Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Considerations in Retirement Income Projections THE CONCERN FOR INCOME SECURITY Historical data on economic well-being for retirees and their families show marked improvements over the post-World War II period. Yet more recent data and projections suggest that the average retiree in future years may face bleaker economic circumstances than the average retiree today. The data also suggest that disparities in income and wealth may widen, not lessen, among the elderly population. The available data and indicators often pose difficult problems of interpreta- tion. For example, there is considerable debate about the meaning of the recent leveling off in rates of pension coverage of the work force and, indeed, about the definition of "coverage;" see Box 2-1. There are also many areas of uncertainty, such as whether the marked postwar trend toward early retirement has ended and whether the recent cost reductions in health care achieved by the spread of man- aged care will prove long lasting or only temporary. Nonetheless, the weight of the evidence supports the conclusion that, with the aging of the population, it will be more difficult to sustain current levels of retirement income security at accept- able costs, and, consequently, that retirement-income-related policies will be the subject of increasing scrutiny and public debate. Current Status On average, retired people and their families today enjoy a reasonable level of 17
8 ASSESSING POLICIES FOR RETIREMENT INCOME ............................................................................................................................. :::::::::::::::::::::::::: ::::::::::: ::::::::::: :::::: ::: :::::::::: ::::: :::::::::: :::::: :::::::: ::::::::::: ::::::::: :::::: ::::::::::::::: ::::::::::::: ::::::: :::::::::: :::::::::::::::::: P::::::::::::::::::::::::::. .:::: i: : :~: :_ : :~ :::::_:_:::~:: : ::,.-A:~: 'd':':':':':':': ~ : : ::: : : : :: :: ~ : :' :_,::::A: : . ::::: :: ::~: :F::~: :~ 1::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: 1 -:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:~-:-:-:-:-:-:-:-:-:-:-:-:-:- B:-:-:-:-:-:-:-~:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-~:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-: l~ it d e-n-~-s~ lint De-d soloing cove-a-e cat oe~ oml-c-d 'I'T""'tO"""l't ~e-rpre~ wltno-d ~ cle-a t.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.: :.:.:.: :.::::::: ::: ::::::: :::::::: ::::::: ::::: :!::::: ::::::: ::: ::: ::::::: ::: ::::: :::: :::::: ::::: ::::::: 1 -:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-----:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:---:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-----:-:-:-:-:-:-:-: I d--e-fi--~-iti-o---n-s - f t P i " Ill Hi the k it e-rs~ e-m-~-l-over~ one-rs~ a ~-ensl-o-n~ --an anal so I-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-.-.-:-:-:-:-:-:-.-.-:-:-:-:-:-.-.-.-:-:-:-:-:-:-:l-.-.-:-:-:-.-.-.-:J-:-:-.-.-.-:-:-:-:-:-:-:-.-.-:-:-:-:-:-:-.-.-.-:-:-:-:-.-.-:-:-:-:-:-.-.-.-:-:-:-:I-:-:-.-.-.-:-:-:-:-:-.-.-.-:-:-:-.-.-:-:-:-:-:-:-:-:-:I-.-.-:-:-: l:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-.-:-:-:-:-:-:-:-:-:-.-.-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-.-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-.-:-: I :::::::::::::::::::::::::::::::::::::::::::::::::::,::d:::::::::i::::::::::::::::::.:: ::::::::::::::::::::::: d:::::::::::::::::::::::::::::::::::::i:::::::::::::::::::::::::::.:: d::::::::::::::::::::d ::::::::::::::::::::::::::d::::::::::::::::::::: IS padlclp~lngln a plan one pension anal vvelTare ~eneTlls Aamlnlsld ~lo-n~ I::::::::::::: :,:::::::::::::::::: :,::::::::::::::: :,,, , 1 1 -:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-.:-------------:-----:-------------:-:-----:-:~-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:---:---:-:-:-:-:-:-:-:-:---:-:-:-:-:---:---:-:-:-:-:-:-:-:-:-:-:-:-----:-:-:-:-- '-:-:-:-:-:-'-:-'-:-'-:-:-:-:-:-:-:-:-:-:-:-'-:-: I ~ t-PWBAJ~ th --1- ~ ---d ti- iti ~ i --i Is-- -d ~ 1- -- t t--th- - 1 -:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-: ~ _ - - .- _ - - - - - ._._ ._ - -. - ~ ~ ~ t t--~ese~ Nat l-- -st--t-- to ~-t-~-~-l-~ tn~ t- -I-l Gig tea f- -Gil Hi l ................. .... ......................... ~_ I ~,,n,,so,,,m,,,n,l,p~ t-m-ploye-r~ s-po-nso- s a plan Tor some or a!! worKem ............................. .............................................................. ........... .............. .............. .. .. .. - , . - . -, , ~sima. w-o-rKe-~. list e-ntlil-ea. To. ece-l- e. s-o met o-e-netlis. from. a. o-l-an. al. t:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: 1 - - - - - - - - - - - - - - - - - - - . - - - - - - - - - - - - - - - .- - - - - - - - - - - - - - - - - - - - .- - - - - %- - - - - - - - - -i - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - i- - - - - - - - - - - - - - - - - - - - I~ n--love-~ n-eo-n-l-e-~ ~ $ it- Hi- . l l:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:.-:d-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-~:-:-:-:-:-:-:-:d-:-:-: :-:_-:_- -:-:-:-_:-:-_:- -_-:-:-:-:-:-:-:-:-:d-:~-:d-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-~:-:-:-:-:-:.-:d-:-:-:-:-:-:-:-:-: it ove-d~n-e~de-rdoalY~YlYy~ alinouan~nesoonsorsnnd en-as- Anne l::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: l-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-: _ ~ :-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-~:-!:-:-:-:-:.-:-:-:-:-:-:-:-:-:~-:.-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-.-:-:-:-:-:-:-:-:~ -:-:-:-:-:-:-:-:-:-:-:-:-:-!:-:-:-:-:~-~:.-:-:-:-:-:-:-:-:-:-:-:-: it l~ne~pa-~-c-~-pall-o-n~ mle~-~c-Iv-~-l-~-an~wo-r e-rs~a-g-ea~l---~a-n-a~ove-~n-as~re~ t-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-::::::::::::: ::::::::::::::::::: :::::::::::::::::::::::::::::::::::: l-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-'-:-:-:-:-:-:-:-:-:-:-:-'-:-:-:-'-'-'-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-'-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-: I """""""" i -~--fI~--a-- b I-4246. I. th ~i d.1979 ED9S th ~d-i E ~ ~~ ~ ~e~ e s--- E se---e- - 1 e- S $ S e- SE 1 l ......................................................... ................................................... ~ ............................................................................................................................ ~ ~ == A.~^ r! ~ ~ ^ ~T r ~.~ ~T~ ~! . ~.r ( E ~ !- -l ~- - -~-~. -~- ~- - - !- !- ~-I - !- I - --- -~ ~-- - - I - --- --~-I -~. -~- I L. ~ t........................ l - - -.-.- - - - - , - t ~ a ~ ~ I - IE-I~ . ~ 1 6 E ~I l............................................................................................................................. ~ + - - - - . - - ., i E trOm~4---~0~p-erce-n~--o e-r---~-n-e~ l . .......... ~ ............................................................................................................................ I i i E h i ~ fr ~2 Q~ E:::::::::::::: :~.- - -.~..:.-: -.::: ~:.-: ~-.:~:.- ~:.-: -.:~:: :...:...~: :..:..: :..~..: ~:~:~:: :. :.: :~. :. :.: :. :-:-~.... :~ ~ ~:~. :~ ~:.-:~. :~ ~.