Dramatic changes in fertility and mortality rates during the 20th century ensure that the world will age rapidly during the 21st century. The speed and pattern of population aging are determined by three principal factors. First and historically most important is the secular decline in fertility rates, which has the effect of producing a gradual but sustained increase in the ratio of older to younger people in a population. Second is the decline in mortality rates, which have fallen substantially in many countries over the past century as the result of a combination of advances in public health (e.g., water purification, antismoking campaigns), medical technology (e.g., the treatment of heart disease), and standards of living (e.g., better nutrition). The final factor is the pronounced but transitory rise in post-World War II fertility rates, usually called the “baby boom,” which appeared with varying intensity and duration in a subset of developed countries. This increase in fertility rates contributed temporarily to younger age distributions, but the aging of the baby boom cohorts will soon accelerate population aging in these countries.
The projected growth in the numbers and proportions of the world's older population poses an array of challenges to policy makers. How do changes in the ratio of workers to retirees affect the ability of societies to fund old-age security systems? Are we living healthier as well as longer lives, or are our added years accompanied by disabilities and generally poor health? In what ways can the structure and the delivery mechanisms of health systems best adapt to the needs of older populations with a higher prevalence of chronic disease? How do changing family struc-
tures affect the demand for public transfers of money, time, and living space? Will population aging lead to lower levels of aggregate saving, investment, and productivity growth? Will health care costs rise or decline relative to other costs?
Three other important features of population aging are also noteworthy. First, there are uncertainties about how some of the demographic forces will play out. For example, will increases in life expectancy accelerate with the development of new technology? Will persistent below-replacement fertility levels compel societies to alter immigration policies? How might the socioeconomic characteristics of tomorrow's elderly population differ from those of today? Second, because population aging generally is a gradual phenomenon, its socioeconomic consequences tend to appear gradually as well, and in some cases with a fair degree of predictability. Thus if policy makers recognize and appreciate the import of the coming changes, they will have a window of opportunity in which to develop policies and programs for coping with the stresses induced by changing population age structures. And third, most statements about aged individuals tend to reflect averages and mask a great deal of diversity in the population. For example, while a greater proportion of the elderly 1 than the nonelderly have some degree of disability, most of the elderly are not disabled. Likewise, there are prominent socioeconomic gradients among older populations in most countries that require explication and policy response.
As we move through the 21st century, countries around the world are apt to face slower growth (or even contraction) of the workforce, rapid increases in the over-65 and especially the over-80 population, potentially larger numbers of disabled persons and greater demand on health care systems, and the increase in poverty likely to accompany rising numbers of widows. Many countries are now in the early stages of adapting to their changing population age structures. Since current and prospective policy responses are likely to differ among countries, a number of natural experiments are, or shortly will be, under way, enabling countries to learn from each other's experience. To take advantage of this opportunity, the U.S. National Institute on Aging asked the National Academies, through its National Research Council, to convene a panel that would provide recommendations for an international research agenda and for the types of data needed to implement that agenda in the context of rapid demographic change.
1There is a growing awareness that the category “elderly” is an inadequate generalization that conceals the diversity of a broad age group spanning more than 40 years of life. For cross-national comparative purposes, however, some chronological demarcation of age categories is required. In this report, the term “elderly” refers to persons aged 65 and over, and the term “oldest old” to persons aged 80 and over.
This study focuses on five domains of research: work and retirement, savings and wealth, family structure and intergenerational transfers, health and disability, and well-being. Recommendations specific to each of these topics are included in the respective report chapters. The panel also developed six major, overarching recommendations that it believes are essential to effective cross-national research and to the generation of policy-relevant data.
- I. The development and use of multidisciplinary research designs are crucial to the production of data on aging populations that can best inform public policies.
The range of topics addressed in this volume illustrates the need for multidisciplinary work that cuts across research domains. Recent demonstrations of the importance of cross-domain relationships—between health and retirement decisions, between economic status and health, between family structure and well-being in older age—amplify the contention that public policy must be guided by an understanding of the interplay of multiple factors. Initiatives in Europe, the United States and Asia that integrate several salient domains of people's lives into single survey instruments have proven to be successful prototype data collection efforts. Study domains include income and wealth, labor force activity and retirement, health status and utilization of health care facilities, cognition, and intergenerational transfers. The panel believes these models can serve as building blocks for research in many countries, including countries that are less developed as well as those more developed.
