The U.S. population is growing older and living longer. Yet older people have been leaving the labor force at younger and younger ages. Moreover, most Americans have saved very little. At the same time, the cost of medical care has been increasing. These demographic trends and changes in individual circumstances will contribute to some of the most important economic transitions and policy challenges for the coming decades. Understanding the determinants of retirement, the nature of saving for retirement, and how to more efficiently provide medical care are perhaps the most critical issues that demographic trends have forced on us. These and related issues make up the activities of the NBER's Program on the Economics of Aging.
Begun in 1986, the Aging Program has developed primarily around large, coordinated research projects that address simultaneously several interrelated issues in the economics of aging. Extensive funding for the program has been provided by the National Institute on Aging (NIA), both through multiple research grants and through a Center grant, which provides centralized infrastructure support to the Program effort.
The major research categories in the NBER's Program on the Economics of Aging are 1) saving and the evolving financial circumstances of older Americans, 2) work and retirement decisions at older ages, 3) health care, and 4) aging around the world. In each of these areas, a major goal of the research is to better understand individual decisions as people age, and how these decisions are affected by individual circumstances and the economic incentive effects of government policies and programs. This article summarizes research in each of these areas.
Much effort also has been directed to attracting young researchers to this field. To that end, our NIA Fellowship Program provides annual fellowships to between five and ten graduate students who are beginning research on the economics of aging. It also provides two or three postdoctoral fellowships each year to recent Ph.D. recipients, enabling them to spend a year at the NBER to do research on issues in the economics of aging and health care. The combination of NIA support for research training and fellowships, new project development, data resource development, and smaller "exploratory" grant support has been instrumental in our efforts to expand the program, and to engage outstanding new scholars in research on aging.
Nearly 100 papers are completed annually on issues in aging by participants in the NBER Program. Some of these appear in a series of books published by the University of Chicago Press.1
THE EVOLVING FINANCIAL CIRCUMSTANCES OF OLDER AMERICANS
The way Americans provide for financial support in retirement is changing rapidly. All three of the traditional pillars of retirement support -- Social Security, employer-provided pensions, and saving --are in transition. Potential changes in Social Security have received the most public attention. The aging of the population has made the continuation of current levels of real benefits from Social Security an uncertain prospect. Employers, too, are reacting to the increasing costs of their retirement benefits, with many of them discontinuing traditional pension benefits and retiree health insurance programs. At the same time, the rapid expansion of 401(k) programs, and the dramatic growth in savings in 401(k) and IRA programs, suggest a transition in personal savings as well. While most households retiring in the past had essentially no financial asset savings, that may not be true in the future. By the mid-1990s, at least one spouse in over half of U.S. families was eligible for a 401(k) plan, and over 70 percent of those who were eligible made contributions. Today, over $100 billion is contributed annually to 401(k) plans.
The Growing Influence of Retirement Savings
A long series of studies by Steven F. Venti, James M. Poterba, and me has considered whether IRAs and 401(k) programs have added to personal saving, or whether they have simply replaced saving that would have already taken place in some other form.2 Using a number of different data sources and analytical approaches, we have consistently demonstrated that the saving taking place in IRA and 401(k) plans is new saving.
Having identified IRA and 401(k) plans as important inducements to saving, we focused more recently on the implications of these savings for the financial status of retirees in the future. A central feature of this work is a detailed summary of eligibility, participation given eligibility, and contribution rates, by earnings decile and age.3 Based on conservative assumptions about future participation and contribution rates, we estimate the 401(k) assets of cohorts who will be age 65 in 2025 and 2035. For all but the lowest earnings decile, the projections suggest that 401(k) saving could represent a very substantial contribution to retirement income. Among those reaching age 65 in 2025, for example, the average level of 401(k) assets is likely to exceed the discounted value of Social Security benefits. Thus their 401(k) plans could contribute more to their retirement support than Social Security. While these large financial asset accumulations are unlikely to be realized by families with the lowest lifetime earnings, 401(k) assets are projected to become a substantial fraction of Social Security wealth for families with lifetime earnings above the two or three lowest deciles.
One widely expressed concern is that employees changing jobs may have the opportunity to remove funds from their pension plans, leaving little for retirement. We studied the incidence and disposition of these lump-sum distributions and found that a substantial number of people with small balances do take money out of these accounts.4 However, most larger accounts paid out as lump-sum distributions are reinvested in ways that preserve the funds for retirement. Thus the vast majority of funds remain in personal retirement saving plans.
More recently we expanded our projections of the 401(k) accumulations of future cohorts of retirees to account for "leakage" caused by cash-out of 401(k) assets when people change jobs. Although cash-outs have some effect on accumulation, the foregone accumulation at retirement attributable to this leakage is only about 4 or 5 percent of assets that are retained until retirement.5
Research by Andrew Samwick and Jonathan S. Skinner has analyzed a closely related trend: the movement among employers away from defined benefit pension plans and toward defined contribution pension plans and employer-sponsored savings plans.6 They ask how this trend will affect the financial security of future retirees. Their simulations show that average and median pension benefits are higher under defined contribution plans than under defined benefit plans, and that minimum benefits are comparable. They argue, therefore, that the growth in defined contribution plans may strengthen retirees' financial security.
