (SOE), privately owned enterprise (POE), foreign-owned enterprise (FOE), self-employed (Self ) and other ownership types (Others ). 7. There are twenty kinds of industry code in the survey. Eight kinds of industry sector dummy variables are conducted on the samples for

3

WORK SKILLS GAP AND THE WAGE DIFFERENTIALS BETWEEN …

51

econometric analysis. They are manufacturing and construction, sale and retail trade, traffic and post, hotel and restaurants, finance and housing, service, social service and social organization, and other industries. 8. Five city dummy variables (Shenyang, Shanghai, Fuzhou, Wuhan, Guangzhou, and Xian) are used to control the regional disparities. Fourth, it is thought that the categories of age, age-squared, having a child, marriage, co-residence with parents may affect the labor force participation, particularly for women. These variables are used as the identification variables in the first step of Heckman two-step model.

3.4

Descriptive Statistics Results

3.4.1

Task Skill Gap for Different Age Groups

Table 3.2 reports the task skill scores for task types and for different age groups. First, a task skill gap is found between different age groups in each task. The task skill level is higher for the group aged 16–49 than for both the groups aged 50–59 and the group age 60 and above in each task. This suggests the task skill level declines as age increases. Second, the results indicate the task skill differences are not the same for the various age groups. For example, in the task skill gap between the group aged 16–49 (the younger group) and the group aged 50–59 (the middle-age group), the skill level is higher for the younger group in Task I (0.086), Task II (0.080), and Task III (0.088), but is lower for the younger group in Task IV (−0.045), while for the gap between the group aged 16–49 Table 3.2 Task skill gap by different age groups in China

Task Task Task Task

I II III IV

Age 16–49

Age 50–59

Age 60+

Gap

(a)

(b)

(c)

(a − b)

(b − c)

(a − c)

0.293 0.463 1.840 0.418

0.207 0.383 1.752 0.463

0.170 0.330 1.833 0.394

0.086 0.080 0.088 −0.045

0.037 0.053 −0.081 0.069

0.123 0.133 0.007 0.024

Note Task I: non-routine cognitive analytical; Task II: non-routine cognitive personal; Task III: routine cognitive; Task IV: routine manual Source Author’s creation based on the 2016 CLUS

52

X. MA AND X. QU

and the group aged 60 and above (the older group), the task skill is higher for the younger group in each task. It can be observed that in the task gap between the middle-age group and the older group, the task skill is higher for the middle-aged group in Task I (0.037), Task II (0.053), and Task IV (0.069) but is lower in Task III (−0.081). It is assumed that the task skill gap may affect the wage gap between the younger, the middle-aged, and the older groups. 3.4.2

Wage Gap for Different Age Groups

Does the wage level change for different age groups? Fig. 3.1 summarizes the average wage level for six kinds of age group. It is found that from age 16 to 39 the wage level goes up with increasing age, but after 39 the wage Task skill

Wage

0.50

40

0.40

30

0.30 20 0.20

10

0.10

0

0.00 age16-24

age25-29 Task I

age30-39 Task IV

age40-49

age50-59

age60+

average wage

Fig. 3.1 Wage gap by task skill types and age group in China (Source Author’s creation based on the 2016 CLUS)

3

WORK SKILLS GAP AND THE WAGE DIFFERENTIALS BETWEEN …

53

level goes down. The wage level is highest for the group aged 30–39. To consider the relationship between task skill and wage, it can be seen that the change pattern of Task I (non-routine cognitive analytical tasks) is similar to that of wage, but the change pattern of Task IV (routine manual) is reversed. It suggests that as age increases, the non-routine cognitive analytical task skill related with the higher wage has a tendency to decline, while the routine manual task skill which is related to the lower wage has a tendency to increase, and this may cause the wage level to decrease with age. Although Table 3.2 and Fig. 3.1 suggest differences of task skills between different age groups it seems likely that the wage level is related to the task skill in each age group and the correlation between task skill and wage level differs for task types and for different age group. However, it should be noted that other factors such as education and industry sector may also affect the wage levels, but these factors are not controlled in these descriptive results. Hence econometric analyses are conducted to investigate the influence of task skill on wage gap between age groups in the following section.

3.5 3.5.1

Econometric Analysis Results

Wage Gap for Different Age Groups and the Impact of Task on Wage

The wage function is used to investigate the wage gap for each age group and to investigate the impact of task skill on wage. To address the sample selection bias problem, we perform the wage function using the Heckman two-step model. Most of the selection items are found to be statistically significant using the OLS model, which indicates that the sample selection bias problem is severe in these wage functions (see Appendix Table 3.8). We distinguish the estimated models from model 1 to model 8 using different variables. These results are summarized in Table 3.3. First, when the influence of task skill is not considered (Model 1 and Model 2) a reversed U-shaped relation between wage and age is found, the peak of the wage-age curve is at age 30–39 and is consistent with the Fig. 3.1. When the other factors (e.g., education, industry) are constant, the wage is lowest for the group aged 60 and over. For example, the wage level is 59.6% lower for the group aged 60 and over than for the group aged 16–24.

Task III*Age 50+

Task II*Age 50+

Task I*Age 50+

Age 50+

Task IV

Task III

Task II

Task I

Age 60+

Age 50–59

Age 40–49

Age 30–39

0.097*** (6.39) −0.001*** (−6.87)

(1)

0.131 (1.51) 0.214** (2.49) 0.156* (1.81) −0.093 (1.03) −0.596*** (2.95)

(2)

0.142* (1.71) 0.207** (2.49) 0.163** (1.96) −0.032 (−0.37) −0.563*** (−2.68) 0.685*** (17.29)

(3)

0.538*** (19.39)

0.110 (1.32) 0.180** (2.17) 0.134* (1.61) −0.085 (−0.99) −0.676*** (−3.25)

(4)

0.150*** (15.84)

0.092 (1.10) 0.161* (1.92) 0.107 (1.27) −0.118 (−1.35) −0.711*** (−3.37)

(5)

−0.292*** (−7.86)

0.140* (1.63) 0.226*** (2.64) 0.169** (1.98) −0.086 (−0.96) −0.747*** (−3.50)

(6)

Wage gap between age groups and the influence of task skill on wages in China

Age (age 16–24) Age 25–29

Age_sq

Age

Table 3.3

0.103 (1.27) 0.161** (1.99) 0.124 (1.53) −0.060 (−0.71) −0.597*** (−2.91) 0.342*** (7.87) 0.356*** (11.62) 0.092*** (9.47) −0.264*** (−7.41)

(7)

0.305*** (6.50) 0.372*** (11.23) 0.096*** (9.01) −0.262*** (−6.72) −0.029 (−0.41) 0.285** (1.97) −0.109 (−1.38) −0.018 (−0.84)

(8)

54 X. MA AND X. QU

0.418*** (4.84) Yes 11,100 5,206 5,894 2575.45 0.000

0.252** (2.24) Yes 11,100 5,206 5,894 2710.08 0.000

(2)

0.188* (0.11) Yes 11,100 5,206 5,894 3120.36 0.000

(3)

0.233** (2.14) Yes 11,100 5,206 5,894 3197.37 0.000

(4)

0.216** (1.97) Yes 11,100 5,206 5,894 3029.08 0.000

(5)

0.257** (2.31) Yes 11,100 5,206 5,894 3442.74 0.000

(6)

0.170* (1.62) Yes 11,100 5,206 5,894 3623.96 0.000

(7) 0.001 (0.01) −0.155*** (−3.59) Yes 11,100 5,206 5,894 3619.73 0.000

(8)

Note 1. *** p < 0.01, **p < 0.05, **p < 0.10 2. Task I: non-routine cognitive analytical; Task II: non-routine cognitive personal; Task III: routine cognitive; Task IV: routine manual 3. Heckman two-step model is utilized. Control variables include female, age, years of schooling, firm size, firm ownership, labor contract, industry, migrant, urban and region dummy variables in the second step; age, years of schooling, having child, married, co-resident with parents, urban and city dummy variables in the first step 4. t-values are in parentheses Source Author’s creation based on the data from 2016 CULS

Control variables Number of obs Censored obs Uncensored obs Wald χ2 (21) Prob > χ2

Selection bias item

Task IV*Age 50+

(1)

3 WORK SKILLS GAP AND THE WAGE DIFFERENTIALS BETWEEN …

55

56

X. MA AND X. QU

Second, it is thought that the wage gap between different age groups may be caused by the task skill. We add the task skill variables in Models 3, 4, 5, 6 and 7. It is indicated that when the influence of task skill is controlled, there is a different wage gap for each age group. For example, in comparison with the group aged 60 and over the group aged 16–24 has a wage level of 56.3% higher (Model 3), 67.6% (Model 4), 71.1% (Model 5), 74.7% (Model 6), and 59.7% (Model 7). Third, the task skills influence the wage level. Task I, Task II, and Task III positively affect the wage, while the Task IV negatively affects the wage. The influence of task skill is greater for non-routine task (Task I and Task II) than for routine tasks (Task III and Task IV). 3.5.2

Differences of the Influence of Task on Wage for Different Age Groups

To compare the task effect on wage for different age groups we employ the wage functions for the total sample (the group aged 16 and above) and separately for the younger group (the group aged 16–49), the middle-aged group (the group aged 50–59), and the older group (the group aged 60 and above). The results are reported in Table 3.4. First, in summary for the total sample, for the younger group and the middle-aged group each task positively affects the wage level, but for the older group, the influence of task skills is not statistically significant for each task. It suggests that the influence of task-specific human capital on wage is small for the older workers most of whom are working beyond retirement. It may be because the wage system differs for different age groups (e.g., workers immediately before retirement and reemployment workers after retirement). Second, to compare the influence of task skills for different task types, it is shown that the influences of Task I, Task II, and Task III are positive, and the influence of Task IV is negative for each age group. The influence is greater for non-routine tasks (Task I and Task II) than for routine tasks (Task III and Task IV). For example, the coefficient value is greater for Task I (0.332 for age 16 and older, 0.284 for age 16–49, 0.507 for age 50–59) than Task III (0.093 for age 16 and over, 0.089 for age 16–49, and 0.092 for age 50–59). Third, to compare the influence of task skill by age groups, the influences of task skill on wage are almost similar for Task III and Task IV,

8.59 11.82 10.52 −8.25 −14.46 14.41 11.18 −11.92 5.15 6.7 8.29 −5.59 −3.73 2.33 −1.24 −0.61 1.67 −4.08 −2.7 −3.82 −0.69

0.332*** 0.327*** 0.093*** −0.264*** −0.212*** 0.046*** 0.033*** −0.001*** 0.109*** 0.163*** 0.227*** −0.117*** −0.097*** 0.082** −0.039 −0.017 0.072* −0.167*** −0.109*** −0.094*** −0.021

−0.086*** −0.035

0.009 0.032 0.107** −0.118*** −0.064

−0.092*** −0.078*** 0.104***

0.100*** 0.153*** 0.220***

0.284*** 0.337*** 0.089*** −0.264*** −0.218*** 0.050*** 0.042*** −0.001***

Coef

Coef

t-value

Age 16–49

Age 16+

−3.34 −1.05

0.26 1.03 2.38 −2.72 −1.48

−4.09 −2.75 2.77

4.55 6.01 7.64

7.04 11.64 9.32 −7.77 −14.32 13.99 10.08 −9.18

t-value

The influence of task on wages by different age groups in China

Task I Task II Task III Task IV Female Years of schooling Exp Exp. Sq Firm size (0–19) 20–99 100–499 More than 500 Labor contact (permanent) Temporary No labor contract Miss value Ownership (Gov.) SOE POE FOE Self Other Industry (Ind1) Ind2 Ind3

Table 3.4

−0.147* −0.020

−0.239*** −0.269*** −0.020 −0.392*** −0.301**

−0.157*** −0.076 0.067

0.134* 0.190** 0.255***

0.507*** 0.286*** 0.092*** −0.264*** −0.181*** 0.026** −0.140** 0.001*

Coef

Age 50–59

−1.81 −0.24

−2.91 −3.39 −0.13 −3.12 −2.48

−2.60 −1.14 0.64

1.91 2.40 2.98

4.04 3.28 3.69 −2.74 −3.51 2.24 −2.02 1.68

t-value

0.67 −0.14

0.34 0.40 1.55 0.06 0.14

−2.22 −2.21 −1.14

1.04 1.41 0.82

0.80 −0.41 0.51 1.37 −0.05 −0.70 −0.62 0.50

t-value

WORK SKILLS GAP AND THE WAGE DIFFERENTIALS BETWEEN …

(continued)

0.342 −0.101

0.200 0.209 1.562 0.037 0.086

−1.023 −0.880 −0.616

0.438 0.786 0.435

0.794 −0.259 0.074 0.760 −0.013 −0.067 −0.291 0.002

Coef

Age 60+ 3

57

(continued)

Coef

t-value −4.33 3.17 −1.31 −3.81 −1.46 5.67 4.59 20.42 11.02 0.91 12.18 1.03 20.93

Coef −0.133*** 0.098*** −0.029 −0.132*** −0.060 0.099*** 0.088*** 0.509*** 0.274*** 0.022 0.291*** 0.026 1.718*** 7,197 0.403

0.523*** 0.287*** 0.030 0.306*** 0.034 1.518*** 6,114 0.407

−0.136*** 0.122*** −0.003 −0.099*** −0.042 0.099*** 0.080***

Age 16–49

Age 16+

19.86 11.01 1.13 12.18 1.28 17.02

−4.21 3.85 −0.11 −2.64 −0.94 5.57 4.06

t-value

0.488*** 0.193** 0.013 0.241*** −0.039 5.994*** 1,000 0.378

−0.116 −0.143 −0.193*** −0.298*** −0.150 0.090 0.128*

Coef

Age 50–59

6.58 2.52 0.19 3.27 −0.52 3.87

−1.23 −1.12 −2.97 −3.20 −1.43 1.39 1.76

t-value

0.281 −0.127 −0.019 0.005 0.401 11.596 83 0.033

0.402 −0.029 0.138 0.086 −0.171 0.223 0.368

Coef

Age 60+

0.65 −0.29 −0.04 0.01 0.83 0.88

0.58 −0.06 0.34 0.14 −0.28 0.64 1.00

t-value

Note 1. *** p < 0.01, **p < 0.05, **p < 0.1 2. Task I: non-routine cognitive analytical; Task II: non-routine cognitive personal; Task III: routine cognitive; Task IV: routine manual 3. Ind1: manufacturing and construction; Ind2: sale and retail trade; Ind3: traffic and post; Ind4: hotel and restaurant; Ind5: financial and housing; Ind6: service; Ind7: social service and social organization; Ind8: Other industry sectors Source Author’s creation based on the data from 2016 CULS

Ind4 Ind5 Ind6 Ind7 Ind8 Migrant Urban Region (Shenyang) Shanghai Fuzhou Wuhan Guangzhou Xian _cons Number of obs Adj R 2

Table 3.4

58 X. MA AND X. QU

3

WORK SKILLS GAP AND THE WAGE DIFFERENTIALS BETWEEN …

59

but the influence of Task I and Task II differs for age groups. The influence of Task I is greater for the middle-aged group (0.284) than for the younger group (0.507), while the influence of Task II is greater for the younger group (0.337) than for the middle-aged group (0.286). 3.5.3

How Does the Work Skill Gap Affect the Wage Gap Between Different Age Groups?

According to the estimated results above, it is indicated that task skills differ between age groups, and the influence of task skills on wage differs for each age group. The wage gap between the younger, the middle-aged, and the older worker groups may be due to these two types of factor. To distinguish the influences of these two types of factor on the wage gap, we employ a decomposition analyses which follows the Blinder-Oaxaca decomposition model. The decomposition results on the wage gap between the younger group (the group aged 16–49) and the middle-aged group (the group aged 50–59) are summarized in Table 3.5. Three main findings follow. First, in general the value of the contribution rate is 53.7% for the explained part, and 46.7% for the unexplained part. The influence of the explained part is greater than the unexplained part. Second, the total contributions of task skill to wage gap are 38.4% in the explained part, and −18.0% in the unexplained part. It suggests that the differences of task skill levels between the younger workers and the middle-aged workers may enlarge the wage gap, while the differences in return to task skills may reduce the wage gap. It is clear that the effects of the explained part and the unexplained part for task are different, and the influence of the differences of task skill levels is greater than the difference of return to task skill. Third, to compare the results for the different task skill types: for the explained part, even though all four types of tasks may enlarge the wage gap, the effect is greater for non-routine tasks (12.9% for Task I, 14.6% for Task II), than for routine tasks (4.6% for Task III, 6.3% for Task IV). For the unexplained part, Task I (−29.3%) and Task IV (−3.5%) contribute to reduce the wage gap, while Task II (9.9%) and Task III (4.9%) seem likely to enlarge the wage gap. This suggests a polarization for the contribution of return to task skills: the return to task skill is higher for both the high wage group (Task I) and the low wage group (Task IV) than for the middle-level wage group (Task II and Task III), which may contribute to

60

X. MA AND X. QU

Table 3.5 Decomposition results of wage gap between workers aged 16–49 and aged 50–59 in China Values

Total Task I Task II Task III Task IV Task total Female Education Firm size Labor contract Ownership Industry Migrant Urban City Constants

Percentage

Explained

Unexplained

Explained

Unexplained

0.100 0.024 0.027 0.009 0.012 0.072 −0.051 0.074 −0.004 −0.019 0.007 0.004 0.018 −0.012 0.010 0.000

0.088 −0.055 0.019 0.009 −0.007 −0.034 −0.018 −0.038 −0.021 0.011 0.194 0.082 −0.004 −0.017 0.047 −0.114

53.3 12.9 14.6 4.6 6.3 38.4 −26.9 39.5 −2.1 −10.2 3.9 2.1 9.5 −6.2 5.3 0.0

46.7 −29.3 9.9 4.9 −3.5 −18.0 −9.4 −20.4 −10.9 5.9 102.9 43.6 −2.4 −8.8 24.7 −60.6

Source Author’s creation based on the data from 2016 CULS

job polarization of employment and wage distributions. The results for China are consistent with that for the USA (Autor et al. 2003; Autor and Dorn 2013; Autor 2013), the UK (Goos and Manning 2007), and Germany (Spitz-Oener 2006; Dustmann et al. 2009). The decomposition results on the wage gap between the group aged 50–59 and the group aged 60 and over are summarized in Table 3.6. The new findings are as follows: First, in general the value of the contribution rate is 33.7% for the explained part and 66.3% for the unexplained part. The influence of the unexplained part is greater than for the explained part. Second, for the total contributions of task skill to wage gap, the total values of task are 3.7% in the explained part, and −35.7% in the unexplained part. This suggests that the differences of task skill levels between the older workers and the middle-aged workers may enlarge the wage gap, while the differences for return of task skills may reduce the wage gap. It is clear that the effects of the explained part and the unexplained part for

3

WORK SKILLS GAP AND THE WAGE DIFFERENTIALS BETWEEN …

61

Table 3.6 Decomposition results of wage gap between workers aged 50–59 and aged 60+ in China Values

Total Task I Task II Task III Task IV Task total Female Education Firm size Labor contract Ownership Industry Migrant Urban City Constants

Percentage

Explained

Unexplained

Explained

Unexplained

0.139 0.023 0.017 −0.006 −0.019 0.015 −0.002 0.066 0.024 −0.003 0.013 0.009 −0.007 0.014 0.010 0.000

0.274 −0.028 0.200 0.050 −0.369 −0.147 −0.034 0.562 −0.172 0.657 −0.343 −0.222 −0.056 −0.159 0.083 0.105

33.7 5.6 4.2 −1.6 −4.5 3.7 −0.5 16.0 5.9 −0.8 3.2 2.1 −1.7 3.3 2.5 0.0

66.3 −6.8 48.3 12.1 −89.4 −35 .7 −8.2 136.0 −41.5 159.1 −83.0 −53.9 −13.7 −38.4 20.2 25.4

Source Author’s creation based on the data from 2016 CULS

task skill are different, and the influence of the differences of return to task skill is greater than that of the task skill levels gap. Third, to compare the results for task types, the effects differ with task type in both the explained part and the unexplained part. For the explained part, both the task skill levels of Task I (5.6%) and Task II (4.2%) are higher for the middle-aged workers which may enlarge the wage gap, but the skill levels of Task III (−1.6%) and Task IV (−4.5%) which contribute to reduce the wage gap are lower for the middle-aged workers than for the older workers. For the unexplained part, the returns to skill of both Task II (48.3%) and Task III (12.1%) which contribute to enlarge the wage gap are higher for the middle-aged workers than for the older workers, but the returns to skill of both Task I (−6.8%) and Task IV (−89.4%) which contribute to reduce the wage gap are lower for the middle-aged workers than the older workers. This suggests that for the workers re-employed after mandatory retirement (the group aged 60 and over), the higher skills of non-routine task may increase the wage levels but the return to task skills decrease greatly for the middle-level

62

X. MA AND X. QU

wage group (the workers occupied Task II and Task III) which causes the wage level to decrease greatly after retirement. It is indicated that the wage systems for workers immediately before retirement and after mandatory retirement may differ within a firm. Detailed survey research on the wage and employment systems for older workers in Chinese firms could usefully be conducted in the future.8

3.6

Conclusions

Using unique survey data from the China Urban Labor Survey (CULS) conducted in 2016, and decomposition models, Chapter 2 investigates the differences of task skill levels and the differences of return to task by task types between the younger group (the group aged 16–49), the middle-aged group (the group aged 50–59), and the older group (the group aged 60 and above). The Blinder-Oaxaca decomposition model is used to estimate the contributions of these two kinds of factors on wage gap between different age groups. Four categories of types of tasks derived from the World Bank’s Skills Measurement Program (STEP) and the questionnaire for CULS are used in this study. They are Task I (nonroutine cognitive analytical tasks), Task II (non-routine cognitive personal tasks), Task III (routine cognitive tasks), and Task IV (manual tasks). The four main findings are as follows. First, differences of skill levels are found for different age groups. The skill level in each task is higher for the younger groups than for the older group. It suggests that task skill declines as age increases. Second, in general the influence of task skills on wage is statistically significant. The influence of task skill is greater for non-routine task (Task I and Task II) than routine task (Task III and Task IV). The influence of task skill on wage differs for each age group. For the younger workers and the middle-aged workers, each task positively affects the wage levels, but for the older workers, the influence of task skills is not statistically significant for each task. This suggests that the influence of task-specific human capital on wage is small for older workers most of whom are working beyond mandatory retirement. Third, the decomposition results of the wage gap between the younger workers and the middle-aged workers indicate that, in general the influence of the explained part (53.7%) is greater than the unexplained part (46.7%). Regarding the contributions of task to wage gap, the differences of task skill levels (38.4%) contribute to enlarge the wage gap, while the

3

63

WORK SKILLS GAP AND THE WAGE DIFFERENTIALS BETWEEN …

differences of return of task skills (−18.0%) contribute to reduce the wage gap, and the influence of the differences of task skill levels is greater than the differences of return to task skills. Fourth, the decomposition results of the wage gap between the middle-age workers and older workers indicate that, in general the influence of the explained part (66.3%) is greater than the unexplained part (33.7%). Regarding the contributions of task to wage gap, the differences of task skill levels (3.7%) contribute to enlarge the wage gap, while the differences of return of task skills (−35.7%) contribute to reduce the wage gap, and the influence of the differences of return to task skills is greater than the differences of task skill levels. The results suggest that the task skill gaps in Task I, Task II, and Task III contribute to expand the wage gap between the younger workers and the middle-age workers, and to expand the wage gap between the middle-aged workers and the older workers. It is thought that task-specific skills may depreciate as age increases. Therefore in order to reduce wage inequality between different age groups and to increase the productivity of older workers, more vocational training for middle-age workers and older workers should be provided by the government.

Appendix See Tables 3.7 and 3.8. Table 3.7 Descriptive statistics of variables (a) Total

Lnwage Task I Task II Task III Task IV Female Years of schooling

(b) Age 16–49

(c) Age 50–59

(d) Age 60+

Mean

S.D

Mean

S.D

0.793 0.210 0.294 0.956 0.226 0.421 3.282

2.555 0.166 0.325 1.815 0.390 0.217 9.855

0.906 0.201 0.307 1.070 0.226 0.415 4.037

Mean

S.D

Mean

3.124 0.279 0.449 1.822 0.424 0.422 12.541

0.738 0.230 0.308 0.930 0.231 0.494 3.420

3.157 0.292 0.461 1.835 0.418 0.456 12.799

S.D 0.720 0.230 0.308 0.923 0.231 0.498 3.365

2.968 0.207 0.383 1.744 0.463 0.230 11.186

(continued)

64

X. MA AND X. QU

Table 3.7 (continued) (a) Total Mean Years of experience Firm size 0–19 20–99 100–499 More than 500 Labor contract Permanent Temporary No labor contract Miss value Ownership Government SOE POE FOE Self Other Industry sector Ind1 Ind2 Ind3 Ind4 Ind5 Ind6 Ind7 Ind8 Migrant Urban Regions Shenyang Shanghai Fuzhou

25.722

S.D

(b) Age 16–49

(c) Age 50–59

(d) Age 60+

Mean

Mean

Mean

11.415 22.610

0.404 0.275 0.176 0.145

S.D

S.D

S.D

9.221 42.583

4.169 51.602

4.434

0.491 0.447 0.381 0.352

0.490 0.434 0.383 0.378

0.518 0.217 0.145 0.120

0.503 0.415 0.354 0.328

0.405 0.271 0.176 0.148

0.491 0.445 0.381 0.356

0.398 0.251 0.178 0.173

0.156 0.405 0.166

0.363 0.143 0.491 0.433 0.372 0.157

0.350 0.235 0.496 0.251 0.364 0.213

0.424 0.181 0.434 0.205 0.410 0.289

0.387 0.406 0.456

0.273

0.445 0.267

0.443 0.301

0.459 0.325

0.471

0.129 0.152 0.399 0.042 0.207 0.071

0.335 0.359 0.490 0.201 0.406 0.258

0.117 0.142 0.422 0.046 0.207 0.067

0.321 0.349 0.494 0.209 0.405 0.250

0.198 0.215 0.262 0.021 0.207 0.097

0.399 0.411 0.440 0.143 0.405 0.296

0.157 0.133 0.361 0.024 0.217 0.108

0.366 0.341 0.483 0.154 0.415 0.313

0.174 0.201 0.068 0.079 0.071 0.284 0.090 0.033 0.484 0.661

0.379 0.401 0.252 0.270 0.257 0.451 0.287 0.178 0.500 0.473

0.169 0.209 0.065 0.080 0.078 0.282 0.086 0.030 0.515 0.644

0.375 0.407 0.247 0.272 0.268 0.450 0.280 0.172 0.500 0.479

0.200 0.154 0.090 0.073 0.029 0.291 0.117 0.046 0.304 0.766

0.400 0.361 0.286 0.260 0.168 0.454 0.322 0.210 0.460 0.424

0.181 0.193 0.048 0.060 0.072 0.277 0.120 0.048 0.373 0.651

0.387 0.397 0.215 0.239 0.261 0.450 0.328 0.215 0.487 0.480

0.140 0.180 0.167

0.347 0.140 0.384 0.178 0.373 0.171

0.348 0.145 0.392 0.181 0.349 0.157

0.354 0.387 0.366

0.347 0.141 0.383 0.190 0.377 0.142

(continued)

3

65

WORK SKILLS GAP AND THE WAGE DIFFERENTIALS BETWEEN …

Table 3.7 (continued) (a) Total

Wuhan Guangzhou Xian Observations

(b) Age 16–49

(c) Age 50–59

(d) Age 60+

Mean

Mean

Mean

Mean

S.D

0.166 0.198 0.149 7,197

0.372 0.160 0.398 0.203 0.356 0.148 6,114

S.D

0.367 0.194 0.402 0.174 0.355 0.159 1,000

S.D

0.396 0.253 0.379 0.133 0.366 0.133 83

S.D 0.437 0.341 0.341

Note 1. The variables used in wage functions by the OLS model are summarized in the table. It should be noticed that the observations may differ by using various models 2. Task I: non-routine cognitive analytical; Task II: non-routine cognitive personal; Task III: routine cognitive; Task IV: routine manual 3. Ind1: manufacturing and construction; Ind2: sale and retail trade; Ind3: traffic and post; Ind4: hotel and restaurant; Ind5: financial and housing; Ind6: service; Ind7: social service and social organization; Ind8: Other industry sectors Source Author’s creation based on the data from 2016 CULS

Notes 1. The task approach is advocated by Autor et al. (2003). For detailed review of the issue, please refer to Acemoglu and Autor (2011) and Autor (2013). For studies on the changes of tasks and task classifications around the world, please refer Acemoglu and Autor (2011), Hardy et al. (2017), and Lewandowski et al. (2017). 2. In order to simplify the expression of equations all constant items are omitted. 3. The published debate suggests an index number problem with the BlinderOaxaca model. Estimated results may vary with the kind of comparison group used (Reimers 1983; Neumark 1988; Cotton 1988). We found the results from these models are similar to those based on the Blinder-Oaxaca model. Due to space constraints and because the two sets of decomposition results are almost identical, only estimated results using Eq. 3.1 are presented in this chapter. The other results are available on request. 4. The retirement age is 45 for a female worker, 50 for male worker, 55 for a female cadre, and 60 for a male cadre. 5. Variable values in the range of the “mean valuethree times S.D.” are defined as abnormal values here. 6. The measurement of task is usually done based on the Occupational Information Network (O*NET) data, which is the survey of occupational demands for the USA started in 2003, and following methodologies proposed by Autor et al. (2003) and Acemoglu and Autor (2011). The O*NET task measures are often merged with country-specific data sources,

Task I*Age 50+

Age 50+

Task IV

Task III

Task II

Task I

Age 60+

Age 50–59

Age 40–49

Age 30–39

0.073*** (13.21) −0.001*** (−13.49)

(1)

0.105*** (2.79) 0.172*** (3.33) 0.140* (1.90) 0.157* (1.66) 0.084 (0.64)

(2)

0.098*** (2.65) 0.143*** (2.82) 0.122* (1.69) 0.160* (1.73) 0.094 (0.73) 0.650*** (18.56)

(3)

Results of wage function in China

Age (Age 16–24) Age 25–29

Age_sq

Age

Table 3.8

0.507*** (20.26)

0.100*** (2.73) 0.148*** (2.95) 0.126* (1.76) 0.152* (1.65) 0.072 (0.56)

(4)

0.147*** (17.18)

0.091*** (2.47) 0.145*** (2.86) 0.122* (1.69) 0.145* (1.57) 0.055 (0.42)

(5)

−0.287*** (−8.60)

0.107*** (2.85) 0.176*** (3.42) 0.147* (2.00) 0.158* (1.68) 0.055 (0.42)

(6)

0.091*** (2.54) 0.128*** (2.61) 0.117* (1.61) 0.148 (1.65) 0.036 (0.29) 0.330*** (8.52) 0.327*** (11.81) 0.092*** (10.45) −0.266*** (−8.29)

(7)

0.326*** (8.47) 0.333*** (12.06) 0.092*** (10.35) −0.271*** (−8.47) 0.009 (0.13) 0.221** (1.97)

(8)

66 X. MA AND X. QU

Yes 7,197 0.342

Yes 7,197 0.344

(2)

Yes 7,197 0.374

(3)

Yes 7,197 0.382

(4)

Yes 7,197 0.369

(5)

Yes 7,197 0.350

(6)

Yes 7,197 0.404

(7)

Note 1. ***p < 0.01, **p < 0.05, **p< 0.10 2. Task I: non-routine cognitive analytical; Task II: non-routine cognitive personal; Task III: routine cognitive; Task IV: routine manual 3. The OLS model is used 4. t-values are in parentheses Source Author’s creation based on the data from 2016 CULS

Control variables Number of obs Adj R 2

Task IV*Age 50+

Task III*Age 50+

Task II*Age 50+

(1) −0.097 (−1.26) −0.006 (−0.28) 0.059 (0.69) Yes 7,197 0.403

(8)

3 WORK SKILLS GAP AND THE WAGE DIFFERENTIALS BETWEEN …

67

68

X. MA AND X. QU

usually labor force surveys, and used to calculated task content in countries other than the USA (Arias et al. 2014; Goos et al. 2009, 2014; Dicarlo et al. 2016; Hardy et al. 2017; Lewandowski et al. 2017). However, the use of O*NET is based on the rich information on occupations detailed in O*NET. Therefore the surveys of skills and skill use at work, such as the OECD’s Programme for the International Assessment of Adult Competencies (PIAAC) and the World Bank’s Skills Measurement Program (STEP), offer an opportunity to measure both the country-specific occupational demands and the within-occupation heterogeneity of tasks performed by workers. The survey of task contents in the 2016 CULS is based on the World Bank’s Skills Measurement Program (STEP). The classification of task type in this study is similar with that in Hardy et al. (2017) and Lewandowski et al. (2017) which based on both PIAAC and STEP. For the small differences between PIAAC and STEP, please refer Hardy et al. (2017). 7. Years of experience = age-years of schooling-6. 8. For the influence of wage system (e.g. seniority wage) on employment of older worker in Japan, please refer Chapter 8 of this book.

References Acemoglu, D., & Autor, D. (2011). Skills, tasks and technologies: Implications for employment and earnings. In D. Card & O. Ashenfelter (Eds.), Handbook of labor economics. Volume 4b. North Holland. Aedo, C., Hentschel, J., Moreno, M., & Luque, J. (2013). From occupations to embedded skills: A cross-country comparison (World Bank Policy Research Working Paper). Antonczyk, D., Fitzenberger, B., & Leuschner, U. (2009). Can a task-based approach explain the recent changes in the German wage structure? (IZA DP No. 4050). Antonczyk, D., DeLeire, T., & Fitzenberger, B. (2010). Polarization and rising wage inequality: Comparing the U. S. and Germany (IZA DP No. 4842). Arias, O. S., Sánchez-Páramo, C., Dávalos, M. E., Santos, I., Tiongson, E. R., Gruen, C., de Andrade Falcão, N., Saiovici, G., & Cancho, C. A. (2014). Back to work: Growing with jobs in Eastern Europe and Central Asia. Europe and Central Asia Reports. Washington, DC: The World Bank. Autor, D. H. (2013). The “task approach” to labor markets: An overview. Journal for Labor Market Research, 46(3), 185–199. Autor, D. H., & Dorn, D. (2009). This job is “getting old”: Measuring changes in job opportunities using occupational age structure. American Economic Review, Paper and Proceedings, 99(2), 45–51.

3

WORK SKILLS GAP AND THE WAGE DIFFERENTIALS BETWEEN …

69

Autor, D. H., & Dorn, D. (2013). The growth of low-skill service jobs and the polarization of the US labor market. American Economic Review, 103(5), 1553–1597. Autor, D. H., Levy, F., & Murnane, R. (2003). The skill content of recent technological change: An empirical exploration. Quarterly Journal of Economics, 118(4), 1279–1333. Autor, D. H., & Price, B. (2013). The changing task composition of the US labor market: An update of Autor, Levy, and Murnane (2003) (MIT Working Paper). Becker, G. S. (1957). The economics of discrimination. Chicago: University of Chicago Press. Becker, G. S. (1964). Human capital: A theoretical and empirical analysis, with special reference to education. New York: Columbia University Press. Blinder, A. (1973). Wage discrimination: Reduced form and structural estimation. Journal of Human Resources, 8(4), 436–455. Cotton, J. (1988). On the decomposition of wage differentials. Review of Economics and Statistics, 70(2), 236–243. Demurger, S., Li, S., & Yang, J. (2012). Earning differentials between the public and private sectors in China: Exploring changes for urban local residents in the 2000s. China Economic Review, 23, 138–153. Desjardins, R., & Warnke, A. J. (2012). Ageing and skills: A review and analysis of skillhain and skill loss over the lifespan and over time (OECD Education Working Papers, No. 72). Dicarlo, E., Bello, S. L., Monroy-Taborda, S., Oviedo A. M., Sanchez-Puerta, M. L., & Santos, I. (2016). The skill content of occupations across low and middle income countries: Evidence from harmonized data (IZA Discussion Paper No. 10224). Dustmann, C., Ludsteck, J., & Schönberg, U. (2009). Revisiting the German wage structure. Quarterly Journal of Economics, 124(2), 809–842. Firpo, S., Fortin, N. M., & Lemieux, T. (2011). Occupational tasks and changes in the wage structure (IZA DP No. 5542). Gathmann, C., & Schonberg, U. (2010). How general is human capital? A taskbased approach. Journal of Labor Economics, 28(1), 1–49. Gimpelson, V., & Kapeliushnikov, R. (2016). Polarization or upgrading? Evolution of employment in transitionary Russia. Russian Journal of Economics, 2(2), 192–218. Goos, M., & Manning, A. (2007). Lousy and lovely jobs: The rising polarization of work in Britain. The Review of Economics and Statistics, 89(1), 118–133. Goos, M., Manning, A., & Salomons, A. (2009). Job polarization in Europe. American Economic Review: Papers & Proceedings, 99(2), 58–63.

70

X. MA AND X. QU

Goos, M., Manning, A., & Salomons, A. (2014). Explaining job polarization: Routine-biased technological change and offshoring. American Economic Review, 104, 2509–2526. Hardy, W., Lewandowski, P., Park, A., & Du, Y. (2017). The global distribution of routine and non-routine work: Findings from PIAAC, STEP and CLUS (IBS working paper 06/2017). Heckman, J. J. (1979). Sample selection bias as a specification error. Econometrica, 47 (1), 153–161. Lewandowski, P., Keister, R., Hardy, W., & Górka, S. (2017). Routine and ageing? The international divide in the deroutinisation of jobs in Europe (IBS Working Paper 01/2017). Ma, X. (2018a). Ownership sector segmentation and gender wage gap in urban China during the 2000s. Post-Communist Economies, 30(6), 775–804. Ma, X. (2018b). Labor market segmentation by industry sectors and wage gaps between migrants and local urban residents in urban China. China Economic Review, 47, 96–115. Ma, X. (2018c). Economic transition and labor market reform in China. Singapore: Palgrave Macmillan. Mincer, J. (1974). Schooling, experience and earning. New York: Columbia University Press. Neumark, D. (1988). Employer’s discriminatory behavior and the estimation of wage discrimination. Journal of Human Resources, 23(3), 279–295. Oaxaca, R. (1973). Male-female wage differentials in urban labor markets. International Economic Review, 14(3), 693–709. Peng, F., Anwar, S., & Kang, L. (2017). New technology and old institution: An empirical analysis of the skill-biased demand for older workers in Europe. Economic Modelling, 64, 1–19. Reimers, C. W. (1983). Labor market discrimination against Hispanic and Black men. Review of Economics and Statistics, 65(4), 570–579. Spitz-Oener, A. (2006). Technical change, job tasks, and rising educational demands: Looking outside the wage structure. Journal of Labor Economics, 24, 235–270. Wang, M. (2005). Work chance and wage differentials in the urban labor market: Work and wage of migrants. Chinese Social Sciences, 5, 36–46. (In Chinese). Ye, L., Li, S., & Luo, C. (2011). Industrial monopoly, ownership and enterprises wage inequality: An empirical research based on the First National Economic Census of Enterprises Data. Management World, 4, 26–36. (In Chinese). Zhang, J., & Xue, X. (2008). State and non-state sector wage differentials and human capital contribution. Economic Research, 4, 15–25. (In Chinese).

CHAPTER 4

The Impact of Social Insurance Contributions on Chinese Firms’ Employment and Wages Xinxin Ma and Jie Cheng

4.1

Introduction

Since 1978, the Chinese government has gradually implemented economy transition and reformed the governmental social security systems. As state-owned enterprises (SOEs) were reformed, the governmental security systems were replaced with new social insurance systems.

This chapter is a revised and developed version of Ma, X., & Cheng, J. (2019). The influence of social insurance on wages in China: An empirical study based on Chinese Employee-Employer Matching data. Emerging Markets Finance and Trade. https://doi.org/10.1080/1540496x.2019.1693363. X. Ma (B) Faculty of Economics, Hosei University, Tokyo, Japan e-mail: [email protected] J. Cheng Institute of Population and Labor Economics, Chinese Academy of Social Sciences, Beijing, China e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 X. Ma (ed.), Employment, Retirement and Lifestyle in Aging East Asia, Social Policy and Development Studies in East Asia, https://doi.org/10.1007/978-981-16-0554-3_4

71

72

X. MA AND J. CHENG

From the 1990s these social insurance systems expanded to cover all firms including SOEs, privately-owned enterprises (POEs), and foreignowned enterprises (FOEs). The regulations and policies of the Ministry of Human Resources and Social Security provide six kinds of social security: (a) pension insurance, (b) medical insurance, (c) unemployment insurance, (d) worker injury compensation insurance, (e) maternity insurance, and (f) the housing fund. Except for unemployment insurance (c) and worker injury compensation insurance (d), the social insurance contributions for firms and workers are determined according to wage levels and social insurance contribution premium rate to wage. The average lowest wage level1 for the social insurance contribution is 60% of the regional average wage level in the prior year, and the highest wage level is three times the regional average wage level in the prior year. The total social insurance contribution premium rate paid by firms, excluding the housing fund, was around 43% in 2015. China has regulated for one of the higher social insurance premiums in the world. According to the rules of perfect market competition in order to gain the maximum profit, a firm may transfer the burden of increased social insurance contributions onto its employees by reducing their wage levels and the number of workers. Most previous empirical studies find that an increased social insurance premium may induce a firm to transfer this burden onto its workers by reducing the workers’ wage level or employment (Hamermesh 1979; Summers 1989; Gruber and Krueger 1991; Gruber 1994, 1997; Fishback and Kantor 1995; Anderson and Meyer 2000; Kugler and Kugler 2003; Iwamoto and Hamaaki 2006; Sakei and Kazegami 2007; Ma and Zhang 2018; Ma and Cheng 2019). Yet these empirical study results are not conclusive and there are insufficient empirical studies on the issue for China. This chapter may help to fill the gap in research knowledge. Using the China Employer-Employee Matching Survey (CEES) data and imputed values based on wage and employment functions this study constructs three indices of social insurance enforcement and four types of combinations of wage and employment to investigate the influence of social insurance contributions on employment and wage in Chinese firms. The innovative features of this chapter are as follows. First, numerous studies investigate the influence of social insurance contributions on wage or employment separately, but they do not consider that firms may simultaneously adjust the workers’ wage level and employment to mitigate the social insurance burden. It is thought that the firm may not make the

4

THE IMPACT OF SOCIAL INSURANCE CONTRIBUTIONS …

73

decision about the wage level separately from the decision about employment level. The firm decides on wage level and number of employees in the same decision-making process: There is a combined decision by the firm on the adjustment of wage and employment. To the authors’ knowledge, there is little empirical study on the issue. Considering the heterogeneity of the behavior of firms, we conducted four types of combinations of firms’ wage and employment based on wage and employment functions. Second, three unique indices are used in this study. In China social security policies are set by central government but these policies are interpreted and applied differently by each local government department. As a result, there is no standardized implementation of policy but a regional patchwork of compliance differentials. For example, a firm’s required social insurance contribution premium rate is lower in Guangdong province in the South Region than in Hubei province in the Central Region. Therefore the differences between local governments in contributions and enforcement of social insurance should be considered in the design of an empirical study of this issue. This study conducts two unique indices to investigate the impact of social insurance contributions at local city government level to investigate local government influences. According to current regulations, the de jure contribution rates for social insurance are more than 40% of the total wage bill. However, there is a huge heterogeneity of de facto contribution rates among firms. These compliance anomalies result in the actual social insurance contributions paid by firms frequently being lower than the social insurance contributions stated in the regulations and legislation. Thus, examination of the effect on firm behaviors of differences in actual implementation of social insurance is an important issue for China. In order to examine the influence of social insurance contributions on workers’ wages and on employment in firms this study uses data from the China Employer-Employee Matching Survey (CEES) to construct three indices of social insurance enforcement: the city contribution rate (CCR), the rate of city enforcement of social insurance (CE), and the firms’ actual contribution rate (FAR) (Cheng et al. 2018).2 The results provide new and relevant evidence for social insurance policy reform. Third, the influence of social insurance contributions may differ for various sectors and groups. For example, there is labor market segmentation of the public sector and private sector,3 and of workers with rural or urban registrations (Hukou). Few studies compare these sectors and

74

X. MA AND J. CHENG

groups. This study analyzes and compares the influence of social insurance enforcement on wages for different sectors and groups. Analysis is conducted according to the type of ownership of the firm (SOEs, POEs, or FOEs), firm size (small, medium, and large), firm technology type (capital-intensive vs. labor-intensive) and region (Guangdong vs. Hubei). These results may provide new evidence to understand the complexities of Chinese social security policies. The two most useful new findings are as follows: First, the results suggest that firms in a region with a high city contribution rate (CCR) or a region with a high rate of city enforcement of social insurance (CE) are likely to employ more workers. Firms with high FAR are more likely to have a high wage and more employment (HWME) or to have a high wage and less employment (HWLE), but the probability of having a low wage and more employment (LWME) is low.4 A bipolarization phenomenon is found in the relation between the firms’ actual social insurance contributions and the adjustments of wage and employment. The impact of the actual contribution rate for social insurance is found to be greater on levels of employment than on wages. The second new finding is that the impact of the level of social insurance contribution on a firm’s wage and employment differs for each group. Privately-owned enterprises are likely to reduce the number of workers to mitigate the social insurance burden; the negative impact of increased social insurance contributions is greater for small-size firms, and labor-intensive firms suffer a larger negative impact from the burden of social insurance. The remainder of this chapter is constructed as follows: Sect. 4.2 introduces the theoretical framework for the influence of social insurance contributions on wage and employment, and summarizes previous empirical studies. Section 4.3 gives the methodological framework for empirical analysis and introduces the data and variables used. Section 4.4 presents and explains the estimated results. Section 4.5 summarizes the conclusions and presents policy implications.

4

THE IMPACT OF SOCIAL INSURANCE CONTRIBUTIONS …

4.2 4.2.1

75

Theoretical Framework and Literature Review

The Theoretical Framework for the Influence of Social Insurance Contributions on Wages and Employment

Following the partial equilibrium model, the impact of social insurance contributions on wage and employment is as follows: The demand for labor is determined by the wage level (w) and the social insurance contribution rate (t f ) charged by the enterprise: It can be expressed by Eq. 4.1:    D = D w 1 + tf (4.1) The labor supply is determined by the wage (w) and the social insurance contribution rate (te ) charged to the individual worker. If the workers consider the benefits of social insurance contributions to be a kind of payment for future security in old age (e.g., pension benefit), the labor supply may increase (Summers 1989). Therefore, the labor supply can be expressed by Eq. 4.2:   (4.2) S = S w(1 − ate ) + qwt f Here, a and q stand for two kinds of benefit from contributions that workers expect: the benefit from the contribution charged to the worker (a), and the benefit from the contribution charged to firms (q). Based on the partial equilibrium model, the labor demand is equal to the labor supply in a perfect competitive market, which can be expressed by Eq. 4.3:       dw  d = −σ + σ s q σ d 1 + t f − σ s 1 − ate + qt f w + dt f − σ s sdte )

(4.3)

Here, σ d stands as the wage elasticity of labor demand, σ s s is the wage elasticity of labor supply. Based on Eq. 4.3, the impact of firms’ social insurance contributions on wage or employment can be expressed by Eq. 4.4: −σ d + σ s q dw    /dt f = d  w σ 1 + t f − σ s 1 − at + qt f

(4.4)

76

X. MA AND J. CHENG

It is indicated that the impact of social insurance contributions on wage and employment is determined by the wage elasticity of labor demand (σ d ) and labor supply (σ s ), and the benefit from contributions that the workers expect (a and q). The labor market, the wage elasticity of labor demand and labor supply, and the social security systems differ in different regions, industry sectors (e.g., capital-intensive vs. labor-intensive) and various groups (e.g., public sector vs. private sector; large-firm vs. smallfirm). Therefore the influence of social insurance contributions on wage and employment is not explicable by conventional theoretical Economics and an empirical study is required (Hamermesh 1979; Kotlikoff and Summer 1987; Gruber and Krueger 1991; Gruber 1997; Bojas 2004; Adhikari et al. 2009). 4.2.2

Empirical Studies

Empirical studies made in the 1970s and early 1980s usually employ the social insurance premium or the ratio of social insurance contribution to wages, as an independent variable in wage functions. Most empirical studies found that social insurance contributions negatively affect wage and employment and this indicates that firms transfer the increased social insurance contribution burden to workers. For example, Hamermesh (1979) finds firms in the UK immediately transfer 33% of the social insurance contribution rise onto workers. Holmlund (1983) uses Swedish time series data to analyze the social insurance reform which led to a social insurance contribution ratio increase from 14% to 40% of the wage: The study finds that firms transferred half of the social insurance contribution burden onto their workers. Triplett (1983), Smith and Ehrenberg (1983), and Asher (1984) argue that there is an endogeneity problem. Later empirical studies address the endogeneity problem. For example, most empirical studies use a quasinatural experiment method (Difference in Difference: DID method) to analyze the impact of social insurance policy change on firms’ wages or employment. Notable studies are: the influence of medical insurance, worker injury compensation law, employment insurance for the USA (Gruber and Krueger 1991; Gruber 1994, 1997; Fishback and Kantor 1995; Anderson and Meyer 2000), social insurance for Japan (Komamura and Yamada 2004; Sakei and Kazegami 2007, Tachibanaki and Yokoyama 2008), and total social insurance contributions for Chile (Gruber 1997) and Columbia (Kugler and Kugler 2003), Ma and Zhang

4

THE IMPACT OF SOCIAL INSURANCE CONTRIBUTIONS …

77

(2018) and Ma and Cheng (2019) for China. However, the results of these empirical studies are not consistent. For example, Fishback and Kantor (1995), Anderson and Meyer (2000), Kugler and Kugler (2003), Sakei and Kazegami (2007), Ma and Zhang (2018), and Ma and Cheng (2019) indicate that the level of social insurance contributions negatively affects workers’ wage levels, whereas Gruber and Krueger (1991), and Gruber (1994, 1997) find that the influence of the rise of social insurance contributions on wages is not statistically significant. As far as the authors are aware there is a paucity of detailed empirical study of the influence of firms’ social insurance contributions on both wage and employment. This study provides new evidence for the issue. 4.2.3

Four Special Features of the Chinese Labor Market

Four features emerge for the labor supply and labor demand situation in China which are not found in the published studies of other countries, which may affect the association between social insurance contributions and firms’ employment/wage in China. First, regarding the wage elasticity of labor supply in China, a surplus of labor persists in rural regions (Minami and Ma 2010, 2014). In 2018 the number of internal migrants is estimated to be around 200 million (NBS 2019). Based on the unlimited supplies of labor hypothesis (Lewis 1954), the wage elasticity of the labor supply for low-skilled migrants is small. In addition, even though pension benefit increased with public pension reform, for example, the annual per capita benefit from the Urban Employee Pension increased from 2,558RMB in 1993 (before reform) to 34,512RMB in 2017, the replacement rate of pension to average wage decreased from 79.0% in 1993 to 46.4% in 2017 (NBS 1993, 2018). It is expected that with the replacement rate of pension to wage decreasing, the labor supply of middle-aged and older adults may increase and cause the decrease in the wage elasticity of the labor supply. Second, with the progress of global value chains numerous privatelyowned enterprises developed due to the advantage of low labor costs in China and therefore the wage elasticity of labor demand in China may be higher. It is thought that compared with the labor supply side, the influence of social insurance on wage and employment may be greater for the labor demand side (firms). This is one reason why this study focuses on the firm and not on individual workers.