~ t::::::::::::::.::::::::::::::::.:::::::::::::::::::.:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: l-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-~-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-~:-:-:-~:-:-:-:-:-:-:-:-:-:-:-~:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:~-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-~:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:d-~:-~-~ E~ ~-pO '-so-rs-n-l-p~ l - - I~ (self e 1 e E e e Et i ate age a d sal ~ E b e pl e E:::::::::::::::-.-.:::::::::::::::::::::::::.-:::::::::.-.-.::::::::::~::::....::::::::::::::::::::::::::::::::::::~:::.-:::::::::::::: t~ s-'ze~ so~ ten-a ~ t-~-n- ES~ I- E~ s-po-nso-rs E-~-p~ rates~ m-~ ~-e~ ane-cte-~ oy~ c- E-an-g-es~ E-n~ E::::::::::::::::::::::::::::::::::::::::::::::::::::::::~::::::::::::::::::::::::::::: :::::::::::::: ::: ::::::: ::::::::: ::::::::::::::::: :::::: ::::::::::::::::::::::::::: ::::: :::::::::::::::::::::::::::::: E -r -r 1- ~-i~ t-l~ t-h~ 1~ r~ r---+h~ ,-- h\~ -t--r d -r -- Id ~d -r-~ r [~ ~!-!- -I ~IV V ~ ~ ! ~- L! I ~I ~ I V =~! V! I ~;! I~ I V ~ I ~ ~~! ~ LI! I I ~ ~! I I ~ I ~ V ~.. =~, d !- -I ~ - - -~!- I - l ~ ~ ............................................................................................................................ E~ nl ee J Ee E e s the h E e i E he e haa e~ i f E ~- ~- -- Y ~ ~y ~y- ~. .. .. .... .... economic security in the United States9 particularly in comparison with earlier generations (although some elderly groups are much less secure). In 1959 the measured poverty rate for people aged 65 and older was 35.2 percent, well above the rate of 22.4 percent for the total population; in 1995 the poverty rate for the
CONSIDERATIONS IN RETIREMENT INCOME PROJECTIONS 19 elderly was just 10.5 percent, below the rate of 13.8 percent for the total popula- tion (Bureau of the Census, 1996:Tables C-1, C-2. The expansion of the U.S. Social Security system, initially established some 60 years ago, played a major role in improving the economic circumstances of retired workers and their families. The system currently covers an estimated 96 percent of all jobs in the United States. From 1940 (when benefits were first paid) to 1972, Congress has increased benefit levels on 10 occasions, although recent legislative changes have curtailed benefits in several ways. Since 1975, benefits have been indexed annually for inflation. Over the post-World War II period, U.S. workers also benefited from provi- sions of the federal income tax code that encouraged the expansion of employer- sponsored pension plans. In 1993, 47 percent of private nonagricultural wage and salary workers participated in a plan, up from 15 percent in 1940 (Employee Benefit Research Institute, 1994:Tables 9, 11~. Most older people also have income from their own savings (although the amounts are usually small) or from other sources, including earnings from part-time and other jobs. However, not all of today's retirees are in comfortable economic circum- stances: 10.5 percent are poor and another 15 percent are categorized as near poor (with incomes between 100% and 150% of the poverty line). Among the poorest groups are elderly women not living with family members (including widows and others), with a poverty rate of 24 percent, and elderly blacks and Hispanics, with poverty rates of 25 percent and 24 percent, respectively (1995 data; Bureau of the Census, 1996:Table 2~. Future Prospects Looking to the future, there is concern that the economic situation of future retirees and their families will not improve, and may worsen, and that disparities in income and wealth levels among the elderly may widen. There are several factors that contribute to the concern, including trends in Social Security costs, employer pensions, personal savings, health care costs, and other demographic and socioeconomic factors. However, other factors, such as increased efforts by 1While the poverty rate is a useful measure of income adequacy, the arbitrariness of the poverty line should be acknowledged, and the official poverty measure is flawed in many respects. Imple- menting the recommendations of a National Research Council panel for an improved measure would somewhat alter the picture of who is poor. However, the elderly poverty rate would not differ markedly today from the total population rate ( 1 or 2 percentage points higher or lower depending on how the measure is altered), and the conclusion that the poverty rate fell disproportionately for the elderly over time would almost certainly not change (see Betson, Citro, and Michael, 1997; Citro and Michael, 1995).
20 ASSESSING POLICIES FOR RETIREMENT INCOME employers and the government to educate workers about the need to save for retirement, may have countervailing effects.2 Social Security The current level of Social Security benefits cannot be sustained indefinitely at the current tax rates. One reason, which has been widely reported in the media, is the demographic transition: because of the aging of the baby boom generation and the relatively small size of the baby bust generation, there will eventually be too few workers relative to the number of retirees to provide the needed level of contributions. The 1996 annual report of the Old- Age, Survivors, and Disability Insurance (OASDI) trustees projects that the old- age and survivors insurance trust fund will start running a deficit by the year 2015 and will be increasingly unbalanced through the rest of the 75-year projection period (to 2070~. The disability insurance trust fund is projected to run a deficit as early as 2003, while the Medicare hospital insurance trust fund has already begun to run a deficit and is expected to be exhausted by 2001. These figures are based on the "intermediate" projections of the Social Secu- rity actuary (Board of Trustees [OASDIi,1996:Table II.F13; see Board of Trust- ees [HIi,1996:Table II.D2 for Medicare). The high-cost (pessimistic) projection shows the old-age and survivors insurance trust fund running a deficit as early as the year 2000 and markedly out of balance thereafter. The low-cost (optimistic) scenario shows the trust fund only modestly out of balance for the period 2020- 2070. The expected Social Security trust fund imbalance must be corrected by some means, which could include reductions in benefits, increases in payroll tax rates, or both. The problems with the Social Security system are particularly important because of pressures on other sources of retirement income and be- cause a significant fraction of the elderly rely almost entirely on Social Security benefits for retirement income. In 1992,14 percent of Social Security beneficia- ries aged 65 and older had no income from other sources, and another 21 percent had other income of less than $2,000 (U.S. House of Representatives, 1994:Table A-9. Pension Coverage Trends in employer-sponsored pensions are mixed. The number of civilian workers aged 16 and over who report that they participated in an employer-sponsored pension plan increased from 53 million in 1979 to 67 million in 1993. However, the percentage of all workers aged 16 and over who participated in a pension plan remained about the same over this period: 46 percent in 1979 and 44 percent in 1993, indicating that the expansion of em 2Recent data indicate that education efforts by employers offering defined contribution pension plans can pay off in terms of the proportion of employees who choose to participate and the level of their contributions (Employee Benefit Research Institute, 1996b).
