Social Security Programs and Retirement around the World: Historical Trends in Mortality and Health, Employment, and Disability Insurance Participation and Reforms 9780226921952

Citation preview

Social Security Programs and Retirement around the World

A National Bureau of Economic Research Conference Report

Social Security Programs and Retirement around the World Historical Trends in Mortality and Health, Employment, and Disability Insurance Participation and Reforms Edited by

David A. Wise

The University of Chicago Press Chicago and London

DAVID A. WISE is the John F. Stambaugh Professor of Political Economy at the Kennedy School of Government at Harvard University. He is the area director of Health and Retirement Programs and the director of the Program on the Economics of Aging at the National Bureau of Economic Research.

The University of Chicago Press, Chicago 60637 The University of Chicago Press, Ltd., London © 2012 by the National Bureau of Economic Research All rights reserved. Published 2012. Printed in the United States of America 21 20 19 18 17 16 15 14 13 12 1 2 3 4 5 ISBN-13: 978-0-226-90309-5 (cloth) ISBN-13: 978-0-226-92195-2 (e-book) ISBN-10: 0-226-90309-5 (cloth) ISBN-10: 0-226-92195-6 (e-book)

Library of Congress Cataloging-in-Publication Data Social security programs and retirement around the world : historical trends in mortality and health, employment, and disability insurance participation and reforms / edited by David A. Wise. pages cm—(National Bureau of Economic Research conference report) Includes bibliographical references and index. ISBN-13: 978-0-226-90309-5 (cloth : alkaline paper) ISBN-13: 978-0-226-92195-2 (e-book) ISBN-10: 0-226-90309-5 (cloth : alkaline paper) ISBN-10: 0-226-92195-6 (e-book) 1. Disability insurance. 2. Older people—Health and hygiene. 3. Retirement—Economic aspects. 4. Social security. I. Wise, David A., editor. II. National Bureau of Economic Research, sponsoring body. III. Series: National Bureau of Economic Research conference report. HD7091.S62443 2012 368.4′3—dc23 2011049671 o This paper meets the requirements of ANSI/NISO Z39.48-1992 (Permanence of Paper).

National Bureau of Economic Research Officers Kathleen B. Cooper, chairman Martin B. Zimmerman, vice-chairman James M. Poterba, president and chief executive officer Robert Mednick, treasurer

Kelly Horak, controller and assistant corporate secretary Alterra Milone, corporate secretary Gerardine Johnson, assistant corporate secretary

Directors at Large Peter C. Aldrich Elizabeth E. Bailey John H. Biggs John S. Clarkeson Don R. Conlan Kathleen B. Cooper Charles H. Dallara George C. Eads Jessica P. Einhorn

Mohamed El-Erian Linda Ewing Jacob A. Frenkel Judith M. Gueron Robert S. Hamada Peter Blair Henry Karen N. Horn John Lipsky Laurence H. Meyer

Michael H. Moskow Alicia H. Munnell Robert T. Parry James M. Poterba John S. Reed Marina v. N. Whitman Martin B. Zimmerman

Directors by University Appointment George Akerlof, California, Berkeley Jagdish Bhagwati, Columbia Timothy Bresnahan, Stanford Alan V. Deardorff, Michigan Ray C. Fair, Yale Franklin Fisher, Massachusetts Institute of Technology John P. Gould, Chicago Mark Grinblatt, California, Los Angeles

Bruce Hansen, Wisconsin–Madison Marjorie B. McElroy, Duke Joel Mokyr, Northwestern Andrew Postlewaite, Pennsylvania Uwe E. Reinhardt, Princeton Craig Swan, Minnesota David B. Yoffie, Harvard

Directors by Appointment of Other Organizations Bart van Ark, The Conference Board Christopher Carroll, American Statistical Association Jean-Paul Chavas, Agricultural and Applied Economics Association Martin Gruber, American Finance Association Ellen L. Hughes-Cromwick, National Association for Business Economics Thea Lee, American Federation of Labor and Congress of Industrial Organizations

William W. Lewis, Committee for Economic Development Robert Mednick, American Institute of Certified Public Accountants Alan L. Olmstead, Economic History Association John J. Siegfried, American Economic Association Gregor W. Smith, Canadian Economics Association

Directors Emeriti Andrew Brimmer Glen G. Cain Carl F. Christ George Hatsopoulos

Saul H. Hymans Lawrence R. Klein Paul W. McCracken

Rudolph A. Oswald Peter G. Peterson Nathan Rosenberg

Relation of the Directors to the Work and Publications of the National Bureau of Economic Research 1. The object of the NBER is to ascertain and present to the economics profession, and to the public more generally, important economic facts and their interpretation in a scientific manner without policy recommendations. The Board of Directors is charged with the responsibility of ensuring that the work of the NBER is carried on in strict conformity with this object. 2. The President shall establish an internal review process to ensure that book manuscripts proposed for publication DO NOT contain policy recommendations. This shall apply both to the proceedings of conferences and to manuscripts by a single author or by one or more coauthors but shall not apply to authors of comments at NBER conferences who are not NBER affiliates. 3. No book manuscript reporting research shall be published by the NBER until the President has sent to each member of the Board a notice that a manuscript is recommended for publication and that in the President’s opinion it is suitable for publication in accordance with the above principles of the NBER. Such notification will include a table of contents and an abstract or summary of the manuscript’s content, a list of contributors if applicable, and a response form for use by Directors who desire a copy of the manuscript for review. Each manuscript shall contain a summary drawing attention to the nature and treatment of the problem studied and the main conclusions reached. 4. No volume shall be published until forty-five days have elapsed from the above notification of intention to publish it. During this period a copy shall be sent to any Director requesting it, and if any Director objects to publication on the grounds that the manuscript contains policy recommendations, the objection will be presented to the author(s) or editor(s). In case of dispute, all members of the Board shall be notified, and the President shall appoint an ad hoc committee of the Board to decide the matter; thirty days additional shall be granted for this purpose. 5. The President shall present annually to the Board a report describing the internal manuscript review process, any objections made by Directors before publication or by anyone after publication, any disputes about such matters, and how they were handled. 6. Publications of the NBER issued for informational purposes concerning the work of the Bureau, or issued to inform the public of the activities at the Bureau, including but not limited to the NBER Digest and Reporter, shall be consistent with the object stated in paragraph 1. They shall contain a specific disclaimer noting that they have not passed through the review procedures required in this resolution. The Executive Committee of the Board is charged with the review of all such publications from time to time. 7. NBER working papers and manuscripts distributed on the Bureau’s web site are not deemed to be publications for the purpose of this resolution, but they shall be consistent with the object stated in paragraph 1. Working papers shall contain a specific disclaimer noting that they have not passed through the review procedures required in this resolution. The NBER’s web site shall contain a similar disclaimer. The President shall establish an internal review process to ensure that the working papers and the web site do not contain policy recommendations, and shall report annually to the Board on this process and any concerns raised in connection with it. 8. Unless otherwise determined by the Board or exempted by the terms of paragraphs 6 and 7, a copy of this resolution shall be printed in each NBER publication as described in paragraph 2 above.

Contents

1.

2.

3.

4.

5.

Acknowledgments

ix

Introduction and Summary Kevin Milligan and David A. Wise

1

Disability, Health, and Retirement in the United Kingdom James Banks, Richard Blundell, Antoine Bozio, and Carl Emmerson Disability Insurance, Population Health, and Employment in Sweden Lisa Jönsson, Mårten Palme, and Ingemar Svensson Health, Disability, and Pathways into Retirement in Spain Pilar García-Gómez, Sergi Jiménez-Martín, and Judit Vall Castelló Health Status, Welfare Programs Participation, and Labor Force Activity in Italy Agar Brugiavini and Franco Peracchi Disability Programs, Health, and Retirement in Denmark since 1960 Paul Bingley, Nabanita Datta Gupta, and Peder J. Pedersen

41

79

127

175

217

vii

viii

6.

7.

8.

9. 10.

11.

12.

Contents

Disability in Belgium: There Is More Than Meets the Eye Alain Jousten, Mathieu Lefebvre, and Sergio Perelman Disability, Pension Reform, and Early Retirement in Germany Axel Börsch-Supan and Hendrik Jürges Disability and Social Security Reforms: The French Case Luc Behaghel, Didier Blanchet, Thierry Debrand, and Muriel Roger Disability Insurance Programs in Canada Michael Baker and Kevin Milligan

251

277

301

327

The Long-Run Growth of Disability Insurance in the United States Kevin Milligan

359

Disability Pension Program and Labor Force Participation in Japan: An Historical Perspective Takashi Oshio and Satoshi Shimizutani

391

Disability Insurance and Labor Market Exit Routes of Older Workers in the Netherlands Klaas de Vos, Arie Kapteyn, and Adriaan Kalwij

419

Contributors Author Index Subject Index

449 453 455

Acknowledgments

Funding for this project was provided by the National Institute on Aging, grant numbers P01-AG005842 and P30-AG012810 to the National Bureau of Economic Research. The views expressed herein are those of the authors and do not necessarily reflect the views of the National Institute on Aging, the National Institutes of Health, or the National Bureau of Economic Research.

ix

Introduction and Summary Kevin Milligan and David A. Wise

Through the coordination of work by a team of analysts in twelve countries, the International Social Security Project has used the vast differences in social security programs across the countries as a natural laboratory to study the effects of program provisions on the labor force participation of older persons. The program results have shown that in most countries the provisions of social security programs per se, along with related programs, provide strong incentive to leave the labor force at young ages, and that reducing the inducement to leave the labor force can yield very large improvements in the financial position of government budgets. The work to date has also made clear that disability insurance programs play an especially large role in the departure of older persons from the labor force, as many pass through disability insurance (DI) on their path from employment to retirement. Thus, with this volume we have begun a series of analyses to focus particular attention on the effect of disability programs on retirement and the potentially large effects that changes in these programs could have on the labor force participation of older workers. This issue is particularly pressing given demographic trends that will increase the cost of social security and health care programs and are likely to increase the need for people in many countries to prolong their working lives. In almost every industrialized country the population is aging rapidly Kevin Milligan is associate professor of economics at the University of British Columbia and a research associate of the National Bureau of Economic Research. David A. Wise is the John F. Stambaugh Professor of Political Economy at the Kennedy School of Government at Harvard University. He is the area director of Health and Retirement Programs and director of the Program on the Economics of Aging at the National Bureau of Economic Research. The introduction has benefited from comments by the country participants. For acknowledgments, sources of research support, and disclosure of the authors’ material financial relationships, if any, please see http://www.nber.org/chapters/c12381.ack.

1

2

Kevin Milligan and David A. Wise

and individuals are living longer. These demographic trends have placed enormous pressure on the financial viability of the pay- as-you- go social security systems in these countries. The financial pressure has been compounded, until recently, by retirement at younger and younger ages, one of the most dramatic features of labor force change over the past several decades. Because of population aging and longer life expectancies, a larger fraction of the population is receiving social security benefits, compared to the fraction of persons in the labor force and paying for the benefits. It seems likely that the social security systems in many countries will have to be reformed to be financially viable. We have found in our work to date that in many countries disability programs essentially provide early retirement benefits before the social security eligibility age. Figure I.1 shows the proportion of men collecting disability benefits at age forty- five and at age sixty- four for eight countries. Within each of the two groups of countries, the proportion collecting disability benefits is about the same at age forty- five—2 or 3 percent in the first group and 4 or 5 percent in the second group. Yet in both groups, by age sixty- four there are large differences in the proportion collecting DI benefits (or other programs such as sickness insurance programs, which in some countries are a gateway to DI). It is implausible that the rate of physical disability varies so much among these industrialized countries.1 This volume presents analysis of historical trends in our group of countries that is intended to set the stage for further, more formal analysis of disability insurance programs. This is the fifth phase of the ongoing project and the first in a series of volumes on disability insurance. The first phase of the project described the retirement incentives inherent in plan provisions and documented the strong relationship across countries between social security incentives to retire and the proportion of older persons out of the labor force (Gruber and Wise 1999). The second phase, based on microeconomic analysis of the relationship between a person’s decision to retire and the program incentives faced by that person, documented the large effects that changing plan provisions would have on the labor force participation of older workers (Gruber and Wise 2004). The third phase (Gruber and Wise 2007) demonstrated the consequent fiscal implications that extending labor force participation would have on net program costs—reducing government social security benefit payments and increasing government tax revenues. The analyses in the first two phases, as well as the analysis in the third phase, are summarized in the introduction to the third phase. In the fourth phase (Gruber and Wise 2010) we directed attention to the 1. The data for France are for ages fifty and fifty- nine. The proportion in France rises from 5 percent at age fifty to 18 percent by age fifty- nine. The proportion falls to 3 percent by age sixty- one and to 1 percent by age sixty- six as retirees take up normal retirement benefits.

Introduction and Summary

3

Fig. I.1 Proportion of men collecting disability benefits, by age (forty-five and sixty-four) Note: France data are for ages fifty and fifty- nine.

oft- claimed proposition that incentives to induce older persons to retire— inherent in the provisions of social security systems—were prompted by youth unemployment. Many have worried that if the incentives to retire were removed, and older persons stayed longer in the labor force, the job opportunities of youth would be reduced. We found no evidence to support this “boxed economy” proposition. In short, we concluded: “the overwhelming weight of the evidence, as well as the evidence from each of the several different methods of estimation, is contrary to the boxed economy proposition. We find no evidence that increasing the employment of older persons will reduce the employment opportunities of youth and no evidence that increasing the employment of older persons will increase the unemployment of youth” (42). The results of the ongoing project are the product of analyses conducted for each country by analysts in that country. Researchers who have participated in the project are named in the following list. The authors of the country chapters in this volume are listed first; others who have participated in one or more of the first four phases are listed second and shown in italics: Belgium: Alain Jousten, Mathieu Lefèbvre, Sergio Perelman, Pierre Pestieau, Raphaël Desmet, Arnaud Dellis, and Jean-Philippe Stijns Canada: Michael Baker, Kevin Milligan, and Jonathan Gruber Denmark: Paul Bingley, Nabanita Datta Gupta, and Peder J. Pedersen

4

Kevin Milligan and David A. Wise

France: Luc Behaghel, Didier Blanchet, Thierry Debrand, Muriel Roger, Melika Ben Salem, Antoine Bozio, Ronan Mahieu, Louis-Paul Pelé, and Emmanuelle Walraet Germany: Axel Börsch-Supan, Hendrik Juerges, Reinhold Schnabel, Simone Kohnz, and Giovanni Mastrobuoni Italy: Agar Brugiavini and Franco Peracchi Japan: Takashi Oshio, Satoshi Shimizutani, Akiko Sato Oishi, and Naohiro Yashiro Netherlands: Adriaan Kalwij, Arie Kapteyn, and Klaas de Vos Spain: Pilar García-Gómez, Sergi Jiménez-Martín, Judit Vall-Castelló, Michele Boldrin, and Franco Peracchi Sweden: Lisa Jönsson, Mårten Palme and Ingemar Svensson United Kingdom: James Banks, Richard Blundell, Antonio Bozio, Carl Emmerson, Paul Johnson, Costas Meghir, and Sarah Smith United States: Kevin Milligan, Jonathan Gruber, Courtney Coile, and Peter Diamond An important goal of the project has been to present results that were as comparable as possible across countries. Thus, the chapters for each phase were prepared according to a detailed template that we prepared in consultation with country participants. In this introduction, we summarize the collective results of the country analyses and borrow freely from the country chapters. In large part, however, the results presented in the introduction could only be conveyed by combined analysis of the data from each of the countries. The country chapters themselves present much more detail for each country and, in addition to template analyses performed by each country, often present country- specific analysis relevant to a particular country. These country- specific analyses are especially important in this phase because the available data varied considerably from country to country. In addition, the country chapters typically present results separately for both men and women. In this introduction, however, we usually present results for men only, except in cases where it is important to emphasize different effects of reforms, for example, by gender. The historical analysis in this volume is intended to set the stage for more formal analysis of disability insurance programs. The key question that future analysis will address is this: Given health status, to what extent are the differences in labor force participation (LFP) across countries determined by the provisions of disability insurance programs? This question raises several issues: (a) we need to develop measures of health that are comparable across countries, (b) we need to understand to what extent disability insurance reforms are prompted by health status in a country, and (c) we need to understand whether disability insurance provisions are prompted by the employment circumstances of older people in each country. To this end, we first consider trends in health status over the past four

Introduction and Summary

5

or five decades. One of the important challenges we face in cross- country comparison of disability insurance (DI) programs is that commonly used health measures are not comparable across countries. For example, selfassessed health (SAH) status is known to be determined in important part by country- specific response effects. One important measure of health status that is comparable across countries and across time within countries, however, is mortality. Thus we begin in section I.1 by considering changes in mortality over time, and in particular the relationship between mortality and labor force participation. In section I.2 we consider how mortality is related to self- assessed health, perhaps the most widely used single measure of health status within countries. In particular we consider how the change in mortality is related to change in SAH. In section I.3 we consider how mortality is related to other indicators of health status. With a view to issue (b), in section I.4 we consider how DI reforms are related to changes in health, in particular measured by changes in mortality. Then in section I.5, with a view to both issues (b) and (c), we consider DI reforms as natural experiments—that are not prompted by changes in the health of older persons or changes in the employment circumstances of older persons—and we show that these “exogenous” reforms can have a very large effect on the labor force participation of older workers.2 I.1

Mortality and Labor Force Participation

We begin with an examination of mortality and how it relates to labor force participation. We take mortality here as an indicator of health, and the analysis aims to uncover information on how labor force participation at older ages is affected by changes in health. Mortality is chosen as a starting point because mortality data is readily available and well measured in each country. Moreover, without reservation, mortality can be compared both across countries and within countries over time. We realize, of course, that mortality is not the same as morbidity or other measures of health status that may be based on summaries of a large number of health indicators. In section I.2 we consider how mortality is related to SAH, perhaps the most commonly used measure of health. The development of health measures that are comparable across countries will be a major part of our further, more formal analysis of disability insurance programs. In this section we explore mortality trends in our twelve industrialized countries, with emphasis on mortality and labor force participation. We begin by documenting long- run trends in mortality, using several graphical 2. The results reported here are all based on within- country historical analysis. As the project progresses, more of the analysis will rely on the nexus of more recent and comparable longitudinal studies—Health and Retirement Study (HRS) in the United States, the English Longitudinal Study of Aging (ELSA), the Survey of Health, Aging and Retirement in Europe (SHARE), and the Japan Study of Aging and Retirement (JSTAR).

6

Kevin Milligan and David A. Wise

approaches. We then compare the mortality trends to labor force participation. Decline in Mortality: We begin with a standard way to show mortality decline. Figure I.2 shows the decline in mortality at age sixty- five over the past six decades in each of the countries. Two features of the data stand out. First, in most of the countries the decline did not begin until about 1970.

Fig. I.2

Decline in mortality at age sixty-five, by country

Introduction and Summary

7

Indeed in several countries—Italy, Denmark, Germany, and the Netherlands—mortality increased between 1950 and 1970. Second, thereafter, the decline is generally greater for countries that had the highest mortality rates in 1970. Thus there seems to be some convergence of mortality rates in these countries since 1970. The country chapters show these data for ages fiftyfive, sixty, and sixty- five. Equivalent Mortality Ages: There are several ways to present mortality data that highlight more clearly the implications of mortality declines for health. One way is to ask how old you had to be today to have the same mortality as a person of a given age in an earlier year. For example, consider the mortality rate of persons aged sixty- five in the 1960s. Then consider the age at which the same mortality rate occurred in later years. Figure I.3 shows the ages of equivalent mortality in the United Kingdom. Men at age seventy- four in 2007 had about the same mortality rate as men aged sixtyfive in the 1960s. The difference is about 9.8 years. That is, by this measure, men aged seventy- four felt about as old as men aged sixty- five four or five decades earlier. Figure I.3 shows the same comparison for the United States and France; the differences are 9.2 and 9.6 years, respectively. Thus, by this measure, there has been a very large improvement in health (a reduction in the mortality rate) over the past several decades. Following, we compare equivalent mortality ages to equivalent self- assessed health ages, which may be closer to healthy equivalent ages. Table I.1 shows the age in 2005 with the same mortality as sixty- fiveyear- old men in the early 1960s for each of the countries. Although in each country the equivalent mortality age in 2005 was substantially greater than age sixty- five mortality in the early 1960s, there is also substantial variation in the 2005 equivalent age—from a low of age 69.7 and an increase of 4.7 years in the Netherlands to a high of age seventy- five and an increase of ten years in Japan. Figure I.4 shows the gain in equivalent mortality years (the equivalent mortality age in 2005 minus sixty- five) was strongly related to initial mortality at age sixty- five in the early 1960s. The increase was greater for countries with the highest mortality in the 1960s. Like figure I.2, this figure suggests some convergence of mortality rates in these countries. Employment by Age: We next present a series of figures to describe the change over time in the relationship between employment and mortality and the relationship between employment and age. We begin with employment by age now, and then turn to employment by mortality and how it has changed over time. Figures I.5 and I.6 show employment by age. The key conclusion is that while the employment rate was similar for most of the countries through age fifty, because of differences in retirement programs, the employment rate varies a great deal by age sixty- five. Figure I.5 shows employment by age in the eleven countries.3 In nine of the eleven countries, 3. Italy is excluded because data for Italy are available only for men and women combined.

A

B

C

Fig. I.3 How old do you need to be to “feel like” a sixty-five-year-old in the 1960s? A, Men in the UK; B, Men in the United States; C, Men in France

Introduction and Summary

9

Table I.1

Gain in equivalent mortality age, early 1960s to 2005, for men aged sixty-five in initial year

Country

First year

Mortality rate in first year (%)

Equivalent mortality age in 2005

Gain in years

1960 1961 1961 1960 1960 1960 1960 1960 1960 1960 1960 1960

3.53 3.26 2.69 3.22 4.15 3.06 3.56 2.35 3.54 2.37 3.53 3.84

72.8 73.4 70.0 73.5 73.2 72.8 75.0 69.7 71.9 71.4 73.4 74.1

7.8 8.4 5.0 8.5 8.2 7.8 10.0 4.7 6.9 6.4 8.4 9.1

Belgium Canada Denmark France Germany Italy Japan Netherlands Spain Sweden United Kingdom United States

Fig. I.4 Gain in equivalent mortality age 1960 to 2005 versus age sixty-five mortality rate in 1960 for men

employment in the forty- five to forty- nine age interval was between 85 and 89 percent, but by age interval sixty to sixty- four the employment rate varied from 18 to 56 percent, and (for seven of the eleven) was between 2.6 to 31 percent in the sixty- five to sixty- nine age interval. Figure I.6 shows employment by year for seven of the twelve countries. In these seven countries the employment rate was between 83 and 90 percent at age fifty, but by age sixtyfour the employment rate varied from about 10 to 59 percent. Mortality by Age: We now turn to employment by mortality and begin by describing mortality by age across the countries. Figure I.7 shows mortality

10

Kevin Milligan and David A. Wise

Fig. I.5

Employment by age interval in 2004, in eleven countries, men

Fig. I.6

Employment by age for seven countries, men

for selected ages in 2005 in the twelve countries. While there are differences across countries, the variation appears small relative to differences in employment by age. Figure I.8 shows mortality by age for each country. The differences across countries are essentially bounded by the lowest mortality rate in Japan and the highest rate in the United States. The country

Introduction and Summary

Fig. I.7

Mortality at selected ages in 2005, by country, for men

Fig. I.8

Mortality by age by country for men, 2005

11

trends are ordered by mortality at age seventy, from high to low as indicated in the legend. Employment by Mortality: We want to understand how employment varies across countries, given health. Again, we use mortality as our measure of health, sticking with a measure that is comparable across countries. We then compare the employment rate across countries for given levels of mortality.

12

Kevin Milligan and David A. Wise

Fig. I.9

Employment versus mortality, selected countries, by one-year intervals

Figure I.9 shows employment by mortality for seven countries by single- year age intervals. At the ages at which the mortality rates in these countries were about 0.5 percent in each of the countries, the employment rates are very similar in the countries, ranging from about 0.82 to 0.90. But as the mortality rate increases, the divergence across the countries increases. For example, at the age at which the mortality rate in each of the countries was 1.5 percent, the employment rates varied from about 7 percent in France to about 50 percent in the United States. Like the increasing divergence in employment rates by age, the divergence as persons age and mortality increases reflects the large variation in the provisions of retirement pathways. Figure I.10 shows mortality rates for ten countries (excluding Sweden and Italy, which are included in figure I.9) by five- year age intervals. The general features of the figure are the same as those of figure I.9. The cross- country employment rates diverge as mortality increases. This pattern is especially apparent for the seven countries with similar employment rates—ranging from 0.85 to 0.89—when the mortality rate is around 0.2 percent. When the mortality rate was about 1.1 percent in each country, the employment rate ranged from 0.17 to 0.58. To understand the cross- country comparisons in figures I.9 and I.10, it helps to consider in more detail the relationship between employment and mortality. Figure I.11 shows the relationship between age and employment in 1977 and 2007. As highlighted, the employment rate at ages sixty- two and sixty- three changed little over this thirty- year period. But at these ages the mortality rates declined substantially between 1977 and 2007—about 1 percent at each age. Figure I.12 presents a different view of the data, showing the employment rate by mortality in 1977 and 2007. Consider first the

Fig. I.10

Employment versus mortality in 2004 by five-year intervals

Fig. I.11 Employment and mortality by age, men in the United States, 1977 and 2007

14

Kevin Milligan and David A. Wise

Fig. I.12

Employment by mortality, men in the United States, 1977 and 2007

age at which 50 percent of men were employed. In 2007, the mortality rate when 50 percent of men were employed was 2.7 percent; thirty years later in 2007, the mortality rate was only 1.5 percent, a decline of 1.2 percent. That is, for the employment rate to be 50 percent in 2007, men had to be much healthier (by the mortality measure) than they were in 1977. Looking at the data another way, at the age at which the mortality rate was 1.5 percent, in 1977 about 80 percent of men were employed but only 50 percent in 2007. Figures I.13 and I.14 show comparable figures for France. Figure I.13 shows that for France, the employment rate at ages sixty- two and sixty- three declined by almost 30 percent. In addition, the mortality rate at these ages declined by about 1 percent as well, about the same as in the United States. Figure I.14 shows that in France at the age at which 30 percent of men were employed the mortality rate was 2.7 percent in 1977 but only 1.1 percent in 2007, a decline of 1.6 percent. That is, for the employment rate to be 30 percent in 2007, men had to be much healthier (by the mortality measure) than they were in 1977. Further, looking at the data another way, at the age at which the mortality rate was 1.1 percent, 90 percent of men were employed in 1977 but only 30 percent in 2007. Figures I.15 and I.16 show comparable figures for the United Kingdom. Figure I.15 shows that for the United Kingdom, the employment rate at ages sixty- two and sixty- three declined by about 20 percent. In addition, the mortality rate at these ages declined by about 1 percent as well, about the same as in the United States and France. Figure I.16 shows that in the

Fig. I.13

Employment and mortality by age, men in France, 1977 and 2007

Fig. I.14

Employment by mortality, men in France, 1997 and 2007

Fig. I.15 Employment and mortality by age, men in the United Kingdom, 1977 and 2007

Fig. I.16

Employment by mortality, men in the United Kingdom, 1977 and 2007

Introduction and Summary

17

Fig. I.17 Proportion of men employed when the mortality rate is 1.5 percent, selected countries, various years

United Kingdom at the age at which 30 percent of men were employed the mortality rate was about 3.1 percent in 1977 but only about 1.5 percent in 2007, a decline of 1.6 percent. That is, for the employment rate to be 30 percent in 2007, men had to be much healthier (by the mortality measure) than they were in 1977. Further, looking at the data another way, at the age at which the mortality rate was 1.5 percent, almost 90 percent of men were employed in 1977 but only about 30 percent in 2007. The data necessary to make figures like these for all countries are not now included in the data files for each country. We can, however, show variation across countries like that shown in the left- hand vertical bars in figures I.12, I.14, and I.16. Figure I.17 shows the employment rate in years in the 2000s for men in eight countries when the mortality rate was 1.5 percent. The rate varies from 4.5 percent in France to 50 percent in the United States. The data in this figure are comparable to the data shown in figures I.9 and I.10, but those figures also show the employment rates when the mortality rate was much lower and the employment rate was similar across countries. This figure shows a very large dispersion of employment rates across countries at the 1.5 percent mortality rate. I.2

Mortality versus Self-Assessed Health

The previous figures highlight the changing relationship between mortality and employment. As emphasized earlier, mortality lends itself to these comparisons because mortality is comparable across time and across time within countries. Other potential measures of health status are not comparable across countries and may not be comparable across time within

18

Kevin Milligan and David A. Wise

countries. Nonetheless, we would like to explore the relationship between mortality and another commonly used summary measure of health status. Perhaps the most commonly used measure is self- assessed health status (SAH). We compare these two measures in several ways. We first compare SAH and future mortality in the United States. Second, we compare mortality equivalent ages over time with SAH equivalent ages over time. We illustrate this idea with data for two countries, the United States and Sweden. Third, we consider within countries the time trend relationship between mortality and SAH. Finally, we combine the latter comparisons to estimate the cross- country relationship between the change in mortality over time with the comparable change in SAH over the same time period. SAH and Future Mortality in the United States: Table I.2 shows that in the United States Health and Retirement Study (HRS) SAH in 1992 is strongly predictive of subsequent mortality. The table shows the proportion of men and women that are deceased by 1996, 2002, and 2008. For example, for men in excellent health only 11.4 percent are deceased by 2008 compared to 57.9 percent for men in poor health. Mortality Equivalent Ages versus SAH Equivalent Ages—United States: Figure I.18 compares these two measures for the United States. First, the figure shows mortality by age in 1977 and in 2007. The figure shows that a person aged seventy- seven in 2007 had about the same mortality rate as a person aged sixty in 1977, a difference of about seven years. Second, the figure shows the proportion of people who reported they were in fair or poor health, by age, in the 1970s and in the 2000s. Comparing these two trends, men who were sixty- nine in the 2000s had about the same SAH as men who Table I.2

Percentage of HRS respondents aged fifty-one to sixty-one in 1992 who are deceased by the beginning of each wave, by SAH in 1992, United States Self- reported health in 1992

Year

Excellent (%)

Very good (%)

Good (%)

Fair (%)

Poor (%)

1996 2002 2008 Percent in category

1.0 5.8 11.4 24.5

Men 1.5 7.2 15.6 29.6

2.1 13.3 25.8 27.8

4.3 22.1 36.5 11.5

10.8 36.9 57.9 6.6

1996 2002 2008 Percent in category

0.4 2.6 6.4 23.1

Women 0.8 5.4 10.3 30.7

0.8 7.1 15.2 25.7

2.1 15.2 28.8 13.6

4.2 24.4 36.8 6.9

Introduction and Summary

Fig. I.18

19

Mortality and self-assessed health 1970s and 2000s, United States

were over nine years younger, aged sixty, in the 1970s. Thus, in this example, there was a greater difference in age equivalent SAH than in age equivalent mortality (about nine versus seven years). Mortality Equivalent Ages versus SAH Equivalent Ages—Sweden: Figure I.19 compares these two, mortality and SAH, for Sweden. Mortality and SAH, for Sweden are only available for age- groups forty- five to fifty- four, fifty- five to fifty- nine, and sixty to sixty- four. The data used in the chapter for Sweden are for 1976 and 2005. For each of these years the average for the age intervals is plotted in figure I.19 at the midpoint of the intervals. These values are shown by the three markers for SAH and mortality for each of the two years. Then the markers are fitted—the mortality markers with a power functional form and the SAH markers with a linear functional form. First, based on these rough approximations, the figure shows that a person aged sixty- three in 2005 had about the same mortality rate as a person aged fifty- five in 1976, a difference of about eight years. Second, the figure shows the proportion of people who reported they were in poor health, by age, in 1976 and in 2005. Comparing these two trends, men who were about sixtyfour in 2005 had about the same SAH as men who were about nine years younger, aged fifty- five, in 1976. Thus, in this example, there is also a greater difference in age equivalent SAH than in age equivalent mortality (about nine versus eight years). Thus, in both the United States and Sweden, there appears to be a sub-

20

Kevin Milligan and David A. Wise

Fig. I.19

Mortality and self-assessed health 1976 and 2005, Sweden

stantial correspondence between age equivalent mortality and age equivalent SAH. Within-Country Trends in Mortality versus SAH: The within- country trends in several countries are shown in figure I.20 for the United States, Sweden, the Netherlands, Canada, and the United Kingdom, respectively. For each of these countries, with the exception of the United Kingdom, there seems to be a close correspondence between the trend in mortality and the trend in SAH. The SAH measure in the United Kingdom, however, differs from the measure used in the other countries—the United Kingdom uses “bad health” versus “fair or poor health” in the other countries. It is important to note that, with the exception of the United Kingdom, even though there seems to be a systematic relationship between mortality and SAH within countries, it can be seen from the levels of the SAH values that the proportion of persons saying they are in fair or poor health varies greatly across countries. Cross-Country Comparison between the Percent Change in Mortality and the Percent Change in SAH: Figure I.21 summarizes the relationship across nine countries for which both series are available over some time period. It compares the percent change in “fair- poor” health with the percent change in mortality. For each country the time period spans the year of the first SAH observation to the year of the last SAH observation. For example, the period for the Netherlands is 1983 to 2008. The percent change in SAH is the percent difference between the 1983 value and the 2008 value. The percent change in mortality is over the same time period. For the nine countries that report the proportion in fair or poor health (or one minus the proportion in better health) there is a close relationship between the change in

A

B

C

Fig. I.20 Mortality versus “fair-poor” self-assessed health: A, Men in the United States; B, Men in Sweden; C, Men in the Netherlands; D, Men in Canada; E, Men in the United Kingdom

22

Kevin Milligan and David A. Wise

D

E

Fig. I.20

(cont.)

mortality and the change in SAH. The SAH is not available for Italy. The data points for France and the United Kingdom use different measures of health—the United Kingdom reports the percent in bad health and France reports the average of a ten- point scale of health status. This suggests a fairly tight within-country relationship between improvements in mortality and improvements in self- assessed health, providing a link between our earlier mortality analysis and one commonly used health measure. We emphasize again, however, that perusal of the previous figures reveals that the level of SAH varies greatly from country to country, consistent with substantial country- specific SAH response effects.

Introduction and Summary

23

Fig. I.21 Percent change in “fair-poor” health versus percent change in mortality, men aged sixty to sixty-four

I.3

Mortality versus Other Health Measures

The evidence presented in section I.2 suggests that on balance there is a substantial relationship between within- country trends in SAH and trends in mortality. Here we consider how mortality is related to other health measures. While most countries have some data on SAH, the data on other health measures varies a great deal from country to country and the evidence on the relationship of the various measures to mortality is mixed. To illustrate the various data series and their relationship to mortality we show examples from several countries—Denmark, Italy, Spain, Sweden, and Japan. Figure I.22 shows mortality rates for cancer, heart, brain, and respiratory causes in Denmark. Among these objective measures of health, only heart disease shows a strong and consistent decline. Figure I.23 shows mortality rates by cause for cancer, circulatory diseases, heart, cirrhosis, and respiratory causes in Italy, with especially large declines in the first three. Figure I.24 shows the trend in the incidence of major illnesses—cholesterol, high blood pressure, allergies, diabetes, heart problems, and respiratory problems—in Spain. The data show an increase in high blood pressure and cholesterol and little change in the others between 1987 and 2006. Figure I.25 shows trends in circulatory disease, back pain, and long- term disease in Sweden, with an increase in circulatory disease and little change in the others. Figure I.26 shows trends for feel not good or sick, Activities of Daily Life, and SAH, with a downward trend in each of them.

Fig. I.22 Mortality rates for selected causes for men aged sixty to sixty-four in Denmark

Fig. I.23

Causes of death of men aged fifty-five to sixty-four in Italy

Fig. I.24

Incidence of major illnesses for men aged sixty to sixty-four in Spain

Fig. I.25

Incidence of major diseases in Sweden for men aged sixty to sixty-four

26

Kevin Milligan and David A. Wise

Fig. I.26

I.4

Health indicators, Japanese men

Mortality versus DI Participation

In this section we emphasize the general absence of a relationship between DI participation and health, as indicated by trends in mortality. The evidence suggests that DI provisions, which determine participation rates, are essentially a train on their own track. That is, DI provisions seem not to be endogenous with respect to health as indicated by mortality. To summarize the evidence across countries, figure I.27 shows the percent change in DI participation against the percent change in mortality between 1980 and 2005. There appears to be little relationship. Although there was a wide variation across countries in the percent reduction in mortality over this period (from about – 30 percent to – 50 percent), the change in DI participation shows little relationship to the reduction in mortality. To provide more detail on the trends within countries, figures I.28, I.29, I.30, I.31, and I.32 show the trends in mortality and DI participation in five countries—the United Kingdom, Canada, Italy, Sweden, and the United States, respectively. It appears that the trend in DI participation in these countries is unrelated to the trend in mortality. Given that SAH trends in the same direction as mortality in most countries for which we have data, as shown in section I.2, it appears that there is also little relationship between SAH and DI participation.

Fig. I.27 Percent change in DI participation versus percent change in mortality, early 1980 to 2005, men

Fig. I.28

Trends in mortality and DI participation in the United Kingdom

28

Kevin Milligan and David A. Wise

Fig. I.29

Trends in mortality and DI participation in Canada

Fig. I.30

Trends in mortality and DI participation in Italy

I.5

“Natural Experiments”: DI Reform and DI Participation

In future analysis, we will consider the relationship between DI provisions and labor force participation. In particular we will ask: Given health status, to what extent are differences across countries in the relationship between health status and LFP determined by the provisions of disability insurance programs? There are two key issues that are important in determining how

Introduction and Summary

Fig. I.31

Trends in mortality and DI participation in Sweden

Fig. I.32

Trends in mortality and DI participation in the United States

29

the results of this analysis can be interpreted. One is whether the relationship between labor force participation and DI provisions can be interpreted to represent the causal effect of the provisions on LFP. If, for example, DI reforms were adopted in part in response to low LFP of older workers, this reverse causality (endogeneity) would imply that to some extent the DI pro-

30

Kevin Milligan and David A. Wise

visions were adopted in response to low LFP rather than low LFP being the result of DI provisions. The second issue relates to DI provisions and health status. Given health status, we will want to estimate the effect of DI provisions on the likelihood that a person is employed. But, if DI provisions in a country were to some extent adopted in response to lower health status, we would overestimate the effect of DI provisions on LFP, given health status. Earlier, we have shown substantial evidence that DI reforms in a country do not seem to follow changes in health status in the country. Here we present several “natural experiments” showing the effect on LFP of reforms that were not prompted by changes in the health of older people or by changes in the LFP of older people. For example, the DI reform in Belgium was prompted by a court ruling. We also emphasize reforms in which the effect was on one gender but not the other or on persons of a given age and not other ages, but in which the reforms were not intended to target these groups. We consider reforms in Canada, Germany, Belgium, Sweden, France, and Denmark. Canada: Several changes occurred to the Canada Pension Plan Disability Insurance program starting in the mid- 1980s, all in line with a trend toward making eligibility and benefits more generous. These reforms included an increase in the benefit level, a relaxation of the number of years one must have worked in order to qualify, and an enhancement of retroactivity in claiming benefits. An important reform in 1989 emphasized socioeconomic factors such as the unemployment rate and job skills in the determination of eligibility for those aged fifty- five to sixty- four. On top of this, efforts to publicize the availability of benefits were made. The trend toward increasing generosity of benefits and eligibility came to an end in 1995, following concerns about the now- high levels of participation and the associated cost. The provision for socioeconomic considerations was removed, existing claimants were systematically reexamined, and a new emphasis on self- sufficiency and returning to work was instituted. This reform had a very substantial effect on the DI participation rate. The change for men aged sixty to sixty- four is shown in figure I.33. The rate was reduced from about 14 percent in 1985 to less than 8 percent by 2007, a reduction of about 40 percent. Germany: The 1984 reform in Germany (effective in 1985) restricted disability pension eligibility and affected women in particular. To be eligible for disability benefits, workers had to have a minimum of three contribution years in the last five years. In effect, this ruled out claims for many women who did not work on a regular basis. Figure I.34 shows the number of new DI claimants by year. The reform had a dramatic effect on the number of women retiring on a DI pension. The number was reduced from 173,000 in 1984 to 67,000 in 1986 (a reduction of over 60 percent), with further reduction to about 60,000 in the next few years. The effect on men was much smaller, from about 163,000 in 1984 to 129,000 in 1986.

Introduction and Summary

31

Fig. I.33 Canada: Effect of 1995 reform on the DI participation of men aged sixty to sixty-four

Fig. I.34

Germany: Effect of 1984 reform on the number of new claimants

The differential effect of the 1984 reform on men and women can also be seen by comparing the pathways to retirement for men and women. Figure I.35 shows pathways to retirement between 1960 and 2005 for women and men, respectively. The figure shows the large effect of the 1984 reform on the proportion of women retiring by way of DI and the relatively small effect on the DI retirement path for men. These figures show differential effects of other reforms on men and women. For example, the 1972 reforms

32

Kevin Milligan and David A. Wise

A

B

Fig. I.35

Pathways to retirement in Germany: A, For women; B, For men

had a much greater effect on men than on women. In general, in addition to differential effects of reform, the pathways to retirement have been quite different for men and women. Belgium: In 1997 the Belgian government was forced by a European Court of Justice ruling to harmonize the normal retirement age and the full career requirements for men and women. This decision resulted in a large increase in the DI participation of women aged sixty to sixty- four. The decision was to align the female full career definition to the definition for men. Since

Introduction and Summary

33

1997, the normal retirement age for women has been raised gradually from sixty to sixty- one in 1997, to sixty- two in 2000, sixty- three in 2003, sixtyfour in 2006, and to sixty- five in 2009. These increases were accompanied by corresponding increases in full career requirements for women. Prior to 1997, with a retirement age of sixty, women over sixty were not eligible for DI benefits. Figure I.36 shows the steep increase in DI participation thereafter. The increase can be explained by the increase in the retirement age and the resulting increase in the LFP of women aged sixty to sixty- four. This example also highlights the importance of considering disability insurance programs as a substitute for standard retirement programs. Sweden: Until the 1990s, there was a trend in Sweden toward more liberal eligibility rules. That trend was reversed with a series of reforms in the 1990s. First, eligibility for DI for labor market reasons was abolished in 1991. Then, in 1991 and 1992, access to sickness absence was restricted through legislative initiatives. Next, a focus on rehabilitation was introduced for employers and local governments. Finally, in 1997, special eligibility rules for older workers were abolished. The impact of these changes can be seen in figure I.37, which shows the reduction in DI participation following these reforms for men and women, respectively. Overall there is a marked decrease in DI participation over the next several years following the reforms. For persons aged sixty to sixty- four, for example, the decline between 1992 and 1998 was 74 percent for men and 66 percent for women. France: Two policy changes in France in the late 1990s and early 2000s demonstrate the importance of reforms in determining sickness benefit par-

Fig. I.36 Belgium: The effect of the 1997 reform on the DI participation of women aged sixty to sixty-four

34

Kevin Milligan and David A. Wise

A

B

Fig. I.37 Sweden: Effect of the 1991 to 1992 reforms on the DI participation by age group: A, By men; B, By women

ticipation. First, some early retirement schemes were being phased out. For example, the number of recipients on an early retirement scheme, at ages fifty- five to fifty- nine, dropped from 230,000 in 1997 to 130,000 in 2005. Unemployment (without search requirement after age fifty- six or fiftyseven) largely acted as a substitute—the number of recipients of unemployment benefits increased from 270,000 to 400,000 over the same period. Older workers may also have sought benefits through long- term sickness leaves. In addition, stricter controls, but without legislated reform, affected sickness benefits in 2004 to 2005. There was an increase in the intensity of reviews for sickness benefit claims following concern over the rapid rise

Introduction and Summary

35

in uptake. In principle, if sickness benefits were being used as a route to retirement by older workers, the stricter sickness benefit regime would have a heavier impact on older workers than on younger workers, who obtain sickness benefits in case of health problems. Figure I.38, for men and women respectively, supports this explanation. The figure shows the evolution of sickness benefits per capita (in 1998 euros) for five age groups—that is, the total amount of sickness benefits received by a given age- group over its size in the full population, in order to correct for

Fig. I.38

France: Effect of 2004 stricter sickness controls: A, Men; B, Women

36

Kevin Milligan and David A. Wise

demographic shifts. Consider panel A, for men. All age- groups display an upward trend over the 1997 to 2008 period, with years 2000 to 2003 standing out above that trend. However, the magnitude of the 2000 to 2003 bump is much larger for older workers in the age- groups fifty- five to fifty- nine and sixty to sixty- four. In particular, the decrease in benefits that occurs in the overall population after 2003 seems to be fully driven by the older age groups while the other groups only display a slowdown in growth after 2004 to 2005. The patterns for men and women are quite similar although the bump is somewhat less apparent for women. Thus these figures show not only the effect of DI (sickness) reforms on DI participation, but also highlight the substitution effects between retirement programs. Denmark: The 1984 reforms of a number of programs in Denmark had a large effect on the number of women on DI, but almost no effect on men. The primary disability program, called the Social Disability Program (SDP), was revamped with three tiers. The lowest of these took social criteria into consideration, meaning that the age and employment prospects of the applicant were taken into account in determining DI eligibility. These reforms led to a large and sudden shift of women from other programs and into the SDP disability insurance program. In short, with the introduction of the 1984 reforms some women became ineligible for DI, but a much larger number became newly eligible, especially women aged sixty and older. Figure I.39 shows the jump in the participation rate of women aged sixty to sixty- four between 1973 and 1995. The participation rate increased from about 18 percent in 1983 to almost 35 percent in 1985. Figure I.40 shows essentially the same information with respect to DI participation, but in the form of a large shift in pathways to retirement. In 1983 about 34 percent of women entered retirement by way of disability insurance. In 1985 this percent jumped to about 69 percent.4 The figure also highlights the substantial changes in the pathways to retirement over the past forty years. The Danish example highlights both the importance of substitution between programs for older workers and the nonhealth reasons for some of the shifts in disability insurance uptake. The Netherlands: Since the mid- 1980s there have been a series of DI reforms aimed at reducing the number of DI recipients. The analysis for the Netherlands makes an important point: the success of these reforms depended in part on the (relative) attractiveness of alternative routes to retirement such as early retirement and unemployment insurance options. These programs also became less generous. Taking this into account, an 4. The other programs are: an Unemployment Program, the PEW, the sixty- five to sixty- six Old-Age Pension (OAP), the Widows Pension and the Early Old-Age Pension (EOAP), the Transitional Benefit Program (TBP) for long- term unemployed, and the Flex Job (FJ) program.

Introduction and Summary

37

Fig. I.39 Denmark: Effect of the 1984 reform on DI participation of women aged sixty to sixty-four

Fig. I.40

Denmark: Pathways to retirement for women aged fifty-five to sixty-four

empirical evaluation of several reforms in DI, early retirement, and unemployment programs shows that, on balance, the reforms were associated with reductions in the inflow into DI and early retirement programs and an increase in inflow into unemployment insurance programs. Such interaction among programs is important for other countries as well.

38

I.6

Kevin Milligan and David A. Wise

Summary and Conclusion

The historical analysis in this volume is intended to set the stage for and inform more formal analysis of disability insurance programs. The key question that future analysis will address is this: Given health status, to what extent are the differences in LFP across countries determined by the provisions of disability insurance programs? To answer this question we need to develop measures of health status that are comparable across countries, we need to understand to what extent DI insurance reforms are prompted by health status in a country, and we need to understand whether DI provisions are prompted by the employment circumstances of older people in each country. Thus we have undertaken this historical analysis to help inform our future analysis. We began by considering changes in mortality over time and in particular the relationship between mortality and labor force participation. In this context we think of mortality as one measure of health status that is comparable across countries and comparable over time within countries. We gave particular attention to the relationship between mortality and labor force participation. We find that even though LFP rates are quite similar at low mortality rates, they diverge substantially with increasing mortality rates. The divergence is very similar to the increasing divergence in LFP across countries at older ages. Having in mind how mortality may be related to other measures of health, we then consider the relationship between the trends in mortality and selfassessed health within a country. In general, we find a rather strong relationship between the two trends, although with very different reported levels of self- assessed health across countries. We then consider how mortality is related to other indicators of health status. Here the available data varies greatly from country to country and the relationship of the different measures to mortality trends also varies from country to country. This is even true of objective measures such as heart disease or respiratory disease. Then we consider how trends in DI participation are related to changes in health, in particular measured by the change in mortality. We find little relationship over time within a country. In this respect we conclude that DI insurance reforms are largely a train on their own track and not endogenously determined with respect to health. Finally we consider natural experiments in which the disability insurance reforms were not prompted by changes in health status or by changes in the employment circumstances of older workers. We find that these exogenous reforms can have a very large effect on the labor force participation of older workers.

Introduction and Summary

39

References Gruber, Jonathan, and David A. Wise, eds. 1999. Social Security Programs and Retirement around the World. Chicago: University of Chicago Press. ———. 2004. Social Security Programs and Retirement around the World: MicroEstimation. Chicago: University of Chicago Press. ———. 2007. Social Security Programs and Retirement around the World: Fiscal Implications. Chicago: University of Chicago Press. ———. 2010. Social Security Programs and Retirement around the World: The Relationship to Youth Employment. Chicago: University of Chicago Press.

1 Disability, Health, and Retirement in the United Kingdom James Banks, Richard Blundell, Antoine Bozio, and Carl Emmerson

1.1

Introduction

Two potentially contradictory trends have been identified as populations around the world have been aging in recent years. On the one hand, improvement in health has led to nonabated increases in life expectancies. On the other, health conditions and disability have become seen, more than ever, as the main obstacle to longer working lives. This apparent paradox is at the core of policies aiming to encourage longer working life as various institutional settings (state pensions, disability benefits, and unemployment insurance) interact with changes in health status and labor market conditions. Previous research has highlighted the impact of financial incentives of pension systems across a number of developed economies (Gruber and Wise 1999, 2004) but much less is known on the role that other pathways to retirement and changes in health conditions have played. James Banks is professor of economics at the University of Manchester and deputy research director of the Institute for Fiscal Studies. Richard Blundell is the Ricardo Professor of Political Economy at University College London and research director of the Institute for Fiscal Studies. Antoine Bozio is a research fellow at the Institute for Fiscal Studies and director of the Institut des politiques publiques (IPP) at the Paris School of Economics. Carl Emmerson is deputy director of the Institute for Fiscal Studies. This chapter forms part of the International Social Security Project at the NBER. The authors are grateful to the other participants of that project for useful comments and advice. We are also grateful to the ESRC- funded Centre for the Microeconomic Analysis of Public Policy at IFS (grant number RES- 544-28-5001) for funding this project. Material from the Family Expenditure Survey and the Labour Force Survey was made available by the UK Data Archive and data on mortality rates from the Human Mortality Database (available at www .mortality.org) and from the Government Actuary’s Department (GAD). Any errors are the responsibility of the authors alone. For acknowledgments, sources of research support, and disclosure of the authors’ material financial relationships, if any, please see http://www.nber .org/chapters/c12382.ack.

41

42

James Banks, Richard Blundell, Antoine Bozio, and Carl Emmerson

The United Kingdom is a fine example of these interactions. With stricter unemployment benefits and relatively few early retirement schemes (Banks et al. 2010), disability benefits have over time come to represent an important pathway to retirement. At the same time, life expectancy has been rising continuously while measures of self- reported health or disability do not seem to exhibit similar improvements. As a result, disability benefits have come to the top of the policy agenda with reforms following each other at a very rapid pace since the mid- 1990s: a major reform in 1995 was followed by important changes in 2000, 2001, 2003, 2006, 2008, and most recently 2010. When one considers the degree of policy interest for this issue, one could be surprised at the limited literature on the subject in the United Kingdom. The main reason behind this is not the lack of interest from economists, but more the lack of suitable data that combine information on the labor market situation and comprehensive measures of health and disability. Most early research had to rely on self- reported measures of incapacity for work and benefit receipts. The obvious problem is that self- reported measures of disability could be affected by benefit receipt and therefore offers limited explanatory power (Myers 1982; Bound 1991). The main result from this early literature (Doherty 1979; Fenn 1981; Piachaud 1986; Disney and Webb 1991) was that both disability benefits and self- reported disability were linked to the labor market conditions: increased unemployment seemed to lead to an increased number of claimants of disability benefits and increased self- reported disability. More recent research (Benítez-Silva, Disney, and Jimenez-Martin 2010) has confirmed this relationship between the business cycle and the incidence of self- reported disability and provided more insights to the mechanisms involved, showing that unemployment had a large impact on the outflow rate out of disability benefits. Increasingly, researchers have tried to go beyond measures of self- reported health to capture the impact of more objective measures of health shocks. Disney, Emmerson, and Wakefield (2006) have, for instance, used panel data to construct instruments for self- reported health, showing that health shocks were important predictors of movements in and out of paid work among those approaching the state pension age in the United Kingdom. In an alternative approach, anchoring vignettes have been used to try and control for group or country- specific reporting effects on subjective health and work disability, with particular application to international comparisons (see Kapteyn, Smith, and van Soest [2007] or Banks et al. [2008], for example). This chapter examines changes in health and disability- related transfers in the United Kingdom over the last thirty years, and describes how they are related to changes in labor force participation. The objective is to present a comprehensive description of the reforms to the institutional setting, along with available time series coming from administrative data on benefit receipt, cross- section or panel data on self- reported health, and their interactions with labor force status. By providing systematic evidence on institutions and

Disability, Health, and Retirement in the United Kingdom

43

data, we hope to help future research by providing a fuller picture of the trends over this period. We also present evidence on the impact of two large reforms to disability benefits that help shed light on the long- term changes in disability prevalence in the United Kingdom. Section 1.2 presents the evolution of transfers targeted toward people with disabilities in the United Kingdom, focusing on recent reforms and the distinctive features of these benefits compared to their equivalent in other countries. Section 1.3 shows the evidence available on the different pathways to retirement in the United Kingdom, while section 1.4 presents evidence on various health measures, including mortality and self- reported health, and contrasts these evidences with labor market outcomes. Section 1.5 presents evidence on two major reforms of the UK disability benefit system, the 1995 reform and the more recent “Pathways- to-Work” program. Section 1.6 concludes. 1.2

History of Transfers Targeted Toward People with Disability in the United Kingdom

Disability is a difficult characteristic to define. The traditional approach in the literature has rested on the pioneering work from Nagi (1965, 1991) who identified three components of disability: a pathology, an impairment, and an inability to perform expected activities.1 This approach leads to the view of disability as a permanent condition, completely separated from sickness, which is defined as a temporary incapacity. This distinction between permanent and temporary conditions has not been instrumental in the design of the UK benefit system. Historically, as this section will describe in more detail, sick and disabled individuals were all covered by sickness benefits, the only distinction coming from the duration of claims. As a result, the focus has been more on long- term sickness than on disability. In order to facilitate the comparison with other countries, we present the benefits available both to the short- term sick and to the long- term sick or disabled. Transfers targeted toward the long- term sick or disabled in the United Kingdom are a complex set of benefits that have evolved over time and have been relabeled multiple times. To clarify this institutional setting with a jungle of acronyms, it is helpful to distinguish four types of disability benefits: work- related injury benefits, disability insurance, non- contributory benefits, and means- tested benefits (Creedy and Disney 1985; Burchardt 1999). 1.2.1

Work-Related Injury Benefits

Compensatory benefits, for injuries at work or during wars, were historically the first ones to be implemented in the United Kingdom with the 1. See Bound and Burkhauser (1999) for a review on these definitions and the implications for the measurement of the population with disabilities.

44

James Banks, Richard Blundell, Antoine Bozio, and Carl Emmerson

enactment in 1897 of the Workmen’s Compensation Act, which established the legal liability of employers to compensate employees for loss of earnings capacity as a result of an accident or disease linked to employment (Walker 1981; Walker and Walker 1991). During World War I a state scheme, the War Disablement Pension, was introduced to offer compensation to veterans of Her Majesty’s (H.M.) Armed Forces. It was followed in 1948 by the Industrial Injuries Disablement Benefit (IIDB), set up by the National Insurance Industrial Injury Act 1946.2 Both schemes still exist today and have only been marginally changed over time.3 They offer more generous benefits than other disability benefits, are not means- tested, and can be cumulated with other benefits. 1.2.2

Disability and Sickness Insurance

The second type of disability benefits is earnings replacement benefits. The UK schemes share some characteristics of other countries’ sickness and disability insurance but also have two defining features inherited from their origin. First, they are not really insurance schemes, as generally understood. The welfare system put in place in the United Kingdom in 1948 largely followed the design of the Beveridge report (Beveridge 1942). It relied on an insurance principle, whereby eligibility to benefits was determined by contribution requirements, but benefits were not earnings related, unlike the US Social Security Disability Insurance (SSDI) or examples in Continental Europe. As a result, the system has largely been targeted at low income individuals for whom flat- rate benefits represented a large replacement rate.4 Second, the UK system has not formally recognized permanent disability conditions. The benefit set up in 1948 was called Sickness Benefit and offered a benefit with unlimited duration.5 Hence the coverage for disability was not distinguished from short- term sickness, and only duration of claim could distinguish the long- term sick from the short- term sick. Table 1.1 presents the evolution of these schemes from 1948 to 2010 ac2. The rate of the IIDB in 2009 to 2010 was £143.60 per week (or $12,000 annually) for an extent of disablement of 100 percent and those over eighteen. The benefit is reduced proportionally with the disablement. 3. The IIDB was originally split into Industrial Injury Benefit (IIB) for the first twenty- six weeks of sickness and Industrial Disablement Benefit for longer durations. In April 1983, IIB was abolished and replaced for the first eight weeks by employers’ Statutory Sick Pay (SSP) and the sickness benefit for durations between nine and twenty- five weeks (see section 1.2.2 for more details on SSP). 4. There is a short period between 1966 and 1980 when earnings- related sickness benefits were introduced, but this social insurance experiment was both limited and short- lived. 5. The system introduced after World War II is also largely the heir of the general sickness insurance introduced by the National Insurance Act 1911. It provided sickness benefits payable for twenty- six weeks along with a disability benefit and some health care benefit. All these benefits were distributed through approved Friendly Societies, but the scheme largely paved the way for further state interventions (see chapter 2 of Creedy and Disney [1985], and Gilbert [1965]).

Disability, Health, and Retirement in the United Kingdom Table 1.1

45

Structure of benefits in the United Kingdom by duration of incapacity (1948–2010) Duration of incapacity to work 1– 8 weeks

9– 28 weeks

29– 52 weeks

More than 1 year

Sickness Benefit Invalidity Benefit (IVB) IVB

Sickness Benefit Invalidity Benefit (IVB) IVB

SSP/Sickness Benefit (IB) short- term lower rate

IVB

IVB

IB short- term higher rate

IB long- term rate

SSP/ESA

Employment Support Allowance (ESA)

(ESA)

1948– 1971 1971– 1982

Sickness Benefit Sickness Benefit

Sickness Benefit Sickness Benefit

1983– 1985

Statutory Sick Pay (SSP) SSP/Sickness Benefit

Sickness Benefit

1986– 1995 1995– 2008

2008–

Table 1.2

SSP/Incapacity Benefit (IB) shortterm lower rate SSP/ESA

Reforms to the UK disability insurance system, 1948 to present day

1948 1966 1971 1972 reform 1980 1983/1986 1995 reform

2001 reform Pathways- to-work expansion 2003– 2008 2008 reform 2010 reform

Introduction of Sickness Benefit. Flat- rate benefit, no distinction by duration of claims. Introduction of earnings- related Sickness Benefit. Introduction of Invalidity Benefit (IVB). Higher rate for duration above six months. Introduction of Invalidity Allowances. Supplements for becoming disabled at younger age. Abolition of earnings- related Sickness Benefit. Introduction of Statutory Sick Pay. Incapacity Benefit (IB) replaces IVB. New claimants receive less generous Incapacity Benefit, which is taxable (unlike IVB). Own occupation test replaced by any occupation test. Regional medical test instead of personal doctor. No longer paid to people over state pension age. Increased contribution requirement to qualify for IB. Introduction of means testing with regard to pension income. Piloting of a package of reforms consisting in increased conditionality, increased support, and increased financial incentives to return to work. Employment support allowance (ESA) replaces IB for new claimants. ESA is applied to all existing IB claimants.

cording to duration of incapacity, while table 1.2 summarizes the changes to the generosity of these sickness and disability schemes. In 1971 Invalidity Benefit (IVB) was split from the Sickness Benefit but still followed the structure inherited from the previous scheme, whereby entry to IVB would be offered to those who had been on sickness benefits for longer than twentyeight weeks. The IVB offered a higher level of benefit than the Sickness Benefit but without imposing another health test when entering IVB. The

46

James Banks, Richard Blundell, Antoine Bozio, and Carl Emmerson

screening process at the time relied on a medical assessment by a personal doctor of the ability to conduct “suitable work.” In 1983, a major reform that was introduced to transfer administration of sick pay claims from Sickness Pay to employers for the first eight weeks of sickness, was increased to twenty- eight weeks in 1986. Employers were mandated to pay Statutory Sick Pay (SSP), payments that would be reimbursed by the government through lower National Insurance contributions.6 For those who would not qualify for SSP, the Sickness Benefit was still available. The number of claimants increased slowly until the mid- 1980s for the older working- age individuals, when a sharp increase of IVB recipients was registered for all age- groups. One can see in figure 1.1 and figure 1.2 the number of IVB recipients as a share of the fifty- five to fifty- nine, sixty to sixtyfour, and sixty- five to sixty- nine age- groups for men and women. Between 1985 and 1996, the share of the fifty- five to fifty- nine- year- old men on IVB almost doubled, from 10.9 percent to 20.0 percent. In 1995 a reform was introduced that replaced the IVB and the Sickness Benefit schemes with the Incapacity Benefit (IB). This maintained the “own occupation test” to qualify for the first twenty- eight weeks of incapacity, but replaced the “suitable work test” of IVB with an “all work test” to qualify for the higher rate IB. This new medical screening was also removed from personal doctors and was instead administered by medical staff at the regional level and commissioned by the scheme’s administration. The growth of the IB roll was stopped, even slightly reversed, but the stock remained high, especially for younger individuals. In addition to these changes, IB was no longer paid to new claimants above the state pension age (sixty- five for men and sixty for women, at the time). Previously, individuals typically preferred to stay on IVB than to receive the basic state pension, as the latter is taxable whereas the former was not. The new IB benefit excludes those above the state pension age (at the time sixty for women and sixty- five for men) and is treated as taxable income. This is why the number of claimants of IB aged above the state pension age drops markedly after the 1995 reform in figures 1.1 and 1.2. The 1999 Welfare Reform and Pensions Act introduced further changes, with a tightening of the health test from April 2000 onward and a reduction in the generosity of IB from April 2001. The new health test is called Personal Capability Assessment, which is designed to assess capacity for paid work instead of checking incapacity for work and is therefore supposed to foster a return to work. The reform also increased the eligibility requirement for IB from having paid contributions in any year before the start of incapacity to having paid sufficient contributions in one of the last three years. 6. Control of SSP was made by self- certification of sickness from the part of employees, which has raised concerns when expenditures on SSP turned out to increase more rapidly than the sickness benefit (Creedy and Disney [1985], page 127).

Fig. 1.1

The IVB/IB recipients as a share of population (males), by age-group

Note: The IV/IB claimants’ data are from Anyadike-Danes and McVicar (2007), and the working- age population is from the Family Expenditure Survey.

Fig. 1.2

The IVB/IB recipients as a share of population (females), by age-group

Note: The IV/IB claimants’ data are from Anyadike-Danes and McVicar (2007), and the working- age population is from the Family Expenditure Survey.

48

James Banks, Richard Blundell, Antoine Bozio, and Carl Emmerson

Finally, it introduced means testing of IB with regard to individual private pension income at a rate of 50 percent above £85 a week. In 2003 the New Labour government decided to pilot an ambitious, and expensive, program to incentivize IB claimants to return to work called Pathways-to-Work. The program included increased conditionality with mandatory work- focused interviews, increased financial incentives to return to work, and increased support with the provision of voluntary schemes designed to help disabled individuals to return to work. The scheme was evaluated in pilot areas and then expanded to the rest of the country (Adam, Bozio, and Emmerson 2012). In 2008 the government announced a new scheme to replace IB, the Employment Support Allowance (ESA) for new claimants. This new scheme incorporated a stricter eligibility health test along with a redesign of the benefit rates. In the first thirteen weeks of claim, the claimant is subjected to a Work Capacity Assessment, which determines whether the individual is entitled to ESA. Among those found eligible for ESA, the Work Capacity Assessment distinguishes between those who have “limited capacity to work and are unable to follow work- related activities” and the remainder who have “limited capacity to work but are able to follow work related activities.” For the last group claimants are mandated to attend the Pathways- to-Work program. The ESA will be progressively applied to all existing IB claimants; that is, existing claimants are going to be retested for the stricter eligibility between October 2010 and 2014. 1.2.3

Non-contributory Benefits

Whereas the previous disability benefits are only available to those who have a sufficient National Insurance contribution record, a set of benefits were created in the 1970s for individuals of working age, with congenital disabilities, and who did not qualify for the contributory scheme. In 1975 the Non-Contributory Invalidity Pension (NCIP) was introduced, offering a benefit of 60 percent of IVB to men or single women. In 1977 the scheme was extended to married women who were “incapable of performing normal household duties” under the name of Housewife Non-Contributory Invalidity Pension (HNCIP), but at a lower rate than the NCIP. Both NCIP and HNCIP were replaced in 1984 by the Severe Disablement Allowance (SDA), which stopped the distinction that it was deemed discriminatory against women. It was subsequently abolished in 2001 for new claimants. In the 1970s a number of schemes were also designed to offer benefits to compensate the extra cost endured by disabled individuals, either in the form of carers or the extra cost of mobility. In 1971 the Attendance Allowance (AA) was created for those who required personal assistance and in 1976 a Mobility Allowance (MA) was introduced for those who had difficulty moving around. Also in 1976 an Invalid Care Allowance (ICA) was introduced for those who could not work because they had to stay at home to care for

Disability, Health, and Retirement in the United Kingdom

49

a disabled relative.7 In April 1992 the Disability Living Allowance (DLA) replaced MA and AA for those who had become disabled before the age of sixty- five, while AA was kept for those aged over sixty- five. In terms of total expenditure, DLA represents the biggest transfer targeted toward people with disability in the United Kingdom. In 2006 to 2007 it represented £9 billion of expenditure, approximately 0.7 percent of national income. If one adds the £4 billion of AA and £1.2 billion of CA, the non- contributory disability benefits represent more than 1.0 percent of national income. In the June 2010 budget, the government announced plans to cut DLA spending significantly by reassessing the health of existing claimants. 1.2.4

Means-Tested Benefits

A number of means- tested benefits targeting poor households have provisions that include premiums for disability. Income Support (IS) on grounds of disability, for instance, offers a premium for low- income households containing at least one disabled individual. Another example, the Working Tax Credit (WTC), the United Kingdom’s equivalent of the US Earned Income Tax credit (EITC), also has a supplement for disabled workers, and has a less onerous hours rule than that applied to nondisabled childless adults, with a further premium for the severely disabled. Housing Benefit (HB) is another means- tested benefit with additional income for those with disability and increased premium for those with severe disability. 1.3

Pathways into Retirement and Program Reforms

Given the complexity of pathways into retirement, it is important to put the changes to disability schemes in the wider context of other reforms to state pension schemes and unemployment schemes. Presenting data on pathways into retirement requires long panel data sets where each individual can be followed from work into retirement status. The United Kingdom does not have comprehensive administrative data such as the ones available for Germany (see Borsch-Supan and Jurges, chapter 7, this volume), but we can shed light on these transitions using three approaches: cross- sections from Family Expenditure Survey (FES) and Labour Force Survey (LFS), oneyear economic transitions from LFS, and the longer panel from the British Household Panel Survey (BHPS). 1.3.1

Cross-Section Evidence on Economic Activity

Two representative surveys provide good information on participation in the labor market in the United Kingdom. The FES goes back to 1968, and from 1975 onward the LFS offers large samples of British households with 7. In 2003 ICA was renamed Carer’s Allowance (CA).

50

James Banks, Richard Blundell, Antoine Bozio, and Carl Emmerson

a full description of their labor market status. The employment rate of older males by three age categories (fifty- five to fifty- nine, sixty to sixty- four, and sixty- five to sixty- nine) over a forty- year period, from 1968 to 2008, is shown in figure 1.3. We also add in the figure the main reforms to disability benefits in the United Kingdom over that period; that is, the introduction of IVB in 1971 and the introduction of IB in 1995. No obvious relationship stands out from these time series. The introduction of a more generous IVB in 1971 does not seem to have led, at least immediately, to a change in the employment rate of older workers, while the more restrictive reform of 1995 is also hardly visible. The progressive introduction of the Pathways- to-Work program over the 2003 to 2008 period is associated with an increase in employment for the older workers, but given that the program only affected a small share of the country until 2006, it is difficult to ascribe this increase to this reform (we return to this issue in section 1.5). Another way to look at the change in labor market status over the long term is to look at reasons given by survey respondents for not being in work. We present in figure 1.4 cross- sections of fifty- five to sixty- four- year- old men by self- reported economic activity. We cannot split those who report a health problem between the short- term sick and long- term sick, but we still capture the changes in nonemployment between those who actively look for paid work (the official unemployed), those who report being inactive because

Fig. 1.3

Employment rate and IVB/IB reform (males)

Sources: 1968 to 1983 Family Expenditure Survey; 1983 onward, Labour Force Survey.

Disability, Health, and Retirement in the United Kingdom

Fig. 1.4

51

Economic activity of fifty-five to sixty-four-year-old men (1968–2009)

Source: 1968 to 2009 Family Expenditure Survey.

they are retired, and those who report being inactive because of sickness. Given the way the questions in the survey are structured, an individual who is not working because of a temporary illness but has kept his job will be classified as employed. Therefore, those who report being sick are both not employed and not looking for work. Two facts stand out from this figure. First, the big drop in older male employment in the late 1970s and early 1980s was associated with a large increase in the unemployed and the retired. The share of those reporting being sick did not increase immediately. However, starting in the mid- 1980s, the share of fifty- five to sixty- four- year- olds reporting being inactive because of sickness increased markedly, in line with the increase in disability benefits recipients observed in figure 1.1. Over the last ten years the increase in the employment rate of this group has largely been at the expense of the unemployed, and only marginally at the expense of those reporting health problems. As a general remark, the share of those inactive because of sickness is always much larger than those looking for work, even when the official unemployment rate reached its highest level in the 1980s. 1.3.2

One-Year Transitions from LFS

One advantage of the LFS since 1992 is that survey respondents are asked about their economic position quarterly in five successive waves. This pro-

52

James Banks, Richard Blundell, Antoine Bozio, and Carl Emmerson

Fig. 1.5 One-year transition rates to inactivity from employment (sixty to sixtyfour-year-old men) Source: Quarterly Labour Force Survey 1992– 2006.

vides us with a one- year panel data set from 1993 onward, allowing us to present evidence on transitions from employment into inactivity. Figure 1.5 presents the evolution of these short- term transition rates for sixty to sixtyfour- year- old men. The most striking fact over the period, especially since the late 1990s, is the reduction in the transition rate from employment into retirement. This coincides with the significant increase in the employment rate of this group over the period. Transitions to unemployment and disability have declined over the early 1990s and stabilized at a low level since. There is hardly any evidence from these statistics that the 1995 reform has had much impact on the transitions through disability and the dominant factor over the period remains the change in retirement behavior of this age group which, over this period, is increasingly occurring at an older age. Figure 1.6 presents similar evidence by looking reversely to the previous activity of newly retired individuals, that is, individuals who declare that they are retired in one year but were not in the previous year. From the mid- 1990s to the days just prior to the financial crisis, direct transition from employment to retirement increased markedly: whereas in 1994 only 54 percent of newly retired men were coming directly from employment, this share reached 67 percent in 2008. This has been matched by a similar decrease of

Disability, Health, and Retirement in the United Kingdom

Fig. 1.6

53

Previous economic activity of newly retired men

Source: Quarterly Labour Force Survey 1993– 2008.

newly retired men coming from unemployment, whose proportions were halved from 20 percent to 10 percent. On the other hand, there is only limited evidence of reductions in those coming from long- term sickness or disability. From 1994 to 2001 the proportion increased, from 25 percent to 35 percent, while a decline is evident in the more recent years, down to 23 percent in 2008. 1.3.3

Evidence from BHPS

The short- term transition rates from the LFS provide a good but limited description of the pathways to retirement that individuals might experience. It is possible to imagine that transitions to unemployment cascade into disability before retirement and that short- term transitions do not capture these effects. In order to shed light on these long- term transitions, we used a long panel data set, the British Household Panel Survey, which surveyed 10,000 individuals every year since 1991 and up to 2007. Although we have access to seventeen waves of BHPS, there are only a few cohorts that we can follow from age fifty through retirement. We have selected the cohort born between 1938 and 1942 who were aged forty- nine and fifty- three in 1991 and aged sixty- five to sixty- nine in 2007. In figure 1.7 we present the evolution of self- reported economic activity for a sample of men from this cohort

54

James Banks, Richard Blundell, Antoine Bozio, and Carl Emmerson

Fig. 1.7

Subsequent activity of men born between 1938 and 1942 in work in 1991

Source: British Household Panel Survey 1991– 2007.

who were in paid work in 1991. Between age fifty and fifty- five, inactivity is largely dominated by unemployment, whereas disability becomes a more substantial aspect from age fifty- five onward up to much older ages. Nonetheless the decrease in employment over the fifty to sixty- nine age- group is still largely dominated by the increase in the other status; that is, retirement. In figure 1.8 we present similar statistics to figure 1.6, but using the long panel of the BHPS as opposed to the short panel of LFS data. Those who were retired in 2007 largely transited directly from employment: 64.5 percent of retired men aged sixty- five to sixty- nine in 2007 were in employment before retiring compared to 63.3 percent for women. This still leaves a significant share that transit through unemployment and disability: 25.0 percent of men aged sixty- five to sixty- nine came from disability compared to 13.8 percent for women. Disability is an ever more important transition for women as the increase in labor force participation of women has reduced the other form of inactivity while increasing eligibility to disability benefits. Figure 1.9 takes full advantage of the long panel from the BHPS by presenting evidence on transitions from employment into retirement and distinguishing the different pathways. The large majority of men and women aged sixty- five to sixty- nine are either still in work or have transited directly from employment to retirement (75.6 percent of men and 78.4 percent of women). This is not to say that spells of unemployment or disability are rare, as a significant proportion of men transit through unemployment (11.3 percent) and disability (8.9 percent). It is, however, much less common to experience

Fig. 1.8

Last activity of those retired in 2007, cohort born 1938–1942

Source: British Household Panel Survey 1991– 2007. Note: Those who are inactive throughout the panel are included in “other.”

Fig. 1.9

Pathways from work into retirement, cohort born 1938–1942

Source: British Household Panel Survey 1991– 2007. Note: The sample includes all those aged forty- nine to fifty- three and in work at the start of the panel (in 1991) and retired at the end (aged sixty- five to sixty- nine in 2007). Less than 1 percent of the forty- nine to fifty- three- year- olds working in 1991 ends up unemployed or disabled in the last wave of the panel.

56

James Banks, Richard Blundell, Antoine Bozio, and Carl Emmerson

multiple transitions from unemployment to disability before retiring as these two options seem to be alternative pathways. 1.4

Evidence on Long-Term Trends in Health and Labor Participation

This section aims to provide evidence on long- term trends in health using measures of mortality rates at different ages and self- reported measures of disability. We then attempt to relate these changes to changes in the labor force participation. 1.4.1

Mortality Data

There are two advantages in using mortality data. First, mortality is a well- defined concept and it is therefore easy to present comparable information across countries. Second, mortality rates are available over long periods and can be matched with historical data on labor force participation. However, mortality data also have very obvious limitations for our purpose: they are not individual data, and do not allow assessing individual- specific health shocks to labor force participation. And perhaps even more importantly, morbidity is a very different issue from disability or incapacity to work, which is more likely to matter when explaining trends in labor force participation. England and Wales life tables are available from 1841 onward by age and sex and by period and cohort.8 We use in this section period data for ease of comparison with other countries. Figures 1.10 and 1.11 show the evolution of period mortality rate of English and Welsh men and women at age fifty- five, sixty, and 65. Until the 1970s, there was only a minor reduction in mortality rates for men at age fifty- five and sixty and almost no improvement at age sixty- five. During that decade male mortality rates started falling rapidly, especially at older ages. The fall in mortality rates is less impressive for women, but as figure 1.11 makes clear, women have experienced much lower mortality rates than men and a much earlier decline in mortality at older ages. Figure 1.12 presents two- year mortality rates by age for both men and women comparing the period data from 1960 and 2005. Mortality rates increase steeply by age and are higher for women but the gap between men and women has got ten smaller since 1960, men having enjoyed a somewhat larger reduction in mortality than women. Whereas the 5 percent two- year mortality rate was reached at sixty- one for men in 1960, it was only attained at age seventy in 2005. For women, the age of the 5 percent two- year mortality rate increased from sixty- eight to seventy- five over the same period. Figures 1.13 and 1.14 show the age of equal mortality rate over time 8. Mortality rates calculated on a period basis do not account for future changes (typically improvements) in mortality rates, whereas those calculated on a cohort basis do allow for such changes.

Fig. 1.10

Age-specific mortality rate for English and Welsh men

Sources: England and Wales life tables, GAD.

Fig. 1.11

Age-specific mortality rate for English and Welsh women

Sources: England and Wales life tables, GAD.

Fig. 1.12

Two-year mortality rate for men and women

Sources: England and Wales life tables, GAD; computations from the authors.

Fig. 1.13

Isomorts: Age of equal period mortality rate, English and Welsh men

Sources: England and Wales life tables, GAD; computations from the authors. Note: p represents the mortality rate of the isomorts.

Disability, Health, and Retirement in the United Kingdom

Fig. 1.14

59

Isomorts: Age of equal period mortality rate, English and Welsh women

Sources: England and Wales life tables, GAD; computations from the authors. Note: p represents the mortality rate of the isomorts.

computed using one- year mortality rates (“isomorts”). This graphically illustrates the aging process as an increase in the age where individuals face the same probability of death: being a British sixty- five- year- old man in 1960—when state pension age was already sixty- five—is equivalent in terms of mortality risk to being aged seventy- four today. Or reversely, being sixtyfive today is like being fifty- five in 1960. The increase is less pronounced for women, reflecting as before the larger reduction in mortality for men, but is nonetheless impressive. For instance, being a sixty- year- old woman in 1960—the then state pension age—is today equivalent in terms of mortality risk to being seventy years old. 1.4.2

Measures of Self-Reported Disability

Although the previous section highlights the large improvement in average life expectancy, the ability to continue economic activity at an older age is more likely to be affected by health conditions that are not obviously related to morbidity. Objective measures of disability are particularly rare over long historical time series as they have only been recently added systematically to surveys on aging. As a result, analysis of such measures, over the time period we are looking at here, is not possible. Going forward, however, the fact that aging studies such as the Health and Retirement Study, the English Longitudinal Study of Ageing, and the Survey of Health, Ageing, and Retirement

60

James Banks, Richard Blundell, Antoine Bozio, and Carl Emmerson

in Europe, now routinely collect objective measures of physical functioning such as walking speed, grip strength, chair stands, balance tests and lung function, along with cognitive performance tests and huge batteries of questions on doctor- diagnosed diseases and limitations in activities of everyday living, means that an analysis based on objective measures of health and functioning should be a priority for future research. For our purpose here, however, there is useful information on self- reported health from the General Household Survey (GHS), which surveyed annually 10,000 households in the United Kingdom from 1971 to 2006. In figure 1.15 we show the proportion of men reporting limiting long- standing illness, the notion closest to the accepted definition of disability, by different agegroups. Two facts are striking. First, over this thirty- year period the share of men reporting some disability is relatively flat, despite the large improvement in health (at least as measured by the improvements in mortality rates). Second, the proportion of individuals saying that they have some limiting longstanding illness is increasing up to at least age sixty- four at every period. The presentation of time- series averages by age- group, however, tends to mask the systematic age variation in the data across later working ages. In figure 1.16 we present the proportion of men reporting limiting longstanding illness by age at ten- year intervals. The share of self- reported disability was steeply increasing by age at every period but, although it was increasing at every age between 1977 and 1997, the latest year in our data

Fig. 1.15

Proportion of men reporting limiting long-standing illness (1972–2006)

Source: General Household Survey 1972– 2006.