78

X. MA AND J. CHENG

Third, in China the trade unions are usually managed by the government, and the ratio of worker participation in the trade unions is low in the private sector (e.g., POEs, FOEs), and the influence of collective bargaining on wages is weak. Therefore firms in the private sector may easily transfer the social insurance contribution burden onto their workers, which causes the level of workers’ wages and the number in employment to decrease. Fourth, the Chinese government favored gradualism in implementing reforms,5 thus the labor market is divided into different sectors and groups and the influences of social insurance on wage may differ for each sector and group (e.g., public sector and private sector; large firm and small firm; labor-intensive firm and capital-intensive firm). Therefore further detailed empirical study is required to examine these varied features in the Chinese labor market, particularly the estimations for public sector and private sector are not employed in the published literature. Although many features of the labor market in China are different from those in developed countries, these features are not analyzed in published studies. In additions, firms may link or combine their decision on wage and employment: an element of this study which is not analyzed in the published literature. Regarding these features of Chinese labor market, using a unique employer-employee survey data this study conducts three indices to investigate the influence of social insurance contributions on both employment and wage in Chinese firms.

4.3

Methodology and Data 4.3.1

Model

First, in order to construct a set of situations for the joint decisions, we calculate the reference standards of wage and employment which are expected to be the market equilibrium wage and employment based on firm characteristics and the local labor market situations. A firm’s wage and employment functions are estimated based on the Ordinary Least Squares (OLS) regression model shown by Eqs. 4.5 and 4.6. LnWic = a + β F Fic + βc Cc + vi

(4.5)

4

THE IMPACT OF SOCIAL INSURANCE CONTRIBUTIONS …





Em ic = b + β F Fic + βC Cc + u i

79

(4.6)

In Eqs. 4.5 and 4.6, LnW stands for the logarithm of workers’ average wage in a firm, Em stands for the number of workers in a firm, F is firm level factors (e.g., firm size, industry sector), C is city level factors (e.g., per capita GDP, population density), i for firm, c for city, the constant terms are a and b, and v and u are the error terms. β an β  represent the estimated coefficient for each variable. Using the estimated results of Eqs. 4.5 and 4.6, we calculated the imputed wage level (LnWˆ ) and employment (E m) ˆ for each firm. These are the appropriate values for each firm based on the same wage and employment determinate mechanism. Then we calculated the gap of actual value and imputed value (wage gap = LnW − LnWˆ ; employment gap = Em − E m), ˆ and constructed four kinds of combinations of wage and employment: (i) high wage and more employment (HWME), (ii) high wage and less employment (HWLE), (iii) low wage and more employment (LWME), (iv) low wage and less employment (LWLE). There are two advantages in using this method. First, in comparison with other methods such as using median and mean values of wage and employment, the imputed values based on wage and employment functions are thought to be the appropriate values for each firm based on the similar wage and employment determinate mechanism. The imputed value can be thought of as an optimum value after the heterogeneity of firms and the market mechanism are considered because in wage and employment functions, we control the factors (e.g., firm size, industry, ownership) which may affect wage and employment. Second, we use the city level variables as identification variables in the first step estimations (wage and employment functions), which is similar to the approach of the instrument variables (IV) method to address in part the endogeneity problem. Second, to investigate the impact of social insurance on the probabilities of joint decision on both wage and employment, the multinomial logistic regression (MLR) model is used: ∗ = am + β S I m S Imi + β Fm Fmi + εmi Ymi

Yni∗ = an + β S I n S Ini + β Fn Fni + εni

(i = 1, 2 . . . N ) (i = 1, 2 . . . N )

80

X. MA AND J. CHENG

Y∗ = n

if β S I n S Ini + β Fn Fni − β S I m S Imi

− β Fm Fmi > (am − an ) + (εmi − εni ) (m = n) (m = n)   exp(an + β S I n S Ini + β Fn Fni ) Pr Yi∗ = n = r m=1 exp(am + β S I m S Imi + β Fm Fmi )

(m = 1, n . . . r ) (4.7)

  In Eq. 4.7, Pr Yi∗ = n stands for the probability of adopting one kind of combination for the wage and employment joint decisions in a firm (e.g., a firm decides to reduce both workers’ wage level and employment), n stands for the four kinds of combination for the wage and employment joint decisions: (i) high wage and more employment (HWME); (ii) high wage and less employment (HWLE); (iii) low wage and more employment (LWME); and (iv) low wage and less employment (LWLE). SI denotes the social insurance index including the city contribution rate (CCR); the rate of city enforcement of social insurance (CE); and the firms’ actual contribution rate (FAR). We use low wage and less employment (LWLE) as a reference group in the MLR model. When the coefficient of social insurance (β S I n ) is a negative value and is statistically significant, it indicates that an increased social insurance contribution may negatively affect the probability of one kind of wage and employment joint decision occurring. For example, compared with the probability of becoming an LWLE, a firm is more likely to make one of the other kinds of joint decision (HWME, HWLE, or LWME). It can be argued that the endogeneity problem may remain in Eqs. 4.5 and 4.6, for example, the variables of F such as firm size may influence social insurance. Notably, it can be thought that larger firms may have more influence over local government enforcement actions or they can get away with skimping on their mandated contributions due to their regional employment market power in developed countries. However in China, because there is one party for which the Communist Party of China (CPC) is the guiding organization, the collective bargaining power of large firms to local (or central) government is small (Ma 2019; Ma and Iwasaki 2019, 2021), therefore it is expected that the influence of

4

THE IMPACT OF SOCIAL INSURANCE CONTRIBUTIONS …

81

the reverse causality problem between social insurance contributions and firm size on results is small. We estimate the influence of firm size in t − 1 year on social insurance contribution in t year to check the reverse causality problem, the results indicate that the coefficients are not statistically significant, it is shown statistically that the reverse causality problem does not remain. Finally, to examine the heterogeneities for different groups, a set of subsamples (groups by firm sizes, ownership types, technology types, and regions) are used to facilitate more detailed analyses based on Eqs. 4.5, 4.6, and 4.7. 4.3.2

Data

Matched datasets with both the firms and the workers’ information are rarely available today. Indeed, only few small developed countries such as Denmark have an employer-employee survey matched dataset (Yu 2017). This paper uses the latest data collected in a firm sampling survey, the China Employer-Employee Matching Survey (CEES), for firms in two provinces of China. The CEES was jointly conducted by the Chinese Academy of Social Sciences, Hong Kong University of Science and Technology, Stanford University, and Wuhan University for 2014, 2015, and 2016. The survey objects are focused on firms in the manufacturing industry sector. Sampling was conducted in two stages, each using probability proportionate-to-size sampling, with the size defined as manufacturing employment. Thus, the firm sample is representative of the firm employment size in China. First, firms were sampled according to the size of their labor force as categorized in the third National Economic Census Database of China in 2013. Second, teams of enumerators visited selected firms to gather data. The firms were located in 19 different counties across 13 prefecture-level cities in Guangdong province (South Region) and 20 different counties across 13 prefecture-level cities in Hubei province (Central Region). Between 6 and 10 employees at each firm were sampled using stratified random sampling. The firm was asked to provide a list of all employees enrolled at the end of the previous year. Approximately 30% of the sampled employees were middle and seniorlevel managers, and 70% of the samples were front-line workers. The firm survey instrument gathers useful information including the firm’s characteristics (e.g., ownership type, firm size, and industrial sector), wage,

82

X. MA AND J. CHENG

employment and social insurance contributions et cetera. The worker survey dataset includes employee’s individual information including the individual characteristics (e.g., gender, education, age, and health), and working conditions (e.g., wage, working hours, occupation, and industry) et cetera. We use only the information about firms from the CEES data because this study focuses on the wage and employment levels set by firms.6 After excluding the missing variable and abnormal value samples, 570 firms were surveyed in Guangdong province in the first wave survey in 2015. The response rate in the firm survey was 82% in Guangdong. In the second wave survey in 2016 data was collected for about 531 firms in Guangdong province. The tracking samples are 486 firms, the response rate in the firm survey was 82% in Guangdong. 548 firms were surveyed in Hubei province in 2016. Combining the samples, 84% of the firms were successfully interviewed in Guangdong and Hubei provinces. Although the firm sample is not large, because the sampling of the survey in wave 1 is based on the third National Economic Census Database of China in 2013 which is a national governmental survey, and the attrition rates of panel survey in both wave 2 and wave 3 are low, it is thought that the samples of firms in manufacturing industry from wave 1 to wave 3 are representative of Guangdong and Hubei provinces. 4.3.3

Variables

Using the CEES data, the dependent variables7 are constructed as follows: First, for the wage function, the logarithm of firm average wage was calculated by dividing the total wage by the number of workers in a firm. Second, for the employment function, the number of workers is used. Third, in the multinomial logistic regression model, a category variable is used. It is constructed using the four types of combinations of wage and employment shown in Table 4.1.8 The main independent variables are three indices of social insurance; they are the city contribution rate (CCR) which is defined as the contribution rate based on a city’s policies, city enforcement of social insurance (CE) which calculated based on CCR and wage base levels,9 and firms’ actual contribution rate (FAR) which is defined as the proportion of total payments to social insurance from contributors to total labor compensation for employees. CCR and CE are city level social insurance indices,

4

THE IMPACT OF SOCIAL INSURANCE CONTRIBUTIONS …

Table 4.1 Four combinations of wages and employment

Group type

Classifications

i. High wage and more employment (HWME)

LnW − LnWˆ ≥ 0 and Em − E mˆ LnW − LnWˆ ≥ 0 and Em − E mˆ LnW − LnWˆ < 0 and Em − E mˆ LnW − LnWˆ < 0 and Em − E mˆ

ii. High wage and less employment (HWLE) iii. Low wage and more employment (LWME) iv. Low wage and less employment (LWLE)

83

≥0 G y , sub-income k is mainly distributed in the highincome segment. The increase in the share of sub-income k will lead to an increase in the total income inequality. The contribution of each sub-income to the total income inequality can be calculated as   n+1 k i i − 2 yi (5.4) sk =   n+1 i i − 2 yi

5.2.2

Theil Index and Decomposition

This chapter uses Theil index decomposition to analyze the inequality between and within groups with various types of pension benefit. The purpose is to understand the severity of the gap created by different pension types between different groups. Since other factors are not considered, the focus of this subsection is to examine the statistical characteristics to facilitate the identification and comparison of the impact of different attributes in the pension benefit inequality. The formula is ⎛ ⎞ K K    yi yk yi /yk ⎠ + yk log yk ⎝ log (5.5) T = Tb + Tw = n k /n yk 1/n k k=1

k=1

i∈gk

Assume that n is the sample size. The sample size in group gk is n k , K n k = n. yk is the average income in group gk . Tb and Tw are and k=1 the income gap between and within groups, respectively. The ratio Tb /T is the contribution of inequality between groups to the overall inequality.

5

PENSION BENEFIT INEQUALITY FOR THE ELDERLY IN CHINA

5.2.3

111

Inequality Decomposition Based on Regression

To further examine the contribution rates of various factors to the pension benefit inequality in the same framework, this chapter adopts the decomposition method based on a regression equation (Wan 2004). The basic idea is to first estimate the logarithmic income equation in which the explanatory variable is the logarithm of the personal pension benefit. The explanatory variables include gender, age, age squared, years of education, urban and rural area, region, ownership of pre-retirement work, pre-retirement occupation, and pre-retirement industry. The regression equation is lnyi = β0 + β1 x1i + β2 x2i + · · · + βk xki + μi

(5.6)

Since the semi-logarithmic model is used in the income equation, if the logarithm of the income is used for the decomposition, the distribution of income variables will be distorted. Therefore, the exponent e is used for both sides when decomposing the equation (Chen et al. 2010): y = eβ0 ∗ e(β1 x1i +β2 x2i +···+βk xki ) ∗ eμi

(5.7)

where eβ0 is used as the constant term of the multiplication coefficient. When using related indicators of income disparity such as the Gini coefficient, the constant term can be removed from the above equation without impacting the results. The difference between the inequality index of the observed income y and the inequality index assuming μi = 0 is taken as the impact of μi on the overall income inequality. After obtaining this effect, the difference between the overall income inequality and the inequality caused by the residual is the influence of the explanatory variable in the income determining equation. Therefore, the role of residuals represents the part of the income inequality that cannot be explained by the control variables in the equation. 5.2.4

Data

The data used in this chapter comes from the Chinese Household Income Survey (CHIPs2013), which contains information on residents across 15 provinces. In China, the mandatory retirement ages are 60 for males, 55 for female white-collar workers, and 50 for female blue-collar workers. Furthermore, there are many early retirees. In accordance with the topic

112

P. ZHAN AND H. JIA

of the pension benefit gap that is the focus of this chapter, the sample selected included those aged 50 and above who should receive pension income. The specific screening steps taken were (1) 17,407 samples of 50-year-olds and above were retained; (2) 7,571 samples that were in employment (including reemployment after retirement) and had no pension benefit were deleted; and (3) a sample involving 9,836 elderly people was derived.

5.3 Basic Facts about Pension Income Inequality among the Elderly 5.3.1

The Pension System Has Not Yet Achieved Full Coverage

The implementation of social pension insurance does not mean that pensions will achieve 100% coverage among the elderly; rather, the coverage rate in some areas, especially rural areas, is relatively low (see Table 5.1). In 2013, 19.55% of the elderly who should receive pension Table 5.1 Proportion of the elderly who did not receive pension benefit in 2013

All Region East Center West Age groups 50–59 60–64 65–69 70–74 75–79 80+ Education Never in school or primary school Junior middle school Senior middle school College and above Source Author’s creation

Nation (%)

Rural (%)

Urban (%)

19.55

21.97

15.88

10.19 19.87 19.13

12.77 23.86 24.29

7.73 14.19 12.92

12.51 15.97 13.14 15.28 18.27 27.90

16.95 18.90 15.57 18.40 24.24 35.90

10.88 11.70 9.10 10.82 10.03 14.59

20.49 10.48 9.01 5.16

21.34 13.76 16.67

18.00 8.35 7.55 5.34

5

PENSION BENEFIT INEQUALITY FOR THE ELDERLY IN CHINA

113

benefit did not receive pensions; the proportion of elderly in rural areas who did not receive pensions reached 21.97%, and that in urban areas was 15.88%. In addition, in both rural and urban areas, the coverage rate in the central and western regions was significantly lower than that in the east. Only 10% of the elderly in the east did not receive a pension, but more than 19% of the elderly in the central and western regions did not receive a pension. The pension coverage rate for those aged over 70 years was lower than that for people under 70 years old. Moreover, as much as 27% of those over 80 years old did not receive pensions. The pension coverage rate for the elderly with low education levels was even lower. Among them, the non-receiving proportion for the elderly with a primary school education or less was as high as 20%, nearly twice that of the elderly with a junior middle school education and 3–4 times that of the elderly with a college degree. 5.3.2

The Overall Situation Surrounding the Pension Benefit Inequality

There is significant inequality in the pension benefit levels among the elderly. According to the calculation results in Table 5.2, the Gini coefficients for pension benefit are as high as 0.59. The important reason for this is that the pension benefit inequality among rural elderly is very large, with the Gini coefficient reaching 0.89, which is in sharp contrast to the coefficient of 0.59 for the urban elderly. In addition, there was a huge gap in pension benefit between urban and rural elderly. The average pension benefit of the urban elderly is nearly 10 times that of the rural elderly. Among the urban elderly, the gap between the UEBP and URSP is also very large, which is the main reason for the urban pension inequality. The pension benefit inequality for the rural elderly is mainly reflected in the difference in whether the elderly can enjoy the pension benefits and the low benefit level of the NRRSP. Agricultural hukou residents can only enjoy the NRRSP or the NRRSP, which offers very low benefits. However, some non-agricultural hukou residents in rural areas have the opportunity to enjoy the UEBP. According to the data, the pension benefits of the UEBP are higher than those of the NRRSP, but the proportions of the elderly who received these two benefits are very different. Only about 11% of the rural elderly received the UEBP, whereas 58.7% of the rural elderly received the NRRSP.

77.96 38.07 11.17 36.12 13.21

13,396 12,047 908 332 109

0.59 0.84 0.71 0.43 0.57

70.47 10.60 7.98 58.74 10.31

Note Rural and urban are defined based on the place of permanent residence Source Author’s creation

Pension income UEBP URSP NRRSP Others

% of receiving

Gini/C.C.

% of receiving

Average

Rural

Nation

Inequality and decomposition of pension benefit

Pension sources

Table 5.2

2,253 1,229 21 615 129

Average 0.89 0.96 0.91 0.74 0.87

Gini/C.C.

87.68 73.70 15.30 6.79 16.97

% of receiving

Urban

21,773 20,180 1,379 120 95

0.59 0.62 0.36 −0.04 0.16

Average Gini/C.C.

114 P. ZHAN AND H. JIA

5

PENSION BENEFIT INEQUALITY FOR THE ELDERLY IN CHINA

115

In addition, the kind of pension system that the elderly enjoy is closely related to their type of work before retirement, especially the type of ownership. Table 5.3 reports the average pension benefit of the elderly as grouped into different enterprise ownership or occupation types before retirement. This shows that pension benefit differs greatly as a result of the elderly’s work situation before retirement. Among them, the pensions for the urban elderly who worked in the Communist Party of China and government agencies and institutions before retirement was 2–3 times that of others and pensions for the elderly who were in charge of government organs and institutions or were professional technicians before retirement were significantly higher. Table 5.3 Pension benefit based on different enterprise ownership and occupation type before retirement Enterprise ownerships before retirement

Pension benefit (yuan)

Party and government agencies

42,427

Institutions

36,589

State-owned enterprises Collective enterprises

26,231

Sino-foreign joint venture or wholly foreign-owned

27,126

Self-employment

13,782

Private enterprise Others

19,275 13,494

Source Author’s creation

21,579

Occupation before retirement

Pension benefits (yuan)

Persons in charge of government organs and institutions Professional and technical personnel Clerks

36,335

Commercial and service personnel Production personnel of agriculture, forestry, animal husbandry and fishery Production and transportation personnel Other

22,032

32,769 29,596

13,594

23,462 21,354

116

P. ZHAN AND H. JIA

5.3.3

Regional Differences and Individual Characteristics of Pension Benefit

The regional differences in pension benefit reflect the influence of the level of economic development and the differences in personal characteristics reflect the differences in personal willing and work conditions before retirement. The regional disparities of pension benefit is mainly manifested through the fact that the old-age security level of the elderly in the eastern region is significantly better than that of people in the central and western regions (see Table 5.4). The main reason for this result is that the UEBP for the elderly in the eastern region is significantly higher than that for the central and western regions, the URSP in the eastern and central regions is higher than that for the western regions, and the NRRSP in the eastern regions is more than twice that for the central and western regions. In addition, the urban–rural gap between the central and western regions is relatively large. The pension benefit level for the elderly in central cities is 11 times that of those in rural areas, 12 times that of the west, and 6.7 times that of the east. From the perspective of personal characteristics, the higher the education level of the elderly, the higher the pension benefit level and pension benefit inequality (see Table 5.5). Among them, the average value of the UEBP is significantly positively correlated with education level; the URSP and NRRSP are related to the “inverted U-shaped” type of education level. The main reason for this result is probably because a higher education level corresponds to a better occupation or a better work unit; as a Table 5.4 Regional gap and urban–rural gap of pension benefit (Unit: Yuan) East

All Pension income UEBP URSP NRRSP Others

Center

West

Rural

Urban

Rural

Urban

Rural

Urban

3,843 2,233 315 1,057 237

25,578 23,987 1,378 116 97

1,728 1,090 144 450 45

19,093 17,284 1,701 45 63

1,662 669 427 438 128

19,514 18,154 1,083 174 103

Note The sample sizes in west-rural, west-urban, center-rural, center-urban, east-rural, and east-urban are 1,630, 1,171, 2,304, 1,433, 2,256, and 2,168, respectively Source Author’s creation

5

PENSION BENEFIT INEQUALITY FOR THE ELDERLY IN CHINA

117

Table 5.5 Educational gap of pension benefit (Unit: Yuan) Never in school or primary school

Junior high school

1,883 13,190

3,638 21,069

Rural Urban

Senior high school College and above 8,269 27,248

30,381 36,273

Source Author’s creation

result, a higher proportion of the UEBP and a relatively lower proportion of residents’ social pension coverage are obtained. Among them, the higher the education level of the elderly, the more likely it is for them to have worked in institutions or state-owned enterprises with better welfare before retirement. The elderly with relatively low levels of education are mainly farmers or are in informal sector. For various reasons, their probability of obtaining the UEBP is low, so they can only participate in the URSP and NRRSP, which offer lower treatment levels. There are also obvious pension benefit differences between genders, as female pension coverage rates and their per capita pensions are lower than for men. From Table 5.6, the UEBP for rural elderly men is higher than that for women, which is 2.7 times that for female retirees, and the coverage rate for men is 1.7 times that for women. There is almost no gender difference between the NRRSP coverage rate and the average pension benefit in rural areas, and the coverage rates are both >60%. More Table 5.6 Gender gap in pension benefit Rural

Urban

Female

Male

Female

Male

Coverage Average Coverage Average Coverage Average Coverage Average (%) (Yuan) (%) (Yuan) (%) (Yuan) (%) (yuan) All Pension income UEBP URSP NRRSP Others

75.42

1,886

83.93

3,138

87.82

20,154

90.05

25,445

8.65 8.84 65.20 8.73

752 313 677 143

14.79 8.27 68.83 10.39

2,053 274 670 141

72.20 17.17 7.62 4.60

18,525 1,415 125 89

79.24 12.53 5.51 4.39

23,909 1,349 96 91

Source Author’s creation

118

P. ZHAN AND H. JIA

than 70% of the urban elderly receive UEBP. The coverage rate for the male elderly is slightly higher than that for females. The average pension benefit for the male elderly is nearly 1.3 times that for females, which is relatively large. While the proportion of women participating in the URSP is slightly higher than that of urban men, the difference is not significant. Therefore, the reason for the gender difference in pension benefit mainly comes from the gender gap in the UEBP. The pension benefit also shows a significant difference by age group. In general, pension for the urban elderly increases with age, and the pension benefit level of those aged 75 and above was significantly greater than that for the elderly aged younger than 75 (Table 5.7). However, the rural areas showed the opposite result. The pensions for the rural elderly aged younger than 60 were much higher than for other rural elderly. The reason may come from the different system establishment histories by urban and rural areas. The NRRSP, which has the ability to cover all rural elderly, was only introduced in 2009, and it was stipulated that in 2009, people over 60 years old (women over 55 years old) could receive the NRRSP directly. Therefore, NRRSPs for different ages would not differ greatly. Due to labor mobility and the gradual improvement of social security systems in enterprises, the coverage of the UEBP and URSP in rural residents has gradually increased. This increase process mainly began as a result of the payment of pension contributions by young people, which shows that the proportion of young people participating in the UEBP and URSP and the benefits are greater than those for old people. This reason has also led to the continuous expansion of the Table 5.7 Age gap of pension benefit (Unit: Yuan) Age group

50–59 60–64 65–69 70–74 75–79 80+

All Pension benefit

UEBP

Rural

Urban

Rural

Urban

4,304 2,289 2,368 2,586 2,607 1,982

21,893 21,114 22,224 22,742 25,981 26,078

3,004 1,204 1,208 1,557 1,478 954

20,784 19,344 20,273 20,775 24,620 24,470

Source Author’s creation

URSP

NRRSP

Others

Rural Urban Rural Urban Rural Urban 656 229 289 250 340 293

1,046 1,559 1,611 1,630 1,145 1,460

508 704 728 662 635 598

29 85 215 239 114 75

137 152 142 117 154 138

34 125 124 99 101 73

5

PENSION BENEFIT INEQUALITY FOR THE ELDERLY IN CHINA

119

pension benefit inequality with increases in age between the elderly living in urban and rural areas.

5.4 Decompositions of Pension Benefit Inequality 5.4.1

The Findings from Theil Decomposition

What causes the large inequality in pension benefit? This section uses Theil index decomposition to analyze the contribution of various characteristics to the inequality of national, rural, and urban pension benefit separately. Table 5.8 reports the inequality within and between specific groups according to the Theil decomposition approach. Tables 5.11 and 5.12 report the urban and rural results. According to Table 5.8, the three factors that contribute the most to pension benefit inequality are pension type, differences between urban and rural areas, and differences in education. Their contribution rates between the groups are 54.1%, 44.2%, and 32.6%, respectively. The factor that contributes the most to the inequality of UEBP and URSP is education, with contribution rates of 19.1% and 10%, respectively. The factor with the largest contribution to NRRSP inequality is regional differences, and the contribution rate is 9.2%. However, when we separately measured rural elderly and urban elderly, the results showed some differences (see Tables 5.11 and 5.12). The contribution of pension type to pension benefit inequality is greater in rural than in urban areas, with the former contributing 41% and the latter only 18.2%. The reason for this is that many rural residents have no pension benefit or very low pension benefit, and only a few rural elderly people have obtained UEBP, which has a higher pension benefit. In addition, the contributions of education to rural and urban pension inequality are both approximately 11.7%, which indicates that education is a very important factor in the pension benefit inequality between urban and rural elderly. Regional differences contribute only 5% to the total pension benefit inequality of the rural elderly. However, if only new rural pensions are considered, the contribution reaches 13%. This shows that the differences in NRRSP caused by different economic development levels are very significant. This has a greater relationship with the formulation of the NRRSP benefit.

120

P. ZHAN AND H. JIA

Table 5.8 Theil decomposition of pension benefit in 2013 for China Within groups

All pension benefit Pension type Urban-rural Region Gender Education level UEBP Urban-rural Region level Gender Education level URSP Urban-rural Region Gender Education level NRRSP Urban-rural Region Gender Education level Other pension benefit Urban-rural Region Gender Education level

Between groups

Theil

Contribution (%)

Theil

Contribution (%)

0.26 0.31 0.54 0.55 0.37

45.91 55.79 96.93 99.52 67.40

0.30 0.25 0.02 0.00 0.18

54.09 44.21 3.07 0.48 32.60

0.20 0.21 0.22 0.18

91.21 96.45 98.60 80.90

0.02 0.01 0.00 0.04

8.79 3.55 1.40 19.10

0.62 0.66 0.69 0.62

90.14 96.01 99.63 89.98

0.07 0.03 0.00 0.07

9.86 3.99 0.37 10.02

0.63 0.59 0.65 0.64

96.86 90.95 99.62 98.69

0.02 0.06 0.00 0.01

3.14 9.05 0.38 1.31

0.71 0.69 0.72 0.72

98.00 95.07 99.79 99.47

0.01 0.04 0.00 0.00

2.00 4.93 0.21 0.53

Source Author’s creation

In urban areas, enterprise ownership before retirement contributes greatly to pension inequality, exceeding 80%. This is mainly due to the large difference in the distribution of pension insurance types between government agencies, institutions, and private enterprises, particularly the distribution of UEBP. The occupational difference before retirement is also an important factor in the inequality of pension benefit for the urban elderly, with a contribution rate of 27%.

5

5.4.2

PENSION BENEFIT INEQUALITY FOR THE ELDERLY IN CHINA

121

Further Findings from Inequality Decomposition Based on Regression Model

In the result of Theil index decompositions, the contribution of many features may be due to the influence of other features. For example, regional differences and urban–rural differences include the results of pension insurance type. The more concerning question for this chapter is, if other factors are controlled, does the influence of certain key factors remain significant? The regression-based decomposition method introduced by Wan (2004) can answer this question well. Based on previous studies and discussions, this study selected the following variables in the regression model as explanatory variables: pension insurance type, gender, age, years of education, dummy variable of urban residence, and region. In the urban questionnaire, there were further inquiries to identify the work features before retirement. Therefore, in the urban model, the regression equation also included pre-retirement enterprise ownership, pre-retirement industry, and preretirement occupation as explanatory variables. Table 5.13 shows the estimated coefficients. According to the estimated values, pension insurance type, urban–rural differences, and personal characteristics significantly affect elderly pension income. If the urban and rural samples are estimated separately, their results are similar; furthermore, pre-retirement enterprise ownership and pre-retirement show significant impacts. Table 5.9 reports each variable’s contribution rate based on the regression-based decomposition method. Evidently, when controlling for other conditions, the most important cause of the pension benefit inequality is the difference in pension insurance type, with a contribution rate of 70.7%. Among them, UEBP has the greatest impact, followed by NRRSP. According to the results in columns 2–5, UEBP is an important reason for the gap within urban areas, and NRRSP is an important reason for the gap within rural areas. When the urban and rural areas are separated and decomposed, it is also found that the causes of urban pension benefit inequality are pension type, enterprise ownership, and occupation before retirement. The main causes of pension benefit inequality for the rural elderly are the gaps in pension insurance type and regional differences. The contribution of education is relatively large throughout the whole nation and within urban and rural areas. The contribution rate of education in regression-based decomposition is lower than that obtained by

122

P. ZHAN AND H. JIA

Table 5.9 Decomposition results of contribution of each variable to pension benefit inequality based on the regression model

Pension type UEBP URSP NRRSP Gender Age Age squared Education years Urban–rural Region Ownership pre-retirement Occupation pre-retirement Industry pre-retirement

(1) Nation (%)

(2) Rural (%)

(3) Urban A (%)

(4) Urban B (%)

(5) Urban C (%)

70.72 50.70 6.57 13.46 0.17 2.29 2.23 4.49 8.84 2.23

75.47 18.99 8.46 48.02 0.01 0.07 0.36 2.77

72.14 60.99 9.68 1.48 0.27 5.82 1.56 2.60

74.20 62.64 9.96 1.59 0.31 6.03 1.57 3.17

77.77 65.39 10.67 1.71 0.24 7.54 2.45 4.79

4.84

2.42 14.08

2.41

3.16

11.41 1.03

Source Author’s creation

Theil decomposition in the previous section. This shows that there is a correlation between the difference in education level and the distribution of pension insurance type. People with higher education levels are more likely to get UEBP with higher benefits, which is consistent with the descriptive statistics above. According to the coefficient estimation of the variable urban in column 1, the contribution rate of urban–rural differences to the national pension benefit inequality is up to 8%. This is the contribution rate after controlling for the difference in pension type, which means that even if covered by the same type of pension, the pension benefit within urban and rural regions is still large. According to the second and third columns, the contribution of the region is greater in rural areas and relatively small in urban areas. The reason for this is that there are big differences in NRRSP in different regions, which is related to their different regional economic development. The further question that should be considered here is how can we improve the level of rural pensions in areas with poor economic development? Columns 3–5 in Table 5.9 contain the decomposition of the urban pension benefit inequality. The results in column 3 control the ownership

5

PENSION BENEFIT INEQUALITY FOR THE ELDERLY IN CHINA

123

of work before retirement. Referring to Zhang et al. (2016), ownership of work is divided into four categories: state-owned enterprises (SOE), collectively-owned enterprises (COE), privately-owned enterprises (POE), and other. Clearly, the difference in ownership before retirement contributes significantly to urban pension benefit inequality, with a contribution rate of 14.1%. This is mainly because early UEBP was primarily distributed in the state-owned enterprises, and the relevant regulations in privately-owned enterprises were imperfect. Column 4 controls the occupational differences before retirement. Occupations are classified as white-collar, blue-collar, service industry, and other. Occupation before retirement is also a main contributor to the urban pension benefit inequality, with a contribution rate of 11.4%. Column 5 controls the industry type of pre-retirement work and shows that industry differences are not the main reason for the urban pension benefit inequality. Table 5.10 reports the contribution of various factors to specific types of pension benefit. Column 1 reports the results of NRRSP. The inequality within the NRRSP mainly comes from regional gaps, and the contribution rate to regional inequality is as high as 18.67%. In addition, age is an important source. Column 3 reports the UEBP decomposition results. Among them, the contribution of work attributes before retirement is very large. When the characteristics of work before retirement are decomposed separately, the contribution of enterprise ownership to Table 5.10 Decomposition results of contribution of each variable to pension benefit inequality based on the regression model, specific pension type

Gender Age Age squared Education years Region Urban-rural Ownership pre-retirement Occupation pre-retirement Industry pre-retirement Source Author’s creation

(1) NRRSP (%)

(2) URSP (%)

(3) UEBP A (%)

(4) UEBP B (%)

(5) UEBP C (%)

0.94 8.19 6.38 3.21 18.67

0.57 6.98 0.46 2.24 5.32

0.51 2.28 0.27 21.49 3.78 7.43 51.13

0.42 0.40 1.83 22.23 3.65 14.73

0.43 4.04 1.05 32.47 3.75 35.66

42.34 4.29

124

P. ZHAN AND H. JIA

pre-retirement work is as high as 51%, and the contribution of occupation pre-retirement reaches 42%. In addition, the contribution rate of education is 20–32%, and the contribution of the gap between urban and rural is 7–36%. This means that the young-age employee pension status and personal income status have an important impact on an individual’s entire lifecycle income, indicating the complexity of the employee pension research. In comparison, the contributions of age, region, and gender in column 2 to URSP inequality are very small, no more than 10%. This shows that the disparity in URSP mainly comes from other reasons and may not have much to do with common economic factors. Due to differences in the design rules of basic pension insurance for employees and the social pension insurance for residents, the benefits of UEBP will definitely be higher than those of URSP and NRRSP. However, the current gap between the two pension system types seems too large, so the pension benefit of the elderly remains relatively large. As a result, under the limited level of pension funds, it is more difficult to solve welfare problems for the large proportion of the elderly whose main source of income is their social pension. In addition, even if the types of pension are controlled, the differences between urban and rural areas in regions remain large. This has a great correlation with the economic development level between regions and between urban and rural areas. Determining how to fill this gap and improve the level of security in low-income areas is a question worthy of in-depth consideration.

5.5

Conclusions

Taking into account the particularity of the elderly group, this chapter studied the inequality of pension benefit among the elderly. The main data source was the 2013 data of the Chinese Household Income Survey (CHIPs2013). The basic findings include the following: First, there is significant inequality in pension benefit in China. The Gini coefficient of national pension benefit reached 0.59. Pension inequality is significantly higher in rural than in urban areas, and its Gini coefficient is as high as 0.89. Comparing different types of pension schemes, UEBP benefit is much higher than the other types, and URSP benefit is significantly higher than NRRSP benefit. The pension coverage rates and pension benefit inequalities within urban and rural areas show different characteristics in terms of gender, region, education level, and age. Second, from a national perspective, the main source of the pension benefit inequality is the type

5

PENSION BENEFIT INEQUALITY FOR THE ELDERLY IN CHINA

125

of pension insurance, with a contribution rate more than 70%. In addition, the urban–rural gap and the gap in education level are also greatly affected. Looking solely at rural areas, gaps between pension insurance types, in education level, and between regions are the main reasons. The main source of the pension benefit inequality in urban areas is the gap caused by the pension insurance type and the gap in education level and enterprise ownership of work and occupation before retirement. In response to the above conclusions, the merging of the social pension insurance system for urban and rural residents since 2014 is a policy reform that meets the current needs. In addition, it is still necessary to continue promoting the reduction of the gap caused by the pension insurance type, particularly considering how to balance the gap between urban employee pension insurance and other types of pension insurance. It is necessary to consider increasing the pension coverage of the informal employment labor force and migrant population so that more of the labor force who can only participate in residents’ pensions can participate in employee pensions. Solving these problems will require further improving social security policies in the labor market, improving the social security system for the elderly in rural areas, and narrowing the gap between urban and rural areas in public service levels.

Appendix See Tables 5.11, 5.12, and 5.13. Table 5.11 Theil decomposition of pension benefit for 2013 in rural China Within groups

All pension benefit Pension type Region Gender Education level

Between groups

Theil

Contribution (%)

Theil

Contribution (%)

0.64 1.03 1.07 0.96

58.67 94.77 98.70 88.29

0.45 0.06 0.01 0.13

41.33 5.23 1.30 11.71

(continued)

126

P. ZHAN AND H. JIA

Table 5.11 (continued) Within groups

UEBP Region Gender Education level URSP Region Gender Education level NRRSP Region Gender Education level Other pension benefit Region Gender Education level

Between groups

Theil

Contribution (%)

Theil

Contribution (%)

0.46 0.46 0.44

94.55 94.83 91.29

0.03 0.03 0.04

5.45 5.17 8.71

0.74 0.73 0.72

99.99 99.58 98.56

0.00 0.00 0.01

0.01 0.42 1.44

0.50 0.57 0.57

86.74 99.74 98.82

0.08 0.00 0.01

13.26 0.26 1.18

0.96 1.03 1.03

92.57 99.54 99.54

0.08 0.00 0.00

7.43 0.46 0.46

Source Author’s creation

Table 5.12 Theil decomposition of pension benefit for 2013 in urban China Within groups

All pension benefit Pension type Region Gender Education level Enterprise ownership pre-retirement Occupation pre-retirement UEBP Region Gender Education level

Between groups

Theil

Contribution (%)

Theil

Contribution (%)

0.21 0.24 0.24 0.96 0.03

81.82 97.69 97.35 88.29 10.35

0.05 0.01 0.01 0.13 0.29

18.18 2.31 2.65 11.71 89.65

0.23

73.12

0.09

26.88

0.18 0.18 0.44

96.14 97.94 91.29

0.01 0.00 0.04

3.86 2.06 8.71

(continued)

5

PENSION BENEFIT INEQUALITY FOR THE ELDERLY IN CHINA

127

Table 5.12 (continued) Within groups

Enterprise ownership pre-retirement Occupation pre-retirement URSP Region Gender Education level Enterprise ownership pre-retirement Occupation pre-retirement NRRSP Region Gender Education level Enterprise ownership pre-retirement Occupation pre-retirement Other pension benefit Region Gender Education level Enterprise ownership pre-retirement Occupation pre-retirement Source Author’s creation

Between groups

Theil

Contribution (%)

Theil

Contribution (%)

0.04

16.06

0.23

83.94

0.21

76.29

0.06

23.71

0.58 0.60 0.72 0.07

95.79 98.48 98.56 81.34

0.03 0.01 0.01 0.02

4.21 1.52 1.44 18.66

0.10

99.40

0.00

0.60

0.70 0.82 0.57 0.11

85.19 99.99 98.82 70.91

0.12 0.00 0.01 0.05

14.81 0.01 1.18 29.09

0.11

71.61

0.04

28.39

1.83 1.85 1.03 0.00

98.53 99.68 99.54 19.25

0.03 0.01 0.00 0.00

1.47 0.32 0.46 80.75

0.00

79.60

0.00

20.40

Ownership pre-retirement

Center

Region East

Urban*male

Urban

0.5147*** (0.0476) 0.0554 (0.0526)

2.7836*** (0.0597) 3.6496*** (0.0587) 6.3465*** (0.0545) 0.0205 (0.0403) 0.1844*** (0.0285) −0.0012*** (0.0002) 0.0909*** (0.0062) 0.7985*** (0.0596)

(1) Nation

0.5900*** (0.0656) −0.0737 (0.0684)

2.6582*** (0.0918) 4.1651*** (0.0602) 5.1509*** (0.0858) 0.1739*** (0.0560) 0.1427*** (0.0456) −0.0010*** (0.0003) 0.0482*** (0.0097)

(2) Rural

0.4947*** (0.0623) 0.2993*** (0.0723)

2.8628*** (0.0874) 0.1902 (0.1950) 5.6956*** (0.0891) −0.1221** (0.0535) 0.1697*** (0.0368) −0.0010*** (0.0003) 0.0392*** (0.0086)

(3) Urban

OLS model results of pension benefit equations for 2013

Education years

Age squared

Age

Male

UEBP

NRRSP

Pension type URSP

Table 5.13

−0.2089*** (0.0705) −0.3638*** (0.0779)

−0.3154*** (0.0592) 0.1252*** (0.0416) −0.0009*** (0.0003) −0.0142* (0.0078)

(4) URSP

0.4982*** (0.0708) 0.2535*** (0.0782)

0.4465*** (0.0594) 0.7001*** (0.0417) −0.0051*** (0.0003) −0.2942*** (0.0079)

(5) NRRSP

0.0675 (0.1357) −0.6847*** (0.1571)

3.4545** (1.6381) 0.2478*** (0.0796) −0.0016*** (0.0006) 0.0706*** (0.0186) 3.8612*** (1.1171) −3.6801** (1.6409)

(6) UEBP

128 P. ZHAN AND H. JIA

(3) Urban

(4) URSP

(5) NRRSP

(6) UEBP

0.7290*** 2.9330*** (0.1933) (0.4154) 0.6559*** 2.7333*** (0.1521) (0.3266) 0.5586*** 2.2297*** (0.1368) (0.2930) 0.3267** 2.0489*** (0.1388) (0.2975) 0.2075 2.0697*** (0.3626) (0.7759) 0.0152 0.3906 (0.1857) (0.3994) Y Y Y Y −5.3339*** −3.4492** −3.8586*** −2.7764** −19.9833*** −6.2620** (0.9622) (1.5855) (1.2417) (1.4111) (1.4164) (2.9135) 0.720 0.644 0.586 0.007 0.182 0.097 9,046 4,641 3,460 9,046 9,046 3,486

(2) Rural

Note 1. * p < 0.10, ** p < 0.05, *** p < 0.01. The values in the brackets are the standard errors 2. The dependent variable is the logarithm of (pension income + 1) 3. Standard errors are in parentheses Source Author’s creation

Adj-R 2 Obs.

Occupation pre-retirement Industry type pre-retirement Constant

Sino–foreign joint venture or wholly foreign-owned Private enterprise

Collective enterprises

State-owned enterprise

Institutions

Party and government agencies

(1) Nation

5 PENSION BENEFIT INEQUALITY FOR THE ELDERLY IN CHINA

129

130

P. ZHAN AND H. JIA

References Chen, Z., Wan, G. H., & Lu, M. (2010). Inter-industry income inequality: An increasingly important cause of income disparity in urban China—A regression-based decomposition. China Social Science, 3, 65–76. (In Chinese). Fang, H., & Feng, J. (2018). The Chinese pension system (NBER working paper 25088). Fei, J., Ranis, G., & Kuo, S. W. Y. (1978). Growth and the family distribution of income by factor components. The Quarterly Journal of Economics, 92(1), 17–53. Huang, W., & Zhang, C. (2021). The power of social pensions: Evidence from China’s new rural pension scheme. American Economic JournalApplied Economics (forthcoming). https://www.aeaweb.org/articles?id=10. 1257/app.20170789. Jiao, N. (2016). Does public pension affect intergenerational support in rural China? Population Research, 40(04), 88–102. (In Chinese). Li, J., Wang, X., Xu, J., & Yuan, C. (2020). The role of public pensions in income inequality among elderly households in China 1988–2013. China Economic Review, 61(6), Li, S., & Zhu, M. (2018). Changes in the income gap of Chinese residents in the 40 years of China’s economic transition. Management World, 12, 19–28. (In Chinese). Li, Z. (2017). Social security theory (4th ed.). China Labor and Social Security Press. (In Chinese). Luo, C. L. (2019). The missing top income group in household surveys and the underestimation of income inequality. Economic Perspectives, 1, 15–27. (In Chinese). Pyatt, G., Chen, C., & Fei, J. (1980). The distribution of income by factor components. The Quarterly Journal of Economics, 95(3), 451–473. Sicular, T., Sato, H., Li, S., & Yue, X. (2020). Changing trends in China’s inequality: Evidence, analysis and prospects. Oxford University Press. Song, X. W. (2016). Research on major issues of social security system during the “13th Five-Year Plan” in China. China Labor and Social Security Press. (In Chinese). Wan, G. (2004). Accounting for income inequality in rural China: A regressionbased approach. Journal of Comparative Economics, 32(2), 348–363. Whitaker, E., & Bokemeier, J. (2013). Patterns in income source expectations for retirement among preretirees. Research on Aging, 13, 467–496. Zhang, L., Wu, B. B., Li, S., & Démurger, S. (2016). The impact of household registration barriers for sectoral entry on income household registration discrimination: The decomposition of income gap based on micro-simulation methods. Chinese Rural Economy, 2, 36–51. (In Chinese).

PART II

Employment, Retirement, and Lifestyles of the Elderly in Japan

CHAPTER 6

Health Capacity to Work and Its Long-Term Trend Among the Japanese Elderly Takashi Oshio

6.1

Introduction

A combination of shrinking labor force and large fiscal deficits are urgent and common challenges among developed countries. The main driving force behind these two serious concerns is the rapid pace of population aging, which dampens labor force participation (LFP) with continued low fertility and expands fiscal deficits under a pay-as-you-go public pension program. A natural and simultaneous solution to these two policy challenges is to encourage the elderly to continue working as late as possible in terms of age. Thus, the main visible target of recent pension reforms

This study was conducted as a part of the Project “Study on the Medium- to Long-Term Social Security System” undertaken at Research Institute of Economy, Trade and Industry (RIETI). The author is grateful for helpful comments and suggestions by Discussion Paper seminar participants at RIETI on October 29, 2018. T. Oshio (B) Institution of Economics Research, Hitotsubashi University, Tokyo, Japan e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 X. Ma (ed.), Employment, Retirement and Lifestyle in Aging East Asia, Social Policy and Development Studies in East Asia, https://doi.org/10.1007/978-981-16-0554-3_6

133

134

T. OSHIO

has been to raise the pensionable age, although pension reforms are often accompanied by revisions in various aspects, such as coverage, adequacy, and sustainability, as well as work incentives (OECD 2017). In fact, many developed countries have implemented or will execute public pension reforms to extend the pensionable age. Although the LFP rate in Japan is higher than in most other developed countries, there have been many policy debates on enhancing the normal eligibility of pensionable age (Oshio et al. 2018). Recently, Japan has been extending the pensionable age. For male pensioners, the eligibility age for the flat-rate benefit of the Employees’ Pension Insurance (EPI) program has increased by one year every three years, rising from 60 years in 2001 to 65 years in 2013. Furthermore, the eligibility age for the EPI wageproportional benefit was scheduled to increase by one year every three years from 2013 to 2025, in order to reach 65 years at the end of this period. For female pensioners, while keeping a five-year lag relative to that for men, the eligibility age for the flat-rate benefit started to be raised in 2006, and that for the wage-proportional benefit started be raised in 2018 in the same manner (Oshio et al. 2011). However, it is possible that a simple extension of the pensionable age may not work, because not all older adults are able to work regardless of whether they are willing to do so. In particular, a major constraint on working is health, either physical or mental, which may also be associated with a decline in cognitive functions. In this context, a simple extension of the pensionable age, which ignores heterogeneity among the elderly, may result in increased inequality between healthy and unhealthy individuals and exacerbate the overall living standard of the elderly. Hence, one of the major concerns to be addressed is whether elderly workers are healthy enough to work longer. This study aims to estimate the additional health capacity of the elderly to work in Japan and its long-term trend over the past 30 years (i.e., between 1986 and 2016), which provides the first such evidence to the best of our knowledge. This study relies on Cutler et al.’s (2012) model (referred to as the CMR model hereafter), a study that estimated the work capacity of the elderly in the United States. Questioning how it would be if people with a given level of health worked as long as they could, the CMR model simulates the work capacity of the age group entitled to receive social security benefits based on the estimated association between the health and work statuses of the age group immediately below the eligibility age.

6

HEALTH CAPACITY TO WORK AND ITS LONG-TERM …

135

Based on simulation results, Cutler et al. (2012) conclude that the elderly’s work capacity is substantial. The CMR model implicitly assumes that the relationship between the health and work statuses of age groups below the eligibility age is stable and holds for the age group above the eligibility age. Correspondingly, any simulated decline in work given the same level of health status is attributable to factors other than health deterioration, particularly to social security benefits. Using this model commonly, the NBER International Social Security Project found substantial slack of work capacity among the elderly in 12 developed countries (Wise 2017), including Japan (Usui et al. 2017).1 In this study, we apply the CMR model to individual-level data obtained from the nationwide, population-based survey “Comprehensive Survey of the Living Conditions” (CSLC), which was conducted and released by the Ministry of Health, Labour and Welfare (MHLW) of the Japanese Government. This study uses data collected from the household and health datasets of the 2016 CSLC, along with its 1986 and 2001 surveys. This study differs from previous ones in three aspects. First, when applying the CMR model, this study considers the characteristics of the elderly LFP in Japan. As a benchmark analysis for assessing overall work capacity, this study first divides work status into “work” and “no work,” and focuses on the elderly’s decision to choose between these two options. Then, this study expands the analysis by dividing work into “fulltime” and “part-time,” a division motivated by the fact that a substantial portion of Japanese male employees shift to part-time work after retiring from their primary full-time work, rather than completely abandoning the labor force (Shimizutani 2011; Shimizutani and Oshio 2010). Second, this study examines the long-term trend of the health capacity to work over the past 30 years, using data from the CSLC conducted in 1986, 2001, and 2016. Considering the improving trend of the elderly health together with a wider coverage of public pension programs, it is reasonable to hypothesize that the additional work capacity has increased over the past 30 years. Lastly, this study examines the distribution of work capacity among the elderly. Even if verified, a high health capacity to work among the elderly would not directly underscore an increase in the age of mandatory retirement or eligibility for pension benefits. As there may be a portion of people who are not able to work due to health conditions, policies should address heterogeneity among the elderly. This study thus examines

136

T. OSHIO

the distribution of the expected probabilities of work and computes the proportion of those who are unable to work due to health conditions. The remainder of this chapter is organized as follows. Section 6.2 describes the data, Sect. 6.3 explains the analytic strategy, Sect. 6.4 presents the estimation results, and Sect. 6.5 presents a conclusion.

