Click for next page ( 2


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 1
Executive Summary When considering proposed legislation, government decision makers regularly call for estimates of the projected costs and other effects of current and alterna- tive policies. The more complex and far-reaching a proposed change, the more needed are accurate estimates of its likely consequences but the more difficult it is to develop such estimates. The grueling debate in 1993-1994 over reforms to the nation's health care system was handicapped by highly uncertain estimates of the effects of alternative proposals. Another major debate is pending over the retirement income security of older Americans and the costs to workers and the economy of providing that security. Questions such as the following are being raised: Can the U.S. Social Security system continue to provide an adequate level of benefits in view of demographic trends that imply a steadily increasing ratio of beneficiaries to contributors? With growing competitive pressures on employers to constrain employee compensa- tion costs, will maintaining pension and retiree health care benefits (or extending them to uncovered workers) take lower priority than maintaining wages? Will most employees gain or lose from the increased opportunities to direct their own pension investments? Will income from personal savings and transfers from family members be adequate to help finance consumption in retirement and make up for any shortfalls in Social Security or employer benefits? Will the changing form of retirement benefits, particularly the increased use of lump-sum payouts at retirement that may not provide lifetime income, adversely affect certain groups of the elderly, such as widows? If more people decide to work to older ages in order to provide for their retirement, will jobs be available for them? Even if 1

OCR for page 1
2 retirees have adequate ASSESSING POLICIES FOR RETIREMENT INCOME lid incomes, will they still be secure if they are not adequately insured against rising costs of health care services? The policy debate about retirement income security will likely consider a wide range of possible changes to the Social Security, Medicare, and Medicaid programs, as well as changes in tax provisions and regulations covering employer pensions and other forms of savings, such as Individual Retirement Accounts (IRAs). Anticipating the debate, the Pension and Welfare Benefits Administra- tion in the U.S. Department of Labor asked the Committee on National Statistics to establish a Panel on Retirement Income Modeling. The National Institute on Aging, the Pension Benefit Guaranty Corporation, the Social Security Adminis- tration, and TIAA-CREF also provided support for the project. The panel was charged to answer the question: What do we need to know, and what kinds of models do we need to have, in order to estimate the short-run and long-run implications of current retirement-income-related policies and proposed changes to them? PRIMARY CONCLUSION: THE IMPORTANCE OF DATA AND RESEARCH Assessing policy options requires the use of projection models, research, and data. Projection models, which take a variety of forms, are the tools that estimate the likely effects of current and alternative policies. They answer "what if" questions and project future outcomes by using data and research from a wide range of sources. Research findings, in turn, come from analytical models that seek to answer "why" such as why people save or fail to save and to estimate the strength of key relationships by applying theoretical constructs to empirical data. Data from panel surveys, which follow the same individuals over time, are particularly important to answer key research questions. One impetus for the work of our panel was to give policy analysts working on retirement income security a head start on developing reliable broad-based projection models. The goal was to avoid the chaotic conditions of the 1993- 1994 health care debate, in which policy modelers scrambled to keep up with the need for estimates and in which estimates developed by different models for the same proposal varied widely. On the basis of published reports and interviews with analysts, however, we determined that the problems with estimating the likely effects of proposed health care changes were due in only small part to the deficiencies of projection models as such: rather, models were significantly handicapped by the poor quality and lack of relevant data and behavioral re- search. For example, many proposals relied heavily on employer-based health insurance coverage, yet little information was available on employers' current health care costs or on their responses to changes in costs. Thus, it was hard to model the baseline situation, let alone develop widely shared estimates of em- ployer behavior under alternative policy regimes.

