WPC# 2BVTZ #|x)?xxx,/Xx6X@DQX@HP LaserJet IIIes IIHPLASIII.WRSx  @,,,4a(HX@2@VR Z Courier New (TT)?xxx,Xx6X@DQX@HP LaserJet IIIes IIHPLASIII.WRSx  @,,,4a(HX@(26FO3#|x SleekNetTable of contents WP Defaults/$` hp x(#$#x6X@CX@#footnote tex#'#x\  PCXP#2gXcl@'endnote text#'#x\  PCXP#endnote refeDefault Paragraph FoDefault Paragraph Font headerheader` hp x (# (# 2d be @`  page numberpage number endnote referenceendnote reference footerfooter ` hp x (# (#  ;1#XO\  PUXP##X\ P:+QXP#2  vj  O heading 1heading 1 C9#g2PkCP# #C\  P6QP#footnote textfootnote text  footnote referencefootnote reference 44#XP\  P6QXP##C\  P6QP#biblio d` #n\  PC&P#2! ?< 3'3'Standard3' 3'StandardHPLASIII.WRSx  (   ЊX1Í.X1Í. 199495 ADVISORY COUNCIL ON SOCIAL SECURITY pTECHNICAL PANEL ON TRENDS AND ISSUES IN RETIREMENT SAVING $FINAL REPORT " 0*0*0*" T  199495 Advisory Council on Social Security T pTechnical Panel on Trends and Issues in Retirement Saving T $Final Report TP  September 29, 1995 T T TABLE OF CONTENTS TPT TP  ? <X` hp x (#%'0*,.8135@8: S<  #Xx6X@DQ/X@#э The inclusion of discouraged workers adds over a third to the unemployment rate of men aged 55 or older and adds over 70 percent for analogous women, but has only a very small effect on the unemployment rate overall (Herz and Rones 1989). > Although older workers (in the mid1980s) were about as likely as younger workers to be laid off, those who were laid off were much more likely to end up out of the labor force about onethird of those aged 55 to 64, threequarters of those 65 or older, compared with fewer than 10 percent of those 20 through 54 (Herz and Rones 1989). Older workers who are laid off or fired are more likely than others to experience long spells of unemployment, and they suffer greater earnings reductions if they do find work (Shapiro and Sandell 1987). Older workers may face declining job prospects as they age for several reasons. Some may suffer age discrimination per se, but others face sources of labor market adversity, sometimes privately efficient, that happen to increase with age. For example, firms with large fixed costs of hiring or training may prefer younger workers who offer more potential years of employment over which to amortize the fixed costs. In some cases, longterm, mutually beneficial implicit contracts between employers and employees may require cessation of employment or a significant cut in pay late in life. Lazear (1979) has argued that firms might lower turnover costs by paying workers less than their contribution to the firm during the early years of employment, and then more than their contribution near the end. Some compensation is delayed to encourage long tenure with the firm. If this pattern of compensation reduces turnover costs and the gains are distributed between the employer and the employees, then both parties can benefit. But at the point when lifetime compensation and lifetime contribution are equal, the employment contract must be terminated or the worker's compensation decreased to the level of his or her contribution. Although this agreement looks discriminatory at the end, it may have benefited the workers over the life cycle. This theory offers an explanation both for mandatory retirement, now generally outlawed, and for pension plans that penalize workers who stay on the job "too long." "'$0*((@@/"Ԍ An additional problem may be the fulltime/parttime mix of jobs that are available. Many older workers would like to retire gradually, which often means a period of parttime work before complete labor force withdrawal. In a recent survey of older Americans, 21 percent of the working men (aged 5564) and 43 percent of the working women (aged 5059) wanted to work part time (and the proportion increased with age), but only 6 and 19 percent, respectively, were actually doing so (Quinn and Burkhauser 1994b: table 1). Given these preferences, why are more older Americans not working part time? There are several possible reasons. Compensation is often poor on the jobs that are available, and many employers are reluctant to hire older workers at all. Workers who move from full time to part time usually do so at considerably lower wage rates and with fewer benefits (Gustman and Steinmeier 1985; Jondrow, Brechling and Marcus 1987; Quinn et al. 1990; Kramer 1995). Lower pay for these older workers is not necessarily evidence of discrimination. Workers who switch jobs lose specific human capital the expertise acquired on the old job that is not relevant on the new one (Shapiro and Sandell 1985). Their productivity can therefore decline when they move, but for reasons having nothing to do with age. Fixed hiring, training and employee benefit costs are amortized over fewer hours, both because of the worker's parttime status and because older workers have fewer years of service to offer to the firm. If so, firms may have to offer lower wages in order to make the arrangement profitable. As a result, firms often offer parttime employment in jobs that require relatively little supervision or training, and exclude fixedcost benefits from the compensation package. A final obstacle is employer attitudes toward older workers. Although employers speak highly of older workers' strong work ethic, loyalty and dedication, many fear that older workers are difficult to train and do not cope well with the technological aspects of many jobs (American Association of Retired Persons 1989; Belous 1990). In summary, older workers do face labor market obstacles as they age. Many who remain on their career jobs face declining compensation, as definedbenefit pensions or social security wealth decline in value with continued employment. Switching jobs is difficult because many employers are not eager to hire older workers, for reasons that may or may not be accurate. Parttime work usually results in lower pay, fewer benefits and reduced economic status. In a detailed survey of these issues, Straka (1992) concludes that without the elimination some of these demand side obstacles, attempts to increase the working lives of older Americans through antidiscrimination legislation or supply side measures may be ineffective. "'%0*((@@/"Ԍ ?< Conclusions: The labor market for older Americans has undergone significant change in the postWorld War II period. The most dramatic change has been the trend toward earlier labor force withdrawal for men, which lasted until the mid1980s. For women, the early retirement trend has largely been offset by the simultaneous increase in the participation rates of married women. Considerable evidence indicates that the growth of and financial incentives imbedded in social security and many (definedbenefit) employer pension plans played an important role in this movement toward earlier retirement. Many older Americans do not retire in one move, directly from full time career employment to complete labor force withdrawal. Rather, some retire more gradually, taking bridge jobs in the interim. These are often part time jobs, sometimes involving selfemployment, and usually involving a change of industry and/or occupation. The bridge jobs generally pay less than the career jobs did, and are lower down the socioeconomic scale.  ?< B. The Changing Nature of Employment   ?0< Broad trends in the labor market will affect those approaching retirement age today as well as those whose retirement decisions are far in the future. Job mobility can influence the relative attractiveness of different types of pensions, and changes in the distribution of earnings can have implications for future elderly poverty rates. The Panel discusses several relevant labor market trends and asks what they might suggest for the future. The recent and current emphasis on corporate downsizing, combined with a recession in the early 1990s, has contributed to a perception that the American workforce is increasingly mobile and that jobs are less stable than they were in the past. Recent layoffs, including middle management positions in bluechip firms once noted for their loyalty to workers, have strengthened these perceptions. The evidence backing these perceptions is mixed. Although there does appear to be a secular growth in the importance of parttime and contingent workers (those hired for a fixed term), aggregate evidence does not indicate that average job tenure for the American workforce is on the decline.  ?(#< Contingent and parttime workers: These workers lack a permanent fulltime attachment to a single job. They may work part time on one or more jobs, may work as temporary employees through an employment agency, or may be subcontractors or independent professionals. "'&0*((@@."Ԍ Temporary employees, probably fewer than 3 percent of the workforce, fill in for a short time on jobs that have been  ?<vacated or created to meet a shortterm increase in demand.2 V ?<  Ѝ` hp x There are no official statistics on the number of contingent workers. An important subset is the temporary help industry, which "has grown from onethird of 1 percent of total employment in the early 1970s to nearly 1.3 percent today (1992). While growth has been explosive, the fraction of the workforce employed on a contingent basis is probably still less than 3 percent." (Council of Economic Advisors 1994:123). See  ?x<also Belous (1989).footnote tex#x\  PC DXP# 2 Many are supplied by agencies that employ the individuals and contract out their services. Agencies that originally specialized in providing clerical help now provide a wide range of skilled and semiskilled workers. Leased employees are placed in jobs lasting a year or more, and have become a popular way for firms to fill jobs without providing benefits or making longterm commitments. Independent contractors are selfemployed people who work on a contract basis for one or more employers. Growth in these categories reflects employers' interest in having greater labor supply flexibility without the employee benefit obligations associated with career workers. Selfemployment, which overlaps with the independent contractor sector, accounts for about 9 percent of the total workforce, a proportion which has changed little over the past  ?<several decades (Quinn 1995).K V ?<  #Xx6X@DQ/X@#э` hp x#Xx6X@DQ/X@# The official statistics on the number of selfemployed exclude both the owners of incorporated businesses, even when the owner is the only employee or one of very few, as well as wage and salary workers who also own side businesses. Haber, Lamas and Lichtenstein (1987: 18) estimate that the "percentage of workers who owned businesses was 60 percent (SIPP) to 75 percent (CPS) larger than the percentage reported as self ?x<employed."#XP\  P6Q DXP# K The selfemployed have to provide their own benefits beyond social security. The selfemployed today have the lowest rates of pension coverage (16 percent in 1992) of any broad category of American workers.  The proportion of the workforce working part time has stabilized in recent years following significant growth during the prior decades. It grew from 15 to 20 percent between 1969 and 1983, and has remained about 19 percent since then (Saltford and Snider 1994: Table 1).  ?< Job tenure and mobility: Despite the broad press coverage that corporate downsizing and interrupted careers have received, aggregate government statistics do not support the notion that job tenure is declining. For example, over the past 15 years, the fraction of workers remaining on the same job for more than eight years has remained constant 30 percent in 1979 and 31 percent in 1983, 1987, and 1991 (Council of Economic Advisors 1994: 126; see also Diebold et al. 1994). Americans are mobile, but they have always been so. These aggregate numbers, however, might be concealing offsetting changes underneath. For example, the increasing proportion of older workers (who tend to have longer job tenure) might be masking declining tenure statistics within age cohorts. Age and genderspecific tenure statistics suggest otherwise, however, because they appear to be fairly stable for men in all age groups and slightly increasing for women (Yakoboski and Silverman 1994: table 16). It is still possible that the aggregate statistics conceal offsetting changes within certain demographic of socioeconomic subgroups of the population. For example, there is evidence that black workers, lowseniority workers and workers without any"''0*((@@/" college education have experienced a decline in job stability, but there are much more conflicting signals on the collegeeducated population (Diebold et al. 1994; Farber 1995; Marcotte  ?X<1995; Swinnerton and Wial 1995).j V ?<  Ѝ` hp x Marcotte (1995) attempts to reconcile the divergent research results on trends in job stability. Among the explanations are changes in the wording of a key CPS question in the early 1980s, and correction techniques for survey non ? <respondents.footnote tex#x\  PC DXP# j The most recent recession (199091) appears to have affected college educated workers and older workers more than previous recessions had (Farber 1993; Gardner 1995). This raises the possibility that the public perception of increasing mobility is caused by the recent experience of members of the labor force who traditionally enjoyed the most stable employment patterns. If so, the aggregate data may be concealing an erosion of the kind of job attachment that is associated with retirement income accumulation.  The Council of Economic Advisors (1994:126) has summed this issue up nicely. "Whether or not job security is decreasing, two things are clear. First, there has always been a great deal of instability in the U.S. labor market. Second, there is no  ?<question that there is a perception that job security is decreasing. This may be due entirely to the normal increases in job losses during the recent recession, to media accounts of mass layoffs at companies that used to offer unusually stable jobs, or to increases in job stability that simply are not reflected in aggregate statistics." Additional research is certainly needed on this issue.   ?P< Quality of jobs: Many analysts have argued that traditional middle class jobs are on the decline in America, and that recent job growth has been concentrated at the extremes lowwage, lowskill personal service workers at one end and highlyskilled, professional and technical workers at the other (e.g., Bluestone and Harrison 1982; Harrison and Bluestone 1988; Levy and Murnane  ?<1992; Rosenthal 1995).' U ?<  #Xx6X@DQ/X@#Ѝ` hp x Kosters and Ross (1987) challenge this view, arguing that the share of workers in the middle of the distribution has been stable, while the share at the upper end has increased. Levy and Murnane (1992) present a discussion of this debate. ' The middle seems to be shrinking, with the decline of averagewage, bluecollar, unionized workers in manufacturing jobs. As late as 1965, nearly 30 percent of the fulltime workforce was in manufacturing jobs while fewer than 15 percent were in services. Twentyfive years later, the positions were reversed: only about 17 percent of employees were in manufacturing and 25 percent were in services in 1991 (Anzick 1993). Jobs in the professional service sector have grown rapidly. These include professional and technical positions that require high levels of skill and creative problemsolving ability, much of which is provided through college and graduate school education, rather than through onthejob training. These are engineers, lawyers, investment bankers, consultants, writers, designers, and a host of other occupations involved with the manipulation, analysis and communication of abstract concepts. This group has experienced real pay increases over the past decade. Many can maintain flexible working careers, controlling"'(0*((@@." their hours and extending their productive worklife into old age if they choose. Gittleman and Howell (1995) use 17 measures of job quality and cluster analysis to group more than 600 jobs covering 94 percent of the American work force into 6 categories, or  ?<contours./U ?<#Xx6X@DQ/X@#  Ѝ` hp x The 17 measures of job quality fell into 5 categories: earnings and benefits, institutional setting (percent unionized; percent in the public sector), employment status (including hours and weeks worked), skill requirements and working conditions. / They find that "the distribution of employment over the period 197390 shifted sharply away from the two middlequality contours toward the two highestquality contours.... [Although] the two lowestquality contours showed no decline in employment share... [there was] a sharp drop in the quality of lowskill jobs." (p. 420) Rosenthal (1995) analyzed data on 278 occupations and found that the majority of jobs created between 1983 and 1993 were in the highest and the lowest paying quartiles.   ? < Racial diversity : The population reaching retirement age in the next century will be much more diverse racially than are current cohorts of retirees. The Census Bureau projects that between 1990 and 2030 the older (65+) white population will grow by about 90 percent, while the older black population will increase by almost 250 percent, and the older Hispanic population (of any race) by nearly 400 percent (U.S. Senate 1991:14). As a result, the proportion of elderly who are minority will double from 14 percent in 1990 to a quarter by 2030, and then increase to nearly a third by the year 2050 (ibid., chart 18). As shown below, minority populations have been underrepresented among those covered by pensions and overrepresented in the poverty populations.  ?8< Income distribution: One result of the growing divisions in the U.S. workforce has been an increase in income inequality, which is found for a variety of income concepts, household units and measures of inequality (Karoly 1993). There is general agreement that the primary cause of this has been increased labor market inequality a widening gap between the hourly wages and annual earnings of those at the extremes of the pay distribution (Levy and Murnane 1992; Danziger and Gottschalk 1993; Gottschalk and Danziger 1995).  Over the past two decades, median real family income in the United States has been stagnant, growing at only 0.2 percent in the entire 20 years between 1973 and 1993, after having more than  ?`"<doubled between 1947 and 1973. (CEA 1994:115; CEA 1995:176).!V ?<#Xx6X@DQ/X@#  Ѝ` hp x These real time series utilize government costofliving indices, which some think overstate the true rate of inflation. If true, some growth would appear over the past 2 decades, but the rising inequality story would be unaffected. ! This overall stagnation, however, conceals dramatic changes. Between 1973 and 1993, the real incomes of the richest quintile of American families grew by 25 percent, while those of the middle quintile were virtually constant and those of the lowest quintile actually fell 15 percent (CEA 1995: 178).  "')0*((@@."Ԍ Studies of wage growth reveal similar patterns. Among both men and women, the distribution of real wages has become more unequal. "Between 1973 and 1993 real hourly wages of fulltime male workers at the 10th percentile...declined 16 percent, while real hourly wages at the median fell 12 percent. Over the same  ?<two decades, workers at the 90th percentile eked out a wage gain of 2 percent. The net effect is that levels of wage inequality for men have been greater in recent years that at any time since 1940. Women received wage increases throughout the wage distribution, but the gains were concentrated at the top. Women at the 10th percentile earned 6 percent higher wages, while those at the 90th percentile had gains of 24 percent" (CEA 1995:176). A major cause of this widening of earnings and therefore income inequality has been increasing returns to education and experience during the 1980s. The collegetohighschool wage premium for young workers increased more than 100 percent between 1974 and 1992, and the ratio of the wages of experienced to inexperienced workers has risen as well (CEA 1994: 117; see also Murphy and Welch 1993). In addition, inequality has increased in  ?<the earnings of those within education, experience and skill classes. The ratio of 90th to 10th percentile wages has increased within the distributions of high school graduates, college graduates, young (inexperienced) and old (experienced) workers. Those in the upper percentiles have experienced significant growth in real wages while those in the lower ends have seen only slight growth or declines (Gottschalk and Danziger 1995).  Many factors have been cited to explain these changes in earnings inequality, including skillbiased technical change (for example, the importance of computer literacy), declining unionization, the erosion of the real minimum wage, changes in industrial structure (the deindustrialization mentioned above), immigration of low skilled labor, and increased international  ?X<competition in product markets.V ?<#Xx6X@DQ/X@#  Ѝ ` hp x A recent volume by Danziger and Gottschalk (1993) addresses many of these hypotheses about why the U.S. earnings distribution has become more unequal. In it, Murphy and Welch (1993) argue that changing industrial structure played only a modest role, and that most of the increased returns to skills occurred within industries. Freeman (1993) finds substantial effects of the decline in unionization, and attributes about 20 percent of the increased earnings inequality among men to this  X@<factor.footnote tex#x\  PC DXP#  Although debate continues about the relative importance of these factors, a general consensus exists that the earnings and therefore the income distribution in the United States has changed dramatically over the past two decades. The earnings distribution has became more unequal both because of a growth at the top and a decline in both absolute and relative earnings at the bottom.  ?!<x Implications for pension coverage and retirement income  ?`"<adequacy: With the exception of the increasing labor force participation of married women, little in the labor market trends discussed suggests that pension plan coverage will increase dramatically in the near future. Middle class unionized jobs have declined, and they were frequently associated with pension coverage. In service industry jobs, coverage is much less  ?'<likely.V ?<#Xx6X@DQ/X@#  Ѝ` hp x Among civilian, nonagricultural wage and salary workers in 1993, nearly 80 percent of those covered by a union contract participated in a pension, compared to only 40 percent of those not covered by a union contract. Less than 20 percent of those in business, personal and entertainment services participated in a pension plan, along with less than a quarter of those in retail trade and 43 percent of those in professional services, compared to 56 and 64 percent of those in nondurable and durable manufacturing, respectively (Yakoboski and Silverman  ?<1994: table 12).footnote tex#x\  PC DXP#  If the importance of the contingent economy and parttime work increase further, longterm job attachment may"'*0*((@@/" diminish, signaling a shift in the obligation to prepare for retirement from employers to individuals (Allen and Freeman  ?<1994).V ?<  Ѝ` hp x For instance, temporary employees are explicitly precluded from participating in employer benefits where they work. Independent contractors and the selfemployed are even more dependent upon themselves to provide retirement and health benefits. In cases where an employer has little interest in or expectation of worker loyalty, it is unlikely that deferred compensation (i.e., pensions) will be used to encourage a longterm commitment to the firm.  This shift can already be seen in the dramatic growth of definedcontribution pension plans, discussed in Section IIC below. Finally, the increase in earnings and income inequality is a concern, both because of its current implications and because it bodes ill for retirement income adequacy in the future.  ?@< C. Trends in Employer!Sponsored Pensions and Retiree Health Benefits  ?< Recent changes in employer!provided pensions are examined in this section. Although the proportion of people in the labor force who participate in a pension has changed very little, a shift toward definedcontribution plans as the primary form of employersponsored retirement plan has occurred, along with a dramatic increase in the number of workers participating in 401(k) plans. Funding and tax regulations are discussed along with concerns about the allocation of risk between employers and employees, which differs by pension plan type. New federal regulations along with changing economic conditions explain many of these trends.  ?</$WP Defaults` hp x(#$#x6X@C/X@#  ?<"WP Defaults"endnote refe The tax and regulatory environment of employer!sponsored  ?<pensions : The Internal Revenue Code provides favorable tax treatment for savings in the form of employer!sponsored qualified  ?<pension plans.1H ?<  Ѝ` hp x(#$` hp x There are many limits on this tax preferred treatment. For example, a qualified definedbenefit plan cannot provide an annual benefit in excess of a certain amount, originally legislated as $90,000, but indexed, and up to about $115,000 by 1993. The contribution maximum on definedcontribution plans is the smaller of $30,000 per year or 25 percent of pay. (This $30,000 limit will also be indexed at 25 percent of the definedcontribution limit once the latter reaches $120,000.) If a definedbenefit plan has assets equal to 150 percent or more of liabilities, determined as if the plan were to terminate, additional contributions are not tax deductible. (For details of this type, see Gulotta 1994; Hubbard 1994; or VanDerhei 1994.)  Employer contributions to a qualified retirement plan are deductible as a business expense. Neither these contributions nor the earnings of the pension fund are counted as part of the employee's taxable income until the pension funds are distributed. Employees may sometimes also make after tax contributions to qualified plans. In that case, the contributions are taxed when earned, not when distributed; however, fund earnings on these contributions are not taxed until they are distributed. If employers establish plans under sections 401(k) or 403(b) of the Code, employee contributions avoid current taxation. #oendnote refe#endnote refe` hp x(#$` hp x(#$ #-endnote refe# The federal government encourages pensions through this preferential tax treatment, the objectives of which include increasing individual savings for retirement and encouraging national savings. To qualify for this preferential treatment, pension plans must conform to a series of government regulations concerning vesting, participation, funding limits, discrimination in terms of coverage and benefit levels, and the purchase of insurance from the Pension Benefit Guaranty Corporation (PBGC). These issues are discussed below. endnote refe #yendnote refe#` hp x(#$` hp x(#$% Tax preferences decrease tax revenues. This loss in government revenue is called a tax expenditure. The pension tax expenditure is the largest calculated by the government, and was"'+0*((@@/" estimated at $57 billion in the Fiscal Year 1993 federal budget, about half of which was due to publicsector plans (Salisbury  ?<1993, 1994).g/S ?<#Xx6X@DQ/X@#  Ѝ` hp x(#$%` hp x There is debate on the size of the federal tax expenditure associated with the preferential treatment of pensions (see Munnell 1989, 1991; Schieber 1990; Clark and Wolper 1993; Goodfellow and Schieber 1993; and Schieber and Goodfellow 1994b). Goodfellow and Schieber (1993:153) argue that at least half of the tax expenditure estimated by the government is attributable to gains due to inflation, and that half of that remaining (i.e., the tax on the real gains) are attributable to public employee plans, where the employers are not taxable entities anyway. g Some think that this tax expenditure is too large for a program that covers only half of the labor force, and that pension contributions or at least the earnings of pension funds should be taxed currently. Others respond that employer pensions encourage saving and provide a major component of retirement income for millions of retirees and therefore should continue to be encouraged. Pension tax expenditures are predominantly a middle to upper income class benefit. Those with incomes below $30,000 filed 42 percent of the taxable returns in 1992, but received less than 9 percent of the pension tax expenditures (Salisbury 1993; table 7). (They also had less than 8 percent of the total tax liability in 1992.) Whether the middle or upper class benefits more depends on the comparison made. If these tax preferences are viewed as simple expenditures, they appear to go disproportionately to the upper class. For example, the 5 percent of the (taxable) households with (1992) income above $100,000 received 20 percent of the pension tax expenditure; the 24 percent with incomes between $50,000 and $100,000 received 43 percent. But the value of the income exclusion is higher for upper income taxpayers not only because they are more likely to participate in employer pensions, but also because they face higher marginal tax rates and pay more taxes. When the tax preferences are viewed not as expenditures but as tax reductions, a different story emerges. Those at the bottom end receive little or no tax relief because they already pay little or no federal income tax. Those with income above $100,000 reaped 20 percent of the pension tax expenditure, but paid more than 40 percent of the taxes paid (ibid.). Those earning $50,000 to $100,000 enjoyed 43 percent of the tax expenditure, but had only 33 percent of the tax liability. And those earning between $30,000 and $50,000 paid 18 percent of the taxes, but enjoyed 28 percent of the tax liability. The ratio of pension tax expenditures to total income taxes paid is the highest for those with incomes between $30,000 and $50,000 per year. Had these pension incentives been eliminated and nothing else changed, those earning between $30,000 and $50,000 would have seen their federal income taxes increase by 18 percent; those in the highest category ($200,000+) would have suffered a 3 percent tax increase. Tax changes enacted in 1993 reduced the maximum compensation that any employee can be deemed to receive for tax deferral purposes from $235,840 to $150,000, which will lower the relative tax expenditure for high earners (Liston and LaBombarde,  ?%<1994).l/S ?<#Xx6X@DQ/X@#  Ѝ` hp x(#$%` hp x For employees earning above this amount, pension benefits or contributions, if left unchanged, would become a greater percentage of deemed compensation. Because of the discrimination test, benefits would have to be increased for the lower paid or decreased for the higher paid (Liston and LaBombarde, 1994). l endnote refe` hp x(#$%` hp x(#$% "H&,0*((@@-"     ?<#w.endnote refe#` hp x(#$%` hp x(#$% Changes in pension plan coverage and type: Employer! provided pension plans are an important source of retirement income to millions of older Americans. Using income data from the Current Population Survey, Grad (1992) reports that income from public and private pensions and annuities provided 18 percent of the aggregate income of couples and unmarried persons aged 65 or older in 1990 (exactly half the proportion provided by social security), an increase of 2 percentage points since  ?<1976.QS ?<#Xx6X@DQ/X@#  Ѝ` hp x(#$%` hp x In contrast, the National Income and Product Accounts suggest that employer pensions (federal, state, local and private) were paying out almost exactly as much as social security was in 1990. See Schieber (1995) for a discussion of  X <this discrepancy.footnote tex#x\  PC DXP#  The importance of pension income to the elderly generally rises with income level from 3 and 8 percent in the lowest two quintiles to 16 percent in the middle quintile, and then to 22 and 20 percent in the top two quintiles. The percentage of the elderly receiving any pension and annuity income tells a similar story. It has risen over time, from 31 to 44 percent between 1976 and 1990, and it increases with income level, rising from 8 to 50 percent between the 1st and 3rd quintiles, and then to 67 percent in the highest 2 quintiles  ?<(ibid.).OQS ?<#Xx6X@DQ/X@#  Ѝ` hp x(#$%` hp x Using a very different data source, the Bureau of Labor Statistics Survey of Consumer Expenditure, Hitschler (1993) reports that private pension and annuity income represented 18 percent of income for households with heads aged 65 to 74 in 1990, and 17 percent of income for households with heads age 75 or older. These are increases of 3 to 4 percentage points over the percentages from the same data source in 1980. Hitschler reports that much of this increase is due to the rise in the proportion of households receiving some pension income. Among the younger elderly households (heads aged 65 to 74), the proportion with pension income increased from 35 to 47 percent between 1980 and 1990, while the proportion of those 75 or over with pension income increased from 29 to 40 percent. O endnote refe` hp x(#$%` hp x(#$%  ?<#<endnote refe# Coverage and participation among current workers: There are several sources of data on pension coverage (Beller and Lawrence 1992). The most commonly used is the Current Population Survey, which often includes special supplements on pension issues. Another is the Form 5500 report sent annually by all private pension plans to the Internal Revenue Service. A third is the IRS Annual Tax File, a sample of those who have filed tax returns in a given year (Schieber 1995). Estimates of pension coverage rates depend on the data source used, the definition of coverage adopted (usually, currently participating in a plan) and the population being  ?<described.pQS ?<  #Xx6X@DQ/X@#э` hp x(#$%` hp x Concerning coverage, the "Form 5500 active participant count includes nonvested participants who have terminated employment but have not yet incurred a break in service generally one year under the plan." (Beller and Lawrence 1992: 61) These "participants" would not be counted as covered in the CPS data. p Populations of interest include all workers, all wage and salary workers (excluding the selfemployed), all private wage and salary workers (excluding government employees), all fulltime private wage and salary workers (excluding parttime workers) and the ERISA work force (which includes all those aged 21 or older who have worked for an employer for at least a year, and who work at least 1,000 hours annually). endnote refe` hp x(#$%` hp x(#$% #Dendnote refe#endnote refe The proportion of workers participating in a pension plan increased rapidly between the end of World War II and the mid1970s, but has remained relatively stable since then (Andrews 1985; Bloom and Freeman 1992; Turner and Beller 1992; Silverman and Yakoboski 1994). Analysis of Current Population Surveys between 1972 and 1993 indicates that pension participation among fulltime private wage and salary workers ranged between 48 and 50 percent during this period (U.S. Department of Labor 1994b). Participation among fulltime male workers has declined slightly from 54 to 51, percent while coverage of fulltime females increased from 38 to 48 percent. #QEendnote refe# Yakoboski et al. (1994: Tables 1 and 2), using 1993 CPS data, report participation rates of 44 percent for all civilian "h)-,**1"   workers and 47 percent for civilian, nonagricultural, wage and  ?<salary workers.k5 ?<#Xx6X@DQ/X@#  Ѝ` hp x(#$%` hp x All these figures are for workers aged 16 or older. k Among the latter, the participation rate is 43 percent for those working between 1500 and 199endnote refe9 hours per year, and 58 percent for those working 2000 or more hours. Of the ERISA work force (see definition above), 56 percent participated in an employersponsored plan in 1993. Coverage is higher among men (50 percent of the nonagricultural wage and salary work force) than among women (44 percent), and higher among those covered by a union contract (79 percent) than others (40 percent). Participation increases dramatically with education (Reno 1993: Table 2.9) and with annual earnings, from only 8 percent of those earning less than $10,000 per year, to 61 percent of those earning between $20,000 and $25,000, and 80 percent of those earning $50,000 or more annually. #,Kendnote refe# endnote refe Participation in pension plans also rises with age and with tenure as workers meet eligibility conditions and advance into better jobs where employers are more likely to offer pensions. Among civilian, nonagricultural wage and salary workers with less than one year of tenure in 1993, only 11 percent participated in an employer pension (Yakoboski et al. 1994: Table 2). This increases to 35 percent for those with 1 to 4 years of tenure, 60 percent for those with 5 to 9 years, and reaches 80 percent for those with 15 or more years of service. In 1993, the participation rate for wage and salary workers between the ages of 21 and 30 was only 34 percent, compared to 53 percent for those aged 31 to 40 and over 60 percent for those aged 41 to 60 (ibid.). The fact that pension participation increases with age suggests that lifetime pension participation rates will exceed the pointintime coverage rates shown in any crosssectional data, since the latter includes many younger workers currently uncovered who are likely to enjoy pension coverage as they  ?<progress in their careers.F5 ?<#Xx6X@DQ/X@#  Ѝ/$WP Defaults` hp x(#$%` hp x(#$#x6X@C/X@# Comparing the pension participation rate for persons in the 1979 Current Population Survey to the participation rate for the same birth cohort a decade later confirms this implication. In 1979, 31 percent of fulltime workers aged 15 to 23 were participating in a pension. By 1988 when they were 24 to 32, the participation rate for this cohort had increased to 48 percent (Clark and Wolper 1993)."VWP Defaults" F #5Nendnote refe#endnote refe` hp x(#$%` hp x(#$% #5Vendnote refe# Pension coverage is also related to firm size. Only 10 percent of workers in firms with fewer than 10 employees participated in a pension in 1993, compared with 50 percent for persons in firms with 100 to 249 employees, and 67 percent for those with more than 1,000 employees (ibid.). Pension coverage is much higher in the government sector. In 1988, for example, while 48 percent of the fulltime private sector employees were covered by pensions on their current jobs, 87 percent of the fulltime federal government workers were covered, as were 86 percent of the fulltime state and 88 percent of fulltime local government employees (Turner and Beller 1992: Table B2). Yakoboski et al. (1994) estimate that 79 percent of all federal workers participated in an employersponsored pension plan in 1993, as did 74 percent of all state and local workers. endnote refe #Zendnote refe#endnote refe In 1993, 63 percent of all private full!time wage and salary workers were employed by firms offering a pension and 50 percent"'.0*((@@/" were actually participating in a pension (U.S. Department of Labor 1994b). This implies that 80 percent of these workers employed by firms with pension plans actually participated in these plans. Among those offered a pension but not participating, 39 percent did not participate because they had not yet met the service conditions while another 31 percent chose not to contribute to the plan. Other reasons for lack of participation included 6 percent who were in jobs not covered by the employer!sponsored plan and 7 percent who did not work enough hours, weeks or months per year to qualify for participation. #-[endnote refe#endnote refe #^endnote refe#` hp x(#$%` hp x In a recent paper, Schieber (1995) argues that the CPS data on which so much the pension information is based significantly understate the receipt of employersponsored retirement income. There are several possible reasons for this. One is that persons receiving lumpsum pension distributions may invest these funds in financial instruments, and then report the income in subsequent years as interest or dividends rather than as pension  ?<income.z ?<#Xx6X@DQ/X@#  Ѝ Reno (1993) notes that income from assets has had the largest increase in the share of aggregate income of the aged over the past 15 years, and wonders whether this may be due to the increased importance of definedcontribution plans, whose distributions are typically lumpsum. z Alternatively, recipients of lumpsum pension payments could use the funds to pay off consumer debt or a mortgage, and report no "income" in later years at all. Another possibility is that some people who receive pension annuity checks from a thirdparty payers (such as insurance companies) do not report them as pension income on surveys. Schieber uses the IRS Annual Tax File to estimate the receipt of pension, annuity and IRA income for elderly (over 65)  ?<federal income tax filers.X ?<  Ѝ This data set has problems of its own. It contains almost no information about the characteristics of the tax filer (fortunately, it does identify those units with a filer over age 65), and underrepresents the low income households, who often do not file federal income tax returns. X His research is most useful at the middle and upper parts of the income distribution, where most households would be expected to file federal income tax forms. Schieber finds that slightly over threequarters of the filing units in the upper two IRS quintiles reported pension income, compared to an estimate of twothirds from the CPS as reported by  ?<Grad (1992). ?<#Xx6X@DQ/X@#  Ѝ See Schieber (1995) for a discussion of differences between the household unit as defined by Grad and the IRS filing unit, and differences in the definitions of the unit's income.  In addition, Schieber finds that the amounts of pension income reported in the CPS is smaller than in the IRS files, and claims that "the CPS fails to measure as much as onethird of the total income that is being paid out in the form of pensions and annuities directly and also fails to attribute other income for the elderly that comes from the employer based retirement system to the plans responsible for that income (p. 29)." Although the differences in samples and income definitions make the sources of these discrepancies impossible to identify, the IRS data suggest, as do the National Income and Product Accounts data, that the CPS estimates understate the importance of the pension income received by older Americans. endnote refe` hp x` hp x(#$%  ?#<#lendnote refe# Types of pension plans: Employerprovided pension plans are typically divided into two types, defined benefit and defined contribution, although recent changes in the structure of some pension plans have created hybrid plans that blur this traditional distinction. Definedbenefit plans promise retirement benefits based on a predetermined formula, while"'/0*((@@/" definedcontribution plans provide for a specified contribution each pay period. These plans differ with regard to the types of financial risks borne by the employer and the employee, the effect of mobility on the value of the pension, the effect of changes in pension rules on the value of benefits, responsibility for the investment of pension funds, their influence on worker behavior and the effect of government regulation on the administrative cost of the plans. One of the most important trends in employer pensions is the movement towards greater use of definedcontribution plans as the primary employersponsored retirement plan (Clark and McDermed 1990; Turner and Beller 1992; U.S. Department of Labor 1994b). The proportion of primary pension plans with more than 100 participants that are definedbenefit has declined in all industries and in all size groups (Clark and McDermed 1990;  ? <Clark et al. 1994).  The proportion of all plans that were definedbenefit fell from 28 percent in 1984 to only 14 percent in 1991 (U.S. Department of Labor 1994a). Non!401(k) definedcontribution plans have remained at approximately 70 percent of all plans while 401(k) plans have increased from 3 percent of all  ?h<plans in 1984 to 16 percent in 1991.z5 ?<  Ѝ` hp x(#$%` hp x Under 401(k) plans, the employee determines whether or not to participate and the level of contributions. z The increasing importance of 401(k) plans is even more dramatic when the number of participants is examined. As a percentage of all pension participants, those in 401(k) plans have increased from 12 percent in 1984 to 31 percent in 1991, while definedbenefit participants have declined from 50 percent of the total to 42 percent. Participants in non!401(k) definedcontribution plans declined from 38 to 27 percent during the same period. The dramatic growth in 401(k) plans is further shown in the actual number of plans and participants. In 1984, about 17,000 such plans covered 7.5 million workers. By 1991, almost 115,000 401(k) plans covered nearly 20 million participants. endnote refe` hp x(#$%` hp x(#$% #5xendnote refe# Participation rates in 401(k) plans are relatively high. In 1993, more than onethird of all private sector workers were offered the opportunity to participate in a 401(k) plan, and of those offered a plan, two!thirds choose to participate. The proportion of those offered who participate increases sharply with age, rising from 34 percent of those under 25 to 57 percent of those aged 25 to 29, 69 percent among those aged 30 to 34, and between 70 and 79 percent of workers between the ages of 35 and 64. This rate also rises with years of service and annual earnings. The median percent of pay contributed to 401(k) plans is 6 percent for both men and women, and does not vary much by age or annual earnings. Despite the relatively high participation rate for a voluntary plan, there are many workers covered by these plans who choose not to participate the other third. If 401(k) plans are the only retirement plans offered by an employer, their optional nature leads to substantially lower participation than if all eligible workers are required to participate, as is the case in many definedbenefit plans."'00*((@@/"Ԍ  ?