The Heritage Foundation

Center for Data Analysis Report #98-08

December 11, 1998

December 11, 1998 | Center for Data Analysis Report on

Social Security's Rate of Return: A Reply to Our Critics

In January 1998, The Heritage Foundation published the first paper in a series analyzing Social Security's rate of return.1 We presented our findings from a detailed study of the retirement income that typical groups of Americans could expect from the retirement portion of their payroll taxes, and we compared this income with the likely return that could be generated by investing those taxes instead in a conservative portfolio of stocks or bonds.

Experts across a wide spectrum of political opinion now concede that Social Security's retirement program provides a poor return for a lifetime of tax payments--the conclusion of the Heritage study as well. Indeed, President Bill Clinton has argued that Social Security's rate of return needs to be higher.2 Much of the current debate discussing Social Security reform focuses on ways to improve its retirement rate of return--an objective rarely heard just a few years ago.

This new emphasis on Social Security's rate of return has reshaped the Social Security reform debate by connecting the interests of taxpaying workers to such arcane but important concepts as "trust fund balances," "dependency ratios," and other elements of a technical analysis of Social Security's long-term problems. But it also triggered criticisms of Heritage's rate of return analysis. By the time the second Heritage report appeared,3 for instance, the Commissioner of the Social Security Administration, Kenneth Apfel, had given congressional testimony on its alleged methodological shortcomings, and left-leaning think tanks had begun issuing studies criticizing our work.

As the authors of the Heritage study, we responded promptly to several of these criticisms. Meanwhile, Heritage's Center for Data Analysis continued to offer workers in various age, income, and ethnic groupings information about their publicly funded retirement program--information that the Social Security Administration (SSA) often refuses to produce, even when asked by the presidentially appointed Social Security Advisory Council.4 Given the current emphasis on Social Security reform, it is both timely and useful to address specific criticisms of our study and offer a more detailed response.

CRITICISMS AND RESPONSES

The following criticisms either paraphrase or, where appropriate, quote from specific objections to our published rate of return studies.5

On Transition Costs

To impose an arbitrary rule on the model would serve to undermine the vailidity of the analysis as an examination of the "pure" opportunity costs of the current system.


 

 

 


Criticism
The Heritage analysis does not take into account the cost of the transition to a system of private Social Security accounts. The rates of return cited fail to acknowledge that workers entering a private system would have to pay for their own retirement as well as support the benefits paid to those who are currently retired and close to retirement.

Response
The purpose of Heritage's rate of return analysis is to apply a yardstick to measure the performance of the current Social Security system, not to propose or cost out an alternative plan. To that end, the comparison of outcomes under Social Security today with outcomes under a hypothetical private system illustrates the opportunity costs of the current program instead of setting out a specific blueprint for reform. In other words, the Heritage analysis provides a benchmark for comparing alternative reforms.

Rate of return outcomes vary enormously, of course, depending on the transition rules that are adopted. Interim financing could be raised through tax increases, benefit cuts, and the issuance of debt--which pose widely different implications for the rates of return of different groups. To impose an arbitrary transition rule on the model would serve to undermine the validity of the analysis as an examination of the "pure" opportunity cost of the current system.

Moreover, it is far from certain that including transition costs would significantly alter the differences in rates of return between the current system and a private system, since maintaining the current system as a viable long-term program also involves large costs. Nevertheless, Mark Weisbrot of the Institute for America's Future has claimed that "as soon as we take into account the real world costs of moving from Social Security to a system of private accounts, the superior return that the [Heritage] authors calculate for private savings vanishes, and in fact becomes negative."6 In support of his criticism, he cites the increased taxes contained in the 1994-1996 Social Security Advisory Council's Personal Security Account (PSA) proposal to fund the transition to a partially privatized Social Security system.

However, Weisbrot fails to note that the SSA's Office of the Chief Actuary analyzed this PSA proposal and found that, even when transition costs are included, it actually offers a higher rate of return to virtually all participants than the current Social Security system does.7

Table 1 shows the returns calculated by the SSA for a low-income single male worker who made $11,000 in 1995, under both the current system (fully funded, using the SSA's own assumptions) and the Personal Security Account proposal of Carolyn L. Weaver, Sylvester J. Schieber, and several other members of the Social Security Advisory Council.

 

 

 

 

 

 

 

 

On Rate of Return Methods of Calculation

Criticism: Steve Goss, Deputy Chief Actuary of the Social Security Administration's Office of the Chief Actuary, has charged that

[T]he Heritage study erroneously analyzes a single outcome where an individual is assumed to know how long he or she will live.... This approach consistently overestimates the expected number of years of work and consistently underestimates the expected number of years after reaching retirement age. As a result, it grossly underestimates the expected rates of return from Social Security retirement benefits.... Clearly, computed rates of return for all men will be much higher for all men [sic], and, moreover the difference between rates of return for black and white men will be dramatically smaller than if the erroneous Heritage method is used.8

Response
This criticism will be addressed directly later in this section, but it is worth noting here that rates of return for 20-year-old white and black male workers--based on Goss's own data and calculating method--are 0.59 percent and -0.15 percent, respectively. When Goss calculates rates of return for whites and blacks in this same age group, he will find that the return for blacks is below that for whites and is negative.

We chose our method for calculating Social Security's rate of return after careful consideration of the advantages and disadvantages of three alternatives:

  • The "expected value" method involves summing the expected (or "probability adjusted") value of benefits and taxes on a year-by-year basis.

  • The "median value" return method calculates the return to the 50th percentile in a population's mortality distribution, and essentially yields the return below which half of a population would receive less.

  • The "average life expectancy" method involves first calculating a group's life expectancy and then calculating the return from Social Security for a worker who lives to that life expectancy. This method, which we selected, usually yields results that lie between the "expected" return and the "median" return.