OCR for page 1
EXECUTIVE SUMMARY 3 As with health care, there are large gaps in needed data and research for retirement-income-related modeling, with regard both to individual behavior and the behavior of employers. It is critical that these data and knowledge gaps begin to be filled. Generally speaking, it takes more time to collect new data and analyze them than it does to build a model to use data and research results. There are more than a few instances in the history of policy analysis when complex projection models were built in a span of weeks or months. It is rare that a critical unmet data or analysis need can be supplied in that short a time. We conclude that, given limited budget resources, the U.S. Department of Labor and other agencies should give priority to funding retirement- income-related data collection and behavioral research in preference to significant investments in large-scale projection models. It is particu- larly important for the federal government to take the lead in data collection, so that high-quality, comprehensive databases are available for policy analysis use in all sectors government, private, and aca- demic. PRIMARY RECOMMENDATIONS In keeping with our primary conclusion, our primary recommendations are for data and data-based analytical research. Much of the needed information about individual behavior can be obtained from continuing and analyzing existing panel surveys that are linked to selected administrative records. However, information with which to analyze the behavior of employers will require significant enhance- ment of existing databases and, possibly, new data collection as well. When relevant data and research insights are more complete, it will be important to develop improved large-scale models for projecting the likely effects of a variety of alternative policy proposals on employers, workers, and retirees. Until then, it makes little sense to allocate scarce resources for the development of complex new projection models, and we recommend against doing so. In the short term, agencies are best advised to address specific policy questions with limited, spe- cial-purpose projection models and the best available data and research findings. The Research Base The long-run outcomes of retirement-income-related policies depend crucially on how workers and employers modify their behavior. For example, to understand the ramifications of pension simplification, one needs to understand the extent to which employers will set up new pension plans and the extent to which workers will elect to participate in them. One also needs answers to such questions as whether newly participating workers will save more or less on their own than

OCR for page 1
4 ASSESSING POLICIES FOR RETIREMENT INCOME they did previously and whether they will obtain retirement income from their pensions or take and use the benefits as lump sums. Research is needed for retirement-income-related policy analysis in such areas as: factors that affect employers' demand for older workers and their deci- sions to adopt, modify, or terminate pension, disability, and retiree health care benefit plans; factors that affect people's decisions to consume or save and their choices of labor and leisure over the life cycle, including the decision to retire; the demographic processes of fertility, immigration, emigration, and par- ticularly mortality, in order to project the numbers of people each year by age who form the population of potential workers or retirees at any given time; trends in marital status and family composition of older people, including interactions with trends in the form of retirement benefits (e.g., increased use of lump-sum payouts), and the implications for people's incomes; and trends in health status, health care costs, and health care financing ar- rangements of older people and the implications for income adequacy for other needs. The knowledge base is deficient in many of these areas, largely due to deficiencies in data. In some instances, the data deficiencies have hampered the development of theory; in other instances, they have impeded the development of robust estimates of behavioral parameters. Data development and basic research need to go forward hand in hand. Data on Individuals Fundamental understanding of the determinants of individual savings and con- sumption decisions is lacking. For example, there is considerable uncertainty about the reasons for the decline in personal savings rates over the past decade and for the wide disparity in savings levels across population groups. Thus, there is little basis on which to project the likely effects on savings decisions and, hence, on retirement income security, of changes in employer pensions and gov- ernment policies. Considerably more is known about labor supply and retirement behavior. However, the primary data source for analysis, the Retirement History Survey, is over 15 years out of date and does not provide adequate coverage of women in the work force. New panel surveys sponsored by the National Institute on Aging, the Health and Retirement Survey (HRS) and its companion, Asset and Health Dynamics Among the Oldest Old (AHEAD) survey, are now beginning to provide appropri- ate longitudinal data for research on savings and for refinement of labor supply

