Click for next page ( 40


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 39
~2 Key Research Issues The economic security of retirees depends on their lifetime experiences of work, savings, and family ties and their health care and other consumption needs. Such experiences depend on individual choices; they also depend on decisions of em- ployers that affect the provision of jobs, earnings, savings vehicles, and retire- ment and health care benefits. Government policies and programs constrain and shape all these individual and employer decisions. Consequently, projecting the implications for retirement income security of proposed changes in government policy requires basic research knowledge about the likely behavioral responses of people and employers and a capability for using it to estimate future outcomes. Knowledge to project behavioral effects is not needed for every retirement- income-related policy question, but it is essential for many questions, particularly when projections are needed for the medium and long term. Contrast short-term versus long-term projections of the implications of reducing the annual cost-of- living adjustment (COLA) for Social Security benefits. In the short term, say for a 5-year period, the projection is straightforward: the Social Security benefits of current retirees and those expected to retire during the next 5 years are increased each year by a reduced COLA instead of by the full estimated amount of inflation as is currently done. If the 5 years are expected to be a period of low inflation, the effect of reduced benefits will not be great for many people, although some people will be moved below the poverty line. Over a longer projection period and particularly if the COLA is cut significantly, the real value of Social Security benefits will decline significantly over people' s retirement years, with potentially severe effects on the income security of many retirees. Given this prospect, one would expect more and more people to change their behavior, for example, by 39

OCR for page 39
40 ASSESSING POLICIES FOR RETIREMENT INCOME delaying retirement or increasing savings. Employers may also change their pension plan provisions to compensate for workers' expected lower Social Secu- rity income. To the extent that such behavioral responses occur, projections need to take them into account to be useful for policy making. Some policy proposals represent major, systemic changes for which there are no historical parallels and, hence, little possibility of obtaining appropriate be- havioral parameters from available (or potentially available) data to use in projec- tions. In these instances, it can still be useful to develop projections with a range of assumptions, reflecting expert judgment. Such estimates may indicate the extremes of possible outcomes, although they remain highly uncertain even then. Many proposals, however, can be expected to influence behavior in ways that research can illuminate. In addition, for some major changes with which the United States has no experience, it may be possible to obtain useful data by analyzing the experience of other countries (such as the experience in several countries with various forms of privatization of social security systems; see World Bank, 1994~. Although the cultural and social milieus are different, there may still be knowledge from comparative cross-national research that can contribute to U.S. projections. In this chapter, we summarize what is known and not known about factors that influence key retirement-income-related behaviors of individuals and em- ployers. For employers, we look at research about decisions on pensions and other benefits and demand for older workers. For individuals, we look at re- search about savings, consumption, and labor supply. Our reviews in the chapter text are quite brief, highlighting key knowledge gaps. They draw from extensive reviews of the literature and research issues in a set of papers we commissioned for our study (Hanushek and Maritato, 1996; see Appendix A for contents).] For projecting the likely effects of retirement-income-related policy changes, particularly over the long term, it is also important to understand likely demo- graphic and health-related trends. We thus look at research about factors and trends in basic demographic processes, particularly mortality, that will affect the size and makeup of the population of workers and retirees at future times; the health status of workers and retirees that will determine their needs for health care services and affect their decisions about work and savings; and health care costs and financing arrangements that will affect retirees' living standards. Our reviews of these areas are also brief.2 Finally, we touch on factors and trends in 1 The papers by Lumsdaine, Parsons, and Poterba review the literature on labor supply, employer behavior, and savings and consumption, respectively; the paper by Gustman and Juster looks at the distribution and sources of income and wealth of households with retired workers and reviews analytical models of labor supply, savings, and pension decisions that contribute to income and wealth at retirement. 2For more extensive discussion and literature reviews, see the paper by Lee and Skinner (in Hanushek and Maritato, 1996; see also Moon, 1995).

OCR for page 39
KEY RESEARCH ISSUES 41 marriage and divorce, which can have important consequences for the economic well-being of the elderly. Our discussion covers a large number of areas, but it is not comprehensive. For example, we do not consider factors that influence trends in worker produc- tivity, even though such trends are key to growth in real earnings, which, in turn, affects Social Security and pension entitlements. Also, we do not discuss factors that influence the financial return to pension investments and personal savings, although the distribution of returns is of growing importance for retirement in- come security and will become even more important if such policy changes as privatizing all or part of Social Security contributions are implemented. None- theless, we cover most of the critical areas for which a strong research base is needed for retirement-income-related policy analysis. We identify important strengths and weaknesses in available knowledge and recommend improvements. Our discussion of research topics is ordered by our assessment of which areas are most deficient in terms of basic knowledge that is relevant for retire- ment income security policy analysis. These deficiencies are largely due to deficiencies in data, which in some instances have hampered the development of theory and in other instances have impeded the development of robust estimates of behavioral parameters. Hence, improvements in data (discussed in Chapter 4) will be required to carry out much of the research agenda that we recommend. However, basic research should not stand still. Although some research cannot proceed very far without better data, other research can go forward with im- proved data that are or will shortly become available or with the use of methods (e.g., case studies) that require much less investment in new or better data. Generally, research should proceed to refine and improve analytical models with the best available data. Indeed, data development and basic research go hand in hand: new findings from data suggest the need to rethink theories and models, and, in turn, analytical developments suggest further improvements to data. We stress in our report the need for investments in data and analytical modeling and research in areas that most need an improved base of knowledge with which to support retirement-income-related policy work. Given constrained resources, however, we recommend against major investments in complex, new projection models for policy purposes until investments in data and basic re- search bear fruit. We begin our review with research on employer behavior. Employers play an important role in the provision of retirement income security through their decisions about personnel and benefits. Moreover, their behavior can change rapidly, as evidenced by the marked increase in recent years in the number of employers offering managed care health insurance plans. Yet very little is known about the strength of the factors that may influence employers' decisions. We next consider research on individuals' choices of savings and consumption, an- other area about which relatively little is known, and individuals' labor supply and retirement decisions. We then consider trends in relevant demographic char

OCR for page 39
42 ASSESSING POLICIES FOR RETIREMENT INCOME acteristics. Lastly, we comment briefly on key questions related to health care costs. EMPLOYER BEHAVIOR Employer pensions represent a significant source of retirement income for many workers, and their availability and characteristics are important determinants of retirement decisions and other relevant behaviors. The availability of other kinds of employer benefits (e.g., disability insurance and retiree health insurance) also plays a role in the retirement-related behavior of workers. More generally, em- ployers' demand for older workers affects retirement income security, both di- rectly in terms of the labor market opportunities for older workers and indirectly in terms of the effects on employer decisions about benefits. Employers' decisions to offer a pension plan or plans (or other benefits) and of what typos) depend on several factors: expected benefits in terms of work force productivity, worker recruitment and retention, and retirement of older workers; federal tax law provisions, such as tax deductions for qualified plans, which are an incentive to provide benefits, and nondiscrimination rules for quali- fied plans, which are a disincentive; other government policies, including those for Social Security and Medi- care, pension insurance laws and regulations, and antidiscrimination laws with respect to age and disability; employers' financial objectives and concerns, including tax liabilities, administrative costs, and costs of compliance with regulations; trends in benefit policies by similar employers; and the demand for benefits by employees. There are a large number of competing theories about why employers offer pension plans, which differ in the importance accorded to the various factors listed above. For example, some theories and models emphasize the importance of tax deferral, others stress the use of pension plans as a worker selection device or as a productivity enhancement mechanism, while others stress the importance of pension plans as incentives for workers to retire. To date, there is no agree- ment on which of these theories best explains employers' behavior or on the extent to which different types of employers may have different mixes of mo- tives. In a literature review commissioned by the panel, Parsons (1996) emphasizes the importance of transaction costs in explaining the well-documented phenom- enon that large, unionized employers are much more likely to offer pensions than are smaller employers: large employers can benefit from economies of scale in administering pension plans that are not available to small employers. He at