Each of these methods has strengths and weaknesses. Goss favors the expected value method. In his discussion of the Heritage analysis, Goss chose to characterize the method we selected as "erroneous" while failing to note some of the disadvantages of the expected value method as a measure of the typical return for members of a demographic group. The expected value method in particular is susceptible to distortion by skewed data. This can make it an unsuitable estimator of the likely return from Social Security for a typical member of a population.

A simple analysis of an imaginary lottery will illustrate this point. Consider a lottery with a single prize of $1,000,000. There are 1,000 contestants, each of whom pays a stake of $900. According to the method suggested by Goss, the expected price for each individual from this lottery would be $1,000, implying an overall positive (net) return of $100. Yet 99.9 percent of the entrants would actually lose $900. It would be misleading to suggest to potential buyers of these lottery tickets that they will receive $100--based on the expected return method.

Although this is an extreme example, there is evidence that the returns from the current Social Security system, and those for African-Americans in particular, are highly skewed in a similar fashion. Preliminary calculations made by Heritage (which will be the subject of a future publication) suggest that, while the calculated expected return for a group of recipients may be positive, a large majority of the members of this group (up to 70 percent in the case of African-Americans) may in fact receive negative returns from the Social Security program.

Thus, while the expected rate of return may be useful to the actuary who is responsible for administering an entire program (such as the administrator of the lottery mentioned above) and must account for all participants (including exceptional cases like the single winner above), it often is a less useful tool for those charged with advising individual participants on how they likely will fare in the program. This is why many actuaries, especially in the private sector, have long recognized the weaknesses associated with the expected value method. In offering investment advice to their clients, actuaries routinely use the average life expectancy method that we employed in our study. Since our objective was to enable ordinary Americans to compare the likely consequences of remaining within today's Social Security program with their likely returns realized from a reform that incorporates some private investments, it was also logical to adopt the average life expectancy method.

Critics not only characterize the nature of Heritage's methodology, but in some cases mischaracterize or misunderstand the data we used. One such critic, former Chief Actuary of the Social Security Administration Robert Myers, mistakenly claimed that Heritage used a life expectancy of exactly 69 years for a 21-year-old African-American male. In fact, we used a life expectancy of 73.81 years, which was based on projections made by the U.S. Census Bureau and the Social Security Administration and takes into account future improvements in longevity.

Perhaps the most flagrant example of mischaracterization of the Heritage approach was the use by the Center on Budget and Policy Priorities (CBPP)9 of a table created by Steve Goss of life expectancies for 20-year-old white and black males in 1997. This table featured prominently in a paper attacking the Heritage rate of return studies. The use of this table was misleading on a number of levels. Among them:

  • The table referred to examples that were not even computed in our study. For example, we did not calculate the rates of return for any white males at all, or for any African-American males born after 1975.

  • The data presented in the Goss table were drawn from a different source (the 1992 Life Tables of the United States10 ) than the one we used and were inappropriate for calculating rates of return from Social Security. In particular, the Life Tables figures are based solely on demographic conditions prevailing in 1992 and, unlike the data used by Heritage, do not take into account likely improvements in life expectancy in the future.

Ironically, despite these shortcomings, the data presented by Goss in this table and prominently featured in the CBPP study can be used to illustrate both the shortcomings of the expected value method favored by Goss and the robustness of the general results calculated in the Heritage study.

According to the data in the 1992 Life Tables, half of all 20-year-old black males who enter the labor force will die before they reach the age of 69.7. Half of all white 20-year-old males will die by age 77. If the retirement age is 65, this means that half of all black male workers will die before receiving Old-Age and Survivors' Insurance (OASI) benefits for 4.7 years, and half of all white male workers will die before receiving OASI benefits for 12 years. According to Goss's expected value method, however, "typical" black and white males would receive, respectively, 8.1 years and 12.1 years of benefits. In reality, over 60 percent of black males and 50 percent of white males will die before
collecting benefits for this length of time.

The expected value method produces results that do not represent the experiences of African-American males. As Table 2 shows, the Goss method suggests that an "average" black male worker fares much better from Social Security (paying taxes for only 4.8 years for each year of benefits) than the median black worker (paying taxes for 9.6 years for each year of benefits). In statistical terms, this difference is due to the concentration of very high rates of return among a very few individuals. But, as noted above, far fewer than half of all black males will receive a rate of return as favorable as the average rate of return estimated by Goss's method. The racial disparity between the return received by the 50th white worker and the return received by the 50th black worker is also much greater than the disparity revealed in Goss's "expected value" method.

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Even if the expected value methodology and data cited by Goss are used to evaluate the rate of return from Social Security, the major conclusions of the Heritage study remain unrefuted. To show this, we calculated the expected rate of return from Social Security for the two men described in the Goss memorandum using his "expected value" method. In line with U.S. Department of Labor data, we assumed that the white worker would earn 118 percent of the national average wage and the black earner would earn 89 percent of the average wage.11 The results are shown in Chart 1.12 Chart 1 shows that a black 20-year-old worker in 1998 can look forward to an inflation-adjusted rate of return of -0.15 percent. His white counterpart, however, will "enjoy" a return of 0.59 percent--better, but nothing that should make him too excited. These calculations show that the real rate of return from Social Security remains well below the measures of the opportunity rate of return, even when the expected value method is used (this is the case whether one uses the 2 percent discount rate used by SSA analysts, the 2.5 to 3 percent available from long-term government securities, or the 7 percent real rate of return that the Social Security Advisory Council estimates to be available from equities). In short, regardless of the method used to measure its return, Social Security remains a poor retirement investment for either minority or non-minority Americans.