<endnote refe Cost of employer pensions : Economists generally agree that, over the long run, workers bear most of the cost of employer  ?X<pension contributions through lower wage earnings.5 ?<  Ѝ` hp x(#$%` hp x This generalization is not meant to apply on a worker by worker basis in the short run. Rather, it is a longrun, equilibrium contention, that wages and employee benefits are both components of compensation, and that more of one, other things held constant, will mean less of another. Hamermesh (1993: 166173) reviews the empirical literature on the incidence of payroll taxes, and concludes that "we must tentatively infer that most of the burden of payroll taxes is on wages."  The tradeoff between wages and employee benefits has long been recognized in collective bargaining, compensation decisions made by business managers, and the selection of jobs by workers (Rosen 1974; Smith 1979; Brown 1980; Woodbury 1983). Research suggests that workers accept a portion of their total compensation in the form of future pension benefits when the effective price of these benefits if bought through the employer is less than what it  ?<would cost the worker. Greater value from employerprovided pensions is attributable to quantity discounts, risk pooling, and the preferential tax treatment of both pension contributions and the investment returns. Factors that increase the cost of employer pensions (like costly government regulations) will tend to reduce the demand for this benefit. #~endnote refe#endnote refe` hp x(#$%` hp x(#$%  ?<#nendnote refe#endnote refe Characteristics of definedbenefit and definedcontribution  ?<plans : In a definedbenefit pension plan, the plan sponsor promises to pay a retirement benefit based on a specified formula. The size of the benefit depends on eligibility and participation requirements, vesting standards, the benefit formula, and often the employee's recent earnings history (see below). In definedcontribution plans, employers and employees make periodic contributions into a pension account for individual workers. #endnote refe#endnote refe #"endnote refe# In 1991, about 70 percent of those in definedbenefit plans offered by medium and large firms had to meet some age and/or service requirement in order to become eligible to participate in  ?8<the plan. 5 ?<  Ѝ/$WP Defaults` hp x(#$%` hp x(#$#x6X@C/X@# The pension data that follow all apply to 1991 and are taken from the U.S. Bureau of Labor Statistics, Employee Benefit Survey of Medium and Large Firms (1993). Trends in these plan characteristics are discussed by Mitchell (1992). Benefit plans in small firms are described in U.S. Bureau of Labor Statistics (1994a)."/WP Defaults"  The most frequently used eligibility standard, covering about a third of participants in definedbenefit plans, was age 21 with one year of service, the maximum permitted by federal regulations. The current participation requirements are reasonable guidelines that do not significantly affect the retirement income of those who do not achieve coverage while limiting the record keeping associated with very short term employees. Most definedbenefit plans require 5 years of service before workers become vested in the plan (that is, before they have legal rights to benefits even if they leave the firm). This period has been reduced by government regulations, which currently require that firms provide vesting standards at least as generous as 100 percent after 5 years or a graded vesting schedule with partial vesting after 3 years and 100 percent after 7 years. Because, under current requirements, all fulltime covered workers will ultimately receive a pension even if they remain with the firm for only 5 years, further reductions in vesting standards would have only a limited effect on the size of retirement incomes."'10*((@@/"Ԍendnote refe #ёendnote refe#endnote refe The normal retirement age of a plan is the age at which a retiree can begin receiving unreduced retirement benefits or benefits based on the prescribed benefit formula in the plan. In most definedbenefit plans, this occurs at a particular age (most commonly, age 65), or when some combination of age and service with the firm (e.g., age 62 and 10 years of service, or age plus years of service equals 85) is achieved (Mitchell 1992: Table 9.4) About 8 percent of participants are in plans that allow retirement with unreduced benefits after a specified number of years of service (usually 30), regardless of age (ibid.). #0endnote refe# In virtually all plans, workers face early retirement options and can begin receiving benefits prior to the normal retirement age. The most prevalent early retirement age is 55 with 10 years of service. These early retirement features typically provide strong incentives for workers to retire from their career employer prior to the normal retirement age, often at the earliest age of eligibility (Kotlikoff and Wise 1989). With the continued aging of the population, the age for full social security benefits (100 percent of the worker's Primary Insurance Amount) is scheduled to be raised, first to age 66 and later to 67. The age at which maximum benefits can be received from a definedbenefit plan is often tied to the age for full social security benefits. If retaining older workers in the labor force is important, consideration might be given to establishing a minimum retirement age for taxqualified employer pension plans. endnote refe #endnote refe#endnote refe In nearly all definedbenefit plans, the benefit depends on years of service. In addition, for more than twothirds of all participants in definedbenefit plans sponsored by medium and large firms in 1991, benefits also depended on earnings. Among these workers, 80 percent were in finalearnings plans (most commonly, the average of the highest consecutive 35 years) and 20 percent were in plans using career average earnings formulas. Participants typically receive a specified percentage of average  ?<earnings (however defined) for each year of service. Another  ?<quarter of all participants in definedbenefit plans (mostly covered by collectively bargained contracts) were in plans that pay a specified dollar amount per year of service, independent of earnings. Other types of formulas were used in plans that cover about 7 percent of definedbenefit participants. #endnote refe#endnote refe #endnote refe# In final earnings plans, for workers who stay with the firm until retirement, the value of initial retirement benefits is protected against changes in the cost of living as long as earnings growth matches or exceeds the rate of inflation. Where retirement benefits are specified in absolute dollar amounts, more common in collective bargaining environments, the real value of initial benefits will decline unless it is periodically"'20*((@@." increased to reflect changes in the price level. These adjustments are often a topic of negotiation. Participants in definedbenefit plans who leave their career employers prior to retirement age suffer pension losses relative to long tenured workers. This loss in the lifetime value of pension benefits results from the use of final earnings in the benefit formula and the lack of indexation of vested benefits to future inflation or wage growth. This loss in the value of a pension discourages job turnover among persons covered by definedbenefit plans (Allen et al. 1993; see Gustman and Steinmeier 1993 for a contrasting view). Often, early retirement benefits are not fully actuarially reduced relative to normal retirement benefits. In this case, the present value of expected benefits is highest at the early retirement age, and then declines. The magnitude of the subsequent decline depends, in part, on the response of future wages and pension benefits to inflation. The decline in pension "compensation" along with the early availability of pension benefits tends to encourage retirement. The discontinuities in pension compensation as workers age (for example, the increases in pension accrual during the years just prior to eligibility for early retirement benefits and the sharp declines in accruals (and therefore compensation) thereafter) change the financial reward for working each additional year and thereby influence labor supply decisions (Burkhauser 1979; Fields and Mitchell 1984a; Kotlikoff and Wise 1989; Quinn et al. 1990; Ruhm forthcoming). endnote refe #zendnote refe# In 1991, 54 percent of all participants in private definedbenefit plans were integrated with social security in the sense that the employer pension provides a smaller proportional benefit to the lower paid, whose social security benefits will represent a greater proportion of their preretirement earnings (U.S. Bureau of Labor Statistics 1993). The most prevalent form of integration (covering about twothirds of those in integrated plans) was the excess method, in which a less generous benefit formula applies to earnings below some specified dollar amount. The remainder were in plans that use the offset method of integration, which reduces pension benefits by some proportion of social security benefits received. The federal government has specified maximum reductions permitted under both types of integration. Definedcontribution plans include savings and thrift plans, profit sharing plans, employee stock ownership plans, money purchase plans, and 401(k) and 403(b) plans. The benefits at retirement depend on past contributions and the rate of return on accumulated pension funds. Many plans allow for individuallydirected accounts so that participants can decide how pension funds are invested. Definedcontribution plans tend to have less restrictive vesting standards: 31 percent of definedcontribution"'30*((@@/" participants are in plans that allow for immediate vesting (this is required for 401(k) plans and common in 403(b) plans). Virtually all plans allow for lump!sum distributions at retirement or at termination of employment, which employees can (but often do not) then roll over into other plans or into individual retirement accounts. Pension compensation in definedcontribution plans is more explicit than it is in definedbenefit plans, because it is simply the employer's contribution to the pension fund. Therefore, these employer pension costs are more visible to workers than are the employer costs associated with definedbenefit plans. Traditionally, employer contributions are a percent of annual earnings and are unaffected by age and years of service, although basing contributions on age or service is becoming more common. Funds are generally deposited in individual accounts whose value is disclosed to the worker, and they are viewed as belonging to the worker. A worker who leaves the firm prior to retirement retains ownership of the entire vested value of the pension fund. Pension assets can either remain with the plan sponsor or be distributed to the departing worker as a lumpsum that can then be rolled over. In either case, the worker continues to receive all future returns on the pension assets. Because of these factors, individuals with definedcontribution plans tend not to suffer losses in pension wealth with job changes that those with finalpay definedbenefit plans do, and mobility rates are therefore less likely to be affected. Annual benefits are determined by the size of the pension account and the age of the recipient when benefits commence. To hold the asset value of the expected benefit stream constant, definedcontribution plans have implicit actuarial adjustments for different retirement ages. (Of course, the size of the pension account continues to grow if the worker remains on the job and new contributions are made.) Thus, definedcontribution plans are less likely than definedbenefit plans to influence retirement decisions. One problem associated with any plan with a lumpsum payout is that the lumpsum must be turned into an annuity if it is to provide a guaranteed (nominal) income flow over the recipient's remaining lifetime. Annuities indexed to the Consumer Price Index are unavailable (other than social security), and the ability to buy even fairly priced nominal annuities in the private market is a concern given the problems of adverse selection and group size. In addition, life expectancies vary by gender, ethnicity and other personal characteristics. Therefore, a key factor in the conversion of a pension accumulation to an annuity is the size and composition of the pool of persons with whom one is grouped. endnote refe "'40*((@@/"Ԍ ?<#Mendnote refe# Risks associated with pension plans : Definedbenefit plans differ from definedcontribution plans in who bears various types of risk. Some of the risks are specific to the individual worker, and some are broader, applying to the firm, the industry or the economy as a whole. For example, workers with earningsdependent pensions face the risk of uncertain future wages, especially those whose benefits will depend on their average wages during their last few years of work. Those whose benefits will depend on wages averaged over their careers face the risks of inflation during the work life, which can seriously erode the real value of early years' earnings. The risk to pensions from job turnover is borne by the worker in a definedbenefit plan, because turnover tends to reduce lifetime retirement benefits. This loss in pension wealth occurs even if the worker finds immediate employment at identical earnings with another firm with an identical pension. The loss occurs because benefits are based on earnings up to or at the time of separation, and are fixed in nominal terms at that time. Thus, a worker who switches employers will have lower total pension benefits than one who remains with a single firm, even if they have the same earnings profile. Both the number and the timing of moves and the rate of inflation thereafter will influence the size of the pension loss (Clark and McDermed 1988; Gustman et al. 1994). The problem is the lack of portability of pension credits across employers, and it is a significant risk factor for the participant in definedbenefit plans.  ?< ` ` Workers in smaller firms may have less opportunity for job advancement and therefore a higher probability of voluntary job separation. If so, such employees are likely to prefer definedcontribution plans. Small firms also are much more likely to go out of business than large firms, which would lead to the same preference. The risk of employer bankruptcy or the lack of adequate pension funding should also be considered by participants in a definedbenefit plan. Termination of a fully funded pension plan (whether because of employer bankruptcy or other economic conditions) has an effect similar to any other form of employment separation; lifetime pension benefits are reduced because workers are unable to complete their working careers at higher levels of earnings in the same pension plan. If workers correctly perceive termination risks, firms with a higher probability of bankruptcy or likelihood of canceling the pension plan will have to offer higher cash earnings to attract comparable workers. Inadequately funded pensions that are terminated at bankruptcy tend to exacerbate the losses of pension wealth. To limit such losses, the Employee Retirement Income Security Act (ERISA) requires that definedbenefit plans purchase pension insurance from the Pension Benefit Guaranty Corporation (PBGC). "'50*((@@/" If the assets in the pension fund are inadequate to pay the vested benefits, the PBGC guarantees the payment of such benefits within specified limits. Pension insurance reduces the risk that workers will not receive vested benefits in the event of plan termination. Since the PBGC now bears the risk associated with underfunding, the government has an increased interest in the full funding of pension funds and in the setting of PBGC premiums so that they provide adequate revenues to cover defaults. Determining appropriate premiums that allow for risk differentials has become an important component of pension policy. These premiums are costs to the plan sponsors and therefore raise the cost of providing pension coverage to workers. At the same time, they increase the value of the pension promises to the covered workers. Another risk faced by definedbenefit participants is the possibility that the firm might reduce the generosity of the plan for future years of service. This outcome would also impose a loss in expected pension benefits on workers in final earnings plans. endnote refe # endnote refe# Financial market risks are borne directly by the participants in definedcontribution plans and only very  ?<indirectly by participants in definedbenefit plans.!5 ?<#Xx6X@DQ/X@#  Ѝ/$WP Defaults` hp x(#$%` hp x(#$#x6X@C/X@# In a definedbenefit plan, it is the firm's responsibility to provide adequate funding for the promised benefits, so the firm bears the rate of return risk. But in the long run, the cost of retirement benefits is a component of labor costs, and higher pension contributions are likely to come at the expense of lower wages. Therefore, a decrease in the longrun rate of return to pension assets (like any other form of financial difficulty by the firm) may well be borne by the workers via lower wages or pension plan amendments that lower the benefit formula."VWP Defaults"  In the former, differences in rates of return to pension assets directly affect the size of an individual's pension account at retirement. Portfolio allocation decisions that yield lower returns lead to lower lifetime benefits. If workers systematically select portfolios that promise less risk but lower returns than do plan managers in definedbenefit plans, then participants in definedcontribution plans will have lower average benefits in retirement  ?8<than those in equally costly definedbenefit plans.o"5 ?<  Ѝ Some analysts (Merton 1969; Bodie 1991; Samuelson 1994) have argued that pension plans should hold portfolios dominated by bonds despite their lower expected returns. This is a hedging argument, that the pension assets should match their liabilities, which are nominal. One counterargument is that these liabilities are only the legal liabilities at a point in time, and that the pension should anticipate the change in liabilities due to future changes in wages. These changes will be affected by inflation, against which nominal assets are not a good hedge. o Greater risk aversion by participants would likely be reflected in individuallydirected pension accounts being more heavily weighted towards bonds rather than equities. As longrun investments, bonds and fixedreturn instruments have consistently yielded lower returns than equities (Ibbotson Associates 1993). The participant must also be concerned about the effect of shortrun fluctuations. Declines in asset values may significantly reduce the value of the pension fund just when the worker had planned to retire. The risk of significant declines in asset values near the planned retirement date can be partially offset by moving funds into less risky assets as the expected retirement age approaches. Inflation risk confronts participants in both types of plans. Those in definedcontribution plans must be concerned with the responsiveness of the rate of return of their pension funds to changes in the rate of inflation while they are still working. If contribution rates are constant, real employer contributions will not decline if earnings keep up with the rate of inflation. After retirement, the effect of inflation on the"'60*((@@/" real retirement benefit will depend on whether the pension funds have been converted into a fixed nominal annuity, an annuity indexed to market returns or whether the assets continue to be actively managed with inflationary increases in nominal returns accruing to the retiree. The issuance of government bonds indexed to inflation would provide considerable additional protection against postretirement inflation risk (see Section IVF below). Participants in earningsrelated definedbenefit plans must consider the impact of inflation on the rate of growth of their earnings. For workers in finalaveragepay plans whose earnings rise with inflation, future real pension benefits are not much affected by moderate inflation as long as they remain with their employers. But if earnings lag behind prices, then the real value of future retirement benefits will decline. For workers who have left their employers and are currently in deferred vested status, inflation systematically erodes the real value of future pension benefits that are frozen in nominal terms at the time the worker left the firm. Even moderate rates of inflation cause major reductions in the real value of deferred pension benefits earned early in the work life. This inflation effect substantially increases the detrimental effect of job changes for workers covered by definedbenefit plans. For participants in dollar amount plans, the inflation risk to benefits at retirement is much more direct. Benefits are specified in fixed dollar terms. As a result, if regular plan amendments do not increase the dollar amounts, the real value of these promised benefits will steadily erode. These plans are nearly all collectively bargained, and the dollar amount is typically set as part of the negotiations. After retirement, the effect of inflation on real pension benefits depends on postretirement benefit adjustments. Many public employee plans provide automatic costofliving adjustments to retirees: for example, federal civil servant pension plans have annual automatic costofliving adjustments tied to increases in the consumer price index. Adjustments under the older federal employee plan are equal to the annual increase in consumer prices. Adjustments under the newer plan are less than the full increase in the price index, however, and increases are granted only to retirees aged 62 or older. In 1992, just over half of the participants in state and local plans were covered by automatic costofliving adjustment provisions, although about 60 percent of those so covered were in plans that limited the increase to 3 percent or less per year (U.S. Bureau of Labor Statistics 1994b). In contrast, only 5 percent of participants in private definedbenefit pension plans had automatic costofliving adjustments in 1991. However, many private plans have provided ad hoc increases after retirement; of those raising benefits in"'70*((@@/" this manner, the increase has been on the order of 40 percent of inflation during the 1970s but less since then (Allen et al. 1986, 1992). Increases in pension benefits for retirees have been more prevalent in large, collectively bargained plans and have been larger for individuals who had greater preretirement tenure and who have been retired for a long period. Although the size of the ad hoc postretirement increases has varied over time, they have generally fallen well short of the rate of inflation. For most workers with defined benefits, then, the real value of the benefits declines during retirement. Individuals differ in their risk preferences and in their understanding of the risks they face. As a result, some workers will prefer definedbenefit pensions while other will prefer definedcontribution plans. Neither of the plans is "better" for all workers. However, it is important to understand what factors have stimulated the movement away from definedbenefit and towards definedcontribution plans and to examine how these trends will affect the economic wellbeing of future retirees.  ?< Questions regarding the future role of pensions in the  ?h<United States. In evaluating the future of employer pensions, several important questions must be addressed. They are outlined below, along with the Panel's current understanding of how they might play out under alternative scenarios. 1. Why did the growth in pension coverage cease in the 1970s? Without further government action, how will pension coverage change in the future? Most researchers conclude that the spread of pensions prior to 1970 was largely driven by the preferential tax treatment given to pension contributions and earnings (Ippolito 1986; Woodbury and Huang 1991). Other factors included rising real income that increased the demand for deferred consumption and the  ?X<1949 Supreme Court decision in the Inland Steel case that made pensions a mandatory issue for collective bargaining. Evaluation of the current status of pension contributions and the decision to offer pensions reveals that employers are sensitive to the aftertax cost of providing pension benefits (Long and Scott 1982; Woodbury 1983; Sloan and Adamache 1986; Woodbury and Huang 1991). The end of pension expansion coincided with a period of stagnant real wages, new government regulations that raised the cost of providing pensions, changes in tax policy and industrial mix, declines in unionization and a changing composition of the labor force. Workers are likely to bear most if not all of the cost of employer pension contributions in the form of lower wages. Given the progressive benefit structure of social security and current marginal income tax rates, it is plausible"'80*((@@." to argue that pensions now cover most workers who are willing to  ?<"buy" taxdeferred retirement income at current prices.d#5 ?<  Ѝ/$WP Defaults` hp x(#$%` hp x(#$#x6X@C/X@# Of course, employer objectives also determine whether the firm offers a pension plan. In most cases, the workers must select an employer and compensation package as a single choice rather than negotiating individually with the employer concerning the pension. Therefore, it is certainly possible that some individuals currently employed by firms that do not provide a pension would be willing to pay for an employer!sponsored pension in the form of reduced wages."/0WP Defaults" d endnote refe` hp x(#$%` hp x(#$% #@endnote refe# Significant growth in the proportion of the labor force participating in employer!sponsored pensions is unlikely to occur given current policies. Mandating that employers offer and workers participate in a plan would, of course, expand coverage. Mandating could be done in conjunction with the existing social security system or in response to a major modification of social security. 2. Why has there been movement away from definedbenefit plans and toward defined contribution plans? This trend is in part the result of government regulations that have increased the relative administrative cost of defined ? <benefit plans.$5 ?<  Ѝ Hay/Huggins (1990) estimate that for definedbenefit plans with 75 participants, ongoing administrative costs per participant increased 125 percent in real terms between 1981 and 1991, while similar costs for definedcontribution plans increased only 51 percent. Over this decade, the ratio of definedcontribution to definedbenefit per participant administrative costs dropped from 70 to 50 percent in small plans (15 or fewer participants), and from 110 to 74 percent in large plans (10,000 or more participants) (see Silverman 1993). Onetime costs associated with responding to frequent regulatory changes were also higher for definedbenefit plans. X1Í.X1Í. Continuing increases in the insurance premium that must be paid to the Pension Benefit Guaranty Corporation may also have contributed to the declining desirability of definedbenefit plans. In addition, new rules have encouraged firms to offer 401(k) plans. The decline in the proportion of pension participants covered by definedbenefit plans (as opposed to firms offering plans) has also been affected by structural changes in the economy (Clark and McDermed 1990; Gustman and Steinmeier 1992; Clark et al. 1994; Silverman and Yakoboski 1994). If Congress continues to enact changes that increase the relative cost of definedbenefit plans, the trend toward definedcontribution coverage will continue. These higher operating costs are also expected to reduce the proportion of the labor force employed in firms that offer pension plans. 3. To what extent will the growth in definedcontribution plans adversely affect retirement incomes in the future? Some analysts have expressed concerns about the shift towards definedcontribution plans. First, some inexperienced participants in selfdirected definedcontribution plans may follow investment strategies that are more conservative (with lower yields and therefore lower eventual retirement benefits) or more risky (with a high probability of loss of capital) than they would pursue if they knew more. Second, some plans require employee matching before the firm contributes to the plan, and some workers temporarily or permanently opt not to participate. Third, separated workers who have participated in definedcontribution plans may be more likely to accept and spend lumpsum distributions rather than roll them over into another  ?#<retirement plan.i%5 ?<#Xx6X@DQ/X@##Xx6X@DQ/X@#Ѝ Not enough is known about the allocation of lumpsum distributions. Yakoboski (1994) reports that about threefifths of all lumpsum total distributions were premature; that is, prior to age 591/2, and that these premature distributions accounted for about twofifths of all the lumpsum funds distributed. Of all these funds (not just the premature portion) about 57 percent was rolled over into Individual Retirement Accounts. Some of the rest may have been used for purposes that enhance retirement wellbeing (e.g., the purchase of an annuity by a retiree, or the purchase of a house by a worker), some for economic emergencies and some unknown portion for discretionary spending. xHerz (1995: table 3), using the 1993 CPS, estimates that onethird of lumpsum distributions were put into retirement savings, other savings, or investments. Another quarter was put into a business or a house, or used to reduce debts. The proportion rolled over into retirement or other savings instruments increased with the age of the recipient, as did the size of the lumpsum distribution. See also Yakoboski et al (1994), pp. 1517, who use the 1993 CPS employee benefit supplement to estimate the frequency, size and disposition of lumpsum distributions. i If these concerns are valid, the shift towards greater use of definedcontribution plans may result in fewer people receiving pension benefits and lower benefits for some who ultimately do receive benefits. "'90*((@@."Ԍ Several offsetting issues should be considered. First, the administrative costs of definedcontribution plans are usually lower than those of definedbenefit plans. These lower costs may result in more firms offering pension plans. Thus, without definedcontribution plans, firms deciding not to offer definedbenefit plans may opt for no plan at all. Second, more immediate vesting and greater portability of pension funds under definedcontribution plans should result in greater benefits in retirement, if the funds are preserved until then. And finally, education by plan sponsors can help employees make wise savings and investment decisions (Milne et al. 1995). 4. How will firms react to increased social security taxes or reduced social security benefits? If the social security tax and benefit structure is changed, there will be an automatic response in those pension plans that are integrated with social security. For example, in offset plans which reduce pensions by some proportion of social security benefits, pension benefits would rise by some fraction of any social security cuts. The exact changes in pension benefits will depend on the details of integration and whether the current integration rules are altered in response to the social security changes. Other reactions of employers to future social security changes are difficult to predict. If social security taxes are raised while benefits are maintained, other employee compensation will be reduced to the extent that workers bear the cost of this payroll tax. Whether this occurs through lower wages or reductions in pension compensation depends on worker preferences for current versus future compensation. If social security benefits are reduced in the future, workers may demand more employerprovided retirement income. It is uncertain, however, whether workers would be willing to give up sufficient current consumption to offset fully declines in social security benefits. Another interesting and important issue is how employers will respond to the changes in the social security environment already legislated. Both the delay scheduled for the age for full social security benefits (from 65 to 67) and the increases in the delayed retirement credit, already underway, will tend to induce later retirement. Will employers attempt to offset the reductions in social security retirement incentives by increasing analogous incentives in their own pension plans? Or will they go along with the social security changes by raising their normal retirement ages as well? How would employers who offer pensions respond to a delay in social security's early age of entitlement (62)? Should the government take a position on this issue by mandating minimum retirement ages for tax qualified pension plans? Answers to these questions require further consideration and debate."':0*((@@/"Ԍ 5. To what extent would the encouragement of portable pensions for workers enhance their retirement income? The preceding analysis has shown how workers who regularly change employers suffer reductions in their pension benefits relative to workers who do not. Many workers have only loose attachments to any particular firm. Workers with high turnover rates would have improved retirement income if they could participate in an occupational pension associated with their profession or occupation rather than with a specific employer or job. In this case, job changes would not require changes in their pension plans. One example of such a pension is TIAACREF, which allows many college and universities staff to move among employers without suffering breaks in coverage or loss in pension benefits. The development of similar pension plans for other occupations would allow workers to have contributions from different employers placed in a single pension account. Pensions are difficult to provide for lowwage and contingent workers because of their high mobility and multiple employers. Their low compensation and relatively low marginal tax rates makes it less likely that they would prefer future benefits to current compensation. A national pension plan might be established that permits irregular contributions, contributions from multiple sources and full portability. Employers of parttime, seasonal, temporary, or domestic workers, as well as fulltime lowwage workers, could then contribute regularly or irregularly to workers' pension accounts.  ?<x Retiree health benefits : The availability of employersponsored health insurance for retired workers is a growing concern for workers, retirees, employers and policy makers. Ever increasing health care costs and recent changes in accounting rules have caused many employers to reexamine their role in  ?<providing health benefits for current and future retirees.&5 ?<#Xx6X@DQ/X@##Xx6X@DQ/X@#Ѝ In December 1990, the Financial Accounting Standards Board approved Statement No. 106, requiring many employers to record a liability for retiree health benefits on their balance sheet in order to comply with generally accepted accounting standards, beginning with fiscal years after December 15, 1992.  The economic security of workers will be affected if fewer employers offer retiree health insurance or more employers require contributions from workers or retirees to fund it. In addition, the Medicare hospital insurance trust fund is forecast in the 1995 Trustees' Report to be exhausted by 2002 under intermediate actuarial economic assumptions (Board of Trustees 1995). The availability of retiree health insurance is likely to become an increasingly important income security issue as the baby boom generation moves into retirement. xIn 1992, 52 percent of the (mostly large) employers surveyed by A. Foster Higgins & Co. (1993) provided health care benefits  ?'<to retirees under age 65, down from 62 percent in 1988.'5 ?<#Xx6X@DQ/X@##Xx6X@DQ/X@#Ѝ These data are heavily weighted toward large employers, who are much more likely than small firms to offer coverage.  The two largest declines were in 1991 and 1992. In addition, the"';0*((@@/" percentage of these employers offering coverage who fully financed their retiree health insurance declined from 38 percent (1988) to 32 percent (1992). The percentage requiring full financing by the retirees fell slightly from 25 to 23 percent in 1992, while the percentage of employers sharing the cost with the retirees increased significantly, from 37 to 45 percent. xMany employers integrate their retiree health insurance packages with Medicare and offer supplemental insurance for  ?<retirees aged 65 or older. This practice is slightly less common than offering health benefits to retirees under age 65 and has also been falling in recent years. A. Foster Higgins survey data indicate that 46 percent of a sample of medium and large employers offered and contributed to health benefits to retirees aged 65 or older in 1992, down from 55 percent in 1988. While  ? <these data suggest a decline in the proportion of firms offering retiree health insurance both before and after age 65, U.S. Bureau of Labor Statistics (BLS) data indicate that the  ?<percentage of fulltime workers in medium and large private establishments offered retiree health insurance has been relatively stable at about 45 percent since 1988 (ibid.). The BLS surveys also show a decrease in the percentage of workers with retiree health benefits available only prior to age 65, and a small increase in the percentage of workers with benefits available throughout the retirement years. xLess is known about fulltime workers employed in small private establishments, but the BLS data indicate that about 18 percent of these workers had employersponsored coverage in 1992, up slightly from 1990 (U.S. Bureau of Labor Statistics 1994a). With respect to fulltime state and local government workers, the proportion with fully or partially employerfunded retiree health insurance coverage dropped from 58 to 51 percent between 1990 and 1992 (U.S. Bureau of Labor Statistics 1994b). Other workers may have been able to purchase health insurance at group rates after retirement because of Consolidated Omnibus Budget Reconciliation  ? <Act (COBRA) regulations.+(5 ?<#Xx6X@DQ/X@##Xx6X@DQ/X@#Ѝ The Consolidated Omnibus Budget Reconciliation Act of 1985 (COBRA) requires firms that employ 20 or more workers and that have health insurance plans to offer continued access to beneficiaries for up to 18 months (29 months if the beneficiary is disabled) if they lose coverage under the plan as a result of a qualifying event, such as retirement. The employee or dependent may be required to pay up to 102 percent of the premium. + ` hp x(#$%` (# xThe first wave of the Health and Retirement Survey (HRS) provides a crosssectional snapshot of retiree health coverage in 1991. Preliminary tabulations suggest that about 60 percent of wage and salary workers aged 5161 had retiree health insurance available, either through their current jobs or through their  ?!<spouses' employers.;) ?<#Xx6X@DQ/X@##Xx6X@DQ/X@#Ѝ The data do not reveal whether or not the insurance is funded by employer. Some respondents may just be reporting their ability to purchase insurance through their employers after they leave. ; Coverage was higher for workers in unions and for workers participating in a pension, and it increased with both firm size and the individual's earnings (Fronstin 1995). xThe overall estimates are slightly lower than those suggested by retirees in the 1988 Current Population Survey. Gustman and Steinmeier (1994) estimate that 72 percent of men and 65 percent of women who retired from jobs lasting 5 years or more had potential health insurance coverage from that job, and that"'<0*((@@/" 60 percent and 43 percent, respectively, actually took the coverage. xA restructuring of retiree benefits appears to be under way. According to a 1988 survey by Johns Hopkins University and the Health Insurance Association of America, over onethird of surveyed firms offering retiree health benefits expected to increase employee contributions (de Lissovoy, et al. 1990). These firms also intend to expand the use of managed care health plans to provide retiree health benefits, to limit health plan coverage and/or benefits, and to tighten eligibility requirements. xAs employers reduce their sponsorship of retiree health insurance, the economic security of retirees (especially early retirees, leaving employment prior to 65, the age of Medicare eligibility) will become less certain. This may induce changes in the labor force behavior of older workers. Recent research suggests that older workers contemplating retirement are sensitive to the availability of retiree health insurance. Karoly and Rogowski (1994), using the Survey of Income and Program Participation (SIPP), find that the availability of retiree health insurance doubles the probability of retiring for  ?<men aged 55 to 62.* ?<#Xx6X@DQ/X@##Xx6X@DQ/X@#Ѝ ` (#"For a 60 year old employed man, for example, the probability of retiring (leaving the labor force) over a 2 year period increased from 12 to 24 percent with the availability of retiree health benefits (Karoly and Rogowski 1994:121). This is liable to be an overestimate, however, because the authors were not able to control adequately for pension and social security wealth, which are likely to be correlated with the availability of retiree health insurance.  Gustman and Steinmeier (1994) find that it discourages retirement prior to the age of eligibility, and encourages it thereafter, with an overall effect that is very modest. Madrian (1994) finds a more substantial effect, and estimates that individuals with employerprovided postretirement health insurance retire substantially earlier on average, about a year earlier than those without. She also estimates that the increased availability of retiree health insurance between the early 1960s (when about 25 percent of retirees had it) and the 1980s (when close to half did) might explain between 10 and 20 percent of the decline in the labor force participation rates of men aged 5564. xThere is agreement that employer sponsored health insurance increased dramatically in the 1960s and 1970s (before it was very expensive to provide), that employers are currently cutting back on this benefit and/or requiring retirees to pay more of the cost and that employers are contemplating further reductions in generosity in the future. There is also consensus that older workers considering retirement are sensitive to the availability of health insurance after retirement, but considerable debate exists about the magnitude of the behavioral effects.  ?#<x Conclusions : Employer provided retirement benefits represent a major source of economic wellbeing for many retirees. They provide an important component of retirement income to many workers who have remained with the same company for most of their careers and smaller amounts to more mobile workers. Retirees from larger firms, unionized firms, and higher"'=0*((@@/" paying firms are more likely to receive pension benefits. Retiree health insurance is also more likely for these retirees. xThe proportion of workers covered by a pension plan or a retiree insurance health plan increased until the end of the 1970s. Pension coverage has remained relatively stable during the past 15 years, while retiree health coverage declined during the 1980s. These retirement benefits are critical to those who receive them, and the elimination or curtailment of these benefits would considerably reduce these individuals' retirement income. xIn the absence of major institutional change, like mandating employer coverage, current trends do not suggest any expansion of pension or health coverage on the horizon. Quite the opposite, current levels of pension and health coverage and benefits are threatened by government regulations and policies aimed at reducing the federal deficit. Because employer provided benefits are important components of the economic wellbeing of older persons, this forecast is a matter of concern, especially if combined with reductions in the level of support from government health and retirement programs.  ?< D. Trends in National Savings xThis section examines U.S. saving rates, and finds them to be low in aggregate compared to U.S. historical trends and to international standards. The microeconomic, disaggregate data are also discouraging: most Americans reach retirement age with insufficient savings to maintain preretirement consumption. Several explanations for this phenomenon are discussed, as are policy options to encourage growth in personal saving.  ?<x Why more saving is desirable: The current rate of saving in the United States is low relative to those of the past. The decline appears especially large when saving is measured on a net basis using national income and product account (NIPA) tabulations. But it is also seen to be lower with alternative measures, including gross NIPA saving, net saving adjusted for revaluations of assets, NIPA saving adjusted for consumer durable purchases, and NIPA saving adjusted for government capital investments. The drop in saving is more ambiguous if research and development spending and education and training investment are included. Part of the ambiguity arises, however, because it is difficult to allocate educational spending between "investment" and "consumption." xCompared with other rich countries, net saving in the United States is low alarmingly low according to some. This has implications for the future growth of income and consumption in the United States relative to other countries, and it may have"'>0*((@@." implications for the relative influence of the United States in world affairs. xOne reason that low saving rates are a concern is that the social return to saving is probably higher than the private return enjoyed by individual savers. Private savers receive an aftertax return, and the taxes on the returns to savings generate benefits for other members of society. Hence, when deciding to allocate lifetime incomes between current and future periods, private savers systematically undervalue the benefits of deferring consumption because they ignore an important part of the future benefits. xMore pertinent for social security analysis is the argument that the drop in saving has reduced the potential rate of growth in national income, which in turn reduces the potential consumption of future workers and retirees. Because future workers will be asked to finance social security and Medicare benefits for the current generation of workers, the redistribution of potential consumption from the future to the present raises questions about the equity of the intergenerational transfer. If real wages in the future turn out to be the same or only slightly higher than current wages, then the increased burden of paying for future social security and Medicare benefits may actually reduce the future net wage below the current aftertax wage. Future workers may thus face the prospect of lower wages (and consumption) than current workers, in part because the current generation of workers and retirees has saved so little. Future workers do not get a voice in current decisions, so it is important that current decision makers keep the interests of this group in mind. x If voters and policy makers believe the U.S. retirement income system faces serious peril, then one way to reduce the peril is to increase the size of the economic pie available in the next century. The retirement consumption of current workers must ultimately be derived from the output produced by future workers. The future output of the country must be divided between the consumption of future workers, future retirees, and investment. If the relative living standards of the retired elderly are to be protected, then the share of national income devoted to the consumption needs of the future elderly must rise, because a larger proportion of the population will be elderly in the next century. The most certain way to boost future income is to raise the future productive capacity of the nation, which can be accomplished through higher saving rates today.  ?$< xWhy do Americans save so little? Why has saving declined so much, especially since the early 1980s? Saving the correct amount for retirement requires a difficult set of calculations, and these computations are only performed, and their results seen, once in a lifetime. Some people make errors and save too"'?0*((@@/" little. Even if they make the correct calculation, many people find it hard to refrain from consuming today in order to ensure adequate income in the distant and uncertain future. In addition, many institutions have been established that diminish the need to save for costly emergencies or predictable events that will occur before retirement, reducing the need for nonretirement savings. Social security and Medicare, by insuring income and medical consumption in retirement, reduce the need for retirement saving. Medical and other kinds of insurance reduce the need to maintain large savings balances for unexpected emergencies. In addition, needbased programs, like public assistance and college aid, create disincentives to save. If two families receive similar incomes, needbased programs offer more generous treatment to the family with less accumulated savings. Finally, the improvement of credit institutions, like mortgage lending and consumer credit, has made it easier for many people to borrow. Americans have taken advantage of this and increased their borrowing. In some cases, the availability of credit has reduced the need to save.  ?< xIn addition, an increased fraction of retirement income has been annuitized; that is, converted into an asset that lasts until the death of the worker or spouse. Employersponsored pensions represent one form of annuitized asset. These pensions provide a higher percentage of retirement income than they did before 1975. Many consumers may therefore leave fewer assets to their surviving heirs in the form of unintended bequests. Saving may be lower than it would be if individual consumers had less  ?<access to annuitized retirement income.+ ?<#Xx6X@DQ/X@##Xx6X@DQ/X@#Ѝ The net impact of increased annuitization on saving depends also on the consumption behavior of the inheritors of the bequests. If they save more out of their own incomes because they expect to inherit less, then the impact of annuitization on savings is reduced. Unfortunately, the quantitative effect of anticipated and actual bequests on consumption is not known.  x Even if these institutional changes have improved consumer welfare, they have reduced the need for precautionary and retirement saving. How much they have reduced the rate of private saving is a matter of controversy among economists.  ?X< xPolicy Options for Increasing Savings: There are really two questions here: what policies can raise private saving, and can they also raise aggregate national saving? A policy that succeeds in raising private saving may fail to increase national saving if it simultaneously reduces public saving; that is, increases the government deficit. The Panel examines specific policies that have been suggested to raise private and/or public saving. Where necessary, the analysis will distinguish between the effects on saving in the short, medium and long run.  ?`"<  ?(#< xTo preview the Panel's findings, the literature suggests few  ?#<simple answers to these questions.n, ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#Ѝ This lack of agreement results partly from the scarcity of wellmeasured saving statistics. One issue is that high net worth individuals tend to be reluctant to cooperate with survey questionnaires regarding their assets and liabilities. A second complexity is that researchers ideally want longitudinal  ?<information about saving patterns for the same people and employers, to track how policy changes affect retirement accounts, housing, capital asset accumulation, and other financial holdings. In practice, such data are difficult to obtain. In the future, new longitudinal surveys such as the Health and Retirement Survey promise to fill in many of these gaps (Juster and Smith 1994). n Assessing the effects of policy on savings patterns is difficult because policies sometimes have effects on households' savings environments that are hard to identify. For example, measuring how tax changes affect household or pension savings requires information on savings responses to different tax regimes, holding other things"'@0*((@@/" constant. The ideal "experimental" data do not exist, of course, so empirical researchers rely on actual historical records to try to identify what a particular tax change accomplished while controlling for other factors that varied over time and across households or firms. Journal debates rage over whether any given set of "control variables" is sufficient to isolate the true effect of key policy variables, and whether observed policy effects actually reflect some other unmeasured but crucially important factor correlated with a policy tool (for example, unobserved health problems or saving proclivities). xA more serious source of professional disagreement is that researchers do not yet have a fully unified theory of why people save. Simple economic models of saving are inadequate representations of real world behavior, but more complex models are difficult to work with and often do not produce clearcut predictions about responses to policy changes. The lifecycle framework is the foundation for most economic savings models. In its simplest variant, it posits that people save out of their (perfectly forecast) earnings to ensure a smooth consumption stream between retirement and their (perfectly forecast) date of death. Not surprisingly, some of the predictions generated by this perfectcertainty model are contradicted by empirical studies of U.S. savers. For example, the simple theory suggests that people dissave as they near death; in this light, elderly people would be expected to draw down their assets as they age, and those who expect to live longest should save the most. In fact, many elderly have virtually no retirement assets (other than their expected social security benefits), and there is considerable debate about whether those with retirement assets draw them down as they age (Hurd 1990). Another ambiguity arises regarding the effect of the return on savings. In the simple model, higher anticipated earnings on retirement savings (or lower taxes on retirement savings) increase peoples' incentives to save, because the rewards from doing so are larger. Nevertheless, higher expected returns (or lower expected taxes) might encourage more current consumption or earlier retirement, because people are wealthier. Hence, even the simple lifecycle model fails to offer simple predictions about the determinants of retirement savings patterns. xImportant extensions of the lifecycle framework have sought to make behavioral models more realistic by recognizing additional savings motives, particularly due to uncertainty (Hubbard et al. 1994). Some studies emphasize the wide range of approaches households can adopt to cope with these risks, although their increased complexity sometimes makes it more difficult to predict how household savings patterns will respond to policy changes. For the wealthier population, for example, oldage savings can take on a strategic role if parents use their prospective bequests to exact a commitment from their adult children to provide for them in old age (Pauly 1990). If this"'A0*((@@/" proposition is widely true, savings patterns should respond not so much to changes in tax rates, but rather to changes in longterm care policy. Another model emphasizes the role of annuities in protecting against longevity risk; here, precautionary saving in old age is strongly affected by insurance market characteristics (Kotlikoff and Spivak 1981). A third class of studies suggests that poor people may not save for old age because they rely on "safety net" government programs such as Aid to Families with Dependent Children (AFDC), Medicaid, Supplemental Security Income (SSI), and Food Stamps, all of which penalize private savings (Hubbard et al. 1995). Consequently, an explanation for the lack of savings observed among the old may be less attributable to myopia than to rational responses to social insurance programs. xOn the whole, secondgeneration savings models acknowledge that many environmental and institutional factors shape savings patterns, in more complicated ways than first anticipated by the simple lifecycle studies. As a result, numerous empirical studies have been conducted examining what factors seem to influence savings patterns, particularly the specific types of assets believed to be vehicles for retirement savings. The consensus is that many Americans reach retirement age with relatively little savings, and what they do have is concentrated in the form of housing equity, anticipated pension benefits, including 401(k) plans, and anticipated social security benefits. A brief review of what is known about the factors that influence the accumulation of each of these three assets is useful, along with other factors the influence savings.  ?p< xDeterminants of saving in the form of housing: Housing is the single most important private asset held by older Americans. About threequarters of households with heads aged 4569 own their homes, and many have substantial housing equity (Hendershott 1994). In 1991, for example, the median value of home equity by elderly households aged 6569 was $50,000 (including those with 0 equity), over half of the median net worth ($96,600, excluding social security and pension wealth) of  ?<these households (Poterba, Venti and Wise, 1994: table 1).-- ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#э The Poterba, Venti and Wise (1994) results presented here and below are derived from the Survey of Income and Program Participation (SIPP). -  ?@< xGiven these statistics, it could be argued that increasing home equity might be a successful strategy to increase savings for retirement. From this vantage point, the question then becomes what policies might be pursued to increase housing wealth. Before turning to that matter, it is important to note the controversy and policy debate about whether housing can and should be considered an asset to sustain life in retirement. For instance, state Medicaid policies do not usually force elderly couples to sell their homes before becoming eligible for publicly provided longterm care coverage. Some analysts have also argued that housing assets should not be "counted" as retirement wealth because few people reduce their housing stock as they move into"'B0*((@@/" retirement (Venti and Wise 1990a). On the other hand, owning a house and living in it does provide housing services, which is inkind income. In addition, at advanced ages and especially after widowhood, people apparently do reduce their housing stock, implying greater housing equity might well improve retirement wellbeing (Hoynes and McFadden 1994). xIf housing equity is an appropriate vehicle for retirement savings, what then are the policy tools available to encourage it? One issue is the responsiveness of housing investment to interest rates. A recent review of the literature concludes that U.S. housing purchases are not very sensitive to changes in interest rates, and that the insensitivity is growing more pronounced with the advent of adjustable rate mortgages (Hendershott 1994). This finding implies that the ability to influence savings in housing by altering interest rates is limited, at least in the short run. xAlthough governments have adopted numerous other policy instruments to influence housing patterns, remarkably little information exists on their effectiveness. Issues that have been studied include the effect of tax preferences and the special capital gains exemption offered on one home sale after age 55. Because inflation has subsided and because fewer people itemize deductions now than previously, the tax subsidy effect has apparently declined. Skinner (1994) and Poterba (1994b) both conclude that the tax subsidy has had only modest effects on the nation's housing stock, increasing it only by 10 percent above what it would have been otherwise. Whether changes in bequest taxes and capital gains exemptions have affected real estate holdings has received less attention. xSeveral other government policies have been designed to influence housing investments including mortgage subsidies, subsidies to those facilitating secondary mortgage markets, and benefits offered to savings and loan institutions. In most cases, little evidence exists about how effective these policies are in influencing net new housing stock (Poterba 1994b). These subsidies are also less common now than a decade ago. Few of these policies have been evaluated in terms of their net effect on household saving, so it is still very much in question whether these subsidies increased private saving or merely reallocated people's saving toward housing. And even if they did increase private savings, the net effect on national saving depends also on eventual impact of the reduced tax revenues (the tax expenditure) on other government taxing and spending decisions, and therefore on eventual government saving. This is an even more difficult question. xEven if policy tools can be identified that influence net housing saving, the question remains as to whether increasing housing equity is necessarily good public policy. Housing prices"'C0*((@@/" are quite volatile in the United States, as demonstrated by the runup in housing prices during the 1970s followed by their poor performance in the 1980s. Because many home owners already have the majority of their total assets in housing, it may not be sensible public policy to encourage even more such investment, particularly were it to come at the expense of other more diversified assets. xThis problem of volatile housing values may become even worse in the future, according to several recent forecasts (Mankiw and Weil 1989; Schieber and Shoven 1994). These models relate timeseries data on the size of demographic cohorts to housing prices, and predict that when the baby boomer generation retires and sells its homes, housing values will fall precipitously. Not everyone agrees with these gloomy forecasts, because they depend on the particular time periods and empirical models used. Indeed Hendershott (1994) claims that these predictions of severe asset value decline are eliminated by adding controls for income and real posttax interest rates to the earlier models, and also by varying the endpoints of the data series. Nevertheless, because the prognosis for future housing values is uncertain at best, dramatic changes in national housing policy are probably not warranted on the grounds of increasing economic security in retirement.  ?< xDeterminants of saving in the form of pensions : Employer sponsored and individually held pensions are the second most important privately held asset available for retirement purposes. Nonetheless, the median level of retirement savings in pensions is quite low $16,000 (including those with 0 pension wealth) for households with heads aged 6569 in 1991, about onethird as large as median housing equity. Pension asset holding is very skewed, however. Average (mean) pension wealth was $62,300 in the same year, about the same as mean housing equity ($65,000) (Poterba, Venti and Wise, 1994; table 1). xOne question of keen interest to policy makers is whether it is possible to increase pension saving without having adverse effects on other components of compensation, employment, and the federal government budget deficit. The answer appears to be no. xBefore turning to that discussion, however, it is important to mention that specialists disagree about whether pension saving should be increased via public tax incentive. One reason is an equity one, because the tax expenditures go disproportionately to higherpaid workers, who are also more likely to receive  ?#<government guarantees on future benefit promises via the PBGC.  Another reason is that this strategy increases national savings by less than it does private savings, because of the increased tax expenditures. In addition, some argue that pension plan investors tend to earn belowmarket rates of return on assets (Lakonishok et al. 1992). Nevertheless, there is a common"'D0*((@@/" perception that less saving would occur were it not for the seemingly automatic nature of employer (and often employee) pension contributions (Thaler 1994). xOn the assumption that more pension savings is desirable, the question arises as to what policy actions would increase pension savings. Perhaps the most important policy instrument is the tax code pension contributions and investment earnings have received tax protection in the United States for at least 40 years, which enhances the appeal of pension saving relative to other forms of saving. The conditions under which this tax protection is offered require a pension plan to meet certain coverage and benefit requirements (known as nondiscrimination rules), as well as reporting requirements to participants and the government. xHow effective is the tax incentive "carrot" in inducing more pension saving? Has the "stick" of nondiscrimination requirements been effective in spreading pension participation to more (lower paid) workers? These questions have received much attention in recent years. One approach is to investigate benefit plan responses to changes in marginal income tax rates. A study that examined the effects of tax cuts under the 1986 Tax Reform Act concluded that employer pension contributions were quite sensitive to tax policy. The results imply that eliminating the tax preference for benefits would cut employer contributions to pensions by onehalf, with lowwage workers feeling the greatest reduction in pension savings (Woodbury and Huang 1991:139). More recent studies tend to confirm that pension savings are sensitive to tax policy, although the estimates are imprecise because of the difficulty of obtaining individuals' marginal income tax rates along with pension savings information. xData problems also plague assessments of the effect of pension regulations, such as nondiscrimination rules, on the willingness of companies to offer pensions as well as the amount of the contributions. One report concluded that about half of the nationwide decline in definedbenefit pension coverage was attributable to the cost and complexity of these nondiscrimination regulations (Clark and McDermed 1990), while other analysts find a smaller regulatory burden (Gustman and Steinmeier 1992; Chang 1993). Some studies suggest that one important deterrent to pension growth has been rising levels of premiums which definedbenefit plans must pay to the Pension Benefit Guaranty Corporation (Ippolito 1989). xWithout focusing on particular point estimates, it seems reasonable to conclude that requiring pension plans to pay insurance premiums and cover lowerpaid workers, and limiting the taxfavored pension savings of higherpaid workers, had a depressing effect on the number of pension plans in the past two"'E0*((@@/" decades. On the other hand, because of the nondiscrimination rules, lowerpaid workers are now more likely to be covered where plans exist. Whether these regulatory changes resulted in a net decrease in overall retirement savings or simply caused substitution of nonpension for pension savings is not known. x Because many of the studies of traditional pension plans have been inconclusive, researchers have turned to two other types of retirement savings Individual Retirement Accounts (IRAs) and 401(k) plans to examine the responsiveness of savings to incentives. Legislation permitting taxdeferred IRAs for workers without employer pensions was first passed in 1974. Eligibility was extended to all workers in 1981 and then restricted in 1986. For a 5year period, 1982 through 1986, most workers were eligible to deposit up to $2,000 of pretax income per year into a taxprotected account, with additional amounts available for spouses. When contribution and eligibility limits made IRA plans less popular in 1987, 401(k) plans began to grow. More than a quarter of the entire U.S. population aged 3554 participates in a 401(k) plan today, with a median account balance of about $6,000 (Engen, Gale and Scholz 1994). xAre these new types of plans really retirement savings vehicles? If so, are they merely substitutes for saving that would have occurred in other forms, or do they generate net new  ?<savings? Research studies are not yet conclusive.c. ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#Ѝ See Gravelle (1991) and Hubbard and Skinner (1995) for reviews of this literature. Studies by Feenberg and Skinner (1989), Venti and Wise (1990b, forthcoming), and Poterba, Venti and Wise (forthcoming) suggest that the existence of taxincentive saving plans has induced significant new saving, while studies by Gale and Scholz (1994) and Engen, Gale and Scholz (1994) suggest the opposite. Joines and Manigold (1995) find an intermediate result. c Demonstrating that people with 401(k) plans save more than similar people without them requires a good "control group," which is difficult to obtain. One approach is to observe that families with access to a 401(k) plan have more savings accumulated than do apparently similar families, with comparable incomes, but without access to a 401(k) (Poterba, Venti and Wise, forthcoming). Their findings suggest that encouraging 401(k) plans would increase private saving. A contrary view is that people with strong preferences for saving choose to work for firms where 401(k) plans are offered, while people who cannot or do not wish to save work in companies without such options (Engen et al. 1994). The latter perspective implies that making 401(k)type plans more appealing (for example, by raising savings limits) would have little impact on current nonsavers (Ippolito 1993). xThe current consensus is that 401(k) and IRA plans have had some positive effect on boosting retirement wealth (but how much remains a topic of active debate), and that they may increase savings more in the long run than in the short run, as people exhaust their opportunities for asset shuffling. The conservative assessment is that the advent of 401(k) plans mainly promoted a reallocation of funds from IRAs and definedbenefit plans, with at most a 1 percentage point net increase in national saving (Engen et al. 1994). This view is based in part on the high rate of withdrawals prior to retirement, and the fact that"'F0*((@@/" many who withdraw funds devote the money to current consumption (Atkins 1986; Andrews 1992). The prohibition of these withdrawals would probably increase eventual retirement savings for those who chose to participate in the plans, but fewer people would contribute because of the reduced accessibility of the money.  ?x< xSocial security and saving for retirement: It is widely agreed that social security policy is the single most potent federal method of influencing retiree wellbeing. In 1991, older households (aged 6569) anticipated a stream of social security benefits totalling almost $100,000 (the median value), which is higher than the median value of the rest of their (nonsocial security and nonpension) net worth (Poterba, Venti and Wise 1994). This benefit stream is fully indexed, providing retirees with inflation protection for their most important retirement  ? <income source.s/ ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#Ѝ We do not focus on medical insurance during retirement in this discussion, though it is estimated that a lowerbound value of future Medicare benefits would add at least another equal amount to retiree wealth. s xTo what extent do social security benefits offset other retirement saving? Would changes in social security, such as benefit cuts or delays in the ages of eligibility, induce offsets in private saving? These questions have been the focus of intense research for two decades; however, as Poterba (1994a:13) concludes, "the existing empirical literature in this area is weak and there is relatively limited prospect for improvement." Timeseries aggregate data do suggest a negative correlation between social security promises and private savings, but many alternative explanations for this finding are available. A positive correlation between social security benefits and savings appears in household crosssectional studies, but this estimate of the effect of social security on savings is not credible; rather, both social security benefits and accumulated savings are positive functions of lifetime earnings. This empirical finding does not predict how individuals would behave, all other things equal, if social security benefits were cut. xEvidence on social security's effects on private pensions is more conclusive, and not particularly encouraging. Early research by Munnell (1974) concluded that higher social security does reduce private pension saving, a finding strongly corroborated in a recent study by Gullason et al. (1993), who concluded that pension wealth is affected, but not other types of savings. However, the results suggest that the effects are small. xSome have argued that social security might not have an observable effect on savings if people reacted to more generous retirement benefits by retiring earlier, instead of by changing their saving behavior. Although this explanation is plausible, data suggest that retirement patterns are not very responsive to changes in benefits of the sort enacted over the past two decades (Fields and Mitchell 1984b; Quinn et al. 1990; Kreuger and"'G0*((@@/" Pischke 1992; Gustman, Anderson and Steinmeier 1995). Hence, moderate social security benefit changes within the existing eligibility structure would probably have only modest effects on labor supply decisions at the end of the worklife.  ?< xIn contrast, it is far from clear how aggregate savings and retirement patterns would be affected by radical changes in social security; for instance, if the age of eligibility for early social security benefits were delayed, or all or part of the social security system were replaced with a national definedcontribution system (Diamond 1993; Myers 1993). The effects of the latter would obviously depend on program characteristics (including whether the system were mandated, its size and administrative costs), whether the contributions were invested in government bonds or private holdings on an internationally diversified capital market, and how much the government protected those with low pensions with an oldage poverty program.  ?< xOther factors influencing savings : Several other factors that affect saving policy should be mentioned. One is how wellinformed consumers are about financial issues and the need to save for retirement. Recent surveys give U.S. consumers low scores. Many consumers apparently do not understand compound interest, making it difficult to estimate how much to save for retirement (Bernheim 1994; Kotlikoff and Auerbach 1994). Other studies have noted that older workers have errorladen estimates of their likely social security benefits (Bernheim 1993, 1994), and many do not understand their employerprovided pension plans (Mitchell 1988). Accurate retirement savings calculations also require an estimate of life expectancy. Hamermesh (1985) and Hurd and McGarry (1993) find that people's expectations about life expectancy and its determinants are roughly accurate, although Hamermesh finds that the distributions implied by survey responses have a wider variance than the true distributions. One way to boost savings may be to educate people about the magnitude  ?X<of their total consumption needs in retirement.H0 ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#Ѝ The U.S. Department of Labor is developing a public education campaign on pensions and retirement savings. For details, see Employee Benefit Research Institute 1995. H  ?< xConclusions: The Panel finds that most Americans hold three major forms of retirement wealth: houses, private pensions, and social security. People's holdings of net home equity are not strongly influenced by policy tools, and even if they were, there would not be a strong impact on aggregate national saving. The Panel also concludes that it is probably not possible to enhance net aggregate saving significantly by easing pension regulations or sweetening tax inducements. It appears that, because pensions and social security are substitutes to some extent, reducing social security benefits would probably encourage pension saving modestly.  ?H&< "H&H0*((@@-"  ?< HH E. Implications for Future Retiree Wellbeing xThis section discusses forecasts about future retiree wellbeing. The Panel asks how income and assets other than social security might be expected to change over the next several decades, and what recent trends in the health status of older Americans imply. The Panel concludes that future retirees will probably have higher real income in retirement than current retirees do, but probably not enough to maintain their preretirement consumption standards unless savings or retirement behavior changes significantly. xSix areas are discussed: (1) the economic status of today's elderly, and how that has changed over the past 35 years; (2) the retirement prospects of today's baby boomers; (3) the socioeconomic diversity of the elderly and the identification of groups at risk in retirement; (4) alternative views about the economy over the next several decades; (5) prospects for asset prices when the baby boomers retire, and (6) the health of older Americans. xThe literature reaches some consensus on the first three topics and the last. Current retirees are, on average, considerably better off financially than prior cohorts of retirees, but this improvement "on average" conceals significant pockets of economic distress. Future retirees, again on average, are likely to exceed their parents' current standard of living in retirement, as they have done at all points in the life cycle to date, but the pockets of economic distress are unlikely to disappear. Single Americans (especially single mothers), those with low levels of education or nonstable employment patterns, minorities, and nonhomeowners are most at risk in retirement. The major area of disagreement here concerns the appropriate standard to use in evaluating the baby boomers' retirement prospects. If they are likely to exceed their parents' real incomes in retirement, yet be unable to maintain their own preretirement standard of living, is this to be judged a success to be lauded, a problem to be addressed, or both? Finally, the life expectancy of older Americans continues to rise, and their health, which does not have to move handinhand with life expectancy, appears to be improving in the aggregate, although variation occurs across income classes and racial groups. xThere is considerably less consensus on the other two topics the future course of the economy over the next several decades (including the impact of demographic shifts on economic growth) and the impact of the retirement of the baby boomers on future asset prices.  ?'< xThe economic status of current retirees : There is general agreement by researchers on several key points about current and  ?(<recent cohorts of retirees 1 ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#Ѝ See Quinn (1987), Hurd (1990), Radner (1992) and Quinn and Smeeding (1993) for reviews of the economic status of the elderly. . "h)I-p+p+1"Ԍ HH xFirst, poverty rates among those aged 65 or older have fallen dramatically over the past 35 years, from 35 percent in 1959 to a low of 11.4 percent in 1989 and back to 13 percent by 1992 (U.S. Bureau of the Census 1993a: table 3). As seen in figure II8, most of the decline occurred by the mid 1970s. The elderly poverty rate had dropped to under 15 percent by 1974, just after legislated increases in real social security benefits on the order of 50 percent. It should be noted, however, that, while the poverty thresholds are increased annually for the cost of living, they are not adjusted for increases in the general standard of living: these official statistics measure absolute,  ?<not relative, poverty.62 ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#Ѝ A recent National Academy of Sciences volume edited by Citro and Michael (1995) discusses the issue of absolute versus relative poverty extensively. 6 xSecond, the official statistics ignore inkind benefits, which are substantial for older Americans. Were they included and the poverty thresholds left unchanged, official poverty for the elderly and for others would be lower. The Census Bureau estimates that the 1992 poverty rate for those aged 65 or older would have been 10.4 percent (rather than the official rate of 12.9 percent) were taxes subtracted and the value of Medicare, Medicaid and other meanstested government noncash transfers  ?h<included.3 ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#Ѝ These estimates depend critically on the value assigned to the noncash benefits (see Quinn, 1987). With respect to Medicare and Medicaid coverage, the Census Bureau has adopted a "fungible value" approach the benefits are counted as income to the extent they free up resources that could have been spent on medical care (i.e., resources available to the family in excess of basic food and housing needs). If the family's income is less than basic food and housing needs, the coverage is treated as though it has no income value. If family income exceeds these basic needs, the value of Medicare and Medicaid coverage is set equal to this excess, up to the market value of the insurance coverage (U.S. Bureau of the Census 1993b: viiiix).  When an imputed return on home equity is included, that elderly poverty rate drops to 6.2 percent (U.S. Bureau of the Census 1993b: table 2). xThird, social security and other federal programs are credited with much of the improvement in the economic wellbeing of the aged, which took place even as the labor force participation rates of older men were dropping significantly. Both coverage and the generosity of social security benefits increased dramatically over the past 35 years. Estimates of the incidence of poverty in the absence of these transfers (assuming, quite unrealistically, no changes in labor supply or savings behavior) suggest the tremendous importance of social security  ?<income among older Americans.4 ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#Ѝ This is an overestimate of the true effect, because it assumes that retirement patterns would be unaffected by the loss in social security benefits. In fact, as we argue in this report, retirement behavior is very much affected by the existence of the social security system.  The Census Bureau estimates that the elderly poverty rate would have been 50 percent in 1992 were government transfers subtracted from money income (U.S. Bureau of the Census 1993b: table 2). xFourth, the elderly are a very heterogeneous group. While the overall poverty rate for those aged 65 or older was about 13 percent in 1992, it was under 11 percent for those aged 6574, 15 percent for those 7584, and 20 percent for those aged 85 or older, the fastest growing age group in America (U.S. House of Representatives 1994: table A7). Nine percent of elderly men were poor, compared with almost 16 percent of elderly women. Only 6 percent of elderly in married couple families were poor, compared to 25 percent of those living alone (more than 80 percent of whom are women). Hispanic elderly are twice as likely to be poor as white elderly (22 versus 11 percent), and black elderly are three times as likely (33 vs. 11 percent). Poverty is the norm in some subgroups; for example, 51 percent of"'J0*((@@/" Hispanic elderly women living alone were poor in 1992, as were 57 percent of analogous black women (U.S. Bureau of the Census 1993a: table 3). xFifth, although actual poverty rates for the elderly are lower than for the entire population, the proportion of the elderly living near poverty is disproportionately high. In 1992, while 13 percent of the elderly were poor, 20 percent had incomes less than 1.25 times the poverty threshold, and nearly a quarter were below 1.5 times the threshold (U.S. Bureau of the Census 1993a: table 17). xIn summary, the economic status of the elderly has improved significantly over time, to a great extent because of large Federal cash and inkind transfers targeted at this group. Nonetheless, serious pockets of poverty continue to exist, primarily among elderly widows and minorities, and a disproportionate number of the elderly who are not poor are not far from the poverty threshold.  ?<  ?<xRetirement prospects of the baby boomers: The messages presented by the U.S. House of Representatives (1987), Easterlin et al. (1993), Sabelhaus and Manchester (1993), Manchester (1994) and Yakoboski and Silverman (1994) and even, to a great extent, Kotlikoff and Auerbach (1994) and Merrill Lynch (1994) are similar with regard to the baby boomers' expected levels of retirement income and the relation of these levels to those of their parents.  ?<xOverall income levels: Sabelhaus and Manchester (1993) analyze real income levels on a per household, per capita, per adult and per adult equivalent basis and show that in all four cases, in aggregate, by income quintile, age (2534 and 3544) and marital status, baby boomers are considerably better off than their parents were at the same ages. For example, the baby boomers' average household income was over $42,000 in 1989, 46 percent higher than their parents' was in 1960 (see table II1). The difference is higher at the upper end of the income scale (50 percent in the highest quintile) than at the lowest (25 percent), and even higher when measured on per capita, per adult or per equivalent adult basis. The authors conclude that "as long as real wages continue to grow and assuming that social security and private pensions remain intact and that health care expenditures do not swamp other gains, most baby boomers are likely to enjoy higher real incomes in retirement than their parents.'' It is worth noting that real wage growth and future health care expenditures are both causes of concern for many policy analysts. xEasterlin et al. (1993: 501, 503) reach similar conclusions: "The evidence clearly belies the notion that on average the baby boomers are doing less well than their parents. . . . Is the baby boomers' advantage over their parents likely to be maintained"'K0*((@@/" into retirement? The life cycle income profiles suggest that the answer is yes. . . .The improvement of the boomers over their parents at retirement, however, may prove to be less than it is currently. Although the data for younger ages are incomplete, the baby boomers' profiles appear to rise less steeply between ages 2529 and 4044 than their parents', and projecting this slower rate of increase to retirement would yield somewhat diminished relative improvement. But, the present comparison clearly does not support the view that the boomers will, on  ?