OCR for page 39
KEY RESEARCH ISSUES 43 tributes the fact that small employers that do offer pension plans almost invari- ably offer a 401(k) or other type of defined contribution plan, rather than a defined benefit plan, to the same reason, namely, lower administrative costs. However, Parsons (1996:179) acknowledges that "alternative hypotheses . . . can explain many of the same observations." Moreover, such analysis cannot explain why some relatively large firms do not offer pensions and some relatively small firms do, nor the variety of pension plan provisions that characterize firms in the same size class. Nor does the research yet provide agreed-upon values of behav- ioral parameters that could be used to estimate the strength of employer responses to policy and other changes: for example, what degree of reduction in plan administrative costs or in regulatory burden would induce a specified percentage of small employers to set up pension plans of a particular type. Employer demand for older workers is also affected by many factors, includ ~ng: desired work force characteristics, in terms of retention, skill levels, pro- ductivity, and other attributes; . employer perceptions of relative productivity of workers of different ages; employer financial objectives and concerns (such as the costs of provid- ing benefits, costs of providing job training); . employer personnel practices, such as flexibility in reassigning workers; government policies and regulations, including anti-age discrimination laws and restrictions on mandatory retirement; and older worker supply. Again, a pervasive finding is that the larger the employer, the lower the share of older workers (age 55 and older) in an employer's work force. Moreover, when mandatory retirement rules were legally permitted in the United States, larger, unionized employers were more likely to have them. Considerable analy- sis has been conducted of differences among employers in mandatory retirement provisions, but there is no agreement on the underlying behavioral mechanisms. Theories to explain this phenomenon include: the propensity of large employers to have more formal rules of all types, reflecting their higher costs of making idiosyncratic decisions; the "representative worker" model, which posits that employers that en- gage in long-term contracting with workers (predominantly larger employers) pay workers more as they age than they are worth in terms of productivity and hence that these employers cap total compensation by mandating retirement by a specific age; and the desire of employers to limit the propensity of people who are hired later in life to work into their less productive older years in order to accumulate more generous pension rights.

OCR for page 39
44 ASSESSING POLICIES FOR RETIREMENT INCOME Alternatively, workers' decisions could play a large role in the age profile of employers' work forces: that is, the smaller share of older workers at large employers could be attributed to the propensity of large employers to offer pen- sions, which, in turn, encourage retirement. Untangling the mix of factors that influence personnel and benefit policies of different kinds of employers is critical for reliable projections of the likely effects of government policy changes, such as the effect of changes in Social Security and Medicare provisions on employer hiring and compensation policies and the consequences for workers' overall retirement income security. However, Par- sons (1996:179-180) concludes that "we are not close to having the level of understanding that would permit us to make quantitative estimates on employer behavior." There is yet another problem for estimating employer responses to policy changes in projection models, that of projecting trends in the mix of employer types. Given the association of pension offerings and fewer older workers with larger employers, a continued shift of employment from the large, highly union- ized sector of the economy to the service sector, with many more small employ- ers, may mean that older workers in the future have greater access to jobs but less access to pensions. Whether this trend will continue, and at what rate, and whether other significant shifts in employer mix will occur are difficult but important questions to answer, as is the question of whether observed relation- ships between employer type and pension and employment practices will con- tinue to hold. Even more than in the case of personal consumption and savings behavior (see below), data gaps and measurement problems greatly constrain researchers' ability to develop reasonable analytical models of employer behavior with regard to pension and other retirement-related benefits. Available cross-sectional data sets suffer from several limitations. The Form 5500 database of the U.S. Depart- ment of Labor, which characterizes employer benefit plans from annual filings to the Internal Revenue Service, does not cover public employers; it also does not provide information on the full range of private employer benefits, which include nonqualified as well as qualified pension plans, retirement window opportunities, retiree health insurance, disability insurance, and other relevant benefits. The Form 5500 database also contains limited information for analyzing differences in benefit packages as a function of employer or work force characteristics. The Employee Benefits Survey (EBS) of the Bureau of Labor Statistics provides extensive information on types of public and private employer benefits for workers in broad occupational categories. However, the survey is based on small samples, has at present no data on benefit costs, and has very little informa- tion on employer characteristics. The National Employer Health Insurance Sur- vey of the U.S. Department of Health and Human Services has large sample sizes, but it is limited to health care benefits and costs. Private sector surveys of

OCR for page 39
KEY RESEARCH ISSUES 45 pension and health care benefit offerings by consulting firms and others are generally limited to larger employers and often to the clients of the survey spon- sor. Case studies by researchers with data from one or a few companies have supported innovative analysis,3 but the results of such studies are not readily generalizable because of the heterogeneity of employers. There are almost no panel data on employers that could be used to trace changes in benefit packages over time or to determine the factors that influence employer behavior in this regard. (The Form 5500 database can be used for only limited kinds of longitudinal analysis.) Similarly, there are no panel (or cross- sectional) data that could support analysis of interactions among employer and employee characteristics that in turn affect benefit plan decisions and, ultimately, retirement income security. Yet such analysis is critically important given the prominent role of employer pensions and other benefits in retirement income security and the evidence that employer behavior with regard to the extent and type of such benefits is dynamic and sensitive to public policies as well as broader economic factors. Finally, there are almost no data with which to analyze employer demand for older workers. Repeated cross-sectional data from the decennial census and the Current Population Survey (CPS) have been used to study the employment of older workers by industrial sector, and panel data from the Retirement History Survey (RHS) have been used to study retirement paths of older workers (e.g., retiring from a career job versus moving from a career to a "bridge" job before exiting the labor force completely). However, no nationally representative data set exists that permits direct analysis of the factors that influence employers' hiring and retention decisions. A longitudinal employer database that the Census Bureau constructed from census and survey data is limited to manufacturing companies, and it contains no information on work force age structure except for a subsample of establishments with records matched to 1990 population census records for their workers. Researchers have made innovative use of personnel and other records of one or a few employers to study workers' compensation in relation to measures of productivity and other characteristics (e.g., Kotlikoff and Gokhale, 1992; Medoff and Abraham, 1981), but such studies are few and limited in generalizability. In short, there are glaring gaps and deficiencies in data about employers (and their workers) with which to develop behavioral parameters for projecting em- ployer responses to government policy changes and other factors that may affect personnel and benefit decisions. In Chapter 4, we describe in more detail the problems of existing data sets on employers and recommend improvements. 3An example is Mitchell and Luzadis (1988), who analyzed pension policies of 14 employers before and after legislation curtailing mandatory retirement.

OCR for page 39
46 ASSESSING POLICIES FOR RETIREMENT INCOME CHOICES OF FAMILIES AND INDIVIDUALS Savings and Consumption People's decisions about allocating income between savings and consumption in their pre-retirement years have important implications for the level of consump- tion that they will be able to sustain after retirement. In this context, personal savings includes after-tax investments as well as tax-sheltered investments, in- cluding Individual Retirement Accounts (IRAs) and voluntary employee contri- butions to 401(k) and other pension vehicles. The theoretical basis for savings and consumption choices has long been posited in terms of a life-cycle model, in which younger people, by borrowing, and older people, by spending down assets, exhibit high consumption-to-income ratios, while middle-aged people with the highest earnings potential exhibit rela- tively low consumption-to-income ratios. However, a "pure" life-cycle model does not explain observed macroeconomic and individual behavior, so the basic model has been elaborated in various ways (see Poterba, 1996~. Three well- worked-out modifications include: a model that attributes precautionary savings motives to individuals (e.g., workers with pensions may save additional amounts to guard against possible future job loss); a model that attributes bequest motives (i.e., the desire to leave assets to descendants); and a model that imposes liquidity constraints (e.g., young people may not be able to obtain affordable loans). All three types of analytical models fit reasonably well with cross-sectional data on the distribution of wealth at retirement, and there is no basis as yet to choose among them or to determine whether and what kind of mixed-motive model best explains savings behavior. Moreover, none of the existing models explains well the trends in personal savings rates over time. Yet other types of models have been posited, in which behavioral or psycho- logical elements play a significant role. One such model assumes that people hold different assets in distinct "mental accounts," which implies that changes in the level of one asset may have relatively small substitution effects on the hold- ings of other assets, contrary to the life-cycle model assumption. Other models posit that people use rules of thumb or other simple heuristics to make savings decisions (e.g., deciding to save a fixed percentage of earnings each year, regard- less of the expected return on saving). These models are intriguing but, to date, have not been well specified or tested. There are major unanswered questions in the area of consumption and sav- ings behavior:

