October 2008, Number 8-17 ARE OLDER MEN HEALTHY ENOUGH TO WORK? By Alicia H. Munnell, Mauricio Soto, and Alex Golub-Sass* Introduction Since the mid-1960s, the median retirement age for This brief uses the National Health Interview Survey men has declined from 66 to 63. If Americans con- to estimate trends in disability-free life expectancy for tinue to retire at age 63, a great many will risk income men at age 50. The first section calculates trends in shortfalls, especially at older ages. This risk is even disability-free life expectancy for the population as a greater for those currently nearing retirement who whole, revealing an increase between 1970 and 2000 have recently seen a large portion of their nest eggs of almost three years. The second section estimates evaporate. the trends in disability-free life expectancy by race and Work directly increases current income, Social Se- educational attainment, showing that the three-year- curity benefits, and retirement saving, and decreases increase is attributable primarily to movement up the the length of retirement. But are Americans healthy education ladder, with minimal increases within edu- enough to work longer? Life expectancy has been cational groups. Moreover, major disparities remain steadily increasing, but disparities in health and mor- between those in the bottom and top quartiles of the tality outcomes have widened and the improvement population. The third section looks to the future, in health outcomes for the population in general may suggesting that the improvement in health outcomes have slowed or even reversed. for the population in general may have slowed or even In determining whether people will be able to reversed and that increases in educational attainment work longer, it is not simply measuring how long they may have ceased. The final section concludes that the will live, but rather how much longer they will be ca- level and dispersion in disability-free life expectancy pable of working. Life expectancy may be increasing, that we have today may be with us for a long time and but can the same be said for healthy, disability-free that a vulnerable portion of the population – perhaps life expectancy? those who most need to work longer – might not be able to extend their work lives. * Alicia H. Munnell is the Peter F. Drucker Professor of Management Sciences in Boston College’s Carroll School of Man- agement and Director of the Center for Retirement Research at Boston College (CRR). Mauricio Soto is a research associate at the Urban Institute. Alex Golub-Sass is a research associate at the CRR. This brief is based on a longer paper (Munnell, Soto, and Golub-Sass 2008). 2 Center for Retirement Research Recent Trends in Disability- the National Health Interview Survey (NHIS).4 This Free Life Expectancy survey of about 100,000 people has been conducted annually since 1959 by the U.S. National Center for Health Statistics to monitor the population’s health In the last 35 years, virtually all studies show that poor and health care utilization. The NHIS asks a series of health has a negative effect on the likelihood of being questions to identify individuals who have a “limita- in the labor force and on the expected retirement age, tion of activities.” Based on the response to those as well as hours worked and wages earned.1 There- questions, Table 1 shows that the percent of men age fore, the health of the older working-age population is 50-54 and 55-59 with an activity limitation was higher a key prerequisite to extending the retirement age. in 2000 than in 1970. Survival expectancies Table 1. Percent of the Non-Institutionalized Male Population with Limitation of Activity, Death is the end point, so a natural starting place for 1970-2000 exploring the ability of older people to work is life expectancy. An increase in life expectancy raises the Age 1970 1980 1990 2000 possibility of a longer worklife with the potential of some period of retirement at the end. Thus, it is the 50-54 18.7 % 20.9 % 21.4 % 21.2 % first step for establishing that people are able to work 55-59 23.3 28.1 23.4 24.3 longer. Between 1970 and 2000, life expectancy for 60-64 30.8 36.1 31.8 29.3 a 50-year-old man increased from 23.2 years to 27.5 years, suggesting that men are capable of working Note: Figures for 1990 and 2000 are adjusted to account longer.2 for survey redesign. See Munnell, Soto, and Golub-Sass (2008) for details. Source: Authors’ calculations using U.S. Department of Disability-free life expectancy Health and Human Services, National Health Interview Survey (NHIS), 1969-2001. Death is not the only relevant end point for how long people can work. Many non-fatal conditions The data on activity limitations can be combined may make it difficult for people to stay in the labor with period life tables to determine the number of