In related work, John B. Shoven and I documented the very high tax rates on distributions from retirement saving plans -- particularly those for savings distributions passing through an estate.7 Marginal tax rates on nonestate distributions were as high as 62 percent, while marginal tax rates on distributions passing through an estate were as high as 99 percent. Moreover, with moderate ongoing contributions to a 401(k) plan over a working career, these high tax rates could be faced by savers who did not have extraordinarily high incomes. However, perhaps in part because of these studies, these "success" taxes were eliminated as part of the Taxpayer Relief Act of 1997.
Alternative Theories of Saving
David Laibson's work has focused on nontraditional theories of saving, drawing on behavioral findings in experimental psychology.8 Laibson suggests that savings decisions are influenced in part by "hyperbolic" discounting -- a form of discounting that uses relatively low discount rates to evaluate distant events, and increasingly high discount rates to evaluate more proximate events. The implication is that people want to plan for the future (such as by saving for retirement), but their long-term plans are impeded by temptation in the present. Laibson's work suggests that actual consumption and saving decisions over the course of life are in fact quite consistent with hyperbolic discounting.
Laibson's work also suggests that commitment mechanisms such as payroll deductions and early withdrawal penalties tend to increase saving among people making decisions based on hyperbolic discounting. These mechanisms are part of why 401(k) plans are so successful in promoting saving. Specifically, by using payroll deductions and by imposing a financial penalty on early withdrawals, 401(k) plans tend to be more effective than other approaches in encouraging saving. Simulations suggest that 401(k) plans may raise the national saving rate by at least 89 percent.
The Composition and Annuitization of Retirement Saving
Samwick has initiated a number of research projects using the rich pension, earnings, and wealth data in the Health and Retirement Survey (HRS). A collaborative analysis by Alan L. Gustman, Olivia S. Mitchell, Samwick, and Tom Steinmeier finds that HRS households have a large proportion of their wealth in the form of entitlement benefits, rather than real and financial assets.9 The combination of pensions, Social Security, and health insurance accounts for half of the wealth held by all households in the HRS, for 60 percent of total wealth of the median HRS household, and for as much as 48 percent of the wealth of households between the 90th and 95th percentiles of the wealth distribution.
Jeffrey Brown, Mitchell, Poterba, and Mark Warshawsky have explored the role of annuities for older households, focusing initially on how annuities are priced.10 Using the Social Security Administration's mortality tables, which reflect death rates of the general population, they find that the expected present value of payouts associated with single-premium, immediate life annuities is approximately 80 cents per premium dollar. However, the mortality rates of those who actually purchase annuities are lower than the mortality rates for the entire population, which raises the expected future payout for these purchasers. When valued using the mortality rates of actual annuity buyers, the (expected present discounted) value of the payouts rises to between 90 and 95 cents per dollar of annuity premium. Thus the pure administrative costs of purchasing an annuity appear to be less than 10 percent of the premium value for current annuitants.
Follow-up work by Poterba and Warshawsky has compared the pricing of annuities in the public market with the pricing in large group retirement saving plans.11 They find even higher payout rates in the large group retirement savings programs, particularly in the annuity products offered by TIAA-CREF. Other follow-up work by Brown, Mitchell, and Poterba has considered the trade-offs among alternative annuity options for paying out retirement savings later in life, and the potential role of inflation-adjusted annuities.12 Using plausible measures of risk aversion, they find that people would be expected to value a variable payout equity-linked annuity more highly than a real annuity because the additional real returns associated with common stocks more than compensate for the volatility of prospective payouts.
The Distribution of Social Security Benefits
Jeffrey B. Liebman has explored the effective returns to Social Security among individuals with different demographic and economic characteristics. As part of this project, he received special permission to use a confidential version of the Survey of Income and Program Participation that is linked to Social Security earnings histories. In preliminary work using these data, Liebman found higher rates of return from Social Security among those with lower incomes; those whose earnings were concentrated later in life; and married couples, especially married couples with unequal incomes. The large majority of Social Security recipients had rates of return of between zero and 4 percent. About 19 percent of the sample had rates of return of between 4 and 8 percent; and only 4 percent of the sample had a rate of return above 8 percent.
Related work by Kathleen M. McGarry and Robert Schoeni shows that Social Security has had a dramatic influence on the living arrangements of elderly widows.13 Using data from the last six Census surveys, McGarry and Schoeni estimate the effects of rising income, particularly of rising income from Social Security, on living arrangements among widows. The share of elderly widows living alone rose from 18 percent in 1940 to 62 percent in 1990. McGarry and Schoeni find that almost two-thirds of this change can be attributed to increases in Social Security benefits over this period.
Demographic Change and Asset Prices
Poterba has explored the impact of changes in population age structure on the rates of return earned on bonds and stocks.14 Several recent popular articles have suggested that the rise in stock prices during the 1990s is partly attributable to the fact that baby boomers have entered "high saving" years. These studies also suggest that stock and bond prices may decline in the years ahead, when baby boomers in retirement begin to liquidate their assets. Poterba finds that households with significant asset holdings appear to decumulate their assets slowly (if at all) after retirement. This casts doubt on the "sell off" view, at least for the share of assets (excluding defined benefit pension assets) that households control directly. Poterba also considers the historical relationship between demographic structure and real returns on Treasury bills, long-term government bonds, and corporate stock. The results do not suggest any robust relationship between demographic structure and asset returns, although the ability to establish a statistically significant relationship is limited in historical data.