OCR for page 39
KEY RESEARCH ISSUES 47 Why have personal savings rates declined in recent years? As just noted, none of the existing models explains this trend. It may be that cohort factors for example, the baby boom generation was not exposed to the Depression are involved or that the development of a government safety net has had an effect. Why are so many middle-income people approaching retirement with very low wealth levels, when Social Security is not a sufficient income source to maintain their standard of living? . To what extent does saving in IRAs and other voluntary pension plans offset other saving? There is continuing controversy about whether IRA-type investments are made with dollars that people would have invested in any case. A key policy question is whether making such accounts more substitutable with other saving (e.g., by permitting withdrawals) would increase or decrease net saving in the long run. (In the short run, the effect would most likely decrease net saving.) More broadly, how much is personal saving influenced by taxes, Social Security, and pension coverage? How much do behavioral elements, such as mental accounts and rules of thumb, influence saving behavior? As with retirement and pension acceptance decisions (see below), adding such factors to models would complicate analysis. Yet there is evidence from anomalous savings behavior (e.g., low rates of savings for middle-income families) that it may be important to take account of such factors. Relatedly, how much do families know about their future financial needs and potential sources of support, and would more knowledge influence their behavior? . The above questions refer to accumulation of savings until retirement age. Further questions arise about consumption and savings patterns after retirement: . Looking at older people, does wealth decline in retirement as much as the life-cycle model would predict? There is conflicting evidence about the extent to which wealth, particularly housing equity, is spent down. What determines the demand for annuities of different types, and, specifi- cally, how do couples decide between single and joint life annuities? . What are the effects of changing the form of retirement benefits lump sums as opposed to annuities, or nominal annuities as opposed to indexed annu- ities on income security over retirees' life spans? What are the effects of inflation and nominal interest rates on real con- sumption patterns in retirement? What is the relationship between retirement income programs and support of the elderly by their families? There is another problem in projecting the retirement income security impli

OCR for page 39
48 ASSESSING POLICIES FOR RETIREMENT INCOME cations of people's savings and consumption choices, namely, that of projecting rates of return on different assets. A major unanswered question is whether rates of return will be influenced by demographic changes. For example, will baby boomers' housing lose value, and what will be the likely effects on their post- retirement standard of living? Also, how much variation will there be in baby boomers' returns on housing and other assets and what will be the likely distribu- tional effects on the adequacy of their retirement income? Data gaps and measurement problems are a major impediment to addressing all of these questions. People's consumption, savings, and wealth are notoriously difficult to measure in surveys, although two relatively new surveys the Health and Retirement Survey (HRS) and the Asset and Health Dynamics Among the Oldest Old (AHEAD) survey have made significant progress in improving the measurement of sample members' financial holdings. To the extent that these surveys accurately measure income and change in net worth, then an estimate of consumption can be obtained by subtraction. Ideally, direct measures of con- sumption would be used in order to more accurately assess pre-retirement living standards, estimate the implications for post-retirement living standards of people's current savings rates, and help estimate likely future savings rates con- sistent with the life-cycle model. The third round of HRS and AHEAD includes a question on total expenditures that may prove useful for such analyses, although the data will require careful evaluation of their quality. Better data are also needed on people's information about likely available sources of retirement income (e.g., their pension rights and anticipated savings) and their expectations about likely future events (e.g., their own life expectancy, the likelihood they will continue in good health, the likelihood they will receive Social Security or pension benefits or an inheritance). More detailed information on pension plan provisions of workers would also be helpful in determining factors that influence savings behavior. Linked family data would help deter- mine the strength of the bequest motive and the factors influencing it. (Some analysis of savings behavior has been conducted with linked family data from the Panel Study of Income Dynamics.) This data need may be of lower priority if policy interest remains generally focused on people with low-to-middle levels of earnings for whom the prospects of significant bequests are relatively low. How- ever, linked family data are important for other purposes, such as understanding care-giving responsibilities and intrafamilial sources of support. HRS and AHEAD are designed to remedy these kinds of data gaps, and the two surveys will need to be continued if they are to make possible the develop- ment of a broadly accepted model of savings and consumption behavior with high explanatory power. Such a model will most likely retain a basic life-cycle approach; however, the evidence of substantial heterogeneity among the popula- tion with regard to savings behavior suggests that a satisfactory model will need to incorporate multiple savings motives or distinguish among motives for differ- ent kinds of people.

OCR for page 39
KEY RESEARCH ISSUES 49 Retirement income projection models need to take account of personal sav- ings, particularly given the possibility of cutbacks in Social Security and em- ployer pension benefits. Although such cutbacks may not occur, they will very likely be considered, and, hence, policy makers will want estimates of their likely effects. The lack of agreement on the most appropriate analytical model con- strains the ability of projection models to estimate the likely effects of policy changes on personal savings. In the absence of clear directions from research, it will be very important for projection models to provide sensitivity analyses under alternative behavioral assumptions. Labor and Leisure A great deal has been learned about labor supply and retirement behavior in the last 20 years; indeed, more is known about labor supply and pension acceptance decisions than about almost any other aspect of retirement behavior (see Lumsdaine, 1996~. One reason is that such behavior is easier to measure than, for example, consumption and savings. Another reason is the availability of rich longitudinal panel data sets for analysis of labor-leisure choices. Earlier panels, such as the Retirement History Survey (RHS), lacked detailed information about workers' pension and health care coverage, but this weakness has been corrected in the new HRS. Repeated cross-sectional surveys, such as the March Current Population Survey (CPS), have also provided valuable information on labor sup- ply trends for population groups.4 What is known about men's retirement behavior underscores the extent of heterogeneity among workers.5 A large fraction of men (40-50%) work full time at a career job until their early 60s and then remain out of the labor force for the rest of their lives. Typically, they apply for Social Security benefits and a pen- sion, if they have one, at age 62 or at age 65. (Some apply for a pension even before they are eligible for Social Security at age 62.) Another large fraction (over 40%) never retire or have complicated in-and-out labor supply patterns. 4The availability of rich data sets in this area resulted from concerns with the trend toward early retirement in the 1960s that led to support for panel surveys, such as the RHS and the National Longitudinal surveys of Labor Market Experience (NLS). Two decades later, HRS and AHEAD were initiated to update the picture on retirement and to respond to concerns about savings behavior and the health status of an aging population. we argue that a concern with changes in employer behavior should motivate support for employer-based surveys, in addition to the continuation of panel surveys of individuals. 5This summary description is based largely on the RHS. More recent studies support the general characterization, although some trends already evident in the RHS such as the shift in the modal age of retirement from age 65 to age 62 are more pronounced in later data (see Huron, Haveman, and O'Donnell, 1995; Karoly and Rogowski, 1994; Peracchi and Welch, 1994). Less is known about women s retirement behavior because, historically, fewer data have been available.

OCR for page 39
so ASSESSING POLICIES FOR RETIREMENT INCOME Very few individuals (less than 5%) phase out of the labor market by gradually reducing their hours of work, probably because of employer constraints on work hours. Factors that influence age at retirement include the availability of em- ployer pensions and retiree health insurance coverage, Social Security provi- sions, eligibility for Medicare, and the individual's health status. There is general agreement among researchers that Social Security and Medi- care provisions have important effects on retirement behavior, at least for the subset of workers who are not at the upper tail of the income and wealth distribu- tion and whose Social Security benefits are not dwarfed by employer pensions. However, there is a large range of uncertainty about the exact magnitude of individual responses to various policies. Much of this uncertainty stems from disagreement about the appropriate strategy for modeling individual behavior whether to use reduced-form statistical models or structural econometric models (see Lumsdaine, 1996:70-75~. Reduced-form models do not require the researcher to impose any underly- ing theory of individual behavior. Hence, they are much easier to formulate and estimate than are structural models, which are derived from an explicit theory of individual behavior and make strong, a priori assumptions. Reduced-form mod- els allow for flexibility in the choice of functional form, which, in turn, makes it easier than in structural models to learn about the data. However, unless policies have changed a great deal in the past, reduced-form models cannot estimate the independent effect of policy parameters on behavior and hence have great diffi- culty in forecasting how behavior will change under alternative policy regimes. Structural models can provide estimates of policy effects, even when there has been little historical variation, because they impose a priori identifying as- sumptions on the data, which typically involve strong restrictions on the nature of individual preferences. However, if the model's assumptions are incorrect, then its predictions are likely to be incorrect. Moreover, there will generally be several sets of assumptions about preferences that "explain" any given body of historical data. If each set gives different predictions about the likely effect of a policy change, there will be little objective basis to choose among them. Despite these problems, the structural approach does appear to yield accurate predictions of the effects of policy changes on retirement behavior in the limited number of out-of-sample predictive tests that have been performed to date. For example, Lumsdaine, Stock, and Wise (1990) estimated a dynamic structural model of retirement decisions at a Fortune 500 firm by using data prior to the introduction of a temporary window plan that created substantial incentives for workers to leave the firm. The model did a reasonably good job of predicting the large increase in retirement rates for most people of the relevant ages after the introduction of the window plan, whereas a variety of reduced-form models performed poorly in this regard. Questions about future directions for labor supply and pension acceptance research include in what ways to pursue the use of complex structural models.