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
Treasury Department Findings
A number of critics have referred to a series of studies carried out by U.S. Treasury Department researchers James Duggan, Robert Gillingham, and John Greenlees.13 For example, Steve Goss claimed that

[I]n fact more careful research reflecting actual work histories for workers by race indicates that the non-white population actually enjoys the same or better expected rates of return from Social Security than for the white population. (See Duggan et al., "The Returns Paid to Early Social Security Cohorts," Contemporary Policy Issues (October, pp. 1-13).14

Rather, the aim of our analysis was to compare workers of similar age, income level, and family structure. In this respect, the result of the U.S. Treasury Department studies is unequivocal: For the African-American worker, Social Security offers a worse deal than it does for a white worker with an identical income and family structure.
 
 
 
 
 
 
 
 
 
 
 
 
The evidence from this valuable study, however, has been misused and distorted. For one thing, the studies carried out by Duggan, Gillingham, and Greenlees refer only to workers born in the period before the one covered in our Heritage study. In particular, the report cited by Goss is based on workers who were born between 1895 and 1922 and who retired between the early 1950s and the mid-1980s. By contrast, the Heritage study calculates returns for workers born after 1932 and retiring from 1997 until 2042. These two periods have seen extensive changes, both in the structure of Social Security taxes and benefits and in socioeconomic differentials in life expectancy. For example, recent trends and projections suggest that the longevity gap between African-Americans and whites, and between rich and poor, is growing.15

The other mistake in the use of the Duggan et al. study is that Goss implies we calculated a general weighted average rate of return for all African-Americans and all whites. This is not the case. Such an average is almost impossible to calculate and in practice is meaningless, requiring as it does an amalgamation of workers of all income levels, marital status, ages, etc. Rather, the aim of our analysis was to compare workers of similar age, income level, and family structure.16 In this respect, the result of the U.S. Treasury Department studies is unequivocal: For the African-American worker, Social Security offers a worse deal than it does for a white worker with an identical income and family structure.

Chart 2, which is based on data from the most recent study by Duggan, Gillingham, and Greenlees, shows that black workers born in 1918 can expect a real rate of return from Social Security that is 0.75 percent below that which a white worker with an identical income will receive.17

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

On the Exclusion of Disability Insurance

Criticism
The Heritage study ignores Disability Insurance (DI). Disability Insurance taxes are included, but not disability benefits. When this is corrected, many of the findings are reversed. This is especially true regarding the result that African-Americans have particularly low rates of return from Social Security.

Response
This common objection is simply wrong and is based on a failure to read our study carefully. DI is a separate program within the Social Security system that has its own tax rate and trust fund. Heritage's study explicitly examined only the Old-Age and Survivors' Insurance program within Social Security, ignoring DI taxes as well as benefits.

It is possible to reform the OASI program and leave the Disability Insurance program untouched. With this in mind, both DI taxes and benefits were excluded from the analysis. We carefully accounted for pre-retirement Survivors' Insurance by excluding the taxes necessary to purchase this insurance.

The Heritage study thus constitutes a complete and consistent analysis of the retirement portion of Social Security--and only this portion of Social Security. In effect, it assumes that, in the hypothetical partly private system, Disability Insurance and pre-retirement Survivors' Insurance are retained exactly as they exist under current law.

Moreover, no empirical study exists to support the claim of Social Security's defenders that including the DI program in rate of return calculations will offset the racial differentials embedded within the OASI program.18 Many advocates of the current Social Security system cite higher than average DI payments to black workers as a defense against the criticism that Social Security yields a lower than average retirement rate of return for blacks. Besides the fact that DI payments are made to workers and not retirees,19 the argument that Disability Insurance is the principal means by which Social Security makes up for poor retirement rates of return is a particularly tortured defense of the current system. It is like telling people whose bank gives a poor return on their savings accounts that they should not worry because their homes are insured.

Even if a study of the combined OASDI program as a whole were conducted and led to a narrowing of racial differentials in rates of return, such a study would itself be vulnerable to the criticism that it failed to include the effects of Hospital Insurance (HI)--more commonly known as the Medicare program. Chart 3 shows that medical expenditures are highly concentrated among the very old.

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

The inclusion of HI is likely to increase racial differentials in Social Security's rates of return. Compared with the general population, African-Americans have a much lower probability of reaching the very old ages at which medical costs tend to escalate. For example, according to the 1992 Life Tables cited by Goss, a white male has a 40.1 percent chance of living to the age of 80, while a black male has only a 24.3 percent chance.20

On the Risk of Private Rates of Return

Criticism
Private investments, unlike Social Security, are highly risky. Given that most people are risk-averse, if the returns from a private system are adjusted for uncertainty, they will compare much less favorably with those from Social Security.

Response
Before addressing the risk associated with private investments, it is important to recognize that Social Security is not inherently less risky than private investments. There are at least two major risks associated with Social Security: a demographic risk and a political risk.

  • Social Security's Demographic Risk. Every participant in the Social Security retirement program faces the risk of dying before reaching retirement age. In the event of death, Social Security pays a monthly benefit to a worker's children who are under the age of 18 and to the spouse who cares for these children. However, if a worker is childless or has adult children, the family receives no such pre-retirement Survivors' Insurance benefits, other than a one-time-only death benefit of $255.

Widowed retired spouses sometimes collect Old-Age benefits based on the taxes paid by their husband or wife. If they do so, they receive nothing in return for the taxes they themselves have paid. Thus, when one partner of a married couple dies without leaving children under the age of 18, at least one spouse ultimately loses all of the taxes he or she has paid into the system.

Most workers who die between ages 50 and 70 face a high risk of receiving little or nothing in return for a lifetime of paying Social Security taxes. In most cases, their children, if any, are older than age 18 when they die and are ineligible for pre-retirement Survivors' benefits. Those who die in a slightly narrower age band (ages 50 to 65) are not eligible to collect full Social Security retirement benefits. Those dying at age 70 are eligible to collect less than five years' worth of full Old-Age benefits.

Chart 4, using the National Center for Health Statistics data cited by Goss,21 shows that 13 percent of white males and 22 percent of African-American males will die between the ages of 50 and 65. Another 8 percent of all white males and 11 percent of all African-American males will die between the ages of 65 and 70. Thus, one in three African-American males and one in five white males will die between ages 50 and 70.

 

 

 

 

 

 

 

 


Stanford University economist Daniel Garrett drew on such data and calculated the variation in returns from Social Security for a single cohort of individuals with the same average life expectancy and income. These variations are shown in Chart 5. For this set of workers, the lifetime net present value of participation in Social Security ranges from -$92,259 for the worst-performing percentile to $85,993 for the best-performing percentile, in terms of 1988 dollars in 1990 present values.22

 
 
 















  • Social Security's Political Risk. The political risk in Social Security arises because workers and families do not enjoy secure property rights, which are enforceable in court, over their future Social Security benefits. The U.S. Supreme Court has ruled in Fleming v. Nestor that a worker's claim to Social Security benefits is "non-contractual and cannot be soundly analogized to that of a holder of an annuity, whose right to benefits are [sic] bottomed [based] on his contractual premium payments.... To engraft upon the Social Security system a concept of accrued property rights would deprive it of the flexibility and boldness in adjustment to ever-changing conditions which it demands."23

In other words, the future benefits of retirees are completely dependent upon future voters and politicians. Given the tax burden needed to fund promised benefits under the current system, it seems appropriate to assign a considerable degree of political risk to future Social Security benefits.

On Figuring the Private Rates of Return

Criticism
The rates of return on private investments assumed in the Heritage study are too high. This exaggerates the benefits of a privately held individual account.

Response
We used very cautious assumptions regarding the rates of return paid on private investments. For the years up to 1997, we used the actual annual historical rates of return on bonds and equities. For 1998 and future years, the real rate of return on equities was assumed to be 5.7 percent, and the real rate of return on bonds was projected to be 2.8 percent.

The 5.7 percent real rate of return on equities lies well below the long-term rates found in the professional literature. For example, the Social Security Administration's own 1994-1996 Advisory Council used a projected return of 7 percent on equities after considering a wide range of expert testimony.24 During the 1926 to 1997 period, large company stock returns averaged 7.7 percent after inflation, while small company stocks yielded an average post-inflation return of 9.3 percent.25 Heritage reduced even these returns on equities and used a return of 5.7 percent.

The 2.8 percent return on U.S. government bonds is the same as the long-term rate used by the Social Security Administration in the 1998 Report of the Trustees of the Federal Old-Age and Survivors' Insurance and Disability Insurance Trust Funds.

However, even if ultra-pessimistic predictions regarding the returns on stocks are adopted, the major conclusions of the Heritage study would be unaffected. One critic of the study cited a report by Dean Baker of the Economic Policy Institute26 in which the claim was made that economic growth, as projected in the Social Security Trustees' Report (whose assumptions were used as the basis for the Heritage study), was consistent with a real rate of return on stocks of only 4.5 percentage points. Citing this rate of return on equities does not, however, indict the Heritage analysis: Our assumed rate of return is even lower, at a very cautious 4.25 percent. In the great majority of cases, returns from a private account exceeded returns from Social Security, even where taxes were invested wholly in ultra-low-risk U.S. government bonds.

In our study, we assumed that individuals were extremely risk-averse in their investment strategies and would concentrate their investments among low-yield, ultra-secure investments. The riskiest portfolio we used was one in which half of all investments were made in long-term government bonds and the remainder in a broad market equity index. The projected future rate of return on this portfolio is 4.25 percent, with the bond component returning only 2.8 percent annually.

On Administrative Costs and Private Rates of Return

Criticism
Administrative costs would eat up 1.5 percent to 2 percent of all private funds annually. This would remove much or all of the gains from privatization for most workers.

Response
Heritage's first rate of return study did not consider administrative costs explicitly. Instead, these costs were taken into account implicitly through an assumption of extremely low rates of return on private assets. However, both a Social Security Administration study and empirical data show that administrative fees will be much lower than the critics' 1.5 percent to 2 percent projection. A study by the Actuary's Office for the 1994-1996 Social Security Advisory Council estimated that administrative costs for the Personal Security Accounts (PSA) plan, which would privatize a substantial part of Social Security, would be only 1.0 percent of fund assets.27

In actual practice, costs are even lower. A 1996 U.S. Department of Labor study showed that the administrative costs for private-sector, multi-employer defined contribution plans were only 0.82 percent of assets. The mean administrative cost for Standard & Poor's 500 Index mutual funds was lower still--0.39 percent, according to Lipper Analytical Services.28 And the Thrift Savings Plan, a privatized retirement plan run by the federal government for its employees, has costs for its three funds that range from 0.08 percent to 0.10 percent.

These lower estimates are supported by data from Australia's privatized social security system, in which annual administrative costs average 0.8 percent of fund assets.29 The structure of the plan is also important. Limiting investment options and creating larger investment pools will hold costs down. These are features of most privatization plans. Also, costs decline rapidly after the plan starts. For instance, administrative costs for the Thrift Savings Plan are 76 percent lower than they were when the plan began operations in 1988.30

One low-cost option would be to allow individuals to invest their Social Security taxes in the new 30-year Series I Savings Bonds, which currently pay a return of 3.3 percent over the inflation rate. These bonds can be obtained at virtually no cost, and they pay a substantially higher rate of return than does the current Social Security system.

On the Employer's Share of Payroll Taxes

Criticism
The Heritage study included not only the employee's share of taxes, but also those paid by the employer. This overestimates the costs of the program to workers.

Response
Glen Lane, district manager of the Social Security Field Office in Cedar Rapids, Iowa, was among those who criticized our inclusion of the employer's share of the Social Security tax burden in our study.31 However, the "employer's share" of Social Security taxes is part of the total amount an employer expends on employee compensation, which includes the worker's wages and employer-provided benefits. The ascription of the term "employer's share" is an accounting label, rather than a meaningful distinction. In the absence of Social Security taxes, this money from the employee's paycheck would be available for the worker to invest in a private account or to use as an addition to take-home pay. As Dean Leimer, chief author of the Social Security Administration's own calculations of its rate of return, has noted:

In any event ignoring the employer share of the tax is clearly inappropriate, because it results in the comparison of benefits with taxes that are insufficient to fund those benefits; as a consequence, Social Security appears to be a much better deal than it actually is when all taxes required to fund the program are considered.32

On Judging Social Security's Effectiveness by Its Rate of Return

Criticism
The rate of return is not a proper measure of the effectiveness of the Social Security program. Rather, the system should be judged on social criteria, such as its success in reducing the poverty rate among the elderly.

When the rate of return from Social Security for lower-income workers is below the rate available from alternative investments, the program actually may add to poverty--or at least slow wealth accumulation--by reducing the resources available to a family over their lifetime.

 
 
 
 
 
 
 
 
Response
To be judged effective, a retirement social insurance program not only must protect all workers from the threat of poverty when they are elderly, but also must provide an efficient level of retirement income for the taxes paid. The rate of return measures the difference between the money that Social Security takes from a family and the money that the family receives from Social Security. A low or negative rate of return means that individual families are foregoing higher retirement income because Social Security is returning less to them than they could have accumulated had they been able to invest their payroll taxes in private accounts. When the rate of return from Social Security for lower-income workers is below the rate available from alternative investments, the program actually may add to poverty--or at least slow wealth accumulation--by reducing the resources available to a family over their lifetime.

The founders of Social Security recognized the importance of the program's rate of return. Arthur J. Altameyer, chairman of the Social Security Board from 1937 to 1946 and the first Commissioner of the Social Security Administration, argued against policies that would lead to the evolution of a social security system that robbed workers of the chance of higher lifetime incomes or a more elaborate safety net by subjecting them to rates of return below those available from private markets. As Altameyer stated in 1945,

Therefore, the indefinite continuation of the current contribution rate will eventually necessitate raising employees' contributions later to a point where future beneficiaries will be obliged to pay more for their benefits than if they had obtained this insurance from a private insurance company.... I say it is inequitable to compel them to pay more under this system than they would have to pay to a private insurance company, and I think that Congress would be confronted with that embarrassing situation.33

On Payroll Tax Assumptions

Criticism
Heritage inappropriately assumes that if Social Security is not partially privatized, it will be restored to balance entirely by raising payroll taxes and that this tax increase will begin in 2015, a decade earlier than the Social Security actuaries project would be necessary.34

Response
There are several ways to balance the Social Security system within its current framework. In addition to increases in payroll taxes, Congress could cut benefits, increase the retirement age, and require all state and local government workers to participate. Each of these proposals would have a different impact on workers of different ages and income levels. For example, extending Social Security coverage to all state and local government workers would create a massive unfunded liability among existing state and local employee retirement funds that would have to be corrected either by cuts in payments to retired state and local employees or by increased taxes.35

In their calculations of the rate of return to the current system, Social Security's own actuaries used two assumptions to reflect the financial imbalance in the system. The first of these assumes that the system is balanced through across-the-board cuts in Social Security benefits. The second assumes that balance is achieved by increases in payroll tax rates. Dean Leimer, who authored SSA's rate of return calculations, found that the rate of return from Social Security for workers born between 1932 and 1975 is higher under a regime of payroll tax increases than in a scenario where benefit cuts are used to balance the system.36 This higher return occurs because current workers bear the full costs of benefit cuts while bearing only a partial share of future tax increases.

We used one of the two assumptions adopted by Social Security in its examination of the current system, and the assumption that we selected for Social Security's rate of return was the one that yielded the higher rate of return. Had we chosen the assumption of reduced future benefits, the rate of return would have been even lower.

The Social Security trust funds are composed entirely of U.S. government bonds, which means they are a set of IOUs that one part of the federal government (the U.S. Treasury Department) has written to another branch of the federal government (the Social Security Administration). When the Social Security system starts taking in less money than it needs to pay its promised benefits (as it is scheduled to do in 2013),37 then the federal government as a whole will have to meet the shortfall. It can do this either by redeeming the IOUs in the Social Security trust fund (which would mean raising non-Social Security taxes or cutting non-Social Security spending) or by cutting promised Social Security benefits or raising payroll taxes.

In each case, Social Security participants will have to bear the burden of this shortfall through increased federal non-Social Security taxes, reduced federal non-Social Security spending, Social Security benefit cuts, or Social Security tax hikes. In making their projections, Social Security's actuaries merely assume that the IOUs in the trust fund are redeemed, and do not take into account the non-Social Security tax hikes and spending cuts that the rest of the federal government will have to implement should it repay these IOUs. The day of financial reckoning is easily within the lifetime of the baby boomers and their children. Unless Congress raises taxes or cuts benefits and other spending, the Social Security Trustees will begin calling in their loans to the U.S. Treasury by about 2012. By about 2030, the Trustees will have been paid back all of their loans and will have to begin making sharp reductions in Social Security's basic programs.

A LACK OF COMPETING ANALYSES BY OUR CRITICS

The criticisms leveled at Heritage's rate of return analysis have not succeeded in altering our finding: Social Security offers a very low rate of return for most Americans, including minorities and low-income families. Not only does a low rate of return reduce a family's potential retirement income, but it also diminishes the ability of families to pass wealth on to children.

That Heritage's major finding remains unrefuted is perhaps best underscored by the failure of any of its critics to publish their own estimates of Social Security's rate of return. In advancing their criticisms, neither the Center on Budget and Policy Priorities, nor the American Association of Retired Persons, nor Robert Myers, nor the Institute for America's Future has produced their own estimates of the rate of return for Social Security or the degree to which our estimate is affected by the alleged errors in its analysis.

However, one major question remains: Why has the Social Security Administration itself not published calculations of the impact of the current program (or any of the major reform alternatives) on minorities, especially in light of the fact that it readily answers rate of return questions based on age and income? This stunning silence is puzzling, given that Social Security constitutes the federal government's largest domestic program, that the mortality and income data required to complete such a study are readily available to federal researchers,38 and that the impact on minorities of almost every other federal program has been subjected to extensive analysis.

William W. Beach is John M. Olin Senior Fellow in Economics and Director of The Center for Data Analysis at The Heritage Foundation. Gareth G. Davis is a former Policy Analyst in The Center for Data Analysis at The Heritage Foundation.

In January 1998, The Heritage Foundation published the first paper in a series analyzing Social Security's rate of return.1 We presented our findings from a detailed study of the retirement income that typical groups of Americans could expect from the retirement portion of their payroll taxes, and we compared this income with the likely return that could be generated by investing those taxes instead in a conservative portfolio of stocks or bonds.

Experts across a wide spectrum of political opinion now concede that Social Security's retirement program provides a poor return for a lifetime of tax payments--the conclusion of the Heritage study as well. Indeed, President Bill Clinton has argued that Social Security's rate of return needs to be higher.2 Much of the current debate discussing Social Security reform focuses on ways to improve its retirement rate of return--an objective rarely heard just a few years ago.

This new emphasis on Social Security's rate of return has reshaped the Social Security reform debate by connecting the interests of taxpaying workers to such arcane but important concepts as "trust fund balances," "dependency ratios," and other elements of a technical analysis of Social Security's long-term problems. But it also triggered criticisms of Heritage's rate of return analysis. By the time the second Heritage report appeared,3 for instance, the Commissioner of the Social Security Administration, Kenneth Apfel, had given congressional testimony on its alleged methodological shortcomings, and left-leaning think tanks had begun issuing studies criticizing our work.

As the authors of the Heritage study, we responded promptly to several of these criticisms. Meanwhile, Heritage's Center for Data Analysis continued to offer workers in various age, income, and ethnic groupings information about their publicly funded retirement program--information that the Social Security Administration (SSA) often refuses to produce, even when asked by the presidentially appointed Social Security Advisory Council.4 Given the current emphasis on Social Security reform, it is both timely and useful to address specific criticisms of our study and offer a more detailed response.

CRITICISMS AND RESPONSES

The following criticisms either paraphrase or, where appropriate, quote from specific objections to our published rate of return studies.5

On Transition Costs



To impose an arbitrary rule on the model would serve to undermine the vailidity of the analysis as an examination of the "pure" opportunity costs of the current system.




Criticism
The Heritage analysis does not take into account the cost of the transition to a system of private Social Security accounts. The rates of return cited fail to acknowledge that workers entering a private system would have to pay for their own retirement as well as support the benefits paid to those who are currently retired and close to retirement.

Response
The purpose of Heritage's rate of return analysis is to apply a yardstick to measure the performance of the current Social Security system, not to propose or cost out an alternative plan. To that end, the comparison of outcomes under Social Security today with outcomes under a hypothetical private system illustrates the opportunity costs of the current program instead of setting out a specific blueprint for reform. In other words, the Heritage analysis provides a benchmark for comparing alternative reforms.

Rate of return outcomes vary enormously, of course, depending on the transition rules that are adopted. Interim financing could be raised through tax increases, benefit cuts, and the issuance of debt--which pose widely different implications for the rates of return of different groups. To impose an arbitrary transition rule on the model would serve to undermine the validity of the analysis as an examination of the "pure" opportunity cost of the current system.

Moreover, it is far from certain that including transition costs would significantly alter the differences in rates of return between the current system and a private system, since maintaining the current system as a viable long-term program also involves large costs. Nevertheless, Mark Weisbrot of the Institute for America's Future has claimed that "as soon as we take into account the real world costs of moving from Social Security to a system of private accounts, the superior return that the [Heritage] authors calculate for private savings vanishes, and in fact becomes negative."6 In support of his criticism, he cites the increased taxes contained in the 1994-1996 Social Security Advisory Council's Personal Security Account (PSA) proposal to fund the transition to a partially privatized Social Security system.

However, Weisbrot fails to note that the SSA's Office of the Chief Actuary analyzed this PSA proposal and found that, even when transition costs are included, it actually offers a higher rate of return to virtually all participants than the current Social Security system does.7

Table 1 shows the returns calculated by the SSA for a low-income single male worker who made $11,000 in 1995, under both the current system (fully funded, using the SSA's own assumptions) and the Personal Security Account proposal of Carolyn L. Weaver, Sylvester J. Schieber, and several other members of the Social Security Advisory Council.

On Rate of Return Methods of Calculation

Criticism: Steve Goss, Deputy Chief Actuary of the Social Security Administration's Office of the Chief Actuary, has charged that

[T]he Heritage study erroneously analyzes a single outcome where an individual is assumed to know how long he or she will live.... This approach consistently overestimates the expected number of years of work and consistently underestimates the expected number of years after reaching retirement age. As a result, it grossly underestimates the expected rates of return from Social Security retirement benefits.... Clearly, computed rates of return for all men will be much higher for all men [sic], and, moreover the difference between rates of return for black and white men will be dramatically smaller than if the erroneous Heritage method is used.8

Response
This criticism will be addressed directly later in this section, but it is worth noting here that rates of return for 20-year-old white and black male workers--based on Goss's own data and calculating method--are 0.59 percent and -0.15 percent, respectively. When Goss calculates rates of return for whites and blacks in this same age group, he will find that the return for blacks is below that for whites and is negative.

We chose our method for calculating Social Security's rate of return after careful consideration of the advantages and disadvantages of three alternatives:

  • The "expected value" method involves summing the expected (or "probability adjusted") value of benefits and taxes on a year-by-year basis.

  • The "median value" return method calculates the return to the 50th percentile in a population's mortality distribution, and essentially yields the return below which half of a population would receive less.

  • The "average life expectancy" method involves first calculating a group's life expectancy and then calculating the return from Social Security for a worker who lives to that life expectancy. This method, which we selected, usually yields results that lie between the "expected" return and the "median" return.

Each of these methods has strengths and weaknesses. Goss favors the expected value method. In his discussion of the Heritage analysis, Goss chose to characterize the method we selected as "erroneous" while failing to note some of the disadvantages of the expected value method as a measure of the typical return for members of a demographic group. The expected value method in particular is susceptible to distortion by skewed data. This can make it an unsuitable estimator of the likely return from Social Security for a typical member of a population.

A simple analysis of an imaginary lottery will illustrate this point. Consider a lottery with a single prize of $1,000,000. There are 1,000 contestants, each of whom pays a stake of $900. According to the method suggested by Goss, the expected price for each individual from this lottery would be $1,000, implying an overall positive (net) return of $100. Yet 99.9 percent of the entrants would actually lose $900. It would be misleading to suggest to potential buyers of these lottery tickets that they will receive $100--based on the expected return method.

Although this is an extreme example, there is evidence that the returns from the current Social Security system, and those for African-Americans in particular, are highly skewed in a similar fashion. Preliminary calculations made by Heritage (which will be the subject of a future publication) suggest that, while the calculated expected return for a group of recipients may be positive, a large majority of the members of this group (up to 70 percent in the case of African-Americans) may in fact receive negative returns from the Social Security program.

Thus, while the expected rate of return may be useful to the actuary who is responsible for administering an entire program (such as the administrator of the lottery mentioned above) and must account for all participants (including exceptional cases like the single winner above), it often is a less useful tool for those charged with advising individual participants on how they likely will fare in the program. This is why many actuaries, especially in the private sector, have long recognized the weaknesses associated with the expected value method. In offering investment advice to their clients, actuaries routinely use the average life expectancy method that we employed in our study. Since our objective was to enable ordinary Americans to compare the likely consequences of remaining within today's Social Security program with their likely returns realized from a reform that incorporates some private investments, it was also logical to adopt the average life expectancy method.

Critics not only characterize the nature of Heritage's methodology, but in some cases mischaracterize or misunderstand the data we used. One such critic, former Chief Actuary of the Social Security Administration Robert Myers, mistakenly claimed that Heritage used a life expectancy of exactly 69 years for a 21-year-old African-American male. In fact, we used a life expectancy of 73.81 years, which was based on projections made by the U.S. Census Bureau and the Social Security Administration and takes into account future improvements in longevity.

Perhaps the most flagrant example of mischaracterization of the Heritage approach was the use by the Center on Budget and Policy Priorities (CBPP)9 of a table created by Steve Goss of life expectancies for 20-year-old white and black males in 1997. This table featured prominently in a paper attacking the Heritage rate of return studies. The use of this table was misleading on a number of levels. Among them:

  • The table referred to examples that were not even computed in our study. For example, we did not calculate the rates of return for any white males at all, or for any African-American males born after 1975.

  • The data presented in the Goss table were drawn from a different source (the 1992 Life Tables of the United States10 ) than the one we used and were inappropriate for calculating rates of return from Social Security. In particular, the Life Tables figures are based solely on demographic conditions prevailing in 1992 and, unlike the data used by Heritage, do not take into account likely improvements in life expectancy in the future.

Ironically, despite these shortcomings, the data presented by Goss in this table and prominently featured in the CBPP study can be used to illustrate both the shortcomings of the expected value method favored by Goss and the robustness of the general results calculated in the Heritage study.

According to the data in the 1992 Life Tables, half of all 20-year-old black males who enter the labor force will die before they reach the age of 69.7. Half of all white 20-year-old males will die by age 77. If the retirement age is 65, this means that half of all black male workers will die before receiving Old-Age and Survivors' Insurance (OASI) benefits for 4.7 years, and half of all white male workers will die before receiving OASI benefits for 12 years. According to Goss's expected value method, however, "typical" black and white males would receive, respectively, 8.1 years and 12.1 years of benefits. In reality, over 60 percent of black males and 50 percent of white males will die before
collecting benefits for this length of time.

The expected value method produces results that do not represent the experiences of African-American males. As Table 2 shows, the Goss method suggests that an "average" black male worker fares much better from Social Security (paying taxes for only 4.8 years for each year of benefits) than the median black worker (paying taxes for 9.6 years for each year of benefits). In statistical terms, this difference is due to the concentration of very high rates of return among a very few individuals. But, as noted above, far fewer than half of all black males will receive a rate of return as favorable as the average rate of return estimated by Goss's method. The racial disparity between the return received by the 50th white worker and the return received by the 50th black worker is also much greater than the disparity revealed in Goss's "expected value" method.

Even if the expected value methodology and data cited by Goss are used to evaluate the rate of return from Social Security, the major conclusions of the Heritage study remain unrefuted. To show this, we calculated the expected rate of return from Social Security for the two men described in the Goss memorandum using his "expected value" method. In line with U.S. Department of Labor data, we assumed that the white worker would earn 118 percent of the national average wage and the black earner would earn 89 percent of the average wage.11 The results are shown in Chart 1.12 Chart 1 shows that a black 20-year-old worker in 1998 can look forward to an inflation-adjusted rate of return of -0.15 percent. His white counterpart, however, will "enjoy" a return of 0.59 percent--better, but nothing that should make him too excited. These calculations show that the real rate of return from Social Security remains well below the measures of the opportunity rate of return, even when the expected value method is used (this is the case whether one uses the 2 percent discount rate used by SSA analysts, the 2.5 to 3 percent available from long-term government securities, or the 7 percent real rate of return that the Social Security Advisory Council estimates to be available from equities). In short, regardless of the method used to measure its return, Social Security remains a poor retirement investment for either minority or non-minority Americans.