Aging, Saving, and Inequality
Using cohort data from numerous countries, Angus S. Deaton and Christina Paxson find that many people continue to save after retirement.15 This result is contrary to the traditional "lifecycle" explanation of saving in which people save only during their working years in order to support themselves (through dissaving) in retirement. The implication is that population aging will have a less pronounced effect on aggregate saving rates than otherwise might be anticipated. However, the effects of aging on distribution appear to be more significant: because of the more unequal distribution of income and wealth at older, as compared with younger, ages, population aging appears to result in a more unequal distribution in the population as a whole. In some countries, the distributional consequences of population aging may lead to more individuals living at very low income levels.
Deaton and Paxson also have examined how to measure poverty among the elderly, and more generally how to measure living standards of people at different ages, given that they live in households of different sizes and age structures.16 Their work is motivated by the fact that the official rate of poverty among the elderly in the United States is lower than the rate among younger age groups. They find, however, that the measurement of poverty of one age group relative to another is quite sensitive to assumptions made when defining poverty. Thus the "result" of lower poverty among the elderly can easily be turned around when different but equally plausible assumptions are made. For example, Deaton and Paxson experiment with different assumptions about the cost of children relative to adults, and the extent of "economies of scale" in larger households. While their research does not go as far as to propose different measures of poverty, it does highlight the fact that the measurement of an individual's economic welfare is no simple task.
Deaton and Paxson's work on aging and distribution most recently has been extended to analyses of health and the distribution of health status, as measured by body-mass index and by self-reported health status.17 As with the economic measures, the health measures become more widely dispersed with age. In addition to measuring the distribution of health over the life cycle, Deaton and Paxson look at the correlation between income and health over the life cycle. The positive correlation between health and socio-economic status is well documented. What Deaton and Paxson discover, however, is that the correlation is strongest during the highest income periods of life, rather than in retirement. Such an effect would be predicted by a model in which health status "causes" income -- for example, through absence from work.
Innovative Survey Measures
A number of NBER investigators are working with the HRS and the Survey of Asset and Health Dynamics among the Oldest Old (AHEAD), two innovative new longitudinal surveys of Americans age 51 and older. One of the innovative components of these surveys is the use of a "bracketing" technique to measure wealth and consumption variables. Michael D. Hurd, Daniel L. McFadden, and their coauthors have analyzed biases in the data that these methods may cause.18 They find very substantial anchoring effects (biases toward an initial bracketing value), enough so that changing the question format within reasonable ranges could induce variations by a factor of two in average savings balances or average consumption levels in the elderly population. This is a red flag that should alert researchers that response bias is a serious issue in surveys of the elderly, and that these surveys should build in sufficient experimental control so that any bias resulting from question format can be determined and possibly corrected.
Another innovative component of the HRS and AHEAD surveys is a series of questions about people's subjective expectations about their likelihood of surviving to particular ages. Hurd, McFadden, and Gan find that the distribution of responses is remarkably close to life tables, and that responses co-vary as one would expect with other variables, such as gender and smoking status.19 At more advanced ages (80 and older), the subjective mortality measures are found to be progressively more optimistic than are the life tables. The investigators find preliminary evidence that the subjective mortality measures help explain individual saving decisions, although their explanatory power is limited by the tendency of respondents to choose "round number" probabilities.
LABOR FORCE PARTICIPATION
Younger retirement has been a long-term trend in the United States. The widespread availability of post-retirement benefits is an important aspect of this national trend. Eligibility for employer-provided benefits can begin as young as age 50, and quite frequently occurs at age 55. Eligibility for Social Security benefits begins at age 62. Eligibility for Medicare begins at age 65. Likely reforms in these benefit programs will change their benefit structure, and their associated incentives for continued work versus retirement. Employers, too, will make adjustments in their employment practices in response to changing population demographics. What does this mean for the work and retirement decisions of older Americans in the future?
Pensions, Social Security, and Retirement
Robin L. Lumsdaine, James H. Stock, and I have conducted a series of studies demonstrating how traditional employer-provided pension plans encourage early retirement and penalize continuation in the labor force at older ages.20 For example, company pension plans commonly provide early retirement benefits beginning at age 55. Although employees become eligible for larger pension payments by continuing to work, the increase is rarely large enough to compensate for the delay in receiving benefits. Because of this economic structure, pension plans have an enormous effect in inducing retirement at much younger ages than would occur without the plans.
Similar incentives for retirement exist in Social Security, although eligibility for Social Security occurs somewhat later than the early retirement age of most traditional pension plans. Samwick explores the combined influence of pensions and Social Security, using data that link the economic and demographic information of households with details of their pension formulas.21 He, too, finds that the most significant economic determinant of retirement is the accrual of retirement wealth that results from continued work. As much as one-fourth of the decline in labor force participation among older Americans in the early postwar period can be attributed to the growth in pension coverage.
Leora Friedberg also has explored the decline in labor force participation among older workers, focusing on the Old Age Assistance (OAA) program -- a means-tested retirement benefit program that was far larger than Social Security during the 1940s. She finds that OAA benefits had a strong influence in inducing earlier retirement during this period.22
Courtney Coile and Jonathan Gruber look at the work and retirement incentives in the current Social Security program. Using HRS data, they estimate the implicit tax or subsidy on work at each age. They find substantial variation in incentives across the population. At the median, they find a small subsidy on continued work between ages 62 and 64 -- equivalent to between 2.5 and 5.4 percent of earnings. However, because of the variability across individuals, about one-third of workers at age 62 have an implicit tax. The addition of private pension incentives does little to affect the work and retirement incentives at the median, but it substantially increases the variability in incentives across individuals.23
In preliminary work using Current Population Survey data, Friedberg finds a substantial increase in part-time work among older Americans, particularly among men. Between 1980 and 1997, the percentage of workers who worked part time increased from 5.5 to 8.7 percent among men ages 55 to 59; from 9.9 to 18 percent among men ages 60 to 64; and from 44.7 to 46.4 percent among men ages 65 to 69. Using 1992 and 1994 data in the longitudinal HRS survey, Friedberg has also started to look at part-time work as a potential transition to retirement. For example, about 12 percent of working men age 60 to 61 reported working part time in the 1992 survey. Among those still working in 1994, 22.5 percent were working part time. Thus part-time work appears to be an increasingly important approach to phasing into retirement. Continuing HRS re-interviews will enable us to learn more about these labor force transitions at older ages.