OCR for page 39
KEY RESEARCH ISSUES 51 Important issues to be addressed are whether such models should be static or dynamic, what kinds of uncertainties need to be modeled, and how comprehen- sive the overall model needs to be. Another important question is whether structural models should continue to assume completely optimizing behavior (as defined by economic theory) on the part of individual workers. Such an assumption can be modified (e.g., by assum- ing that people use various rules of thumb, such as work until age 65 and then retire); however, doing so is likely to make the modeling task even more difficult. Moreover, it is not clear how to specify the ways in which people who are using rules of thumb or other psychologically influenced decision rules will respond to policy changes.6 Important aspects of retirement behavior that are not addressed in current analytical models include workers' decisions to retire or apply for disability benefits and joint retirement decisions of spouses. Another area for work is to integrate models of retirement and savings behavior. The usual approach is to ignore or drastically simplify the nature of consumption and savings choices as they relate to retirement behavior (e.g., assuming that consumption equals in- come). There is some theoretical and empirical justification for this approach (except for the very wealthy), and the practical difficulties of doing otherwise are formidable. Nonetheless, it may be increasingly important to tackle this problem in order to answer questions about future trade-offs among savings, consumption, and work: for example, the extent to which people will save more, consume less, or work longer if Social Security, employer pensions, or health care benefits become less generous or more costly. Finally, it is important to determine the applicability of complex structural models for policy use. At present, it may not be feasible to incorporate a full- blown behavioral dynamic programming model of retirement into a microsimu- lation projection model for estimating the likely effects of alternative Social Security and employer pension policies on retirement income security. The question then becomes what kinds of simplifications are necessary in order to have a practical and usable projection model. The ability to refine and extend behavioral models of labor supply and retire- ment depends on the continued availability of rich panel data sets. Data have been much more plentiful on these topics than on consumption and savings or employer behavior. Also, the new HRS promises to fill important data gaps in earlier retirement surveys. But HRS and related surveys must be continued and enhanced to permit the development of more robust estimates of behavioral pa- rameters for use in projecting workers' labor supply responses to policy changes. 6Questions may be added to HRS and AHEAD to learn about this issue by asking people directly how they make decisions and what information they use.

OCR for page 39
52 ASSESSING POLICIES FOR RETIREMENT INCOME DEMOGRAPHIC VARIABLES Size and Composition of the Population Projecting the likely costs and effects of current and alternative retirement-in- come-related policies requires estimating the number of people alive over the projection period and their distribution by age and other characteristics. In turn, to generate population projections requires estimating births, net immigration, and deaths. For many retirement-income-related analyses, it is not necessary to estimate future fertility levels because the concern is with the number of retirees (for which projections can extend as long as 50-60 years on the basis of the population alive at year 1) or with the numbers of retirees and workers (for which projections can extend as long as 20 years on that basis). Net immigration is important, but it is heavily influenced by legislation, which means that immigration can be included in projection models as a parameter that can readily be given different values to reflect expected policies.7 Estimates of future mortality levels, on the other hand, are important deter- minants of Social Security trust fund balances in both the short and long term. They also affect the viability of employer pension plans. Simulations by the Social Security actuary show 75-year trust fund balance projections to be more sensitive to assumptions about the future course of mortality than to any other demographic or economic variable in the actuary's cost model (see Board of Trustees [OASDIi, 1994:131-132~. However, the projections do not allow for extreme assumptions, such as a return to the high birth rates of the 1950s, in which case mortality might not be the driving variable. Reasonably good data are available on mortality rates by age and sex. The problem is what assumptions to apply to historical data to project mortality rates into the future and how to estimate the uncertainty in the projections, particularly over the long term. The Social Security Administration (SSA) essentially devel- ops mortality projections by extrapolating rates of decline in age-specific death rates for specific causes of death over the previous 20 years. The results, which are inherent in the methodology, imply a sharp slowing of the rates of decline of mortality at all ages, relative both to the previous two decades and to longer run historical trends back to 1900. Lee and Carter (1992), in contrast, project the rates of decline in age-specific mortality rates observed over the twentieth cen- tury (not disaggregated by cause of death), which have been fairly steady despite periods of faster and slower progress. Hence, the Lee and Carter projections (and those developed by other researchers) imply a larger retirement-age population than do the SSA projections (or those developed by the Census Bureau). 7The task is somewhat more complicated than indicated in that net immigration must be distrib- uted into immigration and emigration by such characteristics as age and sex.

OCR for page 39
KEY RESEARCH ISSUES 53 A major advantage of the Lee and Carter approach is that it provides prob- ability intervals describing the uncertainty in their extrapolative method. It does not, however, allow for the possibility of "structural breaks," such as the sharp decline in age-specific mortality rates that occurred between the nineteenth and twentieth centuries, and thus it probably underestimates the extent of uncertainty. Nonetheless, the stochastic models used by Lee and Carter and others to estimate the uncertainty in population forecasts represent a major step forward over the approach that is used by SSA and the Census Bureau to convey estimates of uncertainty to policy makers. In the SSA and Census Bureau approach, "high" and "low" scenarios are developed to bound the "intermediate" or expected fore- cast. For many policy purposes, it would be highly useful to carry out research that could support projections of mortality rates for other characteristics on which mortality is known to vary, in addition to age and sex. In particular, in order to answer distributional questions, such as the retirement income security of wid- ows relative to married couples or of low versus high earners, it is important to have mortality projections by such characteristics as marital status, income level and other indicators of socioeconomic status, and health or disability status. However, little work has been done to develop such projections. Recent studies show that social class differentials in mortality (measured by educational levels) have widened sharply for men at all adult ages since 1960, somewhat less so for women (Preston and Elo, 1995~. Data from Social Security administrative records could provide the basis for an authoritative study of mor- tality variation over time by earnings history for people of retirement age. It would also be valuable to use Social Security data to study mortality variation by marital status. However, at present, marital status is recorded only for people who are receiving benefits as a spouse or widow or widower; it is not recorded for people who are receiving benefits on the basis of their own earnings, who may be married or unmarried. SSA is currently appending information on Social Secu- rity benefits and mortality to several panels of the Survey of Income and Program Participation. These files could provide the basis for a study of the relationship of mortality to income and marital status for a limited sampled Family History Marital history can have important effects on the income and wealth of the elderly, particularly for women. Research has documented significant drops in economic well-being for women after the divorce or death of a spouse (see, e.g., Holden, 1991), and women are more at risk of being widowed and of not remar- rying after either widowhood or divorce. The result, cross-sectionally, is a highly 8Another source of data on mortality differentials by marital status is the National Longitudinal Mortality Study, which Preston and Elo (1995) used. It has income data but only for the last year.