force. Increases in total expected years of life are not years individuals are expected to be alive in the com- necessarily accompanied by increases in expected munity with no activity limitations, or the disability- disability-free life.3 Thus, it is important to examine free life expectancy at age 50. Between 1970 and how disability-free life has changed over recent de- 2000, while life expectancy at age 50 increased by 4.2 cades. The calculation of disability-free life combines years, disability-free life expectancy increased by only data on life expectancy with data on disability from 2.7 years (see Table 2). The pattern was one of virtu- Table 2. Expectations at Age 50 of Years Spent in Various States of Health, 1970-2000 Years Change Expectation of life 1970- 1980- 1990- 1970- 1970 1980 1990 2000 1980 1990 2000 2000 Total 23.2 24.8 26.3 27.5 1.6 1.4 1.2 4.2 Free of disability 15.2 15.2 16.7 17.9 0.1 1.4 1.2 2.7 With disability 7.6 9.1 9.2 9.1 1.5 0.1 0.0 1.5 Institutionalized 0.5 0.5 0.5 0.5 0.0 -0.1 0.0 0.0 Sources: Authors’ calculations using Bell and Miller (2005); U.S. Bureau of the Census (1973a, 1973b, 1983, 1984, 1991, and 2001); and 1969-2001 NHIS. Issue in Brief 3 ally no gains in disability-free years in the 1970s, and Table 3. Relative Mortality Rates and Life roughly equivalent gains in the 1980s and 1990s. Expectancy for Males At Age 50 by Race and The overall conclusion is that men, on average, Education, 1979-1989 can expect more disability-free years than they could in the 1960s when the average retirement age was Race and Relative Implied life 66. But averages may not tell the whole story because education* mortality rates expectancy health status and life expectancy both vary by socio- White economic status. <High school (HS) 1.33 24.2 HS + 0.95 26.3 Recent Trends in Disability- College + 0.57 28.7 Free Life Expectancy by Black Education and Race <HS HS + 2.55 1.75 20.9 22.0 Numerous studies have documented a strong link * <HS is for individuals without a high school diploma. between health and mortality and socioeconomic sta- HS + is for those with a high school diploma and perhaps tus. Education, occupation, and income are the most some college. College + is for those with a college degree widely used measures. The relationship between edu- and perhaps an advanced degree. cation and mortality and health appears to be particu- Sources: Brown, Liebman, and Pollet (2002) and authors’ larly strong, even after accounting for other factors.5 calculations using Bell and Miller (2005). Studies have also shown that health status varies by race. Thus, it is not enough to assert that because life The next step is to use the NHIS to document the expectancy and even years of disability-free life have percent of the population by education and race with increased on average since the 1960s that everyone activity limitations. These percentages are shown can work longer. The following calculations docu- in Table 4 for each subgroup. With the exception of ment the trends in disability-free life expectancy by white college-educated men, the prevalence of disabil- education and race. ity has increased over time. This exercise is more complicated than that for the entire population because it requires life tables for each education and race group. Data are available Table 4. Percent of the Non-Institutionalized on the ratios of the mortality of each education-race Male Population Age 50-64 with Limitation of group relative to the general population mortality for Activity, 1970-2000 the period around 1990. As shown in Table 3, the dis- crepancies are enormous, implying life expectancy at White Black age 50 varying from 28.7 years for a college-educated Year <HS HS + College + <HS HS + white male to 20.9 years for a black male with less than high school. It is not possible to use these same 1970 22.5 % 16.7 % 13.2 % 28.1 % 17.5% relative mortality ratios for all years, however, because 1980 27.3 18.7 13.2 32.3 19.6 the literature shows that the relationship between 1990 28.8 18.7 14.0 33.1 19.0 mortality and education has increased over time.6 Therefore, the relative mortality tables for 1990 are 2000 34.0 20.0 11.4 36.7 25.0 adjusted using the changes in the 10-year death rates Note: Figures for 1990 and 2000 are adjusted to account for by education and race from the decennial censuses.7 survey redesign, based on methodology in Crimmins, Saito, The result is a set of relative mortality ratios for 1970, and Ingegneri (1997). Figures for blacks with high school 1980, 1990, and 2000 for each available education- or more are adjusted for 1970.8 See Munnell, Soto, and race group. These ratios are applied to the period Golub-Sass (2008) for further details. life tables to obtain life tables by education and race Source: Authors’ calculations using the 1969-2001 NHIS. group. 4 Center for Retirement Research The final step is to estimate years of disability-free Table 6. Disability-Free Life Expectancy for life for each group.9 The results show that, with the Males at Age 50, by Quartile, 1970-2000 exception of college-educated whites, disability-free life expectancy has remained virtually unchanged or Quartile of worsened for each group (see Table 5). disability-free life 1970 1980 1990 2000 expectancy Table 5. Total Life Expectancy and Disability-Free Bottom quartile 12.5 12.0 12.8 14.1 Life Expectancy for Males at Age 50, by 2nd quartile 13.7 14.8 17.3 17.9 Education and Race, 1970-2000 3rd quartile 16.7 16.5 17.7 19.1 White Black Top quartile 17.8 20.3 21.0 22.8 Year <HS HS + College + <HS HS + Total 15.2 15.9 17.2 18.4 Total life expectancy Note: This table combines the disability-free life expectancy at age 50 by race-education with the race-education distribu- 1970 21.3 23.6 25.4 17.3 17.7 tion of men aged 50-54 for each year. 1980 22.8 24.9 29.7 19.2 22.9 Source: Authors’ calculations using the 1969-2001 NHIS. 1990 24.2 26.3 28.7 20.9 22.0 2000 25.2 27.2 30.1 22.3 23.4 A Look to the Future Disability-free life expectancy An important question is how disability-free life 1970 13.3 16.7 19.2 10.3 15.4 expectancy will change over time. The outcome 1980 12.7 16.5 21.5 9.6 15.2 depends on two factors – the general health of the 1990 13.0 17.7 21.3 10.7 14.5 population and changing patterns of educational attainment. Recent developments suggest some con- 2000 13.3 17.9 22.8 11.4 14.8 cern on both fronts. Note: Figures for blacks with high school or more are ad- justed for 1970. See endnote 8. Source: Authors’ calculations using the 1969-2001 NHIS. Recent health trends A number of recent studies suggest that continued Of course, educational attainment has increased improvement in the health of the older working-age over the last 30 years, so information by sub-group population may not continue. For example, one study does not give a comprehensive picture of what has reported data from the Health and Retirement Study happened to disability-free life expectancy for the (HRS) on the health status of those age 51-56 from population as a whole and for different quartiles of three different cohorts: the original HRS cohort born the population. Table 6 shows the disability-free 1936-41; the so-called War Babies born 1942-47; and life expectancy for the population as a whole and for the Early Baby Boomers born 1948-53.10 Despite each quartile of the population. This table results enormous advances in diagnosis and treatment, Early from combining information on the distribution of Baby Boomers and War Babies are much less likely to the population by educational attainment and race assess their overall health as “excellent or very good.” at each point in time with the data on disability-free These cohorts also suffer more than the original life expectancy. The numbers for the total population HRS sample from pain, chronic diseases, psychiatric are very similar to those reported in Table 2, but the problems, and alcohol issues.11 And the deterioration estimates by quartile vary enormously. appears to be increasing with each cohort (see Table 7 Two important points emerge from this analysis. on the next page).12 First, since relatively little improvement has occurred In addition, the reductions in risk factors that have in disability-free life expectancy within individual race contributed to the decline in mortality (and presum- and educational groups, most of the overall improve- ably improved health) over the last 30 years may well ment has occurred because people have moved up be offset by the increase in obesity going forward. the educational ladder. Second, enormous disparities One study estimated that the population age-adjusted exist in disability-free life expectancy between those probability of dying in ten years declined from 9.8 in the bottom and the top quartiles of the population. percent to 8.4 percent between the 1970s and the Thus, people vary enormously in terms of their ability 2000s.13 The largest contributors to this reduction to continue working. Issue in Brief 5 Table 7. Self-Reported Health Status of Males in Figure 1. Percent of Men Age 50-54 with a High the Health and Retirement Study, Ages 51-56 by School Diploma or a College Degree, 1969–2020 Birth Cohort Original War Early 100% Health Historical High school cohort Babies Boomers Future status 80% 1992 1998 2004 Excellent/ 60% 57 % 54 % 50 % very good 40% Problem reported: Pain 17 23 29 20% College Chronic 53 54 60 0% Psychiatric 8 17 21 1969 1979 1989 1999 2009 2019 Alcohol 21 23 28 Source: Authors’ calculations using the 1969-2006 NHIS. Source: Soldo et al. (2006). were the decline in smoking and better control of Conclusion blood pressure. But by the early 2020s, rising Body Mass Index (BMI) could more than offset any con- A series of conclusions emerge from this brief. First, tinued reduction in smoking. With two-thirds of the on average, a 50-year-old man could expect almost population overweight or obese, continued improve- three more years of healthy life in 2000 than in ments in health may be an unrealistic expectation.14 1970. Second, disability-free life expectancy varies In short, health and mortality trends may not be significantly by race and education. Third, with the improving. exception of college graduates, little improvement has occurred within each race-education group. Fourth, when collapsing the race and educational groups into Trends in educational attainment quartiles of the population, disability-free life expec- tancy averages 14 years for the lowest quartile, 18 to Increases in disability-free life expectancy over 1970- 19 years for quartiles two and three, and 23 years for 2000 coincided with large increases in educational the highest quartile. Finally, given the leveling of attainment. Between 1970 and 2000, the percent of male educational attainment and the idea that obesity men 50-54 with a high school degree went from 55 may slow or reverse health gains, disability-free life percent to about 84 percent and the percent with a expectancy may not continue to improve in the future. college degree increased from 13 percent to about 32 These conclusions have implications for policy- percent. makers who may be seeking ways to encourage longer Improvements in educational attainment, howev- worklives, particularly in light of the current financial er, have recently plateaued. Men in their 30s and 40s crisis. Physical limitations should not inhibit the bulk today have similar levels of education as those 50-54. of older Americans from working at least until their These trends imply that the education achievement mid-sixties. However, at least a quarter of the popu- of men 50-54 in the future will not be better than it lation may find continued employment extremely is today. In fact, for the next 20 years, the percent of difficult. And employment prospects are unlikely to men 50-54 with high school and college degrees will improve given the plateauing of educational achieve- remain around 85 percent and 30 percent respectively ment and the growing incidence of obesity. (see Figure 1).15 Thus, looking forward, the lack of continued increases in educational attainment might slow down improvements in health. 6 Center for Retirement Research Endnotes 1 For a survey of the literature, see Currie and rates to the death rate of the general population. The Madrian (1999) and Deschryvere (2005). ratio of the 10-year death rate to the general popula- tion goes from 1.02 in 1960 to 1.06 in 1990 and 1.09 2 The calculations in this brief are based on “period” in 2000 for those with less than high school and life expectancy, a measure that assumes age-specific from 0.71 in 1960 to 0.67 in 1990 and 0.64 in 2000 mortality rates for a specific year remain constant in for those with college or more. We apply the changes the future. An alternative measure is “cohort” life in these ratios over time to the 1990 relative mortality expectancy, which includes expected future mortal- tables. For example, in 1990, the mortality rate for a ity improvements that generally produce higher white male with less than high school at age 50 is 1.33 estimates than period life expectancy. Ideally, we times the mortality of the general population (Brown, would like to use cohort life tables jointly with cohort Liebman, and Pollet 2002). The adjustment means disability rates to estimate the changes in disability- that this ratio decreases to 1.27 (1.33*1.02/1.06) in free life expectancy for different cohorts. The deci- 1960 and increases to 1.36 (1.33*1.09/1.06) in 2000. sion to use period life tables is based on two practi- We used the Census calculations as a conservative cal considerations. First, previous literature uses measure of the growing disparities in mortality. An period life tables, and we wanted to make our figures alternative specification using the data on self-report- comparable with the existing literature. And second, ed health instead of the 10-year death rates from the constructing cohort disability rates requires assump- Census generates qualitatively equivalent results, al- tions about the potential changes in disability rates though the implied speed of growth of the disparities of the elderly that might obscure the analysis. We in health across education groups is much larger. conducted alternative analyses using cohort life tables and cohort disability rates (assuming the improve- 8 Figures for blacks with high school or more are ment in disability rates by age group for each cohort adjusted for 1970 because of the small number of ob- equals the average improvements in disability rates servations in this category. The adjustment assumes for similar age groups in the last 40 years), and the that the ratio of the percent of black males with high results from this alternative scenario are qualitatively school or more to white males with high school or comparable to the results presented in this brief. some college in 1970 is the same as the ratio observed for 1980. This imputation does not affect the overall 3 See Crimmins, Saito, and Ingegneri (1997) and results because blacks with high school or more were Fries (1983). only about 2.5 percent of the population in 1970 for the age groups analyzed in this brief. 4 The following analysis builds on the work of Crim- mins, Saito, and Ingegneri (1989 and 1997). 9 As before, the period life tables are used to esti- mate the number of years individuals age 50 in each 5 For a discussion of the theories on why education race-education group are expected to be alive in five plays such an important role, see Cutler and Lleras- age ranges: 50-54, 55-59, 60-64, 65-69, and 70 and Muney (2006). Education has also become the older. The sum of these values is the life expectancy favorite measure because it can be determined for all at age 50. The values for each age category are mul- individuals, and education generally avoids the pos- tiplied by the percent non-institutionalized to get the sibility of reverse causation – for example, poor health years individuals age 50 are expected to be alive in may lead to low income. In contrast, most people the community. These figures are multiplied by one complete their education by their early adult years, so minus the disability rates in Table 4 for each race- educational attainment is unlikely to be affected by education-age category to determine the number of the health impairments that occur later in life. years individuals age 50 are expected to be alive in the community with no disabilities. The sum of these 6 See Crimmins and Saito (2001). values is the years of disability-free life expectancy at age 50. 7 We calculated the 10-year death rates for men 50 and over from the IPUMS-Census data for each edu- 10 See Soldo et al. (2006). cation group and estimated the ratio of these death Issue in Brief 7 11 Respondents were coded as having a potential drinking problem if they responded positively to more than 1 out of the 4 items: ever felt should cut down on drinking, ever criticized for drinking, felt bad or guilty about drinking, or ever taken a drink first thing in the morning. This measure is used clinically to screen for alcoholism. 12 See Cutler, Glaeser, and Rosen (2007). Another suggestion that middle-aged people in the United States are facing serious health problems comes from a study that compares the self-reported rates of several chronic diseases related to diabetes and heart disease, adjusted for age and health behavior risk factors, of non-Hispanic white individuals 55-64 in the United States and the United Kingdom (Banks et al., 2006). The results showed that the U.S. population in late middle age is less healthy than the equivalent U.K. population for diabetes, hypertension, heart disease, myocardial infarction, stroke, lung disease, and can- cer. These results hold even controlling for behavioral risk, including smoking, overweight, obesity, and alcohol drinking, which explain very little of these health differences. These differences are not due to biases in self-reporting disease, because biological markers of disease exhibit exactly the same patterns. And they are not solely driven by the bottom of the socioeconomic distribution; in many diseases, the top of the distribution is less healthy in the United States as well. 13 Cutler, Glaeser, and Rosen (2007) use the 1971-75 and 1999-2002 National Health and Nutrition Exami- nation Survey. 14 If everyone took medication for hypertension and high cholesterol, the impact of rising obesity as mea- sured by BMI could be almost eliminated, but it is not clear that will happen. 15 The figure shows the three-year moving average of the percent with a high school diploma and percent with a college degree. Men with a GED are classified as not having a diploma. The graduation rates after 2006 are constructed from the graduation rates for men 35-39, 40-44, and 45-49 between 2001-2006. 8 Center for Retirement Research References Banks, James, Michael Marmot, Zoe Oldfield, and Deschryvere, Matthias. 2005. “Health and Retirement James P. Smith. 2006. “Disease and Disadvantage Decisions: An Update of the Literature.” Research in the United States and in England.” Journal of Report. European Network of Economic Policy the American Medical Association 295(17): 2037-45. Research Institutes. Brown, Jeffrey R., Jeffrey B. Liebman, and Joshua Fries, James F. 1983. “The Compression of Morbid- Pollet. 2002. “Estimating Life Tables that Reflect ity.” Milbank Memorial Fund/Quarterly/Health and Socioeconomic Differences in Mortality.” In The Society 61(3): 397-419. Distributional Effects of Social Security Reform, eds. M. Feldstein and J. Liebman. University of Chi- Munnell, Alicia H., Mauricio Soto, and Alex Golub- cago Press: Chicago, IL. Sass. 2008. “Will People Be Healthy Enough to Work Longer?” Working Paper 2008-11. Chestnut Bell, Felicitie C. and Michael L. Miller. 2005. Life Ta- Hill, MA: Boston College Center for Retirement bles for the United States Social Security Area 1900- Research. 2100. Actuarial Study No. 116. Office of the Chief Actuary. U.S. Social Security Administration. Soldo, Beth, Olivia S. Mitchell, Rania Tfaily, and John F. McCabe. 2006. “Cross-Cohort Differences in Crimmins, Eileen M., Yasuhiko Saito, and Dominique Health on the Verge of Retirement.” Working Ingegneri. 1989. “Changes in Life Expectancy Paper 12762. Cambridge, MA: National Bureau of and Disability-Free Life Expectancy in the United Economic Research. States.” Population and Development Review 15(2): 235-267. U.S. Bureau of the Census. 1973a. Census of Popula- tion: 1970. Vol. I. Washington, DC: Government Crimmins, Eileen M., Yasuhiko Saito, and Dominique Printing Office. Ingegneri. 1997. “Trends in Disability-Free Life Expectancy in the United States, 1970-90.” Popu- U.S. Bureau of the Census. 1973b. Census of Popula- lation and Development Review 23(3): 555-572. tion: 1970. Vol. II. Washington, DC: Government Printing Office. Crimmins, Eileen M. and Yasuhiko Saito. 2001. “Trends in Healthy Life Expectancy in the United U.S. Bureau of the Census. 1983. Census of Population: States, 1970-90: Gender, Racial, and Educational 1980, Vol. I. Characteristics of the Population. Wash- Differences.” Social Science and Medicine 52: 1629- ington, DC: Government Printing Office. 1641. U.S. Bureau of the Census. 1984. Census of Popula- Currie, Janet, and Brigitte C. Madrian. 1999. “Health, tion: 1980. Vol. II. Subject Reports. Washington, Health Insurance, and the Labor Market.” In DC: Government Printing Office. Handbook of Labor Economics, eds. Orley C. Ash- enfelter and David Card. Volume 3C. Amsterdam: U.S. Bureau of the Census. 1991. Census of Population Elsevier Science Publishers BV. and Housing, 1990: Summary Tape File 1. Washing- ton, DC: Government Printing Office. Cutler, David and Edward Glaeser and Allison Rosen. 2007. “Is the US Population Behaving Healthier?” U.S. Bureau of the Census. 2001. Census 2000 Sum- Working Paper 13013. Cambridge, MA: National mary File 1. Washington, DC: Government Print- Bureau of Economic Research. ing Office. Cutler, David, and Adriana Lleras-Muney. 2006. U.S. Department of Health and Human Services. “Education and Health: Evaluating the Theories National Health Interview Survey, 1969-2006. and Evidence.” Working Paper 12352. Cambridge, Public Use Sample, Documentation, and Code- MA: National Bureau of Economic Research. book. Hyattsville, MD: National Center for Health Statistics. About the Center Affiliated Institutions The Center for Retirement Research at Boston Col- The Brookings Institution lege was established in 1998 through a grant from the Massachusetts Institute of Technology Social Security Administration. The Center’s mission Syracuse University is to produce first-class research and forge a strong Urban Institute link between the academic community and decision makers in the public and private sectors around an issue of critical importance to the nation’s future. Contact Information Center for Retirement Research To achieve this mission, the Center sponsors a wide Boston College variety of research projects, transmits new findings to Hovey House a broad audience, trains new scholars, and broadens 140 Commonwealth Avenue access to valuable data sources. Since its inception, Chestnut Hill, MA 02467-3808 the Center has established a reputation as an authori- Phone: (617) 552-1762 tative source of information on all major aspects of Fax: (617) 552-0191 the retirement income debate. E-mail: crr@bc.edu Website: http://www.bc.edu/crr © 2008, by Trustees of Boston College, Center for Retire- The research reported herein was performed pursuant to ment Research. All rights reserved. Short sections of text, a grant from the U.S. Social Security Administration (SSA) not to exceed two paragraphs, may be quoted without ex- funded as part of the Retirement Research Consortium. The plicit permission provided that the authors are identified and opinions and conclusions expressed are solely those of the full credit, including copyright notice, is given to Trustees of authors and do not represent the opinions or policy of SSA, Boston College, Center for Retirement Research. any agency of the Federal Government, or the Center for Retirement Research at Boston College.