Health Insurance and the Labor Market
Brigitte Madrian has continued to advance our knowledge of how the availability and cost of health insurance influences labor market decisions regarding job mobility, retirement, and self-employment, for example. In a recent study done with David M. Cutler, she looks at how rising health insurance costs affect firms' employment decisions.24 The theory is that employers may substitute hours of work per worker for the number of workers as a means of economizing on health insurance costs. The results show that rising health insurance costs over the 1980s increased hours worked of those with health insurance by up to 3 percent. In another recent study, Madrian and Lars Lefgren finds that the availability of health insurance influences transitions to self-employment.25 They find that workers who have a spouse with health insurance are more likely to become self-employed if they have access to more generous continuation coverage; workers who have a larger family are less likely to become self-employed.
The United States stands out as having the highest health care costs in the world. It is also a leader in the advancement of treatment technology. While health care is already a high-priority issue for policy evaluation and an ongoing target for policy reform, the aging of the population will only heighten attention to the costs and benefits associated with our health care system. The NBER Program on the Economics of Aging has made a major investment in developing data resources for health care research. These resources now include comprehensive medical claims records for Medicare beneficiaries and employees covered by employer-provided health insurance plans, hospital discharge records from numerous states, Medicaid claims records, and survey data that link information about people's health and economic circumstances, their health insurance, and the health care they receive. A wide range of research activities is applying these data toward a better understanding of our health care system, the costs and benefits of treatments as they evolve over time, and the implications for costs and outcomes of alternative approaches to health policy.
The Growth and Distribution of Health Care Expenditures
Alan M. Garber, Thomas E. MaCurdy, and Mark B. McClellan have used Medicare claims data to help explore the composition and growth of Medicare expenditures over time. They confirm a high level of concentration in expenditures among relatively few beneficiaries. The top 10 percent of beneficiaries using Medicare services account for 64 percent of annual Medicare expenditures; the top 2 percent receiving services account for more than a quarter of expenditures. Less well known is that expenditure growth has been more concentrated over the past decade: the top 2 percent of beneficiaries accounted for almost one-third of Medicare expenditure growth. Thus Medicare expenditures have become even more concentrated over time. Garber, MaCurdy, and McClellan also find limited persistence in individual Medicare expenditures from one year to the next: that is, the number of people with high expenditures over multiple years is very small.26 Among Medicare beneficiaries with high expenditures in one year, subsequent death or reversion to lower expenditure levels is much more common than continued expenditure at high levels. These data are very similar to comparable data for younger persons covered by employer-provided insurance plans, as shown by Eichner, McClellan, and Wise, as discussed later in this article.
The use of costly medical treatments for the elderly has been a primary determinant of Medicare expenditure growth over time. In one example of this, Cutler, McClellan, and Joseph P. Newhouse explore the role of technology in increasing both costs -- and benefits -- associated with heart attack treatment.27 The cost of treating heart attack patients has increased at an inflation-adjusted rate of about 4 percent annually. This expenditure growth has resulted entirely from the increasing use of intensive cardiac procedures (including cardiac catheterization, bypass surgery, and angioplasty). The costs of any given type of treatment has been essentially unchanged over time, and the incidence of heart attacks actually has been falling. Life expectancy following a heart attack also has improved in recent years, a trend that Cutler, McClellan, and Newhouse attribute largely to new medical interventions and increased use of existing interventions. Indeed, the value of improved health is estimated to outweigh the increased cost of heart attack care.
Another trend in health care expenditures relates to the age composition of spending. Cutler and Ellen Meara find dramatically higher spending growth among older health care recipients.28 Between 1963 and 1987, real per capita medical spending on the elderly increased 8 percent annually, as compared with 4.7 percent for the population aged 1 to 64. Only spending on infants increased as rapidly (9.8 percent per year) as spending on the elderly. Again, Cutler and Meara point to technological advances in treatment as the core factor in the disproportionate growth in costs among these groups.
Another major change in health care is occurring near the end of life. Garber, MaCurdy, and McClellan show that the nature of medical care near death is changing substantially, particularly for the increasing share of the elderly population whose deaths are from chronic conditions or are associated with frailty.29 For the elderly dying of acute conditions like heart attacks, the vast majority of deaths occur in acute hospitals (inpatient and outpatient), and this pattern has not changed appreciably between 1988 and 1995. In contrast, the location of death has changed substantially for the elderly with chronic illnesses, like cancer, for which death is more predictable. For example, while over three-fourths of patients with heart attacks or hemorrhagic strokes near their death still die in the hospital, the proportion of patients with lung cancer dying in the hospital has fallen by more than half in the same period (from over 60 percent to under 30 percent). This decline in acute-hospital deaths was more than offset by growth in deaths in alternative settings that are financed by Medicare, especially hospice care (representing almost one-third of deaths in 1995) and non-acute hospitals. Relatively few elderly die without receiving any Medicare-covered services near the time of death.