<average, end up less well off than their parents." xEasterlin et al. (1993) offer two caveats to these findings. Because of some demographic differences between the baby boomers and their parents, such as delayed or foregone marriage, increased divorce rates and fewer children, "as they move into retirement, [boomers] are less likely than their parents to be living with a spouse, and are likely to have fewer adult  ?H <children" (p. 517).,5 ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#Ѝ On the other hand, they will be more likely than their parents were to have living siblings and cousins, who might be a source of support. , In addition, the economic prospects of the boomers differ markedly by income quintile. "For [the top income quintile], the life cycle profiles give no sign of flattening until the retirement ages are reached. . . .The lowest income segment of each cohort is a different story. It remains true [in this quintile] that those in the baby boom generation are currently doing better than their counterparts in the parental generation. But the percentage income advantage of the boomer generation over the parental generation is much less than for the upper income group." (p. 514) The authors conclude that "while the boomers in retirement will typically do better than their parents in terms of material wellbeing, their advantage from a total welfare viewpoint may be less. Moreover, the segment of the boomers whose current economic prospects are worst, the poorest part of the trailing edge cohorts, may do less well than their parents in terms of both economic wellbeing and total welfare." (p.520) xYakoboski and Silverman (1994) agree that baby boomers in general will enjoy a standard of living higher than that of their parents in retirement, but they may not be able to maintain their own preretirement living standard. But housing wealth may close the gap for the boomers, to the extent that they are willing to tap this source. This study emphasizes that "nonhomeowners, the less educated, the single and the youngest boomers" are at the greatest risk of experiencing financial hardship in retirement. x xKotlikoff and Auerbach (1994) and a study produced by Merrill Lynch (1994) conclude that baby boomers are likely to exceed their parents' standard of living in retirement, but they emphasize that this will not be in keeping with "the American dream" if future retirees are unable to maintain their own preretirement standard of living. In addition, the authors emphasize the unsustainability of current government fiscal policy, and maintain that once required adjustments are made,"'L0*((@@/" such as tax increases and transfer payment reductions, the retirement prospects of the boomers will dim considerably. Both the Manchester (1994) and Yakoboski and Silverman (1994) project that baby boomers' retirement incomes will exceed those of current retirees, but Yakoboski and Silverman suggest that replacement rates (the ratio of retirement to preretirement income) will fall. xThe picture painted by Levy and Michel (1991) is more negative, even about the current economic wellbeing of baby boomers relative to their parents. The major analytic difference is that Levy and Michel focus entirely on male earnings among the baby boomers, ignoring the large improvements in female earnings and the decline in the average number of children per woman. The most pessimistic picture presented by these authors is by Levy and Michel (1985), in which they emphasize the deteriorating fortunes of middle class baby boomers ("a dramatic decline in its ability to pursue the American dream: a home, financial security and education for their children" (p. 1)) and offer a gloomy picture of the boomers' future. The report implies that the growing federal deficit will keep interest rates and inflation high, and productivity and wages low, so that the baby boomers "in the 21st century [will] bear the costs of resolving not only their personal debt but the national and international debts as well." (p. 21)  ?P<xSavings and wealth: As noted in Section IID, American saving rates are low by international and by U.S. historical standards. Although returns on assets (interest and dividends) provide an important source of income for older Americans (about 17 percent of aggregate income for all elderly units in 1992), they are only about half as important as either earnings (30 percent) or social security (33 percent) (U.S. House of Representatives 1994: table A10). Among the elderly poor, interest and dividends are inconsequential, providing a mere 3 percent of total income. Many Americans reach retirement with little accumulated savings (excluding home equity), and have difficulty maintaining their preretirement standards of living. xDespite these discouraging aggregate statistics, a number of studies suggest that baby boomers on average have accumulated more wealth relative to income than their parents had at the same time in the life cycle. Exceptions are single people and the youngest baby boomers (aged 2534) in the bottom income group. "But, especially when attention is focused on the middle 60%, ratios of wealth to income are the same or significantly higher than in 1960. . . .The only groups which show consistent declines in the ratio of wealth to consumption are single people in the bottom and top quintiles. . . . All other groups, including the vast majority of boomers, have accumulated more wealth [relative"'M0*((@@." to consumption] than their parents' generation had at the same age 30 years ago." (Sabelhaus and Manchester 1993:1213) xOn the other hand, the parents of the baby boom generation (many of whom are now retired) enjoyed substantial windfalls that aided them in retirement, such as strong real wage growth in the 1950s and 1960s, substantial increases in real social security benefits in the early 1970s, increased rates of pension coverage and the substantial appreciation of housing assets. These are windfalls that baby boomers are unlikely to experience. Kingson (1992: 37) feels that "the value of equity in baby boomers' homes is likely to grow less than that experienced by current cohorts of elders." And younger boomers may be at a particular disadvantage, having missed the appreciation of house prices in  ? <the 1970s and 1980s. As house prices appreciated, rents rose.6 ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#Ѝ From 1978 to 1985 the proportion of low income households who spent at least 60 percent of their income on housing grew from 44 to 55 percent, and a substantial number of these households were headed by baby boomers (Leonard et al. 1989: xiixiv).  The high rents decreased the abilities of some baby boomers to save (for example, for a down payment), and diminished their prospects for relying on home equity in their retirement years. xOne unknown in this equation is the role that inheritances will play in the future. The current generation of retirees is doing well, on average, having enjoyed significant windfalls from social security and real estate. What ultimately happens to this wealth remains to be seen. One possibility is that it will be bequeathed back to the current generation of workers, and that this, in the end, will be a significant source of their  ?P<retirement income.7 ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#Ѝ Bernheim (1994:79) argues that inheritances are unlikely to affect significantly the retirement prospects of the typical baby boomers for several reasons. Sizable bequests are highly concentrated. In addition, the life expectancy of the older generation has increased significantly, meaning that they may use up many of their resources themselves. Finally, a large proportion of their wealth is in nonbequeathable forms, such as social security and definedbenefit pensions.  Some of their "savings," in other words, may be in their parents' accounts, not (yet) in their own. Barring this, little evidence suggests that the importance of asset income will grow significantly in the future.  ?8<xEarnings: Despite dramatic declines in the labor force participation rates of older American men in the postWorld War II period, earnings remain a very important source of income for those aged 65 or older. Earnings provide 30 percent of aggregate income for elderly households. Earnings are negatively correlated with poverty status; in 1992, fewer than 4 percent of those aged 65 or older who worked during the year (and fewer than 3 percent of those who worked yearround, fulltime) were poor, compared with more than 14 percent of those who did not work at all during the year. Among blacks, the difference is even more pronounced 7 percent of those who worked (and less than 3 percent of those working yearround, fulltime) were poor, compared with more than 37 percent of those who did not work (U.S. Bureau of the Census 1993a). xLabor force participation rates among older Americans have changed very little since the mid 1980s, suggesting that the long term early retirement trend in America is over. Other changes in the work place are under way: a moderate growth in the number of contingent workers, a growing disparity in the quality of jobs and increases in earnings and income inequality. Barring"'N0*((@@/" significant changes in public policy (for example, a delay in the age of eligibility for early social security benefits), little reason exists to believe that the importance of earnings for older Americans will change dramatically in the future.  ?<xPension income: Private and public sector employee pensions are expected to continue providing a modest but important source of retirement income, especially for middle and upper income households, and their importance should grow for women. Pensions currently provide about 16 percent of aggregate income of the elderly, although less than 3 percent for the elderly poor (U.S. House of Representatives 1994: table A10). Coverage rates declined slightly in the 1980s but have since recovered. Some (e.g., Woods 1989) project declining rates of coverage in the future, and others (e.g., Andrews and Chollet 1988; Yakoboski and Silverman 1994) emphasize the small gains during 198791, and expect further increases. Barring major changes in the regulatory environment, however, few expect dramatic improvements in the proportion of the work force covered by pensions. xThe importance of definedcontribution coverage is increasing, and many analysts expect this trend to continue. Some of this change is attributable to increased administrative costs of definedbenefit plans, and some may be caused by the preferences of workers who expect to change jobs frequently or move in and out of the labor force. Definedcontribution accumulations are more portable than are the pension rights in traditional definedbenefit plans. At the same time, as noted in Section IIC, they entail a different set of risks from definedbenefit plans, such as market risk (because only the contributions are guaranteed), lower yields (if participants are riskaverse and follow investment strategies more conservative that traditional pension managers) and the risk that lumpsum distributions prior to retirement (e.g., when changing employers) will be consumed rather than saved. Also, definedcontribution plans are more likely to require employee contributions, which means that some workers will decline to participate. xAn important unknown is how employer pensions will respond to changes in the social security environment for example, the already legislated increase in the normal retirement age from 65 to 67, or a hypothetical delay in the earliest age of eligibility. Will firms attempt to offset this by increasing their incentives for early retirement, or accommodate the change by raising their own early and normal retirement ages? If the latter, the combined effect of social security and pension changes could have a significant effect on future patterns of retirement.  ?H&< xBaby boomers most at risk in retirement: In a comprehensive report on the implications of diversity on the retirement"'O0*((@@." prospects of the baby boom generation, Kingson (1992: 50) identified the most vulnerable groups as Xxsingle parents, lowwage workers, those who are living at or below the poverty level, those who are closed out of home ownership, or those who lack adequate pension coverage. Although their numbers cannot be precisely estimated, we know they are concentrated among the 10.7 million households headed by female baby boomers. We know they can be found disproportionately among the roughly 3 million baby boomers with less than ninth grade education and among the 'one quarter of black men aged 2534' in 1980 'who had not finished high school and could not compete' in the economy in the 1970s (Levy, 1987). We know they can be found among the roughly 8 million baby boomers with incomes below the official poverty line in 1988 including about 2.8 million blacks (22.4% of all black boomers), 1.3 million Hispanics (20.3% of all Hispanic boomers), and 5.2 million whites (7.8% of all white boomers). And we know they are likely to be found among the roughly 50% of fulltime private wage and salary workers who are not covered by private pensions and among the nearly 50% of baby boomers who did not own homes in 1988.(# xApproximately 18 million baby boomers are members of racial and ethnic minorities. The Census Bureau projects that, while 86 percent of the elderly population in 1990 was nonminority, this proportion will drop to 75 percent by 2030 and to 68 percent by 2050 (U.S. Senate 1991: chart 18). In 1992, 26 percent of all blacks aged 2544 and 24 percent of Hispanics this age were below the poverty line, compared with 9 percent of whites (U.S. Bureau of the Census 1993a: table 5). Kingson suggests that the high levels of unemployment experienced by young black men may have a permanent negative effect on their lifetime earnings patterns. He points out that members of minority groups work disproportionately in arduous and potentially disabling working conditions, and so are likely to be disproportionately affected by proposals to raise retirement ages. Feldman (1991) points to a wide diversity in mortality rates and in morbidity among the elderly, which he does not expect to decline, and Kingson (1992: 44) suggests that "retirement age policy that seeks to encourage later retirement by reducing benefits (e.g., by raising the age of eligibility for full benefits) and does not provide substantial offsetting liberalizations in eligibility for disability insurance benefits will undoubtedly have very deleterious effects on those baby boomers with health problems during their later working years (especially those with fewer years of education and fewer employment options)." xManchester (1994) emphasizes that single, poorly educated boomers will fare the worst in retirement, and that baby boomers"'P0*((@@." who do not own homes may be hard pressed to accumulate sufficient wealth for retirement. xThe Congressional Budget Office projects that for workers in the bottom half of the income distribution in 2019, social security will provide 6070 percent of retirement income (U.S. House of Representatives, 1987). For most baby boomers, retirement incomes will be well above those of today's retirees. Poverty rates will be more concentrated among the single, especially single women. And, since official poverty thresholds rise with prices but not with the standard of living, those living below the poverty level in the future may feel more deprived than do poor people today.  ? <x The economy and the retirement prospects of baby boomers : Researchers generally agree that much of the baby boomers' economic progress to date, or lack thereof, can be attributed to the effects of their relative numbers and to macroeconomic trends. But analysts differ widely in terms of the relative weights they attach to these two sets of causal variables, demographic and macroeconomic. They also differ with regard to the underlying causes of the macroeconomic trends, and hence with regard to prognoses on those trends. As a result, the literature seems to fall into three general categories: ` (# hp x xFirst, the extreme demographic position is that most of the negative trends observed over the past 20 years have resulted, directly or indirectly, from the large demographic fluctuations caused by the size of the baby boom cohort. Other factors are viewed as much less important. The demographic effects have been felt throughout the economy, but most severely by the baby boomers themselves. Although the baby boomers will continue to experience less favorable market conditions than those in smaller birth cohorts throughout their lifetimes, the economy is likely to trend upward over the next two decades as the negative effects of the baby boom on the economy (low wages, low productivity, the shift to lowwage, lowskill, servicesector jobs, and low savings rates) are mitigated by the entry of smaller birth cohorts into the household formation stage. xSecond, the moderate demographic position is that most of the negative effects of large cohort size have been and will be experienced only by the baby boomers themselves, and that their relatively poor experience over the past 20 years is attributable not only to their own cohort size but also to independent macroeconomic effects, many of which have also been adverse. Demography, while important, is not destiny. These researchers tend to be more optimistic about future macroeconomic trends. They generally feel that the next two decades will be more favorable than the past two, but for the most part (with the possible exception of savings rates) they do not attribute this turnaround to demographic fluctuations."'Q0*((@@/"Ԍ xThird, according to the macroeconomic pessimists, the baby boomers have experienced adverse conditions, both as a result of their large cohort size and also as a result of poor macroeconomic performance. Researchers in this category see continued negative effects on the baby boomers of their own cohort size (such as the potential for "asset meltdown" next century) and find little reason for hope that macroeconomic conditions will improve over the next two decades.  hp x ` hp x  xAddressing the future of middle and upperwage baby boomers, Kingson (1992), who typifies the moderate demographic position (group 2), writes: "(t)he greatest threats to the economic wellbeing of these baby boomers in retirement are likely to come from health care costs, the possibility that economic productivity will be less than anticipated, legislative changes that could reduce the value of Social Security (by reducing the costofliving protection, for example, or further raising the age of eligibility for full benefits), and the failure of employerbased pensions to meet contractual obligations." (p. 37) x Kingson echoes an opinion of Russell (1982) and Levy and Michel (1991) that the bulk of the baby boomers' problems to date have not reflected adverse birth cohort size, but rather a prolonged slowdown in economic growth. These authors (along with Aaron, Bosworth and Burtless, 1989) link this economic slowdown to low levels of national savings and investment. Other economists view the productivity slowdown as an independent source of slow earnings growth, not merely the effect of low savings. x Kingson takes pains to dispel the notion that the baby boomers are a homogenous group, fated by their cohort size to a lifetime of deprivation and frustrated attempts to better the living standards of their elders. He argues that other factors (e.g., economic investment, biomedical research) can help prepare for the retirement years of the baby boomers, and accepts the arguments of Aaron et al. (1989) that if the federal deficit is reduced and if payroll taxes are increased to keep the OASDI program afloat, then the additional savings created can increase productivity to help finance the baby boomers' retirement. Similarly, Palmer (1989) estimates that a per capita GNP growth rate of only 1.5 percent per year would be sufficient to cover the increased retirement costs of the baby boom generation. Kingson's general message is that the baby boomers' cohort size is not the only determinant of their fate: he rejects the notion that demographic trends alone will undermine the wellbeing of baby boomers in retirement. x Those in the extreme demographic camp argue that the impacts of cohort size are much more pervasive. For example,"'R0*((@@/" McMillan and Baesel (1990), Fair and Dominguez (1991) and Blomquist and Wijkander (1994, using Swedish data) have suggested that the low rates of savings observed over the past 20 years have been largely the result of demographic patterns; namely, the passage of the baby boomers through the minimum savings points of  ?<their life cycles..8) ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#Ѝ` hp x F N V In addition, Mankiw and Weil (1989) and Fair and Dominguez (1991) document strong effects on housing prices, which drove prices up in the 1970s and 1980s and are now depressing prices. . These authors all attempt to document the strong macroeconomic effects of the age structure of the population. McMillan and Baesel (1990:186), for example, use Census Bureau population data to project the ratio of savers to borrowers into the future, and find that "(t)he middle population projection forecasts [the ratio of savers to borrowers] to increase from 0.666 in 1980 to 1.013 in 2005 and then drop rapidly again to 0.796 in 2030. . . . Using the middle [ratio of savers to borrowers] series, output growth per capita is negative through 1990 and then grows rapidly to 2.5% per year in 2010. The steadystate unemployment rate declines from 7.7% in 1985 to 2.7% in 2010 before increasing again. Equilibrium real interest rates decline from 6.1% in 1985 to 1.4% in 2010 before climbing back to 4.2% in 2030.'' ` hp x ` hp x  xMcMillan and Baesel also explored the effects of raising the retirement age on productivity and savings patterns. "We constructed a 'deferred retirement' middle [ratio of savers to borrowers] series by gradually increasing the retirement age from 65 to 70 during the period 20002030. . . .Surprisingly, this reaction exaggerates the projected changes in output growth, unemployment, and interest rates. . . .Why? Because under the deferred retirement scenario both the Baby Boom generation and their children are simultaneously in their prime productivity and savings years." (p.189) xBlomquist and Wijkander (1994: 47) state that "(t)he results of our simulations suggest that it would be fruitful to take demographics into account to a larger extent than done hitherto in models of the economy. The simulations also show that, whether the interest elasticity of individual savings is positive or negative, there exists no stable relationship between the real rate of interest and aggregate household savings. Both the rate of interest and savings are endogenous in our model. Depending on the form of the demographic changes, the changes can for some time periods create a positive correlation between savings and the interest rate, and for other periods a negative relationship." xAuthors in the third category, the macroeconomic pessimists, take exception to these conclusions. For example, Cantor and Yuengert (1994: 76) suggest that "(w)hile the baby boomers' savings rates should rise with age, the impact on the aggregate saving rate will be mitigated by a continuing high rate of early retirement and a rising share of households headed by individuals either over 65 or under 35 households that tend to have relatively low saving rates.'' "'S0*((@@/"Ԍ xAuthors in the third category tend to be pessimistic about baby boomers' prospects, because they see continued high real interest rates and low savings producing low productivity and wage growth. They think that little of the macroeconomic effects experienced by the baby boomers to date have been attributable to their cohort size, and they therefore see little relief as the demographic structure changes. xAs can be seen in this research, demographic changes are easier to predict than are their macroeconomic consequences. This conclusion is frustrating because the state of the economy over the next several decades will be a key determinant of the size of the pie to be allocated among workers and retirees and therefore of the state of financial wellbeing of future retirees.  ?H <x Asset value uncertainty : Funded pension plans, both private and government, face the same future demographic and economic environment as social security. Just as social security finances are predicted to be strained by the retirement of the babyboom generation, funded pension plans may likewise face massive net redemptions, as aggregate benefit payouts exceed contributions and asset earnings. Further, housing values could be depressed by the relative sizes of the elderly cohort and the household formation cohort, as suggested by Mankiw and Weil (1989). One conclusion is that the members of the baby boom generation, because of their sheer numbers, face added uncertainty about all of their retirement assets. ` hp x ` hp x xPrivate pensions have been a major source of net saving in the economy for the entire postWorld War II period. Figure II9 shows the change in real national wealth and the change in real pension wealth (all expressed in 1990 dollars) for 5year intervals since 1950, taken from the National Balance Sheets of the Flow of Funds Division of the Federal Reserve Board. In the 1950s, 1960s, and 1970s, the change in real pension wealth tended to be between 10 and 15 percent of the change in the national wealth. Since 1981, however, the relationship between the growth in real pension assets and the growth in real national wealth has changed dramatically. It should be noted that the Flow of Funds accounts attempt to measure the market value of assets. Pension assets, particularly stocks and bonds, enjoyed large total real rates of return during this period and the change in real pension wealth accelerated somewhat from the levels of the 1970s. But the change in real national wealth was much lower in the 1980s than in the previous three decades and actually turned negative in the first part of the 1990s. xSome of this change is due to the sharp fall in national saving rates discussed above. The change in the real national wealth was also depressed in recent periods by depressed farm and"'T0*((@@/" commercial real estate values. The result is that the growth in the real wealth in pension funds has exceeded the growth in the real wealth of the country since 1981. Although this phenomenon is not likely to continue, it does reinforce the idea that pensions are a tremendously important part of the national saving picture in the United States. xIn a recent study, Schieber and Shoven (1994) superimposed the 1993 social security intermediate demographic and economic forecasts on the private pension system, taking into account the age and gender structure of the working and retired populations, the pattern of pension plan participation and plan design, the mobility of the workforce and observed retirement behavior. They referred to the longrun real rates of return on assets shown below, and then made conservative assumptions about future returns, assuming real rates of return of 5 percent on stocks, 2 percent on longterm corporate bonds and a zero real rate on money market instruments such as Treasury bills. xThe net saving of private pensions (contributions plus real asset returns less benefit payouts) is equal to about 3.7 percent of the total private payroll in the economy, or between 2.5 and 3.0 percent of GDP. Once again, the importance of pensions is apparent since the total net saving rate in the economy has not been above 3.0 percent of GDP since 1985. The available evidence indicates that pension saving is at least the same order of magnitude as national saving. Shoven and Schieber project that while this may continue for another decade or so, the demographic structure is such that pensions are not likely to remain a main source of saving in the economy. Their baseline forecast for private pensions is shown in Figure II10. The analysis assumes that the fraction of payroll that corporations contribute to pensions will remain at approximately current levels. xThe message from this simulation is both clear and dramatic. Pensions, which have been a major acquirer of assets a source of saving for the past several decades, will soon become a major liquidator of assets a source of dissaving. The projection shows aggregate pension benefits exceeding aggregate contributions for the first time in 2006 and aggregate benefits exceeding both contributions and real asset returns by 2024. Funded pensions become an enormous net seller of assets at approximately the same time that the social security system begins to run a negative cash flow and attempts to sell off the assets in its trust fund. One wonders, to say the least, what effect this massive change from buyers to sellers of assets will have on asset prices. xIt is possible that asset prices will by depressed by the selling pressure resulting from the sheer numbers of the retired baby boomers. The housing price inflation of the 1970s could be replayed in reverse in the second and third decades of the 21st"'U0*((@@/" century, although this time the affected assets would likely include stocks, longterm bonds, houses, and other forms of real estate. If asset prices are depressed by the simultaneous liquidation of pension fund and social security trust fund reserves, it could have a major impact on participants in definedcontribution plans (including IRAs, Keoghs, and 401(k) and 403(b) plans), on the sponsors of definedbenefit plans, and on the PBGC. Obviously, people who save outside a pension vehicle would also be disappointed in their realized rates of return if this scenario occurs. It should be emphasized that what is being suggested is that savings will be very scarce, real interest rates high, and thus the value of real assets and financial assets behind them reduced. xThe underlying demographic pattern is a given, and will almost inevitably cause the saving of the pension system to slow, whether or not the economic assumptions used by the Social Security Administration and by Shoven and Schieber turn out to be accurate. The question of interest here is whether the decreased pension savings will cause asset prices to fall and by how much. The answers depend on economic parameters that are not well known. First, there are several reasons why one might not expect the switch of pension funds from net buyers to net sellers to depress prices. Rational expectations theorists would assert that this occurrence has already been foreseen by market participants and therefore is already imbedded into today's asset prices. Because of the foresight of investors, the occurrence of a widely foreseen event should have no contemporaneous effect on prices. This theory will be put to the test in a significant way with the impact of the retirement of the baby boomers on asset markets, and the members of the cohort themselves have more than an academic interest in the outcome. The rational expectations theory, at least in its strongest form, is not widely enough accepted to suggest that baby boomers need not be concerned. xEven if one does not subscribe to the strong rational expectations view, one might hope that the impact of American retirees on world capital markets would not be sufficient to affect global asset prices and interest rates. Even if there are not enough "Generation X" members to buy the assets of the boomers without a significant discount, the assets can be sold to other participants in global capital markets. The problem with this argument is that the demographic structures of Japan and Europe are similar to that of the United States. It is difficult to imagine the elderly in Japan, for instance, buying the assets of America's elderly when the Japanese will also be trying to finance the retirement of their large elderly cohort. Perhaps the problem will be alleviated by China and the other emerging economies in the world, but considerable doubt exists about that. xEstimating how much asset prices might be depressed is difficult and speculative. It is useful to recall that the"'V0*((@@/" United States currently has 3.2 workers per retiree, and that this ratio is expected to fall to approximately 2.0 workers per retiree by about 2030. These figures are often used to describe the pressure that the baby boomers would cause a purely payasyougo social security system. Two workers (rather than 3.2) would have to finance the benefits of each retiree, forcing a reduction in retiree benefits, an increase in worker contributions, or both. Of course, reality is not quite so simple; for instance, the necessary payroll tax rate depends on the level of real wages as well as on the number of retirees per worker. However, the ratio of workers to retirees is very important when assessing social security finances. This same ratio can provide some rough intuition about the potential magnitude of asset price reductions due to the retirement dissaving of the baby boomers. xIn a closed economy model, for example, the workers would have to finance retirees' benefits in a payasyougo system and also buy the assets of the retirees in a funded system (either a private pension system or a funded social security system). If the 2 workers do not save more per capita than the previous generation's 3.2, and if the amount they save is unresponsive to changes in the rate of return offered on assets, then asset values could fall by approximately onethird. Such a fall in the market value of the U.S. capital stock relative to its replacement cost is not unprecedented. In fact, in the late 1970s U.S. capital stock was selling for less than 70 percent of its replacement value. Despite the fact that assets have been depressed in value by this much before, it still would have a major impact on the wellbeing of the baby boomers in retirement. It is even possible, of course, that the 2 workers per retiree will each save less than previous workers because of the higher income taxes and payroll taxes that they may face and because of the other impacts of a higher total dependency ratio in the economy. In that event, the depression in asset prices could be larger. xAs mentioned earlier, this analysis of the magnitude of the price effects is problematic. In fact, the adjustment to the lack of pension saving could take place in other ways. Rather than sell their assets to the generation of workers, the retired could simply consume their capital stock; that is, run them down and extract as much cash as possible. The adjustment to the dissaving of the numerous elderly could be disinvestment of the capital stock lower or even negative net investment in the economy. Although all of this analysis has been for a closed economy, it is not at all clear that the outcomes are significantly different in an open, global economy, given the demographic similarities in the developed nations mentioned above. "'W0*((@@."ԌxA modest and pragmatic conclusion is that the large baby boom generation does face extra uncertainty about their retirement assets, including social security. The government could help the situation by encouraging private saving and reducing the federal deficit (that is, reducing government dissaving).  ?x<` hp x` hp x (#%'0*,.8135@8:x The health and longevity of older Americans : Statistics clearly indicate that the life expectancy of Americans is on the rise. Men and women born in 1940 (just after the creation of the social security system) could expect to live 61.4 and 65.7 years, respectively. By 1990, these life expectancies had risen to 71.1 and 78.8 years, 10 and 13 years longer than in 1940 (U.S. House of Representatives 1994: table A2). By 2010, when the oldest baby boomers will have attained traditional retirement age, the life expectancy of newborns is projected to be 74 and 80 years. xFor those who actually reach age 65, similar trends are observed. In 1940, the life expectancies for men and women who reached aged 65 were 11.9 and 13.4 years; by 2010, they will be 15.8 and 19.7 years, increases of 4 and 6 years, respectively. xThese increases play a prominent role in the debate about the appropriate ages of early and normal retirement in the future. But an equally important issue concerns the health status of older Americans. Does living longer mean living healthier? In this section, the Panel discusses recent trends and projections regarding the health of workers around early and normal retirement age and the recent literature on mortality and morbidity. A key question is whether increases in life expectancy are likely to produce longer productive lives (and if so, for whom) or merely longer periods of morbidity. xBecause the sharply increasing life expectancies in the 1970s appeared to be accompanied by an increasing incidence of selfreported health problems, some researchers (for example, Gruenberg 1977; Verbrugge 1984) felt that the life expectancy advances had resulted from reduced mortality from chronic disease, resulting in a higher proportion of elderly with long periods of morbidity. Others felt that while the first effect of reduced mortality from chronic disease would naturally be higher levels of morbidity, these levels could be reduced as medical science turned its focus toward increasing the age of onset of chronic disease (Fries, 1980). Crimmins et al. (1994) simulated various patterns of reduced mortality and morbidity, and showed that the first effect would be a lengthening of the period of morbidity and dependence, followed by a shortening as age at onset is increased. xData presented by Crimmins and Ingegneri (forthcoming) indicate that after a period of deterioration in selfreported"'X0*((@@/" health in the 1970s, "there is some evidence that the trend toward deterioration in selfreported health among the middleaged and older population was arrested and perhaps even reversed in the 1980s." They reject the hypothesis that the turnaround was caused by changes in people's expectations regarding health, because no reason exists for these expectations to have declined in the 1980s. xCrimmins and Ingegneri's finding of a reversal in the trends of reported health of the elderly during the 1980s is supported by Manton, Corder and Stallard (1993), Waidmann, Bound and Schoenbaum (1994), and Manton, Stallard and Corder (1995). Manton et al. (1993) analyzed samples of the Medicareeligible population found in the National LongTerm Care Surveys and found a modest drop in the prevalence of chronic (90 days or more) disability between 1982 and 1989. Waidmann et al. (1994:1) concluded that "prevalence rates for the chronic conditions most likely to be disabling either remained stable or fell." Finally, in the most recent study, Manton et al. (1995) used the National LongTerm Care Surveys to examine 16 specific medical conditions. They report that the average number of these medical conditions per person among U.S. elderly fell by 11 percent between 1982 and  ?0<1989.9 ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#Ѝ The mean number of conditions in 1982 was standardized to the 1989 age, sex and disabilityspecific population distribution in 1989. Of the 16 conditions studied, the prevalence of seven declined between 1982 and 1989 (arthritis, dementia, stroke, arteriosclerosis, hypertension, circulatory disease and emphysema), four remained about the same (t < 1.96; heart attacks, cancer, diabetes and asthma), while five increased (other heart problems, pneumonia, bronchitis, Parkinson's disease and fractures). Overall, the declines outweighed the increases.  Declines were observed consistently across age, sex and disability level groups. xBound and Waidmann (1992) went a step further and argued that average health among the elderly did not actually deteriorate in the 1970s, despite the declines noted in the selfreported data. Rather, they argue, the changes reported in the 1970s were the result of a combination of social forces affecting the way individuals perceive their health and report on it. More generous disability benefits and easier access to them increased the proportion of the population who described themselves as impaired, and then applied and often received disability benefits (Quinn and Burkhauser, 1994a). Crimmins and Ingegneri (forthcoming) agree. "This explanation certainly fits the time trend for the older working age and retirement age population of deteriorating health during the 1970s, a period when benefits became increasingly available, and improving health during the 1980s, a time when benefits became more difficult to obtain." x  ? <xWaidmann et al.(1994) argue that changes in the social security disability program in 1965 (when workers with temporary as well as permanent disabilities became eligible for benefits) and the establishment of the Supplemental Security Income (SSI) program in 1974, together with increases in real benefits and increasingly liberal eligibility requirements from 1965 through the mid1970s are strongly related to the pattern of selfreported health during this period. "When we include SSI, the changes in the fraction receiving disability income closely match the changes in the fraction reporting that they are limited in their ability to perform their major activity. . . . Furthermore,"'Y0*((@@/" the most dramatic changes in the fraction of individuals reporting activity limitations occur for those under the age of 65 those who are eligible for disability programs. Indeed, for men and women aged 70 or older, who were presumably unaffected by changes in Disability Insurance (DI) and SSI, levels of activity limitation remained effectively unchanged throughout the study period. If trends in activity limitations reflected mortality declines, we would expect to see these age patterns reversed, since mortality declines were greatest for those over 65." (p.14) Finally, Bound and Waidmann (1992) add that efforts at disease prevention and early diagnosis increased dramatically in the 1970s, leading to increased awareness among the elderly population, rather than absolute increases in the incidence of disease. xAs usual with the elderly, the mean or average can be deceptive. Feldman (1991) finds a substantial increase in the gap between the death rates of those with low and high levels of education between 1960 and 1980. He suggests a scenario in which the longevity gains and morbidity declines fall primarily on those of higher socioeconomic class, while those at the other end of the spectrum (even if enjoying small absolute gains) fall further behind. In this case, increases in the ages of eligibility for social security benefits, which might be suggested by the health gains of the elderly on average, should be tempered by a concern for those not sharing in this good fortune. xIn summary, although the jury is still out on these issues, the latest available evidence suggests that the health of the middleaged and older Americans has increased during the 1980s and into the 1990s, and probably more so at the upper socioeconomic levels. These gains are likely to continue, as medical technology is refocused from reducing mortality from chronic diseases (the primary initial focus) to increasing the age of onset of chronic disease. This is expected to compress the period of dependence and illhealth, at the same time that life expectancy is increased. This trend is expected to be reinforced by the observed relationship between mortality and education (Preston and Elo, 1994), given projected increases in the educational levels of the elderly. xIt was the observed pattern of deterioration in selfreported health among the middleaged and elderly in the 1970s that led to the "failure of success'' argument the hypothesis that sharply decreased mortality from chronic disease resulted in increased proportions of frail and sick elderly (Gruenberg 1977). Recent work appears to have undermined that argument, pointing instead to improvements in the early identification of chronic disease, and therefore the earlier accommodation of known health problems. "'Z0*((@@/"Ԍ ?< xConclusions: Research suggests that the economic status of current retirees is, on average, better than that of their parents (largely because of social security), that the current economic position of the baby boom generation is better than that of their parents at the same age, but that the saving rates of the baby boomers are lower than necessary to maintain their standards of living after retirement. Single people, especially single mothers, those with low education or irregular work histories, minorities, those without home ownership, and those without the option to continue working as they age are at the greatest risk with regard to their retirement prospects. Opinions diverge about how the economy will evolve between now and the time of the baby boomers' retirement. One source of concern is the future profile of asset prices, when private and public retirement systems, both here and abroad, change from net buyers to net sellers of assets. On average, both the life expectancy and the general level of health appears to be rising for older Americans, but differences (which may be increasing) occur by socioeconomic class. xThis analysis of future trends takes the social system as given, despite the fact that currently legislated contribution rates are insufficient to pay currently legislated benefits. Something has to give, and the decisions about what, when and to or for whom can have large impacts the future economic wellbeing of older Americans. E ` hp x (#%'0*,.8135@8:` hp x (#  ?< X1Í.X1."[0*((@@P" III. Policy Options for Dealing with Projected Social Security  ?<Imbalances xIn this section the Panel describes several evaluation criteria that it devised for judging the effects of alternative policies to correct projected social security imbalances. The Panel also outlines what this set of policies might includespecifically, increases in payroll taxes and curtailments in benefits. Finally, the Panel evaluates these potential changes using the specific criteria outlined at the outset and indicates where possible tradeoffs lie. The Panel s charge, and its goal in preparing this Report, was to outline and evaluate options rather than to make policy recommendations, on the understanding that some combination of benefit cuts and/or revenue increases are necessary to restore the social security system to actuarial balance. It is the Panel s intention to offer a framework for assessing various policy alternatives but not to make specific recommendations on which particular package to adopt.  ?< A. Criteria for Evaluating Social Security Benefit and Tax  ?