CONSIDERATIONS IN RETIREMENT INCOME PROJECTIONS 21 ployer-sponsored pension coverage in the post-World War II era may have reached a plateau (Employee Benefit Research Institute, 1994:Table 1 1; see also Pension and Welfare Benefits Administration, 1994:Tables B16-B17~. On the positive side, the percentage of workers who were vested in a plan and therefore entitled to receive benefits increased from 24 percent in 1979 to 38 percent in 1993. (The vesting rate increased more rapidly for women than men.) Pension plan participation among workers is in large part a function of how many of them work for employers that sponsor plans. The sponsorship rate for civilian workers aged 16 and over has also leveled off (it was 56% in 1979 and 57% in 1993), due to such factors as the shift of employment from manufacturing to service companies, which are less likely to offer pension plans, and the declin- ing influence of labor unions. In addition, the standards for pensions to qualify for favorable tax treatment may have led some companies to discontinue or scale back their plans, although recent legislation to simplify pension regulation may encourage more small companies to offer plans. Pension plan participation is also affected by a plan's coverage provisions. The growth of part-time and temporary employment reduces participation be- cause such workers often are not eligible to participate. Finally, participation for eligible workers is elective in some plans. Younger workers have shown a declining participation rate in recent years, which is partly due to the increasing propensity of employers to offer elective 401(k) plans that require employee contributions: younger workers are less likely than older workers to participate in such plans. Yet the fact that older workers have historically participated in pension plans at higher rates than younger workers suggests that participation rates will increase for today's younger workers as they age.3 In sum, the picture is mixed. However, should the flat pattern of sponsorship and participation rates continue, there will remain a significant number of work- ers who are not covered by employer pensions, which will only be partly reduced by a continued increase in the vesting rate. Types of Pension Plans Trends in the types of pensions offered to employ- ees have raised questions about the likely contributions of employer pensions to retirement income security. There are two major categories of plans, each with many variants: defined benefit plans, which guarantee a specific retirement benefit, and defined contribution plans (401(k) plans, profit-sharing plans, and others), which provide for contributions by the employee, employer, or both, but do not guarantee a specific benefit. The number of defined contribution plans has grown from 67 percent of total plans in 1975 to 87 percent in 1992. Because many such plans are offered as a supplement to a defined benefit plan and be 3Participation rates peak between ages 40 and 60 (60% in 1993), after which they fall, presumably because employees with vested benefits retire earlier.
22 ASSESSING POLICIES FOR RETIREMENT INCOME cause defined contribution plans are more prevalent among small employers, defined benefit plans are still the primary plan for the majority of active partici- pants. However, the proportion of participants with a primary defined contribu- tion plan is growing (Employee Benefit Research Institute, 1994:Table 14 and unpublished tables).4 There are arguments that the growing popularity of defined contribution plans is increasing the risks to retirement income security. However, much of the growth in defined contribution plans is among workers in small companies who would otherwise have no pension coverage at all (see Silverman, 1993~. More- over, it is not clear which type of plan ultimately provides more security. The benefits from defined contribution plans may be less predictable because of the greater flexibility that many such plans offer workers (e.g., to determine the level of contributions, to allocate contributions among investment vehicles, to make withdrawals for preretirement consumption, etc.), because workers who change jobs may spend their lump-sum distributions before retirement, and because mar- ket fluctuations may adversely affect workers' investment returns. Yet workers covered by defined benefit plans face risks as well, even though their employers (backed up the Pension Benefit Guaranty Corporation) bear the risks from such factors as market downturns. For example, workers may not work long enough to be vested (vesting periods are typically longer for defined benefit than for defined contribution plans), and workers who are laid off in mid- career may receive significantly lower benefits because such plans typically link benefits to years of service and highest years of earnings (see Samwick and Skinner, 1993~.5 Also, because defined benefit plans increasingly offer partial lump-sum payment options, the risks that benefits will not last a lifetime may be increasing for the recipients (and their survivors) of these plans as well as the recipients of defined contribution plans. Indeed, what can perhaps be concluded is that the trend is toward greater flexibility of pension plan provisions generally. This trend may be beneficial (or not detrimental) to retirement income security in the aggregate, but it increases the likelihood of greater disparities in ultimate levels of retirement income. 4These figures are based on administrative data. In contrast, survey data suggest that defined contribution plans are already the primary plan for the majority of workers (Pension and Welfare senents Administration, 1994:Tables sl3-sl4). Moreover, workers who were vested in a defined benefit plan in a previous job may not be aware that they can apply for benefits (although there is a system of notification by the Social security Administration). In general the percentage of retirees who report that they receive benefits is smaller than the percentage of people nearing retirement who report that they participate in a plan. Reasons for the discrepancy may include survey reporting errors, failure to apply, failure to be vested, using lump-sum distributions for other purposes, etc. (Pension and Welfare senents Admin- istration, 1 995b: 12).