Disability, Health, and Retirement in the United Kingdom

Fig. 1.16

61

Proportion of men reporting limiting long-standing illness by age

Source: General Household Survey 1972– 2006.

exhibits a marked reduction for ages above fifty- one. In figure 1.17 we plot the same data with respect to specific age mortality rates for each year and the same time- patterns emerge. Similar evidence for women is presented in figures 1.18 and 1.19. The changing rate of disability for given levels of mortality probability is something that we will return to in later sections of this chapter. 1.4.3

Health Measures and Labor Force Participation

In order to summarize the evidence on labor force participation, benefit receipt, and the health measures we have discussed previously, we present in figures 1.20 and 1.21 indices of these measures alongside each other for men and women, respectively. Both figures look at the evolution from 1972 to 2006 for the age- group fifty- five to fifty- nine. Mortality is declining constantly over the period and does not seem to be related with any other trends. One interesting fact comes from the correlation between the number of claimants of disability benefits, the self- reported limiting long- standing illness, and the overall change in nonemployment that is observed among men. Nonemployment increased sharply in the early 1980s, peaking after the 1992 recession. The IB claimant count increased slowly over the period before a rapid growth in the early 1990s and a strong reversal after the 1995 reform. Although it is difficult to make precise inferences from these correlations,

Fig. 1.17 Proportion of men reporting limiting long-standing illness by mortality risk Sources: General Household Survey 1972– 2006 and GAD mortality tables. Note: Mortality risk is one- year mortality rate at a given age, from period life tables.

Fig. 1.18

Proportion of women reporting limiting long-standing illness by age

Source: General Household Survey 1972– 2006.

Fig. 1.19 Proportion of women reporting limiting long-standing illness by mortality rate Sources: General Household Survey 1972– 2006 and GAD mortality tables. Note: Mortality risk is one- year mortality rate at a given age, from period life tables.

Fig. 1.20 Health measures and labor force participation, men aged fifty-five to fifty-nine Note: Indices 1 = 1972; LLSI stands for limiting long- standing illness.

64

James Banks, Richard Blundell, Antoine Bozio, and Carl Emmerson

Fig. 1.21 Health measures and labor force participation, women aged fifty-five to fifty-nine Note: Indices 1 = 1972; LLSI stands for limiting long- standing illness.

the trend in self- reported disability is also hump- shaped around the 1995 reform, laying grounds for claims that self- reported disabilities reflect as much the impact of being in receipt of a disability benefit as some measure of perceived incapacity. Figure 1.21 presents similar evidence for women. The graph is dominated by the large increase in receipt of disability benefits, reflecting the increased eligibility of women to contributory disability benefits. Labor force participation is clearly on an increasing trend in that age- group, except during the early 1980s when the employment rate of this group declined sharply. Figures 1.22 and 1.23 contrast two ways of presenting aging and labor force participation. The first panel shows the employment rate by age for three years at a ten- year interval, while the second panel presents the same data by the mortality rate at that specific age. In figure 1.22 the employment of British men exhibit the characteristics that we have highlighted previously: a large drop in employment at the time of reaching the state pension age (age sixty- five) and a significant drop at all ages between 1978 and 1988. The recent period appears favorably with an increase in the employment rate at all ages, but especially between sixty- four and sixty- nine. The second panel, on the other hand, highlights that these changes have taken place during a period of rapid decrease in mortality. For a given mortality rate,

Disability, Health, and Retirement in the United Kingdom

Fig. 1.22

65

Employment rate by age and mortality rates (males)

Source: Labour Force Survey.

employment rates are now lower than at any other date, including the lowest point of the late 1980s. This is also the case for women, as shown by figure 1.23. Only at the youngest ages, below age fifty- five, is it possible to see the increasing participation of women counteracting the decline in employment for a given mortality rate. These figures provide a vivid illustration of the meaning of aging in our

66

James Banks, Richard Blundell, Antoine Bozio, and Carl Emmerson

Fig. 1.23

Employment rate by age and mortality rates (females)

Source: Labour Force Survey.

developed societies where age takes, in effect, different meaning, and are related to a recent analysis of Shoven (2010), who discusses using mortality risk or remaining life expectancy as better measures of age than years- sincebirth for the purpose of social security analysis and design. The limit of this approach in our context, however, is that mortality risk measures do not capture fully functioning ability and therefore err on the side of putting too much emphasis on morbidity as opposed to measures of disability. Another more powerful way of looking at the same underlying data from figures 1.15 to 1.22 is to combine them into one graph showing the evolution

Disability, Health, and Retirement in the United Kingdom

67

of employment and self- reported health over time for a given mortality rate. Figure 1.24 presents the nonemployment rate and measures of self- reported health over time for males at the age corresponding to a 1 percent mortality rate in the relevant year. As one would expect from the analysis in earlier sections of this chapter, the reference age for the comparison constantly shifts upward—in 1975 a 1 percent mortality rate was observed for men aged fifty- three, while in 2008 this age had shifted to sixty- one. Both health measures, that is, the share of men reporting long- standing illness and the share reporting a limiting long- standing illness, have increased over time but at a much slower rate than nonemployment. Taking the period 1975 to 2007 as a whole, long- standing illness increased by two- thirds, limiting long- standing illness increased by half, but nonemployment almost quadrupled, holding mortality probabilities constant. It is also worth noting that at the beginning of the period the rate of nonemployment was only half the rate of disability as measured by limiting long- standing illness. Yet, by the end of the period, nonemployment rates were higher than disability rates by 10 percentage points. These diverging trends are particularly apparent toward the beginning of the period (late 1970s and early 1980s) when nonemployment was rising very fast while self- reported health measures were

Fig. 1.24 Nonemployment rate and self-reported health measures for men with 1 percent mortality rate Sources: Labour Force Survey; General Household Survey; computations from the authors. Note: LLSI stands for limiting long- standing illness, and LSI stands for long- standing illness. Both the LLSI and LSI are three- year moving averages.

68

James Banks, Richard Blundell, Antoine Bozio, and Carl Emmerson

not, and also in the more recent years, when self- reported health measures have stopped their increase. Another possibility of using these associations between age- specific mortality rates and employment rates is to compare countries at various points in time. In figures 1.25 and 1.26 we compare the cases of France, the United Kingdom, and the United States between 1968 and 2006. In 1968, the United Kingdom and the United States have very similar employment rates for given mortality rates, whereas by 2006 the United Kingdom experienced

Fig. 1.25 Employment rate by age and mortality rates in 1968 in France, the United Kingdom, and the United States (males) Sources: Enquête Emploi; Labour Force Survey; Current Population Survey; Human Mortality Database.

Disability, Health, and Retirement in the United Kingdom

69

Fig. 1.26 Employment rate by age and mortality rates in 2006 in France, the United Kingdom, and the United States (males) Sources: Enquête Emploi; Labour Force Survey; Current Population Survey; Human Mortality Database.

much lower employment rates than the United States for mortality rates above 1 percent. While in 1968 the United Kingdom had lower employment rates than the United States after age sixty- five, the British males had at that time higher mortality rates, conditioning on age, than American ones. On the other hand, in 2006, British males saw their mortality rates drop to the level of the Americans and therefore experienced much lower employment rates than the United States at any given mortality rate. In 1968 France had relatively high employment rates at older ages, still

70

James Banks, Richard Blundell, Antoine Bozio, and Carl Emmerson

lower than the United States and the United Kingdom, but with a similar pattern. However, already in 1968, French males experienced lower mortality at a given age than American and British males. This leads to much lower employment rates in France than in the United States and the United Kingdom for a given mortality rate in 1968. By 2006 the lower mortality rates of French males is still visible, but employment rates at older age has dropped further, leading to a much bigger difference with the other two countries, especially at low mortality rates. For instance, for a 1 percent mortality rate the French males have, in 2006, an employment rate of 12 percent against 61 percent for the United Kingdom and 72 percent for the United States, and against 80 percent for French males in 1968. 1.5

Evidence from Disability Benefits Reforms

The evidence presented so far relies heavily on times series but does not show any causal impact that policy targeted on disability benefits could have on employment and retirement patterns of individuals, in particular those who report some form of incapacity to work. This section presents evidence from two reforms of the UK disability benefits: the 1995 reform, which intended to make the health test stricter, while the Pathways- to-Work program was designed to help IB claimants move off benefits and return to work. 1.5.1

The 1995 Reform

Incapacity benefit replaced IVB and sickness benefit in April 1995. The effect of the reform was to reduce the benefit’s generosity in a number of ways and to tighten the eligibility requirements.9 The reduction in generosity was realized by a number of different changes. First, the reform reduced the rate of benefit. The IB is paid at three different rates, according to the length of the period of incapacity. Short- term lower rate IB has replaced sickness benefit for people not eligible for SSP. A shortterm higher rate of ICB is payable from week twenty- nine to week fiftytwo. In spite of its name, this is less generous than IVB. Long- term IB, which is as generous as IVB, is only payable from week fifty- two. Second, the generosity of the age additions has been reduced. Previously, someone would have been eligible for an age addition to their invalidity pension if the period of incapacity began before age fifty- nine. Since 1995 they are only eligible for an age addition if the period of incapacity begins before age forty- five. In addition, the age additions are payable after week fifty- two, when long- term ICB begins, rather than after week twenty- eight. Third, IB became taxable from 1995 onward. This brings it into line with the other 9. The changes affected only new claimants after April 1995. Those people already entitled to receive invalidity benefit continued to do so under the old rules.

Disability, Health, and Retirement in the United Kingdom

71

main benefits (retirement pensions and unemployment benefits) and income support, which are subject to income tax. However, compensatory disability benefits (war disability pension and industrial injuries disablement pension) and extra costs disability benefits (disability living allowance, attendance allowance) are not subject to tax. Fourth, unlike IVB, long- term IB is not payable to anyone over the state pension age, although people who start receiving short- term IB before the state pension age can continue to do so for the full fifty- two weeks. The tightening of eligibility requirement mostly came about with the replacement of the “suitable work test” that applied to IVB recipients after twenty- eight weeks by the “all work test.” Instead of an assessment of a person’s ability to perform jobs that it was reasonable to expect them to do given their age, health, and qualifications, the all work test required an assessment of the person’s ability to do any kind of work. The all work test involved an objective assessment of the level of difficulty the person had in performing different physical and mental activities (for example, walking up and down stairs, bending and kneeling, coping with pressure). Points were awarded for the degree of difficulty they had performing each activity, with a minimum total number of points necessary to be deemed incapable of work. A second change is that the all work test is carried out by the government medical service rather than the individual’s own doctor. As with IVB, the claimant has the right to appeal for their case to be heard by a social security appeals tribunal. The first evidence one can gather on the 1995 reform is to look at the change in inflows into the IVB/IB rolls. Given that the reform has made qualifying for the benefit harder and that the generosity of the benefits has been reduced, one could expect to see changes in inflow rates into the scheme. Figure 1.27 represents the number of claimants to IVB and IB whose claim duration is less than one year. This is a relatively good proxy for the inflow rate although it is affected during the 1980s by the introduction of SSP. The latter has led to a decrease in inflows to IVB by shifting short- term sick into the employers’ sickness scheme. In 1992 the recession hit the United Kingdom acutely, and this seems to have led to a peak in inflows onto IVB. The 1995 reform is associated with a dramatic drop in inflows, which subsequently stabilized at the pre- 1992 level. Disney, Emmerson, and Wakefield (2003, 2006) examined the relationship between health and employment in the United Kingdom using panel data from the British Household Panel Survey from 1991 to 1998. They used a fixed- effects conditional logit model, instrumenting self- reported general health by using responses to questions about specific health problems (following Bound et al. 1999). Older age, reaching the state pension age, and deteriorations in health were all found to lead to increased probability of leaving work. They also tested whether the 1995 reform strengthened the relationship between health and employment by estimating how the

72

James Banks, Richard Blundell, Antoine Bozio, and Carl Emmerson

Fig. 1.27 Change in inflows into IB rolls (men, fifty to sixty-four years old, duration < 1 year) Source: The IV/IB claimants’ data are from Anyadike-Danes and McVicar (2007).

coefficient on health stock interacted with the treated group. The estimated coefficients were positive, but not statistically different from zero at conventional levels of statistical significance. As an alternative and using the same data set, we have run a probit retirement model among those in work, controlling for Disney, Emmerson, and Wakefield’s estimated health stock. We plot in figure 1.28 the year dummies before and after the reform. The coefficients for men do drop markedly postreform, with the combined 1995 to 1996 coefficients statistically different from the combined 1993 to 1994 coefficients. No statistically significant effect is found for women. 1.5.2

Pathways- to-Work Reform

Although the 1995 and 2001 reforms were associated with the ending of the increasing trend in numbers receiving disability benefits, the stock of recipients remained at a high level. As a result a new program, called Pathways- to-Work, designed to help claimants return to work, was implemented. It comprised three components: an increase in financial incentives to return to work with the ability to keep (approximately) 50 percent of the disability benefit for up to twelve months after returning to work; increased conditionality of benefits with mandatory work- focused interviews; and voluntary schemes to help beneficiaries manage their health problem more successfully. Initially the program was applied to those moving on to disability benefits (rather than existing claimants), and the impact of this program on new claimants was piloted and has been thoroughly evaluated (see Adam, Bozio, and Emmerson 2012).

Disability, Health, and Retirement in the United Kingdom

Fig. 1.28

73

Effect of the 1995 reform on retirement probability

Note: Figure shows estimated marginal effects of year dummies from a probit retirement model among those in work in the previous wave, also controlling for health stock (from Disney, Emmerson, and Wakefield 2003), a cubic in age, regional unemployment rate, whether own home outright, and dummies for reaching the state pension age and being in a couple. Model estimated on individuals aged fifty to sixty- four in 1991.

The program was first piloted in three large areas in October 2003, and four further large areas in April 2004. Later on the scheme was expanded to other areas of the country, in various phases. We present in figures 1.29 and 1.30 the outflow rate at six months out of IB in the pilot areas and subsequent expansion areas. After the introduction of the program the exit rate out of benefit increased substantially in each of the treated areas. This provides convincing evidence that the program had a decisive impact on movements off benefits, although there is some evidence that the effect became smaller as it was rolled into subsequent areas. Adam, Bozio, and Emmerson (2012) have shown that the impact on exit out of benefit has been concentrated on durations less than one year, suggesting that the program has mostly been successful in bringing forward exit out from benefit among those who would have left within one year of receipt, rather than removing from the disability rolls those who would otherwise have received benefits for longer than a year. Using a difference- in-difference strategy, the authors show that the program has had a significant effect on the probability to return to work in the two groups of pilot areas, but that this positive effect has been limited to those who do not report a mental health problem and was concentrated on women. The evaluation of this program highlights that outflows from benefit,

74

James Banks, Richard Blundell, Antoine Bozio, and Carl Emmerson

Fig. 1.29

Six-months outflow rate from IB, by pilot and nonpilot areas

Source: Administrative data on benefit flows, DWP.

and more specifically back to employment, do matter considerably. Furthermore, they are not necessarily the same: the study shows that while the impact on benefit receipt did not persist beyond twelve months, the employment impact was still significant at eighteen months. Even if policymakers have tended to concentrate on stricter eligibility with the hope of reducing inflows to benefit, the case for an outflow policy remains strong, at least within the UK institutional setting. 1.6

Conclusion

Over the last thirty years pathways to retirement have changed substantially in the United Kingdom. They were dominated by spells of unemployment in the late 1970s, with an increased importance of disability spells from the mid- 1980s onward. Pathways to retirement through unemployment were reduced in the early 1990s, while disability spells started to be less common from the mid- 1990s onward. At the end of the period—before the financial crisis—the direct route from work to retirement was increasingly more common. The empirical evidence on the underlying causes of these changes is still mixed. There is weak evidence of unemployment and disability reforms’ effects on the routes to retirement, but the general economic conditions seem to have been important driving forces during the entire period. Changes in health measures do not provide convincing explanations for these trends: mortality has been falling over the period without any link to the share of

Disability, Health, and Retirement in the United Kingdom

Fig. 1.30

75

Six-months outflow rate out of IB in the expansion areas

Source: Administrative data on benefit flows, DWP.

the population reporting ill health or disability or to the number claiming benefits. There is some evidence though that self- reported disability is associated with changes in the number of disability claimants. There is also evidence that recent reforms have also had an impact. The 1995 reform was associated with, at the very least, the halting of the previous growth in the rate of in-flow onto IB (and possibly also a fall in the percentage describing themselves as having a limiting long- standing illness). Evidence from the pilots of the Pathways- to-Work program suggests that those moving onto disability benefits moved off these benefits faster than they would otherwise have done as a direct result of the program. This program was also found to have an enduring impact on subsequent employment rates. While the recent financial crisis and associated recession is likely to lead to much attention being focused on getting the newly unemployed back in to paid work, those who receive disability benefits and who could potentially return to the labor market may still need assistance.

References Adam, S., A. Bozio, and C. Emmerson. 2012. “Reforming Disability Insurance in the UK: Evaluation of Pathways to Work.” Institute for Fiscal Studies (IFS) Working Paper, forthcoming.

76

James Banks, Richard Blundell, Antoine Bozio, and Carl Emmerson

Anyadike-Danes, M., and D. McVicar. 2007. “Has The Boom in Incapacity Benefit Claimant Numbers Passed its Peak?” Economic Research Institute of Northern Ireland (ERINI) Working Paper, December. Banks, J., R. Blundell, A. Bozio, and C. Emmerson. 2010. “Releasing Jobs for the Young? Early Retirement and Youth Unemployment in the United Kingdom.” In Social Security Programs and Retirement around the World: The Relationship to Youth Employment, edited by David A. Wise and Jonathan Gruber, 319– 44. Chicago: University of Chicago Press. Banks, J., A. Kapteyn, J. Smith, and A. van Soest. 2008. “Work Disability is a Pain in the ****, Especially in England, the Netherlands, and the United States.” In Health at Older Ages: The Causes and Consequences of Declining Disability among the Elderly, edited by D. Cutler and D. Wise, 251– 94. Chicago: University of Chicago Press. Benitez-Silva, H., R. Disney, and S. Jimenez-Martin. 2010. “Disability, Capacity for Work and the Business Cycle: An International Perspective.” Economic Policy 63:483– 536. Beveridge, Sir William. 1942. Social Insurance and Allied Services. Cmnd 6404. London: Her Majesty’s Stationery Office (HMSO). Bound, J. 1991. “Self-Reported Versus Objective Measures of Health in Retirement Models.” Journal of Human Resources 26:106– 38. Bound, J., and R. Burkhauser. 1999. “Economic Analysis of Transfer Programs Targeted on People with Disabilities.” In Handbook of Labor Economics, edited by O. Ashenfelter and D. Card, 3417– 528. Amsterdam: Elsevier. Bound, J., M. Schoenbaum, T. R. Stinebrickner, and T. Waidmann. 1999. “The Dynamic Effects of Health on the Labor Force Transitions of Older Workers.” Labour Economics 6:179– 202. Burchardt, T. 1999. “The Evolution of Disability Benefits in the UK: Re- weighting the Basket.” London School of Economics, Center for Analysis of Social Exclusion (CASE) Paper no. 26, June. Creedy, J., and R. Disney. 1985. Social Insurance in Transition: An Economic Analysis. New York: Oxford University Press. Disney, R., C. Emmerson, and M. Wakefield. 2003. “Ill Health and Retirement in Britain: A Panel Data Based Analysis.” Institute for Fiscal Studies, (IFS) Working Paper no. WP03/02. ———. 2006. “Ill Health and Retirement in Britain: A Panel Data Based Analysis.” Journal of Health Economics 25 (4): 621– 49. Disney, R., and S. Webb. 1991. “Why Are There So Many Long Term Sick in Britain?” Economic Journal 101 (405): 252– 62. Doherty, N. 1979. “National Insurance and Absence from Work.” Economic Journal 89 (353): 50– 65. Fenn, P. 1981. “Sickness Duration, Residual Disability, and Income Replacement: An Empirical Analysis.” Economic Journal 91 (361): 158– 73. Gilbert, B. 1965. “The British National Insurance Act of 1911 and the Commercial Insurance Lobby.” Journal of British Studies 4 (2): 127– 48. Gruber, J., and D. Wise, eds. 1999. Social Security Programs and Retirement around the World. Chicago: The University of Chicago Press. ———. 2004. Social Security Programs and Retirement around the World: MicroEstimation. Chicago: The University of Chicago Press. Kapteyn, A., J. Smith, and A. van Soest. 2007. “Vignettes and Self-Reports of Work Disability in the United States and the Netherlands.” American Economic Review 97 (1): 461– 73.

Disability, Health, and Retirement in the United Kingdom

77

Myers, R. J. 1982. “Why Do People Retire from Work Early?” Aging and Work 5:83– 91. Nagi, S. 1965. “Some Conceptual Issues in Disability and Rehabilitation.” In Sociology and Rehabilitation, edited by M. B. Sussman. Washington, DC: American Sociology Association. ———. 1991. “Disability Concepts Revisited: Implications for Prevention.” In Disability in America: Toward a National Agenda for Prevention, edited by A. M. Pope and A. R. Tarlove, 309– 27. Washington, DC: National Academy Press. Piachaud, D. 1986. “Disability, Retirement and Unemployment of Older Men.” Journal of Social Policy 15 (2): 145– 62. Shoven, J. 2010. “New Age Thinking: Alternative Ways of Measuring Age, Their Relationship to Labor Force Participation, Government Policies and GDP.” In Research Findings in the Economics of Aging, edited by David Wise, 17– 35. Chicago: The University of Chicago Press. Walker, A. 1981. “The Industrial Preference in State Compensation for Industrial Injury and Disease.” Social Policy and Administration 15:54– 71. Walker, A., and L. Walker. 1991. “Disability and Financial Need—The Failure of the Social Security System.” In Disability and Social Policy, edited by G. Dalley. London: Policy Studies Institute.

2 Disability Insurance, Population Health, and Employment in Sweden Lisa Jönsson, Mårten Palme, and Ingemar Svensson

2.1

Introduction

Compared to other industrialized countries, a large share of the Swedish population receives support from the disability insurance program (see Gruber and Wise 2010). In 2009, 20 percent of the males and more than 30 percent of the females in the age- group sixty to sixty- four received disability benefits. The disability insurance is the most common pathway out of the labor force for those who exit before the normal retirement age. In 2009, the expenditures from this program corresponded to 1.8 percent of the gross domestic product (GDP). Despite the extensive usage of the disability insurance, Sweden has a comparatively high employment rate among older workers. About 70 percent of the population aged fifty- five to sixty- four were employed in 2005, as compared to about 50 percent in Germany, 40 percent in Italy, and 60 percent in the United States (see Gruber and Wise 2010). As in most other European countries, however, there has been a dramatic fall in the employment rate of older men in recent decades. For example, the employment rate of males aged sixty to sixty- four has decreased from above 80 percent in the early Lisa Jönsson is a PhD student in economics at Stockholm University. Mårten Palme is professor of economics at Stockholm University. Ingemar Svensson is a researcher at the Swedish Pensions Agency. This chapter is a part of the National Bureau of Economic Research International Social Security Project. We are grateful to Agneta Kruse and Peter Skogman Thoursie as well as participants in seminars at the ISS meeting for comments on previous drafts of the chapter. We acknowledge financial support from the Bank of Sweden Tercentenary Foundation, the Swedish Council for Working Life, the Jan Wallanders and Tom Hedelius Foundation, and the Tore Browaldhs Foundation. For acknowledgments, sources of research support, and disclosure of the authors’ material financial relationships, if any, please see http://www.nber.org/ chapters/c12383.ack.

79

80

Lisa Jönsson, Mårten Palme, and Ingemar Svensson

1960s to slightly above 60 percent today. This development has caused concern in view of future financial burdens of an aging population. An explanation for the comparatively high employment rate among older workers combined with high disability insurance recipiency is that Sweden does not have a generous early retirement program. Many European countries introduced such programs in the 1970s and 1980s, and the large crosscountry differences in employment rates among older workers emerged during this period. For some time, however, the disability insurance program in Sweden developed toward an early retirement scheme. From only awarding disability benefits for health reasons in the 1960s, less strict eligibility criteria, especially for older workers, were introduced in the 1970s. These rules were abolished in the 1990s, and since 1997 an impaired work capacity for health reasons is again the sole eligibility criteria for disability benefits. In this chapter, we study to what extent the evolution of disability insurance utilization can be explained by changes in the population health status and by changes in eligibility rules, respectively. We focus on the age- group forty- five to sixty- four, which is the most important for the utilization of the disability insurance.1 We pose three main research questions. First, is there a relationship between disability insurance utilization and the development of population health status in recent decades? Second, did the changes in eligibility rules for older workers affect disability insurance utilization? Third, did the changes in eligibility rules for older workers affect labor market outcomes such as employment and labor- force participation, or were they “crowded out” by the utilization of other income security programs? Wadensjö (1996) and Hedström (1987) have previously analyzed the effect of changes in eligibility rules of the disability insurance program in Sweden, in particular the introduction of eligibility rules for labor market reasons targeted at older workers in the early 1970s. Karlström, Palme, and Svensson (2008) studied the abolishment of the special eligibility rules for older workers in 1997. In this study, we extend the previous literature by considering the full forty- year history of eligibility changes starting in the early 1970s and by relating it to different labor market outcomes. In addition, we put together a comparatively wide set of population health measures and relate the development of these measures to the development of disability insurance utilization in different demographic groups. Although we strive to have a broad scope for the empirical analysis, we leave out several plausible explanations for the fluctuations in the utilization of the disability insurance. Previous studies have analyzed the effect of economic incentives on the disability insurance in Sweden (e.g., Kruse and Söderström 1989; Skogman Thoursie 1999; Palme and Svensson 1999, 1. Disability insurance recipiency in younger ages has increased over time, which is a source of concern. In this chapter, however, we limit our focus to the utilization of the disability insurance in older ages.

Disability Insurance, Population Health, and Employment in Sweden

81

2004). Changes in social norms regarding the utilization of the sickness insurance program has been studied by Lindbeck, Palme, and Persson (2009), and should also be a plausible explanation for changes in the utilization of the disability insurance. Finally, changes in the demand for labor with disabilities have not yet been properly studied on Swedish data, but is an interesting topic for further research. The rest of the chapter is organized as follows. Section 2.2 gives a brief history of the development of the disability insurance program in Sweden. Section 2.3 describes the development of various population health measures over time. Section 2.4 describes the development of the utilization of disability insurance and the development of labor market outcomes. Section 2.5 studies the relation between population health and disability insurance utilization. Section 2.6 studies the relation between the eligibility reforms for older workers, disability insurance utilization, and various labor market outcomes. Section 2.7 concludes. 2.2

Historical Overview of the Disability Insurance in Sweden

The disability insurance (DI) is one of Sweden’s most important income security programs. Its main objective is to replace foregone earnings for workers below the retirement age with a permanently impaired working ability for health reasons. The related sickness insurance replaces foregone earnings due to a temporarily impaired working ability for health reasons. Disability benefits can be granted part- time or full- time, depending on the extent of the work impairment. Sweden’s disability insurance has a comparatively long history. The first public pension system covering all citizens, including an invalidity pension, was already implemented in 1914. The recent history of Sweden’s disability insurance, which we analyze in this chapter, started when a public incomerelated supplementary pension scheme (ATP) was introduced, following a referendum in 1957. The new scheme came into place in 1960 and the first payments were made in 1963, but since the program was phased in, it did not reach its full maturity until the beginning of the 1990s. The pension benefit under this scheme replaced 60 percent of the average of an individual’s fifteen best years of earnings up to a social security ceiling. The benefit was linearly reduced if the worker contributed less than thirty years to the scheme, and it was financed through payroll taxes. The new income- related pension also included disability insurance. The size of the individual benefit was calculated in the same way as the old- age pension benefit, with the actual earnings history replaced by an assumed earnings profile. Eligibility for disability benefits was initially based on health. Disability benefits were awarded by the local Social Insurance Agency after a physical examination by a medical doctor. If the health status prevented the worker from doing his or her regular job, but not one that

82

Lisa Jönsson, Mårten Palme, and Ingemar Svensson

suited the worker’s general qualifications, the worker was required to go through a retraining program. Eligibility rules for disability benefits changed on several occasions after the new disability insurance was first introduced. Table 2.1 summarizes the main eras in this history. The first major reform took place in 1970 and had two main components. First, special eligibility rules were introduced for workers aged between sixty- three and the normal retirement age (which was sixty- seven at that time). These rules implied that (a) no rehabilitation or retraining for a new occupation was required if the worker’s health status did not permit his or her regular work, (b) the medical requirements for assessing inability to work were substantially lower for this age- group, and (c) functional limitations due to normal aging could also be considered for eligibility for DI. Second, unemployment was made an additional criterion for DI eligibility in all age- groups. Long- term unemployed workers with functional limitations were made eligible for DI after having been unemployed for one to two years. The next reform toward more generous eligibility rules for DI took place in 1972, when pure labor market reasons for older workers were introduced. These rules implied that workers aged between sixty- three and the normal retirement age could become eligible for DI if they were still unemployed when reaching the time limit in the unemployment insurance, even without any health limitations. In 1974, the age limit for pure labor market reasons was lowered from age sixty- three to age sixty, and in 1976 the age limit for the special eligibility rules for older workers, introduced in 1970, were lowered from age sixty- three to age sixty. The latter change was partly made as a consequence of the decrease in the normal retirement age from age sixtyseven to sixty- five in 1976. Two changes led to higher replacement levels in the disability insurance during the 1970s and 1980s. First, the maturity of the supplementary pension scheme (ATP) led to higher replacement levels in general. Second, the introduction of a special supplement in 1969 led to improvements for lowTable 2.1

Changes in eligibility rules for the disability insurance

Period – 1962 1963– 1970(June) 1970(July)– 1972(June) 1972(July)– 1974(June) 1974(July)– 1976(June) 1976(July)– 1991(Sept) 1991(Oct)– 1996 1997–

Medical reasons Yes Yes Yes Yes Yes Yes Yes Yes

Labor market and medical reasons combined Very small Some Yes Yes Yes Yes Yes Very small

Special eligibility rules for older workers

Pure labor market reasons for older workers

No No Yes, aged 63– 66 Yes, aged 63– 66 Yes, aged 63– 66 Yes, aged 60– 64 Yes, aged 60– 64 No

No No No Yes, aged 63– 66 Yes, aged 60– 66 Yes, aged 60– 64 No No

Disability Insurance, Population Health, and Employment in Sweden

83

income pensioners. It was reduced on a one- to-one basis against income from the supplementary pension (ATP). The special supplement applied to all types of pensions and was gradually increased from 1969 to 1981. In 1977, the amount of the special supplement was doubled for DI pensioners only, and the subsequent development of the supplement for DI beneficiaries followed the gradual increase for regular pensioners but at twice as high a level. The policy toward more liberal DI eligibility rules was reversed in the 1990s. The eligibility for DI for pure labor market reasons for older workers, introduced in 1972, was abolished in 1991. In 1991 and 1992, a new legislation was also enacted with the purpose of reducing sickness absence, which required employers to improve the work environment and take responsibility for the rehabilitation of employees. The social insurance agencies were made responsible for the coordination of rehabilitation among the employer, the public health care system, labor market authorities, the local government, and the individual. In 1997, the favorable eligibility rules for older workers and the eligibility for DI for labor market reasons and medical reasons combined, introduced in 1970, were abolished. Since then, an impaired work capacity for health reasons has been the only eligibility criterion for disability insurance. On January 1, 2003, the disability insurance was shifted from the public pension system to the public sickness insurance system, following a major pension reform. Benefits were renamed and the calculation of benefits changed, but the assessment of eligibility remained the same. Benefits were calculated as 64 percent of the assumed income; that is, the income of the best three of the last five- to-eight years, depending on age, up to a social security ceiling. Individuals not qualifying for the income- related insurance received a guarantee benefit. From January 2005 onward a reassessment of the working capacity for granted individuals should be made every third year. Also in 2005, the organization of the Social Insurance Agency changed, when the twentyone regional offices were integrated into one central authority. In 2008, the eligibility for disability benefits was substantially tightened. For all cases granted after July 1, 2008, working capacity had to be permanently reduced in relation to the entire labor market in order to qualify for benefits. 2.3

The Development of Population Health

Changes in population health would be a natural determinant of the evolution of disability insurance recipiency over time. To describe the development of population health in Sweden over the last decades, we use three main groups of health measures: the mortality rate, self- reported health measures from the Survey on Living Conditions and, finally, the utilization of inpatient care from the National Patient Register. For our purposes, each

84

Lisa Jönsson, Mårten Palme, and Ingemar Svensson

of the three measures of population health has its obvious advantages and disadvantages. The main advantage of mortality as a health measure is that it is objective and accurately measured through the population censuses, which makes it easily comparable across countries and time. A disadvantage is that mortality is more of an outcome measure, in part reflecting technological advances within the health care system rather than the average health status of the population. It might also target a slightly inappropriate population for our purposes, since it measures the health of the marginal survivors rather than the marginal workers. Diseases that cause death and diseases that reduce working capacity may be different. Finally, a higher survival rate implies that individuals who would previously have died now survive, although possibly with a bad health status. This could change the composition of the population, and in turn affect the average health status negatively even though mortality decreases. The advantage of the self- reported health measures is that they are better targeted toward the aspects of health that are relevant for the ability to remain in the labor force. The main disadvantage relates to the fact that they are subjective. Fluctuations over time might capture changes in the interpretation of the questions and the general view of health rather than changes in actual health. Another disadvantage is that the self- reported health measures may be state dependent. For example, since an impaired work capacity is an eligibility criterion for disability benefits, individuals may be more likely to report an impaired work capacity as a result of receiving benefits. The advantage of the utilization of inpatient care as a health measure is that it is accurately measured, since it is obtained from registers, and closely related to the health status of the worker. It has, however, the disadvantage of the self- reported measures of being sensitive to changes in the general view of health. Furthermore, the utilization of inpatient care might be influenced by public health care spending, working procedures at the hospitals, and the division of labor between outpatient and inpatient care. 2.3.1

Mortality

The mortality rate is defined as the number of deaths during a year divided by the average size of the population in a particular age- group. We present the development of mortality in Sweden in three different ways. Figure 2.1 shows the development from 1960 to 2009 of the age at which men and women faced the same mortality rate as that which sixty and sixty- fiveyear- olds faced in 1960. Figure 2.2 shows the mortality rate by age in 1960 and 2005 for men and women, respectively. Finally, figure 2.3 shows the development of the mortality rate for men and women, respectively, at the age of fifty- five, sixty, and sixty- five from 1950 to 2009. These figures reveal two interesting results. First, there has been a marked decrease in mortality for both men and women during the period under

Fig. 2.1

Ages of equal mortality probability, 1960 to 2009

Source: Statistics Sweden.

Fig. 2.2

Mortality rates by age, 1960 and 2005

Source: Statistics Sweden.

86

Lisa Jönsson, Mårten Palme, and Ingemar Svensson

Fig. 2.3

Mortality rates by year, 1950 to 2009

Source: Statistics Sweden.

study. Figure 2.1 shows that the age of equal mortality that the sixty- yearolds faced in 1960 increased by 8.4 years for men and 7.1 years for women until 2009, and that the age of equal mortality that the sixty- five- yearolds faced in 1960 increased by about 7.5 years for both genders until 2009. Figure 2.2 shows that the age at which the mortality rate passes 5 percent increased by five years for men and seven years for women between 1960 and 2005. Finally, figure 2.3 shows that the mortality rate has halved from 1950 to 2009 for men and women in all age groups. The second result is that the large decrease in mortality happened much earlier for women. Figure 2.3 shows that the mortality rate for sixty- fiveyear- old women decreased from about 2 percent in 1950 to about 1 percent in 1980. Since then, it has decreased by less than 0.25 percentage points. Also, for women at younger ages, the mortality rate decreased most rapidly before 1980. For men, on the other hand, the main improvement occurred after 1980. Between 1980 and 2009, the mortality rate for sixty- five- year- old men decreased by 1 percentage point to just over 1 percent. For younger men, mortality also decreased most rapidly during this period. The same pattern is visible in the series of equal mortality probability in figure 2.1. The gap between men and women broadened until the mid- 1980s, and thereafter narrowed substantially. The development for the youngest males is so steep after 1980 that it even surpasses the development for women in the mid 2000s. 2.3.2

Self-Reported Health

Self- reported information about the health of the Swedish population is collected by Statistics Sweden through the Survey on Living Conditions (ULF). It is a yearly survey of a random sample of about 7,500 individuals

Disability Insurance, Population Health, and Employment in Sweden

87

aged eighteen to sixty- four that has been produced since 1975. The survey contains a large set of questions about health in general as well as about particular diseases. We present the results from the survey for men and women in the age- groups forty- five to fifty- four, fifty- five to fifty- nine, and sixty to sixty- four. All series show the share of the population in the age- group with a certain condition. To reduce the problem of large stochastic errors due to small sample sizes within each demographic group, we present three- year moving averages. We also focus on the long- run development of the series rather than fluctuations in single years. Figure 2.4 presents the development of a set of general health indicators from the survey. The indicator “Doctor’s visit” shows the share of the population who visited a doctor within the last three months. The indicator “Long- term disease” shows the share of the population with at least one disease in a list of diagnoses, and the indicator “Impaired work capacity” shows the share of the population reporting that the long- term disease causes an impaired work capacity. The indicator “Impaired ability to move” shows the share of the population who are not able to run 100 meters. Finally, the indicators “Poor health” and “Good health” show the self- assessed health status, based on a question where the individual evaluates his or her general health on a particular scale.2 The mortality rate in each demographic group is included as a comparison. As opposed to the development of mortality, figure 2.4 shows no overall trend for the development of the self- reported health indicators over time for neither men nor women. Two of the indicators, the share of the population with a long- term disease and the share of the population that recently visited a doctor, show an adverse or invariant development over time in all demographic groups. Both of these indicators have increased for men and women aged forty- five to fifty- four and women aged fifty- five to fifty- nine, and have remained stable for men aged fifty- five to fifty- nine and men and women aged sixty to sixty- four. In general, these two indicators seem to have developed more adversely for women than for men and for younger than for older age- groups. Other self- reported health indicators have developed in opposite directions in different demographic groups. The share of the population with an impaired work capacity has remained stable over time for men aged fortyfive to fifty- four and fifty- five to fifty- nine, but has decreased by a third for men aged sixty to sixty- four since 1976. For women, there is instead an upward trend, in particular for the forty- five to fifty- four and fifty- five to fifty- nine- year- olds. Also, women aged sixty to sixty- four show a slight upward trend, implying that the share of women with an impaired work capacity has not developed in the same promising way as the share of men with an impaired work capacity in this age- group. 2. See Statistics Sweden (2009) for more information.

Fig. 2.4

Self-reported health indicators and mortality, 1976 to 2005

Source: Statistics Sweden.

Disability Insurance, Population Health, and Employment in Sweden

89

The remaining health indicators show an invariant or advantageous development over time. The health indicator with the most favorable development is the share of the population with an impaired ability to move, which has decreased in all demographic groups. The reduction has been particularly large for men and women aged sixty to sixty- four, where the share has almost halved over the period. This is the only indicator improving over time for all female age- groups. The share of the population in poor health remained stable for men and women aged forty- five to fifty- four and women aged fifty- five to fifty- nine, and decreased slightly for men aged fifty- five to fifty- nine and men and women aged sixty to sixty- four. This broad pattern is supported by the development of the share of the population in good health, which has remained stable for men and women in the age- group forty- five to fifty- four and has improved for the two older age- groups. In the age- group sixty to sixty- four, the share of the population in good health has increased from about 55 to 65 percent for both men and women. In addition to the general health indicators, the Survey on Living Conditions contains information about self- reported disease prevalence for a number of diagnoses. Figure 2.5 shows the development over time for the prevalence of diseases for five diagnosis groups that are of particular importance for the disability insurance. These include circulatory diseases, musculoskeletal diseases, mental disorders, diseases in the nervous system, and endocrine diseases. The development of the mortality rate in each demographic group is included as a comparison. The development of self- reported disease prevalence does not reveal any unambiguous trend. Different diagnosis groups follow different patterns over time. The prevalence of musculoskeletal diseases has increased in all demographic groups, except men aged sixty to sixty- four, since the mid1980s. The increase has been particularly striking for women. Also, the prevalence of endocrine diseases has increased in all demographic groups since the mid- 1980s. This is likely to reflect that problems related to obesity, such as diabetes, have become more common. The prevalence of mental disorders has remained stable for most of the period, but has increased since 1995 in the two youngest age- groups. The prevalence of nervous diseases has remained stable for the two youngest age- groups and has decreased slightly in the age- group sixty to sixty- four. Although fluctuations in the prevalence of circulatory diseases have been large, there are no clear patterns in the long run. 2.3.3

Inpatient Care

The utilization of inpatient care is registered in The National Patient Register at The National Board of Health and Welfare. The register contains all overnight hospital visits in Sweden from 1987 and onward. For a selection of counties, however, the register contains information from as far back as

Fig. 2.5

Self-reported disease prevalence and mortality, 1976 to 2005

Source: Statistics Sweden.

Disability Insurance, Population Health, and Employment in Sweden

91

1968. To get a longer perspective, we use information about inpatient care for four of Sweden’s twenty- one counties from 1968 to 2008. The presented series show the share of the population in the four counties taken together that experienced at least one overnight hospital visit during the year.3 Figure 2.6 shows the development of inpatient care from 1968 to 2008 for men and women in the three age- groups. In the two oldest age- groups, the utilization of inpatient care has been higher for men than for women throughout the period. In the youngest age- group, utilization has been very similar across genders, although slightly higher for women. As expected, the utilization of inpatient care increases by age. Over time, the series show a slightly increasing trend in the utilization of inpatient care up to the mid1980s, followed by a substantial decrease to levels similar to, or even lower than, those in the late 1960s. The decline was much more pronounced for men than for women in the two oldest age- groups, which has lead to a convergence in the utilization of inpatient care across genders. This suggests an improvement in the health of men relative to women in these agegroups since the mid- 1980s, which is consistent with the results from the self- reported health measures and the development of the mortality rates reported previously. 2.3.4

Conclusions about the Development of Population Health

The three main measures of population health give an ambiguous picture of the development of the general health status in Sweden over the past decades. The development of the mortality rate suggests a marked health improvement over time for both men and women. The development of the self- reported health indicators does, however, not confirm this result. Only men aged sixty to sixty- four show an invariant or positive development for all self- reported health indicators over time. Also the development of the utilization of inpatient care does not indicate a clear- cut health improvement over time. Two main conclusions emerge from our analysis. First, health seems to have developed more adversely for females than for males since the mid1970s. Female mortality decreased most rapidly up until the 1970s, while male mortality decreased substantially from 1980 onward. For the selfreported health measures, we only have information from 1976 onward. Since then, these measures show a less advantageous development of the health of females than the health of males. The development of the utilization of inpatient care confirms this pattern. Since the early 1980s, the utilization of inpatient care in the two oldest age- groups has decreased more 3. The selected counties are Dalarna, Gävleborg, Uppsala, and Jämtland. The development of inpatient care in these counties taken together follows that of the entirety of Sweden from 1987 onward.

92

Lisa Jönsson, Mårten Palme, and Ingemar Svensson

Fig. 2.6 The share of the population receiving inpatient care during the year, 1968 to 2008 Source: National Board of Health and Welfare.

for men than for women. The development in the youngest age- group, aged forty- five to fifty- four, however, has been similar for men and women. The second main conclusion is that the health of younger age- groups seems to have deteriorated compared to older ones. The self- reported health indicators suggested a worsened health status over time for both men and women in the youngest age- group, aged forty- five to fifty- four, whereas the health status of the oldest age- group, aged sixty to sixty- four, improved. This is supported by the development of the mortality rate, where the decline

Disability Insurance, Population Health, and Employment in Sweden

93

was steeper at age sixty- five than at ages fifty- five and sixty. It is not apparent, however, in the development of inpatient care utilization. The two main conclusions are highlighted in figure 2.7, which shows the share of individuals with self-reported poor health and the mortality rate in the three age groups in 1976 and 2005. Panel A shows that the health of men in the youngest age group, aged forty-five to fifty-four, has hardly improved over the thirty-year period, while the health of men aged fifty-five to fiftynine and sixty to sixty-four has improved substantially. For women, the gains in mortality are much smaller than for men over the period, and the share of women with self-reported poor health has even increased for the younger

Fig. 2.7

Share of population in poor health and the mortality rate, 1976 and 2005

94

Lisa Jönsson, Mårten Palme, and Ingemar Svensson

age groups, aged forty-five to fifty-four and fifty-five to fifty-nine. For older women, aged sixty to sixty-four, on the other hand, the share in poor health has decreased substantially over the period. 2.4

2.4.1

Disability Insurance Utilization and the Relation to Labor Market Outcomes The Development of Disability Insurance Recipiency

Figure 2.8 shows the prevalence of disability insurance recipiency by the end of the year from 1962 to 2009 for men and women, respectively.4 The prevalence is defined as the share of the population in an age- group that receives full or partial disability benefits in a given year. Panel A reveals a clear upward trend of disability insurance prevalence for men in all agegroups until the early 1990s. The prevalence for men in the oldest age- group, aged sixty to sixty- four, increased from around 10 percent in 1962 to above 35 percent in 1995. The increase from about 5 percent in 1962 to about 18 percent in 1995 for men in the age- group fifty- five to fifty- nine is also notable. After the mid- 1990s, there is a clear trend break for men in the oldest age- group, with a decrease in the disability insurance prevalence from above 35 percent to about 20 percent. A smaller decrease can be seen in the agegroup fifty- five to fifty- nine. The prevalence of disability insurance recipiency for men aged forty- five to fifty- four, however, continued to increase until the late 2000s. Panel B shows a similar increase in the prevalence of disability insurance recipiency for women until the early 1990s. The increase was most rapid for women aged sixty to sixty- four, for whom the prevalence of disability insurance recipiency rose to the same level as that for men. For the two younger age- groups, however, DI recipiency increased to even higher levels than for men of the same age. A similar trend break as that for men can also be seen for women in the early 1990s, but the development since then has been far less favorable than that for men. The prevalence of DI recipiency for women in the oldest age- group has remained above 30 percent, and the prevalence in the younger age- groups continued to rise until the mid- or late 2000s. Compared to the early 1960s, the level of DI recipiency has more than tripled for all women. Figure 2.9 shows the development of the incidence of disability insurance recipiency from 1971 to 2009. The incidence is defined as the share of the population at risk that starts to receive full or partial disability benefits in a given year, whereas the population at risk is defined as all individuals in the 4. The prevalence of DI recipiency is measured in January from 1963 to 1984 and in December from 1985 onward. In our analysis, we let the January figures from 1963 to 1984 represent DI recipiency at the end of the previous year. For example, DI recipiency in January 1963 is presented as DI recipiency in 1962.

Disability Insurance, Population Health, and Employment in Sweden

Fig. 2.8

95

Disability insurance prevalence by gender and age-group, 1962 to 2009

Source: Swedish Social Insurance Agency.

age- group that are not already receiving disability benefits. Due to limited data availability, we redefined the youngest age- group to age fifty to fiftyfour rather than forty- five to fifty- four as in the previous section. Figure 2.9 shows that the incidence of DI recipiency in the two youngest age- groups increased in a similar manner for men and women until the early 1990s. In the oldest age- group, however, the development is somewhat different across genders. The incidence for men aged sixty to sixty- four started at a historically high level in the beginning of the 1970s and decreased until the early 1980s, while the incidence for women aged sixty to sixty- four

96

Lisa Jönsson, Mårten Palme, and Ingemar Svensson

Fig. 2.9

Disability insurance incidence by gender and age-group, 1971 to 2009

Source: Swedish Social Insurance Agency.

was relatively stable until the early 1980s. Thereafter, the incidence of DI recipiency in the oldest age- group increased for both men and women until the early 1990s, although the level for men was still higher than that for women. From the early 1990s onward, the incidence of DI recipiency has developed similarly across demographic groups. The incidence decreased markedly in all age- groups during the 1990s. In the early 2000s, the incidence once more increased, in particular for women, but then declined to historically low levels in all groups until 2009. The recent decline in the incidence of disability insurance recipiency is

Disability Insurance, Population Health, and Employment in Sweden

97

remarkably large. Since the upward trend was broken in 2005, the incidence has steadily declined to very low levels in all age- groups. As described in section 2.2, new regulation came into place in July 2008 introducing stricter eligibility criteria for disability benefits. It is apparent, however, that the decline had begun well before then. The decline in incidence can also be seen in the development of the prevalence of disability insurance recipiency in figure 2.8, although the levels are still high. If the incidence remains at these low levels, we are likely to see a steep decline in the prevalence of DI recipiency over the coming years. Figure 2.10 shows the development of disability insurance incidence by diagnosis. We present the three most common groups of diagnoses for disability benefits award: circulatory diseases, musculoskeletal diseases, and mental disorders. We also include a category for all other diagnoses. For the group aged sixty to sixty- four we add a category for labor market reasons for the part of the period when this was a sufficient criterion for DI eligibility in this age- group, as described in section 2.2. The figure shows that musculoskeletal diseases have been the most common diagnosis for disability benefits award in all demographic groups throughout the period. The large increases in the incidence of DI recipiency until the early 1990s and the subsequent sharp declines were primarily attributed to musculoskeletal diagnoses. The figure also reveals that mental disorders have become increasingly important in recent years in all demographic groups, in particular for the youngest age- group and for women. Circulatory diagnoses have lost importance over time in all demographic groups. Even though musculoskeletal diseases and mental disorders explain a large share of the increase in incidence during the early 2000s, the series capturing other diagnoses also shows a similar increase. For the oldest age- group, the pure labor market reasons, introduced in 1972, did not become important until the 1980s. In the mid- 1980s, labor market reasons were the most common reason for granting disability benefits to sixty to sixty- four- year- olds. Thereafter, the importance of labor market reasons declined substantially while the importance of musculoskeletal diagnoses increased. 2.4.2

The Development of Labor Market Outcomes

To describe the development of some central labor market outcomes we use data from the Labor Force Surveys, collected by Statistics Sweden. We present the development from 1963 to 2007 for men and women in the agegroups forty- five to fifty- four, fifty- five to fifty- nine, and sixty to sixty- four. Figure 2.11 shows the development of the employment rate. Panel A reveals a similar pattern across age- groups for the development of male employment over time. Male employment decreased from the early 1960s until the early 1990s, and then experienced a pronounced dip during the recession in Sweden in the early 1990s. From the late 1990s onward, male employment

Fig. 2.10

Disability insurance incidence by diagnosis, 1971 to 2005

Source: Swedish Social Insurance Agency.

Disability Insurance, Population Health, and Employment in Sweden

Fig. 2.11

99

Employment rates by gender and age-group, 1963 to 2007

Sources: Swedish Labor Force Survey; Statistics Sweden.

again increased. The pattern is most pronounced for the age- group sixty to sixty- four, where the employment rate decreased from above 80 percent in 1963 to around 50 percent in the mid- 1990s, and then increased to about 65 percent in 2009. Changes in employment in the age- group forty- five to fifty- four are much smaller. Employment in this age- group remained above 90 percent until the economic crisis in the early 1990s, and then shifted down to a level just below 90 percent. Panel B in figure 2.11 shows a different development of the female employ-

100

Lisa Jönsson, Mårten Palme, and Ingemar Svensson

ment rate. Until the 1990s, employment increased substantially for women in all age- groups. The increase was most rapid for the two youngest agegroups, where employment increased from 55 to 90 percent in the age- group forty- five to fifty- four and from 40 to 80 percent in the age- group fiftyfive to fifty- nine. By the time of the economic crisis in the beginning of the 1990s, the stable increase in employment ceased and was followed by a slight decrease in all age- groups. In recent years, however, there has been a marked increase in the employment rate of women aged sixty to sixty- four. Between 2000 and 2007, employment increased by 15 percentage points to about 58 percent in this group. Also, the female employment rate in the age- groups forty- five to fifty- four and fifty- five to fifty- nine has increased since the late 1990s. Figure 2.12 shows the development of the unemployment rate. As opposed to the employment rate, the unemployment rate has developed in a similar manner for men and women. There are two notable properties of the development. First, unemployment in the oldest age- group seems to have been more sensitive to business cycle movements than other age- groups before 1990. After 1990, the series for different age- groups correspond remarkably well. Second, the unemployment rate seems to have increased to a permanently higher level in all age- groups after the recession in the beginning of the 1990s. Figure 2.13 shows the share of the population not participating in the labor force. Panel A reveals an increasing trend in the share of the male population out of the labor force in all age- groups. The steepest increase occured for the oldest age- group, aged sixty to sixty- four, from about 15 percent in 1963 to about 45 percent in 2000. Thereafter, however, the trend reversed and non- labor force participation for men aged sixty to sixty- four decreased to a level just above 30 percent in 2009. The increase in non- labor force participation in the two younger age- groups has been much more modest and the series have stabilized rather than declined in the 2000s. Panel B in figure 2.13 reveals a very different development of the nonlabor force participation rate for women compared to men. There has been a decreasing trend in all age- groups and the decrease has been most pronounced in the two younger groups. Non- labor force participation decreased from 45 to 10 percent for the age- group forty- five to fifty- four and from 55 to 20 percent for the age- group fifty- five to fifty- nine between 1963 and 1990. Thereafter, the level has remained stable for the age- group fifty- five to fifty- nine and increased slightly for the age- group forty- five to fifty- four. The non- labor force participation rate for the oldest age- group, aged sixty to sixty- four, decreased less rapidly, from 65 to 40 percent over the full period. Figure 2.14 shows the labor force participation rate in one- year agegroups in 1965, 1985, and 2003 for men and women, respectively. Panel A reveals a marked decrease in the labor force participation of men above the age of fifty- eight between 1965 and 1985, but no visible change between

Disability Insurance, Population Health, and Employment in Sweden

Fig. 2.12

101

Unemployment rates by gender and age-group, 1963 to 2007

Sources: Swedish Labor Force Survey; Statistics Sweden.

1985 and 2003. In the ages below fifty- eight, however, there was a small but visible decrease in the labor force participation both between 1965 and 1985 and between 1985 and 2003. Panel B reveals a large increase in female labor force participation at ages below sixty- five between 1965 and 1985, and a smaller increase between 1985 and 2003. Beyond the age of sixty- five, there was a decrease in labor force participation between 1965 and 1985 that can be explained by a decrease in the normal retirement age from sixty- seven to sixty- five in 1976.

102

Lisa Jönsson, Mårten Palme, and Ingemar Svensson

Fig. 2.13 to 2007

Non-labor force participation rates by gender and age-group, 1963

Sources: Swedish Labor Force Survey; Statistics Sweden.

In section 2.3 we showed that the mortality rate has fallen quite dramatically in all age- groups in recent decades. An increased life expectancy can affect the time an individual spends in the labor force, the time an individual spends as a retiree, or both. To report the changes in labor force participation over time without correcting for changes in mortality, as we have done so far, implicitly implies that all gains in life expectancy are taken out in time as a retiree. The other extreme would be that the time as a retiree is held

Disability Insurance, Population Health, and Employment in Sweden

Fig. 2.14

103

Labor force participation by age and year

Source: Own calculations based on data from Statistics Sweden.

constant and the increased life expectancy only affects the time spent in the labor force. One way of investigating this alternative is to calculate labor force participation rates at constant mortality risks. The results from this exercise are shown in figure 2.15, which shows the labor force participation rate at a given mortality rate. As was apparent in figure 2.14, labor force participation for men decreased even when not taking the decrease in mortality into account. Hence, figure 2.15 reveals an even larger decrease in labor force participation rates for men. In contrast

104

Lisa Jönsson, Mårten Palme, and Ingemar Svensson

Fig. 2.15

Labor force participation by mortality rate and year

Source: Own calculations based on data from Statistics Sweden.

to the results in figure 2.14, however, there is also a large decrease between 1985 and 2003, a period during which the mortality rate of men decreased significantly. Interestingly, the increased labor force participation for females between 1965 and 1985 that was shown in figure 2.14 reverses in figure 2.15, since mortality improves more than labor force participation increases. Only for very low mortality rates did labor force participation still increase between 1965 and 1985. For women at higher mortality rates, labor force participation decreased substantially between 1965 and 1985, and continued to decrease until 2003.

Disability Insurance, Population Health, and Employment in Sweden

2.4.3

105

Pathways to Retirement

Figure 2.16 shows the development of the share of the population receiving disability benefits, the share of nonemployed, and the share not participating in the labor force from 1963 to 2009. Panels C and E show a large increase in nonemployment and non- labor force participation of elderly men between 1970 and 1988, along with a corresponding increase in disability insurance recipiency. In the age- group fifty- five to fifty- nine, nonemployment increased by 5.2 percent between 1970 and 1988, non- labor force participation by 5.3 percent, and disability insurance recipiency by 8.1 percent. For men aged sixty to sixty- four the corresponding figures were 15.0, 15.5, and 15.2. Hence, the disability insurance seems to have been the dominating pathway to retirement for older males until 1990. For men aged forty- five to fifty- four, the same close correspondence between nonemployment and non- labor force participation on the one hand and disability insurance recipiency on the other is not apparent. Nonemployment and non- labor force participation increased at the end of the 1960s and did not increase again until 1990, while disability insurance recipiency was gradually increasing. The right- hand panels in figure 2.16 show a very different pattern for females until 1990. The large gap between disability insurance recipiency and the non- labor force participation rate consists of the diminishing fraction of homemakers. The fact that the gap closes earlier for younger age- groups tells us that this development is primarily a cohort effect. Since the close link between nonemployment, non- labor force participation, and disability insurance recipiency for men was broken in 1990, the development for women has been more similar to that for men. In all age and gender groups, a gap emerged between disability insurance recipiency and nonemployment from 1990 onward. In the oldest age- group, this gap primarily consisted of increased non- labor force participation that was not due to increased disability insurance recipiency. To study the background to this development, we use the annual income statistics from tax returns. Figure 2.17 shows the share of men and women aged fifty- five to fifty- nine and sixty to sixty- four with one of the four main income security programs in Sweden as the main income source. These include the disability insurance, the sickness insurance, the unemployment insurance, and occupational insurance. An individual is classified as receiving his or her main income from a certain program if the benefits from the program account for 50 percent or more of the total yearly income. The upper panels in figure 2.17 show that the gap between disability insurance recipiency and nonemployment in the age- group fifty- five to fifty- nine has been mainly accounted for by the unemployment insurance. Also the sickness insurance has been a main income source for a substantial share of the population in this age- group, and the relative importance of the sickness

Fig. 2.16

DI prevalence, non-labor force participation and nonemployment

Sources: Swedish Social Insurance Agency and Statistics Sweden.

Disability Insurance, Population Health, and Employment in Sweden

Fig. 2.17

107

Main income source, 1990 to 2005

Source: Annual income statistics, Statistics Sweden.

and the unemployment insurance has shifted over time. The lower panels in figure 2.17 show a different pattern for the age- group sixty to sixty- four. A large share of the gap between disability insurance recipiency and non- labor force participation in this age- group after 1990 has been accounted for by benefits from occupational insurance schemes. 2.4.4

Conclusions about the Disability Insurance and Labor Market Outcomes

The results in this section showed some general patterns. The development of disability insurance recipiency and labor market outcomes can be divided into two main eras: before and after 1990. From the early 1960s until 1990, the share of disability insurance recipients increased in all demo-

108

Lisa Jönsson, Mårten Palme, and Ingemar Svensson

graphic groups. For males, the increase was closely accompanied by an equal increase in non- labor force participation and nonemployment. For females, the correspondence between disability insurance recipiency and labor market outcomes was weak, but the gap was closing over time as female labor force participation increased. After 1990, the development of disability insurance recipiency was different across demographic groups. While disability insurance recipiency decreased for men aged fifty- five to fifty- nine and sixty to sixty- four, it continued to increase for men and women aged forty- five to fifty- four and women aged fifty- five to fifty- nine, and only stabilized for women aged sixty to sixty- four. From the early 1990s, a gap emerged between disability insurance recipiency and nonemployment. A closer study showed that the disability insurance program was losing importance as a pathway to permanent exit from the labor force. In the age- group fifty- five to fifty- nine, the unemployment and sickness insurance programs became more important while in the age- group sixty to sixty- four, the role of occupational insurances increased substantially after 1990. Finally, we should note the sharp decrease in the incidence of disability insurance utilization in recent years. The incidence reached a historically low level of below 1 percent of the risk population in all demographic groups in 2009. If these exceptionally low levels of incidence continue, it will lead to a sharp decrease in the prevalence of disability insurance utilization in the coming years. 2.5 2.5.1

Population Health and Disability Insurance Disability Insurance Prevalence and Population Health

Figure 2.18 presents the development of the prevalence of disability insurance recipiency along with the mortality rate and the share of the population with a self- reported impaired work capacity, the share of the population with an impaired ability to move, and the share of the population with selfassessed poor health from the Survey on Living Conditions. The most apparent result from this figure is the lack of correlation between the mortality rate and the prevalence of DI recipiency in all of the demographic groups. The increase in the prevalence of DI recipiency for all groups until the mid1990s instead coincided with decreasing mortality rates. Turning to the self- reported health indicators in figure 2.18, there is a resemblance between the development of impaired work capacity and disability insurance prevalence. The two series shared a similar pattern from the mid- 1980s onward for the two younger age- groups, and from the early 1990s onward for the oldest age- group. Both the prevalence of DI recipiency and the share of the population with an impaired work capacity increased more rapidly for women than for men in the two youngest age- groups. In addition,

Fig. 2.18

Disability insurance prevalence, mortality and health indicators

Sources: Statistics Sweden; Swedish Social Insurance Agency.

110

Lisa Jönsson, Mårten Palme, and Ingemar Svensson

the sharp drop in DI recipency among men aged sixty to sixty- four since the mid- 1990s has coincided with a decrease in the share of the population with an impaired work capacity. The same correspondence is not present between the development of the share of the population in poor health and disability insurance recipiency or the share of the population with an impaired ability to move and disability insurance recipiency. Figure 2.19 presents the development of disability insurance prevalence along with three additional health indicators: the share of the population who visited a doctor during the last three months, the share of the population with a long- term disease, and the share of the population with selfassessed good health. From the development of these indicators it is not possible to reject that disability insurance recipiency and population health are unrelated. The indicators did, however, develop more adversely for women and for younger age- groups, which was also the case for disability insurance recipiency. 2.5.2

Disability Insurance Incidence and Population Health

To further explore the relation between population health and disability insurance recipiency, figures 2.20, 2.21, and 2.22 show the development of three diagnosis- specific health measures: (a) the share of the population receiving inpatient care for a specific diagnosis, (b) the corresponding self- reported prevalence of a specific disease from the Survey on Living Conditions (ULF), and (c) the diagnosis- specific mortality rate, along with diagnosis- specific incidence of DI recipiency. The figures present this information for the three most common diagnoses for DI eligibility: circulatory diseases, musculoskeletal diseases, and mental disorders. We show the development from 1971 to 2005 for men and women in the age- groups fifty- five to fifty- nine and sixty to sixty- four. Figure 2.20 shows the development for circulatory diagnoses. There has been a decreasing importance of circulatory diagnoses as a ground for disability insurance recipiency in all demographic groups over time. This decrease has been shared with the diagnosis- specific mortality rate. Mortality decreased later for men than for women, and for the younger age- group, aged fifty- five to fifty- nine, the decrease in DI incidence showed a similar pattern. For the older age- group, however, the decrease in DI incidence occurred simultaneously across demographic groups, although mortality decreased later for men. The decreasing trend is not as apparent for the utilization of inpatient care or the self- reported prevalence of circulatory diagnoses. For the older age- group, aged sixty to sixty- four, the development of the utilization of inpatient care showed no resemblance with the development of DI incidence. For the younger age- group, aged fifty- five to fifty- nine, however, there was a similarity between the utilization of inpatient care and DI incidence from the mid- 1980s onward. For men, the development of the self- reported

Fig. 2.19

Disability insurance prevalence, mortality and health indicators

Sources: Statistics Sweden; Swedish Social Insurance Agency.

112

Lisa Jönsson, Mårten Palme, and Ingemar Svensson

Fig. 2.20

Circulatory diagnoses by gender and age-group, 1971 to 2005

Note: *ULF/10 = Share of the population reporting the particular disease in The Survey on Living Conditions (ULF), divided by ten.

prevalence of circulatory diseases showed no correspondence with DI incidence. For women, however, there was a resemblance between the two series. Figure 2.21 shows the development for musculoskeletal diagnoses. There was a clear peak in the importance of musculoskeletal diagnoses for DI incidence from the mid- 1980s until the mid- 1990s in all demographic groups. Interestingly, this peak is also visible in the self- reported prevalence of musculoskeletal diseases. A smaller upturn in DI incidence can be seen in the early 2000s. Also this pattern is visible in the self- reported disease prevalence, except for men aged sixty to sixty- four. The development of inpatient care utilization due to musculoskeletal diagnoses has been relatively stable over time, although a slight increase is visible as DI incidence increased in the

Disability Insurance, Population Health, and Employment in Sweden

Fig. 2.21

113

Musculoskeletal diagnoses by gender and age-group, 1971 to 2005

Note: *ULF/10 = Share of the population reporting the particular disease in The Survey on Living Conditions (ULF), divided by ten.

late 1980s. The mortality rate in musculoskeletal diagnoses has been highly volatile but has followed a decreasing trend that did not correspond to the development of the diagnosis- specific DI incidence. Finally, figure 2.22 shows the development for mental disorders. The importance of mental disorders for DI incidence has been invariant or falling in all demographic groups until the late 1990s. This was followed by an enormous increase in the importance of mental disorders for granting disability benefits to women and a slight upturn for men. This increase is well reflected in the self- reported disease prevalence, except for men aged sixty to sixty- four. Also, the stable or falling trend in DI incidence before the 2000s corresponded to the development of self- reported disease prevalence, except for women aged sixty to sixty- four. The utilization of inpatient care due to mental disorders shows no correspondence with diagnosis- specific DI incidence. Inpatient care utilization

114

Lisa Jönsson, Mårten Palme, and Ingemar Svensson

Fig. 2.22

Mental disorders by gender and age-group, 1971 to 2005

Note: *ULF/10 = Share of the population reporting the particular disease in The Survey on Living Conditions (ULF), divided by ten.

increased dramatically in the early 1970s and was then consistently falling over time. Such a consistent development might be due to changing working procedures in the health care system with this type of patient rather than an underlying trend in health. The mortality rate also shows no correspondence with the development of DI incidence. It should be noted, however, that the mortality rates in musculoskeletal diseases and mental disorders are extremely low and, hence, very dependent on diagnosing patterns. 2.5.3

The Relative Health of Disability Insurance Recipients to That of Nonrecipients

An alternative way of examining the role of health for the development of disability insurance recipiency is to compare the health of disability insurance recipients with the health of nonrecipients. With a fixed health

Disability Insurance, Population Health, and Employment in Sweden

115

threshold for DI recipiency, we would expect the relative health of recipients relative to nonrecipients to remain constant even if the prevalence of DI recipiency changes. If changes in the prevalence of DI recipiency are instead induced by, for example, economic incentives, less stringent health requirements for eligibility or a change in demand for workers with health- induced work limitations, we would expect the health of DI recipients relative to the health of nonrecipients to improve as recipiency increases. We divided the respondents in the Survey on Living Conditions into disability insurance recipients and nonrecipients. Figure 2.23 shows the development of the relative health of DI recipients relative to nonrecipients for nine self- reported health indicators from the survey along with the prevalence of DI recipiency. For each indicator, the relative measure shows the prevalence of a particular condition in the DI population as a fraction of the prevalence in the non-DI population. For example, the upper left- hand panel in figure 2.23 shows that in 1976, a ten times larger proportion of those receiving disability benefits reported an impaired work capacity as compared to non-DI recipients. Due to sample size restrictions, the results are presented for the entire age- group of forty- five to sixty- four- year- olds, and the presented series are three- year moving averages. The left- hand panels show the development for men and the right- hand panels for women. Figure 2.23 reveals much volatility, but no obvious trend, for the health of DI recipients relative to nonrecipients before 1995. For men, most indicators also remain constant after 1995, although the share of men in poor health, the share with an impaired ability to move, and the share who recently visited a doctor increased in the 2000s. This suggests, if anything, a worsened health of male DI recipients relative to nonrecipients in recent years. For women, there is an opposite trend as that for men from 1995 to 2005. During this period, disability insurance recipiency for women increased substantially. The health indicators in panel B, showing the prevalence of impaired work capacity, impaired ability to move and poor health, show a downward trend since 1995. Also, the prevalence of long- term disease in panel D has been falling, while the share of women with self- reported good health has increased, for DI recipients relative to nonrecipients. The share of women who visited a doctor, reported in panel F, fell at least initially as DI caseloads took off. This suggests a relative improvement in the health of female DI recipients as compared to nonrecipients between 1995 and 2005. 2.5.4

Conclusions about Health and the Disability Insurance

The analysis in this section showed that there is no relation between mortality and disability insurance recipiency in general in any age and gender group. This result does not, however, preclude that there is an underlying relation between population health and DI recipiency. As discussed in section 2.3, mortality might be a poor measure of the aspects of population health that are important for DI recipiency.

Fig. 2.23 to 2005

Relative health of DI recipients compared to non-DI recipients, 1975

Sources: Survey on Living Conditions and the Swedish Social Insurance Agency.

Disability Insurance, Population Health, and Employment in Sweden

117

For the self- reported health indicators, there are some results indicating that population health indeed is an important determinant of disability insurance utilization. In particular, the share of the population with an impaired work capacity developed in a similar manner as the prevalence of DI recipiency. The health in younger age- groups has declined compared to older groups, and the health of women has declined compared to men. This pattern is consistent with the development of disability insurance recipiency. From the analysis of the diagnosis- specific health indicators, we also saw a corresponding pattern between disability insurance incidence and the development of self- reported diagnosis- specific diseases. The relation between health and disability insurance seems to be strongest for the younger age- groups. For the age- group sixty to sixty- four, a potential relation between health and DI recipiency can be seen only from 1990 onward. For men in this age- group, the drop in DI recipiency during the last decade coincided with a drop in the share of the population with an impaired work capacity. For women in this age- group, however, the development of disability insurance recipiency during the last decade seems to be less health related than for men. As DI recipiency increased, the health of disability insurance recipients relative to nonrecipients improved. This implies that relatively healthier women than before started to receive disability benefits. 2.6

Changes in Disability Insurance Eligibility

The changes in the design of the disability insurance program in Sweden were described in section 2.2 of this chapter. Two major reforms in the history of the program were directed toward older workers only. First, special eligibility rules for older workers were introduced in 1970 and abolished in 1997. Second, a possibility of granting disability benefits based on pure labor market reasons for older workers was introduced in 1972 and abolished in 1991. The age limits were initially set to ages sixty- three to sixty- six, but was changed to sixty to sixty- four in 1974 for pure labor market reasons and in 1976 for the special eligibility rules. The fact that the changes in eligibility affected a limited group only makes the implementation of these rules favorable from an evaluation point of view. It enables us to analyze the effect of changes in eligibility in a demographically defined group and use the younger age- group aged fifty- five to fifty- nine as an unaffected control group. In this section, we analyze the introduction of the special eligibility rules in 1970 and the pure labor market reasons in 1972 and the subsequent abolitions in 1997 and 1991. We analyze the impact of the eligibility reforms on disability insurance recipiency and study to what extent the reforms also affected labor market outcomes. As described in section 2.2, another set of eligibility rules were also in effect between 1970 and 1997. These rules affected all workers, and implied

118

Lisa Jönsson, Mårten Palme, and Ingemar Svensson

that long- term unemployed with functional limitations were made eligible for disability benefits after having been unemployed for one to two years. Since these rules affected all workers, there is no control group to use in order to distinguish the effect of the rules from general time trends. When studying the effect of the special eligibility rules that were in effect during the same period, we implicitly assume that the unemployment as an additional criterion for disability insurance eligibility affected the age- groups fifty- five to fifty- nine and sixty to sixty- four equally. If the unemployment criterion in fact was more important in the older age- group than in the younger, the effect of these rules will be subsumed in the effect of the special eligibility rules for older workers. As was also described in section 2.2, eligibility for disability insurance recipiency was recently changed again. Since July 1, 2008, an individual’s working capacity has to be permanently reduced in relation to the entire labor market in order to qualify for disability benefits. Since this change affected all disability insurance applicants simultaneously, we cannot evaluate the impact of these changes in a control group setup. Figure 2.9 showed a substantial decrease in the incidence of disability insurance recipiency during the last years, and the decrease is particularly steep since 2008. 2.6.1

Program Eligibility and Disability Insurance Recipiency

The upper panels in figure 2.24 show the development of the prevalence of disability insurance recipiency for men and women, respectively. Vertical lines mark the introduction of the special eligibility rules for elderly workers in 1970 and the abolition of these rules in 1997, as well as the introduction of labor market reasons in 1972 and the subsequent abolition in 1991. The lower panels in figure 2.24 show the differences in disability insurance recipiency between the group aged sixty to sixty- four and the younger group aged fifty- five to fifty- nine. Figure 2.25 presents similar panels for the incidence of disability insurance recipiency, that is, the admitted disability insurance recipients as a share of the risk population in each age- group. The upper panels in figure 2.24 show a clear increase in the growth rate of DI recipiency after the 1970 reform. The increase in the growth rate is particularly large for the oldest age- group. The lower panels show that the difference in DI prevalence between age- groups sixty to sixty- four and fifty- five to fifty- nine is fairly constant before 1970, at least for women, but increases rapidly after 1970. This indicates that there was an effect of the special eligibility rules for older workers that were introduced in 1970. Unfortunately, we do not have data on the incidence of disability insurance recipiency in these age- groups before 1971. The number of entrants into DI in all ages, however, almost doubled from around 23,000 in 1968 to around 44,000 in 1970. The next reform is the introduction of pure labor market reasons for older workers in 1972. It is not possible to perceive any effect of this reform

Disability Insurance, Population Health, and Employment in Sweden

119

Fig. 2.24 Prevalence of disability insurance utilization and the timing of DI reforms Source: Swedish Social Insurance Agency.

on the growth rate of DI prevalence in figure 2.24. From panels E and F in figure 2.10, however, we know that pure labor market reasons accounted for an entry rate of around 1 percent of the risk population from its introduction until 1983. The lower panels in figure 2.25 show the difference in the incidence of disability insurance recipiency between the age- groups sixty to sixty- four and fifty- five to fifty- nine. The difference is slightly larger in 1972 and 1973 than in 1971, which might indicate a small immediate effect of the 1972 reform on total DI entry rates. As discussed in section 2.4, there was a clear trend break in disability insurance recipiency in the older age- groups in the early 1990s. This coincided with the abolishment of pure labor market reasons in 1991 and the rehabilitation reform in 1992. It also coincided, however, with a deep recession in Sweden. The upper panels of figure 2.24 show this trend break in

120

Lisa Jönsson, Mårten Palme, and Ingemar Svensson

Fig. 2.25

Incidence of disability insurance utilization and the timing of DI reforms

Source: Swedish Social Insurance Agency.

the prevalence of disability insurance recipiency in age- groups fifty- five to fifty- nine and sixty to sixty- four. The increase in 1992 and 1993 is due to the fact that a large number of recipients of sickness benefits were transferred to the disability insurance program as a consequence of a rehabilitation reform affecting the work of the social insurance agencies. The lower panels in figure 2.25 show that the difference in disability insurance entry rates between age- groups sixty to sixty- four and fifty- five to fifty- nine was substantially lower after the 1991 reform than before. Hence, the abolishment of the pure labor market reasons for those aged sixty to sixty- four in 1991 seems to have had an effect on disability insurance recipiency in the affected age- group. The effect was larger for men than for women. Entry rates into disability insurance were higher for men before the reform, but of the same magnitude as for women after the reform.

Disability Insurance, Population Health, and Employment in Sweden

121

The abolishment of special eligibility rules for older workers in 1997 has been thoroughly analyzed in Karlström, Palme, and Svensson (2008). According to their analysis, there is no significant effect on entry rates into the disability insurance. There is, however, a significant anticipation effect— an increase in entry rates into DI just before the reform—corresponding to almost 2 percent of the labor force in ages sixty to sixty- four. Since the new eligibility rules were announced long before they were implemented, workers who believed they would pass the prereform eligibility rules, but not the postreform ones, could apply under the prereform regime. This effect is seen in panels C and D in figure 2.17 from the increase in the difference in DI entry rates during 1996 and 1997. 2.6.2

Program Eligibility and Labor Market Outcomes

The eligibility reforms for older workers seem to have had an effect on the utilization of the disability insurance. An extended question is to what extent these effects were translated into effects on employment and labor force participation rates. Figure 2.26 shows the development of disability insurance prevalence, non- labor force participation and nonemployment for men and women aged fifty- five to fifty- nine and sixty to sixty- four. Figure 2.27 shows the difference in non- labor force participation and nonemployment rates between the age- groups sixty to sixty- four and fifty- five to fifty- nine, along with the corresponding difference for the prevalence and incidence of disability insurance recipiency. The reforms under study are marked with vertical lines. The left- hand panels in figure 2.26 show that the change in the prevalence of disability insurance recipiency after the reform in 1970 was indeed translated into a correspondingly large increase in nonemployment and non- labor force participation for the male population in both age- groups. Figure 2.27 shows that the differences in non- labor force participation and nonemployment between the age- groups sixty to sixty- four and fifty- five to fifty- nine increased in the same manner as disability insurance recipiency during the 1970s. The pure labor market reasons, introduced in 1972, were not extensively used until the mid- 1980s. When they were used, however, we do see an increase in nonemployment and non- labor force participation that suggests a continuously close relationship between the prevalence of disability insurance recipiency and labor market outcomes also in the 1980s. For the 1970s and 1980s, the utilization of the disability insurance program seems to have translated into effects on nonemployment and non- labor force participation rates. The 1991 reform seems to have had a very different effect. While there was a rapid decrease in disability insurance recipiency, both nonemployment and non- labor force participation increased in the age- group sixty to sixty- four. In the age- group fifty- five to fifty- nine, the decrease in disability insurance recipiency coincided with an increase in nonemployment, but not in non-

122

Lisa Jönsson, Mårten Palme, and Ingemar Svensson

Fig. 2.26

DI prevalence, non-labor force participation and nonemployment

labor force participation. The background to this result might be the deep economic recession in the early 1990s that resulted in a sharp decrease in labor demand. As concluded in section 2.4, disability benefits were replaced by income from unemployment benefits, occupational pension, and sickness benefits. Contrary to the 1991 reform, figure 2.27 suggests that the abolition of the special eligibility rules for workers aged sixty to sixty- four in 1997 was followed by increased employment and labor force participation. A detailed analysis of the effects of this reform on employment and on the utilization of the sickness and unemployment insurance programs is found in Karlström, Palme, and Svensson (2008). They did not, however, find a significant effect of the reform on employment, but did find an effect on both entry and persistence in the unemployment and sickness insurance programs (not considered in figures 2.26 and 2.27). Their conclusion is that the other income security

Disability Insurance, Population Health, and Employment in Sweden

123

Fig. 2.27 Differences between age-groups sixty to sixty-four and fifty-five to fifty-nine

programs worked like communicating vessels that crowded out the employment effect of the stricter eligibility rules enacted in the 1997 reform. Looking closer at figure 2.27, the decrease in nonemployment and non- labor force participation did not come until a few years after this reform. It is therefore difficult to attribute the drop to the reform itself. 2.6.3

Conclusions about Program Eligibility, Disability Insurance, and Labor Market Outcomes

Our analysis of the changes in the eligibility rules shows that the introduction of special eligibility rules for older workers in 1970 seems to have had an effect on the utilization of the disability insurance, and that the effect translated into effects on labor force participation and employment. We did not find support for an additional effect of the introduction of pure labor market reasons for older workers in 1972 on disability insurance recipiency. The pure labor market reasons were not being used extensively until the early and mid- 1980s. At that time, however, there is also an increase in nonemployment and non- labor force participation. For the 1990 and 1997 reforms, the analysis shows that the marked change in utilization of the disability insurance was crowded out by changes in the utilization of other income security programs. However, the long- term relative increase in employment and labor force participation of the age- group

124

Lisa Jönsson, Mårten Palme, and Ingemar Svensson

sixty to sixty- four among both males and females after 1997 suggests that there was an effect of the 1997 reform in prolonging the time before the permanent exit from the labor market of older workers. 2.7

Overall Conclusions

In this chapter, we posed three main research questions. The first question was whether the development of disability insurance recipiency over the past decades can be explained by changes in the health status of the population. We found some support for this hypothesis. We focused on the development for men and women in the age- groups forty- five to fifty- four, fifty- five to fifty- nine, and sixty to sixty- four. The analysis showed that the demographic groups with the least advantageous health development were the same groups with the least advantageous development in disability insurance recipiency. In particular, we found a more adverse development of the health of women compared to men, and the health of younger compared to older. The same pattern can be found in the development of disability insurance recipiency. The relation between population health and disability insurance recipiency was least apparent for the oldest age- group, aged sixty to sixty- four. The second question was whether the changes in disability insurance recipiency can be explained by changes in the eligibility rules in the disability insurance program. We focused on the introduction and abolishment of two sets of eligibility rules that affected the oldest age- group, aged sixty to sixty- four, only. The first were the special eligibility rules for older workers, implying an exemption from rehabilitation and retraining, lowered requirements for the medical assessment of working capacity, and a possibility to consider functional limitations due to normal aging for eligibility to disability benefits. The second was the introduction of pure labor market reasons for older workers, making them eligible for disability benefits if they were still unemployed when reaching the time limit for unemployment benefits. The special eligibility rules for older workers were in effect between 1970 and 1997, while the pure labor market reasons were in effect between 1972 and 1991. For some of the changes in eligibility for older workers, we found evidence of an effect on disability insurance recipiency. The introduction of the special eligibility rules in 1970 seems to have had a large impact on disability insurance recipiency. The introduction of pure labor market reasons in 1972, however, seems to have induced only a small additional increase in the entry rates into the disability insurance program. The abolishment of the pure labor market reasons in 1991 seems to have had a larger effect on disability insurance recipiency. The abolishment of the special eligibility rules for older workers in 1997 did affect disability insurance recipiency, but the effect on employment was crowded out by an increased utilization of the sickness and

Disability Insurance, Population Health, and Employment in Sweden

125

unemployment insurances. In the long run, however, the difference in nonemployment rates between the age- groups sixty to sixty- four and fifty- five to fifty- nine has been decreasing after the reform, which might suggest that the eligibility changes in the disability insurance in 1997 eventually spilled over on employment. The final question was to what extent the changes in eligibility rules for older workers affected employment and labor force participation. The answer is ambiguous. For the 1970 reform, this seems to be true in the sense that the reform opened the disability insurance program as a much more frequent exit route from employment than before. In fact, the disability insurance program became almost the only pathway out of the labor force before the normal retirement age in the decades following the reform. The results are complicated to analyze for the 1991 and 1997 reforms. The 1991 reform coincided with a sharp drop in the employment rate, caused by a labor demand shock from a severe recession. The conclusion for the 1997 reform is that it did not lead to a significant increase in employment. The effect on the disability insurance utilization was crowded out by an increase in the unemployment rate and increased utilization of the sickness insurance. However, several years after the reform, we have seen significant improvements in employment rates among older workers. It is an open question to what extent the new regime within the disability insurance after 1997 contributed to this development. There are several significant changes in the usage of disability benefits that cannot be directly related to either changes in health or reforms of the rules of the program. Throughout the many graphs shown in this chapter we have seen that trends tend to continue, without visible changes in eligibility rules or population health. Possible explanations are (a) changes in the demand for labor with health impairments, (b) formation of norms on elibigility to disability insurance in the social security administration and in the society in general, (c) administrative policies within the social insurance system, or (d) changes in economic incentives the disability insurance program primarily attributed to maturation of the supplementary pension program (the ATP system). The relative importance of these factors is a subject for further research in this area.