It is the panel's conviction that the optimum way to develop both the research agenda and the data needed to deal with the economic and social issues associated with an aging world is through the ongoing interaction of multidisciplinary national scientific communities. Extended interaction among demographers, economists, epidemiologists, social psychologists, sociologists, and statisticians is absolutely essential to (1) the creation and refinement of harmonized measures (conceptually comparable across societies) that are necessary to understand outcomes such as labor force participation, health and disability status, complex family relationships, and economic status; and (2) the development of databases that can be used to maximize the potential of cross-national and cross-temporal research for identifying the determinants of critical outcome variables.
It is important to stress that potential gains will not be realized unless there is a continuing and effective dialogue between the policy community and researchers that results in the design of a program of data collection that can properly inform policy makers. This dialogue must be ongoing
since many of the key dimensions of population aging can be expected to change as socioeconomic circumstances evolve.
- II. Longitudinal research should be undertaken to disentangle and illuminate the complex interrelationships among work, health, economic status, and family structure.
There is a pressing need in most countries for longitudinal microdata that include extensive measures of economic status, financial incentives to retire, various aspects of health status, and intergenerational relations and transfers. Such data are needed to better understand patterns of age-related transition along these dimensions, interrelationships among the dimensions, and ultimately the ways in which these domains contribute to overall well-being. One can anticipate with some degree of certainty the demographic parameters and trends that give rise to broad policy issues. Much less is known about individual responses to policy interventions, for example, the labor supply response of 60-year-old men and women to a restructuring of public pension plans that raise the early retirement age from 62 to 65. Ultimately, policy options are grounded in understanding individual and family behaviors and their responsiveness to changing life circumstances.
From a research standpoint, the variation in response patterns associated with changing circumstances implies the need for panel studies that trace cohorts over time. Studies can be repeated cross sections, single-cohort panel studies, or panel studies that continue to add new cohorts at the bottom end of the age range and are thus continually representative of the study population. If affordable, panel studies that add new cohorts are clearly best, since they not only capture the dynamics of change over time for individuals, but also continue to describe the broader population and not just a single cohort. Because the world is dealing with a phenomenon (population aging) that is likely to require the careful attention of policy makers for at least the next five decades, neither repeated cross sections nor single-cohort designs are very attractive. Interestingly, panel studies that add new cohorts may be less expensive than repeated cross sections with the same frequency, sample size, and length, simply because the cost of reinterviews is much lower than that of initial interviews. (The reason reinterviews are less expensive is that contact costs are lower; for example, telephone calls can be made to set up appointments.)
It is crucial to note that the focus of panel studies on aging should not be restricted to the upper ends of the age spectrum. We know that the characteristics of tomorrow's cohorts of elderly will be very different from those of today and will be determined by lifelong experiences. Within the bounds of practicality, surveys need to capture as much of the life-course experience as possible.
- III. National and international funding agencies should establish mechanisms that facilitate the harmonization (and in some cases standardization) of data collected in different countries.
The panel believes major scientific and policy gains would be possible if a number of countries could be induced to embark on data design and collection activities that would provide a rich set of comparable (i.e., harmonized) data. Advantages would arise from the confluence of several factors: the differential rates of population aging throughout the world that result from differences in fertility and mortality histories and thus provide a unique opportunity for countries to learn from each other's experiences; the concomitant economic and social changes (e.g., in pension reform, marriage and divorce rates, schooling levels, adoption of innovative medical technology) that are occurring differentially throughout the world; and the growing awareness among policy makers that problems resulting from global aging pose what is arguably the most important set of economic and social challenges they will face over the next half-century. To benefit from the possibility of exploiting institutional differences to understand the effects of policy measures, data collection efforts in different countries must be harmonized in the sense that conceptually comparable information is collected, and procedures (e.g., for sampling and quality control) are synchronized to the extent possible. Much of this harmonization can probably be achieved through extensive exchange of information among scientific groups working on new data collection efforts.