<Changes xThe Panel developed six criteria as guidelines for evaluating the effects of alternative methods of correcting projected social security imbalances:  ?P<Criterion 1. Adequacy of retirement income. A primary goal of the social security system is to contribute to an adequate retirement income for older Americans. Many Americans approach their late middle age without adequate sources of retirement income. Some experienced poverty all their lives and therefore were not able to save. Others may have been myopic (thinking that they would never retire), may have refused to think about their consumption needs in old age, may have underestimated their longevity, or simply may have miscalculated what was necessary to maintain their standard of living in retirement. xFrom a policy perspective, there are two ways of thinking about retirement income adequacy. One legitimate goal is the avoidance of poverty among older Americans (that is, meeting a minimum income standard). Another is the provision of a reasonable replacement rate, meaning that retirement income should attain at least some minimum fraction of preretirement income. The Panel understands that both standards are imperfect measures of the adequacy of a particular retirement package, because they examine only current levels of income and ignore the possibility that oldage consumption patterns change over time. Nonetheless, these measures are useful and widely discussed ways of judging the adequacy of retirement income. Using this criterion, a social security change that increases retirement income adequacy would be judged positively. "'\0*((@@/"ԌxIn addition to concern about incomes near the time of retirement, it is also important to examine the continuing value of benefits afterwards. The fact that social security benefits are paid as an annuity indexed to the Consumer Price Index (CPI) plays a crucial role in maintaining the adequacy of retiree benefits as people age. Because employerprovided pensions are not usually fully indexed, they can shrink in real terms over time. Similarly, people may consume their nonannuitized wealth as they age, again exposing them to the risk of outliving their command over resources. Thus adequacy many years into retirement must be considered, as well as benefit adequacy immediately after retirement.  ?( <Criterion 2. Insurance against income fluctuations. Social security is designed to provide some insurance against unforeseen events that interrupt a worker s flow of income. These shocks include death of an earner and events that precipitate early retirement such as disability. One could also include here bad fortune in the choice of an occupation whose skills have become obsolete, or unanticipated shocks to the value of training and education (for example, the result of industry downsizing). Social security also provides risk pooling, providing income to those who are eligible for benefits by virtue of having met benefit eligibility criteria. For example, a single worker who contributes to the system and survives to retirement age receives retired worker benefits, but a single worker who dies before retirement age does not receive payments to his or her estate. Similarly, social security offers insurance by transferring from shorter lived people (men, on average) to those who are longer lived (e.g. women, on average). In this way, the social security system affords a form of insurance by pooling risks across those who do, and do not, eventually meet eligibility criteria. xInsurance that cushions the blow of reduced earnings will inevitably affect people s decisions and behavior. For example, research suggests that the availability of social security retirement benefits induces early retirement and increases the chance that someone who could work will apply for disability benefits. Also, workers are covered by Disability Insurance, and if they have dependents, these receive Survivors Insurance. These components of the social insurance program act as labor market distortions by lowering the worker's effective net labor market earnings (the difference between what the worker could earn on a job and the social benefit payments). Some labor market inefficiency is an inevitable consequence of providing insurance. Thus, the question is not whether labor market distortions exist, but whether the amount of insurance is too great or too small given the accompanying distortions.  ?H&<Criterion 3. Avoidance of market inefficiencies. With respect to market inefficiencies, the Panel focused on two important decisions that individuals and families make, namely:"']0*((@@/"Ԍxi) The laborleisure choice, or how people allocate their time both in their early and middle years, and also at the time of the retirement decision; and xii) The savingsconsumption choice, or how people allocate income between consumption during the worklife and consumption during retirement. xSocial security influences both laborleisure and savingsconsumption behavior, enabling people to retire earlier than they otherwise would. For younger workers, social security payroll taxes reduce the marginal takehome pay of those earning less than maximum covered earnings. People will tend to perceive that they are paying more in social security contributions than they will receive in the form of social security benefits (that is, if payroll taxes are considered to be net taxes), which will reduce the value of working and affect labor supply decisions as do other taxes on earnings. xSavings patterns are affected by social security if promised benefits induce some people to decrease their own private savings. How much they are affected is important to national savings, an issue taken up below. The Panel is also concerned about whether people save what they should given their opportunities and constraints, and about whether these private savings decisions are influenced by government programs such as social security. Although neither assets not asset income directly affect the computation of a retiree s social security benefits, the program does interact with both via the income tax and meanstested benefits programs. These interactions cause some to save at rates different from those they would if not confronted by the distortions inherent in these programs. For example, taxing retirees  social security benefits may discourage private savings by those whose asset income would otherwise be high enough to expose them to benefits taxation. Alternatively, for some who retire early with social security, additional private savings may be generated in anticipation of a longer period out of the labor force. Overall, a social security reform which reduced labor market and savings distortions would be viewed positively according to this criterion.  ?@<Criterion 4. Equity of lifetime social security taxes and  ? <benefits. Social security differs from privatelypurchased insurance in that it includes explicit and implicit redistributional components. For example, those who are high earners over their lifetimes pay more in taxes and receive lower returns on these taxes than do the lifetime poor. In addition, substantial crossgenerational differences exist in benefits received versus taxes paid. Massive intergenerational transfers have been delivered to current and past retirees who contributed to the system in its early years before there were many retirees, financed by the current generation of workers to whom smaller net benefits will accrue under current law. This pattern, often seen"'^0*((@@/" during the start up of certain types of social insurance systems, is no longer viable. xThe Panel recognizes that the past and future of social security are very different. Under this criterion, therefore, the Panel asks how various reform proposals affect the lifetime distribution of social security contributions and benefits. In general, the Panel seeks to identify changes that alter the distribution of lifetime benefits and taxes both: xi) Between generations; and xii) Within generations.  ?( <Criterion 5. Encouragement of national saving. The Panel shares many experts  concern that the national saving rate in the United States is too low. This low saving rate undercuts investment, which in turn is a primary determinant of future output. All forms of saving are important: personal saving, corporate saving and saving (or dissaving) by the government sector. When using this evaluation criterion, the Panel asks whether a specific policy proposal is likely to increase aggregate national saving.  ?0<Criterion 6. Strengthening the financial integrity of retirement  ?<income systems. The Panel s goal is to discuss policy proposals to help eliminate the longrun fiscal imbalance of the social security system. The typical social security analysis forecasts the system s budget 75 years into the future. In the Panel's view, this focus should be broadened to address other concerns influencing the social security system s integrity. The Panel also emphasizes that social security reform proposals should be evaluated in terms of their contribution to strengthening the financial integrity of the private retirement income system as well. In general the Panel believes that:  ?< xi) All proposed reforms should ensure that the Social Security Trust Fund does not approach or dip below zero during the 75year interval; and xii) A reform that offers a better balance at the end of the 75year period and thereafter implies stronger financial integrity. xThe Panel stresses that these criteria be used to design a combination of benefit cuts and/or revenue increases necessary to restore the social security system to actuarial balance.  ?#< B. The Scope of Social Security and the Timing of Adjustments xThis section begins with a description of two baseline scenarios for social security that frame most of the other reforms currently under discussion. In this analysis the Panel compares and contrasts a "larger" versus a "smaller' social"'_0*((@@/" security system. A larger system maintains promised benefits at their currently legislated levels. This scenario implies that taxes must be raised to cover these promised benefits. At the other end of the spectrum, a smaller system would maintain current payroll tax rates while reducing benefits to achieve balance with those implied revenues. The size of the system can be varied continuously between those two limits. x xThe Panel assesses the pros and cons of both the smaller and larger systems using the evaluation criteria developed above. The Panel then addresses the question of timing, focusing on the relationship between the timing of benefit cuts or tax increases and the size of the changes needed to achieve actuarial balance.  ? <x Larger versus Smaller Systems : The Panel believes that the social security system must be restored to actuarial balance. Many combinations of revenue increases and benefit decreases could achieve this. To list them all, comparing each one with each other one, would be tedious and difficult to follow. Instead, the Panel chose to describe what we believe to be the most important types of revenue increases and benefit decreases.  ?h<The Panel selects what it terms a baseline tax increase option,  ?0<along with a companion baseline benefit decrease option, and then proceeds in four logical steps. First, the Panel uses the evaluation criteria described above to discuss the baseline tax increase and benefit decrease. Second, the Panel considers alternative timing paths for the baseline reforms. Third, it examines other ways to cut benefits with the baseline benefit cut. Fourth, other tax increase policies are compared with the baseline tax increase option.  ?8<x Comparing tax increases and benefit decreases : Our  ?<hypothetical baseline benefit decrease is a proportional decrease in the Primary Insurance Amount (PIA) formula. In this scenario, all persons born in or after a particular year, which is arbitrarily taken here to be 1940, will experience the same percentage decrease in their PIA amounts. The hypothetical  ?<baseline tax increase is a rise in the payroll tax rate beginning in the year 2002 when workers born in 1940 attain age 62 with equal increases levied on employers and employees (and a matching increase for the selfemployed). These two baselines are chosen for their simplicity, not because the Panel has selected them as particularly desirable changes to make. xIn the first baseline scenario, the Panel assumes that benefits would be cut by reducing the Primary Insurance Amount (PIA) by the same percentage for all. This reduces the benefit as a fraction of the worker's Average Indexed Monthly Earnings (AIME). For example, in 1995, a worker s PIA is computed as 90 percent of his first $426 of AIME, plus 32 percent of the next $2,567 of AIME, plus 15 percent of AIME above that. (Thus $426 is the "first bend point" and $2,567 the "second bend point" in"'`0*((@@/" the PIA formula.) A 10percent reduction in benefits would be achieved by proportionally reducing the percentages in the PIA formula, such that the smaller PIA would be 81 percent of the first $426 of AIME, plus 28.8 percent of the next $2,567 of AIME,  ? <plus 13.5 percent of AIME above that."endnote reference"X1.X01ÍÍ : ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#59. The point of the present discussion is to compare an acrosstheboard benefit cut sufficient to achieve longrun system balance, with the alternative, a payroll tax increase generating equivalent revenue. Whether any particular PIA percentage reduction or tax increase will be sufficient to bring the system into long term balance given a particular set of assumptions is a  ?<question that is best addressed by the actuaries.#XZ\  P@Q2XP#  This formula change reduces benefits proportionately for all those covered by the new formula, on the presumption that family maximums and disability benefits are treated symmetrically. xThe second baseline scenario increases payroll taxes beginning in the year 2002 by enough to raise the same amount of revenue that is generated by the "baseline" benefit cut described above. It would not alter the maximum earnings subject to tax or the definition of earnings subject to tax (for example, employee benefits), and would maintain the current division of the tax among employers, employees, and the selfemployed.  ?H <x Magnitudes of Benefit Cuts or Tax Increases : Table III.1 shows three combinations of proportional PIA decreases and payroll tax increases that would be sufficient to restore actuarial balance over the next 75 years, assuming they were legislated to begin in the year 2002, and no other changes were effected beyond those already legislated. Looking from left to  ?< Table III.1. Combinations of PIA Decreases and Payroll Tax Increases Adequate  ?<to Restore 75year Actuarial Balance, Three Start Dates ______________________________________________________________  ?< x` `  Size of social security retirement program x` `   x` `  hhLarger..................... Smaller ______________________________________________________________  ?< Start date 2002:  ?X<Percentage cut in benefits@ 0h pp10.25  xx20.5  ? <Tax rate increasehh@ 2.5hpp 1.25  xx 0  ?< Start date 2012:  ?x<Percentage cut in benefits@ 0h pp12.75  xx25.5  ?@<Tax rate increasehh@3.12hpp 1.56   0  ? < Start date 2022:  ?!<Percentage cut in benefits@ 0h pp16.75  xx33.5  ?`"<Tax rate increasehh@4.04hpp 2.02   0 _____________________________________________________________  ?#<Note: Each proposal is structured so as to restore actuarial balance over the next 75 years, assuming changes implemented as of the start date given and no other changes beyond those already legislated. Computations are based on assumptions and time horizons relevant in 1994.  ?'<Source: SSA, Office of the Actuary."'a0*((@@/"Ԍright, benefit decreases become larger and tax increases grow smaller. Now the Panel asks how alternative combinations of changes in taxes and benefits would be assessed by policymakers, applying the criteria developed above.  ?<Criterion 1. Adequacy of retirement income. Changing the PIA formula reduces benefits proportionately for all social security claimants. Such a benefit decrease would reduce some retirees  and survivors  benefits and total retirement incomes below a minimally adequate level. The larger the PIA benefit cut, of course, the larger will be the number of people with inadequate incomes. Replacement rates, or the ratio of social security benefits to preretirement earnings, would also be reduced for all retired workers. xOne group likely to be affected by PIA cuts is very old widows whose deceased husbands had low lifetime earnings: a proportionate reduction in all social security benefits may eventually produce higher levels of hardship among very elderly widows. To the extent the elderly experiencing benefit cuts apply for and receive Supplemental Security Income (SSI) or food stamps, they would be somewhat protected from poverty. Some are reluctant to apply for meanstested benefits after they have worked and contributed to social security over a lengthy career. The SSI and food stamp amounts are also quite low, currently insufficient to lift the indigent over the poverty line. Finally, older people can qualify for SSI on the basis of age only after age 65, but some of the people affected will be between 62 and 65. In general, then, a reduction in benefits thus reduces adequacy of retirement income. x xBy contrast, raising payroll taxes threatens retirees  benefit adequacy far less because most are not earning much income subject to this tax (Leonesio 1990). Some people both work (and pay payroll taxes) and receive benefits, but for the most part, this is not a lowincome group. In consequence, raising payroll tax rates would do little to harm retirement income adequacy.  ?@<Criterion 2. Insurance against income fluctuations. A proportionate cutback in OASI benefits (the baseline scenario) reduces the insurance protection provided by social security for workers and survivors at all income levels. The larger the benefit cut, the greater the loss of income insurance. In contrast, raising payroll taxes threatens retirees  insurance against inadequate incomes far less because most are not earning much income subject to the payroll tax.  ?H&<Criterion 3. Avoidance of market inefficiencies. If retirement benefits were smaller, older workers with good employment prospects would be less likely to withdraw from the labor force"'b0*((@@/" to collect social security benefits. The magnitude of this effect may be modest: research suggests that a 20 percent reduction in the PIA would cause healthy older men to postpone their retirement by fewer than six months (Fields and Mitchell 1984b). Of course, a link exists between labor market inefficiencies and providing insurance against inadequate income. The length of one's earnings career is a critical determinant of lifetime economic wellbeing, subject both to individual decisions and to random events. An inevitable tradeoff arises between the level of insurance the system provides and the degree to which retirement decisions are distorted. xSocial security benefit cuts may also alter the laborleisure choices of younger workers. Peoples' decisions about work and human capital accumulation patterns can be influenced by the social security payroll taxes they pay and the future benefits they expect to receive. For workers with very low lifetime earnings, additional years of covered earnings might increase the lifetime return from social security, with the expected (present) value of future benefits rising by more than an additional year s taxes. For higher paid workers, working longer often raises future benefits by less than extra taxes paid. Similarly, many twoearner couples (along with middle and high income single workers) tend to earn few extra benefits from the social security system for another year of work at the end of the worklife. As a result, a smaller system offering lower benefits is likely to reduce these potential labor market inefficiencies. In contrast, a larger social security system with higher tax rates makes work less appealing, though greater need for income as a result of higher taxes encourages work. On net, the labor market distortions probably encourage earlier  ?8<retirement.()endnote reference("endnote reference"\; ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#` hp x (#` hp x (#60. We recognize that these distortions do not influence all employees, since the net value of working more years under social security depends on life expectancy, interest rate assumptions, actuarial adjustment factors, and individual earnings patterns that vary from person to person. In what follows we frequently presume that delayed claiming of social security benefits tends to reduce the net value of benefits received, a conclusion which applies to many but not all workers. \ xOther problems may arise when a larger system influences peoples' saving and consumption choices. One possibility is that because social security benefits are tied to earnings rather than income from saving, the system does not unduly distort saving decisions. Nevertheless, other programs that are means (or asset) tested do interact with social security benefits, which produces a similar result. For example, Medicaid coverage for nursing home care is only available to those with limited assets and little income, all factors that interact with saving and social security benefit policy. How large these effects are is currently not well known, and the topic requires further study. xAnother way that social security probably influences saving patterns is that the system forces deferred consumption for the subset of (myopic) people who save too little to live comfortably in retirement. From this perspective, a larger social security system represents an improvement in this group's wellbeing. Even for these people, however, the fact that benefits are paid as an indexed annuity can affect saving. Those who desire an"'c0*((@@/" indexed annuity, but cannot obtain it in the private market, may save less since social security offers such an annuity. On the other hand, social security also forces saving in a way that cannot be offset by some people who already save adequately; for this group, having a smaller system diminishes distortions. On net, whether a smaller system is preferred depends on whether Americans would be more likely to save too little, or adequately, on their own. Many studies have sought to answer which behavior is more prevalent, but as yet the evidence remains controversial.  ?<Criterion 4. Distribution of taxes and benefits. Proportional social security benefit cuts or tax increases leave the relative distribution of both taxes and benefits within a generation unchanged. For instance, if a lowwage worker currently receives twothirds of the social security benefit received by an averagewage worker, this ratio would continue were PIAs proportionally reduced. However, the absolute difference between the benefits received by lowwage and averagewage workers would obviously decline. Reduced social security benefits would thus have a smaller redistributive impact on the incomes of older Americans than they do under the current formula. xOther dimensions of withingeneration redistribution have concerned some analysts. For example, the longlived elderly would be better served by reforms that raised taxes more and cut benefits less. Conversely, shortlived workers and survivors would favor larger benefit reductions and smaller tax increases. (A single worker who dies at age 60 gains if payroll taxes are reduced and does not suffer any loss in retirement benefits if OASI pensions are scaled back. By contrast, a worker who lives until age 90 gains from the payroll tax reduction but loses a substantial amount of lifetime benefits if pensions are scaled back.) xSocial security insures workers and their spouses and dependents against the risk of exhausing their assets due to higher than expected longevity. For this reason, many observers do not think social security's redistribution from the shortlived to the longlived is redistribution in any meaningful sense. At age 18, when people enter the work force and begin to contribute to social security, workers are uncertain whether they will live to 60 or to 100. Hence, from the perspective of an 18yearold worker, a reduction in both social security benefits and taxes does not favor or harm the worker in comparison with other workers of the same age. A problem arises, however, when some entering workers expect to live longer than other workers. Healthy workers on average live longer than unhealthy ones; and higher income workers on average live longer than lower income workers. A reduction in both social security benefits and taxes will thus harm some and benefit others, as compared with the status quo. From this perspective, a smaller social security"'d0*((@@/" system would redistribute less within a generation, as compared with a larger system. xRegarding betweengeneration distribution, a smaller system brought about by a proportional reduction in the PIA (initiated in the future and not made retroactive) would leave the entitlements of older people unchanged, though to the extent that they worked past 2002, they would be subject to lower payroll taxes. This older generation would therefore benefit from changes in the baseline package that embody the smaller tax increase. However, these effects are likely to be small. In contrast, for nonretired persons currently age 54 or younger, the younger the worker, the more that he or she would favor having a lower social security tax and lower benefit because younger cohorts will have more years ahead paying taxes than receiving benefits. xWhether this pattern of redistribution is viewed favorably or unfavorably depends on a number of factors. One of the most important is the prospect for real wage growth. If real wages seem likely to rise, younger workers will enjoy higher living standards than the older generations, irrespective of social security benefit cuts. On the other hand, if wage growth prospects are poor, the living standards of younger workers may be similar to or worse than those of older workers. xIn assessing lifetime tax and benefit comparisons, it is worth noting that under current law the normal retirement age will be 67 for workers attaining age 65 after 2024, whereas the normal retirement age is 65 for workers attaining age 65 before 2003. The generation now aged 3054 can therefore expect to receive larger lifetime benefits than will the generation younger than 30. This effect is somewhat counterbalanced, however, by the expectation that currently younger workers will, on average, have longer lives. Moreover, the increase in the payroll tax rate already legislated will have a larger impact on younger workers than older workers. In general, then, a larger social security system is likely to be viewed more favorably by today's retirees and near retirees, but less so by today's younger workers.  ? <Criterion 5. Encouragement of national saving. A small social security system, with lower tax rates and low benefit levels, would probably boost rates of national saving above what they would be in a larger system. This effect is likely to be small, however, reflecting differences in propensities to save of people of different ages and different life expectancies. Also, smaller benefits and taxes may not have a clearcut effect raising national saving. The uncertainty arises because lower benefits could cause workers to retire later, and facing less time in retirement, they might choose to accumulate less retirement saving while at work. This effect would be offset if workers"'e0*((@@/" anticipating smaller social security benefits decide that a larger retirement nest egg is needed. xMany studies have attempted to measure the net impact of social security on saving, but the studies remain inconclusive. For example, some believe saving has been substantially depressed by social security, implying that a cutback in benefits will boost private saving. Others show no consistent negative effect of social security on saving. The Panel concludes that a somewhat smaller social security system would have at best a  ?<small positive effect on saving.(Hendnote reference("endnote reference"%< ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#` hp x (#` hp x (#61. We also note that by construction, the alternative combinations of taxes and benefits considered here all result in the same level of actuarial balance. Therefore, by this construction, public saving is not noticeably affected over the 75 year forecast period, although there is some effect from the  ?<different impacts in timing of the changes in taxes and benefits.#X\ P:+Q2XP# %  ?` <Criterion 6. Financial integrity of retirement income systems. By this construction, the baseline comparisons in this section produce no difference in the social security system s actuarial balance over the next 75 years. Nevertheless additional considerations for financial integrity are important. One key concern, in the Panel's view, arises from the fact that the system s financial problems do not end in 75 years. Indeed, current projections show larger deficits in the distant future than in the near term. Thus with a rolling 75year horizon, the model replaces a financially sound year with a less sound year in each new projection for the foreseeable future. xFor this reason, the Panel believes that the cashflow patterns are important on a yearbyyear basis, particularly the contrast between early and late years, and beyond the traditional 75year projection period. The baseline benefit cut examined here affects a smaller fraction of benefits over the next 75 years (by exempting everyone 55 or over), while the baseline tax increase affects all earnings after 2002. As a result, a larger benefit decrease and smaller tax increase produces a better  ?<balance over the 75year horizon.(m8endnote reference("endnote reference"f= ?<#Xx6X@DQ/X@##Xx6X@DQ/X@#` hp x (#` hp x (##Xx6X@DQ/X@#62. Many on the Panel also believe that there is reason to be concerned about the responsiveness of the system to uncertainties that are projected but by no means accurately forecasted. Nevertheless, it is difficult to assess whether the system would automatically respond to future developments any more or less satisfactorily, comparing these different changes, or whether  ?<future modifications would be more or less straightforward.#X\ P:+Q2XP# f  ?<x Why Timing Matters : The baseline comparisons discussed above considered tax and benefit changes beginning in 2002. Here the Panel reviews four issues pertinent to timing. First is the question of timing of legislation, as opposed to implementation. Second, are the implications of alternative implementation dates. Third, is whether baseline changes should be phased in smoothly or abruptly. Fourth, the Panel considers adjusting benefits for those already retired. xThe Panel finds that any significant changes in social security benefits should be announced with considerable lead time to allow workers to adjust to the changes by adapting their retirement, saving and consumption plans. Implementation delays will require more substantial adjustments in benefits, taxes, or both, which may take a larger economic and political toll in the future. "'f0*((@@."Ԍ r ` hp x (#` hp x xDelaying legislation does not help people plan for the future. Rather, promptly legislated policy changes combined with delayed implementation of some benefit cuts and tax increases better fulfills this goal. Another reason to make changes sooner rather than later is the increased flexibility that comes from making changes when the national dependency rate is lowerwhen fewer retired individuals have their immediate livelihoods at stake. The urgency of prompt legislation is increased by the desirability of having implementation lags incorporated in the legislation. Gradual implementation also reduces the magnitudes of notches (different treatment of cohorts close in age) and the perception of unfairness that notches engender.  r ` hp x ` hp x (# xHaving argued that timing is of the essence, the question arises as to when specific changes could be made. Various alternatives include a delayed start of tax increases and/or benefit cuts, and benefit reductions by year rather than by birth date.  ?<xDelayed start of tax increases and/or benefit cuts: The baseline tax increases in the Panel's hypothetical scenario are scheduled to begin in 2002, while the baseline benefit decreases started for the cohort turning 62 in 2002. Although this date is relatively near for large changes, it is not too soon for either the financial needs of social security or for the beginning of a phasedin set of changes. The implications of delaying the start of the baseline changes by one or two decades are substantial, as is evident from Table III.1 which depicts alternate starting dates of 2012, and 2022. In each calculation, the proposal is to enact immediate legislation with delayed implementation; as a result, the calculations are undertaken for the current 75year planning period, not for a 75year period beginning at the date of change. With an increase in net revenues that happens quickly enough, little concern exists that the Trust Fund will run out of money in the middle of the 75year projection period. As startdates for implementation are pushed later and later, insolvency becomes a greater and greater risk. xEvaluating the effects of delayed implementation is facilitated using the criteria described above. The Panel confines its discussion to the halfandhalf mix of baseline changes.  ?`"<Criterion 1. Adequacy of retirement income. Delaying reform in social security taxes and benefits means that even larger benefit cuts will be required, making more widespread the problem of inadequate income replacement. In other words, delay worsens the impact of the reforms by the adequacy criterion.  ?'<Criterion 2. Insurance against income fluctuations. Decreasing the size of the social security program decreases its insurance"'g0*((@@/" component. Delaying the reforms produces a larger insurance cut for those affected.  ?X<Criterion 3. Avoidance of market inefficiencies. Labor market inefficiencies depend on the size of the implicit net taxes on work due to social security benefit and tax rules. Economic analysis implies that market distortions increase more than in proportion with the size of the distortion; in this instance, smaller cuts starting sooner have smaller labor market distortion effects. Conversely, delaying the date when benefits and taxes are altered makes necessary larger system changes, which in turn produces worse market inefficiencies.  ?( <Criterion 4. Distribution of taxes and benefits. Delaying a tax increase and making it larger later increases the relative lifetime taxation of those with more of their earnings occurring late in life. Because higher income people, on average, have steeper earnings profiles, such a tax is more progressive within a generation. Delaying a benefit decrease at the expense of having to cut benefits more later has no similar intragenerational effect because each cohort will experience benefit cuts depending on their year of birth, not on when they retire. xDelaying implementation of benefit and tax changes also changes the extent of redistribution between generations. Deferring a tax increase favors older workers near the end of their working lives, and means that younger workers will be subject to the higher tax rates over more years of their worklives. Similarly, delaying benefit cuts favors generations whose benefits are spared the cut and hurts later generations who receive larger cuts. Whether this pattern of redistribution is viewed positively or not depends on several things, including future real earnings growth prospects. If the prospects for future wage growth are good, younger workers will enjoy higher living standards than the older generations, however much social security benefits are scaled back. On the other hand, if the prospects for future wage growth are poor, the living standards of younger workers could be similar to or worse than those of older workers. Also, older workers have paid lower lifetime contributions to social security than younger workers can expect to pay in the future. Although real future wage growth cannot be predicted precisely, actual wage growth in the United States has been very low and for some groups even negative for almost two decades. xIn assessing lifetime losses, it is also worth noting that under current law the normal retirement age will be 67 for workers reaching age 65 after 2024, whereas the normal retirement age is 65 for workers reaching 65 before 2003. The generation now aged 3054 can therefore expect to receive larger real lifetime benefits than the generation younger than 30. This"'h0*((@@/" effect is somewhat counterbalanced, however, by the expectation that currently younger workers will have longer lives on average. Moreover, the increase in the payroll tax rate already legislated will have a larger impact on younger workers than older workers.  ?<Criterion 5. Encouragement of national saving. Because these alternative scenarios are designed to have similar impacts on the social security system's actuarial balance over the 75year accounting period, timing differences should not have a significant effect on public saving over this entire period. Of course delays could increase the extent of uncertainty about what reforms will eventually be passed, which might increase private saving for those losing confidence in the social security system. On the other hand, younger workers face larger, if delayed, increases in taxes which could depress their saving. Although existing research does not definitively prove that a smaller social security system would increase national saving, the Panel believes that delaying needed system reforms would probably not increase saving by very much.  ?<Criterion 6. Financial integrity of retirement income systems. The comparisons made here have been structured so they produce the same actuarial balance over 75 years. Nevertheless, some concerns arise for retirement system financial integrity. One is that the financial problems of social security do not end 75 years hence; indeed, some projections have shown larger deficits in the distant future than in the near term. Thus with a rolling 75year horizon, a financially sound year will be successively replaced with a less sound year in each new projection. The Panel is also concerned with the pattern of balances on a yearbyyear basis, particularly the contrast between early and late years. The later the change in benefits and taxes the greater the annual change has to be. As a result, delay increases the size of annual change and so improves this aspect of longrun financial integrity. xSome concerns also arise about the responsiveness of the system to uncertainties inherent in any forecast. In many recent years forecasts of the social security system's integrity have worsened. It is likely that earlier implementation of needed changes (and the fact that changes implemented sooner would be smaller) would make it easier to respond to future adverse developments, should they occur.  ?`"<xPhasein benefit reductions by year rather than birth date: For convenience, the discussion thus far has focused on changes in taxes and benefits that begin in 2002. For reasons of equity and macroeconomic impact, however, it may be more appropriate to implement payroll tax increases and benefit decreases in a series of small stepsusing a gradual phasein approach. An appeal of this gradual approach is that it would smooth the treatment of"'i0*((@@." members of nearby generations, both on an annual and a lifetime  ?<basis.(Aendnote reference("endnote reference"> ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#63. This discussion focuses on cutting real benefits for prospective retirees. A measure of benefit reductions for retired workers would spread the cuts more evenly, and hence warrants serious consideration. Such cuts could be limited by, for example, having them not exceed the COLA adjustment in any year. Obviously benefit increases less than the full costofliving adjustment make it difficult for people to adapt to  ?x<steadily declining real incomes.#X\ P:+Q2XP#   ?X< C. Alternative Means of Lowering Benefits xThis section describes six different methods of reducing the level of social security benefits. Each approach is assessed using the criteria developed earlier in this report, and compared with the baseline benefit cuts. Specifically, the proposals considered are as follows: x"Reducing PIAs of highwage workers; x"Raising the normal retirement age; x"Adjusting automatic costofliving adjustments; x"Meanstesting social security benefits; x"A doubledecker system; and x"A doubledecker system with means testing. These changes all have at least one feature in common: they will reduce social security outlays below the level that would occur if the current benefit formula and eligibility conditions were left unchanged. The earlier discussion focused on a comparatively neutral form of benefit reduction a proportional reduction in the benefits for all future retirees. The remaining discussion focuses on less neutral reforms reforms that will impose larger benefit reductions on some classes of new retirees than on others. xThe Panel's evaluations consider all but one of the criteria enumerated above. The Panel does not ask how the financial integrity of social security is affected by these reforms, because each of the changes is conceptualized so as to achieve the same reduction in future outlays and hence strengthen the system s integrity. For example, the Panel's evaluation of an increase in the retirement age should be conceptualized as follows: compared with a proportional reduction in social security benefits, what are the effects of a retirement age increase that achieves the same reduction in future outlays over the 75year planning horizon?  ?@<x Reducing PIAs of highwage workers : One way to maintain benefit adequacy while reducing benefit outlays is to cut PIAs more for highwage than for lowwage workers. This change in benefits could be accomplished by reducing the earnings replacement rate on AIME above the first or second bend points in the PIA formula, while holding the earnings replacement rate below the first bend point unchanged. The earnings replacement rates on are now 90 percent, 32 percent, and 15 percent. A change might lower the 32 percent to 30 percent, and/or the 15 percent to 10 percent. "'j0*((@@."ԌxAn alternative that might accomplish the same goal is to reduce the dollar amount of the bend points in the PIA formula, for example by linking the annual adjustment in the bend points  ?X<to changes in prices rather than changes in economywide average  ? <wages. Another alternative might be to raise the formula bend points in line with the lesser of price changes and economywide wage changes; in this event social security benefits would never be higher and would frequently be lower than under the present system. The actual amount of future benefit reduction would be uncertain, however, and would depend on the exact future relationship between price and average wage changes. xIn the following discussion, the Panel points out the similarities and differences between the two simple methods of reducing PIAs of highaveragewage workers, one that reduces the earnings replacement rates in the PIA formula and the second that reduces the bend points in the formula.  ?<Criterion 1. Adequacy of retirement income. Both approaches to cutting benefits fully protect the wellbeing of retired workers who had very low lifetime wages, most of whom probably have low retirement incomes. But some low/averagewage workers are members of highincome families (for example, the retired spouses of highly paid workers). Neither method protects the benefits of workers who may qualify for average or above average benefits but who have low retirement incomes. On balance, both methods should be preferred to a proportional benefit reduction if the goal is to provide a minimally adequate retirement income.  ?<Criterion 2. Insurance against income fluctuations. Both methods provide better insurance protection for workers earning less than average, and both provide worse insurance protection for higher paid workers.  ?<Criterion 3. Avoidance of market inefficiencies: Both proposals increase labormarket distortions compared with a proportional PIA reduction for all workers. In this respect, reducing the earnings replacement rate(s) above the first or second bend points probably introduces more distortion than the proposal to reduce the bend points. xDistortions are increased relative to a proportional benefit reduction because highwage workers face a heavier net tax on their lifetime earnings. The marginal net tax imposed on workers by the social security system is (approximately) the difference  ?(#<between the payroll tax they pay on additional earnings and the  ?#<present value of the extra lifetime benefits they and their survivors will receive as a result of the increase in their AIME. How this net tax changes with additional earnings is the marginal net social security tax, which is relevant for many decisions. Fullcareer single workers with AIME above the second bend point currently receive an additional $0.15 in monthly PIA benefits for"'k0*((@@/"  ?<each additional dollar in AIME.(5mendnote reference("endnote reference"b? ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#64. This abstracts from recomputation of benefits that might  ?<take place with increased work.#X\ P:+Q2XP# b A reduction in PIAs that is targeted on highwage workers would reduce the increment in benefits that highwage workers receive as a result of boosting their AIME. It would increase the marginal net tax on their earnings by more than a proportional reduction in PIAs, thus  ?<introducing a larger distortion in their laborleisure choice.(nendnote reference("endnote reference"[@ ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#` hp x (#` hp x (#65. It should be noted that low earners are probably not much more sensitive to incentives than are high earners. Also people pursuing higher paying careers (by investing in education, for instance) change their AIME over a range, which means they face an averaging of the marginal net tax discussed above. The difference between the various proposals is smaller under this  ?