References Gruber, Jonathan, and David A. Wise, eds. 2010. Social Security Programs and Retirement around the World: The Relationship to Youth Employment. Chicago: The University of Chicago Press. Hedström, Peter. 1987. “Disability Pension: Welfare or Misfortune?” International Journal of Sociology 87:1261– 84.

126

Lisa Jönsson, Mårten Palme, and Ingemar Svensson

Karlström, Anders, Mårten Palme, and Ingemar Svensson. 2008. “The Employment Effect of Stricter Rules for Eligibility to DI: Evidence from a Natural Experiment in Sweden.” Journal of Public Economics 92:2071– 82. Kruse, Agneta, and Lars Söderström. 1989. “Early Retirement in Sweden.” In Redefining the Process of Retirement, edited by W. Schmäl. Berlin: Springer. Lindbeck, Assar, Mårten Palme, and Mats Persson. 2009. “Social Interaction and Sickness Absence.” Stockholm University, Department of Economics. Working Paper no. 2009:4. Palme, Mårten, and Ingemar Svensson. 1999 “Social Security, Occupational Pensions and Retirement in Sweden.” In Social Security and Retirement around the World, edited by J. Gruber and D. Wise. Chicago: The University of Chicago Press. Palme, Mårten, and Ingemar Svensson. 2004. “Income Security Programs and Retirement in Sweden.” In Social Security Programs and Retirement around the World: Micro-Estimation, edited by J. Gruber and D. Wise. Chicago: The University of Chicago Press. Statistics Sweden. 2009. “Undersökning om levnadsförhållanden 2009.” Available at: http://www.scb.se/Statistik/LE/LE0101/_dokument/LE0101_DO_2009.pdf. Skogman Thoursie, Peter. 1999. “Disability and Work in Sweden.” PhD diss. Swedish Institute for Social Research, no. 39, Stockholm. Wadensjö, Eskil. 1996. “Early Exit from the Swedish Labour Market.” In The Nordic Labour Markets in the 1990’s, edited by E. Wadensjö. Amsterdam: NorthHolland.

3 Health, Disability, and Pathways into Retirement in Spain Pilar García-Gómez, Sergi Jiménez-Martín, and Judit Vall Castelló

3.1

Introduction

Disability has always been an important social welfare program in developed countries and Spain is not an exception. The program represented an average of 1.5 percent of the gross national product (GDP) in the 1995 to 2010 period, slightly below the European Union (EU) average of 2.2 percent of the GDP. The relative importance of the program with respect to the pension program has varied in the last twenty to twenty- five years. In 1977 the ratio of disability to retirement benefits was 0.44, in 1985, by the end of the crisis of the early 1980s, it reached a maximum of 0.58 and it has slowly decreased since then. By 1997 the ratio was again down to 0.45. Figures from 1998 are difficult to compare as all disability pensions from age 65-plus were, since that year, converted to retirement benefits, but back- of-the- envelope calculations suggest that the ratio decreased in the early 2000s and increased from 2008 onward because of the recent crisis.1 Pilar García-Gómez is assistant professor at the Erasmus School of Economics, Erasmus University Rotterdam. Sergi Jiménez-Martín is associate professor of economics at Universitat Pompeu Fabra in Barcelona, director of the Barcelona Microeconometrics Summer School (BMiSS, Barcelona GSE), and chair of LaCaixa-FEDEA Economía de la Salud y Hábitos de Vida at FEDEA. Judit Vall Castelló is a Robert Solow Postdoctoral Fellow at the Center for Research in Health and Economics (CRES) at Universitat Pompeu Fabra. We thank the Ministerio de Ciencia e Innovación for financial support (research projects ECO2008-06395-C05 and ECO2011-30323-C03-02). García-Gómez thankfully acknowledges financial support from the European Union under the Marie Curie Intra-European Fellowships. Vall Castelló thankfully acknowledges financial support from the Centre Cournot for Economic Research in Paris. For acknowledgments, sources of research support, and disclosure of the authors’ material financial relationships, if any, please see http://www.nber.org/chapters/c12384.ack. 1. Own calculations using data from the Muestra Continua de Vidas Laborales (MCVL), a random sample of administrative records provided by the Spanish Social Security Administration.

127

128

Pilar García-Gómez, Sergi Jiménez-Martín, and Judit Vall Castelló

The previous ratio of benefits has a clear business cycle component because the disability program has often been used to cushion recessions and to alleviate transitions into and from economic inactivity in regions of high unemployment (Disney and Webb 1991; McVicar 2006; Benítez-Silva, Disney, and Jiménez-Martín 2010). In addition, disability programs have often been used as an alternative to early retirement through old- age programs, either because of restrictions on early retirement itself, or because disability programs offer a more attractive after- tax benefit level (Gruber and Wise 1999, 2004; Jiménez-Martín, Labeaga, and Vilaplana Prieto 2006; Jiménez-Martin and Vall Castelló 2009). This alternative route into inactivity for older people proved particularly attractive when employers were seeking to shed older workers in periods of recession such as the early 1980s, early 1990s, and late 2000s. It is common in the literature to distinguish between “health disability,” which arises from clearly diagnosed medical conditions, and “work disability,” which may also have its roots in economic and social circumstances. Although there is a link between work and health disability, economic conditions and the variations in the risk of unemployment over time may play an important part in explaining the dynamics of the disability rolls (BenítezSilva, Disney, and Jiménez-Martín 2010). Other factors, including the trend in the relative generosity of disability benefits relative to unemployment or pension benefits (Autor and Duggan 2003, 2006; Burkhauser and Daly 2001; OECD 2006, 2007 more generally), and underlying demographic and morbidity trends, are also relevant. As in many other Organization for Economic Cooperation and Development (OECD) countries, the underlying demographic trends in Spain are very favorable. The increase in life expectancy at birth has also been translated in increases in life expectancy at age sixty- five. In 1960 women aged sixty- five expected to live 15.3 more years, while the expectations were 21.9 in 2008. Similar improvements are also observed among men (from 13.1 in 1960 to 18.0 in 2008) (OECD Health Data 2010). As a consequence, there have been striking improvements in age- specific mortality rates over the last fifty years in Spain across individuals aged over fifty- five. The extent to which these changes translate into higher labor capacity depends on the evolution of ill health. Thus, further reduction in the age- specific mortality rate will only translate in an increase in the population able to work at older ages if the average age of the onset of a work- related disability increases and, simultaneously, the requirement to work for the disabled are adjusted accordingly. When analyzing trends of disability rolls, Social Security reforms should also be taken into account, as they may change the relative balance between the various exit routes into (retirement) benefits. There have been two main social security reforms since 1990. In 1997 there was a reorganization of the disability assessment system, the medical requirements for temporary

Health, Disability, and Pathways into Retirement in Spain

129

disability were tightened, and the generosity of the old- age pension system was reduced, while in 2002 more incentives to retire later were introduced. The main purpose of this chapter is to analyze the trends in labor force participation and transitions to benefit programs of older workers in relation to health trends. In particular, we explore questions such as: Do mortality improvements at older ages translate into more participation? Do we reach the same conclusions with other health variables? Do health improvements reduce the prevalence of disability among older workers? Have recent social security reforms favored the participation of older workers, and in particular, of disabled workers? Have they affected transitions from employment to unemployment or disability at older ages? Do we find any substitution across programs? The link between program participation and health will be analyzed descriptively. Alternatively, the effect of reforms on program participation will be evaluated using administrative data on the stock and inflow into unemployment, disability, and retirement for individuals approaching the normal retirement age. We complement the analysis using data from the Spanish Labor Force Survey. In order to identify the key parameters of the model we aggregate data by gender and age group (fifty to fifty- four, fifty- five to fifty- nine, sixty to sixty- four) using two levels of regional aggregation. Our preliminary conclusions are pessimistic regarding the effect of health improvements on the labor market attachment of older workers. We show that despite the large improvements in mortality rates among older individuals in Spain, the employment rates of individuals older than fifty- five remain lower than the ones observed in the late 1970s. On the other hand, decreases in mortality rates do not necessarily go hand in hand with improvements in population health. We do not find any conclusive evidence on whether morbidity has improved or worsened during the period of analysis. Regarding the effect of social security reforms we find that both the 1997 and the 2002 reforms decreased the stock into old- age benefits at the cost of an increased share into disability. More interestingly, the magnitude of the two opposite effects is the same, suggesting a clear substitution effect among these two programs in the older age- groups. Finally, we find that none of these two reforms had any effect on the share of these age- groups into unemployment, which is highly explained by the total share of the population out of employment. Regarding the effects of these two reforms on the outflows from employment, we find that there was a significant increase in the outflow from employment into disability after the 2002 reform. The rest of the chapter proceeds as follows. In section 3.2 we present the institutional setting and discuss disability insurance and pension reforms. In section 3.3 we review historical data on mortality, health status, and labor force participation and compare their trends during the last thirty years. We analyze the effect of program reforms on disability rolls and the substitution among the different programs in section 3.4. Finally, section 3.5 concludes.

130

3.2

Pilar García-Gómez, Sergi Jiménez-Martín, and Judit Vall Castelló

Disability Insurance (DI) and Social Security Reforms

The aim of this chapter is to provide descriptive evidence on the relationship between past trends on health status and labor market participation at older ages and the role of social security reforms. Therefore, we first describe the disability system in Spain, as it is the pathway out of employment more closely linked to the individual’s health. However, the transition into a program will also depend on the availability and characteristics of the other programs. Therefore, we also highlight the main changes in the unemployment and old- age systems. In all three cases, social security is responsible for the payment of contributory benefits (old- age, disability, unemployment, temporary sick leave, maternity leave, and survivor’s benefits), while non- contributory benefits are managed by the regional authorities of each Autonomous Community and the IMSERSO (the Institute of Aged People and Social Services) in Ceuta and Melilla. 3.2.1

The Disability Insurance System

In Spain, there are two types of permanent disability benefits: (a) contributory, which are given to individuals who have generally contributed to the social security system before the onset of the disabling condition, and (b) non- contributory, which are given to individuals who are assessed to be disabled but have never contributed to the social security system (or do not reach the minimum contributory requirement to access the contributory system). Non- contributory disability benefits are means- tested and managed at the regional level.2 The size of the non- contributory system is relatively small compared to the contributory system (197,126 individuals received non- contributory disability benefits in 2009, while 920,860 received contributory benefits during the same year). The amount of benefits received is also smaller in the non- contributory case (the average non- contributory pension is 417.09 euros per month compared to an average contributory disability pension of 831.49 euros per month). For these reasons, in the remainder of this chapter, we put more emphasis on the permanent contributory disability system in Spain. Social security defines the permanent contributive disability insurance as the economic benefits to compensate the individual for losing a certain amount of wage or professional earnings when affected by a permanent reduction or complete loss of his or her working ability due to the effects of a pathologic or a traumatic process derived from an illness or an accident. In order to capture the different situations a person can be in after suffering from a disabling condition, the Spanish Social Security Administration uses 2. Income is evaluated yearly. The income threshold in 2010 was set at 4,755.80 euros per year for an individual living alone. This amount is adjusted if the individual lives with other members.

Health, Disability, and Pathways into Retirement in Spain

131

a classification of four main degrees of disability that depend on the working capacity lost: 1. Permanent limited disability for the usual job: the individual loses at least 33 percent of the standard performance for his or her usual job, but the individual is still able to develop the fundamental tasks of his or her usual job or professional activity. Individuals in this level of disability only receive a one- time lump sum payment. 2. Partial disability: the individual is impaired from developing all or the fundamental tasks of his or her usual job or professional activity, but he or she is still capable of developing a different job or professional activity. 3. Total disability: the individual is impaired from the development of any kind of job or professional activity. 4. Severe Disability: Individuals who, as a result of anatomic or functional losses, need the assistance of a third person to develop essential activities of daily living such as eating, moving, and so forth. Figure 3.1 shows the distribution of individuals receiving disability benefits by degree of disability and region in 2009. Note that in all Spanish regions, the percentage of individuals receiving permanent limited disability was rather small. In fact, at a national level, only 917 out of the 920,863 contributory permanent disability pensions were for individuals classified as permanent limited disabled for the usual job in 2009. The first dimension that can be highlighted from figure 3.1 is the strong regional variation, not only in the percentage of the working- age population receiving disability benefits, but also on its distribution across types. Asturias is the region with a higher share of the working- age population receiving contributory disability benefits (4.9 percent) followed by Galicia (3.8 percent), Andalucía (3.7 percent), and Cantabria (3.6 percent). On the other end, Madrid is the region with the lowest percentage of recipients of contributory disability benefits (1.6 percent). As can be observed in the figure, Ceuta and Melilla concentrate the largest share of the population receiving non- contributory disability benefits in Spain. Therefore, the share of the population receiving non- contributory benefits in the country as a whole is smaller when we exclude the special autonomous regions of Ceuta and Melilla. Among the other regions, we find the highest percentages in Canarias, Galicia, and Extremadura. Eligibility and Pension Amount The eligibility requirements and the pension amount depend on the source of the disability (ordinary illness, work related, or unrelated accident or occupational illness), the level of the disability, and the age of the onset of the disability. Table 3.1 summarizes the main parameters of both the eligibility criteria and the pension formula. The two main features to highlight are: (a) there are no contributory requirements if the health impairment is due

132

Pilar García-Gómez, Sergi Jiménez-Martín, and Judit Vall Castelló

Fig. 3.1 Percentage of individuals receiving disability benefits, by region and degree of disability in 2009 Source: Own elaboration using data on disability pensions from the Ministry of Work and Immigration (www.mtin.es) and population from the Spanish Institute of Statistics (www .ine.es). Note: The percentages have been computed as (total number of disability beneficiaries/population aged sixteen to sixty- four) * 100. The percentage receiving permanent limited disability benefits has been excluded as it is smaller than 0.01 percent in all regions.

to either an accident or an occupational illness, and (b) individuals older than fifty- five with a partial disability receive a higher replacement rate if it is considered difficult for them to find a job due to lack of education or the social and labor market conditions of the region where they live. The total amount of the pension is obtained by multiplying a percentage, which varies depending on the type of pension and the degree of disability (as shown in the last rows of table 3.1) to the regulatory base, which depends on the source of the disability and on previous salaries.3 The number of years included in the regulatory base depends on the source of the disability. The income tax rules differ across disability types. Partial disability benefits are taxable under the general income tax rules, while total disability pen3. Benefit = Regulatory Base * Percentage.

Health, Disability, and Pathways into Retirement in Spain Table 3.1

133

Summary of the parameters to calculate permanent disability pensions Work-unrelated accident

Ordinary illness Age > = 31: Contributed 1/4 time between 20 years old and disabling condition. Minimum of 5 years Age < 30: Contributed 1/3 time between 16 years old and disabling condition. No minimum number of years required Average wage last 8 years of work

Eligibility No minimum contributory period required

Regulatory base Average annual wage of 24 months within the last 7 years of work

Work-related accident or professional illness

No minimum contributory period required

Average wage last year of work

Percentage applied to the regulatory base Partial Disability: 55% Individuals older than 55 with difficulties to find a job due to lack of education or characteristics of the social and labor market of the region where they live: 75% Total Disability: 100% Severe Disability: 100% + 50%

sions are always exempted from income taxes. Furthermore, if the individual works while receiving the pension, there is a reduction in the earnings used to calculate the income tax of 2,800 euros per year if their degree of disability is low (between 33 percent and 65 percent), and of 6,200 euros per year if the disability level is higher (more than 65 percent), or if the disabled has reduced mobility. In addition, individuals receiving partial disability benefits can combine the benefits with earnings from work, as long as the type of job is compatible with his or her disability. In general, to be granted a permanent disability benefit, the individual must come from a situation of sick leave (also called temporary disability/ incapacity) and be observed as still presenting anatomic or functional reductions that decrease or cancel his or her capacity to work after following the prescribed medical treatment. The application can be started by the provincial office of the National Institute of Social Security (NISS), by the institutions that collaborate in the process (such as hospitals), or by the individual himself (in which case, more documentation is required). The Disabilities Evaluation Team evaluates the medical report and the professional background of the applicant and, on the basis of this analysis, the directors of the provincial office of the NISS decide on the type of disability pension granted (if any), the benefit level, and the date of the next medical check-up.

134

Pilar García-Gómez, Sergi Jiménez-Martín, and Judit Vall Castelló

All permanent disability pensions are automatically converted to old- age pensions once the individual turns sixty- five.4 Sick Leave or Temporary Disability Sick leave benefits are available to all workers who have contributed for at least 180 days during the five years prior to the onset of the illness in the case of common illness. If the origin of the sick leave is an accident (whether or not working accident) or an occupational illness, no minimum contributory period is required. The amount of benefits also depends on the source of the disability. Individuals who have a disability because of a working accident or an occupational illness are entitled to receive 75 percent of the basic salary (including overtime pay) from the first day of leave. In the other cases, there is a waiting period of three days without benefits unless it is covered by a collective agreement and, from the fourth day until the twentieth, the employee receives 60 percent of the basic salary. After the twenty- first day, the compensation represents 75 percent of the basic salary. Social security is responsible for the payment from the first day in the case of a working accident or occupational illness and from day sixteen otherwise. In this last case, the employer pays from the fourth until the fifteenth day. It is not possible to combine sickness benefits with any kind of paid work, not even part- time work. The duration of the benefits is for a maximum of twelve months with a potential extension of an additional six months when it is foreseeable that the beneficiary will become capable of working within this additional period of time. At the end of this period the individual is either considered nondisabled or can apply for permanent disability benefits. Certification and monitoring of sick leave is ensured by a doctor (GP) of the Public Health Services (or from a doctor from a Mutual Work Fund). Major Health Conditions of Disability Benefits Recipients Figure 3.2 shows the percentage distribution of total contributory and non- contributory disability recipients by health conditions, distinguishing among mental, musculoskeletal, and other health problems by three agegroups in 2004. First of all, it is important to note the importance of the musculoskeletal conditions because 50.3 percent of all the individuals receiving a disability benefit in Spain do so on the basis of a musculoskeletal health problem. This condition is more prevalent among the youngest group, as it represents 59.4 percent of the individuals aged twenty to thirty- four receiving disability benefits. Its importance slightly decreases with age (54.0 per-

4. Most of the outflows from the permanent disability system are due to death or automatic transfer to old- age pensions. Around 4 percent of the outflows are due to improvement of the health condition and 2.7 percent to a judicial process. Monthly outflows in 2010 were around 2,500 to 3,000.

Health, Disability, and Pathways into Retirement in Spain

135

Fig. 3.2 Major health conditions of disability benefit recipients in Spain, 2004 (percentage distribution of total benefit recipients by age-group) Source: Data from OECD (2007). Data refer only to a sample of people officially diagnosed with a disability. These data come from special tabulations provided by the University of Madrid, based on linked NISS and IMSERSO data.

cent of disability pensioners aged thirty- five to forty- nine and 48.2 percent of disability pensioners aged fifty to sixty- four). The figure also shows that 9.7 percent of all the individuals in the disability rolls are diagnosed with a mental condition, the prevalence being higher among the youngest group. The share of claimants due to mental health conditions is among the smallest in Europe. For example, the share of inflows into disability due to mental health diseases was 34.3 percent in the United Kingdom in 2006, 41 percent in Switzerland in 2004, 43.4 percent in Denmark in 2005, and 25.4 percent in Norway in 2004 (OECD 2008). The share of other health conditions varies across age- groups, as it represents from 26.9 percent of the pensioners aged twenty to thirty- four up to 43.3 percent of the pensioners aged fifty to sixty- four. Unfortunately, there is no evidence available of the distribution of the other illnesses. In order to shed some light on its likely distribution, figure 3.3 shows both the percentage of total cases and total days of sickness leave in Spain in 2005. The biggest category within the specific illnesses is diseases of the musculoskeletal system and connective tissue, which corresponds to the main category among the recipients of disability benefits. Next, diseases of the respiratory system have the second highest share among the number of cases, while mental health problems have the second highest share in the number of days. Surprisingly, the share associated with circulatory problems both in days and cases is smaller, and similar to the share of diseases of the digestive and the nervous systems.

Distribution by health problems of temporary disability or sickness leave in Spain in 2004

Source: Own elaboration using data from Oliva (2010).

Fig. 3.3

Health, Disability, and Pathways into Retirement in Spain

137

Reforms of the Disability Insurance System Permanent disability benefits were used extensively as an early retirement mechanism for workers in restructuring industries (such as shipbuilding, steel, mining, etc.) or as substitution for long- term unemployment subsidies in depressed regions during the late 1970s and 1980s (OECD 2001), which resulted in an increase in the inflows into the disability system and permanent disability benefits. These events prompted a number of reforms introduced during the second half of 1980 and the beginning of 1990 that aimed at reversing these trends. The main objective of these reforms was to abolish the incentive effects to permanently leave the labor market before reaching the legal retirement age through the disability system. Here we focus on some distinctive features of the main reforms since the creation of the National Institute of Social Security in 1979, while we refer the reader to table 3.2 for a summary of all the reforms in the disability system in Spain during this period. The biggest reform of the disability system took place in 1997 and it included four main points: 1. Sickness benefits: stricter control of the sickness status by doctors of the social security system, reduction of the level of long- term sickness benefits, replacement of the old own job assessment by a more objective definition of the usual occupation of the individual. Table 3.2 1985 1990

1997

1998 2004–2005 2007

Main reforms of the disability system in Spain The terms of eligibility for disability pensions are tightened. Introduction of a means-tested non-contributory disability pensions for people aged 65+ and for disabled people aged 18+ who satisfy residency requirements. Stricter control of sickness status, reduction of long-term sickness benefit level, usual occupation replaces own job assessment. Permanent disability pensions individuals 65+ are converted to old-age pensions. New INSS disability assessment team to assess permanent disability instead of the GP. Entitlement to non-contributory benefits is not lost if working, and can be collected if losing the job. Possibility for doctors from INSS and mutual insurance companies to review health situation of beneficiaries. Improve monitoring and control of sickness leave with new INSS tool. Possibility to combine non-contributory disability with some earnings. Minimum contributory period to access permanent disability is reduced for young workers. The formula to calculate the regulatory base of the benefit gets closer to the formula for old-age pensions.

138

Pilar García-Gómez, Sergi Jiménez-Martín, and Judit Vall Castelló

2. The permanent disability pensions of individuals aged at least sixtyfive are automatically transferred to the old- age pension system. This is just a change in the classification within the pension system. 3. Organizational reform, as all the permanent disability matters are transferred to the NISS. The permanent disability status was in the past assessed and granted by local GPs and this reform created a group of experts (the disability assessment team inside the NISS) that was in charge of assessing the person’s ability to work on the basis of the available medical files and a special medical assessment done by one of the NISS doctors. 4. The individual does not lose entitlement to non- contributory disability benefits if he or she starts working. He or she will then still be entitled to receive non- contributory disability benefits if he or she loses his or her job. Apart from this major reform in 1997, the 1998 budget law introduced the possibility for doctors from the NISS and mutual insurance companies to review the health situation and status of beneficiaries. However, in reality, very few individuals in the permanent disability system do effectively lose their benefits. In 2004 and 2005 monitoring the use of sick leave was tightened with the creation of a new subdepartment at the NISS and a new monitoring tool to reduce absence rates. In 2005, a general absence control was put in place for cases in which the absenteeism took longer than six months. Finally, in 2007 the minimum contributory period to access permanent disability pensions was reduced for young workers in order to adjust for the current later entrance into the job market of younger workers. At the same time, the formula to calculate the regulatory base of the benefit was slightly modified: the regulatory base of permanent disability due to a common illness since then decreased by 50 percent if the individual had not contributed at least fifteen years, and it is lower the further the individual is from age sixty- five. All these reforms have ensured the financial stability of the disability system in Spain as inflow rates have remained at stable levels and have not experienced any dramatic increases like in other countries. 3.2.2

Reforms in Other Social Security Programs in Spain

The extent to which reforms in the disability system are able to decrease the outflows from employment at older ages will depend on the evolution of other programs that can be used as alternative early retirement routes. Therefore, in this section we summarize other important reforms that have taken place in other social security programs in Spain. In particular, we focus on reforms in the unemployment and old- age systems. Table 3.3 provides a chronological summary of these reforms.5 5. A detailed exposition of the changes in the old- age pension system in Spain is provided in Boldrin, García-Gómez, and Jiménez-Martín (2010).

Health, Disability, and Pathways into Retirement in Spain Table 3.3

1984

1985

1989 1997

2001

2002

2007

139

Main reforms since 1980 of the old-age and unemployment systems in Spain Introduction of temporary contracts. Introduction of unemployment assistance (UA) benefits (non-contributory). Special provision for workers aged 55+; they can receive UA until retirement if comply with requirements to get old-age pension (except age requirement). Increased the minimum mandatory annual contributions from 8 to 15. The number of contributive years used to compute the pension increases from 2 to 8. Several early retirement schemes are introduced; partial retirement and special retirement at age 64. Special scheme of UA (permanent until retirement) extended to workers 52+. The number of contributive years used to compute the pension increases from 8 to 15 (progressively by 2001). The formula for the replacement rate is made less generous. The 8% penalty applied to early retirees between the ages of 60 and 65 is reduced to 7% for individuals with 40 or more contributory years. Introduction of a new permanent contract with reduced severance payments targeted to certain population groups. Lower social security contributions for employer’s for the first two years if one of these new permanent contracts was signed. Broaden the 1997 labor market reform; extension of new permanent contract of 1997 to more population groups. Suppression or reduction of social security contributions to support permanent employment for certain groups of the population. Early retirement only from age 61. Impulse partial retirement; possible to combine it with work. Unemployed aged 61 can retire if contributed for 30 years and the previous 6 months registered in employment offices. Incentives to retire after age 65. Individuals aged 52+ can combine unemployment benefits with a job. Extension of group of individuals that can benefit from the integration contract (program to help integrate the unemployed into the labor market). 15 effective contributory years are used to calculate the pension. Reduction from 8% to 7.5% of the per-year penalty applied to early retirees between 60 and 65 for individuals with 30 contributory years. Broaden incentives to stay employed after age 65. Increase contributions made by the social security administration for individuals receiving the special scheme of UA for 52+ (they will receive a higher old-age pension when retiring).

In 1984, both temporary contracts and non- contributory unemployment benefits (also called unemployment assistance benefits) were introduced. In addition, a special provision was established for workers aged over fifty- five who were allowed to receive unemployment assistance benefits until retirement age. To receive these benefits, individuals had to satisfy the entitlement requirements of the retirement pension except for the age. The subsidy paid 75 percent of the minimum wage until reaching the age to be transferred to an old- age pension. Furthermore, the years spent unemployed under

140

Pilar García-Gómez, Sergi Jiménez-Martín, and Judit Vall Castelló

this special scheme were counted as contributive years toward an old- age pension. In the following year, 1985, an old- age pension reform was passed that increased the minimum mandatory annual contribution to old- age pensions from eight to fifteen years, it also increased the number of years of contribution used to calculate the pension from two to eight years, and introduced several early retirement programs linked to hiring a new worker such as the partial retirement program that allowed part- time retirement at sixty- three combining part- time wages and old- age pension, and special retirement at sixty- four if the employer hired a registered unemployed.6 In 1989 the special provision of unemployment assistance benefits until the retirement age of sixty- five for individuals aged at least fifty- five was extended to individuals aged fifty- two, thus increasing the incentives of older workers to leave the labor market at younger ages. The decrease in the labor force participation rates of older individuals observed in Spain during the 1980s and the early 1990s prompted the government to adopt a change in the strategy, and to start a series of reforms to reverse these negative labor market trends. Therefore, the reforms introduced during the 1990s had the objective of keeping older workers active in the labor market for longer. There have been two main reforms since the mid- 1990s, one in 1997 and the other in 2002. In 1997 the number of contributory years used to compute the benefit base was progressively increased from eight to fifteen years and the formula to calculate the replacement rate was also made less generous.7 On the other hand, the 8 percent penalty applied to early retirees between the ages of sixty and sixty- five was reduced to 7 percent for individuals with forty or more years of contributions at the time of early retirement. Some changes in the incentives on the demand side were also introduced in 1997 to reduce the unemployment rates and the share of temporary contracts among the disadvantaged groups, including individuals aged fortyfive or older who were either unemployed or had a temporary contract. In 2002 changes in both the old- age and the unemployment systems were introduced. Before 2002, only individuals who had contributed to the system earlier than 1967 could benefit from early retirement at sixty, while the rest had to wait until the normal retirement age of sixty- five. In 2002, earlier retirement at age sixty- one was made available for the rest of the population. At the same time, there was an impulse to the partial and flexible retirement 6. The change in the minimum mandatory annual contributions to have access to an oldage pension affected all individuals since 1985, but the number of years used to calculate the pension was progressively increased: during the first year, the last seventy months were used, seventy- two months in the second year, and eighty- four in the third year. 7. In 1997 the last 108 months are included, the last 120 months in 1998, the last 132 months in 1999, the last 144 months in 2000, the last 156 months in 2001, and the last 180 months from 2002 onward.

Health, Disability, and Pathways into Retirement in Spain

141

schemes with the possibility of combining income from work with old- age benefits, and the introduction of incentives for individuals to retire after the legal retirement age of sixty- five (an additional 2 percent per additional year of contribution beyond the age of sixty- five for workers with at least thirty- five years of contributions on top of the 100 percent applied to the regulatory base). At the same time, the possibility to access retirement was extended to individuals who are unemployed for reasons beyond their willingness at sixty- one, and who have contributed for at least thirty years and have been registered in the employment office for the previous six months. On the other hand, the reform in 2002 opened up the possibility for individuals aged fifty- two or more who are receiving unemployment benefits to combine the receipt of these benefits with earnings, as they could receive 50 percent of normal benefits and the employer would pay the remaining quantity in wages. In addition, it extended the program that helps to integrate unemployed persons in the labor market to all individuals aged at least forty- five who have been unemployed for one month and to people with disabilities, among others.8 Last, in 2007 the incentives to retire later than age sixty- five were further increased by providing an additional 3 percent, instead of the 2 percent agreed to in 2002. Moreover, in order to have access to an old- age pension the individual must have contributed for at least two out of the last fifteen years, and the proportional part related to the extra monthly salaries would not be taken into account when computing the number of contributed years. On the other hand, the 8 percent penalty applied to early retirees between the ages of sixty and sixty- five was reduced to 6 to 7.5 percent, depending on the number of years contributed, for those individuals with thirty years of contributions. In addition, the contributions for unemployed workers older than fifty- two were increased so that they would receive a higher old- age pension when retiring. 3.3

Historical Data

Mandatory insurance for job- related accidents was introduced in Spain in 1900 through a bill that also authorized the creation of some funds, for public employees only, for paying disability and retirement pensions. In 1919, mandatory retirement insurance (Retiro Obrero Obligatorio) was introduced for private- sector employees aged sixteen to sixty- five whose total annual salary was below a certain threshold. In 1926, a universal pension system for public employees (Régimen de Clases Pasivas) was established, which still exists under the same name. By the late 1930s, most Spanish employees were covered by some minimal government mandatory retirement insurance program. 8. This program is called Contrato de Integración (Integration Contract).

142

Pilar García-Gómez, Sergi Jiménez-Martín, and Judit Vall Castelló

Since the introduction of the first insurance programs, the life expectancy of the Spanish population has experienced an outstanding improvement; while the life expectancy at birth in 1930 was of 51.13 years for females and 47.46 for males, it reached 71.65 (females) and 66.66 (males) years in 1960, and it still increased up to 84.07 (females) and 77.58 (males) years by 2006 (Human Mortality Database 2010). The labor force participation of older workers has not always gone hand in hand with the evolution of life expectancy. In this section we provide some descriptive evidence of the trends in mortality, health, and labor force participation in order to unravel the existence of any common trends during the last thirty years. 3.3.1

Mortality

The increase in life expectancy at birth has also been translated into increases in life expectancy at age sixty- five. In 1960 women aged sixty- five expected to live 15.3 years more while the expectation was 21.9 in 2008. Similar improvements are also observed among men (from 13.1 in 1960 to 18.0 in 2008) (OECD Health Data 2010). Moreover, the higher reductions in mortality rates achieved since 1960 are concentrated among the population aged sixty- five and older as figure 3.4 shows. Figure 3.5 shows how much steeper the decrease in mortality was among individuals aged sixty- five compared to individuals aged sixty or fifty- five. Figure 3.4 also shows that the decrease in mortality rates observed from 1960 to 1985 is similar to the decrease from 1985 to 2006. In addition, it shows that males’ mortality rates evolve after female’s mortality rates, as the curve for men in 1985 overlaps the curve for women in 1960, and mortality rates of men in 2006 are similar to mortality rates of women in 1985. Therefore, one would expect further improvements in life expectancy and mortality rates, at least for men, in the coming years. One of the conclusions to be derived from figure 3.5 is that individuals in later years reach the mortality rates of previous cohorts at older ages. For example, women aged fifty- five in 1960 had the same mortality risk as women aged sixty in 1980 and women aged sixty- five in 2005 (0.006). If one understands by old age the later part of life with some reference to deterioration, then one would probably agree that the experience of a given high mortality rate should be part of the elements to be considered when classifying a group of individuals as elderly. As recently pointed out by Shoven (2010), this raises some challenges to compare individuals through time.9 For example, if individuals were classified as elderly in the 1960s at age sixty- five, it seems somehow surprising that they were still classified as elderly in 2000 when their mortality rates were like the ones of individuals aged sixty in 1960 among men, and even lower among women. This suggests 9. As Shoven (2010) acknowledges, similar ideas were proposed earlier by others. See, for example, Fuchs (1984) or Cutler and Sheiner (2001).

Health, Disability, and Pathways into Retirement in Spain

143

that age- since- birth is possibly not the best measure to compare the aging of populations across time, and it poses some questions about its convenience to set the rules of the social security system. Figure 3.6 shows how individuals of different ages- since- birth could be considered to have the same age if measured by the same mortality rate. It plots the ages at which cohorts in different years face the mortality risk as

Fig. 3.4

Mortality rates by gender and age (1960, 1985, and 2006)

Source: Own elaboration from data from the Human Mortality Database.

Fig. 3.5

Mortality rates at different ages by year

Source: Own elaboration from data from the Human Mortality Database.

144

Pilar García-Gómez, Sergi Jiménez-Martín, and Judit Vall Castelló

Fig. 3.6 Ages of equal mortality rate in Spain in 1960 and age at which 1960 mortality rates are reached in Spain Source: Own elaboration from data from the Human Mortality Database.

sixty and sixty- five- year- olds in 1960. It can be seen that a man aged 72.7 (68.6) in 2006 had the same mortality risk as a sixty- five- (sixty)- year- old in 1960. Similarly, a woman aged 74.8 (71.5) had the same mortality risk as a sixty- five- (sixty)- year- old in 1960. Then, a mortality- based age system would suggest that a 74.8-year- old woman in 2006 and a sixty- five- year- old woman in 1960 were the same age. Similarly, a 72.7-year- old man in 2006 would have the same age as a sixty- five- year- old man in 1960. 3.3.2

Health Trends

We have previously shown that there have been striking improvements in age- specific mortality rates over the last fifty years in Spain across individuals aged over fifty- five. The extent to which these changes translate into higher labor capacity depends on the evolution of ill health. Thus, further reduction in the age- specific mortality rates will only translate in an increase in the population able to work at older ages if the average age of the onset of a work- related disability increases. This would certainly be the case if the compression of morbidity hypothesis (Fries 1980) was satisfied.10 However, 10. Fries’ (1980) compression of morbidity states that the burden of lifetime illnesses will be concentrated in a shorter period before death as the age of functional impairment due to ill health will be increased.

Health, Disability, and Pathways into Retirement in Spain

145

the international literature is inconclusive in this respect (Mackenbach et al. 2008). For example, the evidence provided by several studies that analyze disability trends in the United States from the 1980s and 1990s suggests that while the prevalence rates for individuals older than sixty have decreased, the rates for the younger age groups have seen no improvement or even a deterioration (for example, Crimmins, Reynolds, and Saito 1999; Lakdawalla, Bhattacharya, and Goldman 2004; Bhattacharya, Choudhry, and Lakdawalla 2008). In order to shed some light on the past trends of health status in Spain, we use data from the 1987, 1993, 1995, 1997, 2001, 2003, and 2006 editions of the Spanish Health Survey (ENS) available from the Ministry of Health and Social Policy (www.msps.es). These are nationwide cross- sectional surveys that collect information on health and socioeconomic characteristics of individuals. The surveys contain separate samples for adults (aged sixteen and older) and children. The following figures are based on the adult samples. We use weighting factors to compute the different averages. Figure 3.7 shows the evolution of self- assessed health status for men and women in the age- groups forty to forty- four, fifty to fifty- four, and sixty to sixty- four based on the question “how would you rate your health during the last twelve months?”. We show the percentage of each age and gender group that report being in good or very good health (good at least) and the percentage that report being in fair, good, or very good health (fair

Fig. 3.7

Evolution of self-assessed health

Source: Own elaboration from data from the Spanish Health Surveys.

146

Fig. 3.8

Pilar García-Gómez, Sergi Jiménez-Martín, and Judit Vall Castelló

Evolution of health limitations

Source: Own elaboration from data from the Spanish Health Surveys.

at least). We cannot conclude that there has been an overall improvement in self- reported health, neither a worsening, although there has been an increase in the percentage of women in at least good health among the three age- groups considered. This improvement in self- assessed health among women is not found when looking at the percentage of women in at least fair health. In addition, older individuals do not report being in worse health on average than the younger counterparts if we compare the percentage who report being in at least fair health. The age- gradient is observed with the other dichotomization used (at least in good health). This could suggest that reporting heterogeneity among age- groups is stronger for reporting being in fair health, as one would expect a clear age- gradient in average health status. In order to hypothesize what could have happened with the percentage of individuals with a health impairment to work, information which is not available for Spain, figure 3.8 depicts the percentage of individuals with a work limitation in the two weeks prior to the survey11 and the percentage of individuals who had any type of accident during the last year. First, the percentage of individuals with a work limitation or who have experienced an accident has increased for all age- groups for both sexes during the last twenty years. On the other hand, while there is a clear age pattern among 11. It is based on the question: “Have you had to reduce your principal activity (work, study, housework) at least half a day due to a health discomfort or symptom in the last two weeks?”

Health, Disability, and Pathways into Retirement in Spain

147

women, that is, a higher prevalence of both work limitations and accidents among women aged sixty to sixty- four compared to women in the other two age groups, the evidence among men is more mixed. Figure 3.9 shows trends in the prevalence of several diagnosed chronic illnesses: cholesterol, high blood pressure, diabetes, heart problems, and asthma or bronchitis (respiratory problems). The data show that the prevalence of cholesterol and high blood pressure have increased both for men and women across all age groups. Moreover, the reported prevalence of men and women in their forties at the end of the time period is similar to the one reported by individuals aged fifty twenty years ago. On the other hand, one should be cautious before concluding from these increasing trends that the

Fig. 3.9

Prevalence of chronic illnesses

Source: Own elaboration from data from the Spanish Health Surveys.

148

Pilar García-Gómez, Sergi Jiménez-Martín, and Judit Vall Castelló

prevalence of these two health problems has increased over the observed period as this could be due to a better awareness of the population, which would translate in higher self- reported rates. Johnston, Propper, and Shields (2009) find that, while the rates of self- reported hypertension in England were 5.5 percent in 1998 and 8.5 percent in 2003, hypertension measured by a nurse on the same sample decreased from 37 percent to 31 percent over the same period. Unfortunately, this objective information is not available for Spain through time. On the other hand, the reported prevalence of diabetes, heart, and respiratory problems remained stable through time with few exceptions: the prevalence of diabetes has increased among men older than fifty, the prevalence of bronchitis or asthma has increased among the youngest and the oldest group of women, and women in their sixties now suffer more heart problems. Figure 3.10 shows the prevalence of overweight and/or obesity in the last twenty years, which are known to be risk factors that could increase the burden of disease in the future. The data show that both the percentage of men with overweight and with obesity have increased during the last twenty years among all age- groups, with the increase in obesity rates being steeper than the increase in overweight rates. In 2006 almost 80 percent of men older than fifty were either overweight or obese compared to 60 percent in 1987. The percentage of women who are overweight is lower compared to men, which is consistent with the evidence found by Andreyeva, Michaud, and van Soest (2007) using data from the 2004 sample of the Survey of Health and Retirement in Europe. On the other hand, little has changed in the

Fig. 3.10

Prevalence of overweight and/or obesity

Source: Own elaboration from data from the Spanish Health Surveys.

Health, Disability, and Pathways into Retirement in Spain

149

weight distribution of women aged forty to fifty- five during the last twenty years. Finally, the prevalence of reported obesity among women in their sixties has increased by 10 percentage points, while the percentage that report being overweight has remained around 45 percent. In order to complement the self- reported descriptive evidence, we provide some information on the number of hospitalizations by type of diseases using administrative data from 1998 to 2007 from the Spanish Ministry of Health and Social Policy. We present data for the same three age- groups, but unfortunately we are not able to show figures for men and women separately. Figure 3.11 shows the number of hospitalizations for each 10,000 inhabi-

Fig. 3.11 Number of hospitalizations per 10,000 inhabitants by major health conditions Source: Own elaboration from administrative data from the Spanish Ministry of Health and Social Policy.

150

Pilar García-Gómez, Sergi Jiménez-Martín, and Judit Vall Castelló

tants by age- group and by major condition. The data show that neoplasm and circulatory diseases are the two main groups of health conditions leading to a hospitalization among all age- groups. Neoplasm represents the first cause among individuals younger than fifty in the overall period, but its incidence has decreased since 2000. On the other hand, while circulatory problems were the main cause of hospitalization among individuals aged fifty to fifty- four ten years ago, it currently ranks second after neoplasm due to a decrease in the number of hospitalizations due to circulatory problems in the last decade. There is a decreasing trend in the number of hospitalizations due to health problems related to the circulatory system among the older age- group, which remains the first cause, but simultaneously there is an increase in the number of hospitalizations due to a neoplasm. It is also worth mentioning the observed increase in the number of hospitalizations due to mental problems, it being steeper among the youngest age- group. The number of hospitalizations related to respiratory health problems has also increased among the younger group, while a decreasing trend is found for the rest of the health problems considered. 3.3.3

Mortality and Health

In the previous two sections we have shown that, while age- specific mortality rates have decreased through time, the evidence on the health status of the individuals aged forty to sixty- five is less conclusive, and it depends on the health measure used. Here we explore further the relationship of mortality with the other health variables. Figure 3.12 (men) and 3.13 (women) compare mortality equivalent ages versus SAH equivalent ages. We plot the average for each age- interval at the

Fig. 3.12

Mortality and self-assessed health 1987 and 2006, men

Health, Disability, and Pathways into Retirement in Spain

Fig. 3.13

151

Mortality and self-assessed health 1987 and 2006, women

midpoint of the intervals and then we fit a linear functional form for SAH and a power function for mortality. First, based on these rough estimates, figure 3.12 shows that men aged 58.7 in 2006 had the same mortality rates as men aged fifty- five in 1987, a difference of almost four years. The difference is about one year higher for women (figure 3.13) as women aged 59.9 in 2006 had the same mortality rates as women aged fifty- five in 1987. Second, both figures also show the proportion of individuals who reported that their health status was less than good in 1987 and 2006. The gain in SAH over this period is smaller than the gain in mortality as men aged 56.9 in 2006 reported similar SAH as men aged fifty- five in 1987 and women aged 57.7 in 2006 had almost the same SAH as women aged fifty- five in 1987. Thus, figures 3.12 and 3.13 show that the large gains in mortality rates have not been translated into better SAH. To reinforce the previous evidence, figures 3.14 (men) and 3.15 (women) depict trends in mortality, self- reported health status, and self- reported work limitations as defined earlier. On the one hand, the percentage of individuals who reported that their health status was less than good in the last year slightly decreased over the last twenty years, while the percentage who reported having to cut their principal activities at least half a day because of a health- related problem increased during the same period. 3.3.4

Trends in Disability and Labor Force Participation

In this section we provide some graphical evidence on labor force trends by age- groups and gender. Data on employment, unemployment, and disability come from the Encuesta de Población Activa (EPA). The EPA is a rotating quarterly survey carried out by the Spanish National Statistical Institute (Instituto Nacional de Estadística [INE]). The planned sample size consists of about 64,000 households with approximately 150,000 adult individuals. Although the survey has been conducted since 1964, publicly released cross-

152

Pilar García-Gómez, Sergi Jiménez-Martín, and Judit Vall Castelló

Fig. 3.14 Mortality rate of men, percent in less than good health and with work limitations, aged sixty to sixty-four

Fig. 3.15 Mortality rate of women, percent in less than good health and with work limitations, aged sixty to sixty-four

sectional files are available only from 1977. The 1977 questionnaire was modified in 1987 (when a set of retrospective questions were introduced), in the first quarter of 1992, in 1999, and 2004. The EPA provides fairly detailed information on labor force status and education and family background variables but, like most of the other European- style labor force surveys, no information on health is provided. The reference period for most questions is the week before the interview.

Health, Disability, and Pathways into Retirement in Spain

153

Figure 3.16 shows the evolution of employment rates by age- group for men and women separately. The data show that there has been an important increase in female labor force participation since the mid- 1980s, although employment rates of women in their forties are still far below the rates of their male counterparts (around 20 percentage points difference). The increase in female participation also translates in an increase in unemployment rates (figure 3.17). The unemployment rates of males and females have moved in parallel since the beginning of the nineties. In addition, the data show that there are no differences across age- groups among men, except for

Fig. 3.16

Employment by age-group, men and women

Note: Own elaboration using data from the Spanish Labor Force Survey.

Fig. 3.17

Unemployment by age-group, men and women

Note: Own elaboration using data from the Spanish Labor Force Survey.

154

Pilar García-Gómez, Sergi Jiménez-Martín, and Judit Vall Castelló

the lower unemployment rate of the older group resulting from lower labor force participation as also suggested by the employment rate. In the same spirit, figures 3.18 (men) and 3.20 (women) show the evolution of the percentage of individuals who classify themselves as permanently disabled when asked about their labor status in the previous week using data from EPA. It should be noticed that there was a change in the survey in 1987 that affects the numbers shown. Before 1987 there is information available about one state, while after 1987 individuals can be seen in up to three different states. This implies that some individuals could report being permanently disabled and doing some volunteer or paid work, for example. We have decided to count an individual as permanently disabled if he reports being so in any of the three possible states. This results in an increase in the percentage of disabled individuals after 1987, the discontinuity being higher among the older age groups. The data show that, for both male and female, the percentage of permanently disabled individuals is higher among the older age- groups, the difference across age- groups being larger among males. Figure 3.19 shows the different enrollment rates into DI for men by age for the last year of data available. The data show that the percentage of men collecting disability benefits increases with age from less than 4 percent among men aged forty to over 12 percent among men aged sixty- four. The share of women who can claim a contributory disability pension has increased through time with their labor market participation. This could explain the lower differences across age- groups among women, as well as the increase in the percentage of women into DI, while the share among men remains stable, except for the group aged over fifty- five. Figures 3.18 and 3.20 illustrate the years in which the two main reforms took place (1997 and 2002). We will investigate their effects in a bit more

Fig. 3.18

Evolution of disability rates by age-group, men

Note: Own elaboration using data from the Spanish Labor Force Survey.

Health, Disability, and Pathways into Retirement in Spain

Fig. 3.19

155

Proportion of men collecting disability benefits at different ages

Note: Own elaboration using data from the Spanish Labor Force Survey.

Fig. 3.20

Evolution of disability rates by age-group, women

Note: Own elaboration using data from the Spanish Labor Force Survey.

detail later. However, it is now worth mentioning that neither the trends nor the levels seem to have changed after the implementation of these reforms. 3.3.5

Disability, Health, and Mortality

We combine in figures 3.21, 3.22, and 3.23 the information on permanent disability shown in the previous section with the health and mortality information shown previously for three age- groups: (a) forty to forty- four (figure 3.21), (b) fifty to fifty- four (figure 3.22), and (c) sixty to sixty- four

Fig. 3.21 Disability, self-reported health, work limitation, and mortality at age forty, individuals aged forty to forty-four, by gender Note: The x- axis is common for the graph of men and women. The axis on the left is for the variables disability, bad self- assessed health, and work limitation, while the axis on the right refers to the mortality rate.

Fig. 3.22 Disability, self-reported health, work limitation, and mortality at age fifty, individuals aged fifty to fifty-four, by gender Note: See figure 3.21 note.

Health, Disability, and Pathways into Retirement in Spain

157

Fig. 3.23 Disability, self-reported health, work limitation, and mortality at age sixty, individuals aged sixty to sixty-four, by gender Note: See figure 3.21 note.

(figure 3.23). The data show that the trend of the percentage of individuals into disability does not follow any of the other health measures. So, despite the reduction in age- specific mortality, the percentage of individuals into disability remains almost constant. As argued earlier, we would expect a reduction of disability coming from an improvement in mortality only if the onset of the disabling condition happens later in life. Otherwise, the share of the population at a given age in ill health could even increase. Unfortunately, we cannot take any conclusive evidence on this regard with the analysis shown here. On the other hand, the stability of the share into disability during the last twenty years suggests that any changes that could have happened in the population’s health have not affected the inflows into disability. Thus, it is likely that other dimensions of the program are more important in explaining its evolution. We focus on the role of the different social security reforms in section 3.4. 3.3.6

Health, Mortality, and Labor Force Participation

The data in figure 3.16 showed that employment rates of men in their late fifties and early sixties in Spain decreased during the 1980s until the mid- 1990s and, although they have slightly increased in the last decade, they are far from the rates observed in 1980. In this section we combine data on health status and labor force participation. We do not show figures for

158

Pilar García-Gómez, Sergi Jiménez-Martín, and Judit Vall Castelló

Fig. 3.24

Evolution of the employment rate by mortality rate, men

women as the continuous increase in female participation rates masks any relationship. We follow Shoven (2010) and look at a different definition of age based on mortality risks and compare the employment rates of individuals at the same mortality risk in different points in time. Then, if we assume that individuals with the same mortality risk experience the same health status, we can evaluate how participation rates change across time for individuals with the same health status. Figure 3.24 illustrates the employment rate in 1980, 1990, 2000, and 2006 (the last year for which both employment and mortality data are available) for each mortality risk for men.12 The conclusions we draw from figure 3.24 are somehow different from the ones drawn before. First of all, we find that participation rates have not increased among the older group of individuals, defined as individuals with higher mortality, in the later years. On the other hand, the employment rates have decreased, not for all the individuals, but among the ones whose mortality risk is at least 0.5 percent. Consistently, the decrease through time is higher among the groups with higher mortality risks to the extent that the employment rates of groups whose mortality risk is at least 0.01 have been halved. Looking at the data in another way, the mortality rate when 60 percent of men were employed was 0.0175 percent in 1980, but it was only 0.01 percent twenty- six years later. Therefore, men in 2006 had to be much 12. We combine the previously presented information on employment rates from the EPA with information on mortality rates from the Human Mortality Database for individuals aged forty to sixty- nine. The rates refer to five age- group averages for both measures.

Health, Disability, and Pathways into Retirement in Spain

159

healthier (by the mortality measure) in 2006 that they were in 1980 for the employment rate to be 60 percent. As argued earlier, individuals with the same mortality risk do not necessarily face the same health status, as health care technology improvements could have helped to decrease the age- specific mortality risks, but at the cost of a higher prevalence of disability. In order to shed some light on this, we focus on the evolution of employment rates using both information on self- assessed health and the existence of a work limitation using ENS data. Figure 3.25 shows the employment rates by an individual’s self- assessed health. We have grouped self- assessed health into two categories: good or very good health (thin lines in the figure) and fair, bad, or very bad health (thick lines in the figure). We have used this division because it was found to better capture the age differences in health status (see figure 3.7). The data show that employment rates of individuals in bad health are much lower compared to their healthy peers among individuals younger than sixty. More specifically, among men aged fifty to fifty- four who report being in bad health, only about 60 percent is at work, while this number is higher than 80 percent among the healthy ones. The employment rates of healthy and unhealthy individuals converge among the group aged at least sixty. The second feature shown by figure 3.25 is that employment rates of individuals older than fifty- five in good health had fallen from 1987 to 1997 and, although some recovery is observed in 2006, employment rates are still below the ones observed in 1987. This is in line with the data shown in figure 3.16. The evidence regarding individuals in bad health is less clear-

Fig. 3.25

Evolution of the employment rate by self-assessed health, men

160

Pilar García-Gómez, Sergi Jiménez-Martín, and Judit Vall Castelló

cut and it seems that employment rates of this group have remained stable through time. All together, the evidence presented here suggests that health status is an important variable in determining labor force participation among individuals younger than sixty years old, but it becomes less important as the social security incentives of the old- age pension system kick in. Therefore, we will now focus on the role played by the different social security reforms in explaining the evolution of labor force participation trends in general, and more specifically, the participation in the disability program. 3.4

Pathways to Retirement and Program Reforms

In the previous section we have shown that trends in labor force participation are not likely to be driven by the evolution of health in the population. Thus, in this section we analyze the relationship with the other usual suspects, that is, the incentives that the social security system poses to individuals to withdraw from the labor market. In particular, we first look at the association between the characteristics of the social security system and the inflow into and the stock of the main contributory social security programs (permanent disability, retirement, and unemployment). After that, we provide a tentative evaluation of the two main reforms of the social security system that have affected the incentives of old- age workers to withdraw from the labor force during the period for which we have data available. 3.4.1

Descriptive Evidence on the Pathways to Retirement

The evidence illustrated in figures 3.16 to 3.20 using aggregate data from the Spanish Labor Force Survey, which uses self- reported labor status, did not show any change in behavior after the 1997 reform, nor after the 2002 reform. Let us recall that the main characteristics of the reform held in 1997 were the reorganization of the disability assessment system, the implementation of stricter medical control to apply for temporary disability benefits, and the decrease in the generosity of the contributory old- age benefits. The 2002 reform provided individuals with more incentives to continue working beyond the age of sixty- five while, at the same time, a more stringent search criteria was required among the unemployed. Figure 3.26 shows administrative data on the inflows into contributory permanent disability benefits obtained from the Spanish National Social Security Institute (www.seg- social.es). The data show that the percentage of individuals going into this system slightly decreased after 1997 for all the age- groups considered (from forty to sixty- four) and it stayed constant thereafter. Figure 3.27 shows comparable data for the inflows into contributory old- age benefits. The trends and levels remain stable through the period despite the different reforms.

Health, Disability, and Pathways into Retirement in Spain

Fig. 3.26

161

Inflow into contributory permanent disability by age-group

Source: Own calculations using administrative data from the Ministry of Employment and Immigration and population figures from the Ministry of Health and Social Policy.

Fig. 3.27 Inflow into old-age benefits, individuals aged sixty to sixty-four and sixty-five-plus Source: See figure 3.26.

In order to look at changes in the postponement of the retirement age, figure 3.28 shows the percentage of the population that enters into contributory old- age benefits by age for individuals aged sixty to sixty- four. Unfortunately, we do not have information on the total number of inflows into old- age contributory pensions by individual ages for individuals aged

162

Pilar García-Gómez, Sergi Jiménez-Martín, and Judit Vall Castelló

Fig. 3.28

Inflow into old-age benefits by age

Source: See figure 3.26.

sixty- five or older so we present two different shares. The first one considers inflow as a share of the total population aged at least sixty- five (as in figure 3.26), while the second one uses the population aged sixty- five to sixty- nine, as retirement later than sixty- nine is anecdotic. In order for the policy reforms to have effects on the sustainability of the system, they should have influenced the stock of individuals in the different programs. From 2002 onward, at least in theory, in order to claim unemployment benefits older workers also had to enroll in active searching at the same time that unemployed individuals aged at least fifty- two could combine unemployment benefits with earnings. A priori, one would expect this reform to have some effects on the number of older individuals claiming unemployment benefits. However, the percentage of the population older than fifty- five that was receiving unemployment benefits did not change thereafter. This is shown in figure 3.29, which uses administrative data from the Ministry of Employment and Immigration. In fact, the share of the population older than fifty- five receiving unemployment benefits continued to grow despite the economic growth. In figure 3.30 we make use of another source of administrative data, the Muestra Continua de Vidas Laborales (MCVL), to contrast the results obtained previously. The MCVL is a microeconomic data set based on administrative records provided by the Spanish Social Security Administration. It contains a random sample of 4 percent of all the individuals who, at

Fig. 3.29 Percentage of the population aged at least fifty-five receiving unemployment benefits, Spain, 1980 to 2009 Source: See figure 3.26.

Fig. 3.30 Number of individuals entering permanent disability benefits each year, Spain, 1970 to 2007 Source: Own calculations using administrative data from the Muestra Continua de Vidas Laborales.

164

Pilar García-Gómez, Sergi Jiménez-Martín, and Judit Vall Castelló

some point during 2007, had contributed toward the social security system either by working, being in an unemployment scheme, or receiving a contributory pension. The random sample selected contains over one million individuals. The MCVL provides rich employment history information at the cost of scarce representativeness as we go back in time. This is particularly important here as the individuals of any cohort entering into disability benefits are expected to have higher mortality rates and, therefore, to have a lower probability of being selected in 2007. In this respect, the information provided here is complementary to the other sources. Figure 3.30 plots the number of individuals entering permanent disability benefits each year from 1970 until 2007. The two vertical lines correspond to the years of the 1997 and 2002 reforms. As for the reform in 1997, a possible anticipation effect is observed as inflows into permanent disability pensions increased in 1997. This could be either due to expectations of potential future higher requirements as a result of the reorganization of the assessment system or to the decrease in the generosity of the old- age pension system. However, inflows into the disability system dropped in the subsequent years. On the other hand, the 2002 reform, which tightened the job search criteria for unemployed individuals, had a clear substitution effect of increasing the inflows into disability benefits. Recall that disability benefits are of a permanent nature and do not have any job search criteria attached to their eligibility requirements and are, thus, much more attractive for older workers already with some previous health- related problems. Figures 3.31 and 3.32 report, for the two- fifty and older groups of working- age individuals, the percentage of that being in each program in the EPA. The figures are only shown for men and cannot be fully compared to the ones shown in figures 3.26 to 3.30; not only because it is self- reported, but also because individuals receiving a non- contributory pension should also report receiving a pension or benefit (either disability, old- age, or unemployment) in the EPA, while they were not included in the previous figures. The data from these figures reinforce the previous evidence; the percentage of the population either unemployed, receiving disability or old- age benefits increases with age. On the other hand, the share of individuals that declare being disabled or retired is higher the older the group, while at the same time, the share of unemployed individuals becomes smaller. In fact, the share of individuals aged sixty to sixty- four that report being unemployed is below 5 percent compared to the rate of 10 percent (except for the last years) of the other group. At the same time, the data do not show any substitution effects across programs after the different reforms or any drop or increase in the participation rates. In order to better approach a measure of the pathways into retirement, we look at the exit routes from employment. We use the retrospective information available in the second quarter of the EPA regarding the labor status

Fig. 3.31 Percentage of individuals in each social security program (self-reported), men aged fifty-five to fifty-nine Note: Own elaboration using data from the Spanish Labor Force Survey.

Fig. 3.32 Percentage of individuals in each social security program (self-reported), men aged sixty to sixty-four Note: Own elaboration using data from the Spanish Labor Force Survey.

166

Pilar García-Gómez, Sergi Jiménez-Martín, and Judit Vall Castelló

of individuals in the previous year. We calculate the percentage that transit from employment to each of the status of interest. This is shown in figures 3.33, 3.34, and 3.35. Unfortunately, the retrospective information does not distinguish between the different jobless statuses, which would have allowed us to identify the individuals that transit from unemployment or disability into retirement. The conclusions reached from these figures are similar to the ones presented earlier. The share of individuals that leave employment and transit into unemployment is higher among the relatively younger individuals than among the older groups (80 percent of men aged fifty to fifty- four that leave employment and transit to one of the statuses of interest go to unemployment, compared to 20 percent among men aged sixty to sixty- four). The main difference with previous figures is the smaller percentage of individuals that leave employment to transit into disability among men older than sixty. This is consistent with the numbers shown in figure 3.26 for contributory permanent benefits, and suggests that the biggest share of individuals into disability in this age- group is mostly due to an accumulation of individuals who left employment and transit into disability earlier in their career and/ or individuals receiving non- contributory disability benefits transiting from another jobless state.

Fig. 3.33 Outflows from employment into unemployment, disability, and old-age, men aged fifty to fifty-four Note: Own elaboration using data from the Spanish Labor Force Survey.

Fig. 3.34 Outflows from employment into unemployment, disability, and old-age, men aged fifty-five to fifty-nine Note: Own elaboration using data from the Spanish Labor Force Survey.

Fig. 3.35 Outflows from employment into unemployment, disability, and old-age, men aged sixty to sixty-four Note: Own elaboration using data from the Spanish Labor Force Survey.

168

Pilar García-Gómez, Sergi Jiménez-Martín, and Judit Vall Castelló

3.4.2

A Tentative Analysis of the Quantitative Effects of the Reforms

The descriptive evidence shown in the previous sections of this chapter points out that the reforms had, if any, small effects on the labor market behavior of older workers. In this section we attempt to provide some estimates of both the effects of the different reforms in participation in different programs, as well as analyze the substitution effects among programs. We first show in figure 3.36 the share of each age- group that goes into permanent contributory disability before and after the set of reforms that took place in 1997 in both the disability and old- age pension systems using the administrative data from the Ministry of Employment and Immigration shown in figure 3.26. The data show that the share of individuals transiting into disability decreases among individuals older than forty, except among individuals aged fifty to fifty- four. The drop is small in absolute terms among all groups, but this is due to the small shares into disability. However, it represents a 9 percent drop for the group with the largest inflow into disability (individuals aged fifty- five to fifty- nine). These estimates represent a before- after analysis in a period of economic growth. As can be derived from the evidence shown earlier, the labor force participation has been increasing since the midnineties for older workers. Therefore, we need to control, at least, for changes in total employment outflows before drawing any conclusion. We provide estimates of how the percentage in each program (disabil-

Fig. 3.36

Inflow into permanent disability before and after the 1997 reform

Note: Data from the Ministry of Employment and Immigration.

Health, Disability, and Pathways into Retirement in Spain

169

ity, unemployment, and old- age) changes after each of the two reforms of interest when we introduce controls for the time trend and the total share of individuals out of employment. We use data from EPA aggregated by age and gender (fifty to fifty- four, fifty- five to fifty- nine, sixty to sixty- four) using two levels of regional aggregation. We first use the same figures shown previously in this chapter, and we later construct aggregates at the regional level (Autonomous Communities). Although there is no variation in the timing of the reforms across regions, this exercise provides us with variation in the shares of individuals out of employment and in per capita GDP. Thus, we estimate the following model: Ssrgat = α s0 + ∑ α sr Dr + ∑ α sg Dg + ∑ α sa Da + λ st + β s1997 D1997 r

g

a

+ β s2002 D2002 + δ sOsrgat + ε srgat ,

where Ssrgat is the share of individuals of gender g and age- group a in region r in year t that are in the status s (s being disability, unemployment, and retirement), and Osrgat is the corresponding share out of employment. The other set of explanatory variables are region- dummies (Dr), gender- dummies (Dg), age- dummies (Da), a time trend (λst) and two dummies capturing the effect of the reforms: D1997 takes value one from 1997 onward and zero otherwise, and D2002 takes value one from 2002 onward and zero otherwise. In order to control for the endogeneity of the total outflow, we use regional GDP per capita as an instrument. When we provide estimates at the countrylevel, we use national values (Spanish Institute of Statistics, www.ine.es). We repeat the same analysis but, instead of using information about the stock of individuals in the different programs, we use information on the outflows from employment as shown in figures 3.33, 3.34, and 3.35. This analysis is only shown using the aggregates at the country- level because the data cells at the regional- level were too small, as only the information on individuals that were working in the previous year is used to obtain the different aggregates. The results for the parameters of interest are shown in table 3.4. Tables 3A.1, 3A.2, and 3A.3 in the appendix show both the first- stage estimates and the coefficients of the other variables. First, notice in tables 3A.1, 3A.2, and 3A.3 that the total outflow and the share out of employment diminished after both the 1997 and the 2002 reforms. The sign of the effects of the two reforms on the participation in each program is the same using the data at the country- level (first three rows of table 3.4) or exploiting the regional variation (second three rows of table 3.4) except for the effect of the 1997 reform on the disability system. However, none of the effects at the country- level are significant, probably due to the small sample size (120 observations). On the other hand, the analysis at the regional- level shows an interesting pattern: both the 1997 and the 2002

170

Pilar García-Gómez, Sergi Jiménez-Martín, and Judit Vall Castelló

Table 3.4

Selected results of the estimate of the 1997 and 2002 reform on the stock of the different social security programs, and the outflows from employment

Stock national level Disability Unemployment Old-age Stock regional level Disability Unemployment Old-age Outflows from employment (national level) Disability Unemployment Old-age

1997 Coef (SE)

2002 Coef (SE)

–0.0059 (0.0036) –0.0035 (0.0020) –0.0048 (0.0026)

0.0010 (0.0019) –0.0007 (0.0026) –0.0061 (0.0042)

0.0085 (0.0045) –0.0045 (0.0045) –0.0111 (0.0048)

0.0209 (0.0057) –0.0068 (0.0058) –0.0193 (0.0066)

0.0002 (0.0002) –0.0024 (0.0015) 0.0001 (0.0007)

0.0008 (0.0002) 0.0002 (0.0009) 0.0020 (0.0011)

N

R2

120

0.649

120

0.051

120

0.839

1,836

0.481

1,836

0.276

1,836

0.733

108

0.626

108

0.781

108

0.846

Note: Results based on national and regional aggregates.

reforms decreased the stock into old- age benefits at the cost of an increased share of the participation into disability. More interestingly, the magnitude of the two opposite effects is the same, suggesting a clear substitution effect among these two programs in the older age- groups. The results also show that these two reforms seem to decrease the participation share of these age- groups into unemployment (although the coefficient is not significant, which is highly explained by the total share of the population out of employment). Regarding the effects of these two reforms on the outflows from employment into the different programs shown in the last three rows of table 3.4, we find that there was a significant increase in the outflow from employment into disability after the 2002 reform. The rest of the coefficients are nonsignificant, although this could be due to the lack of explanatory power of the small sample size (108 observations). 3.5

Conclusions

In this chapter we have shown that despite the large improvements in mortality rates among older individuals in Spain, the employment rates of

Health, Disability, and Pathways into Retirement in Spain

171

individuals older than fifty- five remain lower than the ones observed in the late 1970s, and the decrease in participation is more drastic when comparing different cohorts with the same mortality- based age than with the same age- since- birth. In addition, decreases in mortality rates do not necessarily go hand in hand with improvements in population health. However, the descriptive evidence on health trends provided here remains inconclusive. On one hand, there is some evidence suggesting a deterioration in population health, as the percentage that reports having reduced their principal activity because of a health problem, as well as the prevalence of hypertension, cholesterol, and obesity, and the number of hospitalizations due to mental disorders have increased. On the other hand, the percentage that reports being in good or very good health has also increased, while the number of hospitalizations due to other illnesses except for mental problems has decreased. Health status is an important variable in determining labor force participation among individuals younger than sixty, but it becomes less important as the social security incentives of the old- age pension system kick in. The comparison of trends in mortality, health, and employment and participation in different social security programs shows a lack of an overall association among these dimensions in the last twenty years in Spain. Thus, in this chapter we have tried to disentangle the effect of the two main social security reforms since 1990. In 1997 there was a reorganization of the disability system, the medical requirements for temporary disability were tightened, and the generosity of the old- age pension system was decreased, while in 2002 the job search criteria to receive unemployment benefits was tightened and more incentives to retire later were introduced. Using regional aggregate data, we find that both the 1997 and the 2002 reforms decreased the stock into old- age benefits at the cost of an increased share of the participation into disability. More interestingly, the magnitude of the two opposite effects is the same, suggesting a clear substitution effect among these two programs in the older age- groups. An avenue for further research is the evaluation of the aforementioned reforms using longitudinal individual data in order to follow the different transitions. Moreover, it would be of interest to use the time variation in the implementation of the different old- age reforms in order to disentangle the effects of interest.

172

Pilar García-Gómez, Sergi Jiménez-Martín, and Judit Vall Castelló

Appendix Table 3A.1

Outflows from employment into the different programs Employmentdisability

Out of employment

Out of employment Women 50 to 54 55 to 59 D1997 D2002 GDP Constant N

Coef.

SE.

–0.003 0.012 0.072 –0.015 –0.015 1.5×10–7 –0.028

0.003 0.004 0.004 0.007 0.008 4.2×10–8 0.023

Employmentunemployment

Employmentretirement

Coef.

SE.

Coef.

SE.

Coef.

SE.

0.030 –0.003 0.001 –0.001 0.000 0.001

0.017 0.000 0.000 0.001 0.000 0.000

0.093 –0.012 –0.005 –0.016 –0.002 0.000

0.040 0.002 0.002 0.006 0.002 0.001

0.118 –0.018 0.004 0.014 0.000 0.002

0.033 0.005 0.004 0.007 0.001 0.001

0.017

0.003

0.006

0.006

0.001

0.001 108

Note: Data aggregated at the national level.

Table 3A.2

Stock in the different programs Nonworking

Nonworking Women 50 to 54 55 to 59 D1997 D2002 Trend GDP Constant N

Coef.

SE.

0.391 0.108 0.281 –0.010 –0.003 0.012 –6.5×10–7 0.570

0.010 0.012 0.012 0.021 0.026 0.005 2.5×10–7 0.117

Disability

Unemployed

Retired

Coef.

SE.

Coef.

SE.

Coef.

SE.

–0.191 0.035 0.039 0.089 –0.006 0.001 0.001

0.028 0.011 0.014 0.018 0.004 0.002 0.000

0.370 –0.176 –0.048 –0.143 –0.003 –0.001 0.001

0.142 0.069 0.024 0.045 0.002 0.003 0.001

0.169 –0.215 0.028 0.165 –0.005 –0.006 0.002

0.192 0.108 0.048 0.048 0.003 0.004 0.001

0.017 120

–0.022

0.029

0.049

0.058

Note: Data aggregated at the national level.

0.098

173

Health, Disability, and Pathways into Retirement in Spain Table 3A.3

Stock in the different programs Nonworking

Nonworking Women 50 to 54 55 to 59 D1997 D2002 Trend GDP Constant N

Coef.

SE.

0.403 0.107 0.278 –0.026 –0.035 0.009 –1.1×10–5 0.363

0.004 0.004 0.004 0.007 0.006 0.002 2.1×10–6 0.010

Disability

Unemployed

Retired

Coef.

SE.

Coef.

SE.

Coef.

SE.

0.276 –0.149 –0.014 –0.047 0.008 0.021 0.000

0.124 0.051 0.014 0.035 0.004 0.006 0.000

0.264 –0.135 –0.036 –0.109 –0.005 –0.007 0.001

0.141 0.058 0.015 0.039 0.005 0.006 0.000

0.006 –0.168 0.049 0.215 –0.011 –0.019 0.002

0.148 0.063 0.020 0.041 0.005 0.007 0.000

–0.021

0.043 1836

0.013

0.050

0.099

0.057

Note: Data aggregated at the level of the Autonomous Communities.

References Andreyeva, T., P. C. Michaud, and A. van Soest. 2007. “Obesity and Health in Europeans Aged 50 years and Older.” Public Health 121:497– 509. Autor, D., and M. Duggan. 2003. “The Rise in the Disability Rolls and the Decline in Unemployment.” Quarterly Journal of Economics 118:157– 205. ———. 2006. “The Growth in the Social Security Disability Rolls: A Fiscal Crisis Unfolding.” Journal of Economic Perspectives 20 (3): 71– 96. Benítez-Silva, H., Richard Disney, and Sergi Jiménez-Martín. 2010. “Disability, Capacity for Work, and the Business Cycle: An International Perspective.” Economic Policy 25 (63): 486– 513. Bhattacharya, J., K. Choudhry, and D. N. Lakdawalla. 2008. “Chronic Disease and Severe Disability among Working- age Populations.” Medical Care 46 (1): 92– 100. Boldrin, M., P. García-Gómez, and S. Jiménez-Martín. 2010. “Social Security Incentives, Exit from the Workforce and Entry of the Young.” In Social Security Programs and Retirement around the World. The Relationship to Youth Employment, edited by Jonathan Gruber and David A Wise. Chicago: The University of Chicago Press. Burkhauser, R., and M. Daly. 2001. “United States Disability Policy in a Changing Environment.” Federal Reserve Bank of San Francisco. Working Papers in Applied Economic Theory no. 2002-21. Crimmins, E. M., S. L. Reynolds, and Y. Saito. 1999. “Trends in Health and Ability to Work among the Older Working- age Population.” Journal of Gerontology (Series B) 54 (1): S31–S40. Cutler, D. M., and L. Sheiner. 2001. “Demographics and Medical Care Spending: Standard and Non- standard Effects.” In Demographic Change and Fiscal Policy, edited by A. Auerbach and R. Lee, 253– 91. Cambridge: Cambridge University Press. Disney, R., and S. Webb. 1991. “Why Are There So Many Long- term Sick in Britain?” Economic Journal 101:252– 62.

174

Pilar García-Gómez, Sergi Jiménez-Martín, and Judit Vall Castelló

Fries, J. F. 1980. “Aging, Natural Death, and the Compression of Morbidity.” New England Journal of Medicine 303 (3): 130– 5. Fuchs, V. 1984. “Though Much Is Taken: Reflections on Aging, Health, and Medical Care.” The Milbank Memorial Fund Quarterly: Health and Society (Special Issue, Financing Medicare: Explorations in Controlling Costs and Raising Revenues) 62 (2): 142– 66. Gruber, J., and D. A. Wise, eds. 1999. Social Security and Retirement around the World. Chicago: The University of Chicago Press. ———. 2004. Social Security and Retirement around the World: Micro-Estimation. Chicago: The University of Chicago Press. Human Mortality Database. 2010. University of California, Berkeley (USA), and Max Planck Institute for Demographic Research (Germany). Accessed July. Available at: http://www.mortality.org. Jiménez-Martín, S., J. M. Labeaga, and C. Vilaplana Prieto. 2006. “Award Errors and Permanent Disability Benefits in Spain.” Universitat Pompeu Fabra. Working Paper no. 966. Jiménez-Martín, S., and Judit Vall Castelló. 2009. “Labor Force Transitions and Business Cycle: Evidence for Older Workers in Spain.” FEDEA, DT 2009-25. Johnston, D. W., C. Propper, and M. A. Shields. 2009. “Comparing Subjective and Objective Measures of Health: Evidence from Hypertension for the Income/ Health Gradient.” Journal of Health Economics 28:540– 52. Lakdawalla, D. N., J. Bhattacharya, and D. P. Goldman. 2004. “Are You Becoming More Disabled?” Health Affairs 23 (1): 168– 76. Mackenbach, J., W. Nusselder, S. Polinder, and A. Kunst. 2008. “Compression of Morbidity: A Promising Approach to Alleviate Societal Consequences of Population Aging?” Network for Studies on Pensions, Aging and Retirement. Netspar Panel Paper no. 7. McVicar, D. 2006. “Why Do Disability Benefit Rolls Vary Across Regions? A Review of the Evidence from the USA and the UK.” Regional Studies 40:519– 33. Shoven, J. B. 2010. “New Age Thinking: Alternative Ways of Measuring Age, Their Relationship to Labor Force Participation, Government Policies, and GDP.” In Research Findings in the Economics of Aging, edited by David A. Wise. Chicago: The University of Chicago Press. Organization for Economic Cooperation and Development (OECD). 2001. Economic Survey-Spain, 2001. Paris: OECD. ———. 2006. Sickness, Disability and Work: Breaking the Barriers (Vol. 1): Norway, Poland and Switzerland. Paris: OECD. ———. 2007. Sickness, Disability and Work: Breaking the Barriers (Vol. 2): Australia, Luxembourg, Spain and the United Kingdom. Paris: OECD. ———. 2008. Employment Outlook. Paris: OECD. ———. 2010. OECD Health Data, 2010. Paris: OECD. Oliva, J. 2010. “Pérdidas Laborales Ocasionadas por la Enfermedad y Problemas de Salud en España en el año 2005.” Madrid: Institute of Fiscal Studies. Working Paper no. PTN 5/10.