CONSIDERATIONS IN RETIREMENT INCOME PROJECTIONS 23 Personal Savings It appears unlikely that personal savings will make up for any shortfalls that occur in Social Security and employer pension benefits. Sur- vey data indicate that many people are approaching retirement with virtually no personal savings of any kind (except, perhaps, a home). For example, among people aged 51-61 in 1992, 25 percent of couples had less than $13,000 in financial assets (e.g., stocks, bonds, savings accounts), 25 percent of single men had less than $1,000 in financial assets, and 25 percent of single women had no financial assets. Adding the value of home equity, 25 percent of couples, single men, and single women aged 51-61 had, respectively, less than $56,000, less than $4,000, and less than $1,200 in such assets (Gustman and Juster, 1996:Table 2- A1~. Aggravating the situation is that people with low asset levels are less likely to have employer pension benefits than are people with higher asset levels (Gustman and Juster, 1996:Table 2-5~. In fact, understanding people's motives for saving and why so many people save so little may be one of the key questions for policy makers, given that such motives influence not only personal savings but also elective pension plan contri- butions. Policy alternatives are likely to differ according to the reasons for lack of saving, such as an inability to save due to low earnings or to not understanding the levels of saving that are needed to maintain a comparable life-style in retire- ment. (In all likelihood, reasons differ across groups of the population.) Health Care Costs Despite efforts to control rising health care costs, it is likely that they will continue to grow faster than incomes for the foreseeable future. Although virtually all elderly people have Medicare coverage and many of them have supplementary private insurance coverage or Medicaid coverage, they incur significantly higher out-of-pocket costs for medical care on average than younger people. In 1987 (the most recent data available), 89 percent of people aged 65 or older incurred out-of-pocket medical care costs, not counting insurance premiums or long-term care costs, and their mean expense was $1,005; the corresponding figures for younger people were 74 percent and $391 (Taylor and Banthin,1994:Table 2~. Also, the elderly on average spend a higher propor- tion of their income on medical care than do younger people: in 1987,23 percent of elderly households compared with only 7 percent of nonelderly households- had out-of-pocket medical expenses that exceeded 10 percent of their income (Taylor and Banthin, 1994:Table 8~. Among the reasons the elderly incur higher costs are that they have higher rates of medical care utilization and they face gaps in coverage (e.g., prescription drugs are not paid for by Medicare) and the requirement to pay insurance premi- ums, deductibles, and copayments. If medical care costs continue to increase, a retirement income stream that would previously have been adequate for living expenses, including health care, may no longer be adequate. Moreover, cutbacks in Medicare, Medicaid, and medical insurance provided to retirees by their former
24 ASSESSING POLICIES FOR RETIREMENT INCOME employers would exacerbate the situation, if elderly people cannot reduce their health care utilization or costs. Such cutbacks may already be occurring. The proportion of all current wage and salary workers who were covered by employer-provided health insurance between 1988 and 1993 declined from 65 percent to 61 percent. The decline was concentrated among private-sector workers; it appeared to stem largely from lower rates of provision of health care plans by employers, rather than from higher rates of nonparticipation by workers at employers offering plans (Pension and Welfare Benefits Administration, 1994:4~. Data on trends in retiree health insurance coverage are harder to obtain, but there are indications that a significant proportion of employers are curtailing health insurance coverage for current and future retirees or plan to do so in the near future (see, e.g., Buck Consultants, Inc., 1993; Lohse, 1993~. Demographic and Socioeconomic Trends Several demographic and so- cioeconomic trends are contributing to the concern about future retirement in- come security. First, rising life expectancies mean that retirement income must be stretched out over a longer period a particular concern when retirement income flows are not indexed for inflation. This is the case for employer pension plan payments for which there is no regular cost-of-living adjustment. Second, lower ages at retirement mean that workers have fewer years in which to save for retirement (and fewer years to contribute to the financing of Social Security benefits for current retirees), while having more years of retire- ment consumption to support. The post-World War II period has seen a marked drop in labor force participation rates for older men. The rates for men aged 60 to 64 declined between 1955 and 1993 from 83 percent to 54 percent; the rates for men aged 65 and over declined from 40 percent to 16 percent (U.S. House of Representatives, 1994:856~.6 Recently, the trend toward earlier ages at retire- ment for men has leveled off, and the labor force participation rate for women aged 60 and over, which is significantly below that of men in the same age group, has increased somewhat.7 Whether the 1983 Social Security amendments to gradually reduce initial benefits (sometimes referred to confusingly as an in- crease in the "normal" retirement age) offset for workers older than ages 65 to 67 by an increase in the delayed retirement credit will lead to later ages at retirement will not be known for some years.8 Third, an increasing rate of disabilities among workers lowers labor supply 6see Ransom, Sutch, and Streib (1 g88) for labor force participation rates for men aged 60 and over for 1870-1937; ~umsdaine and wise (1994) continue the series through Also. 7surkhauser and Quinn (1994) argue that the decline in labor force participation rates has bot- tomed out, while Peracchi and Welch (1994) argue the opposite, particularly for low wage earners. 8For differing views on the likely effects of these provisions, see Gustman and steinmeier (1991), Reimers and Honig (1993), and Rust and Phelan (1993).
CONSIDERATIONS IN RETIREMENT INCOME PROJECTIONS 25 and the opportunities to save and increases the need for income support over a longer period. Although surveys do not appear to show higher rates of disability in the elderly population (see, e.g., Freedman and Soldo, 1994; Waidmann, Bound, and Schoenbaum, 1995), higher proportions of workers have been apply- ing for and receiving public and private disability benefits. Workers receiving Social Security disability benefits as a percentage of covered civilian workers increased between 1960 and 1975 from 0.8 percent to 3.2 percent. The number of beneficiaries declined to 2.7 percent of covered workers in 1985, largely due to a decline in the percentage of applications that were approved, but it increased to 3.1 percent in 1991 (calculated from U.S. House of Representatives, 1994:Tables 2-8, 3-5; see also Lewin-VHI, 1994~. Also, there appears to be a trend among workers toward younger ages for first receipt of disability benefits (Rust, 1993~. All of these trends could result in a situation in which, absent changes in policy and behavior, future retirees as a group will be worse off than current retirees. It is even more likely that retirement income levels will vary widely: some people will be relatively well off (e.g., people who retained jobs in indus- tries with good benefits), while many others will be badly off. Of course, the worst case may not develop because of countervailing factors. For example, higher immigration levels could mean increased revenues to the Social Security system; the spread of managed care could achieve significant, sustained reduc- tions in the rate of increase in health care costs; and stepped-up efforts by the government, employers, and financial investment companies to educate workers about the need to increase pension plan contributions and other savings could have a positive effect. Whether a worst-case or best-case situation will occur cannot be known, but policy adjustments (e.g., some changes in Social Security financing) will very likely be necessary to maintain future retirement income security even if such factors as low rates of personal saving show improvement. Moreover, to the extent that needed changes are delayed, they may have to be larger in magnitude.9 The question is how to evaluate the range of policy alternatives that may be considered. NEAR-TERM POLICY OPTIONS Debates about retirement income security in the United States are likely to focus on four broad policy areas: Social Security, employer-provided pensions, other personal savings and wealth, and health care needs and costs. The kinds of policy 9See 1994-95 Advisory Council on Social Security Technical Panel on Trends and Issues in Retirement Saving (1995) for an extensive discussion that reaches similar conclusions. For other reviews, see Congressional Budget Office (1993), Employee Benefit Research Institute (1994), and Reno (1993).
26 ASSESSING POLICIES FOR RETIREMENT INCOME options in each of these areas for which decision makers are likely to want cost and other projections vary widely in scope and complexity and, hence, in the challenges they pose for projection models. Below we identify many possible options in each area, but it is important to note that recent policy debates have shared several primary themes: a focus on containing costs (e.g., of employer pension and health care benefits, Medicare and Medicaid); a concern to maintain a mixed public-private system of benefits in which employers play an important role; and, perhaps most important, a focus on putting more responsibility on individuals to save for their own retirement and health care needs, through their employers and other means. Proposals to stimu- late increased savings by individuals include simplifying pension regulations to encourage small businesses to set up certain kinds of defined contribution benefit plans for their workers, increasing incentives to contribute to Individual Retire- ment Accounts (IRAs), and providing for tax-advantaged medical savings ac- counts that encourage people to limit their health care spending. (Proposals to privatize part or all of Social Security payments would give individual workers choices about fund investments, but not about whether to participate in the pro- gram.) These and similar options place a premium on understanding individual savings behavior, including how people react to perceived risks, and on the behavior of employers with regard to compensation policies two areas in which data currently are very inadequate and research knowledge is weak. These kinds of options also call for capabilities to simulate the effects of random or cyclical shocks (recessions, bull or bear stock markets, etc.) that can be expected to affect employment and savings growth opportunities differently across time and popu- lation groups. Social Security Social Security clearly needs and will receive attention to consider ways to re- store a balance between costs and benefits. It seems likely that the policy debate will look first for incremental solutions that spread the costs broadly. Such a goal will likely require modeling the effects of a combination of several small changes in the system (instead of larger changes in one or two features). Such changes include the following, each of which is framed in terms of the direction that is needed to restore balance to the Social Security trust fund: . raising the "early" retirement age (the age at which people can first collect benefits); · increasing payroll tax rates for employers, employees, or both; · increasing the payroll tax base, perhaps by removing the current cap on the amount of earnings that is subject to the old-age and survivors insurance
CONSIDERATIONS IN RETIREMENT INCOME PROJECTIONS 27 portion of the payroll tax and making all earnings subject to tax (as has already been done for the Medicare portion); · decreasing benefits by changing the benefit formula;l° · adding a means test for benefits, so that retirees with other sources of income or assets receive smaller Social Security payments; · increasing the amount of benefits that is subject to the income tax; · repealing the recent law that increased the amount of earnings that benefi- ciaries can receive without a reduction in benefits; · reducing the yearly cost-of-living adjustment; and · investing some part of current trust fund surpluses in stocks instead of U.S. Treasury notes (as recommended by the Social Security Advisory Council; see Quinn and Mitchell, 1996~. Consideration of changes in Social Security benefit provisions in some in- stances will require consideration of whether the same changes should be made for other components of the system, such as disability insurance. For example, an increase in the Social Security early retirement age or a reduction in benefits will be less effective in restoring balance to the system if disability benefits and the consequent incentives to apply for disability payments are not also reduced. There may also be policy interest in modeling yet more radical alternatives, such as partially privatizing the Social Security system, perhaps with the intent of restoring confidence in the system. (Some members of the Social Security Advi- sory Council recommended partial privatization.) At present, public opinion data indicate that younger workers doubt that they will receive Social Security benefits commensurate with the taxes they are paying into the system (see Burtless, 1996:244; see also Employee Benefit Research Institute, 1995~. Priva- tizing some (or all) contributions could restore the confidence of younger work- ers that their taxes will in due course be returned to them as benefit streams from their "own" retirement accounts. Indeed, a problem with policy changes that tinker with the existing system, instead of attempting radical reform, is that they may increase the perception that the system is at risk and that workers cannot be confident that the rules will stay the same for any length of time. Employer Pensions Social Security is not and was never intended to be the only source of retirement income security: historically, employer-provided pensions have also been ex 10At present, the primary insurance (benefit) amount is calculated with two "bend points": for a worker reaching age 62 in 1996, it was 90 percent of the first $437 of the worker's average indexed monthly earnings (AIME), plus 32 percent of the next $2,198, plus 15 percent of the remaining amount above $2,635, if any (U.S. General Accounting Office, 1996b:52-53). (The AIME is based on covered earnings that are taxable for Social Security, for which the maximum amount in 1996 was $62,700.)
28 ASSESSING POLICIES FOR RETIREMENT INCOME peeled to make an important contribution. There is extensive legislation and government regulation of employer plans and, hence, a need to model the effects of changes to them. Policy changes could be made either to affect workers or employers (or both. Examples of policy changes that could increase retirement income security by requiring workers to maintain their pension contributions for retirement purposes or by encouraging them to increase their contributions in- clude the following: · increasing the penalties for early withdrawals from 401(k) plans and other defined contribution pension plans and restricting loans against such accounts; · allowing workers to contribute to 401(k) and similar plans immediately on employment, without a waiting period (employers can waive a waiting period, but most do not); · requiring workers, when they change jobs, to roll over their pension plan contributions to some other plan (i.e., disallowing lump-sum payments that are not earmarked for retirement purposes); . alternatively, allowing workers who change jobs to carry 401(k) or Sims lar plans to which they and their new employers can contribute; · mandating that workers draw retirement income from employer pensions in the form of regular payments or an annuity (i.e., disallowing any lump-sum benefits); . . raising the age (currently 70.5 years) at which workers must begin draw- ing from their defined contribution pension accounts; · regulating pension investment options; and . plans. increasing tax incentives to contribute to defined contribution pension Perhaps even more important are policy changes that alter employer behav- ior. Examples of policy changes that could increase retirement income security by altering the behavior of employers include the following: · raising the limits on the extent to which employers can prefund defined benefit pension plans; · increasing protections for spousal benefits in pension plans; · equalizing the tax treatment of employee contributions to defined benefit and defined contribution pension plans (contributions to defined benefit plans are i{Recent pension simplification legislation made changes in the direction of some of the policy options listed: for example, increasing the penalty for early withdrawals from a "simple" pension account within the first 2 years after the account is set up (the rules were not changed for 401(k) accounts); removing the requirement for currently employed workers (but not for former employees) that they must begin drawing benefits from their defined contribution pension accounts at age 70.5; and simplifying some administrative rules.
CONSIDERATIONS IN RETIREMENT INCOME PROJECTIONS 29 in after-tax dollars, but contributions to defined contribution plans can be tax deferred); · changing nondiscrimination rules and other regulations so as to encour- age employers to establish and expand pension plans; · changing age-discrimination laws to encourage employers to retain older workers and provide benefits to them (e.g., giving employers flexibility to alter compensation for older workers); · mandating additional pension plan design constraints for employers, such as indexation of pensions and altering the ability to offset employer pensions by Social Security benefits; · allowing employers to opt out of Social Security (as is possible in the United Kingdom) and provide comparable or better benefits through pension plans; and · mandating employer pensions. Some of these policy proposals (e.g., a policy to require all employers to provide pensions) are radical steps that would require careful consideration of a wide range of effects, not only on retirement income security but on the economy more broadly (see "Outcome Criteria" below). In general, it is important to assess whether changes that are proposed to increase retirement security would actually have the desired effect or its opposite effect, for example, by inducing employers to opt out of pension plans. It is also important to consider the interactions of employer pensions with the corporate profits tax and other tax policy provisions and with the Social Security system. Many employer plans are designed to be integrated with Social Security (by paying different amounts of benefits for earnings below the Social Security payroll tax threshold), so that Social Security policy changes may affect these plans. Other Personal Savings and Wealth Personal savings other than pension plan contributions are another potentially important source of retirement income support. IRAs are one type of personal savings that government policy has encouraged in order to increase retirees' income security. There is likely to be policy interest in modeling such options as raising the annual limit on the amount of tax-deferred contributions or allowing withdrawals for other purposes in order to increase incentives to contribute to IRAs, increasing (or decreasing) penalties for early withdrawals, and mandating that income be taken in the form of regular payments or an annuity and not as a lump sum. Generally, it is important to address policy changes that affect people's ability to acquire wealth, broadly defined (including tax-deferred and non-tax- deferred financial investments, property, insurance), and save it for retirement, as well as to pass it on to the next generation. Given the pressures on Social
30 ASSESSING POLICIES FOR RETIREMENT INCOME Security and the leveling off of growth in employer pension coverage, the baby boom generation may need to rely significantly on their own savings or bequests from their parents for retirement income. The direction of policy with regard to parental assets is not clear. There may be reason to encourage older cohorts to hold assets for bequest purposes so as to enhance the retirement income security of their children. In this case, it would be important to model policy changes that protect assets against the risks of long- term health care costs for example, subsidizing long-term-care insurance.l2 Alternatively, there may be reason to encourage current retirees to use their assets to reduce the taxpayer costs of providing health care and other benefits to them. In this case, it would be important to model such policy changes as raising estate tax rates; taxing capital gains on inherited assets; lowering asset limits for pro- gram eligibility (e.g., for Supplemental Security Income) or imposing asset tests (e.g., for Medicare or Social Security); and encouraging reverse-mortgage-type programs to allow the elderly to unlock housing equity. If lack of understanding of needed savings levels is a factor in the low rate of personal savings for current workers, it would be important to model the effects of financial education programs. Such programs include direct education by government agencies (e.g., the Social Security Administration and Department of Labor) and programs to encourage or require employers to provide education to employees about pension plan contribution opportunities and investment strate- gies. Finally, it could also be important to model the effects of changes in federal government fiscal policy on savings and wealth accumulation generally. Health Care Needs and Costs Health care needs and costs are an important, as well as a highly uncertain, aspect of retirement income security and well-being. To the extent that medical care costs continue to rise, particularly for the elderly, no assessment of retirement income security is complete without consideration of health care needs and costs. In this report, we do not address models or modeling issues for the health care system as such. However, for completeness, we identify some of the kinds of policy changes for which there is likely to be interest in modeling. We also believe it is important to have links between projection models of retirement income security and health care costs and benefits. There are a number of possible health care policy changes that relate specifi- cally to retirees. Some of these changes are directed to making adequate health 12The Health Insurance Portability and Accountability Act of 1996 made an important change for current workers by allowing employers to sponsor tax-advantaged long-term-care accounts for their employees.
CONSIDERATIONS IN RETIREMENT INCOME PROJECTIONS 31 insurance coverage more widely available; others are directed to curtailing utili- zation of services and reducing costs: · equalizing the tax treatment of employer-provided and individually pur- chased health insurance; · mandating the provision of health insurance coverage for retirees by former employers; plans; . · providing spousal protection in employer-sponsored health insurance providing government insurance for long-term care (or providing tax in- centives to purchase such insurance); and · making one or a combination of changes to Medicare, such as: changing coinsurance rates and deductibles, changing the mix of benefits (e.g., providing prescription drug coverage but curtailing some other kinds of benefits), increas- ing Medicare payroll taxes, providing incentives for enrollees to move into man- aged care or to set up medical savings accounts, providing vouchers for enrollees to purchase lower cost private insurance, means-testing the receipt of Medicare benefits, improving the monitoring of Medicare claims to reduce unneeded ser- vices and fraudulent claims, or turning Medicare into a catastrophic insurance program with the expectation that private insurance will cover expenses up to the ceiling at which Medicare begins to provide benefits. OUTCOME CRITERIA In order to evaluate the costs and benefits of alternative policy proposals, such as those just listed, decision makers need information on a range of outcomes. In consultation with members of a Social Security Advisory Council technical panel (cochaired by our panel member Olivia Mitchell), we developed a list of outcome criteria (see 1994-95 Advisory Council on Social Security Technical Panel on Trends and Issues in Retirement Saving, 1995~. Retirement Income Adequacy An important evaluation criterion is how a policy change affects the adequacy of retirement income. Measures of income adequacy include the poverty rate for the elderly, which focuses on the most vulnerable subgroup in the population, and the rate at which public or private pension income replaces earnings for all groups of elderly people. Information to apply this criterion requires a capability for disaggregation of projection model inputs and outputs (e.g., disaggregated data on earnings histories to estimate replacement rates for people with lower, average, and higher earnings). Insurance Against Income Fluctuations Another criterion by which to evaluate the entire retirement income system, including Social Security and em- ployer plans, is how well the system provides insurance against unforeseen events
32 ASSESSING POLICIES FOR RETIREMENT INCOME that cut off a worker's flow of income. Thus, Social Security and many employer pensions offer survivor and disability benefits and also permit early retirement for those without appealing job opportunities due to industry downsizing or obsolescence of skills, as well as for those who prefer leisure. (Another pension issue noted above that has an insurance aspect is whether to restrict lump-sum payments of pension benefits and thereby require individuals to insure against the risk of above-average longevity.) A cost of providing such insurance is that its availability undoubtedly intro- duces labor market inefficiencies, such as encouraging early retirement or appli- cation for disability benefits. The question then is whether a proposed retirement income policy change achieves an appropriate balance between increasing insur- ance against unforeseen income shocks and increasing distortions in labor mar- kets. Equity of Taxes and Benefits Within and Across Generations Current retirement income policies often involve transfers within and across generations. For example, in some employer pension plans, healthier people may, on average, receive more lifetime pension benefits than less healthy people with the same contributions because of their longer life expectancy (the same is true of Social Security). A comparison across generations shows that current retirees are the beneficiaries of large net transfers of Social Security benefits from current work- ers who are paying more in payroll taxes and will receive less in benefits relative to their contributions than current (and past) retirees. Assessments of proposed policy changes will often need to consider issues of within- and across-genera- tional equity. Financial Soundness of the Retirement Income System Clearly, an impor- tant outcome criterion for any proposal to change the Social Security system concerns the implications for long-term actuarial balance of taxes and benefits. Reforms that do not provide for balance in the system over the standard 75-year projection period, or even longer, are not likely to be acceptable. Similarly, proposed changes in the rules under which employers maintain their pension plans, and under which individuals save for their retirement, should be consid ered from the perspective of their likely effects on the financial soundness of these plans. Two other issues are the extent to which government guarantees of pension benefits, such as are currently provided by the Pension Benefit Guaranty Corporation, should be provided as a back-up to private-sector guarantees and the appropriate price to charge for these guarantees. Avoidance of Market Inefficiencies Another important evaluation criterion is the extent to which a policy change avoids introducing or exacerbating market distortions with respect to individuals' choices of labor versus leisure and savings versus consumption.
CONSIDERATIONS IN RETIREMENT INCOME PROJECTIONS 33 Although research knowledge is not definitive about the magnitude of the effects on labor-leisure choices, it is clear that Social Security, employer pen- sions, the tax code, and other policy levers (e.g., disability benefits) influence people's choices, including their decisions on hours of work early and late in their careers and their decision about when to retire and whether to retire completely or only partially. It is important to know the extent to which a proposed policy change increases or reduces market distortions in terms of labor-leisure deci- sions. Social Security, employer pensions, the tax code, and other policy levers (e.g., publicly provided health care benefits) also affect people's choices about how they allocate income between savings and consumption during their work years and during their retirement years. It is important to know the extent to which a proposed policy change increases or reduces market distortions with regard to pre-retirement savings in order to support post-retirement consumption. For example, Social Security could move toward ensuring adequate retirement accumulations by making retirement benefit payments depend more on the size of employer pension contributions. But such a policy would lower the value of any employee pension contributions or other personal savings, thus encouraging individuals to save less on their own during their working lives. Such distortions in the savings-consumption choices of individuals throughout their lives can have direct effects on the economy. Encouragement of National Savings It is widely believed that U.S. savings rates (including individual, corporate, and government savings) are too low, ei- ther to provide adequate income support to future retirees or to provide adequate investment levels with which to increase future economic output and the overall standard of living. Hence, one criterion for evaluating a proposed policy change is the extent to which it would increase national savings. For example, current restrictions on the extent to which employers can prefund defined benefit pension plans are designed to increase federal tax revenues in the short run, but these restrictions probably decrease corporate savings, certainly in the short run and perhaps in the long run as well, with the result that the net effect on national savings may be negative (see Schieber and Shoven, 1993; England, 1994~. In this case, it is important to have estimates of the effects, not just on tax revenues, but on aggregate national savings. Encouragement of Overall Economic Efficiency, Competitiveness, and Growth For some kinds of policy changes, it is important to consider effects on the behavior of employers that, in turn, affects the overall efficiency of the national economy, the rate of economic growth, and the ability of the United States to remain competitive in world markets. For example, a policy change to require all or a large proportion of employers to provide health care benefits for their retirees or pensions for their workers could have macroeconomic effects.
34 ASSESSING POLICIES FOR RETIREMENT INCOME Such mandates on employer behavior might affect the industrial structure of the economy or the rate of real economic growth. Cost-Effect~veness of Benefit Program Administration In evaluating the effects of retirement-income-related policy changes, policy makers should recog- nize the existence of important issues involving the efficiency and effectiveness of program administration. For example, "leakage" occurs in the Social Security system to the extent that benefits are paid in the name of deceased people or to disabled people who could return to work. However, minimizing leakage of benefits could entail increased costs for program administration. In the private sector, there is concern about the effects of government regulation of pension and health care benefit plans on employers' costs for plan administration and, in turn, the effects on employer decisions about whether and what kinds of plans to offer. If some or all of the Social Security system were privatized, it would clearly be important to evaluate the likely efficiency and effectiveness of schemes for ad- ministering workers' IRAs. Similarly, administrative costs are an important evaluation criterion for any major change in the health care benefit system, in- cluding changes to Medicare. CHALLENGES TO PROJECTION MODELING Outcomes Evaluating one or more policy proposals on the range of outcome criteria listed above presents formidable challenges for projection models. Although few policy changes require information on every criterion, most changes will need to be evaluated on several criteria. There are often trade-offs to consider: for example, selecting the desired balance between benefit reductions and tax increases to restore actuarial balance to the Social Security system. Evaluating proposals on a range of outcome criteria is also important because policy changes can change behavior in ways that alter the original estimates of costs and benefits. For example, a policy change to maintain the adequacy of income for future retirees by, say, requiring increased contributions to employer pension plans or increased Social Security payroll taxes might induce workers to reduce work hours and thereby decrease their pension plan or payroll tax contributions. Program Interactions Because there are many retirement-income-related policy areas and many differ- ent agencies handle different areas, it is difficult to carry out integrated analyses that consider interactions among components of retirement income security and the overall effects of policy changes. Typically, policy debates focus on only one component of retirement income, such as employer pensions or health insurance
CONSIDERATIONS IN RETIREMENT INCOME PROJECTIONS 35 coverage. And decision makers often ask for answers to policy questions that are relatively narrow in scope: for example, how many small businesses will offer pension plans if certain kinds of regulations are relaxed or how the value of self- directed pension plan contributions will be affected by campaigns to educate workers about investment choices. At times, decision makers focus on entirely unrelated aspects of retirement-income-related policy proposals: for example, recent changes in tax laws on pension funding were evaluated for their potential to increase revenue in the short term without considering the possibly adverse effects for retirement income such as leading employers to drop or scale back their pension plans (see, e.g., Lindeman and Utgoff, 1992~.13 It will often be appropriate and cost-effective to have projection modeling tools that address specific policy levers and issues. Yet it is also critical to have broader based modeling tools (or ways to link issue-specific models) to consider the effects of policy changes on retirement income security overall. Even when the focus is on specific policies and narrow issues, it is important to consider such interactions among components of retirement income as the following: · many employer pension plans are linked to the benefit structure of the Social Security system, so that plan provisions could be expected to change if Social Security changes; · Social Security retirement and disability insurance benefits are competing sources of support for early retirees, so that changes in one program can be expected to affect the other; and · newer forms of employer pension plans that allow workers to determine the level of their contributions and the choice of investment vehicles are very similar to other kinds of personal savings, so that there is a strong reason for analyzing them within the same framework. Heterogeneity Retirement income and its components exhibit considerable heterogeneity among population groups. Almost all workers are covered by Social Security, but there is wide variation in the extent to which workers are covered by employer-pro- vided pensions and in the provisions of such coverage. There is also variation in the amounts and form of pension benefits provided upon retirement: for ex- ample, whether benefits are taken as lump sums or annuities and whether annuity payments are indexed for inflation. Finally, there is wide variation in the popula- tion in personal savings behavior, earnings capacity, labor market opportunities, 1 3Sometimes, the broader implications e.g., U.S. Treasury Department, 1991). of tax law changes are studied after their enactment (see,
36 ASSESSING POLICIES FOR RETIREMENT INCOME the availability of income support from other family members, and health care needs and resources. Consequently, there are significant disparities in income and wealth taking account of all sources among retirees and workers, which means it is important to evaluate policy changes for their distributional as well as aggregate effects. Some models may usefully focus on constructing important overall estimates- for example, projections of the balance in the Social Security trust fund by using aggregate, or relatively aggregate, inputs. However, there must also be projection models that can reflect the heterogeneity in the population and show, for example, whether a proposal helps some retirees but not others. These effects are as important to understand as the central tendencies. Indeed, much of the focus of government retirement policy is on ensuring some minimum living standard for the elderly people who may have been unlucky or unwise in prepar- ing for their post-work years. Models are needed that can produce estimates of the likely numbers of such people, as well as estimates of the average experience. In this regard, as the policy process increasingly looks to encourage people to assume more of the burden for their own retirement saving, models are needed that can simulate the effects of differential risk. Currently, promised benefits from Social Security come from the entire tax-paying population, while defined benefit pension plan entitlements come from employer resources, with backing from the Pension Benefit Guaranty Corporation. In other words, the risks to retirement income from such factors as downturns in markets, productivity de- clines, recessions, and poor business judgment are spread widely. In contrast, individual investments in defined contribution pension plans, IRAs, or privatized Social Security accounts are subject to risks that potentially vary widely. Some people, helped by good judgment and fortunate timing, may achieve windfall gains while others suffer large, unanticipated losses. Models need to be able to plausibly simulate variations in rates of return on investments across time and population groups in order to assess the distributional consequences for retire- ment income security within and across age groups and generations. Time Horizon When assessing retirement income security overall and for population groups, it is critical to have models that can estimate the longer term as well as short-term effects of alternative policies. Many changes for example, in the tax treatment of employee or employer contributions to pension plans are not likely to alter the behavior of people who are near retirement, but such changes may well have significant behavioral effects on younger workers and their employers over the longer run. In turn, behavioral responses to policy changes may have yet further consequences for long-term outcomes. The longer the time horizon for a policy projection, the more difficult it is to build in plausible assumptions about policy interactions, behavioral responses,
CONSIDERATIONS IN RETIREMENT INCOME PROJECTIONS 37 and the like. For example, large, unanticipated changes in fertility behavior in the post-World War II period have periodically upset long-range population fore- casts. Projecting over the long term can be more computationally taxing than projecting over the short term, but much retirement-income-related policy analy- sis, by its nature, requires long-term projections. Consequently, there is a pre- mium on building models with capabilities for sensitivity analysis and other methods of estimating the much greater uncertainty in long-range forecasts. Knowledge About Behavior Many behaviors and processes contribute to the retirement income security of current and future generations. The capability to address important questions about many retirement-income-related policy proposals, particularly over the longer term, requires a solid base of research knowledge and ways to incorpo rate that knowledge effectively in projection models in such areas as: · employers' decisions about whether and what types of pension plans and other benefits (e.g., health insurance, disability benefits) to offer employees; · employers' labor demand choices, particularly for older workers; · performance of capital markets that determine the value of pension hold- ings and other forms of saving and of group and individual annuity markets; · people's choices about how much of their income to devote to savings versus consumption at different stages of the life cycle and what kinds of invest- ment decisions to make; · people's labor supply choices at different stages of the life cycle, includ- ing job search and home production; · basic demographic processes that affect the numbers of working-age and retirement-aged people, especially mortality, as well as net immigration, and depending on the length of the projection period fertility; · people's marital histories and living arrangements, joint decisions with other family members about labor supply and savings, and the likelihood of financial and other support from relatives; . interactions of trends in mortality and marital status with trends in retire- ment benefits (e.g., increased use of lump-sum payouts), which may affect retire- ment income security over the remaining lifetimes of retirees and their survivors; · people's health status and disability, including trends over time and changes in status as people age; · trends in health care costs and health care financing arrangements; and · interactions between microlevel behavior (decisions of workers and em- ployers) and the economy.
38 ASSESSING POLICIES FOR RETIREMENT INCOME Uncertainty A final challenge for retirement income modeling is to provide estimates of the uncertainty in projection model outputs, which stem from errors and variability in data inputs, research parameters, and model specifications. The estimation of uncertainty has been much neglected in policy analysis work. When estimates have been done, the methods used such as constructing "high," "intermediate," and "low" scenarios for long-range forecasts of the elderly population or the balance in the Social Security trust fund have often had little scientific basis. Admittedly, the task of estimating uncertainty is difficult. The difficulties are compounded when a projection model is asked to evaluate a complex policy change on several outcome criteria over a long time period and when the change interacts with other policies and behaviors in complex and varied ways for differ- ent groups. Yet it is precisely in these instances that decision makers most need to be aware of the likely wide range of uncertainty in the estimates. The next three chapters address improvements in retirement-income-related research, data, and projection model capabilities. Research on methods to evalu- ate quality and estimate uncertainty should be a priority in all three areas, as should research on helpful ways to convey information about uncertainty to decision makers. Estimates of uncertainty are vital both to inform the policy process and to make possible continued improvements in research knowledge, data, and projection capabilities that can support informed policy debate.