6.2 6.2.1

Data Study Sample

Conducted by the MHLW, the CSLC started in 1986; its household survey has been conducted every year since then, while its health and income/savings surveys have been conducted every three years. This study mainly used individual-level data obtained from the 2016 CSLC, which was conducted in early June 2016. Samples of the CSLC were collected nationwide through a two-stage random sampling procedure. First, about 5,400 districts were randomly selected from about 940,000 national census districts. Second, about 290,000 households were randomly selected from each selected district, according to its population size. All members of each selected household were asked to complete the questionnaires. A total of 224,641 households and their members (568,425 individuals) responded to the survey. The response rate was 77.6% at the household level. By restricting the study sample to individuals aged between 50 and 74 years and excluding respondents who did not have essential variables, this study was then limited to the use of data of 197,004 individuals (94,083 men and 102,921 women). The key results were then compared with those obtained from the 1986 and 2001 CSLCs, which contain data of 213,826 individuals (99,282 men and 114,544 women) and 234,139 individuals (111,805 men and 122,334 women), respectively. 6.2.2

Variables

6.2.2.1 Work Work status was divided into “work” and “no work” based on the participants’ answers to the question whether they did any paid work in May. Work was further divided into “full-time” and “part-time” based on

6

HEALTH CAPACITY TO WORK AND ITS LONG-TERM …

137

whether the reported hours worked in the week of mid-May were 35 h or longer (full-time) or not (part-time). 6.2.2.2 Health The following health variables were considered: (i) self-rated health, (ii) diagnosed diseases, (iii) functional disabilities, (iv) psychological distress, (v) hospitalization, and (v) smoking. Variables (i) to (iv) were addressed as follows: i. Self-rated health: Respondents were asked about their current health conditions according to a five-point scale (good, somewhat good, average, somewhat poor, and poor). Six binary variables were established for each scale, as well as for unanswered questions. ii. Diagnosed diseases: Respondents were also asked whether they had each of the 41 diagnosed diseases (such as diabetes, obesity, and hyperlipidemia). Binary variables were established for each of them, as well as for unknown and unanswered. iii. Functional disabilities: Respondents were asked whether they had any difficulty in (a) everyday activities, (b) going out, (c) work, housekeeping, and study, (d) exercise and sports, (e) other(s). Binary variables were established for each of these difficulty items, as well as for unanswered. iv. Psychological distress: The participants’ assessments on this item were obtained through the following six-item questionnaire rated on a 5-point scale (0 = not at all to 4 = all the time): “During the past 30 days, approximately how often have you felt a) nervous, b) hopeless, c) restless or fidgety, d) so depressed that nothing could cheer you up, e) that everything required effort, and f) worthless?” The sum of the reported scores were calculated (range: 0–24) and defined as the Kessler 6 (K6) score (Kessler et al. 2002, 2010). Higher K6 scores reflect higher levels of psychological distress. Two binary variables of psychological distress were then determined, to which we allocated “1” to 5 ≤ K6 ≤ 12 and K6 ≥ 13, respectively. K6 ≥ 5 and ≥ 13 indicate a mood/anxiety disorder and severe mental illness in a Japanese sample, as established by Sakurai et al. (2011).

138

T. OSHIO

In addition to (i) to (iv), this study considered whether the respondents were currently hospitalized and whether they were current smokers. Along with these individual-level health variables, we consider life expectancy, which indicates how many additional years individuals are expected to live on average at each age, as a proxy of general health status for those at that age. Life expectancy was included as an explanatory variable in regression models as the estimates that use information only about self-reported, individual-level health status tend to fail to capture full health dimensions and therefore overstate the ability to work at more advanced ages. Data on life expectancy was collected from the Life Tables of each year released by the MHLW.2

6.3 6.3.1

Analytic Strategy

Estimation of the Health Capacity to Work

Following the procedure incorporated in the CMR model, this study first estimated the linear probability model (referred to as Model 1 hereafter) to explain the binary variable of the “no work” status by a full set of health variables for individuals aged 50–59 years. Next, the potential work capacity was calculated for those aged 60–64, 65–69, and 70–74 years, respectively, using the estimated regression parameters obtained from the regression model and a set of actual values of health variables in each age group. The gap between the potential work capacity and the actual employment rate was defined as the additional work capacity. The actual employment rate and potential/additional work capacities are presented in terms of the proportion out of the total number of respondents in each age group. These calculations were conducted separately for men and women. Poterba et al. (2013) propose an alternative regression model, in which the full set of individual-level health variables is replaced by a single health index value. To this end, the authors first obtain the first principal component of the individual-level health variables. Then, they use the coefficients estimated from this analysis to predict a percentile score for each respondent, referred to as the composite health index. They demonstrate that this index is strongly related to mortality and future health events, such as strokes and diabetes onsets. This study thus estimated Poterba et al.’s alternative model (referred to as Model 2 hereafter) to verify the robustness of the Model 1 results.

6

HEALTH CAPACITY TO WORK AND ITS LONG-TERM …

139

Then, this study expanded the CMR model analyses to capture three types of work status: full-time work, part-time work, and no work. Taking full-time work as a reference category, the multinomial logistic model was estimated in order to explain part-time work and no work for individuals in their 50s. Based on the estimated parameters and actual values of each health variables, the potential and, correspondingly, additional capacities for part-time and full-time work were projected for each age group beyond 60 years of age. This analysis focused on the estimation results obtained according to a full set of health variables, as the use of the composite health index lead to much similar results. To examine how health capacity has changed over the past 30 years, the estimations were repeated with dichotomized work status (“work” and “no work”), using the data of the 1986 and 2001 CSLCs. It should be noted that the health variables used in the regression analysis are somewhat different from those used in the 2016 survey; selected diagnosed diseases were slightly different across the three surveys, K6 scores were not collected in the 1986 and 2001 surveys, and the “smoking” variable was not included in the 1986 survey. Therefore, caution should be exercised in comparing the estimation results of the three surveys. Based on the estimation results of the three surveys, the change in the additional capacity over the past 30 years was decomposed. The change in the additional health capacity to work between two years is equal to the difference between the change in the potential capacity and the change in the actual employment rate, given its definition. By applying the so-called Blinder-Oaxaca decomposition technique, the change in the potential health capacity was further decomposed into two components: (i) the component due to the change in the mean health status in each age group and (ii) the component due to the change in the mean behavior of those aged 50–59 years.3 This decomposition used Model 2 results, as this model focuses on two health variables (the composite health index and life expectancy), which were commonly used in regression models in the three surveys.4 6.3.2

Key Assumptions

This analytic strategy is based on some key assumptions. First, while focusing exclusively on the relationship between the health and work statuses, the choice of work status is also likely to be affected by other factors. Notably, the women’s LFP must be affected by their duties

140

T. OSHIO

regarding housekeeping and care of parents, children, or grandchildren. In addition, lump-sum retirement allowances and private (corporate and/or personal) pension benefits, as well as accumulated financial assets, can encourage retirement. The CRM model, which assumes that nonhealth determinants of work status remain intact, ignores any impact of these factors. Second, the CRM model also assumes that the relationship between health and work statuses remains intact in the 50s and beyond. The possibility that the elderly’s decision to work may become more or less sensitive to health status with age should not be ruled out. Considering that simulations are based on estimation using the data of those aged 50–59 years, more caution should be taken in interpreting the simulation results for those aged 70–74 years than the results for those aged 60–64 and 65–69 years. Third, the institutional settings for those aged 50–59 years are different across the three surveys. Notably, the Act on Stabilization of Employment of Elderly Persons forced companies to raise the mandatory retirement age to 60 years as of 1998, meaning that the LFP of respondents of the 1986 survey may have been affected by mandatory retirement. For women, the eligibility age to claim public pension benefits (both flatrate and wage-proportional benefits) was below 60 years for a substantial portion of female respondents aged 50–59 years in the 1986 and 2001 surveys. Finally, the proportion of self-employed elderly has substantially declined over the past three decades.5 It is likely that this change in the labor force has made the elderly’s decision on work more sensitive to mandatory retirement and eligibility for public pension benefits. However, it should be noted that we assume that when calculating the potential and additional capacities in each year any non-health factor remained unchanged from the 50s. This study exclusively examined how individuals aged 50–59 years would behave if their health statuses were replaced by those of older individuals with other characteristics in common. The benchmark for calculating the potential and additional capacities is the actual employment rate for those aged 50–59 years in each year. Therefore, the abovementioned institutional settings do not matter seriously in interpreting the estimated values of the potential additional capacities. In addition, using data of those aged 50–59 years as a common benchmark is helpful to compare the results across the three surveys.

6

HEALTH CAPACITY TO WORK AND ITS LONG-TERM …

6.4 6.4.1

141

Results

Changes in Health and Work Statuses Over the Past 30 Years

Table 6.1 summarizes the proportions of respondents who were working (on paid jobs), having poor SRH (self-rating health as poor or somewhat poor), and having at least one functional disability, as well as the average life expectancy in each age group of each survey. As shown in this table, for both men and women, there have been clear trends of improvement in all health variables among those aged 65–69 and 70–74 years. The improvement trend has been more mixed among younger age groups, who faced a modest deterioration after 2001. In addition, the proportion of those working presents different directions between men and women. While the female employment rate has been consistently increasing among all age groups, men show mixed trends; although men aged 50–54, 55– 59, and 70–74 years show consistent downtrends, men aged 60–64 and 65–69 years show some recoveries after 2001.6 6.4.2

Health Capacity to Work in 2016

To calculate the health capacity to work, this study applied the CMR model analysis to the 2016 CSLC data. As the first step, two types of linear probability models were estimated (Models 1 and 2) to explain the binary variable of “no work” according to a set of health variables. Table 6.A1 in Appendix presents the results of these models for both men and women. Model 1 results show that lower health statuses, including poorer SRH, functional disabilities, psychological distress (only for men), and some diagnosed diseases tend to be positively associated with the probability of no work. However, the presence of some diagnosed diseases, such as hypertension and smoking (only for men), have negative associations with no work. Meanwhile, higher life expectancy, which indicates better overall health status, is negatively associated with no work. The Model 2 results revealed that both the composite health index and life expectancy are negatively associated with work. Based on these regression results, Table 6.2 and Fig. 6.1 summarize the simulation results of the potential and additional work capacities in each age group for men and women. Figure 6.1 graphically presents the Model 1 results. For men, the Model 1 results show that the potential capacity has declined modestly to 84.1% for those aged 70–74 years,

142

T. OSHIO

Table 6.1 Summary statistics of working and health statuses in 1986, 2001, and 2016 Age group Men 50–54

55–59

60–64

65–69

70–74

Women 50–54

55–59

60–64

65–69

70–74

Year

Work (%)

Life expectancy (years)

Poor self-rated health (%)

Functional disabilities (%)

N

1986 2001 2016 1986 2001 2016 1986 2001 2016 1986 2001 2016 1986 2001 2016

95.7 95.3 91.1 88.5 92.4 89.2 67.4 69.6 75.5 53.6 51.9 52.5 39.7 38.0 32.7

26.3 28.4 30.7 22.2 24.1 26.2 18.3 20.2 21.9 14.5 16.3 18.0 11.2 12.9 14.2

13.0 10.2 12.6 14.5 12.1 14.3 16.6 13.0 14.2 19.6 15.8 15.1 24.6 19.0 18.2

N/A 8.3 11.2 N/A 10.4 13.4 N/A 13.1 14.2 N/A 16.5 15.5 22.4 20.0 19.3

26,332 30,779 16,795 24,609 22,188 17,278 18,626 21,026 19,460 13,155 20,393 24,165 11,124 17,419 16,385

1986 2001 2016 1986 2001 2016 1986 2001 2016 1986 2001 2016 1986 2001 2016

56.3 66.8 75.9 44.4 56.9 69.0 31.9 36.3 50.7 22.3 25.3 32.2 13.4 17.6 19.1

30.9 34.4 36.3 26.3 29.7 31.7 22.0 25.4 27.0 17.8 21.0 22.6 13.7 16.9 18.1

16.5 13.0 11.9 16.6 13.3 13.0 18.7 14.3 13.2 22.3 17.1 13.6 27.0 20.7 16.0

N/A 9.8 11.3 N/A 10.9 13.0 N/A 13.1 14.2 N/A 16.1 14.3 N/A 21.0 16.7

28,094 31,343 17,946 26,293 23,022 18,468 22,906 23,375 21,114 18,150 23,520 26,301 15,162 21,074 19,092

Note 1. The proportion of respondents who reported poor and somewhat poor self-rated health 2. The proportion of respondents who reported at least one functional disability Source Author’s creation

6

HEALTH CAPACITY TO WORK AND ITS LONG-TERM …

143

Table 6.2 Estimated potential and additional capacities to work in 2016 (% of total respondents in each age group) Age group Men Actual employment rate Potential capacity Model 1 Model 2 Additional capacity Model 1 Model 2 N Women Actual employment rate Potential capacity Model 1 Model 2 Additional capacity Model 1 Model 2 N

50–59

60–64

65–69

70–74

90.2

75.5

52.5

32.7

90.2 90.2

87.8 87.5

86.2 85.8

84.1 84.0

12.4 12.0 19,460

33.7 33.3 24,165

51.3 51.3 16,385

72.4

50.7

32.2

19.1

72.4 72.4

61.3 61.4

54.3 54.4

47.3 47.2

10.5 10.6 21,114

22.1 22.1 26,301

28.2 28.1 19,092

– – 34,073

– – 36,414

Note Based on the estimation results reported in Table 6.A1 Source Author’s creation

compared to the actual employment rate of 90.2% for those aged 50– 59 years, reflecting a relatively limited deterioration in health status. Meanwhile, the actual employment rate has declined more substantially between these two age groups, reaching 32.7%. As a result, the potential capacity rose to 12.4%, 33.7%, and 51.3% for those aged 60–64, 65–69 and 70–74 years, respectively. Model 2, which replaces a set of health variables with the composite health index, obtained almost similar results. Compared to men, women’s potential capacity has declined more rapidly with age, to 72.4% for those aged 70–74 years. This is probably because women’s work status in their 50s was more sensitive to health status reflecting their more diversified lifestyles, making their potential capacity drop more rapidly in response to a decline in health status with age. Correspondingly, an increase in the additional capacity was relatively limited to 28.2%, according to the Model 1 results. Meanwhile, there is virtually no difference in the results between Models 1 and 2, as with men.

144

T. OSHIO

Fig. 6.1 Estimated potential and additional capacities to work in 2016, based on Model 1 results (Note The numbers in [ ] indicate the potential capacity to work. Source Author’s creation)

6.4.3

Part-Time Work Vs. Full-Time Work

This study then estimated the multinomial logistic models to predict no work and part-time work—with full-time work as a base outcome— through a set of health variables, using the 2016 CSLC data. The regression results for men and women are shown in Table 6.A2 in Appendix. The results are presented in terms of the relative risk ratio (RRR), along with its 95% CI, of no work and part-time work relative to full-time work. Hausman tests were conducted as they support the null hypothesis that the odds are independent of other alternatives for both men and women. As seen in this table, lower health statuses tend to be positively associated with the probabilities of retirement and part-time work. The levels of estimated RRRs, if significantly above one, tend to be higher for no work than for part-time, a result consistent with intuition. Regarding the simulation analysis based on the regression results, Table 6.3 summarizes the estimated potential and additional capacities to full-time and part-time works, respectively, for men and women. For

6

HEALTH CAPACITY TO WORK AND ITS LONG-TERM …

145

Table 6.3 Estimated capacity to full- and part-time work in 2016: multinomial logistic models (% of total respondents in each age group) Type of work

Full-time

Age group

50–59

Men (N = 34,073) Actual 83.5 employment rate Potential 83.5 capacity Additional – capacity Women (N = 36,414) Actual 42.5 employment rate Potential 42.5 capacity Additional – capacity

Part-time

60–64

65–69

70–74

50–59

60–64

58.5

30.5

16.0

6.6

16.9

22.0

16.7

78.6

74.7

69.5

6.6

9.1

11.0

13.5

18.6

41.0

47.9

–

23.4

12.3

7.7

29.9

27.3

20.0

11.4

34.5

29.2

23.9

29.9

25.7

22.4

19.0

11.1

16.9

16.2

–

2.5

7.6

– 7.4

– 1.7

65–69

– 10.0

70–74

– 1.7

Note Based on the estimation results reported in Table 6.A2 Source Author’s creation

men, the proportion of full-time workers declined sharply from 83.5% for those aged 50–59 years to 58.5% for those aged 60–64 years and remained in decline thereafter. Meanwhile, the potential capacity declined relatively modestly for full-time work, resulting in a substantial increase of the additional capacity to 41.0% and 47.9% for those aged 65–69 and 70– 74 years, respectively. Meanwhile, an actual increase in the proportion of part-time work exceeded an increase in its potential capacity. This causes the additional capacity to be negative, indicating excess of part-time work judging by health status. More women work on a part-time basis than men, and the additional full-time work capacity after 60 years of age was much smaller compared to that of men. For part-time work, the actual employment rate and potential capacity were relatively close. 6.4.4

Long-Term Changes in Health Capacity to Work

Potential and additional work capacity was estimated using the data of the 1986 and 2001 CSLC surveys and the results were compared with those

146

T. OSHIO

of the 2016 survey, focusing on the dichotomized work statuses (work and no work). The results are summarized in Table 6.4.7 Table 6.4 Estimated capacity to work in 1986, 2001, and 2016a Men Age group

50–59

Women 60–64

65–69

70–74

50–59

60–64

65–69

70–74

1986 Actual 92.2 67.4 53.6 39.7 50.5 31.9 22.3 13.4 employment rate Potential capacity Model 1 92.2 82.2 75.1 68.4 50.5 34.2 23.7 13.2 Model 2 92.2 82.2 75.7 70.1 50.5 34.4 24.1 14.4 Additional capacity Model 1 – 14.9 21.5 28.7 – 2.4 1.4 – 0.1 Model 2 – 14.8 22.1 30.4 – 2.5 1.8 1.1 N 54,387 22,906 18,150 15,162 54,387 22,906 18,150 15,162 2001 Actual 94.1 69.6 51.9 38.0 62.6 36.3 25.3 17.6 employment rate Potential capacity Model 1 94.1 89.8 86.8 83.9 62.6 48.0 38.6 29.5 Model 2 94.1 90.1 87.7 85.5 62.6 48.2 39.2 30.7 Additional capacity Model 1 – 20.2 34.9 45.9 – 11.7 13.3 11.9 Model 2 – 20.5 35.8 47.6 – 11.9 13.9 13.1 N 54,365 23,375 23,520 21,074 54,365 23,375 23,520 21,074 2016 Actual 90.2 75.5 52.5 32.7 72.4 50.7 32.2 19.1 employment rate Potential capacity Model 1 90.2 87.8 86.2 84.1 72.4 61.3 54.3 47.3 Model 2 90.2 87.5 85.8 84.0 72.4 61.4 54.4 47.2 Additional capacity Model 1 – 12.4 33.7 51.3 – 10.5 22.1 28.2 Model 2 – 12.0 33.3 51.3 – 10.6 22.1 28.1 N 36,414 21,114 26,301 19,092 36,414 21,114 26,301 19,092 Note a Figures in the table indicate the proportion (%) of the total number of respondents in each age group Source Author’s creation

6

HEALTH CAPACITY TO WORK AND ITS LONG-TERM …

147

For men, the actual employment rate declined with age in all surveys, and its level in each age group showed no substantial changes over the past 30 years. Meanwhile, a reduction of the potential capacity with age became more limited in recent years and, correspondingly, the additional work capacity of the older age groups became larger. In particular, Model 1 shows that the potential capacity for men aged 70–74 years remained high, at 84.1% in 2016 and 83.9% in 2001, compared to 68.4% in 1986. With a modest reduction in the actual employment rate throughout the past 30 years, the additional capacity of those aged 70–74 years rose from 28.7% in 1986 to 45.9% in 2001 and to 51.3% in 2016. This pattern is also observed regarding those aged 65–69 years as well. In contrast, the additional capacity for those aged 60–64 years declined somewhat from 2001 to 2016, after rising during 1986–2001. This occurred because an increase in the actual employment rate dominated an increase in the potential capacity. For women, the actual employment rate rose for each age group from 1986 to 2016. In response to a consistent increase in the actual employment rate for those aged 50–59 years, the levels of potential work capacity for the older age groups became higher in recent years. Together with improvement in health status, this raised the additional work capacity, especially compared to the results of 1986, when the additional capacity was quite limited. Moreover, as with men again, the additional capacity for those aged 60 the 64 years declined somewhat according to the 2001 and 2016 results. 6.4.5

Decomposition of the Change in Additional Health Capacity to Work

Based on the Model 2 results, the change in additional health capacity to work was decomposed, as shown in Table 6.5. First, the bottom panel presents the entire change over the past 30 years. For men, a substantial portion of the increases in the additional capacity for those aged 65–69 and 70–74 years over that period (11.2 and 20.9 percentage points) was accounted for by an increase in the potential capacity (10.1 and 14.0 percentage points). In addition, regarding the increases in the potential capacity, the contributions from the improving health status were relatively stable (3.7–4.5 percentage points) across the three age groups, while that from the behavioral changes rose with age (0.9–10.3 percentage points). Another noticeable finding is that enhanced LFP

148

T. OSHIO

Table 6.5 Decomposition of the change in the additional capacity, based on Models 2 results Men Age group Additional capacity (%) 1986 2001 2016 Change from 1986 to 2001 (% point) Due to: (+) Increase in the potential capacity Due to the change in the mean health status Due to the change in the mean behaviors in those aged 50–59 years (−) Increase in the actual employment rate Change from 2001 to 2016 (% point) Due to: (+) Increase in the potential capacity Due to the change in the mean health status Due to the change in the mean behavior in those aged 50–59 years (−) Increase in the actual employment rate Change from 1986 to 2016 (% point) Due to: (+) Increase in potential capacity Due to the change in the mean health status Due to the change in the mean behaviors in those aged 50–59 years (−) Increase in the actual employment rate Source Author’s creation

Women

60–64

65–69

70–74

60–64

65–69

70–74

14.8 20.5 12.0 5.7

22.1 35.8 33.3 13.6

30.4 47.6 51.3 17.1

2.5 11.9 10.6 9.4

1.8 13.9 22.1 12.1

1.1 13.1 28.1 12.0

7.9 2.6

12.0 2.4

15.5 2.2

13.8 8.0

15.1 7.5

16.3 7.1

5.4

9.6

13.3

5.9

7.6

9.2

2.2

– 1.7

– 1.7

4.4

3.0

4.2

– 8.5

– 2.5

3.7

– 1.3

8.2

15.0

– 2.6 1.0

– 1.9 1.1

– 1.5 0.9

13.1 3.3

15.1 3.3

16.5 2.8

– 3.6

– 3.0

– 2.4

9.9

11.8

13.7

5.9

0.6

– 5.2

14.4

6.9

1.5

– 2.8

11.2

20.9

8.1

20.3

27.0

5.4 4.5

10.1 4.3

14.0 3.7

27.0 10.8

30.2 10.5

32.7 9.5

0.9

5.8

10.3

16.2

19.7

23.2

8.1

– 1.1

– 6.9

18.9

9.9

5.7

6

HEALTH CAPACITY TO WORK AND ITS LONG-TERM …

149

reduced the additional capacity for those aged 60–64 years (minus 2.8 percentage points); this change occurred largely in 2001 and 2016. The contribution from an increase in the potential capacity was more substantial for women than for men. Improvement in health status made large and stable contributions (9.5–10.8 percentage points) across the three age groups, while their magnitudes were exceeded by those of the contributions from the behavioral changes (16.2–23.2 percentage points). In addition, an increase in the actual employment rate, especially among those aged 60–64 years, consistently contained an increase in the additional capacity. 6.4.6

Distribution of Estimated Probabilities of Work

Lastly, this study examined the distribution of estimated probabilities of work, based on the Model 1 results, considering that the estimated work capacity ignores the heterogeneity among the elderly. Table 6.6 summarizes the results. As seen in Table 6.6, a larger proportion of individuals tended to have higher probabilities of work in recent years. About 90% or more of the elderly men were expected to work with a probability of 75% Table 6.6 Distribution of the estimated probabilities of work, based on Models 1 results (%) Year

Probability (%) Men

1986 75–100 50–75 25–50 0–25 2001 75–100 50–75 25–50 0–25 2016 75–100 50–75 25–50 0–25 Source Author’s creation

Women

50–59

60–64

65–69

70–74

50–59

60–64

65–69

70–74

94.8 4.8 0.4 0.0 98.2 1.6 0.2 0.0 93.3 6.5 0.2 0.0

88.7 8.6 2.6 0.1 96.1 3.5 0.4 0.0 91.8 7.9 0.3 0.0

72.2 21.5 5.9 0.5 92.9 6.4 0.6 0.0 90.5 9.2 0.3 0.0

32.5 55.4 11.7 0.5 88.5 10.2 1.2 0.1 86.4 12.9 0.6 0.0

0.0 57.0 41.6 1.4 1.1 93.3 5.1 0.5 37.4 61.2 1.3 0.0

0.0 0.0 90.2 9.8 0.0 47.4 50.1 2.4 0.2 94.9 4.7 0.2

0.0 0.0 49.9 50.1 0.0 1.5 92.0 6.6 0.0 84.2 15.3 0.4

0.0 0.0 3.3 96.7 0.0 0.0 80.0 20.0 0.0 35.5 63.0 1.4

150

T. OSHIO

or more in 2001 and 2016, while the proportion of those with a probability of 25% or below has become negligible in recent years. However, a non-negligible portion (6.7% for those aged 50–59 years to 13.6% for those aged 70–74 years) of men were expected to work with a probability between 25% and 75% even in 2016. The same applies in large part to women, although their probabilities of work have been distributed in a much smaller range compared to men, probably reflecting their more important role in family life. Their probability of work has consistently increased over the past 30 years, but a dominant proportion of elderly women were expected to work with a probability below 75%, even in 2016.

6.5

Discussion

This study examined the health capacity to work of the elderly—that is, how much longer the elderly can work judging by their health status— by applying the CMR model to data collected from the CSLC. The key findings and their implications are summarized as follows. First, a substantial slack of work capacity was observed among the Japanese elderly. Assuming that the relationship between the health and work statuses of those aged 50–59 years remains intact over 60 years of age, this study’s simulation results using the data from the 2016 CSLC suggest that additional 33.7% of men and 22.1% of women in their late 60s can work. The additional work capacity for the elderly in their early 70s is even larger: 51.3% of men and 28.2% of women. Taking these estimation results together, the additional capacity for those aged 60–74 are estimated to be approximately 6,650,000, roughly equivalent to 10% of total labor force in 2016, suggesting a potentially substantial impact of enhancing the elderly’ LFP. Such a large additional work capacity is explained by a substantial reduction in the actual employment rate regarding people beyond 60 years of age, compared to a much more limited deterioration in health status. The former change seems to be largely due to public pension benefits, to which individuals become eligible to claim after the age of 60, as well as mandatory retirement. This reasoning is in line with many previous studies that have demonstrated the negative impact of public pension benefits on the elderly’s labor supply (e.g., Gruber and Wise 1999). As mentioned at the end of the Analytic Strategy section, however, it should be noted that this CRM model ignores the potential impact of

6

HEALTH CAPACITY TO WORK AND ITS LONG-TERM …

151

non-health factors on an individual’s decision to work. In addition, the model assumes that the relationship between the health and work statuses remains intact in the 50s and thereafter. Therefore, caution should be exercised in interpreting the simulation results, especially regarding those aged 70–74 years. It should be also noted that this study disregards the potential impact of labor productivity among the elderly; its improvement would effectively add to labor supply even without an actual increase in the elderly LFP. Second, this study’s simulation results about the three types of work statuses suggest that there is some room for a shift to full-time work not only from retirement but also from part-time work. Additional capacity to part-time work is considered negative for men in all age groups, indicating that there is an excess of part-time workers among elderly men, judging by their health status. Meanwhile, there is a substantial slack of full-time work among the elderly. One of the most plausible reasons for this difference is again the public pension programs. The earnings-tested pension benefits program (Zaishoku Rorei Nenkin) may induce pensioners to work on a part-time basis and keep their income low enough to prevent their pension benefits from reducing (Shimizutani and Oshio 2013), regardless of their health status. A reduction or abolishment of the earnings-tested pension benefits programs is expected to encourage a shift from part-time and no work to full-time work, which in general is likely to reduce the overall additional capacity. Third, when comparing the simulation results of the 1989, 2001, and 2016 surveys, an increasing trend of the additional work capacity for both men and women was observed, even recognizing that caution should be taken in interpreting the results of different surveys. The labor supply of elderly men has been reducing (albeit modest recovery in most recent years) despite an increase in the potential capacity, and the labor supply of elderly women has not been increasing as fast as the potential capacity. An increase in the potential capacity is attributable to two factors, as indicated by the results of the Blinder-Oaxaca decomposition analysis. The first is a general trend of improvement in health status, which has been consistently reducing health-related constraints on work. The second is behavioral changes in the age of 50. The elderly’s decisions on work have become less sensitive to health status in recent years. This change is more notable among women, probably because the longterm uptrend in the women’s LFP has been reducing the negative

152

T. OSHIO

age gradient of the employment rate and, correspondingly, reducing a negative association between the work and health statuses. Other factors explaining the increase in the additional work capacity over the past 30 years are the reduction in the actual employment rate among men and the slower increase in relative potential capacity among women. These changes are probably related to a wider coverage of public pension programs and their increasing benefits. In particular, an increase in EPI beneficiaries, which covers those who have been working as private sector employees, has been strengthening the link between work status and eligibility to claim pension benefits. This trend was probably amplified by a structural change in the labor market, away from self-employment. It should be noted, however, that the actual employment rate among men aged 60–64 years has been accelerating since 2001, and the pace of increase among women in the same age group has been well above the older groups since that year. A gradual increase in the eligibility age for public pension benefits, as well as an increase in the mandatory retirement age, seems to have begun to encourage workers to stay in the labor market, eventually leading to a reduction in the additional capacity since 2001 among both men and women aged 60–64 years. Finally, caution should be exercised in interpreting the results from the viewpoint of social policies for the elderly. It is worth recalling that this study addressed the work capacity of the elderly overall, not that of specific individuals. Even if the overall health of the population is improving, there will always be individuals that are too sick to work. Indeed, our analysis of the distribution of the estimated probabilities of work reveals that a non-negligible proportion of men are expected to work only with a probability between 25% and 75%, probably due to health status, even in 2016.

6.6

Conclusions

This study has posed the following question: if older individuals with a given health status worked as much as their younger counterparts, how longer could they work? After evaluating the capacity to work solely on the basis of health, our simulation results revealed that the employment rates of men and women in their late 60s could be more than 30 and 20 percentage points higher, respectively. This study also found that there is some room for the male elderly to shift from part-time work to fulltime work, suggesting there is an excess of part-time work according to

6

HEALTH CAPACITY TO WORK AND ITS LONG-TERM …

153

a health-based evaluation. This study further observed that the elderly’s additional work capacity has increased over the past 30 years, along with the improvement of health status, although health conditions still prevent some individuals from working. Although this study does not address that aspects other than health status may affect the elderly’s ability to work longer, we believe that our simulation results provide new insights into the policy debates about public pension reform. The exact values estimated for potential increases in employment rates are less important than the overall conclusion that the elderly are healthy enough, at least on average, to work longer. The results also underscore that policy measures can utilize the extra work capacity, a part of which seems to have been created by public pension programs and other institutional factors.

Appendix Table 6.A1 Regression results of Models 1 and 2 for those aged 50–59 years in 2016 (dependent variable = no work) Men ( N = 34,073) Coef. Model 1 Self-rated health (ref = 1 [good]) 2 (somewhat good) −0.005 3 (average) 0.004 4 (somewhat poor) 0.032*** 5 (poor) 0.127*** Unanswered 0.003 Diagnosed diseases Diabetes 0.007 Obesity 0.038* Hyperlipidemia −0.014* Thyroid gland −0.009 Psychological distress 0.161*** Dementia −0.030 Parkinson’s disease −0.084 Other neurotic diseases 0.070***

Women ( N = 36,414) (SE)

Coef.

(SE)

(0.005) (0.005) (0.007) (0.014) (0.055)

−0.017* −0.002 0.013 0.062** −0.012

(0.008) (0.007) (0.010) (0.023) (0.036)

(0.006) (0.018) (0.007) (0.018) (0.012) (0.023) (0.044) (0.019)

0.059*** 0.007 −0.001 0.001 0.182*** −0.066* −0.051 0.126***

(0.014) (0.038) (0.011) (0.018) (0.016) (0.029) (0.062) (0.028)

(continued)

154

T. OSHIO

Table 6.A1

(continued) Men ( N = 34,073)

Eye diseases Ear diseases Hypertension Stroke Angina pectoris Other cardiovascular diseases Cold Allergic rhinitis COPD Asthma Other respiratory diseases Gastroduodenal diseases Liver and gall bladder diseases Other digestive diseases Dental diseases Atopic dermatitis Other skin diseases Gout Rheumatoid arthritis Arthropathy Stiff shoulder Backache Osteoporosis Kidney diseases Benign prostatic hyperplasia Menopausal disorders Fracture Other injuries Blood diseases Cancer Pregnancy

Women ( N = 36,414)

Coef.

(SE)

Coef.

(SE)

0.017 −0.027 −0.012** 0.079*** 0.013 0.034**

(0.008) (0.017) (0.004) (0.014) (0.011) (0.012)

−0.011 −0.002 −0.028*** 0.061* −0.027 −0.023

(0.012) (0.027) (0.008) (0.028) (0.026) (0.023)

0.020 −0.012 0.093* 0.014 −0.017

(0.036) (0.013) (0.044) (0.016) (0.015)

−0.038 0.013 0.131 −0.004 0.007

(0.046) (0.016) (0.112) (0.023) (0.030)

−0.007 0.038*

(0.012) (0.015)

0.016 0.027

(0.020) (0.027)

0.022 −0.009 0.003 −0.012 −0.016 −0.007 0.015 0.007 −0.006 0.121*** 0.058*** −0.057***

(0.014) (0.007) (0.024) (0.013) (0.010) (0.022) (0.012) (0.011) (0.008) (0.015) (0.015) (0.017)

0.025 0.013 −0.018 −0.009 0.010 0.066** −0.023 −0.040** −0.032** −0.054** 0.073*

(0.022) (0.011) (0.024) (0.018) (0.046) (0.025) (0.016) (0.014) (0.013) (0.019) (0.030)

0.017 −0.054* 0.063** 0.005

(0.021) (0.022) (0.023) (0.019)

0.029 0.000 −0.026 −0.030 0.061** 0.047

(0.025) (0.030) (0.029) (0.030) (0.023) (0.255)

(continued)

6

Table 6.A1

HEALTH CAPACITY TO WORK AND ITS LONG-TERM …

155

(continued) Men ( N = 34,073) Coef.

Infertility Other 0.014 Unknown 0.102* Unanswered 0.076** Functional disabilities Activities of daily living 0.043*** Going out 0.094*** Work, housekeeping, 0.047*** and study Exercise and sports −0.001 Other 0.091*** Unanswered 0.135** Psychological distress (ref = K6 score 0–4) K6 score 5–12 0.010* K6 score 13–24 0.052*** In hospital 0.188*** Smoking −0.023*** Life expectancy −0.004*** 0.0592 Adjusted R 2 Model 2 Composite health index −0.094*** Life expectancy −0.365*** (year/100) 0.0088 Adjusted R 2 Note *** p < 0.001, ** p < 0.01, * p < 0 Source Author’s creation

Women ( N = 36,414) (SE)

(0.011) (0.043) (0.025)

Coef. 0.270 0.028* −0.080 −0.007

(SE) (0.313) (0.014) (0.066) (0.042)

(0.009) (0.010) (0.008)

0.009 0.071*** 0.013

(0.013) (0.015) (0.012)

(0.009) (0.013) (0.053)

0.038** 0.050** 0.038

(0.014) (0.018) (0.084)

(0.004) (0.009) (0.055) (0.005) (0.001)

0.002 0.022 0.086* −0.004 −0.016*** 0.022

(0.006) (0.012) (0.038) (0.012) (0.001)

(0.006) (0.063)

−0.054*** −1.564***

(0.008) (0.087)

0.0101

156

T. OSHIO

Table 6.A2 Regression results of multinomial logistic models for individuals aged 50–59 years in 2016 (Base outcome = Work) Dependent variable

Men ( N = 34,073)

Women ( N = 36,414)

No work

No work

RRR

Part-time work (SE)

Self-rated health (ref = 1 [good]) 2 (somewhat 0.93 (0.07) good) 3 (average) 1.10 (0.07) 4 (somewhat 1.49*** (0.12) poor) 5 (poor) 2.43*** (0.31) Unanswered 1.06 (0.78) Diagnosed diseases Diabetes 1.09 (0.08) Obesity 1.46* (0.28) Hyperlipidemia 0.81* (0.07) Thyroid gland 0.89 (0.18) Psychological 3.38*** (0.34) distress Dementia 0.78 (0.17) Parkinson’s 0.52 (0.23) disease Other neurotic 1.85*** (0.31) diseases (0.11) Eye diseases 1.24* Ear diseases 0.83 (0.17) Hypertension 0.86** (0.05) Stroke 2.05*** (0.27) Angina pectoris 1.15 (0.14) 1.33* Other (0.16) cardiovascular diseases Cold 1.30 (0.50) Allergic rhinitis 0.85 (0.13) COPD 2.02 (0.77) Asthma 1.16 (0.20) 0.86 (0.15) Other respiratory diseases Gastroduodenal 0.92 (0.12) diseases ** Liver and gall 1.49 (0.22) bladder diseases

RRR

(SE)

RRR

Part-time work (SE)

RRR

(SE)

0.86

(0.07)

0.94

(0.04)

1.06

(0.04)

1.06 1.11

(0.07) (0.10)

0.99 1.02

(0.04) (0.06)

0.99 0.91

(0.03) (0.05)

1.53** 1.22

(0.25) (0.91)

1.35* 0.76

(0.17) (0.15)

1.06 0.54**

(0.14) (0.11)

1.06 1.09 0.75** 0.94 2.45***

(0.09) (0.28) (0.08) (0.23) (0.32)

1.32*** 1.23 0.95 1.09 2.73***

(0.10) (0.26) (0.06) (0.11) (0.24)

1.00 1.45 0.90 1.20 1.56***

(0.08) (0.30) (0.06) (0.12) (0.16)

0.74 0.51

(0.22) (0.31)

0.77 0.73

(0.13) (0.25)

1.17 0.82

(0.18) (0.29)

1.89**

(0.38)

1.86*** (0.28)

1.13

(0.19)

0.87*

1.37** 1.74** 0.84** 1.77*** 1.05 0.91

(0.14) (0.32) (0.05) (0.29) (0.16) (0.15)

0.90 1.07 0.82*** 1.27 0.87 0.89

(0.06) (0.16) (0.04) (0.19) (0.13) (0.11)

1.18 0.87** 0.88 0.97 1.00

(0.06) (0.17) (0.04) (0.15) (0.14) (0.12)

1.65 0.85 0.97 1.07 1.12

(0.68) (0.15) (0.59) (0.23) (0.21)

0.83 1.18 1.63 1.08 1.05

(0.22) (0.11) (0.96) (0.14) (0.18)

1.01 1.24* 0.83 1.23 1.03

(0.25) (0.11) (0.61) (0.15) (0.17)

0.93

(0.15)

1.00

(0.11)

0.81

(0.09)

1.44*

(0.26)

1.08

(0.16)

0.87

(0.14)

(continued)

6

Table 6.A2 Dependent variable

HEALTH CAPACITY TO WORK AND ITS LONG-TERM …

157

(continued) Men ( N = 34,073)

Women ( N = 36,414)

No work

No work

RRR Other digestive 1.27 diseases Dental diseases 0.90 Atopic 1.03 dermatitis Other skin 0.89 diseases Gout 0.78 Rheumatoid 1.00 arthritis Arthropathy 1.13 Stiff shoulder 1.09 Backache 0.94 Osteoporosis 2.43*** Kidney diseases 1.59*** 0.49** Benign prostatic hyperplasia Menopausal disorders Fracture 1.16 Other injuries 0.58 Blood diseases 1.54* Cancer 1.12 Pregnancy Infertility Other 1.22 Unknown 1.60 Unanswered 2.16*** Functional disabilities Everyday 1.42*** activities Going out 1.79*** 1.53*** Work, housekeeping, and study Exercise and 1.02 sports Other 2.01*** Unanswered 2.90* Psychological distress (ref =

Part-time work (SE)

RRR

(SE)

RRR

Part-time work (SE)

RRR

(SE)

(0.18)

1.02

(0.19)

1.19

(0.14)

1.12

(0.14)

(0.08) (0.29)

0.95 0.78

(0.10) (0.29)

1.09 0.96

(0.06) (0.13)

1.04 1.13

(0.06) (0.15)

(0.13)

0.99

(0.17)

0.95

(0.09)

0.99

(0.09)

(0.11) (0.23)

0.83 1.28

(0.13) (0.33)

1.12 1.28

(0.29) (0.17)

1.16 0.87

(0.29) (0.13)

(0.14) (0.13) (0.08) (0.32) (0.22) (0.11)

0.90 1.14 0.93 1.55* 0.95 0.75

(0.14) (0.16) (0.10) (0.29) (0.20) (0.18)

0.92 0.78** 0.83** 0.79* 1.41*

(0.08) (0.06) (0.06) (0.09) (0.23)

1.07 0.92 0.96 1.12 1.01

(0.09) (0.07) (0.06) (0.12) (0.18)

1.25

(0.17)

1.17

(0.16)

(0.17) (0.15) (0.15) (0.16) (2.03) (2.89) (0.09) (0.28) (0.26)

1.06 1.14 1.00 0.88 1.33 0.00 1.09 1.20 1.29

(0.18) (0.18) (0.16) (0.12) (1.89) (0.00) (0.09) (0.42) (0.29)

(0.23) (0.17) (0.33) (0.22)

1.10 1.20 0.89 1.29

(0.28) (0.31) (0.29) (0.29)

(0.14) (0.58) (0.52)

1.26 0.00 1.32

(0.18) (0.00) (0.44)

1.03 0.93 0.85 1.26 1.43 2.03 1.19* 0.73 1.08

(0.12)

1.31*

(0.14)

1.01

(0.07)

0.93

(0.07)

(0.16) (0.12)

1.19 1.67***

(0.15) (0.16)

1.40*** (0.11) 1.09 (0.07)

1.04 1.07

(0.09) (0.07)

(0.09)

1.15

(0.12)

1.17*

(0.09)

0.96

(0.08)

(0.23) (1.14)

1.14 1.06

(0.11) (0.46)

0.78* 0.69

(0.08) (0.34)

(0.22) 1.56** (1.29) 1.80 K6 score 0–4)

(continued)

158

T. OSHIO

Table 6.A2

(continued)

Dependent variable

K6 score 5–12 K6 score 13–24 Hospitalized Smoking Life expectancy Pseudo R 2

Men ( N = 34,073)

Women ( N = 36,414)

No work

No work

Part-time work

RRR

(SE)

RRR

1.14** 1.58*** 4.89* 0.72*** 0.95 0.0520

(0.05) (0.14) (3.62) (0.05) (0.01)

1.03 1.25 1.94 0.92 0.94***

(SE) (0.06) (0.15) (1.45) (0.07) (0.01)

Part-time work

RRR

(SE)

RRR

(SE)

0.99 1.12 1.85** 0.94 0.92 0.0120

(0.03) (0.07) (0.37) (0.06) (0.00)

0.96 1.02 1.77** 0.91 0.99

(0.03) (0.07) (0.39) (0.06) (0.00)

Note *** p < 0.001, ** p < 0.01, * p < 0.05 Source Author’s creation

Notes 1. Matsukura et al. (2018) applied the CMR model to compute an untapped work capacity using JSTAR and simulate the impact of the use of the untapped work capacity upon potential economic growth using the National Transfer Account framework. They insist that the augmented effect of the economic support ratio upon potential economic growth is substantial in the long term, generating a sizable “silver dividend” in Japan. 2. Marital status, family relationships, or other socio-demographic/economic variables were not included as covariates, because their inclusion will make it difficult to distinguish the impact of the change in health status on work status from that of other factors. Educational attainment, which is a fixed attribute for the elderly, is available in the 2016 CSLC. However, we do not use it either, because it is not included in the 1986 or 2001 CSLC, which makes it difficult to consistently compare the results across three years. We also find that the estimation results remain almost intact after including educational attainment as a covariate using the 2016 survey. 3. Based on the Blinder-Oaxaca decomposition, this study allocated the interaction effects of two components into these components by taking an average of the mean of the health variable ( X¯ ) and regression coefficient (β) between two points in time is  the decomposition   (0 and 1). Hence, calculated as (i) (β1 + β0 )/2 × X¯ 1 − X¯ 0 and (ii) X¯ 0 + X¯ 1 /2 × (β1 − β0 ). 4. The Blinder-Oaxaca decomposition cannot be applied to Model 1 as this model comprises different sets of health variables across the three surveys. 5. The proportion of the self-employed from the total population, for those aged 50–59 years in the sample used in this study in 1986, 2001, and 2016, is 31.1%, 21.7%, and 13.4%, respectively, for men and 7.2%, 6.0%, and 4.7%, respectively, for women.

6

HEALTH CAPACITY TO WORK AND ITS LONG-TERM …

159

6. Oshio et al. (2018) reveal a U-shaped recovery of elderly males’ LFP in the 2000s in Japan, somewhat later than in the 1990s, as observed in many other advanced countries. 7. The estimation results are based on the Models 1 and 2 results in each survey year. The entire set of these results are not reported to conserve space, but are available upon request.

References Cutler, D. M., Meara, E., & Richards-Shubik, S. (2012). Health and work capacity of older adults: Estimates and implications for social security policy. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.2577858. Gruber, J., & Wise, D. A. (Eds.). (1999). Social security and retirement around the world. Chicago: The University of Chicago Press. Kessler, R. C., Andrews, G., Colpe, L. J., Hiripi, E., Mroczek, D. K., Normand, S. L., et al. (2002). Short screening scales to monitor population coverages and trends in non-specific psychological distress. Psychological Medicine, 32, 959–976. Kessler, R. C., Green, J. G., Gruber, M. J., Sampson, N. A., Bromet, E., Cuitan, M., et al. (2010). Screening for serious mental illness in the general population with the K6 screening scale: Results from the WHO World Mental Health (WMH) Survey Initiative. International Journal of Methods in Psychiatric Research, 19, 4–22. Matsukura, R., Shimizutani, S., Mitsuyama, N., Lee, S., & Ogawa, N. (2018). Untapped work capacity among old persons and their potential contributions to the “silver dividend” in Japan. Journal of the Economics of Ageing, 12, 236–249. OECD. (2017). Pensions at a Glance 2017: OECD and G20 Indicators. Paris: OECD Publishing. Oshio, T., Oishi, A., & Shimizutani, S. (2011). Social security reforms and labor force participation of the elderly in Japan. Japanese Economic Review, 62, 248– 271. Oshio, T., Usui, E., & Shimizutani, S. (2018). Labor force participation of the elderly in Japan (NBER Working Paper No. 24614). Poterba, J., Venti, S., & Wise, D. A. (2013). Health, education, and the postretirement evolution of household assets. Journal of Human Capital, 7, 297– 339. Sakurai, K., Nishi, A., Kondo, K., Yanagida, K., & Kawakami, N. (2011). Screening performance of K6/K10 and other screening instruments for mood and anxiety disorders in Japan. Psychiatry and Clinical Neurosciences, 65, 434–441.

160

T. OSHIO

Shimizutani, S. (2011). A new anatomy of the retirement process in Japan. Japan and the World Economy, 23, 141–152. Shimizutani, S., & Oshio, T. (2010). New evidence on the initial transition from career job to retirement in Japan. Industrial Relations, 49, 248–274. Shimizutani, S., & Oshio, T. (2013). Revisiting the labor supply effect of social security earnings test: New evidence from its elimination and reinstatement in Japan. Japan and the World Economy, 28, 99–111. Usui, E., Shimizutani, S., & Oshio, T. (2017). Health capacity to work at older ages: Evidence from Japan. In D. Wise (Ed.), Social security programs and retirement around the world: The capacity to work at older ages (pp. 219–241). Chicago: The University of Chicago Press. Wise, D. (Ed.). (2017). Social security programs and retirement around the world: The capacity to work at older ages. Chicago: The University of Chicago Press.

CHAPTER 7

Willingness to Continue Volunteering Among the Middle-Aged and Older Adults in Japan Xinxin Ma

7.1

Introduction

In Japan, as the fertility rate declines and the population progressively ages, problems such as labor force shortages and a burdened government pension fund are becoming severe. However, from a volunteering perspective, because the older adults can receive public pension benefits and have more leisure time, it is thought that unlike the young and middle-aged adults, the older adults are more likely to participate in social activities, including volunteering. From a broader perspective, in order to establish an age-free society, it is important to promote the participation

This chapter is a revised version of: Ma, X. (2017). The determinants of willingness to continue to volunteering among middle-aged and older adults in Japan. In The Japan Institute for Labour Policy and Training (JILPT) (Ed.), Employment of older adults in population decrease society (pp. 230–270). JILPT, Tokyo. (In Japanese). X. Ma (B) Faculty of Economics, Hosei University, Tokyo, Japan e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 X. Ma (ed.), Employment, Retirement and Lifestyle in Aging East Asia, Social Policy and Development Studies in East Asia, https://doi.org/10.1007/978-981-16-0554-3_7

161

162

X. MA

of older adults in various social activities, including volunteering, as well as participation in work. This chapter investigates the determinants of willingness to continue volunteering in Japan and compares these factors between middle-aged and older adults. Regarding the determinants of social activity participation in Japan, although previous studies have found that an individual’s characteristics, family structure, and income factors influence the probability of volunteering, these previous studies did not focus on the older adults (Atoda et al. 1999; Atoda and Fukushige 2000; Yamauchi 2001; Ono 2006; Moriyama 2007; Ma and Ono 2013; Ma 2012a, b, 2014). In addition, these studies did not analyze the determinants of a willingness to continue volunteering. Using a unique survey, the Activities and Working Status of NPOs (nonprofit organizations) Survey of individuals and NPOs conducted by the Japan Institute for Labor Policy and Training (JILPT) in 2014, we constructed an employee-employer matched dataset including individuals and NPOs to investigate the influences of four kinds of factors on the willingness to continue volunteering: (1) human capital, (2) household income, (3) activity motivation, and (4) reward factors, we compare the results among middle-aged and older adults in Japan. The remainder of this chapter is organized as follows. Section 7.2 provides a discussion of the possible channels of influence of the four factors on volunteering and summarizes the results of previous studies. Section 7.3 explains the methodology, including models and data. Section 7.4 presents the estimation results of influences of four kinds of factors on willingness to continue volunteering in Japan, and Sect. 7.5 gives the conclusions of this study.