OCR for page 1
EXECUTIVE SUMMARY s models knowledge that is essential for the development of improved policy projection models. However, these surveys have not yet collected sufficient waves of data for analysis purposes; they must be continued. Other panel surveys (e.g., the National Longitudinal Surveys of Labor Mar- ket Experience) can also support relevant analysis. They need to be continued and regularly reviewed, together with HRS and AHEAD, to determine ways to improve their usefulness for retirement-income-related research. Existing panel surveys of middle-aged and older people, particularly HRS and AHEAD, should receive continued government support. These surveys should be refreshed periodically with new cohorts in order to offer insight into how behavior changes over time. Agencies and researchers involved in the major retirement-income- related panel surveys of individuals should collaborate regularly to iden- tify ways to improve data quality and utility. The National Institute on Aging should facilitate such collaborative efforts. Data on Employers Important information on the characteristics and determinants of employer de- mand for older workers and employer benefit plan offerings is lacking, incom- plete, or not provided in a usable manner. For example, no information is avail- able on how many and what kinds of employers recently offered special incentive plans for early retirement such as "window" plans or intend to do so in the near future. Given the central role of employers in providing retirement income and health care benefits, the lack of an adequate database is a major handicap to evaluation of alternative policy proposals in these areas. Collecting data for employers presents difficulties (e.g., low response rates) that can make it expensive to obtain high-quality results, but it is critical to do so. In order to make the best use of scarce resources, priority should be given to the development of creative ways to refine and extend existing data systems on employers, such as the Bureau of Labor Statistics Employee Benefits Survey, the Department of Labor Form 5500 data series, and the employer components of the HRS and AHEAD surveys. Consideration should also be given to the feasibility and cost-effectiveness of a panel survey (refreshed periodically) that collects detailed information on employers and their workers for behavioral analysis. The development of analytically useful databases on employers for re- tirement-income-related policy analysis and projection purposes should be an important government goal. The U.S. Department of Labor should establish an interagency task force on employer data to specify an integrated data collection plan. The employer data collection plan should include short-term and long-term goals for obtaining improved information on the distribution

OCR for page 1
6 ASSESSING POLICIES FOR RETIREMENT INCOME across employers of benefit plan offerings (including pensions, health insurance, disability insurance, retiree health insurance, life insurance) and on labor demand for older workers and the factors that may affect that demand. Expanded Use of Administrative Data Some of the information that is essential for analysis of savings and retirement decisions and the effect of medical care use and expenditures on retirement income security is most cheaply and accurately obtained from existing adminis- trative records. To be useful for estimation, this information (e.g., Social Secu- rity earnings histories, Medicare and Medicaid benefits) needs to be linked to the broad array of data for individuals that is available from such panel surveys as HRS and AHEAD. Similarly, linkage of employer survey data with administra- tive records could provide enhanced analysis and modeling capability at low marginal cost, although issues of confidentiality protection are more difficult for employer data. Matched files of survey responses and key administrative records should be regularly produced for retirement-income-related policy analysis and projection purposes. Agencies should vigorously explore creative solu- tions for providing research access to exact-match files that safeguard the confidentiality of individual responses. Data Quality Data collection efforts must include resources for evaluation of the quality of the information, both internally (e.g., by experimentation with alternate question wording) and in comparison with other data sources. For example, worker re- ports of pension and health care benefit coverage may differ from employer reports of such coverage, and the differences need to be understood in order to develop reliable analytical and projection models. Budgets for retirement-income-related surveys should include sufficient resources for regular evaluation of data quality. Organizational Concerns Responsibility for major retirement-income-related policy levers (e.g., employer pension regulation, Social Security, tax policy, health care programs) is spread among several agencies in different departments. Relevant data collection and research are also spread across these agencies and others. There are benefits of having many players involved in this complex policy area. However, in our observation, not enough attention has been given to the need for interagency