This emphasis on harmonization does not imply that survey protocols need be identical in all countries. The track record of prior attempts to impose standardized data collection approaches across countries is mixed at best. Each country has unique institutional features and policy priorities that shape data collection and research. To illustrate, while all countries are likely to regard estimates of household wealth as an important element of their data collection activity, they are unlikely to measure the same components; for example, only the United States has 401(k) plans in household wealth portfolios. On the other hand, disability, disease, and functional health need to be measured in a standardized way to enable accurate cross-national analysis. The objective is to enable researchers to estimate accurate models of the incentives to work, to retire, and to save, and make it possible to link these patterns to other important domains of older peoples' lives.
- IV. Cross-national research, organized as a cooperative venture, should be emphasized as a powerful tool that can enhance the ability of policy makers to evaluate institutional and programmatic features of policy related to aging in light of international experi-
ence, and to assess more accurately the impact of potential modifications to existing programs.
Cross-national studies conducted within a framework of comparable measurement can be a substantially more useful tool for the analysis of policy impact than studies of single countries. A cross-national perspective provides a broader and richer set of institutional arrangements within which to understand policy initiatives, and offers opportunities to relate variations in institutional arrangements to the distribution of attributes that determine program eligibility, benefit levels, and ultimately individual and household behaviors.
Sophisticated comparative analyses can exploit differences and changes in policy rules across countries by isolating their impacts from those of other macroeconomic and social changes. One penetrating example of cross-national research on 11 developed countries revealed three important features that could not easily have been discerned from single-country studies. First, the data showed a strong correspondence between early and normal retirement ages and the probability of departure from the labor force. Second, public pension provisions in many countries were found to place a heavy tax burden on work past the age of early retirement eligibility, and therefore to provide a strong incentive for early withdrawal from the labor force. Third, this implicit tax—and hence the incentive to leave the labor force—varied substantially among countries, as did retirement behavior. Thus, considering comparisons across the countries made it possible to draw several general conclusions about the relationship between retirement incentives and retirement behavior.
More generally, at least three conditions must be met to provide an accurate assessment of policy impacts on behavior. Thinking of the policy as a treatment, (1) there must be a sizable comparison (untreated) group with observable characteristics similar to those of the treatment group; (2) the comparison group must be unaffected by the policy (no spillover effects); and (3) the treatment and comparison groups must be subject to the same socioeconomic trends over time. Cross-national comparisons can help on all three counts. Policy interventions typically occur in one country but not elsewhere, meaning that valid comparison groups generally exist across but not within countries. Comparison groups in other countries are unlikely to be affected by a policy intervention in one country, so that spillover effects within countries do not necessarily distort the estimated impact of the intervention. Finally, comparison groups can be selected on the basis of characteristics that suggest relatively similar life experiences; for example, individuals with high incomes and education levels can be compared across countries. And even when within-country
variation is informative, cross-country comparisons can add substantially to the variability in the data and thereby improve the precision of a policy intervention's estimated impact.
- V. Countries should aggressively pursue the consolidation of information from multiple sources to generate linked databases.
The integration of different types of information (e.g., survey, census, administrative, medical) produces a dataset whose depth and explanatory power exceed what is possible for any single source. The advantage of linking survey data with administrative records is that the latter are likely to contain extended histories that could not be obtained from a survey, or if obtainable, would be associated with significantly higher measurement error. Under ideal conditions, therefore, administrative records can provide unbiased measures of change over time for a standard set of concepts. The ability to merge data of this sort with data tailored to the analytic issues addressed by surveys clearly has major advantages.
Beyond the scientific advantages, the linking of administrative and other information with survey data reduces respondent burden, a not-insignificant factor given the complexities of survey research instruments and the sometimes strong cultural reluctance to participate in survey endeavors. And finally, close attention must be paid to novel and potentially revolutionary ways of gathering data. The likelihood that a large majority of households in many countries will soon be connected to the Internet, for example, opens up promising new methods of data collection.
- VI. The scientific community, broadly construed, should have widespread and unconstrained access to the data obtained through the methods and activities recommended in this report.