OCR for page 39
54 ASSESSING POLICIES FOR RETIREMENT INCOME skewed distribution of income and wealth by marital status and sex among eld- erly people aged 65 and over. In 1995, the poverty rate for single elderly women was 5.6 times the rate for married elderly women; the corresponding ratio for elderly men was 3.0. At the same time, elderly women were more than twice as likely as elderly men to live without a spouse or other family member (Bureau of the Census, 1996:Table 2~. Similar patterns are evident for wealth (see Gustman and Juster, 1996:Tables 2-A4, 2-A5~. There are also differences in income and wealth by age: older subgroups of the elderly population are poorer and less wealthy than younger subgroups. It has not been determined how much these patterns reflect dissaving and outliving sources of income by people as they age and how much they reflect cohort differences in initial income and wealth levels. Considerable research has been conducted on models of first marriage and divorce. Less work has been done on remarriage, and relatively little work has been done that is directly relevant for retirement income security policy analysis (see Caldwell, 1993, for a review of the literature; see also McLanahan and Casper, 1995~. Needed work includes the development of models of marital behavior and projections of trends that explicitly account for the risks of mar- riage, divorce, widowhood, and remarriage for people as they approach retire- ment age and beyond. Such models and projections should take account of trends in mortality differences by sex and marital status. Work is also needed to draw out the economic consequences of marital histories for post-retirement income and wealth levels. Such work should take account of women's increased labor force participation, which may result in their accumulating higher levels of pen- sion and other wealth than in the past, and of trends in the form of retirement benefits (e.g., lump sums versus annuities), which may adversely affect the retire- ment income security of surviving spouses in particular. Data on former spouses' rights to pension and other benefits are also important to include in analyses that link marital histories and retirement income security. Another way in which people's family histories affect retirement income security is through the effects on kinship networks. An important policy concern is whether the baby boom generation, which exhibited higher ages at first mar- riage and lower fertility rates than the previous generation, will be supported by as many kin (own children and other relatives). Work on the availability of kin to provide financial support and care-giving for older people has been hampered until recently by the design of household surveys, which historically have not asked about adult children or other relatives not living in the household (see Wolf, 1994, for a review of the literature in this field). Newer surveys, such as HRS, AHEAD, and the National Survey of Families and Households, have at- tempted to remedy this lack with detailed questions about kin networks and intrafamilial transfers.

OCR for page 39
KEY RESEARCH ISSUES 55 Health Status An important factor to include in retirement-income-related policy projections is the health and disability status of workers and retirees and the likely trends over time. A decrease in mortality for older people has different implications for retirement income security if they suffer from disease or disability or if their additional years of life are active and healthy, with few expensive medical care needs and the opportunity to continue to earn income. Similarly, an increase or decrease in disability levels for workers as they approach retirement has implica- tions for the extent to which they retire early or, when this is not possible, apply for benefits from public and private disability programs. Generally, income and wealth differ markedly by health and disability status. While the direction of causality is not established, it is clearly important to have indicators of health and disability status to understand the distributional consequences for retirement in- come security of many kinds of proposed policy changes. Unfortunately, the identification of health and disability status, whether cross- sectionally or longitudinally, is beset with measurement problems and ambigu- ities of classification. Lee and Skinner (1996), in their review commissioned by the panel, consider the evidence on trends in disability status defined in several different ways: objective health measures, incidence of specific diseases, mea- sures of functional ability, and self-reports. They find a mixed picture, although there appears to be a long-term trend toward lower overall disability levels among the elderly. Self-reported health assessments, however, show marked short-term fluctuations, which may be influenced by such factors as improvements in medi- cal diagnosis. Also, there is a close match between increases and decreases in self-reported disability and changes in disability insurance programs, including changes in the intensity of administrative efforts to ascertain eligibility and to follow up cases once enrolled, as well as changes in eligibility requirements. Panel data with improved measures of health and disability status are needed to model relationships with key behaviors, including decisions about savings, consumption, and retirement. (The HRS and the AHEAD survey are designed to serve this purpose.) Panel data are also needed to determine trends in disability levels over time and whether, as some researchers hypothesize, morbidity will be "compressed" into the last years of life. Such a compression could lead to significant reductions in the medical care needs and costs of the elderly, although this effect could be offset by such factors as an increase in the proportion of the "oldest old" in the population or a decrease in the availability of family care- g~vers. For the purpose of establishing trends, panels need to follow large samples of people for long periods of time; large samples are required because relatively few people are disabled. Also, panels of new cohorts must be initiated periodically. Recent studies that find significant declines in the extent of disability among the elderly (e.g., Manton, Stallard, and Liu, 1993, who use a measure of functional

OCR for page 39
56 ASSESSING POLICIES FOR RETIREMENT INCOME status) are based on data spanning only 7 to 10 years. More years of data are required to confirm these results. HEALTH CARE COSTS No assessment of retirement income security is complete without consideration of likely trends in the magnitude and distribution of health care costs. To the extent that older Americans face rising health care costs that they must finance through some combination of higher health insurance premiums, taxes, and direct out-of-pocket outlays, then a retirement income stream that would have been adequate in the past to cover other needed consumption in addition to medical care may no longer be adequate. Projections of likely future trends in aggregate medical care costs and in the availability of medical care benefits are subject to extreme uncertainty, given the large number of actors whose behavior must be modeled federal and state agen- cies, private health insurers, employers, medical care providers, medical care technology developers, and medical care consumers and the complexities of the interactions among them. Economic incentives clearly play a role in medical care consumption. However, there is an argument that in the United States the devel- opment of new technologies and treatments coupled with a strong disinclination on the part of providers and consumers to forgo their use, once introduced, is a driving force for medical care cost increases. The shift to managed care has reduced the rise in costs, but it is not clear that this trend will continue once excess capacity is wrung out of the system (see Moon, 1995~. In short, determining the relative importance of various factors that influence medical care costs and benefit packages and the role of public or private sector policy changes in changing relevant behaviors presents an almost overwhelming research challenge. Research and models that are focused on retirement income security cannot hope to resolve these issues. What seems most fruitful for retire- ment-income-related research and modeling to address is the likely distributional consequences of alternative medical care cost and insurance coverage scenarios for the retirement income security of groups of the elderly population. For this purpose, it is important to develop good estimates of the relationship of health and disability status, insurance coverage, and other individual-level variables (e.g., age, gender, ethnicity, employment status, income level) to medi- cal care costs in a relative sense: that is proportionally how much more is spent on medical care in total and out-of-pocket by people in worse health than by people in better health. The National Medical Expenditure Survey (NMES) is an important data source for this purpose. However, it was last conducted in 1987 (see below), and key relationships may have changed since then as a consequence of major changes that have occurred in health care financing. Panel data are also needed to determine differences in medical care spending patterns across the years of retirement, taking account of both acute and long

OCR for page 39
KEY RESEARCH ISSUES 57 term care costs. The few available studies suggest that spending patterns are correlated across time (the small group of people who are high spenders in one year are high spenders in subsequent years), but that there is a considerable dropoff in the concentration of spending over a long period. There are plans to conduct NMES on a continuing basis, beginning in 1996 and renamed the Na- tional Medical Expenditure Panel Survey (MEPS), but individual sample mem- bers will only be followed for a 2-year period. The HRS and AHEAD surveys, when linked with Medicare and Medicaid records, may develop a capability to provide needed longitudinal data on the distribution of medical care costs across retirement. CONCLUSION This discussion has touched on an array of important topics for understanding and projecting retirement income security and has identified many gaps in basic research knowledge. In some areas, such as employer benefit plan decisions, employer demand for older workers, and savings and consumption choices of individuals, there is no agreement on the underlying behavioral phenomena. The primary reason for the lack of agreement is lack of data: key data elements are missing or grossly inadequate in one or more respects for either cross-sectional or longitudinal analysis. In other areas, such as labor supply and retirement deci- sions, better data have been available and more is known. However, there is still disagreement about the strength of key relationships (e.g., to what extent Social Security or Medicare influences age at retirement), and there are still areas that are not fully explored (e.g., the retirement behavior of women and joint decisions of couples). New panel surveys, such as HRS and AHEAD, promise a rich set of information with which to refine labor supply models and also to unlock some of the puzzles in savings and consumption behavior. However, there has been little opportunity as yet to mine these surveys and determine their power or to identify enhancements that may be needed. In still other areas, such as mortality projec- tions, the need is to develop more sophisticated projection models that exploit existing data and to develop methods for estimating uncertainty in the projec- tions, which typically extrapolate past trends for long periods into the future. We believe that little progress can be made in the development of improved projection modeling tools with which to estimate the likely effects of proposed changes in retirement-income-related government policies until improved ana- lytical models are developed and key knowledge gaps are filled. Filling these gaps requires priority attention to the underlying data needs, the topic of Chapter 4. It also requires systematic research. We end this chapter with a list of priority topics for policy-relevant basic research that should move forward as new and improved data become available; see Box 3-1. Research in many of these areas should be extended to include the experience of other countries with policy initiatives that may be considered in the United States.