The Composition of Firm Health Care Expenditures
Matthew Eichner, McClellan, and I have conducted a series of studies on employer-provided health insurance benefits, the composition and distribution of plan expenditures, and the cost and outcome implications of alternative plan designs. A key finding from this work relates to the concentration of health care expenditures among a small percentage of covered individuals, and the persistence of individual health care expenditures over time. For example, the most expensive 20 percent of plan enrollees in 1990 accounted for over 85 percent of health care costs in 1990, and those same individuals still accounted for almost 50 percent of health care costs in 1992.
Eichner, McClellan, and I have paid particular attention to the implications of persistence for the feasibility of medical savings accounts, or MSAs.30 An MSA is set up in combination with a catastrophic health care plan with a high deductible. Under these plans, a given amount is deposited each year in an employee MSA. All medical expenditures below a deductible -- $4,000 in the illustrative plan we consider -- are paid out of this account. Any expenditures above the deductible are paid by the insurance plan. Such schemes are designed to provide consumers, and their health care providers, incentives not to spend money on care that offers only low marginal benefit. The viability of any MSA-based system, however, depends on the persistence of individual expenditures. The key substantive conclusion of this work is that expenditure persistence does not severely limit the feasibility of MSAs as a way to mitigate moral hazard and thus to more efficiently guide the purchase of health care. While there is clear persistence in health care expenditures, the proportion of persons who have high expenditures in successive years is very small. For example, only about 6.23 percent of plan enrollees had medical expenditures above $1,000 in 1989 and 1991, and only 1.17 percent had expenditures above $5,000 in each of those years. More recently, we have pursued this work using a six-year rather than a three-year panel of individual expenditure data. Eichner and I also consider in some detail the potential balance in an MSA at retirement, accounting for different rates of return on account balances. This work is a precursor to thinking more intensively about the potential for combined MSA and 401(k)-like saving plans.31
Eichner, McClellan, and I have also analyzed the sources of cost differences across firms.32 There are three reasons for cost differences: 1) differences in the demographic attributes -- age and gender - of enrollees, 2) differences in the illnesses that are treated -- the rate of treatment, 3) and differences in the cost of treating illnesses. While the differences in costs across firms is quite large, it becomes even larger after accounting for differences in the demographic characteristics of plan enrollees. Another finding is that the cost of treatment for almost every diagnosis is higher in the plans that have the higher overall costs, but the likelihood of receiving treatment tends to be lower in the high-cost plans. The decomposition analysis also has allowed us to isolate the diagnoses for which the intensity of treatment (dollar cost) differences are the greatest. For example, the average cost of childbirth ranges across plans from a low of $4,046 to a high of $7,119. The proportion of deliveries by cesarean section ranges from 17 to 34 percent.
Isolating Incentive Effects from Selection Effects
Many U.S. employers offer a menu of health insurance options to their employees. Thus employees can choose among plans with different premiums, different deductibles and copayments, different coverage characteristics, and different health care management arrangements. Using the range of health plans offered at one large U.S. company, Eichner has explored the determinants of plan choice and how these determinants relate to plan expenditures.33 He finds that about half of the cost difference between the high-cost plans and the lower-cost plans results from "selection" -- or the tendency of less healthy individuals to choose more generous health insurance. The other half results from covered individuals choosing more health care services when their insurance is more generous -- that is, from plan incentive effects. He finds a rather substantial effect of plan incentive effects on expenditures, in some ways again highlighting the relationship between plan provisions and medical expenditures.
Socioeconomic Status, Health, and Longevity
Two recent studies have documented the strong relationship between wealth and mortality. Hurd, McFadden, and Angela Merrill use AHEAD data to examine the death rates of individuals aged 70 and older during the two-year period between the 1993 and 1995 surveys.34 They find that individuals in the lowest wealth quartile were between two and three times as likely to die during this period as individuals in the highest wealth quartile. Orazio Attanasio and Hilary W. Hoynes find about the same differences in mortality rates using SIPP data.35
One reason for the strong relationship between health and economic status is that adverse health events have adverse economic implications. Using HRS data, Tracy Falba and McClellan have focused on withdrawal from the labor force after an adverse health event.36 They find that employed men who experience a major health event (such as a heart attack or stroke) are about 25 percentage points more likely to transition to zero hours of work than employed men without a health event; the average number of work hours declines by more than 600 per year. If the acute health event is accompanied by a major decline in functional status, then the men are about 75 percentage points more likely to transition to zero hours of work; their average number of work hours declines by about 1,700 more than men with no health events.
AGING AROUND THE WORLD
Population aging is occurring not just in the United States, but in every country of the world. While the specific demographic circumstances and the details of aging-related policies differ across countries, the fundamental challenges of funding social security and health care systems with growing populations of older people remain the same. The NBER Program on the Economics of Aging has initiated significant comparative work on aging in different countries, much of which is being done in collaboration with investigators overseas. These international comparisons are enabling us to study experiences and policy variation that do not exist in any single country.
Social Security and Retirement
Jonathan Gruber and I have been directing a collaborative study on the public retirement income programs in 11 developed countries including the United States, and how these programs affect retirement decisions.37 The project has involved scholars in each of the 11 countries who have expertise in retirement analysis.38 Both the incentive structure of the policies and the changes in retirement behavior in the countries studied suggest a very strong relationship between policy provisions and retirement behavior.
We find a trend toward earlier retirement in every country. In the early 1960s, for example, labor force participation among men between the ages of 60 and 64 was over 70 percent in every country studied. By the mid-1990s, the rate had fallen to below 20 percent in Belgium, Italy, France, and the Netherlands; to about 35 percent in Germany; to about 40 percent in Spain; to 53 percent in the United States; to 57 percent in Sweden; and to 75 percent (from 83 percent) in Japan. The earlier retirement, combined with the growing population of older persons, has placed enormous pressure on the financial solvency of retirement income systems around the world. This set of studies suggests that the retirement policies themselves contribute to the earlier retirement, often providing generous retirement benefits at young ages and imposing large financial penalties on labor earnings after the age of eligibility for retirement benefits. As a result of these incentives, there is a strong correspondence between the age at which benefits are available in each country and the age when people tend to leave the labor force.
The project has produced important findings both within the individual country studies and in the cross-national summaries that integrate them. Within countries, we find that people retire more often when it is financially beneficial, as defined by the benefit provisions of each country's social security system. The policy reforms that have changed the benefit structure in particular countries appear to have caused corresponding changes in the age of retirement in those countries. In comparing countries, we find earlier retirement in those with social security systems that "tax" continued work more heavily. So, in countries where the general tax burden on work at older ages is large, many more older persons choose to retire rather than to continue to work. The implication of these findings is that social security programs themselves, by "taxing" continued work at older ages, actually are exacerbating their own financial problems.
The Economics of Aging in Japan and the United States
I have also been directing a long-term project on the economics of aging in Japan and the United States. This work has been done in collaboration with economists at the Japan Center for Economic Research (JCER).39 Japan already has one of the oldest populations in the world, and the projections for the future elderly population are as high as in any country in the world. Studies have been done on retirement decisions in Japan and the United States, on the wealth accumulations and financial circumstances of older people in the two countries, on housing and living arrangements, on health care systems and costs, and on the implications of population aging on government finances and financial markets. Numerous NBER-JCER joint meetings have been held, stimulating collaborative research and interaction between economic scholars in both countries and resulting in several research volumes.
Provision for Aging in South Africa
David E. Bloom, Anne Case, Deaton, and I have been engaged in a project that brings together a small team of scholars from South Africa and the United States to collect data and conduct research on the pension system in South Africa.40 South Africa's public pension system is unusual in comparison to the systems of other countries, in that prior labor force participation is not a condition of receiving benefits. Access is universal, although means-tested. Although intended for the elderly (women qualify at age 60, men at 65), the pension system serves as a vital welfare program more generally and provides an important source of household income maintenance for a substantial portion of the population. For example, the benefit received by most black recipients is about half of average household income and more than twice the median per capita household income of the black population. The payment is three times the World Bank's cutoff for absolute poverty. Both the generosity and the universality of the system are likely to have significant implications for the health, family dynamics, labor market behavior, and well-being of the elderly, their extended families, and the overall population of South Africa. In particular, by extending such benefits to a large group of poor people, the government has made the presence of a pensioner in the household an important determinant of household well-being and may have significantly affected work and savings incentives for much of the population.
Our project's goal was to devise, test, and apply methods for gathering data about the effects of South Africa's pension system, in a manner appropriate to the cultural, social, and economic conditions of the country; we also south to study the pension system's effects on the health, well-being, family dynamics, and labor supply behavior of the elderly and those who live with them. To date, we have conducted a series of pilot surveys and focus groups. Case and Deaton also have completed a study documenting some of the key characteristics of the pension system in South Africa.41
In future work, we will pursue many of the issues discussed here and will initiate work in new areas as well. For example, the work on social security around the world is continuing, and we will initiate analogous comparisons on health care systems in several countries. In new work we will try to understand more clearly the relationship between wealth and health. We will pay considerable attention to the implications of increased personal retirement saving, and to some of the nontraditional explanations for people's decisions about how much to save. We will also explore changes in labor force behavior at older ages, changes in employment opportunities for older workers, changes in the composition of employer-provided retirement benefits, care-giving demands, and part-time work arrangements. We will continue to study the major components of medical expenditures and how they are changing, and the relationships among trends in disease prevalence, treatment patterns, expenditures, and outcomes. New work will consider the macro implications of health care expenditures, considering whether high-tech expenditures are worth their cost. We also have developed new research treatment variation across the population and issues of differential access to appropriate care. And we will continue to pursue analysis of employer-provided health insurance systems. A new project on survey design and in particular the best way to collect accurate information on the financial status of households is also under way.
*This report has been prepared with the intensive help of Richard Woodbury, NBER and the Maine Center for Policy Research, and the NBER's Program on the Economics of Aging.
1 The Economics of Aging (1989), Issues in the Economics of Aging (1990), Topics in the Economics of Aging (1992), Studies in the Economics of Aging (1994), Aging in the United States and Japan: Economic Trends (1994), Advances in the Economics of Aging (1996), The Economic Effects of Aging in the United States and Japan (1997), Inquiries in the Economics of Aging (1998), Frontiers in the Economics of Aging (1998), Social Security Programs and Retirement around the World (1999), and Issues in Aging in the United States and Japan (forthcoming). An NBER volume was also published this past year on The Evolution of Retirement (1998), by Dora Costa.