Good data are public goods for both policy and research. Scientific advances and policy insights that may emerge from the development of a dataset are greatly enhanced if a broad community of scientific users with different interests, theoretical perspectives, and models have ready access to the information. Moreover, the best way to identify errors in data is through the user community. The Health and Retirement Survey (HRS) in the United States, begun in the early 1990s and soon to be entering its second decade, is a prominent example of how data should be made available to the research community at large. Perhaps the most important reason for the widespread use of HRS data is that they are made available in a timely fashion on the Internet to scientists and policy makers alike. More than 300 scientific papers, many by non-U.S. researchers, have been written using these data.
The track record and protocols for access to data in many countries tend to discourage use of the data. In various European and Asian countries, researchers' access to data is severely limited or unnecessarily costly in terms of time and/or money. Because of restricted access and the limited role of scientists in the design of surveys, scientific innovation in the collection of data is hampered. Moreover, many of the best scientists in these countries often choose to use data from other countries to test their ideas since it is too difficult to use their own national data. The panel recognizes that all surveys involve legitimate and thorny issues of privacy and confidentiality that must be explicitly addressed and resolved. There are, however, statistical and legal methods for preserving confidentiality that can be used without unduly limiting scientific access to data.
FIVE RESEARCH DOMAINS
As noted, an inexorable demographic momentum will produce increasingly aged populations in most if not all nations of the world. There is, however, ample reason to believe that nations will be able to cope with the current and projected demographic changes provided policy makers have access to information about the emerging economic and social forces that will shape future societal well-being. The panel focused on five interrelated research domains in which new international data are required to inform policy making in the coming decades. Their relative importance is likely to differ by society and may change within a society over time. But if countries want to use current natural experiments throughout the world to improve their own adaptations to population aging, these are the five essential domains that must be addressed by international data and research.
Work, Retirement, and Pensions
The declining labor force participation of older persons in many parts of the world is one of the most dramatic economic trends of the past four decades. While the decline appears to have leveled off or even reversed in the last decade in some Organization for Economic Co-Operation and Development (OECD) countries, it has done so at historically low rates. Moreover, the decline has continued to occur in other, quite similar OECD countries. The increasing financial pressure faced by public pension systems around the world is often attributed to demographic trends that have led to aging populations. But decreasing labor force participation rates for a given age structure also contribute to financial imbalances within pension programs, further increasing the number of retired persons relative to those in the workforce. Ironically, public pension schemes
in some countries, while providing protection against income loss in retirement, have themselves offered incentives to leave the labor force early, thus by their very structure exacerbating the financial problems countries face.
Two central features of public pension plans have an important effect on work incentives. The first is the age at which benefits are initially available, called the early retirement age. The extent to which people continue to work after they reach early retirement age is closely related to the second important feature of plan provisions, the pattern of benefit accrual. The most important determinant of benefit accrual is the adjustment to benefits if a person works for another year. In some countries, there is an “actuarial” adjustment such that benefits are increased to offset the fact that they are received for fewer years. In other countries, however, there is no such adjustment. If there is no adjustment, or if the adjustment is not large enough to offset the fewer years of receipt of benefits, the result is an incentive to leave the labor force.
In many countries, disability and unemployment insurance programs effectively provide for early retirement before the explicit early retirement age. Such programs must be considered in conjunction with the public pension program itself when social security reform is contemplated. One must be able to calculate accrual under various programs, in addition to accrual under the public program. Ideally, one should also know which paths to retirement are available to each person. A person who is eligible for disability benefits, for example, typically experiences much stronger early retirement incentives than one not eligible for these benefits. Eligibility usually depends on program provisions and individual circumstances that are often difficult to determine solely from administrative or survey data.
While the public pension system is the principal source of retirement benefits in many countries, in other countries it is only one of several important sources. In the United Kingdom, the Netherlands, the United States, and elsewhere, employer-provided pension plans are a key source of benefits; some retirees receive defined benefits in the form of annuities from employers in addition to payments from the public pension system. Indeed, in some countries (e.g., the United States), employer-provided benefits are often integrated with publicly provided benefits. The incentive effects of these private defined-benefit plans are quite similar to those of public pension programs. For many developing countries that are just designing pension programs with private as well as public components, there are opportunities to learn from the trials and errors of wealthier nations.
Private Wealth and Income Security
Income security during retirement, coupled with an increase in post-retirement years during which individuals can enjoy family and leisure, is one of the primary social achievements of the 20th century in most industrialized nations. At the same time, this accomplishment has introduced some fundamental public policy challenges associated with population aging. The two most basic challenges are (1) that individuals will have sufficient income security during their retirement years so that retirement does not necessarily imply a substantial decline in living standards, and (2) that individuals will have protection against the increasing risk of experiencing periods of poor health. From a policy viewpoint, these challenges involve a number of critical interrelationships, including those between health and economic status (i.e., income and wealth), between public policy and individual behavior (in particular, the extent to which public income support discourages private savings), and between demography and the stock market. The latter interaction is particularly important for countries that rely heavily on individual choices to provide for old age.
Although economists have gained a better understanding of individual saving and investment decisions and how these decisions interact with policy, numerous questions remain. Some are fundamental, such as whether the elderly dissave in old age (they should, according to economic theory) and whether consumption levels change at retirement (by and large they should not, according to economic theory). Others are directly related to policy, such as whether generous public or private pension provisions seriously attenuate individual incentives to save for retirement (they appear to); whether we can explain the current low savings rate in the United States (capital gains appear to be an important part of the explanation); and whether the recent runup of the stock market was driven by demography (there is some evidence that it may have been, and if so, this has important consequences for the wealth position of the baby boomers when they retire).
Clearly, such questions can be answered only if adequate data are available. Partly on the basis of research commissioned for five countries (Japan, the United Kingdom, the Netherlands, Italy, and Germany), we provide an overall assessment of datasets available in various countries and their potential to support investigation of important research and policy questions such as those posed above. In none of the countries except the United States is there substantial information about health in combination with wealth. Given the important interaction among health status, work status, and the amount of wealth accumulated over the life course, this is a serious omission. Although quality problems, including
substantial underreporting of several asset categories (in particular, stocks and bonds) and selective nonresponse, are associated with many surveys, our general verdict is that the quality of the existing wealth surveys is surprisingly good. Thus careful cross-national comparisons using the data from these surveys are warranted. In many other countries, however, such data for addressing key policy issues do not currently exist and will have to be developed. We have identified strategies for dealing with these limitations.
The well-being of older persons depends to a large extent on the content and volume of an intricate set of transfer systems in which they are engaged over their lifetimes. In industrialized countries, the most salient of these systems are associated with individual savings behavior (representing transfers over individual life cycles), exchanges with family members, and intergenerational transfers from current workers to non-workers through many types of social security and pension programs. Individuals usually participate in a variety of transfer systems concurrently, with the mix varying across individuals and societies, as well as over the life course. The older generation in many countries is particularly likely to be involved in a complex series of exchanges involving public programs, employers, and family members, with significant shifts in salience and magnitude often occurring with age.
Although considerable progress has been made in understanding the volume, content, and implications of each major transfer system, many gaps remain. Particularly deserving of attention are interrelationships across systems and how changes in one system (for example, public pensions) affect another (for example, the level and direction of family support). Insofar as different transfer systems overlap and compete, sound policy and program development requires the best possible knowledge of how this marketplace operates, how responsive its different elements are, and what efficiencies and inefficiencies exist.
Many of the key transfer systems can be understood and measured at both the individual and family levels, as well as at the societal or macro level. This complexity poses further challenges to the tracing of interrelationships across systems. A microbehavioral approach is the major perspective for studying family exchanges, while macroeconomic analysis is the dominant strategy for studying public and employer transfer arrangements. The mix of transfer programs and their salience varies considerably between the more and less developed economies and within each group as well. This variation in approaches, incentives, and reactions to new programs and societal changes provides a continuum of experi-
ence whose study can greatly illuminate understanding of the dynamics of each system and the interrelationships across systems.
The Health of Aging Populations
Health plays an important role in the social outcomes associated with the other substantive research areas discussed in this report: retirement, wealth accumulation, transfer systems, and well-being. Hence it is critical that appropriate health data be collected to populate models that can be used to explain differences in these substantive outcomes within and across countries. At the same time, health is a critical policy variable in its own right. Although population aging may or may not result in increasing proportions of elderly persons in poor health, the numbers experiencing that condition are almost certain to increase. Thus as the populations of all industrialized countries and most developing countries age, the social and economic demands on individuals, families, communities, and nations will grow, with a substantial impact on formal and informal medical and social care systems and on the financing of medical services in general.
In conjunction with growing numbers of older persons, most nations face secular changes in health status, as reflected in rates and outcomes of various conditions and disabilities. For example, rates of heart attack and stroke are decreasing in many cases, while such trends may not be obvious for cancers. Some countries are experiencing rising rates of acquired immune deficiency syndrome (AIDS) at older ages. Research increasingly suggests declines in rates of severe physical disability among older persons in developed nations, although the cross-national comparability of such findings remains problematic. Trends in cognitive impairment and dementia have enormous policy implications, but our time-series knowledge base is lacking. Whether changes in disease and disability rates alter the rates of long-term institutionalization within and among nations is uncertain as well.
While all countries must address the changing health needs of their older citizens, their approaches are surprisingly diverse in terms of health system organization, administration, and financing. At the same time, national health systems themselves are in transition. Administratively centralized national health systems are being decentralized. Quality-of-care assurance mechanisms are increasingly in vogue. The diversity among nations in approaches to health system structure, program content, and financing argues for the value and utility of cross-national research for those attempting to improve the levels and quality of both preventive and therapeutic health services while maintaining budget discipline and responsiveness to cultural and social expectations.
However, effective cross-national and cross-cultural research demands suitably comparable data on health status and outcomes, informal and alternative sources of medical and social services, and health system administration and financing. These data must come from a variety of sources, including vital statistics and administrative records; research studies; population censuses; health outcome evaluations; and geographically defined longitudinal population surveys that assess the trajectories of the health, social, and economic status of families with older persons. Such information frequently is nonexistent, inaccessible, or nonstandardized, making it difficult if not impossible to conduct high-quality, informative cross-national research on health.
The supreme criterion by which the success of a government can be judged is the quality of life its citizens experience from birth to death. Questions of well-being are particularly important during major life transitions, both because such transitions provide a specific frame of reference for the assessment of well-being and because they involve significant changes in activity patterns. Proposals for the continuing assessment of well-being have been put forth for at least a half-century. Despite these proposals and a wide range of research efforts by academics, there is no consensus on the conceptual definition, conceptual levels, or measurement of well-being. Thus there is a pressing need for review and integration in this critical area.
The economic distinction between stocks and flows suggests a framework for helping to meet this need. Stocks reflect the cumulative impact of past history on a current state; a person's total assets or wealth is thus a stock. The change in that person's assets or wealth during some specific time period represents the corresponding flow. The analogy to subjective well-being is obvious. A person's satisfaction with his or her job, marriage, or life as a whole is a stock; a change in any of those satisfactions during a specific time period, or a change that results from a particular activity, is a flow.
Employment is a major source of well-being, for both extrinsic and intrinsic reasons. Remuneration determines access to most of the resources society offers—from housing, to clothing, to transportation, to entertainment, to higher education, and on through the long list of goods and services for sale. But one's job is also a source of more intrinsic satisfactions (or dissatisfactions)—the opportunity to use valued skills and abilities and to acquire additional ones, the sense of involvement in a larger productive process, and the cooperative and friendly relationships that may develop in the workplace. For all these reasons, the transition from
paid employment to retirement or from full-time paid employment to part-time or temporary postretirement work is of great importance for individual well-being. The process of aging is in most respects gradual; in contrast, retirement for many people is an abrupt transition. Whether it comes as welcome or unwelcome, retirement signals significant changes in an individual's (or couple's) pattern of activity, use of time, and network of social relations. It is a life event of major importance with implications for both the stock and flow of well-being.
Cross-national research on aging should include the development of overall measures of subjective well-being (life satisfaction), as well as further development and assessment of measures of emotional states and moods. Cross-national comparisons should also include the experience of major life events, especially those stressful events that are increasingly likely in old age, such as illness, bereavement, retirement (complete or partial), and changes in activity patterns.