OCR for page 39
58 ASSESSING POLICIES FOR RETIREMENT INCOME ............................................................................................................................. E- - - - --m-p -ogre else lit ec-ls-l-o-ns~ a-e-c e-m-a-nc or t c -e-r~ wool ce-rs~ . . . . ............................................................................................................................... - . - . ... .. ................................... -. -. -. -. - - -. - - - - -. ............................................................................................................................. ................... .- - - ....... ........... - t-' 'Al'' i' '''''''A'' '''' ' '' 'r' ' '' I'' t' ''' - -- r At- '''' ' ''ill'''' ' ' I'i'l'~' ''' '''' '' ' ~ '' ' +' ' ' ''''''''''''''' V '!'~1'1'1'1~'~'.'''''~!''1'''~1 'V~!''l~'t~''' ~ --I~-vlVl ~ i ! -Iv-Hi- AVAIL lI l~IU-~ ~ ~V! I twit t~;lV'I'!'''''''''''' . ~1:.:.:.:1.:.:.:.:.:.:.:.:.:. 1:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.: ~ of such motives anouac~o so as taxi u-e-~--erra-~ co-n-s-l-u-e-r t!-o-n-s~ annul roe use of :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: . ~ - * ~ ~ oe-n-et-lt~ ants as a---~-e lee tor o-r~e-r~ se-l-ectlo-n~ ---- ~'1' let e-n-n-a-nce-m-e-nt~ ''''''''''''''' ' ' '' ' ' ' ''! - ' '' '' ''' '' ' ' ' ' ' ''' ' ' '''' ' ' ' '' ' ' :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: ~ alsctlmlnallon rules ana olner mgulalo~ p~vlslons ma laws mealcam :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: pen-s-l-o-n~ l-n-s-u--~n-ce- on e-m-p-~-oye-r~ ~-ec-l-s-l-o-ns~ a . . . ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: :::::::::: .................................... - - - - - ................... ... ... ... ... .... ............................................................................................................................. .......... p-lal--n~ e-mp-!-oye-rs~ aecls!-o-n-s~ aDo-ut~ n-l-r!-n-g~ any ele-nTl-o-n~ ..o T~ o-!-a-e-r~ wo-rKe-.r ~ :::::::::::::: :::::::::::::::::::::::::::: :::::::: :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: ::::::::::::::::::::::::::::::::::::::::::::: :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: ~ T -- -i- so fir i--- te-e-st~ i-- - 1- -de---e- -- 1 - - so health a--e as f --- Ide a d .......................................... ~ yo-u-na-e-~ worm- so wove- p-roO-u-cTI-v-lly~ and totals co-m-p-ensal-l-on~ p-mTl--l-es~ rye ~ , ................................. -. - - -. - - -. - - . ---age~ l-o-rm-al~ ant l-n-l--o-rm-al~ pe-~o-n-n-e-l~ ru-le$-;~ p-~-~--ra-m-s~ l-o-r~ err ::::::::::::::::::::::::::: ::::::::::::::::::::::::::: ::: :::::::::::::: :::::::::: ::::::::::::: ::::::::: ::::: :::: ............................................................................................. ............................ ..... ............................................................................................................................ . , ., . i , ... . . .... . ~ J-o-D-s~ me p-e-melvea~ su-D-sl-llulaD-IllTy of yo' -n-ge-r~ any o-la-e-r---wo-rKe-rs~ an-a lintel . ~ . ... ................... , - ... * . , - -., - ... ..... . - - - ~ 1 - Am- ~1 - - Ed '' '' ' " 'A Itl - t I a I E t E ~nt^~mgtl~r ~^ rat! lt'^n on ton E^m^m E tot lends door- 1:-1:-:1-1:-~:1-:1 :-:-:-1:-:1-:: :!:-.. :V:~ I:~ Ll:~ l: :l:: ~:l: l::: Ll: :l A:: :~ l: :l :! ~l: l:~:: ~ ~:l::: ~:l ~ ~:l:::: I -- j - j ......................... :.:.:.:.:.:.:.:.:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:- :~-~:~:-~-~-~:~ -_-~-:-:-:_-:~:~:~:~:i :-:-:~:::-.-.A:~:~::~:~:~:~::::::~::::~:~:~:~::::::::::::::::::: 1---V-I E-Ual---~- EN EE'I'~- ~VE'l-'~l'I'E-'~EV't't'''~C-'l-'I'Vl'~'~''''''''''''''''''' . - . i i - . I as a- E-a co-E -s- E-m-oll-o-E ~ De- E-a ---lo- to~ Dense exo-l-al- ED Tre bale l--E ~ ~-e-rso-n-al~ save . ~.r. ~.' ''' ' i ' '' '''~te'i ' ' ' ''''' ' ' ' ''' 'i ' ''''' ' ~e'' ' i' -fit - --h- i li t ~ Huh- ~ if--- i--l- '' ''' a' ........ TV! - " - ~l l"t ~ Bevy! ' - ~ ~"v'! t - - !V' ~= ECU `;~ - ~ - l ~ Mu! - V' ............. HE- E-m-~ Ul!'!'e'(eE i'''''menta~ aCcOu' -my '!-O'E'-'~-|'-E'-E'e-te- -E- ::::::::::::::::::::::::::::::::::: ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: ~ Itle E mO Eel SE O EI E IE testate COE S E'm~tlOE and sa IE ES Itn la 0( $ Entail - ~ ............................................................................................................................. - ~^I=I^r`= 1::~:1r`~^ baler ~:~:~:::::T ~ mIr Air =~= Em mr`^ ~ A:lO:~:l::::Ti - :::::~:~:~:~::::::l:::l ~ T.-^r l ~E:~:~::^T:::::::::::::: . ~Ew~l-!'-!'- E''- !- -1-w - - Ivy v!- -E I!- -| y E V ~! Ed ~hi y ~= |- w! l ~vv y ~I ! I .,.-,.,= U ! ~ ~! ~- v ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: ~ ~ ~-ele-rog-ene-l-ly~ l--E ~ savl-E-gs~ D-e-E -av.-l-o-E l-ncl-u-a-l-E-g~ onyx .. .... ...................... '"""' '""' ,,,,,,,' E"""' ti- - - t-- ;tb~ I-ittl -- i- - $ ~ t ::::::::::::::::::::::::::: :~: :~:::::::::::::::::::::::::::::.::::::::::::::::::~::::::::::::::::::::::::::::::::: :j - :::::::::: ,,,,,,,,,,,~^l~t~r ~ I --- ,, ~ ~ El-' :' ' '' '' '' j ':':t:' :':':E ' :' ''' :~'E:' E'' '' ''' '' ^''' ' ' ' ' +' t'l' 2 2'''''''''''''~=''~'t'i'~ ='E'E 1 'E~=''l'l'l'''~ l'~'l'= l'~'l'l'~='l'' E'l l' I ='i'= l''l E ==''E'l'l' E 1:'E~'VV :1'= E E E , ~ E | ~:: =^E =~:E :l'V E':l=:':':':':':':~ ~ a~o-u--l~ s-ou-E es o! rell eme' l l Ecome a' u llE ely lulure even'$ llq EIUI~ COE .............. T ........................................... ~ ~n ^---OI--I^~Tl-TI ITO^III^/':':':'^T:':':'^:lTT~ r~nr T^rmO ~T O- E`Jln^~'''''~E ^~' Tl^Oll\, T='V-:-:-':':':':':':':':':':':' ' ' ' ' ' I 'l~' ' ' '~'E '~-'~I 'E 'E '~E ~ !-I -l L~- ~ l ~ l l l ~'L' ~l 'l'E :':':':|'V:|':'|:'L':|'~' ' ' '~ |' ' ' '~'V |: :| ~ ~ ::: ~ ~.- ~|: |: |:V~:l |: -:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-: . .-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:

OCR for page 39
KEY RESEARCH ISSUES 59 ~ -aTT=rn~ ~ :~:,lnn~ an E::::E 'e=~`,lnn~ ln ~" ^~-T_r ET'r~m~nT ~'. r :':':':':':':':':':':':':':':':':':':':':':':':':':':':':':':':':':': :':':':~'b'~-t~ , ~ - - - . :':':':'W' :-:'l: ':'~:~':':':~:':~' E:':':':' ~-'I:-:'~:~':'~:~':':':'~:~':'~:':':':' :':':':':' '':':':':' :':':':':':,: : :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: """"""""""""""'i''nC 'UPC 'E'n'0""" El$pOSE Ion o -o-us-n-~ ant ~ p-e-nsio-n~ asse so O C -e :e- m-l--n - ~ . ~ :::::::::::::::E nits ln E E E nor In hi/= _ Plane ET E ME r ~Eta --------- .................................... ::::::::::::::::::::::::::::::::::::::::::::::::::: ::::::::::::::::::::::::::: :: :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: ::::::::::::::::::::::::::::::: ::::::: :::::::::: a'E ~ ~ ex~e- -c l~ :u--res. o. s-avl-n-as. D-e 1avior a -c COE su-m-ol-l-o E. l-eve-ls. ane-E. ret ~ . . :':':':':':':':':':':':':':':':':'::::::: ::::::::1:::e::l ::::: :I:': :: ': :: .- : :' t ::::1: 1 :~:e:~:l: E h ~Af:: : E E~ A t ~1 ':':':':':':':':':':':':':':':':':':':':':':':':':::::':':':':':':':'=:U:!': 1:'1: - ~U':l':!':'l:':' - :~. ,:1 :'t~:!' - :!':'I:':':!':=' - ~'I: - 'E - - ':':':!'i':I - I:'E '~'I:'I:':!'~':':'I:':!'V:V'v:':':!':'l:'l:"'l:':!':V:':':"'l:':!'~':':'v-v':!':'l:"'E ':':~! ~ ~At~r~= r~ ~-r=~ l-~-~-ln^~ ~-l-l-~-r l~ r-~n=-l-l-l ~---=-n-E-~--- I-r 1=~ ~-n-~-~-E~-~-~=n^~e--- l ^-~---=~^n^-m-~ ~-V 1- E~:~:. ~:1:~:::~:|~1:~: E~:::~-01::|:::~!''1~'l'l'E~'''~1'1' E::':':~|':'|:~':':':'-~'|:':|'-~'~' E~!''| E E t ' ....... '''''''''''''''''~'''''''~'=='~'E En=''7E r''tn=''~l'll'~'r'E "ril'n' E''ln''r ='rO^:n:~:'E:':'O'=~/'I:n:~:':':':':l'n'^l:l:'t^'l:n^':':~:l':'l:~:n:':':':':':':':':':':':':':':' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' '1' ' E ~ ~'' 'l~' ' ,'~'' ' ' '~, ' ''~ ' '~'~'V'~', ~', ' ' ' E~ ~l'' '' ', ~' ' ,'l' E ' '~ ~ ,' ~', ' l'~ l' ' - ~'V-' l', ' l' E~ ' ' ' ', ,' '1 ~'' E ~ ''l' E y ::::::::::::::: :::::::::::::::::::::::::::::::::: :::::::::::::: :::::::::::::::::::: ::::::::::::::::::::::::::::::::: ~ 'fa'-t' 'E -'-'as'-'-th'e-'-'eRe' ~-'-'' f'~ e~ e- t~ fl -I-e~ ~ I - E""' t""~'' ' '1" I se l :::::::::::: :E::::::::::::::::::::: ::::: :::::::::::::::: :::::::::::::::::::::::::::::::::':::':::':':':::'' :':':':':':::::':':':::':':::: :~ T\:-~ :~T^-~-:~:~ ~:~T^T\-J:~ ~:^T~ ~-r~^Er^:~:~:l:::^:~:^E:::T~::::~T~:l~:~::::^T:~ :~:~:~-1:~:f~-~:~ :~:~:1~:~:~ :::::::::::::: :L V:::: : ~1: :1~:1::: :~1: =:~:V::: ~ :1~:~::: 1: :v ~1: "1: 1:1:~: t::: ":1:1:~::: L1: :1~::: :=A:~= 1: :1~1~:1: 1::: :W t::: =:1: 1:1:~ I:v V:~.: 1:::: :~:~1: t=1~:1: 1:: T:':':':':':':':':':':-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:- 1-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-~:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-.:-~:-:-:-:-:-:-:-:-:-:.-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-.:-:-:-:-:-:-:-:-: ~ n^---~-/~-rlml--l~---ns~-E ~---T^---^E'^r T-l-l-~-l~ r-=rl-r^-m-=nr- e--I-I^n O~ ~ E=l-~---^T---~ Elr---~ :::::::::::::::::::::::::::::::::::: ::~: E~::::V:~|::l~:::~:~:l:l~:::tV:::~.:Y:~':~:~: E~| | ~| | ~,:l:!:l:~l::l~:: ~E~' ~ - ~|:~:: ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: ~ se-q- Eences TOE economic weEI-Deln E~ ane-~ rel-l-' -me-' l~ ::::::::::::::::::::::::: ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: ~ [1e ~Ee 01 E OE EmOE eEa~ aSPeVIS V E WV- K~ |- -C-E-U-OIng .............................................................................................................................. ........................................ - - - - - ::::::::::::::: : : : T::: : ::: :;t f ~::: :: :::: :t : ~ : ::::: : ::: :~: : : :t ::~ : ;: :; : : ::::::::::::::::::::::::::::::::::::::::::::::: ~ V l I I ~ I ~ I I ~ "O ~= I ~= I V ~ ~ E V V V V l - = I ~ ~ I I I ~ L! I ~ I I I ~ I I ~ ~ =~I ~ I ~ I l~ ~ :::::::::::::::::::::::: :::::::::::::: ::::::: ::::::: ::::::: ::::::::::: ::::::: :. ::::::: :::::::::: ::::::: ::: 1::: ::::::::::: ::::::: :::::::::::::.::::::::: ::::::: ::::::::: ::: ::::::: ::: ::: ~ - ~eeSIImaIIOn OT UYnaE -IC--SIrE ~-E -ral--mO-Uel-S---OT--Ia~OE--S EPPIY anU re .. :::::::::::::: *::::::::::::::::::::: ::::: ::::::::: :::: :::::::::::::::::::::~: :~:::::::::::::::::::::: :::::~:::: :::::: ~ t E t ~ l l It E I t ~ t ~ l ~ t l l ~ l ~ ~ :.:.:.:.:.:.:.:.:':':':':':':^'~':':'^:m:^l^':l^:r:':' E^n~l'^^''~'~'~''' E^^'lt^'''l^^l''l'r-. E: E^^''m'l~^'''rl''l'l^~.'''^^'~'lt'n'''=t~'tl''l^''''^l^' ''''''''''''''' .:.:.:.:.:.:.:.:.:.:.:':':':'V:!':'I:':':='I:'I:':!'E EV':Y:='l:':':':~w'l:':!'~'l:V'l:':!':':'":!':'l:~:':':'l:':!'w.-'":!'Ll:':!':':':!'l':la'~':!':~l:':!'V=':':'~:!'"l':l:':':l:'~':!w'~'':':':'l:':!'=:"l:~:!':'l:':':~:~"'L'E :::::::::::::::~:~I:l:l~t::::~:n:E :::I:l:lnE =~ :::~:n:E :::'~E ~1:~:1:::~ E^~:l ~ll~ ::~:n:E T:l:l:~:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: :::::::::::::: :~ F:~ :~ ~':'':':':'~:~':'~:~:':':'~:~':~'~:':~'~V'~'':':':'~:~':'~:~:':':'w'~V-~:~-~:-:-:~. :~-~:-:~-~:~ ~:~ ~- -~ ~. -~ -~ ~ . :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: :::::::::::::::::::::::::::::::::: - ::::~:::::::::::::~:::::::::::::::::~:::::::::::*:::::~::::::::::::::::::::::::~:::::~:~::::: ~ txte-n-s-l-o-E OT ex'~I' a an al-wl-cal~ Eooel$ to a Eores$ sEEmultaneo ES""""""""""""""~ :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::~::::::::::::::::::~::::: ::::::::::::::::::::::::::::::::::_::::':::::::::::::.:::::::::::::::: ::::::::::.::::: :.:::::::::::::::::::::::: :::::E::::::: ~ txm-n-s-l-o-n~ ol~ exl-stl--nO~ E -o-a-e-ls~ to~ aa- E-E ss~ l-o-l-nt~ reb-re-m-ent~ ~-e -:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-.-.-:-.-.-:-:-:-:-:-:-.-.-:-:-:-:-.-.-:-:-:-:-:-:-:-:-.-.-.-:-:-:-:-:-:-.-.-:-:-:-:-:-:-:-.-.-.-:-.-.-:-:-:-:-:-:~:-:-:-:-:-:-:-:-:-:-.-.-:-:-.-.-:-:-.-.-.-:-:-:-. ::::::::::::::::::::::::::::::::::::::::::::::::::::::':::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: ............................................................................................................................. ............... - - - - - ~ t tl I l~l ll ~ Il t 1 E E jA 1 t t r : ~- t ............... ~.~E. ! I .~. !." ~ !.~. I V ! I ~.U.~E. l.~.Ll .V. ~:.:.:~:~ I:! ~:! ~:: \ :~:.: - .: ':: :V I: :! ~: :~:: - V V l ! l l ! ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: :~ :=f .~ ~-~=~-~.~ ~.~ n=-n-~-l~n---,-.-~-n-~.-l-,-~-.~ l-n---~.-~-^~---l^~ r~. av ---,-,.~---~,-~-n-^---~-~-~-l-~l-~-l-l~n~ :':':':':':':':':':':':':':':~ E':':':'l:':~:V~,.:!'' :~:':':':V:!':':':':E.'~'I:':!'~'I:V,,:~:'L':':':~'I::!~:!:!:~I: :':':':!'L':~:':':':V:!':'~'~, :!':':':':'E~':':':'I::~:~:':':':~:!':'I:~:':':': E:~!':~'l:':!'Y'''!'~'~'E'I'!'V' ::::::::::::::::::::::::::::::::::::::::::.:::::::::::::::::.::::::::::::::::::::::::::::::::::::::::::::::::::::::..::::::: """"""""'""""""I"E -ol-l-eo~ ~- t Ee rational e nectations~ a~macE tE at IS w~lcall seo~ to~ ~2 2 2 2 2 2 2 2 2 2,E-2 ' ' '' ''' J ::::::::::::::::::::::::::::::::::::::::::::::::::::: :1 :::: ::::::::::::::::::::::::::::::::::-:-:::-:::::::::::::::::-:-::::::-:-:-:-::::::::::-::::-:::::-:-:-:-:-::::::::::::::::-:::::::::::::-:-::::-:: ~ ~ I ~ ~ i E I ~ 1 l E ~ ~ ; ; t . t; E ~ E ~ ~ h ~ t t :::::::::::::::::::::::::.:.:.:::::::~:~::::V|::::: EVvl=|::::~U:1:I:' V :::~:|:|::||:|:|:|~:~:|:~1:V|::|::::|:=VV|::~:~::::~|::|~::::V L1::1~'::::~:~:::t ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: 1 . 1 1 1 E :~ :~:~-:I-:m:~:-~:~ m-~:r.-E:- E~ l:E-~:I:~ :~-. E-E:-~ r-~-l~ :~- E-:~ E-l:n: E^:~ :~:-^f ~ m^:~l:-~: E ~ ~ E ~ ~ l t ~l n ~ ~r ~^ ~ ~ ~ l m :':':':':':':':':':':':':':':~:~!':!':!':'l:"'`=':':':'l:'l':!'V:!':'E l:'l:t,':':':':!':~Lm,-'~':':':'~'Y:':':':='":!':I:'l'l:!':'l:y - ':':':"'l:':! - ':':':'l:':!':l~!':'l:'L~l:':':': E:~t E - ' I! I - .. I .~:!::::LV::::!I:~: .. .. I tl E S t t E I I I I $t~ $ ~ E tl t ............................................................................................................................. ....... '''''''''''''''''~''''''' - ^T'-='r''''m'^ E^l'O''''^T'''Tn'~'''^^T^rm'ln='n'~''''~'n^'''^^'n'e-='m'I'l~n^~O'''^T''''r E Erl'~'i''''''''''''''' ~- ~ L~ ~! ~ l V E ~ l ~ l ~! l ~r~- t~ l ~ ! l ! l ! ~! l &- ~ l ! ~VV !- -' E~, :~: E~ l: ! V~ ~!