7.2 7.2.1

Literature Review Previous Empirical Studies

Why does an individual participate in social activities such as volunteering? Psychological theories emphasize personality traits, self-concepts, and motivation (Atkins et al. 2005; Handy and Cnaan 2007; Erez et al. 2008; Einolf 2008; Finkelstein 2008; Gronlund 2011). Sociological theories focus on the individual’s social economics status (SES) and demography factors such as race, gender, social network, and community characteristics (Messner and Bozada-Dea 2009; Einolf 2010; Rotolo et al. 2010; Taniguchi 2011; Boyle and Sawyer 2010). Economic theories treat social

7

WILLINGNESS TO CONTINUE VOLUNTEERING AMONG …

163

activity participation as a form of leisure or unpaid work (Schram and Dunsing 1981; Vaillancourt 1994; Freeman 1997; Carlin 2001; Segal and Weisbrod 2002).1 Regarding empirical studies of developed countries, Schram and Dunsing (1981), Vaillancourt (1994), Freeman (1997), Carlin (2001), and Segal and Weisbrod (2002) found that income factors (wages, household income, unearned income, working hours), human capital (educational attainment, age, years of experience, occupation), individual characteristics (gender, race, years of marriage, social position, sibling order, health status), family structure (marriage status, children, occupation of householder, spouse), psychological factors (willingness to turn over, job satisfaction), and other factors (donation, housing, city size, region, tax system) influence volunteering. These findings are consistent with theories from the psychological, sociological, and economic perspectives mentioned above. Particularly, from an economic perspective, Menchik and Weisbrod (1987) advocated the consumption model and the human capital model. According to the consumption model, since volunteering is a form of leisure, a preference for leisure may become higher for an individual with a higher unearned income, which may cause a higher probability of volunteering. Based on the human capital model, participation in a social activity can increase human capital and increase the probability of getting a good job in the future; therefore, the supply of volunteers may tend more toward the younger generation than adults who are middle-aged and older. Regarding empirical studies for Japan, Atoda et al. (1999), Atoda and Fukushige (2000), Yamauchi (2001), Ono (2006), Moriyama (2007), Ma and Ono (2013), and Ma (2012a, b, 2014) found that income factors (wage rate, household income, working hours, and working days of householder), human capital factors (education, age, past volunteering experience), individual characteristics (gender), family structure (marriage status, children), NPO characteristics (system, organization characteristics, activity content), and other factors (city size, region) influence volunteering. Particularly, for willingness to volunteer, Ma (2016) investigated the influences of reward factors (wage level, wage gap, change of wage level), and found that willingness to continue participating in social activities (volunteering) is higher both for the higher wage group and the group with greater change of wage levels. Moriyama (2016) pointed out that NPO management and the perceived empathy, value, and purpose

164

X. MA

of the NPO that reflect the individual’s opinion influence the individual’s willingness to continue working with the NPO. 7.2.2

The Influence of Four Factors

Based on these previous studies, using data from a unique survey conducted in 2016 in Japan, this chapter analyzes the willingness of Japanese individuals aged 50 and older to continue volunteering (working in the NPOs). We focus on four kinds of factors: (1) human capital, (2) household income, (3) activity motivation, and (4) reward factors. First, regarding the human capital factor, Menchik and Weisbrod (1987) pointed out that the experience of volunteering can be thought as a kind of human capital that may increase the probability of getting a good job in the future. However, because the elderly will exit the labor market in the near future, the channel advocated by Menchik and Weisbrod (1987) cannot explain the volunteering of the elderly. On the contrary, we consider that the human capital accumulated in the past (acquisition of professional qualifications, past volunteer experiences) can be thought of as human capital, which may influence current volunteering. For example, it can be expected that willingness to continue working in an NPO may be higher for an individual who has volunteered in the past than for an individual who has not participated in past social activities. Education is an important form of an individual’s human capital (Becker 1964; Mincer 1974). When considering the influence of educational attainment, both the positive and negative effects of education must be examined. Concretely, it is thought that wage levels are higher for the well-educated group than for the less-educated group; when the substitution effect of rising wages is greater than the income effect,2 an individual may have a strong desire to concentrate on a job in the labor market, whereas the willingness to participate in social activities may be lower (negative effect). On the contrary, a well-educated individual with greater learning ability and potential trying to compete for a new job by participating in social activities may be more willing to continue volunteering (positive effect). Because it not clear which influence is greater, the negative effect or the positive effect, we should confirm the effect of education based on the results of empirical study. Second, regarding the income factor, according to the consumption model advocated by Menchik and Weisbrod (1987), household

7

WILLINGNESS TO CONTINUE VOLUNTEERING AMONG …

165

income positively affects social participation. Based on the partial equilibrium model, the decision to participate in work depends on the budget constraint (wage3 ) and the leisure preference. When the leisure preference and wage are constant, the total income will increase as unearned income increases, which causes a decrease in working hours and a rise of leisure hours including hours of volunteering. Third, motivations for volunteering are broadly divided into “altruism” and “egoism.” When other factors are constant, it is considered that a group with a strong altruistic spirit will actively participate in social activities and have a high willingness to volunteer. Finally, unlike the motivation to participate in work in the labor market, the main purpose of volunteering is not to obtain earned income; this analysis is not limited to unpaid volunteers, but also applies to paid volunteers and regular or non-regular employees of NPOs. Employees of NPOs, particularly regular workers in NPOs, may need to be paid to maintain their livelihoods; therefore, it is expected that the reward level in NPOs may affect people’s willingness to continue working for the NPO.

7.3

Methodology and Data 7.3.1

Model

An ordered logistic regression model is used to investigate the influence of four factors on willingness to continue volunteering. The model is expressed in Eqs. 7.1 and 7.2. R = a + γ H Hi j + γ M Mi j + γ X X i j + γz Z i j + +vi j

(7.1)

 Pr(U = m) = Pr k(m−1)i j < b + β H Hi j + β I Ii j + β M Mi j  +β R Ri j + β X X i j + εi j < k(m+1)i j

(7.2)

In Eqs. 7.1 and 7.2, U is an ordinal scale variable of willingness to continue volunteering; m denotes ordering options (four types); k expresses utility; i is the individual; j is the NPO; H , I , M, and R express the human capital factor, the income factor, the motivation factor, and the reward factor, respectively. X is other factors that may affect the willingness to continue volunteering. β is the coefficient of each variable, and ε is the error.

166

X. MA

It is thought that an endogeneity problem may remain between willingness to continue volunteering and reward. For example, the willingness to continue volunteering may be higher since the reward is higher; on the contrary, the reward is higher because the individual works hard when his (her) willingness to volunteer is higher. To address the endogeneity problem, the 2SLS model is used in this study. The imputed reward used in Eq. 7.1 is calculated from the first step estimation (reward function) expressed by Eq. 7.1.4 Start year of participation in social activity, age, and region are used as identification variables.5 To address the sample selection bias, the Tobit model is used in the reward function. The results of the reward function are shown in Table 7.5. 7.3.2

Data

This study uses the unique survey, Activities and Working Status of NPOs Survey of individuals and NPO organizations conducted by the Japan Institute for Labor Policy and Training (JILPT) in 2014. We conducted an employee-employer matched dataset including individuals and NPOs. Because a relatively large number of samples related to the disaster area were extracted in this survey, we took the weighted analysis using extracted weights from 47 districts.6 The dependent variable7 is an ordered categorical variable related to the willingness to continue volunteering. We constructed the variables as willingness to “continue in the current NPO as much as possible = 4; to continue in the current NPO for a certain period = 3; to change jobs to another NPO = 2; or I want to stop the activity = 1.” The four kinds of factors as the main independent variables and other variables are constructed as follows. First, for the human capital factor, (1) six kinds of educational attainment dummy variables (junior high school and lower, senior high school, college, university, graduate school, and other); (2) five kinds of qualification dummy variables (legal or tax, education, medical, other qualifications, and no qualifications); (3) a training dummy variable; (4) four kinds of years of work experience dummy variables (regular working years in the current NPO, non-regular working years in the current NPO, CEO working years in the current NPO, working years in another NPO, and working years in a for-profit organization); (5) three kinds of volunteering experience dummy variables (participate voluntarily, experienced in school and company initiatives, no activity); and (6) four kinds of health status

7

WILLINGNESS TO CONTINUE VOLUNTEERING AMONG …

167

dummy variables (poor, not good, good, very good) are used as the indices of human capital factors in this study. Second, unearned income is calculated as “unearned income = total household income-respondent income.” It is expected that unearned income positively affects an individual’s willingness to volunteer. Third, based on the question “What motivated you to start volunteering in the current NPO,” when the respondent selected the option “to help people and contribute to society and the community,” it can be considered that the motivation of the individual is “altruism.” It is expected that the altruism variable may positively affect willingness to volunteer. Fourth, the wages of regular employees and non-regular employees and the rewards of the volunteer are used as the reward in the analysis. Fifth, other individual-level factors—(1) occupation,8 volunteering of other family members,9 male, employment status in the current NPO,10 employment status in a for-profit organization,11 family care,12 marriage, and number of family members13 —are used. (2) NPO activity field,14 organization size,15 change in the NPO’s CEO,16 a high proportion of younger employees, a high proportion of well-educated employees, and a high proportion of male employees17 are constructed as organization level factors. (3) Ono (2006), Ma (2012a, b, 2014), and Ma and Ono (2013) found that the population size in a community influences the probability of volunteering. In addition, it can be considered that the situations of labor supply and demand and the promotion policy of participating in social activities may differ across regions. We constructed four kinds of living region dummy variables: government-designated city, city with a population of 100 thousand or more, city with a population of less than 100 thousand, and towns and villages. The descriptive statistics by different age groups (age 19 and older, age 50–59, age 60–64, and 65 and older) are summarized in Appendix Table 7.4.

168

X. MA

7.4 7.4.1

Results Basic Results

The results of using the total samples are summarized in Table 7.1. First, for the effect of human capital factors, (1) the coefficients of educational attainment level variables are not statistically significant. (2) The coefficient of qualifications of legal or tax, and education are −0.802 and − 0.340, respectively, which are all statistically significant. It suggests that these two kinds of qualifications negatively affect the probability of being willing to continue volunteering. However, the influence of medical and care qualifications is not statistically significant. (3) Training positively affects the willingness to volunteer. These results are consistent with the assumption mentioned above. Some human capital factors (qualifications) may negatively affect willingness to continue volunteering, while some human capital factors (receiving training) may have positive effects. Second, the coefficients of the household income dummy variables are not statistically significant; thus, the consumption hypothesis is not supported. Third, the coefficient of altruism is 0.646, and it is statistically significant at a 1% level. This indicates that an individual with altruism is likely to continue volunteering. Fourth, the influence of the reward level is not observed. It is indicated that the influence of reward on willingness to continue volunteering is small. 7.4.2

Results by Age Group

To compare the influences of the four factors by age group, we used the subsamples of three groups: groups aged 50–59, aged 60–64, and aged 65 and older were used for analyses. Table 7.2 reports these results. First, the influence of human capital factors is greater for the group aged 50–59 and the group aged 60–64 than for the group aged 65 and older. Concretely, for the group aged 65 and older, excepting the medical qualification that is only statistically significant at a 10% level, most human capital factors are not statistically significant. Educational attainment, qualifications, training, years of work experience, and volunteering experience significantly affect the willingness to volunteer for the group aged 50–59, and the educational attainment, qualifications, training, years of work experience, and health status significantly affect the willingness to

7

WILLINGNESS TO CONTINUE VOLUNTEERING AMONG …

169

Table 7.1 Results of willingness to continue volunteering in Japan Variables Human capital factor

Income factor

Motivation factor Reward factor

Education (University) Junior high school and lower Senior high school College Graduate school Other Qualification (Non-qualification) Legal or tax Education Medical Other types of qualification No training Experience years (Working in profit organization) Regular working years in the current NPO Non-regular working years in the current NPO CEO working years in the current NPO Working years in other NPO Activity in the past Participate nonvoluntarily No activity Health (poor) Very good Good Not good Non-earned income (1st quintile) 2nd quintile 3rd quintile 4th quintile 5th quintile Altruism Reward Reward squared Individual level variables

Coef.

z-value

0.357 −0.154 0.013 0.134 −0.238

0.63 −0.86 0.06 0.43 −0.95

−0.802** −0.340** −0.267 0.212 −0.429**

−2.17 −1.81 −1.45 1.33 −2.40

−0.016*

−1.76

0.021

1.21

−0.008

−0.73

0.011

1.40

0.108 −0.220

0.55 −1.44

0.042 −0.109 −0.637

0.08 −0.21 −1.19

−0.195 0.114 −0.014 0.065 0.646*** 0.383 −0.009 Yes

−0.94 0.60 −0.07 0.31 3.26 1.24 −0.43

(continued)

170

X. MA

Table 7.1 (continued) Variables NPO organization-level variables Regional population variables Observations Maximum log likelihood Pseudo R 2

Coef.

z-value

Yes Yes 2,313 −2520.576 0.079

Note ***p < 0.01, **p < 0.05, *p < 0.10 Source Author’s creation based on the data from 2016 Activities and Working Status in NPOs Survey

continue volunteering for the group aged 60–64. It should be noticed that the positive influence of health on the willingness to continue volunteering is only found for the group aged 60–64. This suggests that when a healthy individual aged 60–64 exits the labor market, for example, due to mandatory retirement, the probability of participating in social activities seems higher than that of individuals with poor health. In considering the findings about the influence of health on the labor force participation of the elderly, it can be said that health is a key human capital factor that can influence both the participation in the labor market and volunteering of the elderly. Second, for the group aged 50–59 and the group aged 65 and older, it is shown that the probability of willingness to volunteer is higher as the income increases. Concretely, the coefficients of income in the 3rd quintile are 1.057 for the group aged 50–59, and the coefficients of income in the 5th quintile are 1.634 for the group aged 60 and older, which are statistically significant at 5% levels. This suggests that in comparing the low-income group (income 1st quintile), the probability of being willing to continue volunteering is higher for the middle-income group (the group aged 50–59) and the high-income group (the group aged 65 and older). On the contrary, for the group aged 60–64, the coefficient of income in the 5th quintile is −2.091, which is statistically significant at a 5% level. It is shown that for the group aged 60–64, to compare with lowincome group, the probability of being willing to continue volunteering is lower for the high-income group. This indicates that the influences of income differ by age group; high household income may increase the probability of participating in social activities for the group aged 50–59

Education (University) Junior high school and lower Senior high school College Graduate school Other Qualification (Non-qualification) Legal or tax Education Medical Other types of qualification No training Experience years (Working in profit organization) Regular working years in the current NPO Non-regular working years in the current NPO CEO working years in the current NPO Working years in other NPO Volunteering experience (Participation voluntarily)

Variables −2.59 −1.23 −0.25 −0.34 −2.03 −0.84 0.10 −2.08 1.99 −1.45 −1.05 1.28 −1.62 0.35

−0.529 −0.104 −0.256 −1.081** −0.692 0.040 −0.690** 0.666** −0.670 −0.025 0.045 −0.047* 0.006

z-value

−4.262***

Coef.

Age 50–59

0.006

0.111**

0.110

0.003

−1.977***

0.253 −1.419* −1.671** −0.393

1.035 0.719 −0.384 0.326

5.988***

Coef.

Age 60–64

0.29

2.12

1.38

0.09

−3.56

0.17 −1.68 −2.39 −0.58

1.06 0.66 −0.34 0.33

7.84

z-value

Results of probabilities of willingness to continue volunteering by age in Japan

Human capital factor

Table 7.2

−0.32

−0.004

WILLINGNESS TO CONTINUE VOLUNTEERING AMONG …

(continued)

0.20

0.17

−0.40

−0.92

0.74 −0.77 −1.87 −1.18

1.43 0.64 −0.27 1.50

0.43

z-value

0.004

0.007

−0.007

−0.422

0.823 −0.342 −0.815* −0.429

0.573 0.406 −0.269 1.261

0.315

Coef.

Age 65+

7

171

−0.20 1.30 0.83 0.40

−0.064 1.356 0.809 0.417

622 −562.470 0.184

Yes

−0.73 2.23 −0.11 −0.54 1.35 1.46 −1.24

−2.24

−0.852**

−0.406 1.057** −0.043 −0.204 0.672 1.231 −0.069 Yes Yes

z-value

Coef.

Age 50–59

259 −195.841 0.407

Yes

1.750* −0.405 0.915 −2.091** 0.859 −1.315 0.193** Yes Yes

5.461*** 4.808*** 3.304*

−0.802

−0.461

Coef.

Age 60–64

Note ***p < 0.01, **p < 0.05, *p < 0.10 Source Author’s creation based on the data from 2016 Activities and Working Status in NPOs Survey

Observations Maximum log likelihood Pseudo R 2

Motivation factor Reward factor

Participate nonvoluntarily No activity Health (Poor) Very good Good not good Non-earned income (1st quintile) 2nd quintile 3rd quintile 4th quintile 5th quintile Altruism Reward Reward squared Individual-level variables NPO organization-level variables Regional population variables

Variables

(continued)

Income factor

Table 7.2

1.81 −0.49 1.29 −2.17 0.74 −1.01 2.10

2.96 3.20 1.93

−1.29

−0.56

z-value

491 −393.977 0.231

Yes

0.211 0.039 0.547 1.634** 2.1886*** 0.2513 −0.0159 Yes Yes

0.015 0.140 −0.327

−0.762

0.224

Coef.

Age 65+

0.53 0.09 1.20 2.50 3.70 0.31 −0.33

0.01 0.13 −0.29

−2.43

0.47

z-value

172 X. MA

7

WILLINGNESS TO CONTINUE VOLUNTEERING AMONG …

173

and the group aged 65 and older, which is consistent with the consumption hypothesis advocated by Menchik and Weisbrod (1987); however, the consumption hypothesis is not supported in the group aged 60–64. Third, for the group aged 65 and older, the coefficient of altruism is 2.189, and it is statistically significant at a 1% level, whereas they are not statistically significant for the groups aged 50–59 and 60–64. This suggests that the influence of altruism motivation on social activity participation becomes greater with increasing age, particularly for the group aged 65 and older. Fourth, for the group aged 60–64, the coefficient of reward squared is 0.193, and it is statistically significant at a 5% level, which suggests that when the reward exceeds a certain level, the willingness to continue volunteering increases as the reward rises. However, the influences of reward are not statistically significant for both groups aged 50–59 and 65 and older. It is indicated that the influence of reward differs by age group.

7.5

Conclusions

Using the data from a unique survey—Activities and Working Status of NPOs Survey conducted by the Japan Institute for Labor Policy and Training (JILPT) in 2014, we conducted an employee-employer matched dataset including individuals and NPOs to investigate the influence of four factors (human capital, household income, activity motivation, and reward) on willingness to continue volunteering. The main findings are as follows (see Table 7.3). First, in general, the human capital factor and motivation factor significantly affect willingness to continue volunteering, but the influences Table 7.3 Summary of results

Human capital factor Income factor Motivation factor Reward factor

Total (19+)

Age 50–59

Age 60–64

Age 65+

◯ × ◯ ×

◯ ◯ × ×

◯ × × ◯

× ◯ ◯ ×

Note ◯: statistically significant; ×: statistically insignificant Source Author’s creation

174

X. MA

of household income and reward from the NPO are not statistically significant. Second, the influences of four factors on willingness to continue volunteering differ by group. Concretely, the influences of human capital and income factors are confirmed for the group aged 50–59, and the influences of human capital and reward factors are significant for the group aged 60–64, while the influences of income and motivation factors are confirmed for the group aged 65 and older. To consider the policy implications, first, the positive influence of altruism motivation on willingness to continue volunteering is shown to become greater as age increased for the group aged 65 and older, but its influence is small for the group younger than 65. This suggests that other factors, except the motivation factor, may greatly influence volunteering. It is necessary to educate middle-aged individuals to promote participation in social activities, as well as to implement policies that may promote participation or continued social activity, such as a system of flexible working hours and leave for volunteering. Second, for the middle-aged group (the group aged 50–59) and the elderly group (the group aged 65 and older), the influence of household income is greater; it is clear that the probability of participating in social activity is greater for middle-income and high-income groups than for the low-income group. To establish an active society to address the aging of the population for the low-income group and implement income-redistribution policies, such as a minimum income protection policy, it will also be necessary to consider policies that support and promote volunteering, as social participation, including volunteering, is expected to improve health and increase the social capital of the elderly (Kanamori et al. 2014; Amagasa et al. 2017; Oshio and Kan 2019; Ma et al. 2020).

Appendix See Tables 7.4 and 7.5.

Variables

Descriptive statistics of variables

【Individual level variables】 】 Education Human capital factor Junior high school and lower Senior high school College University Graduate school Other Qualification Law or tax Education Medical Other types of qualification Non-qualification No training Age Experience years Regular working years in the current NPO Non-regular working years in the current NPO CEO working years in the current NPO Working years in other NPO Working in profit organization

Table 7.4

10.9% 45.8% 33.3% 49.4% 20.9% 29.5% 15.1% 37.2% 46.8% 42.9% 11.3% 37.0% 13 14 5 8 0 15

2.3% 16.5% 32.3% 75.7% 1.3% 16.3% 53 17 5 5 0 24

S.D.

1.2% 29.9% 12.7% 41.9% 4.6% 9.7%

Mean

Total (19+)

0 24

6

5

16

2.5% 20.4% 33.1% 76.6% 1.6% 13.0% 55

0.6% 30.0% 15.9% 38.5% 3.9% 11.1%

Mean

Age 50–59

0 13

7

6

12

15.8% 40.3% 47.1% 42.4% 12.7% 33.7% 3

7.7% 45.9% 36.6% 48.7% 19.4% 31.4%

S.D.

0 33

6

5

25

2.5% 15.5% 28.2% 75.0% 1.3% 16.5% 64

1.6% 32.3% 9.6% 46.6% 4.1% 5.8%

Mean

Age 60–64

0 14

9

6

14

15.5% 36.2% 45.0% 43.4% 11.4% 37.1% 2

12.7% 46.8% 29.4% 49.9% 19.9% 23.3%

S.D.

0 16

11

5

15

15.4% 34.5% 39.9% 46.2% 0.0% 29.3% 3

18.7% 48.9% 19.7% 50.0% 17.7% 16.6%

S.D.

WILLINGNESS TO CONTINUE VOLUNTEERING AMONG …

(continued)

0 36

8

4

29

2.4% 13.7% 19.8% 69.4% 0.0% 9.5% 73

3.6% 39.1% 4.0% 47.2% 3.2% 2.8%

Mean

Age 65+

7

175

Other variables

Motivation factor

Activity in the past Participate voluntarily Participate nonvoluntarily No activity Health status Poor Not good Good Very good Non-earned income 1st quintile 2nd quintile 3rd quintile 4th quintile 5th quintile Altruism Participation into social activity of other family member Participating Participated in the past Not participated Male Working status in NPO

Variables

(continued)

Income factor

Table 7.4

35.4% 14.8% 49.8% 49.9%

21.0% 12.7% 17.3% 16.3% 17.3% 88.9%

47.8% 35.5% 50.0% 50.0%

40.8% 33.3% 37.9% 36.9% 37.9% 31.4%

39.6% 13.9% 46.5% 38.1%

17.7% 10.0% 16.5% 14.7% 25.9% 90.0%

11.5% 76.5% 10.9% 1.0%

35.4% 45.3% 33.3% 11.8%

14.7% 71.2% 12.7% 1.4%

Mean

38.6% 13.7% 47.7%

S.D.

Age 50–59

34.7% 47.6% 14.1% 34.8% 51.2% 50.0%

Mean

Total (19+)

48.9% 34.7% 49.9% 48.6%

38.2% 30.1% 37.1% 35.4% 43.9% 30.1%

32.0% 42.5% 31.2% 10.2%

48.7% 34.4% 50.0%

S.D.

41.7% 10.7% 47.6% 64.3%

19.6% 15.5% 19.6% 16.5% 11.7% 93.1%

10.4% 77.6% 10.7% 1.3%

38.5% 9.9% 51.6%

Mean

Age 60–64

49.3% 30.9% 50.0% 48.0%

39.7% 36.2% 39.7% 37.1% 32.2% 25.4%

30.5% 41.7% 30.9% 11.4%

48.7% 29.9% 50.0%

S.D.

41.5% 10.1% 48.4% 79.0%

14.1% 68.1% 16.5% 1.2% 10.1% 16.1% 18.5% 16.5% 12.9% 11.3% 96.0%

42.3% 8.5% 49.3%

Mean

Age 65+

49.4% 30.2% 50.1% 40.8%

34.9% 46.7% 37.2% 11.0% 30.2% 36.9% 38.9% 37.2% 33.6% 31.7% 19.7%

49.4% 27.8% 50.0%

S.D.

176 X. MA

Regular employee Non-regular employee Paid volunteer Unpaid volunteer in executive office Other types of unpaid volunteer Employment status in non-NPO organizations Regular employee Non-regular employee CEO Housewife Non-working Others Occupation CEO/Executive director Manager Clerk Accounting/HR Professional job Volunteer Other occupation

Variables

39.0% 29.0% 35.1% 32.2% 49.6% 15.3% 48.1% 23.7% 28.7% 32.9% 31.7% 37.5% 27.4%

18.7% 9.3% 14.3% 11.8% 43.5% 2.4% 36.2% 6.0% 9.1% 12.4% 11.3% 16.9% 8.2%

37.0% 6.7% 7.3% 10.8% 12.0% 16.5% 9.6%

19.8% 10.8% 16.6% 14.4% 36.1% 2.2%

5.1%

6.9% 25.3%

Mean 47.2% 20.5% 9.4% 17.7%

S.D.

Age 50–59

49.9% 39.1% 27.8% 39.6%

46.5% 18.8% 8.4% 19.4%

Mean

Total (19+)

48.3% 25.1% 26.1% 31.1% 32.5% 37.1% 29.5%

39.9% 31.1% 37.3% 35.1% 48.1% 14.8%

22.0%

50.0% 40.4% 29.3% 38.2%

S.D.

47.1% 4.0% 6.4% 15.2% 6.8% 14.2% 6.4%

16.6% 8.7% 15.0% 9.9% 46.3% 3.5%

9.6%

38.6% 16.6% 10.7% 24.5%

Mean

Age 60–64

50.0% 19.5% 24.5% 35.9% 25.1% 34.9% 24.5%

37.3% 28.3% 35.7% 29.9% 49.9% 18.3%

29.4%

48.7% 37.3% 30.9% 43.1%

S.D.

49.2% 19.7% 17.7% 28.5% 22.3% 32.7% 26.0%

32.7% 23.1% 39.3% 26.7% 50.1% 20.6%

38.6%

40.8% 28.5% 36.9% 48.1%

S.D.

(continued)

59.3% 4.0% 3.2% 8.9% 5.2% 12.1% 7.3%

12.1% 5.6% 19.0% 7.7% 51.2% 4.4%

18.1%

21.0% 8.9% 16.1% 35.9%

Mean

Age 65+

7 WILLINGNESS TO CONTINUE VOLUNTEERING AMONG …

177

Other variables

Married Family number Co-resident with parents Family care Caring Cared No experience Experience of in patient Suffered from the Great East Japan Earthquake Experience of retirement 【NPO organization level】 】 Change of reward Increased Decreased No change Average wage of regular employee Average wage of non-regular employee NPO organization size Less than 10 10–49 50–99

Variables

(continued)

Reward factor

Table 7.4

S.D.

51.9%

57.3% 49.5%

50.0%

46.9% 50.0% 40.9% 43.7%

37.3% 1 43.9% 38.1%

S.D.

68.8%

17.3% 42.2% 40.5% 26.9% 25.7%

87.3% 3 15.5%

Mean

Age 60–64

46.4%

37.9% 49.4% 49.1% 44.4% 43.7%

33.3% 1 36.2%

S.D.

69.5%

8.5% 45.2% 46.4% 27.4% 19.4%

89.9% 3 4.8%

Mean

Age 65+

46.1%

27.9% 49.9% 50.0% 44.7% 39.6%

30.2% 1 21.5%

S.D.

8.7% 28.1% 64.0% 48.0% 17.4% 37.9%

9.6% 65.2% 18.0%

29.5% 47.7% 38.4%

9.9% 60.5% 17.3%

29.9% 48.9% 37.9%

9.7% 62.5% 19.4%

29.6% 48.5% 39.6%

40.8% 49.2% 38.8% 48.8% 30.2% 46.0% 37.2% 48.9% 6.7% 25.1% 8.7% 28.2% 10.8% 31.1% 9.3% 29.4% 52.4% 50.0% 52.5% 50.0% 59.1% 49.3% 53.5% 50.5% 253 75 250 80 243 79 231 101 960 410 941 195 999 537 1,109 936

32.5% 49.9% 21.1% 25.6%

83.4% 3 26.1% 17.5%

Mean

Age 50–59

34.1% 46.9% 49.8% 40.8% 43.1%

13.4% 32.6% 54.0% 21.1% 24.6%

77.0% 42.1% 3 1 23.1% 42.2%

Mean

Total (19+)

178 X. MA

43.1% 28.0%

28.8%

50.1% 8.8% 7.9% 4.8% 9.1% 19.2%

25.3% 41.5% 24.6%

30.7% 46.1%

50.0% 28.9% 28.2% 23.8% 28.5% 39.0%

48.5% 9.2% 8.7% 6.0% 8.9% 18.7%

29.0% 45.4% 39.5% 48.9% 22.1% 41.5%

9.4% 29.2% 8.5% 2,313 622

43.5% 49.3%

50.0% 28.4% 27.1% 21.4% 28.8% 39.4%

45.3%

46.7%

32.1%

25.9% 49.2% 28.2%

S.D.

36.7% 48.2%

Mean 7.2% 58.9% 8.7%

S.D.

9.9% 29.9% 53.6% 49.9% 12.7% 33.3%

Mean

Age 50–59

Source Author’s creation based on the data from 2016 Activities and Working Status in NPOs Survey

Observations

More than 100 No change of CEO Hight proportion of younger employee High proportion of male employee High proportion of well-educated employee Activity field Medical and care Community Academic, culture, art or sports Environmental protection Child development Other 【Regional variable】 】 Community size Government-designated city City with population of 100 thousand or more, City with population less than 100 thousand Town and village

Variables

Total (19+)

10.4% 259

21.9%

32.3% 35.4%

44.0% 10.0% 8.9% 8.6% 8.7% 19.8%

31.0%

42.0%

12.4% 47.1% 7.7%

Mean

Age 60–64

30.5%

41.4%

46.8% 47.9%

49.7% 30.1% 28.5% 28.0% 28.3% 39.9%

46.3%

49.4%

32.9% 50.0% 26.7%

S.D.

8.1% 491

16.9%

34.7% 40.3%

36.7% 14.5% 12.9% 8.1% 7.7% 20.2%

30.2%

45.2%

8.5% 51.6% 2.0%

Mean

Age 65+

27.3%

37.6%

47.7% 49.2%

48.3% 35.3% 33.6% 27.3% 26.7% 40.2%

46.0%

49.9%

27.9% 50.1% 14.1%

S.D.

7 WILLINGNESS TO CONTINUE VOLUNTEERING AMONG …

179

180

X. MA

Table 7.5 Results of volunteering reward function in Japan Age 50–59 Coef. 【Individual level variables】 】 Motivation Altruism Activity start years (before 1990) 2000–2004 2005–2009 2010–February 2011 After March 2011 Age Age squared Education (Senior high school and lower) College University/graduate school Other Training Qualification (Non-qualification) Legal or tax Education Medical Other types of qualification Grade points Working status in NPO (Regular employee) Non-regular employee Paid volunteer Unpaid volunteer in executive office Other types of unpaid volunteer Other

Age 60–64 z-value

Age 65+

Coef.

z-value

Coef.

z-value

−0.203

−0.50

−0.479

−0.70

−0.185 −0.066 0.606

−0.33 −0.11 0.81

0.148

0.63

−0.427 −0.454 −0.346

−1.23 −1.28 −0.81

0.529 1.133** 0.118

1.08 2.31 0.19

−0.607* 0.526 −0.005

−1.69 0.47 −0.47

0.267 2.360** −0.164**

0.52 2.38 −2.37

−1.112* −1.401 0.010

−1.76 −1.30 1.28

0.088 0.014

0.37 0.06

0.033 0.232

0.08 0.76

−1.219* −0.593*

−1.79 −1.63

0.306 −0.085

1.16 −0.38

0.460 0.398

1.06 1.26

−0.016 0.679*

−0.02 1.61

−0.162 0.175 0.159 0.331*

−0.28 0.86 0.87 1.79

−0.323 −0.608* −0.618** 0.656**

−0.47 −1.72 −2.30 2.32

2.96

−0.294

−1.17

0.110

0.35

−1.03

0.066

0.22

0.131

0.34

0.541***

−0.211

3.314*** 3.69 0.172 0.39 0.104 0.28 −0.341 −0.98

−6.630*** −20.39 −10.857*** −17.08

−7.103*** −11.87 −8.535*** −15.96

−5.995***−13.10 −10.596***−16.64

−17.790*** −4.76

−8.692*** −10.73

−9.870***−12.55

(continued)

7

WILLINGNESS TO CONTINUE VOLUNTEERING AMONG …

181

Table 7.5 (continued) Age 50–59

Employment status in non-NPO organizations (Regular employee) Non-regular employee CEO Housewife Non-working Others Working hours in non-NPO organization Male Health (Poor) Very good Good No good Experience of retirement 【Organizationlevel variables 】 Organization size (Less than 10) 10–49 50–99 More than 100 Average productivity No change of CEO High proportion of younger employee High proportion of male employee High proportion of well-educated employee High proportion of non-regular employee

Age 60–64 Coef.

Age 65+

Coef.

z-value

z-value

Coef.

z-value

−0.081

−0.21

0.500

0.81

0.391

−1.027*** −1.093** −1.048** 0.142 −0.024***

−2.93 −2.45 −2.51 0.22 −2.59

0.297 0.898 0.593 0.523 0.005

0.62 1.37 1.05 0.56 0.37

1.886** −0.077 0.994 1.801 0.031

0.529***

2.66

−0.385

−1.26

0.441

0.96

0.36 2.29 −0.06 0.90 1.37 1.21

−0.165 −0.569 −0.651 0.286*

−0.23 −0.84 −0.92 1.70

1.748* 0.243 0.789 0.333

1.86 0.28 0.87 1.23

−0.911 −0.416 −1.299 −0.492

−0.45 −0.21 −0.64 −1.43

−0.464 −0.679* −0.757* 0.000 −0.114 −0.104

−1.34 −1.77 −1.73 −0.16 −0.71 −0.38

0.886* 1.130** 0.619 0.000 0.201 0.238

1.83 2.12 1.06 0.11 0.83 0.58

0.108 0.162 −0.394 0.000 −0.352 0.660

0.15 0.20 −0.45 −0.01 −1.13 1.07

−0.475**

−2.19

−0.313

−1.07

0.145

0.72

0.415

1.39

−0.101

−0.27

0.323*

1.76

−0.342

−1.17

0.185

0.52

−1.374*** −3.73

(continued)

182

X. MA

Table 7.5 (continued) Age 50–59 Coef. Activity field (Medical or care) Community Academic, culture, art or sports Environmental protection Child development Other Constant Non-censoring variables Censoring variables Observations Maximum log likelihood Pseudo R 2

Age 60–64

Age 65+

Coef.

z-value

1.51 1.05

−0.047 0.247

−0.08 0.53

−0.815*

−1.92

−0.015

−0.03

0.492*

1.74

−0.419

−0.88

−0.309 −27.274** 324

−0.85 −2.36

0.522 0.363

0.020 −5.898 457

z-value

0.09 −0.19

Coef.

0.413 1.733*** −0.328 0.991* 0.506 56.971 235

160 617 −932.100

98 313 −475.899

219 454 −616.503

0.368

0.377

0.338

z-value

0.61 3.07 −0.46 1.70 1.16 1.49

Note ***p < 0.01, **p < 0.05, *p < 0.10 Source Author’s creation based on the data from 2016 Activities and Working Status in NPOs Survey

Notes 1. For the survey on volunteerism, please refer to Smith (1994), Bekkers and Wiepking (2011), and Wilson (2012). 2. According to the partial equilibrium model for labor force participation decisions, when the market wage rises, the market wage will be higher than the reservation wage, work in labor market will increase (longer working hours), and leisure time may decrease (substitution effect). On the contrary, if the market wage rises, when the expected income is constant, leisure becomes a senior good, work in the labor market may decrease (to reduce working hours) and leisure time may increase (income effect). 3. According to the partial equilibrium mode, the slope of the budget constraint is determined by the market wage rate, and the intercept of the budget constraint is determined by unearned income. 4. For results of the reward function, please see Appendix Table 7.5.

7

WILLINGNESS TO CONTINUE VOLUNTEERING AMONG …

183

5. Although the instrument variable (IV) method can be used to address the endogeneity problem, it is difficult to find appropriate instrumental variables. The estimation to address the endogeneity problem becomes our future research issue. Empirical analysis using more rigorous manipulated variables is a topic for the future. In this study, we use the start year of participation in social activity, age, and region as identification variables in the 2SLS model. 6. For details on extraction weights for 47 districts, please refer to the Figure and Table 1-1-1 on page 3 in Activity and Work Status Survey (Organization Survey and Individual Survey): Considering the Support Activities of Great East Japan Earthquake Reconstruction Support Activities (JILPT Survey Series No. 139, May 2019) (In Japanese). For the classifications of 47 districts in Japan, please refer https://www.stat.go.jp/data/shugyou/ 1997/3-1.html. 7. For the descriptive statistics of variables, please see Appendix Table 7.4. 8. This study used seven kinds of occupation dummy variables: CEO/executive director, volunteer, manager, clerk, accounting/HR, professional job, and other occupations. 9. For the activity status of family members, three kinds of dummy variables are constructed: (1) participating in social activity now; (2) participated in the past; (3) have not participated. 10. For the work status in the current NPO, five kinds of dummy variables are used: (1) regular employee; (2) non-regular employee; (3) paid volunteer; (4) unpaid volunteer in executive office; and (5) other types of unpaid volunteers. 11. For the employment status in non-NPOs, we used six kinds of dummy variables as follows: (1) regular employee; (2) non-regular employee; (3) CEO; (4) housewife; (5) non-working; and (6) others. 12. Three kinds of family care dummy variables are used: (1) caring; (2) cared; and (3) did not care. 13. For the number of family members, six kinds of dummy variables from number one to number six are used. 14. According to the questionnaire of the NPO survey, the activity fields are composed of 20 kinds; in order to secure a sample that can be analyzed, the activity fields were reclassified, and six kinds of dummy variable were used in the study: (1) medical and care; (2) community; (3) academic, culture, art, or sports; (4) environmental protection; (5) child development; and (6) other activities. 15. For NPO size, four kinds of dummy variables are used: (1) less than 10 employees; (2) 10–49 employees; (3) 50–99 employees; and (4) more than 100 employees. 16. It can be thought that when the management of an NPO is unstable, the willingness to continue social activity will decrease. To control the

184

X. MA

influence of management change, the change of executive director dummy (1 = change, 0 = no change) variable is used. 17. Based on answers to questions about the human resource structure (age, education, gender) of the NPO, we used three kinds of dummy variables: (1) high proportion of younger employees; (2) high proportion of welleducated employees; and (3) high proportion of male employees. For example, for the age structure, the survey asks respondents to choose whether “A. there are more younger employees (employees younger than 35); or B. there are more middle-aged and elderly employees.” There are four options for the answer: (1) close to A; (2) somewhat close to A; (3) somewhat close to B; and (4) close to B. The NPO belongs to the group with a “high proportion of younger employees” when the NPO chose “(1) close to A.”

References Atkins, R., Hart, D., & Donnelly, T. (2005). The association of childhood personality type with volunteering during adolescence. Merrill-Palmer Quarterly, 51, 145–162. Amagasa, S., Fukushima, N., Kikuchi, H., Oka, K., Takamiya, T., Odagiri, Y., et al. (2017). Types of social participation and psychological distress in Japanese older adults: A five-year cohort study. PLoS ONE, 12(4), Atoda, N., Kim, L., & Mekawa, K. (1999). Social welfare and volunteer: case study of japan and Korea. Social Security Research, 35(3), 264–275. (In Japanese). Atoda, N., & Fukushige, M. (2000). Participation structure of volunteer activity of middle-aged adults and elders: Determinants analysis based on survey data. Social Security Research, 36(2), 246–255. (In Japanese). Becker, G. S. (1964). Human capital: A theoretical and empirical analysis, with special reference to education. New York: Columbia University Press. Bekkers, R., & Wiepking, P. (2011). A literature review of empirical studies of philanthropy: Eight mechanisms that drive charitable giving. Nonprofit and Voluntary Sector Quarterly, 40(5), 924–973. Boyle, M., & Sawyer, J. (2010). Defining volunteering for community campaigns: An exploration of race, self-perceptions, and campaign practices. Journal of Community Practice, 18, 40–57. Carlin, P. S. (2001). Evidence on the volunteer labor supply of married women. Southern Economic Journal, 67 (4), 801–824. Einolf, C. (2008). Empathic concern and prosocial behaviors: A test of experimental results using survey data. Social Science Research, 37, 1267–1279. Einolf, C. (2010). Gender differences in the correlates of volunteering and charitable giving. Nonprofit and Voluntary Sector Quarterly, 40(6), 1092–1112.

7

WILLINGNESS TO CONTINUE VOLUNTEERING AMONG …

185

Erez, A., Mikulincer, M., van Ijzendoorn, M., & Kroonenberg, P. (2008). Attachment, personality, and volunteering: Placing volunteerism in an attachment-theoretical framework. Personality and Individual Differences, 44, 64–74. Finkelstein, M. (2008). Predictors of volunteer time: The changing contributions of motive fulfillment and role identity. Social Behavior and Personality, 36, 1353–1364. Freeman, R. B. (1997). Working for nothing: The supply of volunteer labor. Journal of Labor Economics, 15(1), 140–166. Gronlund, H. (2011). Identity and volunteering intertwined: Reflections on the values of young adults. VOLUNTAS: International Journal of Voluntary and Nonprofit Organizations, 22(4), 852–874. Handy, F., & Cnaan, R. (2007). The role of social anxiety in volunteering. Nonprofit Management and Leadership, 18(1), 41–57. Kanamori, S., Kai, Y., Aida, J., Kondo, K., Kawachi, I., & Hirai, H. (2014). Social participation and the prevention of functional disability in older Japanese: The JAGES cohort study. PLoS ONE, 9(6), e99638. Ma, X. (2012a). The determinants of social activity participation of elderly: Focus on volunteer activity supply. In JILPT (Ed.), Research on social contribution activity of elderly: Based on quantitative and qualitative analysis (JILPT Labor Policy Research Report No. 142, pp. 39–72). (In Japanese). Ma, X. (2012b). The motivation of social contribution activity of middle-aged adults and elders and its influence on activity status. In JILPT (Ed.), Research on social contribution activity of elderly: Based on quantitative and qualitative analysis (JILPT Labor Policy Research Report No. 142, pp. 73–102). (In Japanese). Ma, X. (2014). An empirical study on determinant of volunteer supply of elderly. Japanese Labor Research, 643, 70–80. (In Japanese). Ma, X. (2016) Wage structure and its influences on wage satisfaction and willingness to continue activity. In JIPLT (Ed.), Research on NPO working: Capture changes driven by constant growth and the disaster (JILPT Labor Policy Research Report No. 183, pp. 54–97). (In Japanese). Ma, X., & Ono, A. (2013). Determining factors in middle-aged and older person’s participation in volunteer activity and willingness to participate. Japan Labor Review, 10(4), 90–119. Ma, X., Piao, X., & Oshio, T. (2020). The impact of social participation on health among middle-aged and elderly adults: Evidence from longitudinal survey data in China. BMC Public Health, 20, 502. Menchik, P. L., & Weisbrod, B. A. (1987). Volunteer labor supply. Journal of Public Economics, 32, 159–183.

186

X. MA

Messner, M., & Bozada-Dea, S. (2009). Separating the men from the Moms: The making of adult gender segregation in youth sports. Gender and Society, 23(1), 49–71. Mincer, J. (1974). Schooling, experience and earning. New York: Columbia University Press. Moriyama, T. (2007). The career, role and work status of executive director. In JIPLT (Ed.), Path to employment development of NPO: Thought from human resources, finances and legal systems (JILPT Labor Policy Research Report No. 82, pp. 64–93). (In Japanese). Moriyama, T. (2016). NPO as a carrier: Comparative analysis based on the 2004 and 2015 surveys. In JIPLT (Ed.), Research on NPO working: Capture changes driven by constant growth and the disaster (JILPT Labor Policy Research Report No. 183, pp. 98–120). (In Japanese). Ono, A. (2006). Work status and consciousness of paid volunteers: Does reward affect the continue activity? In JILPT (Ed.), Paid employee and volunteer in NPO: Work status and consciousness (JILPT Labor Policy Research Report No. 60, pp. 103–141). (In Japanese). Oshio, T., & Kan, M. (2019). Preventive impact of social participation on the onset of non-communicable diseases among middle-aged adults: A 10-wave hazards-model analysis in Japan. Preventive Medicine, 118, 272–278. Rotolo, T., Wilson, J., & Hughes, M. E. (2010). Homeownership and volunteering: An alternative approach to studying social inequality and civic engagement. Sociological Forum, 25, 570–587. Schram, V. R., & Dunsing, M. M. (1981). Influences on married women’s volunteer work participation. The Journal of Consumer Research, 7 (4), 372–379. Segal, L. M., & Weisbrod, B. A. (2002). Volunteer labor sorting across industries. Journal of Policy Analysis and Management, 21(3), 427–447. Smith, D. H. (1994). Determinants of voluntary association participation and volunteering: A literature review. Nonprofit and Voluntary Sector Quarterly, 23(3), 243–263. Taniguchi, H. (2011). The determinants of formal and informal volunteering: Evidence from the American Time Use Survey. VOLUNTAS: International Journal of Voluntary and Nonprofit Organizations, 23, 920–939. Vaillancourt, F. (1994). To Volunteer or not: Canada, 1987. Canadian Journal of Economics, 27 (4), 813–826. Wilson, J. (2012). Volunteerism research: A review essay. Nonprofit and Voluntary Sector Quarterly, 4(2), 176–212. Yamauchi, N. (2001). Non-profit market from gender perspective: Why housewives aim for NPOs. Japanese Labor Research, 493, 30–41. (In Japanese).

CHAPTER 8

Seniority Wage and Employment of the Older Workers in Japanese Firms Atsushi Seike and Xinxin Ma

8.1

Introduction

As mentioned in the former chapter, Japan has experienced an aging population; it has become the country with the highest proportion of

This chapter is a revised version of: Seike, A., & Ma, X. (2010). The determinants of employment age setting of older adults: An empirical study on the influences of the seniority wage. In The Japan Institute for Labour Policy and Training (JILPT) (Ed.), Situations and issues of continued employment of older adults (Reports of Labor Policy Research No. 100). Tokyo: JILPT (In Japanese). A. Seike Faculty of Business and Commerce, Keio University, Tokyo, Japan e-mail: [email protected] X. Ma (B) Faculty of Economics, Hosei University, Tokyo, Japan e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 X. Ma (ed.), Employment, Retirement and Lifestyle in Aging East Asia, Social Policy and Development Studies in East Asia, https://doi.org/10.1007/978-981-16-0554-3_8

187

188

A. SEIKE AND X. MA

older adults in the world. Therefore, increasing the labor force participation of older adults has become an important issue for the Japanese government (Seike and Ma 2008). In fact, to address the problem of labor force decreased as the population aging and to reduce the government financial burden of public pension, since the 1980s, the Japanese government has implemented a set of policies. For example, the mandatory retirement age in most Japanese firms is around 60 years. In 2007, the Japanese government published the Elderly Employment Security Act, and firms are obliged to maintain the employment of workers aged 60 years and older. Considering the employment of older adults, from the labor demand perspective,1 it is argued that a firm’s wage system may have a significant impact. In most Japanese firms, particularly in large-sized firms, the seniority wage,2 long-term employment, and mandatory retirement systems are performed, which are called as “Japanese style management.” Ohashi (2005), Higuchi and Yamamoto (2002), and Mitani (2001) found that Japanese firms have adjusted the age-wage profile (the relationship between wage and age) to respond to the policy of raising the mandatory retirement age. However, empirical studies on the influence of seniority wage on Japanese firms’ employment of older workers are scarce. This study fills this gap. Using the unique Japanese firms’ survey data, this chapter investigates the influence of seniority wage system on the mandatory retirement age and reemployment age in Japanese firms. The empirical results suggest that seniority wage significantly affects mandatory retirement age and reemployment age in Japan. The remainder of this chapter is organized as follows. Section 8.2 introduces the Lazear model to explain the influence of seniority wage on mandatory retirement age and summarize previous empirical studies on this issue. Section 8.3 gives the framework of the empirical analysis, including models and datasets. Section 8.4 introduces the results of the influence of seniority wage on mandatory retirement age and reemployment age. Section 8.5 summarizes the conclusions and policy implications.

8

SENIORITY WAGE AND EMPLOYMENT OF THE OLDER WORKERS …

8.2 8.2.1

189

Literature Review

Lazear Model on Seniority Wage and Mandatory Retirement Age

Regarding the relationship between seniority wage3 and mandatory retirement age, the implicit contract hypothesis was advocated by Lazear (1979, 1981). Under the assumption that labor productivity is constant from younger to older age, when the wage level increases with age (the seniority wage), based on the zero profit-wage path,4 a worker will obtain a wage lower than the productivity at a younger age and obtain a wage higher than the productivity at an older age. When the worker is fired for some reasons (e.g., when he or she does not put in effort at work) at a younger age, the worker may lose a higher wage that he or she will obtain at an older age. Therefore, the seniority wage system may prevent negligence and promote workers to work within a firm for a long term. The Lazear (1979, 1981) model indicates the rationality of the existence of a seniority wage system and mandatory retirement system. Figure 8.1 expresses the association between the seniority wage system and mandatory retirement system. AB and AB’ are the age-wage curves of firms M and N , respectively, and the gradient of AB is larger than that of AB’. Based on the zero profit-wage path, for a firm, it is desirable that a worker’s lifetime contribution (labor productivity) is equal to the lifetime wage

B” B'

B

C

O O’

P

P'

A

RA1(M)

RA2(N) mandatory retirement age

Fig. 8.1 Age-wage profile (seniority wage) and mandatory retirement age (Source Author’s creation based on Lazear [1979])

190

A. SEIKE AND X. MA

wage. When ACO = OPB for firm M or ACO’=O’P’B’ for firm N , both firms M and N can employ the worker. Since the gradient of the agewage profile differs between firms M and N , the mandatory retirement age differs for these two firms: RA1 for M and RA2 for N , respectively. This suggests that the wage structure (the relationship between wage and age) may influence the mandatory retirement age; when the age-wage profile gradient is lower, the mandatory retirement age or reemployment age of elders may rise, as in firm N . 8.2.2

Empirical Studies on Seniority Wage and Employment of Older Workers in Japan

For the seniority wage system, Ono (1989) investigated the relationship between wage and age by using survey data from the Basic Wage Structure Survey conducted in 1970, 1975, and 1980 by the Ministry of Health, Labor and Welfare Japan. He estimated the wage function using variables, including years of tenure, age, and years of occupational work experience and found that the influence of age on wages is greater than that of years of tenure. He pointed out that for the reasons of seniority wage in Japanese firms, the influence of the living security effect may be greater than the human capital effect. Koike (2005) compared the relationship of wage and age between Japan and other developed countries, such as the US and European countries. He found that in Japan, the association between wage and age of blue-collar workers is like that of white-collar workers. He pointed out that the seniority wage system also covered blue-collar workers in Japan, but not in other countries. Kawaguchi and Kanbayashi (2007) conducted matching panel data based on the Basic Wage Structure Survey from 1993 to 2003 and Industrial Statistics Survey of Firms to estimate the relationship between productivity and age and the relationship between wage level and age. They found that the gradient is higher for the age-wage profile than the ageproductivity curve for the older worker group, suggesting that workers obtain wages lower than their productivities at younger ages, while they obtain wages higher than their productivities at older ages. These previous studies suggest that in Japanese firms, there is a seniority wage system that covers both blue-collar workers and white-collar workers, the wage levels increased with age (or years of tenure, years of work experience). How does the seniority wage system influence the employment of older workers in Japan? Seike (1994) investigated the influence of the

8

SENIORITY WAGE AND EMPLOYMENT OF THE OLDER WORKERS …

191

relationship between wage and contribution at 55 years on reemployment by using the Survey on Wage System Models Adapted to the Aging Era conducted by the Elderly Employment Development Association in 1994. He found that the probability of performing the elderly reemployment system is lower for firms that answered that the wages of employees are higher than their contributions. Kubo (1994) pointed out that the wage-age profile is flatter for firms in which the average age of employees is higher by using the same survey data. Using the Firm Continuing Employment Survey conducted by the Japan Institute for Labor Policy and Training (JILPT) in 2007, Yamada (2007) found that the average wage level and the rise range of wage level influence the employment of workers aged 60 years. Using the Elderly Employment Status Survey (Firm Survey) conducted by the Ministry of Health, Labor and Welfare Japan in 1992, 1996, and 2000, Higuchi and Yamamoto (2002) indicated that the proportion of implementation of mandatory retirement system is larger for firms in the industry sector in which the influence of age on the wage level is greater (with a higher gradient of the age-wage profile). Although previous studies have investigated the senior wage system and its influence on the employment of older workers in Japan, the influence of the seniority wage system on mandatory retirement age is scarce. This study fills this gap.

8.3

Methodology and Data 8.3.1

Model

To investigate the influence of the seniority wage system on firms’ mandatory retirement age and reemployment age setting, two problems should be considered. The first is the sample selection bias problem in the results using the OLS model. This study uses the Maddala model (Maddala 1983) to address the sample selection bias problem. The second is the endogeneity problem between the mandatory retirement age or reemployment age setting and the seniority wage system. We used an imputed gradient of seniority wage which is similar to the 2SLS method to address the endogeneity problem. A set of variables, such as firm operation age, average age squared of permanent worker, and average years of tenure and its squared of permanent worker, are used as the identification variables in the wage function. These models are expressed as follows:

192

A. SEIKE AND X. MA

First, we perform the probit regression model to calculate the adjustment item. The probit regression model is expressed by Eq. 8.1. Pr(y = 1) = Pr(a + γ1 SWi + γ2 Hi + vi > 0)

(8.1)

In Eq. 8.1 Pr(y) denotes the probability of performing the mandatory retirement system or reemployment of older workers in a firm, SW expresses the index of seniority wage in a firm, H is a factor that influences the probability of setting the mandatory retirement age, or setting the reemployment age in a firm, and γ is the coefficient of each independent variable. Based on the results of the probit regression model, the selection bias correction item can be calculated using Eq. 8.2. φ(·) expresses the density function, and Φ(·) expresses the distribution function. λ=

φ(γ1 SW + γ2 H ) Φ(γ1 SW + γ2 H )

(8.2)

The mandatory retirement age function is expressed by Eq. 8.3. M R A = b + β1 SWi + β2 λi + β3 X i + εi

(8.3)

In Eq. 8.3, M R A indicates the mandatory retirement age or reemployment age set by a firm, λ is the selection bias correction item, X indicates the factors that influence the mandatory retirement age or reemployment age setting in a firm, and β indicates the coefficient of each variable. 8.3.2

Data and Variables

This study used a unique survey—Elderly Employment and Recruitment Survey (EERS)—conducted by the JILPT in September 2008. The samples were 5,000 Japanese firms. We can obtain rich information, such as the seniority wage system, mandatory retirement age, reemployment age, and firm characteristics, from the EERS, which can be used to investigate the influence of the seniority wage system on mandatory retirement age or reemployment age in Japanese firms. The main dependent variables are mandatory retirement age and reemployment age (see Table 8.8). They are conducted based on the questions “How old is the permanent worker’s mandatory retirement age in your firm?” and “How old is the oldest age in firm’s elderly reemployment system?” In the estimations of the probability of the mandatory retirement system and elderly reemployment system, the binary variables (1 =

8

SENIORITY WAGE AND EMPLOYMENT OF THE OLDER WORKERS …

193

performing the mandatory retirement system or the elderly reemployment system, 0 = otherwise) are used.5 For the independent variables (see Appendix Table 8.8), first, the main independent variable is the index of seniority wage. Based on the questions in the questionnaire “How much is the average monthly salary of new workers in your firm?” and “When the average monthly salary of new workers is 100, how much are the average wage levels for each age group in your firm?”, we calculated the average wage levels for the groups aged 30, 40, 45, 50, 55, 59, and 60–64. Then, we constructed the three kinds of indices of seniority wage in a firm as follows: i. Index 1 is the imputed wage increased rate by each age, which was used in Lazear (1979). Index 1 is calculated using Index 1 = W¯ 50 −W¯ 30 ¯ . W50 and W¯ 30 are imputed average wages of the group aged 20 50 years and the group aged 30 years based on the wage functions shown in Appendix Table 8.9.6 ii. Index 2 is the actual wage increased rate by each age, which was used in Yamada (2007). Index 2 is calculated using Index 2 = W50 −W30 . W50 and W30 are the actual wage levels of the group aged 20 50 years and the group aged 30 years. iii. Index 3 is the ratio of average wage of workers aged 30 years to average wage of workers aged 50 years, which was used in Kubo (1994) and Higuchi and Yamamoto (2002). Index 3 is calculated 50 using Index 3 = W W30 . Second, for the other variables, (1) the relationship between wage and contribution at 55 years may influence the firm’s retirement age setting, and the three dummy variables are used to control the influences: (i) wage is higher than contribution; (ii) wage is equal to contribution; and (iii) wage is lower than contribution; (2) regarding the influences of the starting salary of a new worker, the starting wage level variable is used; (3) the weekly working hours are used to control the influence of working hours management. Third, for the firm characteristics, (1) five kinds of industry sector dummy variables—construction, manufacturing,7 wholesale and retail, service, and others8 ; (2) three kinds of firm size dummy variable9 —firm with employees from 1 to 99, firm with employees from 100 to 299, and firm with employees more than 300; (3) the trade union dummy variable,

194

A. SEIKE AND X. MA

which is equal to 1 when there is a trade union in a firm, 0 when otherwise, are used as control variables; and (4) firm operation year10 is used as an identification variable. Fourth, for the firms’ employee characteristics, five kinds of variables—(1) the average age, (2) years of tenure of permanent employees, (3) the proportion of female workers, (4) the proportion of permanent employees, and (5) the education level of most employees—are also used in this study. Fifth, to control the influences of other systems in a firm, five dummy variables—(1) holding the seminar on elderly career and life, (2) vocational training system, (3) turnover and job search system, (4) advantage for early retirement system, and (5) support for start-up initiatives are conducted. Sixth, the following nine kinds of regional dummy variables are used to control the influence of regional disparity in labor markets: Hokaido, Tohoku, Hokuliku, Tokai, Kinki, Shikoku, and Kyushu.

8.4 Results of Descriptive Statistics of the Relationship of Age and Wage 8.4.1

Age-Wage Profile by Education

The association between age and wage by education is shown in Fig. 8.2. It is observed that in each education group, the wage level increases from starting point to 59 years of age. Compared to the group aged 40– 59 years, the wage level is lower for the group aged 60 years and older. In addition, the gradient of the age-wage profile from the starting point to 59 years of age is higher for the well-educated (college) group than the less-educated group (senior high school). 8.4.2

Age-Wage Profile by Firm Size

The relationship between age and wage by firm size is presented in Fig. 8.3. This suggests that in each group, the wage level increases from the starting point to 59 years of age; compared to the group aged 40– 59 years, the wage level is lower for the group aged 60 years and above. It also observed that the gradient of the wage-age profile from the starting point to 59 years of age is higher for large firms (the firm with employees

8

SENIORITY WAGE AND EMPLOYMENT OF THE OLDER WORKERS …

195

Unit: 10,000 Japanese Yen monthly

50 45 40 35 30 25 20 15 10

starting

age30 Total

age40

age50

Senior high school

age59

60 and older

College

Fig. 8.2 Age-wage profile by education in Japan (Source Author’s creation based on Elderly Employment and Recruitment Survey) Unit: 10,000 Japanese Yen monthly

50 45 40 35 30 25

20 15 10

starting less than 99

age30 100-299

age40 300-499

age50 500-999

age59

60 and older

more than 1,000

Fig. 8.3 Age-wage profile by firm size in Japan (Source Author’s creation based on Elderly Employment and Recruitment Survey)

196

A. SEIKE AND X. MA Unit: 10,000 Japanese Yen monthly 50 45 40 35 30 25 20 15 10

starting

age30

age40

age50

Construction

Manufacturing

Finance/Housing

Service

age59

60 and older

Wholesale and retail

Fig. 8.4 Age-wage profile by industrial sectors in Japan (Source Author’s creation based on Elderly Employment and Recruitment Survey)

from 500 to 999 and the firm with employees more than 1,000) than the medium-size and small-sized firms. 8.4.3

Age-Wage Profile by Industrial Sectors

Figure 8.4 expresses the relationship between age and wage by industrial sectors. This indicates that in each firm size group, the wage level increases from the starting point to 59 years of age; compared to the group aged 40–59 years, the wage level is lower for the group aged 60 years and above. It also observed that the gradient of wage-age profile from starting point to 59 years of age is higher for the wholesale and retail, construction, and service industry sectors, while it is lower for the finance/housing and manufacturing industry sectors.

8.5

Results of Econometric Analyses

8.5.1 Does the Seniority Wage Influence the Probability of Performing the Mandatory Retirement System in Firms? Does the seniority wage system influence the probability of performing mandatory retirement systems in firms? To answer the question, we

8

SENIORITY WAGE AND EMPLOYMENT OF THE OLDER WORKERS …

197

employ the estimation by using probit regression model, and the results are summarized in Table 8.1. First, the results indicate that Index 3 positively affects the probability of performing the mandatory retirement system in a firm. Although the rate of firms in which the mandatory retirement system is performed in the survey is 98%, it is shown that when the increased range of wages with age is larger, the probability of performing the mandatory retirement system is higher. Second, a firm with a large proportion of male employees tends to perform the mandatory retirement system. Third, the influences of the operation year, industry sector, firm size, trade union, average age of employees, average tenure years of employees, proportion of permanent employees, average education attainment level of workers, and the relationship between wage and productivity at 55 years on the probability of performing the retirement system are not statistically significant. 8.5.2

Does the Seniority Wage Influence the Mandatory Retirement Age?

Table 8.2 reports the results of mandatory retirement age functions using Maddala model. The selection items for model 1 (Index 1 model) and model 3 (Index 3 model) are statistically significant and are the positive values. The results indicate that there remains a sample selection bias in the results using the OLS model. The coefficient of Index 1 is −4.232, and it is statistically significant at 1% level. It suggests that when the annual average wage increases 10,000 Japanese Yen (JY), the mandatory retirement age may be lowered by 4 years in a firm. The result is consistent with the Lazear (1979) model. As mentioned above, according to Lazear (1979) model, to enforce a worker working for a long term and enhance the motivation for long term, a firm may perform the seniority system, which lets the worker obtain higher wage as they age. To address the problem of the gap between wage and labor productivity, the firm fires the worker at some time point that is the mandatory retirement age. This study corresponds to the Lazear (1979) model and suggests that the seniority wage system, particularly the increased range of wages with age, does influence the mandatory retirement age. Table 8.3 summarizes the results for different firm sizes. The coefficients of Index 1 are −4.814 for firms with employees from 1 to 99 (small

198

A. SEIKE AND X. MA

Table 8.1 Probability of performing the mandatory retirement system in Japanese firms Index 1

Seniority wage Industry sector (Manufacturing) Construction Wholesale and retail Service Other Firm operation years Firm size (1–99) 100–299 More than 300 Trade union Average age of employees Average age sq. Average years of tenure Average years of tenure sq. Proportion of permanent employees Proportion of female employees Education (Lower than senior high school) College and higher Relation between wage and productivity (Wage > productivity) Wage = productivity Wage < productivity Region Constant Observations Log likelihood Pseudo R 2

Index 2

Index 3

Coef.

z-value

Coef.

−1.801

−0.70

0.334

−0.299 −0.100 −0.102 −0.225 0.010

−0.63 −0.28 −0.35 −0.89 1.54

0.208 0.157 0.378 0.163

1.00 0.46 1.24 0.88

0.218 0.367 0.515 0.111

0.99 1.04 1.32 0.78

0.227 0.343 0.520 0.117

1.02 0.97 1.32 0.81

−0.002 −0.039

−0.87 −0.44

−0.001 −0.082

−0.70 −1.19

−0.001 −0.091

−0.71 −1.29

0.002

0.66

0.002

0.96

0.003

1.03

0.673

1.39

0.487

1.15

0.479

1.13

−0.968*

−1.73

−0.589* −0.034 −0.070 −0.217 0.016***

−0.947*

z-value

Coef.

z-value

1.35

0.826***

2.21

−1.82 −0.11 −0.23 −0.81 2.28

−0.575* −0.020 −0.065 −0.193 0.016***

−1.75 −0.06 −0.22 −0.72 2.28

−1.67

−0.951*

−1.68

0.801

1.08

0.186

0.80

0.143

0.61

−0.174

−0.76

−0.241

−0.91

−0.222

−0.83

−0.105

−0.40

−0.258

−0.88

−0.258

−0.87

Yes −0.328 1,914 −116.94 0.0931

−0.11

Yes −0.707 1,285 −101.39 0.1255

−0.24

Yes −1.810 1,285 −99.58 0.1412

Note ***p < 0.01, **p < 0.05, *p < 0.10 Source Author’s creation based on Elderly Employment and Recruitment Survey

−0.60

8

SENIORITY WAGE AND EMPLOYMENT OF THE OLDER WORKERS …

199

Table 8.2 Results of mandatory retirement age function in Japan Index 1 Coef. Seniority wage Relation between wage and productivity (Wage > productivity) Wage = productivity Wage < productivity Starting wage Working hours Industry sector (Manufacturing) Construction Wholesale and retail Service Other Firm size (1–99) 100–299 More than 300 Trade union Average age of employees Average years of tenure Proportion of regular employees Proportion of female employees Education (Lower than senior high school) College and higher Seminar Vocational training Turnover system Early retirement system Support for starts-up Support for job search

Index 2

Index 3

t-value

Coef.

t-value

Coef.

t-value

−5.85

−0.117

−1.28

−0.096

−0.72

0.074

0.87

0.057

0.60

0.061

0.65

0.052

0.55

0.013

0.12

0.012

0.12

1.613*** 0.006

6.30 1.32

1.199*** 0.007

4.73 1.43

1.110*** 0.007

4.35 1.44

0.487*** 0.170

2.83 1.51

−0.066 −0.062

−0.39 −0.55

−0.073 −0.066

−0.44 −0.58

0.267** 0.077

2.43 0.76

0.189 0.164

1.58 1.50

0.188 0.166

1.58 1.52

−0.012 0.243** 0.132 −0.011

−0.15 1.99 1.41 −1.12

−0.048 0.040 0.032 0.034***

−0.54 0.31 0.32 4.22

−0.044 −0.50 0.044 0.34 0.031 0.31 0.035*** 4.30

0.014

1.58

−0.024***

−3.32

−0.024*** −3.38

0.270

1.36

−0.023

−0.12

−0.018

−0.09

0.060

0.21

0.326

1.02

0.308

0.97

−4.232***

0.888*** −0.172* 0.054 −0.172 0.040

4.18 −1.73 0.34 −0.63 0.25

−0.242* −0.238** 0.088 −0.278 0.057

−2.60 −2.17 0.51 −0.96 0.32

−0.247*** −2.67 −0.239** −2.18 0.085 0.49 −0.285 −0.98 0.045 0.26

0.430 −0.228

0.79 −0.50

0.526 −0.410

0.94 −0.84

0.526 −0.407

0.94 −0.83

(continued)

200

A. SEIKE AND X. MA

Table 8.2 (continued) Index 1

Index 2

Coef. Region Selection adjust item Constant Observations Adjusted R 2

t-value

Yes 19.250** 52.383*** 1,764 0.0802

1.96 17.99

Index 3

Coef.

t-value

Yes 11.634 52.282** 1,466 0.0593

1.56 21.68

Coef.

t-value

Yes 12.273* 52.398** 1,466 0.0641

1.84 23.29

Note ***p < 0.01, **p < 0.05, *p < 0.10 Source Author’s creation based on Elderly Employment and Recruitment Survey

Table 8.3 Results of retirement age function by firm size in Japan Index 1

Employees 1–99 Employees 100–299 Employees more than 300

Index 2

Index 3

Coef.

t-value

Coef.

t-value

Coef.

t-value

−4.814*** −5.080*** −5.080***

−3.99 −4.58 −4.58

−0.060 −0.205 −0.205

−0.40 −1.53 −1.53

−0.050 −0.189 −0.189

−0.23 −0.95 −0.95

Note 1. ***p < 0.01, **p < 0.05, *p < 0.10 2. The other variables are similar to those in Table 8.1. They are not expressed in this table Source Author’s creation based on Elderly Employment and Recruitment Survey

firm) and −5.080 for firms with employees from 100 to 299 (medium firm), and they are statistically significant at 1% level. Although the statistical significance is at 10% level, the coefficient of Index 1 for firms with more than 300 employees (large firms) is −2.785. These results indicate that when the annual average wage increases 10,000 JY, the retirement age may be reduced by 5 years for small firms, 5 years for medium firms, and 3 years for large firms. This suggests that the influence of the seniority wage system on mandatory retirement age is greater for smalland medium-sized firms than for large firms in Japan.

8

SENIORITY WAGE AND EMPLOYMENT OF THE OLDER WORKERS …

201

8.5.3 Does the Seniority Wage Influence the Probability of Performing the Elderly Reemployment System in Firms? Does the seniority wage system influence the probability of performing an elderly reemployment system? Does the seniority wage influence the probability of setting the oldest age in the elderly reemployment system? The results obtained using the probit regression model are summarized in Tables 8.4 and 8.5. First, considering the seniority wage system on the probability of performing the elderly reemployment system (Table 8.4), the coefficients are −3.740 for Index 1 and −0.086 for Index 2. They are statistically significant at 5% level. This suggests that seniority wage negatively affects the probability of performing the elderly reemployment system in a firm. When the increased range of wages with age is larger, the probability of performing a reemployment system will be lower. Second, regarding the seniority wage system on the probability of setting the oldest age in the elderly reemployment system (Table 8.5), the coefficients are 1.536 for Index 1, 0.212 for Index 2, and 0.352 for Index 3. They are statistically significant at 10%, 5%, and 1%, respectively. This indicates that when the increased range of wages with age is larger, the probability of setting the oldest age in the elderly reemployment system increases. 8.5.4

Does the Seniority Wage Influence the Reemployment Age?

Table 8.6 reports the results of reemployment age functions using Maddala model. Two kinds of selection items calculated based on the results of Tables 8.4 and 8.5 are used. It is shown that the selection items are statistically significant at 1% or 10% levels and are positive values. It suggests that when the sample selection bias is not considered, the results will be underestimated. The coefficient of Index 1 is −2.405, and it is statistically significant at 1% level. It suggests that when the annual average wage increases 10,000 JY, the retirement age will be reduced by almost 2 years in a firm. The result is consistent with the Lazear (1979) model. Table 8.7 summarizes the results for different firm sizes. The coefficient of Index 1 is −5.404, and it is statistically significant at 5% level for firms with employees more than 300 (large firm), which indicates that when the annual average wage increases 10,000 JY, the older worker’s reemployment age will be reduced by almost 5 years. However,

202

A. SEIKE AND X. MA

Table 8.4 Probability of performing the elderly reemployment system in Japanese firms Index 1 Coef. Seniority wage −3.740** Industry (Manufacturing) Construction 0.052 Wholesale and 0.768*** retail Service −0.089 Other −0.283* Firm operation years 0.001 Firm size (1–99) 100–299 0.178 More than 300 0.555** Trade union 0.189 Average age of 0.287** employees Average age sq. −0.004** Average years of 0.121** tenure Average years of −0.002* tenure sq. Proportion of 0.672 regular employees Proportion of −0.514 female employees Education (Lower than senior high school) College and higher 1.070** Relation between wage and productivity (Wage > productivity) Wage = −0.123 productivity Wage < −0.078 productivity Region Yes Constant −2.821

Index 2

Index 3

z-value

Coef.

z-value

Coef.

z-value

−2.29

−0.086

−0.60

−0.023

−0.12

−0.450** 0.437*

−2.02 1.70

−0.463** 0.429*

−2.08 1.67

−0.48 −1.73 0.38

−0.065 −0.251 0.002

−0.33 −1.46 0.57

−0.063 −0.250 0.002

−0.32 −1.45 0.58

1.34 2.29 1.03 2.40

0.157 0.290 0.126 0.103

1.12 1.34 0.67 1.05

0.155 0.286 0.124 0.102

1.11 1.33 0.66 1.04

−2.41 2.42

−0.001 0.042

−0.93 1.08

−0.001 0.042

−0.92 1.06

−1.61

−0.001

−0.42

−0.001

−0.42

2.16

0.416

1.49

0.411

1.48

−1.31

−0.367

−0.91

−0.363

−0.90

2.29

0.188

1.21

0.163

1.07

−0.81

−0.067

−0.44

−0.064

−0.42

−0.45

0.047

0.26

0.045

0.25

−1.43

Yes −1.248

−0.62

0.17 2.71

Yes −1.255***

−0.62

(continued)

8

SENIORITY WAGE AND EMPLOYMENT OF THE OLDER WORKERS …

203

Table 8.4 (continued) Index 1

Index 2

Coef. Observations Log likelihood Pseudo R 2

1,986 −264.56 0.10

z-value

Coef.

Index 3 z-value

1,796 −236.88 0.09

Coef.

z-value

1,796 −237.04 0.09

Note ***p < 0.01, **p < 0.05, *p < 0.10 Source Author’s creation based on Elderly Employment and Recruitment Survey

the coefficients of index are not statistically significant for firms with employees from 1 to 99 (small firm) and those with employees from 100 to 299 (medium firm). This suggests that the influence of seniority wage on reemployment age is greater for large firms than for small- and medium-sized firms in Japan.

8.6

Conclusions

Using the Elderly Employment and Recruit Survey conducted by the Japan Institute for Labor Policy and Training in September 2008, this study investigates the influences of the seniority wage system on mandatory retirement age and reemployment age of older workers in Japanese firms. The Maddala and 2SLS models were used to address the sample selection bias and endogeneity problems. The main findings are as follows. First, it is shown that the seniority wage influences the employment of older workers. Concretely, for a firm with a larger increased range of wages with age (typical seniority wage structure), the probability of performing a mandatory retirement system will increase, while the probability of performing an elderly reemployment system will become lower. When the annual average wage per employee increases 10,000 JY, the mandatory retirement age will be reduced by 4 years, and the oldest reemployment age will be reduced by 2 years. Second, the influence of the seniority wage system on the employment of older workers differs by firm size. Concretely, (1) when the annual average wage increases 10,000 JY, the mandatory retirement age will be reduced by about 5 years for small firms (firms with employees from 1 to 99), 3 years for medium firms (firms with employees from 100 to 299), and 3 years for large firms (firms with employees more than 300); (2)

204

A. SEIKE AND X. MA

Table 8.5 Probability of setting the oldest age in the elderly reemployment system in Japanese firms Index 1

Seniority wage Starting wage Working hours Industry (Manufacturing) Construction Wholesale and retail Service Other Firm operation years Firm size (1–99) 100–299 More than 300 Trade union Average age of employees Average age sq. Average years of tenure Average years of tenure sq. Proportion of regular employees Proportion of female employees Education (Lower than senior high school) College and higher Relation between wage and productivity (Wage > productivity) Wage = productivity

Index 2

Coef.

t-value

1.536 −1.188 −0.001

0.25 −0.97 −0.26

0.212** −0.905*** 0.000

2.21 −3.63 −0.06

0.352*** −0.797*** 0.000

2.62 −3.19 −0.08

−0.329 −0.099

−0.42 −0.2

−0.127 0.044

−0.86 0.38

−0.132 0.044

−0.90 0.38

−0.034 0.021 −0.003

−0.29 0.14 −0.59

0.035 0.050 −0.005**

0.29 0.45 −2.31

0.034 0.050 −0.005*

0.28 0.46 −2.28

0.274** 0.444 0.086 0.108

Coef.

Index 3 t-value

Coef.

t-value

2.32 0.93 0.49 0.33

0.286*** 0.553*** 0.174* 0.182***

3.31 4.07 1.67 2.70

0.283*** 0.548*** 0.174* 0.180***

3.28 4.03 1.67 2.68

−0.001 −0.005

−0.25 −0.04

−0.002*** 0.032

−2.77 1.39

−0.002*** 0.031

−2.74 1.37

0.000

0.11

0.000

0.305

0.52

0.453**

−0.246

−0.8

−0.123

−0.08

0.103

1.15

−0.388

−0.65

2.45 −1.45

−0.63

0.000 0.446** −0.386

2.41 −1.44

0.175*

1.84

0.166*

1.75

0.083

0.87

0.085

0.89

(continued)

8

SENIORITY WAGE AND EMPLOYMENT OF THE OLDER WORKERS …

205

Table 8.5 (continued) Index 1

Wage < productivity Region Constant Observations Log likelihood Pseudo R 2

Index 2

Index 3

Coef.

t-value

Coef.

t-value

Coef.

t-value

−0.081

−0.81

−0.115

−1.10

−0.116

−1.11

Yes 0.856** 1,836 −799.78 0.06

0.13

Yes −0.833 1,663 −722.33 0.07

−0.51

Yes −1.499 1,663 −721.31 0.07

−0.91

Note ***p < 0.01, **p < 0.05, *p < 0.10 Source Author’s creation based on Elderly Employment and Recruitment Survey

when the annual average wage increases 10,000 JY, the oldest reemployment age will be reduced by 5 years for large firms, but the negative effect on reemployment age was not found for small firms and medium firms. This suggests that the influence of seniority wage on the employment of older workers is greater for large-sized firms than for mediumand small-sized firms. The policy implications can be considered as follows: Although the seniority wage may increase the motivation of workers to work in firms for a long term, the seniority wage system negatively affects the employment of older workers. To address the problem of labor force decreased as the population aging in Japan, to build an “age-free” career society has become an important issue. The wage system reform from the traditional seniority wage to the ability wage (or performance payment) should be considered. However, it should be considered that when the association between wage and age becomes smaller (the wage-age profile becomes flatter), the motivation of employees may decrease, and the monitor cost of firm may increase. Therefore, the transformation of seniority wage to the performance pay system should be established and implemented stepby-step. To motivate workers to work in firms for a long term, other human resource management systems, such as career ability development and fringe benefit systems, should be considered. In addition, it indicated that the negative effect of seniority wage on reemployment age is greater for large-sized firms. In Japan, the traditional seniority wage system and mandatory retirement system are usually implemented in most

206

A. SEIKE AND X. MA

Table 8.6 Results of reemployment age function in Japan Index 1 Coef. Seniority wage Relation between wage and productivity (Wage > productivity) Wage = productivity Wage < productivity Starting wage Working hours Industry sector (Manufacturing) Construction Wholesale and retail Service Other Firm size (1–99) 100–299 More than 300 Trade union Average age of employees Average years of tenure Proportion of regular employees Proportion of female employees Education (Lower than senior high school) College and higher Firm systems Seminar Vocational training Turnover system Early retirement system Support for starts-up Support for job search

Index 2

Index 3

t-value

Coef.

−2.045**

−2.24

0.164

0.174** −0.038 0.201 −0.003

2.40 −0.47 0.43 −0.65

0.159** −0.130 −0.587** −0.001

2.21 −1.54 −2.13 −0.17

0.158** −0.130 −0.492* −0.001

2.20 −1.54 −1.89 −0.19

0.241 0.238** 0.202** 0.112

1.23 2.25 2.19 1.28

−0.214 0.170* 0.253*** 0.131

−1.54 1.82 2.70 1.38

−0.220 0.173* 0.250*** 0.126

−1.56 1.86 2.68 1.32

0.321*** 0.419*** 0.085 0.012

3.49 3.38 1.14 1.11

0.414*** 0.513*** 0.113 0.036***

5.05 4.14 1.45 5.25

0.410*** 0.504*** 0.113 0.036***

5.02 4.10 1.45 5.23

0.006 0.431***

0.82 2.66

0.001 0.440***

0.08 2.59

0.001 0.437***

0.10 2.58

−0.145

t-value 1.03

Coef. 0.234

t-value 1.09

−0.64

−0.367

−1.52

−0.368

−1.53

2.64

0.044

0.54

0.041

0.51

−0.084 0.051 0.075 −0.028

−1.09 0.41 0.37 −0.23

−0.120 0.041 0.031 −0.052

−1.51 0.34 0.15 −0.41

−0.118 0.040 0.033 −0.047

−1.49 0.33 0.16 −0.37

0.141 −0.241***

0.31 −0.68

0.094 −0.314

0.21 −0.90

0.088 −0.312

0.20 −0.90

0.523***

(continued)

8

207

SENIORITY WAGE AND EMPLOYMENT OF THE OLDER WORKERS …

Table 8.6 (continued) Index 1

Index 2

Coef. Region Selection adjust item Item 1 Item 2 Constant Observations Adjusted R 2

t-value

Yes

Coef.

Index 3 t-value

Coef.

Yes

−5.928 6.499*** 64.049*** 1,486 0.068

−1.33 2.73 44.72

2.210 10.065*** 60.316*** 1,348 0.068

t-value

Yes 0.39 5.88 31.38

2.665 9.951*** 59.704*** 1,348 0.069

0.47 5.84 30.38

Note ***p < 0.01, **p < 0.05, *p < 0.10 Source Author’s creation based on Elderly Employment and Recruitment Survey

Table 8.7 Results of reemployment age function by firm size in Japan Index 1

Employees 1–99 Employees 100–299 Employees more than 300

Index 2

Index 3

Coef.

t-value

Coef.

t-value

Coef.

t-value

0.946 −2.084 −5.404**

0.58 −1.22 −2.06

0.377*** 0.082 0.283

2.92 0.68 1.16

0.510*** 0.119 0.358

2.88 0.69 1.12

Note 1. ***p < 0.01, **p < 0.05, *p < 0.10 2. The other variables are similar to those in Table 8.1. They are not expressed in this table Source Author’s creation based on Elderly Employment and Recruitment Survey

large firms; therefore, the wage system reform in large-sized firms should be considered carefully.

Appendix See Tables 8.8 and 8.9.

208

A. SEIKE AND X. MA

Table 8.8 Descriptive statistics of variables Variables

Obs.

Means

S.D.

Max.

Min.

Retirenment age Reemployment age Wage-age curve indices Index1 Index2 Index3 Relation between wage and productivity wage > productivity wage = productivity wage < productivity Working hours weekly Industry sector Construction Manufacturing Wholesale and retail Service Other industry sector Operation years Firm size Employee 1-99 Employee 100-299 Employee more than 300 Trade union Starting wage level Average age of employees Average tenure years Proprotion of regular employees Proprotion of female employees Education attainment Senior high school Higher than college Other education attainment Other systems in firms Holding the seminar Vocational training system Turnover and job search system Advantage for early retirement system Support for turnover system Support for starts-up Regions Hokaidao

3,648 2,794

61 65

2 1

60 60

68 70

2,690 2,469 2,469

0.647 0.642 1.504

0.199 0.457 0.326

−0.086 −0.995 0.500

1.172 6.405 5.357

2,517 2,517 2,517 3,567

0.255 0.488 0.256 38

7

0 0 0 6

1 1 1 48

3,777 3,777 3,777 3,777 3,777 3,754

0.082 0.274 0.197 0.167 0.279 40

22

0 0 0 0 0 1

1 1 1 1 1 267

3,803 3,803 3,803 3,776 3,221 3,581 3,456 3,737 3,611

0.488 0.368 0.144 0.231 19 40 12 0.781 0.184

4 6 6 0.244 0.164

0 0 0 0 9 18 1 0 0

1 1 1 1 53 64 40 1 1

3,582 3,582 3,582

0.615 0.322 0.063

0 0 0

1 1 1

3,809 3,832 3,846 3,867 3,867 3,867

0.146 0.049 0.016 0.047 0.006 0.007

0 0 0 0 0 0

1 1 1 1 1 1

3,828

0.038

0

1

(continued)

8

209

SENIORITY WAGE AND EMPLOYMENT OF THE OLDER WORKERS …

Table 8.8 (continued) Variables Tohoku Kando Hokuliku Tokai Kinki Chiokoku Shikoku Kyushu

Obs.

Means

3,828 3,828 3,828 3,828 3,828 3,828 3,828 3,828

0.074 0.382 0.092 0.149 0.159 0.057 0.024 0.025

S.D.

Max.

Min.

0 0 0 0 0 0 0 0

1 1 1 1 1 1 1 1

Table 8.9 Wage function by age in Japan Age 30 Coef. Wage level at starting point 1.080*** Industry sector (Manufacturing) Construction 1.040** Wholesale and retail 0.738** Service 0.172 Other 0.037 Operation years −0.004 Firm size (1–99) 100–299 −0.102 More than 300 0.230 Trade union −0.186 Average age of employees −0.023 Average age squared −0.002 Average years of tenure 0.199*** Average years of tenure squared −0.003** Proportion of regular employees 1.041** Proportion of female employees −2.268** Education (Lower than senior high school) Higher than college 0.306 Regions Yes Constant 6.578* Observations 2,228 F -values 72.12 Adjusted R 2 0.434

Age 50 t-value

Coef.

t-value

32.98

1.266**

16.93

2.54 2.37 0.52 0.13 −0.82

3.527** 2.264*** 0.203 −0.331 −0.020*

3.79 3.20 0.27 −0.50 −1.63

−0.44 0.69 −0.69 −0.13 −0.74 3.37 −2.10 2.17 −3.19

0.174 1.688** 0.254 1.008** −0.017*** 0.645*** −0.012*** 2.855*** −2.884*

0.33 2.20 0.41 2.50 −3.42 4.75 −3.26 2.57 −1.74

1.28

5.401*** Yes −6.545 2,168 41.04 0.307

9.76

1.86

Source Author’s creation based on Elderly Employment and Recruitment Survey Note ***p < 0.01, **p < 0.05, *p < 0.10

−0.78

210

A. SEIKE AND X. MA

Notes 1. Numerous studies have focused on the determinants of participation in the work of Japanese older adults from the labor supply perspective. It was found that the individual characteristics (e.g., education, gender), family factors (e.g., child, spouse working status), income factors (e.g., household income, wealth, owning housing), and work career in the past affect the participation in the work of elders in Japan. For the survey on the issue, please refer Seike and Ma (2008). 2. Seniority wage is defined as that the wage level rise with age. 3. For the rationality of the seniority wage system, according to human capital theory (Becker 1964; Mincer 1974), wage is determined by the general human capital (e.g., education attainment, etc.) and the special human capital (e.g., job training for a special job). Years of tenure can be thought of as a kind of special human capital; therefore, the wage should be increased as one grows older. Excepting the human capital theory and implicit contract hypothesis (Lazear 1979, 1981), Ono (1989) advocated that the living security hypothesis can explain the seniority wage in Japan and Republic of Korea. He found that the age effect is greater than the tenure-year effect in wage functions. In addition, the job matching model (Jovanovic 1979), tournament hypothesis (Lazear and Rosen 1981), and job itinerary hypothesis (Koike 1966, 1981) can also explain the rationality of seniority wage in firms. 4. According to the zero profit-wage path, even if wages and work (or contributions, labor productivity, etc.) do not match at each point in the employment period, the total wages paid should be equal to the total labor productivity in the total employment period. 5. Concretely, the binary variables of the mandatory retirement system or elderly reemployment system are equal to 1 when a firm performs the system, which is equal to when not. 6. The reasons for the selection of ages 30 and 50 years in the calculations can be considered as follows. First, the age of highest wage level in most Japanese firms is between 45 and 50 years; therefore, we chose 50 years as the age with the highest wage level. Second, because we could not obtain information about the age at which they started to work, we chose the age of 30 years in the survey as the youngest age in the calculations. 7. The manufacturing industry includes five kinds: (i) general machinery and equipment manufacturing, (ii) transportation machinery and equipment manufacturing, (iii) precision machinery and equipment manufacturing, (iv) electrical machinery and equipment manufacturing, and (v) other manufacturing. 8. The other industries include electricity, gas, heat supply, water supply, information and communication, transportation, finance and insurance,

8

SENIORITY WAGE AND EMPLOYMENT OF THE OLDER WORKERS …

211

real estate, restaurant and hotel, medical and care, education, and learning support industries. 9. For the dummy variables of firm size, the total number of firm samples is 3,803 firms, the proportions are 48.80% (1,856 firms) for firms with employees from 1 to 99, 36.76% (1,398 firms) for firms with employees from 100 to 299, 6.57% (250 firms) for firms with employees from 300 to 499, 4.44% (169 firms) for firms with employees from 500 to 999, and 3.42% (130 firms) for firms with employees more than 1,000. The samples of firms with more than 500 employees are smaller; therefore, we conducted three kinds of firm size dummy variables in this study. 10. The firm operation year is calculated as “firm operation years = 2008-firm established year.”

References Becker, G. S. (1964). Human capital: A theoretical and empirical analysis, with special reference to education. New York: Columbia University Press. Higuchi, Y., & Yamamoto, I. (2002). Labor supply behavior mechanism of elderly men in Japan: Analysis of the effect of the pension and wage system and the future image of employment for the elderly. Institute for Financial Research Bank of Japan. Financial Research, 2002(10), 31–77. (In Japanese). Jovanovic, B. (1979). Job matching and the theory of turnover. Journal of Political Economy, 87 (5), 972–990. Kawaguchi, D., & Kanbayashi, L. (2007). Does seniority wages diverge from productivity? Empirical analysis using industrial statistics survey and wage structure statistics survey data. Economics Research, 58(1), 61–90. (In Japanese). Koike, K. (1966). Wage: Theory and analysis. Tokyo: Diamond Press. (In Japanese). Koike, K. (1981). Skill in Japan: Excellent human resources economic system. Tokyo: Toyo Keizai Shiho Press. (In Japanese). Koike, K. (2005). Economics of work. Tokyo: Toyo Keizai Shiho Press. (In Japanese). Kubo, K. (1994). Changes in wage profiles due to aging. In Elderly Employment Development Association (Ed.), Research report on wage system model adapted to aging era. Tokyo: Elderly Employment Development Association. (In Japanese). Lazear, E. P. (1979). Why it there mandatory retirement. Journal of Political Economy, 87 (6), 261–1284. Lazear, E. P. (1981). Agency, earnings profiles, productivity, and hours restrictions. American Economic Review, 71, 606–620.

212

A. SEIKE AND X. MA

Lazear, E. P., & Rosen, S. (1981). Rank-order tournaments as optimum labor contracts. Journal of Political Economy, 89(5), 841–864. Maddala, G. S. (1983). Limited-dependent and qualitative variables in economic. Cambridge: Cambridge University Press. Mincer, J. (1974). Schooling, experience and earning. New York: Columbia University Press. Mitani, N. (2001). Employment policy of the elderly and labor demand. In T. Inoki & F. Otake (Eds.), Economics analyses on employment policy. Tokyo: University of Tokyo Press. (In Japanese). Ohashi, I. (2005). Wage, working hours and job satisfaction. In The trend of modern economics 2005. Tokyo: Toyo Keizai Shiho Press. (In Japanese). Ono, A. (1989). Skill or living security? The determinants of senior wage. In A. Ono (Ed.), Japanese employment practices and labor market. Tokyo: Toyo Keizai Shiho Press. (In Japanese). Seike, A. (1994). Wage system for lifetime active employment. In Development Association (Ed.), Research report on wage system model adapted to aging era. Elderly Employment Development Association. (In Japanese). Seike, A., & Ma, X. (2008). Determinants of elderly male labor participation: 1980–2004. In Japan Institute for Labor Policy and Training (Ed.), Research on employment of the elderly: Interim Report on the promotion of employment for the elderly (Labor Policy Report No. 100). Japan Institute for Labor Policy and Training. (In Japanese). Yamada, A. (2007). Firm response to elderly’s continuous employment obligation: Focusing on wage and annual income adjustment. In Japan Institute for Labor Policy and Training (Ed.), Current status and issues of human resource management for continuous employment of the elderly (Labor Policy Report No. 83). Japan Institute for Labor Policy and Training. (In Japanese).

CHAPTER 9

Living Arrangement and Well-Being of the Middle-Aged and Older Adults in Japan Tsukasa Matsuura and Xinxin Ma

9.1

Introduction

In Japan, population aging has progressed rapidly since the 1970s, and now it has become the country with the highest population aging rate in the world. The aging of the Japanese population is expected to continue to progress in the future. Figure 9.1 shows that in 2015, the aging rate was 26.5% for Japan, while in 2060, it will become 38.1%, which is higher than the other developed countries (e.g., France, UK, USA.) As population aging progresses, elderly single-person households are predicted to increase. Table 9.1 shows the proportion of each type of living arrangement and family structure for the elderly in Japan. For household heads aged 65 and older, the proportion of single-person households was 32.5% in 2015, and it is expected to be 40% in 2040;

T. Matsuura (B) Department of Economics, Chuo University, Tokyo, Japan e-mail: [email protected] X. Ma Faculty of Economics, Hosei University, Tokyo, Japan e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 X. Ma (ed.), Employment, Retirement and Lifestyle in Aging East Asia, Social Policy and Development Studies in East Asia, https://doi.org/10.1007/978-981-16-0554-3_9

213

214

T. MATSUURA AND X. MA

% 45 Japan

40

France

UK

US

35 30 25 20 15 10 5 2060

2055

2045

2050

2035

2040

2030

2020

2025

2015

2010

2005

1995

2000

1985

1990

1980

1975

1970

1965

1955

1960

1950

0

Fig. 9.1 Proportion of older adults: comparison with Japan and other countries (Note The proportions of population aged 65 and older to total population are shown in the figure. Source Author’s creation based on data from World Population Prospects: The 2017 Revision published by United Nations. Japan: 1950–2015 Census data, Ministry of Internal Affairs and Communications, Japan: 2020– 2060 data calculated by National Institute of Population and Social Security Research)

for household heads aged 75 and older, the proportion of single-person households was 37.9% in 2015, and it is expected to be 42.1% in 2040. Can the rising proportion of elderly single-person households be considered a social problem? It is argued that the probability of becoming poor is higher for elderly single-person households than for younger single-person households. For Japan, previous studies have suggested that the rise in elderly single-person households may be a factor of elderly poverty (Sun et al. 2007; Tachinabaki and Urakawa 2009; Tachinabaki 2011; Matsuura 2020). However, when the elderly chooses voluntarily to have a single-person household, it may not become a social problem. It can be considered that, due to various restrictions, when one must live with other family members, he (she) may feel more stress, which may

9

LIVING ARRANGEMENT AND WELL-BEING OF THE MIDDLE-AGED …

215

Table 9.1 Proportions of each types of family structure of household head aged 65 and older and aged 75 and older in Japan (Unit: %) Year

Household Total

Other Single

Nuclear family household Total

Household head aged 65 and older 2015 100.0 32.6 56.3 2020 100.0 34.0 56.0 2025 100.0 35.7 55.1 2030 100.0 37.4 54.0 2035 100.0 39.0 53.0 2040 100.0 40.0 52.4 Household head aged 75 and older 2015 100.0 37.9 51.5 2020 100.0 38.0 53.0 2025 100.0 38.4 53.2 2030 100.0 39.5 52.4 2035 100.0 40.9 51.4 2040 100.0 42.1 50.6

Couple only

Parents and child

Single parent and child

32.7 32.6 32.2 31.5 30.9 30.6

14.9 14.5 13.9 13.4 13.0 13.0

8.7 8.8 9.1 9.2 9.1 8.8

11.1 10.0 9.2 8.5 8.0 7.6

30.8 31.5 31.7 31.2 30.3 29.9

10.9 11.5 11.7 11.4 10.9 10.7

9.8 10.0 9.8 9.9 10.1 10.0

10.6 9.1 8.4 8.0 7.7 7.4

Note Total values are not match due to rounding Source Author’s creation based on data from Future Estimation of the Future Number of Japanese Households (National Estimation) (Estimation in 2018)

hinder well-being. Therefore, someone may prefer to choose to live alone based on his (her) own lifestyle and life values. The choice to voluntarily live alone (live independently) may have a positive effect on an individual’s well-being. Klinenberg (2012) and Jamieson and Simpson (2013) criticized the viewpoint that emphasizes the negative aspects of living alone and pointed out the positive effects of living alone. Ueno (2007) also emphasized the positive effect of living alone for Japanese older adults. How do their living arrangements (e.g., living alone) affect the well-being of elderly in Japan? This chapter employs an empirical study to answer the question. We focus on the influence of living alone on Subjective WellBeing (SWB) measured by happiness using Japanese longitudinal survey data. The features of this study can be summarized as follows: first, most previous studies used cross-sectional data; therefore, there endogeneity and heterogeneity problems may remain in the results. Although some

216

T. MATSUURA AND X. MA

studies used the instrumental variables (IV) method to address the endogeneity problem, it is difficult to find an appropriate instrumental variable that is exogenous. This study uses Japanese longitudinal survey data and the fixed-effects model to address the endogeneity problem caused by individual heterogeneity. Second, this study compares the differences in the effect of living alone on well-being by gender. Life expectancy varies by gender; women live longer than men.1 Due to gender gaps in the labor market and pension benefits, the proportion of co-residence with a child is higher for women than men after separation or bereavement of a spouse. Therefore, the impact of living alone on well-being can be considered to differ by gender. Especially in Japan, it is predicted that the number of elderly single-person households will increase rapidly in the future; thus, it is meaningful to investigate the difference by gender. The remainder of this chapter is organized as follows. Section 9.2 summarizes the results of previous studies regarding living arrangements and well-being. Section 9.3 explains the methodology, including the models and data. Section 9.4 presents the estimation results, and the discussion and conclusions are presented in Sect. 9.5.

9.2

Literature Review

Regarding the issue analyzed in this study, we mainly summarize the results of empirical studies on the impact of living arrangements on the well-being of the Japanese elderly in the following.2 Brown et al. (2002) used a panel survey data to estimate the determinants of health status. He found that a person’s living arrangement three years earlier may affect health status. The influence of living alone on mortality after three years is not statistically significant, but the mortality after three years is significantly higher for the group living alone. The study by Raymo et al. (2008) is most similar with this study, as they investigated the influences of living arrangements and spouse status on well-being (self-rated health: SRH, life satisfaction) and compared the differences by gender. They found that marriage enhances both SRH and life satisfaction for men, but the positive effect of marriage (having a spouse) on well-being is not confirmed for women. Oshio (2012) also targeted elderly Japanese and investigated the impact of living arrangements on well-being by gender, which is similar to this current study. He found that when co-residence with a child or parents was controlled, living with a spouse may enhance life satisfaction for men, while the

9

LIVING ARRANGEMENT AND WELL-BEING OF THE MIDDLE-AGED …

217

influence of living with a spouse on well-being is not statistically significant for women. He also found that elders living with an unmarried son are likely to feel dissatisfied. Using cross-sectional data of the Japanese General Social Surveys (JGSS), Raymo (2015) focused on adults aged 20–39 years, and found that, as compared with the cohabitation group, individuals in the group living alone are likely to feel unhappy, while the effect of living alone on well-being is not statistically significant when social capital factors such as the frequency of communication with friends were controlled. However, the married group is likely to feel much happier, even when social capital factors such as the frequency of communication with friends are controlled. Although some of the empirical studies mentioned above focused on Japan, some issues should be investigated. For example, most previous studies used cross-sectional data; therefore, the heterogeneity problem may persist in the results. No empirical study has compared the results by gender. This current study can fill this void.

9.3

Methodology and Data 9.3.1

Model

First, we use ordered probit regression model to investigate the association between living arrangement and happiness of middle-aged and older Japanese adults. For the correlations of errors of individuals, clusterrobust standard errors are used.3 The model is expressed by Eqs. 9.1.1 and 9.1.2: Yi = a + βAlonei + γ X i + vi Yt = 0

(9.1.1)

if Yi∗ ≤ 0

= 1 if 0 ≤ Yi∗ ≤ μ1 = 2 if μ1 ≤ Yi∗ ≤ μ2 .. . =N

if μ N −1 ≤ Yi∗

(9.1.2)

In Eq. 9.1.1, Yit expresses well-being (happiness), Alone is the dummy variable of living alone, and β denotes the coefficient of Alone, which indicates the impact of living alone on an individual’s well-being. β is

218

T. MATSUURA AND X. MA

an important quantity in this study. X it expresses a set of control variables including age, health status, household income, and marital status (having a spouse); we use different control variables in analysis. ci is an item related to individual-specific and time-invariant factors, and vit is the error. In Eq. 9.1.2, μ is the rank of happiness (N ranges 0–10). Yi∗ is the expected probability of an individual’s happiness. Second, because vit includes the error related to individual-specific and time-invariant factors, there may be a heterogeneity problem in the results using the ordered probit regression model. To address the heterogeneity problem, we use a random effects probit model. The model is expressed by Eq. 9.2: Yit = a + βAloneit + γ X it + ci + u it

(9.2)

In Eq. 9.2, ci is an item related to individual-specific and time-invariant factors, and u it is the idiosyncratic error. In Eq. 9.2, all effects are random—apart from possibly an overall intercept term a, the observations are no longer uncorrelated, but instead have a covariance that depends on the variance of the random effects. Using the random effects model, we can address the individual heterogeneity (ci ) and obtain the real error (u it ). Regarding how the impact of living alone on well-being may differ by age, we employ the estimations for two age groups—(i) the group aged 45 and older; (ii) the group aged 60 and older—and compare the results for these two groups. 9.3.2

Data and Variable Setting

Seven waves of longitudinal data were used from the Japanese Household Panel Survey (JHPS) and the Keio Household Panel Survey (KHPS), which was conducted by Keio University in Japan from 2011 to 2018. The samples of middle-aged and older adults aged 45 and above in the baseline survey are used. Missing values are excluded. The total samples included 16,091 women, and 15,263 men. The dependent variable is an individual’s well-being status. Two kinds of variables are used: (i) happiness in the past one-year and (ii) lifetime happiness. The results for happiness in the past year are used as the basic analysis, while the results for lifetime happiness are used for a robustness check shown in Tables 9.7, 9.8, 9.9, and 9.10. For the ordered probit

9

LIVING ARRANGEMENT AND WELL-BEING OF THE MIDDLE-AGED …

219

regression model shown in Eqs. 9.1.1 and 9.1.2, the ordered category variable is treated as the dependent variable using a scale of values from 0 to 11 (0 = never happy, 11 = very happy); for the random effects model shown by Eq. 9.2, we conducted a happiness binary variable that is equal to 1 when the value of happiness is from to 6–11, equal to 0 when the values are from to 0–5. The main independent variable is the dummy variable of living alone. Using the questionnaire item of the number of family members, including the respondent him (her) self, it is equal to 1 when the number of family members is one and equal to 0 when the number of family member is two or more. Variables of age, household income,4 health status, and having a spouse5 are also constructed. Objective health status variables are constructed as follows: based on the questionnaire item, the number of diseases is calculated based on options including (a) headache and dizziness; (b) palpitations or shortness of breath; (c) something wrong with stomach and intestines; (d) back, waist, and shoulders hurt; (e) tire easily; and (f) catch a cold more easily. It is equal to 1 when the answer is “often” or “occasionally” and otherwise is equal to 0. The total values of these items are used as health status indices for China and Japan, separately. The descriptive statistics for individuals aged 45 and above are shown in Table 9.2. The proportion of individuals living alone is 11.6% for women, slightly higher than for men (9.8%). The average age is 61.5 years old for women and men. The average household income is around 5 million Japanese Yen (JY). The proportion living with a spouse is 74.6% for women and 84.6% for men—slightly higher for men than women. The Table 9.2 Descriptive statistics of variables

Living alone Age Income (10 thousand JY) Health status Spouse

Observations

Women Mean

S.D.

Observations

Men Mean

S.D.

16,091 16,091 13,656

0.116 61.539 488.779

0.320 10.095 354.115

15,263 15,263 13,806

0.098 61.506 516.102

0.298 10.128 349.323

16,091 16,091

2.497 0.746

1.573 0.435

15,263 15,263

2.591 0.846

1.688 0.361

Source Author’s creation based on JHPS/KHPS 2011–2018

220

T. MATSUURA AND X. MA

proportion of people living with a spouse is lower for women than men, which may be caused by women’s longer life expectancy. Figures 9.2 and 9.3 show the results of living alone and well-being of middle aged and older women and men. Two groups—individuals aged 45 and older, and those aged 60 and older—are calculated separately. The proportion of women who said they were “very happy” (selecting 10 on a scale of 0–11) is higher for individuals living alone (7.2% for individuals aged 45 and above, 7.73% for individuals aged 60 and older) than for those not living alone (5.52% for individuals aged 45 and older, 5.70% for individuals aged 60 and older). In addition, the proportion of individuals who answered “very unhappy” (selecting 0 on a scale of 0– 11) is higher for individuals living alone (2.1% for individuals aged 45 and older, 2.45% for individuals aged 60 and older) than for those not living alone (1.17% for individuals aged 45 and older, 0.92% for individuals aged 60 and older). This suggests that the correlation between living alone and well-being is polarizing for Japanese women, which may be caused by the heterogeneity of individuals. For example, when Japanese women are living alone (living independently), some of them report being the happiest, while some report being the least happy. Therefore, it is necessary to address the individual heterogeneity problem in econometric analyses. For men, the proportion of individuals who reported being “very happy” is lower for individuals living alone (2.66% for individuals aged 45 and older, 3.09% for individuals aged 60 and older) than for those not living alone (5.09% for individuals aged 45 and older, 5.93% for individuals aged 60 and older), while the proportion of individuals who answered “very unhappy” (selecting 0 on a scale of 0–11) is higher than for those living alone (3.33% for individuals aged 45 and older, 4.12% for individuals aged 60 and older) than for those not living alone (1.21% for individuals aged 45 and older, 1.06% for individuals aged 60 and older). These results indicate that for men, a linear correlation remains between living alone and well-being, but for women, the correlation is likely obscured by polarization. It should be noticed that these crosstabulation results did not control other factors that may affect individuals’ well-being, such as age, household income, health status, and marital status. The results considering the influences of these factors are shown in the following section.

10(Happy)

9

8

Alone Aged45

7

6

Alone Aged60

5

4

3

Non-Alone Aged45

2

1

0(Unhappy)

Non-Alone Aged60

Fig. 9.2 Life satisfaction by living arrangement type for Japanese women (Source Author’s creation)

0

5

10

15

20

25

30

9 LIVING ARRANGEMENT AND WELL-BEING OF THE MIDDLE-AGED …

221

10(Happy)

9

Alone Aged45

8

7

6

Alone Aged60

5

4

Non-Alone Aged45

3

2

1

Non-Alone Aged60

Fig. 9.3 Life satisfaction by living arrangement type for Japanese men (Source Author’s creation)

0

5

10

15

20

25

30

35

0(Unhappy)

222 T. MATSUURA AND X. MA

9

LIVING ARRANGEMENT AND WELL-BEING OF THE MIDDLE-AGED …

9.4

223

Results

In this section, we report the results of the econometric analysis of the determinants of well-being of middle-aged and older Japanese individuals. The results for the individuals aged 45 and older are shown in Tables 9.3 and 9.4, and the results for the individuals aged 60 and older are shown in Tables 9.5 and 9.6. Four models are used, distinguished by the dependent variables used: Models 1 and 5 only use the living-alone dummy variable; Models 2 and 6 use the age variable as a controlled variable; Models 3 and 7 use age, household income, and health status as controlled variables; and Models 4 and 8 use age, household income, health status, and spouse variables as controlled variables. Using different controlled variables allows us to compare the differences in the effects of living alone among these models. We summarize the results by gender (Models 1– 4 for women, Models 5–8 for men) and focus on the results for living arrangement as follows. 9.4.1

Results for Individuals Aged 45 and Older

First, regarding the results of Table 9.3, it is indicated that the influence of living alone on the well-being of middle-aged and older individuals also differs by gender. Specifically, based on the results of Models 1 and 5 in Table 9.3, the coefficients of living alone on happiness are not statistically significant for women aged 45 and older, whereas they are significantly negative for men. This suggests that when living alone, a Japanese man may feel unhappiness, but a Japanese woman may not. However, it should be noted that other factors that may affect happiness, such as age, income, health status, and spouse, were not controlled in these results. The results after controlling for the other factors can be summarized as follows: for women (see Models 3–4), the coefficients of living alone are positive and statistically significant. These results suggest that, for women aged 45 and older, when other factors are held constant, individuals living alone are likely to feel happiness. Contrariwise, for men (see Models 6–7), the coefficients of living alone have significant negative values, which indicates that, when other factors are held constant, middle-aged and older men living alone are likely to feel unhappiness. This shows that having a spouse influences well-being, and the effects differ by gender. For example, for men, when controlling for having a spouse and other factors (see Model 8), the coefficient of living alone

−0.018 (0.0518)

−27970.60 13,657

16,091

0.111* (0.0573) 0.010*** (0.0017) 0.195*** (0.0218) 0.027*** (0.0060)

(3)

−33200.32

−0.069 (0.0531) 0.006*** (0.0016)

(2)

Note 1. Clustered robust standard errors in parentheses 2. ***p < 0.01, **p < 0.05, *p < 0.10 Source Author’s creation based on JHPS/KHPS 2011–2018

Log −33230.98 likelihood Observations 16,091

Household income Health status Spouse

Age

Living alone

(1) Female Male

(5)

(6)

(7)

(8)

13,657

15,263

15,263

13,806

13,806

0.248*** −0.445*** −0.451*** −0.263*** 0.030 (0.0662) (0.0567) (0.0563) (0.0572) (0.0753) 0.010*** 0.010*** 0.015*** 0.014*** (0.0017) (0.0016) (0.0017) (0.0017) 0.173*** 0.291*** 0.270*** (0.0216) (0.0224) (0.0225) 0.027*** 0.019*** 0.020*** (0.0060) (0.0056) (0.0056) 0.197*** 0.382*** (0.0495) (0.0653) −27941.15 −31459.58 −31377.85 −28076.45 −28009.02

(4)

Table 9.3 Results of living alone and well-being of Japanese individuals aged 45 and older using ordered probit regression model (Dependent variable: Happiness in the past one year)

224 T. MATSUURA AND X. MA

−0.058 (0.0706)

−7364.10 13,657 2,701

16,091 2,849

0.048 (0.0790) 0.005* (0.0029) 0.198*** (0.0292) 0.025*** (0.0097)

(3)

−8631.97

−0.069 (0.0721) 0.002 (0.0027)

(2)

Note 1. Clustered robust standard errors in parentheses 2. ***p < 0.01, **p < 0.05, *p < 0.10 Source Author’s creation based on JHPS/KHPS 2011–2018

Log −8632.22 likelihood Observations 16,091 Number of 2,849 groups

Household income Health status Spouse

Age

Living alone

(1) Female Male

(5)

(6)

(7)

(8)

13,657 2,701

15,263 2,688

15,263 2,688

13,806 2,611

13,806 2,611

0.226** −0.530*** −0.543*** −0.361*** 0.076 (0.0894) (0.0855) (0.0850) (0.0863) (0.104) 0.006** 0.012*** 0.016*** 0.014*** (0.0029) (0.0028) (0.0028) (0.0028) 0.177*** 0.322*** 0.301*** (0.0296) (0.0310) (0.0311) 0.026*** 0.006 0.007 (0.0097) (0.0090) (0.0090) 0.328*** 0.752*** (0.0768) (0.0989) −7354.99 −8149.51 −8140.80 −7390.83 −7361.38

(4)

Table 9.4 Results of living alone and well-being of Japanese individuals aged 45 and older using random effects probit model (Dependent variable: Happiness in the past one year)

9 LIVING ARRANGEMENT AND WELL-BEING OF THE MIDDLE-AGED …

225

−0.064 (0.0591)

−14955.38 7,373

8,975

0.110* (0.0661) 0.007* (0.0036) 0.198*** (0.0296) 0.018** (0.0078)

(3)

−18350.49

−0.078 (0.0604) 0.004 (0.0034)

(2)

Note 1. Clustered robust standard errors in parentheses 2. ***p < 0.01, **p < 0.05, *p < 0.10 Source Author’s creation based on JHPS/KHPS 2011–2018

Log 18353.14 likelihood Observations 8,975

Household income Health status Spouse

Living alone Age

(1) Female Male

(5)

(6)

(7)

(8)

7,373

8,544

8,544

7,658

7,658

0.250*** −0.470*** −0.463*** −0.244*** 0.016 (0.0800) (0.0782) (0.0776) (0.0785) (0.124) 0.008** 0.011*** 0.016*** 0.016*** (0.0036) (0.0033) (0.0034) (0.0034) 0.188*** 0.301*** 0.292*** (0.0296) (0.0304) (0.0308) 0.018** 0.007 0.008 (0.0079) (0.0073) (0.0074) 0.187*** 0.298*** (0.0640) (0.1080) −14941.11 −17462.51 −17445.02 −15480.28 −15465.89

(4)

Table 9.5 Results of living alone and well-being of Japanese individuals aged 60 and older using ordered probit regression model (Dependent variable: Happiness in the past one year)

226 T. MATSUURA AND X. MA

−0.121 (0.0819)

−3946.66 7,373 1,600

8,975 1,700

0.036 (0.0941) −0.007 (0.0060) 0.210*** (0.0415) 0.013 (0.0131)

(3)

−4794.84

−0.096 (0.0836) −0.008 (0.0054)

(2)

Note 1. Clustered robust standard errors in parentheses 2. ***p < 0.01, **p < 0.05, *p < 0.10 Source Author’s creation based on JHPS/KHPS 2011–2018

Log −4796.05 likelihood Observations 8,975 Number of 1,700 groups

Household income Health status Spouse

Living alone Age

(1) Female Male

(5)

(6)

(7)

(8)

7,373 1,600

8,544 1,598

8,544 1,598

7,658 1,549

7,658 1,549

0.240** −0.801*** −0.800*** −0.556*** −0.131 (0.1140) (0.1240) (0.1240) (0.1270) (0.1730) −0.005 0.010* 0.020*** 0.019*** (0.0060) (0.0060) (0.0060) (0.006) 0.195*** 0.365*** 0.355*** (0.0417) (0.0454) (0.0455) 0.013 −0.008 −0.007 (0.0131) (0.0128) (0.0128) 0.330*** 0.591*** (0.1030) (0.1650) −3941.48 −4442.74 −4441.11 −3961.04 −3954.61

(4)

Table 9.6 Results of living alone and well-being of Japanese individuals aged 60 and older using random effects probit model (Dependent variable: Happiness in the past one year)

9 LIVING ARRANGEMENT AND WELL-BEING OF THE MIDDLE-AGED …

227

228

T. MATSUURA AND X. MA

had a positive value and was not statistically significant. The coefficients of the variable for having a spouse (Model 4 for women, Model 8 for men) are significant for both women and men, but the effect of having a spouse on well-being is greater for men than for women. These results can be summarized as follows: For women aged 45 and older, when other factors are held constant, women living alone are likely to feel happiness, while men living alone are likely to feel unhappiness. Living alone negatively affects the well-being of men aged 45 and older, and spouse status (having a spouse or not) significantly affects the association between living alone and happiness. Second, regarding the heterogeneity problem, the results obtained using the random effects probit model are presented in Table 9.4. These results are similar to those obtained using the ordered probit regression model (see Table 9.3). Concretely, for women, when the other factors are not controlled, the coefficient of living alone is not statistically significant, but when the other factors are held constant (see Model 4), the coefficient becomes positive and statistically significant, which suggests that women live alone are likely to feel happiness. Contrariwise, for men, the coefficients of living alone are negative and statistically significant (see Models 5–7), which indicates that men living alone are likely to feel unhappiness. It should be noted that for men, when spouse status was controlled, the coefficient of living alone was not statistically significant, suggesting that the influence of having a spouse is significant for men aged 45 and older. We use lifetime happiness as the dependent variable for a robustness check. These results are summarized in Tables 9.7 and 9.8. These results are consistent with those using the past one-year happiness as the dependent variable shown in Tables 9.3 and 9.4. 9.4.2

Results for Individuals Aged 60 and Older

We then focused on the older age group (individuals aged 60 and older). The results are presented in Tables 9.5 and 9.6. The main findings are summarized as follows: First, regarding the results of the ordered probit regression model (Table 9.5), for women, the coefficients of living alone in Models 3 and 4 are significant and positive, which indicates that for older women, when other factors such as socioeconomic status factors and spouse status are held constant, living alone may have significantly positive effects on their well-being. These results are consistent with the findings for women aged

9

LIVING ARRANGEMENT AND WELL-BEING OF THE MIDDLE-AGED …

229

45 years and older (see Table 9.3). This suggests that when we focus on the group of older women, we again find that women who live alone are likely to feel happiness. For men, the coefficients of living alone in Models 5–7 have significantly negative values, confirming again that for older men, living alone may have a significantly negative effect on their well-being. However, when we controlled for spouse status (see Model 8), the coefficient of living alone was not statistically significant. The results of having a spouse are similar to those in Table 9.3. In addition, although having a spouse positively affects the well-being of both women and men, the effect is greater for men than for women. Second, the results using the random effects probit model are presented in Table 9.6. These results are similar to those obtained using the ordered probit regression model (Table 9.5). More specifically, for women, when the other factors are controlled (see Model 4), the coefficient of living alone has a positive, statistically significant value, indicating that older women living alone are likely to feel happiness. In contrast, for older men, the coefficients of living alone are negative and statistically significant (see Models 5–7), which indicates that older men live alone are likely to feel unhappiness. When controlling for spouse status, the coefficient of living alone was not statistically significant, suggesting that the influence of having a spouse is significant for older men. In addition, although the tendencies of these results are similar with those of men aged 45 and older, when we make comparisons with the group aged 45 and older, the positive effect of living alone on well-being is smaller for older women, while the negative effect is greater for older men. We also use lifetime happiness as the dependent variable for a robustness check. These results are summarized in Tables 9.9 and 9.10. These results are consistent with those using the past one-year happiness as the dependent variable shown in Tables 9.5 and 9.6.

9.5

Discussion and Conclusion

Using longitudinal data from the Japanese Household Panel Survey (JHPS) and the Keio Household Panel Survey (KHPS) from 2011 to 2018, this study investigates the influence of living alone on the wellbeing of the middle-aged and older adults in Japan. The main findings are as follows. First, for women, those living alone are likely to feel happy significantly for both age groups—aged 45 and above and aged 60 and above. This

230

T. MATSUURA AND X. MA

suggests that the positive effect of living alone on the well-being of older women is greater. Second, living alone negatively affects the happiness of men aged 60 and above. However, the negative effect of living alone disappears when marital status is controlled. The results indicate that the positive effect of having a spouse on well-being is greater for men. It is clear that in Japan the influence of living alone on the well-being of the elderly differs by gender. Let us consider the results of this study as compared with those of previous studies. First, Ishigawa (1999) analyzed the mortality rate by marital status based on the life table and found the average life expectancy is longer for men with spouses than for men without spouses (unmarried, separated, bereaved).6 He argued that the results are related to improvements in diet and health and physical and mental stability due to marriage. The results of this study for men support this explanation; it can be assumed that the positive effect of having a spouse on well-being is caused by improved mental stability. Second, living alone positively affects women’s well-being. This result is consistent with that of Raymo (2015), who pointed out that there are two reasons for these results: first, women have a larger social network than men. Second, co-residence may increase the responsibility for housework, which may reduce the well-being of Japanese women. Living with their child may increase stress. Using the Japanese panel survey data, Urakawa and Matsuura (2007) found that living with a parent negatively affects the life satisfaction of women. However, it should be noticed that the results in Fig. 9.2 show a polarized relationship between living alone and well-being for women, suggesting that living alone has two effects on the well-being of Japanese women: (i) it may reduce the stress of co-residence (positive effect); (ii) it may increase the probability of becoming impoverished, which reduces well-being (negative effect). Detailed research regarding these two effects should be conducted in the future. This study may deepen the discussion and understanding of some issues of family economics. Most previous studies found that the number of children negatively affects parents’ happiness or life satisfaction (Hansen, 2012; Nelson et al., 2014). These results suggest a puzzle—why do people have children even when it may reduce their well-being? The intergenerational support hypothesis may explain it. Concretely, parents invest in their children as a good investment during their young and middle-age years, expecting to be supported by their children when they

9

LIVING ARRANGEMENT AND WELL-BEING OF THE MIDDLE-AGED …

231

grow old. When the results indicate that living alone negatively affects the well-being of the elderly, this suggests that the intergenerational support hypothesis is indirectly supported—although caring for children negatively affects an individual’s well-being during the young and middleaged years, in order to avoid the negative impact of living alone in their elderly years, they may choose to have a child or have more children. The results of this study indicated that living alone positively affects the well-being of Japanese women. This suggests that the intergenerational support hypothesis is not supported for Japanese women. Policy implications based on these results can be considered as follows. In general, the increase of elderly single-person households is often argued to be a problem because it is associated with poverty and the isolation of the elderly (Yeung and Cheung 2015). However, singleperson households of middle-aged adults and older adults may be not a severe problem. As has been pointed out by Klinenberg (2012) and Jamieson and Simpson (2013), living alone (or becoming a single-person household) may be the result of an individual’s voluntary choice, and it may positively affect an individual’s well-being. The results in this current study suggested that living alone positively affects the happiness of Japanese women. However, we should be cautious to argue the positive effects of living alone by two reasons: first, the proportion of households receiving welfare in Japan is higher for the elderly single households, because the poverty ratio is higher for elderly single-person households than for other groups. Second, the proportion of those who reported themselves to be “never happy” is higher for those living alone, even among Japanese women (see Fig. 9.2).

Appendix See Tables 9.7, 9.8, 9.9, and 9.10.

−0.059 (0.0501)

−26558.92 13,657

16,091

0.120** (0.0564) 0.007*** (0.0018) 0.229*** (0.0239) 0.025*** (0.0058)

(3)

−31643.00

−0.079 (0.0520) 0.003 (0.0017)

(2)

Note 1. Clustered robust standard errors in parentheses 2. ***p < 0.01, **p < 0.05, *p < 0.10 Source Author’s creation based on JHPS/KHPS 2011–2018

Log −31647.65 likelihood Observations 16,091

Household income Health status Spouse

Age

Living alone

(1) Female Male

(5)

(6)

(7)

(8)

13,657

15,263

15,263

13,806

13,806

0.280*** −0.447*** −0.451*** −0.257*** 0.062 (0.0673) (0.0590) (0.0584) (0.0610) (0.0801) 0.007*** 0.008*** 0.012*** 0.010*** (0.0018) (0.0016) (0.0017) (0.0017) 0.204*** 0.285*** 0.262*** (0.0236) (0.0251) (0.0256) 0.025*** 0.022*** 0.023*** (0.0058) (0.0056) (0.0056) 0.231*** 0.417*** (0.0532) (0.0672) −26518.83 −29849.42 −29805.80 −26635.02 −26555.44

(4)

Table 9.7 Results of living alone and well-being of Japanese individuals aged 45 and older using ordered probit regression model (Dependent variable: Lifetime happiness)

232 T. MATSUURA AND X. MA

0.112 (0.0784)

−6734.61 13,657 2,701

16,091 2,849

0.214** (0.0876) −0.001 (0.0033) 0.212*** (0.0315) 0.032*** (0.0104)

(3)

−7911.29

0.141* (0.0801) −0.006* (0.0031)

(2)

Note 1. Clustered robust standard errors in parentheses 2. ***p < 0.01, **p < 0.05, *p < 0.10 Source Author’s creation based on JHPS/KHPS 2011–2018

Log −7912.97 likelihood Observations 16,091 2,849 Number of groups

Household income Health status Spouse

Age

Living alone

(1) Female Male

(5)

(6)

(7)

(8)

13,657 2,701

15,263 2,688

15,263 2,688

13,806 2,611

13,806 2,611

0.438*** −0.547*** −0.556*** −0.399*** 0.086 (0.0990) (0.0881) (0.0878) (0.0900) (0.1100) −0.001 0.009*** 0.011*** 0.010*** (0.0033) (0.0029) (0.0030) (0.0030) 0.188*** 0.269*** 0.249*** (0.0318) (0.0305) (0.0305) 0.032*** 0.010 0.012 (0.0104) (0.0093) (0.0093) 0.426*** 0.799*** (0.0871) (0.1020) −6722.70 −7779.95 −7775.55 −7028.31 −6997.40

(4)

Table 9.8 Results of living alone and well-being of Japanese individuals aged 45 and older using random effects probit model (Dependent variable: Lifetime happiness)

9 LIVING ARRANGEMENT AND WELL-BEING OF THE MIDDLE-AGED …

233

−0.075 (0.0553)

−14290.82 7,373

8,975

0.098 (0.0625) 0.006* (0.0037) 0.207*** (0.0317) 0.021*** (0.0077)

(3)

−17559.31

−0.085 (0.0571) 0.003 (0.0036)

(2)

Note 1. Clustered robust standard errors in parentheses 2. ***p < 0.01, **p < 0.05, *p < 0.10 Source Author’s creation based on JHPS/KHPS 2011–2018

Log −17560.79 likelihood Observations 8,975

Household income Health status Spouse

Living alone Age

(1) Female Male

(5)

(6)

(7)

(8)

7,373

8,544

8,544

7,658

7,658

0.214*** −0.422*** −0.415*** −0.205** 0.050 (0.0774) (0.0780) (0.0769) (0.0817) (0.1380) 0.007* 0.012*** 0.017*** 0.017*** (0.0038) (0.0033) (0.0034) (0.0034) 0.199*** 0.269*** 0.260*** (0.0317) (0.0374) (0.0381) 0.021*** 0.004 0.005 (0.0077) (0.0074) (0.0074) 0.155** 0.293** (0.0669) (0.1180) −14281.04 −16645.48 −16623.36 −14747.45 −14733.58

(4)

Table 9.9 Results of living alone and well-being of Japanese individuals aged 60 and older using ordered probit regression model (Dependent variable: Lifetime happiness)

234 T. MATSUURA AND X. MA

0.082 (0.0898)

−3702.31 7,373 1,600

8,975 1,700

0.199* (0.1020) −0.006 (0.0065) 0.227*** (0.0436) 0.026* (0.0137)

(3)

−4496.35

0.115 (0.0916) −0.011* (0.0060)

(2)

Note 1. Clustered robust standard errors in parentheses 2. ***p < 0.01, **p < 0.05, *p < 0.10 Source Author’s creation based on JHPS/KHPS 2011–2018

Log −4498.10 likelihood Observations 8,975 Number of 1,700 groups

Household income Health status Spouse

Living alone Age

(1) Female Male

(5)

(6)

(7)

(8)

7,373 1,600

8,544 1,598

8,544 1,598

7,658 1,549

7,658 1,549

0.396*** −0.709*** −0.706*** −0.531*** −0.117 (0.1230) (0.1200) (0.1190) (0.1250) (0.1740) −0.004 0.014** 0.020*** 0.019*** (0.0065) (0.0055) (0.0059) (0.0059) 0.214*** 0.265*** 0.255*** (0.0438) (0.0425) (0.0426) 0.025* −0.001 −0.001 (0.0137) (0.0127) (0.0128) 0.327*** 0.556*** (0.1130) (0.1640) −3698.13 −4338.15 −4335.09 −3865.22 −3859.45

(4)

Table 9.10 Results of living alone and well-being of Japanese individuals aged 60 and older using random effects probit model (Dependent variable: Lifetime happiness)

9 LIVING ARRANGEMENT AND WELL-BEING OF THE MIDDLE-AGED …

235

236

T. MATSUURA AND X. MA

Notes 1. Based on data from the WHO in 2018, the average life expectancy is 75.0 years for Chinese men, 80.2 years for Chinese women, 81.25 years for Japanese men, and 87.32 years for Japanese women. 2. There are several empirical studies on the issue for other countries. For, example, using the data of the elderly in Vietnam, Yamada, and Teerawichitchainan (2015) analyzed living alone and the well-being of the elderly in Vietnam and compared the differences by gender. They found that intergenerational co-residence significantly increases the psychological wellbeing of older adults in Vietnam; for men, co-residence positively affects well-being, while it is not significant when including the quasi-co-residence samples; for women, the influence of co-residence on well-being is not statistically significant, but co-residence positively affects well-being when including the quasi-co-residence samples. 3. The fixed-effects model using cluster-robust standard errors can address the heterogeneity problem, but the individual-specific and time-invariant factors such as educational attainment cannot be estimated. 4. Household income is adjusted by CPI (consumption price indices) for China and Japan. 5. The having a spouse dummy variable is equal to 1 when the individual has a spouse; otherwise, it is equal to 0. 6. For women, the difference between being unmarried and married (with a spouse, separated from a spouse, bereaved of a spouse) is greater than that between the group with spouses and the group without spouses. However, the difference between the unmarried and married groups has become smaller in the current period.

References Brown, J. W., Liang, J., Krause, N., Akiyama, H., Sugisawa, H., & Fukaya, T. (2002). Transitions in living arrangements among elders in Japan: Does health make a difference? Journal of Gerontology: Social Sciences, 57, S.209–S.220. Hansen, T. (2012). Parenthood and happiness: A review of folk theories versus empirical evidence. Social Indicator Research, 108(1), 29–64. Ishigawa, A. (1999). Marital life table: 1995. Journal of Population Problems, 55(1), 35–60. (In Japanese). Jamieson, L., & Simpson, R. (2013). Living alone: Globalization, identity, and belonging. Basingstone: Palgrave Macmillan. Klinenberg, E. (2012). Going solo: The extraordinary rise and surprising appeal of living alone. New York: Penguin Press. Matsuura, T. (2020). Modern population economics. Tokyo: Nippon Hyoron Sha Co. Ltd. (In Japanese).

9

LIVING ARRANGEMENT AND WELL-BEING OF THE MIDDLE-AGED …

237

Nelson, S. K., Kushlev, K., & Lyubomirsky, S. (2014). The pains and pleasures of parenting: When, why, and how is parenthood associated with more or less well-being? Psychological Bulletin, 140(3), 846–859. Oshio, T. (2012). Gender differences in the associations of life satisfaction with family and social relations among the Japanese elderly. Journal of Cross Cultural Gerontology, 27, 259–274. Raymo, J. M. (2015). Living alone in Japan: Relationships with happiness and health. Demographic Research, 46, 1267–1298. Raymo, J. M., Kikuzawa, S., Liang, J., & Kobayashi, E. (2008). Family structure and well-being at older ages in Japan. Journal of Population Research, 25(3), 379–400. Sun, W., Watanabe, M., Tanimoto, Y., Shibutani, T., Kono, R., Saito, M., et al. (2007). Factors associated with good self-rated health of nondisabled elderly living alone in Japan: A cross-sectional study. BMC Public Health, 7, 297–305. Tachibanaki, T. (2011). The identity of an unrelated society: How the blood relations, ground relations, and social relations collapsed. Tokyo: PHP Institute. (In Japanese). Tachinabaki, T., & Urakawa, K. (2009). Japanese poverty research. Tokyo: University of Tokyo Press. (In Japanese). Ueno, C. (2007). Old age of one person. Tokyo: T Houken CORP. (In Japanese). Urakawa, K., & Matsuura, T. (2007). Effect of relative disparity on life satisfaction: An analysis based on Panel Survey on Consumer Life. Japanese Journal of Research on Household Economics, 73, 61–70. (In Japanese). Yamada, K., & Teerawichitchainan, B. (2015). Living arrangements and psychological well-being of the older adults after the economic transition in Vietnam. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 70(6), 957–968. Yeung, W. J., & Cheung, A. K. (2015). Living alone: One-person households in Asia. Demographic Research, 32, 1099–1112.

PART III

Population Aging, Work, and Lifestyles of the Elderly in Other East Asian Regions

CHAPTER 10

Population Aging, Labor Force Participation, and Family Structure in the Republic of Korea Toru Suzuki

10.1

Introduction

East Asian economies including the Republic of Korea (hereafter, abbreviated as Korea) have been the top runners of the world in both economic and demographic changes since the late twentieth century. Korea started from an extremely underdeveloped position after the Korean War (1950– 1953) to be a noticeable economic power through the miracle on the Han River. Four East Asian “dragons” (Korea, Taiwan, Hong Kong, and Singapore) have achieved impressive economic success since the 1970s to disprove the proposition of dependency theory that the North–South divide is perpetual. Although China’s development was slower than East Asian newly industrializing economies (NIEs), economic growth has accelerated since the 1980s, and China replaced Japan as the second largest economy in 2011. Demographically, East Asia is characterized by the lowest level of fertility and the anticipated highest percentage of elderly persons. A Total

T. Suzuki (B) Graduate School of Public Health, Seoul National University, Seoul, Korea e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 X. Ma (ed.), Employment, Retirement and Lifestyle in Aging East Asia, Social Policy and Development Studies in East Asia, https://doi.org/10.1007/978-981-16-0554-3_10

241

242

T. SUZUKI

Fertility Rate (TFR) of less than one was observed only in Taiwan and Korea except for metropolitan countries or regions. An incredibly low TFR of 0.895 was recorded in Taiwan in 2010, and a TFR of only 0.92 was recorded in Korea in 2019. While Japan is presently the oldest country in the world, it will be replaced by Korea and Taiwan by the middle of this century. It is also possible that the life expectancy at birth of Korea and Taiwan will catch up with that of Japan, which may further accelerate population aging in those countries. It is understandable that some Eastern European countries recorded extremely low fertility rates because of the socioeconomic turmoil they experienced during the transition to capitalism and a market economy. However, the TFR values of Korea and Taiwan have dropped below those of Eastern Europe, although Korea and Taiwan did not experience such chaos. It seems that cultural disorders derived from the Confucian family pattern played a role. Such cultural disorders may have included lower gender equity within family than in other sectors, strong parent–child ties and lateness in achieving economic independence, and heated competitions to enter high-ranking universities and corporations. Korea’s TFR has shown further acute decline since 2016 to record values lower than 1.0. This new decline can be interpreted with a disappointment toward the Moon Jae-In government. Although drastic population aging is expected, the present values of the percentage of elderly population and dependency ratio are not as high in Japan and in some European countries. However, the situation of elderly Koreans is already serious as seen in an extremely high elderly poverty rate and suicide rate. Such a situation could be explained by weak familial support and immature social security system.

10.2

Lowest Fertility in the World

When below-replacement fertility emerged in Northern/Western Europe in the 1980s, the second demographic transition theory (van de Kaa 1987) interpreted the trend as a symptom of a value-change syndrome from familism to individualism, along with an increase in cohabitation, extramarital births, divorce, female labor force participation, and living alone. In the 1990s, however, among Southern/Eastern European countries the TFR dropped to “lowest-low” level of 1.3 or less (Kohler et al. 2002). The emergence of lowest-low fertility drastically changed the correlation between fertility level and family variables. Today, countries

10

POPULATION AGING, LABOR FORCE PARTICIPATION …

243

with robust marriage institutions, traditional gender roles, and strong familism tend to show lower fertility. Even when Southern/Eastern European countries experienced historically low fertility levels of less than 1.3 during the 1990s, demographers could not have imagined that East Asian advanced countries would eventually be at the forefront of fertility decline. Although Korea and Taiwan showed TFR values lower than that in Japan during the late 1980s, these countries sustained higher levels than Japan throughout the 1990s (Fig. 10.1). However, the sudden acceleration of fertility decline in Korea and Taiwan following the small millennial baby boom resulted in lower TFR values than in Japan. Korea arrived at a TFR of 1.3 in 2001, followed by Japan and Taiwan in 2003. While Japan, as well as many European countries, escaped lowest-low fertility after 2005, the TFR values in Korea and Taiwan stayed at the lowest-low level. Korea recorded a TFR of 1.08 in 2005, and as mentioned above, that of Taiwan was 0.895 in 2010, supposedly the record-lowest value for a country with a rural area. The proposition “when gender equity rises to high levels in individualoriented institutions while remaining low in family-oriented institutions, fertility will fall to very low levels” (McDonald 2000) refers to the gap 3.0 2.8 2.6 2.4 2.2 2.0 1.8 1.6 1.4 1.2 1.0 0.8 1980

1985

1990

Japan

1995

2000

Korea

2005

2010

2015

Taiwan

Fig. 10.1 International comparison of total fertility rate (TFR) (Source Author’s creation)

244

T. SUZUKI

between the family and non-family systems in the gender issue. However, the gap between familial and non-familial institutions is not necessarily limited to the gender issue. For example, a stronger parent–child tie than in Northern/Western European societies (Reher 1998) is supposed to promote a delay in the transition to adulthood, which in turn results in lowest-low fertility, as discussed for Southern Europe in the 1990s (Dalla Zuanna 2001; Livi-Bacci 2001). The notion that it is a mother’s monopolistic role to care for a young child would restrict the supply and use of childcare services, limiting fertility and mothers’ labor force participation simultaneously. Educational fever and the low prestige of manual labor in the Confucian tradition create a furious competition, raising the educational cost and promoting a sentiment of hopelessness. These cultural disorders, in addition to the gap in gender equity, are presumed to have reinforced fertility decline in Taiwan and Korea (Suzuki 2013, 2018). It is supposed that the discrepancy between family and non-family sectors is largest in Taiwan. Taiwan has the highest level of gender equity in East Asia both according to the Gender Inequity Index and Global Gender Gap (DGBAS 2016a, b). In 2019, Taiwan became the first country in Asia to legalize homosexual marriage. While the government is comparatively liberal, it seems that Taiwanese people are more conservative than Japanese and Koreans when it comes to family relationship. Taiwanese respondents showed the most conservative attitudes in the East Asian Social Survey in 2006 (Iwai and Yasuda 2009). While the sex ratio at birth has been normalized in Korea, the ratio in Taiwan suggests that there remains a strong son preference. Taiwan is the only country that participated in National Transfer Account studies where the familial transfer played the most important role in the life of elderly persons (Lee et al. 2012). Such wider discrepancies between the familial and non-familial sectors in Taiwan suggest that fertility in Taiwan should be lower than in Korea. In fact, Taiwan showed lower TFR scores after 2006, even if fertility decline of Korea started earlier. Since 2016, however, Korea has entered a new phase of fertility decline. Korea recorded TFR values as low as 0.98 in 2018 and 0.92 in 2019. It is quite certain that the rate for 2020 remains less than 1.0 under the COVID-19 pandemic. This requires another ad-hoc interpretation. A possible interpretation is that the disappointment toward the Moon Jae-In government caused the recent acute decline of fertility. Young Koreans have suffered from the difficulty in finding a stable job, working

10

POPULATION AGING, LABOR FORCE PARTICIPATION …

245

poor status, widening economic disparity, and generational succession of poverty. Neoliberal economic policies favoring Chaebols (family-owned conglomerates) adopted by the conservative governments of Lee MyungBak (2008–2013) and Park Geun-Hye (2013–2017) were blamed for such problems. Young people expected that the progressive government of Moon Jae-In, who took power in May 2017, would change the situation. However, Moon’s economic policies, such as raising the minimum wage, limiting the maximum hours that could be worked per week, or enforcing permanent employment, have resulted in slow economic growth, widening economic disparity, and a decline in job opportunities for young workers. Thus, it turned out that the change from a conservative to a progressive government could not improve the situation of young Koreans. Such a disappointment supposedly caused the recent acute decline in marriages and childbirths in Korea.

10.3

Population Aging of the Republic of Korea

According to the OECD Family Database, lowest-low fertility lasted for eleven years in Italy, Spain, and Slovenia. However, it is quite certain that Korea and Taiwan will stay at that level for more than two decades. Such long-lasting extremely low fertility implies that drastic population aging will hit these East Asian economies in the predictable future. While Japan is the oldest country in the world today, Korea and Taiwan will certainly overtake Japan by the middle of this century. As shown in Table 10.1, the share of elderly population in Korea and Taiwan is around 16% in 2020, which is still considerably lower than in Japan (28–29%). However, it is projected that the percentage will rise at a significantly higher pace than Japan. According to the United Nations Population Division (UNPD), Korea and Taiwan will overtake Japan in 2045–2050 and 2060–2065, respectively. While the UNPD assumed that the TFR of low fertility countries will gradually recover to the replacement level, national official projections by the National Institute of Population and Social Security Research (IPSS), Statistics Korea, and the Council for Economic Planning and Development (CEPD) were not optimistic. Assumed TFR values for 2060–2065 in the medium variant of the UNPD (2019) were 1.60 for Japan, 1.52 for Korea, and 1.58 for Taiwan. On the other hand, the assumed values for 2065 in the medium variants of official projections were 1.44 for Japan, 1.27 for Korea, and 1.20 for Taiwan. Because the difference is larger

246

T. SUZUKI

Table 10.1 Projected percentages of elderly population aged 65+ Year

2020 2025 2030 2035 2040 2045 2050 2055 2060 2065

United Nations

Official projections

Japan

Korea

Taiwan

Japan

Korea

Taiwan

28.4 29.6 30.9 32.5 35.2 36.7 37.7 38.3 38.3 38.2

15.8 20.2 24.7 29.0 32.9 35.8 38.1 39.2 40.9 42.1

15.8 19.6 23.4 26.5 29.1 32.4 35.0 36.2 37.5 38.3

28.9 30.0 31.2 32.8 35.3 36.8 37.7 38.0 38.1 38.4

15.7 20.3 25.0 29.5 33.9 37.0 39.8 41.4 43.9 46.1

16.0 19.9 23.9 27.3 30.1 33.7 36.5 38.0 39.7 41.2

Note Values are those of medium variants Source Author’s creation based on UNDP (2019), IPSS (2018), Statistics Korea (2019), and CEPD (2018)

for Korea and Taiwan than for Japan, the tempo of population aging is far faster than the UNPD expected. A comparison of official projections reveals that Korea and Taiwan will have higher percentages of elderly persons than Japan in 2040–2045 and 2050–2055, respectively.1 Although fertility is the most important determinant of the population age structure, mortality and migration also have significant impacts. Table 10.2 compares the major demographic indicators for 2065 in medium Table 10.2 Demographic indicators in 2065

Japan Korea Taiwan

United Nations IPSS United Nations Statistics Korea United Nations CEPD

TFR Life expectancy at birth

Total population

Elderly

Male

Female

(1,000)

(%)

86.6 85.0 85.7 88.3 84.7 81.9

92.8 90.2 91.5 91.5 88.3 88.6

94,366 88,077 40,565 40,293 20,242 17,353

38.2 38.4 42.1 46.1 38.3 41.2

1.60 1.44 1.52 1.27 1.58 1.20

Note Values are those of medium variants Source Author’s creation based on UNDP (2019), IPSS (2018), Statistics Korea (2019), and CEPD (2018)

10

POPULATION AGING, LABOR FORCE PARTICIPATION …

247

variants of the UNPD and official projections. While assumed TFR values for Japan are considerably different between the UNPD and IPSS, the projected percentages of elderly are close. This is because the UNPD assumed much lower mortality than the IPSS did. Since lower mortality produces a larger elderly population in the future, the projected total populations are substantially different between two projections. Unlike in the case of Japan, the official projection by Statistics Korea assumes noticeably lower mortality for males than that for the UNPD. In fact, a comparison between national official projections reveals that the life expectancy at birth for males and females in Korea will exceed those of Japan in 2029 and 2062, respectively. This is one reason why the population aging projected by Statistics Korea is so drastic. If the mortality decline is slower and does not overtake that Japan, as in the UNPD assumption, the population aging of Korea will be somewhat milder, and the total population will be smaller than the official projection. Unlike Statistics Korea, mortality assumption for males by the CEPD of Taiwan is much less optimistic than the UNPD. If male mortality improves faster than assumed, the difference between Korea and Taiwan will be smaller than the projected results for population decrease and aging. Presently, East Asia is characterized by low percentages of foreigners in the population. In 2015, the share of the population with foreign nationality accounted for only 1.4% in Japan, 2.7% in Korea, and 3.0% in Taiwan. However, if the government decides to accept a large number of immigrants, international migration can have a significant impact on the population age structure in the future.

10.4

Problems in Supporting Elderly

While Korea is still several decades away from being the oldest society, the situation of elderly Koreans is already the worst among advanced countries. Onishi (2014) pointed out that the poverty rate of elderly Koreans was the highest among 31 countries where data in 2009 were available. This situation is confirmed by the latest data. According to the OECD (2020), Korea sustained the highest elderly poverty rate among 41 countries in 2016 (Table 10.3). Korea is also notorious for having the highest suicide rate in the developed world. While the WHO (2020) estimates age-specific suicide

248

T. SUZUKI

Table 10.3 Elderly poverty rate and suicide rates Country

Korea China Estonia Latvia Lithuania Mexico Australia United States India South Africa Israel Japan Switzerland Chile Turkey United Kingdom Russia Slovenia Canada Sweden New Zealand Italy Germany Portugal Spain Poland Austria Belgium Greece Luxembourg Brazil Finland Ireland Hungary

Old age income poverty, 65+ (%)

Suicide (per 100,000) (2016)

(2016)

60–69

70–79

80+

43.8 39.0 35.7 32.7 25.1 24.7 23.2 23.1 22.9 20.7 19.9 19.6 19.5 17.6 17.0 15.3 14.1 12.3 12.2 11.3 10.6 10.3 9.6 9.5 9.4 9.3 8.7 8.2 7.8 7.7 7.7 6.3 6.0 5.2

34.2 23.4 21.0 25.6 40.4 6.1 28.8 17.9 17.5 23.9 7.9 20.8 21.5 12.9 3.0 8.5 30.8 26.7 13.6 17.3 8.1 10.2 16.1 18.1 10.6 21.7 17.8 24.2 7.1 21.1 9.2 16.5 12.4 28.1

64.5 44.0 28.3 23.7 37.5 7.9 13.1 18.4 21.9 28.1 9.8 25.7 28.1 15.5 5.6 7.4 35.2 35.8 12.1 18.4 9.0 13.3 23.2 28.7 14.2 16.0 25.1 27.5 6.4 38.5 11.9 18.7 8.2 37.3

83.6 61.3 36.2 31.7 46.0 15.7 12.9 25.1 24.5 33.3 19.0 27.8 84.6 15.3 10.1 15.4 46.7 38.9 21.8 45.3 21.7 22.9 39.2 47.3 21.2 14.5 40.9 53.4 7.2 32.8 19.5 22.0 11.0 47.4

(continued)

10

POPULATION AGING, LABOR FORCE PARTICIPATION …

249

Table 10.3 (continued) Country

Czech Republic Norway Slovak Republic France Netherlands Denmark Iceland

Old age income poverty, 65+ (%)

Suicide (per 100,000) (2016)

(2016)

60–69

70–79

80+

4.5 4.3 4.3 3.4 3.1 3.0 2.8

17.0 16.3 17.6 20.3 17.3 17.1 24.0

18.8 12.4 19.5 27.2 16.9 20.0 8.5

22.4 32.1 21.7 73.4 32.8 49.6 8.0

Source Author’s creation based on OECD (2020) and WHO (2020)

rates for 183 countries worldwide, suicide rates for 60–69 year-olds, 70– 79 year-olds, and 80 years old and older are shown in Table 10.3 for 41 countries listed in OECD (2020). Korea’s suicide rate for 70–79 yearsold was the highest, and rates for 60–69 years-old and 80 and older were second-highest in 2016. Such a serious condition seems to reflect the unstable financial basis of elderly Koreans. The framework of the National Transfer Account (NTA) studies (Lee et al. 2012) divides financial sources other than labor income into (1) public transfers, such as pensions; (2) private transfers, especially from children; and (3) asset reallocation, such as dissaving and debt. For elderly Japanese, as for elderly persons in other developed countries in Europe and North America, the public transfers played an important role. In Taiwan, private transfers occupied the largest part in elderly’s income. In Korea, however, asset reallocation was the most important income source as in developing countries in Southeast Asia and Latin America. This suggests that familial support has declined more rapidly in Korea than in Taiwan and that the social security institution developed more slowly in Korea than in Japan. While this comparison was based on national sample surveys conducted around 2005, it seems that recent developments in social security in Korea still cannot compensate for the decline in familial support. One indicator that reflects weak familial support in Korea is the percentage of elderly persons living alone (Table 10.4). Although the percentage of such persons in Korea declined in 2010–2015, it was still

250

T. SUZUKI

Table 10.4 Percentage of elderly aged 65+ living alone

Total By Age

By Sex By Area

65–69 70–74 75–79 80+ Male Female Urban Rural

Korea (2015)

Korea (2010)

Japan (2015)

Japan (2010)

Taiwan (2010)

China (2010)

18.5 14.6 17.4 21.1 23.3 10.7 24.1 17.0 22.1

19.7 15.1 19.2 24.2 24.0 9.8 26.3 17.3 24.4

17.7 15.3 16.3 18.7 20.6 13.3 21.1 18.1 14.5

16.4 13.7 15.5 18.2 18.6 11.1 20.3 16.8 13.1

14.3 12.5 13.8 15.4 16.2 12.3 16.2 14.4 14.2

12.1

10.5 13.7 11.8 12.3

Source Author’s creation based on Census of each country

higher than Japan. While age and sex differences in Korea were larger than those of other East Asian countries, the urban–rural difference deserves a special attention. In Japan, the propensity to live alone was higher in urban areas for all ages. In China and Taiwan, the urban–rural difference was relatively small. In Korea, however, the percentage was significantly higher in rural areas than in urban areas. This implies that the living arrangements of elderly Koreans have been strongly affected by the compressed urbanization. According to the 2015 census, 49.5% of the South Korean population lived in the capital area, including Seoul, Incheon, and Gyeonggi-do, which accounts for only 11.8% of the land area. The degree of concentration in Seoul was so intensive that capital relocation has been underway since 2012, with many governmental functions being moved to the newly built Sejong City. Whatever the future impact of relocation may be, it is assumed that the drastic urbanization in Korea has resulted in a larger elderly population living alone and weaker familial support than in Taiwan or China. While Japan established universal public pension and health insurance programs in 1961, it was in 1999 that the public pension scheme covered all workers in Korea. Early and occupation-specific pension programs in Korea included the Government Employees Pension, established in 1960; the Military Serviceman Pension, in 1963; and the Private School Teacher’s Pension, in 1975. When the National Pension of Korea was established in 1988, it covered only employees at factories that employed 10 or more workers. The coverage expanded gradually to become

10

POPULATION AGING, LABOR FORCE PARTICIPATION …

251

universal in 1999. While the percentage of elderly persons was 5.8% when Japan introduced a universal pension program in 1961, the figure for Korea was 6.8% in 1999. Thus, the development of social welfare institutions relative to the demographic change in Korea was slower than that in Japan. Because having access to full benefits requires 20 years of contribution, the National Pension of Korea only matured in 2019. Although the number of elderly Koreans receiving full pension benefits will increase significantly in the near future, there will remain older elderlies without sufficient public transfers. In addition, extremely low fertility and drastic population aging in Korea stimulate anxiety about the sustainability of the public pension system. In Korea, the low labor productivity of old workers and the pressure for early retirement are supposed to be the causes of poor conditions of elderly people. According to An, et al. (2011), older workers in Korea suffer from a disadvantage, given the country’s rapid changes in technological progress and education attainment. According to newspapers, the age discrimination against older workers is expressed in such Korean phrases as “sa-o-jeong” (“retiring at age 45 is usual”) or “o-ryuk-tu” (“staying until age 56 makes one a thief”). Figure 10.2 compares the rate of male labor force participation in the 2015 census for Korea and Japan. In Japan, the rate was constant until ages 55–59 years and suddenly starts to decline after age 60. This suggests that most Japanese male workers remain at the workplace until the official retirement age of 60. On the other hand, the male labor participation rates in Korea began to decline for the workers in their 50s. This suggests that lifetime employment is not as prevalent as in Japan, and that older workers find it difficult to secure employment because of pressure to retire early. While the labor force participation rates of Korea were lower than Japan in age 50–69, the rate for those age 70 and older was higher than that of Japan. This implies that older Koreans need to stay in the labor force because of insufficient familial support and safety nets. The practice of early retirement also promotes income instability. A comparison of the 2015 census reveals that the percentages of self-employed persons and family workers among male workers aged 50–59 in Korea and Japan were 38.1% and 12.0%, respectively. It is assumed that elderly Koreans are at a higher risk than their counterparts in Japan.

252

T. SUZUKI

% 100 80 60 40 20

65-69

60-64

55-59

50-54

45-49

40-44

35-39

30-34

25-29

20-24

15-19

Korea

70以上

0

Japan

Fig. 10.2 Male labor force participation: 2015 (Korea and Japan) (Source Author’s creation)

Another source of anxiety about the future concerns an increase in never-married elderly people living alone. Those people cannot expect any financial or care support from their spouse or children and are assumed to be more vulnerable than ever-married people. As with fertility decline, nuptiality decline has also been more drastic in Korea than in Japan. As shown in Table 10.5, the percentages of single males and females age 34 or younger have already exceeded those of Japan in 2015. When these cohorts reach age 65, the percentage of single elderly persons will be significantly higher than today. Furthermore, East Asian Confucian societies have a problem of male marriage squeeze that Japan does not. Korea, Taiwan, and China began to show abnormally high sex ratios at birth since the early 1980s because of a strong son preference and availability of sex-selective abortions. In Korea, the ratio has been normalized since around 2010. However, Korean males born in 1980–2010 may have difficulty finding a spouse. Then, there should be more single old men without familial support in Korea than in Japan.

10

POPULATION AGING, LABOR FORCE PARTICIPATION …

253

Table 10.5 Percentage of single based on 2015 Census in Korea and Japan Male

15–19 20–24 25–29 30–34 35–39 40–44 45–49 50–54 55–59 60–64 65+

Female Korea

Japan

Korea

Japan

99.8 98.8 90.0 55.8 33.0 22.5 13.9 7.9 4.7 2.6 1.1

99.7 95.1 72.8 47.3 35.2 30.0 25.9 20.9 16.6 13.5 9.3

99.8 96.8 77.3 37.5 19.2 11.3 6.4 3.7 2.8 2.2 1.2

99.5 91.5 61.7 34.9 24.0 19.4 16.2 12.0 8.3 6.2 5.2

Note Foreign populations are not included Source Author’s creation based on data from 2015 Population Census in Korea and Japan

10.5

Conclusion

Even when Southern/Eastern European countries in the 1990s experienced historically low fertility levels of less than 1.3, demographers could not have imagined that East Asian advanced economies would eventually be at the forefront of fertility decline. While it is regrettable that no demographic theory could predict this development, we still need to seek interpretations of such an emergent change. This chapter has presented a cultural deterministic view of fertility decline. Confucian societies experience more serious contradictions between rapidly changing social, economic, and political systems and a slowly changing family system than feudal societies in Japan and Europe (Suzuki 2013, 2018). The distance at the starting point of family change causes a larger discrepancy between the family and non-family systems, which results in more drastic fertility decline. While this view would predict lower fertility in Taiwan than in Korea, a sudden fertility decline in Korea has reversed the situation. This chapter has presented another ad-hoc interpretation that disappointment with the Moon Jae-In government caused a recent acute decline. Given the lowest level of fertility worldwide, it is quite certain that Korea and Taiwan will overtake Japan to be the oldest and second-oldest

254

T. SUZUKI

society. Thus, the three oldest societies will be located in East Asia by the middle of this century. While such an acute demographic change creates a big challenge for the sustainability of living standards and social security institutions, the condition of elderly Koreans is already poor. This chapter has interpreted the situation as involving a combination of weak familial support and immature social safety nets. The high percentage of elderly Koreans living alone was shown to be an indicator of weak familial support. It was pointed out that Korea was slower than Japan to launch a universal public pension. Age discrimination against older workers and the pressure of early retirement also raised the risk facing elderly Koreans. Even if the proportion of never-married elderly remains low today, a rapid increase is expected in the future, especially for males. This implies a further decline of already weak familial support for elderly people in Korea. According to the OECD Family Database, only 8 of 51 countries had TFR scores higher than the replacement level in 2018, Israel, South Africa, Saudi Arabia, Indonesia, Argentina, Peru, India, and Mexico. All developed countries in Europe, North America, and East Asia had below-replacement fertility. It seems apparent that the lifestyles in today’s developed countries cannot secure the replacement level of fertility to sustain populations. Such lifestyles may include individualism that rejects the traditional authority of the kinship group, gender equity that refuses the division of labor between paid works and house works, and the demand for privacy that gives up the scale merit of household economy. Extremely low fertility implies an extremely rapid population decline in addition to drastic population aging. Thus, East Asian advanced economies should face greater difficulty than other developed countries in sustaining economic development and their standards of living. While governments of Japan, Korea, and Taiwan are still reluctant to accept a large number of foreign workers, in the near future, they will be forced to compete over inviting immigrants.

Note 1. Accurate percentages in 2055 are 38.01480% for Japan and 38.01483% for Taiwan.

10

POPULATION AGING, LABOR FORCE PARTICIPATION …

255

References An, C. B., Chun, Y. J., Gim, E. S., Hwang, N., & San Lee, S. H. (2011). Intergenerational resource allocation in the Republic of Korea. In R. Lee & A. Mason (Eds.), Population aging and the generational economy (pp. 381–393). Cheltenham, UK: Edward Elgar. Council for Economic Planning and Development. (2018). Population projections for the Republic of China: 2018–2065. (In Chinese). Directorate-General of Budget, Accounting and Statistics, Executive Yuan (Xingzehngyan Zhuji Zongsuo). (2016a). Guozing Tongji Tongbao, No. 9. (In Chinese). Directorate-General of Budget, Accounting and Statistics, Executive Yuan (Xingzehngyan Zhuji Zongsuo). (2016b). Guozing Tongji Tongbao, No. 57. (In Chinese). Dalla Zuanna, G. (2001). The banquet of Aeolus: A familistic interpretation of Italy’s lowest low fertility. In D. Z. Gianpiero & G. A. Micheli (Eds.), Strong family and low fertility: A paradox? (pp. 105–125). Dordrecht: Kluwer. Iwai, N., & Yasuda, Y. (2009). Data-de Miru Higashi Asia-no Kazokukan: Higashi Asia Shakai Chosa-ni yoru Nikkanchutai no Hikaku. Nakanishiya Shuppan. (In Japanese). Kohler, H. P., Billari, F. C., & Ortega, J. A. (2002). The emergence of lowestlow fertility in Europe during the 1990s. Population and Development Review, 28(4), 641–681. Lee, S. H., Mason, A., & Donghyun, P. (2012). Overview: Why does population aging matter so much for Asia? Population aging, economic growth, and economic security in Asia. In P. Donghyun, S. H. Lee, & A. Mason (Eds.), Aging, economic growth, and old-age security in Asia (pp. 1–31). Edward Elgar. Livi-Bacci, M. (2001). Too few children and too much family. Daedalus, 130(3), 139–156. McDonald, P. (2000). Gender equity in theories of fertility transition. Population and Development Review, 26(3), 427–440. National Institute of Population and Social Security Research. (2018). Population Projections for Japan: 2016–2065. (In Japanese). OECD. (2020). Pensions at a glance: Income and poverty of older people. https:// stats.oecd.org/index.aspx?queryid=69414 (Seen on December 1, 2020). Onishi, H. (2014). Senshinkoku Kankoku-no Yuutsu. Chuo Shinsho. (In Japanese). Reher, D. S. (1998). Family ties in Western Europe: Persistent contrasts. Population and Development Review, 24–2, 203–234. Statistics Korea. (2019). Population projections for the Republic of Korea: 2017– 2067 . (In Korean).

256

T. SUZUKI

Suzuki, T. (2013). Low fertility and population aging in Japan and Eastern Asia. Springer. Suzuki, T. (2018). Eastern Asian population history and contemporary population issues. Springer. United Nations Population Division. (2019). World population prospects, 2019 revision. van de Kaa, D. (1987). Europe’s second demographic transition. Population Bulletin, 42(1), 1–59. WHO. (2020). Global health observatory data repository. https://apps.who. int/gho/data/node.main.MENTALHEALTH?lang=en (Seen on December 1, 2020).

CHAPTER 11

Retirement Timing and Post-retirement Employment in Taiwan Ruoh-Rong Yu and Ming-Chang Tsai

11.1

Introduction

Taiwan has been experiencing rapid population aging. The proportions of population aged 65 or over in 1989, 1999, 2009, and 2019 are 6.0%, 8.4%, 10.6%, and 15.3%, respectively.1 This figure is projected to exceed 20% in 2025. Taiwan will become a super-aging society in a short time, with grave impacts on public policy.2 Rapid population aging has led to deficits in all kinds of occupation-based pension funds. Although the eligibility age for retirement pensions and old-age insurance benefits has been raised, the average age of retirement has only slightly increased. Men’s and women’s average effective ages of retirement during 2002–2007 are 62.2 and 59.4, respectively, while those during 2013–2018 are 64.6 and 61.0.3

R.-R. Yu (B) · M.-C. Tsai Research Center for Humanities and Social Sciences, Academia Sinica, Taipei, Taiwan e-mail: [email protected] M.-C. Tsai e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 X. Ma (ed.), Employment, Retirement and Lifestyle in Aging East Asia, Social Policy and Development Studies in East Asia, https://doi.org/10.1007/978-981-16-0554-3_11

257

258

R.-R. YU AND M.-C. TSAI

Compared to other East Asian societies with similar trends of population aging, Taiwan’s workers indeed retire much earlier: For the period of 2013 to 2018, it is age 70.8 and 69.1 for men and women in Japan and 72.3 and 72.3 in Korea.4 It is not surprising to find that the labor force participation rate for those aged 65 or above is low at 8.3% in Taiwan, while the rates are 25.3% and 34.0% in Japan and South Korea.5 Early withdrawal from the labor market in Taiwan constitutes a peculiar case in this region. This chapter has two research goals. The first one is to examine the determinants of early retirement in Taiwan using data from the Panel Study of Family Dynamics (PSFD). The PSFD is a longitudinal survey project initiated in 1999. The core modules have been repeated in the subsequent waves of panel survey and are suitable for observing retirement behaviors over time. The second is to investigate the patterns and determinants of post-retirement employment in this island society. As suggested by researchers, working in retirement has increasingly become more common due to the erosion of family support and pension systems (Johnson 2009; Pleau 2010). It has been suggested that retirement behaviors differ across genders. Men and women have been found to have different concerns for the timing of retirement (Damman et al. 2015; Dentinger and Clarkberg 2002; Kridahl 2017; Loretto and Vickerstaff 2013; Lumsdaine and Vermeer 2015; O’Rand and Farkas 2002) as well as whether to work after retiring (Henretta et al. 1993a; Kim and Moen 2002; Moen et al. 2001; Pleau 2010). In particular, men are usually regarded as the main breadwinner in a family. They hence are more likely to stay in the labor market after retirement (Dentinger and Clarkberg 2002; Drobniˇc 2002). Women, on the hand, are more likely to have a different work trajectory due to family concerns (e.g., childbearing and rearing). In addition, women’s roles as caregiver and homemaker might also affect their retirement decisions (Damman et al. 2015; Lumsdaine and Vermeer 2015; O’Rand and Farkas 2002). Detailed examination of differences between males and females facilitates our understanding of gender differences in the decisions of retirement timing and post-retirement employment. The retirement systems in Taiwan will be briefly described in the next section. The relevant literature is to be discussed in the third section. The fourth section explains the data, variables, and analytical methods. In the fifth section, the findings will be presented and discussed. The final section concludes.

11

RETIREMENT TIMING AND POST-RETIREMENT EMPLOYMENT …

11.2

259

Institutional Background

There is a two-tier occupation-based income support system for retirees in Taiwan, comprising retirement pensions and old-age insurance benefits. Almost all individuals who are employed in the public and private sectors are covered by some form of retirement pension system. A majority of workers who are employed in the private sector are covered under the Labor Standards Act enforced in the early 1980s. This Act contains regulations on retirement pension and eligibility requirements (age and/or years of service) for voluntary and mandatory retirement.6 To be eligible for voluntary retirement, a worker should meet any one of the following conditions: (1) reaching age 55 and having completed 15 years of service; (2) having completed 25 years of service; and (3) reaching age 60 and having completed 10 years of service. The mandatory retirement age is set at 65. A new labor pension system under the Labor Pension Act was initiated in 2005.7 The main difference between the old and new labor pension systems is that a worker’s years of service are not portable among employers under the old system, while the new system removes this restriction. The public service pension system of Taiwan was established in the early 1940s. The public service pension system has reformed in 1995 from a purely government-financed system to a pension fund that is supported jointly by the government and employees (including public servants, public school faculty, and military personnel). The mandatory retirement age for civil servants and public school teachers is 65, while that of military personnel varies according to their rank in the army. As to non-civil servants employed in government organizations or public schools and the faculty employed in private schools, the corresponding pension systems were established in 1995 and 2010, respectively. The old-age insurance benefits are applicable to workers who participate in occupation-specific insurance such as labor insurance, civil servant and school faculty insurance, and military personnel insurance. A common feature of these occupation-based insurance schemes is that the retirees have to meet the age and/or years of service requirements specified in the relevant regulations.8 Yet, the eligibility requirements for old-age insurance benefits differ by insurance scheme. In the past few decades, the government in Taiwan has broadened the coverage of workers who are eligible for the retirement pension and oldage insurance benefits. At the same time, the eligibility requirements of

260

R.-R. YU AND M.-C. TSAI

age and years of service for retirement pensions and old-age insurance benefits have become more stringent. For example, for a civil servant to be eligible for monthly retirement pensions, the original requirement is that the sum of age and years in service should be 75 at a minimum. The threshold has been gradually increased to 84 in 2020, and it will be further increased to 94 in 2030. The civil servant’s eligibility age for old-age insurance benefits will also be raised from 50 in 2020 to 65 in 2030. Workers covered by labor insurance face a similar trend, in that the eligibility age for old-age insurance benefits will be gradually increased from 60 to 65 by 2026.9 Despite these policy changes, the eligibility criteria of retirement pensions and old-age insurance benefits for public sector workers are more generous than those in the private sector. Within the private sector, company size is crucial for a worker’s eligibility for voluntary retirement and retirement pensions. It is noteworthy that the small and medium-sized companies account for over 78% of employment. Their average duration of survival is 13 years as of 2019, much shorter than in large enterprises.10 Under such institutional and organizational backgrounds, one can speculate that whether one is employed in the public sector, a large enterprise, or small or medium-sized company matters for retirement options and timing.

11.3 11.3.1

Literature Review Studies on Retirement Timing

Retirement timing has been regarded as one of the most important topics in the studies of retirement (Beehr 1986; Fisher et al. 2016; Scharn et al. 2018) and has drawn much attention from rational choice and life course theories to explain retirement decisions and the determinants of retirement timing (De Preter et al. 2013; Pleau and Shauman 2013). The earlier studies treated retirement as a personal decision, while more recent ones have noticed spousal and familial factors in operation (De Preter et al. 2013; Drobniˇc 2002; Kubicek et al. 2010; Madero-Cabib et al. 2016). In the following, we first discuss how personal traits and work characteristics may affect retirement timing, then turn to assess spousal and the other familial factors. While age is a strong predictor for retirement timing (Jex and Grosch 2013), the influence of education has been less easy to pin down. People with more education are found to retire earlier in some previous studies

11

RETIREMENT TIMING AND POST-RETIREMENT EMPLOYMENT …

261

(De Preter et al. 2014; Siegrist et al. 2007), while some others reported an opposite effect (Schils 2008).11 Health condition is conceived as an important factor for early retirement (McGarry 2004). Nevertheless, a few studies found insignificant effects in some societies (Schils 2008). Partnered men and women are more divergent than unmarried men and women with regard to the patterns and determinants of retirement timing (Drobniˇc 2002). With respect to work characteristics, researchers suggested that a pension system which offers generous retirement options would encourage early retirement (Engelhardt 2012). People with higher income tend to delay retirement to sustain their material well-being. However, the findings are mixed, because a few studies have found that workers with higher earnings are more likely to retire early (Fisher et al. 2015; Dahl et al. 2000). Married partners’ retirement timing is often found to be interdependent. Researchers have suggested that this might be owing to the spouses’ desire to spend time together (Gustman and Steinmeier 2004; Michaud et al. 2020), financial needs in case of spousal retirement (Henretta and O’Rand 1983), age and/or educational homogamy (Gustafson 2018; Henkens and Siegers 1991; Henkens et al. 1993), and gender role attitudes (Talaga and Beehr 1995). Even though many studies have documented correlation of partners’ retirement decisions, the direction and magnitude are still undetermined (Gustman and Steinmeier 2000; Henretta et al. 1993b; Ho and Raymo 2009; O’Rand and Farkas 2002; Smith and Moen 1998). Gustman and Steinmeier (2000, 2004), for instance, found that married individuals’ retirement timing is related to that of their spouses, and the relationship is stronger for married men than for married women. Syse et al.’s (2014) study reported that marital partners’ retirement decisions are correlated, but the gender differences were less clear. Researchers have suggested that other familial factors might affect one’s retirement timing (e.g., Dentinger and Clarkberg 2002; Fisher et al. 2016; Matthews and Fisher 2013). The caring needs of family members can trigger early retirement (Dentinger and Clarkberg 2002; Van Bavel and De Winter 2013). Financial needs of family members, on the hand, are usually associated with late retirement (Drobniˇc 2002; Fisher et al. 2016).

262

R.-R. YU AND M.-C. TSAI

11.3.2

Studies on Post-retirement Employment

Retirement is more of a prolonged process than an abrupt event (Hayward et al. 1994; Pleau 2010; Pleau and Shauman 2013; Van Solinge 2014). Post-retirement employment thus can be regarded as partial retirement (employment with reduced working hours) or “bridge employment” (transitional employment between career employment and complete withdrawal from the labor force).12 Post-retirement employment can still be a full-time job when retirement is defined as receipt of retirement pension or as a self-identified status (Beehr and Bennett 2015; Pleau 2010). Compared to retirement timing, post-retirement employment has been an understudied issue (Beehr and Bennett 2015; Wang et al. 2008). Postretirement employment is often regarded as driven by extrinsic motives such as financial needs and economic payoffs (Hurd 1990; Van Solinge 2014). On the other hand, intrinsic motives (such as role identities and personal values and goals) are also highlighted in the decisions of working after retirement (Van Solinge 2014; Wang and Shi 2014). Retirees with higher education and better health are more likely to work after retirement (Platts et al. 2019; Pleau 2010; Pleau and Shauman 2013; Wang et al. 2008). As for gender, studies have found that men tend to engage in post-retirement employment more than women (Davis 2003; Pleau and Shauman 2013; Singh and Verma 2003). This gender difference has been attributed to the traditional gender roles of men as breadwinners and women as homemakers and caregivers in a family. However, some studies have shown different findings (Han and Moen 1999; Ruhm 1990). When it comes to marital status, Pleau and Shauman (2013) showed that males who have a partner are less likely to reenter the labor force than those who have none (being divorced, separated, widowed, or never married), while women reveal the opposite pattern. Besides these demographic traits, spousal and familial factors are reported to have influence on post-retirement employment. Workers with a working spouse are more likely to engage in post-retirement employment (Platts et al. 2019). Having dependent children prevents exit from the labor market (Choi 2002; Cahill et al. 2006). Caregiving needs of family members are associated with a lower probability of continuing working or reentering the workforce after retirement (Pleau 2010). Wealth and family income might also affect the decision to work after retiring (Cahill et al. 2006). In general, wealthy retirees are more likely

11

RETIREMENT TIMING AND POST-RETIREMENT EMPLOYMENT …

263

to withdraw from the labor force (Beehr and Bennett 2015; Hardy 1991; Schellenberg et al. 2005). However, Platts et al. (2019) found that this pattern is not so obvious. Pleau and Shauman’s (2013) study reported no association between family income and post-retirement employment for women, but among men, they found a positive association (Pleau 2010).

11.4

Data and Analytical Methods

The data are drawn from the first two main samples of the PSFD. These two groups of respondents were born in 1935–1954 and 1953– 1964, with their first-wave data collected in 2000 and 1999, respectively. The first-wave survey was conducted by face-to-face interview, with the target population being Taiwan citizens who were born in the abovespecified years. A stratified multi-stage PPS (probability proportionate to size) sampling method was used to select the sample, while the sampling frame is household registration records. For the main respondents who completed the first-wave questionnaire, the follow-ups were conducted on an annual basis till 2012. Since 2012, the follow-ups have been implemented biennially. The panel data collected from the first wave till the 2018 survey are employed for analysis. 11.4.1

Retirement Timing: Sample, Variables, and Method

To analyze the transition into retirement, the analytical sample is limited to the respondents who were between 49 and 64 years of age and employed full time, with weekly working hours exceeding 30 hours at the time of the survey. The choice of lower and upper bounds of age is made based on the eligibility age for retirement pension and the mandatory retirement age of the ongoing retirement systems in Taiwan. Full-time employment is defined as jobs with more than 30 working hours per week, following the definition of OECD.13 The employers and self-employed are excluded from analysis because most of them are not covered by retirement systems. Two definitions of retirement are adopted in our analyses.14 One is whether the respondent has retired from a full-time job covered by any of the ongoing retirement systems, regardless of receiving retirement benefits. After deleting the individual-time observations with missing covariates, the number of observations left for the analysis of “transition into retirement” is 3,632, with the number of respondents being 960.

264

R.-R. YU AND M.-C. TSAI

The other definition is based on whether the respondent has ever received any pension, either in lump sum or on a monthly basis. The number of observations for the model of “transition into pension” is 3,947, while the corresponding number of respondents is 1,027. The Cox proportional hazards model (Cox 1972) is used to analyze transition into retirement or pension, with cluster robust standard errors being estimated to correct for heteroscedasticity. The covariates include personal traits, work characteristics, familial factors, and household characteristics. 11.4.1.1 Demographic and Other Personal Traits Age is measured by the year of survey minus the year of birth. Education is categorized into three dummies: senior high school or below (reference group), vocational school or college, and university or above. Marital status is grouped into three dummies, including unmarried (reference), married or cohabiting, and separated, divorced, or widowed. Health is coded as 1 when the self-reported health condition is poor or very poor, and 0 otherwise. 11.4.1.2 Job Characteristics The sectors in which the respondents work are divided into the private sector, public sector, and non-profit organizations. The private sector is further distinguished into large enterprises (hiring more than 100 workers) and small and medium-sized companies (less than 100 workers). Occupation is grouped into four categories: (1) elementary and craft workers (reference); (2) clerical, service, and sales workers; (3) professionals (including technicians and associate professionals); and (4) managers (including chief executives, senior officials, and legislators). Earnings are the logarithmic values of the self-reported monthly income from the main job (in thousand Taiwan dollars), being deflated by the consumer price index (CPI) with 2016 as the base year. 11.4.1.3 Familial Factors and Household-Level Information The spouse’s retirement and working status might affect a partnered individual’s retirement decision. Therefore, two interaction variables are included as additional covariates. The first one is the interaction term of married/cohabiting status and spousal retirement status (spouse retired or not). The second is the interaction term of married/cohabiting status and spousal working status (spouse in paid employment or not).

11

RETIREMENT TIMING AND POST-RETIREMENT EMPLOYMENT …

265

To take into account the caregiving and financial needs of parents and/or children, three variables are employed. One is the health condition of parents. If all the parents are in poor or very poor health, it is coded as 1; otherwise 0. The other two child-related covariates are: having children aged 22 or below (= 1), and having children aged over 22 (= 1). Age 22 is chosen as the threshold due to the high enrollment ratio of higher education in recent decades.15 The household-level factors used in analysis include non-labor household income and residential area. Non-labor household income is the sum of annual business income, rents, interests, dividends, and capital gains (in thousand Taiwan dollars), being deflated by CPI and transformed to log values. The residential area dummy is measured by whether the respondent resides in a metropolitan area (versus rural residence). Within these covariates, only education is time invariant, while all the other covariates are time varying. The mean values of the covariates for male and female samples are listed in Table 11.1. Approximately 66% and 56% (57% and 48%) of the male and female respondents reported they retired (received pensions) during the observation period.16 Men have a clearly higher probability of transition into retirement or pension than women, probably because women’s employment is more likely to be interrupted for various reasons (e.g., childbearing and rearing) and thus they were less likely to be eligible for voluntary retirement or retirement pension than men at a comparable age. 11.4.2

Post-retirement Employment: Sample, Variables, and Method

In the analysis of post-retirement employment, the sample is confined to those who had retired and whose ages were between 50 and 70 when interviewed. After deleting missing values, the number of observations left is 8,354 from 1,494 respondents. Note that the number of observations (respondents) exceeds that of the analysis of retirement timing. This is because the respondents who retired before the first wave of survey are also included in the analysis of the post-retirement employment. A random-effects multinomial logit model is used to analyze retirees’ choice among full-time employment, part-time employment, and not working. Cluster robust standard errors are estimated to adjust for heteroscedasticity.17 The covariates contain personal traits, pension status, familial factors, and other family-level information. Pension status is a dummy indicating whether the respondent was receiving any pension

266

R.-R. YU AND M.-C. TSAI

Table 11.1 Means of variables for analyses on retirement timing: by gender Male

Age (in years) Education Senior high school or below Vocational school or college University or above Marital status Unmarried Married or cohabiting Separated/divorced or widowed Poor health Sector and company size Small or medium-sized company Large enterprise Public sector Non-profit organization Occupation Elementary and craft workers Clerical, service, and sales workers Professionals Managers Log earnings (in thousand Taiwan dollars) Married/cohabiting × spouse retired Married/cohabiting × spouse working Parents in poor health

Female

Transition into retirement

Transition into pension

Transition into retirement

Transition into pension

54.7

54.8

54.0

54.2

68.4 (%)

68.9 (%)

75.9 (%)

77.8 (%)

13.4 (%)

13.3 (%)

10.1 (%)

9.2 (%)

18.2 (%)

17.8 (%)

13.9 (%)

13.0 (%)

5.0 (%) 90.9 (%)

4.7 (%) 91.1 (%)

9.8 (%) 75.9 (%)

8.9 (%) 75.8 (%)

4.2 (%)

4.2 (%)

14.3 (%)

15.3 (%)

8.0 (%)

8.3 (%)

10.3 (%)

10.4 (%)

43.6 (%)

44.6 (%)

54.3 (%)

56.3 (%)

19.7 (%) 33.6 (%) 3.1 (%)

19.4 (%) 33.0 (%) 3.0 (%)

19.7 (%) 22.0 (%) 4.0 (%)

19.4 (%) 20.5 (%) 3.9 (%)

47.6 (%)

47.8 (%)

42.0 (%)

42.6 (%)

12.5 (%)

13.3 (%)

31.9 (%)

32.9 (%)

21.5 (%) 18.5 (%) 3.9

21.2 (%) 17.7 (%) 3.9

18.1 (%) 8.0 (%) 3.3

16.9 (%) 7.6 (%) 3.3

8.6 (%)

9.1 (%)

28.1 (%)

29.7 (%)

45.9 (%)

46.7 (%)

45.3 (%)

44.7 (%)

18.1 (%)

18.1 (%)

20.6 (%)

20.2 (%)

(continued)

11

RETIREMENT TIMING AND POST-RETIREMENT EMPLOYMENT …

267

Table 11.1 (continued) Male

Presence of children under 22 Presence of children over 22 Log household income (in thousand dollars) Metropolitan area Number of observations Number of respondents Number of failures (retirement)

Female

Transition into retirement

Transition into pension

Transition into retirement

Transition into pension

50.3 (%)

49.3 (%)

29.4 (%)

28.9 (%)

69.5 (%)

70.3 (%)

82.7 (%)

83.5 (%)

1.5

1.5

1.7

1.8

11.7 (%) 2,081

11.6 (%) 2,194

11.9 (%) 1,736

11.2 (%) 1,946

518

531

451

504

340

302

251

241

Source Author’s creation

benefit at the time of the survey. The definition of the other covariates is the same as that of the covariates in the retirement timing model, and thus is not repeated here. The means of variables in the model of post-retirement employment are listed in Table 11.2. This table indicates that 25%, 9%, and 66% of the male observations are in the statuses of full-time employment, part-time employment, and not working. The corresponding figures for females are 17%, 6%, and 77%. These statistics reveal that, relative to men, women are less likely to engage in full-time or part-time employment after retirement.

11.5 11.5.1

Findings

Findings for Retirement Timing

The findings for the Cox proportional hazards model are presented in Table 11.3. We first discuss the results for men’s and women’s “transition into retirement” (columns 1 and 3). One common finding between genders is that the probability of transition into retirement is increasing with age. The other is that the workers with a higher household income are more likely to delay their retirement.

268

R.-R. YU AND M.-C. TSAI

Table 11.2 Means of variables for analyses on post-retirement employment: by gender

Working status Not working Part-time work Full-time work Age (in years) Education Senior high school or below Vocational school or college University or above Marital status Unmarried Married or cohabiting Separated/divorced or widowed Poor health Any retirement pension Married/cohabiting × spouse working Married/cohabiting × spouse with pension Parents in poor health Presence of children under 22 Presence of children over 22 Log household income (in thousand Taiwan dollars) Metropolitan area Number of observations Number of respondents

Male

Female

66.1 (%) 9.0 (%) 24.9 (%) 62.5

76.5 (%) 6.2 (%) 17.3 (%) 59.3

80.3 (%) 8.8 (%) 10.9 (%)

89.1 (%) 5.4 (%) 5.5 (%)

3.9 (%) 88.4 (%) 7.7 (%) 19.4 (%) 12.2 (%) 23.3 (%) 7.7 (%) 9.0 (%) 14.0 (%) 92.2 (%) 2.5 11.4 (%) 3,647 677

5.7 (%) 75.0 (%) 19.3 (%) 21.8 (%) 8.8 (%) 28.1 (%) 12.4 (%) 10.9 (%) 20.3 (%) 84.9 (%) 2.4 11.4 (%) 4,707 820

Source Author’s creation

Marital status exhibits different effects between genders. Relative to unmarried women, married or cohabiting women are marginally less likely to retire from their main career, but the effect is not observed among men. This is probably because partnered women’s career trajectory is more likely to be interrupted by family factors. Separated, divorced, or widowed men, compared to unmarried men, have a higher chance of transition into retirement. The female sample reveals an opposite result, that unmarried women tend to retire earlier than separated, divorced, or widowed women. One possible explanation for the late retirement of separated, divorced, or widowed women is the heavier financial burden faced by them, while their male counterparts’ earlier retirement is probably due to the responsibilities of homemaking and caregiving. Health

11

RETIREMENT TIMING AND POST-RETIREMENT EMPLOYMENT …

269

Table 11.3 Cox model on retirement transition: by gender and retirement definition Male

Age Education (Ref: Senior high school or below) Vocational school or college University or above Marital status (Ref: Unmarried) Married or cohabiting Separated/divorced or widowed Poor health Sector and company size (Ref: Small or medium-sized company) Large enterprise Public sector Non-profit organization Occupation (Ref: Elementary and craft workers) Clerical, service, and sales workers Professionals Managers

Female

Transition into retirement

Transition into pension

Transition into retirement

Transition into pension

1.127*** (0.016)

1.089*** (0.017)

1.087*** (0.018)

1.060*** (0.018)

1.140 (0.185) 1.081 (0.191)

0.990 (0.169) 1.233 (0.228)

0.781 (0.172) 0.752 (0.163)

0.926 (0.197) 0.896 (0.195)

1.395 (0.349) 1.892* (0.605) 0.979 (0.177)

1.346 (0.353) 1.613 (0.553) 1.096 (0.200)

0.660† (0.165) 0.471** (0.122) 1.300† (0.204)

0.495** (0.134) 0.502* (0.138) 1.422* (0.220)

1.476** (0.210) 1.331* (0.175) 1.134 (0.390)

1.373* (0.210) 1.391* (0.186) 1.209 (0.422)

1.416* (0.210) 1.273 (0.211) 1.242 (0.376)

1.276 (0.206) 1.250 (0.204) 1.133 (0.311)

1.190 (0.184) 0.974 (0.161) 1.025 (0.182)

1.170 (0.187) 1.004 (0.171) 1.004 (0.193)

1.063 (0.160) 1.446* (0.256) 1.344 (0.320)

1.003 (0.143) 1.357† (0.236) 1.117 (0.299)

(continued)

270

R.-R. YU AND M.-C. TSAI

Table 11.3 (continued) Male Transition into retirement Log earnings

1.105 (0.086) Married/cohabiting × 1.349† (0.207) spouse retired Married/cohabiting × 0.849 spouse working (0.097) Parents in poor health 0.719* (0.109) Presence of children 0.862 under 22 (0.117) Presence of children over 1.034 22 (0.165) Log household income 0.933** (0.024) Metropolitan area 1.296† (0.191) Log likelihood −1,727.66 Number of observations 2,081 Number of respondents 518 Number of failures 340

Female Transition into pension

Transition into retirement

Transition into pension

1.137 (0.099) 1.387* (0.217) 0.818† (0.094) 0.673* (0.113) 0.883 (0.121) 1.111 (0.182) 0.926** (0.024) 1.211 (0.195) −1,546.88 2,194 531 302

1.228* (0.125) 1.121 (0.174) 0.918 (0.136) 0.784 (0.139) 0.972 (0.155) 1.545† (0.348) 0.902*** (0.024) 1.331 (0.252) −1,243.55 1,736 451 251

1.304* (0.139) 1.025 (0.169) 1.142 (0.177) 0.830 (0.142) 1.046 (0.174) 1.511† (0.335) 0.928** (0.024) 1.298 (0.249) −1,220.97 1,946 504 241

Note 1. Hazard ratios are in the table. Standard errors are in parentheses 2. ***p < 0.001, **p < 0.01, *p < 0.05, † p < 0.1 Source Author’s creation

condition exhibits different effects between genders, such that women in poor health are marginally more likely to transition into retirement, while the effect is not found among men. This reflects women’s role as the secondary worker in a family. Men and women employed in large enterprises tend to retire earlier than those who are working for small or medium-sized companies. However, working in the public sector is a beneficial factor for men’s early retirement, while no significant effect is found among women. These findings are in general consistent with our conjecture that institutional and organizational factors are crucial for one’s timing of retirement.

11

RETIREMENT TIMING AND POST-RETIREMENT EMPLOYMENT …

271

With regard to the effects of the other work characteristics, women in professional or higher-income jobs are more likely to retire earlier than women in blue-collar or lower-income jobs. Yet, neither occupation nor earnings matter for men’s retirement timing. This reveals that men in Taiwan, under the role as the main breadwinner in families, tend to be more attached to their main jobs regardless of their occupation or level of earnings. Some of the familial factors reveal divergent effects between men and women. Married or cohabiting men with a retired spouse are marginally more likely to transition into retirement, while the pattern is not found among partnered women. This finding is partly consistent with Gustman and Steinmeier’s finding (2000, 2004) that married people’s retirement timing is contingent on that of their spouses, and the relationship is stronger for men than for women. The other difference is the association between parental health condition and retirement timing. Men tend to retire later if their parents are in poor health, while the association is not found among women. This reflects the fact that men and women have different concerns for retirement timing when facing the financial and caregiving needs of parents in poor health. Women with children aged 22 or over are marginally more likely to transition into retirement, yet the association is not found among men. This is also consistent with women’s role as the secondary worker in a family. We now turn to the model of “transition into pension” (columns 2 and 4 of Table 11.3) and compare its results with the model of “transition into retirement.” It can be seen that the findings are only slightly different between models, for either the male or female sample. This indicates that the findings for retirement timing are by and large robust to the definition of retirement. 11.5.2

Findings for Post-retirement Employment

The results for the random-effects multinomial logit model on retirees’ post-retirement employment status (including full-time working, parttime working, and not working) are presented in Table 11.4, in which not working is the reference group. Male and female retirees share a commonality that the older they become, the less likely they are to work full time, as compared to the reference group of not working. Also for both men and women, retirees with children over 22 years of age are more likely to have a full-time job.

272

R.-R. YU AND M.-C. TSAI

Table 11.4 Random-effects multinomial logit model on post-retirement employment: by gender Male

Age Education (ref: Senior high school or below) Vocational school or college University or above Marital status (ref: Unmarried) Married or cohabiting Separated/divorced or widowed Poor health

Female

Part-time working

Full-time working

Part-time working

Full-time working

0.992 (0.027)

0.885*** (0.025)

1.031 (0.022)

0.949* (0.021)

0.422 (0.258) 1.667 (0.864)

1.627 (0.874) 0.959 (0.545)

1.142 (0.694) 1.066 (0.587)

1.143 (0.689) 0.592 (0.422)

0.545 (0.357) 0.490 (0.331) 0.583* (0.130) 0.946 (0.263) 1.476 (0.367) 2.218** (0.664) 1.537 (0.442) 2.034* (0.638) 4.681** (2.492) 1.070* (0.036) 1.200 (0.696) 61.377* (115.988)

0.802 (0.368) 1.370 (0.588) 0.722 (0.182) 0.898 (0.289) 1.999* (0.543) 1.046 (0.341) 0.763 (0.233) 2.164* (0.732) 1.859† (0.692) 1.030 (0.038) 0.712 (0.323) 0.002*** (0.003) −2,361.03 4,707

1.076 (0.521) 1.694 (0.660) 0.542** (0.110) 0.575* (0.162) 2.182** (0.549) 1.098 (0.299) 1.000 (0.259) 1.345 (0.459) 2.782** (0.981) 1.068* (0.032) 0.243** (0.133) 0.243 (0.343)

0.371 (0.244) 0.245† (0.186) 0.443** (0.125) Any retirement pension 0.751 (0.233) Married/cohabiting × 2.438*** spouse working (0.586) Married/cohabiting × 2.563** spouse with pension (0.896) Parents in poor health 1.080 (0.373) Presence of children 1.135 under 22 (0.453) Presence of children 1.898 over 22 (1.242) Log household income 1.139*** (0.043) Metropolitan area 0.726 (0.374) Constant 0.070 (0.129) Log likelihood −2,164.30 Number of 3,647 observations

(continued)

11

RETIREMENT TIMING AND POST-RETIREMENT EMPLOYMENT …

273

Table 11.4 (continued) Male Part-time working Number of respondents

667

Female Full-time working

Part-time working

Full-time working

820

Note 1. Odds ratios are in the table. Standard errors are in parentheses 2. ***p < 0.001, **p < 0.01, *p < 0.05, † p < 0.1 Source Author’s creation

Poor health is linked to a lower likelihood of full-time employment for both men and women, and it also lowers the chance of men having a part-time job. Female retirees with a retirement pension are less likely to engage in full-time employment than those without pension. This effect is insignificant among the male counterparts. As to the working status of the spouses of partnered respondents, men who have a working spouse are more likely to work part-time than those with a non-working spouse. For women who have a partner working, their odds of working part-time or full-time are higher. These findings suggest that the spouse’s working status has an influence on partnered women’s work decisions after retirement, but less so for partnered men. The findings also indicate that if the spouse is receiving a pension, partnered men have a higher probability of engaging in part-time or fulltime work. But the effect is not significant among partnered women. The presence of children aged 22 or under is found to be associated with a higher probability of men’s (woman’s) engaging in full-time (part-time) employment. This result reveals the role of dependent children as a factor in prolonging working life. Non-labor household income is positively associated with men’s and women’s likelihood of having a full-time job. It also increases men’s chance of working part-time after retirement. Retired women who reside in urban areas are less likely to engage in full-time employment, probably because full-time working opportunities for aging women are relatively scarce in urban areas.

274

R.-R. YU AND M.-C. TSAI

11.6

Conclusions

This chapter examines retirement timing and post-retirement employment among males and females using multi-wave longitudinal data collected in Taiwan. We highlight the importance of gender roles. The descriptive statistics indicate that the proportion of men who transition into retirement (pension) is higher than that of women, probably because men are more likely to have a continuous work trajectory and thus a higher likelihood of voluntary retirement (receipt of pension) than women. In addition, the proportions of men engaging in full-time or part-time employment after retirement are higher than women’s counterparts, which might be attributed to the different gender roles of men and women. For the Cox proportional hazards model on retirement timing, one important finding is that women working in professional or higherincome jobs are inclined to retire early, while men do not show similar patterns. This suggests that men’s role as the primary breadwinner results in stronger attachment to their jobs, while women’s role as the supplementary income earner for families lends their retirement decisions more subject to the working conditions before retirement. Secondly, men who are separated, divorced, or widowed are likely to retire earlier than unmarried men, while an opposite pattern is found among women. Separated, divorced, or widowed men, because they have no spouse to share housework and caregiving, tend to spend more time in families, which is associated with a higher probability of early retirement. Women in separation, divorce, or widowhood may in turn assume the obligations of supporting the family’s expenses and therefore tend to retire later from their main jobs. Thirdly, men with a retired spouse have a higher likelihood of transition into retirement, while the female counterpart does not show such a pattern. This is consistent with the gender role perspective that men, as the main breadwinner of the family, tend to stay longer in their main jobs and delay retirement until their spouses have retired. Fourthly, the finding that men tend to postpone their retirement timing when their parents are in poor health also reflects men’s role as the economic provider when their family members are in financial need. Lastly, women who have children aged 22 or over are more likely to retire early, while this correlation is trivial for men. This is also consistent with the gender roles of men and women.

11

RETIREMENT TIMING AND POST-RETIREMENT EMPLOYMENT …

275

The findings on post-retirement employment indicate the importance of gender roles as well. Female retirees with retirement pension are less likely to engage in full-time employment, while the effect is not significant among males. In addition, spousal working status affects partnered women’s post-retirement working decisions, but its effect is weaker for partnered men. Furthermore, partnered men with spouse receiving pension are more likely to work after retirement, while the effect is insignificant among partnered women. All these findings indicate that gender roles are not only important for men’s and women’s decisions on retirement timing, but also crucial for their choice of whether to work after retirement. Our findings echo the studies which stress the importance of gender roles on men’s and women’s retirement decisions (Dentinger and Clarkberg 2002; Worts et al. 2016). Men’s breadwinning roles make them more attached to labor market, with regard to either retirement timing or post-retirement employment. In contrast, women’s time-intensive caregiving roles make their decisions on retirement timing and working after retirement more subject to their employment and familial economic situations. Nevertheless, separation, divorce, or widowhood might somehow reverse the traditional gender roles of men and women, which is in line with the arguments of Pleau (2010) and Stoiko and Strough (2019) that when men (women) enact female-typical (male-typical) caregiving (breadwinning) roles, their retirement decisions resemble that of women (men). This chapter contributes to our understanding of how gender roles might shape men’s and women’s retirement timing and post-retirement employment decisions. Under the trend that gender role attitudes have become more egalitarian in Taiwan, future research should explore whether gender differences in men’s and women’s retirement decisions can be narrowing rather than widening.

Notes 1. The relevant statistics are retrieved from the website of the National Development Council, Taiwan (https://pop-proj.ndc.gov.tw/main_en/ dataSearch.aspx?uid=78&pid=78). 2. Please refer to the website cited in the previous note for relevant information.

276

R.-R. YU AND M.-C. TSAI

3. The average effective age of retirement is the weighted average of (net) withdrawals from the labor market at different ages over a 5-year period for workers initially aged 40 and over. The statistics for Taiwan are retrieved from the 2009 and 2019 International Labor Statistics published by the Ministry of Labor, Taiwan (https://www.mol.gov.tw/statistics/ 2452/2457/18350/). 4. The relevant figures are retrieved from the OECD website (https://www. oecd.org/els/emp/average-effective-age-of-retirement.htm). 5. These figures are retrieved from the website cited in Note 3. 6. Contents of the Labor Standards Act can be found on the website of the Laws and Regulations Database, Taiwan (https://law.moj.gov.tw/ENG/ LawClass/LawAll.aspx?pcode=N0030001). 7. Interested readers can refer to the Laws and Regulations Database website for contents of the Labor Pension Act (https://law.moj.gov.tw/ENG/ LawClass/LawAll.aspx?pcode=N0030020). 8. The contents of the Labor Insurance Act, Civil Servant and School Faculty Insurance Act, and Act of Insurance for Military Personnel can be found on the following webpages. a. https://law.moj.gov.tw/ENG/LawClass/LawAll.aspx?pcode=N00 50001 (Labor Insurance Act) b. https://law.moj.gov.tw/LawClass/LawAll.aspx?PCode=S0070001 (Civil Servant and School Faculty Insurance Act) c. https://law.moj.gov.tw/ENG/LawClass/LawAll.aspx?pcode=F00 50001 (Act of Insurance for Military Personnel) 9. Information for relevant reforms is retrieved from a government webpage (in Chinese): https://www.lkjh.chc.edu.tw/resource/openfid. php?id=419. 10. Relevant statistics are retrieved from the website of the Small and Medium Enterprise Administration, Ministry of Economic Affairs, Taiwan (https://book.moeasmea.gov.tw/book/doc_detail.jsp?pub_SerialNo= 2020A01653&click=2020A01653). 11. On the one hand, higher education is often associated with more favorable occupations or working conditions, which might lead to earlier retirement. On the other hand, people with higher education are more likely to retire later due to the longer time spent in schooling (Fisher et al. 2016). 12. See, for example, Beehr and Bennett (2015), Quinn and Kozy (1996), Ruhm (1990), and Van Solinge (2014) for the concepts of partial employment and bridge employment. 13. See the OECD website for details (https://data.oecd.org/emp/parttime-employment-rate.htm). 14. For possible definitions of retirement, interested readers can refer to, for example, Beehr and Bowling (2013) and Denton and Spencer (2009).

11

RETIREMENT TIMING AND POST-RETIREMENT EMPLOYMENT …

277

15. The rapid expansion of higher education in Taiwan has led to higher chances for women to be admitted into universities. Since 1997, the maleto-female enrollment ratio in higher education institutions has persisted at approximately 1:1 (Luo and Chen 2018, p. 89). 16. The failure rates are computed from the last two rows of Table 11.1. For example, 66% is the number of failures (340) divided by the number of respondents (518). 17. For the random-effects multinomial logit model, interested readers can refer to, for example, Hartzel et al. (2001) for more details.

References Beehr, T. A. (1986). The process of retirement: A review and recommendations for future investigation. Personnel Psychology, 39(1), 31–55. Beehr, T. A., & Bennett, M. M. (2015). Working after retirement: Features of bridge employment and research directions. Work, Aging and Retirement, 1(1), 112–128. Beehr, T. A., & Bowling, N. A. (2013). Variations on a retirement theme: Conceptual and operational definitions of retirement. In M. Wang (Ed.), The Oxford handbook of retirement. New York: Oxford University Press. Cahill, K. E., Giandrea, M. D., & Quinn, J. F. (2006). Retirement patterns from career employment. The Gerontologist, 46(4), 514–523. Choi, N. G. (2002). Self-defined retirement status and engagement in paid work among older working-age women: Comparison between childless women and mothers. Sociological Inquiry, 72(1), 43–71. Cox, D. R. (1972). Regression models and life tables. Journal of the Royal Statistical Society, 34(2), 187–202. Dahl, S. A., Nilsen, O. A., IV., & Vaage, K. (2000). Work or retirement? Exit routes for Norwegian elderly. Applied Economics, 32(14), 1865–1876. Damman, M., Henkens, K., & Kalmijn, M. (2015). Women’s retirement intentions and behavior: The role of childbearing and marital histories. European Journal of Population, 31(4), 339–363. Davis, M. A. (2003). Factors related to bridge employment participation among private sector early retirees. Journal of Vocational Behavior, 63(1), 55–71. Dentinger, E., & Clarkberg, M. (2002). Informal caregiving and retirement timing among men and women gender and caregiving relationships in late midlife. Journal of Family Issues, 23(7), 857–879. Denton, F. T., & Spencer, B. G. (2009). What is retirement? A review and assessment of alternative concepts and measures. Canadian Journal on Aging, 28(1), 63–76.

278

R.-R. YU AND M.-C. TSAI

De Preter, H., Van Looy, D., & Mortelmans, D. (2014). Retirement timing of dual-earner couples in 11 European countries? A comparison of Cox and shared frailty models. Journal of Family and Economic Issues, 36(3), 396–407. De Preter, H., Van Looy, D., Mortelmans, D., & Denaeghel, K. (2013). Retirement timing in Europe: The influence of individual work and life factors. Social Science Journal, 50(2), 145–151. Drobniˇc, S. (2002). Retirement timing in Germany: The impact of household characteristics. International Journal of Sociology, 32(2), 75–102. Engelhardt, H. (2012). Late careers in Europe: Effects of individual and institutional factors. European Sociological Review, 28(4), 550–563. Fisher, G. G., Chaffee, D. S., & Sonnega, A. (2016). Retirement timing: A review and recommendations for future research. Work, Aging and Retirement, 2(2), 230–261. Fisher, G. G., Ryan, L. H., & Sonnega, A. (2015). Prolonged working years: Consequences and directions for interventions. In J. Vuori, R. Blonk, & R. H. Price (Eds.), Sustainable working lives: Managing work transitions and health throughout the life course (pp. 269–288). Dordrecht, The Netherlands: Springer. Gustafson, P. (2018). The gendered economics of synchronized retirement. Research on Aging, 40(7), 623–644. Gustman, A. L., & Steinmeier, T. L. (2000). Retirement in dual-career families: A structural model. Journal of Labor Economics, 18(3), 503–545. Gustman, A. L., & Steinmeier, T. L. (2004). Social security, pensions and retirement behaviour within the family. Journal of Applied Econometrics, 19(6), 723–737. Han, S. K., & Moen, P. (1999). Clocking out: Temporal patterning of retirement. American Journal of Sociology, 105(1), 191–236. Hardy, M. A. (1991). Employment after retirement: Who gets back in? Research on Aging, 13(3), 267–288. Hartzel, J., Agresti, A., & Caffo, B. (2001). Multinomial logit random effects models. Statistical Modelling, 1(2), 81–102. Hayward, M. D., Hardy, M. A., & Liu, M. C. (1994). Work after retirement: The experiences of older men in the United States. Social Science Research, 23(1), 82–107. Henkens, K., Kraaykamp, G., & Siegers, J. (1993). Married couples and their labour market status: A study of the relationship between the labour market status of partners. European Sociological Review, 9(1), 67–78. Henkens, K., & Siegers, J. (1991). The decision to retire: The case of Dutch men aged 50–64. European Journal of Population, 7 (3), 231–249. Henretta, J. C., & O’Rand, A. M. (1983). Joint retirement in the dual worker family. Social Forces, 62(2), 504–520.

11

RETIREMENT TIMING AND POST-RETIREMENT EMPLOYMENT …

279

Henretta, J. C., O’Rand, A. M., & Chan, C. G. (1993a). Gender differences in employment after spouse’s retirement. Research on Aging, 15(2), 148–169. Henretta, J. C., O’Rand, A. M., & Chan, C. G. (1993b). Joint role investments and synchronization of retirement: A sequential approach to couples’ retirement timing. Social Forces, 71(4), 981–1000. Ho, J. H., & Raymo, J. M. (2009). Expectations and realization of joint retirement among dual-worker couples. Research on Aging, 31(2), 153–179. Hurd, M. D. (1990). Research on the elderly: Economic status, retirement, and consumption and saving. Journal of Economic Literature, 28(2), 565–637. Jex, S. M., & Grosch, J. (2013). Retirement decision making. In M. Wang (Ed.), The Oxford handbook of retirement (pp. 267–279). New York: Oxford University Press. Johnson, R. W. (2009). Employment opportunities at older ages: Introduction to the special issue. Research on Aging, 31(1), 3–16. Kim, J. E., & Moen, P. (2002). Retirement transitions, gender, and psychological well-being: A life-course, ecological model. Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 57 (3), P212–P222. Kridahl, L. (2017). Retirement timing and grandparenthood in Sweden: Evidence from population-based register data. Demographic Research, 37, 957–994. Kubicek, B., Korunka, C., Hoonakker, P., & Raymo, J. M. (2010). Work and family characteristics as predictors of early retirement in married men and women. Research on Aging, 32(4), 467–498. Loretto, W., & Vickerstaff, S. (2013). The domestic and gendered context for retirement. Human Relations, 66(1), 65–86. Lumsdaine, R. L., & Vermeer, S. J. (2015). Retirement timing of women and the role of care responsibilities for grandchildren. Demography, 52(2), 433–454. Luo, Y. H., & Chen, K. H. (2018). Education expansion and its effects on gender gaps in educational attainment and political knowledge in Taiwan from 1992 to 2012. International Journal of Educational Development, 60, 88–99. Madero-Cabib, I., Gauthier, J. A., & Le Goff, J. M. (2016). The influence of interlocked employment: Family trajectories on retirement timing. Work, Aging and Retirement, 2(1), 38–53. Matthews, R. A., & Fisher, G. G. (2013). Family, work, and the retirement process: A review and new directions. In M. Wang (Ed.), The Oxford handbook of retirement (pp. 354–370). New York: Oxford University Press. McGarry, K. (2004). Health and retirement: Do changes in health affect retirement expectations? Journal of Human Resources, 39(3), 624–648. Michaud, P. C., Van Soest, A., & Bissonnette, L. (2020). Understanding joint retirement. Journal of Economic Behavior & Organization, 173, 386–401.

280

R.-R. YU AND M.-C. TSAI

Moen, P., Kim, J. E., & Hofmeister, H. (2001). Couples’ work retirement transitions, gender, and marital quality. Social Psychology Quarterly, 64(1), 55–71. O’Rand, A. M., & Farkas, J. I. (2002). Couples’ retirement timing in the United States in the 1990s: The impact of market and family role demands on joint work exits. International Journal of Sociology, 32(2), 11–29. Quinn, J. F., & Kozy, M. (1996). The role of bridge jobs in the retirement transition: Gender, race, and ethnicity. The Gerontologist, 36(3), 363–372. Platts, L. G., Corna, L. M., Worts, D., McDonough, P., Price, D., & Glaser, K. (2019). Returns to work after retirement: A prospective study of unretirement in the United Kingdom. Ageing & Society, 39(3), 439–464. Pleau, R. L. (2010). Gender differences in postretirement employment. Research on Aging, 32(3), 267–303. Pleau, R. L., & Shauman, K. (2013). Trends and correlates of post-retirement employment, 1977–2009. Human Relations, 66(1), 113–141. Ruhm, C. J. (1990). Bridge jobs and partial retirement. Journal of Labor Economics, 8(4), 482–501. Scharn, M., Sewdas, R., Boot, C. R., Huisman, M., Lindeboom, M., & Van Der Beek, A. J. (2018). Domains and determinants of retirement timing: A systematic review of longitudinal studies. BMC Public Health, 18(1), 1083. Schellenberg, G., Turcotte, M., & Ram, B. (2005). Post-retirement employment. Statistics Canada. Schils, T. (2008). Early retirement in Germany, The Netherlands, and the United Kingdom: A longitudinal analysis of individual factors and institutional regimes. European Sociological Review, 24(3), 315–329. Siegrist, J., Wahrendorf, M., Von Dem Knesebeck, O., Jürges, H., & BörschSupan, A. (2007). Quality of work, well-being, and intended early retirement of older employees—Baseline results from the SHARE Study. European Journal of Public Health, 17 (1), 62–68. Singh, G., & Verma, A. (2003). Work history and later-life labor force participation: Evidence from a large telecommunications firm. Industrial and Labor Relations Review, 56(4), 699–715. Smith, D. B., & Moen, P. (1998). Spousal influence on retirement: His, her, and their perceptions. Journal of Marriage and Family, 60(3), 734–744. Stoiko, R. R., & Strough, J. (2019). His and her retirement: Effects of gender and familial caregiving profiles on retirement timing. The International Journal of Aging and Human Development, 89(2), 131–150. Syse, A., Solem, P. E., Ugreninov, E., Mykletun, R., & Furunes, T. (2014). Do spouses coordinate their work exits? A combined survey and register analysis from Norway. Research on Aging, 36(5), 625–650. Talaga, J. A., & Beehr, T. A. (1995). Are there gender differences in predicting retirement decisions? Journal of Applied Psychology, 80(1), 16–28.

11

RETIREMENT TIMING AND POST-RETIREMENT EMPLOYMENT …

281

Van Bavel, J., & De Winter, T. (2013). Becoming a grandparent and early retirement in Europe. European Sociological Review, 29(6), 1295–1308. Van Solinge, H. (2014). Who opts for self-employment after retirement? A longitudinal study in The Netherlands. European Journal of Ageing, 11(3), 261–272. Wang, M., & Shi, J. (2014). Psychological research on retirement. Annual Review of Psychology, 65, 209–233. Wang, M., Zhan, Y., Liu, S., & Shultz, K. S. (2008). Antecedents of bridge employment: A longitudinal investigation. Journal of Applied Psychology, 93(4), 818–830. Worts, D., Corna, L., Sacker, A., McMunn, A., & McDonough, P. (2016). Understanding older adults’ labour market trajectories: A comparative gendered life course perspective. Longitudinal and Life Course Studies, 7 (4), 347–367.

CHAPTER 12

Population Aging, Low Fertility, and Social Security in Russia Kazuhiro Kumo

12.1

Introduction

It goes without saying that social security policies are implemented because of poverty arising from problems centered on injury/illness, childbirth, and aging. The same was/is true for both Soviet Russia and the modern Russian Federation. However, it has to be said that the economic background as well as social systems were very different between two regimes. For example, in the Soviet Union, unemployment was not supposed to exist, so there was no system of employment insurance. Yet injury/illness, childbirth, and aging are events that occur

This chapter is one of the outcomes of a grant-in-aid for scientific research from the Ministry of Education, Culture, Sports, Science and Technology of Japan (B) (19H01478) and a joint-use/joint-research base project administered by the Institute of Economic Research, Hitotsubashi University. K. Kumo (B) Institute of Economic Research, Hitotsubashi University, Tokyo, Japan e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 X. Ma (ed.), Employment, Retirement and Lifestyle in Aging East Asia, Social Policy and Development Studies in East Asia, https://doi.org/10.1007/978-981-16-0554-3_12

283

284

K. KUMO

regardless of the economic systems, so there has always been a need to protect individuals from them. In this chapter, the author offers an overview of the economic circumstances and social environment that provide the background to social security policy in modern Russia. Although the chapter does not touch on the Soviet era in detail, it must at least refer to the sorts of socioeconomic shocks that Russia experienced following the collapse of the Soviet Union. Therefore, the chapter summarizes the changes in the socioeconomic environment that took place from the end of the 1980s until after the Soviet collapse, as well as trends after 2000, when sustained economic growth began to be seen. Despite the risk of repeating oneself, the author should state that the ultimate goal of social security is to reduce poverty. So to begin with, the author looks at the trend with poverty levels from the end of the Soviet period and in Russia subsequently, and trends in the area of economic disparities. After that, the author describes the aspects of injury/illness and aging from among the main causes of poverty.

12.2

Poverty and Economic Disparities in Russia

It is fair to say that it is widely known that economic disparities were small and poverty levels were low in the former Soviet Union and other socialist countries. In the Soviet Union, income redistribution was conducted broadly, the state set uniform wage rates, and social security measures such as medical coverage and pensions were generous. These policies kept poverty at low levels (McAuley 1979). But as the system began to be transformed, this situation underwent major changes. The well-known Milanovic (1997) estimated the total number of people with income less than the poverty line based on a variety of data. Milanovic (1997) found that in 1987–1988, the number of people in poverty in Russia was no more than 2.2 million people (1.5%) out of a total population of 146 million (1987), but that once the systemic transformation had got underway, 66 million people, or 44% of the total population of 148.5 million (1993), were in poverty. This meant that the total number of people in poverty had increased by 30 times (Fig. 12.1). This put the poverty line at an income of USD 4 per person per day in terms of purchasing power parity in 1993, which can be said to be quite a high estimate. However, this does not alter the overall trend.

12

285

POPULATION AGING, LOW FERTILITY, AND SOCIAL …

(Period) 1993-1995 1987-1988 0

10

20

30

40

50

60

70

(million people) Fig. 12.1 Number of people in poverty in Russia (Source Author’s creation based on data from Milanovic [1997])

Of course, even in the Soviet Union, which operated under a system of socialism, it is doubtful that poverty did not exist at all. Because it was impossible to look at data on household incomes and household consumption, the situation was just such that investigations could not be performed. But at the same time, poverty expanded in Russia in conjunction with the systemic transformation, and it is reasonable to say that this was seen on a broader scale. The expansion in poverty pointed out by Milanovic (1997) and shown here has been described as “sudden poverty” in previous research on poverty in Russia (Ruminska-Zimny 1997). This choice of expression is indicative of the view that once the socialist system, with its generous social security, collapsed, poverty expanded rapidly. In fact, when the poverty headcount during the socialist era and the period after the start of systemic transformation are compared, a major shift can be observed. That being said, hardly any data for the socialist era exists, which is something the author mentioned earlier. At present, various estimate series can be used, and relying on those, the trend in the poverty headcount is presented in Fig. 12.2, which shows Russia’s poverty headcount (defined in the Russian Federation as the proportion of the population earning less income than the “basic cost of living”) and per-capita gross domestic product (GDP) between 1980, prior to the Soviet collapse, and 2018. Here, Russia’s poverty headcount, which was 11.4% in 1991, had reached 31.5% in 1993, after the systemic transformation, which had begun at the end of 1991, had already got underway. This attests to the truth of the “sudden poverty” in transition economies described by Ruminska-Zimny (1997).

286

K. KUMO

(1=100%)

(1000 R in 1997)

Poverty headcount (left axis) Per-capita GDP (right axis)

2018

2016

2014

2010

1300 2012

0.05 2006

3300

2008

0.1

2002

5300

2004

0.15

2000

7300

1998

0.2

1996

9300

1992

0.25

1994

11300

1988

0.3

1990

13300

1980

0.35

Fig. 12.2 Russia’s poverty headcount and per-capita GDP, 1989–2018 (Source Author’s creation based on data from Rosstat, Sotsial’noe polozhenie i uroven zhisni naseleniya Rossii, various years; Rosstat, Regiony Rossii, various years; World Bank, World Development Indicators )

On the other hand, it is easy to see that the poverty rate jumps dramatically in the 1990s before falling back in the 2000s, which indicates contrastive dynamics depending on the time period. Here, it can be pointed out that there is a close relationship between economic conditions and the poverty rate. This is obvious when one compares the poverty rate and per-capita GDP, as the coefficient of correlation between the poverty headcount and per-capita GDP as shown in Fig. 12.2 is −0.82, which shows that the poverty headcount falls as per-capita GDP expands. Similarly, as shown in Fig. 12.3, the Gini coefficient, which is an indicator of income disparities, spiked from 0.265 in 1991 to 0.398 in 1993. It has stayed at a high level since, but after showing signs of rising once again in the middle of the 2000s, it fell back again, and has been stable since the second half of the 2000s. At the beginning of the systemic transformation, it is certainly true that both the poverty rate and income disparities broke with the previous trend and started to rise discontinuously. But it can be confirmed that this was in line with the sharp drop in economic output (Figs. 12.2 and 12.3). However, when one looks at the situation after 1999, when

12

POPULATION AGING, LOW FERTILITY, AND SOCIAL …

287

(1000 R in 1997) 0.45

13000

0.43

12000

0.41

11000

0.39 0.37

10000

0.35

9000

0.33

8000

0.31

7000

0.29

2018

2014

2016

2012

2008

2010

2004

2006

2002

2000

1998

1996

1994

1992

5000 1990

0.25 1988

6000 1980

0.27

Gini coefficient (left axis)

Percapita GDP (right axis)

(Years)

Fig. 12.3 Russia’s income disparities and per-capita GDP, 1980–2013 (Source Author’s creation based on data from Braithwaite [1995]; Rosstat, Sotsial’noe polozhenie i uroven zhisni naseleniya Rossii, various years; Rosstat, Regiony Rossii, various years)

economic growth began, one sees that the poverty rate clearly contracted, while economic disparities did not necessarily widen. It is not hard to imagine that economic growth, by raising income levels among all classes of people, would lead to a drop in the poverty rate. But besides that, it is also feasible that social security effectiveness would increase on a wider scale, resulting in income redistribution, which would curtail any expansion in economic disparities. In Russia in the first half of the 1990s, a certain degree of progress had been made with the establishment of a legal framework, but it was impossible to effectively reduce poverty levels because, for example, funding from the federal budget was restricted and the value of the benefits provided were so measly. It therefore seems likely that the sustained economic growth seen in the 2000s was what made it possible to implement effective social security policies. As mentioned earlier, the chapter was able to confirm time-series estimates for poverty and economic disparities in Russia. What should be examined next are trends with the various factors that determine such poverty levels. Among these factors, the author looks at the central

288

K. KUMO

ones, namely healthcare and pensions, as well as aspects relating to childrearing/childbirth.

12.3

Longevity and Injury/Illness

During discussions of Russian health policy, what comes up frequently is the issue of healthcare levels and the issue of diets and lifestyles. As a result of such factors, average life expectancy at birth in Russia has exhibited a startling trend. Figure 12.4 shows the average life expectancy at birth of males not only in Russia and a number of other former socialist countries, but also in Western European countries from 1961 to 2018. From the middle of the 1960s, a distinctly different trend can be seen with the former socialist countries (Bulgaria, Hungary, Poland, and Russia) on the one hand and the Western advanced countries on the other. In general, the curves for the Western countries climb continuously from left to right. However, it can be said that those for the socialist countries (Age) 85

Russia United Kingdom

80

Japan Hungary

75

Austria Denmark

70

Finland France

65

Italy New Zealand

60

Norway

1960 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008 2011 2014 2017

55

319

Bulgaria Poland Netherlands

Fig. 12.4 Average male life expectancy at birth (Source Author’s creation based on data from World Bank, World Development Indicators; Rosstat, Demograficheskii ezhegodnik Rossii, various years)

12

POPULATION AGING, LOW FERTILITY, AND SOCIAL …

289

did not rise at all from the mid-1960s until the systemic transformation that occurred in 1989–1991. Among the socialist countries, Russia’s divergence is observed to be especially large. It could be even said that the curve was trending downwards. But what is behind this high death rate in Russia? Because the death rate was already high during the Soviet era (average life expectancy at birth was low), the conclusion cannot be drawn that the deterioration of healthcare levels and the collapse of the social security system following the demise of the Soviet Union was a direct cause of the rise in the death rate. The principal controversialists in the field of Russian demographics explain what happened using such reasons as a sharp rise in stress levels in conjunction with the systemic transformation (Vishnevsky and Bobylev 2009). Supporting this logic is the relative frequency of different causes of death. Figure 12.5 shows causes of death for men only, and the proportion of deaths that occurred due to each cause. Between 1965 and 1990, the proportion of deaths from “cardiovascular disease” increased. Furthermore, the proportion for “external Infectious disease

100% 90%

Digestive disease

80% 70%

Respiratory disease

60% 50%

Malignant growth

40% 30% 20%

External causes (accident, suicide, murder, etc.)

10% 0%

Cardiovascular disease

(Year) Fig. 12.5 Causes of death among Russian males, 1965–2018 (Source Author’s creation based on data from Rosstat, Demograficheskii ezhegodnik Rossii, various years)

290

K. KUMO

causes” was high in 1965–1980. The data strongly suggests that these high figures for “cardiovascular disease” and “external causes” could be connected with the high overall death rate in the Soviet Union and Russia, the low average male life expectancy at birth, and standards of living, or more specifically, alcohol consumption (Nemstov 2002). What clearly appears following the collapse of the Soviet Union at the end of 1991 is a jump in the proportion of deaths from “external causes,” which had fallen back in 1985–1990. The proportion remained high until the early 2000s. Also clear is the fact that the proportion of deaths from “cardiovascular disease” climbed rapidly after 1995 and stayed at a high level thereafter. This would be consistent with the interpretation that stress resulting from the systemic transformation triggered an increase in alcohol consumption, and that this led directly to a rise in the death rate. This view is also shared with various analyses employing microdata, which hold that until the middle of the 2000s alcohol intake harmed the health of Russians and was one of the factors behind the increase in the death rate. The statement that “Russians drink too much alcohol” might come across as a joke, but the insights accumulated from previous research indicate that the statement is factual. However, as one moves into the second half of the 2000s, one sees a clear downward trend in the proportion of deaths attributed to “external causes,” and this can also be said to be consistent with the fact that the economy became stable. Indeed, as was seen in Fig. 12.4, the average male life expectancy at birth has increased continuously since 2005 and at a speed never observed previously. The issue of lifestyles that was seen in the 1990s, along with the problem of Russia’s traditional approach to health, namely medical care that is focused on treatment rather than prevention, could not be ignored. However, the economy at the end of the 1990s had shrunk in size to just a little over half that seen at the end of the Soviet era, and under these circumstances it was difficult to successfully implement measures that were suited to the conditions. For these circumstances to undergo changes, and the Russian health authorities to deliver recognizable improvements, what was needed was the rapid economic growth that occurred from beginning of the 2000s, and which also helped to reduce poverty levels. In 2005, a national priority project called “Public Health” was established with the goals of not only improving advanced medical care through the improvement of

12

POPULATION AGING, LOW FERTILITY, AND SOCIAL …

291

frontline healthcare levels and the deployment of more medical equipment, but also stepping up action against traffic accidents and cardiovascular disease, transforming the healthcare system, recommending lifestyle improvements, shifting the focus to preventative medicine, and so on.1 As a result, government funds began to be invested in the healthcare field on a large scale for the first time since the collapse of the Soviet Union.

12.4

Aging and Pensions

There is no argument that population is what is important when considering a social security system, and especially a pension system. One of the biggest problems facing Russia was, alongside the high death rate discussed in the previous section, falling fertility (see next section), which meant, as can be seen from Fig. 12.6, that during the more than 20year period between 1992, just after the Soviet collapse, and 2012, the natural rate of population increase was a negative, meaning that the total (Per mille) 17 16 15 Crude Birth Rate

14 13 12

Crude Death rate

11 10 9 8

(year) Fig. 12.6 Crude birth rate and crude death rate in Russia, 1960–2018 (Source Author’s creation based on data from Rosstat, Demograficheskii ezhegodnik Rossii, various years)

292

K. KUMO

population was falling.2 The drop in population was caused mainly by the declining birth rate, but as the number of children decreases, the proportion of the population who are elderly also increases. Given the extremely low average life expectancy at birth, which the chapter saw in the previous section, it might seem odd that the population of Russia is aging. However, due to the impact of the age structure of the population (e.g., the baby-boom generation born after the Second World War becoming part of the elderly population) and differences in the definition of working age, the population of Russia is aging, too. Figure 12.7 shows calculations for age composition indexes along based on the ages at which Russians become eligible to receive pensions. Here, one can see that the elderly population index is increasing. Furthermore, the elderly population index rose sharply from 32.6 in 2005 during the following 15 years, reaching 46.7 at the beginning of 2019. This, alongside the figure for Japan, is the highest in the world (the figure for Japan was 47.2 in 2020 according to annual estimates of population from the Statistics Bureau of Japan [SBJ]). Aging is one of the main causes of poverty. Asset disparities expand as the age group rises, and, as is widely known, when elderly people without assets become unable to work, they are at higher risk of falling 100%

50

90% 80% 70%

45 40

Working age population ratio (left axis, males under 60 years / females under 55 years)

35

Child population ratio (left axis, under 15 years)

60% 50%

Elderly population ratio (left axis, males 60 years or over / females 55 years or over)

40% 30% 20%

30 25

10% 0%

Elderly population index (right axis)

20

Child population index (right axis)

(Year) Fig. 12.7 Proportion of population by age group/age composition indexes, 1989–2019 (Source Author’s creation based on data from Rosstat, Demograficheskii ezhegodnik Rossii, various years)

12

POPULATION AGING, LOW FERTILITY, AND SOCIAL …

293

into poverty. However, the poverty risk among pensioners in Russia is not remarkably high. This is partly because the average annual pension benefit in Russia has generally exceeded the “basic cost of living” determined by the federal government. Refer to Fig. 12.8. From the 1999 Russian financial crisis until 2001, there was rapid inflation, and indexation failed to keep up, so the average annual pension benefit temporarily dipped below the “basic cost of living,” but apart from that, it has usually been in the vicinity of or higher than the basic cost of living. Furthermore, since 2010, the average annual pension benefit has stayed steady at 1.5 times the basic cost of living. While pensioners may not regard the amount they receive as adequate, their benefits appear to be generous, at least in light of the economic circumstances and scale of fiscal expenditure since 2009. Starting with an amount more or less equal to the basic cost of living, pensions were increased rapidly from 2007 onward.3 What has made this level of pension benefits possible is, of course, the increase in government revenue from (%) 180 55.00

160

50.00

140

Average pension as a percentage of basic cost of living (left axis)

45.00

120

Pensioner index (pensioner population * 100 / working age population, right axis)

2018

2016

2014

2012

2010

2008

2006

2004

30.00 2002

60 2000

35.00

1998

80

1996

40.00

1994

100

(Year) Fig. 12.8 Average pension benefit relative to basic cost of living and pensioner index, 1994–2018 (Source Author’s creation based on data from Rosstat, Sotsial’noe polozhenie i uroven zhisni naseleniya Rossii, various years)

294

K. KUMO

the export of resources, mainly oil and gas. So generous pensions can also be said to have been gifted by economic growth. Nevertheless, the continued rise in such indicators as the elderly population index (Fig. 12.7) and the ratio of pensioners to the working-age population (Fig. 12.8) makes it necessary to redesign systems relating to the use of government funds, the supply of which is not inexhaustible and which are affected by energy market conditions. Until recently, people have begun receiving pension benefits at the extremely young ages of 60 years for men and 55 years for women, but because the pension fund account is already in the red, in October 2018 the decision was made to increase the age of eligibility to receive pension benefits, and this measure took effect in January 2019.4 As a result of this law, the pension eligibility age would be gradually raised to 65 for men and 60 for women. Given that the population is going to continue to age in the future, improving pension finance is a pressing issue for Russia, too.5

12.5

Childbirth/Childrearing

Childbirth and then childrearing lead at least to the short-term withdrawal of the main care providers from the labor market or to them no longer being fully employed. The risk of them falling into poverty may therefore be higher. It is commonly understood that in high-income countries, demand for “quality” in terms of children increases, and that demand for “quantity” of children shrinks in response, and that this has led to a declining birth rate, or in other words, fewer children (Becker 1960). Because the Soviet Union lost a huge number of lives during the Second World War, having children was always encouraged in the postwar Soviet Union. From the 1960s onwards, birth rates in Western advanced countries declined rapidly, while the socialist countries maintained a birth rate of just over 2.0, enough to sustain the population, until 1989, partly as a result of plentiful social childcare facilities (nursery schools and kindergartens under the control of companies or government organizations). After the collapse of the Soviet Union, however, the network of social childcare facilities (nursery schools and kindergartens) weakened swiftly. Those that had been operated for companies for their employees, and which were almost free of charge, were either closed or charges for them

12

POPULATION AGING, LOW FERTILITY, AND SOCIAL …

295

were introduced. This led directly to an increase in the cost of childcare.6 Furthermore, the economic crisis that accompanied the systemic transformation resulted repeatedly in sharp decreases in the size of the economy. Because of this, the ability of the new generation to bear the cost of childrearing declined. The Soviet Union was known for its generous social security system (McAuley 1979). However, the systemic transformation destroyed the foundation of the system. The Soviet labor market was also characterized by stable employment, an absence of unemployment, and stable, though not especially high, wages. But such features were lost with the economic transition. Factors like these compounded one another, and the end result was rapid drop in Russia’s total fertility rate, which slumped to below 1.20 in 1999 and 2000 (Fig. 12.9). The Russian government came out with various measures for addressing this situation. In “Population Development Concept for the Russian Federation by 2015,”7 a document that the Russian federal (1000 R in 1997)

(0/00)

13000

2.1

12000

2

11000

1.9 1.8

10000

1.7

9000

1.6

8000

1.5

7000

1.4 1.3

6000

1.2

5000

1.1

(Year) Per-capita GDP (in 2010 USD, left axis)

Total fertility rate (right axis)

Fig. 12.9 Russia’s total fertility rate and per-capita GDP, 1989–2018 (Source Author’s creation based on data from Rosstat, Demograficheskii ezhegodnik Rossii, various years; Rosstat, Regiony Rossii, various years)

296

K. KUMO

government produced in 2001, the government promised to take steps to improve the health of citizens and increase the birth rate. At that time, however, no new measures to tackle the falling birth rate and the rising death rate were taken. In other words, the document did not have any real meaning. In the policy perspective, a turning point arrived during the latter half of the first Putin administration, after sustained economic growth had begun. In 2005/2006, President (at that time) Putin, in his annual addresses to the Federal Assembly, mentioned the issue of the slumping birth rate, and stated that increasing it was a governmental goal. Following this, in December 2006 the childcare allowance, etc. was raised,8 while a new “mothers’ fund”9 was established as a government-funded scheme that would provide large sums of money for having children. Income redistribution in the form of support for childbirth/childrearing was designed to reduce the risk of people of reproductive age falling into poverty. However, what needs to be kept in mind here is that, as Fig. 12.9 shows, the rise in the birth rate can be seen to have begun in 2001, prior to the introduction, in 2006, of the government-funded scheme that can be viewed as a measure aimed at encouraging people to have children. In other words, the inflection from a declining to a rising birth rate can be regarded as matching the start of economic growth, and this can also be seen here. Attention also needs to be paid to the fact that measures to encourage childbearing are implemented as a means of transferring income to the childrearing generation, but that the execution of such measures requires a fiscal foundation. Here, too, economic growth itself can be seen to have enabled social policies to be implemented. Between the 1990s and the early 2000s, it would not have been an overstatement to say that the Russian government’s social policies were nothing more than words of encouragement. However, this situation underwent substantial changes from the mid2000s onwards, as measures that were actually accompanied by fiscal resources began to be introduced.

12.6

Conclusions

In this chapter, an overview of the socioeconomic background behind social security policy in Russia was provided. The rapid economic growth seen in the 2000s provided the foundation that allowed Russia’s social

12

POPULATION AGING, LOW FERTILITY, AND SOCIAL …

297

security system, which had become fragile following the collapse of the Soviet Union, to be strengthened. However, the modes of behavior of individuals are up to individuals themselves to decide, and, of course, it is by no means certain, even in Russia, that the wishes of the government will be fulfilled. Lifestyles and the shock of the systemic transformation pushed up the death rate in Russia. Once into the 2000s, economic growth began to be observed on a sustained basis, and the federal government has started to invest its new resources in strengthening the public health system. With the population aging, a reinforced fiscal foundation has provided the background to efforts to boost pension benefits, but it would be difficult to argue that this will be sustainable over the long term, so the consequences of the pension system reforms that have been introduced will be an issue in the future. Post-Soviet Russia, which experienced an ultra-low birth rate, is using massive government revenues obtained from oil and gas to implement extremely aggressive measures to promote childbirth/childrearing, and the success of these measures may be being demonstrated in the form of an increase in the birth rate. Against a backdrop of continued economic and social stability, Russia can be regarded as having finally started to expand and reform its social security policy. As for the end result, however, the reforms themselves are still at a nascent stage, and the evaluation of policies will require the observation of the trend over a certain period of time. It will therefore be some time before the outcomes of the policies can be ascertained.

Notes 1. The website of the steering committee for national priority projects (https://www.rost.ru), which reports directly to the President of the Russian Federation, provides detailed information about various “national priority projects.” The “Public Health” project is described in detail in the section of the aforementioned website for specific projects (https://www. rost.ru/projects/health/health_main.shtml). 2. In 2013, the natural rate of increase turned positive for the first time in more than 20 years, but was heavily impacted by the age structure, within which women of reproductive age accounted for a large proportion of the total population. In fact, since 2016 the natural rate of increase has once again been negative.

298

K. KUMO

3. Poslanie federalnomu sobraniyu Rossiiskoi Federatsii, 26 aprerya 2007. https://archive.kremlin.ru/text/appears/2007/04/125339.shtml (checked on March 25, 2020). 4. Federal’nyy zakon ot 3 oktyabrya 2018 goda № 350-FZ “O vnesenii izmeneniy v otdel’nyye zakonodatel’nyye akty Rossiyskoy Federatsii po voprosam naznacheniya i vyplaty pensiy”. 5. The basics of Russia’s pension system were defined in “Federalnyi zakon ot 15 dekabrya 2001g. N167-FZ ” and “Federalnyi zakon ot 17 dekabrya 2001g. N173-FZ ,” but since then numerous alterations have been made. 6. Vechernaya Moskva, No.37, February 3, 2007; Vechernii Peterburg, August 25, 2009. 7. Rasporyazhenie pravitel’stva RF ot 24.09.2001 No. 1270-r. 8. At that time, the childcare allowance, etc. was a flat 700 rubles (approx. 30 USD at that time), but this was increased to 1,500 rubles (approx. 70 USD at that time) for the first child and 3,000 rubles (just under 140 USD at that time) for the second child and subsequent children. As stated in “Federal’nyi zakon ot 1 marta 2008, No.18-FZ o vnesenii izmenenii v otdel’nye zakonodatel’nye akty Rossiiskoi Federatsii v tselyakh povysheniya razmerov otdel’nykh vidov sotsial’nykh vyplat i stoimosti nabora sotsial’nykh uslug,” these amounts are normally revised based on the rate of inflation. 9. Federal’nyi zakon ot 29 dekabrya 2006, No. 256-FZ o dopolnitel’nykh merakh gosudarstvennoi podderzhki semei, imeyushchikh detei.

References Becker, G. (1960). An economic analysis of fertility. In Demographic and economic change in developed countries (pp. 209–231). Princeton: Princeton University Press. Braithwaite, J. (1995). The old and new poor in Russia: Trends in poiverty (ESP Discussion Paper Series 21227). World Bank. McAuley, A. (1979). Economic welfare in the Soviet Union: Poverty, living standards, and equality. Madison: University of Wisconsin Press and Hertfordshire: George Allen & Unwin. Milanovic, B. (1997). Income, inequality, and poverty during the transition from planned to market economy. World Bank. Nemtsov, A. (2002). Alcohol-related human losses in Russia in the 1980s and 1990s. Addiction, 97 (11), 1413–1425. Ruminska-Zimny, E. (1997). Human poverty in transition economies: Regional overview for HDR 1997 . Human Development Report Office, United Nations Development Programme.

12

POPULATION AGING, LOW FERTILITY, AND SOCIAL …

299

Vishnevsky, A., & Bobylev, S. (Eds.). (2009). National human development report: Russian Federation 2008. Russia facing demographic challenges. Moscow: The United Nations Development Programme in Russian Federation.

Index

0–9 2013 Chinese Household Income Survey (CHIPs2013), 9, 107, 111, 124

B bipolarization, 8, 74, 88, 93, 94 Blinder-Oaxaca(decomposition) model, 46, 59, 62, 139, 151, 158

A activity motivation, 10, 162, 164, 173 actual employment rate, 138–140, 143, 145–150, 152 additional work capacity, 10, 135, 138, 145, 147, 150–153 age group, 7–9, 12, 20, 21, 23, 30, 42, 43, 45–47, 51–54, 56, 57, 59, 62, 63, 109, 112, 118, 134, 135, 138, 139, 141–143, 145–149, 151, 152, 167, 168, 170, 173, 218, 228, 229, 292 age-wage profile, 11, 188–191, 194–196 aging, 2, 3, 5, 10–12, 14, 15, 34, 44, 45, 107, 174, 187, 213, 247, 273, 283, 284, 292

C capacity, 10, 15, 134, 135, 139, 141, 143–153, 158 capital-intensive, 74, 76, 78, 91, 92, 94 caregiving, 262, 265, 268, 271, 274, 275 (child)birth, 14, 245, 283, 288, 294, 296, 297 childcare, 25, 244, 294–296, 298 childrearing, 288, 294–297 China, 1–9, 14, 20, 22, 27, 28, 30, 32–34, 36, 42, 47, 50–52, 54, 57, 60, 61, 66, 72, 73, 77, 78, 80–82, 85–87, 89–92, 94, 99, 101, 107, 108, 111, 120, 124–126, 219, 236, 241, 248, 250, 252

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 X. Ma (ed.), Employment, Retirement and Lifestyle in Aging East Asia, Social Policy and Development Studies in East Asia, https://doi.org/10.1007/978-981-16-0554-3

301

302

INDEX

China Employer-Employee Matching Survey (CEES), 8, 72, 73, 81–83, 85, 89–93, 95, 98, 99, 101 China Health and Nutrition Survey (CHNS), 6, 24, 27, 32, 35, 36 China Urban Labor Survey (CULS), 7, 42, 47, 49, 50, 55, 58, 60–62, 65, 67, 68 Chinese firm(s), 7, 8, 62, 72, 78, 85–87, 94 city contribution rate (CCR), 8, 73, 74, 80, 82–84, 88–93, 96, 97, 101 CMR model, 134, 135, 138, 139, 141, 150, 158 collectively-owned enterprises (COE), 123 Comprehensive Survey of the Living Conditions (CSLC), 9, 135, 136, 139, 141, 144, 145, 150, 158 Confucian, 13, 242, 244, 252, 253 continue working, 10, 133, 164, 165 Cox proportional hazards model, 13, 264, 267, 274 cultural disorder, 242, 244

D death, 14, 289–291, 296, 297 decomposition, 7, 9, 42, 44, 47, 48, 59–63, 65, 109, 111, 114, 119–123, 125, 126, 139, 148, 158 demographics, 13, 24, 241, 242, 246, 251, 253, 254, 262, 264, 289 diagnosed diseases, 137, 139, 141, 153, 156 discrimination hypothesis, 43, 47

E East Asia, 1–3, 5, 14, 15, 241, 243–245, 247, 250, 252–254, 258 economic disparity, 245 economic growth, 14, 94, 158, 241, 245, 284, 287, 290, 294, 296, 297 economic resources, 13 economy transition, 5, 8, 71 education, 25, 27, 35, 53, 60, 61, 82, 111–113, 116, 117, 119–128, 163, 164, 166, 168, 169, 171, 175, 180, 184, 194, 195, 197– 199, 202, 204, 206, 209–211, 251, 260, 262, 264–266, 268, 269, 272, 276, 277 the elderly, 3, 6, 9–12, 20, 107–109, 112, 113, 115–119, 124, 125, 133–135, 140, 149–153, 158, 164, 170, 174, 191, 193, 201, 202, 204, 213, 214, 230, 231, 236, 292 Elderly Employment and Recruitment Survey (EERS), 11, 192, 195, 196, 198, 200, 203, 205, 207, 209 elderly population, 242, 245–247, 250, 292, 294 eligibility age, 134, 135, 140, 152, 257, 260, 263 employment, 3, 5–15, 20–22, 24, 26, 41, 43–45, 50, 60, 62, 68, 72–95, 112, 125, 141, 152, 153, 167, 177, 181, 183, 187, 188, 190, 191, 203, 205, 210, 245, 251, 260, 262–265, 267, 273–276, 283, 295 employment function, 8, 72, 73, 78, 79, 82, 83, 99, 101 endogeneity, 20–22, 215

INDEX

endogeneity problem, 11, 20, 23, 76, 79, 80, 166, 183, 191, 203, 216 explained part, 8, 47, 59–63 external causes, 290 F familial and non-familial, 244 familial factors, 260, 262, 264, 265, 271 familial support, 242, 249–252, 254 family(household) income, 25 family structure, 13, 15, 23, 162, 163, 213, 215 family support, 9, 13, 258 female labor force participation, 242 fertility, 12, 13, 133, 161, 241–246, 251–254, 291, 295 firms’ actual contribution rate (FAR), 8, 73, 74, 80, 82–84, 87–94, 97, 101 firms’ actual social insurance contribution, 8, 74 firm size, 11, 50, 55, 57, 60, 61, 64, 74, 79–81, 83, 90, 97–99, 193–204, 206, 207, 209, 211 fixed effects (FE), 20, 22, 23, 26–30, 32, 33, 36 fixed-effects model, 6, 216, 236 foreign-owned enterprise (FOE), 5, 50, 57, 64, 72, 83, 89 F-test, 26, 36 full-time, 10, 14, 44, 135–137, 139, 144, 145, 151, 152, 262, 263, 265, 267, 268, 271–275 functional disabilities, 137, 141, 142, 155, 157 G gender, 12, 21–23, 25–28, 46, 50, 82, 109, 111, 117, 118, 120–127, 162, 163, 184, 210, 216, 217,

303

223, 230, 236, 242–244, 254, 258, 261, 262, 266–270, 272, 275 gender gap, 59, 117, 118, 216 gender role, 14, 243, 261, 262, 274, 275 Gini coefficient, 9, 107, 109, 111, 113, 124, 286 H happiness, 12, 215, 217–219, 223–235 Hausman test, 26, 36, 101, 144 Health, 6, 7, 9, 10, 19–24, 26–30, 32–34, 82, 134–145, 147–153, 158, 163, 166, 168–170, 172, 174, 181, 216, 218–220, 223–227, 230, 232–235, 250, 261, 262, 264–266, 268–274, 288, 290, 296, 297 health-based capacity, 9 healthcare, 6, 19, 34, 101, 288, 289, 291 health index, 138, 139, 141, 143, 155 Heckman two-step model, 6, 7, 22, 23, 26, 29, 32, 36, 46, 51, 53, 55 heterogeneity, 21, 22, 29, 68, 73, 79, 90, 134, 135, 149, 216, 218, 220 heterogeneity problem, 6, 20, 22, 23, 26, 29, 32–34, 215, 217, 218, 220, 228, 236 high-income, 110, 294 high-income group, 8–10, 170, 174 high wage and more employment (HWME), 8, 74, 79, 80, 83–94, 101 hospitalization, 137 household income, 10, 23, 25, 27, 35, 162–165, 167, 168, 170, 173, 174, 210, 218–220, 223–227,

304

INDEX

232–236, 265, 267, 268, 270, 272, 273, 285 Hukou, 5, 23, 25, 46, 50, 73 human capital, 3, 7, 10, 25, 43–48, 50, 56, 62, 162–171, 173–175, 190, 210 human capital theory, 19, 25, 43, 46, 47, 210

I income inequality, 5, 6, 9, 14, 41, 107–111, 122 income redistribution, 9, 10, 107, 108, 284, 287, 296 income-sharing, 9 industrial sector, 81, 196 inequality, 9, 14, 41, 63, 109–111, 113, 114, 116, 119–124, 134 instrument variables (IV), 21, 36, 79, 95, 101, 183, 216 inverse Mills ratio, 23

J Japanese firm, 188 Japanese Household Panel Survey (JHPS), 12, 218, 219, 224–227, 229, 232–235 Japan (Japanese), 1–6, 9–15, 68, 76, 134, 135, 137, 150, 158, 159, 161–164, 169, 171, 178, 180, 183, 187, 188, 190–192, 195, 196, 198–200, 202–207, 209, 210, 213–227, 229–236, 241–254, 258, 292

K Keio Household Panel Survey (KHPS), 12, 218, 219, 224–227, 229, 232–235 Korea, 12, 13, 241–254, 258

L labor demand, 3, 11, 15, 75–77, 188 labor force, 2, 3, 68, 81, 125, 133, 135, 140, 150, 188, 205, 251, 262, 263 labor force participation (LFP), 2–4, 12, 13, 19–24, 26–28, 30, 32–34, 36, 51, 133–135, 139, 140, 147, 150, 151, 159, 170, 182, 188, 244, 251, 252, 258 labor force shortage, 2, 11 labor insurance, 259, 260 labor-intensive, 74, 76, 91 labor-intensive firm, 8, 74, 78, 91, 92, 94, 95 labor market, 10, 12, 25, 43, 44, 50, 76–78, 125, 152, 164, 165, 170, 182, 194, 216, 258, 262, 275, 276, 294, 295 labor market segmentation, 25, 73, 101 Labor Pension Act, 259, 276 labor productivity, 12, 20, 43, 151, 189, 197, 210, 251 labor supply, 3, 7, 10, 20, 22, 25, 26, 34, 75–77, 150, 151, 167, 210 lagged variable (LV), 6, 22, 23, 26–30, 32, 33 Lazear (model), 188, 189, 193, 197, 201, 210 life expectancy, 13, 138, 139, 141, 142, 155, 158, 216, 220, 230, 236, 242, 246, 247, 288–290, 292 lifestyle, 5, 6, 12, 14, 15, 25, 143, 215, 254, 288, 290, 291, 297 linear probability model, 138, 141 live alone/living alone, 12, 215–220, 223–236, 242, 249, 250, 252, 254 living arrangement, 12, 14, 213, 215–217, 221–223, 250

INDEX

longitudinal data, 12, 218, 229, 274 low-income, 8, 10, 124, 170, 174 low wage and more employment (LWME), 8, 74, 79, 80, 83–94, 101 low wages and less employment (LWLE), 8, 79, 80, 83–85, 88–94, 101

M Maddala model, 11, 191, 197, 201 mandatory retirement, 15, 32, 61, 62, 135, 140, 150, 188, 189, 191, 192, 197, 198, 203, 205, 210, 259 mandatory retirement age, 11, 13, 25, 32, 36, 140, 152, 188–192, 197, 199, 200, 203, 259, 263 manual task, 43, 53, 62 market-oriented economy, 5 marriage (married), 13, 51, 55, 163, 167, 178, 216, 217, 230, 236, 243–245, 252, 261, 262, 264, 266, 268–272 medical insurance, 3, 5, 7, 34, 76 men, 6, 12–14, 21, 25, 26, 28, 29, 34, 117, 118, 134, 136, 138, 141–155, 158, 216, 218–220, 222, 223, 228–230, 236, 252, 257, 258, 261–263, 265, 267, 268, 270, 271, 273–275, 289, 294 middle-aged, 5–7, 10, 12, 15, 20, 22, 29, 32, 42, 43, 45, 52, 56, 59, 62, 161–163, 174, 184, 217, 218, 223, 229, 231 middle-aged worker, 7, 43, 47, 59–63 multinomial logistic regression (MLR) model, 79, 80, 82, 85, 89–92 multinomial logit model, 265, 271, 272, 277

305

N New Rural Cooperative Medical Scheme (NRCMS), 5 New Rural Residents’ Social Pension Insurance (NRRSP), 108, 113, 114, 116–119, 121–124, 126–129 New Rural Social Pension Scheme (NRSP), 5, 9, 109 nonprofit organizations (NPO), 10, 162–167, 169–176, 178–180, 182–184 non-regular worker, 20, 24 non-routine cognitive analytical task, 48, 53, 62 non-routine cognitive interpersonal task, 45, 48, 49 non-routine task, 42, 43, 56, 59, 61, 62

O occupation-specific insurance, 259 old-age insurance, 257, 259, 260 older adult(s), 2, 3, 6, 10–12, 14, 15, 22, 134, 161, 162, 187, 188, 210, 214, 215, 218, 229, 231, 236 older worker, 3, 7, 8, 11, 12, 15, 42, 43, 45, 47, 56, 59–63, 68, 188, 190–192, 201, 203, 205, 251, 254 ordered logistic regression model, 165 ordered probit regression model, 217–219, 224, 226, 228, 229, 232, 234 ordinary least squares (OLS), 23, 45, 46, 53, 65, 67, 78, 128, 191, 197 ownership, 25, 36, 50, 55, 57, 60, 61, 64, 74, 79, 81, 83, 88–92, 97, 98, 111, 115, 120–123, 125–127

306

INDEX

P Panel Study of Family Dynamics (PSFD), 13, 258, 263 partial equilibrium model, 75, 165, 182 part-time, 10, 14, 135–137, 139, 144, 145, 151, 152, 156–158, 265, 267, 268, 271–274 pension, 9, 10, 13, 14, 95, 108–129, 133, 161, 249–251, 257–259, 261, 264–270, 272–275, 284, 288, 291–294, 297, 298 pension benefit, 8, 9, 75, 77, 108, 109, 112, 113, 115–124, 135, 140, 150–152, 161, 216, 251, 293, 294, 297 pension benefit gap, 108–110, 112, 119, 121, 124, 125 pension benefit gap (inequality)/inequality of pension benefit, 107 pension coverage rate, 113, 117, 124 pension eligibility age, 11, 294 pension insurance, 72, 94, 98, 100, 101, 108, 124, 125 pension (insurance) system, 125 pension insurance type, 120–122, 125 pension types, 108–110, 119–123, 125, 126, 128 planned economy, 5 population, 1–7, 10–15, 20, 22, 34, 79, 107, 125, 136, 152, 158, 161, 167, 170, 172, 174, 179, 187, 213, 246, 247, 250, 254, 257, 263, 284, 285, 291, 292, 294, 297 population aging, 1, 2, 5, 6, 11, 12, 14, 19, 133, 188, 205, 213, 242, 245–247, 251, 254, 257, 258, 297

post-retirement employment, 13, 15, 258, 262, 263, 265, 267, 268, 271, 272, 274, 275 potential work capacity, 138, 147 poverty, 5, 14, 214, 231, 245, 248, 249, 283–287, 290, 292–294, 296 poverty rate, 242, 247, 248, 286, 287 pre-retirement job, 9, 13 pre-retirement work, 111, 123, 124 privately owned enterprise (POE), 50, 57, 64, 72, 83, 123 private sector, 50, 73, 76, 78, 88, 93, 95, 101, 152, 259, 260, 264 probit regression model, 6, 26, 27, 32, 46, 192, 197, 201 psychological distress, 137, 141, 153, 155–157 public pension, 3, 5, 9, 11, 32, 77, 133–135, 151–153, 188, 250, 251, 254 public sector, 13, 32, 47, 50, 73, 76, 78, 88, 93, 101, 260, 264, 266, 269, 270 public transfer, 9, 249, 251

R random effects (RE), 23, 218, 219, 225, 227–229, 233, 235 rate of city enforcement of social insurance (CE), 8, 73, 74, 80, 82, 84, 86, 88–93, 97, 101 reemployment age, 11, 188, 190–192, 201, 203, 205–207 regional disparities, 9, 51 regional gap (inequality), 116, 123 regular worker, 6, 7, 20, 22, 24, 26–30, 32–34, 165 replacement rate of pension, 77 Republic of Korea, 1–5, 14, 210 the residual, 111

INDEX

retirement, 6, 11–14, 47, 56, 62, 65, 111, 112, 115–117, 120, 121, 123, 125, 140, 144, 151, 170, 178, 181, 193, 251, 254, 257, 258, 260–264, 266–271, 273–276 retirement pension, 118, 257, 259, 260, 262, 263, 265, 268, 272, 273, 275 retirement system, 11, 13, 108, 193, 194, 196, 197, 199, 206, 258, 263 retirement timing, 13, 258, 260–263, 265–267, 271, 274, 275 return to task skill, 59, 61, 63 reverse causality problem, 6, 22, 32, 81 reverse causal relation problem, 20 reward, 10, 162–169, 172–174, 178, 180, 182 robustness checks, 6, 25, 32, 34, 218, 228, 229 routine cognitive task, 43, 62 routine task, 42, 44, 56, 59, 62 rural areas, 9, 108, 112, 113, 116–119, 121, 122, 124, 125, 250 rural elderly, 113, 117–119, 121 rural residents, 20, 22, 23, 118, 119, 125 Russia, 3, 5, 14, 44, 248, 283–285, 287–298 S sample selection bias, 11, 166, 197, 201, 203 sample selection (bias) problem, 22, 23, 46, 53, 191 self-employed, 12, 24, 50, 140, 158, 251, 263 self-rated health, 36, 137, 142, 153, 156

307

self-reported health (SRH), 24, 26–28, 30, 33, 141, 216, 264 seniority wage, 11, 12, 15, 68, 187–193, 196–198, 200–206, 210 simulation, 10, 135, 140, 141, 144, 150–153 single-person household, 12, 213, 214, 216, 231 skill gap, 7 smoking, 137, 139, 141, 155, 158 social activity, 162, 163, 173, 174, 176, 183 social insurance contribution, 8, 72–78, 80–82, 88, 90, 92–96, 100, 101 social insurance contribution rate, 75, 101 social insurance premium, 8, 72, 76, 90–95 social security, 5, 6, 8, 9, 12, 14, 71–74, 76, 84, 94, 95, 107, 108, 118, 125, 134, 135, 242, 249, 254, 283–285, 287, 289, 291, 295–297 socio-economic environment, 14 state-owned enterprises (SOE), 5, 32, 36, 50, 57, 64, 71, 72, 74, 83, 84, 88–90, 93, 108, 115, 117, 123 Subjective Well-Being (SWB), 215 T Taiwan, 5, 12–14, 241–247, 249, 250, 252–254, 257–259, 263–266, 268, 271, 274–277 Task I, 7, 46, 48, 49, 51–56, 58–63, 65, 67 Task II, 7, 46, 48, 49, 51, 52, 54–56, 58–63, 65, 67 Task III, 7, 46, 48, 49, 51, 52, 54–56, 58–63, 65, 67

308

INDEX

Task IV, 7, 46, 48, 49, 51–56, 58–63, 65, 67 task skill, 7, 8, 42, 43, 46, 51–54, 56, 59–63 task skill gap, 43, 45, 51, 52, 63 task type, 41, 42, 44–46, 51, 53, 56, 61, 62, 68 Theil index decomposition, 110, 119, 121 timing of retirement, 13, 15, 258, 270 Total Fertility Rate (TFR), 12, 242–247, 254 transition into pension, 264, 271 transition into retirement, 13, 263–265, 267, 268, 270, 271, 274

volunteering, 10, 161–171, 173, 174, 180

U unexplained part, 7, 47, 59–63 urban–rural gap, 116, 125 urban areas, 9, 108, 113, 119–122, 124, 125, 250, 273 urban elderly, 113, 115, 118–120 Urban Employee Basic Pension Insurance (UEBP), 108, 109, 113, 114, 116–124, 126, 128, 129 urban residents, 23, 25, 47, 50 Urban Residents’ Social Pension Insurance (URSP), 108, 109, 113, 114, 116–119, 122–124, 126–129 Urban-Rural Residents’ Social Pension Insurrance (URRSP), 108

W wage, 7, 8, 11, 15, 25, 41, 43–48, 50, 52–54, 56, 57, 59–63, 68, 71–96, 98–101, 134, 140, 163–165, 167, 178, 182, 188–191, 193, 194, 196–207, 209, 210, 245, 284, 295 wage function, 45–48, 53, 56, 65, 66, 76, 82, 190, 191, 193, 209, 210 wage gap, 7, 8, 41–47, 50, 52–54, 56, 59–63, 79, 163 well-being, 5, 11, 12, 14, 215–218, 220, 223–236, 261 willingness, 3, 10, 161–171, 173, 174, 183 women, 12–14, 21, 25, 26, 28, 29, 51, 117, 118, 136, 138–155, 158, 216–221, 223, 228–231, 236, 257, 258, 261–263, 265, 267, 268, 270, 271, 273–275, 277, 294, 297 working hour, 6, 7, 20–29, 31–34, 36, 82, 163, 165, 174, 181, 182, 193, 199, 204, 206, 262, 263 work (no work), 135, 136, 138, 139, 141, 144, 146, 151 work participation (participation in work), 6, 7, 20–30, 32–34, 162 work skill, 7, 8, 42 work skill gap, 7, 15, 42

V voluntary retirement, 13, 14, 259, 260, 265, 274

Y younger worker, 7, 45, 59, 62, 63