Conclusion and Major Recommendations
The inexorable momentum toward increasingly aged populations around the world may well become the most significant demographic process of the 21st century. Sustained shifts in population age structure will require innovative national and international policy responses. For these responses to be effective in optimizing societal well-being, they must be based on an enhanced scientific understanding of the critical dynamics associated with population aging, such as the determinants of retirement decisions; the links among labor force participation, health status, and economic status; the relationship between retirement decisions and the specific features of both public and private pension plans; the impact of changes in public transfer systems on private transfers; the relationships among aging, income, and private savings; and the impact of medical technology on health, disability, and longevity. The preceding chapters have focused in turn on these domains and their important interrelationships and offered recommendations for data and research in each area. Beyond those domain-specific recommendations, the panel developed six major, overarching recommendations that we believe are essential to effective cross-national research and to the generation of policy-relevant data for an aging world.
- I. The development and use of multidisciplinary research designs are crucial to the production of data on aging populations that can best inform public policy.
The range of topics covered by the preceding chapters illustrates the
need for a multidisciplinary approach 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—support the contention that public policy must be guided by an understanding of the interplay among multiple factors. Initiatives in the United States, Europe, and Asia that integrate several salient domains of people's lives into single survey instruments have proven to be successful prototype data collection efforts. Examples of such endeavors include the Berlin Aging Study, the U.S. Health and Retirement Study (HRS), the Taiwan Study of the Elderly, the National Survey of Midlife Development in the United States, and the German Socio-Economic Panel. From their inception, these studies have included some or all of the following domains: income and wealth, labor force activity and retirement, health status (including biologic measurement) and utilization of health care facilities, cognition, and intergenerational transfers. The panel believes these models can be (and in some cases have been) successively adapted and used in many countries, both more and less industrialized.
It is the panel's conviction that the optimum way to develop both the research agenda and the data needed to address the economic and social issues associated with an aging world is through ongoing interaction among multidisciplinary national scientific communities. We believe extended interaction among sociologists, economists, demographers, epidemiologists, social psychologists, and statisticians is essential to (1) the creation and refinement of harmonized measures (conceptually comparable across societies) needed 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 maximize the potential of cross-country and cross-time research for identifying the determinants of critical outcome variables. To deal effectively with differences among countries in policies, institutions, and incentive structures, it is equally essential that the multidisciplinary dialogue be driven by appropriate theories and models and that the data requirements of these theories and models be the main criteria used to select the empirical content of studies on aging populations.
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, leading to 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 shift as socioeconomic circumstances change.
- 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 benefits that raises 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 and sample size simply because the cost of reinterviews is much lower than that of initial interviews.
It is crucial to note that the focus of panel studies of 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 are 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 to be identical in all countries. Each country has unique institutional features and policy priorities that should help 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 if useful cross-national analysis is to be possible.
The track record of prior attempts to impose standardized data collection approaches across countries is mixed at best. Thus, the goal can accurately be described as harmonization rather than standardization (although, as noted, standardization is essential in some cases). 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 experience,
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, as described in Chapter 3, 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, similar to those noted in Chapter 4 with regard to the Netherlands CentER panel.
- 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 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 the data.
In summary, the enhanced scientific understanding needed to provide effective guidance for public policy in many countries will depend on the generation of longitudinal databases that contain representations of the critical sets of variables needed to model aging processes. The beginnings of such rich longitudinal and multidisciplinary data systems are available in the designs of various surveys mentioned throughout this report. While these are good models from which to start, what is clearly needed is a multinational version(s) of these models that takes account of differences in the nature and structure of institutions in both developed and developing countries.