4 Health Status, Welfare Programs Participation, and Labor Force Activity in Italy Agar Brugiavini and Franco Peracchi

4.1

Introduction

Social protection is a central theme in the public policy debate, both because of a concern for protecting the welfare of individuals at different stages of their life cycles while reaching a certain degree of coverage and because financial sustainability of the social security system, under the demographic pressure, has become a major challenge. The present chapter describes the different Italian social protection programs that have provided resources to the elderly, trying to establish a link with longevity and the evolution of health conditions. Along with a description of the institutional setup and its relevant reforms, we describe the evolution of the main drivers: mortality and labor force participation. In the Italian context, access to oldage and early pensions has shaped the labor supply decisions of individuals, so we relate the evolution of health status to disability insurance participation as well as retirement pensions. Our evidence, based on time series, cross- sectional and panel data, shows that the spectacular increase in longevity experienced by the Italian population in the last thirty to forty years has not lead to an increase in labor market participation. In fact, the “Italian paradox” is that higher longevity and improved health conditions are associated with a widespread detachment from the labor force. The main drivers of this behavior are the incentives Agar Brugiavini is professor of economics at University Ca’ Foscari, Venice. Franco Peracchi is professor of econometrics at the University of Rome Tor Vergata. We are grateful to Fondazione Rodolfo De Benedetti (FRDB) for letting us use the INPS sample and to the Italian National Statistical Institute (ISTAT) for letting us use the MARSS data. For acknowledgments, sources of research support, and disclosure of the authors’ material financial relationships, if any, please see http://www.nber.org/chapters/c12385.ack.

175

176

Agar Brugiavini and Franco Peracchi

embedded in the social security system that encouraged retirement, even at very early ages. This is clearly seen in the response to the pension reforms introduced during the 1990s, when a “retirement run” is observed as reforms are implemented—or even in the response to a reform announcement. It is worth stressing at the outset that, after the changes introduced in the mid1980s, disability pensions play a relatively minor role in Italy, representing less than 10 percent of the annual flow into retirement, the main and most easily accessible exit route from employment being early retirement. The remainder of this chapter is organized as follows. Section 4.2 introduces the institutional characteristics of the welfare system in Italy and describes the main reforms in the Italian social security system. Section 4.3 presents some historical data on mortality and health status. This section provides a comprehensive overview of the patterns of mortality in Italy by gender and age- group, and their evolution over time, both at the national and the regional level. The analysis of the mortality patterns by birth cohorts and causes of death also tries to emphasize the possible sources of the gains in longevity. Section 4.4 presents labor force and social security program participation in Italy over time and by age- group. The aim is to establish how the reforms of the social security system affected participation behavior and the take-up of the different benefits. We exploit information on new entries into the public pension and the disability insurance system to better understand the immediate and the long- term effects of the reforms on workers’ retirement decisions. This section also analyzes the various pathways to retirement, making use of the panel dimension of the data. We follow different cohorts of individuals who, throughout their working lives, experienced different social security arrangements and different reforms. Section 4.5 relates historical data on employment, disability, and retirement to mortality and evolution of the health status. Finally, section 4.6 concludes. 4.2

History of Reforms

The Italian welfare system is characterized by high spending on retirement pensions relative to unemployment benefits and disability provisions. Pension expenditures represent 15.2 percent of the gross domestic product (GDP) in 2007, a fraction that is still increasing partly as a result of the recent recession (RGS 2009). Old- age and early retirement pensions alone account for 10 percent of GDP, survivor pensions for 2 percent, total disability and invalidity pensions together for just 1.7 percent, with the rest being other forms of income support for the elderly. Unemployment benefits represent only 0.5 percent to 0.6 percent of GDP, and other social areas are also negligible (ISTAT 2010; OECD 2009). It is clear that, in Italy, old- age pensions crowd out other social expenditures such as unemployment benefits. Several reforms of the welfare system have taken place over the years, but

Italy

177

the most important ones have been in the area of public pensions. In this section we briefly review these reforms.1 4.2.1

Institutional Features of the Social Security System

Because the basic historical aspects of the system are well documented (see, for example, Brugiavini, Peracchi, and Wise [2003]; Brugiavini and Peracchi [2004]), in this chapter we only describe briefly its main rules and recent developments (see also OECD 2007). The Italian social security system is based on a variety of institutions administering pension programs for different types of workers (privatesector employees, public- sector employees, self- employed, professional workers).2 All programs are of the unfunded pay- as-you go (PAYG) type. Despite a process toward convergence during the 1990s, the various programs maintain quite different rules. In terms of pillars, the first pillar is the most relevant. It ensures against three types of risks: longevity (old age and early retirement benefits), death (survivor benefits), and disability (disability benefits).3 The second and third pillars have been encouraged by some legislative acts, but never actually implemented. Currently, about two- thirds of the labor force is insured with the National Social Security Institute (INPS). The Institute is responsible for a number of separate funds, of which the most important (Fondo Pensioni Lavoratori Dipendenti [FPLD]) covers the private- sector, nonagricultural employees. The system started before World War I, but was redesigned in 1969: social security benefits changed from a defined- contribution to an earningsrelated social insurance system. The aim was to guarantee to every retired worker an income comparable to that earned during the working life. At the same time, coverage increased. An early retirement pension was introduced, which also guaranteed a retirement income to individuals with at least thirty- five years of contribution, irrespective of age. The way benefits were computed changed over time, but was essentially of a final salary type, as average earnings over the last years of work (pensionable earnings) formed the basis for the calculation. To pensionable earnings, a rate of return (representing the accrual factor) was applied: a parameter approximately equal to 2 percent for each year of contribution.4 Until 1968, an old- age pension could not be combined with earnings, but this restriction was lifted in 1969, albeit within limits. Again in 1969, pension benefits were automati1. “Social security” and “pensions” in Italy have the same meaning, as there is essentially no second or third pillar yet (and certainly not in the period under investigation). 2. Social security system and pension system are used as synonyms in this chapter. In fact, social security is the main source of publicly provided income in old age in Italy. 3. In Italy, disability pension refers to more than one type of pension. See section 4.2.2 for more details. 4. The rate was initially lower and grew over time.

178

Agar Brugiavini and Franco Peracchi

cally indexed to both inflation and part of the wage growth. In 1976, they were automatically indexed to the contractual wages of the workers in the industrial sector. The eligibility age for old- age benefits was sixty for men and fifty- five for women though, as already mentioned, a generous early retirement option was available. As for public- sector employees, benefits calculations were even more generous than in the private sector, as the earnings immediately prior to retirement formed the basis of the benefit. Further, men and women could take early retirement with just twenty or fifteen years of contributions, respectively. 4.2.2

Reforms of the Social Security System in the 1990s

While minor changes in the rules took place very frequently throughout the 1970s and 1980s (almost invariably increasing the generosity of the system), major reforms of the social security system took place in 1992, 1995, and 1997.5 They are known, respectively, as the Amato, Dini, and Prodi reforms, from the names of the prime ministers at the time. Further changes to the system have been made nearly every year since 1992. Of the three main reforms of the 1990s, the Dini reform appears as the most radical, because it completely redesigns the system by modifying the eligibility rules and by changing the benefit formula from defined- benefits to notionally definedcontributions. However, because the changes are only introduced gradually, through a very long transitional period, the direct effects of the Dini reform were lower compared to the less radical Amato reform. An important aspect of the reforms of the 1990s is the differential treatment of different cohorts of workers. The Amato (1992) reform explicitly distinguishes between workers with at least fifteen years of contributions at the end of 1992 and all other workers. The old (1969) system applied to the former group, whereas the new system applied to the latter. Under the new system, the eligibility age for an old- age pension was to increase gradually by one year of age every two years, starting from 1994, until reaching age sixty- five for men and age sixty for women in the year 2000. The eligibility age was instead immediately set at age sixty- five for state employees and age sixty for local government employees, irrespective of gender. The number of years of contribution required for an old- age pension was to increase gradually by one every two years starting from 1993, until reaching twenty years of contributions in 2001. For workers with less than fifteen years of contributions at the end of 1992, the reference period for computing pensionable earnings was to increase gradually to include the whole working life, with past wages adjusted to inflation on the basis of the annual rate of change of the cost- of-living index plus an additional 1 percent. One important change, 5. See also Brugiavini and Peracchi (2004), Brugiavini and Galasso (2004), and Brugiavini (2009).

Italy

179

aimed at curtailing the generosity of the system, was indexation of pension benefits only to price inflation, not to wage growth. The differential treatment of older and younger workers was maintained in the subsequent Dini (1995) and Prodi (1997) reforms. The Dini reform distinguished between three categories of workers depending on their number of years of contribution at the end of 1995: (a) those with at least eighteen years, (b) those with less than eighteen years, and (c) new entrants to the labor force. Except for the changes to the eligibility rules introduced by the Amato reform, very little changed for workers with at least eighteen years of contributions. On the other hand, for new entrants, the Dini reform changed dramatically the benefit computation method from definedbenefit (DB) to notionally defined- contribution (NDC). For workers with less than eighteen years of contributions, benefits would be calculated on a pro rata basis according to the number of years of contributions under the two regimes (contributions paid after 1995 count under the new regime). The Dini reform also modified the existing rules on early retirement by introducing a dual requirement in terms of age and years of contribution: starting from 1996, a worker with thirty- five years of contribution could retire only if aged fifty- two or older. This limit was raised gradually until reaching age fifty- seven in 2002 for both men and women (similar, though slightly different rules applied to public sector employees). The minimum number of years of contribution was also increased gradually until reaching forty years in 2008. Since the eligibility rules for old- age and early retirement pensions are driving the behavior of workers during the period that we consider, tables 4.1 and 4.2 summarize their year- by- year changes for the two main funds of INPS, private employees (FPLD), and artisans and traders. By far the most important change of the Dini reform was the change in the benefit computation method from DB to NDC for the new entrants into the labor market. For these workers, the pension is to be based on a notional contribution equal to 33 percent of covered annual earnings in the case of employees (only 20 percent in the case of the self- employed). Cumulated notional contributions earn an annual return equal to the average rate of increase of gross domestic product (GDP) during the past five years. At retirement, the lifetime accrued notional stock of contributions is converted into a pension through an actuarial coefficient that varies depending on the worker’s age (from 4.720 percent at age fifty- seven to 6.136 percent at age sixty- five). Since a pro rata system applies to the benefit calculation for workers with less than eighteen years of contributions at the end of 1995, actual benefits also vary depending on the year of entry into the labor market. As for collecting benefits while working, the new rules allowed combining benefits with earned income, subject to some limitations. The pre- 1995 regime also included a minimum pension, which was later abolished for workers entering the labor market after 1995 (under the new regime, retirees

60 60 60 60 60 60 60 60 61 61 62 63 63 64 65 65 65 65

1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003

55 55 55 55 55 55 55 55 56 56 57 58 58 59 60 60 60 60

Age female

15 15 15 15 15 15 15 16 16 17 17 18 18 19 19 20 20 20

Contributive years 65 65 65 65 65 65 65 65 65 65 65 65 65 65 65 65 65 65

Age male 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60

Age female 15 15 15 15 15 15 15 16 16 17 17 18 18 19 19 20 20 20

Contributive years — — — — — — — — — — 52 + 35 52 + 35 54 + 35 55 + 35 55 + 35 56 + 35 57 + 35 57 + 35

Age and seniority 35 35 35 35 35 35 35 35 35 35 36 36 36 37 37 37 37 37

Seniority only

Early-retirement (private employees)

— — — — — — — — — — 56 + 35 56 + 35 57 + 35 57 + 35 57 + 35 58 + 35 58 + 35 58 + 35

Age and seniority

35 35 35 35 35 35 35 35 35 35 40 40 40 40 40 40 40 40

Seniority only

Early-retirement (artisans and traders)

Note: It should be noted that the Dini (1995) reform introduced a window of ages for eligibility to an old-age pension (between fifty-seven and sixty-five for both genders) if five years of contributions were completed. However, in practice these rules never took place because they would apply only to workers with less than eighteen years contributions in 1996 (typically workers born between 1955 and 1965). Since the flexible window for retirement was abandoned in 2007, no worker of these cohorts would have reached age fifty-seven before 2007. Note that rules in 1985 were the same as in 1986.

Age male

Old-age (artisans and traders)

Eligibility criteria for old-age and early retirement pensions 1986 to 2003

Old-age (private employees)

Year

Table 4.1

Italy

181

with total incomes below the social- assistance level can claim a meanstested social benefit from age sixty- five). 4.2.3

Disability Insurance

A public invalidity pension was introduced in 1919 for individuals who, because of reductions in their working capacity, were unable to earn income below a certain amount, equal to one- third of the normal wage for a worker in the same activity in the same area of the country. In 1984, the public invalidity pension was modified as to cover individuals with a certified mental or physical impairment, which reduces their working capacity by twothirds. Also, a public inability pension was introduced for private- sector employees and self- employed workers with a physical or mental disease, certified by a medical test, which is permanent and makes it impossible to carry out any job. With the Dini (1995) reform, the public inability pension was extended to public employees under the proviso that the impairments are not caused by work (a special inability pension for the public sector already existed, which provides for income support in the case of workcaused impairments). Currently, the Italian social security system provides two types of disability benefits: 1. Ordinary (“civilian”) disability benefit, granted to all citizens under certain health impairments (including deaf or dumb people older than eighteen). In some cases, a monthly attendance benefit is also paid in order for the beneficiary to receive home care. For people older than sixty- five, a non- contributory pension is paid ( pensione sociale). Since 1971, an inability pension is also envisaged for individuals who cannot carry out any type of work, at ages between eighteen and sixty- five. The handicap should be certified to be 100 percent of the working capacity of the individual. 2. Invalidity benefit, granted to workers registered with the Italian social security administration (INPS) whose working capacity is permanently reduced by at least two- thirds because of physical or mental impairments. The important landmark in terms of reforms was a law passed in 1984 establishing that an invalidity pension could be granted only if the physical or mental disease was certified by a medical test. Furthermore, at least five years of social security contributions were necessary, of which at least three were paid during the five years before applying for benefits. The important restriction introduced in 1984 was that the invalidity pension was no longer a permanent pension but should be renewed every three years, becoming permanent only after three renewals. It would also be automatically converted into an old- age pension at the legal retirement age and could not be paid along with unemployment benefits. On the other hand, inability pensions are permanent (not subject to renewal) and never become old- age pensions. Further small restrictions were introduced later on. For example,

182

Agar Brugiavini and Franco Peracchi

starting from 1995 the disability pension cannot be collected along with the life annuity (INAIL life annuity) granted in case of job injury. 4.2.4

Unemployment Insurance

The Italian unemployment insurance system differs substantially from that of other countries in both the eligibility conditions and the amount and duration of the benefit. The system currently provides income only to workers who lost their job, not to first- time job seekers. Contributions to the system are paid by workers and employers at differential rates (in the industrial sector 4.71 percent of gross wage, of which 4.41 percent falls on the employer). The system offers two different schemes: Comprehensive insurance (UI) Ordinary unemployment benefit Special unemployment benefit Mobility benefits Partial Insurance Cassa Integrazione Guadagni Ordinaria (CIGo) Cassa Integrazione Guadagni Straordinaria (CIGs) Unemployment insurance (UI) benefits are paid to workers individually laid off in the private sector or collectively laid off, not eligible for other benefits, and who have paid contributions for at least fifty- two weeks during the twoyear period preceding unemployment.6 Some benefits may also be claimed by those who worked at least seventy- eight days over the last year (reduced requirement). Ordinary UI benefits are paid on a seven- day- a- week basis, for a maximum of one hundred eighty days (extended to eight months since January 2008). For unemployed aged fifty and older, the duration has recently been extended from nine to twelve months. Furthermore, the benefit is stopped if the beneficiary gets a new job or refuses a job similar to the lost one, or does not accept to be employed in a socially useful job. Benefits are paid as a percentage of the average wage in the last three- months- wage, with some ceilings imposed. For example, ordinary UI benefits are 60 percent of the average wage for the first six months, 50 percent for the seventh and eighth month, and 40 percent for the subsequent months, up to a maximum level of €886 for wages below €1,917, and €1,065 for wages above that figure. Cassa Integrazione Guadagni (CIG) is a wage supplementation fund for workers in industrial firms with fifteen or more employees (workers in industrial firms with less than fifteen employees and in most of the services are excluded). Workers on CIG formally retain an ongoing work relationship, as their contract has not been terminated, so they do not enter the official unemployed count. Still, we think that CIG is relevant for two reasons. First, 6. Young workers in vocational training are excluded.

Italy

183

in survey data, individuals may report themselves as unemployed if they receive CIG benefits. Second, a broad definition of unemployment should also include these beneficiaries. Ordinary benefits (CIGo) are paid to workers for foregone hours of work (also at zero hours) due to the temporary reduction of the firm’s activity. The CIGo benefits are normally paid for a maximum of thirteen months (under special circumstances, up to fifty- two weeks). The replacement rate is 80 percent of the wages foregone. Special benefits (CIGs) are paid when reduction of the activity is not temporary, but is due to sector- or area- specific firm restructuring. The CIGs benefits are normally paid for twelve to twenty- four months (the length depending on the type of difficulties that the firm faces as well as on the restructuring strategy). Some extension may be obtained if restructuring lasts for more than twenty- four months. In any case, CIGs benefits cannot be paid for more than thirty- six months over five years. Workers (except those in the construction industry) who formerly benefited from CIGs are available for the mobility benefits under the UI scheme. Hence, mobility benefits represent a follow up to CIG. 4.3 4.3.1

Historical Data Mortality

This section reviews the historical trends and the current patterns of mortality among the elderly in Italy. Our data come from two sources. The first is the Human Mortality Database (HMD), which contains annual mortality rates broken down by gender and single year of age. The HMD data are derived from official vital statistics and census counts published by the Italian National Statistical Institute (ISTAT) and refer to the country as a whole. They cover the period from 1872 to 2006, thus offering a long- term view of the time trends of mortality in Italy. However, because the quality of the data for the early period 1872 to 1905 is lower than in later years, we confine ourselves to the period 1906 to 2006. Our second data source is the sequence of annual life tables for the period 1974 to 2006, broken down by gender and region of residence, compiled by ISTAT. These data offer a detailed geographical description of the changing patterns of mortality over the last three decades. Secular Trends Figure 4.1 shows the secular trend of mortality at four different ages (forty, fifty, sixty, and seventy), respectively for females and males. Mortality rates are in percentage terms and the vertical lines in the figures mark the periods 1915 to 1918 and 1940 to 1945, which correspond to World War I and World War II, respectively (notice that Italy entered both wars one year later). Figure 4.2 displays the same data in a different format by showing the

184

Fig. 4.1

Agar Brugiavini and Franco Peracchi

Mortality rates over time (women left, men right)

Fig. 4.2 Mortality rates by birth cohort at ages forty to seventy (women left, men right)

profiles of mortality for females and males belonging to six different birth cohorts (born in 1866, 1880, 1894, 1908, 1922, and 1936). This figure helps us appreciate the changes across cohorts in the profile of mortality between age forty and age seventy. If we ignore the sharp spike corresponding to the last years of World War I and the post- war influenza pandemic and the spike corresponding to World War II, the figures are dominated by the dramatic decline of mortality for both females and males. It is worth pointing out at the outset that, except for its timing and intensity, mortality decline in Italy follows a pattern that is qualitatively similar to that observed in all industrialized countries. As a consequence of the mortality decline, life expectancy at birth has risen from about forty- five years for both men and women in 1906 to seventy- eight years for men and eighty- three years for women in 2006, with a gain of thirty- three life years for men and thirty- eight for women. At age sixty the gain in life expectancy is 7.6 years for men and 11.5 years for women, while at age eighty the gain is 3.2 life years for men and 4.9 for women. In 1906, both men and women reached the 5 percent mortality rate at age sixty- nine, while in 2006 the 5 percent threshold is reached by men at age seventy- eight,

Italy

185

that is seven years later, and by women at age eighty- three, that is fourteen years later. Mortality first began to shrink in the early decades of the 1900s. Between 1906 and 1931 men gained 8.9 life years at birth and women about 10.5. Between 1931 and 1956, survival takes another major leap, with men gaining 11.7 extra life years at birth and women 14 extra years. During this period, the main contribution to the trend comes from the decline in infant and youth mortality, largely due to the progressive eradication of infectious diseases. The rapid decline in infant and youth mortality in the first part of the 1900s contrasts sharply with the marginal gains in male survival at later ages: for adult males very little changes are registered before World War II. The reduction in adult male mortality begins only later, during the 1950s and 1960s for the younger adults, and generalizes at later ages between the 1970s and the 1980s. The pattern observed for adult females is quite different. Their mortality risk is similar to that of males at the beginning of the period, but their mortality decline begins much earlier and gives them an increasing advantage over men. Until the 1970s, mortality rates among adult males hardly change, especially at ages above sixty. On the contrary, the decline of adult female mortality shows no sign of slowing down. From the mid1970s, mortality rates decline for both genders even at older ages, although at a slower pace for men than for women. Figure 4.3 shows the behavior of the ratio between male and female mortality, respectively, over time for selected ages and by cohort. Over time, the two genders diverge to such an extent that the mortality risk of a sixtyyear- old Italian woman today is only about half the risk of a male of the same age. Differences in lifestyle and work environment help explain this huge difference. Traffic accidents among the young, lung cancer and heart attacks for adults, and cardiovascular diseases for the elderly are the main causes of the gender gap that grows over most of the 1900s. These causes are clearly linked to risk factors identified with being male—speeding, smoking, working conditions, eating habits, and alcohol abuse. The figure

Fig. 4.3

Relative male mortality, over time (left) and by birth cohort (right)

186

Agar Brugiavini and Franco Peracchi

suggests a reversal of the trend in recent years with the gap narrowing, albeit slowly. Regional Differences Italy is one case where going beyond the aggregate trends is important because of the considerable degree of heterogeneity across regions. In this section we briefly consider the evidence obtained from the life tables produced by ISTAT for each single year from 1974 to 2006, and broken down by age, gender, and region of residence. At the beginning of the 1900s, adult and elderly male mortality in Southern Italy was on average much lower than in the North, and especially much lower than in the Alpine areas, while differences in female mortality were smaller (Caselli et al. 2003). The gap in adult male mortality between the North and the South grew over time, reaching a maximum in the early 1970s. Taking life expectancy at age sixty as reference, a sixty- year- old man from Lombardy could expect to live about three years less than someone of the same age from Sardinia or Calabria. The life tables for the early 1990s show that, for men, the geographical features of mortality remain qualitatively unchanged but the size of regional differences is considerably narrower, largely due to the gains made in the North compared to the South. The life tables for the early 2000s show that, with the exception of Campania, regional convergence in male mortality has essentially being reached. The southern region of Campania is an exception because it continues to exhibit relatively high levels of mortality, a disadvantage that is actually slowly increasing in relative terms compared to the other regions. The geographical features of mortality in the early 2000s are the opposite for older women, with the Southern regions now exhibiting higher mortality rates than the North or the Center, although regional differences are smaller for females than for males. Once again, Campania tops the list, with sixty- year- old women having a life expectancy that is two years shorter than the Italian average, followed by Sicily, Basilicata, and Calabria. Figure 4.4 summarizes these trends by showing the time profile of female and male mortality between 1974 and 2006 for the three macro regions of the country, namely the North, the Center, and the South and Islands. For both men and women, the largest gains in survival during the last thirty years have been made in the Northern regions. The different behavior of mortality in the North and the South between the 1970s and the 1990s is largely explained by the different evolution of mortality for cardiovascular diseases. The decline of this specific cause of death, which is the real protagonist of the positive trend of the last thirty years, has been especially strong in the Northern regions, partly because of the better quality of the health services and partly because of the success of prevention campaigns in changing eating habits and lifestyles at the individual level, which is of fundamental importance in order to control the risk factors that cause these kinds of diseases.

Italy

Fig. 4.4

187

Mortality rates over time by region (women left, men right)

Recent Trends among the Elderly A noticeable feature of recent years is the acceleration of the trend toward falling mortality among people over age sixty. In terms of life expectancy at age sixty, males gained 1.5 life years between 1906 and 1956, and 6.1 life years between 1956 and 2006, while women gained 3.3 life years between 1906 and 1956, and 8.3 life years between 1956 and 2006, giving them an advantage of nearly 4.2 life years over men of the same age (figure 4.5). Perhaps more striking is the sharp decline in mortality of older people (aged eighty and above) over the last fifty years. In terms of life expectancy at age eighty, men gained only 0.23 life years between 1906 and 1956 but a full three life years between 1956 and 2006, while women gained only 0.5 life years between 1906 and 1956 but 4.4 life years between 1956 and 2006, giving them an advantage of nearly two life years over men of the same age. Figure 4.5 shows recent trends in mortality for three age- groups that are of interest in this chapter. For both genders there is a steady decline in the mortality rate since the 1970s, but men start at a much higher level (almost thirty out of 1,000 men aged sixty- five in 1970) and their decline is more substantial. Since women have already gained longevity in the past decades, women in the younger age- group of fifty- five show a negligible decline in the mortality rate. Mortality is increasingly concentrated among the oldest old. If we consider the Italian women, most of the deaths now occur after age seventy and, in particular, after age eighty. According to some demographers, we are witnessing the opening up of new frontiers in elderly survival. It is widely held that primary prevention, closely linked to the large investments in education and, more generally, in the human capital of the new cohorts, will play a crucial role in increased longevity. Opponents of this optimistic view stress instead the negative effects of pollution and modern lifestyles, and also argue that lack of selection at early ages may increase frailty and compromise health during old age. To further appreciate the recent gains in longevity, figure 4.6 presents the ages of equal mortality probability with respect to a reference age for

188

Agar Brugiavini and Franco Peracchi

Fig. 4.5

Mortality rates at different ages (women left, men right)

Fig. 4.6

Ages of equal mortality probability (women left, men right)

females and males separately. In 2006 a woman aged sixty- eight has the same mortality probability as a woman aged sixty in 1974 (and a woman aged seventy- two has the same as a woman aged sixty- five in 1974). For men the gains in survival are even more important in the recent years (a man aged 68.4 in 2006 has the same mortality probability of a man aged sixty in 1974). Figure 4.7 provides the same evidence in a more direct way by showing two- year mortality rates by age and gender. At a mortality rate of 5 percent, both men and women experienced gains between 1974 and 2006 of six years (from seventy- two to seventy- eight for men and from seventy- six to eighty- two for women). 4.3.2

Mortality by Cause of Death

Unfortunately, Italy does not have any readily available measure of health status (such as self- reported measures of health) over a sufficiently long period. The only evidence available for a sufficient time- span is mortality rates by causes of death (figure 4.8). The mortality decline observed throughout the 1900s is largely driven by the steady decline of mortality due to cardiovascular diseases. This appears

Italy

Fig. 4.7

189

Two-year mortality rates by age and sex

to be true for both genders. Mortality by cancer is also falling or remaining constant, especially in the sixty to seventy- nine age range, particularly for males. During the 1990s, the reduction of the gap in mortality between males and females is especially noticeable for the cohorts born after World War II, and reflects the sharper reduction of cardiovascular diseases and cancer among males (Caselli and Egidi 2010). This trend is likely to continue in the future. Caselli and Egidi (2010) point out that the better performance of females in terms of survival probabilities masks their relative disadvantage in terms of health and functional limitations, a disadvantage that grows rapidly between sixty and seventy- nine years of age and becomes especially important after eighty years of age. They also argue that this contradiction between the longer life expectancy of females and their worse health is likely to remain in the future. An important issue is the age when functional limitations first appear. Caselli and Egidi (2010) show that this age has been increasing steadily over time, especially for women. Thus, the large gains in life expectancy observed during the last few decades have been accompanied by substantial gains in the quality of life at older ages. Despite this progress, the mere increase in the size of the elderly population implies a dramatic increase in the number of people affected by functional limitations. According to Caselli and Egidi (2010), the number of people aged sixty and above with partially reduced functional abilities would reach eight million, of whom four million would be aged eighty and above. 4.3.3

Labor Force Status and Health Conditions

We now turn our attention to the relation between labor force status and health conditions. Figure 4.9 plots the age- profile of employment rates in Italy in three different years (1977, 1990, and 2003), separately for women and men. The figure makes use of the Modello di Analisi Regionale della Spesa Sociale (MARSS) data set, which provides comparable figures on

Fig. 4.8

Mortality rates by cause of death

Italy

Fig. 4.9

191

Employment rates by age (women left, men right)

employment, unemployment, and nonemployment by age, gender, and region from 1977 to 2003 and has been constructed by ISTAT by putting together the cross- sectional information from the various waves of the labor force survey. The figure reveals several striking features of the Italian labor market. First, employment rates at older ages are much higher for men than for women: it is only in 2003 that the employment rate of women aged fifty and fifty- one reaches 50 percent, while for all other ages it is well below this percentage. Second, as widely documented elsewhere (Brugiavini and Peracchi 2004, 2007), after age fifty employment rates decline monotonically, with a sharp drop at age fifty- five and again at ages sixty and sixty- five. This is especially true for men: by age sixty, their employment rate has dropped from the value of 90 percent at age fifty to below 50 percent. Third, the 2003 profile of employment rates is entirely above the 1997 and 1990 profiles for women (i.e., there is an increasing trend in employment rates at all ages considered), but is entirely below for men (i.e., there is a declining trend in employment rates at all ages considered). To see whether there is any link between these patterns and the health conditions of workers, we examine the relationship between employment rates and mortality rates in three different years (1977, 1990, and 2003), separately for men and women (figure 4.10). For both genders, this relationship is negative, with higher mortality associated with lower employment rates, suggesting a possible link between falling employment rates and declining health status. Notice, however, that women exhibit a much lower attachment to the labor market than men, despite their higher longevity. In fact, the negative relationship between employment rates and mortality is just spurious, being driven entirely by the age trend. This is shown clearly in figure 4.11, which presents the scatter plot of employment rates and mortality rates at three different ages (fifty- five, sixty, and 65), separately for females and males. The scatter plots show that, over time, the relationship between employment and mortality rates is actually positive at all ages for men and for older women, while it is negative for younger women. This

192

Agar Brugiavini and Franco Peracchi

Fig. 4.10

Employment and mortality rates over time (women left, men right)

Fig. 4.11

Employment and mortality rates by age (women left, men right)

suggests that, at least for men and older women, one cannot associate the declining trend in employment rates to worse health conditions. In order to relate this evidence to social security programs, we compare time trends in three dimensions: (a) mortality rates, based on the ISTAT data; (b) employment rates, based on the MARSS data; and (c) pension recipiency rates (the fraction of people of a given age who receive DI or early retirement benefits), based on the INPS sample. The evidence, presented in Figures 4.12 and 4.13, should be interpreted with care, as the first two dimensions are based on nationally representative data, while the latter is a sample from the administrative archives of the main Italian institution administering public pensions. Hence, for the INPS sample, the reference population differs from the general population. Also notice that, in the INPS data, people with an early retirement pension are registered as early retirees even past the normal retirement age. To allow for gender comparisons, graphs for women and men are presented on the same scales, the left- scale for employment rates and pension recipiency rates, and the right- scale for mortality rates. The evidence from figures 4.12 and 4.13 is in line with our previous findings: mortality rates have declined steadily over time, particularly for the

Fig. 4.12 Employment rates, mortality rates, and pension recipiency rates over time (women)

Fig. 4.13 Employment rates, mortality rates, and pension recipiency rates over time (men)

Italy

195

age- groups fifty- five to fifty- nine and sixty to sixty- four, employment rates have also declined (except for the younger women), while recipiency rates of DI benefits have remained stable or increased only slightly. In contrast, recipiency rates of early retirement pensions have increased significantly over time for both men and women, at all ages. This prima facie evidence suggests that, while health conditions have improved over time (as measured by mortality), employment of the elderly has fallen, the typical exit route from the labor force being early retirement. In the next section we look in more detail at the relationship between labor force participation and participation to social security programs. 4.4 4.4.1

Exit from the Labor Force and Program Reforms Disability and Social Security Reforms versus LFP and Pathways to Retirement

The evidence discussed so far poses the important question of how the length of the working lives of Italians has adapted to the increases in longevity during the last decades. Brugiavini and Peracchi (2010) show that, not unlike other countries, the working life of Italian workers has been shrinking for some time after World War II. Figure 4.14 shows aggregate labor force participation rates based on the MARSS data.7 The time span is 1977 to 2003 and we distinguish between six age- groups (forty to forty- four, forty- five to forty- nine, fifty to fiftyfive, fifty- five to fifty- nine, sixty to sixty- four, and sixty- five to seventy). As discussed in previous contributions (Brugiavini and Peracchi 2010), aggregate labor force participation is increasing for the younger age- group (age forty to fifty), and constant or even declining for the older age- groups (sixty and above). However, there are important gender differences, as the upward trends are totally explained by female labor supply, especially for the younger cohorts. Focusing on the ages at risk of exit from the labor force (ages fifty- five and older), figure 4.15 shows that men have substantially reduced their participation over time, although a reversal of the trend can be observed after the Dini reform in 1995, especially for the fifty- five to fifty- nine age- group. In order to document the effect of the different welfare provisions on labor supply and program participation at older ages, we again use the INPS sample. These data contain two types of information. The first is the stock information on the number of workers paying social security contributions at the end of each year, along with the number of beneficiaries by type of benefit (old- age, early retirement, DI, or survivor pension).8 The second is 7. See footnote 9. 8. These beneficiaries are mostly widows who, although previously or currently working, have their husband’s survivor pensions as the main source of income.

196

Agar Brugiavini and Franco Peracchi

Fig. 4.14

Aggregate labor force participation rates by age-group

Fig. 4.15

Labor force participation rate by age-group and gender

the flow of information on the number of new pension awards each year. We focus on the period from 1990 to 2004. Figure 4.16 shows the changes over time in the composition of the INPS sample by pension recipiency status for four age- groups (fifty to fifty- four, fifty- five to fifty- nine, sixty to sixty- four, and sixty- five to sixty- nine). In constructing the graphs, we drop benefit recipients who could not be linked with their earnings history (for example, former public- sector employees or widows who had no earnings ever in their lives but currently collect survivor benefits). We also drop people who receive either “pre- pensions” ( prepensionamenti ), which are a form of early retirement available to employees of selected firms in specific industries undergoing severe crisis, or benefits that are not work- related, such as income maintenance for the elderly ( pensione sociale).9 There are two main reasons for our sample selection criteria. 9. Pre- pensions will be discussed in section 4.4.2 when looking at trends in newly awarded benefits.

Italy

Fig. 4.16

197

Program participation rates over time (both genders)

First, we are interested in the transition from work to other nonemployment states and want to focus on former (or current) workers only. Second, it is possible for people to work and collect some benefit (for example, disability benefit), or to collect more than one benefit at the same time; in these cases we assign individuals to one category or the other on the basis of an income criterion (e.g., individuals are disability- claimants if the main source of income is the disability pension). Hence we need to have information on earnings as well as benefits for each individual. At younger ages (fifty to fifty- four), recipients of either early retirement or survivor benefits represent a significant percentage of the sample, but this percentage has been falling since the year 2000, partly in response to the pension reforms and partly because of cohort effects. Program participation rates at ages fifty- five to fifty- nine and sixty to sixty- four are very interesting: for this age- group, early retirement is becoming an increasingly important exit route (at least until about the year 2000), replacing the old- age exit route. This is explained by the tighter entry- age conditions gradually applied after the Amato (1992) reform, which are particularly biting for women (see table 4.1 in section 4.2). For the older age- group (sixty- five to sixty- nine) the pattern is similar but old- age pensions remain the most important exit route.

198

Agar Brugiavini and Franco Peracchi

The DI pensions are not particularly relevant and show no clear pattern over the period considered. 4.4.2

Effects of Pension Reforms on Labor Force Participation

Because the timing and content of the Italian reforms of the 1990s were hardly anticipated (this is especially true for the Amato reform), these pension reforms may be regarded as natural experiments that provide the exogenous variation that is needed in order to evaluate the effects of specific policy changes on labor force participation. By focusing the attention on the 1992 and 1995 pension reforms, we compare labor market participation before and after the reforms for different cohorts. The data set that we use is again MARSS. Results are presented in table 4.2 in the form of exit rates (in percentages), which are obtained by taking the between- years rate of change in the labor force participation for a given cohort and gender. The first column (rate 1989 to 1986) shows the exit rates before the main reforms of 1992 and 1995, while the last column shows the same exit rates well after the reforms.10 While we cannot draw sharp conclusions on the basis of these results, we can distinguish two significant effects of the two main pension reforms of the 1990s, the Amato (1992) and the Dini (1995) reforms. The first, which we call an “announcement effect,” consists of a clear “run to retirement.” We observe this effect immediately after the pension reforms were announced and approved by the Parliament. Since the reforms were characterized by more restrictive eligibility conditions, many workers who satisfied the eligibility requirement for retirement in 1992 and 1995 decided to retire as soon as possible because of the fear of being trapped in their jobs. Such an effect is visible for all the cohorts that were between fifty- five and sixty- three years of age in 1992 to 1995, both men and women. Table 4.2 documents that the exit rate of men in the cohort born between 1932 and 1934 (aged fiftyeight to sixty in 1992 and sixty- one to sixty- three in 1995) was 45.3 percent between 1992 and 1995, that is almost 7 percent higher than the exit rate of the cohort 1929 to 1931 when reaching the same age (between 1989 and 1992). Also, between 1992 and 1995 and between 1995 and 1998 the exit rate of men aged fifty- eight to sixty (cohorts 1935 to 1937 and 1938 to 1940) was 29.0 percent and 30.2 percent respectively, versus 21.8 percent of the cohort 1932 to 1934 when of the same age. Similar effects are also observed for women. For example, the exit rate of women aged fifty- eight to sixty in 1995 (fifty- five to fifty- seven in 1992) is 7 percent higher than for women of the same age in 1992. The second effect is a direct “mechanical effect” of the change in the rules 10. There have been some minor changes in 1997 and in the following years that have tightened eligibility conditions for early retirement. However, while the 1992 reform (and to some extent also the 1995 reform) was unexpected, these changes were by and large predictable.

Italy Table 4.2

Men Men Men Men Men Men Men Women Women Women Women Women Women Women

199

Exit rates by cohorts and gender (percent)

Cohort

Exit rate 89–86

Exit rate 92–89

Exit rate 95–92

Exit rate 98–95

26–28 29–31 32–34 35–37 38–40 41–43 44–46 26–28 29–31 32–34 35–37 38–40 41–43 44–46

40.59 24.78 20.24 9.10 1.81 –0.21 –0.39 41.82 32.92 29.79 13.09 4.49 2.50 –4.70

41.23 37.82 21.85 15.46 11.79 5.63 1.72 45.49 42.51 28.33 27.25 13.24 5.30 6.41

53.64 47.31 45.30 29.01 20.67 16.08 8.83 56.68 47.03 52.39 35.71 30.24 15.99 6.51

55.30 51.47 41.77 37.87 30.23 23.19 13.16 43.41 51.90 35.83 45.20 32.71 21.29 14.00

Exit rate 01–98

59.17 47.78 44.48 31.87 25.89 22.12 47.20 52.06 47.59 47.12 31.63 20.06

Source: The MARSS data on labor force, ISTAT. Notes: The exit rate is obtained, for example, as (lfpr86-lfpr89)/lfpr86, where lfpr86 is the labor force participation rate in 1986. Bold grey characters: exit rates between age-groups 49–51 and 52–54. Grey characters: exit rates between age-groups 52–54 and 55–57. Bold characters: exit rates between age-groups 55–57 and 58–60. Italic characters: exit rates between agegroups 58–60 and 61–63. Underlined characters: exit rates between age-groups 61–63 and 64–66.

that forces workers to delay retirement. This effect has a smaller order of magnitude if compared with the announcement effect, and it also appears more diluted on a longer period of time over the years that follow the implementation of the reforms. The exit rates fall after 1996 for the younger age- groups and increase for the older ones. We can see that for men aged sixty- one to sixty- three the exit rate decreases in 1998 from 45.3 percent to 37.87 percent and further to 31.87 percent in 2001. A similar pattern is observed for men in the age- group fifty- eight to sixty and for women in the age- groups fifty- eight to sixty and fifty- five to fifty- seven. Basically what one can see in table 4.2 is that a decreasing number of individuals in these groups (cohorts) satisfy the new, more restrictive eligibility conditions for retirement after the year 1996. This interpretation is also supported by the increased exit rates of men aged sixty- four to sixty- six and of women aged sixty- one to sixty- three after 1998. One has to be cautious in drawing a final conclusion on the effects of these pension reforms on labor market participation as, at the same time, other underlying trends could partly interact with our natural experiment, especially in the case of women. Further, our aggregate data do not properly account for the variety of possible exit routes that individuals experience, which do not necessarily respond to the same incentives brought about by the reforms.

200

Agar Brugiavini and Franco Peracchi

Fig. 4.17

New entries by benefit type (women left, men right)

However, one general conclusion that we can draw from this very descriptive evidence is that pension reforms in Italy created huge disparities between cohorts, as also documented by the contributions of Attanasio and Brugiavini (2003) and Bottazzi, Jappelli, and Padula (2006). Instead of focusing on changes in the stocks, figure 4.17 looks for evidence of the effects of the pension reforms of the 1990s by focusing on changes in the annual flows of newly awarded pensions by benefit type. We refer to these flows as “new entries” into the pool of pensioners. The graph uses the counts of individuals who, in each year, receive for the first time their pension from the three main INPS funds (namely FPLD and the Artisans and Traders Funds), obviously excluding from the sample those who never receive any benefit (which leaves us with approximately 145,000 individuals). The percentages are based on the date when the person received her or his first pension payment, which is the date at which the individual is recorded as entering the archive. The graph describes the routes of entry into the public pension system, but does not allow us to keep track of subsequent changes from one type of benefit to another. We observe that before 1993, that is, before the Amato reform, the prevailing channel for entering the welfare system was old- age retirement. After this date, old- age retirement shows decreasing percentages while early retirement increases until 1999, when the two exit routes reach some sort of steady state. Figure 4.17 looks at men and women separately. In the case of men, oldage pensions after the reforms represent only about 22 percent of the total versus about 65 percent of entries through early retirement pensions. It is also striking to see a spike of early retirement entries in 1993, just after the Amato reform, followed by a dip in 1994 and again a spike in 1995, the year of the Dini reform. In our data we can also distinguish a particular type of early retirement, called pre- pension, available only to the employees of selected firms in specific industries undergoing severe crisis. This particular exit route, which was very popular in the 1980s, survived through the 1990s

Italy

201

but practically disappeared after the year 2001, when it was replaced by early retirement pensions. Finally, DI pensions are steadily declining over time, from a starting value of approximately 20 percent in 1985 to about 10 percent in 2004. This is because the strict entry requirements introduced in 1984 ended up being quite effective. In 1993, when the percentage of new awards for early retirement diminishes abruptly, disability pensions show a spike together with old- age pensions. Because the results presented in the previous figures may be affected by cohort effects, figures 4.18 and 4.19 look separately at two birth cohorts: people born in 1930 to 1939 and people born in 1940 to 1949. It is clear from the figures that the two cohorts followed very different routes in exiting the labor force. The earlier cohort exited predominantly through the

Fig. 4.18 New entries by benefit type and birth cohort (men born 1930 to 1939 top, men born 1940 to 1949 bottom)

202

Agar Brugiavini and Franco Peracchi

Fig. 4.19 New entries by benefit type and birth cohort (women born 1930 to 1939 top, women born 1940 to 1949 bottom)

old- age pension route, although the DI pension route was also relevant. The later cohort, instead, exited almost entirely through the early pre- retirement route, especially in the case of men. This vividly illustrates the effect of the Amato and Dini reforms. Notice that, although the trends for men and women are similar, the levels are quite different, as the early retirement route is much less important for women. This difference reflects three main factors. First, labor market participation for women in the two cohorts is very different. Second, the requirements for an old- age pension are different for men and women (age sixty for women, age sixty- five for men). Third, women hardly qualify for early retirement, as this benefit is awarded conditionally on a seniority requirement, that is, on the basis of the number of years of contributions. Because

Italy

203

Italian female workers typically have short contributive histories (Boeri and Brugiavini 2009), this exit route is of secondary importance for them. Another way of displaying the same information is to compute the rate of change in the newly awarded benefits. We do this in two ways: first with respect to the base year 1985, then year- by- year. While the first way is affected by the composition of the sample in the base year, the second way is strongly affected by some important changes taking place in the reform years, which cause huge variations in the denominator of the rate itself, making it very hard to properly interpret the hazards. Figure 4.20 presents the percentage rates of change in entries in the public pension and disability insurance in the index- form, using the year 1985 as the base year.11 We used only the sample of individuals who were receiving their pension from the three main INPS funds (FPLD, Artisans, and Traders). We restrict the sample to birth- cohorts 1915 to 1950; that is, we consider individuals aged thirty- five to seventy in 1985, so that these individuals are potentially workers in the base year. The two dotted lines mark the years 1992 and 1995, which are the years of the two major pension reforms in Italy. We observe that entries in the pension system through prepensions diminish constantly starting in 1986 (except for the year 1992). On the other hand, entries through early retirement pensions exhibit a steady increase with two important peaks: one in 1992 and the other one in 1994. In addition, the old- age retirement route shows an almost constant path with a peak in 1992 followed by a slight decrease in 1993. It is important to note that a feature that is common to all three indexes is the changes in 1992 and in 1994. This is because the reforms had a strong announcement effect and Italy witnessed a run to retirement. Many workers who fulfilled the less restrictive retirement conditions in 1992 (or 1994) decided to retire immediately in order to avoid the risk of being trapped in their job, through whatever mode of retirement. At that point the easiest way out was through early retirement. The same features characterize the paths of retirement for men, while for women the variability is larger, but this should be explained by the different participation of women to the labor market. Unfortunately, due to the fact that the denominator changes abruptly from one year to the next, it is hard to interpret these figures, particularly when the absolute numbers are small, as in the case of pre- pensions. The important spikes are observed in 1992 and 1994 as expected.

11. The rates of change in retirement in the year n with respect to 1985, are: Pn − P1 9 8 5 × 100, P1 9 8 5 where Pn represents the number of retirees in year n while P1985 represents the number of retirees in the base year.

Fig. 4.20

Rate of change of entry into the pension system (women left, men right)

Italy

4.4.3

205

Pathways into Retirement

More compelling evidence on the effect of the different eligibility conditions applying to the different exit routes can be obtained by exploiting the panel dimension of the INPS sample. In order to follow the different pathways we focus on those individuals registered with INPS (i.e., paying contributions) who were working in 1985. Similar to the previous figures, we restrict the sample to individuals who were receiving some form of benefit from the three main INPS funds, and we add the condition that they were working in the year 1985 (approximately 78,000 individuals). Then, starting in 1985, we follow this sample during their working lives until exiting from the employment status and entering the social security archive. Figure 4.21 shows the pathways as percentages of individuals who are in any of the mutually exclusive states: employees, early pensions, old- age pensions, pre- pensions, and so forth, by year and age- group. Figure 4.22

Fig. 4.21 Pathways into retirement: Participation to the different welfare programs for the panel sample of people working in 1985, by age-group (both genders)

206

Agar Brugiavini and Franco Peracchi

Fig. 4.22 Pathways into retirement: Participation to the different welfare programs for the panel sample (men left, women right)

compares pathways across genders. Since in some cases the same individual could be employed and receive a benefit (or could receive more than one benefit), we impute individuals to a labor market status on an income basis: the individual is defined as old- age retired if his or her old- age pension provides the highest income. Hence, in figures 4.21 and 4.22 we are not effectively looking at transitions from one state to another state, but simply showing participation in the different programs in a panel sample, conditional on being employed in 1985. It should be stressed that in our data an early retirement benefit is always recorded as such even if the beneficiary becomes older (aged sixty- five or above), while disability benefits are mostly automatically converted into oldage pensions when the disabled person reaches age sixty- five. This explains why, even at older ages, early retirement is a prevalent mode of being nonemployed. Once again we find that early retirement substitutes for old- age pensions as eligibility requirements for old- age pensions become more stringent. Disability pensions are not particularly relevant and they are also quite stable over time. Proper transitions across states can be measured in the panel sample through transition matrices. Table 4.3 shows transition matrices for two groups of individuals distinguished by year of birth. Transitions are measured over a period of ten years. Table 4.3 clearly shows that in Italy the main exit route is directly via retirement (old- age pension or early retirement pension). Very few workers go into disability, and in fact, not many transit from work into retirement via disability. 4.5

Trends in Labor Force Status and Program Participation

To complete the analysis we provide historical data on individuals employed, unemployed, and not in the labor force. We make use of the

Italy Table 4.3

207

Transition matrices (percent)

Employment

Job pension

Disability pension

Social pension

Non-contr. pension

Survivor pension

A. Cohorts 1931–1940 Transitions 85–95 Employment Job pension Disability pension Social pension Non-contr. pension Survivor pension Transitions 95–04 Employment Job pension Disability pension Social pension Non-contr. pension Survivor pension

97.10 0.00 0.00 0.00 0.00 0.00

0.85 100.00 0.00 0.00 0.00 0.00

1.70 0.00 100.00 0.00 0.00 0.00

0.00 0.00 0.00 0.00 0.00 0.00

0.32 0.00 0.00 0.00 100.00 0.00

0.35 0.00 0.00 0.00 0.00 0.00

97.10 0.00 0.00 0.00 0.00 0.00

0.85 100.00 0.00 0.00 0.00 0.00

1.70 0.00 100.00 0.00 0.00 0.00

0.00 0.00 0.00 0.00 0.00 0.00

0.32 0.00 0.00 0.00 100.00 0.00

0.35 0.00 0.00 0.00 0.00 0.00

B. Cohorts 1941–1950 Transitions 85–95 Employment Job pension Disability pension Social pension Non-contr. pension Survivor pension Transitions 95–04 Employment Job pension Disability pension Social pension Non-contr. pension Survivor pension

55.68 0.00 0.00 0.00 0.00 0.00

39.05 100.00 10.00 0.00 0.00 66.67

4.00 0.00 80.00 0.00 0.00 0.00

0.08 0.00 0.00 0.00 0.00 0.00

0.16 0.00 0.00 0.00 100.00 0.00

1.02 0.00 10.00 0.00 0.00 33.33

9.85 0.06 1.42 0.00 0.00 1.05

83.42 99.03 35.65 12.50 8.82 18.95

2.89 0.03 60.57 6.25 2.94 0.00

0.80 0.00 0.95 75.00 0.00 0.00

0.92 0.03 0.79 0.00 82.35 0.53

2.12 0.86 0.63 6.25 5.88 79.47

labor force survey data provided by ISTAT through the MARSS time series. Although we cannot gain much detail of the type of nonemployment, the data allow us to distinguish different age- groups and gender for the three main components. 4.5.1

Trends in Labor Force Status

As previously pointed out, employment rates of men are much higher than for women, and this is a rather stable pattern for all age- groups (figure 4.23). However, younger cohorts of women have a higher employment rate and show a marked difference with respect to older cohorts. The other striking feature of figure 4.23 is that, at age sixty to sixty- four, the employment rate of men goes from 40 percent in 1977 to 30 percent in 2004. Unemployment rates are much less stable, particularly for the youngest age- group, mainly

208

Agar Brugiavini and Franco Peracchi

Fig. 4.23

Trends in labor force status by gender and age-group

as a result of business cycle fluctuations (see Brugiavini and Peracchi 2010). Comparing the trends in “out of the labor force” and “unemployed” rates it is clear that unemployment is not a major source component of nonemployment. Indeed the unemployment rate figures are very small, especially for the age- group sixty to sixty- four: the evidence presented in the previous sections makes clear that retirement (particularly early retirement) has been the main explanation for exits from the labor force. Also, there seems to be nonsubstitutability between unemployment and other forms of exits at older ages.

Italy

4.5.2

209

Trends in Recipiency of DI Benefits

In order to provide an explanation for the nonemployment patterns at older ages we look at trends in recipiency of DI benefits: we have to resort to a different data set (the INPS sample) because the labor force survey does not contain details on the type of nonemployment. Given the compelling evidence on the relevance of early retirement we present a comparison of program participation rates to disability and to early retirement. Because for the early years older individuals are underrepresented, we restrict the analysis to the period 1990 to 2004. First it should be noted that participation rates for disability are very small. They are mostly declining over time, apart from the age- group sixty to sixty- four (figure 4.24). Participation rates to early retirement are of considerable size and increasing at all ages. This is further evidence that Italian workers transit from work to nonemployment mainly through early retirement. A main objective of this chapter is to relate the marked changes in health conditions and survival with the labor market behavior of older workers in Italy. Because we do not have available long time series documenting trends in health conditions, we propose the following comparisons: (a) trends in mortality versus trends in disability (or early retirement), (b) heights of young men (at the medical examination for army conscripts) versus disability take up and early retirement take up, and (c) mortality rates by cause of death versus disability rates. The period of observation is 1990 to 2003. Figure 4.25 shows once again that mortality rates are steadily declining during the period, suggesting that health conditions of the elderly improve. At the same time, disability participation rates also decline for this agegroup. However, we have shown that health improvements are not leading to higher participation in the labor market: in fact, men in that age- group exhibit strongly increasing trends in early retirement pension take up (which

Fig. 4.24 Male participation rates to disability (left) and early retirement (right) by age-group

210

Agar Brugiavini and Franco Peracchi

Fig. 4.25

Male mortality rates versus disability (left) and early retirement (right)

are approximately ten times higher than disability rates). For the age- group sixty to sixty- four disability rates are also increasing in the sample period, hence for this age- group more and more workers have left the labor market at a time when their life expectancy increases. The second comparison is based on the idea that different cohorts started their lifetimes with different health stocks (as measured by height). We have available height measures of young men at the army- conscription medical test for several years (hence, cohorts), all recorded at age nineteen (see A’Hearn, Peracchi, and Vecchi 2009).12 We want to establish if the initial health stock could explain the labor market behavior later in life; of course, we cannot control for the different lifestyles and health shocks experienced by these men between age nineteen and the age at which they left the labor market. Figure 4.26 shows that while the initial health stock increases over time for all cohorts, disability rates exhibit very unclear patterns: decreasing over time for most age- groups apart from the oldest group, sixty to sixty- four. 12. For the cohorts of men included in the sample period, serving in the army was compulsory.

Italy

Fig. 4.26

211

Male height at age nineteen (in cm) versus disability rates

It is hard to draw a conclusion on the actual relationship between these patterns. Figure 4.27 provides the same evidence relating heights to early retirement participation rates. Once again we find that while health conditions (as measured by the initial health stock) improve, individuals tend to leave the labor market at higher rates. Figure 4.28 looks at mortality rates by cause of death and early retirement rates. This is to assess whether gains due to changes in specific conditions (such as heart conditions) lead to changes in labor market behavior. Since we have provided plenty of evidence on the scarce significance of disability pensions, we focus on early retirement pensions, which are the most important exit route in Italy in the sample period. Once again the correlation is puzzling as reductions in mortality rates for important conditions (such as heart conditions) do not imply changes in early retirement participation. 4.6

Conclusions

This chapter describes the relationship between the characteristics of the Italian social security system (and its reforms) and the working life of

212

Agar Brugiavini and Franco Peracchi

Fig. 4.27

Male height at age nineteen (in cm) versus early retirement rates

individuals by considering a relevant nexus, which had been previously neglected, of these variables with trends in health conditions. We first analyze the relevant aspects of the social security system, by focusing on the main provisions of the Italian welfare system for old- age insurance. We find that eligibility conditions for access to early retirement and old- age benefits have shaped the institutional landscape of the past decades. At the same time we present the main demographic trends in relation to mortality and health along with labor market trends (employment, unemployment, and labor force participation). The evidence points to a potential paradox: the spectacular increase in longevity experienced by the Italian population in the past decades is correlated with increasing exits from the labor force of older workers. The historical data on mortality trends document an important increase in life expectancy and longevity. Mortality fell dramatically in Italy between 1906 and 2007. The decline in mortality evolved differently for men and women; more precisely, it was more important for women in the early decades. As a result, life expectancy increased considerably. For example, at the age of sixty it augmented by 7.6 years for men and 11.5 years for women,

Italy

Fig. 4.28

213

Male mortality rates by cause of death versus early retirement rates

while at the age of eighty it improved by 3.3 years for men and 4.9 years for women. A decreasing path of the mortality is also observed at the regional level, but we remark different patterns between the three main geographical areas (North, Center, and South). In the long run these differences led to an almost total compensation of the initial existing gaps between these regions. Potential explanations for these important improvements in longevity are a better quality of the health services and a more responsible behavior at the individual level in terms of lifestyle. The reduction in mortality is to a large extent the result of the decline in mortality due to cardiovascular and respiratory diseases for men and to diabetes for women. When looking at labor market behavior in Italy we observe that labor force participation exhibits a declining path for all workers in the age- groups fiftyfive and over, though women show very different trends from men because of cohort effects in the period 1997 to 2003. At the same time, participation in the different social security programs (old age, early retirement, disability, etc.) is dominated by retirement. For the age- groups sixty to sixty- four and fifty- five to fifty- nine old- age retirement exhibits a decreasing trend, which starts in 1992 (the date of a major pension reform). The old- age exit route is

214

Agar Brugiavini and Franco Peracchi

gradually substituted by the early retirement route, while disability increases very slightly in the same period of time. The most striking result on the effects of the reforms of the social security system in Italy is provided by the trends in the new entries in the public pension and disability insurance, as well as by the pathways to retirement. We show that after the 1992 pension reform the prevailing exit route switched from old- age pension to the early retirement pension, especially for men who had sufficient seniority to become eligible for early retirement. The same conclusion is reached by analyzing the pathways to retirement and the transition matrices that we computed on the panel dimension of our data. This body of evidence raises the questions of what the relationship is between health status (longevity) and the length of the working life. On the one hand, there seems to be a negative correlation between the employment rate and the mortality rate, which is much more important for men with respect to women. On the other hand, the steady decline of the mortality rates along time is accompanied by a consistent increase in the early retirement rates. Disability does not seem to play an important role in Italy.

References A’Hearn, B., F. Peracchi, and G. Vecchi. 2009. “Height and the Normal Distribution: Evidence from Italian Military Data.” Demography 46:1– 25. Attanasio, O. P., and A. Brugiavini. 2003. “Social Security and Households’ Saving.” Quarterly Journal of Economics 118 (3): 1075– 119. Boeri, T., and A. Brugiavini. 2009. “Pension Reforms and Women Retirement Plans.” Journal of Population Ageing 1:7– 30. Bottazzi, R., T. Jappelli, and M. Padula. 2006. “Retirement Expectations, Pension Reforms, and Their Impact on Private Wealth Accumulation.” Journal of Public Economics 90 (12): 2187– 212. Brugiavini, A. 2009. “Welfare Reforms and Labour Supply in Italy.” Working Paper Series 2009:29. Uppsala: IFAU-Institute for Labour Market Policy Evaluation. Brugiavini, A., and V. Galasso. 2004. “The Social Security Reform Process in Italy: Where Do We Stand?” Journal of Pension Economics and Finance 3 (2): 165– 95. Brugiavini, A., and F. Peracchi. 2004. “Micro-Modeling of Retirement Behavior in Italy.” In Social Security Programs and Retirement around the World: MicroEstimation, edited by J. Gruber and D. A. Wise, 345– 99. Chicago: The University of Chicago Press. ———. 2007. “Fiscal Implications of Pension Reforms in Italy.” In Social Security Programs and Retirement around the World: Fiscal Implications of Reforms, edited by J. Gruber and D. A. Wise, 253– 94. Chicago: The University of Chicago Press. ———. 2010. “Youth Unemployment and Retirement of the Elderly: The Case of Italy.” In Social Security Programs and Retirement around the World: The Relationship to Youth Employment, edited by J. Gruber and D. A. Wise, 167– 215. Chicago: The University of Chicago Press. Brugiavini, A., F. Peracchi, and D. Wise. 2003. “Pensions and Retirement Incentives,

Italy

215

A Tale of Three Countries: Italy, Spain and USA.” Giornale degli Economisti e Annali di Economia 61:131– 70. Caselli, G., and V. Egidi. 2010. “Una Vita Più Lunga e Più Sana. Riflessioni sul Passato per Immaginare il Futuro.” Unpublished manuscript. Caselli, G., F. Peracchi, E. Barbi, and L. M. Lipsi. 2003. “Differential Mortality and the Design of the Italian System of Public Pensions.” Special issue, Labour 17: 45– 68. Istituto Italiano di Statistica (ISTAT). 2010. “Trattamenti Pensionistici e Beneficiari al 31 Dicembre 2008.” Available at: http://www.istat.it/salastampa/comunicati/ non_calendario/20100611_00/. Organization for Economic Cooperation and Development (OECD). 2007. Social Expenditure 1980– 2003. Interpretative Guide. Available at: http://stats.oecd.org/ OECDStatDownloadFiles/_OECDSOCX2007InterpretativeGuide_En.pdf. ———. 2009. Social Policy Division, Directorate of Employment, Labour and Social Affairs, Country Chapter, Benefits and Wages. Available at: http://www.oecd.org/ els/social/workincentives. Ragioneria Generale dello Stato (RGS). 2009. Relazione Unificata sull’Economia e sulla Finanza Pubblica. Available at: http://www.rgs.mef.gov.it/VERSIONE-I/ Finanza-Pu/Relazione- 1/2009/index.asp.

5 Disability Programs, Health, and Retirement in Denmark since 1960 Paul Bingley, Nabanita Datta Gupta, and Peder J. Pedersen

5.1

Introduction

Over the last fifty to sixty years Danish society has undergone quite a fundamental change in terms of the share of the population eighteen to sixty- five years old being provided for in one of several transfer programs. The increase has been from about 5 percent to 20 percent of the population in the labor market active age- group. This development has occurred throughout quite different cyclical situations and it can be characterized by changes in programs that existed back in the 1950s and by the introduction of new programs—some permanent and some temporary. In the present chapter the main emphasis is given to Social Disability Pension (SDP) along with other programs for early retirement from the labor force. The SDP is not an insurance- based program but is financed from general tax revenues with eligibility originally depending on medical, and later on a mix of medical and social criteria. The long- run development in take-up of SDP reflects a number of different factors. Over the last fifty to sixty years a number of reforms of SDP have been enacted with different motives and impact on the contents and aspirations of the program. Over this time, the economy has undergone a number of big cyclical swings with Paul Bingley is research professor at SFI-The Danish National Centre for Social Research. Nabanita Datta Gupta is professor in the Department of Economics and Business at Aarhus University. Peder J. Pedersen is professor in the Department of Economics and Business at Aarhus University. We are grateful to participants at several ISS meetings for comments on previous drafts of the paper. Thanks to Chalotte Bøgesvang and Katrine Pedersen for excellent research assistance. Financial support was provided by the Danish Social Sciences Research Council (grant FSE- 09-063859). For acknowledgments, sources of research support, and disclosure of the authors’ material financial relationships, if any, please see http://www.nber.org/chapters/ c12386.ack.

217

218

Paul Bingley, Nabanita Datta Gupta, and Peder J. Pedersen

the possible consequence that take-up of SDP to some extent might be influenced by movements in unemployment. Finally, as the primary criteria for entry to SDP are medical, take-up could be affected by trends in mortality or in the prevalence of more serious health conditions. In the following, we present in section 5.2 the background for further analyses in the subsequent sections by looking into some long- run historical trends. We focus on some aggregate long- run trends in SDP take-up, in unemployment, in labor force participation, and in mortality. Finally, section 5.2 outlines the main questions or hypotheses being subsequently analyzed. The purpose of section 5.3 is to introduce and describe trends in mortality over the period since 1960 as one element in an attempt to understand the development in the relationship between health—with mortality as the ultimate indicator—SDP and trends in the labor market regarding employment and unemployment by gender and age- groups. Section 5.4 introduces a mix of health indicators over all or part of the period since 1960 consisting of register- based data for the years since 1980 and indicators of self- assessed health for a number of years between 1987 and 2005. The main purpose in section 5.5 is to describe the specific programs with the main emphasis on SDP and the reforms that were enacted since the 1960s. The section further contains descriptions of other early retirement programs that have had an impact on labor force participation for older workers. For these programs, the description focuses on the motives behind the introduction and subsequent reforms, followed by data for the take-up of each of the programs. Further, section 5.5 illustrates the pathways from the labor force to retirement as well as evidence of substitution between programs. Section 5.6 describes trends in employment and unemployment with a special emphasis on the timing of permanent as well as temporary policy changes in the area of early retirement. Next, the purpose of section 5.7 is an attempt to identify the relationship—or the lack of such—between the changes we find in the labor market regarding employment for the age- groups most relevant in an early retirement perspective, and the trends we have found in the earlier sections looking into the development in mortality and health along with policy reforms. Finally, section 5.8 concludes and summarizes. 5.2

Historical Overview—Some Aggregate Trends

In this section we briefly summarize some aggregate trends before moving on to more disaggregate analyses. It is well known that a gray zone may exist between unemployment and disability insurance. The idea of a competing risk- setting between SDP and unemployment programs is the topic in Black, Daniel, and Sanders (2002) and Autor and Duggan (2003). The possible interaction between disability insurance (DI) and labor supply is the topic in Gruber (2000) and Campoliteti (2004) with focus on the US labor market. In Bratsberg (1999) the focus is on DI in the setting of a Scandinavian welfare

Disability Programs, Health, and Retirement in Denmark since 1960

219

state. In a recent contribution, Bratsberg, Fevang, and Røed (2010) takes up explicitly the relationship between unemployment and DI in the Norwegian welfare state. Many studies in this area take the specific approach of analyzing the consequences of big firm closures with focus on what happens to older laid off workers. When panel data are available the laid off workers can be followed through an initial phase of unemployment to either a new job or to an exit to DI or another program for early retirement. Other obvious possibilities of program substitution are between SDP and alternative programs for early retirement beyond unemployment, especially programs without the medical or social eligibility criteria valid for SDP. Figure 5.1 shows the aggregate profile in Denmark 1966 to 2008 in the unemployment rate and in the number of people in the SDP program relative to the population eighteen to sixty- five years old. There is no obvious relationship between SDP and unemployment at the aggregate level. The unemployment rate follows an inverted U profile peaking slightly above 12 percent in 1994. The SDP ratio shows a completely different profile, moving for most of the period between 6 and 8 percent without any simple correlation with the business cycle represented by the unemployment rate. It should be emphasized, however, that other programs were introduced in the period as pathways out of the labor force for older workers. Behind the aggregate lack of any relationship shown in figure 5.1, the unemployment SDP interaction is more complicated when we consider program substitution. The next aggregate profiles shown in figure 5.2 are the mortality rates for women and men aged sixty- five years over the last fifty years. Roughly, mortality for the sixty- five- year- olds is constant until a decline is seen from

Fig. 5.1 to 2008

Unemployment rate and DI/population share (eighteen to sixty-five), 1966

Source: Statistics Denmark.

220

Fig. 5.2 to 2008

Paul Bingley, Nabanita Datta Gupta, and Peder J. Pedersen

Mortality rate per 100,000, men and women, sixty-five years old, 1961

Source: Statistics Denmark.

Fig. 5.3 Labor force participation rates, sixty to sixty-four years old, men and women, 1972 to 2008 Source: Statistics Denmark.

around 1990. Here we have, at least at the aggregate level, the same lack of any simple relationship between mortality and the SDP share. The final aggregate profile is the labor force participation rates for the “critical” age- group in the present context, the sixty to sixty- four- year- olds shown in figure 5.3 for women and men, using consistent data since 1972.

Disability Programs, Health, and Retirement in Denmark since 1960

221

Looking at the profile for men it is obvious that factors other than SDP and mortality are behind the kinks and trends. We return to this in section 5.6. For women the profile appears much more smooth in reflecting that the impact from program innovations, to a big extent, is counteracted by strong cohort increases in female labor force participation. 5.3

Mortality Trends

The main emphasis in the present chapter is on the interaction between measures of health, DI(SDP) take-up, and labor market performance, including the impact of policy reforms and changes in the area of early retirement. As an introduction to the treatment of SDP and alternative retirement programs, we begin with a brief description of available indicators of population health in Denmark over recent decades. In this section the focus is on mortality as an indicator of the general health status in the population. The reduction in mortality over the last fifty years is among the lowest in the Organization for Economic Cooperation and Development (OECD) area. We show the mortality rates for men and women at ages fifty- five, sixty, and sixty- five in figures 5.4 and 5.5. It is evident that there were essentially no gains in this area until about 1990. From then on we observe a fairly strong decline in mortality at all three ages, most pronounced for the sixty- fiveyear- olds. The level is higher for men at all three ages and the reduction in mortality in absolute as well as in relative terms is greater for men.1 The average retirement age has gone down in Denmark in the period since 1960. It is interesting to relate this to the increase in longevity shown in figures 5.6, 5.7, and 5.8. Figure 5.6 shows mortality by age, from forty to eighty- five, for men and women in 1960 and nearly fifty years later in 2008. Figure 5.6 illustrates by gender the age at which mortality is equal to 2 percent in 1960 and in 2008, respectively. For men this occurs in 2008 at age sixty- seven instead of sixty- two, and for women mortality reaches this level at age seventy- one in 2008 compared with age sixty- six back in 1960. So, for both men and women the 2 percent mortality level has moved up with five years of age since 1960. From figures 5.4 and 5.5 we know that this increase occurred in the last fifteen to twenty years. Figures 5.7 and 5.8 illustrate the decline in mortality in a different way. From 1950 to 2009, we show which age a person should have in each of the years to have the same mortality as a sixty- year- old, respectively a sixtyfive- year- old person in 1960. For men in figure 5.7 and for women in figure 5.8 we find that the overall decline in mortality after 1990 in this illustration 1. Data are available for expected lifetime at birth back to 1840. It is interesting to note that the female- male difference in expected longevity is about two years for about the first 100 years of observations. It then increases from a minimum slightly below two years in the first half of the 1920s to a peak of six years in the second half of the 1980s. From then on the difference in expected lifetime goes down to 4.5 years in 2008.

222

Fig. 5.4

Paul Bingley, Nabanita Datta Gupta, and Peder J. Pedersen

Male mortality rate per 100,000, 1961 to 2008

Source: Statistics Denmark.

Fig. 5.5

Female mortality rate per 100,000, 1961 to 2008

Source: Statistics Denmark.

corresponds with an increase to, respectively, the mid- sixties or to around age seventy for having the same mortality at the end of the period as a sixtyyear- old, and respectively, a sixty- five- year- old had in 1960. Finally, we summarize in this section in figures 5.9 and 5.10 how four major causes of death have evolved since 1977 for sixty to sixty- four- yearold men and women. For men, it is evident from Figure 5.9 that a strong decline in myocardial infarction and related diseases is the main factor

Disability Programs, Health, and Retirement in Denmark since 1960

Fig. 5.6

223

Two-year mortality rates by age and gender, 1960 and 2007, 2008

Source: Statistics Denmark.

Fig. 5.7

Ages of equal mortality probability for men, 1960 to 2009

Source: Statistics Denmark.

behind the decline in mortality since the years around 1990. The number of deaths due to cancer has also gone down, although not as much as myocardial infarction. The two other major causes, strokes and related diseases and diseases in the respiratory system, have been fairly stable over the period. For women, figure 5.10 shows a different picture. The number of deaths due to cancer is fairly stable over the period while myocardial infarction goes

224

Fig. 5.8

Paul Bingley, Nabanita Datta Gupta, and Peder J. Pedersen

Ages of equal mortality probability for women, 1960 to 2009

Source: Statistics Denmark.

Fig. 5.9 Four major causes of death per 100,000, men aged sixty to sixty-four years old, 1977 to 2005 Source: Statistics Denmark.

down as found for men. Comparing causes of death for men and women sixty to sixty- four years old, we find excess mortality for men due to cancer, myocardial infarction and related diseases, and strokes and related diseases. For strokes, excess mortality is constant since 1977. For cancer it is falling until no excess mortality was found by the end of the period, and for

Disability Programs, Health, and Retirement in Denmark since 1960

225

Fig. 5.10 Four major causes of death, women aged sixty to sixty-four years old, 1977 to 2005 Source: Statistics Denmark.

myocardial infarction the decline is much stronger for men. There is only one exception to this pattern, that is, for diseases of the respiratory system, initially with a clear excess mortality for men but with a shift to female excess mortality from around 1990. 5.4

Trends in Population Health

It is straightforward to find reliable mortality data. It is more difficult to find long, consistent time series for the topics in focus in this section, that is, self- reported health and register data- based health indicators. Looking first at indicators for self- reported health, we have two sources that are both used in this section. The first source is a survey run by the National Institute of Public Health (NIPH), collected five times between 1987 and 2005. The other source is the Danish panel in the European Community Household Panel (ECHP), collected annually from 1994 to 2001. The share reporting their health status as “good” in the NIPH survey seems to be at the same level as the aggregate share of respondents finding their health status “very good” or “good” in the ECHP survey. In figure 5.11 we show the results from the NIPH survey separately for women and men, aged twenty- five to forty- four, and forty- five to sixty- four years, respectively. The surveys cover a time span of eighteen years so—with some caution—it seems we can conclude that satisfaction with health status is falling or about constant for the twenty- five to forty- four- year- olds and seems to be increasing for the forty- five to sixty- four- year- olds. A tentative

226

Paul Bingley, Nabanita Datta Gupta, and Peder J. Pedersen

Fig. 5.11

Share with self-reported health “good”

Sources: OECD Health Data, June 2010; National Institute of Social Health.

conclusion is that self- reported health (and mortality, cf. section 5.3) seems to become better for the groups most relevant for SDP and the other early retirement programs being introduced in the period in focus here. When we compare with the shorter period covered by the ECHP, we find in figure 5.12 approximately the same trend for the twenty- five to forty- fouryear- olds as in figure 5.11, while the share for the forty- five to sixty- fouryear- olds, is approximately constant. The Danish panel in the ECHP has been used also in figure 5.13. We have pooled data for all eight waves and show in the graph the share by age, fortyfive to seventy, and gender who find their health status to be “very good” or “good.” As expected, the share is falling with age but at varying speed. It seems that two points—with caution—can be concluded from figure 5.13. First, while the share is at the same level for women and men in their midforties the decline is much faster for women. Secondly, the share seems to be flat or only falling quite slowly from the midfifties to the midsixties, that is, the age interval where many were eligible for one or another of the early retirement programs that were open in the years covered by the ECHP (cf. the detailed description of these programs in section 5.5). Next, we present a few health indicators based on register data. First, figure 5.14 shows the total number of new cases (per 100,000) of all kinds of cancer found annually from 1978 to 2008, separately for women and men. Notice that the number increases throughout the period and more so as mortality goes down, also due to cancer in the sixty to sixty- four- year- old group (cf. figures 5.9 and 5.10).

Disability Programs, Health, and Retirement in Denmark since 1960

Fig. 5.12

227

Share with self-reported health “good” and “very good”

Source: Calculations from European Community Household Panel.

Fig. 5.13 Share of respondents with self-reported health “very good” or “good” by age and gender, average values 1994 to 2001 Source: Calculations from European Community Household Panel, pooled over eight waves.

Figures 5.15 and 5.16 show for the same age- group and separately by gender the (scaled) prevalence of annual diagnoses of three major disease groups, that is, malignant tumors, myocardial infarction and related diseases, and psychiatric diagnoses along with mortality. Due to a new classification system for diagnoses introduced in 1991, this year is the first in figures

228

Paul Bingley, Nabanita Datta Gupta, and Peder J. Pedersen

Fig. 5.14

New cases of cancer per 100,000, 1978 to 2008

Source: Statistics Denmark.

Fig. 5.15 Mortality and prevalence of three major diagnoses, men aged sixty to sixty-four years Source: Statistics Denmark.

5.15 and 5.16. The data shown in figures 5.15 and 5.16 are illustrative. A cautious interpretation is that diagnoses and mortality are on different tracks. For instance, for men sixty to sixty- four years old, the number of diagnoses of cases with serious diseases of the heart goes up quite strongly while heart diseases as cause of death goes down equally strongly (cf. figure 5.9).

Disability Programs, Health, and Retirement in Denmark since 1960

229

Fig. 5.16 Mortality and prevalence of three major diagnoses, women aged sixty to sixty-four years Source: Statistics Denmark.

5.5 5.5.1

DI and Other Early Retirement Programs Social Disability Pension (SDP)

The historical origin of SDP is a “classic” disability pension granted on medical criteria ending with a transition to Old-Age Pension ([OAP], folkepension) at age sixty- seven (decreased to sixty- five years in 2004). The number of participants in SDP relative to the population is shown in figure 5.17. Until 1983 there were, along with SDP, a number of smaller social security pension programs available before the OAP age granted on a mix of medical and social criteria. These small programs consisted of a program for widows’ pension conditional on being aged fifty- five years or older, and programs for granting early OAP before age sixty- seven to single women, and to women above a certain age with an older spouse receiving OAP. Further, a small group of men could be granted early OAP on specific social criteria. Policy changes and reforms have been enacted on several occasions. The first policy change—in the period we consider—occurred in 1965. The decline in the number of participants from 1965 to 1967—after a strong increase from 1960—was, however, due to purely administrative reclassifications (Bengtsson 1989). The next policy change was in 1967, making the eligibility criteria less restrictive regarding health. From 1974, married women sixty- two to sixtysix years old with an older spouse receiving OAP no longer received an independent early OAP. This was replaced with an increase in pension benefits to the retired spouse. The decline in the number of DI recipients from 1974 to

230

Paul Bingley, Nabanita Datta Gupta, and Peder J. Pedersen

Fig. 5.17 Participants in DI, aged eighteen to sixty-six/sixty-four as share of population, 1960 to 2008 Source: Statistics Denmark. Note: Policy reforms in 1967, 1974, 1983, 1998, and 2003.

1975 was thus artificial, that is, the number of people in the labor force was not affected and the expenditures for the program were not affected either. A major change of SDP was enacted in 1984 through a major reform of social security programs regarding early retirement. The reformed SDP was intended to encompass a number of earlier programs. The biggest among these by far was the classical disability pension. Other programs, which were included into the new SDP, were, as before, a public- financed program for widows’ pension, a program for early OAP for specific groups, and a program for disabled persons with a low- level compensation. The data break in 1984 occurred as a consequence of new groups who became eligible at the same time as some among the previously eligible groups lost their eligibility. New groups that became eligible were fifty to sixty- six- year- olds on social criteria and all eighteen to sixty- six- year- olds on a broad mix of social and/or health criteria. Those now excluded were a number of widows fiftyfive years and older and single women, sixty- two to sixty- six years old, excluded on a new means test. The group of newly eligible was significantly bigger than the group who lost eligibility. From 1984, SDP on medical or social criteria could be granted on three levels. The highest level was applicable to persons younger than sixty whose work capacity had been reduced to almost nothing. The intermediate level SDP was open for those younger than sixty with a work capacity reduced to one- third of the normal level, and to people sixty to sixty- six years old with almost no remaining work capacity. Eligibility for the highest and the

Disability Programs, Health, and Retirement in Denmark since 1960

231

intermediate levels of SDP was decided on medical criteria. Finally, eligibility for the lowest level, the so-called ordinary level SDP, depended on work capacity having been reduced to below half the normal level. The evaluation of this was based on health criteria or on a combination of health and social criteria. Recipients of the ordinary level of SDP that were younger than age sixty were entitled to a supplementary amount. From 1999 the granting of the ordinary level pension was dependent on rehabilitation having been considered or tried without success. Recipients of benefits in the number of smaller programs that were merged with the classical disability pension from 1984 were all entitled to the lowest level of SDP. It is consequently only possible to avoid the data break in 1984 if the focus is on the group of people entitled to the highest and the intermediate level of benefits. For this group a data series is consistent until a reform making the program more simple was enacted in 2003. In 1992, 1997, 1998, and 1999 administrative and financial incentives were tightened up for the municipalities who are responsible for granting SDP. The purpose of these changes was to restrict/reduce entry to the program. A reform was, as mentioned, enacted in 2003. The main contents were a tightening of the eligibility criteria and a reduction of benefits for future SDP pensioners. The three levels of benefits (plus different ad hoc benefits) were replaced by one benefit set at maximum unemployment insurance benefits for married/cohabiting, and 85 percent of this amount for singles. In figures 5.18 and 5.19 we disaggregate by gender and age and show the relative participation in SDP since the 1984 reform. In this period alternative early retirement programs were introduced—or already in operation—to be described later. One of these, the Post Employment Wage (efterlon, hereafter

Fig. 5.18

The SDP/population, men, 1984 to 2006

Source: Statistics Denmark.

232

Paul Bingley, Nabanita Datta Gupta, and Peder J. Pedersen

Fig. 5.19

The SDP/population, women, 1984 to 2006

Source: Statistics Denmark.

[PEW]) program, is relevant from 1979 for the sixty to sixty- six (sixty- four)year- olds while another one, the Transitional Benefit Program (overgangsydelse, hereafter [TBP]), was relevant for the fifty to fifty- nine- year- olds.2 This transitional program was open from 1992 to 1996 and is described later. For the forty to forty- nine- year- olds, SDP was the only potentially available early retirement program and for this age- group we find an increase in the share in contrast to the two older groups. For the fifty to fifty- nine- yearolds, and especially for the sixty to sixty- four- year- olds, it seems evident that program substitution is occurring. We return to this later. 5.5.2

The Flex Job Program

This program was introduced in 1998. One explicit purpose was to reduce the entry to SDP by creating a new type of work opportunity for people with permanently reduced work capacity, but with a residual work capacity above the level for becoming eligible to SDP. The employer pays the wage to a flex job employee and is reimbursed from the public sector with either half or two- thirds of the collectively agreed minimum wage in the relevant part of the labor market, the share depending on the work capacity of the individual employee. The program is supplemented with an Unemployment Compensation Benefit (not part of the standard unemployment insurance program) designed to provide an income for persons unable to be admitted to the flex job program until a flex job is available, and to provide an income during unemployment spells between flex jobs. The program quickly became popular 2. The PEW program was relevant for the sixty to sixty- six- year- olds until 2004, but data for figures 5.14 and 5.15 are only available for the age interval sixty to sixty- four.

Disability Programs, Health, and Retirement in Denmark since 1960

233

Fig. 5.20 Number of participants in the Flex Job Program, women and men, by age, relative to population in relevant age groups, 2000 to 2009 Source: Statistics Denmark.

with entry rates far above predictions in the policy preparation phase and seemingly without much impact on entry rates to the SDP program. By mid2009, about 60,000 persons corresponding to about 2 percent of the labor force were enrolled (cf. figure 5.20). 5.5.3

The Post Employment Wage (PEW) Program

In 1979 a new early retirement program, the PEW, was introduced with eligibility based solely on being at least sixty years old and having sufficient tenure as member of an unemployment insurance fund. Entry to the program was high compared to estimates made during the policy preparation phase. There were two main arguments behind the introduction of the program. One was that certain groups of manual workers with a labor market history of hard physical work should have an option for early retirement without having to fulfill the formal medical criteria for the disability pension program. The other was to release jobs during a deep recession for younger unemployed workers. In 1979 the OAP age was sixty- seven so the relevant age- group was the sixty to sixty- six- year- olds in the labor force. Initially, eligibility was conditional on membership of an unemployment insurance fund for five out of the most recent ten years. Benefits in the program for workers coming from a job were set at unemployment insurance benefit levels for the first 2.5 years, and at 82 percent of maximum unemployment benefits for the remaining period until age sixty- seven. Participants were allowed to work 200 hours per year at most, reflecting the original motive of creating jobs for the young

234

Paul Bingley, Nabanita Datta Gupta, and Peder J. Pedersen

unemployed. A recent study, Bingley, Gupta, and Pedersen (2010), concluded that the program did not have the intended impact on youth employment or unemployment. During the thirty years since its introduction, the PEW program has undergone a number of changes. The first came in 1980 when eligibility became conditional on unemployment insurance fund membership in ten out of the most recent fifteen years. Like later reforms of the program this was, however, grandfathered so the more restrictive rule was applied only for workers younger than fifty at the time of the change. The initial five out of ten years rule was thus effective for all entrants over the first ten years with the program. The next major change was in 1992 with the purpose of reducing entry, which became conditional on unemployment insurance fund membership in twenty out of the previous twenty- five years. However, this was again grandfathered to apply only to persons younger than forty in 1992. The 1992 reform introduced a “sixty- three years rule,” meaning that entry from age sixty- three implied PEW at 100 percent of unemployment insurance benefit level through to age sixty- six. Also in 1992 another temporary program for early retirement for individuals fifty to fifty- nine- years- old was introduced. This, the so-called Temporary Benefits Program (TBP), is described later. The next reform of PEW came in 1999, introducing the “Flexible PEW.” The main elements in the reform were that entry before age sixty- two implied means- testing against private pension plans, also those with delayed benefits, and further PEW benefits were 91 percent of maximum unemployment insurance benefits for the whole duration of PEW. Entry at sixty- two years and older implied no means- testing against individual pension schemes. Further, postponing entry implied accumulation of a quarterly tax- free premium up to the duration of the whole PEW period. Next, the 200 hours limit on paid work was replaced by a reduction of benefits by a fixed amount for each hour worked. Another new element was the introduction of voluntary PEW contribution; however, at far lower than a fair actuarial contribution. These were previously bundled with unemployment insurance contributions. Finally, the number of years of unemployment insurance fund membership was changed again to twenty- five out of the last thirty years. This was, however, also grandfathered. The most important part of the pension reform in 1999 was a reduction of the OAP age from sixty- seven to sixty- five, effective from 2004 for those born July 1939 or later. The impact on labor supply from this was relatively small as most of the sixty- five to sixty- six- year- olds were in the PEW program or in other early retirement programs. For the public purse, financial savings were that for two years individuals could no longer collect PEW but instead became eligible to OAP, which was less generous. On the other hand, a number of people, either working or being provided for by their family, could now collect OAP for two more years.

Disability Programs, Health, and Retirement in Denmark since 1960

235

Overall, the motive in reforms in this area has been to change incentives for entering PEW, thus keeping more people in the labor force. We return to look into whether these intentions have succeeded or not. We show in figures 5.21 and 5.22 the number of men and women in the PEW program relative to the number of people in the relevant part of the population. In both figures we attempt to correct for the OAP reform by including in the graphs from 2005 the number of people sixty- five to sixty- six years old who would have collected PEW assuming the same take-up rate for those two cohorts as in 2004. For men, we find a very strong increase in the share collecting PEW in the first period after 1979 until all eligible sixty to sixtysix- year- olds, could collect benefits. A new increase occurs throughout the 1990s followed by a 5 percentage points decline when adjusting for the OAP change from age sixty- seven to sixty- five. For women, figure 5.17 shows a nearly uninterrupted increase in the PEW share until a peak in 2003 followed by a very moderate decline after adjustment. From around 2000 the share of women surpasses the share of men collecting PEW. For SDP the share of women has been significantly higher than the share of men ever since 1984 (cf. figures 5.14 and 5.15). 5.5.4

The Transitional Benefits Program

Besides the PEW program, a more restricted early retirement program in the social security area, the Transitional Benefit Program (TBP), was opened for long- term unemployed people in their fifties in 1992. Eligibility conditions were stricter than to the PEW program as it was made additionally

Fig. 5.21 Number of participants in PEW relative to population in relevant age group, men, 1979 to 2008 Source: Ministry of Labor and Statistics Denmark.

236

Paul Bingley, Nabanita Datta Gupta, and Peder J. Pedersen

Fig. 5.22 Number of participants in PEW relative to population in relevant agegroup, women, 1979 to 2008 Source: Ministry of Labor and Statistics Denmark.

dependent on twelve months in unemployment out of the most recent fifteen months. In 1992, long- term unemployed people fifty- five to fifty- nine years old could enter the program. After entry they collected reduced unemployment insurance benefits and no longer had to search actively and be available for a job offer. Entry was thus an effective exit from the labor force. From age sixty participants were transferred to PEW. From 1994 the program was open also for long- term unemployed fifty to fifty- four years old. Entry to the program was closed again in 1996. In spite of being open for entry only four years, the TBP had an impact for several years on the labor market for people fifty to fifty- nine years old. A person entering aged fifty years old in 1996, just before the closing of the program, would thus transfer to PEW in 2006. The number of participants relative to the population in relevant ages is shown in figure 5.23. For the fifty- five to fifty- nine- year- olds the increase was a dramatic one, going from 3 percent entering the first year to a stock of 9 percent of all fifty- five to fifty- nine- year- olds in the third year. Like for PEW, the entry far exceeded estimates in the policy preparing phase. No less than eight programs for early retirement—not all in operation in all the years—have been available in Denmark in the period 1970 to 2008. In figures 5.24 and 5.25 we attempt to illustrate the shifting importance of these different pathways over the period. This is a pragmatic approach as the ideal flow data are not available for the period. Instead we have separately,

Fig. 5.23 Number of participants in the TBP program by age relative to population, 1992 to 2006 Source: Statistics Denmark.

Fig. 5.24 Pathways to retirement, men aged fifty-five to sixty-six years, Denmark, 1970 to 2008 Source: Calculations from Statistics Denmark.

238

Paul Bingley, Nabanita Datta Gupta, and Peder J. Pedersen

Fig. 5.25 Pathways to retirement, women aged fifty-five to sixty-six years, Denmark, 1970 to 2008 Source: Calculations from Statistics Denmark.

for men and women fifty- five to sixty- six years old, calculated the relative distribution on the different programs annually. Taking a specific year as example, the graphs show the distribution of men, and respectively women, on the included programs for early retirement.3 The importance of opening up new programs and subsequent crowding out, at least in relative terms, of existing programs is evident from figures 5.24 and 5.25. The opening of new programs has not only a substitution effect but also an effect on the total number of people fifty- five to sixty- six years old being provided for in the different programs. This effect on the absolute numbers is not visible from the two graphs. For the period 1984 to 2000 data are available (Larsen and Pedersen 2008) on origin and destination for people moving into early retirement programs. Annual average values are shown for men and women in figures 5.26 and 5.27 for the composition by origin for people having PEW, SDP, other early retirement programs and OAP as destinations, with coming from a job as dominant for all destinations. The importance of coming from unemploy3. The relative importance of unemployment as a pathway is exaggerated as a share of unemployment spells in the age- group ends with employment in a new job.

Disability Programs, Health, and Retirement in Denmark since 1960

239

Fig. 5.26 Relative distribution on flows into early retirement and OAP, average values, men, 1984 to 2000 Source: Calculations from Larsen and Pedersen 2008.

Fig. 5.27 Relative distribution on flows into early retirement and OAP, average values, women, 1984 to 2000 Source: Calculations from Larsen and Pedersen 2008.

ment varies between destinations with least importance for SDP as destination. For women we see the destination “other early retirement” differ by origin relative to other early retirement states. The relative importance of unemployment presumably reflects the TBP program where long- term unemployment was a condition for eligibility. For the two big programs SDP and PEW, we show in figures 5.28 and 5.29

240

Paul Bingley, Nabanita Datta Gupta, and Peder J. Pedersen

Fig. 5.28 Share of male population aged sixty to sixty-six years old in SDP and PEW programs, 1979 to 2008 Source: Statistics Denmark.

Fig. 5.29 Share of female population aged sixty to sixty-six years old in SDP and PEW programs, 1979 to 2008 Source: Statistics Denmark.

the shares being participants in each program by gender for the period from the opening of PEW in 1979 until 2008. For men sixty to sixty- six years old the SDP share is stable until the late 1990s when a decline of 5 percentage points is seen at the same time as the PEW share goes down. For women, on the other hand, the SDP share goes down with some 20 percentage points at the same time as the PEW share goes up. There is, however, no evidence of a more direct program substitution.

Disability Programs, Health, and Retirement in Denmark since 1960

5.6

241

Labor Force Participation, Unemployment, and Policy Reforms

The opening of new programs has had a clear impact on labor force participation in the eligible age- groups. Labor force participation rates for men fifty- five to fifty- nine, sixty to sixty- four, and sixty- five to sixty- nine years old are shown in figure 5.30, including indications for major policy changes. The impact from the introduction of PEW is very clearly illustrated while no impact is found in 1984 from the SDP reform. The opening of TBP in 1994 for the fifty- five to fifty- nine- year- olds is equally visible. The participation rate for the sixty to sixty- four- year- olds shows a fairly strong increase after the 1999 reform for both men and women. Part of this may, however, reflect statistical conventions as earnings above a certain level in the most recent years (from 2003) can result in a participant in the PEW program being classified as a labor force participant if earnings are sufficiently high. For women, shown in figure 5.31, the reaction to PEW is weaker than for men due to much lower participation rates for women in their sixties. For the fifty to fifty- nine- year- olds, on the other hand, the TBP reaction is much stronger than for men. The 1984 SDP reform is, however, not visible in the participation rates. Policy changes are also reflected when we look at employment and unemployment rates for men in the age intervals fifty- five to fifty- nine and sixty to sixty- four from 1972 to 2007 (cf. figures 5.32 and 5.33). Employment rates for the fifty- five to fifty- nine- year- olds return to the level of about 0.80, found during the 1980s after the TBP generated dip in the the 1990s. Looking

Fig. 5.30 Labor force participation, men aged fifty-five to fifty-nine, sixty to sixtyfour, and sixty-five to sixty-nine years old Source: Statistics Denmark.

Fig. 5.31 Labor force participation, women aged fifty-five to fifty-nine, sixty to sixty-four, and sixty-five to sixty-nine years old Source: Statistics Denmark.

Fig. 5.32 Male employment rates, fifty-five to fifty-nine and sixty to sixty-four in Denmark Source: Statistics Denmark.

Disability Programs, Health, and Retirement in Denmark since 1960

243

Fig. 5.33 Male unemployment rates, fifty-five to fifty-nine and sixty to sixty-four in Denmark Source: Statistics Denmark.

at unemployment rates in figure 5.33, the impact from policy changes is even more clearly found than in the labor force participation data. In the late 1970s, after the first oil price shock and before introduction of PEW, unemployment was significantly higher for the sixty to sixty- four- year- olds than for those aged fifty- five to fifty- nine. The introduction of PEW results—as expected—in unemployment falling to half the pre- 1979 level for the sixty to sixty- four- year- olds. Unemployment remains low until the early 1990s while it increases from about 6 to about 12 percent for the fifty- five to fiftynine- year- olds. The TBP program results in a large decline in unemployment for this age group. 5.7

Health, Policy Reforms, and the Labor Market

The earlier focus has been on the impact on labor force participation, employment, and unemployment from a number of changes in labor market and retirement policies. We found no visible impact from reforms in the SDP program as an indicator for changes in health and mortality. The purpose in this section is to relate labor market changes in a more explicit way to changes in mortality and SDP participation. In the first illustration, we show in figure 5.34 the employment rate for men fifty to seventy years old at three points in time, 1981, 1994, and 2008. We find the ranking between the years as expected, that is, the highest employment rate in 1981 where PEW not yet had its full effect, a strong decline in the peak unemployment year 1994, and finally a return to a higher employment rate in the near full employment year of 2008.

244

Paul Bingley, Nabanita Datta Gupta, and Peder J. Pedersen

Fig. 5.34

Employment rate, men aged fifty to seventy years, 1981, 1994, and 2008

Source: Calculations from Statistics Denmark.

What we do next is for each age between fifty and seventy in each of the years to find mortality and plot that against the employment rate. This is done in figure 5.35 for men. The result is an unstable relationship over time between mortality and employment at given ages. Instead, we find mortality declining at a given employment rate or employment declining at a given rate of mortality. Taking mortality as a health indicator, we thus have that an improvement of health is accompanied by a decline in employment rates. A possible explanation is that non- health- related early retirement programs have a crowding out effect relative to the impact from health in isolation. For women, figure 5.36 reflects a strong cohort effect regarding employment until the early sixties. Still, we find in figure 5.37 a shift of the relationship to the left indicating as for men that reduced mortality (improved health) correlates with lower employment rates. In the last four figures, 5.38, 5.39, 5.40, and 5.41, we include in the same graphs for men and women fifty- five to fifty- nine and sixty to sixty- six years old, mortality rates, (scaled) relative participation in the SDP program, and labor force participation rates for the period 1973 to 2008. For expositional reasons disability rates are multiplied with 200. For all four gender/ age groups mortality as a health indicator is falling from the first half of the 1990s. Due to the introduction of PEW, SDP participation is not comparable between the fifty- five to fifty- nine and the sixty to sixty- six- year- olds. Furthermore, the impact from TBP makes it difficult to interpret SDP as

Fig. 5.35 Employment rate by mortality risk * 100, men aged fifty to seventy years old, 1981, 1994, and 2008 Source: Calculations from Statistics Denmark.

Fig. 5.36 Employment rate, women aged fifty to seventy years, 1981, 1994, and 2008 Source: Calculations from Statistics Denmark.

Fig. 5.37 Employment rate by mortality risk * 100, women aged fifty to seventy years, 1981, 1994, and 2008 Source: Calculations from Statistics Denmark.

Fig. 5.38 Labor force participation, mortality and (scaled) disability, men aged fifty-five to fifty-nine years Source: Calculations from Statistics Denmark.

Fig. 5.39 Labor force participation, mortality and (scaled) disability, men aged sixty to sixty-four years Source: Calculations from Statistics Denmark.

Fig. 5.40 Labor force participation, mortality and (scaled) disability, women aged fifty-five to fifty-nine years Source: Calculations from Statistics Denmark.

248

Paul Bingley, Nabanita Datta Gupta, and Peder J. Pedersen

Fig. 5.41 Labor force participation, mortality and (scaled) disability, women aged sixty to sixty-six years Source: Calculations from Statistics Denmark.

a health indicator for the fifty- five to fifty- nine- year- olds from the mid1990s. With these reservations in mind SDP is flat until the impact from the SDP reform and introduction of competing programs makes it difficult to interpret SDP take-up as a genuine health indicator. Labor force participation has the profile for the four gender/age groups described earlier, that is, heavily influenced by program innovations and reforms since the early 1980s. 5.8

Summary and Conclusions

The main conclusion from the previous analyses is the strong relationship between labor force participation, employment, and unemployment on one hand, and non- health- related programs for early retirement on the other hand for everybody sixty years or older for most of the period for which we have consistent data series. For the last twenty years this also applies for people in their fifties. Only fairly incomplete evidence exists for self- reported health status. We combine this with register- based evidence for the health sector, that is, prevalence of some major diseases and some main causes of death, and finally we include mortality for the most relevant age groups. However, no clear relationship appears between labor market performance and these health indicators. Changes in transfer programs unrelated to health status tend to dominate any relationship between labor market attachment and health measures. It is fairly difficult, based on available data, to construct a long consistent

Disability Programs, Health, and Retirement in Denmark since 1960

249

series of take-up of the health- related part of SDP. This might be part of the reason that no evident relationship appears between health indicators and SDP take-up. A very important reason for, or explanation of, the lack of a number of expected correlations is most probably that SDP is “on its own track” due to program innovations and reforms creating competing risks or program substitution dominating the picture in the most recent decades for those older than fifty in focus in the present study. While it seems to be difficult to relate SDP take-up to available health indicators, it is still a highly relevant question to continue analyses of the factors behind the development in this very expensive public sector program for early retirement.

References Autor, D., and M. G. Duggan. 2003. “The Rise in the Disability Rolls and the Decline in Unemployment.” Quarterly Journal of Economics 118:157–205. Bengtsson, S. 1989. Førtidspension eller . . . ? (Disability Pension Or . . . ?). Copenhagen: Socialforskningsinstituttet. Bingley, P., N. D. Gupta, and P. J. Pedersen. 2010. “Social Security, Retirement, and Employment of the Young in Denmark.” In Social Security Programs and Retirement around the World. The Relationship to Youth Employment, edited by J. Gruber and D. A. Wise. Chicago: The University of Chicago Press. Black, D., K. Daniel, and S. Sanders. 2002. “The Impact of Economic Conditions on Participation in Disability Programs: Evidence from the Coal Boom and Bust.” American Economic Review 92:27–50. Bratsberg, B. 1999. “Disability Retirement in a Welfare State.” Scandinavian Journal of Economics 101 (1): 97–114. Bratsberg, B., E. Fevang, and K. Røed. 2010. Disability in the Welfare State: An Unemployment Problem in Disguise? Institute for the Study of Labor (IZA). Discussion Paper no. 4897. Bonn: IZA. Campoliteti, M. 2004. “Disability Insurance Benefits and Labor Supply: Some Additional Evidence.” Journal of Labor Economics 22 (4): 863–89. Gruber, J. 2000. “Disability Insurance Benefits and Labor Supply.” Journal of Political Economy 108 (6): 1162–83. Larsen, M., and P. J. Pedersen. 2008. “Pathways to Early Retirement in Denmark, 1984– 2000.” International Journal of Manpower 29 (5): 384–409.

6 Disability in Belgium There Is More Than Meets the Eye Alain Jousten, Mathieu Lefebvre, and Sergio Perelman

6.1

Introduction

The economic literature has produced a wide body of research on the determinants of labor supply at older ages. It suggests that the generous early retirement routes opened up within the setting of various social protection systems are largely responsible for the low participation of elderly. There is no doubt that the strongest factors driving early retirement are the financial incentives provided by social security systems, but health issues might be of importance as well.1 Specifically, disability insurance (DI) is a key component of the social security system in most advanced countries. Belgium is no exception to this role with the public DI system protecting workers in case of longer- term loss of work capacity. However, few studies on Belgium have focused on DI as one of several systems allowing elderly to leave the labor force. As compared to the substantially larger health and retirement programs, DI has only attracted a limited interest in the Belgian public policy debate.

Alain Jousten is professor of economics at the University of Liège and a research fellow with the Institute for the Study of Labor (IZA) and with Netspar. Mathieu Lefebvre is a postdoctoral researcher at the University of Liège. Sergio Perelman is professor of economics at the University of Liège. The authors wish to thank Michael Turco (CIMIRE), Lut Vanden Meersch (INAMIRIZIV), and Khiêm Nguyen (ONEM-RVA). Financial support from the Communauté Française de Belgique ARC contract (ARC 05/10-332) is gratefully acknowledged. All errors are our own. For acknowledgments, sources of research support, and disclosure of the authors’ material financial relationships, if any, please see http://www.nber.org/chapters/c12387.ack. 1. See Pestieau and Stijns (1999); Dellis et al. (2003); Desmet et al. (2007); Disney, Emmerson, and Wakefield (2006); and Kalwij and Vermeulen (2008).

251

252

Alain Jousten, Mathieu Lefebvre, and Sergio Perelman

According to Eurostat, spending on DI accounted for only 1.8 percent of gross domestic product (GDP) in 2007. Two key factors are at play. On the one hand, while there is no doubt that poor health is a key determinant of inactivity, the multiplicity of exit routes available to older workers in Belgium makes it difficult to clearly identify the relationship between health condition and DI participation and, hence, also the influence of DI on retirement decisions. On the other hand, poor health may also affect a worker’s productivity and, hence, affect his or her risk of job loss. In both cases, there exist various schemes that allow the workers to withdraw from the labor market. In light of these issues, we focus on two specific questions in this chapter. First, we evaluate whether there is a systematic link between indicators of population health—both objective and subjective—and program participation. We research whether the improvement observed in health status and mortality in the last decades is associated with a reduction of programs participation—with a special focus on the DI program. We would expect such a link to prevail, though causality of this link could obviously run both ways. Second, we describe the role of the DI system in the wider social security system. Previous work on retirement incentives has clearly illustrated the need to take a comprehensive approach when looking at the behavioral consequences of social protection systems—particularly in a context of reform over time (see Dellis et al. 2004, Desmet et al. 2007). Indeed, Belgium excels in the use of a variety of early retirement programs if the average age of retirement was about 61.4 in 2007; it was 57 in 2000. The chapter is structured as follows. Section 6.2 describes the institutional landscape of the Belgian social insurance and assistance schemes. We provide a description of the changes to the different early retirement schemes over time—with an emphasis on those relevant in the disability context. Section 6.3 provides an overview of historical data on health trends. Relying both on administrative and survey data, we show the link between health status and mortality trends, on the one side, and labor market and disability participation, on the other. Section 6.4 analyzes program participation, with a special emphasis on the link between behavioral trends and changes in the program design. This is of prime importance because the coexistence of several schemes raises a problem of communicating vessels, wherein a change in the rules of one scheme may have a direct impact on the participation to another scheme. Section 6.5 concludes the chapter. 6.2

Institutional Background

The aim of this section is to describe the most important social protection schemes that can serve as a source of replacement income when retiring from the labor market within a disability context. As in previous work on retirement incentives and labor market outcomes (see Jousten et al. 2005, 2010),

Disability in Belgium: There Is More Than Meets the Eye Table 6.1

253

List of exit routes DI UI CER SER RPP OAU

Disability insurance system Unemployment insurance system Conventional early retirement system Special early retirement system Retirement prepension system Old-age unemployment system

for reasons of data availability, our focus will be the wage- earner scheme.2 (See table 6.1.) 6.2.1

Disability Insurance (DI)

The DI system is the main program targeted at people withdrawing from the labor market for disability reasons. To be eligible for the benefit, the worker has to satisfy a series of conditions. First, there is a condition in terms of the loss of earnings capacity in the usual job. In order to be eligible, the worker has to suffer from a loss of earnings capacity of 66 percent over a period of at least twelve months—as certified by the national DI administration (INAMI-RIZIV). Continued benefit eligibility is validated using periodic medical and administrative controls. Beyond this loss of earnings- capacity requirement, workers also have to satisfy minimum contributory requirements to qualify for benefits. The coverage under the system prior to the onset of the disability has to be assured for at least two quarters, combined with at least 120 days of actual or assimilated work (or 400 hours for part- time workers) before the covered event occurs. The system is financed on a pure pay- as-you- go (PAYG) basis. Benefits are payable up to a maximum age of sixty- five—corresponding to the normal retirement age. At this age, people become eligible for regular retirement benefits and are automatically transferred into the retirement system. In line with other social insurance replacement income programs, such as unemployment insurance (UI), health insurance (HI), and conventional early retirement (CER), periods of benefit receipt fully count toward the buildup of pensions—though no contributions are due. Periods on DI are fully assimilated to work in pension records, with fictive (indexed) wages being imputed into the earnings history at a level corresponding to real earnings in the year prior to disability onset (see the following). The benefit level is a function of the household status of the worker and is paid out free of income taxes. The benefit is equal to 65 percent of reference earnings if the insured has dependents, 40 percent if the insured cohabits 2. The wage earner social insurance system is by the far the largest one in terms of enrollment terms, also representing the largest share of public spending on social security. It is less generous than the scheme for civil servants and more generous than the one for the self- employed.

254

Alain Jousten, Mathieu Lefebvre, and Sergio Perelman

but has no dependents, and 53 percent if the insured lives alone. Reference earnings are based on real observed earnings in the period leading up to the onset of the insured event. A cap on benefits applies for higher levels of income. Benefits are adjusted periodically for changes in the consumer prices using the so-called health- index—a slowed- down version of the Consumer Price Index (CPI) serving as a basis of indexation notably for rents and social benefits. In the twelve months leading up to the benefit eligibility, workers suffering from a loss of work capacity are not left without income. During this socalled primary incapacity, people are covered under the rules of the HI— with slightly different eligibility rules that are administered in a decentralized way by the various mutual health insurance organizations. During the first month of incapacity, the benefit is equal to 60 percent of the reference income. Over the next eleven months, the benefit is equal to 60 percent if the insured has dependents or lives alone and 55 percent if he or she cohabits. When the DI system was introduced in 1963, the benefit was equivalent to 60 percent of reference earnings if the insured has dependents and 33 percent otherwise. These replacement rates were adjusted along time. In 1986, a differentiation between individuals living alone and cohabitants was introduced and replacement rates adjusted to 65 percent for insured with dependents, 40 percent if cohabiting with no dependents, and 45 percent if the disabled person lives alone. In 2002, the replacement rate of the disabled living alone was further increased to 50 percent before finally reaching current levels in 2004. Aside from changes in the replacement rates, an important reform occurred in 1998, significantly enlarging the notion of “dependents” by including the members of the spouse’s family. The aim was to align the definition with the one used in the unemployment rules in order to make systems more coherent. 6.2.2

Pension System

The wage- earner pension system is the main program covering the elderly. The program is financed by tax- deductible employer and employee contributions as well as by transfers from the federal budget. The system runs on a pure PAYG basis. Individuals are eligible to full benefits at the age of sixty- five, with early retirement possible at age sixty for both women and men. A full career corresponds to forty- five years of earnings or assimilated periods. Assimilated periods correspond to periods on DI, UI, HI, CER, as well as various other arrangements such as career breaks and some other forms of leave. No actuarial adjustment is applied for early retirement beyond the effect possibly playing through the incompleteness of the earnings history. Social security benefits correspond to 75 percent of average lifetime earnings for one- earner couples and to 60 percent for others. Benefits have a

Disability in Belgium: There Is More Than Meets the Eye

255

preferential tax treatment and are regularly indexed using the health index as well as occasionally uprated on a purely discretionary basis to prevent a severe decoupling of benefits from average wages in the economy. Special regimes also remain in place for some specific sectors, such as most notably for coal mining and the maritime sector. Survivor benefits are available since 1968. Originally, the basis for the current- day system was laid in two separate steps for blue- and white- collar employees in 1953 and 1957. Since its creation, the system has undergone a series of major changes. In 1967, the systems were merged and key changes introduced. Male retirement age was set at sixty- five and female retirement at sixty, with corresponding full- career lengths of forty- five and forty years—combined with a possibility to retire early by five years with a 5 percent reduction per year of anticipation. A reform of 1977 can be seen as the first step toward an active policy of shifting people into inactive early retirement program—even though this specific program was only short- lived and phased out in 1979 and 1982.3 The reform allows for early retirement within the preceding general pension system without financial penalties. The so-called special early retirement (SER) gives disabled workers or long- term unemployed aged of at least sixty for men and of fifty- five for women a bonus equal to the difference between the actual pension and the one they would receive if it was not reduced by the 5 percent rule. The ensuing 1983 creates a new (and longer- lived) scheme of early retirement within the framework of the wage- earner pension system. The program of retirement prepension (RPP) allows male workers to retire a maximum of five years early without any benefit reduction if the employer commits to replacing the worker by an unemployed who is benefiting from a full- time compensation from the unemployment insurance. The system only ended in 1991 with the formal end of a normal retirement age. In 1991, a major reform of the system modifies the landscape in retirement ages. While female early retirement before the age of sixty was already eliminated in 1987, the 1991 reform generalized the concept of a window of retirement ages: subject to career requirements, workers of both sexes are free to choose the retirement age between sixty and sixty- five. The normal retirement ages for women and men do, however, remain different at ages sixty and sixty- five. The 5 percent rule as well as the RPP both lose their rationale and are eliminated. It is, however, only in 1997 that the Belgian government, forced by a European court of justice ruling, harmonized the normal retirement age and the career- length requirements for women and men. The decision was taken to 3. The program was the first of a series of programs put in place to shift older workers out of the labor market to free up jobs for younger workers. E.g., Walker (2007) and Jousten et al. (2010) discuss these “lump- of-labor” issues in more detail and with diverging conclusions.

256

Alain Jousten, Mathieu Lefebvre, and Sergio Perelman

align the female full career condition to the one applicable to men. Since 1997, the normal retirement age for women has gradually been raised to sixty- one in 1997, sixty- two in 2000, sixty- three in 2003, sixty- four in 2006, and, finally, attained the now steady- steady of sixty- five in 2009, with the corresponding increases in full- career requirements. While aimed at eliminating discrimination, the measure has also a clearly beneficial impact in budgetary terms and contributes positively to the longer- term viability of the regime. The most recent reform to retirement incentives under the pension system dates back to 2007. Under the so-called Intergenerational Solidarity Pact (ISP) negotiated between the social partners and the government in late 2005, workers working beyond the age of sixty- two or beyond forty- four years of career can benefit from a pension supplement.4 The pension bonus of an amount of €2 per day worked beyond these limits, augments the annual benefit payable, and this independently of the wage earned or the contributions accumulated. As such, it can be seen as a much stronger relative incentive for lower- wage earners than for higher- wage earners. 6.2.3

Conventional Early Retirement

Next to the pension system, a parallel system of complementary benefits for early retirement is created in 1973 for the old workers in case of firing: the conventional early retirement (CER). In order to be eligible, the worker has to have been laid off by his or her employer and also satisfy some career length requirements.5 Benefits correspond to the ones payable under the unemployment insurance system (see the following) topped up by a complementary benefit paid by the employer—the latter being equal to half the difference between net wage and the unemployment benefit. A particularity of the system is that these early retirees are exempted from job search and—as for all unemployed—the time spent in the early retirement program is fully credited in the earnings file for pension purposes. At the outset, the system is only accessible if the employer recruits a replacement worker benefiting from full- time unemployment benefits as a replacement for the worker. Further, the initial age of eligibility is set at sixty. But both criteria are hollowed out over time by means of legislative change and collective bargaining agreements within industries. As a result, there ultimately exists a variety of different regimes with different career requirements, minimum ages, re4. The Belgian Intergenerational Solidarity Pact includes some thirty measures aimed at reducing early retirement without changing the legal retirement age or current benefits already granted. Measures include limiting the number of people taking early retirement, stimulating employers to retain or hire older workers, as well as making early retirement less attractive for both workers and employers. 5. Initially, the minimum length was set to twenty- five years. It has been gradually increased to thirty years for men and twenty- six years for women. In 2012, it it set to increase to thirtyfive and twenty- eight years for men and women, respectively.

Disability in Belgium: There Is More Than Meets the Eye

257

placement of the worker, and so on for different sectors and companies. Though legislation from 1986 as well as the ISP of 2005 tried to harmonize and put constraints onto this situation, numerous exceptions persist with respect to the general rules. Even today, while workers below the age of sixty do in theory have to be replaced when they are put onto early retirement, this is not the case for companies that are considered in economic difficulty or in restructuring—with the age sometimes being as low as fifty. The same applies if the company is closing or unable to find a suitable replacement and so on. Similarly, while the age of fifty- eight is currently the statutory minimum access age, a lower age is possible in some sectors (steel, glass, textile, etc.) depending on more stringent career conditions or conditions on hardship of the job. 6.2.4

Unemployment and Old-Age Unemployment

The last exit path out of the labor market works through the unemployment insurance system. Two alternatives exist, the specific old- age unemployment (OAU) system and the regular UI system. Eligibility for the regular UI system is essentially based on having paid contributions during at least twelve months in paid employment or assimilated periods in the last eighteen months. The basic UI system is available to workers of all ages and pays out benefits of 60 percent of the last gross wage. Both the reference wage and the benefits have caps and floors. As a result, the replacement rate of the unemployment benefits was 30 percent of average gross wages in 2004 (Faniel 2008). Benefits are not subject to income tax and are not limited in time with payments ending upon reaching the normal retirement age of sixty- five. At this age, the transit into the pension system follows rules similar to the DI system. Under UI rules, individuals have to be available for the job market and actively look for employment. A specificity of the system is that benefits, as for the rest of social security, are paid out by a federal agency, while job placement is only in the hands of the various regional authorities. The OAU status was introduced back in 1985 into the Belgian social insurance landscape. In line with the previously discussed programs, its motivation was essentially one of freeing up space on the job market in a lump- of-labor logic (see Jousten et al. 2010). As compared to the regular UI system, several key factors of differentiation exist. First, the beneficiary of OAU benefits does not have to be actively looking for a job, nor does he or she have to be available for the labor market. Second, benefits are more generous than UI. In addition to the UI benefits, the old- age unemployed receive a complement that is a function of age and family status. To qualify for OAU, workers have to satisfy age and unemployment criteria. Initially, the system was limited to those people aged fifty- five and above who were unemployed for at least two years; as of 1996 the age was lowered to fifty and the UI duration to one year. In 2004, faced with the

258

Alain Jousten, Mathieu Lefebvre, and Sergio Perelman

growing importance of these arrangements, the government progressively introduced changes to the legislation. New entrants currently have to satisfy more stringent conditions to benefit from the job search and availability waiver. Under the new rules, only workers aged fifty- eight and above or with long careers still benefit from a full waiver, younger workers benefiting from an intermediate system as compared to the basic UI system.6 In its most extreme version, the system has been used in what is known as Canada Dry retirement arrangements (Cremer, Lozachmeur, and Pestieau 2009), where employers separate from older workers by sweetening the firing with lump- sum compensations or side payments—this way avoiding the more complicated and stringent severance pay. 6.3

Historical Health Trends

After having reviewed the institutional setting, it is useful to review some indicators of outcomes in terms of health and mortality. Figure 6.1 presents the general trend in terms of mortality for the population aged fifty- five to sixty- five in Belgium. While there has been some variation across sexes and age groups, the dominating trend is clearly a steep decrease in mortality or, expressed differently, a steep increase in life expectancy. Though encouraging, it does, however, not mean that morbidity also decreased. Two additional figures give a snapshot picture of the strong improvements in mortality. The first indicates the age at which members of the age sixty and sixty- five cohorts in the various calendar years will suffer from a mortality rate that their 1960 counterparts suffered when sixty and sixty- five, respectively. The increasing trend displayed in figure 6.2 indicates a clear shift toward longer life expectancies as higher mortality rates occur much later in life, about ten years from 1960 to 2006. The second compares mortality rate for men and women in 1960 and 2005. Considering a 5 percent mortality rate, figure 6.3 shows that the age at which this rate is attained shifted from sixty- nine to seventy- six for men and from seventy- three to eighty for women, a gain of seven years in both cases. Beyond pure mortality indicators, we only have very limited health data available. The first one is the Eurobarometer—a cross- national survey carried out over a wide range of European countries. A question on subjective health has been asked periodically for the years 1987, 1989, 1990, 1993, and 2001. However, a major drawback is that the question is not expressed in the same way, and the scale of answers has also changed overtime. The second survey is the Belgian Health Interview Survey (HIS) that exists since 1997

6. Starting in 2002, the age at which people can claim the old- age unemployed status is increased to fifty- eight, with a transitory period in which it is first fifty- six then fifty- seven. To exit the labor market before these ages, the workers have to justify a career of thirty- eight years.

Fig. 6.1

Mortality rates for male (left) and female (right)

Source: Human Mortality Database.

Fig. 6.2

Ages of equal mortality

Source: Human Mortality Database.

Fig. 6.3

Mortality rates by age

Source: Human Mortality Database.

260

Alain Jousten, Mathieu Lefebvre, and Sergio Perelman

Fig. 6.4 Good health (Eurobarometer) and mortality for male (left) and female (right) Sources: Eurobarometer and Human Mortality Database. Note: The subjective health indicator gives the percentage of people who declare to be in fairly good, good, or very good health.

and is conducted every four years. For the latter, we only have access to data by ten- year age cohorts. Looking first at the Eurobarometer data of figure 6.4 and comparing with the underlying mortality trends also plotted on the graphs, it is clear that this data has only a very limited use—if only because of the changing questions over the different waxes. While the mortality rates show a clear improvement along the period, the perceived health displays no clear pattern. The HIS provides self- reported subjective and objective health measures. The advantage is that the HIS allows to link mortality to morbidity. Figure 6.5 displays the percentage of people aged fifty- five to sixty- four who declare being in a bad health, those having functional limitations, those having chronic conditions, and the average body mass index in this population. It show that, contrary to the drop in mortality also reported, health measures look rather flat—except maybe for the subjective health measure that seems to display a slightly decreasing trend, especially for women. Given the sparse information on direct health outcomes, we conclude that mortality is likely the only meaningful indicator of health improvements in Belgium. Also, it is the only available indicator covering a rather long period and showing consistent estimation along time. To get a sense of the importance of longevity changes on labor force participation, we present cross- tabulations of two key labor force indicators against mortality. Implicitly, we want to verify whether changes in the average health and well- being of the population affect the labor market. In theory, if the DI system were to play its role as a protection against bad health, both variables should be affected—particularly for workers close to the retirement age. Figure 6.6 plots the trends of mortality against the employment rate and DI receipt. The employment rate is derived using the

Health measures (HIS) and mortality for male (left) and female (right)

Sources: Belgian Health Interview Survey and Human Mortality Database. Note: The subjective health indicator gives the percentage of people who declare to be in bad or very bad health. The chronic condition indicator gives the percentage of people who declare to have one or more long- standing health condition. The functional limitation indicator gives the percentage of people who have functional limitations. The BMI is the average body mass index in the specific population.

Fig. 6.5

Employment versus mortality versus disability at age sixty to sixty-four for male (left) and female (right)

Sources: Human Mortality Database; INAMI-RIZIV; and Labor Force Survey.

Fig. 6.6

Disability in Belgium: There Is More Than Meets the Eye

263

data from the Labor Force Survey that has been conducted since the early 1980s, while the DI rate is based on data from the federal DI administration.7 For men, mortality has faced a steady decline over this twenty- five- year period, while the employment rates have started to edge upward since the end of the 1990s—a trend more broadly observed all across the Organization for Economic Cooperation and Development (OECD). The same is true, though to a lesser extent, for DI in the second half of the first decade of the new millennium. For women, the upward trend in employment is much more sustained and can largely be explained by an increasing fraction of women joining the labor market and becoming eligible for benefits. Furthermore, the inexistence of DI recipiency until the late 1990s and the steep increase thereof at the end of the observation period can be explained by the increase of the retirement age from a departure level of sixty—hence originally excluding all women above sixty from the benefits of the DI system. One conclusion emerges: DI receipt and employment rates do not correlate well with mortality or health indicators. 6.4

Pathways to Retirement and Program Reform

This section extends the previous analysis and looks beyond the narrow concept of DI benefit receipt as an indicator of program participation. It also expands the focus beyond the age group sixty to sixty- four as similar effects on activity levels can also affect younger and older cohorts. 6.4.1

Key Labor Market Indicators

Figures 6.7, 6.8, and 6.9 give an overview of labor market outcomes over the period of time covered by the Labor Force Survey. These more extensive labor force data comfort the previous observations of long- term trends on the labor market, with male rates being relatively stable over the entire period of observation, while female employment and labor force participation strongly trend upward. While unemployment rates vary considerably over the economic cycle, there is little impact on overall employment and labor force participation levels. Beyond these traditional sets of labor market indicators, we propose two additional measures. Figure 6.10 graphs the same Labor Force Survey employment data for the ages ranging from fifty to seventy in three different calendar years. It confirms the preceding observations. A final tabulation links the employment rate to mortality rates in different calendar years. Every point represents the position of a five- year age bracket in the given year. The results of figure 6.11 shows that employment rates at any given mortality rate have steadily dropped for men. The reason for this 7. We rely on administrative data given that Labor Force Survey data on disability are only available for the period starting in 1992 and are rather erratic.

Fig. 6.7

Labor force participation for male (left) and female (right)

Source: Labor Force Survey.

Fig. 6.8

Employment rate for male (left) and female (right)

Source: Labor Force Survey.

Fig. 6.9

Unemployment rate for male (left) and female (right)

Source: Labor Force Survey.

Disability in Belgium: There Is More Than Meets the Eye

Fig. 6.10

265

Employment rate five-year age bracket for male (left) and female (right)

Source: Labor Force Survey.

Fig. 6.11

Employment rate and mortality rate for male (left) and female (right)

Source: Labor Force Survey and Human Mortality Database.

is that as time went by, any given mortality level corresponded to a higher and higher age, which, in turns, is associated with a lower employment rate. For women, beyond this reduced mortality effect, there is a second strong generational factor at work, involving a steeply increasing labor force participation over the observation period. 6.4.2

Social Security Program Participation

Disability Insurance While the preceding Labor Force Survey data have the distinct advantage of giving a close reflection of the individuals’ true economic status, they are not able to reflect the richness of the Belgian institutional landscape. Therefore, it should not have come as a surprise that the aggregate data of the previous subsection do not seem to reflect any of the specific social security reforms we discussed in section 6.2.

266

Alain Jousten, Mathieu Lefebvre, and Sergio Perelman

In order to investigate the effect of institutions on labor market participation, we draw on additional administrative data. In this context, program participation or “take-up” is defined as the number of individuals receiving benefits within the context of each program divided by the population in the corresponding cohort. To identify the various reforms to the social security schemes described in section 6.2, we draw vertical lines in the graphs plotting participation by age across time. These horizontal lines serve as a guide as to when the various reforms should have their effects or, at the very least, start having effects. Looking at the DI program participation presented in figure 6.12, we observe different trends for male and female. Male DI rate for the age groups forty to forty- four, fifty to fifty- four, and fifty- five to fifty- nine is almost constant, while disability for those aged sixty to sixty- four is decreasing. None of the reforms to generosity and eligibility of DI seem to have had an impact on participation. However, reforms concerning access to other program may well explain part of the decrease—particularly for those aged sixty to sixtyfour. For example, this could have been the case with the opening up of the OAU route in 1985 and the introduction of a flexible age of retirement without actuarial adjustment factor in 1991. The positive trend for women of all age cohorts follows a different rationale mostly because of a strong expansion of the share of women in the labor market. While reforms to OAU and other early retirement routes affect women less due to insufficient work histories to qualify for such benefits, the increase of the female normal retirement age was not without effects such as previously discussed. The Belgian disability administration prefers to use a different reference population when determining participation. Their reference population is composed of those individuals that are eligible to claim benefits because they contribute, minus those on CER because they do not have any economic incentives to switch to the DI system.

Fig. 6.12

DI participation for male (left) and female (right)

Source: INAMI-RIZIV.

Disability in Belgium: There Is More Than Meets the Eye

Fig. 6.13

267

Adjusted DI participation for male (left) and female (right)

Sources: INAMI-RIZIV; ONEM; and BNB-Belgostat.

If one follows this logic of excluding those that are either not eligible or have no economic interest to join the DI ranks, then this definition of the population is still too broad. For example, it is highly likely that the individuals on UI and OAU also do not have much of an economic incentive to switch—given the previously discussed absence of explicit time limits in benefit receipt. Using this adjusted population figure as a reference point, we derive equivalent participation trends over time (figure 6.13). The figures reveal much more sharply the effect of the reforms discussed in section 6.2. The increase in the age groups fifty- five to fifty- nine after 1986 corresponds to the introduction of the old- age unemployment status, which strongly reduces the denominator of the disability rate by shrinking the population that has an economic interest in claiming DI. Also, the initial trend observed for the age group sixty to sixty- four can be traced back to the increase in the number of retirees until 1987, thus shrinking the reference group. A similar logic applies to the early 1990s with its major reforms to the retirement system. Summing up the information of figures 6.12 and 6.13, we can draw two conclusions. First, the trends in DI participation over time are not confined to the age group close to retirement, but also younger cohorts are affected. Second, while aggregate DI participation rates have stayed unchanged, there have been substantial changes to the people “at economic risk” of filing for DI benefits, notably because of reforms to the eligibility and generosity of other social security programs targeted at the older segments of the labor force. Other Social Security Programs The effect of other program is, however, not limited to the one of restricting the population economically at risk of filing for benefits. Indeed, the multiplicity and generosity of other exit routes make it possible that truly

268

Alain Jousten, Mathieu Lefebvre, and Sergio Perelman

disabled people file for other social security benefits such as CER, OAU, or UI in lieu of filing for DI. In this sense, the question of evaluating the participation and the changes in participation over time is of prime importance. We first focus on the regular UI program. As with the just discussed DI data, the UI data comes from the federal UI administration. The number of unemployed thus corresponds to the number of individuals currently claiming UI benefits under the rules in place. Given the institutional setup, this means that some truly unemployed people may not be eligible for benefits, while some segments of those receiving benefits would not be considered unemployed according to the standard definition of the International Labor Organization (ILO). The time pattern of UI participation of figure 6.14 displays three clearly separated phases over time. Contrary to DI numbers, these rates seem to be independent of the demographic trend and are more heavily exposed to cyclical variations in the economy. Economic cycles don’t, however, provide a complete picture. The stark decrease of UI participation after 1984 can be traced back to the introduction of OAU such as identified by the first vertical line. The shift into OAU is, however, not total, if only because some people don’t satisfy the somewhat more stringent entry criteria for the OAU system as compared to UI. Similarly, the further liberalization and expansion of the system in 1995 corresponds to the next drop of people on the UI rolls—an effect further reinforced by the arrival of the Canada Dry pensions in 1997. Finally, the third phase corresponds to the tightening of the access to OAU in 2002—with an ensuing communicating vessels effect increasing UI as OAU decreases (see also figure 6.15). However, the decrease in the number of unemployed is lower than the increase in the number of old- age unemployed—as a quick glance at the rates presented in figures 6.14 and 6.15 document. Thus, OAU clearly ap-

Fig. 6.14

UI participation for male (left) and female (right)

Source: BNB-Belgostat.

Disability in Belgium: There Is More Than Meets the Eye

Fig. 6.15

269

OAU participation for male (left) and female (right)

Source: ONEM-RVA.

pears as a new exit path with a net addition of exits and not only reallocations between programs. As compared to their male counterparts, the situation for women is not very different, as the preceding figures 6.14 and 6.15 clearly illustrate. Unsurprisingly, the pattern for the sixty to sixty- four age group displays a somewhat different pattern than for men, again for the same reason as before: the initially lower retirement and early retirement ages for women, with the ensuing labor market effects after their progressive increase starting in 1997. Finally, figures 6.16 and 6.17 present the trends regarding CER and the regular pension system participation. The CER trends for men of all age groups are characterized by a strong responsiveness to changes in the CER and OAU eligibility affecting the relative attractiveness of this system. Women benefit much less from the system, given the career length requirements that are often a nontrivial hurdle in case of incomplete careers. The two reforms to the main pension system modifying the early and normal retirement ages for women have the expected effects on CER and the pension scheme—with shifts toward CER in the affected groups. For men, the trends in the regular pension system reflect reforms rather closely until the end of the last millennium. For example, the increase for ages sixty to sixty- four over the first half of the 1980s is related to the SER and RPP measures. The observed path after 2000 cannot be traced back to any specific legislative change and coincides with general improvements on the labor market leading to an increase in the effective age of retirement. Similarly, the trends for the female age group sixty to sixty- four before the progressive increase of the retirement age as of 1997 is difficult to link to any specific change: the participation rates being stocks, they reflect the cumulative weight of past reforms. Thus, changes to the OAU, the UI, and the CER have no doubt had an influence. However, they are insufficient to explain the sheer size of fluctuations observed in the pension scheme that might partially

270

Alain Jousten, Mathieu Lefebvre, and Sergio Perelman

Fig. 6.16

CER participation for male (left) and female (right)

Source: INAMI-RIZIV.

Fig. 6.17

Retirement participation for male (left) and female (right)

Source: ONP-RVP.

also be due to interactions of incentives and benefits for couples as well as the sometimes unpredictable effect of incomplete careers on benefit eligibility. Figure 6.18 presents the information on program participation in a synthetic way for all available social security systems. Figure 6.18 highlights the connected vessels aspect of the various social security schemes with a strong role for OAU. Also, by now, OAU and UI combined account for more than 35 percent of all workers in the fifty to sixty- four age bracket making the unemployment benefit receipt the single most prominent retirement status. The fraction even rises to well beyond 60 percent when integrating the CER schemes that are also administered through the UI scheme. 6.4.3

Mortality, Health, and DI

Figure 6.19 displays the overall enrollment or participation in the different systems on a year- by- year basis. The results show the relative constancy of

Distribution of exit paths (%) for male (left) and female (right)—aged fifty to sixty-four

Sources: INAMI-RIZIV; ONEM; BNB-Belgostat; and ONP-RZV. Note: Based on the number of people in benefit receipt in the various schemes.

Fig. 6.18

Participation in the systems for male (left) and female (right)—aged fifty to sixty-four

Sources: INAMI-RIZIV; ONEM; BNB-Belgostat; and ONP-RZV. Note: Based on the number of people in benefit receipt in the various schemes.

Fig. 6.19

Disability in Belgium: There Is More Than Meets the Eye

273

the DI scheme—with enrollment rather stable for males and growing progressively for females. Two interpretations are possible. First, the optimistic one would argue that entry into the DI program is so severely limited by powerful and effective gatekeeping mechanisms making it close to impossible to enter the system unless a serious medical or psychological condition exists. In this vein, stability is the result of good management, and the growth of the female participation is attributed to the increase in the normal retirement age, which mechanically increases the number of DI beneficiaries. The second and less optimistic interpretation is that other social insurance programs are sufficiently generous in terms of access and benefit levels, thereby rendering the more stringent DI system less attractive. The data of figure 6.20 raise further doubts with respect to a contemplative view of DI as a severely controlled and rather steady system. When decomposing by primary cause of DI benefit payment, one notices that mental health issues become a more and more prevalent issue as compared to purely physical ones. This leads to interesting avenues for future research: could DI payments be lower if more severely tested? Do other programs also have a changing pool of beneficiaries, maybe including increasing numbers of truly disabled people? This idea is further reinforced when looking at the relative prevalence of DI participation in the populations aged forty to forty- four and sixty to sixty- four, the former not being eligible for other early retirement programs. These data are represented in figure 6.21 and show steep decline in the difference between the two groups—hence indicating a relatively lower prevalence of DI cases in the older population as time went by. Though these data don’t allow us to conclude formally on the shifting of true DI recipients onto other programs, such a phenomenon is highly likely. 6.5

Conclusions

Disability insurance has attracted a limited interest in the Belgian policy debate. This is no doubt influenced by the relatively limited budgetary cost of this program, when comparing it to its bigger health and public pension siblings. The present chapter shows that disability insurance indeed seems to have a rather unspectacular time trend over the last decades. We show that the other social security programs allowing an early exit from the labor market play the key role in the Belgian retirement context. As compared to younger cohorts, older workers tend to benefit from disability payments less and less frequently. Furthermore, the composition of the disabled population is changing over time, with increasing numbers of benefits being based on mental problems that are intrinsically harder to check than physical impairments. This has two policy implications. First, in spite of its appearance, the

Percentage of disabled individuals by cause, male (left) and female (right)—aged fifty to sixty-four

Source: INAMI-RIZIV.

Fig. 6.20

Disability in Belgium: There Is More Than Meets the Eye

Fig. 6.21 for men

275

DI participation aged sixty to sixty-four less aged forty to forty-four

Source: INAMI-RIZIV.

increasing role of harder- to-check benefit decisions may well mean that the disability program is serving as an early retirement route. Second, substantial numbers of disabled may, in fact, be using other social security programs to exit the labor market. The latter would mean that the various social insurance programs are getting more and more diverted from their original aims—in the Belgian context, particularly the unemployment system. Additional research on the underlying health characteristics of the beneficiaries of the various social security programs would be warranted to guide any future reforms. A further area of investigation could be a systematic analysis of the work environment that people face during their active life. This environment has no doubt undergone profound changes over the last decades, both because of a seminal trend toward the service sector and stricter general work- safety requirements in place.

References Cremer, H., J. M. Lozachmeur, and P. Pestieau. 2009. “Use and Misuse of Unemployment Benefits for Early Retirement.” European Journal of Political Economy 27 (2): 174– 85. Dellis, A., R. Desmet, A. Jousten, and S. Perelman. 2004. “Micro-Modelling of Retirement in Belgium.” In Social Security and Retirement Around the World: Micro-Estimation, edited by J. Gruber and D. Wise, 41– 98. Chicago: University Chicago Press. Desmet, R., A. Jousten, S. Perelman, and P. Pestieau. 2007. “Micro-Simulation of Social Security in Belgium.” In Social Security Programs and Retirement Around

276

Alain Jousten, Mathieu Lefebvre, and Sergio Perelman

the World: Fiscal Implications of Reform, edited by J. Gruber and D. Wise, 43– 82. Chicago: University Chicago Press. Disney, R., C. Emmerson, and M. Wakefield. 2006. “Ill Health and Retirement in Britain: A Panel Data-Based Analysis.” Journal of Health Economics 25 (4): 621– 49. Faniel, J. 2008. “Unemployment Insurance between European Pressures and Regional Polemics.” Chroniques internationales de l’IRES 115 (November): 52– 64. Jousten, A., M. Lefebvre, S. Perelman, and P. Pestieau. 2005. “Social Security in Belgium: Distributive Outcomes.” IZA Discussion Paper no. 1486. Bonn, Germany: Institute for the Study of Labor. ———. 2010. “The Effect of Early Retirement on Youth Unemployment: The Case of Belgium.” In Social Security Programs and Retirement Around the World: The Relationship to Youth Employment, edited by J. Gruber and D. Wise, 47– 76. Chicago: University Chicago Press. Kalwij, A., and F. Vermeulen. 2008. “Health and Labour Force Participation of Older People in Europe: What Do Objective Health Indicators Add to the Analysis.” Health Economics 17 (5): 619– 38. Pestieau, P., and J. P. Stijns. 1999. “Social Security and Retirement in Belgium.” In Social Security Programs and Retirement Around the World, edited by J. Gruber and D. Wise, 37– 71. Chicago: University Chicago Press. Walker, T. 2007. “Why Economists Dislike a Lump of Labor.” Review of Social Economy 65 (3): 279–91.

7 Disability, Pension Reform, and Early Retirement in Germany Axel Börsch-Supan and Hendrik Jürges

7.1

Introduction

Disability insurance—the insurance against the loss of the ability to work—is a substantial part of social security expenditures and an important part of the welfare state regime in all developed countries (Aarts, Burkhauser, and de Jong 1996). Like almost all elements of modern social security systems, disability insurance faces a trade- off. On the one hand, disability insurance protects unhealthy people who are not able to work from falling into poverty before they are eligible for normal retirement benefits. On the other hand, however, disability insurance creates incentives to exit the labor force early and may act as another pathway to early retirement without the incidence of a major health loss. The recipiency rates of disability insurance (DI) benefits vary strikingly across European countries; see Börsch-Supan and Roth (2010). Germany is in between countries of very high recipiency rates such as Sweden, Denmark, and the Netherlands (more than 12 percent), but with 6.5 percent of all fifty- to sixty- four- year- old persons substantially higher than in France (less than 2 percent). Three candidate causes of this international variation come to mind: cross- national differences in the age structure, cross- national differences in health, and cross- national differences in the early retirement incentives created by the DI system. Based on the 2004 waves of the Survey of Health, Ageing and Retirement in Europe (SHARE), the English LonAxel Börsch-Supan is director of the Mannheim Research Institute for the Economics of Aging at the University of Mannheim and a research associate of the National Bureau of Economic Research. Hendrik Jürges is professor of health economics and management at the Schumpeter School of Economics and Business, University of Wuppertal. For acknowledgments, sources of research support, and disclosure of the authors’ material financial relationships, if any, please see http://www.nber.org/chapters/c12388.ack.

277

278

Axel Börsch-Supan and Hendrik Jürges

gitudinal Study of Ageing (ELSA), and the Health and Retirement Study (HRS), we showed that cross- sectional differences in demographic structure and current health status cannot explain the cross- national differences in DI recipiency (Börsch-Supan 2005) although health explains a great deal of the within- country variation (see also Avendano and Mackenbach 2010). Using the same data, we showed that adverse health events in a two- year observation window do not significantly trigger a higher probability of becoming a DI benefit recipient at the end of this window (Börsch-Supan 2008). Finally, we exploited data on childhood and midlife health only to find that also these health measures, while capturing a large share of the within- country variation, do not explain the cross- national differences in DI recipiency (Börsch-Supan and Roth 2010). The current chapter concentrates on time series variation in health and DI uptake or early retirement in Germany. It follows the structure of the other chapters in this volume and has two aims: • To provide historical information on mortality and “health” status measures in each country. • To understand the relationship between changes over time in disability program reforms in particular—as well as other social security programs—on the one hand and DI program participation on the other hand. Section 7.2 describes the most relevant features of the German pension system and which reforms the system has undergone since the 1970s. Section 7.3 shows long- term trends in mortality, primary diagnoses for DI benefit recipients, and subjective health measures together with long- term trends in labor force participation. In section 7.4, we analyze how selected reforms of the pension system, in particular introductions of or changes in the generosity of various early retirement options, have affected the retirement behavior in Germany. Section 7.5 tracks early retirement behavior by birth cohorts and shows that cohorts born during World War I have retired early than cohorts born before or after the war. Section 7.6 concludes the chapter. 7.2

Regimes of Retirement Policies in Germany

The German pension system began as a funded disability insurance scheme some 120 years ago, but was quickly broadened into a general oldage pension system and was transformed into a pay- as-you- go system in 1957 after about half of the capital stock was lost in two world wars and a hyperinflation. As opposed to other countries such as the United Kingdom and the Netherlands, which originally adopted a Beveridgian social security system that provided only a base pension, public pensions in Germany are designed to extend the standard of living that was achieved during work life also to the

Disability, Pension Reform, and Early Retirement in Germany

279

time after retirement: individual pension benefits are, therefore, proportional to individual labor income averaged over the entire course of the working life and feature only few redistributive properties, in particular a minimum pension at the social assistance level. Benefits in the disability branch are identical to benefits for old- age pensions. They are, however, calculated as if the working life had extended to the early retirement age. Subsection 7.2.1 adds further details to this bird’s- eye view of the German pension system, and subsection 7.2.2 describes the changes in legislation that will be used to identify the incentives generated by the system. 7.2.1

Pathways to Retirement in Germany

From the very beginning, the German system has distinguished “oldage pensions” from “disability pensions.” Old- age pensions have an earliest retirement age as fundamental eligibility criterion, independent of the ability to work, while disability pensions require an assessment of the inability to do work, independent of age. In practice, the difference is less clear cut. Disability pensions require a work history of at least five years of contribution, of which three years have to be during the last five years. The assessment of the inability to work is typically record- based, that is, there will be no specific medical exam but a submission of medical records by a general practitioner or a company doctor, potentially but rarely audited by a government official. Old- age pensions have various pathways, summarized in table 7.1, including an “old- age disability pension” characterized by a much more lenient medical check. In addition, there are several preretirement options, such as the “unemployment channel,” where a workers first receives unemployment benefits (possibly augmented by a severance payment by his former employer) until age fifty- nine, then switches to an old- age pension for the unemployed, which is converted to a standard pension after age sixty- five. In the sequel of the chapter, we will refer to “retirement” in a strict sense as receiving an old- age or disability pension and not in the sense of exiting the labor force, for example, via the unemployment channel. While, in principle, all old- age and disability pathways to retirement genTable 7.1

Standard Long-term insured Disabled workers Unemployed Part-time retirement Women

Pathways to old-age pensions Earliest age

Years of contribution

Other

65 63 60 60 60 60

5 35 35 15 (8 in last 10 years) 15 (8 in last 10 years) 15 (10 after age 40)

None Actuarial adjustment No actuarial adjustment At least 52 weeks unemployed 2 years part-time

280

Axel Börsch-Supan and Hendrik Jürges

erate the same retirement income, the pathways differ by the eligibility age and the various extra requirements indicated in table 7.1. Moreover, disability pensions are attractive because they feature an upper limit of actuarial adjustments. These are 3.6 percent per year but are limited to 10.8 percent in case of a disability. The choice among these pathways thus has three components that are well- defined and straightforward to model (earliest retirement age, minimum employment requirement, and actuarial adjustment) plus “soft” factors such as the lenience of the medical exam and a possible deal between employer and employee regarding severance pay and a recommendation by the company doctor that are hard to detect in the usual data sources. 7.2.2

Regimes of Retirement Policies in Germany

The system has experienced many changes, and the pathways in table 7.1 refer essentially to the time between 1972 and 2002, during which most of our analysis takes place. There were, however, several important changes that we will exploit to identify the incentives exerted by the German pension system. It should be noted that the multitude of pathways is currently being simplified while the standard retirement age is increased to age sixty- seven. The following brief history of the German pension system distinguishes five phases:1 (a) a relatively stable phase after the introduction of the payas-you- go system until 1972; (b) a phase of increasing generosity precipitated by the 1972 pension reform; (c) a phase of modest retrenchment, especially affecting disability benefits in the mid- 1980s; (d) a phase of cost- cutting reforms after 1992 leading to a sustainable pension system by 2007; and (e) first signs that we may actually experience a phase of reform backlash. Phase 1 (1957 to 1972): Stability Initially, the pay- as-you- go system introduced in 1957 had a single eligibility age for old- age pension: age sixty- five for men and age sixty for women (conditioned on a minimum number of years of service). Earlier retirement was impossible unless one could prove a disability. Disability rates were very high after World War II and then declined. As figure 7.1 shows, disability insurance was the main entry path into the German pension system until 1972 for both men and women. Phase 2 (1972 to 1984): Increasing Generosity The 1972 reform was a major change in policy. It introduced “flexible retirement for the long- term insured” by providing old age pension benefits at age sixty- three, given that workers had a minimum of thirty- five years in which they contributed to the system. These benefits were not actuarially 1. For a detailed description of the evolution of the German pension system, see BörschSupan and Wilke (2006).

Disability, Pension Reform, and Early Retirement in Germany

Fig. 7.1

281

Pathways to retirement

Source: Deutsche Rentenversicherung Bund, Rentenzugangsstatistik.

adjusted. Average retirement age dropped by more than two years (BörschSupan 2000), and the “flexible retirement” pathway partly substituted for the disability pathway into retirement (see figure 7.1). At the same time, the “old- age pension for disabled workers” was introduced, first with a earliest entry age of sixty- two, then, after 1978, in two steps to age sixty. Between 1984 and 1987, early retirement was further extended by creating a “bridge to retirement.” The government introduced more generous unemployment insurance benefits for older workers that were especially attractive in the age rage from fifty- five to fifty- nine years: up to thirty- two months of unemployment insurance benefits at 63 or 68 percent of former net wages. These benefits were neither means tested nor were job- search activities required for those unemployed who were aged fifty- five and older. In addition, severance pay became tax advantaged for the employers. As a result of the “bridge to retirement,” the pathways to retirement changed again: registered unemployment of elderly (age fifty- five to fifty- nine) rose—particularly dramatically between 1991 and 1996—and the uptake of disability benefits declined (see figure 7.1). Phase 3 (1984 to 1992): Modest Retrenchment In 1984, the balance between old- age and disability pensions was changed by reducing the eligibility requirement for old- age pensions (at regular retirement age sixty- five) from fifteen to five contribution years. At the same

282

Axel Börsch-Supan and Hendrik Jürges

time, restrictions on the eligibility for disability pension were strengthened. This included the introduction of a minimum of three contribution years in last five years and stricter medical examinations. Phase 4 (1992 to 2007): Sustainability Reforms Threatened by demographic change, Germany began in the early 1990s a fifteen- year lasting process of reform steps. These reform steps were not masterminded; some “happened” due to budget crises and new political constellations. Seen from hindsight, however, the reform steps follow an astoundingly consistent red threat. Step 1: Toward Actuarial Adjustments (1992). The first step in the long German reform process was the 1992 reform. It anchored benefits to net rather than to gross wages. This removed an odd mechanism that would have created a vicious cycle of increasing pension benefits in response to increasing contribution rates. At the same time, credits for higher education were abolished and survivor benefits reduced. The second important element in the 1992 reform was the introduction of “actuarial” adjustments to benefits to retirement age. Actuarial is set in quotes because the adjustments factors have been set discretionarily at 3.6 percent for each year of earlier retirement and are not directly linked to changes in life expectancy. They are about 1.5 percentage points lower than current life tables and a 3 percent discount rate would imply.2 Nevertheless, their gradual introduction between 1998 and 2006 reduced incentives to retire early, and retirement age and labor force participation of older individuals has indeed increased since then, almost symmetrically to the decline after the 1972 reform (see Börsch-Supan [1992] for an early prediction of this effect). Step 2: Toward a Genuine Multipillar System (2001). The financial situation of the pension system worsened rather quickly after the 1998 elections that brought the Social Democrats to power in Germany. As a remarkable irony in politics, the former union leader then secretary of labor Walter Riester successfully passed a major reform bill through parliament in 2001.3 The Riester reform is a major change of the German public pension system. It changed the monolithic pay- as-you- go retirement insurance to a genuine multipillar system by partially substituting pay- as-you- go financed pensions with funded pensions. The reform aimed to achieve three main objectives. First, the reform was to stabilize contribution rates. The Riester reform law actually states that contribution rates to the public retirement 2. Actuarial computations depend on a discount or interest rate that makes payments made or received at different points in time commensurable. Usually, a rate of 3 percent is assumed, sometimes 4 or 5 percent. The German computations rest on a discount rate of about 1 percent. 3. The 2001 reform is, therefore, popularly referred to as the Riester reform.

Disability, Pension Reform, and Early Retirement in Germany

283

insurance scheme must stay below 20 percent until 2020 and below 22 percent until 2030, while the net replacement rate must stay above 67 percent. Failure must precipitate further government action. Second, a new pillar of supplementary funded pensions was introduced. Contributions to this pillar are subsidized, either by tax deferral and tax deduction, or by direct subsidies. These supplementary pensions are, however, not mandatory. Third, benefits of the pay- as-you- go system were scheduled to be gradually reduced in proportion to the maximum subsidized contribution to the new supplementary pensions. Step 3: Toward Sustainability (2004). Although praised as a “century reform,” it quickly became obvious that the cost- cutting measures of the Riester reform would not suffice to meet the contribution rate targets. A new reform commission, the Commission for Sustainability in Financing the German Social Insurance Systems, was established in November 2002.4 Its twin objectives were those of the Riester reform: to stabilize contribution rates while at the same time ensuring appropriate future benefit levels. The Commission met in 2003 very different circumstances than Riester faced just a few years earlier. Unexpectedly high unemployment rates and the poor performance of the German economy with extremely low growth rates precipitated a short- run financial crisis of the pension system and created a sense of urgency for reform. Moreover, the electorate became increasingly aware that stabilizing social security contributions and thus limiting the increase of total labor compensation will be essential for enhancing future growth. This paradigm shift away from thinking in pension claims toward thinking in financing possibilities had a noticeable impact on the Commission’s reform proposals. The Commission proposed an entire reform package (Commission 2003). In addition to a gradual shift of the retirement age in proportion to the expected change of life length after retirement, the key element of the Commission’s reform proposal was a new pension benefit indexation formula linking benefits to the system dependency ratio, called “sustainability formula.”5 It will lead to further decreases in pension benefits vis-à-vis the path planned by the Riester reform. Most of the Commission proposals, and most significantly the introduction of the sustainability formula, were quickly passed by the German parliament in May 2004. In parallel, the government also passed major changes in the unemployment insurance system, called “Hartz reforms.”6 They dramatically short4. Popularly referred to as the Rürup commission after its chairman, Bert Rürup, the Commission was in charge of making reform proposals for the pension system, health care, and long- term care insurance. We only refer to the proposals of the pension group that was cochaired by one of the authors of this chapter. 5. Technical details are described in Börsch-Supan and Wilke (2006). 6. Peter Hartz, former chief personnel officer at Volkswagen, headed the commission.

284

Axel Börsch-Supan and Hendrik Jürges

ened the duration of unemployment benefits, especially for older individuals, to eighteen months (rather than thirty- two months) and made unemployment insurance much less attractive as a substitute for early retirement and disability insurance benefits. This was accompanied by shifting the age limit for “old- age pensions due to unemployment” to age sixty- three. Step 4: Toward Later Retirement Ages (2007). The Commission also proposed an increase of the normal retirement age from sixty- five to sixtyseven years according to a schedule from 2011 to 2035 reflecting expected future changes in life expectancy. The underlying rationale was to divide the lifetime gained in proportion to the current division between lifetime in work and in retirement, namely two- to-one. In order to prevent substitution into early retirement and disability pensions as a result of the increase in the retirement age, the Commission also proposed to increase the early retirement ages (to the same extent and on the same schedule as the normal retirement age) and to increase the actuarial adjustments for disabled and long- term insured workers. The shift in the retirement age was deemed politically too dangerous and was excluded from the legislation package in March 2004. The unions heavily opposed this adaptation of retirement age to life expectancy, using the argument that it would lead to higher unemployment and take jobs away from the young. Nevertheless, in yet another ironic move, just two years later, with population aging high on the political agenda, the then labor secretary Müntefering unilaterally announced an accelerated increase of the retirement age, being fully effective in 2029. It was legislated in March 2007. The age limit for “old- age pensions for disabled” was shifted to sixty- five years and the “old- pension for women” effectively phased out. Phase 5: Reform Backlash? The increase of the retirement age irritated the left wing and was watered down by exemptions for those workers who have forty- five years of service. This may be the beginning of a period of reform backlash. Under increasing pressure from the newly founded “Left Party,” the grand coalition government reverted the decision to shorten the duration of unemployment insurance benefits for older workers that was part of the “Hartz-IV” labor market reform. Moreover, the government decided in the spring of 2008 to make a two- year exemption from the sustainability formula to increase pension benefits in 2008 and 2009 when federal elections will be held. Finally, the issue of “blockwise partial retirement”—essentially an early retirement device—is back on the agenda. It is too early to judge whether these changes will end the phase of sustainability reform and begin a phase of reform rollbacks.

Disability, Pension Reform, and Early Retirement in Germany

7.3

285

Trends in Mortality, Health, DI Uptake, and Labor Force Participation

We start the empirical analysis of the relationship between health, DI uptake, and labor force participation by showing long- term trends in mortality, other measures of health, and labor force participation at older ages in Germany from the 1960s to today. We first document the substantial increases in life expectancy in the last fifty years. Then we relate those increases (as proxies for improvements in general health and—by extension— ability to work) to patterns of labor force participation. All numbers refer to West Germany only, unless stated otherwise. 7.3.1

Trends in Mortality

In this section we describe secular changes in mortality in Germany in the last four decades. All mortality data used in this study were drawn from the Human Mortality Database (www.mortality.org). Figure 7.2 shows trends in ages of equal mortality rates from 1957 to 2006. References are given by one- year mortality probabilities of sixty- and sixty- five- year- olds in 1960. Thus, sixty- (sixty- five)- year- olds have the mortality of sixty- (sixty- five)year- olds in 1960 by definition. Positive trends in age of equal mortality rates mean that individuals become older and older when they reach the mortality probability of a sixty- (sixty- five)- year- old in 1960. Presumably, mortality probabilities are a good proxy for current health; thus, rising trends in ages of equal mortality probability indicate improvements in population health over time.

Fig. 7.2

Ages of equal mortality probability, Germany 1957 to 2006

286

Axel Börsch-Supan and Hendrik Jürges

Figure 7.2 reveals a number of remarkable facts about mortality in Germany. First, mortality decreases (health improves) substantially between the 1960s and the 2000s. In 2006, a seventy- three- year- old man or woman had the same probability of dying within the next year as a sixty- five- year- old in 1960. Second, relatively speaking, men and women have fared equally well. Compared to 1960 mortality levels, neither sex has experienced a larger increase than the other. The time pattern, however, differs somewhat. Until 1970, men experienced hardly any improvement in mortality. Sixty- and sixty- five- year- olds in 1970 had practically the same mortality probabilities as sixty- and sixty- five- year- olds in the late 1950s. For women, we observe an increase of approximately 1.5 years. For both sexes, improvements in health or mortality have gained considerable momentum in 1970, after which we observe a nearly linear increase in ages of equal mortality. Finally, most observers will notice seemingly erratic fluctuations around the linear trend in the early 1980s (for reference age sixty) and the late 1980s (for reference age sixty- five). These fluctuations reflect changing mortality patterns among cohorts born during and after World War I. We will come back to the possible long- term effects of birth during times of war and hunger on health and disability uptake in section 7.5. In addition to just showing the improvement in longevity in the last fifty years, figure 7.2 can also be used for a simple thought experiment to put current discussions about raising the mandatory retirement age to sixty- seven in perspective. Let us imagine that—in order to create a sustainable pension system—the German government had decided in 1960 to link the future mandatory retirement age to the mortality risk at age sixty- five for men and sixty for women (the mandatory retirement age in the baseline year). The rationale would be to link the retirement age not to chronological age but to “functional age,” average health or life expectancy. Figure 7.2 shows how the mandatory retirement age would have increased in the last five decades. In 2006, men would have had to work until they were seventy- three years and six months old. Women would have had to work until they were sixty- nine years and four months old. Figure 7.3 gives a different representation of the gains in longevity. It shows one- year mortality rates by age for men and women in 1960 and 2005. Note that mortality rates are depicted on a logarithmic scale. The nearly linear increase in log mortality rates between age forty- five and age ninety thus reflects an exponential increase in mortality rates by age. Mortality rates of men are higher at all ages than mortality rates of women. Further, mortality rates at all ages are substantially smaller in 2005 than 1960. In 1960, a sixty- nine- year- old man had a probability of dying within the next year of 5 percent. In 2005, such a mortality rate was reached only at age seventy- six, representing a gain of seven years. For women, the respective numbers were seventy- three and eighty- one, representing an eight- year gain. Figure 7.4 shows relative improvements in mortality rates by age and

Disability, Pension Reform, and Early Retirement in Germany

Fig. 7.3

287

One-year mortality rates in Germany, by age and sex, 1960 and 2005

Fig. 7.4 Ratio of one-year mortality rates in Germany, by age and sex, 1960 and 2005

sex. Note that smaller numbers indicate larger improvements. Across the whole age range, women have gained more than men. However, there is hardly any difference in relative changes between age fifty and sixty- three, where mortality rates have roughly halved since 1960. The biggest relative improvements for women have been between ages sixty- four and eighty- five. Mortality rates in 2005 were nearly as small as one- third of the mortality rates in 1960. For men, the ratio of 2005 to 1960 mortality rates is substantially higher. Two findings in figure 7.4 are relevant for the explanation (or lack of

288

Axel Börsch-Supan and Hendrik Jürges

explanation) of trends in labor force participation at older ages in Germany. First, in the relevant age range, secular increases in life expectancy (and health) have favored neither sex. Second, the biggest relative improvements in life expectancy (and health) for men have been exactly at the relevant ages of fifty- nine to sixty- six. 7.3.2

Trends in Other Health Measures

Although mortality rates are arguably a good measure of health of a specific cohort at some point in time, it might not be the best proxy for work disability. Death is usually the endpoint of a longer- lasting disease that might have reduced work capacity of individuals’ years before their death. Further, certain diseases and conditions are known to reduce the ability to work in the labor market without being lethal. Depression and musculoskeletal diseases are two prominent examples. Moreover, as mortality declines further among cohorts that are still of working age, the relative importance of nonlethal and lethal conditions for early retirement behavior and disability uptake may shift dramatically. A weak or missing relationship between mortality and health does not necessarily imply that health is irrelevant for early retirement behavior. Figure 7.5 shows the development of the proportion of primary diagnoses for disability pension uptake between 1983 and 2008. The importance of cardiovascular diseases has diminished dramatically. In 1983, 37 percent of all men entered DI with cardiovascular disease as primary diagnosis. This percentage has dropped to 14 percent in 2008. Among women, the drop was even larger (from 38 percent to 6 percent). During the same period, the relative importance of mental illness has nearly quadrupled for men and quintupled for women. In 2008, mental illness was by far the most frequent

Fig. 7.5

DI uptake rates, by primary diagnosis and sex

Disability, Pension Reform, and Early Retirement in Germany

289

primary diagnosis (31 and 43 percent, respectively). Detailed statistics by International Statistical Classification of Diseases and Related Health Problems (ICD) codes reveal that, among women, 43 percent of those retiring early because of mental illness suffered from affective disorders (i.e., depression). Among men, alcohol- related psychiatric problems play an important role. In 2008, more men retired early because of alcohol- related psychiatric problems than because of ischemic heart disease (Deutsche Rentenversicherung 2009). Another noteworthy finding in figure 7.5 is the humpshaped trend for musculoskeletal diseases (mostly osteoarthritis and back problems). While being on the increase as primary diagnosis until the early 1990s, reaching primary diagnosis rates of about one- third, musculoskeletal diseases have become much less prevalent among both men and women and are now at 15 percent. Finally, cancer as primary diagnosis for DI uptake has nearly doubled between 1983 and 2008, increasing from some 7 percent to 15 percent for both sexes. To summarize, figure 7.5 clearly shows that, with the exception of cancer, nonlethal conditions are important drivers of work disability in Germany. Mental illness and musculoskeletal diseases account for about half of all men and three- fifths of all women retiring on DI. Mortality might thus not in all circumstances be the best proxy for work disability. Figures 7.6 and 7.7 show—by age and sex—trends in health satisfaction and self- rated health drawn from the German Socio-Economic Panel (SOEP). Data on satisfaction with health have been collected annually in the same format since the first wave of the SOEP in 1984. Data on selfrated health are available only since 1992. With few exceptions, it is hard to see any salient trends in the data. Health dissatisfaction (defined as rating one’s satisfaction with health below 5 on a 0-to-10 scale) and poor health (defined as rating one’s health as poor or very poor) are basically flat. Good

Fig. 7.6

Health dissatisfaction by sex and age category, Germany 1984 to 2006

290

Fig. 7.7

Axel Börsch-Supan and Hendrik Jürges

Self-rated health, by sex and age group, Germany 1992 to 2006

health (defined as rating one’s health as good or very good) seems to be increasing among sixty- to sixty- four- year- old men and women and fiftyfive- to fifty- nine- year- old women. While this pattern seems largely compatible with the secular increase in life expectancy, the downward trend in good health among fifty- to fifty- four- year- old men does not fit the general picture of improving health status. Moreover, due to sampling variability and other reasons, the data on health satisfaction and general health appear quite volatile. It is unlikely that these data could be usefully employed to explain more short- term trends in disability uptake and labor force participation. 7.3.3

Trends in Labor Force Participation

Trends in old- age labor force participation rates for men and women are shown in figure 7.8. The thin vertical lines indicate the years of various reforms. Looking first at the trend in labor force participation among men, it becomes obvious that the largest changes have happened in the sixty to sixty- four age range. Labor force participation has dropped from more than 80 percent in 1966 to less than 35 percent in 1980s and 1990s. This was not a long- term process, though. Between 1973 and 1979, labor force participation dropped from 71 percent to 39 percent. Of course, this should not come as a surprise. The 1972 reform addressed exactly the age groups before the official retirement age of sixty- five years. Since about the year 2000, we observe an increase in labor force participation in this age group. This is most likely a result of the 1992 reform, which is phased in only slowly and that raised legal retirement ages in almost all early retirement options (cf. Börsch-Supan 1992, 2000). Among men aged fifty to fifty- four and men

Disability, Pension Reform, and Early Retirement in Germany

Fig. 7.8

291

Labor force participation rates, by sex and age group

aged fifty- five to fifty- nine, figure 7.8 shows a steady decline in labor force participation rates between the years 1965 and 2000. Nothing happens on a similar scale as for sixty- to sixty- four- year- olds, and it is hard to see any connection between the various pension reforms (as indicated by the vertical lines) and changes in labor force participation at ages fifty to fifty- nine. Labor force participation of older women clearly follows a secular trend toward higher participation rates. Between 1960 and 2006, participation rates of women aged fifty to fifty- four doubled from 38 to 76 percent, and participation rates of women aged fifty- five to fifty- nine nearly doubled from 33 to 63 percent. Among women aged sixty to sixty- four, we observe a similar pattern as for men in that age range (although on a lower level). The highest labor force participation rate was reached in 1966 (24 percent) and dropped to around 12 percent in the 1980s and 1990s. Again, the largest part of the decline happened as a results of the 1972 reform. In recent years, however, we see a steep increase in labor force participation due to various changes in the course of the 1992 reform. Börsch-Supan (2000) shows that this change can be traced to the introduction of actuarial adjustments enacted with this reform. 7.3.4

Mortality and Labor Force Participation

Using mortality rates at a specific age as a proxy for health at a specific age, we now compare long- term trends in mortality and early retirement or DI uptake. We shift perspective by looking at the mortality and retirement at ages fifty to fifty- nine for German cohorts born between 1906 and 1943. The

292

Axel Börsch-Supan and Hendrik Jürges

retirement data are aggregated administrative figures collected by the German Pension Fund (available at www.fdz- rv.de). For cohorts born between 1906 and 1943, we know what proportion retired at what age. The data do not show which type of pension was drawn when individuals retired, but this can be more or less inferred from the retirement age. Men and women retiring before age sixty, which is what we look at in the following, can only draw DI benefits. If long- term trends in disability pension update were driven by long- term trends in health, and given the vast improvements in life expectancy or health seen in the preceding, we should expect to see a secular decline in early retirement rates. Figure 7.9 contrasts trends in mortality and early retirement rates for men and women. Ten- year mortality rates at age fifty have gone down from 10.3 to 6.5 percent among men and 5.7 to 3.3. percent among women. Again, we observe a nearly linear decrease with somewhat erratic movement for births around the time of World War I. In contrast, the longterm changes in early retirement are best described by a “down-up- down” pattern: first a decline for cohorts born before 1913, then a steep increase for cohorts born 1913 to 1925, and another decline for cohorts born after 1925. Among men, the proportion who retired aged fifty to fifty- nine decreased from 18.5 for the 1908 cohort to 16.8 percent for the 1912 cohort, then shot up to 22.9 percent for the 1925 cohort and decreased to 14.7 percent for the 1943 cohort. Among women, the proportion who retired aged fifty to fiftynine decreased from 16.4 (1908 cohort) to 12.7 (1912 cohort), surged to 18.8 (1924 cohort), and then fell to 9.6 percent (1943 cohort).

Fig. 7.9 Retirement and mortality rates at age fifty to fifty-nine, German cohorts born 1904 to 1943

Disability, Pension Reform, and Early Retirement in Germany

293

Fig. 7.10 Labor force participation rates, by age and mortality risk, men, Germany

Whereas the trends before 1913 and after 1925 are in line with the notion that early retirement behavior is largely driven by health, the increase between 1913 and 1925 clearly isn’t. On the contrary, cohorts were getting healthier and healthier while retiring earlier and earlier. As we show in the following (section 7.4.3), trends in labor force participation can at least partly be explained by changes in eligibility rules for disability pensions. Another way to show the relationship between age, mortality risk or health, and labor force participation over the years is presented in figure 7.10, which shows labor force participation rates by one- year mortality rates for 1970, 1980, 1990, and 2000, and contrasts this with participation rates by age. Panel A shows that labor force participation rates have decreased at all ages. The biggest shifts have happened between 1970 and 1980 and in the sixty to sixty- four age range (due to the 1972 and 1978 reforms). For instance, in 1970, 62 percent of all sixty- three- year- old men were in the labor force. In 1980, these were 36 percent and 24 percent in 2000. We know from the earlier analysis that life expectancy has increased substantially at all ages since 1970. Thus, despite better health, more and more men have left the labor market. To account for both types of changes, reductions in labor force participation and gains in longevity, at the same time, panel B shows labor force participation by mortality risk (and thus mutatis mutandis health levels). For instance, a man facing a 1 percent mortality risk had a likelihood of being in the labor force of 91 percent in 1970 but only 57 percent in 2000.

294

Axel Börsch-Supan and Hendrik Jürges

A man facing a 2 percent mortality rate had a likelihood of being in the labor force of about 80 percent in 1970 and 14 percent in 2000. Thus, conditional on health (as measured by mortality risk), old- age labor force participation has dropped dramatically in the three decades between 1970 and 2000. 7.4

How Specific Reforms Have Affected the Retirement Age of Different Cohorts

In this section, we illustrate by three examples how changing social security and disability insurance laws have changed the retirement behavior of German workers. Specifically, we look at the 1972, 1978, and 1984 reforms already described in section 7.2. We also look at retirement behavior from a different angle than before by exploring how reforms have affected the retirement age of different cohorts. Often reforms take effect sharply and by birth cohort, that is, individuals born before a certain cutoff date are not affected by a reform, while all individuals born after the cutoff date are affected. Straightforward before- after comparisons of retirement age by cohort thus allow us to assess the effect of those reforms. Because the changes in early retirement rules have such predictable and sharp temporary and permanent effects on early retirement and DI uptake, it is highly unlikely that sharp health changes happening simultaneously can explain those effect. As shown in the previous section, mortality rates have decreased steadily and fairly smoothly. 7.4.1

The 1972 Reform

As shown in the preceding, the 1972 reform, which took effect in 1973, increased the generosity of the German pension system by a substantial margin. The reform had two components: first, the retirement age of workers with a minimum contribution of thirty- five years (long- term employed) was reduced from sixty- five to sixty- three (without actuarial reductions). Second, disabled workers could draw old- age pensions (instead of disability pensions) from age sixty- two onward. The eligibility criteria and the year the reform took effect taken together allow us to identify the cohorts that have been affected by each reform component and to predict how the retirement age has been affected. The last cohort not affected by the reduction of the retirement age for the long- term employed from sixty- five to sixty- three years in 1973 are individuals born in 1908, that is, those who were already sixty- five in 1973. The first cohort affected by the reform are those born in 1909 who “suddenly” became eligible for old- age pensions at age sixty- four if they contributed for at least thirty- five years to the pension system. Cohorts born 1910 and later became eligible at sixty- three. Consequently, we expect to see a sudden jump in proportion of workers retiring at age sixty- four exactly for the

Disability, Pension Reform, and Early Retirement in Germany

295

Fig. 7.11 Retirement of men at age sixty-two to sixty-four, by cohort, Germany 1904 to 1943

1909 cohort. This should only be a transitory phenomenon, as every cohort born later would be given the option of retiring at age sixty- three if they had worked long enough. The above prediction is borne out by the figure 7.11, which shows for each cohort from 1904 to 1943 the hazard rate of retiring at ages sixty- four, sixty- three, and sixty- two. While the retirement hazard of workers at age sixty- four was around 10 percent in the cohorts up to 1908, it jumped to 41 percent for the 1909 cohort and dropped back to less than 24 percent for the 1910 cohort. In contrast, the retirement hazard rate at age sixty- three jumps from 10 percent in the 1909 cohort to 40 percent in the 1910 cohort and even increases to nearly 60 percent for the 1915 cohort. The second component of the 1972 reform was the introduction of oldage pensions for disabled workers aged sixty- two. Thus, cohorts born from 1911 onward had been given one more early retirement option, and we should see a permanent jump in retirement hazard rates at age sixty- two. However, this is not what we see in figure 7.11. Retirement hazard rates at age sixty- two start to increase only from the 1913 cohort onward until the 1917 cohorts, probably due to the delays in the medical assessment process. Afterward, the retirement hazards at age sixty- two decline sharply, which is due to the next reform we discuss in this section. 7.4.2

The 1978 Reform

In 1978, it was decided to reduce the age limit for old- age pensions for disabled workers from sixty- two to sixty years (in two steps in 1979 and 1980). Thus, workers born in 1918 fulfilling all eligibility criteria could draw

296

Axel Börsch-Supan and Hendrik Jürges

Fig. 7.12 to 1943

Retirement of men at age sixty and sixty-one, by cohort, Germany 1904

this pension in 1979 at the age of sixty- one. Workers born in 1919 became first eligible in 1980 also at the age of sixty- one. Finally, workers born 1920 or later could draw this pension from 1980 onward at the age of sixty. Consequently, we expect to see a temporary jump in retirement hazards at age sixty- one for cohorts 1918 and 1919 and a permanent jump in retirement hazards at age sixty from the 1920 cohort onward. Again, this expectation is confirmed by the numbers shown in figure 7.12. Retirement hazards at age sixty- one increase from around 12 percent for the 1917 cohort to around 20 percent for those born 1918 and 1919 and then drop back again to 16 percent (and declining) for workers born 1920 or later. The latter group shows a marked increase in retirement at age sixty (from 13 to 23 percent) 7.4.3

The 1984 Reform

The 1984 reform, which took effect in 1985, was characterized by restrictions in disability pension eligibility. To be eligible for disability benefits, workers had to have a minimum of three contribution years in last five years before they could draw a pension. In effect, this has ruled out claiming DI for many women who did not work on a regular basis. It is more easy to illustrate how the 1984 reform has affected retirement behavior by changing the graphical representation. Figure 7.13 shows the absolute number of workers entering disability insurance by calendar year. It is obvious that the 1984 reform has dramatically changed the number of women retiring on DI pensions. The number has practically halved from 173,000 in 1984 to 86,000 in 1985 (with further decreases in 1986). Among men, the number has dropped from 163,000 in 1984 to 129,000 in 1986.

Disability, Pension Reform, and Early Retirement in Germany

Fig. 7.13

7.5

297

New disability benefit claimants, Germany 1960 to 2005

Early Childhood Environment and Very Early Retirement—An Illustrative Case Study

Although the preceding sections have clearly suggested that health is not the main driver behind long- term trends and major short- term changes in early retirement behavior, in particular on disability pensions, it would be surprising if health played no role at all in explaining early retirement. To illustrate possible causal effect of health on retirement, we draw on the current literature on early childhood conditions and long- term health effects. In particular, we show how adverse early childhood conditions that presumably affect health and work capacity in the long run might be responsible for retirement behavior some fifty years onward. We show this using again the data on cohorts born 1904 to 1943, for which we know at what age people entered retirement. Within these cohorts are those born during World War I (1914 to 1918). In particular, the last two years of war (1917 and 1918) have been characterized by widespread hunger among the civilian population. Our question is, does being born during wartime (as a health- shock) explain some of the time series variation in retirement behavior? Figure 7.14 shows the proportion of each cohort born 1904 to 1943 retiring below age fifty- five. Because no other early retirement schemes exist before age fifty- five, these “very early retirees” are exclusively on disability insurance when they retire. In the cohorts 1904 to 1914, the proportion of very early retirees is fairly stable at about 12 to 13 percent. For cohorts 1915 to 1918, this proportion increases steadily until it reaches 17.2 percent in 1918 (an increase of nearly 36 percent). The first two postwar cohorts show a marked decrease in very early retirement. Beginning at cohorts born

298

Axel Börsch-Supan and Hendrik Jürges

Fig. 7.14

Retirement below age fifty-five, by cohort, Germany 1904 to 1943

1921, another wave of very early retirement sets in. At this point, we can only speculate about the reasons. The picture for women is even clearer. We observe a continuous decline in very early retirement rates for women born between 1904 and the beginning of World War I. Then very early retirement rates rise by approximately 3 percentage points or 29 percent. For the 1920 cohort, retirement rates are back to the prewar level of slighly less than 10 percent before they increase again slowly. However, they never reach the 1918 levels again. The purpose of this short section was to illustrate how short- term fluctuations in disability that are independent of reforms may indeed be explained by health shocks. However, the key is to find the “right” health measure. Mortality rates have not proved to be very successful in explaining early retirement and disability rates. Being born during wartime and times of hunger may have had long- term effects on health and ability to work, and the effects of World War I appear to show in the retirement behavior of German cohorts. Although birth cohort is a very “distant” proxy for health in later adulthood, it has the advantage of being presumably exogenous to disability pension rules. 7.6

Conclusion

This chapter exploited the history of the German pension system with its phases of increasing generosity and subsequent retrenchment in form of several “case studies” and low- level “natural experiments.” These reforms suddenly affected many people (mostly by cohort). Simple before- after comparisons suggest that these reforms had substantial effects on labor

Disability, Pension Reform, and Early Retirement in Germany

299

force participation and retirement. Changes in rules had predictable and sharp temporary and permanent effects on early retirement and DI uptake. Because changes in mortality have generally been very smooth (with the exception of war and immediate postwar cohorts), it is highly unlikely that sharp health changes happening simultaneously are plausible alternative explanations of the effects of institutional change. Specifically, there is hardly any cross- temporal correlation between DI uptake rates (or, more generally, old- age labor force participation) and available measures of health. We also know from other studies, such as Börsch-Supan (2005, 2008), that there is hardly any cross-national correlation between DI uptake rates and measures of health. This result holds notwithstanding that health does explain within country variation at a specific point in time. Especially mental illness (e.g., depression) has emerged as a driver for early retirement via the disability pathway. However, this is largely unrelated to mortality. Hence, in summary, disability insurance appears to be mostly a train on its own track, and it is largely unrelated to mortality or measures of health status over time and in a cross- national comparison.

Appendix Data Sources • Mortality Data: Human Mortality Database (www.mortality.org) • Self- rated health and Health satisfaction: German SOEP, various waves (www.diw.de/soep/en) • Primary diagnosis for disability pensions, retirement age by cohort: Forschungsportal der Deutschen Rentenversicherung (http://forschung .deutsche- rentenversicherung.de/ ) • Labor force participation: German Microcensus (Research Data Centre of the German Federal Statistical Office [http://www.forschungsdaten zentrum.de/en/index.asp])

References Aarts, L. J. M., R. V. Burkhauser, and P. R. de Jong, eds. 1996. Curing the Dutch Disease. An International Perspective on Disability Policy Reform. Aldershot, UK: Avebury. Avendano, M., and J. P. Mackenbach. 2010. “Life-Course Health and Labour Market Exit in 13 European Countries: Results from SHARELIFE.” In The Individual and the Welfare State: Life Histories in Europe, edited by A. Börsch-Supan, M. Brandt, K. Hank, and M. Schröder, 203– 14. Berlin: Heidelberg.

300

Axel Börsch-Supan and Hendrik Jürges

Banks, J., A. Kapteyn, J. P. Smith, and A. van Soest. 2004. “International Comparisons of Work Disability.” RAND Working Paper no. WP- 155. Santa Monica, CA: RAND. Börsch-Supan, A. 1992. “Population Aging, Social Security Design, and Early Retirement.” Zeitschrift für die gesamte Staatswissenschaft 148:533– 57. ———. 2000. “Incentive Effects of Social Security on Labor Force Participation: Evidence in Germany and across Europe.” Journal of Public Economics 78:25– 49. ———. 2005. “Work Disability and Health.” In Health, Ageing, and Retirement in Europe: First Results from the Survey of Health, Ageing and Retirement in Europe, edited by A. Börsch-Supan, A. Brugiavini, H. Jürges, J. Mackenbach, J. Siegrist, and G. Weber, 253– 58. Mannheim, Germany: Mannheim Research Institute for the Economics of Aging. ———. 2008. “Changes in Health Status and Work Disability.” In Health, Ageing, and Retirement in Europe: Starting the Longitudinal Dimension, edited by A. Börsch-Supan, A. Brugiavini, H. Jürges, A. Kapteyn, J. Mackenbach, J. Siegrist, and G. Weber, 228– 36. Mannheim, Germany: Mannheim Research Institute for the Economics of Aging. Börsch-Supan, A., and H. Roth. 2010. “Work Disability and Health over the Life Course.” In The Individual and the Welfare State: Life Histories in Europe, edited by A. Börsch-Supan, M. Brandt, K. Hank, and M. Schröder, 215– 24. Berlin: Heidelberg. Börsch-Supan, A., and C. B. Wilke. 2006. “The German Public Pension System: How It Will Become an NDC System Look-Alike.” In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, edited by Robert Holzmann and Edward Palmer, 573– 610. Washington, DC: World Bank. Commission for Sustainability in Financing the Social Security Systems (Kommission für dieNachhaltigkeit in der Finanzierung der Sozialen Sicherungssysteme). 2003. Final Report (Abschlußbericht). Berlin: Bundesministerium für Gesundheit und Soziale Sicherheit. http://www.bmgs.bund.de/deu/gra/themen/sicherheit/ kommission/index.cfm.

8 Disability and Social Security Reforms The French Case Luc Behaghel, Didier Blanchet, Thierry Debrand, and Muriel Roger

8.1

Introduction

It is well known that France is characterized by low levels of employment in the fifty- five to sixty- four age bracket, compared to other developed countries. Until recently, the specific role of disability benefits in explaining these early exits did not attract a lot of academic attention. One obvious reason for this lack of interest is that this role is quantitatively limited. The French pattern of early transitions out of employment is basically explained by the low age at “normal” retirement and by the importance of transitions through unemployment insurance (UI) and preretirement schemes (PR) before access to normal retirement. The role of these various routes has been repeatedly demonstrated in previous contributions to these International Social Security (ISS) series volumes and again in the last one (Ben Salem et al. 2010). These routes have exempted French workers from massively relying on disability motives for early exits, contrarily to the situation that prevails in some other countries where normal ages are high, unemployment benefits low, and preretirement schemes almost nonexistent. Yet this role of disability remains interesting to examine in this French case, at least for prospective reasons. The current tendency is toward increasing the normal retirement age and toward a more restricted access to unemLuc Behaghel is associate professor at the Paris School of Economics and an INRA researcher at UMR Paris-Jourdan Sciences Économiques. Didier Blanchet belongs to the French National Statistical Institute (INSEE) and is affiliated with the Research Center in Economics and Statistics (CREST). Thierry Debrand is research director at the Institute for Research and Information in Health Economics (IRDES). Muriel Roger is associate professor at the Paris School of Economics. For acknowledgments, sources of research support, and disclosure of the authors’ material financial relationships, if any, please see http://www.nber.org/chapters/c12389.ack.

301

302

Luc Behaghel, Didier Blanchet, Thierry Debrand, and Muriel Roger

ployment benefits or preretirement for senior workers. In such a context, the disability route may gain some renewed importance in the future, acting as a substitute to these other early retirement schemes whose coverage will progressively decline. Investigating whether such substitution effects have been already observed in the past can shed some light on their potential impact for the future. More generally, we observe an increasing interest in the French pension debate for the connection between retirement behavior and health (Struillou 2003; Volkoff and Bardot 2004; Molinié 2006; Blanchet and Debrand 2008). We all know that different categories of workers end up their working lives with very uneven health conditions and very uneven life expectancies, sometimes directly linked to past working conditions. How should pension reforms take this kind of inequality into account, especially when poor health directly results from work conditions? This question remains currently unanswered. The pension reform that took place in 2003 was expected to be followed by a negotiation between social partners on this specific issue, but this negotiation has been unfortunately unable to provide any form of tangible result, and this issue is again central in debates generated by the new reform in 2010. At the time this chapter is written, the main components of this reform project are a progressive shift of the minimum retirement age from sixty to sixty- two, and a parallel shift from sixty- five to sixty- seven for the age at which a full rate pension can be obtained whatever the length of one’s career. Consequences of such a shift for workers with bad health, low life expectancies, or who have been exposed to hard working conditions have been one of the main objections raised by opponents to the reform. The proposed answer to this concern has consisted in maintaining the threshold of sixty for people with a level of impairment of at least 10 percent, and one wonders in which proportions this possibility will be used as a new instrument of early exits. It is with these considerations in mind that we shall examine the French experience of interaction between pension reforms and take-up of disability benefits and, more generally, the importance of health considerations in the design of pension policies. We shall first concentrate on substitution effects. Even if the disability route has always played a limited role in France, its importance is not negligible (Barnay and Jeger 2006), and it has fluctuated over the past decades. These fluctuations can be used to test the substitutability hypothesis: disability loses importance when other routes become more widely accessible and, conversely, when these routes are reformed in a more restrictive way. After a brief presentation of the trend of the French old workers participation to the labor market in section 8.2, section 8.3 will first recall the main historical changes that occurred in the pension system, in early retirement schemes, and in disability benefits and focus more specifically on interactions that took place until the early eighties. We shall more specifically focus

Disability and Social Security Reforms: The French Case

303

on two major historical changes going in the direction of more generous pensions and a lower take-up of disability benefits: the first one took place during the 1970s (the Boulin law) and the second one in 1983, when the normal retirement age was lowered to sixty. The consequences of these changes have been large enough to allow observation of substitution effects at the macro level. Section 8.4 will go on along the same vein focusing on two post- 1990 changes going in the opposite direction. The first one is the 1993 pension reform that restricted the access to full pension benefits at sixty. Up to now, this reform did not affect a very large proportion of the population, and this implies that substitution effects are difficult to observe on macro- series. But such effects exist and can be identified on micro data. We shall present results from this quasi experiment following Bozio (2006). The second episode relates to sickness leaves covered by health insurance. The first half of the 2000s witnessed an increase in these sickness leaves that has been sometimes interpreted as another manifestation of the substitution effects between early retirement schemes and the disability route to retirement. Strengthened control over these sickness leaves after 2003 dramatically reduced their role for older workers. In short, the message of sections 8.3 and 8.4 is that, whatever the episode considered, institutional changes or changes in the intensity of controls are sufficient to explain changes in the take-up of disability benefits. Given this evidence, it might appear hardly necessary to check that these changes are uncorrelated with global health trends. We shall, nevertheless, examine this question in section 8.5. One good reason for exploring this question is that an absence of correlation at the macro level can still go along with a significant interaction at a more micro level. Moreover, the fact that the take-up of disability benefits essentially reflects institutional changes does not mean that those who claim these benefits do not actually face real health problems. At the macro level, section 8.5 confirms the lack of correlation between health and labor market status: the past decades have been a period of uninterrupted improvement in average health. But micro data can help build a more detailed story that reconciles this macro view with the increasing concern for health considerations in the pension debate. An improvement of average health can go along with the persistence of a significant health dimension of pension problems if the dispersion of health status is large and a fortiori if this dispersion is increasing. Micro data suggest that such a phenomenon may be at work. If this trend is confirmed, the question that we have raised in the preceding retains all its importance: if health is an obstacle to remaining on the labor market for an important share of the population, how should pension policy address the case of these people? The general improvement of health and the lack of historical connection between health and the take up of disability benefits do not mean that health can be neglected in the design of retirement policies.

304

Luc Behaghel, Didier Blanchet, Thierry Debrand, and Muriel Roger

8.2

General Labor Force Participation Trends and Pension Reforms

To have a clear view of the role of disability benefits in the general pattern of retirement routes of older workers in France, we first give a brief recall on the general labor force participation (LFP) trends in the past decades. Moreover, we underline the link between LFP and the pension reforms implemented during the same period. Data on labor force participation and pathways to retirement are provided by labor force surveys (LFS) conducted by the French National Statistical Institute (INSEE) since 1950. We use the 1968 to 2005 waves of this LFS. Trends in labor force participation for men and women are given in figures 8.1 and 8.2. Participation rates of senior workers aged sixty- five and over are quite low. For men aged sixty to sixty- four, we note a large decrease over the period. We observe the same trend for women but in smaller proportions due to the counteracting influence of increased attachment to the labor force for successive cohorts. The difference between men and women is still more pronounced for the fifty- five to fifty- nine age group. For men, after a period of relative stability at the beginning of the period, we observe a drop just after 1982 and then a new period of stability. For women, the arrival in the fifty- five to fifty- nine group of women increasingly attached to the labor market fully dominates the trend toward earlier exits. Previous work from Ben Salem et al. (2010) has underlined the correlation between changes in senior labor force participation and the key dates of

Fig. 8.1

Labor force participation for men, 1968 to 2005

Source: INSEE.

Disability and Social Security Reforms: The French Case

Fig. 8.2

305

Labor force participation for women, 1968 to 2005

Source: INSEE.

retirement reforms or labor market policies toward senior workers in France. Until 1983, the decline in LFP for the sixty to sixty- four age group has been due to the opening of retirement at sixty for some specific subgroups of the population and to the development of early retirement schemes essentially targeted to this age bracket. For this age group, the small accident that is observed in 1980 is a pure statistical artifact due to the transition through the age group of the small cohorts born between 1915 and 1919. Because LFP rates are higher at sixty than they are at sixty- four, the arrival of small cohorts first depresses the average LFP rate for the whole five- year group and, conversely, when the bulk of these small cohorts are in the second half of the age group. If we abstract from this accident, we see that the trend has been almost continuous. The impact of the 1983 reform has been essentially to replace early retired people by normal retirees. Since then, the 1993 and 2003 reform have introduced changes that, in the long run, should reincrease LFP rates for the sixty to sixty- four group, but consequences of these reforms are still negligible. The reason is that these reforms did not suppress the possibility to get a pension at sixty and relied instead on another parameter, which is the number of years of contributions requested for this pension to be a full rate one. This parameter had been fixed at 37.5 years of contributions in 1983. The 1993 reform made it increase gradually from 37.5 years to 40 years between cohorts 1933 and 1943, but only for private- sector employees. The 2003 reform extended this forty- year condition to public- sector employees (progressively between 2003 and 2008) and now imposes a further strengthening for all workers, first from forty to

306

Luc Behaghel, Didier Blanchet, Thierry Debrand, and Muriel Roger

forty- one years between 2008 and 2012 and then in line with changes in life expectancy, all that combined with changes in the incentives to retire either before or after this full rate. If these reforms did not result until now in large changes in participation rates, it is because these new conditions, until now, have affected only a limited proportion of the population: cohorts currently retiring generally started working before twenty and are still frequently able to validate forty or forty- one years of contribution at sixty. There is also the fact that changes in incentives to retire before the full rate have changed in a way that makes it less penalizing, at least for private- sector employees. There is also the fact that, since 2003, some of these employees—those having started working very early—have benefited from the possibility to retire as soon as they get their forty years of contributions, without needing to wait until sixty. This aspect of the reform has been introduced to make the reform more acceptable to the public opinion, and it has been precisely targeted to low- skilled workers with long careers who are more likely to reach the fifty- five to fifty- nine brackets in bad health. The consequences of this latter disposition are visible at the very end of the lines for the fifty- five to fifty- nine age group on figures 8.1 and 8.2, with significant drops both for men and women although, here again, some perturbations could be attributed to the transition, through the age group, of cohorts of very uneven size born during and just after World War II and also to a break in the series in 2003 due to the shift from an annual to a continuous survey.1 Then, going backward in the past with this fifty- five to fifty- nine year age group and focusing on men for whom the impact of retirement and early retirement policies is not blurred by other trends, we see that the major event for this age group is the drop during the first half of the 1980s. This drop can be understood as a temporary consequence of lowering the normal retirement age to sixty. This lowering mechanically resulted in the disappearance of early retirement in the sixty to sixty- four bracket but also led to the development of new flows of early retirees in the fifty- five to fifty- nine group. However, this time, public authorities decided to put these new flows under stricter control, and the drop in the LFP rate stopped in 1985. Since then, with the exception of the post- 2003 episode, policies have been able to just stabilize the LFP rate. 8.3

The Role of Disability Benefits: Long-Term Trends

What has been the relative importance of the disability route over these decades? 1. From 1968 to 2002, the households included in the Labor Force Survey sample were interviewed in March of three consecutive years with one- third of the households replaced each year. Since 2003, the households included in the French LFS are interviewed six consecutive quarters with one- sixth of the households replaced each quarter.

Disability and Social Security Reforms: The French Case

307

Figure 8.3 uses labor force surveys and some other data sources to give an evaluation of the relative importance of this route compared to other possible pathways out of the labor force. The proxy that is used is the status of the individual the year before his or her sixtieth birthday. Data go back to 1983, that is, they only cover the period where the normal retirement age is sixty. Moreover, data on sickness leave by age are only available after 1997; before that date, workers in sickness leave are recorded as employed. All over this period, we see a decrease in the share of people still in employment just before their sixtieth birthday and an increase in the shares of people in early retirement or benefiting from unemployment insurance benefits. Yet, all over the period, the number of people going through disability insurance or sickness leave is not negligible, amounting to between 5 percent and 8 percent of the population. We will now focus on this route, using additional statistics specific to this route and covering a longer time period. First of all, some explanations are required concerning the organization of this disability route. Even if disability is not a predominant route out of the labor force in the French system, the idea that disability is a basic underlying motive for access to retirement has been historically important in the

Fig. 8.3

Pathways to retirement, men and women

Sources: Enquête Emploi; Unédic; DARES; CNAM; and CNAV. Note: Pathways are proxied by the situation at age fifty- nine (see Enquête Emploi) corrected with administrative data on unemployment (see Unédic), early retirement (see DARES) and complemented with administrative data on sickness leave (see CNAM) and inflows from “pension d’invalidité” to “pensions d’inaptitude” (see CNAV). The break in the series indicates that data on sickness leave are missing before 1997; before that date, workers in sickness leave are recorded as employed.

308

Luc Behaghel, Didier Blanchet, Thierry Debrand, and Muriel Roger

foundation of French social security. This goes back to consequences of the Great Depression. It had resulted in very high unemployment rates for workers above fifty, and this has led, in 1941, under the Vichy government, to the creation of the AVTS (Allocation aux Vieux Travailleurs Salariés) providing early retirement for workers above sixty excluded from the labor force either for health or economic reasons. The employability issue was also at the heart of the creation of the current system of Social Security in 1945. At that time, the risque vieillesse (oldage risk) was clearly characterized as the risk of living beyond an age at which the individual becomes unable to maintain his or her standard of living through labor force participation, be it for health or other reasons. The designers of the system had initially considered that the age of sixty- five was an acceptable proxy for this age but already allowed an earlier access, as soon as sixty, either for people already benefiting from disability insurance before this age (retraite pour ex-invalides) or for people declared inapt for work at this age even if they were not previously benefiting from disability benefits (retraite pour inaptitude). This second category remained, however, highly selective: it required a disability rate of 100 percent, it was limited to people having worked for at least thirty years and provided a benefit being, at the maximum, equal to 40 percent of the average of past wages.2 This was higher than the rate of replacement for people claiming early retirement at sixty without this disability motive but remained little attractive. Until the beginning of the 1970s, this system was, therefore, little used, mostly by men, with women more frequently resorting to the retraite pour ex-invalides. A more dynamic phase for the pension d’inaptitude took place during the 1970s. From a general point of view, this period has been a period of increasing generosity for the pension system, in particular with the aim of reducing the prevalence of poverty among elderly people (it was one of the aspects of the “new society” program of the post- 1968 Chaban-Delmas government). This period was also marked by the strong union pressure in favor of lowering the normal retirement age to sixty for the entire population that ultimately led to retirement at sixty in 1983. During this period, that demand remained unsatisfied, but the Boulin reform in 1971 opened several possibilities for earlier exits for various categories of population, including a move toward less selective and more generous rules for the pension d’inaptitude: the threshold for the rate of disability was lowered to 50 percent, the condition of having worked thirty years or more was suppressed, and benefits were increased to 50 percent of the average wage, that is, in line with a normal full rate pension. This mechanically led to an increased importance of this route, but, as shown on figure 8.4, the incidence of this change was more pronounced for 2. The disability rate measures the intensity of limitations encountered by the disabled person.

Disability and Social Security Reforms: The French Case

Fig. 8.4

309

“Pensions d’inaptitude”: Total flows

Source: CNAV in Omnès (2006).

women, because men, at the same period, started benefiting from the development of the early retirement route, financially more attractive garantie de ressources and allocation spécifique du fonds national pour l’emploi, especially at the end of the 1970s. On the whole, the share of pensions d’inaptitude in the yearly flow of new pensioners increased over this period, albeit with oscillations, and finally peaked at 30 percent of total exits in 1983 (see figure 8.5). It is just after that peak, that access to a full pension at age sixty became possible for the large majority of the population, the only condition being to have contributed to the pension system for at least 37.5 years. In this context, claiming for a pension d’inaptitude became useless for a large share of people. We note in figure 8.6 that the average pension amount for normal retirement became at least twice higher than disability benefits. Those who still had a reason to rely on this route were people reaching the age of sixty in bad health and with incomplete careers. Their share in the total flow of new pensioners is now comprised between 10 and 15 percent. Did subsequent pension reforms alter this picture? As mentioned, the 1993 reform gradually increased the number of years of contribution required for a full pension. The 2003 reform went further in the same direction. As already shown, these two reforms have had only small upward effects on retirement ages until now. It is, therefore, not surprising that figure 8.4 does not exhibit significant moves in flows of disabled people. To analyze whether these reforms have started reorienting part of the flows toward the disability route, we must look at a micro level and analyze whether disability take-up

310

Fig. 8.5

Luc Behaghel, Didier Blanchet, Thierry Debrand, and Muriel Roger

“Pensions d’inaptitude”: Percentage of the flow of new retirees

Source: CNAV.

Fig. 8.6 Average pension amount: Normal pensions and “pensions d’inaptitude,” 1963 to 2003 (in euros) Source: CNAV in Omnès (2006).

has significantly increased among these people who have been hurt by the pension reform. Figure 8.3 also suggests looking at another related route in more details because there seems to be a move in the number of people benefiting from sickness leaves around the 2003 reform. This is the subject of the next section.

Disability and Social Security Reforms: The French Case

8.4 8.4.1

311

Two Quasi Experiments The 1993 Pension Reform: Disability Pensions as a Substitute to Normal Pensions?

The 1993 reform is the first reform in France that aimed at delaying retirement. It reduced the average full- rate pension amount (by introducing a longer period to compute the reference wage) and gradually increased the length of contribution required to get the full replacement rate. Access to disability pensions ( pensions d’inaptitude) remained unchanged. As a consequence, the reform increased the relative attractiveness of disability pensions for workers who were no longer eligible for a full- rate pension at age sixty in the normal pension scheme (due to an insufficient contribution length) but were potentially eligible for a disability pension. These workers basically had three options: (a) keep claiming a normal pension at age sixty at the cost of a high penalty;3 (b) delay claiming until reaching the required number of quarters for a normal pension; (c) apply for a disability pension at age sixty. Our goal is to assess the role played by the third option. The impact of the 1993 reform is, unfortunately, hardly detectable from aggregate data (such as in figure figures 8.3 to 8.5), as the reform only affected about 8 percent of the workers (the wage earners of the private sector with 131 to 160 quarters of contribution at age sixty retiring through a normal pension). However, Bozio (2006) shows that it can be detected from micro data, applying a difference- in-difference approach to an administrative, exhaustive data set. In what follows, we summarize and discuss his results. The identification of the reform rests on the comparison of workers with the same completed contribution length at age sixty over different birth cohorts. With a given contribution length (e.g., 151 quarters), older cohorts (e.g., born in 1934) are eligible for a full- rate pension at the first retirement age (sixty), whereas younger cohorts (e.g., born in 1935) need additional contribution quarters. Comparing the retirement behavior of these two groups, therefore, identifies the impact of increased contribution requirements. The identification is made robust to possible cohort trends using workers unaffected by the reform given their completed contribution lengths as a control (e.g., workers having contributed 152 quarters at age sixty). For a representative worker aged sixty who has contributed the required number of quarters for a full- rate pension, increasing the contribution requirement by a quarter (a) increases the average retirement age through the normal pension scheme by about two months and (b) increases the probability of receiving a disability pension by 13 percentage points (Bozio 2006, 3. The replacement ratio was reduced by 10 percentage point per year of missing contribution.

312

Luc Behaghel, Didier Blanchet, Thierry Debrand, and Muriel Roger

tables 2.4 and 2.7). In other words, the three types of responses are observed; some workers keep retiring at sixty through the normal pension scheme; some delay claiming and retirement;4 last, a significant share of workers switch to disability pensions. Interestingly, the effects of the 1993 reform are smaller for workers missing more than a quarter of contribution at age sixty. Even before the reform, these workers did not have access to normal pensions at a full rate at age sixty; they were, therefore, already induced to apply for a disability pension, and the impact of the 1993 reform may be small. On average, Bozio (2006) finds that a one- quarter increase in required contribution increases the probability of receiving a disability pension by 2 percentage points. Overall, even though they are not visible on aggregate data, these substitution effects are sizeable. The 1993 natural experiment is, therefore, useful from a prospective viewpoint. For instance, if the same substitution effects persists for cohorts born in the 1960s and 1970s for which the 1993 reform should be binding for one worker out of two (due to the increase in age at labor market entry), disability pensions may become a significant route to retirement. However, the interpretation of these effects remains an open question: one cannot conclude from the results whether the reform induced a pure disclosure effect—workers with health problems who used to rely on the normal pension scheme come to use disability schemes that are aimed at them—or whether it generated a moral hazard problem—workers that are not targeted by disability pensions use them as a way to escape the new contribution requirements. 8.4.2

The Impact of Tighter Control on Long-Term Sickness Leaves for Older Workers

The sharp increase in sickness benefits that occurred at the beginning of the 2000s in France has been a matter of debate (Lê and Reynaud 2007; Kusnik-Joinville et al. 2006). Between 1997 and 2003, aggregate sickness benefits rose from an index of 100 to 140 before falling back to 125 by 2005. Changes in the population size and structure as well as the decline of unemployment can explain about half of the rise (the rest remaining unexplained); as for the drop after 2003, it has been widely attributed to the tighter controls that were implemented (Kusnik-Joinville et al. 2006). Indeed, the number of controls for short- term sickness leaves increased from 34,000 in 2003 to 250,000 in 2005; and since 2005, long- term leaves (above sixty days) undergo systematic control. None of the previous studies has explored whether this rise and fall of sickness benefits in France has varied by age. This is the question we consider here, using original administrative data from the French public health insurance administration (Caisse Nationale d’Assurance Maladie [CNAM]). Indeed, older workers in the late 1990s and early 2000s had specific incen4. Until recently, workers should retire when they claim for pension benefits.

Disability and Social Security Reforms: The French Case

313

tives to rely on sickness benefits, as some of the early retirement schemes were being phased down quite rapidly (the number of recipients of an early retirement scheme, at age fifty- five to fifty- nine, dropped from 230,000 in 1997 to 130,000 in 2005). Unemployment (without search requirement after age fifty- six or fifty- seven) largely acted as a substitute (the number of recipients increasing from 270,000 to 400,000 over the same period), but long- term sickness leaves may also have played a role. Moreover, the sudden restriction to sickness benefits in 2004 to 2005 through tighter control may have had a stronger impact on older workers if they were using sickness benefits as one route to retirement among many, as opposed to younger workers who could only rely on sickness benefits in case of health problems. Figure 8.7 provides evidence in favor of these two hypotheses. It shows the evolution of sickness benefits per capita (in 1998 euros), for different age groups among men (i.e., the total amounts of sickness benefits received by a given age group over its size in the full population in order to correct for demographic shifts). All age groups display an upward trend over the 1997 to 2008 period, with years 2000 to 2003 standing out above that trend. However, the magnitude of the 2000 to 2003 “bump” is much larger for workers aged fifty- five to sixty- four. In particular, the decrease in benefits that occurs in the overall population after 2003 seems to be fully driven by the older age groups: other groups only display a slowdown in growth after 2004 to 2005. Figure 8.8 shows quite similar patterns for women, though the bump is slightly less apparent. In order to control for fixed differences across birth cohorts and for age effects, we conduct a difference- in-difference analysis. This analysis uses the

Fig. 8.7

Sickness benefits for men by age group, 1997 to 2008

Note: Benefits per capita are computed as the total spending on sickness benefits for a given age group, divided by the population size of that group.

314

Fig. 8.8

Luc Behaghel, Didier Blanchet, Thierry Debrand, and Muriel Roger

Sickness benefits for women by age group, 1997 to 2008

Note: Benefits per capita are computed as the total spending on sickness benefits for a given age group, divided by the population size of that group.

same administrative data as in the previous graphs, that is, data grouped by five- year age groups. This data limitation implies that we observe a given five- year birth cohort only every five years. For instance, we observe the 1944 to 1948 birth cohorts in 1998, when they are fifty to fifty- four years old, in 2003, when they are fifty- five to fifty- nine, and in 2008, when they are sixty to sixty- four. As “controls,” we use the adjacent cohorts, that is, the 1949 to 1953 birth cohorts (observed in 2003 and 2008 when they are fifty to fifty- four and fifty- five to fifty- nine, respectively), and the 1939 to 1943 birth cohorts (observed in 1998 and 2003 when they are fifty- five to fifty- nine and sixty to sixty- four, respectively). We start by computing the impact of an easy access to sickness benefits for workers aged fifty- five to fifty- nine. To do so, we compare the evolution of benefits for the 1944 to 1948 cohorts (that reach age fifty- five to fifty- nine in 2003 at a time of loose controls) to the following cohorts (who reach age fifty- five to fifty- nine after the restrictions in 2008). That is, we contrast two cohorts that reach age fifty- five to fifty- nine (a decisive age in terms of retirement pathways) at a time of loose controls versus tightened controls on sickness benefits. The benefits per capita are displayed in panel A of table 8.1 (left part). For the 1944 to 1948 cohorts, benefits per capita increased from 288 to 470 euros between 1998 and 2003. As 2003 was a year when benefits were easily accessible for all ages, part of this rise may indicate the impact of loose controls over benefits for workers aged fifty- five to fifty- nine. Of course, part of it may also be due to the mere effect of age.

182

1st difference

116

1st difference

3rd difference

189 305

40–44 45–49

1954–1958 cohort

288 470

1944–1948 cohorts

91

267 357

1959–1964 cohort

27

388 416

1949–1954 cohorts

–384

45–49 50–54

127

305 432

1954–1958 cohort

C. 55–64 versus 45–54

2nd difference 25

129

1st difference

470 86

B. Younger age group: 45–54

2nd difference 155

55–59 60–64

1944–1948 cohorts

A. Age group of interest: 55–64

Sickness benefits by cohort and age group: Men (euros per capita)

50–54 55–59

Table 8.1

166

222 388

1949–1954 cohorts

–165

323 158

1939–1944 cohorts

–180

2nd difference –39

2nd difference –219

316

Luc Behaghel, Didier Blanchet, Thierry Debrand, and Muriel Roger

We control for this age effect using the 1949 to 1954 cohorts, observed in 2003 and 2008. At the age of fifty- five to fifty- nine, these cohorts witnessed tighter controls. Correspondingly, the increase of benefits is much lower; the difference in differences is 155 euros. In other words, the difference in differences shows that when workers are in their fifties, the reliance on sickness benefits increases much faster with age when controls are loose (for the 1944 to 1948 cohorts) than when they are getting tightened (for the 1949 to 1953 cohorts). Consider now what happens between the fifty- five to fifty- nine and the sixty to sixty- four age ranges when controls get tighter. To do so, follow the 1944 to 1948 cohort again in 2003 and 2008 (a period of control tightening) and compare it to the 1939 to 1943 cohort in 1998 and 2003 (a period of loose controls). The results are given on the right part of table 8.l, panel A. For the 1944 to 1948 cohorts, benefits per capita strongly decreased between 2003 and 2008. Part of this is obviously due to an age effect: a large share of workers had retired by 2008 and was no longer eligible for sickness benefits. However, the comparison with individuals born between 1939 and 1943 shows that the drop is specifically large for the 1944 to 1948 cohorts: the difference in differences is – 219 euros.5 Overall, this shows that sickness benefits display a specific age pattern for the 1944 to 1948 cohorts that distinguishes them from the two adjacent cohorts: a very large increase between the ages of fifty to fifty- four and fifty- five to fifty- nine, followed by a dramatic fall when reaching the ages of sixty to sixty- four. As these cohorts reached the age of fifty- five to fiftynine just before a strong tightening of sickness benefit controls, a plausible explanation for this age pattern is the sensitivity of workers in the “crucial age” of fifty- five to fifty- nine to the rules governing the access to different pathways to retirement. One may, of course, wonder whether similar patterns have occurred for younger cohorts on the same period. In panel B of table 8.1, we analyze the 1954 to 1958 cohorts using the two adjacent cohorts to control for age effects (in the same way as for panel A). We do find a rise in benefits between 1998 and 2003 (+25 euros, adjusting for age) and a fall between 2003 and 2008 (– 39 euros, net of age effects). However, the magnitude of these movements is much lower than for the 1944 to 1948 cohorts. This is confirmed by the third difference in panel C. Table 8.2 provides quite similar evidence for women, although the magnitude of the effects is somewhat lower. Overall, this analysis confirm that 5. Note that the 1939 to 1943 cohorts are better described as a “control” group than the 1949 to 1953 cohorts, as the former remained under a regime of loose controls, whereas the latter were impacted by the policy change: they reached the age of fifty- five to fifty- nine under a regime of tight controls. In that sense, the first difference- in-differences combines the effects of two opposite “treatments”: increased access for the 1944 to 1948 cohorts (loose controls at age fifty- five to fifty- nine) contrasted to tightening of controls for the 1949 to 1953 cohorts.

105

1st difference

3rd difference

126 231 87

200 288

1959–1964 cohort

1954–1958 cohort

40–44 45–49

25

98

1st difference

279 304

1949–1954 cohorts

177 275

1944–1948 cohorts

45–49 50–54

–217

275 58

103

231 334

1954–1958 cohort

C. 55–64 versus 45–54

2nd difference 18

55

1st difference

B. Younger age group: 45–54

2nd difference 73

55–59 60–64

1944–1948 cohorts

A. Age group of interest: 55–64

Sickness benefits by cohort and age group: Women (euros per capita)

50–54 55–59

Table 8.2

130

149 279

1949–1954 cohorts

–67

156 89

1939–1944 cohorts

–123

2nd difference –27

2nd difference –150

318

Luc Behaghel, Didier Blanchet, Thierry Debrand, and Muriel Roger

workers aged above fifty- five have “overreacted” to the rise and fall of sickness benefits over 2003. As this occurred in a context of restrictions of access to early retirement schemes, this evidence can be interpreted as a sign of the specific sensitivity of older workers to the rules of access to sickness benefits. To conclude this section, we find robust evidence that sickness benefits and disability pensions can act as substitutes for other retirement routes, suggesting that older workers are quite sensitive to the financial incentives to use this route rather than other ones. Combined with the historical evidence on the important role played by disability routes before 1980, this suggests that today’s low use of the disability route is mostly due to its relative unattractiveness. 8.5

Increasing Health Inequalities among Seniors

Previous sections have shown that disability can be used as a substitute for other retirement routes. This suggests that its apparent prevalence is not necessarily connected with changes in actual health status, and this also raises doubts about the role of adverse health conditions as a major explanation for the development of early retirement: the expansion of early retirement would have resulted from other factors, and the fact that it has sometimes taken the form of increased disability rates would only reflect country- or period- specific institutional arrangements. But this may be only part of the story. A complete diagnosis of the link or the nonlink between early exits and health status requires a more direct examination. This last section will provide some elements on this issue both at the micro and the macro level, relying on various measures of health. Messages will require more or less qualification according to the perspective that is retained. 8.5.1

Mortality Rates and Life Expectancy

We first study the link between retirement and health trends proxying health by mortality. Mortality is an imperfect indicator of health or disability; however, it remains an important indicator of the change in health status of a population. Figures 8.9 and 8.10 provide mortality rates at different ages for men and women. The mortality rate at seventy has been divided by more than two between 1962 and 2000, but each age is concerned by the decrease. More than the mortality rate, the ages of equal mortality probability given in figure 8.11 provide an idea of the increase in the average length of life over the past decades. What is computed at each period is the age at which an individual reaches the same mortality rate as an individual aged, respectively, fifty- five, sixty, or sixty- five. On the average, for both men and women, these “equal mortality ages” have shifted by about ten years between 1960 and 2008. For instance, you need to be a little more than seventy today to face the instant mortality risk to which an individual aged sixty was exposed in 1960.

Disability and Social Security Reforms: The French Case

Fig. 8.9

319

Mortality rates at different ages, men

Source: INSEE.

Fig. 8.10

Mortality rates at different ages, women

Source: INSEE.

This is more or less consistent with the global shift in life expectancy at birth that has gained 10.6 years for men and 10.7 years for women over the same period (respectively, from 67.0 to 77.6 years, and from 73.6 to 84.3 years). Unsurprisingly, the comparison of labor force participation with such mortality data on figure 8.12 gives no indication of a link between health and labor force participation of older workers. If health is proxied by mortality, this approach sends a clear message of a full disconnection between health and retirement trends.

320

Luc Behaghel, Didier Blanchet, Thierry Debrand, and Muriel Roger

Fig. 8.11

Ages of equivalent mortality

Sources: INSEE and authors’ computations.

Fig. 8.12

Ages of equivalent mortality and labor force participation rates, men

Sources: INSEE and authors’ computations.

However, aggregate mortality rates suffer from limitations when assessing health trends. Although life expectancy in France is increasing, it is not clear whether these extra years are spent in good health. For Jagger et al. (2008), at fifty years of age, life expectancy in France is 29.6 years for a man and 35.4 years for a woman, whereas healthy life expectancy is 18.0 years for a man and 19.7 years for a woman.6 Moreover, the aggregate trend may hide changes in the dispersion of health. 6. These authors calculate the differences between life expectancy and healthy life years in all the countries using mortality data in the form of life tables and age specific prevalence of activity limitation. The results showed that in twenty- five countries, a fifty- year- old man can expect, on average, to live until 67.3 years old without activity limitation and a woman until 68.1 years old, whereas a man’s life expectancy is 78.9 years and a woman’s is 83.5 years. The

Disability and Social Security Reforms: The French Case

321

Several alternative health indicators can be used to study these changes in health status both at the macro and micro levels over the last decades. We shall concentrate here on two of them: self- reported health and the body mass index (BMI), both available since 1992 from the Enquête Santé et Protection Sociale (ESPS), which is a biennal survey on a panel of beneficiaries from health insurance, performed by IRDES (Institut de Recherches et de Documentation en Economie de la Santé). 8.5.2

Self-Reported Health Status

Figures 8.13 and 8.14 show the average self- reported health level on a scale of 10, for various age groups, for men and women. They do not show any clear trend over the last twenty- years but point at a slight deterioration over recent years (2002 to 2008). At the micro level, it can also be checked that subjective health indicators are clearly lower for disabled. Looking by employment status shows also an increasing gap between unemployed and employed older workers. For men, we note a monotonic decline in subjective health status of unemployed over the period, health status of other categories being almost constant. Results are less clear for women even if there is a decrease at the end of the period for the unemployed ones. Globally, messages delivered by these subjective health indicators do not run completely opposite to those obtained with mortality data. We do not find a massive global deterioration of health status that could account for observed declines in employment rates. Global trends rather suggest that health factors have been globally neutral over the period. But micro data suggest that that health factors may have eventually fed pressure for maintaining early retirement over the very recent years. Even if there is no strong correlation between health status and retirement along the time dimension, this correlation does exist at the micro level. Bad health is correlated to nonemployment and is, therefore, a reasonable candidate for explaining part of early exits. This micro link can come under various forms. It is straightforward when exit takes place through the disability route. It will be more indirect but still easy to figure out when exit takes place through unemployment. On the demand side, workers in bad health are more likely to suffer from belowaverage productivity and are more likely to be laid off by firms confronted to excess capacity or in search of productivity gains. The probability of being hired, already low for all senior job seekers, will also be still lower for those in bad health. On the supply side, these workers in bad health are more likely to find attractive and accept financial conditions offered in case of individual differences or inequalities between countries are considerable both in terms of life expectancy (9.1 years for men and 6.1 years for women) and healthy life expectancy (14.5 years for men and 13.7 years for women). On the basis of a cross- country econometric analysis, the authors demonstrate that disability- free life expectancy is related to a country’s gross domestic product (GDP) and health expenditures in the care for the elderly of both sexes. For men, it is positively correlated to the long- term unemployment rate and negatively correlated to lifelong learning.

322

Luc Behaghel, Didier Blanchet, Thierry Debrand, and Muriel Roger

Fig. 8.13

Self-reported health status

or collective layoffs and, when out of the labor force, will be less motivated for trying to return to employment. We must recall, however, that quantification of these links may be affected by various biases or measurement errors. For instance, the causal impact of health on employment can be either moderated or amplified by symmetrical effects of labour force status on health: being in employment can be the cause or an aggravating factor of some health problems (Strauss and Thomas 1998), especially for people working under conditions of physical or psychological pressure, but unemployment can also be a negative factor for health. Depending upon which effect dominates, the observed relationship between health and employment will underestimate or overestimate the true causal impact of health. Biases can also come from declarative behaviour. An increasing sensitivity to a health problem can lead to a stationary or deteriorating index of self- reported health, even when objective health improves. The correlation observed at the micro level between nonemployment and bad health can also result from self- justification bias, that is, people preferring to attribute

Disability and Social Security Reforms: The French Case

Fig. 8.14

323

Self-reported health status by labor market status

their early exit from the labor force to health problems rather than to any other possible cause. 8.5.3

The Body Mass Index

This latter category of biases due to self- declaration can be avoided by relying on more objective measures. The one we favor here is the body mass index, widely accepted in the medical and public health literature as an index of bad health because it represents a risk factor of health disorders and premature mortality. Body mass indexes by age and gender are given on table 8.3. On average, the BMI increases consistently during the period 1992 to 2008. For men aged between fifty and sixty- five years old, it increases from 25.9 in 1992, to 26.3 in 2000 and to 26.4 in 2008. For women aged between fifty and sixty- five years old, it was lower at the beginning of the period (24.7

324 Table 8.3

Luc Behaghel, Didier Blanchet, Thierry Debrand, and Muriel Roger Body mass index (BMI) by period, gender, and age group

Mean

Standard deviation

% with BMI > 25

% with BMI > 30

50–65 age group Men 1992 2000 2008 Women 1992 2000 2008

26.0 26.2 26.4

3.5 3.5 3.9

57.0 56.3 61.5

11.8 12.3 16.1

24.7 25.0 25.2

4.1 4.4 4.7

40.1 42.7 43.3

10.8 13.7 15.4

All ages Men 1992 2000 2008 Women 1992 2000 2008

22.7 23.0 23.5

4.5 4.8 5.0

29.4 32.9 37.1

5.0 6.5 9.1

21.9 22.2 22.9

4.5 4.8 5.3

20.8 23.4 29.3

5.3 6.9 9.9

Sources: ESPS surveys and authors’ calculations.

against 25.9 for the men), but it increases also from 24.7 in 1992, to 25.0 in 2000, and to 25.2 in 2008. This increase does not seem to be that large: averages still do not contradict the view that health factors have been relatively neutral over the period. But other distributional characteristics deliver a less favorable message, especially at the top of the distribution. If we define a BMI > 30 as obesity, its prevalence between 50 and 65 has increased by about 30 to 40 percent between 1992 and 2008, from 11.8 to 16.1 for men and from 10.8 to 15.4 for women. The phenomenon is still more striking for younger ages, with a prevalence that has been almost multiplied by two for the entire population. This suggests a generation effect that may accelerate health deterioration in the fifty to sixty- five group over the decades to come. This provides one good reason to believe that some pressure will persist or even increase in favor of specific measures in favor of people reaching retirement ages under adverse health conditions, and it points to the necessity of looking simultaneously at means and at dispersions if we want to measure properly the magnitude of this problem. Finally, correlations between BMI levels and employment status exhibit a less clear pattern than for subjective health. However, it shows that the BMI, for both men and women, is almost always higher for disabled individuals and lower for individuals in employment. Except for men at the end of the period, the BMI indicator is higher for unemployed than for employed older workers. This confirms the results we had with the subjective health

Disability and Social Security Reforms: The French Case

Fig. 8.15

325

Body mass index indicator by labor market status

indicator. Unemployed people seem to be in a worth health than employed people. As previously, unemployment may act as a substitution to disability routes but acting on people with health difficulties, and the question is to know how pension policy must address the case of these people. Again, the question is all the more important that BMI trends among younger cohorts suggest that the problem may gain further importance over the decades to come. (See figure 8.15.) 8.6

Conclusion

Even if disability doesn’t seem to be a key component in the retirement history of old workers in France, discussions on disability as a retirement route remain essential in a context of a general decrease in the generosity of the pension scheme.

326

Luc Behaghel, Didier Blanchet, Thierry Debrand, and Muriel Roger

The study of the past reforms of the pension system underlines that disability routes have often acted as a substitute to other retirement routes. Changes in the claiming of disability benefits seem to match changes in pension schemes or controls more than changes in such health indicators as the mortality rates. However, the last section suggests increases in average health levels over the past two decades have come along with increased disparities. In that context, less generous pensions may induce an increase in the claiming of disability benefits partly because of substitution effects but also because the share of people with poor health increases.

References Barnay, T., and F. Jeger. 2006. “Quels dispositifs de cessation d’activité pour les personnes en mauvaise santé?” (“Which Exit Routes for People in Bad Health?”) Drees-Etudes et Résultats 492. Ben Salem, M., D. Blanchet, A. Bozio, and M. Roger. 2010. “Labor Force Participation by the Elderly and Employment of the Young: The Case of France.” In Social Security Programs and Retirement around the World: The Relationship to Youth Employment, edited by J. Gruber and D. Wise, 119– 46. Chicago: University of Chicago Press. Blanchet, D., and T. Debrand. 2008. “The Sooner, the Better? Analyzing Preferences for Early Retirement in European Countries.” IRDES Working Paper no. 13. Bozio, A. 2006. Réforme des retraites: Estimation sur données françaises.” (“Social Security Reforms: Estimation on French Data.”) PhD diss., EHESS, Paris. Jagger, C., C. Gillies, F. Moscone, E. Cambois, H. van Oyen, W. Nusselder, and J. M. Robine. 2008. “Inequalities in Healthy Life Expectancies in EU25.” Lancet 14:1– 8. Kusnik-Joinville, O., C. Lamy, Y. Merlière, and D. Polton. 2006. “Déterminants de l’évolution des indemnités journalières maladie.” (“Determinants of Changes in Sickness Benefits.”) CNAMTS Points de Repère 5. Lê, F., and D. Reynaud. 2007. “Les indemnités journalières.” (“Sickness Benefits.”) Drees-Etudes et résultats 592. Molinié, A.-F. 2006. “La santé au travail des salariés de plus de 50 ans.” Données Sociales, INSEE. Omnès, C. 2006. “Hommes et femmes face à la retraite pour inaptitude de 1945 à aujourd’hui.” (“Disability Pensions for Men and Women since 1945.”) Retraite et Société 49:77– 97. Strauss, J., and D. Thomas. 1998. “Health, Nutrition, and Economic Development.” Journal of Economic Literature 36 (2): 766– 817. Struillou, Y. 2003. Pénibilité et retraites. (Hard Working Conditions and Retirement.) Rapport remis au Conseil d’Orientation des Retraites. Volkoff, S., and F. Bardot. 2004. “Départs en retraite, précoces ou tardifs: À quoi tiennent les projets des salariés quinquagénaires?” “Early or Late Retirement Ages: What Determines the Plans of Workers above Fifty?”) Gérontologie et Société 111:71– 94.

9 Disability Insurance Programs in Canada Michael Baker and Kevin Milligan

9.1

Introduction

Disability insurance is now an important labor and health program in many countries. This development stems from a number of factors: changing attitudes about the nature of disability (such as the treatment of mental health); broad trends in the labor market; deliberate acts of policy. This latter factor includes not only direct changes to the eligibility or benefit rules of disability insurance programs, but also by changes to other programs (such as early retirement provisions in public pensions, unemployment insurance, or welfare) that might be substitute sources of income for those whose health is fading. As countries grapple with increasing stress on their retirement systems, the need for coordination between reforms of these programs and disability insurance programs is clear. In this chapter, we begin an examination of the impact of disability insurance on elderly Canadians. To do so, we pursue an analysis of the long- run trends in Canada Pension Plan Disability Insurance participation. We relate these trends to observed changes in different measures of health and an institutional analysis of policy changes in order to understand what is driving the changes in disability insurance receipt through time. The results provide clear evidence that changes in program rules have a Michael Baker is professor of economics at the University of Toronto and a research associate of the National Bureau of Economic Research. Kevin Milligan is associate professor of economics at the University of British Columbia and a research associate of the National Bureau of Economic Research. We thank the organizers of the NBER International Social Security Project and other country teams for their suggestions. For acknowledgments, sources of research support, and disclosure of the authors’ material financial relationships, if any, please see http://www.nber .org/chapters/c12390.ack.

327

328

Michael Baker and Kevin Milligan

large impact on receipt of disability insurance. By lining up the observed long- run trends with the timing of institutional changes, these relationships are uncovered. In contrast, the long- run trends in the measures of health that we are able to observe do not appear to have a strong relationship with the trends in disability insurance participation. We begin our investigation by describing the development of the Canada Pension Plan Disability Insurance program (and its sister program in the province of Quebec). We then review existing evidence on disability insurance in Canada. Next, we describe the data sources available to us for the analysis and then proceed to graph the time series and some cross- section patterns of health and program participation. We close with some thoughts on interpretation of the evidence. 9.2

Development of Disability Insurance in Canada

Until 1970, public long- term disability insurance was provided by provinces through Workers’ Compensation programs that covered employmentrelated accidents and disability claims. Private insurance supplemented these programs. As a backstop, the disabled might resort to provincial social assistance (welfare) programs. There was also some short- term sickness insurance available through the federal Unemployment Insurance program.1 The disability insurance landscape changed dramatically with the introduction of the Disability Insurance component of the Canada Pension Plan (CPP). The Canada Pension Plan passed into law in 1966 and comprised both a retirement benefits and disability benefits components. The Canada Pension Plan covered Canadians in nine provinces and the northern territories. The province of Quebec opted out of this initiative, instead setting up a sister program—the Quebec Pension Plan. The Canada Pension Plan is administered by the federal government, but it operates with the approval of the provinces. Changes must be supported by two- thirds of the “included” provinces with two- thirds of the total “included” population. The disability insurance component of the Canada Pension Plan began paying benefits in 1970. Entitlement depends on the finding of a disability that is “severe” and “prolonged”.2 Note that, unlike Workers’ Compensation, the source of the disability does not matter. In addition to the disability test, there is a requirement for attachment to the workforce. To be eligible, an individual must have contributed to the program either in four 1. See Campolieti and Lavis (2000) for some details on the roles of each of these programs, and Canada Pension Plan Disability Insurance, from 1970 to 1996. A history of the Disability Insurance component of the Canada Pension Plan is provided by Torjman (2002) and Prince (2002). Our discussion here draws on these sources. 2. According to the Canada Pension Plan Act section 42(2), “severe” is defined as being “incapable regularly of pursuing any substantial gainful occupation,” and “prolonged” means “likely to be long continued and of indefinite duration or is likely to result in death.”

Disability Insurance Programs in Canada

329

of the previous six years, or three out of the previous six years if he or she has made contributions to the CPP for twenty- five years or more. Workers’ Compensation programs continued to insure workplace accidents after the introduction of the Canada Pension Plan but were residual insurers—that is, Canada Pension Plan payments were subtracted from any Workers’ Compensation entitlements. Benefits under the Canada Pension Plan Disability Insurance program are comprised of three parts. First is a fixed amount—currently $426.13 per month.3 To this is added an earnings- related component. Earnings only up to a cap (set at approximately the economywide average earnings level) in each year of the earnings history are considered, with provisions for discarding low- earning months in the calculation. The base replacement rate amounts to 18.75 percent of average capped earnings. This reaches a current maximum of $700.63.4 Finally, the third component affects those with children under the age of eighteen, with a fixed monthly payment of $214.85. The average total payment made in July 2010 was $809.54 per month. At age sixty- five, benefits are transformed into retirement benefits. The parallel Quebec Pension Plan Disability Insurance program is administered separately, but the program parameters are very similar.5 There have been important differences between the Canada and Quebec Pension Plans through time in the definition of disability and how it is implemented. 9.2.1

Reforms through the 1980s, 1990s, and 2000s

Several changes occurred to the program starting in the middle of the 1980s. These changes are important for understanding the times series trends we uncover in the data. We describe these developments in the following. In 1987, the Canada Pension Plan Disability Insurance was reformed to increase the flat benefit to match the rate in the Quebec Pension Plan. Along with this change, eligibility was made easier, now requiring only work in two of three last years rather than five of the last ten. Finally, the ability to make claims retroactively was enhanced at this time. In general, this 1987 reform made the program more generous. A further reform in 1989 came in the form of an administrative policy guideline that instituted consideration of factors such as the local unemploy3. This amount is for the October to December quarter of 2010. It is in 2010 Canadian dollars. One Canadian dollar is worth 98 US cents in October 2010. 4. The cap is called the “Year’s Maximum Pensionable Earnings.” For 2010, it is set at $47,200. 5. The Quebec Pension Plan (QPP) requires that the disability be “permanent” and “severe,” which means that an individual is unable to work enough to earn more than an earnings threshold that was $13,521 in 2009. Applicants between the ages of sixty and sixty- four may be eligible if they have had to leave their usual employment as a result of their health and are no longer able to perform their usual work. QPP eligibility requires contributions in two of the previous three years, or five of the previous ten years, or in at least half of the individual’s contributory period (two- year minimum), which starts at age eighteen and ends either (a) the month preceding a retirement benefit claim, (b) the month of the seventieth birthday, or (c) the month of death.

330

Michael Baker and Kevin Milligan

ment rate, job skills, and socioeconomic factors in determining eligibility for those aged fifty- five to sixty- four in the Canada Pension Plan. There was also an increase to the flexibility of medical guidelines. Finally, there was a concern that Canadians were not well aware of the benefits available to them, so an information campaign was pursued to make people more aware of the program. A strong reversal occurred in 1995. The administrative guidelines on socioeconomic considerations were repealed. There was a reassessment of many existing claimants. Also, there was a facilitation of “self- sufficiency,” making it easier to return to work. In 1998, a major reform of the Canada and Quebec Pension Plan was implemented. On the financing side, the payroll tax to fund the plans was set on a sharply increasing schedule allowing future benefits to be partially prefunded. The benefit formulas were changed slightly as well.6 Most relevant to disability insurance, the eligibility requirement was changed to having worked in four of the last six years. In 2008, a slight change to eligibility led to the current rules. The change only affected those with more than twenty- five years of contributions over their lifetimes. Previous to 2008, one needed to have worked four of the last six years. As of March 3, 2008, however, those with twenty- five years or more of contributions were able to be eligible with work in three of the last six years. These developments can be seen in figure 9.1. We take the number of Disability Insurance beneficiaries between the ages twenty- five and sixtyfour and divide it by the total population in that age range. The source of the data is administrative reports on the Canada Pension Plan, as described in the next section. The participation rate grows steadily and sharply from 1971 to 1987, where a vertical line indicates the timing of the reforms that made Disability Insurance more generous. The growth continues to 1995, indicated by the second vertical line. In 1995, the eligibility criteria were tightened, resulting in a topping out of the growth of program participation at a rate just below 2.5 percent. Since then, the participation rate has settled at a level just over 2 percent and has been quite steady. 9.2.2

Research on Canada and Quebec Pension Plan Disability Insurance

Research on the Canada Pension Plan Disability Insurance program has touched on two topics of relevance for our focus. First, there is a literature examining the impact of disability insurance on the labor supply of older workers. Gruber (2000) exploits differences in the fixed component of the 6. Rather than using the past three years’ average value of the Year’s Maximum Pensionable Earnings in order to update lifetime earnings to current levels, the new formula used the past five years’ average value.

Disability Insurance Programs in Canada

Fig. 9.1

331

Growth of Canada Pension Plan Disability Insurance participation

Source: Canada Pension Plan Statistical Bulletin. Notes: Ages twenty- five to sixty- four, both sexes. Number of disability insurance pensions being paid divided by total population in the age range.

benefit formula between the Quebec and Canada Pension Plans in the 1980s, finding sizeable elasticities of work with respect to benefits. Campolieti (2004) performs a similar exercise for an earlier Quebec-Canada differential in the 1970s but finds small effects. He hypothesizes that the very tight screening on disability in place in the 1970s may have decreased the responsiveness to program parameters. Campolieti (2001) also looks at replacement rates, finding large effects both for men and women. The second strand of the literature of interest here is the impact of administrative rules and medical screening on program participation. Campolieti (2002) again compares Canada and Quebec Pension Plans, finding that the increase in the Canada Pension Plan flat benefit in 1987 led to an increase in hard- to-diagnose soft- tissue and musculoskeletal claims. He also finds that administrative tightening of the screening criteria in the Canada Pension Plan in 1995 decreases claims for disability from soft- tissue and musculoskeletal problems. Campolieti (2006) goes into greater depth on the 1995 reform, finding no evidence that easier- to-diagnose disabilities were affected by the tighter screening. As a summary of this evidence, it appears that there is substantial scope for substitution between the labor market and disability receipt for older workers. Moreover, eligibility and screening rules can have a large influence on participation as well.

332

9.3

Michael Baker and Kevin Milligan

Data

We pursue our analysis using several survey and some administrative data sources. In total, there are five sources of data that we bring together. Compared to the data available for other countries, Canada lacks panel data on elderly households (such as the Health and Retirement Study-English Longitudinal Study of Ageing-Survey of Health, Ageing and Retirement in Europe [HRS-ELSA-SHARE] data available elsewhere). This limits somewhat the panel dimension to the analysis we can perform. We begin with administrative data on usage of the Canada Pension Plan Disability Insurance benefits. We draw these data from a monthly publication called the Canada Pension Plan Statistical Bulletin. This publication contains comparable tables from 1971 to 2010 that allow us to construct time series of Canada Pension Plan Disability Insurance benefit receipt by sex and five- year age group. The second administrative data source is for mortality. Mortality data are collected as part of vital statistics separately by each province. Statistics Canada aggregates these into national numbers. We gather deaths and population counts by age, sex, and year using the data available for Canada in the Human Mortality Database.7 We combine the death and population data to form age- sex- year mortality rates. Our survey data start with the Labour Force Survey. Similar to the Current Population Study in the United States, the Labour Force Survey is a monthly survey used for high- frequency information about the state of the labor market. Around 50,000 individuals are in the monthly data. The Labour Force Survey lacks information on program participation and does not have complete information on why respondents are absent from the labor market. It does have excellent information on current labor market status, and we use it for that purpose here. We also use the Survey of Consumer Finances and the Survey of Labour and Income Dynamics for income information. These surveys span most years from 1971 to 1997 (Survey of Consumer Finances) and 1998 to 2007 (Survey of Labour and Income Dynamics). From this income information, we can derive some measures of program participation. As well, there is a detailed question for the reason someone is not working. These surveys are annual. The final source of survey data is the General Social Survey. In Canada, the General Social Survey asks a common core of questions each year, along with questions on one from a set of themes. Information on self- assessed health and the types of activity limitation is available sporadically across the time period 1985 to 2006. For all three of these surveys, we construct age- sex- year samples, using 7. This was collected from http://www.mortality.org.

Disability Insurance Programs in Canada

333

five- year aggregated age groups. We also employ the provided sample weights to ensure representativeness. 9.4

Results

The following results proceed through four steps. The first step is to examine the broad time trends on measures of health and mortality. The second set of results documents the participation in the different government programs that are used by older workers for income. Next, we uncover the relationships between disability and labor market participation. Finally, there are some graphs relating disability and other measures of health. 9.4.1

Mortality and Health

The analysis of mortality and health has two goals. Not only are the time trends interesting here, but we are also interested in how well health and mortality trends correspond. This correspondence is important because mortality is the best measured and most internationally comparable measure of health that we have available. To the extent that mortality and other measures of health are aligned, we can have greater confidence in using mortality trends as general indicators of health. The first mortality graph presented in figure 9.2 attempts to measure the progress in mortality since 1961. We begin with the mortality rates at age sixty and sixty- five in 1961, separately for males and females. The initial mortality rate in 1961 for males is 2.1 percent at age sixty and 3.2 percent

Fig. 9.2

Age at which 1961 aged sixty and sixty-five mortality is reached

Source: Statistics Canada mortality data.

334

Michael Baker and Kevin Milligan

for age sixty- five; for females 1.1 percent at age sixty and 1.8 percent at age sixty- five. The lines in figure 9.2 show the age at which mortality in years after 1961 reaches the age sixty and age sixty- five levels seen in 1961. For example, in 1971 for sixty- five- year- old males, it took until age 66.2 to hit the 3.2 percent mortality rate seen for sixty- five- year- olds in 1961.8 That is, it took an extra 1.2 years of age to reach the same mortality rate, so that 66.2 year old males in 1971 are like sixty- five- year- olds in 1961 in terms of mortality. Mortality improvements were typically very strong starting in the 1970s. The line for women is above the line for men at both ages sixty and sixty- five until the mid- 1990s, when improvements for men became sharper. By the end of our data period, there is a remarkable improvement in mortality. The mortality rate seen for sixty- year- old males in 1961 was not reached until age 69.5 in 2007—an increase of almost a decade of life. For women, the increase was slightly less at 67.8. Given the mortality advantage of females, this served to close slightly the gap between male and female mortality. Age sixty- five mortality rates for men and women improved to ages 73.9 for men and 72.9 for women. These developments show a very tangible improvement in mortality over this forty- six- year period. These same trends can be presented in a different way in figure 9.3. In this graph, we show the cross- sectional mortality rates for the two end- point years in our data, being 1961 and 2007. The drops in the curves for each sex between the two years indicate substantial mortality improvements. For men, the improvement at age sixty is 58 percent; at age seventy is 54 percent; and at age eighty is 42 percent. For females at age sixty the improvement is 54 percent; at age seventy is 50 percent; at age eighty is 54 percent. This suggests a fairly similar improvement at different points in the mortality curve and across sexes. A third way to visualize these mortality changes is to graph age- specific mortality rates through time. Figures 9.4 and 9.5 do so for men and for women, at ages fifty- five, sixty, and sixty- five. While the magnitude of the drop through time for age sixty- five is larger than for age sixty or age fiftyfive, the percentage drop for all three is around 50 percent. Women in figure 9.5 display a similar pattern. Taken together, this evidence suggests a strong and fairly consistent improvement in mortality rates over the period 1961 to 2007. The gains were of comparable percentage changes across ages and sexes. We next turn

8. We only have mortality by single year of age so the exact age at which the mortality target is reached is determined by linear interpolation. For example, the age sixty- five male target of 3.2 percent mortality sits about 20 percent of the way between the mortality rates seen for ages sixty- six and sixty- seven in 1971, yielding a mortality equivalent age of 66.2. (We perform these operations with several decimal points, but round them to one decimal point here for ease of exposition.)

Disability Insurance Programs in Canada

Fig. 9.3

335

Cross-sectional mortality in 1961 and 2007

Source: Statistics Canada mortality data.

Fig. 9.4

Mortality at specific ages for men

Source: Statistics Canada mortality data.

to an examination of how well these mortality improvements correspond to subtler measures of health. For the health analysis, we combine three measures of self- assessed health with our previously graphed mortality rates. All three are taken from the General Social Survey. The first year of data available is 1985, with more frequent responses then available after 1990.

336

Fig. 9.5

Michael Baker and Kevin Milligan

Mortality at specific ages for women

Source: Statistics Canada mortality data.

The first measure we use is from a question on self- assessed health.9 We code those responding that their health is “fair” or “poor” as a 1, and those feeling “excellent,” “good,” or “very good” as a 0. In this way, the measure is an indicator of worse health and should go in the same direction through time as mortality if general health improvements are behind the mortality trends. The second measure is based on a question of long- term activity limitation.10 Those who are limited are coded with a 1; those not limited are coded with a 0. Note that this definition is broader than the employmentbased definition used for eligibility for the Canada Pension Plan. The third measure looks at satisfaction with health. It is only available for three years in the sample, but we include it to capture—albeit weakly—any trend in this variable. We code it to 1 if someone expresses that they are “not satisfied” with their health, comprising the categories “very” and “somewhat” unsatisfied.11 Together, these three variables capture more subjective and more subtle elements of health than does mortality. Figures 9.6 and 9.7 show the evolution of our three subjective health measures compared against mortality, defined as the number of deaths over the population in the given age range. We use only the time period between 1985 9. The specific question is “In general, would you say your health is:.” The five possible responses are excellent, very good, good, fair, and poor. 10. The specific question is “Are you limited in the amount or kind of activity you can do at home, at work, or at school because of a long- term physical or mental condition or health problem?” 11. The specific question is “How do you feel about your health?” The allowed answers are very satisfied, somewhat satisfied, somewhat dissatisfied, and very dissatisfied.

Disability Insurance Programs in Canada

Fig. 9.6

337

Subjective health and mortality at ages sixty to sixty-four for men

Sources: Statistics Canada for the mortality data and the General Social Survey for subjective health data.

Fig. 9.7

Subjective health and mortality at ages sixty to sixty-four for women

Sources: Statistics Canada for the mortality data and the General Social Survey for subjective health data.

and 2007, which spans the available years of the General Social Survey. We take the age group sixty to sixty- four in all cases. The mortality rates are graphed against the right- hand- side vertical axis, while the subjective health measures are graphed against the left- hand- side vertical axis. Figure 9.6 for men shows a clear downward trend for self- assessed fair

338

Michael Baker and Kevin Milligan

or poor health. From over 33 percent in 1985, the level drops to 11 percent in 2007, with steady improvements through the 1990s. Dissatisfaction with health goes down for the three years observable. Activity limitations, however, are quite steady at around 20 percent over the years available in the General Social Survey. For women in figure 9.7, the patterns are similar, but weaker than for men. There is a decline in self- assessed fair or poor health, but it is less of a decline than for men. Health dissatisfaction increases over the three years shown. Activity limitations, like for men, are quite steady. This analysis of subjective health measures has uncovered some evidence that more subtle measures of health—specifically self- assessed fair or poor health—show an improvement over the 1985 to 2007 period. However, activity limitations do not show this same change. Overall, these results do not contradict the strong improvements seen in the more objective mortality measures. 9.4.2

Programs and Participation

We now turn our attention to trends in program participation through time. The goal here is to look for any impact of the reforms we have outlined in the preceding as well as for any possible substitution among different government programs. In this analysis, we draw mostly on the annual income source data in the Survey of Consumer Finances/Survey of Labour and Income Dynamics. We also use the administrative data from the Canada Pension Plan and some questions on why there was a departure from a job from the Labour Force Survey. We begin with cross- sectional participation rates by age in 1981 for men and women. We show indicators of receipt of social assistance, Canada/ Quebec Pension Plan benefits, and Unemployment Insurance benefits from the survey data, as well as the rate of Canada Pension Plan Disability Insurance recipiency from the administrative data. For social assistance, the variable in the Survey of Consumer Finances includes both social assistance (welfare) and also provincial supplements that are paid to low- income seniors age sixty- five plus. For the Canada/Quebec Pension Plan payments, the data cannot distinguish between retirement, survivor, and disability payments. Figure 9.8 shows the cross- section by five year age groups in 1981. In 1981, retirement benefits through the Canada and Quebec Pension Plans could not be taken until age sixty- five. Furthermore, few men receive survivor benefits because few men have lost their spouse before age sixty- five. So we expect most of those receiving Canada/Quebec Pension Plan income are receiving Disability Insurance benefits. In figure 9.8, the Canada Pension Plan Disability Insurance line from the administrative data is very close to the Canada/Quebec Pension Plan benefit receipt line, until age sixty- five is reached. The level of benefit receipt for the Canada/Quebec Pension Plan

Disability Insurance Programs in Canada

Fig. 9.8

339

Program participation in 1981 for men

Sources: CPP/QPP benefits received, unemployment benefits, and social assistance come from the Survey of Consumer Finances. Disability rate, CPP, comes from the Canada Pension Plan Statistical Bulletin.

at ages after sixty- five is 79 percent in the sixty- five to sixty- nine age range. This is less than 100 percent because some men may still be working and have not yet started their benefits. For the seventy to ninety- nine age range, the rate is even lower at 67 percent. This reflects the fact that those retiring before 1967 received no Canada Pension Plan benefits because the program only began in that year. In 1981, that would include everyone age seventynine and older. Unemployment Insurance benefits are fairly steady between ages forty and sixty- four, giving little indication that Unemployment Insurance benefits are being used as a substitute for early retirement benefits. Finally, social assistance rates are fairly low until age sixty- five, when they leap upward corresponding to the availability of provincial income supplements for low- income seniors. Figure 9.9 repeats the exercise for women. Women are less likely to be on Unemployment Insurance benefits because they are less likely to be eligible because of lower labor- force attachment. There is a substantial gap between the Canada Pension Plan Disability Insurance indicator from administrative data and the indicator for receipt of Canada/Quebec Pension Plan income for women. This makes sense, however, because it is much more common for women in this age range to be receiving survivor benefits following the death of a spouse than it was for men. Participation rates at ages sixty- five to sixty- nine and seventy to seventy- four are lower than for men, again because the labor- market attachment of women is lower, meaning less eligi-

340

Fig. 9.9

Michael Baker and Kevin Milligan

Program participation in 1981 for women

Sources: CPP/QPP benefits received, unemployment benefits, and social assistance come from the Survey of Consumer Finances. Disability rate, CPP, comes from the Canada Pension Plan Statistical Bulletin.

bility. The social assistance and Unemployment Insurance benefits appear similar in trend to the men. We next move forward to 2007. The largest difference in 2007 is the availability of early retirement benefits through the Canada/Quebec Pension Plan at ages sixty to sixty- four, instituted in 1984 in the Quebec Pension Plan and 1987 in the Canada Pension Plan.12 In figure 9.10, we graph the crosssectional program participation rates for men by five- year age group. Figure 9.11 does the same for women. The largest difference in the graph is the jump in Canada/Quebec Pension Plan participation at ages sixty to sixty- four. With the availability of an early retirement option through the Canada/ Quebec Pension Plan, will there be a decrease in uptake of Disability Insurance benefits? The graph doesn’t show any—the disability rate taken from Canada Pension Plan administrative data comes in at approximately the same level as was the case for 1981. There is no sign that the new early retirements were drawing people out of disability insurance claims. The previous graphs showed the programs from which Canadians were drawing income. These showed the “stocks” at each age and year. To get a more complete picture, we now attempt to put together some information on the “flows” out of work. To do so, we turn to the Labour Force Survey. The Labour Force Survey asks respondents who are not employed the reason 12. See Baker and Benjamin (1999) for details and evidence on the introduction of early retirement in Canada.

Fig. 9.10

Program participation in 2007 for men

Sources: CPP/QPP benefits received, unemployment benefits, and social assistance come from the Survey of Labour and Income Dynamics. Disability rate, CPP, comes from the Canada Pension Plan Statistical Bulletin.

Fig. 9.11

Program participation in 2007 for women

Sources: CPP/QPP benefits received, unemployment benefits, and social assistance come from the Survey of Labour and Income Dynamics. Disability rate, CPP, comes from the Canada Pension Plan Statistical Bulletin.

342

Michael Baker and Kevin Milligan

for leaving their last job. Unfortunately, this question is only asked of those who worked in the previous year, so those not working for a year or more are not asked the question. Still, this question can give us an indication if the flow out of jobs into retirement or into disability has shown any large changes through time. Figure 9.12 shows the flows out of the labor market for men and figure 9.13 for women. Both graphs show age groups fifty- five to fifty- nine and sixty to sixty- four. In all cases, the denominator for the calculation of this rate is all survey respondents of that sex and age. For men, there is an increase in the rate of departures to retirement at ages fifty- five to fiftynine during the 1980s. For both ages fifty- five to fifty- nine and sixty to sixty- four, there are declines starting around 1995 in the flow out of work and into retirement. If retirement and disability were substitutes, we might expect a mirrored response for departures to disability. However, this is not evident here, with departure rates to disability dropping slightly from more than 1 percent to under 1 percent. No clear correspondence with retirement flows is evident. Women in figure 9.13 show an increase departure rate to retirement thought time. This may reflect the fact that more women in these later years are in the labor force, meaning that more of them have a job from which they can retire. Again, there is no clear correspondence to the departure rates to disability. The final graphs for this analysis of program participation compares different measures of disability to get a better view of how they may be related. In figure 9.14 for men and figure 9.15 for women, we compare the disability rate from the Canada Pension Plan administrative data, the flow

Fig. 9.12

Flows out of the labor market, men

Source: Labour Force Survey.

Fig. 9.13

Flows out of the labor market, women

Source: Labour Force Survey.

Fig. 9.14

Comparing disability measures, men aged sixty to sixty-four

Sources: Labour Force Survey; the Survey of Consumer Finances; and the Canada Pension Plan Statistical Bulletin.

344

Michael Baker and Kevin Milligan

Fig. 9.15

Comparing disability measures, women aged sixty to sixty-four

Sources: Labour Force Survey; the Survey of Consumer Finances; and the Canada Pension Plan Statistical Bulletin.

of people leaving their work because of disability from the Labour Force Survey, and the proportion who are not currently working because of disability from the Survey of Consumer Finances. We do this analysis for those aged sixty to sixty- four. For neither men nor women does there appear to be any interesting trends in the flow into disability through time. For women, the uptrend in those not working because of disability in the administrative data matches very closely the trend in the survey data. For men, there is a similar uptrend, but the correspondence isn’t as close. Unfortunately, the Survey of Labour and Income Dynamics that replaced the Survey of Consumer Finances after 1997 did not contain a comparable question on disability, so we cannot observe if the downward trend in the disability rate in the administrative data is mirrored in the survey data after 1997. Our analysis of program participation reveals two important findings. First, there is very little indication of disability program participation being a substitute for early retirement or unemployment benefits. Second, there is a strong correspondence between survey and administrative data sources for the measurement of disability insurance participation. 9.4.3

Labor Markets and Disability

We next address the question of how overall labor market decisions are affected by changes in receipt of disability insurance. This relates directly to the literature on disability insurance generosity and employment discussed earlier. We graph employment, unemployment, and nonattachment to the

Disability Insurance Programs in Canada

345

labor force for men and women through time. We then look at trends in the reasons not working—including disability—before looking again at the trends in disability uptake. Figures 9.16 and 9.17 display the employment rate for men and women, respectively, across three different age groups using Labour Force Survey data. For men aged forty to forty- four and fifty to fifty- four in figure 9.16,

Fig. 9.16

Employment across age groups, men

Source: Labour Force Survey.

Fig. 9.17

Employment across age groups, women

Source: Labour Force Survey.

346

Michael Baker and Kevin Milligan

most work. There are some slight business cycle fluctuations across time, but the magnitude of these fluctuations is not large. For men aged sixty to sixty- four, there is a downward trend in employment from 63.6 percent in 1976 down to 39.9 percent by 1995. This largely reflects the introduction of early retirement options under the Canada and Quebec Pension Plans in the 1980s. The downward trend and the changing point in 1995, though, line up quite neatly with the patterns of participation in Canada Pension Plan Disability Insurance seen earlier. Since 1995, around half of this drop in employment has been recovered, leaving employment at 52.6 percent in 2009. One explanation for this upswing is provided in Schirle (2008), who argues that the increase in older male labor market participation was driven largely by a preference for joint retirement with their wives—and women of the cohorts in this age range since 1995 were much more likely to work than earlier cohorts. For females in figure 9.17, any business cycle effects are dominated by an upward trend in employment driven by cohort differences in lifetime employment attachment. The pickup in female employment in the 1970s and 1980s at ages forty to forty- four and fifty to fifty- four is echoed by these same cohorts in the age range sixty to sixty- four after 2000. This increase in employment by older women is consistent with the story in Schirle (2008) mentioned in the preceding. Unemployment is graphed in figures 9.18 and 9.19 for men and women. Unemployment for both sexes at older ages is low. There are obvious business cycle effects, but no clear correspondence to developments in disability

Fig. 9.18

Unemployment across age groups, men

Source: Labour Force Survey.

Disability Insurance Programs in Canada

Fig. 9.19

347

Unemployment across age groups, women

Source: Labour Force Survey.

insurance. This suggests that there is no large- scale substitution between disability insurance and unemployment. The proportion of men and women out of the labor force in figures 9.20 and 9.21 across ages look quite similar to what was seen for employment, but in reverse. This suggests that the broad trends in employment discussed in the preceding reflected moves from being out of the labor force into employment, more so than between being unemployed and employed. The reasons for being out of the labor market can be decomposed using data from the Survey of Consumer Finances and the Survey of Labour and Income Dynamics. In figure 9.22, we graph the proportion of men aged sixty to sixty- four who reported that their primary activity over the previous year was not working or looking for work. The top line shows all those not working or looking for work. This measure does not show the sharp turnaround in 1995 that was evident with the Labour Force Survey data in figure 9.20. This difference may be driven by the fact that the Labour Force Survey asks about employment activity in a reference week, while the Survey of Consumer Finances focuses on activity over the whole year. The second and third lines in figure 9.22 show the proportion that is not working or looking for work because of retirement or because of disability.13 The disability variable is not available after 1998 because of changes in the way disability was measured in the Survey of Labour and Income Dynamics 13. By “disability” here, we mean that respondents reported that they were “permanently unable to work.”

348

Michael Baker and Kevin Milligan

Fig. 9.20

Out of the labor force across age groups, men

Source: Labour Force Survey.

Fig. 9.21

Out of the labor force across age groups, women

Source: Labour Force Survey.

in that year. A large majority of those not working are not working because of retirement. There is a fairly large percentage increase in those not working because of disability from 1971 through 1995 (as seen earlier in figure 9.14), but this does not explain a lot of the overall trend in not working because the proportion out of work for disability remains fairly small as a proportion of all of those out of work.

Disability Insurance Programs in Canada

Fig. 9.22

349

Reasons not working or looking for work, men aged sixty to sixty-four

Sources: Survey of Consumer Finances and the Survey of Labour and Income Dynamics.

Fig. 9.23

Reasons not working or looking for work, women aged sixty to sixty-four

Sources: Survey of Consumer Finances and the Survey of Labour and Income Dynamics.

Figure 9.23 repeats the decomposition of not working for women. Here, the gap between the total proportion not working or looking for work and those who are retired is very large—especially at the earlier years of the time series. This reflects the fact that a large proportion of women in these cohorts during these years were at home rather than in the paid workforce. Retirement increases through time because more women had work from

350

Michael Baker and Kevin Milligan

which to retire. The increase in disability evident in the bottom line is again large as a percentage, but small in terms of how much it can explain overall labor force trends. This section has found several interesting relationships between labor market behavior and the receipt of disability benefits. First, there is no evidence of using unemployment insurance as a substitute for disability insurance benefits. Second, for women, the large increase across birth cohorts in labor market attachment is the dominant force in the labor market over this time period and disability insurance plays a relatively minor role. Finally, for men there is an interesting correspondence of the trends in employment and disability insurance receipt at ages sixty to sixty- four, before and after 1995. This suggests that there might have been some substitution between disability insurance and employment for men aged sixty to sixty- four over this time period. 9.4.4

Employment and Mortality

Another way to look at the relationship between employment and health is to ask how much work is done for a given level of health. To do so, we graph in figure 9.24 the employment rate and mortality rate of men in 1976 and 2007. The mortality rate is on the horizontal axis, and the employment rate is on the vertical axis. Each point represents the average value for a five- year age group, as labeled. At a mortality rate of 0.01, around 85 percent of males were employed in 1976. However, by 2007 at the same level of mortality, only half of males were employed. Looking the other way, it

Fig. 9.24

Employment and mortality, men

Sources: Labour Force Survey for employment and Statistics Canada for the mortality data.

Disability Insurance Programs in Canada

Fig. 9.25

351

Employment and mortality, women

Sources: Labour Force Survey for employment and Statistics Canada for the mortality data.

took a mortality rate of 0.025 to reach a point of 50 percent employment in 1976, but only 0.01 to reach that same employment rate in 2007. This suggests that, given similar health levels (as measured by mortality), there was lower employment in 2007 than 1976. Figure 9.25 repeats the same analysis for females. Here, the most important trend is the great increase in employment by females with younger cohorts of women. In this graph, this manifests as a higher employment rate at lower levels of mortality although the lines cross at older ages. This suggest that, for the same level of mortality, we see more employment by females in 2007 than 1976. 9.4.5

Disability and Health

In our last set of results, we look for patterns across age, sex, and time in the relationship between health and disability. We begin by looking at receipt of disability pensions in the Canada Pension Plan and then compare these findings to our measures of mortality and self- assessed health. The first two graphs of this analysis of labor markets and disability show disability rates by age group and sex using the Canada Pension Plan administrative data. For men (figure 9.26), the peak in 1995 is evident both for those aged sixty to sixty- four (as in figure 9.14) and those at younger ages. The peak at younger ages is comparable in terms of percentage increases, but as a share of the male population at ages forty to forty- four and fifty to fifty- four, disability insurance plays a small role.

352

Michael Baker and Kevin Milligan

Fig. 9.26

Canada Pension Plan Disability Insurance benefits recipients, men

Source: Canada Pension Plan Statistical Bulletin.

The data for women appear in figure 9.27. Over this time period, more women are entering the workforce, as seen in figure 9.17. This means that more women became eligible for Canada Pension Plan Disability Insurance from 1971 onward. This leads to a steeper upward trend—not only is uptake among those who are working growing through this time period, but also the proportion working is growing. After 1995, the proportion receiving a Disability Insurance benefit drops a bit before becoming quite constant at all ages. Figures 9.28 and 9.29 focus on the disability rates of men and women aged forty- five to forty- nine and compare them to the mortality rates prevailing in those age groups. In figure 9.28, mortality rates trend clearly down from 1970 to 2007, while disability rates follow first an upward trend, then down after the policy changes of 1995. For women in figure 9.17, there is a clear, sharp upward jump in disability pension receipt in the early 1990s, but the rate soon stabilized around 2 percent. In neither the case of men or women is there an obvious relationship between mortality and Canada Pension Plan Disability Insurance benefit receipt. The next two figures repeat the analysis for the age sixty to sixty- four age group. Again, the upward trend in disability benefit participation is contrasted by trending improvements in mortality. For men in figure 9.30, there is also a fairly large downward trend after the policy reforms of 1995. There

Fig. 9.27

Canada Pension Plan Disability Insurance benefits recipients, women

Source: Canada Pension Plan Statistical Bulletin.

Fig. 9.28 Canada Pension Plan Disability Insurance benefits recipients and mortality rate, men aged forty-five to forty-nine Sources: Canada Pension Plan Statistical Bulletin and Statistics Canada mortality data.

Fig. 9.29 Canada Pension Plan Disability Insurance benefits recipients and mortality rate, women aged forty-five to forty-nine Sources: Canada Pension Plan Statistical Bulletin and Statistics Canada mortality data.

Fig. 9.30 Canada Pension Plan Disability Insurance benefits recipients and mortality rate, men aged sixty to sixty-four Sources: Canada Pension Plan Statistical Bulletin and Statistics Canada mortality data.

Disability Insurance Programs in Canada

355

Fig. 9.31 Canada Pension Plan Disability Insurance benefits recipients and mortality rate, women aged sixty to sixty-four Sources: Canada Pension Plan Statistical Bulletin and Statistics Canada mortality data.

is no seeming correspondence to mortality. With women in figure 9.31, the picture looks largely similar as it did for younger women in figure 9.29—no clear relationship is evident. We now turn to self- assessed health in the final four figures. At ages fortyfive to forty- nine for men in figure 9.32, the initial data point for 1985 for being in fair or poor health is at 2 percent. However, for each of the other years, the data bounce in a tight band between 0.005 and 0.01. The same pattern is broadly true for women in figure 9.33 although more of a downward trajectory is evident. In neither case, however, do these trends line up easily with what is observed for Canada Pension Plan Disability Insurance benefit receipt. In figures 9.34 and 9.35, we show the same results for Canadians aged sixty to sixty- four. A stronger downward trend for self- assessed fair or poor health is evident, but again this does not conform with the trends seen for disability benefit receipt. This analysis of receipt of Canada Pension Plan Disability Insurance benefit receipt and observable measures of health exposes no clear relationship between disability benefit receipt and either of the observable health measures. This evidence makes it hard to suggest that the trends in disability benefits receipt are related to underlying actual health trends.

Fig. 9.32 Canada Pension Plan Disability Insurance benefits recipients and self-assessed health, men aged forty-five to forty-nine Sources: Canada Pension Plan Statistical Bulletin and General Social Survey.

Fig. 9.33 Canada Pension Plan Disability Insurance benefits recipients and self-assessed health, women aged forty-five to forty-nine Sources: Canada Pension Plan Statistical Bulletin and General Social Survey.

Fig. 9.34 Canada Pension Plan Disability Insurance benefits recipients and self-assessed health, men aged sixty to sixty-four Sources: Canada Pension Plan Statistical Bulletin and General Social Survey.

Fig. 9.35 Canada Pension Plan Disability Insurance benefits recipients and self-assessed health, women aged sixty to sixty-four Sources: Canada Pension Plan Statistical Bulletin and General Social Survey.

358

9.5

Michael Baker and Kevin Milligan

Conclusion

This paper examines disability insurance programs in Canada using a variety of data sources spanning the years 1961 to 2009—almost a half century. Over this period, we document substantial and steady improvements in mortality and some improvements in measures of subjective health. We also see large movements of women into the labor force and significant trend to early retirement for men. Disability insurance receipt doesn’t appear to relate well to these trends in health or mortality, which suggests that something other than underlying health is driving disability benefit receipt. In the labor market, we don’t find any evidence of substitution between unemployment insurance and disability insurance receipt, but there are some time series correspondences between early retirement at ages sixty to sixty- four and disability insurance trends—especially around the time of a major Canada Pension Plan reform in 1995.

References Baker, Michael, and Dwayne Benjamin. 1999. “Early Retirement Provisions and the Labor Force Behavior of Older Men: Some Evidence from Canada.” Journal of Labor Economics 17 (4): 724– 56. Campolieti, Michele. 2001. “Disability Insurance and the Labour Force Participation of Older Men and Women in Canada.” Canadian Public Policy 27 (2): 179– 94. ———. 2002. “Moral Hazard and Disability Insurance: On the Incidence of Hardto-Diagnose Medical Conditions in the Canada/Quebec Pension Plan Disability Program.” Canadian Public Policy 28 (3): 419– 41. ———. 2004. “Disability Insurance Benefits and Labor Supply: Some Additional Evidence.” Journal of Labor Economics 22 (4): 863– 89. ———. 2006. “Disability Insurance Adjudication Criteria and the Incidence of Hard- to-Diagnose Medical Conditions.” Contributions to Economic Analysis and Policy 5 (1): article 15. Berkeley, CA: Berkeley Electronic Press. Campolieti, Michele, and John N. Lavis. 2000. “Disability Expenditures in Canada, 1970– 1996: Trends, Reform Efforts, and a Path for the Future.” Canadian Public Policy 26 (2): 241– 64. Gruber, Jonathan. 2000. “Disability Insurance Benefits and Labor Supply.” Journal of Political Economy 108 (6): 1162– 83. Prince, Michael J. 2002. “Wrestling with the Poor Cousin: Canada Pension Plan Disability Policy and Practice, 1964– 2001.” Office of the Commissioner of Review Tribunals, Government of Canada. Unpublished Manuscript. Schirle, Tammy. 2008. “Why Have the Labor Force Participation Rates of Older Men Increased Since the Mid- 1990s?” Journal of Labor Economics 26 (4): 549– 94. Torjman, Sherri. 2002. “The Canada Pension Plan Disability Benefit.” Caledon Institute of Social Policy Report. http://www.ocrt- bctr.gc.ca/dap- dep/r012002/ r012002-eng.pdf.

10 The Long-Run Growth of Disability Insurance in the United States Kevin Milligan

10.1

Introduction

The Social Security Disability Insurance program provided benefits for over 8 million Americans in 2010. The pace of growth of public expenditures in this program, however, is pushing hard against fiscal constraints. Autor and Duggan (2006) project a “crisis” if no serious reforms are attempted. It is in this context that the sources of growth of Social Security Disability Insurance are being examined. In this chapter, I pursue two goals. First, I provide long time series of data from a variety of sources related to Social Security Disability Insurance. These long time series, with data stretching almost sixty years from 1950 to 2009, allow for perspective and context on the different potential causes of the growth of the program. The second goal is to provide data that are comparable to what is available in other countries in this project. This comparability will help place the experience in the United States in context, and the cross- country comparisons may also yield important insights. There are three major findings of the chapter. First, health (as measured by mortality and self- assessed health) seems to have little relationship to any observed trend in the Social Security Disability Insurance program. Second, legislative and administrative changes such as benefit formulas, screening rules, and eligibility criteria have a very clear and large impact on recipiency. Finally, labor market trends, such as the emergence of cohorts of women Kevin Milligan is associate professor of economics at the University of British Columbia and a research associate of the National Bureau of Economic Research. I thank the organizers of the NBER International Social Security Project and other country teams for their suggestions. For acknowledgments, sources of research support, and disclosure of the author’s material financial relationships, if any, please see http://www.nber.org/chapters /c12391.ack.

359

360

Kevin Milligan

with long attachment to the labor force, can also make a large difference to long- run trends in Disability Insurance recipiency. The chapter is laid out as follows. First, I review the details of the current working of the Social Security Disability Insurance program and how the program has developed through time. Next is a brief review of the relevant research. I then turn to a description of the data sources used here, followed by a thorough discussion of the time series and cross- sectional graphs that make up the analysis of this chapter. I conclude with a summary of the findings. 10.2

The Development of Social Security Disability Insurance

I begin with a description of the current state of the Social Security Disability Insurance (hereafter SSDI) program, including eligibility and benefit calculation rules. I also briefly describe related programs. Following this exposition of the current rules, I go through the history of the program, pointing out major changes that may have an impact on the behavior of the beneficiaries of the program. 10.2.1

Current Benefit Rules for SSDI and Related Programs

Benefits under SSDI are earnings- related. Qualification depends on a system of “credits.” Credits depend on meeting a minimum level of earnings, and up to four credits are available each year. Generally, an applicant needs to have earned twenty credits over the last ten years, and up to forty overall.1 The latter qualification rule depends on age, though, so those under age sixty- two can qualify with fewer than forty overall credits.2 The disability must be found to be “severe” and expected to last at least twelve months (or be expected to result in death). The benefit formula uses many of the same features as for Social Security Old Age benefits. Monthly earnings over the lifetime are updated to current dollar levels using an inflator that depends on the average earnings growth in the economy.3 Some periods of low earnings may be discarded, and covered earnings are capped at a maximum level.4 The result of this calculation is the Average Indexed Monthly Earnings, or AIME. The AIME is then put through a nonlinear benefit schedule that replaces 90 percent of AIME up to a first “bend point,” 32 percent up to a second bend point, and 15 percent of AIME after that. The bend points in 2010 are $761 and $4,516 per 1. For 2010, a credit is earned for each $1,120 of wages in a year, up to four credits per year. This amount is indexed to average wages. 2. There is a sliding scale. Those under age forty- two need only twenty overall credits, and this amount increases by age until it reaches forty required credits at age sixty- two. Under age thirty- one, there are also lower qualifying conditions for work over the last ten years. 3. Specifically, earnings at time t are inflated to time T using the ratio of economywide average earnings at times T – 2 and t. 4. The Social Security cap for 2010 is $106,800.

The Long-Run Growth of Disability Insurance in the United States

361

month. Benefits are partially taxable, depending on other sources of income. Overall, only one- third of all Social Security recipients pay any tax on their benefits. The average monthly benefit as of August 2010 was $1,066. When the recipient reaches the Federal Retirement Age, the benefit is transformed into a Social Security Old Age benefit.5 Social Security Disability Insurance is funded as part of the Federal Insurance Contributions Act (FICA) package of payroll taxes on employers and employees.6 There is an explicit amount, currently 0.9 percent on employers and employees, set aside for SSDI. Payment for SSDI is made for earnings up to the annual cap, currently at $106,800. The self- employed must pay both employer and employee portions, totaling 1.8 percent. The inflows and outflows on the overall Social Security and the SSDI accounts are kept separately but are also included on the unified federal government budget. There have been periodic changes to the FICA rates, most recently in 1990. Beyond SSDI, there are other recourses for those unable to work. First, workplace injuries are insured by state- level workers’ compensation systems. These systems vary dramatically by state, incorporating different mixes of private and public provision, employer mandates, and regulation. Second, those who are disabled (or blind) and have quite low incomes can be eligible for Supplementary Security Income. This program has strict asset and income tests, and the benefit levels are quite modest ($674 per month for singles and $1,011 per month for couples in 2010). There are nearly the same number of beneficiaries of Supplementary Security Income as SSDI (although Supplemental Security Income also covers lower- income people over sixty- five), but the average benefit of $498 is around half that for SSDI. These benefits may be combined with SSDI benefits, providing income and asset tests are met. Finally, state- run unemployment insurance and welfare programs may also pick up those not qualifying for other programs. In particular, Autor and Duggan (2003) find substantial substitution between unemployment insurance and SSDI among older, lower- educated workers. 10.2.2

Historical Development of SSDI

The Social Security Disability program was added to the existing Social Security old age retirement benefits in 1956.7 The idea of disability insurance had been an active area of discussion throughout the 1940s and 1950s, but concerns over the definition of disability, how to administer the program, 5. The Federal Retirement Age is being moved from age sixty- five to age sixty- seven across cohorts. It was sixty- five for those hitting age sixty- five before 2003 and reached sixty- six for those hitting sixty- six in 2009. By 2027, it will be at age sixty- seven. 6. FICA stands for Federal Insurance Contributions Act. Also included in the package are amounts for Social Security retirement and survivors benefits and Medicare. 7. This discussion draws on Autor and Duggan (2006), Berkowitz (1989, 2000), and Social Security Administration (1986, 2009).

362

Kevin Milligan

Fig. 10.1

Social Security Disability Insurance recipiency, men and women

Sources: Social Security Administration and Census data. Note: Total number of SSDI recipients who were workers (i.e., not dependents or spouses) divided by population age twenty- five to sixty- four.

and how to structure the benefits slowed its introduction. In the following, I first describe the evolution of eligibility rules, then discuss how the benefit formula has changed. The overall trends in uptake are shown in figure 10.1, which graphs the number of Social Security Disability Insurance recipients who were workers, divided by the population age twenty- five to sixty- four, both sexes combined.8 As part of the compromise to get the bill passed in 1956, SSDI was limited to those aged fifty to sixty- four with an “inability to engage in substantial gainful activity because of any medically determinable permanent physical or mental impairment.” The disability had to be experienced for a continuous period of six months. Eligibility was expanded to those under age fifty in 1960, and, in 1965, the requirement for permanent disability was relaxed— one only had to have a disability expected to last a year or more. Specifically, the legislation now stated that the disability “be expected to last for a continuous period of not less than 12 months.” This led to some increase in uptake of SSDI, and a 1967 amendment reemphasized the medical nature of disability and changed the definition to a “disability that precludes engagement in any substantial gainful work existing in the national economy.” In 1972, the continuous months of disability requirement was relaxed from six to five months. 8. Only a small fraction of SSDI recipients are under age twenty- five. The magnitudes on this graph match those in Autor and Duggan (2006) quite closely.

The Long-Run Growth of Disability Insurance in the United States

363

Large increases in the SSDI rolls through the 1970s led to concerns about the fiscal cost of the program. Starting in 1977, there was a move toward tightening the medical screening done by the state- level boards in charge of the initial screening for SSDI. Gruber and Kubik (1997) estimate that denial rates increased from 53.8 percent to 69.3 percent over the 1977 to 1980 period. This pressure on the program continued with legislation in 1980 that made it harder to get benefits, changed the structure of benefits, and imposed more frequent medical reviews on those who were already receiving benefits. These measures were effective in reversing the upward trend in SSDI receipt. However, they spawned a political backlash resulting in the Social Security Disability Benefits Reform Act of 1984. This legislation had a major impact on the program. The eligibility criteria shifted from medical to functional, as the test for disability became the “ability to function in a work- like setting.” Assessments were also changed substantially, with pain and discomfort given greater weight, allowing for multiple minor impairments to “add up” to a disability and relaxed assessment of mental health and musculoskeletal damage. There was also a shift in the emphasis away from the Social Security medical assessment toward the person’s own health care provider. In sum, these reforms made the system more complex, more liberal in eligibility, and more subjective. On the benefit side, SSDI eligibility brought eligibility to Medicare coverage as well, starting in 1973. The last major legislative change was in 1978. In that year, the current system of calculating the AIME and putting it through the “bend point” formula was established. The replacement rates used previously were modified in this reform, and also the indexing of the “bend points” to average wages was regularized. The impact of this change has been greater than was likely anticipated because of the growth in earnings inequality over the last thirty years. Higher earnings for those at the top pulled up average earnings more quickly than the growth of earnings at lower levels of the earnings distribution. Autor and Duggan (2003) calculate that between 1979 and 1995, average real earnings for those without a high school degree fell by 19.5 percent, but the average wage calculated by the Social Security Administration went up by 21.6 percent. This effectively raised replacement rates because earnings were updated using the high average wage index, and the bend points moved up quickly, which allowed more of the AIME to fall in the 90 percent replacement rate range. For those at the 10th percentile of earnings, Autor and Duggan (2003) estimate that this difference in earnings growth had the impact of raising replacement rates from 52 percent to 74 percent. A number of papers have addressed the growth in SSDI. Most recently, Duggan and Imberman (2009) set out five potential reasons for the growth. First are the 1984 screening and eligibility changes. Second, replacement rates have increased substantially for those at lower earnings levels, as described in the preceding. Third, as more females joined the workforce, they

364

Kevin Milligan

became newly eligible for SSDI. Fourth, the overall population is aging. Fifth and finally, economic conditions may have provided incentives to shift into SSDI. In their simulations, Duggan and Imberman (2009) find that the eligibility changes, economic conditions, and increased replacement rates are most important for men. For women, economic conditions are not as important, but increased disability insurance (DI) coverage because of higher employment levels is important. 10.3

Other Research

There is an extremely large and extensive literature on the impact of SSDI on labor market decisions. Instead of an extensive review of the literature, I aim here to point out the major findings of the literature. Research in the 1970s and 1980s tended to find large disincentive effects. For example, Parsons (1980) can attribute the entire decline in the employment of older male workers to increasing DI uptake. Bound (1989), however, finds smaller difference in the work behavior of those who applied and received DI and those who applied and were rejected. His findings suggest that DI can explain less than half of the nonparticipation of older males. Gruber and Kubik (1997) find that the state- varying tightening of medical screening in the late 1970s did have a substantial effect on DI uptake, especially on those who were seemingly less disabled. Autor and Duggan (2003) exploit cross- state differences in the industrial composition of wages to investigate how benefit replacement rates influence DI applications and uptake. They find, for lower- educated men and women, that higher replacement rates can lead negative labor demand shocks to translate into larger DI case loads. Moreover, some of this comes out of what would have been observed previously as unemployment, suggesting some substitution across benefit programs for those not currently working. Autor and Duggan (2006) provide a recent survey of the literature, with projections suggesting that, without reforms, the program will continue to grow in size. 10.4

Data Sources

One of the goals of this chapter is to provide the longest- possible time series of data across measures of employment, SSDI usage, and health. This goal means some compromises between the detail available and the length of the time period that may be analyzed when choosing the appropriate data sets. There are three primary data sets used in this analysis. For both the labor market and program participation data, I employ the March Current Population Survey (CPS). This survey extends back to 1962 and has detailed information on employment and income sources. The level of detail of the data improves through time with the CPS. There are also some questions on

The Long-Run Growth of Disability Insurance in the United States

365

self- assessed health that I use. The mortality data comes from the Human Mortality Database, which combines data from the US Census and the National Center for Health Statistics.9 Finally, the General Social Survey has subjective health questions on self- assessed health and activity limitations that are useful for the analysis. I use the combined 1972 to 2008 database prepared by the National Opinion Research Center.10 For both the CPS and the General Social Survey, I use the provided survey weights to account for survey design and sampling. To these surveys, I add data taken from administrative publications of the Social Security Administration. 10.5

Results

The analysis examines the long- run trends in health, employment, and program participation. I do so by creating time series graphs of older Americans, separately by sex. The first set of results looks at mortality and at health. The second group of results studies program participation. Third, I search for links between labor market behavior and disability. Finally, the relationship between health and disability is explored. 10.5.1

Mortality and Health

Mortality is the first focus of the analysis. Mortality has the advantage of being an objective measure of health. It is available for long time periods, is comparable across countries, and comes from population data rather than a survey. Mortality is, however, a coarse measure of health. After graphically characterizing the trends in mortality in several ways, I compare the trends in mortality to those seen in more subjective and subtle measures of health. The mortality rate at a given age is determined by dividing the number of deaths by the population at that age. I do this separately by sex for each year available in the data. I use data spanning the fifty- six years from 1950 to 2006 in order to examine the long- run trends. The first analysis of mortality takes the mortality rate of sixty- and sixtyfive- year- olds in 1950. For subsequent years, I denote the age at which the 1950 level of mortality is reached. That is, how much older do you have to be to have the same level of mortality as someone in 1950? These data are graphed in figure 10.2 separately for males and females. In 1950, men aged sixty had a mortality rate of 2.52 percent. By 1970, it took to age 61.03 to hit this same 2.52 percent level of mortality—an extra year of life.11 Over the next twenty- year period, however, the age at which 1950 levels of age sixty mortality are seen is not until age 65.31. By the last 9. See http://www.mortality.org. 10. See http://www.norc.uchicago.edu/GSS+Website/. 11. I use linear interpolation to determine the age at which age sixty mortality is reached. For example, the target age sixty mortality of 2.52 percent is 3 percent of the way between ages sixty- one and sixty- two, so I calculate that the age- sixty equivalent age is 61.03.

366

Kevin Milligan

Fig. 10.2

Age at which aged sixty and sixty-five 1950 mortality level is reached

Sources: Mortality data from National Center for Health Statistics and US Census.

year in the data, 2006, this had risen to 69.48. This is a substantial change in mortality over this half- century time period. Figure 10.2 also shows the results for aged sixty women. In 1950, the mortality rate for women aged sixty was 1.48 percent. The female mortality rate dropped more quickly in the 1950s and 1960s than it did for men, meaning that by 1970 the age sixty level of mortality was not reached until age 63.35, which is an improvement more than two years greater than what was seen for men. However, the men close the gap through the 1970s and 1980s. By the 1990s, the gains in female mortality seem to level out, reaching an age of 68.1 by 2006. At age sixty- five, the pattern is very similar. The beginning mortality rate for men is 3.62 percent and for women is 2.22 percent. Men’s mortality is actually worse in 1970 than in 1950, with those at age 64.84 having the equivalent mortality as an age sixty- five man in 1950. However, after 1970 substantial and consistent gains for men take the age at which age sixty- five mortality is reached to 73.60 by 2006. This is a gain of 8.6 years, or 13.23 percent over age sixty- five. For age sixty male mortality, the gain was only slightly more at 15.84 percent. For women, there were gains in the 1950s and 1960s, but not over the next twenty- year period. By 2006, the age at which age sixty- five 1950 mortality was reached was 72.80. Figure 10.3 explores how cross- sectional mortality has changed from 1950 to 2006, for both males and females. This helps to understand if the gains in mortality were across all ages or concentrated at certain age ranges.

The Long-Run Growth of Disability Insurance in the United States

Fig. 10.3

367

Cross-sectional mortality in 1950 and 2006

Sources: Mortality data from National Center for Health Statistics and US Census.

Male mortality seems to drop fairly consistently across all age ranges. At age eighty, the drop is from 11.6 percent to 6.6 percent, or a 43.2 percent drop. At age sixty, it is a drop from 2.52 percent to 1.20 percent, or a 53.5 percent drop. Women also saw consistent gains across all ages, with an improvement of 50 percent at age eighty; from a 9.00 to a 4.50 percent mortality rate. Other ages saw similar improvements. Figures 10.4 and 10.5 show the age- specific mortality across time for ages fifty- five, sixty, and sixty- five. These graphs confirm what was seen earlier in figure 10.2. Mortality rates of men in figure 10.4 stay roughly constant until 1970, when serious improvements start at all ages. For women in figure 10.5, the improvements are sharpest in the first twenty years of the data. To see how well these mortality trends reflect more subtle measures of health, I now add some self- assessed measures of health to the analysis. From the General Social Survey, I take the proportion of respondents who answer “fair” or “poor” to the question on self- assessed health.12 This is available for the years 1972 to 2008. From the Current Population Survey, I use a similar self- assessed heath question, available only from 1995 onward.13 Finally, I form an indicator for respondents having a limit on the amount of work they can do because of a disability. This variable is also from 12. The question is the GSS is “Would you say your own health, in general, is excellent, good, fair, or poor?” 13. The question in the CPS is “Would you say . . .’ s health in general is.” Valid responses are excellent, very good, good, fair, and poor.

Fig. 10.4

Mortality at specific ages for men

Sources: Mortality data from National Center for Health Statistics and US Census.

Fig. 10.5

Mortality at specific ages for women

Sources: Mortality data from National Center for Health Statistics and US Census.

The Long-Run Growth of Disability Insurance in the United States

Fig. 10.6

369

Health and mortality for men aged sixty to sixty-four

Sources: Mortality data from National Center for Health Statistics and US Census. Health data from General Social Survey and Current Population Survey.

the CPS and seems to be broader in definition than what is used by SSDI for determining eligibility.14 It is available from 1989 onward. Figure 10.6 graphs these three subjective health measures on the left axis and mortality on the right axis for men aged sixty to sixty- four. Mortality declines almost by half over the time period 1965 to 2006. The General Social Survey (GSS) measure of self- assessed health is quite noisy from year to year, most likely because of sampling variation related to the smaller sample size. There does appear to be a downward trend in this variable through time, however, with the 1970s, on average, showing worse health. The two CPS measures do not display any noticeable trend, but they are available only for shorter time windows. For women, the same data are graphed in figure 10.7. Over this time period, the improvement in mortality for women is not as large, as the period of greatest mortality improvement for women was 1950 to 1970, which is mostly before the time window covered in this graph. There is a more clear decline in the GSS measure of fair or poor health than there was for men. In the 1970s, this measure showed levels around 0.5, but, by the 1990s, this had declined to around 0.3. The CPS measure of fair or poor health shows 14. The question in the CPS is “Does . . . have a health problem or a disability which prevents work or which limits the kind or amount of work.”

370

Kevin Milligan

Fig. 10.7

Health and mortality for women aged sixty to sixty-four

Sources: Mortality data from National Center for Health Statistics and US Census Bureau. Health data from General Social Survey and Current Population Survey.

declines over the time period it is available, while the measure of health limitations goes up until 1994 before declining. Another way to look at the relationship between self- assessed health and mortality is to look at how the age patterns have changed through time. In figure 10.8, I show mortality and the proportion with fair or poor health for men in the 1970s and the 2000s. The health data here come from the National Health Interview Survey. At age sixty in the 1970s, mortality is 0.022 and the incidence of fair or poor health is 0.215. By the 2000s, however, it took until age sixty- seven to reach that same level of mortality and until age sixty- eight to reach that level of fair or poor health. In this way, it seems clear that there have been similar improvements in both measures of health through time. For females in figure 10.9, the gain in mortality through time isn’t as strong as it is for the incidence of fair or poor health. However, both show strong improvements over this thirty- year period. In summary, the data clearly show large improvements in mortality rates both for men and women since 1950. A correspondence of mortality with subjective measures of health is present, as both are seen to be improving through time. 10.5.2

Programs and Participation

I now turn to analysis of program participation using data from two sources. The primary source is the CPS, which started to report the receipt of different types of benefit income starting in 1970. Five variables are formed

The Long-Run Growth of Disability Insurance in the United States

Fig. 10.8

371

Health and mortality for men by age, 1970s and 2000s

Sources: Mortality data from National Center for Health Statistics and US Census Bureau. Health data from National Health Interview Survey.

Fig. 10.9

Health and mortality for women by age, 1970s and 2000s

Sources: Mortality data from National Center for Health Statistics and US Census Bureau. Health data from National Health Interview Survey.

from this source, including dummies indicating receipt of Social Security, Unemployment Insurance, public assistance or welfare, veterans’ benefits, and workers’ compensation. For the Social Security dummy, the data do not distinguish the type of benefit (i.e., Old Age, Survivor, Disability, etc.). The second source used here is the administrative data from the Social Security Administration (SSA) containing counts of all SSDI beneficiaries by age

372

Kevin Milligan

group and year. For this, I use receipt of the Disabled Workers category of benefit, rather than children and widow benefits. These counts are normalized by corresponding age- sex- year population estimates from the United States Census Bureau. Figure 10.10 shows the receipt of these six measures of public benefits by age group for males in 1970. Until age sixty, Veterans’ Benefits constitute the largest proportion take-up of any of the benefits. Unemployment Insurance take-up is fairly constant across age groups until age sixty- five at around 4 percent. Social Security benefits (from the CPS) and SSDI take-up (from the SSA) are very close to each other and rising across age groups. At age sixty to sixty- four, Social Security participation jumps higher as benefits can be taken at the Early Retirement Age of sixty- two. The women are graphed in figure 10.11. There are several striking differences. First, Veterans’ Benefits play a much smaller role for women, owing to the sex difference in rates of military service. Social Security benefit receipt (from the CPS) is distinctly higher than SSDI receipt, reflecting the fact that many women receive survivor benefits. The level of SSDI receipt is much lower, and the slope less steep, than for men. This in part is driven by the lower labor force participation of women in 1970—many do not qualify for SSDI because of the work requirements. For similar reasons, the proportion of women on Unemployment Insurance is noticeably lower than for men in figure 10.10. The next two figures turn the clock forward to 2009 and repeat the same analysis. Figure 10.12 shows the men. Veterans’ Benefits now play a much

Fig. 10.10

Program participation for men, 1970

Sources: Current Population Survey and Social Security Administration.

Fig. 10.11

Program participation for women, 1970

Sources: Current Population Survey and Social Security Administration.

Fig. 10.12

Program participation for men, 2009

Sources: Current Population Survey and Social Security Administration.

374

Kevin Milligan

Fig. 10.13

Program participation for women, 2009

Sources: Current Population Survey and Social Security Administration.

smaller role than in 1970, reflecting the cohort pattern of military participation. The SSDI participation rates are double what was seen in 1970. Unemployment Insurance now decreases across age groups, possibly reflecting the kind of substitution found in Autor and Duggan (2003). For the women in figure 10.13, SSDI rates are much higher than in 1970 because not only are more working women claiming SSDI, but there are many more working women. Unemployment Insurance receipt declines after age fifty- four, but other benefits show little participation. I now isolate just recipiency of SSDI. In figure 10.14 I graph the SSDI recipiency rate for three different age groups for men. The broad trends of receipt are the same across the three age groups, although at different levels. There is a peak at 1980, reflecting the tightening of medical screening starting in the late 1970s and the legislated changes in 1980. The loosening of rules in 1984 doesn’t appear to have an immediate impact here, with SSDI rates for sixty to sixty- four- year- olds not reaching their 1980 height again until the second half of the 2000s. Growth in the rolls at younger ages is evident, though, as the 1980 peak is surpassed in the early 1990s at ages forty to forty- four and fifty to fifty- four. This suggests that much of the growth for males in the 1980s and 1990s came from younger ages. The women are graphed in figure 10.15. The rate of SSDI recipiency peaks at 1980, just like for the men. However, the growth rate takes off after 1990—especially for those aged sixty to sixty- four. By 2009, recipiency rates at all ages for women were more than double what was seen in 1980. The analysis of program participation has revealed several interesting findings. First, SSDI has moved from one program among many, to being

Fig. 10.14

SSDI receipt for men

Sources: Social Security Administration and US Census Bureau population estimates.

Fig. 10.15

SSDI receipt for women

Sources: Social Security Administration and US Census Bureau population estimates.

376

Kevin Milligan

the dominant program supporting nonworking older people in the United States. The growth in SSDI receipt has been accompanied by dips in Veterans’ Benefits and a possible offset with Unemployment Insurance receipt at ages sixty to sixty- four. The largest growth is found for women and for men at middle ages. 10.5.3

Labor Markets and Disability

It is clear from the preceding analysis that labor market trends are very important for understanding what is going on with SSDI receipt. For example, the exceptional growth of receipt for women likely reflects the increasing workforce attachment of the cohorts of women arriving at older ages in the 1980s, 1990s, and 2000s. In this set of results, I graph employment, unemployment, and labor force nonattachment by sex and age groups. I also decompose the reason not working into retirement and disability components. All of this analysis is performed with the CPS. The employment rates for the age groups forty to forty- four, fifty to fiftyfour, and sixty to sixty- four are shown in figure 10.16 for men. The first two age groups show a very slight downward trend, but the level of employment remains very high. Business cycle effects do not appear terribly large at these ages for men. For ages sixty to sixty- four, there is a steady drop in the employment rate from the 1960s to the mid- 1990s. The total drop is 39 percent from the 1962 starting point for age sixty to sixty- four men. After 1995, the employment rate for men in this age group recovers slightly. Schirle

Fig. 10.16

Employment rates by age for men

Source: Current Population Survey.

The Long-Run Growth of Disability Insurance in the United States

Fig. 10.17

377

Employment rates by age for women

Source: Current Population Survey.

(2008) presents evidence that this in part reflects the possibility of joint retirement for men in this cohort, who are married to women from cohorts with much stronger lifetime workforce attachment. These increasing female cohort employment patterns can be seen in figure 10.17, which shows the employment rates for women for the same three age groups. The breakout in employment at age forty to forty- four in the 1970s and 1980s leads to higher employment in the 1980s and 1990s for ages fifty to fifty- four, and then the 1990s and 2000s for ages sixty to sixty- four. This higher rate of employment not only helps to explain the pattern of male employment at ages sixty to sixty- four, but also has very important implications for SSDI. The rate of sixty to sixty- four- year- old female employment is 36.7 percent higher in 2009 than it was in 1989. This suggests that many more women are likely eligible to claim SSDI than was the case in the 1970s and 1980s. Indeed, Duggan and Imberman (2009) find that increasing eligibility can explain around one quarter of total SSDI growth for women. Figures 10.18 and 10.19 examine unemployment rates at different ages. Here, business cycle effects can be seen clearly, with peaks in the early 1980s, 1990s, 2000s, and in 2009. The level of unemployment for women is lower, but the age dispersion of unemployment rates appears larger than for men. This may stem from the large cohort differences in labor market attachment for women. There are strong sex differences for being out of the labor force. The rate of nonparticipation in the labor market is graphed in figure 10.20 for men

Fig. 10.18

Unemployment rates by age for men

Source: Current Population Survey.

Fig. 10.19

Unemployment rates by age for women

Source: Current Population Survey.

The Long-Run Growth of Disability Insurance in the United States

Fig. 10.20

379

Not in the labor force by age for men

Source: Current Population Survey.

and figure 10.21 for women. In both cases, however, these trends provide near mirror image graphs for the trends discussed previously for employment. More informative, however, are the reasons why men and women are not working. The CPS asks those who haven’t worked at all in the past year the main reason why that person did not work. Two of the coded responses are “ill or disabled” and “retired.”15 Figure 10.22 graphs these data for men from 1968 to 2009. For men, the proportion of men not working at ages sixty to sixty- four more than doubles from the late 1960s to the mid- 1990s. Most of this growth, however, comes from retirement. The proportion of men not working because of illness or disability is relatively constant. In fact, from the late 1970s onward, it declines slightly. There is no sign of the increase seen in SSDI recipiency starting in the late 1980s and continuing on forward. Here, not working because of disability is flat through time. The responses of women are graphed in figure 10.23. The nature of the graph is very different for women because the category “taking care of home or family” is the most important until the 1990s for these age sixty to sixtyfour- year- old women. By the 2000s, like men, most women not working are doing so because either of retirement or disability. Most curious is the relatively constant rate of not working because of disability. In figure 10.15, 15. The other valid responses in 2009 are “taking care of home or family,” “going to school,” and “could not find work.” There have been slight changes to the question since.

Fig. 10.21

Not in the labor force by age for women

Source: Current Population Survey.

Fig. 10.22

Reasons not working, aged sixty to sixty-four men

Source: Current Population Survey.

The Long-Run Growth of Disability Insurance in the United States

Fig. 10.23

381

Reasons not working, aged sixty to sixty-four women

Source: Current Population Survey.

SSDI recipiency for women more than doubled from 1985 to 2009. Here, there is no sign of an increase. For another view on employment rates, I look in figure 10.24 at the relationship between employment and mortality. Each point on the graph is a plot of employment versus mortality for a given age. I use ages fifty to seventy- five. Taking the years 1996 and 2006, I compare how much employment is observed at different levels of mortality. In 1966, it took a mortality rate of 0.04 to reduce employment to around 50 percent of males. In 2006, the same 50 percent employment rate was reached at a much lower mortality rate—around 0.018. This suggests that, for the same level of health (as measured by mortality), there was much more work in 1966 than in 2006. Figure 10.25 repeats this analysis for women. The graph here is more difficult to interpret because of the great increase in employment rates among females over the decades. Still, there are signs of a small change in employment for a given level of mortality in the middle of the chart. The analysis of employment behavior uncovers two important findings. First, the emergence of widescale female labor force participation is a dominant factor in understanding the trends in SSDI over this past thirty years. Second, despite large increases in SSDI recipiency, the proportion of both men and women reporting that their absence from the labor market is caused by illness or disability is relatively constant. This could suggest that the upswing in SSDI recipiency is due to SSDI taking an increasing share of

Fig. 10.24

Employment and mortality, men

Sources: Current Population Survey for employment and mortality data from National Center for Health Statistics and US Census Bureau.

Fig. 10.25

Employment and mortality, women

Sources: Current Population Survey for employment and mortality data from National Center for Health Statistics and US Census Bureau.

The Long-Run Growth of Disability Insurance in the United States

383

those who were not working because of disability, but who previously had no resort to SSDI because of the type of illness or other eligibility factors. 10.5.4

Disability and Health

The final set of results looks more deeply at the relationship between SSDI, other measures of disability, and health. This continues the investigation of the finding immediately in the preceding that the proportion out of the labor force for disability does not seem to line up well with trends in SSDI uptake. In the following, I graph two measures of disability against mortality rates and self- assessed health measures. The goal here is to ascertain what relationship measured disability has with measures of health. Figure 10.26 looks at mortality, not working because of a disability, and SSDI receipt for men aged forty- five to forty- nine. All of these variables have been used previously in this chapter, but putting them together in one place brings new insight. Specifically, the strong improvements in mortality over the period from 1972 to 2009 move in a starkly opposite direction to the two disability measures. To the extent that mortality captures important aspects of health, this suggests that heath has been getting better, while disability measures show increasing disability. For the two disability measures, they both trend up sharply over this time period. Social Security Disability

Fig. 10.26 Disability, SSDI receipt, and mortality for aged forty-five to forty-nine men Sources: Not working because of a disability is from the Current Population Survey. SSDI receipt is from the Social Security Administration and US Census Bureau. Mortality data from National Center for Health Statistics and US Census Bureau.

384

Kevin Milligan

Insurance receipt is below the proportion this is not working because of a disability. This may be because of eligibility issues—not everyone who can’t work is able to draw on SSDI if, for example, they haven’t worked sufficiently to gain the number of credits needed for SSDI benefits. The other possibility is that they are simply measuring different things. For women aged forty- five to forty- nine, figure 10.27 shows a very similar pattern. The largest difference between women and men is the magnitude of the gap between those not working because of a disability and SSDI receipt. This larger gap is likely rooted in the larger proportion of women who have not earned sufficient credits to qualify for SSDI. Of course, the overall rates here are quite low as most forty- five to forty- nine- year- olds are relatively healthy. I now repeat the analysis, using men aged sixty to sixty- four. Figure 10.28 reveals the same steady downward trend in mortality seen at younger ages. For disability, SSDI receipt and not working because of a disability are quite close to each other until 2000. This might mean that most men of this age are eligible so that SSDI captures almost all of the men who are disabled. There is a divergence after 2000, with SSDI receipt taking off higher than being out of work with a disability. This is hard to reconcile with eligibility because more seem to be receiving SSDI than are reporting not working because of a disability.

Fig. 10.27 Disability, SSDI receipt, and mortality for aged forty-five to forty-nine women Sources: Not working because of a disability is from the Current Population Survey. SSDI receipt is from the Social Security Administration and US Census Bureau. Mortality data from National Center for Health Statistics and US Census Bureau.

The Long-Run Growth of Disability Insurance in the United States

Fig. 10.28

385

Disability, SSDI receipt, and mortality for aged sixty to sixty-four men

Sources: Not working because of a disability is from the Current Population Survey. SSDI receipt is from the Social Security Administration and US Census Bureau. Mortality data from National Center for Health Statistics and US Census Bureau.

For women aged sixty to sixty- four, the gap between the measures of disability in figure 10.29 closes over the fifteen years from 1990 to 2005. Again, this may reflect increased eligibility or different underlying phenomena for the two measures. However, neither measure of disability shows any relationship with the mortality rate. As discussed earlier, the mortality rate is an abrupt measure of health. With this in mind, I repeat the preceding analysis but include the selfassessed fair or poor health measure taken from the GSS, as used earlier. Figures 10.30 and 10.31 show men and women at ages forty- five to fortynine. In both cases, there is a gentle improvement in health as measured by those suffering from fair or poor health over this time period. Both measures of disability move up, which is at odds with the self- assessed health trend. The final two figures have men and women ages sixty to sixty- four and compare fair or poor health to measures of disability. Figure 10.32 shows a slight downward trend for fair or poor health for men, but figure 10.33 has a clear downward trend for the health of women in this age range. The disability measures show no obvious correspondence with the health measure. This section of results investigated the relationship between health and disability measures. The data revealed starkly opposite trends both for mortality and self- assessed health when compared to disability. For the two disability measures, the proportion reporting not working because of dis-

Fig. 10.29 Disability, SSDI receipt, and mortality for aged sixty to sixty-four women Sources: Not working because of a disability is from the Current Population Survey. SSDI receipt is from the Social Security Administration and US Census Bureau. Mortality data from National Center for Health Statistics and US Census Bureau.

Fig. 10.30 Disability, SSDI receipt, and fair or poor health for aged forty-five to forty-nine men Sources: Not working because of a disability is from the Current Population Survey. SSDI receipt is from the Social Security Administration and US Census Bureau. Health data from General Social Survey.

Fig. 10.31 Disability, SSDI receipt, and fair or poor health for aged forty-five to forty-nine women Sources: Not working because of a disability is from the Current Population Survey. SSDI receipt is from the Social Security Administration and US Census Bureau. Health data from General Social Survey.

Fig. 10.32 Disability, SSDI receipt, and fair or poor health for aged sixty to sixty-four men Sources: Not working because of a disability is from the Current Population Survey. SSDI receipt is from the Social Security Administration and US Census Bureau. Health data from General Social Survey.

388

Kevin Milligan

Fig. 10.33 Disability, SSDI receipt, and fair or poor health for aged sixty to sixty-four women Sources: Not working because of a disability is from the Current Population Survey. SSDI receipt is from the Social Security Administration and US Census Bureau. Health data from General Social Survey.

ability was often higher than the proportion receiving SSDI benefits. This gap may indicate the extent of ineligibility among the disabled, or it may indicate that movements in SSDI benefit recipiency are not strongly related to work- limiting disability. 10.6

Conclusion

This chapter has examined the long- run trends in health, employment, and disability among older Americans. By taking a very long time horizon for the analysis, several interesting findings emerge. First and foremost, the long- run trends in takeup of SSDI seem to bear little obvious relationship to trends in health, as measured by mortality and subject self- assessed health. Instead, the movements in SSDI recipiency seem most related to changes in the legislative and administrative framework as well as the increasing employment (and, thus, potential eligibility) of women. Also, the correspondence of SSDI recipiency with being out of the labor market because of illness or disability wasn’t as strong as one might expect— at ages sixty to sixty- four, SSDI recipiency has increased substantially without a similar increase in the proportion not working because of disability. This may be due to changes in eligibility, meaning that SSDI captures an increasing fraction of those not working because of disability. On the other

The Long-Run Growth of Disability Insurance in the United States

389

hand, it may indicate that the increases in SSDI are not driven by those not working because of a disability. Autor and Duggan (2006) and Duggan and Imberman (2009) have forecasted serious trouble for the SSDI program on the horizon. The longrun analysis in this chapter, along with the experience of other countries reported in other parts of this project, should provide some perspective as Americans grapple with the challenges with SSDI.

References Autor, David H., and Mark G. Duggan. 2003. “The Rise in the Disability Rolls and the Decline in Unemployment.” Quarterly Journal of Economics 118 (1): 157– 205. ———. 2006. “The Growth in the Social Security Disability Rolls: A Fiscal Crisis Unfolding.” Journal of Economic Perspectives 20 (3): 71– 96. Berkowitz, Edward D. 1989. Disabled Policy: America’s Programs for the Handicapped. Cambridge: Cambridge University Press. ———. 2000. “Statement before the Subcommittee on Social Security of the Committee on Ways and Means.” June 13. https://socialsecurity.gov/history/edberkdib .html. Bound, John. 1989. “The Health and Earnings of Rejected Disability Insurance Applicants.” American Economic Review 79 (3): 482– 503. Duggan, Mark, and Scott A. Imberman. 2009. “Why Are the Disability Rolls Skyrocketing? The Contribution of Population Characteristics, Economic Conditions, and Program Generosity.” In Health at Older Ages: The Causes and Consequences of Declining Disability, edited by David M. Cutler and David A. Wise, 337– 79. Chicago: University of Chicago Press. Gruber, Jonathan, and Jeffery D. Kubik. 1997. “Disability Insurance Rejection Rates and the Labor Supply of Older Workers.” Journal of Public Economics 64 (1): 1– 23. Parsons, Donald O. 1980. “The Decline of Male Labor Force Participation.” Journal of Political Economy 88 (1): 117– 34. Schirle, Tammy. 2008. “Why Have the Labor Force Participation Rates of Older Men Increased Since the mid- 1990s?” Journal of Labor Economics 26 (4): 549– 94. Social Security Administration. 1986. “A History of the Social Security Disability Programs.” http://segurosocial.gov/history/1986dibhistory.html. ———. Various years. Annual Statistical Supplement to the Social Security Bulletin. Woodlawn, MD: US Social Security Administration.

11 Disability Pension Program and Labor Force Participation in Japan An Historical Perspective Takashi Oshio and Satoshi Shimizutani

11.1

Introduction

In Japan and other developed countries, disability pension programs commonly have a long history in public pension programs. The Japanese disability pension program, which started in 1944, has gradually expanded over several major revisions and has contributed to improving the living standards of the disabled. Indeed, the eligibility criteria have been generally eased, and the number of program recipients has shown a trend of modest increase. In contrast to some European countries (Börsch-Supan 2005), however, the disability pension program is rarely related with labor force participation (of the elderly) in Japan. However, the fact that the disability pension program has so far been unlikely to be detrimental to labor force participation in Japan does not imply that the disability pension will not be relevant for retirement decisions in Japan. Indeed, the number of recipients of disability pension benefits has increased in recent years. It has been established that the aging population in Japan is growing at a remarkable speed, and the proportion of the elderly aged sixty- five and over in the total population had reached 20 percent in 2005 and will reach 30 percent in 2025 and approximately 40 percent in Takashi Oshio is professor at the Institute for Economic Research, Hitotsubashi University. Satoshi Shimizutani is a senior research fellow at the Institute for International Policy Studies. We are grateful for the useful comments from participants at the conference on the International Social Security Project (Phase VI) organized by the NBER in Lisbon, Portugal, on May 28– 29, 2010, and financial support via a Grant- in-Aid for Specially Promoted Research from Japan’s Ministry of Education, Culture, Sports, Science, and Technology (grant number 22000001). For acknowledgments, sources of research support, and disclosure of the authors’ material financial relationships, if any, please see http://www.nber.org/chapters/c12392.ack.

391

392

Takashi Oshio and Satoshi Shimizutani

2050 (National Institute of Population and Social Security Research 2006). In the future, together with the historically lower fertility rate, population aging will cause a decline in the numbers of the labor force, and the situation might be exacerbated by expanding disability program participation. From this aspect, it is worthwhile to investigate the reason for the low take-up rate in Japan, focusing on a distinction between institutional aspects (i.e., tight eligibility conditions) and noninstitutional ones (i.e., better health conditions), which have not been sufficiently explored in Japan. To this end, this chapter utilizes historical information to explore the relationship between the disability pension program and labor force participation in Japan. In particular, we examine the time series data to identify what has determined the trend in disability program participation. Specifically, this chapter performs twofold analyses to separate institutional factors from noninstitutional factors accounting for program participation. First, we provide historical information on mortality and health status measures in Japan. A change in health status is sure to affect participation in disability insurance programs even without any revisions. Hence, understanding the trends in health status over time is critical to distill the effect of health conditions on the program participation. This task is not easy, however, because there is no “true” measure of health status. Thus, we must explore, over time, the relationships among representative measures of health, mortality rates, and subjective health status. Second, we attempt to understand the relationship between changes over time in the disability program and program participation. We identify three major revisions of the program in the postwar period and examine the relationship between these changes and the program participation rate so that we identify how trends in the disability program have been related with institutional revisions. Our discussion proceeds as follows. Section 11.2 provides a historical overview of the disability program in Japan, focusing on major revisions that are to be examined in later sections. Section 11.3 investigates historical representative data on health status, that is, mortality, subjective health status, and other measures and relates them with each other in Japan. Section 11.4 describes the development of labor force participation, examines the relationship with program revisions, and quantifies the effect of the disability program revisions on “activity” measured by labor force participation rate. The last section concludes. 11.2 11.2.1

Disability Program and Other Social Security Program Reforms Historical Overview

This section gives an overview from an historical perspective of the disability program and other related reforms to the social security program

Disability Pension Program and Labor Force Participation in Japan

393

in Japan. At the outset, we need to clarify that what is often referred to as the “disability insurance (DI) program” in other countries corresponds to the “disability pension program” in Japan; the program was constructed in the public pension scheme, and all revisions to the disability program have been linked to those to the core pension programs. Among several programs to assist the disabled, the disability pension plays the most important role in terms of income compensation; therefore, we will focus on the description of the disability pension program and briefly mention related programs at the end of this section. The Japanese public pension program consists of three programs: the Employees’ Pension Insurance (EPI, Kosei Nenkin), whose pensioners are private employees; the National Pension Insurance (NPI, Kokumin Nenkin), whose pensioners are self- employed or agriculture, forestry, and fishery cooperative employees; and the Mutual Aid Insurance (Kyosai Nenkin) covering employees in the public sector and private schools. In terms of the number of pensioners, the EPI and the NPI contribute to the total by slightly less than half, respectively, and the MAI occupies the remaining small portion.1 We will describe in the following the revisions of the disability pension program over time, focusing on the EPI and NPI (see table 11.1). When the EPI was launched in 1944 as the first social insurance style public pension program, it contained the disability pension program.2 The initial program rated the disabled using two grades (Grade 1 and Grade 2) once qualified. The grading depended on functional ability to perform activities of daily living, rather than on loss of earning ability. Grade 1 referred to a condition in which a person was unable to perform activities of daily living (e.g., severe disability affecting both hands or complete blindness). Grade 2 referred to a condition in which a person faced very severe limitations in performing activities of daily living (any severe disability affecting either hand). We need to pay attention to the fact that the program insured persons with mental disorders from the beginning, via the EPI. The revision of 1954 introduced Grade 3 to cover more disabled persons with less- severe conditions than those in Grade 2.3 After establishment in the EPI, the disability pension program has expanded in some ways. To date, there have been four major revisions during 1. See Oshio, Shimizutani, and Oishi (2010) for a detailed description of the Japanese public pension program and its historical development. 2. A brief review of development of the disability pension program was provided by the Ministry of Health, Labour and Welfare (2009). The previous program of the EPI, which was called the “Workers Pension Insurance (Roudou Sha Nenkin Hoken Seido)” was launched in 1941 and covered only male and nonoffice workers. In 1944, the name of the program was changed to the EPI, and it began to cover office workers or females; this was very similar to the current system in terms of coverage. 3. The EPI had only a single layer of a wage proportional benefit before 1954 in the old- age pension program and was reconstructed to a double- tier structure (fixed rate part as the first tier and a wage proportional part as the second tier) in 1954. Even after 1954, however, the disability pension program had a single- tier structure until the 1985 revision.

394 Table 11.1

Takashi Oshio and Satoshi Shimizutani Development of disability pension program in Japan National Pension Insurance (self-employed, agricultural, forestry, and fishery sector)

Disability Pension (with contribution)

Disability Welfare Pension (without contribution)

1944 1954 1959 1964–1965 1974 1986–

Employees’ Pension Insurance (private firm employees) Grade 1 and Grade 2 (including mental diseases) Grade 3 was added

Grade 1 and Grade 2 Grade 1 Covered mental diseases Grade 2 was added Merged to Disability Basic Pension

Disability Basic Pension + wage proportional benefit

its development. We will consider them, focusing on who has been most affected in each revision. The first revision was the introduction of the NPI effective in 1961. The NPI began to insure those who were not covered by the EPI, that is, self- employed workers and agricultural, forestry, and fisherysector workers. The introduction of the NPI was important, as it launched the universal pension system in the Japanese public pension program. The NPI drastically expanded the coverage of the disability pension program to more groups than just employees in the private sector. Unlike the EPI, the NPI had not covered mental disease at the time of its introduction. The NPI had two types of disability programs for recipients with premium contribution and for those without. The first was the “Disability Pension Program (Shogai Nenkin),” which was designed for those who contributed the premium. The NPI was motivated by the spirit of social insurance and, thus, required the recipients to contribute the insurance premium to receive benefits. The second was the “Disability Welfare Pension Program (Shogai Fukushi Nenkin),” which was designed for those who did not make premium contribution. Eligibility to receive disability pension benefit was judged at the time of the first doctor’s visit to survey the extent of the disease that made the person disabled. Thus, those who had received the first visit before reaching the age of twenty (the minimum age for NPI participation is twenty) or before 1961 were not insured by the disability pension program under the NPI. They were covered by the disability welfare pension program, which was financed by the government. The eligibility for this program was meanstested, and the amount of benefit was lower than that of the disability pension program. The second revision took place in 1974. It called for expanding the coverage for mental disease. The NPI began to insure mental disorders in 1964 and

Disability Pension Program and Labor Force Participation in Japan

395

mental deficiency in 1965.4 However, the coverage for mental disability was very limited. While those who paid the premium were eligible to receive the disability pension benefits once qualified as Grade 1 or 2 (note that there was no Grade 3 in the NPI), the disability welfare program insured the disabled only if rated as Grade 1. In 1974, the disability welfare program began to cover Grade 2 as well, and many patients with mental disorders or deficiencies became eligible to receive the benefit. The third revision was implemented as a part of the major revision of core public pension programs in 1985 (effective from 1986). This revision was most drastic in recent years, as it harmonized all the public pension programs into an integrated form (see Oshio, Oishi, and Shimizutani 2011). It reduced the benefit multiplier and flat- rate benefit in the old- age pension program for the first time, aiming to restrain an increase in total pension benefits. Three revisions were implemented with respect to the disability pension programs. First, a double- tier structure was introduced. The NPI pensioners, both with and without premium contributions, were entitled to receive (a) the flat rate “Disability Basic Pension (Shogai Kiso Nenkin)” benefit as the first tier, which was linked to the Basic Pension Benefit, and (b) the wage proportional “Disability Employees’ Pension (Shogai Kosei Nenkin)” program as the second tier. The Disability Welfare Pension, which was funded by the government before the revision, was replaced by the Disability Basic Pension, which was funded by the government and premium contributions of the NPI pensioners. Second, the benefits for the disabled without premium contribution were raised to be at the same level as those for the disabled with premium contribution in the NPI. Both groups of the disabled were entitled to receive the same Disability Basic Pension benefit, and the amount doubled for the recipients of the disability welfare pension benefit. This is a remarkable revision for those who received the disability welfare pension, given that the 1985 revision reduced old- age pension benefits in general. Third, grading of disability conditions was harmonized across programs. Before the revision, there was disparity in qualification criteria for the disabled even if the disability condition was the same. However, even after the harmonization of the grading, the Disability Basic Pension covered the disabled only in Grades 1 and 2. The EPI program covers the disabled in Grade 3, too, and provides “disability compensation” for a disabled pensioner with a disability less severe than Grade 3 if the disabled condition is fixed.5 4. In 1966, the NPI began to cover all diseases including liver and kidney diseases among other internal disorders. 5. Because the EPI pensioners were required to join the NPI in the 1985 reform, the entitlement to receive disability pension became contingent on the grading of the NPI (Disability Basic Pension), even if a disabled person had been approved to receive disability pension benefits in the EPI or MAI program. The MAI program has a Grade 3 as well.

396

Takashi Oshio and Satoshi Shimizutani

Last, the government allowed the Disability Basic Pension recipients aged sixty- five years or above to additionally receive EPI benefits if they had made any EPI contributions in the past. This revision became effective as of 2006, most probably providing the elderly with incentives to apply for disability pension benefits. In sum, the disability pension programs in Japan have a long history starting in 1944. The disability pension program for private firm employees (EPI pensioners) was introduced relatively early and was generous in that it covered mental diseases as well as patients who were less severely disabled. The coverage of the disability pension has expanded. From 1961, NPI pensioners were entitled to receive the disability pension benefits, a move that included the self- employed. From 1974, the disabled without premium contribution were entitled to receive the benefits if rated as Grade 2, thus including many persons who were mentally deficient or had mental disorders. From 1986, the disabled without contribution were entitled to receive the disability basic pension benefit, which was same as for those with contribution. Finally, the Disability Pension Benefit recipients with any EPI contributions were entitled to additionally receive EPI benefits in 2006. Despite the domestic expansion, however, the size of the disability pension benefit is still relatively small in terms of economic size and public expenditure from an international perspective. According to Organization for Economic Cooperation and Development’s (OECD’s) Social Expenditure Database, the share of the expenditure on disability pension benefits out of gross domestic product (GDP) was 0.3 percent in Japan in 2005, much lower than that in Denmark, Sweden, and the United Kingdom, which register about 2 percent. Further, the share of the expenditure on disability pension benefits out of total public expenditure is about 2 percent, which is again much lower than that in other countries. 11.2.2

Current Scheme

Under the current scheme, a person who visited a doctor for the first time for consultation about the cause of the disability when he or she was under the age of twenty or when he or she was an NPI pensioner is entitled to receive the Disability Basic Pension benefit. Note that there is no limitation in terms of age for receiving disability pension benefits, unlike in some other countries where the recipients of disability pension benefits are converted to eligibility to receive old- age pension benefits after attaining the eligible age (i.e., age sixty- five). The formula to compute the benefit is as follows. Grade 1 = Basic Pension benefit × 1.25 + additional benefit for dependent children. Grade 2 = Basic Pension benefit + additional benefit for dependent children. The amount of the Basic Pension benefit is 792,100 yen per year, and the additional child benefit is 227,000 yen each for the first and second children and 75,900 yen each for the third and subsequent children.

Disability Pension Program and Labor Force Participation in Japan

397

In addition to the Disability Basic Pension, a person who consulted a doctor to identify the cause of the disability when he or she was an EPI pensioner is entitled to receive wage- proportional Disability Employees’ Pension benefit or Disability Mutual Aid Pension benefit (for the MAI recipients). The formula to compute the benefit of the second tier is as follows. Grade 1 = Wage proportional benefit × 1.25 + additional benefit for a spouse. Grade 2 = Wage proportional benefit + additional benefit for a spouse. Grade 3 = Max [wage proportional benefit, 594,000 yen]. The amount of additional benefit for a spouse is 227,900 yen per year.6 Figure 11.1 reports the number of the recipients who received the disability pension benefits between 1970 and 2006.7 The data source is the Annual Report of Social Security Administration (Shakai Hoken Jigyo Nenpo) published by the Social Security Agency. Unfortunately, there is no data available by gender or age. The number of recipients was about 0.5 million in 1970 and increased to 2.0 million in 2006; it expanded four times over thirty- six years. As seen from the figure, the dominant recipients are the NPI pensioners, who share about 80 percent of the total. Because most of them are self- employed, their labor supply is less likely to be associated with the generosity of the disability pension program and its institutional changes. In contrast, the EPI recipients have occupied less than 20 percent of the total. As discussed in the following, their labor supply is likely to be affected by institutional changes in eligibility of the disability pensions, but their proportion is relatively small. Finally, the number of MAI pensioners to receive the disability pension, who are also likely to be affected by institutional factors, has been very small, 2 to 3 percent in all years.8 The impact of past revisions to disability pension programs on DI participation are illustrated more clearly in figure 11.2, which shows the growth rate of disability pension recipients. We observe three jumps: in 1974 to 1975, 1985 to 1986, and 2005 to 2006. As described in section 11.2.1, the 1974 revision added a Grade 2 level for the NPI Disability Pension program. The 1985 revision raised the benefit for the NPI pensioners without premium contributions (i.e., Disability Welfare Pension recipients) to the level for 6. Momose (2008) used the purchasing power parity to compare the amount of benefits in Japan with that in the United States and Sweden. While the amount of benefits of the disability employee pension (Grade 1 or 2) is larger than that in the United States and Sweden, that of the disability basic pension (Grade 1) is much smaller and that for Grade 2 is half of the standard benefit in the United States and Sweden. 7. The fiscal year starts in April and ends in March in Japan. The figures are measured as of the end of the fiscal year. 8. The number of MAI pensioners to receive the disability pension is not available; that of MAI pensioners eligible to receive the benefits is available in the Annual Report on Social Security Statistics (Shakai Hoken Tokei Nenpo) compiled by the National Institute of Population and Social Security Research. We compute that the number of MAI pensioners to receive the disability pension, assuming the proportion of those to receive out of those eligible, both of which are available in the Annual Report of Social Security Administration, is the same for the EPI and the MAI programs.

Fig. 11.1

The number of recipients of disability pension benefits

Fig. 11.2

Growth rate of disability pension recipients

Disability Pension Program and Labor Force Participation in Japan

399

those with contribution. Finally, the 2006 revision allowed Disability Basic Pension recipients to receive the EPI benefit as well if they had made EPI contributions. These jumps, albeit with limited impact on the total labor force, confirm that the DI participation is affected more directly by institutional changes than they are by changes in health status. 11.2.3

Other Programs for the Disabled

In addition to the disability pension, there are some other programs to assist the disabled. One is employers’ compensation for employees who were injured, diseased, disabled, or killed during work- related activities, including a disaster while commuting. Even a firm employing one employee is required to join the insurance scheme by law at the firm’s cost. The benefits include compensation for the treatment, labor in absence, and a disabled status after a treatment as well as benefits for family members if the employee is killed. While the employers’ compensation covers the mentally disabled, the eligibility is very limited, and the number of those approved for the mentally disabled benefit, which is relevant to the rapid increase in the number of the recipients of the disability insurance in Europe, has been small.9 Another program is public assistance. While eligibility does not require a disabling condition to receive public assistance, the proportion of those who receive public assistance is larger for the recipients of the disability pension, suggesting that the disability pension benefits are not sufficient to compensate for the minimum living standards. This is particularly the case for those with mental disorders, and the share of the disabled to receive public assistance has increased for the physically disabled and those with mental disorder.10 11.3

Historical Data on Health

This section reviews some long- term time series data on representative measures of health status. First, the trend of mortality in Japan is examined. 9. According to The Current Condition of Work Disaster Compensation for Mentally Disabled (Seishin Shougai nado no Rousai Hoshou Jyokyo) annually released by the Ministry of Health, Labour, and Welfare, the number of the approved was very small (nine between 1983 and 1996 [fourteen years]) but increased to 100 in 2002 and around 270 in 2007 or 2008 (http:// www.mhlw.go.jp/bunya/roudoukijun/rousaihoken04/090316.html). 10. Momose (2008) remarked that 250,000 persons received an additional allowance (Shougai Sha Kasan) for the disabled among the public assistant recipients in 2006, and the number has substantially increased, compared to 100,000 in the mid- 1990s. According to a survey by the Tokyo Metropolitan Government (Shougai Sha no Seikatsu Jittai), the proportion of the recipients of public assistance is 6.4 percent (in 2003 and 7.0 percent in 2008) for the physically disabled, 3.4 percent (in 2003 and 2.7 percent in 2008) for the mentally retarded, and 25.7 percent (in 2003 and 31.0 percent in 2008) for those with mental disorders, all of which are larger than the average proportion of the public assistance recipients. The high proportion of the mentally disabled is accounted for by the low employment rate, the high rate of remaining single, and the high rate of nonrecipients of pension benefit.

400

Takashi Oshio and Satoshi Shimizutani

Second, another measure of health status, a subjective health status is used to describe the development of health status in Japan. Third, we will relate the two measures—mortality and subjective health status in Japan. 11.3.1

Mortality

Mortality trends are, of course, not identical with health status and, thus, with disability trends, but are probably the only historical measure available in Japan and that comparable with other countries. We present the mortality trends in three ways. Figures 11.3 and 11.4 illustrate the ages of equal mortality probability for males and females, respectively. The historical data on mortality by age and gender in every year from 1960 to 2007 is available in “Simplified Life Table (Kan-I Seimei Hyo).”11 We set the mortality at age sixty and sixty- five in 1960 as the reference points in the base year and computed the corresponding figures in the subsequent years, assuming the mortality increases linearly with the ages. As seen in figure 11.3, for males, the mortality at age sixty in 1960 corresponds to that at age 71.1 in 2007 (11.1 years extension), and the mortality at age sixty- five in 1960 does so to that at age 75.6 in 2007 (10.6 years extension). Those results show that the mortality in Japan drastically declined over fifty years. Figure 11.4 reveals that the extension is more remarkable for females. The mortality at age sixty in 1960 corresponds to that at age 74.0 in 2007 (14.0 years extension), and the mortality at age sixty- five in 1960 does so to that at age 78.1 in 2007 (13.1 years extension). On a closer look, it can be observed that the tempo of extension for males has accelerated in the 1970s and has become slightly stagnant in the 1990s. A similar pattern is observed for females, too, but the weak trend in the 1990s is less evident. Figure 11.5 presents the mortality trends in a different manner. It shows the two- year mortality rates by age and gender in 1960 and 2005. If we take the 5 percent level of morality, the corresponding age for males in 1960 and 2005 was 68.7 and 78.9, respectively, while that for females in 1960 and 2005 was 72.5 and 84.0, respectively. The gap in the two years is larger for males probably because the starting age in the initial year is lower than that for females. This figure also demonstrates that the mortality rate has substantially declined over forty- five years. Figures 11.6 and 11.7 report the mortality rates at ages fifty- five, sixty, and sixty- five in every year for males and females, respectively. This again shows that the mortality rate has declined over half a century. The mortality has been monotonically declining and has always been lower for females than it has been for males. As observed in figure 11.5, the decline in the male mortality rate is greater in the 1970s but slower in the 1990s, while that for the female mortality rate is linear. 11. Unfortunately, data are available only for every five year since 1995. We interpolated a linear trend for the five years to obtain the data in every year.

Disability Pension Program and Labor Force Participation in Japan

Fig. 11.3

Ages of equal mortality probability for men

Fig. 11.4

Ages of equal mortality probability for women

11.3.2

401

Self-Rated Health

Next, we move on to discussing other health measures. A representative measure of health status is a self- rated health status, which has been widely used to stand for the state of progress of the condition. While there are some critiques for the measure because of its subjective nature, it is accepted as a popular aggregate health measure that is easily obtained at a low cost. A standard version of self- reported health status is the North American version with five answer categories ranging from “excellent” to

402

Takashi Oshio and Satoshi Shimizutani

Fig. 11.5

Two-year mortality rates by age and gender, 1960 and 2005

Fig. 11.6

Male mortality

“poor.” Another is the European version with five answer categories ranging from “very good” to “very poor.” The North American version is more popular and employed in some large data sets in Japan, including the Comprehensive Survey of Living Conditions of People (CPSLCP) compiled by the Ministry of Health, Labour, and Welfare starting in 1986 and JSTAR (Japanese Study of Ageing and Retirement), which is internationally com-

Disability Pension Program and Labor Force Participation in Japan

Fig. 11.7

403

Female mortality

parable to the Health and Retirement Study-English Longitudinal Study of Ageing-Survey of Health, Ageing and Retirement in Europe (HRS-ELSASHARE; Ichimura, Hashimoto, and Shimizutani 2009), starting in 2007. We utilize the data of self- rated health on a five- point scale from the CPSLCP— “excellent,” “very good,” “good,” “fair,” and “poor.” Figures 11.8 and 11.9 illustrate the developments of the self- rated health status in males and females, respectively—the shares of those assessing their heath conditions “excellent” or “very good” and those assessing them as “very good” or “good”—as well as the shares of respondents reporting subjective symptoms and of those who visit a hospital or clinic among those aged fifty- five to sixty- four (per thousand persons), along with the mortality rate for the same age group over the period between 1986 and 2007. The CPSLCP began in 1986 and is performed every three years. While the morality rate has been steadily declining, the share of those reporting good health increased up to the middle of the 1990s, although it has displayed a declining trend to date. With no clear trend observed from the other two health measures, it suggests that there is a negligible relationship between the health measure and the decline in mortality. 11.3.3

LFP and DI versus Mortality

This section examines the relationship between LFP and DI in comparison to mortality. To this end, we first collected data of the number of DI recipients by age group and gender from the Review of Public Pension Finances (Zaisei Saikeisan), which the Ministry of Health, Labour, and Welfare releases almost every five years. Next, we linearly interpolated the

404

Takashi Oshio and Satoshi Shimizutani

Fig. 11.8

Male health measures and mortality

Fig. 11.9

Female health measures and mortality

figures for other years with some adjustment.12 Figures 11.10 and 11.11 show the trends of employment and disability along with the mortality rate for 12. In the Review of Public Pension Finances, the number of recipients by age and gender for EPI pensions is available in 1979, 1983, and 1986 and that of the eligible is in 1991, 1996, 2001, and 2007. We use the shares of the recipients (or the eligible) by gender and age group (multiplied by the total number available in the Annual Report on Social Security Administration) to estimate the number of the recipients in those years. We assume that the shares for MAI pensioners are identical with those of EPI pensioners. The number of recipients of NPI

Disability Pension Program and Labor Force Participation in Japan

405

Fig. 11.10 Male employment versus mortality versus disability, aged sixty to sixty-four

Fig. 11.11 Female employment versus mortality versus disability, aged sixty to sixty-four

those aged sixty to sixty- four for males and females, respectively. While the mortality rate has been declining steadily, the trend of the employment rate has been slightly downward for males and almost flat for females, with some pensioners under the old program before 1985 is available in 1978, 1981, 1986, 1991, 1996, 2001, and 2007 and those under the new program after 1985 in 1986, 1991, 1996, 2001, and 2007 (the number of the eligible instead of the number of recipients since 1991 for both). We applied the same method to compute the number of recipients in those years. After those computations, we performed linear interpolation.

406

Takashi Oshio and Satoshi Shimizutani

cyclical movements for both. More important, the share of DI recipients has remained very low for both males and females, albeit with a slight upward trend, and had no clear comovements with the employment rate and the mortality rate. Figures 11.12 and 11.13 show the relationships between the employment rate and age and between the employment rate and mortality risk for males, where mortality risk is one- year mortality rate at a given age. The employment rate by age declined slightly from 1980 to 2005 but that by mortality risk declined more substantially. More specifically, the employment rate was 50 percent at an approximate age of seventy in 1980 and at an approximate

Fig. 11.12

Male employment rate by age

Fig. 11.13

Male employment rate by mortality risk

Disability Pension Program and Labor Force Participation in Japan

407

age of sixty- six in 2005, while it was 50 percent at the mortality rate of about 3.5 percent in 1980 and only about 1.4 percent in 2005. Although the mortality rate declined substantially over the past twenty- five years, the elderly have become more inclined to retire. See figures 11.14 and 11.15 for females. The employment rate by age shifted upward up to age sixty and remained almost unchanged beyond that between 1980 and 2005. The mortality- employment curves skewed to the vertical axis, and the level of the mortality rate that corresponds to a 50 percent employment rate stayed approximately 0.5 percent over the same period.

Fig. 11.14

Female employment rate by age

Fig. 11.15

Female employment rate by mortality risk

408

11.4

Takashi Oshio and Satoshi Shimizutani

Historical Data on Activity and Program Participation Rates

This section explores the historical relationship between labor force participation and disability pension programs.13 The goal is to assess the effect of each revision in the disability pension program on the labor market by discovering trends that may be explained by the revisions. 11.4.1

Historical Labor Force Data

Figures 11.16 to 11.21 examine the long- term trends in the labor force, focusing on the rates of employment, unemployment, and not being in the labor force for three age groups: forty to forty- four, fifty to fifty- four, and sixty to sixty- four over 1970 to 2010 for males and females. For males, the employment rate shows cyclical movements with a slightly downward trend for those aged sixty to sixty- four, while it has remained stable at a high level for the younger two age groups. Correspondingly, the oldest age group shows clearer cyclicality of the rates of those unemployed and of those not in the labor force as compared to the other age groups. For females, the rates of those employed and of those not in the labor force show a modest uptrend and downtrend, respectively, for the two younger age groups, while they are stable for those aged sixty to sixty- four. The unemployment rate has been moving almost the same way across age groups. 11.4.2

Historical DI Data

Figures 11.22 and 11.23 present long- term trends of the shares of DI beneficiaries in total population for three age groups: forty to forty- five, fifty to fifty- four, and sixty to sixty- four for males and females, respectively, between 1970 and 2010. The figures also indicate the three major revisions in 1974, 1986, and 2006. We observe that the share of DI recipients has been modestly increasing for all the age groups in both males and females, though the share remains low. For males, the share is 1 to 2 percent for those aged forty to forty- five, whereas that for those aged fifty to fifty- four and sixty to sixty- four increased from 1 to 2 percent in 1970 to 3 to 4 percent in 2010. Taking a closer look, we see that there are small jumps in 1974 to 1975, 1985 to 1986, and 2005 to 2006—albeit not for all age groups—consistent with figure 11.2. All of them are caused by the revisions to disability pension programs. In particular, for both males and females aged sixty to sixtyfour, we observe a remarkable increase in the DI beneficiaries following the 1975 revision and the jump in 2006. In 1974, the disability welfare program began to cover Grade 2 as well, and many patients with mental disorders or deficiencies became eligible to receive the benefit. The 2006 revision allowed Disability Basic Pension recipients aged sixty- five or above to additionally 13. It would be excellent if direct data were available on pathways to retirement, but, unfortunately, no such data are available in Japan.

Fig. 11.16

Male employment

Fig. 11.17

Female employment

Fig. 11.18

Male unemployment

Fig. 11.19

Female unemployment

Fig. 11.20

Male not in labor force (NLF)

Fig. 11.21

Female not in labor force (NLF)

Disability Pension Program and Labor Force Participation in Japan

Fig. 11.22

Male not in labor force (NLF), disability

Fig. 11.23

Female not in labor force (NLF), disability

411

receive EPI benefits, providing the elderly with more incentives to apply for Disability Basic Pension. However, these trends in the DI participation rates are unrelated to those in the unemployment rates in all age groups (see figures 11.18 and 11.19), suggesting that there is no trade- off between DI and unemployment benefit receipts. 11.4.3

Historical Data on DI Participation versus Mortality and Health

Next, we try to relate the DI participation and health measures. Figures 11.24 to 11.27 present long- term trends in mortality at age forty- five and sixty and DI participation rates at ages forty to forty- four and sixty to sixtyfour for males and females. It is clear that there is no reasonable relation

Fig. 11.24

Male mortality versus disability, aged forty-five

Fig. 11.25

Female mortality versus disability, aged forty-five

Fig. 11.26

Male mortality versus disability, aged sixty

Disability Pension Program and Labor Force Participation in Japan

Fig. 11.27

413

Female mortality versus disability, aged sixty

between the two series in each figure. Instead, we observe a contradicting pattern: the DI participation rate has increased despite the lower mortality implying that people have become healthier. Figures 11.24 to 11.27 show that there are two small bumps in the DI participation rate after the 1974, 1985, and 2006 revisions, which are evident for males aged fifty to fifty- four and sixty to sixty- four. In contrast, there is no jump in the mortality rate in the same period, reinforcing the idea that the small increases in the DI participation rate were motivated by the DI program revisions. This justifies the observations that there is no discontinuity in the DI participation rate after some disturbances in the mortality rates in the second half of the 1990s and the beginning of the 2000s, which is the case for both males and females. Figures 11.28 to 11.33 illustrate long- term trends of self- rated health status at age fifty- five to sixty- four and DI participation rates at age sixty to sixty- four for males and females, respectively. Figures 11.28 and 11.29 show that there is little correspondence between DI participation and the shares of those who reported their health was “excellent” or “very good” and those who reported it as “very good” or “good.” The DI participation rate dropped in 2004 for males and showed an uptrend for females, while the share of those who reported good or modestly good health remained virtually the same for both males and females. Figures 11.30 and 11.31 also show little association between DI participation and the share of those who reported bad health (“poor” or “very poor”). The share of people suffering from bad health declined until the mid- 1990s but increased from the end of the 1990s for both males and females in a way not correlated with the trend of the DI participation rate for both genders. Figures 11.32 and 11.33 computed a score of self- rated health status, which assigns “five” to the best of the five choices and “one” to the worst.

Fig. 11.28

Male good health versus disability, aged sixty

Fig. 11.29

Female good health versus disability, aged sixty

Fig. 11.30

Male bad health versus disability, aged sixty

Fig. 11.31

Female bad health versus disability, aged sixty

Fig. 11.32

Male self-rated health versus disability, aged sixty

Fig. 11.33

Female self-rated health versus disability, aged sixty

416

Takashi Oshio and Satoshi Shimizutani

Again, there is no clear relationship between DI participation and the score of self- perceived health status. These figures show that the trend of DI participation is unrelated with self- rated health status, the same as for the relationship between DI participation and mortality, although we need to be careful in interpreting the comparison of self- rated health status across individuals and time. 11.5

Concluding Remarks

This chapter investigates historical information to explore the relationship between labor force participation of middle aged and old people and the disability program in Japan. In particular, we explore the time series dimension to identify what has determined the trend in disability program participation over time and relate it with the labor supply. We find that the mortality and other health measures have been largely unrelated with the disability program participation rates. While major revisions of the disability program have slightly expanded the eligibility for the DI program, the program participation is still very low; thus, the effect on labor force participation is very limited in Japan in contrast to some European countries with a high take-up rate inducing early retirement.

References Börsch-Supan, Axel. 2005. “Work Disability and Health.” In Health, Ageing and Retirement in Europe: First Results from the Survey of Health, Ageing and Retirement in Europe, edited by Axel Börsch-Supan, Agar Brugiavini, Hendrik Jürges, Johan Mackenbach, Johannes Siegrist, and Guglielmo Weber, 253– 58. Mannheim, Germany: Mannheim Research Institute for the Economics of Aging (MEA). Ichimura, Hidehiko, Hideki Hashimoto, and Satoshi Shimizutani. 2009. “JSTAR First Results: 2009 Report.” RIETI Discussion Paper Series 09-E- 047. Research Institute on Economy, Trade and Industry, Project on Intergenerational Equity and Center for Intergenerational Studies (CIS), Hitotsubashi University, Discussion Paper no. 443-447. Ministry of Health, Labour and Welfare. 2009. “Textbook for the Study Programme for the Senior Social Insurance Administrators (abstract).” http://www.mhlw. go.jp/english/org/policy/p36-37a.html. Ministry of Health, Labour and Welfare. 1980, 1985, 1989, 1995, 1999, 2004, 2009. Review of Public Pension Finances. Momose, Yu. 2008. “Income Security Systems for the Disabled with a Special Reference to the Disability Pension Programs (in Japanese). Kikan Shakai Hosho Kenkyu 44 (2): 171– 85. National Institute of Population and Social Security Research. 2006. Population Projection for Japan: 2006–2055. Tokyo: National Institute of Population and Social Security Research.

Disability Pension Program and Labor Force Participation in Japan

417

Oshio, Takashi, Akiko Sato Oishi, and Satoshi Shimizutani. 2011. “Social Security Reforms and Labor Force Participation of the Elderly in Japan.” Japanese Economic Review 62 (22): 248– 71. Oshio, Takashi, Satoshi Shimizutani, and Akiko Sato Oishi. 2010. “Does Social Security Induce Withdrawal of the Old from the Labor Force and Create Jobs for the Young? The Case of Japan.” In Social Security Programs and Retirement around the World: The Relationship to Youth Employment, edited by Jonathan Gruber and David Wise, 217– 41. Chicago: University of Chicago Press.

12 Disability Insurance and Labor Market Exit Routes of Older Workers in the Netherlands Klaas de Vos, Arie Kapteyn, and Adriaan Kalwij

12.1

Introduction

Earlier research suggests that the low participation rate of persons well below the statutory retirement age of sixty- five in the Netherlands can, at least in part, be attributed to generous incentives to retire (cf. Kapteyn and de Vos 1999). Until recently, the incentive to retire early was particularly strong if the employer offered an Early Retirement (ER) scheme, but an alternative route to retire via Disability Insurance (DI) was also quite attractive. A less attractive route to retire for workers is via Unemployment Insurance (UI) or Social Assistance (SA). The attractiveness of DI and its relatively easy access for elderly workers can be seen as significant contributing factors for the high prevalence of DI receipt in the Netherlands, as compared to most other developed countries. There is a large body of research showing that a considerable part of the DI recipients in the Netherlands are not really (totally) unfit to work (cf. Aarts and de Jong 1990). De Gier, Henke, and Vijgen (2003) conclude however that most of the results of this research have hardly affected DI policy making. In this chapter, we aim to contribute to the knowledge on the relationship between DI and retirement by focusing on two questions. The first question is whether, next to or instead of the strong incentives to retire via DI, we can find indications that trends in health and mortality actually played a role in the increase of the numbers of DI recipients until the early 1990s and their Klaas de Vos is a senior researcher in the Quantitative Analysis Department at CentERdata. Arie Kapteyn is a senior economist at the RAND Corporation. Adriaan Kalwij is assistant professor at Utrecht University and a research fellow at Netspar. For acknowledgments, sources of research support, and disclosure of the authors’ material financial relationships, if any, please see http://www.nber.org/chapters/c12393.ack.

419

420

Klaas de Vos, Arie Kapteyn, and Adriaan Kalwij

decrease since then. To that end, we will present various graphs relating health, mortality, DI receipt, and labor market participation. The second question concerns the effect of reforms on DI and other income maintenance and early retirement programs on the number of older workers exiting the labor market and the pathways chosen to exit and effectively retire. To assess possible effects of reforms, we will present graphs that show the dates of various reforms in combination with the trends in the pathways to retirement. In addition, we will estimate a model in which the reforms to DI and other programs are included as explanatory variables for changes in the pathways to retirement. The chapter proceeds as follows. Section 12.2 contains a brief overview of DI and other programs and a chronological overview of reforms during the past four decades. Section 12.3 presents graphs on health, mortality, DI, and labor force participation. Section 12.4 presents graphs on the pathways to retirement. Because the data on pathways only cover a limited period, section 12.5 presents a number of additional graphs covering a longer period. Section 12.6 aims to quantify the effects of reforms on the pathways to retirement. Section 12.7 concludes. 12.2

12.2.1

Disability Insurance and Other Social Security Programs and Reforms Disability Insurance (DI)

Introduced in 1967, the Dutch Disability Insurance (WAO, Wet op de Arbeidsongeschiktheidsverzekering) aimed to insure employees against loss of earnings as a result of long- term inability to work as a result of illness or incapacity. If, after having been ill for a period of one year, the employee could not resume work, he or she would be entitled to an earnings- related DI benefit that could last until the employee reached the statutory retirement age of sixty- five. Starting in the 1970s, the numbers of individuals on DI in the Netherlands showed a continuous increase until the 1990s. These numbers were much higher than expected when the new DI legislation was introduced and much higher than might be expected given the average health status of the population. In fact, in the mid- 1970s, when unemployment was rising dramatically, the route to DI was generally used by employers as a path of least resistance to shed superfluous employees. For the employee, DI was both socially more acceptable and more attractive than UI, in particular because the benefit could be received until age sixty- five when the old- age pension would kick in. 12.2.2

DI Reforms

With the increase in the number of benefit recipients, expenditures on DI started to rise dramatically, and since the start of the 1980s, government

Disability Insurance and Labor Market Exit Routes in the Netherlands

421

policy has sought to reverse the trend of the ever increasing DI expenditures by various reforms to limit access to DI, increase the number of persons exiting DI, and lower the average DI benefit (see table 12.1). In 1985, the replacement rate of DI was lowered from 80 to 70 percent in the wake of the recession of the early 1980s. This affected both new entrants to DI and “existing cases.” In 1987, access to the full DI was limited for partially disabled unemployed new entrants. In the early 1990s, again government saw the need to do something about the high numbers of DI recipients, and this resulted in a series of measures: the duration of the full DI benefit was limited for new entrants younger than fifty, stricter disability criteria were introduced for entry into DI, and younger DI recipients were to be retested. For individuals younger than forty- five, the new disability criteria were to be used, and, for persons between forty- five and fifty, retesting was to be done on the basis of the old disability criteria. Mainly because most employees took out a private insurance to compensate for the shorter duration of the full DI benefit for younger persons, DI remained an attractive option. Next to limiting the access and the generosity of the benefit, policies were also introduced to shift the costs to firms with high numbers of employees exiting to DI. First, the costs of sickness benefits were charged directly to the employer for two to six weeks (1994), and later on for a full year preceding the exit to DI. Second, in 1998, experience rating was introduced: for large firms in particular, the DI contributions were partly based on the DI record of the firm in question. A high exit rate into DI resulted in higher contributions. Because all these reforms did not succeed in substantially reducing the numbers of DI recipients, by 2002, the feeling was that enough was enough and the time had come for a more radical approach. As from 2002, during the year of sickness preceding exit to DI, employer and employee are jointly responsible for taking sufficient action for reintegration into the workforce. Moreover, this sickness period can be extended if insufficient reintegration measures have been taken. As from 2004, exit to DI only happens after two years of sickness, during which time the employer pays sickness benefits. As from 2006, the new DI law (WIA, Wet werk en inkomen naar arbeidsvermogen) makes a strict distinction between fully and permanently disabled and partially or temporarily disabled. The former receive a generous 75 percent of their previous earnings until age sixty- five (IVA, Inkomensvoorziening Volledig Arbeidsongeschikten). The latter receive a less generous benefit (WGA, regeling Werkhervatting Gedeeltelijk Arbeidsongeschikten), which depends on the previous earnings, the number of weeks worked before, and the current earnings (if any) and the percentage of previous earnings that the employee is deemed to be capable of earning. Furthermore, once again, a retest procedure is set up for existing DI beneficiaries younger than fifty (whose DI remains unchanged). All in all, despite the decreased generosity of the benefit, for a long time, DI remained an attractive route to exit the labor force, and while the levels of

Table 12.1

Timeline of reforms to Disability Insurance (DI), State Pension (SP), Early Retirement (ER), Occupational Pension (PP), and Unemployment Insurance (UI) DI/(long-term) sickness insurance

1967

Introduction of DI (WAO; 20,000 beneficiaries expected)

1974 1975–82

1985 1987

764,000 beneficiaries, replacement rate reduced from 80% to 70% No more (full) DI for (partially) unemployed

1988 1991 1993

1994

(i) Persons younger than 50 receive DI for a limited period (ii) Stricter disability criteria (iii) Retesting of younger DI recipients Introducing employer-paid periods of sickness (2–6 weeks)

1995 1996

1998

2004

2006

2008

SP: Benefit raised and linked to net minimum wage ER: Gradual introduction by sector/firm/ department; 1980 ER contribution tax deductible UI: Replacement rate reduced to 70% SP: Married women get independent claim UI: Changes in eligibility and benefit period; earnings-related benefit followed by continuation benefit SP: Distinction between married and cohabiting couples abolished; introduction single parent allowance SP: Earnings tested supplement when partner is younger than 65 UI: Eligibility revised

SP: Earnings tested supplement adapted UI: Eligibility revised, introduction short-term benefit

(i) Sickness benefit privatized: employer pays 70% of earnings (1 year) (ii) Exemptions for earnings tested supplement abolished (i) Introduction of (limited) experience rating DI contributions employer (ii) Public employees included in DI

2000–05 2001 2002 2003

SP, ER, PP, UI

ER/PP: Trend toward actuarially fairer flexible ER age UI: Public employees included Stricter reintegration rules in case of sickness Experience rating for small employers abolished (i) Sickness benefit period extended to 2 years (ii) Strict reevaluation DI recipients younger than 50 Introduction of new DI: strict distinction between partially and fully, permanently disabled

UI: Abolition of continuation benefit

ER: Fiscal friendly treatment of ER contributions repealed; introduction of life course saving UI: benefit period shortened, higher benefit first two months

Experience rating DI abolished

Main source: Kroniek van de sociale verzekeringen 2008, www.uwv.nl. Note: WAO = Wet op de Arbeidsongeschiktheidsverzekering.

Disability Insurance and Labor Market Exit Routes in the Netherlands

423

DI expenditures as a percentage of gross domestic product (GDP) have been decreasing considerably since the early 1980s, the number of DI recipients peaked much later, in the mid- 1990s. Only the most recent series of major reforms, starting in 2002, appears to have resulted in a clear downward trend in the numbers of DI recipients. 12.2.3

Other Programs

As mentioned in the preceding, DI benefits end when the recipient turns sixty- five. More in general, the Dutch benefit system is characterized by a clear distinction between benefits for persons younger than sixty- five and benefits for persons aged sixty- five or above. The latter receive the flat rate state pension (SP), in most cases supplemented by an occupational pension (PP). The former may be entitled to Disability Insurance (DI), Unemployment Insurance (UI), means- tested Social Assistance (SA), or, when having retired before age sixty- five from a firm offering an occupational pension plan, an Early Retirement benefit (ER). Early Retirement was introduced in most sectors during the 1970s, mainly in reaction to rising unemployment (cf. Kapteyn, de Vos, and Kalwij 2010) and, at least until the end of the 1990s in most cases, consisted of an offer too good to refuse. In particular, workers retiring later than the earliest possible ER date were not compensated by higher benefits or lower taxes so that, in fact, they faced an implicit tax rate of more than 100 percent (cf. Kapteyn and de Vos 1999). For workers approaching sixty who were not entitled to ER (e.g., because of an insufficiently long employment history or because they worked in a firm that did not offer ER) and who could not plausibly be retired via DI, Unemployment Insurance (UI) offered a third pathway to retirement before the statutory retirement age of sixty- five. In most cases, it offered a replacement rate of 70 percent, and, furthermore, until recently, no obligation to search for employment after the age of 57.5. Upon reaching the age of sixty- five, all residents are entitled to the flat rate state pension (AOW), which is financed by a pay- as-you- go social insurance contribution. Moreover, most employees accumulate fully funded occupational pension rights and supplement their state pension to (ideally) 70 percent of previous earnings. Notably, traditionally in the Netherlands, substantial numbers of women used to leave the labor market (long) before age sixty- five without an earnings replacing benefit. In recent decades, labor market participation of older women has increased rapidly, and this pathway to retirement has become less common. 12.2.4

Reforms to Other Programs

The prospect of exploding costs once the large baby boom cohorts start to reach the ER age turned out to be sufficiently threatening for effective

424

Klaas de Vos, Arie Kapteyn, and Adriaan Kalwij

reforms to be put in place by the end of the 1990s. In most cases, a cost reduction as a result of reducing the effective ER entitlement was combined with the introduction of a more or less actuarially fair system. As a result, the employee could still opt for retiring early, but with a reduced pension, or later, but with an increased pension. By 2006, the government dealt a final blow to the old ER systems by effectively terminating the tax exemption for ER contributions that would enable a retirement age lower than sixty- five. Only systems offering a replacement rate of at most 70 percent at the pension age of sixty- five, and actuarially fair reductions when an earlier pension age is chosen can still collect tax exempt contributions. This reform does not affect workers who are already close to sixty- five and have accumulated more generous entitlements. The Dutch Unemployment Insurance (UI) system saw a major reform in 1987, affecting both eligibility and benefit period. However, since 1984, workers aged 57.5 or older could receive UI until the state pension age of sixty- five, and this did not change in 1987. As of 2004, persons aged 57.5 or older receiving UI are no longer exempt from the requirement to seek work. In other words, they are no longer “automatically” receiving UI until age sixty- five but have to try to find work and, in theory, accept a job offer. Moreover, as of October 1, 2006, the maximum duration of UI is thirtyeight months. After that period, all that is left is a means- tested entitlement to Social Assistance (SA) with a benefit equal to the net minimum wage. By and large, since 1974, for most single individuals and couples older than sixty- five, changes in the flat rate state pension are indexed by changes in the after tax minimum wage. Revisions include the introduction of an independent pension entitlement for married women in 1985 and the entitlement of supplementary state pension for persons with a spouse younger than sixty- five (1987, revised 1994). A proposed legislation to increase the statutory pension age by two years (to sixty- seven) by 2025 is still in discussion. In addition to the universal right of a state pension from the age of sixtyfive onward, an increasing number of persons reaching the age of sixty- five (approximately 96 percent of men and 67 percent of women in 2007. See Knoef, Alessie, and Kalwij 2010) have been accumulating supplementary defined benefit occupational pension rights (PP). However, since the early 2000s, the generosity of these occupational pensions has gradually been reduced because most of the pension funds have shifted from calculating the pension on the basis of final earnings to calculations based on average earnings. Moreover, in general the indexation of the benefits, which used to be based on the wage index, has become less generous following the successive crises on the stock market, affecting the investments of pension funds. Recently proposed pension reforms aimed at ensuring a sustainable Dutch pension system are still in discussion (cf. Stichting van de Arbeid 2010).

Disability Insurance and Labor Market Exit Routes in the Netherlands

425

12.3 Historical Data: Mortality, Health Measures, Labor Force Participation (LFP), and Disability Insurance (DI) In this section, we address the first question formulated in the introduction on whether we can find indications that trends in health actually played a role in the increase of the numbers of DI recipients until the early 1990s and their decrease since then. For this purpose, we present statistics on labor force participation, the use of DI, self- reported health, and mortality rates of the population. These statistics point to the conclusion that there is not much of a relationship between the numbers of persons using DI and the general health of the population. 12.3.1

Mortality

We present three types of figures on mortality. Figures 12.1 (men) and 12.2 (women) take the mortality rates at age sixty and sixty- five of 1960 as their starting point and show the ages with the same mortality rate in the period up to 2008. For men, the age of equal mortality was actually lower than in 1960 until the early 1980s. From 1970, it shows an increase of more than seven years up to 2008. For women, unlike for men, the age of equal mortality did not show a clear decrease in the 1960s. However, on balance, the increase up to 2008 was only marginally larger than that of men. Consistent with the findings in figures 12.1 and 12.2, figure 12.3 shows that both sexes saw their two- year mortality rates decline between 1960 and 2007, especially at higher ages. The decrease of the mortality rate for older women was substantially larger than that of men. Figures 12.4 and 12.5 show the time paths of mortality rates for men and women ages fifty- five, sixty, and sixty- five, respectively. They confirm that

Fig. 12.1 to 2008

Ages of equal mortality probability in the Netherlands for men, 1960

426

Klaas de Vos, Arie Kapteyn, and Adriaan Kalwij

Fig. 12.2 to 2008

Ages of equal mortality probability in the Netherlands for women, 1960

Fig. 12.3 Two-year mortality rate in the Netherlands by age and sex, 1960 and 2007

for men, mortality in 1980 was about equal to mortality in 1960, but it also shows a clear decrease between 1970 and 2008. Women show a more uniform decline, starting from much lower levels. 12.3.2

Mortality and Health

None of the figures on mortality would lead one to suspect that, starting in the 1970s, the Netherlands saw an unprecedented increase in the numbers of disability benefit recipients and that, in fact, until recently, the Netherlands had one of the highest levels of DI receipt in the world. Conceivably, however, the introduction of a generous DI system may have contributed to the marked fall in male mortality during this period. Apart from mortality, not many health statistics can be found that can be used to track health over time. One of the few health statistics that is more or less comparable over a

Disability Insurance and Labor Market Exit Routes in the Netherlands

Fig. 12.4

Male mortality (/1000)

Fig. 12.5

Female mortality (/1000)

427

fairly long period of time is the series of self- reported good health collected by the Organization for Economic Cooperation and Development (OECD), available from 1983 until 2008. For both men and women aged forty- five to sixty- four, this statistic shows a somewhat erratic pattern (figures 12.6 and 12.7). For men, we observe on balance an increase in the percentage reporting good health or better between 1983 and 1995 and, with a few exceptions, a fairly stable rate (of about 75 percent) since then. For women, there is even less of a trend, and, in most years, the percentage of women reporting a good health or better hovers around 70 percent. 12.3.3

LFP and DI versus Mortality

It appears from figures 12.6 and 12.7 that self- reported health and mortality do not show a clear (negative) correlation. However, neither self- reported health nor mortality show a clear relation with DI prevalence or labor force

428

Klaas de Vos, Arie Kapteyn, and Adriaan Kalwij

Fig. 12.6 Male mortality (/1000, left axis) and self-reported good health (%, right axis)

Fig. 12.7 Female mortality (/1000, left axis) and self-reported good health (%, right axis)

participation either. As an example, we show employment and disability for men aged sixty to sixty- four in one figure with mortality at age sixty and sixty- five (figure 12.8) from 1981 until 2007. For women, we present the same information in figure 12.9. For men, the continuous decrease in the mortality rates has been accompanied by decreasing employment and increasing disability prevalence during the first part of the period and increasing employment and decreasing disability since approximately 1995. For women, decreasing mortality rates accompany more or less stable employment and somewhat increasing DI during the first half of the period and rapidly increasing employment and stable DI prevalence during the second half of the period. It is hard to interpret these patterns just based on health trends. During the first part of the period, one could argue that the increase in DI receipt has contributed

Disability Insurance and Labor Market Exit Routes in the Netherlands

429

Fig. 12.8 Male employment and disability (%, left axis) versus mortality (/1000, right axis), aged sixty to sixty-four

Fig. 12.9 Female employment and disability (%, left axis) versus mortality (/1000, right axis), aged sixty to sixty-four

to longer lives, by allowing workers an escape from jobs that are detrimental to health. During the second part of the period, however, we see that both mortality and DI receipt are falling, so then we would have to assume that causality has changed direction and that a healthier population (at least for the male population) has less need to apply for DI benefits. Altogether, it seems more plausible that the changes in DI receipt are largely the result of the various reforms discussed before. It is to the effects of these reforms that we now turn. 12.4

Pathways to Retirement and Program Reforms

In this section, we address the second question formulated in the introduction and assess to what extent reforms in DI and other income maintenance

430

Klaas de Vos, Arie Kapteyn, and Adriaan Kalwij

and early retirement programs have affected the pathways to retirement and the labor force participation of the population approaching the retirement age. Detailed data on the pathways to retirement are only available for a relatively short period (1989 to 2007). Therefore, we will also present some results that do not depend on the availability of pathways data (section 12.5). The focus of the present section is the information on the pathways to retirement from the Income Panel Survey, which we first briefly describe in the following. 12.4.1

IPO Data

The data are taken from the 1981, 1985, and 1989 to 2007 Income Panel Study of the Netherlands (IPO [Inkomens Panel Onderzoek], Centraal Bureau voor de Statistiek [CBS] 2009) gathered by Statistics Netherlands. The IPO, a representative sample of the Dutch population, consists of an administrative panel data set of, on average, about 95,000 individuals per year. Sampling is based on individuals’ national security numbers, and the selected individuals are followed for as long as they are residing in the Netherlands on December 31 of the sample year. The data set includes individuals living in institutions for the elderly, such as nursing homes. Individuals born in the Netherlands enter the panel for the first time in the year of their birth and immigrants to the Netherlands in the year of their arrival. An individual only exits the panel by death or emigration from the Netherlands. The IPO contains data on the demographic characteristics, income, and labor market status for each member of a selected individual’s household. 12.4.2

IPO Definitions of Labor Market Status

The IPO income data are based primarily on records from the tax office and institutions that pay out (insurance) benefits and contain detailed information on all income components at an individual level. Based on the largest income component, Statistics Netherlands assigns a labor market status to an individual. An individual can be (self-)employed, unemployed (UI, receives unemployment insurance or social assistance benefits), on disability (DI, receives disability or [long- term] sickness insurance benefits), early retired (ER, when receiving pension income), or nonparticipating (NP, no labor income, pension, or benefits).1 12.4.3

Policy Reforms

In the figures presented in the following and in the next section, we indicate the dates of the main reforms in DI and other income maintenance and

1. The disability insurance scheme is referred to as WAO or WIA and the (long- term) sickness insurance scheme as ZW (see www.uwv.nl). Data limitations prevent a more refined classification. For instance, if an individual is partially disabled and the main source of income is from employment, this individual is classified as employed.

Disability Insurance and Labor Market Exit Routes in the Netherlands

431

early retirement programs from 1980 onward. More specifically, we indicate the following reforms: • 1985: a reduction in the DI and unemployment benefits replacement rates from 80 to 70 percent • 1993: stricter DI eligibility criteria • 1996: introduction of one- year sickness benefits paid by employer before entering DI • 1998: introduction of experience rating of DI contributions by employers • 2000: actuarially fairer adjustments of ER pensions. • 2002: Stricter gatekeeping for DI • 2004: introduction of two- year sickness benefits paid for by employer before entering DI. • 2006: several reforms introduced at the same time, with strict distinctions between partial, full, and permanent disability, a shortening of the UI benefit period, and the abolition of fiscal- friendly treatment of ER contributions As discussed in section 12.2, table 12.1 provides a complete overview of reforms. It needs to be noted that there have been a more or less continuous series of Social Security reforms in the various programs since the early 1990s, and it is, therefore, hard to assess whether some of the trends we will discuss in the following can be attributed to particular reforms. In addition, business cycle fluctuations may blur the pictures, and we return to this issue in section 12.6. 12.4.4

(Self-)Employment

We start with two figures presenting the employment rates (including selfemployment) for men and women in three age groups (fifty to fifty- four, fifty- five to fifty- nine, and sixty to sixty- four). Notably, the employment rate of older persons (sixty- five+) can be assumed to be fairly negligible in the Netherlands. Figure 12.10 illustrates the dramatic decline in the employment rate of older men as well as the reversal of this trend starting about 1995. The pattern is particularly striking in the age group sixty to sixty- four, where the employment rate dropped from almost 60 percent in 1981 to less than 25 percent in 1994 and then picked up to more than 40 percent by 2007. For women, on the other hand, we hardly observe decreasing employment rates (see figure 12.11). On the contrary, in the group of women aged fifty to fifty- four, the employment rate has been increasing since 1985, and the older age groups followed this trend some years later. Notably, at the end of the observation period, in all three age groups, the employment rate is still substantially below the figure for men. As mentioned in the preceding, given the almost continuous series of

432

Klaas de Vos, Arie Kapteyn, and Adriaan Kalwij

Fig. 12.10

Male employment rate by age group

Fig. 12.11

Female employment rate by age group

reforms since 1993, it is hard to assess whether some of the changes in employment may be attributed to particular reforms. It can be assumed that at least part of the increase in the employment rate of older men would not have been observed if the implicit incentives to enter the respective retirement programs (ER, DI, UI) would have remained as generous as they were in the early 1990s. Cohort effects underlie large parts of the increases in the employment rates for women. However, here we can also assume that the employment rate would have been lower if access to DI, ER, and UI would not have been made more difficult and financially less attractive.

Disability Insurance and Labor Market Exit Routes in the Netherlands

12.4.5

433

Pathways to Retirement

Did the reforms affect the pathways to retirement? Undoubtedly, they did. However, it is difficult to distinguish the effects of reforms from other factors causing variation in the pathways. Moreover, even if the pathways remain stable, reforms may have effects on retirement (if, e.g., everyone chooses to retire one year later, the distribution across pathways may remain unchanged). Figures 12.12 and 12.13 show the pathways to retirement of men and women aged fifty to sixty- four between 1989 and 2006. A pathway

Fig. 12.12

Pathways to retirement, men, aged fifty to sixty-four

Fig. 12.13

Pathways to retirement, women, aged fifty to sixty-four

434

Klaas de Vos, Arie Kapteyn, and Adriaan Kalwij

is a particular exit route from employment between two subsequent years. For example, figure 12.12 shows that about 20 percent of the male workers aged fifty to sixty- four in 1989 who leave the labor market that year are on DI in 1990. Looking at these figures, we see that the pattern is fairly stable given the many reforms to DI and other Social Security programs during this period. Nevertheless, some trends can be observed. Over this period, DI becomes a relatively less frequently used pathway to retirement for both men and women, in particular from the late 1990s onward. For men, ER is the dominant route to retirement throughout the whole period. The year with the lowest share of ER (and the highest share of unemployment) is 1993, a year with low economic growth and rapidly increasing unemployment. For women, leaving the labor market without own source of replacement income (“Other”) and relying on the income or pension of the spouse remains a popular pathway to retirement but has been overtaken by ER.2 This is likely due to changes in the rules of the occupational pension schemes and the increased participation in terms of hours of work. For both men and women, retirement at the statutory retirement age of sixty- five remains quite exceptional. Unemployment and Disability Insurance together make up 30 to 40 percent of the pathways to retirement of the persons retiring in any given year among men and about 20 to 25 percent among women. 12.5 12.5.1

Historical Data on Activity and Health Employed, Unemployed, Not in the Labor Force

In this section, we present labor force data for three age groups (forty to forty- four, fifty to fifty- four, and sixty to sixty- four) for men and women for the period 1981 to 2007 on the basis of the Income Panel Survey of Statistics Netherlands (see section 12.4). Figures 12.14 and 12.15 present employment rates, figures 12.16 and 12.17 unemployment rates, and figures 12.18 and 12.19 present the complement of employment and unemployment: not in the labor force. For men, we once again note the steep decline of the employment rate in the older age group (sixty to sixty- four), from almost 60 percent in 1981 to less than 25 percent in 1994, picking up to more than 40 percent in 2007. Unemployment peaked at less than 15 percent in this age group in the 1990s, so, for the most part, the decline and subsequent increase in employment was complemented by a considerable increase in the percentage of men not in the labor force, peaking at more than 60 percent in 1994 and showing a clear decrease to less than 50 percent in 2007. The other two age groups 2. In fact, recipients of a state pension (sixty- five +) are entitled to a supplement if they have a younger spouse without income.

Disability Insurance and Labor Market Exit Routes in the Netherlands

Fig. 12.14

Male employment rate by age group

Fig. 12.15

Female employment rate by age group

435

show much less variation, with employment hovering close to 90 percent in the youngest age group and slightly lower in the age group fifty to fifty- four. For women, the employment rates show clear cohort effects: employment starts its spectacular increase during the 1980s in the younger age groups, followed with some delay, and at a much lower level, by the oldest age group. Unemployment, starting off at very low levels, hovers between 5 and 10 percent after 1989 in all three age groups. The percentages not in the labor force again mainly reflect the trends in the employment rates by showing the opposite pattern.

436

Klaas de Vos, Arie Kapteyn, and Adriaan Kalwij

Fig. 12.16

Male unemployment rate by age group

Fig. 12.17

Female unemployment rate by age group

If we separate out the DI recipients from men not in the labor force (figure 12.20), we note that whereas the latter statistic showed a steep increase in the oldest age group between 1981 and 1994, the number of DI recipients only increased moderately (from about 20 to 23 percent). On the other hand, a large part of the decrease in the number of men not in the labor force after 1994 in this age group is made up by a decrease in the number of DI recipients. For women, the DI rates do not show decreasing trends: they more or less appear to reflect the increasing labor force participation rates (see figure 12.21).

Disability Insurance and Labor Market Exit Routes in the Netherlands

Fig. 12.18

Male not in labor force (NLF) by age group

Fig. 12.19

Female not in labor force (NLF) by age group

12.5.2

437

Historical Data on DI versus Health, Mortality

Comparing mortality trends and DI receipt of men aged forty to fortyfour (figure 12.22), we notice that this age group showed an increase in the number of DI recipients between 1981 and 1994, despite the slightly decreasing mortality rate. After 1994, the decreasing mortality rate could be seen as one of the contributing factors to the decrease in the number of DI recipients. For the women in this age group, mortality and DI receipt show a negative correlation (figure 12.23), if any. In the age group sixty to sixty- four, the pattern is comparable: for men

438

Klaas de Vos, Arie Kapteyn, and Adriaan Kalwij

Fig. 12.20

Male disability rate by age group

Fig. 12.21

Female disability rate by age group

(figure 12.24), DI and mortality show opposing trends until 1994 but both move downward afterward. For women (figure 12.25), DI and mortality show a negative correlation throughout the whole period. Both for this age group and for the age group forty to forty- four, the negative correlation would be the direct result of the cohort effects in female labor force participation. Female employment rises steadily during the period, and DI receipt rises with it. Because at the same time mortality decreases steadily, a negative correlation is the result. For health, we only have a crude measure (the percentage of people with self- reported health good or better) for a broad age group (forty- five to

Disability Insurance and Labor Market Exit Routes in the Netherlands

Fig. 12.22

Male mortality versus disability, aged forty to forty-four

Fig. 12.23

Female mortality versus disability, aged forty to forty-four

439

sixty- four). For men, there is some indication that an improvement of the health measure coincides with a decrease in DI receipt. For women, there is no clear trend in the health measure, while DI receipt tends to increase slightly (see figures 12.26 and 12.27). A final perspective is given by figure 12.28, which shows the difference between DI receipt in the age groups sixty to sixty- four and forty to fortyfour for men and women. Notably, for men, this difference is much higher than for women throughout the whole period. This probably does not really mean that disability prevalence increases much faster with age among men

440

Klaas de Vos, Arie Kapteyn, and Adriaan Kalwij

Fig. 12.24

Male mortality versus disability, aged sixty to sixty-four

Fig. 12.25

Female mortality versus disability, aged sixty to sixty-four

than among women. More likely it is caused by the fact that women who left the labor market relying on the income of their spouse have no access to DI. The declining difference between men and women is indicative of the fact that more and more older women stay on the labor market. The fact that the difference between forty to forty- four- year- old men and sixty to sixty- fouryear- old men decreases at the end of the period suggests that the earlier, larger difference between these age groups is not really the result of older

Fig. 12.26

Male self-rated health versus disability

Fig. 12.27

Female self-rated health versus disability

Fig. 12.28 Male and female disability, aged sixty to sixty-four less aged forty to forty-four

442

Klaas de Vos, Arie Kapteyn, and Adriaan Kalwij

people having a lower ability to work but is largely caused by the fact that it used to be easy to get access to DI for older men who wanted or were forced to leave the labor market before the statutory retirement age of sixty- five. In summary, health and mortality do not seem to be strongly related to DI prevalence. At the end of the period, the reforms appear to be successful in causing a decrease in DI prevalence. In the next section, we will see if this result can be confirmed by a more rigorous quantitative analysis. 12.6

An Explorative Empirical Analysis

The empirical evidence presented in the preceding sections seems to suggest that DI receipt is unrelated to the general health of the population and that DI reforms may have led to relatively fewer older workers exiting the labor market through DI. In this section, we empirically investigate these issues using a regression model. Such an approach makes it possible to control for cyclical variation in the labor market exit rate and to quantify the impacts of health and policy reforms on the various pathways to retirement conditional on age and gender. The policy reforms are modeled using yeardummy variables indicating the reforms that have also been highlighted in the figures in the previous sections (see also table 12.1). The data are taken from the Income Panel Study of the Netherlands (see section 12.4). We select individuals who are fifty to sixty- three years of age and aggregate these data based on year, gender, and age category (fifty to fifty- four, fifty- five to fifty- nine, and sixty to sixty- three).3 Over the time period 1989 to 2006, we have 108 observations. 12.6.1

Empirical Framework: A Simple Pathway Model

Individuals can exit employment through the pathways Disability Insurance (DI), early retirement (ER), Unemployment Insurance (UI), or nonparticipation (NP). With individual level data, this can be modeled using a competing risk model. We use, however, aggregated data to carry out the empirical analysis. This means we do not have individual pathways to retirement but proportions of (groups of) people that choose certain pathways. c We define pagt as the proportion of employed individuals with age a and gender g that leave employment through pathway c in year t. For a logittype model for each pathway, we can specify the following linear regression model: c k ⎛ pagt ⎞ ln ⎜ + =  ⎟ ∑ cj D jt + 0c X ag + 1c Hagt + 2c ln(Eagt ) + εcagt , 0c c 1− p j =1 ⎝ agt ⎠

3. We exclude sixty- four- year- old persons as most often their employment contract is automatically terminated when they reach the statutory retirement age of sixty- five and they start receiving a state pension.

Disability Insurance and Labor Market Exit Routes in the Netherlands

443

where c ∈ (DI, ER, UI, NP), t ∈ {1989, . . , 2006},4 the Djt s are dummy variables indicating reforms (see table 12.1), Xag are explanatory variables (gender and age), Hagt is remaining healthy life expectancy,5 and Eagt is the total outflow from employment. In the estimation, we use the pathway nonparticipation (NP) as the reference category.6 The error term εcagt is heteroskedastic, and this is taken into account when calculating the standard errors. For details on the estimation, we refer to Cameron and Trivedi (2005). We use an instrumental variables estimator to take into account that the total outflow from employment Eagt may, for instance, be influenced by unobserved variables also affecting a specific pathway.7 The additional instrument is total employment by age, gender, and year obtained from OECD statistics. The justification for choosing this additional instrument is that the outflow from employment will be positively related with total employment because the more people are employed, the more people are at risk of leaving employment, while we do not expect total employment to affect the different pathways once we control for total outflow from employment.8 12.6.2

Empirical Results

The first column of table 12.2 presents the estimation results of a linear regression model explaining the (logarithm of) total flow from employment. The remaining columns present estimation results for the model discussed in the preceding for each of the pathways. The total outflow from employment increases with age and is higher among women than men. The total outflow appears unrelated to health. The DI reform in 1996 and the reforms in 2006 are significantly and negatively associated with the outflow from employment, but there is no significant association between the other reforms and the outflow from employment. Conditional on the total flow from employment and compared to male workers, female workers are less likely to exit employment through DI, ER, and UI and more likely to exit employment without own source of replacement income. With age, ER becomes relatively more important, and DI and UI become less important exit routes. In line with our conclusions drawn from the various figures in the preceding sections, we find no significant relationship between healthy life expectancy and the pathways to retirement. 4. Outflow is from t – > t + 1. We do not observe the outflow in 1981 and 1985 because the subsequent years are not in our data. Hence, 1989 is our first year, and 2006 is the last year in our “flow” sample. 5. Remaining healthy life expectancy is defined as the number of years a person of certain age and in a particular year is expected to live in good health where good health is defined as absence of chronic disease, absence of physical limitations, or perceived good health. 6. We use the normalization pNagtP = 1 – pDagtI – paEgtR – pUagtI . 7. Another possible bias may come from measurement error in IPO data. For this reason, we use an instrument based on OECD data. 8. We follow Bound, Jaeger, and Baker (1995) and present a test for whether the instrument has sufficient explanatory power (see table 12.2).

Table 12.2

Estimation results ln(P(pathway c)/(1-P(pathway c))) Ln(total outflow from employment)

Intercept Woman Age 55–59a Age 60–63a ln(healthy life expectancy) 1993, Stricter DI criteria 1996, DI, 1 year sickness benefits 1998, Experience rating employer 2000, ER actuarially fairer 2002, DI, gatekeeper improvement 2004, DI, 2 years sickness benefits 2006, new DI, ER/UI changes Ln(level of employment)b

0.279** (3.54) 0.307** (2.38) 0.745** (3.38) 1.493** (2.56) –0.47 (–0.54) 0.055 (0.58) –0.241** (–2.52) 0.112 (1.13) –0.073 (–0.68) –0.056 (–0.64) 0.022 (0.27) –0.292* (–1.70) 0.896** (4.62)

Ln(total outflow from employment) Exogeneity test (p-value)c Test for the explanatory power of the additional instrument (p-value)d R2 No. of observations

Disability Insurance

Early Retirement

–0.745** (–4.20) –0.688** (–5.48) –0.568** (–2.13) –1.078 (–1.55) 1.25 (0.68) –0.433** (–2.58) 0.234 (1.54) 0.386** (2.80) –0.297** (–2.36) –0.345** (–2.44) 0.031 (0.20) –0.259 (–1.50)

–1.683** (–13.36) –0.468** (–4.26) 1.398** (8.39) 3.021** (7.01) 0.892 (0.81) –0.317** (–2.72) 0.091 (0.77) –0.285** (–2.11) 0.231* –1.8 –0.008 (–0.07) –0.083 (–0.71) 0.114 (0.85)

0.221 (0.63) 0.049**

0.769** (2.28) 0.779

Unemployment Insurance –0.792** (–3.30) –0.477** (–3.11) –0.496* (–1.73) –2.163** (–2.52) –0.732 (–0.29) 0.637** (3.75) –0.364** (–2.75) –0.652* (–1.75) 0.22 –1.06 0.256 (1.53) 0.089 (0.50) 0.051 (0.32)

–0.003 (–0.00) 0.573

0.000** 0.656 108

0.755 108

0.945 108

0.76 108

Note: In parentheses are t-values. a Reference age 50–54. The age dummies are jointly significant (p-value F-test is 0.001). b OECD statistics, by age and gender. c H0: total outflow from employment is exogenous. d The null-hypothesis is that the additional instrument Ln(level of employment) has no explanatory power. **p < 0.05. *p < 0.10.

Disability Insurance and Labor Market Exit Routes in the Netherlands

445

The 1993 and 2002 DI reforms are associated with a lower exit probability into DI, while the 1998 reform (the introduction of experience rating) appears to have affected the exit probability to DI positively. We have no explanation for this latter finding, but the termination of experience rating in 2008 hints toward this indeed being an ineffective instrument for reducing DI inflow. A counterintuitive finding is the positive effect of the 2000 reform on ER and its negative impact on DI inflow. This may be the result of workers (and employers) making more use of ER in anticipation of further limitations on ER eligibility. In most cases, the coefficients of the reforms on the estimated probability to exit via unemployment have the opposite sign as the ones found for the probability to exit to DI, suggesting a possible substitution between these two routes. To obtain further insights into these empirical results, table 12.3 presents the predicted percentage changes in the total outflow and the pathway probabilities due to a change in the explanatory variables. The total outflow from employment is 36 percent higher among women than men (first row). The outflow from employment more than quadruples for workers aged sixty to sixty- three compared to workers aged fifty to fifty- four, and, with age, workers are increasingly more likely to flow into early retirement than into disability or unemployment (rows [2] and [3]). Again, health appears to have no effect on the total outflow from employment and the pathways to retirement (row [4]). Total outflow from employment is not significantly associated with the outflow from employment into DI but is positively associated with the flow into ER. This suggests that when the total outflow from employment is high, for instance, as a result of a downturn in the economy, the exit route ER is relatively more frequently used. Finally, the last row suggests that, cumulatively, the reforms have had substantial effects: they have reduced the total outflow from employment with 38 percent and, moreover, reduced the shares of the pathways into DI and ER with, respectively, 40 percent and 20 percent at the expense of an increase in the outflows into UI and nonparticipation (“Other”) with, respectively, 17 percent and 53 percent. 12.7

Conclusion

The relatively generous and easily accessible DI benefit can be seen as an important institution contributing to the large numbers of workers retiring from the labor market at relatively early ages in the Netherlands. There are no indications that the high numbers of DI are related to high mortality rates or high percentages of persons living in bad health. By a continuous series of reforms, the government has attempted to reverse the trend of ever expanding numbers of DI recipients. Until recently, these reforms did not result in much of a decrease in the number of DI recipients.

2.05 2.38 1.33 –0.55

–2.33

35.91 110.69 345.17

–4.60

–37.70

t-value

0.65 0.62 –2.61

1.51 –39.97

–6.72 –2.53 –2.25

t-value

8.66

–40.17 –34.16 –56.81

Prediction (% change)

32.84%

Disability Insurance

Men aged 50–54 in 1989. The reforms listed in table 12.1 and included in the analysis (table 12.2).

b

a

Women relative to men Age 55–59, relative to age 50–54 Age 60–63, relative to age 50–54 10% increase in healthy life expectancy 10% increase in the flow out of employment Cumulative effect of the reformsb

Prediction (% change)

3.46%

Total outflow from employment

Predicted changes in the pathway probabilities

Probability of outflow for the reference groupa

Table 12.3

6.66 –19.87

7.76

–33.49 173.84 405.39

Prediction (% change)

15.23%

2.14 –1.12

0.78

–5.27 5.96 5.43

t-value

Early Retirement

31.29%

–0.02 16.92

–4.96

–29.60 –30.64 –84.11

0.00 0.50

–0.30

–3.50 –2.05 –6.88

t-value

Unemployment Insurance

Prediction (% change)

Pathway

–7.28 52.63

–11.98

133.54 –27.47 –81.22

Prediction (% change)

20.64%

–3.10 1.54

0.79

5.25 –0.13 –0.90

t-value

Nonparticipation

Disability Insurance and Labor Market Exit Routes in the Netherlands

447

Despite the reforms, DI remained attractive and accessible. However, it is likely that without reforms, the number of DI recipients would have kept on increasing much further. The empirical analysis shows that several reforms to the DI, ER, and UI programs are related to the pathways to retirement. Notably, these reforms are found to be associated with reductions in the labor market exit rate of older workers and the percentage inflow into DI and ER. This has, arguably, contributed to the observed increased labor force participation rate among the fifty- five to sixty- four population from the mid- 1990s onward.

References Aarts, L. J. M., and P. R. de Jong. 1990. “Economic Aspects of Disability Behaviour.” PhD diss., Erasmus University, Rotterdam. Bound, J., D. A. Jaeger, and R. M. Baker. 1995. “Problems with Instrumental Variables Estimation When the Correlation between the Instruments and the Endogenous Explanatory Variable Is Weak.” Journal of the American Statistical Association 90:443– 50. Cameron, A. C., and P. K. Trivedi. 2005. Microeconometrics, Methods and Applications. New York: Cambridge University Press. Centraal Bureau voor de Statistiek (CBS). 2009. Documentatierapport Inkomenspanelonderzoek (IPO) (Documentation Income Panel Survey). Voorburg, the Netherlands: Centrum voor Beleidsstatistiek. De Gier, E., R. Henke, and J. Vijgen. 2003. “The Dutch Disability Insurance Act (WAO) and the Role of Research in Policy Change.” Amsterdam School for Social Science Research Working Paper no. 03/02. Kapteyn, A., and K. de Vos. 1999. “Social Security and Retirement in the Netherlands.” In Social Security and Retirement around the World, edited by Jonathan Gruber and David A. Wise, 269– 304. Chicago: University of Chicago Press. Kapteyn, A., K. de Vos, and A. Kalwij. 2010. “Early Retirement and Employment of the Young in the Netherlands.” In Social Security Programs and Retirement around the World: The Relationship to Youth Employment, edited by Jonathan Gruber and David A. Wise, 243– 59. Chicago: Chicago University Press. Knoef, M., R. J. M. Alessie, and A. Kalwij. 2010. “Changes in the Income Distribution of the Dutch Elderly between 1989– 2020: A Dynamic Microsimulation.” Netspar Working Paper. Stichting van de Arbeid. 2010. Pension Accord Spring 2010 (Pensioen Akkoord). www .stvda.nl.

Contributors

Michael Baker Department of Economics University of Toronto 150 St. George Street Toronto, ON M5S 3G7, Canada

Richard Blundell Department of Economics University College London Gower Street London WC1E 6BT, England

James Banks School of Social Sciences The University of Manchester Manchester M13 9PL, England

Axel Börsch-Supan Mannheim Research Institute for the Economics of Aging University of Mannheim L13, 17 68131 Mannheim, Germany

Luc Behaghel Paris School of Economics, INRA and CREST 48, Bd Jourdan 75014 Paris, France Paul Bingley SFI-The Danish National Center for Social Research Hwerluf Trolles Gade 11 1052 Copenhagen, K., Denmark Didier Blanchet CREST and INSEE Timbre H230 18 Bd A. Pinard 75675 Paris Cedex 14, France

Antoine Bozio Institute for Fiscal Studies 7 Ridgmount Street London WC1E 7AE, England Agar Brugiavini Department of Economics Università Ca’ Foscari, Venice Cannaregio, Fondamenta S. Giobbe, 873 30121 Venice, Italy Thierry Debrand IRDES, Institute for Research and Information in Health Economics 10 Rue Vauvenargues 75018 Paris, France

449

450

Contributors

Klaas de Vos CentERdata P.O. Box 90153 5000 LE Tilburg, The Netherlands

Arie Kapteyn RAND Corporation 1776 Main Street, P.O. Box 2138 Santa Monica, CA 90407-2138

Carl Emmerson The Institute for Fiscal Studies 7 Ridgmount Street London WC1E 7AE, England

Mathieu Lefebvre CREPP University of Liège 7, Bd Rectorat (B31) Liège 4000, Belgium

Pilar García-Gómez Department of Applied Economics Erasmus School of Economics Erasmus University Rotterdam P.O. Box 1738 3000 DR Rotterdam, The Netherlands Nabanita Datta Gupta Department of Economics and Business Aarhus University Frichshuset, Hermodsvej 22 8230 Aabyhoej, Denmark Sergi Jiménez-Martín Department of Economics Universitat Pompeu Fabra Ramon Trias Fargas 25-27 08005 Barcelona, Spain Lisa Jönsson Department of Economics Stockholm University SE- 106 91 Stockholm, Sweden Alain Jousten Department of Economics University of Liège 7, Bd Rectorat (B31) Liège 4000, Belgium Hendrik Jürges Schumpeter School of Economics and Business University of Wuppertal Rainer-Gruenter-Str. 21 (FN.01) D-42119 Wuppertal, Germany Adriaan Kalwij School of Economics Utrecht University P.O. Box 80125 3508 TC Utrecht, The Netherlands

Kevin Milligan Department of Economics University of British Columbia #997-1873 East Mall Vancouver, BC V6T 1Z1, Canada Takashi Oshio Institute of Economic Research Hitotsubashi University 2-1 Naka, Kunitachi City Tokyo 186-8603, Japan Mårten Palme Department of Economics Stockholm University SE- 106 91 Stockholm, Sweden Peder J. Pedersen Department of Economics and Business Aarhus University 8000 Aarhus, C., Denmark Franco Peracchi Faculty of Economics University of Rome “Tor Vergata” via Columbia 2 00133 Rome, Italy Sergio Perelman Department of Economics University of Liège 7, Bd Rectorat (B31) Liège 4000, Belgium Muriel Roger Paris School of Economics 48, Bd Jourdan 75014 Paris, France

Contributors Satoshi Shimizutani Institute for International Policy Studies Toranomon 30 Mori Bldg., 6F 3-2-2 Toranomon, Minato- ku Tokyo 105-0001, Japan

Judit Vall Castelló Research Centre for Economy and Health (CRES) Universitat Pompeu Fabra Plaça de la Mercè, 10-12 08002 Barcelona, Spain

Ingemar Svensson Swedish Pensions Agency Box 38190 100 64 Stockholm, Sweden

David A. Wise Kennedy School of Government Harvard University 79 John F. Kennedy Street Cambridge, MA 02138

451

Author Index

Page numbers followed by the letter f refer to figures. Aarts, L. J. M., 277, 419 Adam, S., 48, 72, 73 A’Hearn, B., 210 Alessie, R. J. M., 424 Andreyeva, T., 148 Anyadike-Danes, M., 47f Attanasio, O. P., 200 Autor, D., 128, 218, 361n7, 362n8, 363, 364, 374 Avendano, M., 278 Baker, R. M., 443 Banks, J., 42 Bardot, F., 302 Barnay, T., 302 Benítez-Silva, H., 42, 128 Ben Salem, M., 301, 304 Berkowitz, E. D., 361n7 Beveridge, Sir W., 44 Bhattacharya, J., 145 Black, D., 218 Blanchet, D., 302 Boeri, T., 203 Boldrin, M., 138n5 Börsch-Supan, A., 277, 278, 280n1, 281, 282, 283n5, 290, 291, 299, 391 Bottazzi, R., 200 Bound, J., 42, 43n1, 71, 364, 443n8 Bozio, A., 48, 72, 73, 303, 311, 312 Bratsberg, B., 218, 219

Brugiavini, A., 177, 178n5, 191, 195, 200, 203, 208 Burchardt, T., 43 Burkhauser, R., 43n1, 128, 277 Cameron, A. C., 443 Campoliteti, M., 218, 328n1, 331 Caselli, G., 189 Choudhry, K., 145 Creedy, J., 43, 44n5, 46 Cremer, H., 258 Crimmins, E. M., 145 Cutler, D. M., 142n9 Daly, M., 128 Daniel, K., 218 Debrand, T., 302 De Gier, E., 419 De Jong, P. R., 277, 419 Dellis, A., 251n1, 252 Desmet, R., 251n1, 252 De Vos, K., 419, 423 Disney, R., 42, 43, 44n5, 46, 71, 128, 251n1 Doherty, N., 42 Duggan, M., 128, 218, 361n7, 362n8, 363, 364, 374, 377 Egidi, V., 189 Emmerson, C., 42, 48, 71, 72, 73, 251n1

453

454

Author Index

Faniel, J., 257 Fenn, P., 42 Fevang, E., 219 Fries, J. F., 144, 144n10 Fuchs, V., 142n9 Galasso, V., 178n5 García-Gómez, P., 138n5 Gilbert, B., 44n5 Goldman, D. P., 145 Gruber, J., 2, 41, 79, 128, 218, 330, 363, 364 Hashimoto, H., 403 Hedström, P., 80 Henke, R., 419 Ichimura, H., 403 Imberman, S. A., 363, 364, 377 Jaeger, D. A., 443 Jagger, C., 320 Jappelli, T., 200 Jeger, F., 302 Jiménez-Martin, S., 42, 128, 138n5 Johnston, D. W., 148 Jousten, A., 252, 257 Kalwij, A., 251n1, 423, 424 Kapteyn, A., 42, 419, 423 Karlström, A., 80, 121, 122 Knoef, M., 424 Kruse, A., 80 Kubik, J. D., 363, 364 Kusnik-Joinville, O., 312

Oishi, A., 393, 395 Oshio, T., 393n1, 395 Padula, M., 200 Palme, M., 80, 81, 121, 122 Pedersen, P. J., 238 Peracchi, F., 177, 178n5, 191, 195, 208, 210 Persson, M., 81 Pestieau, P., 251n1, 258 Piachaud, D., 42 Prince, M. J., 328n1 Propper, C., 148 Reynaud, D., 312 Reynolds, S. L., 145 Røed, K., 219 Roth, H., 277, 278 Saito, Y., 145 Sanders, S., 218 Schirle, T., 376–77 Sheiner, L., 142n9 Shields, M. A., 148 Shimizutani, S., 393, 395, 403 Shoven, J., 66, 142, 142n9, 158 Skogman Thoursie, P., 80 Smith, J., 42 Söderström, L., 80 Stijns, J. P., 251n1 Struillou, Y., 302 Svensson, I., 80, 121, 122 Torjman, S., 328n1 Trivedi, P. K., 443

Labeaga, J. M., 128 Lakdawalla, D. N., 145 Larsen, M., 238 Lavis, J. N., 328n1 Lê, F., 312 Lindbeck, A., 81 Lozachmeur, J. M., 258

Vall Castelló, J., 128 Van Soest, A., 42, 148 Vecchi, G., 210 Vermeulen, F., 251n1 Vijgen, J., 419 Vilaplana Prieto, C., 128 Volkoff, S., 302

Mackenbach, J., 145, 278 McVicar, D., 47f, 128 Michaud, P. C., 148 Molinié, A.-F., 302 Momose, Y., 397n6, 399n10 Myers, R. J., 42

Wadensjö, E., 80 Wakefield, M., 42, 71, 251n1 Walker, A., 44 Walker, L., 44 Webb, S., 42, 128 Wilke, C. B., 280n1, 283n5 Wise, D. A., 2, 41, 79, 128, 177

Nagi, S., 42

Subject Index

Page numbers followed by the letter f or t refer to figures or tables, respectively. Age groups: Dutch employment rates by, 431–32, 432f; Dutch labor force data by, 434–37, 435–37f Aging, population, 1–2 Attendance Allowance (AA) (UK), 48 Belgium: conventional early retirement (CER) in, 256–57; conventional early retirement (CER) participation in, 269–70; disability insurance (DI) participation in, 265–67, 270–73, 274f, 275f; disability insurance (DI) system in, 251–54; distribution of exit paths in, 270, 271f; historical health trends in, 258–63; Intergenerational Solidarity Pact, 256; labor market indicators, 263– 65; “natural experiments” in DI reform and DI participation in, 32–33; old-age unemployment (OAU) participation in, 268–69; old-age unemployment (OAU) system in, 257–58; participation in different systems in, 270–73, 272f; pension system in, 254–56; unemployment insurance participation in, 268; unemployment insurance system in, 257–58 Beveridge report (UK), 44 Body mass indexes (BMIs), in France, 323–25 Britain. See United Kingdom (UK)

British Household Panel Survey (BHPS), 49, 53–56 Canada: development of disability insurance in, 328–29; “natural experiments” in DI reform and DI participation in, 30; trends in mortality and DI participation in, 28f Canada Pension Plan (CPP), 328 Canada Pension Plan (CPP) Disability Insurance program, 328–29; data, 332– 33; employment and mortality, 350– 51; employment/unemployment and receipt of disability, 344–50; growth of, 330, 331f; health and receipt of disability, 351–57; labor markets and, 344–50; mortality and health results, 333–38; program participation results, 338–44; reforms, 329–30; research on, 330–31 Demographic trends, pay-as-you-go social security systems and, 1–2 Denmark: aggregate trends in, 218–21; employment rates in, 243–46; Flex Job Program, 232–33; labor force participation in, 241–42, 244, 246f, 247f, 248f; mortality rates vs. other health measures in, 23, 24f; mortality trends in, 221–25; “natural experiments” in DI

455

456

Subject Index

Denmark (continued) reform and DI participation in, 36–37; population health trends in, 225–29; Post Employment Wage (PEW) program, 233–35; Social Disability Pension (SDP) in, 217–18, 229–32; Temporary Benefits Program (TBP), 234, 235–40; unemployment rates for, 241–43 Disability insurance (DI): departure of older persons from labor force and, 1; early retirement benefits and, 2; “natural experiments” in reform and participation, 28–37; participation vs. mortality, 26–28; phases of study of, 2–3. See also Reforms, disability insurance; and specific countries Disability insurance (DI) participation: in Belgium, 265–67, 270–73, 274f, 275f; in Canada, 328–29; in Sweden, 94–97 Disability insurance (DI) programs: in Belgium, 241–54; in Canada, 328–33; challenges of cross-country comparison of, 5; in Germany, 277–78; in Italy, 209–11; in Japan, 408, 411–16; in the Netherlands, 420–23, 422t, 430–31, 437–42; in Quebec, 330–31; in Spain, 127–28, 130– 38, 137t; in Sweden, 79, 81–83, 94–97, 108–14, 117–24; in United Kingdom, 45t; in United States, 360–64 Disability insurance (DI) systems. See Disability insurance (DI) programs Disability Living Allowance (DLA) (UK), 49 Disability pension programs, in Japan, 391–99, 394t Early childhood environment, and German very early retirement, case study of, 297–98 Employment: in Denmark, 243–46; mortality and, 7–9, 11–17; mortality in Canada and, 351–52; mortality in France and, 14, 15f; mortality in United Kingdom and, 14–17, 16f; in the Netherlands, by age groups, 431–32, 432f; of older males in United Kingdom, 50–51, 50f, 51f; among older workers in Sweden, 79–80; receipt of disability in Canada and, 334–50; receipt of disability in United States and, 376–83. See also Labor force participation; Unemployment

Employment Support Allowance (ESA), 48 England. See United Kingdom (UK) Family Expenditure Survey (FES), 49–50 Flex Job Program (Denmark), 232–33 France: body mass indexes, 323–25; creation of social security system, 308–9; employment and mortality in, 14, 15f; general labor force participation trends, 304–6; impact of tighter control of sickness benefits, 312–18; mortality rates and life expectancy, 318–21; “natural experiments” in DI reform and DI participation in, 33–36; 1993 pension reform, 311–12; pension d’inaptitude, 308–10; role of disability benefits and early retirement in, 301–2; self-reported health status, 321–23; trends in role of disability benefits, 306–10 Germany: case study of early childhood environment and very early retirement, 297–98; disability insurance in, 277–78; Hartz reforms, 283–84; health measure trends, 288–90; labor force participation trends, 290–91; mortality and labor force participation, 291–94; mortality trends, 285–88; “natural experiments” in DI reform and DI participation in, 30–32; 1972 reform and effect on retirement age, 294–95; 1978 reform and effect on retirement age, 295–96; 1984 reform and effect on retirement age, 296; pathways to retirement in, 279–80; pension systems in, 278–79; regimes of retirement policies in, 280–84; Riester reform, 282–83 Great Britain. See United Kingdom (UK) Hartz reforms (Germany), 283–84 Health measures: cross-country comparisons and, 5; mortality vs. other, 23–26; self-assessed, 5, 17–23; self-reported, 86–89, 321–23; and UK labor force participation, 61–70 Health trends: in Belgium, 258–63; in Denmark, 225–29; in Germany, 288–90; in the Netherlands, 425–427, 437–42; in Spain, 144–50 Height: disability rates in Italy and, 210–11, 211f; early retirement rates in Italy and, 211

Subject Index Holland. See Netherlands, the Housewife Non-Contributory Invalidity Pension (HNCIP) (UK), 48 Housing Benefit (HB) (UK), 49 Incapacity Benefit (IB) (UK), 46 Income Support (IS) (UK), 49 Industrial Injuries Disablement Benefit (IIDB), 44, 44n2, 44n3 International Social Security Project, 1 Invalid Care Allowance (ICA) (UK), 48–49 Invalidity Benefit (IVB), 45–46 Italy: effects of pension reforms on labor force participation, 198–205; height and disability rates in, 210–11, 211f; height and early retirement rates and, 211; historical mortality data, 183–89; history of social security reforms in, 176–83; labor force participation rates in, 195– 98; labor force status and health conditions in, 189–95; mortality rates vs. other health measures in, 23, 24f; pathways to retirement in, 205–6; trends in labor force status, 207–8; trends in mortality and DI participation in, 28f; trends in recipiency of disability insurance benefits in, 209–11 Japan: current scheme of disability pension program, 396–99; development of disability pension program in, 394t; disability pension programs in, 391–92; Employees’ Pension Insurance (EPI), 393–94, 393n3; health indicators for men in, 23, 26f; historical data on disability insurance participation vs. mortality and health, 411–16; historical disability insurance data, 408; historical health data, 399–407; historical labor force data, 403; historical overview of disability pension programs in, 392–96; labor force participation vs. mortality, 403–7; mortality trends, 400, 401f, 402f, 403f; Mutual Aid Insurance (MAI), 393; National Pension Insurance (NPI), 393, 394–95; other programs for disabled in, 399; self-rated health measures, 401–3 Labor force participation: in Denmark, 241–42, 244, 246f, 247f, 248f; vs. mortality, in Japan, 403–7; in the Nether-

457

lands, 427–29, 434–37, 435–37f; rates, in Italy, 195–98; in Spain, 151–55, 157– 60; trends, in France, 304–6; trends, in Germany, 290–94; in United Kingdom, 61–70. See also Employment Labor markets. See Employment; Unemployment Labour Force Survey (LFS), 49–50 Life expectancy. See Mortality Mobility Allowance (MA), 48 Mortality: by age, 9–11; in Canada, 28f, 333–38, 350–51; decline in, 6–7; in Denmark, 23, 24f, 221–25; disability insurance participation vs., 26–28; by employment, 11–17; employment and, 7–9; in France, 14, 15f, 318–21; in Germany, 285–88, 291–94; in Italy, 23, 24f, 28f, 183–89; in Japan, 400, 401f, 402f, 403–7, 403f, 411–16; long-run trends in, 5–6; in the Netherlands, 425–29, 437– 42; other health measures vs., 23–26; presenting data on, 7–8; self-assessed health (SAH) vs., 17–23; in Spain, 23, 25f, 142–44, 150–51, 155–60; in Sweden, 23, 25f, 29f, 84–86; in United Kingdom, 14–17, 16f, 27f; UK data, 56–59; in United States, 29f, 365–70, 383–85, 383f, 384 National Insurance Act (1911) (UK), 44n5 Netherlands, the: benefit system, 423; disability insurance in, 420; disability insurance reforms, 420–23, 430–31; disability insurance timeline, 422t; Dutch Unemployment Insurance system reforms, 424; empirical analysis and results for retirement pathway in, 442– 45; employment rates by age groups in, 431–32, 432f; historical data on disability insurance vs. health, mortality, 437–42; historical mortality and health data, 425–27; historical mortality data for, 425–26, 425–26f; labor force data by age groups for, 434–37, 435–37f; labor force participation and disability vs. mortality data for, 427–29; “natural experiments” in DI reform and DI participation in, 36–37; pathways to retirement in, 429–30, 433–34 Non-Contributory Invalidity Pension (NCIP) (UK), 48

458

Subject Index

Pathways to retirement: in Germany, 279– 80; in Italy, 205–6; in the Netherlands, 429–30, 433–34; in Spain, 160–67; in Sweden, 105–7; in United Kingdom, 41–43. See also Retirement Pathways-to-Work program (UK), 48, 50, 70, 72–74 Pay-as-you-go social security systems, pressure of demographic trends on, 1–2 Pensions. See Disability insurance (DI) programs; Retirement; and specific countries Personal Capability Assessment, 46–47 Population aging, 1–2 Post Employment Wage (PEW) program (Denmark), 233–35 Quebec Pension Plan (QPP), 328, 329n5. See also Canada Pension Plan (CPP) Quebec Pension Plan (QPP) Disability Insurance program, 329; research on, 330–31. See also Canada Pension Plan (CPP) Disability Insurance program Reforms, disability insurance: in Belgium, 32–33; in Canada, 30, 329–30; in Denmark, 36–37; in France, 33–36, 311–12; in Germany, 30–32, 282–84, 294–96; in Italy, 176–83, 198–205; in the Netherlands, 36–37, 420–23, 430–31; in Spain, 137–38, 137t; in Sweden, 33; in United Kingdom, 45t, 70–74. See also Disability insurance (DI) Reforms, unemployment system: in the Netherlands, 424; in Spain, 138–41. See also Unemployment Riester reform (Germany), 282–83 Retirement: in Belgium, 256–57, 269–70; early benefits and disability insurance, 2; in France, 301–2; in Germany, 279– 84, 294–98; in Italy, 205–6, 211; in the Netherlands, 429–30, 433–34, 442–45; in Spain, 160–67; in Sweden, 105–7; in United Kingdom, 41–43. See also Pathways to retirement Self-assessed health (SAH) status, 5; mortality vs., 17–23. See also Health measures Self-reported health: in France, 321–23; in Sweden, 86–89. See also Health measures

Severe Disablement Allowance (SDA) (UK), 48 Sickness Benefit (UK), 44–45 Social Disability Pension (SDP) (Denmark), 217–18, 229–32. See also Denmark Social Security Disability Insurance (SSDI) (US): current rule benefits for, 360–61; data sources of analysis of, 364–65; employment/unemployment rates and receipt of, 376–83; health and, 385–88, 386–88f; historical development of, 361–64; mortality rates and, 383–85, 383f, 384f; mortality trends and, 29f; participation, 370–73; program findings, 359–60; research on impact of, 364. See also United States (US) Social security systems, pay-as-you-go, pressures of demographic trends on, 1–2 Spain: disability, health, and mortality data, 155–57; disability and labor force participation trends in, 151–55; disability insurance system in, 130–36; disability programs in, 127–28; health, mortality, and labor force participation data, 157– 60; health trends in, 144–50; mortality and health in, 150–51; mortality rates in, 142–44; mortality rates vs. other health measures in, 23, 25f; pathways to retirement in, 160–67; reforms of disability system in, 137–38, 137t; reforms of old-age and unemployment systems in, 138–41, 139t; tentative analysis of quantitative effects of reforms in, 168–70 SSDI. See Social Security Disability Insurance (SSDI) (US) Statutory Sick Pay (SSP) (UK), 46, 46n6 Survey of Living Conditions (ULF) (Sweden), 86–89 Sweden: changes in disability insurance eligibility in, 117–24; development of disability insurance recipiency in, 94–97; development of labor market outcomes in, 97–104; development of population health in, 83–84, 91–94; disability insurance and labor market outcomes in, 107–8; disability insurance in, 79; disability insurance incidence and population health in, 110–14; disability insurance prevalence and population health in, 108–10; employment rate among

Subject Index older workers in, 79–80; inpatient care in, 89–91; mortality rates in, 84–86; mortality rates vs. other health measures in, 23, 25f; “natural experiments” in DI reform and DI participation in, 33; overview of disability insurance in, 81–83; pathways to retirement in, 105–7; relative health of disability insurance recipients compared to nonrecipients in, 114–15; self-reported health in, 86– 89; trends in mortality and DI participation in, 29f Temporary Benefits Program (TBP) (Denmark), 234, 235–40 Unemployment: in Denmark, 241–342; insurance system in Belgium, 257–58, 268f–69; receipt of disability in Canada and, 334–50; receipt of disability in United States and, 376–83; systems in Spain, 138–41, 139t. See also Employment United Kingdom (UK): BHPS data on transitions in, 53–56; disability and sickness insurance in, 44–48; disability benefits as pathway to retirement in, 41–43; disability benefits reforms in, 70–74; earnings replacement benefits in, 44–48; employment and mortality in, 14–17, 16f; employment rate of older males in, 50–51, 50f, 51f; health measures and labor force participation in, 61–70; means-tested benefits in, 49; measures of self-reported disability, 59–61; mor-

459

tality data, 56–59; 1995 disability benefits reform, 70–72; non-contributory benefits in, 48–49; one-year transitions from LFS, 51–53; reforms to disability insurance system of (1948 to present day), 45t; structure of benefits in, by duration of incapacity (1948–2010), 45t; surveys on participation in labor market in, 49–51; trends in mortality and DI participation in, 27f; workrelated injury benefits in, 43–44 United States (US): analysis of mortality, 365–70; current benefit rules for Social Security Disability Insurance, 360–61; data sources for analysis of Social Security Disability Insurance, 364–65; employment/unemployment rates and receipt of Social Security Disability Insurance, 376–83; historical development of Social Security Disability Insurance, 361–64; research on impact of Social Security Disability Insurance, 364; Social Security Disability Insurance and health, 385–88, 386–88f; Social Security Disability Insurance and mortality rates, 383–85, 383f, 384f; Social Security Disability Insurance participation, 370–76; Social Security Disability Insurance program findings, 359–60; trends in mortality and DI participation in, 29f Work Capacity Assessment (UK), 48 Working Tax Credit (WTC) (UK), 49