2 J. Poterba, S. Venti, and D. Wise, "Personal Retirement Savings Programs and Asset Accumulation: Reconciling the Evidence," in Frontiers in the Economics of Aging, D. Wise, ed. Chicago: University of Chicago Press, 1998.
3 J. Poterba, S. Venti, and D. Wise, "Implications of Rising Personal Retirement Saving," in Frontiers in the Economics of Aging, D. Wise, ed. Chicago: University of Chicago Press, 1998.
4 J. Poterba, S. Venti, and D. Wise, "Lump-Sum Distributions from Retirement Saving Plans: Receipt and Utilization," in Inquiries in the Economics of Aging, D. Wise, ed. Chicago: University of Chicago Press, 1998.
5 J. Poterba, S. Venti, and D. Wise, "Employment Mobility and the Accumulation of Retirement Assets," unpublished working paper, 1999.
6 A. Samwick and J. Skinner, "How Will Defined Contribution Pension Plans Affect Retirement Income?" NBER Working Paper No. 6645, June 1998.
7 J. Shoven and D. Wise, "The Taxation of Pensions: A Shelter Can Become a Trap," in Frontiers in the Economics of Aging, D. Wise, ed. Chicago: University of Chicago Press, 1998.
8 D. Laibson, "Comment: Psychological Perspectives on 401(k)s," in Frontiers in the Economics of Aging, D. Wise, ed. Chicago: University of Chicago Press, 1998; D. Laibson, A. Repetto, and J. Tobacman, "Self-Control and Saving for Retirement," Brookings Papers on Economic Activity, No. 1 (1998), pp. 91-106; D. Laibson, "Life Cycle Consumption and Hyperbolic Discount Functions," European Economic Review, 42 (1998), p. 861.
9 A. Gustman, O. Mitchell, A. Samwick, and T. Steinmeier, "Pension and Social Security Wealth in the Health and Retirement Study," NBER Working Paper No. 5912, February 1997; forthcoming in Wealth, Work, and Health: Innovations in Measurement in the Social Sciences.
10 J. Brown and J. Poterba, "Joint Life Annuities and the Demand for Annuities by Married Couples," unpublished working paper, 1998; J. Brown, O. Mitchell, J. Poterba, and M. Warshawsky, "New Evidence on the Money's Worth of Individual Annuities," American Economic Review, forthcoming.
11 J. Poterba and M. Warshawsky, "The Costs of Annuitizing Retirement Payouts from Individual Accounts," NBER Working Paper No. 6918, January 1999.
12 F. Brown, O. Mitchell, and J. Poterba, "The Role of Real Annuities and Indexed Bonds in an Individual Accounts Retirement Program," NBER Working Paper No. 7005, March 1999.
13 K. McGarry and R. Schoeni, "Social Security, Economic Growth, and the Rise in Independence of Elderly Widows in the 20th Century," NBER Working Paper No. 6511, April 1998.
14 J. Poterba, "Population Age Structure and Asset Returns: An Empirical Investigation," NBER Working Paper No. 6774, October 1998.
15 C. Paxson, "Saving and Growth: Evidence from Micro Data," European Economic Review, 1996; and A. Deaton and C. Paxson, "The Effects of Economic and Population Growth on National Saving and Inequality," Demography, 1997.
16 A. Deaton and C. Paxson, "Measuring Poverty among the Elderly," in Inquiries in the Economics of Aging, D. Wise, ed. Chicago: University of Chicago Press, 1998; A. Deaton and C. Paxson, "Economies of Scale, Household Size, and the Demand for Food," Journal of Political Economy, 1998.
17 A. Deaton and C. Paxson, "Health, Income, and Inequality over the Life-Cycle" in Frontiers in the Economics of Aging, D. Wise, ed. Chicago: University of Chicago Press, 1998; and A. Deaton and C. Paxson, "Aging and Inequality in Income and Health," American Economic Review, 1998.
18 M. Hurd, D. McFadden, H. Chand, L. Gan, A. Merrill, and M. Roberts, "Consumption and Savings of the Elderly: Experimental Evidence on Survey Response Bias," in Frontiers in the Economics of Aging, D. Wise, ed. Chicago: University of Chicago Press, 1998.
19 M. Hurd, D. McFadden, Y. Deng, and L. Gan, "Subjective Survival Curves and Lifecycle Behavior," in Inquiries in the Economics of Aging, D. Wise, ed. Chicago: University of Chicago Press, 1998.
20 R. Lumsdaine, J. Stock, and D. Wise, "Retirement Incentives: The Interaction between Employer-Provided Pensions, Social Security, and Retiree Health Benefits," in The Economics of Aging in the United States and Japan, M. Hurd and N. Yashiro, eds. Chicago: University of Chicago Press, 1997.
21 A Samwick, "New Evidence on Pensions, Social Security, and the Timing of Retirement," NBER Working Paper No. 6534, April 1998.
22 L. Friedberg, "The Effects of Old Age Assistance on Retirement," NBER Working Paper No. 6548, May 1998.
23 C. Coile and J. Gruber, "Social Security Retirement Incentives," Working Paper, 1999.
24 D. Cutler and B. Madrian, "Labor Market Responses to Rising Health Insurance Costs: Evidence on Hours Worked," RAND Journal of Economics, forthcoming.
25 L. Lefgren and B. Madrian, "The Effect of Health Insurance on Transitions to Self Employment," Working Paper, 1998.
26 A. Garber, T. MaCurdy, and M. McClellan, "Persistence and Medicare Expenditures among Elderly Beneficiaries," Frontiers in Health Policy Research, Vol. 1. Cambridge, Mass.: MIT Press, 1998.