<scenario.#X\ P:+Q2XP# [ xA reduction in the bend points introduces a similar distortion in the laborleisure choice, but for a narrower class of workers. If the second bend point is reduced, for example, fewer workers would see their PIAs raised by $0.32 for each $1.00 increase in AIME; more workers would see their PIAs raised by only $0.15 for each $1.00 increase in AIME. The net social security tax faced by the affected workers would rise substantially. Reducing the bend points increases the net social security tax by a very large amount for each affected worker, but the tax rate is increased for a smaller number of workers than would be the case if the earnings replacement rate above the first and/or second bend points were cut. xOf course, a proportional reduction in PIAs raises the net marginal social security tax faced by all workers who expect to collect social security benefits, including lowwage workers. The distortion is larger and more economically significant if the PIA reductions are concentrated on highwage workers, however. This is true for two reasons. First, many lowwage workers do not presently pay a net marginal social security tax at all. Many workers with average lifetime earnings below the first bend point receive a net subsidy from social security for each $1.00 increase in AIME they will receive more additional discounted benefits than they will pay in additional payroll taxes. A proportional reduction in their PIA will reduce the net subsidy; it will not impose a heavier marginal tax. Second, the amount of the labor supply distortion rises faster with a 1 percentage point change in taxes when the tax rate is initially high than it does when the tax rate is low. Highwage workers currently face higher net tax rates than do lowwage workers. A reduction in PIAs that is concentrated on highwage workers would yield a larger increase in the net tax increase on those workers than a proportional PIA reduction for all covered workers. Reducing the PIAformula bend points is slightly favored over reducing the earnings replacement rate above the second bend point because the top net social security tax is left unchanged by the first reform. A larger distortion does not necessarily imply that highwage workers will work less than they would if PIAs were reduced proportionally for all workers. They might delay their retirement because they are forced to accept lower social security pensions than they could obtain if PIAs are reduced proportionally. (Lowwage workers on the other hand, would probably retire earlier under the disproportionate cuts, because their benefits would not change.) "'l0*((@@/"Ԍ ?<Criterion 4. Distribution of taxes and benefits. The previous discussion makes clear that either alternative to a proportional reduction in PIAs favors lowaveragewage workers at the expense of highaveragewage workers. So either alternative redistributes more within a generation than does the baseline case. Between generations, all three proposals favor cohorts currently age 60 or older, in comparison with the generations under age 60. Neither proposal, however, favors the older generation any more or less than the proposal to impose proportional reductions in PIAs on future generations of  ?<retirees.(endnote reference("endnote reference"SA ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#66. Arguments for and against phasing in a benefit reduction are independent of whether the benefit reductions will be imposed proportionately on all workers or disproportionately on workers with high lifetime earnings. If the changes occur gradually instead of immediately, the burden on the generation under age 30 would rise, and the burden of the generation age 3059 would decline. (In order to achieve the same change in the 75year cost rate, workers under age 30 will have to accept a larger reduction in their benefits than they would face if benefits are immediately reduced, including benefits payable to the generation aged 3059.) The argument against implementing a benefit reduction immediately is that workers who are nearing retirement age are offered little opportunity to adjust their saving or work behavior in response to a significant change in their retirement income prospects. S  ?` <Criterion 5. Encouragement of national saving. A reduction in PIAs targeted on highwage workers might be expected to boost saving if those affected offset lower promised social security retirement benefits with higher personal saving. Moving the PIAformula bend points places more of the benefit burden on lower earners, as compared with cutting the lowering the earnings replacement rate above the first or second bend point.  ?<x Raising the normal retirement age : The most common proposal to reduce future social security benefits is to raise the normal retirement age (NRA), that is, the age at which full social security retirement benefits (100 percent of the PIA) can be claimed. The NRA is already scheduled to increase from the current age 65 to 66 (for those reaching age 62 in 2005) and then to 67 (for those reaching age 62 in 2022). What is under discussion here is an additional rise in the NRA. (A proposal to raise the age for early retirement benefit eligibility is discussed in the next section along with eligibility ages for a variety of programs including SSI and DI). xLifting the retirement age while leaving the early entitlement age (EEA) unchanged produces almost exactly the same effect on retired workers' social security benefits as a proportional reduction in PIAs. (Some differences exist between the two proposals for workers who attain the NRA before 2009. But for workers who attain the NRA after 2008, when the delayed retirement credit will equal 8 percent per year of delay, the difference between the two methods of reducing benefits is small.) xTo see why, compare a 7 percent reduction in the PIA with a oneyear increase in the NRA for workers who attain age 66 after 2009. If the PIA is reduced 7 percent, all workers claiming benefits will receive a monthly benefit check that is 7 percent smaller than promised under current law. If, instead, the NRA is raised by one year but the PIA formula is left unchanged, persons claiming benefits at age 66 in 2010 will not be able to receive an unreduced pension. Instead, their monthly benefit will be subject to an actuarial reduction and will be approximately 6.67 percent less than their benefits would have been if the NRA had"'m0*((@@/" remained unchanged. The percentage reduction in benefits below what benefits would be under current law is similar to the reduction when PIAs are reduced 7 percent. Similar calculations for workers who choose to claim benefits at ages 62, 65, and 70 show that advancing the retirement age by one year has an effect on benefits that is similar to a proportional reduction in PIAs. xUsing the Panel's evaluation criteria, raising the retirement age may be seen to be as attractive, or as unattractive, as proportionally reducing PIAs. Two potential noneconomic differences arise between the two proposals. First, raising the NRA signals to workers that the same monthly benefit can be obtained by postponing retirement, encouraging some to delay retirement rather than accept a lower pension. Employersponsored pension plans might also be modified to encourage retirement at the later age; on the other hand they might be refashioned to encourage earlyouts. Second, many Americans may find increases in the retirement age to be more understandable and perhaps "fairer" than equivalent reductions in PIAs, given increases in life expectancy. The message would be that the benefit amount is appropriate, but the timing no longer is. xAn important difference between the two proposals is that under current rules, DI benefits received at any age are linked to the worker's unreduced PIA. If the PIA formula were cut in relation to the AIME, DI benefits would also be reduced. By contrast, raising the NRA while leaving the PIA formula unchanged would not change DI benefits. Instead, early retirees' OASI monthly benefits would fall relative to DI benefits. As a result, older workers below the NRA and in poor health might be more likely to apply for DI benefits, preferring a (relatively larger) DI benefit as compared with a (reduced) OASI benefit. Such workers would be better off if the retirement age were raised and the PIA formula left unchanged. Reducing PIAs versus raising the normal retirement age would therefore affect mainly retirees who could meet the disability criteria of the DI program, and they would prefer the latter option. People in poor health who did not meet the eligibility test for DI would have to accept a reduced early OASI pension, which would be true if the PIA were reduced or the NRA were increased. A general point worth emphasis is that for a given revenue gain, larger DI cuts imply smaller OASI cuts and vice versa.  ?!<Criteria 1 and 2. Adequacy of retirement income and income  ?`"<insurance. Raising the NRA while leaving the EEA unchanged contrasts with the baseline scenario of reducing PIAs uniformly. Workers in good health, ineligible for DI benefits, would receive less adequate protection from the social security system at some ages and better income protection at other ages. Insurance protection against the hazard of very long life is improved, because workers are discouraged from claiming early, but low, OASI benefits. Workers in poor health who can meet the DI"'n0*((@@/" eligibility test receive disability benefits unaffected by the NRA; under the baseline scenario, as pointed out above, DI benefits would be reduced by a change in the PIA formula. Hence, raising the NRA decreases income insurance for workers who do not meet the test. An issue about which relatively little is known is whether DI recipients are needier than OASI recipients and therefore less well able to absorb a cut.  ?@<Criterion 3. Avoidance of market inefficiencies. In terms of the proposal's effect on the savingconsumption choice, raising the normal OASI retirement ages would require people to rely more on their own earnings, pensions, and private saving to finance consumption if they desired to retire prior to their NRA. In addition, people might work longer, offsetting the need to increase private saving. On the other hand, the increased appeal of DI benefits may produce a more distorted laborleisure pattern: workers in marginal health might reduce their labor supply so as to enhance their eligibility.  ?<Criterion 4. Distribution of taxes and benefits. Advancing the NRA reduces lifetime benefits paid to workers and survivors who die young, and increases the amount of lifetime benefits available to workers and survivors who live to an advanced age. On the whole, this change favors workers with high lifetime incomes relative to lowwage workers, because higher earners on average live longer than lower earners. Nevertheless, raising the retirement age protects DI payments which helps lower earners in as much as disabled workers tend to have low average and lifetime incomes. As a result, the protection of DI payments tends to favor some workers who have low wages as compared with a plan that reduces PIAs.  ?<Criteria 5 and 6. Encouragement of national saving and  ?<strengthening retirement income systems. Compared with a proportional cutback in PIAs, increasing the normal retirement age may boost private and national saving slightly. If DI benefits were made more accessible, this could reduce the need for precautionary saving. To the extent that private and employersponsored pension saving would be raised to pay for retirement prior to the NRA, other nonOASI elements of the retirement income system might be strengthened.  ? <x Adjusting automatic costofliving adjustments : Some policy makers have recently proposed reevaluating the costofliving (COLA) mechanism that links annual increases in social security benefits to changes in the Current Price Index for All Urban Wage Earners and Clerical Workers (CPIW). Critics of the price index believe it is inappropriate as an index for social security benefits because it overstates inflation. One proposed remedy for this bias is to limit social security benefit adjustments to the annual change in the CPIW minus 1 or 2 percentage points. A preferred remedy, in the Panel's view, is to seek technical"'o0*((@@/" improvements in the measurement of price changes and their influence on consumer wellbeing, with particular attention to the market basket consumed by the elderly. xReform in the Cost of Living Adjustment (COLA) procedure has also been suggested as a method of reducing social security outlays, regardless of the proposal's technical merits. In the following discussion it is assumed that COLA reform occurs in such a way that benefits fall in real terms, and that the wage indexing method used to calculate workers' PIAs is not altered; only the adjustment in benefits in force is assumed to be affected. Once again, this method of reducing benefits is compared with a proportional reduction in PIAs that achieves the same reduction in the 75year time frame used by social  ? <security.(endnote reference("endnote reference"B ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#"endnote reference"` hp x (#` hp x (#67.(endnote reference( Some observers have proposed that COLA adjustments be trimmed or deferred for one or two years, as a method of temporarily reducing benefit growth. Current and future retirees would never make up the loss in benefits caused by the temporary suspension of full indexation. This reform is not equivalent to proportionally reducing PIAs, however. Indexation of AIME up to age 60 is based on wages, not prices, so that anyone younger than age 60 when the COLA is deferred will be unaffected. A temporary suspension of COLA adjustments therefore imposes benefit cuts on  ?<the retired generation.#X\ P:+Q2XP#  The remainder of this section assesses the comparative effects of permanently reducing COLA adjustments below the amount that would be needed to keep the purchasing power of benefits from falling.  ?<Criterion 1. Adequacy of retirement income. Decreasing the social security COLA reduces benefit adequacy for workers and survivors who collect benefits for many years. Of course, if the real social security benefit declined far enough, some lowincome beneficiaries would become eligible for SSI and food stamp benefits. The adequacy of retirement income would be improved during the first few years of disability or retirement, when real benefits would be somewhat higher than they would otherwise be, because the PIA formula was not cut. Poverty and inadequate income appear to be more severe problems among the longlived than for new retirees, so it is plausible that COLA reductions will increase poverty more, especially among widows, than would a proportional cutback in PIAs that achieved the same total benefit reduction. COLA reduction thus is rated as harmful to the goal of ensuring adequate retirement income, except to the recently retired. x  ? <Criterion 2. Insurance against income fluctuations. Economists believe that real annuities are an efficient form of insurance for retirees. Reducing the indexation of social security benefits makes retirees' income more subject to fluctuations in real value, which reduces retirement wellbeing.  ? <Criterion 3. Avoidance of market inefficiencies. Because social security is the only CPI indexed annuity available to most workers, those who understand its value have an incentive to work longer to receive more of this valuable insurance. Therefore reducing the indexation of benefits makes this insurance less valuable, and is less of an encouragement for continued work. Workers who are shortsighted might be more tempted to myopically retire with initially high but eroding benefits. Reducing benefit indexation induces more saving among those who have reason to believe that they face a long retirement period, unless"'p0*((@@/" they would be willing to accept lower consumption if and when they survived into their 80s and 90s. Conversely, workers in poor health who do not expect to survive much past age 70 might save less.  ?<Criterion 4. Distribution of taxes and benefits. Within a generation, cutting COLAs rather than PIAs probably makes the OASI system more redistributive toward the lifetime poor since reducing COLAs increases the relative benefits paid to those who die young and highwage workers on average live longer than lowwage workers. Focusing on betweengeneration changes, if COLAs were reduced immediately, some of the burden would be borne by the current retired generation which has contributed much less to the social security system than it will receive in benefits. Because the proposal achieves some costsaving from benefit reductions from current elderly, younger cohorts will not have to accept as large a loss in benefits to achieve a target reduction in the 75year cost rate. In this way, the intergenerational consequences of this proposal differ from those of the proposals discussed earlier; it is the only one to reduce payments (in real, not nominal terms) paid to people now collecting benefits. x  ?0<Criteria 5 and 6. Encouragement of national saving and  ?<strengthening retirement income systems. Compared with a proportional cutback in PIAs, reducing COLAs may increase private and national saving slightly, thus strengthening nonOASI elements of the retirement income system somewhat.  ?<x Meanstesting social security benefits : Another way to reduce social security outlays would be to introduce an explicit means test, which restricts benefits to persons with low or moderate retirement incomes (or wealth). Highincome (or high wealth) retirees and survivors would be prevented from receiving benefits or would receive substantially lower benefits than they presently do. One option is to phase out social security as family income (or wealth) rises above a certain threshold, say $50,000 per year. For example, benefits might be reduced by 40 cents for every dollar of income in excess of the threshold. A person who is entitled to $10,000 in social security benefits under current law would receive just $6,000 in benefits if countable income apart from social security was $60,000. The same person would receive no social security at all if countable income exceeded $75,000. xAnother option would be to permit people to collect a specified percentage of their entitlement regardless of income, with the remainder of their social security benefits subjected to a means test. This option reduces rather than eliminates benefits available to highincome families. xSome people regard benefit taxation as a means test. Up to 85 percent of social security benefits are already subject to"'q0*((@@/" income taxes, so it would be difficult to achieve significant additional tax revenue from this source unless benefits received by lower income families were also taxed. In other words, further increases in social security taxation will come at the expense of moderate income families rather than the welltodo. The remainder of this discussion assumes that meanstesting will involve reductions in social security benefits to persons or families with high incomes apart from their social security benefits.  ?<Criterion 1. Adequacy of retirement income. The adequacy of retirement income will be substantially improved in comparison with a proportional PIA reduction that achieved the same reduction in the 75year cost rate. By definition, the only people who suffer losses in benefits under a means test are those with high incomes from other sources. In fact, no proposal the Panel considered provides such strong protection against low retirement income, even the proposals that disproportionately reduced PIAs among workers with high lifetime wages. Under those proposals, workers with moderately high lifetime earnings but low retirement incomes from other sources could suffer large proportional reductions in retirement incomes. x  ?<Criterion 2. Insurance against income fluctuations. If a higher PIA is accomplished by implementing means testing, all workers who are not subject to the means test would receive better insurance protection. Those who would be means tested because of high levels of income or assets receive less income insurance, though a meanstested benefit still provides insurance to affluent workers if they live to a very advanced age and exhaust their retirement assets.  ?<Criterion 3. Avoidance of market inefficiencies. The effects of means testing on the laborleisure choice are complicated and depend on a worker's circumstances and age. Some workers would anticipate receiving higher social security benefits under a meanstested system, and consequently would work less and retire earlier than under the current system. On the other hand these same workers might pay lower net lifetime taxes under social security, because their net benefits are higher than they would be if PIAs were proportionally reduced. Consequently their labor supply decisions as young workers may thus be less distorted than they would be if PIAs were cut. People having assets that place them at or above the means test threshold face a sharply higher tax, and on balance this higher tax could reduce work over the entire lifetime. Additionally, the extent of labor supply distortions would probably be worse, since high earners are already more heavily taxed by social security over their lifetimes and means testing exacerbates this effect. xOn the savingconsumption choice, the most serious problem created by a means test is that it strongly discourages employees"'r0*((@@/" from saving in any form that generates periodic repoted income. People who anticipate being near the threshold might reason that it is better to consume while young so as to fall below the asset threshold when old. Also people might be induced to save in assets that are less likely to count in an asset test, or which are more easily hidden. xOne group would certainly face stronger incentives to save under this reform: workers who believe they will have too much nonsocial security income as retirees to qualify for social security benefits. They would have to save more than they now do to achieve the same level of retirement consumption. Their saving decisions will be less distorted by the availability of social security than they would be if PIAs were proportionally reduced.  ? <Criterion 4. Distribution of taxes and benefits. Within generations, a means test clearly benefits lowincome retirees and survivors who would not suffer benefit reductions. Highincome retirees and survivors, by contrast, would lose all or a substantial portion of the benefits otherwise available to them. xIf means testing were imposed soon, older retirees would be affected in addition to those from younger generations. Such a policy contrasts with most of the other options discussed in this Report, which are structured to affect only the younger cohorts. Whether such targeting makes a proposal politically more appealing is difficult to say. To the extent that the proposal achieves some saving from benefit cutbacks imposed on the current elderly, future generations will not have to accept as large a loss in benefits in order to achieve a target reduction in the 75year cost rate. In any event, comparing a meanstested option implemented across the board with a proportional reduction in PIAs that affects only future retirees, the means test on social security benefits allocates the burden of social security cost saving more evenly across generations.  ?<Criterion 5. Encouragement of national saving. Compared with a proportional cutback in PIAs, a means test on social security benefits has ambiguous effects on private and national saving. The proposal would depress employer pensions and assets that count in an asset test. Betteroff workers who believe they will be eliminated from the meanstested payment will seek to save more privately to achieve the same level of retirement consumption. On the other hand, people expecting to be near the threshold might reason that it is better to consume rather than save while young, so as to fall below the asset threshold in retirement.  ?H&<Criterion 6. Strengthening the financial integrity of retirement  ?'<income systems. Compared with a proportional cut in PIAs, means testing would tend to weaken other nonOASI elements of the"'s0*((@@/" retirement income system. In addition, means testing might seriously erode political support for the system, although it is unclear whether this would be any more problematic in a meanstested regime than in a regime that cut PIAs at the top.  ?<x A doubledecker system : The final set of benefit reforms the Panel considered involve fundamental restructuring of the PIA formula, restructuring that would probably also require significant changes in the role of SSI in providing income to the old and disabled. Instead of retaining the structure of the current PIA rules, these alternative formulas provide a fixed basic benefit (perhaps subject to a means test, and probably varying with years of service) and a constant rate of increase in the social security benefit for each $1 increase in a worker's AIME. xThree examples of a nonmeanstested benefit formula help illustrate the essential idea of benefit restructuring under this heading. Consider three hypothetical ways of determining the retiree s full benefits:  ?h<x1)` ` B1 = $400 + (0.22 x AIME);  ?<x2)` ` B2 = $550 + (0.15 x AIME); or  ?<x3)` ` B3 = $800. Under each option, part or all of a worker's benefit consists of a flatrate component. In two of the formulas, (1) and (2 ), the benefit also includes a component that is strictly proportional to the worker's AIME. Formulas such as 1 and 2 are sometimes referred to as "doubledecker" social security systems. (The first "deck" consists of the flatrate pension; the second consists of a benefit that is strictly proportional to covered earnings.) Formula 1 is the most generous formula for workers with high lifetime incomes, because it is the formula where the benefit rises fastest with increases in AIME. Formula 3 is the most generous to workers with low lifetime incomes, because the entire benefit consists of the flatrate component. Of course, formula 3 is also least advantageous for workers with high lifetime earnings because they receive benefits that are no higher than those received by lowwage workers.  ?!<Criterion 1. Adequacy of retirement income. Available evidence is sparse, but suggests that adequacy of retirement income under doubledecker formulas could improve relative to a proportional cutback in PIAs that achieved the same reduction in the 75year cost rate. The alternative benefit formulas outlined here ensure all retirees a fixed minimum benefit, regardless of their average earnings, so social security would assure disabled and retired workers and survivors at least a minimal income under all circumstances in which they become entitled to receive benefits. "'t0*((@@/" The particular parameters of the system matter a great deal, however, and those at the bottom may or may not do better under a doubledecker system, depending on how it is constructed. Replacement rates for low earners would probably increase (depending on the size of the bottom deck), while replacement  ?<rates for higher earners would likely fall.(endnote reference("endnote reference"C ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#` hp x (#` hp x (#68. In order to determine how a double decker system would differ from a change in the PIA formula that tilts toward low earners, it will be necessary to identify precisely the type of low earners helped more or less by the flat amount. Research is required to indicate whether the first deck would raise benefits primarily for shortcareer or lowage fullcareer workers, and what the effect would be of linking years of service under the system to entitlement to the full flat amount.  X-#X\ P:+Q2XP#њ x  ?x<Criterion 2. Insurance against income fluctuations. The double decker benefit formulas would offer better insurance protection to workers with low average earnings, unchanged insurance protection to many workers with lowtomoderate earnings, and worse protection to workers with moderate or high lifetime earnings. The current PIA formula creates a strong link between benefits and average earnings, so that earnings losses after disability, death, or retirement are replaced by benefits that offset a large percentage of the loss, especially in the case of low and moderateincome workers. Formula (3) completely severs the link between benefits and previous earnings; formulas (1) and (2) retain the link but in many cases reduce the percentage of earnings loss that is replaced by benefits.  ?h<Criterion 3. Avoidance of market inefficiencies. A twotier reform such as this would probably reduce work incentives for lowwage employees when they reach the early or normal retirement age, and also would impose a sharply higher "net" social security tax on earnings when young. This point can be illustrated in the hypothetical case of a worker whose AIME is below the first bend point. On the margin, a $1 increase in AIME could boost his PIA by $0.90 under the current formula, by just $0.22 under alternative formula (1), by $0.15 under formula (2), and by  ?p<nothing at all under formula (3).(uendnote reference("endnote reference"D ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#"endnote reference"` hp x (#` hp x (#69. (Pendnote reference( Few individuals with a full work career will have an AIME in the 90 percent band, and even for such a person, additional work at higher earnings levels might displace low earnings years so that benefits would be increased with delayed retirement. The point is that additional work generates a lower net increase in benefits under the double decker plans than under alternative  ?<plans.#X\ P:+Q2XP#  Thus, in comparison with the current formula, all three of the alternative formulas provide a smaller reward on the margin for increases in work effort and earnings when workers are young, implying that work decisions of low and moderatewage workers could become more distorted than if PIAs were cut. xHighwage workers could anticipate receiving substantially lower social security benefits under these doubledecker options that will probably increase work incentives around the early and normal retirement ages. These workers may also face higher "net" lifetime taxes for social security (or receive lower net benefits), depending on the alternative formula selected. If the alternative formula offered a marginal replacement rate that is much below 15 percent, the "net" lifetime tax on social security will be higher than it would be if PIAs were proportionally reduced. This is certainly the case under alternative formula (3), where the marginal replacement rate is 0 percent. On the other hand, if the marginal replacement rate in the reformed formula is 15 percent or higher, as it is under formulas (1) and (2), the "net" tax imposed by social security (on the margin) would actually be lower than under a reform with proportional PIA"'u0*((@@/" reductions. Labor supply distortions for highwage workers could be reduced under these circumstances. xOn balance, the alternative social security benefit formulas probably lead to larger labor market distortions overall. A larger fraction of social security contributions is used to finance fixed basic benefits that are not linked to workers' contributions into the system. For a much higher percentage of workers, extra contributions into the system are not matched by increases in future benefits that are as large or larger than the increase in contributions. A larger fraction of the social security tax will therefore represent a pure tax on labor earnings, increasing the labor supply distortions caused by the program. xIn terms of the savingconsumption choice, little reason exists either to favor or oppose the alternative PIA formulas in comparison with a proportional reduction in all PIAs.  ?<Criterion 4. Distribution of taxes and benefits. The discussion thus far suggests that the doubledecker formulas favor low to averagewage workers at the expense of highaveragewage workers, in comparison with the proposal to reduce PIAs proportionally. Also, each of the benefit formulas favors the current generation that is age 60 or older, relative to the generations under age 60, but neither proposal favors the older generation any more or less than the proposal to impose proportional reductions in PIAs on future generations of retirees.  ?<Criterion 5. Encouragement of national saving. Compared with a proportional cutback in PIAs, these alternative formulas might boost private and national saving. Little evidence exists on the question of whether workers with high lifetime earnings would save more in response to a change in social security benefits than would those with low lifetime earnings. If they did, the highly paid would raise their pension and other saving assuming the reduction in retirement benefits is concentrated on highwage workers. Among the three, formula (3)least generous to highwage workersmight induce the largest increase in national saving.  ? <Criterion 6. Strengthening the financial integrity of retirement  ? <income systems. To the extent that pension saving was increased to offset the redistributive formula, other nonOASI elements of the retirement income system might be strengthened. As before, implementing a doubledecker system could erode the system's political support, although it is unclear whether political difficulties are any greater than those under a meanstested regime.  ?'<x A doubledecker system with means testing : The doubledecker PIA formulas considered above (formulas 1 and 2) each"'v0*((@@/" include a flatrate component and a second component proportional to the worker s AIME. To reduce the net cost of the flatrate component, this portion could be subject to a means test. Retirees with large assets or high current incomes apart from social security might receive reduced flatrate pensions or no flatrate pensions at all. The earningsrelated component of their pension would not be subjected to a means test. xSuch a partial meanstesting scheme would substantially reduce the cost of the flatrate component of the social security benefit. Because fewer retirees would receive this portion of the benefit (and many fewer would receive the full amount of the flatrate component), at any given contribution rate for social security, the system could afford to offer either a larger flatrate component in the benefit formula or a larger earningsrelated component. If the flatrate component were made more generous, the income adequacy of social security benefits would be improved. If the earningsrelated component were made more generous, the insurance value of social security would be improved for highwage workers. xAn important disadvantage of meanstesting has already been mentioned. The labor supply decisions of low to moderateincome retirees would be much more distorted than they are under the current system, and the labor supply decisions of low to moderatewage workers would also be more distorted. Work and retirement decisions of highwage and highincome workers would be less distorted than under the current system, if the earningsrelated component of social security benefits gives workers at least $0.15 in extra benefits for each $1 in extra AIME; otherwise, the decisions of highwage and highincome workers would be at least as distorted as under the current system. The saving decisions of workers would be more distorted than they are under a proportional reduction in PIAs. The larger the flatrate component in the social security benefit formula, the larger these distortions would be. As a result, national saving would probably fall. xThe current system is similar to the meanstested doubledecker system in one crucial respect. An elderly or disabled household is eligible for a SSI and food stamp benefits if it has low enough assets and income, and these benefits are phased out as the household's income rises. The current system thus enjoys some of the same advantages that are claimed for the meanstested doubledecker system. In particular, it ensures minimal incomes to lowincome people who are old enough or disabled enough to claim benefits. Combined SSI, social security, and food stamp benefits currently amount to about 85 percent of the poverty line for a person 65 or older living alone, and more than 100 percent  ?H&<of the poverty line for an aged couple.(W endnote reference("endnote reference"E ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#"endnote reference"` hp x (#` hp x (#70. (pendnote reference( SSI benefits are reduced dollar for dollar for social security payments, after a $20 per month disregard; U.S. House  ?<1994, Tables 69 and 610.#X\ P:+Q2XP#  "'w0*((@@."ԌxDifferences arise between the current system and a meanstested doubledecker system, of course. SSI and food stamps are currently financed out of general revenues rather than payroll taxes. SSI benefits are phased out quite rapidly as an eligible person's income from other sources (including social security) rises. For that reason, the percentage of old and disabled persons who are eligible for SSI benefits is small. In addition, people who may be eligible for these benefits must apply for them in a process separate from their application for social security. As a result, many eligible people do not apply for SSI or food stamps. (Just over half of the elderly poor participate in the SSI program; some nonparticipants may be assetineligible. See U.S. House of Representatives 1994, pp. 241242) In a meanstested doubledecker system, the application for the flatrate and earningsrelated components of social security benefits would occur at the same time and could be combined., resulting in a substantially higher take up rate. Finally, the earningsrelated component of the current social security formula is more complicated than the earningsrelated component of the formulas we have described.  ?h< D. Alternative Means of Increasing Revenues xAn earlier subsection described how social security's long run fiscal gap might be closed either by raising taxes or reducing benefits that is, by expanding or reducing the overall size of the system. Here the Panel considers other methods of increasing revenues to the social security trust funds, evaluating them relative to the simple increase in the payroll tax using the criteria described at the outset. Three proposals are considered: x(1) Raising the earnings limit on which payroll taxes apply;  ?<x(2) Expanding the definition of taxable income; and 0 x(3) Introducing additional general revenues. xIt should be noted that evaluation Criteria 1 and 2 are not particularly relevant in this discussion. Adequacy of retirement income and insurance against income fluctuations are primarily benefit and not tax issues, although higher net lifetime taxes do lower disposable income and therefore diminish households' ability to save for retirement. In addition, the Panel is considering proposals that raise the same amount of revenue over the 75year projection period, so they will have the same effect on social security financing (Criterion 6). For this reason, the  ?(#<Panel concentrates here on the remaining three criteria.(+endnote reference("endnote reference"gF ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#"endnote reference"` hp x (#` hp x (#71.(xendnote reference( This discussion depends on whether workers view OASDI contributions as "taxes" or as mandatory savings to be returned as benefits later. For some (for example, a working woman planning to claim 50 percent of her husband's benefit), the contributions are pure taxes with no payoff later and would be expected to have the same effects as any other tax on earnings. For many past, current and nearfuture social security recipients, however, what may have looked like taxes turned out to be subsidies, because the benefits generated were (and still are) multiples of the present value of what the workers and their employers contributed. Nonetheless, the workers might have viewed these taxes as taxes when they paid them, with the attendant economic effects. Calculations of rates of return can help show how workers should have viewed or should now view these taxes, but not how they actually did or do view them. The critical point here is that an increase in taxes, holding benefits constant, is presumably still viewed by all as a tax  ?H <increase.#X\ P:+Q2XP# g  ?$<x Raising the earnings limit on which payroll taxes apply : The proposal examined here is an increase of the taxable payroll ceiling instead of an increase in the payroll tax rate as described in the baseline case. In 1995, social security OASDI payroll taxes were paid by employers and by employees (6.2"'x0*((@@/" percent each) on the first $61,200 of each worker's earnings. After this maximum taxable amount, raised annually by the rate of increase of average wages, the OASDI payroll tax rate drops to zero. The HI tax of 1.45 percent each is paid on all earnings. The selfemployed pay both parts (12.4 percent for OASI; 2.9 percent for HI) on the same taxable amounts. xThe proposal here is to raise the real level of income subject to the taxable limit or to eliminate the tax ceiling altogether. Unless there were a simultaneous change in the benefit formula, increasing the taxable limit would also increase the social security benefits eventually paid to those affected, because benefits are based on average covered earnings, which would rise under this proposal. Benefits would not rise in the baseline case, because covered earnings would not change. Specifically, the Panel compares two payroll tax changes: raising the baseline increase from 6.2 percent (for employer and employee) to something higher for all those earning less than the maximum taxable amount, versus an increase from zero to 6.2 percent (for employer and employee) on all those earning more than the old and less than the new maximum taxable amount (or on everyone earning more than maximum taxable amount, were the ceiling eliminated altogether). How do these two proposals differ?  ?<Criterion 3. Avoidance of market inefficiencies. The baseline case is a pure tax increase because it would not result in higher social security benefits later. In contrast, raising the taxable income limit would be only a partial tax, because individuals' AIMEs and therefore subsequent benefits would also rise. (Most of those paying the increment would be beyond the 15 percent bendpoint, where the return on additional earnings taxed is the lowest.) Moreover, because of the increase in future benefits under this taxable ceiling proposal, more taxes would have to be raised than under the baseline to have the same net effect on social security finances. xThe number of people affected at the margin differs significantly under the two scenarios. The baseline case increases the marginal tax rate of all those under the maximum taxable limit. Raising the taxable limit, on the other hand, affects only some of those over it at the margin. Currently, fewer than 15 percent of workers covered by social security earn above the maximum (Steurle and Bakija 1994). Some of them would have their marginal tax rates increased by the full tax rate (depending on how high maximum taxable earnings were raised), but far fewer than the 85 percent or more whose marginal tax rates would be increased by an increase in the payroll tax rate. Also, the magnitudes of the increases in marginal tax rates are very different in the two cases. Raising the taxable pay limit increases the marginal tax rate from 0 to 12.4 percent (assuming the employer's share is eventually taken from workers' pay). "'y0*((@@/" This increase is on top of the marginal federal and state taxes these individuals already pay, which tend to be at the high end, given the earnings levels of those affected. By contrast, in the baseline case, the increase in the marginal tax rate would be much more modest, and much of it would fall on those who face  ?<lower combined federal and state marginal tax rates.(8endnote reference("endnote reference"[G ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#` hp x (#` hp x (#72. But note that the difference in marginal tax rates for those two groups is less dramatic than it first appears, because the current OASDI rate (12.4 percent, if we count both halves) is not part of the marginal tax rate of those earning above the taxable limit; only the 2.9 percent HI tax is. Those under the limit face the full OASDHI tax (15.3 percent) at the margin, through  ?<generally lower state and federal taxes.#X\ P:+Q2XP# [  ?x< xThe tradeoff here is between a large marginal tax rate increase on a small number of high earners, and a smaller increase on a much larger number of low and middlelevel earners. Both of these tax increases will probably increase laborleisure inefficiencies for younger workers. Because the magnitudes of these effects are not firm, it is difficult is quantify the size of the two distortions and decide which of the proposals does more damage. Finally, for people near retirement, the two plans have different effects. Increasing the taxable pay limit reduces net earnings for some highly paid workers, while raising the tax rate lowers net earnings for all workers (though not at the margin for those earning more than the new limit). Both changes would be expected to induce earlier retirement, on net, although more would be affected by the baseline case.  ?0<Criterion 4. Equity of lifetime social security taxes and  ?<benefits. The distributional impacts of these two proposals are very different. The baseline plan, which raises the payroll tax rate, represents a pure tax increase on all those with taxable earnings since taxes rise but eventual benefits do not. Raising the taxable limit, by contract, is a tax on some or all of those earning more than the old taxable limit, even factoring in the eventual increase in benefits. Raising the taxable limit, therefore, is more consistent with the progressive nature of social security's tax and benefit structure. xOn the other hand, because of this progressive nature of social security, workers with high earnings are already more likely to be suffering negative expected returns on their social security contributions now and in the future. An increase in the taxable limit would make the returns even more negative, further weakening the link between contributions and benefits for the upper income population. To make the same point at the other end of the income distribution, low wage workers are more likely to receive positive rates of return, both because of the social security benefit structure and because of the Earned Income Tax Credit (EITC), which was specifically designed to compensate for the regressivity of social security taxes. Those preferring that social security imitate a pure insurance scheme, with less intergenerational redistribution, might favor the baseline case (increasing the payroll tax) for this reason.  ?H&<Criterion 5. Encouragement of national saving. Both tax increase proposals reduce the social security deficit by the amount of the net taxes raised. The question is where these taxes would come"'z0*((@@/" from peoples' consumption or saving. To the extent that taxes are paid out of private saving, then national saving may not change at all; instead, there is merely a change in the party doing the saving. Alternatively, the loss in disposable income may come from consumption, resulting in a larger increase in net national saving. Some analysts argue that raising the taxable limit will affect saving negatively since upper income households are most targeted by the tax increase, and they do much of the nation's saving. Relatively little is known about this proposition empirically.  ?< xExpanding the definition of taxable income : Social security payroll taxes are paid only on money income, because employerprovided benefits are (for the most part) not subject to either income or payroll taxes (elective contributions for 401(k) and 403(b) plans are included in the payroll tax base). Total employer payments for benefits amount to roughly onethird of payroll, up from close to zero in the 1920s. The erosion of the tax base over time has meant that payroll taxes on the portion of income that is taxable were set at a higher rate than would be the case otherwise, to raise a given amount of revenue. As a result, some policymakers have proposed expanding the definition of taxable income, subjecting all currently taxprotected employerprovided benefits to both income and payroll tax.  ?< xOne practical problem is that significant measurement problems arise in evaluating employee benefits. While determining the values of some is relatively straightforward (for example, an employer's contribution to a 401(k) pension plan), other benefits are more difficult to price, for example, projected benefits in a definedbenefit plan or health insurance premiums for a given employee. Health insurance is particularly problematic to value for younger workers, because many group plans crosssubsidize the high health costs of older workers by charging more for relatively healthy (and hence inexpensive) younger workers. If health insurance costs were simply allocated per capita, young workers would be forced to pay taxes on a health benefit amount that far exceeded their actuarially fair premium. Despite these practical difficulties, the Panel  ?x<proceeds with an evaluation of this proposal.(oNendnote reference("endnote reference"H ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#"endnote reference"` hp x (#` hp x (#73.(endnote reference( A proposasl has be made to lift the taxable ceiling only on the employer half of social security payroll taxes (with selfemployed matching), which would raise system revenues without increasing promised benefits. However this option is not considered seriously here since economic analysis implies that this form of tax increase is sooner or later passed on to workers in the form of lower wages and/or lower employment; this is discussed in Part IV, below.    x xNontaxable employee benefits are more common among higher paid workers, and thus from a vertical equity perspective, many have argued that these benefits should be included in the income tax base. The argument is less persuasive for the OASI payroll tax, becaise social security benefits were designed to replace money earnings, but not other nonwage benefits. When retirees lose their activeworker benefits by retiring prior to age 65, some replace their former medical coverage by purchasing it, while others receive extended coverage from their previous employers. One might therefore argue that some benefits could be included in the tax base and also in the eventual social security"'{0*((@@/" benefit calculation. This proposal would decrease the favorable impact of the tax inclusion on social security system finances, because future benefits would rise as well.  ? <Criterion 3. Avoidance of market inefficiencies. Expanding the definition of taxable income to include employee benefits would raise the covered income of everyone with employee benefits. For those already over the OASI taxable limit, neither total payroll taxes paid nor the marginal tax rate would change. For those with income under the limit, total payroll taxes (and eventual benefits) would increase, but their marginal tax rate would not. Some people, pushed over the limit by the redefinition, would face increases in total payroll taxes (and eventual benefits), but their marginal payroll tax rates would drop to zero. On the whole, this proposal would result in less labor market distortion than the baseline payroll tax increase, because it keeps constant or lowers the marginal tax rate, whereas the baseline proposal increases the tax rate for all those under the taxable limit. xSubjecting benefits to tax could have a more potent effect on peoples' saving and consumption choices. For example, the fact that (many) employee benefits are not subject to payroll and income tax has lowered the price of group health and retirement insurance, encouraging employees to buy more than they would have obtained if they could only be purchased at aftertax prices. The tax subsidy has increased consumption of health insurance, and raised the amount of private saving held in the form of pensions. Removing the tax protected status of pensions would make pensions less attractive to those under the taxable limit, and would probably decrease the total amount saved. Whether this result is viewed as a beneficial or negative outcome depends to a large extent on whether people would be likely to underconsume the benefit without the subsidy. For example, those believing that retirement saving are too low see the tax qualified status enjoyed by pensions as a beneficial public policy. Others who think there is too much health insurance think that including health benefits in taxable earnings would be reducing a distortion.  ?x<Criterion 4. Equity of lifetime social security taxes and  ?@<benefits. This proposal would reallocate taxes and benefits within a generation, because the costs fall primarily on the middle class (for whom social security taxes would rise). Compared with the baseline increase in the payroll tax, this proposal is more favorable for lowincome workers without significant employee benefits and those already earning over the taxable limit, but worse for those in the middleincome category. Within an income bracket, the proposal would tax those with benefits more heavily than those without benefits, probably a move toward horizontal equity. In general, the proposal does not radically change the distribution of taxes and benefits across generations. "'|0*((@@/"Ԍ ?<ԙCriterion 5. Encouragement of national saving. As noted above, making pension contributions taxable would probably decrease the total amount saved. Whether this effect is larger or smaller than the baseline tax increase is not clear, but the Panel believes that saving would probably be decreased. Taxing health benefits might not have this effect.  ?x<x Introducing additional general revenues : Some portion of OASDHI is currently finance from general revenue, namely Medicare Part B (Supplemental Medical Insurance) for physicians and other care outside of hospitals, and the income tax collected on taxable social security benefits (at each individual's highest marginal tax rate), some of which goes to the OASDI Trust Funds and some of which goes to Medicare. The proposal examined here is to add an extra flow of general revenues from the federal income tax to pay for retirement benefits. xThe use of general revenues rather than the payroll tax has certain attractive features. It is administratively simple and is potentially more redistributive within generations than is an increase in the payroll tax. It would also generate a contribution from those already retired, for whom expected benefit payments over the retirement years far exceed payroll taxes over the worklife. On the other hand introducing general revenues to balance the system would substantially change the nature of social security, and it might also eventually erode public support. The additional use of general revenues would change the public's perception of social security benefits as earned rights, and might further politicize social security by drawing it more explicitly into annual budget debates. There are also analysts who argue that the redistributional component of social security should be explicitly identified and recognized, and funded from the same sources used for other public programs. xDespite these political economy objections, it is useful to employ the Panel's bynow familiar evaluation criteria to this proposal.  ?<Criterion 3. Avoidance of market inefficiencies. This proposal would have labor market consequences because some workers' marginal income tax rates would rise to generate these additional revenues. Exactly whose taxes would rise, and by how much, would affect how this proposal compares with the baseline payroll tax increase. xThe baseline payroll tax increase would raise the marginal tax rates of all those under the taxable limit. (Payroll tax payments on earnings below the threshold would rise for those above this limit, but their marginal OASDI tax rate would remain at zero.) By contrast, suppose that new revenues were raised by a proportionate increase on all federal income tax payers. Because the federal tax base is larger than the social security"'}0*((@@/" tax base, a smaller rate increase would be required in the federal income tax. Also, unlike the payroll tax, this income tax rate increase would affect those above the social security taxable limit, those whose total marginal tax rates tend to be the highest already. Assessing what effect raising payroll taxes would have on work and retirement is unclear, because higher marginal tax rates not only lower net earnings (encouraging less work and earlier retirement), but also reduce income (inducing more work). On net, the policy might encourage earlier retirement (Fields and Mitchell 1984b), although the effects of both this and the baseline proposal would probably be small. xOther decisions would also be affected by this proposal, including the way people allocate their income to saving versus consumption. Compared with the baseline proposal, raising federal income taxes would probably reduce saving more: the increase in the federal income tax is levied on asset earnings and therefore discourages saving and asset accumulation, whereas the payroll tax does not. The size of this effect is difficult to determine.  ?0<Criterion 4. Equity of lifetime taxes and benefits. Raising social security revenue through the federal income tax system increases withincohort redistribution more than the baseline approach, inasmuch as the federal income tax system is more progressive than the payroll tax. By using additional revenues unrelated to the earnings base on which social security benefits are calculated, this proposal will probably be seen as weakening the connection between individuals' social security taxes paid  ?p<and benefits received.(dfendnote reference("endnote reference"I ?<#Xx6X@DQ/X@##Xx6X@DQ/X@##Xx6X@DQ/X@#"endnote reference"` hp x (#` hp x (#74. (endnote reference(If benefits paid are then linked to additional income tax revenue gathered, the connection between taxes paid and benefits  ?<received may be strengthened rather than weakened.#X\ P:+Q2XP#  This perception would arise because payroll taxes are levied on a tax base directly tied to social security benefits, while increases in general revenue are not. If one turns attention to crossgenerational redistribution, it must be concluded that raising the income tax is also more redistributive. The burden of payroll tax increase falls entirely on future earners, exempting those already retired. Raising general revenues would distribute the social security financing burden more widely, requiring support from all those with taxable income, rather than just those with earnings.  ? <Criterion 5. Encouragement of national saving. As mentioned above, this proposal will change the way people allocate their income to saving versus consumption and might reduce private saving more than the baseline case.  ?#< E. Extending Social Security Coverage to State and Local Workers x xFew groups of employees mainly selected state and local government employees remain outside the social security system today. An argument in favor of including these is that of attaining universal coverage. Also people with careers spanning"'~0*((@@/" both public and private employment would be treated more uniformly. Objections to placing these workers in the social security system are that the federal government would be seen as intruding into the business of states (which may raise constitutional problems) and the fact that covering these workers will generate some additional payroll tax revenue now, at the expense of having to pay them social security benefits later. As a result, this change alone will not solve the social security system's solvency problems.  ?< F. Investing Trust Funds in Private Capital Markets xBy law, Social Security Trust Funds are currently invested in special issue Treasury securities. These are government obligations with maturities ranging up to 15 years, and the interest rate paid on them is equal to the average market interest rate for all marketable interestbearing Treasury obligations that are not due or callable in the next 4 years. Technically, these are redeemable at par. Because these are government securities, interest and principal are virtually free of default risk. (The Treasury obligations are subject to inflation risk since they are not indexed bonds). xSome analysts have proposed that some or all of the Social Security Trust Funds be invested in private capital markets, in particular in the U.S. stock market, on the theory that doing so could earn a higher rate of return. In this discussion the Panel examines some of the pros and cons of this course of action. xA judgment requires an assessment of the opportunities and the costs of this investment strategy, as compared with the current policy. If investing the Trust Funds in stocks carried no risk and provided a higher return in every state of the world, then a stock portfolio would always dominate the current investment portolio. This argument is made by some analysts who examine long holding period returns over the past 70190 years in the United States (Siegel 1994). For example, if one examines holding periods of 22 years, stocks outperformed longterm government bonds, intermediate term government bonds, and 3month Treasurybills in every 22year period beginning in 1926 (Zeldes, in progress). If this held true in the future, it would make sense for SSA to hold its entire portfolio in stocks rather than in Treasury securities of any maturity; in fact, the SSA in this case should borrow in order to purchase additional stocks. More generally, under this scenario, the entire government sector and in fact every individual with the appropriate length horizon would put all available funds in stocks. (Of course liquidity needs would still have to be met). xThe fact that all individuals with long horizons do not invest their entire portfolios in stocks suggests that there is a riskreturn tradeoff, despite the past history of returns. That"'0*((@@/" is, the future cannot be predicted with much precision, and even if the stochastic processes generating future capital market outcomes were the same as in the past, past outcomes are unlikely to be representative of the entire distribution of potential returns. There are fewer than four independent 22year historical periods since 1926, and it is difficult to estimate confidently the empirical distribution using four observations. An additional concern is that riskreturn characteristics of capital markets could change. xIf stocks are not certain to outperform bonds in every possible state of the world, then sensible policy must take into  ?` <account tradeoffs between risk and return.(endnote reference("endnote reference"ZJ ?<#Xx6X@DQ/X@##Xx6X@DQ/X@#"endnote reference"` hp x (#` hp x (#75. (lendnote reference(It is possible that some of the premium for holding stocks arises not because of risk, but because stocks are less liquid than Tbonds. If the SSA did not need this liquidity, this would be an argument in favor of investing some of the portfolio in equities. Unfortunately, this is not something that has been  ?<carefully analyzed in the literature.#X\ P:+Q2XP# Z To analyze this, the Panel must investigate social security's goals and constraints, to assess what the agency would take into account in determining the riskreturn tradeoff after which one could derive the optimal portfolio for the SSAthe ideal fraction of the portfolio to invest in each available asset. Although it seems unlikely that the optimal portfolio would hold 100 percent in nominal intermediateterm government bonds, it is possible that the fraction in equities could be positive or negative. Only if the optimal fraction were positive would it make sense to invest in the stock market. xThere has been little public discussion of SSA's goals and objectives under conditions of uncertainty. One simple way to think about how the agency might make decisions is that it seeks to minimize some function (such as a weighted average) of the mean payroll tax rate and the variation in that tax rate, while at the same time maintaining a constant real average social security benefit level. A large economic literature on tax smoothing shows that the government will tend to smooth tax rates so as to reduce the larger distortions resulting from proportionally higher tax rates (Barro 1979; Bohn 1990). Whether the share in stocks is positive would then depend on the relationship between payroll tax revenue and stock and bond returns over some horizon (as well as liquidity needs). xFor example, imagine that, relative to trend, real wages and payroll tax revenues tend to fall at the same time that the excess return of stocks over intermediate term Treasury bonds is low. In this case, investing the Trust Funds in equities rather than current government securities could make it necessary to raise tax rates even further when real wages fall. This positive correlation between wages and the excess return of stocks could increase the variation in tax schedules and add to the risk faced by the system. Conversely, if, relative to trend, real wages rise when the excess return on the stock market is low, then holding stocks in the Trust Funds could reduce the risk. What this correlation has been in the past (and what it might be in the future) remains to be investigated, but more knowledge of this relationship is critical in evaluating the proposal. In"'0*((@@/" addition, the relevant horizon for examining these correlations is not clear. Intuition suggests that what might be relevant is the correlation over long horizons, such as decades. xWhether investing a positive fraction of the social security Trust Funds in stocks is optimal in this model depends both on the risk, as just discussed, and also on the difference between the expected returns on stocks and bonds. If covered earnings and stocks were positively correlated and the risk premium from holding stocks were small enough, then SSA would optimally hold no stocks at all in its portfolioin fact, it would prefer to shortsell stocks. Alternatively, if covered earnings and stock returns were negatively correlated (or if the risk premium were large enough), the optimal SSA portfolio would include some holdings of stocks. Bohn (1990) explored this risk from the perspective of the government as a whole (rather than that of a single agency), and concluded based on an examination of quarterly data for the U.S. that the covariances are such that the optimal holding of a U.S. stock market index (the Standard & Poor 500) would be negative, at least in the absence of any risk premium. The government should take a short position in the stock market. Taking into consideration a large equity premium might reverse this theoretical prescription, but the finding makes it clear that the government's optimal holdings of stock are not necessarily positive, and the work gives the Panel an additional reason to question whether stock market investment is appropriate for the Trust Funds. xTo determine the optimal portfolio for the Social Security Administration, the agency should explicitly evaluate how it weights risk versus return, something that has not been done systematically to date. More broadly, the agency should undertake explicit stochastic modeling. Currently the agency examines only a handful of different scenarios (low, intermediate, and high cost), a strategy that implicitly assumes each economic and demographic variable can take on only three values and that these variables are perfectly correlated. A more comprehensive treatment of uncertainty would require taking a stand on the joint probability distribution of these variables, including allowing these variables to be less than perfectly correlated, and explicitly writing down the objective function of the agency. Such an undertaking is necessary before deciding whether Trust Funds managers should be encouraged to invest in equities. xA different way to think about the proposal is to realize that the risks and returns in the social security agency s portfolio would be passed on to households, perhaps with some lag. When portfolio returns are high, this improves social security benefits or reduces social security taxes in the future; when returns are low, this approach reduces social security benefits or increases taxes in the future. As a result,"'0*((@@/" households receiving these benefits and paying these taxes may take the Trust Funds portfolio into account in forming their own portfolios. One hypothetical but useful benchmark is the case where households consider SSA's portfolio to be entirely their own. In this instance, as long as households have sufficient private wealth, any change in the SSA portfolio would give rise to an offsetting adjustment in their private holdings. Thus, if the trust fund reduced holdings of Treasury bonds and increased holdings of equities, households could offset this by increasing their holdings of Tbonds and decreasing their holdings of equities. A key implication of such a benchmark scenario is that social security trust fund portfolio investment policy is neutral: that is, it does not change the riskreturn characteristics of households' overall portfolios. It follows that, if households behave in this manner, no reason exists to oppose equity investment by the Trust Funds, but there is also no particular reason exists to favor it. xThis neutrality result does not hold true if people are prevented from investing in stocks to the extent they would like to: in this instance, holding equities in the Trust Funds could make them better off. One such case would arise if people have no wealth other than expected future social security benefits. Indeed recent evidence suggests that a significant fraction of the population enters retirement with little other than housing and social security wealth (Smith 1994). Another possibility is that trust fund risk will fall on as yet unborn generations, who  ?<have no opportunity to hold stocks now on their own accounts.(gendnote reference("endnote reference"OK ?<#Xx6X@DQ/X@##Xx6X@DQ/X@#"endnote reference"` hp x (#` hp x (#76. (lendnote reference(In a world with bequests generated by altruism, this effect  ?<would not be present.#X\ P:+Q2XP# O Holding equities in the fund could give future generations exposure to current stock returns, thus improving risk sharing across generations. In both cases, as long as households are not implicitly being exposed to stock market risk through other channels, adding some equities into their social security portfolio would make them better off. Of course, this policy leaves open the question of how large the fraction held in equities should be. Increasing this fraction too much would make some households (those relatively more risk averse) worse off. xThe next question that must be addressed is whether investing the Trust Funds in equities will increase national saving. Focusing first on government saving, proponents assert that holding social security funds in private assets will increase pressure on politicians to reduce the government deficit. The importance of this political argument is difficult to gauge. It is plausible, however, that if legislation allowed the Trust Funds to invest in equities, legislation could also require that the purchase of equities would not count as an expenditure and thus would not alter the measured unified budget deficit. Turning next to private saving, whether placing equities in the Trust Funds would alter private household saving is far from clear. People might respond to having a riskier asset in their social security portfolio by increasing"'0*((@@/" precautionary saving, although this effect migh be offset by the higher expected return on equities. And for people who have no wealth, investing the Funds in equities would have no effect on saving. They are unable to save any less, and unwilling to save any more. Overall, this change is unlikely to have a large effect on national saving. xFinally, some critics argue that the proposal to have equities in the Trust Funds would expose people to additional political risk, in that government officials would have to select investment options and might not make these decisions purely on riskreturn criteria. If the SSA were allowed to choose its investments, it might not remain immune from pressure to direct investments toward economically targeted firms, industries, and nations. This problem becomes more serious the larger is the size of the Trust Funds. One solution is to require SSA to invest in an index fund, thus reducing discretion on the part of the SSA. Another option is to subcontract with private money managers, although again selection of these managers might not be determined purely on riskreturn grounds. Another option would be to transfer these funds to individual accounts, a topic to which the Panel turns in the next section. Several other questions also remain to be investigated in future research, including the matter of how quickly a shift to equities should occur to avoid massively altering capital market prices. xThe Panel concludes that a judgment on the proposal to invest some of the Trust Fund in equities should depend on an assessment of its opportunities and its costs. Given the lack of literature on many of these questions, we did not reach a consensus on the proposal, but instead propose that this issue deserves additional study before a clear policy prescription can be made.  ?< G. Individual Social Security Retirement Accounts xIn this section the Panel takes up a different policy proposal attracting debate of late, namely that of converting a portion of the social security payroll tax into individual pension accounts. The purpose of the discussion is to identify the most important advantages and disadvantages of such a policy change. xAssuming that the necessary adjustments have been made in order to make the social security system solvent, the system will run a current annual surplus for a number of years. One approach would be to redirect a portion of this social security surplus to individual accounts (Porter, forthcoming). Of course, paying this surplus to workers in the form of individual accounts would require compensating reductions in future social security benefits. The contributions to and income from the private"'0*((@@." system would replace a portion of the taxes paid to and income  ?<promised under social security.(endnote reference("endnote reference"L ?<#Xx6X@DQ/X@##Xx6X@DQ/X@#"endnote reference"` hp x (#` hp x (#77. (lendnote reference( This analysis assumes that these changes would be mandatory; an alternative would be to make the switch voluntary, but this raises a multitude of implementation problems that are avoided  ?X<with the mandatory approach.#X\ P:+Q2XP#   ?X< xA substantial departure from the existing approach would entail a move to a twopillar system. As an example of a twopillar model, the first pillar could be designed to guarantee a minimum flow of retirement income after some age, perhaps an amount that was higher if the retiree had spent more years contributing to the system. This bottom tier would provide minimum or povertyline income to the aged, and might be payroll taxfinanced or perhaps paid for out of general revenue. The second pillar of such a revamped retirement income system could be thought of as a mandatory definedcontribution plan funded by payroll taxes that depended on each worker's earnings. This second pillar would therefore entail a conversion of part of workers' social security payroll taxes into a mandatory individual account system. Participants' funds in this second pillar plan would be invested in a range of capital market assets, presumably directed by the individual contributors themselves. xVariants of a twotier system have become popular in this hemisphere over the past decade and a half. The most widely discussed example is that of Chile, where the national payasyougo social security system began to be phased out in 1981, and replaced with a national, mandatory, definedcontribution (DC) system paid for by a new 13 percent payroll tax on covered workers  earnings. Of course even after the old system stopped accruing new benefit promises, retirees had to continue to be paid and vested active workers still held benefit promises from that system. In addition, this DC plan is backstopped by a generalrevenuefinanced social safety net system that provides a minimum retirement benefit. To bring about the transition, the Chilean government decreed substantial cuts in promised social security benefits, but as it still faced a major revenue shortfall to the social security system, it turned to government budget surpluses totaling about 4 percent of GDP to finance the switchover. This surplus continued since the 1981 startup of the system and is projected to continue for another decade, falling after that for the next fifteen years. The fact that Chile had a budget surplus at the onset of the reform contrasts with the U.S. economic situation where budget deficits have been a fact of life for decades. xDespite the fact that Chile s situation was, and remains, somewhat unique, the rapid growth of Chile's economy and the fact that the pension fund shared in this growth has gained that system many supporters (World Bank 1994). Others following in Chile's footsteps include Argentina, Bolivia, Colombia, Peru and Mexico, and other nations expressing an interest include Brazil and some of the nations of the former Soviet Union (Davis 1995). "'0*((@@/"ԌxOne argument offered in support of a privatelymanaged mandated pension pillar in the United States is that it would be perceived as reducing political risk surrounding retirement income. In the current environment, workers increasingly realize that promised social security benefits are unsustainable, and believe that benefits will fall or taxes rise to maintain system solvency. Giving people individual private plans would probably reduce uncertainty regarding the politics of the future social security system, although it must be recognized that a private system still faces certain political risks as well. If political risk were in fact reduced with a twotier plan, some of the benefits flowing from this change could be captured by social security to improve the long term solvency of the system. This result could occur, for instance, if middleaged and younger workers would be willing to pay social security payroll taxes required to maintain a larger system for current retirees, in exchange for some level of contributions devoted to individual plans they could manage themselves. The government might reduce its long term promises to social security system participants by cutting future retirees' definedbenefit promises, in exchange for individuallymanaged definedcontribution plans owned by workers themselves. xAnother argument in favor of allowing workers to invest part of their payroll tax in an individual retirement savings account is that it affords participants the opportunity to allocate investments as they see fit. Some people have also suggested that this program would promote financial literacy, because households would be given control over some of their social security funds and would have more of an incentive to learn how saving and portfolio choices affect their standard of living in retirement. xThere are also those who offer more macroeconomic rationales for favoring the proposal. One is that individual accounts could create political pressure to reduce the deficit, because the unified budget would show a larger measured deficit and thus would put pressure on politicians to reduce it. Another is that a second pillar pension might create a source of new investment funds. For example, advocates of the Chilean pension reform believe that Chile's private pension pillar increased net national saving because the pension system created capital markets that did not previously exist in that nation; this is probably not a potent argument here in the United States. Whether net new national saving would occur in the United States under a mandatory redirection of the payroll tax would depend on whether saving outside of these accounts increasedby both households and governments. If the reduction of future benefits were relatively small, other private saving would likely fall due to a reduction in future benefits relative to what was previously expected. "'0*((@@."Ԍendnote refe  ` hp x (#` hp x(#xA number of disadvantages have been noted with the type of twopillar system described above. One is that the individual account secondtier plan might not provide adequate income to those with low lifetime earnings. This might occur because the twopillar plan is likely to be less redistributive than the current social security system, or because political support for the first pillar might weaken once it is explicitly separated from the second pillar. x` ` #endnote refe#  ` hp x(#` hp x (#USUSxAnother concern is that the second pillar plan exposes participants to capital market risk, as do definedcontribution pension plans more generally (see Section II). That is, portfolio allocations yielding low returns reduce the individual's pension account at retirement. Additionally, there is a question as to whether workers are wellequipped to direct their own retirement accounts, particularly if the social safety net of the social security system is weaker than it is now. xA different problem with individual mandated pension accounts is that workers might want to take lumpsum cashouts prior to retirement, but permitting people to spend their individual account accumulations while young would no doubt leave some in poor economic condition during their retirement years. There is a related concern that some workers would, if permitted, take lumpsum cashouts at the time of retirementperhaps because they would like to include these assets in bequeathable wealth. Unfortunately, allowing such cashouts erodes the market for retirement annuities. Under current social security system rules, benefits are paid out only as a life annuity and the current risk pool for that annuity is national, mandatory, and multigenerational. People with shorterthanaverage life expectancies subsidize via the national risk pool the annuities for those who expect to live past their cohort's life expectancy. If individual pension accounts were permitted to be cashed out, those expecting to die younger than average would not annuitize their pensions, which in turn would lead to a potentially large problem with adverse selection. x xAnother insurancerelated concern regarding individualaccount pensions is that social security benefits are currently indexed to the CPI, while private annuities in the United States are typically defined in nominal terms or tied to asset market performance (which does not guarantee real retirement income). In the Panel s view, there are serious risks from replacing indexed social security benefits with mandated individual pensions, unless a mechanism were available to obtain indexed private annuities. One answer might be for the Treasury to offer indexed federal government debt, which would help private insurers create indexed annuities. (Indexed bonds are taken up in Section IV below). It is again interesting to note what happened under the Chilean reform. There, retirees purchasing annuities"'0*((@@." (rather than making phased withdrawals from their accumulations) are permitted to purchase only CPIindexed annuities. xExperts discussing the design of a twopillar retirement system also raise questions about the costs of a mandatory second pillar pension costs which could vary depending on the precise way a secondtier individual account plan is formulated. Tax collection costs could rise substantially if contributions had to be collected person by person, as in Chile, instead of by the government as is now the case. Clearly the scale economies enjoyed by the current system, due to the central role of the Internal Revenue Service which oversees collection of the payroll tax, could be eroded under a decentralized collection scheme which would also increase the possibility for evasion. Money management costs would also rise in an individual account pension system, inasmuch as the Social Security Administration currently spends little on the money management and investment function. Also it is worth noting that money management costs are often considerably higher for individual accounts, as compared with group pension plans. xA different, and potentially more significant, concern is that of the remaining political risk that a secondtier individual account pension plan might face. A worry is that the individual accounts could become the basis for assetbased means testing for other government programs such as Medicaid or SSI, or even programs that are not currently means tested such as unemployment insurance and Medicare. Accumulations in these accounts could make households ineligible for other government programs, producing incentives for evasion and retirement insecurity (Hubbard, Skinner, and Zeldes, 1995). Also, workers likely to receive firstpillar social security benefits might face a substantial incentive to avoid contributing to the second tier, and evasion is a possibility that cannot be ignored. xThe Panel discussed several other points raised by analysts evaluating a twopillar proposal. One is that making individual accounts more prevalent might crowd out employersponsored pensions, though this is less likely if a portion of current payroll taxes were redirected toward a small secondtier pension. A second is that private definedcontribution holdings might fail to provide investors a good inflation hedge. This concern underscores the appeal of governmentissued indexed bonds, a topic taken up in Part IV of this report. A third is that people might find individual social security accounts more appealing if they were permitted to deposit their funds into a governmentrun indexed passively managed mutual fund, which offered a lowcost option competing with private money managers. In addition it seems clear that it would be necessary to require money managers to disclose and report individual investors' costs and returns in a standard format with comparable information across programs to"'0*((@@." help investors make betterinformed retirement account investment decisions. xSome of the disadvantages raised in this discussion could possibly be overcome if the government were to offer a simple indexed, passively managed, lowcost investment option. Concerns arising from imperfections in private annuity markets remain to be resolved. Regarding the distributional concerns mentioned above, it might be feasible to design the second pillar in such a way that, though it was a definedcontribution system, it credited a smaller rate of return to highearners' pension contributions, and a higher return to those earning low wages. Such a modified second pillar plan might be adequate to address concerns about income inadequacy and to some extent income insurance. In addition, mandatory reporting of fund management costs might help alleviate concerns over high administrative costs experienced by some individual account pension plans. On balance, many Panel members find this approach potentially promising, though it urges additional study of all of these issues.  ?h< H. Conclusions and Assessment xThe Panel was charged with outlining and evaluating a range of options to return the nation's social security system to long term balance. Here we collect our assessment of the various benefit and revenue changes examined in this section of the Report. xFirst, the Panel believes that some combination of benefit cuts and/or revenue increases is necessary to restore the social security system to actuarial balance. Second, we believe that timing is crucial: appropriate legislation should be enacted promptly. The Panel also suggests that it is important to combine immediate legislation with delayed implementation of some benefit cuts and/or tax increases. Indeed, the urgency of prompt legislation is increased by the desirability of implementation lags. Gradual implementation reduces the magnitudes of notches (different treatment of cohorts close in age) and the perception of unfairness that notches engender. Significant changes in social security benefits should be announced with considerable lead time to allow workers to adjust to the changes by adapting their retirement, saving and consumption plans. xThe analysis examines two baseline reform options that span the range of options most commonly discussed in public policy circles. One scenario reduces benefits (PIAs) by a fixed percent for all persons born in and after some year, which we arbitrarily take to be 1940. The size of the PIA cut would have to be sufficient to bring the system into longterm balance. The alternative baseline scenario would impose an acrosstheboard payroll tax increase beginning in the year 2002, when those born"'0*((@@/" in 1940 turn age 62. These two alternatives frame most of the other reforms currently under discussion, and are useful as comparison standards in assessing several other changes as well. ` hp x (#` hp x x` ` xThe Panel finds that the criteria adopted do not unequivocally favor either raising taxes or decreasing benefits. Rather, some criteria such as adequate retirement income favored tax increases, while others such as equity of lifetime social security taxes and benefits between generations favored benefit cuts. Thus closing the fiscal imbalance with additional revenues rather than benefit decreases is suggested if one emphasizes the first two criteria, adequate retirement income and insurance against unforeseen income fluctuations. Social security benefit cuts would increase the number of Americans with inadequate retirement income, and lower the insurance protection offered to workers, survivors and dependents. Within a generation, the use of tax increases rather than general benefit cuts favors those with the longest life expectancies those most likely to receive benefits for a long time and those with lower incomes for any given life expectancy. x xOn the other hand, closing the fiscal imbalance with benefit decreases rather than tax increases is suggested if one  ?