OCR for page 39
60 ASSESSING POLICIES FOR RETIREMENT INCOME .................................. - - - - - - - - 't' ''' ' -'-f' ~ '' I-d'' -''-'-' ' '' ' -1'' -'-'a-' -d-'-'th' '-'-'if' ' t' ~ ' Tim ' '' t-'-'i-'' ' ' '' '' -'-'' ' -d-- It-h-- fly -:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-.-:-.-.-.-:-.-:-.-:-:-.-.-.-.-:-:-:-.-:-:-.-.-:-.-:-:-:-:-.-.-.-:-.-:-.-.-.-:-.-.-.-:-.-:-:-:-1.-.-:-:-.-.-.-:-.-.-.-:1-.-.-:-.-:-.-.-.-:-:-:-.-.-.-.-.-:-:-.-:-.-.-.-:-:-:-.-.-.-:-.-:-.-.-.-:-:-:-.-.-.-:-.-:-.-:-:-.-.-: ..................................................................... .. ......... . - ........ .. ......... . -. -. ...... _ ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: .................................. _ _ ............ .- .......... - - .... . ~ e-- 'O'S"""I""' - ............................................................................................................................. .............. .............................. 4 ....... ............................................. ... ............. . ~J---~ Ann I for -inn Mel lT~ Ann --- -nnm_ ~rm---~=r~--~. ^e~--~J=-r--T Aim--= ~-n---^ --ran A 1 l! lo ~ 1 ~ l y ~ Eve t~ Al a`` lo. I l y ~ ~ l l l v~! ~vv~ ~ - ~ Y ~ 1 ~ 1 ~ ~-~1 1 ~ 1 ~ . ... ... - ~ 'ti-' '' '' -' to Add''' ' ' -'-'t-h~ hi'' '' ' tin 'I-'-'' h' th'' ~ 'I-i'f' ~t' '' ' ' '-'-i'' '' '' ' ' ' ' ~ ............................................................ d ........................................................ :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: ~ mo-r ala-lly~ and a-lsa-D-I-l-lW~ wl--l-l~ l-l-Kely~ co-m-p-ress~ l-nlo~ Tewe-~ years ::::::::::::::::::::::::::::::::::::::::: :::::::::::::::::::::::::::::::::::::::::::: ::::::::::::::::::::::::::::: ::::::: ................................... -. -. -. -. -. - - - -. - - -. - - - - ............................................................................................................................. ........................................... - - - - - - ~ . ~ I ~ ~ t . ~ 16^ . t tl ~ ~+^ ~'~ t ~t t tl ~ '''''''''''''''O'U~'E'I'''I'=~ [V I'~'''"O'''I''E=. "'E LI''I''~'L=LU'O'''''~ =.'E'I'~'= I''I'''= LI''I I''I'EW*'L V'l'''=l'l''l ~IV V'l'l''l=. l'*'L'''O'L"LU~,'''M'E'l'~''''''' ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: T . . . . ~ e' ~ I ta Eli il E E E E m E t a E~ tE 1l el ~--t E v . .:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.~:.:.:.:.~.:.:.:.:.:.:.:.:-:-:-:-:-:-:-:-:-:-:-:,-:,-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:~-:-:-:-:-:-:-:-:- 1 -:-:-:-:-:-:-:~-~-:-:-:-:-:':':':':':'.:':':':':':':':':':':':':':':':':':':':':':':':':' ' ' givlE ~ pal~er* s l* ClUalE ~ ! _::::::::::::::::::::: :1::::::::::::::::::::::::::::::~::::::::::::::::::::::::::::::::::::::: . -^rr~r~r~r~r~C~1~^ 4~mC~^C~rr~r~TC~ ~ ~^rr~r~r~^r~TC~ m~ Ir~ :mr~ ~At~ITr~ -- ~1 1~1 =1 1~1 I~lV ~"~.~1 1~1 16 ~Vl ~ll l~vl l~l lL ~Vl ll lvV1 1 I ~"1 1 - VV="IL I ............... -. - - -.- - - - - - ::::::::::::::T a T:: :A -: l:::~: :: :: T:::^IT -. : T:: ~ : T ::: E . I l ::T:~: :1:::: ::::::::::::::::::::::::::::::::::::::::: ~-~ -= l- -l -l w-l - l-~ - - v v w-, l~-~-~-l - l- E - - V -~ - - --I I I ~ I ~ I I ~ wY I I V I L~ ~I I ~E l v ~ ~- ~= ~ l l~ Y ~ E w. . ~ - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - ~ - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -