In the Social Security reform debate, the issue of how minorities fare under the program's rules has surfaced intermittently. Two recent Heritage Foundation papers have returned to the question of how Social Security treats minorities (Beach and Davis, 1998a, 1998b). These two papers argue that African-Americans and Hispanics receive poorer rates of return from Social Security than they would get if they were allowed to invest their Social Security contributions in private retirement funds made up of 50 percent equities and 50 percent government bonds.
While the primary focus of the Heritage report is the higher rates of return African American and Hispanic minorities might receive from a privatized Social Security system, by singling out two visible minorities the authors also imply that these two minorities are especially disadvantaged by Social Security now relative to the rest of the population. For example, they argue that African Americans' lower life expectancies reduce their rates of return by lowering their lifetime benefits relative to their lifetime contributions.
The authors cannot make a similar argument about Hispanics because their own data show higher rates of return for Hispanics, who have higher life expectancies and lower wages than the general population. In fact, one wonders why the authors single out Hispanics rather than focusing on the general population, since rates of return are lower for the latter than the former. Whatever their motivation, the conclusions the authors reach depend in part on a number of important methodological choices regarding life expectancy and other between-group differences, as well as on assumptions they make regarding the measurement of returns from private investments.
This Issue Brief explores the treatment of African American and Hispanic populations under the Social Security program. After identifying the impact of certain factors that affect how minorities fare under Social Security, we examine the methodological treatment of some of these factors in the Heritage Foundation reports and how that treatment influences their findings. Next, we examine the assumptions in the Heritage reports regarding rates of return on private contributions and how omissions in this part of the analysis might influence the findings of the authors. Finally, we examine other studies that have addressed the impact of Social Security on minorities and discuss their findings.
Factors That Affect Returns
Although Social Security's rules are neutral with respect to sex and race, the program's rules differentiate among individuals based on such characteristics as wage levels, life expectancy, ability to work, and marital status. To the extent race or ethnicity is associated with any of these characteristics, their interaction with program rules may disadvantage or advantage a particular racial or ethnic minority.
Benefit Formula and Wage Levels
Social Security's benefit formula provides proportionally greater benefits to lower-wage workers than to higher-wage earners, i.e., their benefits are higher relative to their contributions. Because both African Americans and Hispanics have lower wages than the general population, they will receive higher replacement rates and higher rates of return on their Social Security contributions than white beneficiaries, other things being equal. Higher-wage workers will still receive higher benefits in retirement than lower-wage workers, but the latter will receive higher lifetime "money's worth" ratios (total lifetime benefits divided by total lifetime contributions, both adjusted for inflation).
African Americans have shorter life expectancies than whites in the United States. However, life expectancy varies with age, and in order to assess the rates of return from Social Security, life expectancy at birth is a less useful measure than life expectancy at age 20, when one typically enters the workforce, or life expectancy at 65, the normal retirement age.
An average white male has a life expectancy of 74 years at birth, 55 additional years at age 20, 36 additional years at age 40, and almost 16 additional years at age 65. The average black male has a life expectancy of 66 years at birth, 48 additional years at age 20, 31 additional years at age 40, and 14 additional years at age 65. Thus, the life expectancy differential between white and black males narrows from eight years to less than two between birth and retirement age. Among women, the differences are smaller between races, from just over five years at birth to under two at retirement. These differences narrow even more when we compare races at the same income level (HHS, 1998).
These differences suggest that African Americans who survive to retirement age are likely to experience similar treatment from Social Security as the general population. Although a higher percentage of African Americans will contribute to Social Security without receiving retirement benefits, a higher percentage of them will receive disability benefits than the general population (see below). Overall, life expectancy differentials will disadvantage African Americans relative to the general population in terms of their Social Security retirement benefits.
African Americans are more likely than whites to receive Social Security Disability Insurance benefits before retirement age. In this case, disability benefits provide them with an advantage that reduces their disadvantage due to shorter life spans. Those who become disabled prior to attaining age 65 and are forced to stop working due to their disability receive benefits sooner and cease making contributions sooner than retirees, so their rates of return are higher. About 18 percent of disability beneficiaries are African American, a level about 50 percent higher than their representation in the total population. This disparity is even more acute at younger ages (SSA, 1998), so that the shorter life expectancies of African Americans at younger ages noted earlier is partly offset by their greater tendency at younger ages to be eligible for disability benefits, rather than having to wait until a retirement age they may not reach.
Social Security provides benefits to spouses of retired workers equal to 50 percent of the benefit received by the worker, even if the spouse has made no contribution into Social Security. The occurrence of married-couple households is substantially lower among African Americans and somewhat lower among Hispanics than among whites. To the extent that minorities are less often married than whites, their rates of return from Social Security may be lower.
Of course, the one-earner household is far less frequently observed today than when Social Security was enacted, and two-earner couples are far more prevalent. But for two-earner couples, the benefit entitlement of one spouse (usually the wife) is often less than half the entitlement of the primary working spouse, in which case the lower-earning spouse will receive the higher spousal benefit. Since there are still substantial percentages of couples receiving spousal benefits, married couples tend to receive higher rates of return relative to contributions than do individual workers, which may affect rates of return to demographic groups differently.
The studies that examine Social Security's treatment of minorities can be divided into those that use actual survey data, such as the Retirement History Survey or the Social Security Continuous Work History Sample, and those that use a "representative worker" or stylized individual case approach. The latter attempts to generalize to the entire population from single cases.
The Beach and Davis studies fall into the latter category, and therefore have the same shortcoming in that they use one case to characterize an entire population distribution.
In addition to the inherent limitations of the approach, a number of methodological flaws or omissions seriously undermine the conclusions reached by Beach and Davis. They have treated life expectancy differences incorrectly, they have not adjusted for the correlation between life expectancy and income and marital status, they have omitted disability benefits, and they have made other assumptions that bias the results in favor of their argument. These issues are discussed in turn below.
Life Expectancy Projections
Beach and Davis erroneously assume that, for individuals of any given age, their average life expectancy is exactly how long each one will live. If male life expectancy at a given age extended past age 65, Beach and Davis assumed that all workers of that age worked to age 65 and then spent the years remaining above age 65 in retirement. As the Deputy Chief Actuary of the Social Security Administration points out, "this approach consistently overestimates the expected number of years of work and consistently underestimates the expected number of years after reaching retirement age" (Goss, 1998).
For example, if 20-year-old males have an average life expectancy of 56 additional years, the Heritage report assumes they all live to age 76, working for 45 years and spending 11 years in retirement. In reality, not all 20-year-olds reach 65. They do so with a specific probability, and even if they reach age 65, they don't necessarily work for 45 years. Furthermore, their number of years in retirement depends on their life expectancy at 65, not at age 20.
As Goss observes, instead of all 20-year-old white and black males working a full 45 years, the expected years of work is 42.2 for white men and 39.1 years for black men (see Table 1). Notice that this difference of 3.1 years implies higher rates of return for African Americans because they contribute less, whereas the Beach and Davis assumption of equal years of work suggests no difference between races. The Heritage study estimated the years in retirement for white men and for black men to be 9.3 and 2.2 years respectively, a difference of over seven years. However, the average expected years in retirement for 20-year-olds depends on the conditional probability of their reaching age 65. Taking this into account, the expected years in retirement for white males is 12.1 years and for 20-year-old black males is 8.1 years, a difference of only four years.
Expected Years of Work and Retirement Years for 20-Year-Old Males in 1997
|White Men||Black Men||White Men||Black Men|
Source: Goss, 1998.
Thus Beach and Davis assume that African American, Hispanic, and white males work longer than actual experience indicates they will work and that they have shorter retirements than actually expected. This gives all males lower rates of return than they actually receive, and it increases the differential between African Americans and whites by assuming their years of work to be more equal while assuming their years in retirement to be more divergent than they actually are.
Correlation of Life Expectancy With Income and Marital Status
A second major flaw of the Heritage Foundation study is that it did not adjust for the fact that life expectancy is highly correlated with both income and marital status, both of which are correlated with race (Panis and Lillard, 1996). The Social Security Office of the Actuary estimates that about 20 percent of the difference in rates of return between low- and high-income workers is accounted for by mortality differences between income groups (Goss, 1998).
Because black males have lower incomes and are more likely to be unmarried than white males, these variables confound the life expectancy results. In other words, if we compared African American and white men with comparable income and marital status, the remaining differences accounted for by race would diminish or disappear.
Although this does not change the fact that African Americans have lower life expectancies which may disadvantage them in terms of their Social Security benefits, it does say that the disadvantage is not mainly due to race.
Few studies have made the adjustment in life expectancy to reflect differences in income and marital status. Aaron (1977) and Garrett (1995) found lower mortality rates among better educated, higher income persons which has the effect of offsetting the progressivity of the Social Security benefit formula. Another study has shown that adjusting for income and marital status reduces the mortality differences between whites and African Americans by more than two-thirds. In that study, for an unmarried average earner, a 4.4 year differential between African American and white males in expected age of death at age 25 was reduced to a difference of 1.4 years when marital status and income were controlled (Panis and Lillard, 1996, p. 8).
Exclusion of Disability Benefits and Taxes
A third significant flaw in the Heritage Foundation report is that its authors excluded disability contributions and benefits from the analysis. The exclusion of disability would erroneously widen the rate of return between races because disability benefits are received disproportionately by African Americans. As noted earlier, about 18 percent of disability beneficiaries are African American, a level about 50 percent higher than their representation in the total population. Few studies have taken disability benefits into account in calculating returns to Social Security, and to that extent they all underestimate the rate of return received by African Americans. Because Beach and Davis claim that African Americans are particularly disadvantaged, the omission of disability benefits seriously qualifies their findings.
Raising Taxes in 2015 to Achieve Actuarial Balance
In setting up their base case to compare with privately invested Social Security contributions, Beach and Davis assume that long-term solvency for Social Security would be achieved by raising payroll taxes, rather than cutting benefits or some combination of both. The report's analysis assumes that taxes will increase beginning in 2015, several years earlier and more than is necessary to achieve solvency. The practical effect of this choice is twofold. First, it increases Social Security contributions relative to benefits by more than is necessary, and therefore makes rates of return in Social Security looklower for both the general population and for minorities than they would otherwise be.
Second, since Beach and Davis have, as noted in Table 1, underestimated the difference between whites and African Americans in terms of number of years worked, the premature increase in taxes in the study's base case exaggerates the amounts that African Americans would contribute relative to whites, and therefore makes African Americans' relative rates of return look worse than they would if taxes were not raised prematurely.
Flaws in the Treatment of Private Accounts
While the Heritage Foundation study is critical of Social Security for being a bad deal for African Americans and Hispanics, it enthusiastically supports privatization of Social Security as offering a better, fairer deal for these minorities than the existing Social Security program. Here, too, Beach and Davis have made assumptions that bias the results against Social Security and in favor of their preferred private accounts approach.
Proponents of privatization, like Beach and Davis, have touted the potential returns from investing in equities over investments in Treasuries or fixed-rate securities. But with higher returns comes greater risk. Typically, privatization advocates have assumed that the average return on equities is an appropriate measure by which to evaluate individual returns. But few or none will actually receive the average return on stocks. Some people will do better than average, and some will do worse. Market analysts are well aware of the fact that investment returns must be adjusted by the risk associated with the volatility of the particular investment, and the risk from equities is much greater than Social Security risk. The omission of any discussion of this issue is a major flaw in the papers.
Because the Social Security system is a huge, centralized system with substantial economies of scale, the administrative costs of Social Security are very low. Administrative costs for Old-Age and Survivors Insurance (OASI), which have fallen fairly steadily since the 1980s, have been less than one percent of contributions for the past decade. The costs for Disability Insurance (DI) have been higher, but still represent only about 2 percent of contributions since1995. Combined, the OASDI administrative costs relative to contributions are less than one percent.
The Employee Benefits Research Institute (EBRI) reports on a study that found administrative costs in firms with more than 10,000 participants to be $49 per participant (EBRI, 1998). Since nearly half of all workers reported taxable earnings of less than $15,000 in 1996, at a contribution rate of between two and five percent (the range which has been proposed in most privatization plans), total annual contributions for the median worker would range from $300 to $750 per year. If administrative costs were $49, they would represent anywhere from 7 percent to 16 percent of contributions for the median worker. Since half the workforce is by definition below the median wage, at this high a rate, administration would absorb much or all of the return for the bottom half of the workforce.
Beach and Davis do not account for the very large transition costs that would be imposed by any shift to a private account plan. Transition costs are the costs to workers of paying benefits for current Social Security beneficiaries while contributing to their own retirement accounts. Beach and Davis estimate what individuals would receive if they only contribute to their own private investment accounts, but they ignore transition costs. Since the government is unlikely to abruptly renege on its obligations to current Social Security beneficiaries, a true estimate of the total costs to workers of a privatization plan would have to add the costs of one's own account and the existing taxes that are currently paid to support today's retirees. In ignoring the transition effects of a shift to private accounts, the report suffers from a shortcoming that characterizes numerous other privatization studies.
Conservative Investment Strategies
Beach and Davis assume that African Americans and Hispanics can receive higher returns on contributions by investing in a conservative mix (50-50) of equities and government bonds. What they and many other privatization advocates ignore is that people are highly risk averse and would probably tend not to invest in riskier options (stocks or stock mutual funds), thus lowering their expected returns. A substantial body of evidence from studies of actual investment behavior suggests that women, people with lower incomes, lower wealth, and members of minorities are more conservative investors and therefore might do less well under a privatized Social Security system (Bajtelsmit and VanDerhei, 1997; Hinz, et al., 1997; Jianakoplos and Bernasek, forthcoming; Riley and Chow, 1992).
The Federal Reserve Board's Survey of Consumer Finances finds that boomers, the first real test case for privatization of Social Security, are not willing to take much risk. Nearly 40 percent of them say they are unwilling to take any risk in investing, and another 40 percent or so are only willing to take average financial risks in expectation of earning average returns (Gist and Wu, forthcoming). We should therefore expect that the returns assumed in the Heritage studies are overstated relative to what people are likely to realize.
Even among more affluent minorities, who have the resources to invest and who can afford to take more risks and invest in some kind of idealized life cycle pattern, risk aversion seems to be fairly high. A recent survey by Yankelovich Partners (done for Ariel Mutual Funds and Charles Schwab) that included only households with over $50,000 income found that 81 percent of whites invested in the stock market, as compared with only 57 percent of African Americans (Harper, 1998).
Evidence from other studies of investment choices in pension plans suggests that, if opportunities were extended to those with lower incomes, they would invest even more conservatively than the Yankelovich study suggests. Conservative investment strategies can offset the putative advantages of individual accounts. A recent EBRI study suggests that a conservative investment (bonds only) strategy would result in inferior net benefits and rates of return for both men and women under a privatization plan compared with Social Security (Olsen et al., 1998).
What Other Literature Tells Us
Studies such as that by Beach and Davis that use a "representative worker" approach are inherently limited because they attempt to generalize from single cases. Several previous studies have estimated the rates of return to whites and minorities from Social Security based on actual population samples (Aaron, 1977; Duggan et al., 1993; Freiden et al., 1976; Hurd and Shoven, 1985; Meyer and Wolff, 1987; Panis and Lillard, 1996; Caldwell, 1998).
Most previous studies have shortcomings that bear on the conclusions regarding Social Security's treatment of minorities. Some do not use actual Social Security earnings histories (Aaron, 1977; Panis and Lillard, 1996), or they exclude survivors and/or disability benefits (Meyer and Wolff, 1987; Hurd and Shoven, 1985; Panis and Lillard, 1996). Perhaps the most careful study is that of Duggan et al. (1993). They used the Continuous Work History Sample, which included actual wage histories of Social Security beneficiaries for a very large sample of workers, including those born between 1895 and 1922 (who retired between 1960 and 1987). They also included survivors benefits in their analysis. They found that African Americans in these cohorts received a slightly higher rate of return from Social Security than whites.
While all of the above studies were based on past cohorts of workers, simulations using the CORSIM dynamic microsimulation model (Caldwell, 1998) have estimated rates of return by sex and race for future beneficiaries (birth cohorts from 1936 to 1970). These cohorts, which will retire between 1998 and 2032, show little difference in median rates of return between African American and white males, and declining differences between African American and white females (see Figure 1). The rates of return for women exceed those for men because women are advantaged by longer life expectancies and by a greater likelihood of receiving dependent or survivor benefits than men.
Summary and Conclusion
The issue of how minorities are treated by Social Security has again come under scrutiny in the debate over Social Security reform. The most thorough research into this subject suggests that, while life expectancy differences may disadvantage minorities, other elements of Social Security advantageous to minorities partially or totally offset this disadvantage. Those who suggest, as the Beach and Davis Heritage Foundation studies have, that minorities will benefit more from a privatized Social Security than from the present system tend to ignore several important factors. They do not adjust the presumed higher returns from equity investment for the increased risk they carry, they tend to ignore increased administrative costs, they generally ignore transition costs, and they overlook the generally high aversion to risk which characterizes most Americans' attitudes toward investment. Given the opportunity, most Americans choose fairly conservative investments. Recent research has shown that such investment strategies could yield poorer, and certainly riskier, results than Social Security provides (Olsen et al., 1998).
Aaron, H. (1977) "Demographic Effects on the Equity of Social Security Benefits," in Feldstein, M. and Inman., R. eds., The Economics of Public Services (London: MacMillan), 151-73.
Bajtelsmit, V. and VanDerhei, J. (1997) "Risk Aversion and Pension Investment Choices," in Gordon, M. et al. (eds.) , Positioning Pensions for the Twenty-First Century, Philadelphia: U. of Pennsylvania Press.
Beach, M. and Davis, G. (1998), "Social Security's Rate of Return," Heritage Center for Data Analysis, Heritage Foundation, Washington, DC.
_____. (1998) "Social Security's Rate of Return for Hispanic Americans," Heritage Center for Data Analysis, Heritage Foundation, Washington, DC.
Caldwell, S. (1998) Strategic Forecasting, Inc., unpublished simulation results from CORSIM model.
Duggan, A. et al. (1993) "Returns Paid to Early Social Security Cohorts," Contemporary Policy Issues, 11 (October): 1-13.
Employee Benefits Research Institute (1998) "Are Individual Accounts Administratively Feasible?" EBRI website.
Freiden, A. et al. (1976) "Internal Rates of Return to Retired Worker-Only Beneficiaries Under Social Security, 1967-70,"Studies in Income Distribution, No. 5, U.S. Dept. of HEW, (October).
Garrett, D. (1995) "The Effects of Differential Mortality Rates on the Progressivity of Social Security," Economic Inquiry, 33 (July): 457-75.
Gist, J. and Wu, K. (forthcoming) "Do Boomers Save and, If So, What For?" Public Policy Institute Issue Brief, AARP.
Goss, Steve. (1998) "Problems with 'Social Security's Rate of Return,' A Report of the Heritage Center for Data Analysis," February 4, memorandum.
Harper, P. A. (1998) "Survey Shows Blacks and Whites Tend to Invest Money Differently," The Washington Post, April 9.
Hinz, R., McCarthy, D., and Turner, J. (1997) "Are Women Conservative Investors?" in Gordon, M. et al. (eds.) , Positioning Pensions for the Twenty-First Century, Philadelphia: U. of Pennsylvania Press.
Hurd, M. and Shoven, J. (1985) "The Distributional Impact of Social Security," in D. Wise, Pensions, Labor, and Individual Choice (Chicago: University of Chicago Press), 193-221.
Jianakoplos, N. and Bernasek, A., (forthcoming), "Are Women More Risk Averse?" Economic Inquiry.
Meyer, C. and Wolff, N. (1987) "Intercohort and Intracohort Redistribution Under Old Age Insurance: The 1962-1972 Retirement Cohorts," Public Finance Quarterly, 15 (July): 259-81.
Olsen, K., et al. (1998) "How Do Individual Social Security Accounts Stack Up?" EBRI Issue Brief No. 195, (March).
Panis, C. and Lillard, L. (1996) "Socioeconomic Differentials in the Returns to Social Security," Working Paper 96-05, Labor and Population Program, RAND, (February).
Riley, W. and Chow, K. (1992) "Asset Allocation and Individual Risk Aversion," Financial Analysts' Journal, (November/December), 32-37.
Social Security Administration, (1998) "Social Security and Hispanic Americans: Some Basic Facts," mimeo.
Social Security Administration, (1998) "Social Security Administration Fact Sheet," mimeo.
U.S. Department of Health and Human Services (1998) Health, United States, 1998, National Center for Health Statistics.
- Because the data on Hispanics are more limited or not available (e.g., the Office of the Actuary of SSA does not even provide "money's worth" estimates for Hispanics because of data limitations), we focus more on African Americans.
- There are no data reported on Hispanics.
- The spouse receives 100 percent of the retired worker's benefit upon the death of the worker.
Written by John R. Gist, AARP Public Policy Institute
May be copied only for noncommercial purposes and with attribution; permission required for all other purposes.
Public Policy Institute, AARP, 601 E Street, NW, Washington, DC 20049