<emphasizes the fourth and fifth criteria, equity between generations and the encouragement of private saving. The expected return on social security contributions is already going to be lower for baby boomers than for past, current, and nearfuture recipients (and this return will decline even further when either social security taxes are raised or future benefits are cut). Younger participants would pay the higher taxes for many more years than would participants planning to retire soon. A smaller social security system (lower benefits) would also encourage some individuals to offset part of the loss through their own saving behavior (see below). x xFocusing on the other criteria, the Panel concludes that social security retirement benefits induce some older workers to leave the labor force earlier than they otherwise would. Benefit cuts, especially if combined with an increase in the early age of entitlement (now age 62), are likely to reduce this effect. In addition, payroll taxes may discourage the labor supply of younger workers, a labor market distortion that is more likely to decline if benefits are cut than if payroll taxes are increased. The Panel finds little professional consensus on the size of the impact of social security on private saving. To the extent that social security benefits substitute for private saving, a smaller system (benefit cuts rather than tax increases) would encourage private saving. But many workers with little or no saving beyond their home equity are unlikely to make significant changes in their saving behavior in response to the changes in social security benefits being contemplated. The Panel concludes that"'0*((@@/" reducing benefits might have a small positive effect on private saving.  ?<` hp x ` hp x  ` hp x ` hp x xThe Panel also compares the effects of the baseline benefit cut (an acrosstheboard decrease in the PIA formula) with those of several alternatives, including reducing disproportionately the benefits of highwage workers, delaying retirement ages, reducing the costofliving adjustment, and means testing benefits. If benefits were to be reduced, strong arguments support increasing the ages of eligibility for early and normal social security benefits. Delaying these retirement ages is a sensible response to increases in life expectancy, and one that prevents lifetime benefits from automatically increasing as recipients live longer. Specifically, if benefits were to be reduced, most Panel members agree that the normal retirement age for social security benefits could be increased and eventually indexed to life expectancy. A suggestion along these lines is to eliminate the scheduled hiatus in the normal retirement age rise to age 66 (20002005) and 67 (20172022). xIf benefits were to be reduced by means other than, or in addition to, increases in the NRA (for example, if the PIA formula became less generous), there is some support for the notion of disproportionate cuts at the top, versus an acrosstheboard decrease. The Panel also discussed how to allocate the burden of benefit reductions across different cohorts those already retired, those about to retire (for example, within 5 years), and those further away from retirement. If benefits were cut, it is difficult to argue that people already retired or about to retire should be exempted entirely; however, there was support for smaller benefit reductions for these groups than for future retirees. xThe Panel also studied two other methods of cutting benefits: means testing, and lessthanfull inflation indexation. There are strong arguments against meanstesting social security benefits on the basis of other retirement income or accumulated wealth. To avoid loss of social security benefits, some workers might reduce their own retirement saving or persuade employers to shift compensation from pension contributions to earnings. Either response would lower saving and private retirement incomes. Focusing on indexing, the Panel recognizes that social security benefits are the only fully indexed annuity available to (nearly) all workers. The threat of inflation would be a very serious concern to retirees, especially those with long lives after retirement, if full indexation were eliminated. For this reason, the Panel opposes permanently indexing social security benefits by less than the costofliving. On the other hand, the Panel urges that the Bureau of Labor Statistics investigate whether the specific Consumer Price Index currently used to adjust benefits correctly measures the costofliving. If this measure is found to be biased, the Panel"'0*((@@/" would support corrective changes in the method of calculation.  ?<There was also extensive discussion on whether a temporary delay or reduction in the costofliving adjustment would be desirable; with some arguing that if benefit were decreased for future retirees, a temporary delay in indexation for current recipients would spread the burden more equitably.  ?x<x` hp x ` hp x  ` hp x ` hp x The Panel then compared the effects of several different tax increase proposals with the baseline revenue increase (a simple increase in the payroll tax rate). Three options received the most attention: raising the earnings limit on which payroll taxes apply, expanding the definition of taxable income to include employee benefits, and infusing additional general revenues into the Social Security Trust Fund. On the basis of our analysis, most Panel members would favor raising the payroll tax rate rather than increasing the taxable earnings threshold, if additional revenues were to be raised. The threshold increase, unless applied only to the employers' portion or combined with a change in the benefit formula, would increase future benefits for those at the upper end of the income distribution, which a payroll tax increase would not. There was little enthusiasm for including employee benefits in the taxable wage base primarily due to significant measurement problems. Furthermore, the Panel's review of the evidence suggested little advantage to seeking an additional direct infusion of general revenues, instead supporting the maintainance of a link between social security contributions made and benefits received.  ?<` hp x ` hp x  ` hp x ` hp x xTurning next to the issue of the Trust Funds, the Panel suggests that they should continue to be at least partly funded, meaning that a significant and stable margin of income over annual expenditures should be maintained over the foreseeable future. The Trust Funds currently have assets exceeding one year's outlays, an amount projected to grow for about two decades as revenues exceed benefit payments. The Funds are currently invested in special issue Treasury securities, whose interest and principal are virtually free of default risk. The Panel examined the question of whether the Trust Funds should be invested in private capital markets, in the hope that investments would earn a higher rate of return than if invested solely in Treasury securities. The Panel believes that a judgment on this proposal should depend on an assessment of its opportunities and its costs. Given the lack of literature on some of these questions, the Panel concluded that this issue deserves additional study before a clear policy prescription can be made.  ?(#<` hp x ` hp x   ?#<` hp x ` hp x x The Panel also discussed the pros and cons of converting all or part of the social security Trust Funds (or the annual surplus) to individual social security accounts, over which participants would exercise some investment discretion. The Panel noted that distributing the annual surpluses to individual accounts would require additional adjustments to benefits and/or"'0*((@@/" taxes beyond what would be required to achieve system solvency without this distribution. Several advantages of this proposal were identified, along with some disadvantages. xDespite these questions and concerns, the Panel is open to the idea of converting part of the Trust Funds to individual accounts, if the remainder of the social security system could be made solvent. The Panel recognizes the need to coordinate the pattern of any benefit cuts with the pattern of benefits that would be received from these individual accounts. Many Panel members would recommend prohibiting access to the funds for any reason other than retirement, and most would recommend mandating that the benefits be wholly or in part distributed in the form of an annuity, rather than permitting a full lumpsum cashout. There is a division of opinion on whether the annuity could be best managed by the government or the private sector, though there are strong arguments for having the government offer one of several possible investment vehicles that workers might chose. ` hp x ` hp x (# "0*((@@"  ?<Gjr X-#XP\  P6Q DXP#  U 1 1 #C\  P6QP# #XP\  P6Q DXP#  X-  ` hp x (# (#  #C\  P6QP# #XP\  P6Q DXP#  IV. Other Policy Options Regarding Future Retirees' Incomes xHaving considered a series of changes in social security benefits and taxes, we now discuss additional ways to improve retiree wellbeing including changes in eligibility ages and proposals to alter private pension incentives and regulations. Several other options to improve retiree wellbeing are also examined, including expanding the role of the SSI program, issuing indexed bonds, and utilizing reverse annuity mortgages. These proposals were not evaluated in as much detail as were the policy options investigated earlier, but the Panel thought it important to go on record on several of these items nonetheless. x  ?( <  ` hp x (#r4` hp x  A. Integrating Social Security Eligibility Ages rrn44   r4` hp x ` hp x (#xThe Panel believes that several changes in social security are important enough on their own merits that they should be considered as highpriority reform areas whether or not they contribute to improving system balance. The most salient of these is the development of an integrated approach to coordinating the ages at which Americans become eligible for benefits under different components of the social security system, including disability, survivors, early entitlement, and normal as well as delayed retirement benefits, as well as the supplemental income program (SSI). The other topic that the Panel deemed important to discuss was social security rule changes that could equalize the treatment of one and twoearner couples, widows, and changes in indexation policy after age 60. This section takes up each in turn.  ?p<x Integration of eligibility ages in social security benefit  ?8<calculations : The normal retirement age (NRA) plays two roles in the process of determining benefits. First, retirement benefits depend on the number of years before or after the NRA that they are first claimed. Second, the NRA is also the maximum age at which a worker can apply for disability benefits; that is, the maximum disability age (MDA) is set to equal the NRA. Under current law, the MDA is scheduled to increase along with the NRA for future cohorts. This lockstep linkage should be contrasted with the way the system handles the earliest eligibility age (EEA) for retirement benefits; under current rules, the EEA will not increase beyond 62. Similarly, the maximum age for applying the earnings test (ETM) does not increase beyond 70 under current legislation, and the minimum age for receipt of SSI (based on age) is scheduled to remain fixed at age 65. Finally, under current law, the earliest age at which survivor benefits (ASB) can be received is age 60. "%0*''JJ-" xIn this section, the Panel discusses linkages among the NRA, EEA, MDA, SSI, ETM, and ASB. The values for these ages under current rules and currently legislated changes for the future are listed as follows:  ?<xAges` ` Present Future  ?<xEEA` `  62hh@ 62  ?x<xSSI` `  65hh@ 65  ?@<xMDA` `  65hh@ 67  ?<xNRA` `  65hh@ 67  ?<xETM` `  70hh@ 70  ?<xASB` `  60hh@ 60 The discussion proceeds by examining why minimum and maximum ages exist at all under the social security system, versus extending benefits to the entire population that now are available to only an age-restricted subsection of the population.  ?<xNormal Retirement Age (NRA): As noted above, raising the NRA to age 67 or later is a method of reducing benefits for future retirees, and as such was compared with other benefit reduction options in Section III. Nevertheless the Panel agrees that raising the NRA has considerable political appeal since it avoids the vocabulary of benefit cuts, and may alter the public view of the "socially acceptable" age at which to leave the workforce. Because most people do not retire at the NRA, a more potent change in the nation's retirement vocabulary would probably result from raising the early entitlement or early retirement age.  ?p<xEarly Entitlement Age (EEA): If there were no minimum age at which one could apply for actuarially reduced social security benefits, then a worker of any age with enough quarters of coverage and sufficiently low earnings could claim and receive retirement benefits. This arrangement would involve a number of differences from the current situation. Lifetime low earners might receive benefits for much of their working lives without cutting back on work very much. Individuals who stopped working (for child care, for example) could receive benefits irrespective of whether single or married (this possibility would hold even though another member of a couple had high earnings). Workers retiring early could choose lower benefits for the rest of their lives rather than having to wait until age 62. xIn contrast, under current social security rules, workers with lifetime low earnings (and no saving or pension) must wait until age 62 to receive benefits. If people filing for early benefits received true actuarial reductions, then the only policy concern would be that some people might have too little income in later years because they had chosen to take benefits too early. "H&0*''JJ." It is because of this form of myopia that social security does not permit tooearly retirement. There remains the problem that the current system does not reduce benefits exactly actuarially, so that people may retire earlier than optimal in order to benefit from early retirement subsidies. Because recomputation rules are complex, some people may be making early retirement choices erroneously. Additionally, people vary in both their perceived and actuarially calculated life expectancies, which makes it impossible for a uniform actuarial reduction to be actuarially neutral for every single individual. This fact creates retirement incentives that vary across members of the population. xPermitting people to retire early under a nonactuarially neutral benefit formula induces labor market distortions by encouraging earlier retirement than would be the case in a more neutral environment. This effect is offset by the fact that having access to early retirement benefits is a useful form of insurance for those who experience seriously curtailed earnings opportunities through no fault of their own, who cannot qualify for disability benefits, and who have no other sources of saving. The key question is how to weigh the market distortions offered by early retirement against the societal gains of having an early retirement option that is not actuarially neutral. The choice of an early entitlement age (EEA) depends on balancing these two considerations. xNo research has estimated the relative size of these two groups, making it in principle impossible to assess the ideal EEA. If health and functioning improve in step with life expectancy, then a healthier population has less need for an early EEA. According to this view, the early entitlement age should be raised over time, although perhaps not in lockstep with life expectancy because morbidity rates probably improve more slowly than mortality rates fall. Analysts agree that much of the longterm trend toward earlier retirement in the United States is probably not attributable to poor health; indeed people in their 60s appear to be at least as healthy, and probably more able to work than in the past. For this reason, increases in the EEA might be feasible. Offsetting this is the view that retirement is a normal good, and a social security system reflective of the preferences of wealthier cohorts over time should allow earlier access to benefits, so that early retirees need not live solely from privately accumulated wealth and pension benefits.  ?#<xEligibility Age for Supplemental Security Income (SSI): The Supplemental Security Income (SSI) program is a meanstested, federally administered income assistance program authorized by  ?H&<Title XVI of the Social Security Act (U.S. House, 1994).(endnote reference("endnote reference"(lendnote reference( It"H&0*''JJ-" provides monthly cash payments in accordance with uniform, nationwide eligibility requirements to needy, aged, blind, or disabled persons. To qualify for SSI payment based on age, a person must be 65 or older and a resident of the United States and either a U.S. citizen or a legal immigrant. SSI currently pays what is necessary to bring an individual to a statutorily prescribed income 'floor," and benefits are indexed by statute to the Consumer Price Index (CPI). Most states supplement the federal benefit for at least some participants. SSI plus food stamps ensures an income floor near the official poverty line for most older people. The mix of people receiving program benefits has changed over time, mainly because retirement benefits have increased faster than SSI payments. During the early days of the SSI program, more than half of all SSI recipients were aged; now only about onequarter are aged, and the majority of SSI recipients are now categorized as disabled. xIf SSI benefits could be received at any age without being blind or disabled, SSI would be in effect a guaranteed minimum income program for everyone, but would subject all to very high implicit taxes for work. Such a hypothetical system is not in place for everyone because of the expense, and because of concerns that work would be strongly discouraged. In contrast, SSI is made available to older persons currently those age 65 and older without being disabled for two reasons. One is that the fraction of the eligible population for whom work is a reasonable alternative is probably smaller than for the population as a whole. Second, the presence of social security benefits makes the program less expensive, because claiming social security benefits is a necessary condition for receipt of SSI benefits. xSome have proposed that this latter argument is just as relevant for people younger than age 65; that is, SSI could be made available earlier say, age 62 as a mechanism for protecting older people against poverty if the early retirement age were raised. How much would such a change cost, and how large a population would likely apply for SSI if it were made available at a younger age? Approximately 2.5 million people aged 62 to 64 received retired worker benefits in December 1993, and many of these early retirees would probably have been continued working or would have had other income sufficient to rule out SSI eligibility if the early retirement age were raised to, say, age 65. But for others, lowering SSI eligibility to age 62 would increase retirement income adequacy and insurance against low retirement income, because current social security benefits do not guarantee an income floor. Having access to SSI benefits at a younger age might discourage work and possibly private saving because of the high implicit tax rate on earnings and asset accumulation. In terms of the criteria developed"H&0*''JJ-" earlier, these market distortions and costs would have to be weighed against improved protection against oldage poverty. Also this type of system reform would probably not raise private saving, and would increase the burden on general revenue.  ?<xMaximum Age for Disability Insurance (MDA): Under current social security rules, disability benefits are equal to retirement benefits at the normal retirement age. Under current law, as the NRA rises in the future so will the maximum age at which a worker may file for disability benefits (MDA). Thus someone older than the NRA has no reason to apply for DI (unless he or she were to receive DI benefits offsetting an early retirement adjustment for leaving work prior to the NRA). In particular, a worker who retires and then becomes disabled can apply for DI benefits, although not beyond age the NRA (and limited by the need for recent labor force attachment). There is no apparent logic in allowing increased benefits for disabilities that occur after retirement. xOne question worth pondering when restructuring the system is whether the NRA is a reasonable age to collapse the disability and regular retirement plans into a single program. Current practice has evolved as a result of having a benefit structure for the disabled that is independent of the age at which the retiree claimed benefits. It may be that the right range for the MDA is between the EEA and NRA. xOne rationale for having two separate benefit programs is that it permits the government to give larger benefits to some workers, those for whom the labor disincentives associated with the benefits are believed to be fairly small (Diamond and Sheshinski, 1995). The age at which the programs should be combined should be set by comparing the administrative and economic costs and benefits of running a separate disability program, including labor market disincentives and inequities associated with inevitable errors in classification of applicants, versus the costs and benefits of having them operate separately. Most broadly, the MDA could be set in terms of the social desirability of having people work, at the age where it is difficult to distinguish between people who are truly disabled and those who are not. A problem is that experts disagree about how to measure disability: some focus on health problems, while others wish to include the willingness of people to take available jobs given their age, training, skill level, and working conditions. After that age when a majority of workers chooses retirement, it is not clear that further applications for DI would allocate funds any better than would general benefits. It should be recognized that choosing an age less than the NRA will result in some disabled workers suffering an actuarial reduction, equal to that experienced by those "voluntarily""H&0*''JJ-" retiring at that age. Alternatively, all disability benefits could be adjusted for assumed retirement at the MDA. x xOn balance, the Panel suggests that a comprehensive rethinking of the DI entitlement and benefit structure is necessary, with greater recognition of the wide range of ages at which people claim benefits.  ?@<xMaximum Age for the Earnings Test (ETM): At present, workers attaining age 70 apply for and can begin receiving social security benefits irrespective of their level of earnings. In addition, there is no delayed retirement increment for work beyond age 70, although the AIME may be recalculated for a worker with sufficiently high earnings. Under current law, the maximum age for the earnings test, the ETM, will remain at age 70. ` hp x (#r4` hp x (#rrnThere have been frequent proposals for lowering the ETM to the NRA. To investigate this proposal, the Panel begins by asking why there is an ETM at all. Those continuing work beyond the NRA who earn above the earnings limit are better off, on average, than those retiring earlier. This fact might imply that the ETM should be above the NRA, but does not lead directly to any particular choice of age. Indeed the fact that people employed beyond the NRA are better off seems more relevant for the choice of the delayed retirement credit, than for the choice of the ETM itself. Instead, the focus should probably be on the incentive to continue working. With no ETM, some workers would get little or no benefits if they continued working for as long as they were able. With an ETM, all benefits are being foregone by the decision to work; that is, having an ETM reduces incentives to retire earlier. Thus the choice of an ETM must balance the need for incentives for continued work against the cost of giving benefits to a group with higher than average earnings. The current setting at 70 seems reasonable based on  ?<available empirical evidence of these tradeoffs."endnote reference"X01ÍÍ X01ÍÍ My ?<#Xx6X@DQ/X@##Xx6X@DQ/X@#"endnote reference"r4` hp x (#` hp x (#78. (lendnote reference( This discussion suggests a possible modification of the earnings test. Recognizing that those working beyond NRA are probably better off than many, one could consider removing the delayed retirement increment, replacing it with the incentive that comes from paying part of benefits without an earnings test. This part could shift smoothly from, say, 20 percent at age 65 to 100 percent at age 70. That is, a 65 year old worker would receive 20 percent of PIA independent of earnings but would have the remaining 80 percent subjected to the current earnings test. Such a modification in the workings of the combination of the delayed retirement increment and the ETM would represent a  ?<smoother set of labor market incentives.#B\  PUP#  r4` hp x (#` hp x (#xPrivate pension experts also raise the concern that the social security NRA is currently the age at which maximum allowable benefits from a private definedbenefit plan may legally be paid without actuarial reductions. A full review of the links between pension and social security ages should be part  ? <of the reform process. "!0*''JJ_("  ?<B. Reducing Older Widows' Risk of Poverty (endnote reference("endnote reference" (endnote reference( xA notable achievement of U.S. social policy over the past three decades has been that of reducing poverty in old age. For most married couples, the risk of poverty is small, even many years after retiring. When one partner dies, however, the surviving spouse often faces a much greater risk of falling into poverty. And this risk disproportionately affects older women, who are nearly three times as likely as older men to be widowed (49 percent vs 14 percent) and can expect to remain widowed an average of 17 years (Population Resource Council, 1994). Currently, older women are more likely to be poor than older men (16 percent versus 9 percent) mainly because the average older woman is older than the average older man. At all ages, women are less likely to be married, and single people in general have lower incomes than do married couples. xOne suggestion to diminish widow's high risk of oldage income insecurity, is to redistribute social security benefits across members of a married couple, so as to produce relatively lower incomes while both parties are alive, and relatively higher incomes for widowed survivors. This view is motivated by the observation that work patterns have changed dramatically since the social security system was instituted while benefit payout rules have not. As a result, differences between two and oneearner households have emerged that become more pronounced after  ?P<the death of a spouse."endnote reference" N> ?<#Xx6X@DQ/X@##Xx6X@DQ/X@#"endnote reference"` hp x (#` hp x (#79.(rendnote reference( This becomes clear if one compares two hypothetical couples with identical combined social security covered average yearly earnings of $60,000, but with a different division of earnings between the wife and the husband. In Couple A, all earnings were the husband's; after retirement both partners receive social security benefits of $21,600 per year if he retires at 65 (the retired worker receives $14,400 and his wife receives a spouse benefit equal to onehalf of her husband's retiredworker benefit or $7,200). After the husband's death, his widow receives the husband's workerbenefit of $14,400 in place of her spouse benefit. In this "traditional family" the survivor's benefit is twothirds of the total amount previously paid to the couple. In contrast, Couple B has earnings equally split between the (sameaged) husband and wife, pays the same amount of taxes into the system, but receives lower yearly retirement benefits of $19,272 while the husband is alive (each member of Couple B receives an identical retiredworker benefit of $9,636). Note that the higher total worker benefits paid to the two members of Couple B relative to Couple A ($9,636 x 2 versus $14,400) is due to the progressive nature of the social security benefit formula. But this advantage is more than offset by Couple B receiving no spouse benefit, because it is less than each person's own retiredworker benefit ($4,818 versus $9,636 in this case). Furthermore, if the husband died, Widow B may receive either a survivor benefit of $9,636 based on her husband's earnings record, or continue to receive her own benefit of $9,636, but will not receive both. Hence benefits for the widow in the twoearner couple amount to only half of the already lower total benefits paid to both when the husband was alive. This is despite the fact that twoearner couples pay the same amount of taxes into the system as oneearner couples. (See U.S. House of Representatives 1985; Congressional Budget Office 1986;  ?<Burkhauser and Holden 1982; Steuerle and Bakija 1994).#B\  PUP#   One option is an "earnings sharing" approach, in which households allocate social securitycovered earnings of a married couple equally between them to calculate retirement benefits (Burkhauser and Holden, 1982; Burkhauser, 1984). At retirement, each married partner would receive a benefit based on his or her individual covered earnings record, and the spouse benefit under current social security rules would  ?<be abolished.(endnote reference("endnote reference"O> ?<#Xx6X@DQ/X@##Xx6X@DQ/X@#"endnote reference"` hp x (#` hp x (#80. (lendnote reference( For Couples A and B, this method of benefit calculation would provide exactly the same benefits ($19,272 while married) by lowering benefits for the oneearner couple.  yOX-#B\  PUP#  These benefit reductions could then be used to raise all survivor benefits to threequarters of total social security benefits while married. A threequarter survivor benefit is consistent with the official Department of Health and Human Services equivalence scale used to evaluate the needs of two and oneperson families for all its transfer programs. This plan would decrease poverty among older widows and widowers, slightly increase poverty among couples, end the arbitrary relationship between couple benefits and survivor benefits that results in replacement rates of between .5 to .67 for survivors, and treat equally one and twoearner couples who pay the same  ?`"<amount into the system.(*endnote reference("endnote reference"YP> ?<#Xx6X@DQ/X@##Xx6X@DQ/X@#"endnote reference"` hp x (#` hp x (#81.(lendnote reference( Of course ending the spouse benefit will severely reduce benefits for "traditional" households, because the progressive PIA formula would lead to smaller benefits for a single earner couple ($14,400 in our example) relative to the twoearner couple ($19,272 in our example). The spouse benefit equalizes payments  ?<to one and twoearner couples.#B\  PUP# Y xOther proposals to shift benefits from couples to survivors include reducing the spouse benefit as a fraction of the PIA, reducing the PIA for all married couples, and sharing AIMEs instead of earnings. Although the Panel did not review all"H&0*''JJ-" proposals, most agree that mechanisms should be considered to raise the ratio of survivor's to couple's benefits.   ` hp x (#r4` hp x   ?X<rrn Linkages between the Age for Survivor Benefits (ASB) and  ? <Other Ages : At present, widows (and widowers, although for convenience both will be referred to here as widows) can claim early retirement benefits at age 60, with a benefit reduction factor determined by a linear interpolation between age 60 and the NRA. Benefits are computed as 71.5 percent of the PIA for filing at age 60 and 100 percent at the NRA. For this reason, raising the NRA delays the age at which widows can receive full benefits, and raising the EEA lengthens the period during which widows must receive reduced survivor payments. Both proposals therefore reduce survivor benefit protection, and lower the value of income insurance provided by the system. Labor market incentives for widows are also altered, although the precise effects are difficult to predict. If the early age of entitlement for widows' benefits is increased, then the calculation of actuarial reductions for widows could be changed to preserve benefit levels or limit benefit cuts for this population. In general, the Panel suggests careful attention should be devoted to linkages between other ages used in social security benefit programs and those applying to widows' (and widowers') benefits.  ?<  r4` hp x ` hp x (#  ?P<C. Expanding the SSI program xThe Panel has discussed above the SSI program in general terms and the pros and cons of lowering the age of SSI eligibility. In addition, other proposals for changing the SSI system have been offered, with the intention of raising the level of income support offered by this plan. xSocial security benefits are currently the single greatest source of income for SSI recipients. The SSI program considers social security benefits as unearned income and thus counts all but $20 of it in determining monthly SSI benefit amounts. This policy imposes a 100 percent tax on all social security benefits over the $20 disregard. In contrast, the first $65 of monthly earned income plus onehalf of all remaining earnings are excluded from counted income.  ? < xOne reform some have suggested is to change the SSI's monthly $20 disregard amount established in 1972 for unearned income by raising it to its inflationadjusted equivalent in 1994 (approximately $70 per month) and then to index this protected amount in the future. This change would increase income adequacy for the 65 percent of aged SSI recipients now receiving social security benefits, and is extremely target"H&0*''JJ-" efficient because all beneficiaries are in or near poverty. This proposal would also increase the income floor for people with low social security benefits, affording them more retirement income insurance. Because SSI is paid out of general revenues, its revenue burden is the same as all other government expenditures paid via the federal income tax. The benefit increase would also go to social security DI recipients, so that benefits would be spread across age cohorts. It is unlikely to have much of an effect on national saving, but would place a small additional burden on general revenues. If one considers OASDI and SSI as two parts of a system for providing transfers and social insurance to older people and people with disabilities, then an alternative method of payment would be a small decrease in PIA. Any general reduction in PIA that shifted money to SSI would mean a shift from higher to lower income recipients. Transfers (as opposed to benefits linked in an actuarially fair way on contributions) made via SSI rather than social security will always be more target efficient.  ?< xA different proposal would entail cutting SSI benefit reduction rates from 100 to 50 percent, by treating social security benefits as earned income. An argument in favor of this proposal is that some believe that work at younger ages "earns" them eventual social security benefits, to which they remain entitled irrespective of SSI recipiency. All the consequences discussed under the previous reform hold true here as well, but it reduces oldage poverty more effectively than do increases in the first tier of social security benefits (Burkhauser and Smeeding 1981).  ?<  ?p<D. Mandating Private EmployerSponsored Pensions xAmericans on the whole save very little; in fact housing and employersponsored pensions are the major sources of private saving, and saving in these assets have not increased over the past several decades. As was demonstrated, about half the workforce is not covered by a pension at any given time, and many workers reach retirement with small (or no) pension benefits. Some policy makers suggest that this pattern of uneven coverage would be rectified by mandating employersponsored private pensions, and this view has received support as the prospect of cuts in social security benefits looms larger. xWhether mandated private pensions would help increase retirement income security was a topic discussed in Sections III.F and G of this report, where we evaluated arguments for and against converting a portion of the social security system to individual accounts. In this section we focus specifically on the question of whether employer pensions would be an efficient and/or equitable way to handle pension mandating, if it were"H&0*''JJ-" deemed beneficial to require additional private saving in the form of a pension. Two foreign experiences relevant in this regard are that of Australia where a system of mandated private pensions is in its early stages, and Switzerland which also has mandated private pensions (Davis 1995). A major contrast between the experiences of pensions in these countries and that of Chile, described above, is that the former countries have placed the mandate on employers to collect the funds, rather than on individual employees. This has the advantage of using current payroll systems to collect the funds, which is presumably easier for the government to monitor and less expensive than having pension plans collect contributions employeebyemployee. If such a system were implemented in the United States, employers might be able to obtain pension group services less expensively than could individual workers, could research and select pension money managers more efficiently, and might resolve contested claims and queries more effectively. On the other hand, small and medium size employers would almost certainly find that overall labor costs would rise under a mandated employerprovided pension system, particularly those firms which do not currently offer a pension plan. xOne factor the Panel deems worthy of emphasis is that a mandated employer pension is not a "free lunch". Rather, a pension mandate would raise labor compensation, implying that wages and other benefits would have to fall to offset to the extent this raised labor costs. Hence, after some adjustment period, a pension mandate would be expected to result in a relatively full passthrough to other forms of compensation for most uncovered workers, and some disemployment for employees at or around the minimum wage. Most employers would not experience a change in total labor costs, but on the whole, employees would tend to see work as less attractive. This could result in some net reduction in labor supply (given small estimated labor supply  ?<elasticities), and perhaps some decrease in employment.(endnote reference("endnote reference"dQ> ?<#Xx6X@DQ/X@##Xx6X@DQ/X@#"endnote reference"` hp x (#` hp x (#82. (lendnote reference( This discussion assumes the mandated pension is of the definedcontribution variety. The analysis becomes significantly more complicated when the mandated pension is of the definedbenefit variety since the expected value of the current cost to the firm is difficult to estimate, and its actual value is random. Future wage growth, job tenure and interest rates are difficult to predict, so under a definedbenefit formula firms' pension costs are uncertain. Further complicating the analysis is the fact that a rise in the market value of pension funds decreases the current cost of past promises. While this might be seen as simply a working out of the allocation of risk inherent in the previous promise, these fluctuations influence ongoing labor/management negotiations. A further complication comes from the impact of the mandate on firms that already have pension plans with voluntary worker contributions. Here a pension mandate will affect some but not all of the workers and will change the relative labor cost of particular employees at the firm. These issues are not satisfactorily resolved in existing  ?<research.#B\  PUP# d Of course the average impact would not hold for all workers and would depend on whether all employees were uniformly covered. For instance, if employers were exempted from making pension contributions for workers near retirement (e.g. workers older than the normal retirement age under social security), then this group would become relatively less expensive and might benefit from increased employment opportunities.  ?!< xMandating might also have a variety of other effects on the labor market (Gustman, Mitchell, and Steinmeier 1994). Currently, employerbased pensions serve as a sorting device as some employees prefer relatively high nonwage employee benefits, while other employees who have a choice select higher wages and lower benefits. A pension mandate would shrink crossemployer differences in benefit packages, and would decrease this type of"H&0*''JJ-" sorting. The extent to which this effect is good or bad for the economy depends on whether workers are making an informed tradeoff between current income and retirement benefits when selecting their jobs and pay/benefits packages, and whether employers can still design compensation packages that attract and keep the most appropriate workers. There is also the possibility that mandating employersponsored pensions could strengthen aspects of the private retirement income systemon the assumption that individuals would not be permitted to take lumpsum cashouts of their pension accumulations prior to retirement (see Section III.G, above).  ?` <x An OptOut Variant : An alternative approach popular in some circles would allow voluntary, rather than mandatory, switching from social security to employerprovided pensions. The Panel saw this option as not manageable. This is because a great deal of the implicit taxation that occurs under the existing social security benefit and tax structure would be undermined if voluntary opting out were permitted. x xThere are two dimensions to the implicit taxation that occurs under the national mandatory social security system. Substantial intragenerational taxation takes place. First, the social security PIA benefit formula is progressive relative to lifetime earnings (AIMEs). As a result, if opting out were permitted, highwage earners facing relatively high payroll taxes compared with expected social security benefits would be anticipated to opt out of the system. In addition, different groupings for annuitization are part of the system s implicit taxation; that is, people with different expected lengths of lives have different lifetime needs, earning capacity held constant. Therefore, based on optimal social insurance principles, a wellfunctioning system redistributes toward those with longer expected lives, lifetime earnings held constant. Allowing an optout would undermine withingenerational redistribution because the funds would simply not be there to ensure retirement benefits for poor people and those likely to live longer than average. xA second type of redistribution in social security occurs across generations. Today's retirees will receive much more in benefits than they paid into the system under current rules, and additional sums have been promised to those near retirement in excess of the (present) value of their contributions. These promises can only be sustained by taxing future enrollees of social security, who, if faced with an optout possibility, might bankrupt the system by pulling out. Of course this implicit national debt could be shifted to other groups of taxpayers, but it cannot be made to vanish without wiping out promised benefits in excess of current accumulations. "H&0*''JJ-"ԌxIn general, the Panel believes that an optout plan is not manageable without acknowledging and making explicit the implicit tax liability of current and future cohorts from lifetime participation in social security.  ?<` hp x (#` hp x  E. Encouraging Employer Pensions and Private Saving xAs noted in Section II above, many workers reach retirement with little or no employer pension benefits. A full analysis of private pension reform proposals is beyond the scope of this group s charge, but many Panel members find persuasive the argument that tax rules under which employer pension plans operate should be simplified, in order to reduce the administrative costs discouraging small firms interested in establishing companysponsored pension plans. Many also support the idea of having a streamlined set of regulations that companies can follow when establishing a taxqualified definedbenefit or definedcontribution plan. Some Panel members favor raising both contribution and benefit limits permitted under employerprovided pensions, as a means to spur pension coverage and saving, and most advocate better coordination between permissible contribution and benefit levels across definedbenefit and definedcontribution pensions offered by the same firm. Nendnote refe` hp x