27 D. Cutler and M. McClellan, "What Is Technological Change?" in Inquiries in the Economics of Aging, D. Wise, ed. Chicago: University of Chicago Press, 1998; D. Cutler, M. McClellan, and J. Newhouse, "The Costs and Benefits of Intensive Treatment for Cardiovascular Disease," NBER Working Paper No. 6514, April 1998.
28 D. Cutler and E. Meara, "The Medical Costs of the Young and Old: A Forty Year Perspective," in Frontiers in the Economics of Aging, D. Wise, ed. Chicago: University of Chicago Press, 1998.
29 A. Garber, T. MaCurdy, and M. McClellan, "Medical Care at the End of Life: Diseases, Treatment Patterns, and Costs," NBER Working Paper No. 6748, October 1998.
30 M. Eichner, M. McClellan, and D. Wise, "Insurance or Self-Insurance? Variation, Persistence, and Individual Health Accounts," in Inquiries in the Economics of Aging, D. Wise, ed. Chicago: University of Chicago Press, 1998; and M. Eichner, M. McClellan, and D. Wise, "Health Persistence and the Feasibility of Medical Savings Accounts," in Tax Policy and the Economy. Cambridge, Mass.: MIT Press, 1997.
31 M. Eichner and D. Wise, "Little Saving and Too Much Medical Insurance: Medical Saving Accounts Could Help," in Personal Saving, Personal Choice, D. Wise, ed. Hoover Institution Press, 1999.
32 M. McClellan and D. Wise, "Where the Money Goes: Medical Expenditures in a Large Corporation," in Issues in Health Care in the United States and Japan, A. Garber and S. Ogura, eds. Chicago: University of Chicago Press, forthcoming; and M. Eichner, M. McClellan, and D. Wise, "The Sources of Cost Difference in Health Insurance Plans: a Decomposition Analysis," forthcoming.
33 M. Eichner, "The Impact of Intrafamily Correlations on the Viability of Catastrophic Insurance," in Frontiers in the Economics of Aging, D. Wise, ed. Chicago: University of Chicago Press, 1998; M. Eichner, "The Demand for Medical Care: What People Pay Does Matter," American Economic Review, 1998; and M. Eichner, "Choice among Employer-Provided Insurance Plans: Some Preliminary Evidence," in Issues in Aging in the United States and Japan, S. Ogura, T. Tachibanaki, and D. Wise, eds. Chicago: University of Chicago Press, forthcoming.
34 M. Hurd, D. McFadden, and A. Merrill, "Healthy, Wealthy, and Wise? Socioeconomic Status, Morbidity and Mortality among the Elderly," unpublished working paper, 1998.
35 O. Attanasio and H. Hoynes, "Differential Mortality and Wealth Accumulation," Journal of Human Resources, forthcoming.
36 M. McClellan, "Health Events, Health Insurance, and Labor Supply: Evidence from the Health and Retirement Study," in Frontiers in the Economics of Aging, D. Wise, ed. Chicago: University of Chicago Press; and T. Falba and M. McClellan, "The Costs of Illness in Middle Age," unpublished working paper, 1998.
37 J. Gruber and D. Wise, eds., Social Security Programs and Retirement around the World. Chicago: University of Chicago Press, 1999; and J. Gruber and D. Wise, "Social Security and Retirement around the World," American Economic Review, No. 6134 (August 1998).
38 Participants in the project to date include Pierre Pestieau and Jean-Philippe Stijns (Belgium), Didier Blanchet and Louis-Paul Pelé (France), Axel Börsch-Supan and Reinhold Schnabel (Germany), Agar Brugiavini (Italy), Takashi Oshio and Naohiro Yashiro (Japan), Arie Kapteyn and Klaas de Vos (Netherlands), Michelle Boldrin, Sergi Jimenez and Franco Peracchi (Spain), Mårten Palme and Ingemar Svensson (Sweden), and Richard Blundell and Paul Johnson (United Kingdom).
39 This group has included Yukiko Abe (Kyoto University), Seki Asano (Tokyo Metropolitan University), Hiroyuki Chuma (Hitotsubashi University), Tetsuo Fukawa (Institute of Public Health), Tatsuo Hatta (Osaka University), Naoki Ikegami (Keio University), Hiroki Kawai (JCER), Makoto Kawamura (Hosei University), Yukio Noguchi (Hitotsubashi University), Noriyoshi Oguchi (Sunsyu University), Seiritsu Ogura (Hosei University), Akiko Oishi (JCER), Atsushi Seike (Keio University), Haruo Shimada (Keio University), Kenji Shuto (Ministry of Health and Welfare), Reiko Suzuki (JCER), Toshiaki Tachibanaki (Kyoto University), Noriyuki Takayama (Hitotsubashi University), Takeshi Yamada (Health Care Science Institute), Naoto Yamauchi (Osaka University), Naohiro Yashiro (Sophia University), and Fumiaki Yasukawa (Health Care Science Institute).
40 Merton Dagut (Witwatersrand University), Monica Ferreira (University of Cape Town), and Valerie Möller are the key participants from South Africa.
41 A. Case and A. Deaton, "Large Cash Transfers to the Elderly in South Africa," Economic Journal, 108, No. 450 (September 1998), pp. 1330-61.
42 D. Costa, The Evolution of Retirement: An American Economic History, 1880-1999. Chicago: University of Chicago Press.