OCR for page 1
EXECUTIVE SUMMARY 7 coordination to identify data gaps and overlaps and to help establish priorities in light of resource constraints. We earlier recommended an interagency employer data collection task force and collaborative efforts to coordinate retirement-re- lated panel surveys of individuals. Such efforts could usefully relate to the Interagency Forum on Aging-Related Statistics. Coordination mechanisms should recognize the need for flexibility and experimentation and not impose added bureaucratic requirements. Relevant agencies should establish coordination mechanisms to help im- prove the quality and utility of retirement-income-related data, reduce unnecessary duplication of effort, and identify priorities. Finally, closer collaboration between government agencies, private sector analysts and the academic research community is needed to spur advances in retirement-income-related knowledge and data that can support the development of improved projection models. Large numbers of researchers from several disciplines have worked effectively with agency people on the design, content, and analysis of the HRS and AHEAD surveys. This model should be extended to employer data collection systems, which should involve private sector represen- tatives as well. Closer collaboration expands the community of people who can contribute to cost-effective, policy-relevant data collection and analytical re- search and to the development of projection models that appropriately use data and research results. GOALS FOR PROJECTION MODELS Although data and research are the most important short-term priorities for gov- ernment agencies that work on retirement income, it is important not to lose sight of the longer term need to use those data and research findings in projection models. Projection models are needed that can estimate the overall costs and other effects of specific policy options. More comprehensive models are also needed that can estimate such effects as: . interactions among policies, such as how changes in Social Security or Medicare may affect coverage and benefits of employer pension plans and health insurance for retirees; heterogeneity or variability among workers, employers, and retirees- such as how a proposal may affect people of different income levels or couples in comparison with single people, and how fluctuations in markets and interest rates may produce winners and losers for retirement income from personal savings and pensions invested in IRA-type accounts; and long-term effects, such as the extent to which changes in the tax treatment of employee or employer contributions to pension plans may alter the behavior of

OCR for page 1
8 ASSESSING POLICIES FOR RETIREMENT INCOME younger workers and employers, even if there is little immediate effect on people near retirement. All of these needs pose formidable challenges. There is also the challenge of estimating the extent of uncertainty in model projections. Policy simulations are subject to errors and variability in data and research inputs and errors in model specifications. The longer the time period, the more complex a proposed policy change, and the more interactions it has with other policies and behaviors, the wider the margins of error are likely to be. Estimating the uncertainty of projec- tions (even if very crudely) and conveying those estimates to decision makers is essential to informing the policy process. Two important long-term goals for projection model development are the following: To develop employer-based models (using cell-based or microsimulation techniques) that can project the likely future distribution of benefit plan offerings and the demand for older workers. Such models are needed to answer questions about employer responses to policy changes and to interact with individual-level models to determine the consequences of employer decisions for workers and retirees. No such models currently exist. To develop a new large-scale individual-level microsimulation model to project the distributional effects of a range of policy proposals within a consistent framework. Existing models do not provide an adequate structure or database on which to build an improved model, although some of their components could contribute to such a model. Important capabilities for a new model are to esti- mate dynamically the effects of policy changes on individual behavior, to interact with employer-based models of worker demand and benefit offerings and with macro models of the economy, and to simulate random and cyclical shocks to employment and savings from such forces as market ups and downs. Consider- able thought will be required to determine potentially useful approaches to imple- menting these capabilities for example, how best to capture the heterogeneity of employer and worker behavior in linkages between individual and employer- based models. These two goals should be kept in mind when determining priorities for data and research; however, no significant investments in development of large-scale projection models should be made until there is better understanding of key behaviors and better data with which to project trends in those behaviors and other factors. Of course, the Department of Labor and other agencies cannot put off re- quests for analysis of retirement-income-related policy proposals because devel- opment of new large-scale models is being deferred. To respond to policy mak- ers' legitimate information demands, agencies should use limited, special-purpose

OCR for page 1
EXECUTIVE SUMMARY 9 models together with the best available data and research findings to answer particular questions. Depending on the issue, a special-purpose model may al- ready exist, or, more likely, it will be necessary to modify another model or construct a model on the spot. Agencies should seek opportunities for cost- effective improvements to special-purpose models and to make them known and available. Such models may not provide very accurate estimates. However, trying to develop complex new individual-level microsimulation or employer models in advance of improvements in data and research has little prospect of producing better results and will likely represent a misuse of scarce resources. Relevant agencies should consider the development of employer models and a new integrated individual-level microsimulation model for retirement-income-related policy analysis as important long-term goals. Construction of such models would be premature until better data, re- search knowledge, and computational methods are available. To respond to immediate policy needs, agencies should use limited, special-purpose models with the best available data and research find- ings to answer specific policy questions. For the full text of these recommendations and for additional recommendations and suggestions on specific points, see Chapter 3 (research), Chapter 4 (data), Chapter 5 (projection models), and Chapter 6 (furthering collaborative arrange- ments among agencies and with the private sector and academia for data, re- search, and model development).