Treasury Department Findings
A number of critics have referred to a series of studies carried out by U.S. Treasury Department researchers James Duggan, Robert Gillingham, and John Greenlees.13 For example, Steve Goss claimed that

[I]n fact more careful research reflecting actual work histories for workers by race indicates that the non-white population actually enjoys the same or better expected rates of return from Social Security than for the white population. (See Duggan et al., "The Returns Paid to Early Social Security Cohorts," Contemporary Policy Issues (October, pp. 1-13).14

Rather, the aim of our analysis was to compare workers of similar age, income level, and family structure. In this respect, the result of the U.S. Treasury Department studies is unequivocal: For the African-American worker, Social Security offers a worse deal than it does for a white worker with an identical income and family structure.
The evidence from this valuable study, however, has been misused and distorted. For one thing, the studies carried out by Duggan, Gillingham, and Greenlees refer only to workers born in the period before the one covered in our Heritage study. In particular, the report cited by Goss is based on workers who were born between 1895 and 1922 and who retired between the early 1950s and the mid-1980s. By contrast, the Heritage study calculates returns for workers born after 1932 and retiring from 1997 until 2042. These two periods have seen extensive changes, both in the structure of Social Security taxes and benefits and in socioeconomic differentials in life expectancy. For example, recent trends and projections suggest that the longevity gap between African-Americans and whites, and between rich and poor, is growing.15

The other mistake in the use of the Duggan et al. study is that Goss implies we calculated a general weighted average rate of return for all African-Americans and all whites. This is not the case. Such an average is almost impossible to calculate and in practice is meaningless, requiring as it does an amalgamation of workers of all income levels, marital status, ages, etc. Rather, the aim of our analysis was to compare workers of similar age, income level, and family structure.16 In this respect, the result of the U.S. Treasury Department studies is unequivocal: For the African-American worker, Social Security offers a worse deal than it does for a white worker with an identical income and family structure.

Chart 2, which is based on data from the most recent study by Duggan, Gillingham, and Greenlees, shows that black workers born in 1918 can expect a real rate of return from Social Security that is 0.75 percent below that which a white worker with an identical income will receive.17

On the Exclusion of Disability Insurance

Criticism
The Heritage study ignores Disability Insurance (DI). Disability Insurance taxes are included, but not disability benefits. When this is corrected, many of the findings are reversed. This is especially true regarding the result that African-Americans have particularly low rates of return from Social Security.

Response
This common objection is simply wrong and is based on a failure to read our study carefully. DI is a separate program within the Social Security system that has its own tax rate and trust fund. Heritage's study explicitly examined only the Old-Age and Survivors' Insurance program within Social Security, ignoring DI taxes as well as benefits.

It is possible to reform the OASI program and leave the Disability Insurance program untouched. With this in mind, both DI taxes and benefits were excluded from the analysis. We carefully accounted for pre-retirement Survivors' Insurance by excluding the taxes necessary to purchase this insurance.

The Heritage study thus constitutes a complete and consistent analysis of the retirement portion of Social Security--and only this portion of Social Security. In effect, it assumes that, in the hypothetical partly private system, Disability Insurance and pre-retirement Survivors' Insurance are retained exactly as they exist under current law.

Moreover, no empirical study exists to support the claim of Social Security's defenders that including the DI program in rate of return calculations will offset the racial differentials embedded within the OASI program.18 Many advocates of the current Social Security system cite higher than average DI payments to black workers as a defense against the criticism that Social Security yields a lower than average retirement rate of return for blacks. Besides the fact that DI payments are made to workers and not retirees,19 the argument that Disability Insurance is the principal means by which Social Security makes up for poor retirement rates of return is a particularly tortured defense of the current system. It is like telling people whose bank gives a poor return on their savings accounts that they should not worry because their homes are insured.

Even if a study of the combined OASDI program as a whole were conducted and led to a narrowing of racial differentials in rates of return, such a study would itself be vulnerable to the criticism that it failed to include the effects of Hospital Insurance (HI)--more commonly known as the Medicare program. Chart 3 shows that medical expenditures are highly concentrated among the very old.

 

The inclusion of HI is likely to increase racial differentials in Social Security's rates of return. Compared with the general population, African-Americans have a much lower probability of reaching the very old ages at which medical costs tend to escalate. For example, according to the 1992 Life Tables cited by Goss, a white male has a 40.1 percent chance of living to the age of 80, while a black male has only a 24.3 percent chance.20

On the Risk of Private Rates of Return

Criticism
Private investments, unlike Social Security, are highly risky. Given that most people are risk-averse, if the returns from a private system are adjusted for uncertainty, they will compare much less favorably with those from Social Security.

Response
Before addressing the risk associated with private investments, it is important to recognize that Social Security is not inherently less risky than private investments. There are at least two major risks associated with Social Security: a demographic risk and a political risk.

  • Social Security's Demographic Risk. Every participant in the Social Security retirement program faces the risk of dying before reaching retirement age. In the event of death, Social Security pays a monthly benefit to a worker's children who are under the age of 18 and to the spouse who cares for these children. However, if a worker is childless or has adult children, the family receives no such pre-retirement Survivors' Insurance benefits, other than a one-time-only death benefit of $255.

Widowed retired spouses sometimes collect Old-Age benefits based on the taxes paid by their husband or wife. If they do so, they receive nothing in return for the taxes they themselves have paid. Thus, when one partner of a married couple dies without leaving children under the age of 18, at least one spouse ultimately loses all of the taxes he or she has paid into the system.

Most workers who die between ages 50 and 70 face a high risk of receiving little or nothing in return for a lifetime of paying Social Security taxes. In most cases, their children, if any, are older than age 18 when they die and are ineligible for pre-retirement Survivors' benefits. Those who die in a slightly narrower age band (ages 50 to 65) are not eligible to collect full Social Security retirement benefits. Those dying at age 70 are eligible to collect less than five years' worth of full Old-Age benefits.

Chart 4, using the National Center for Health Statistics data cited by Goss,21 shows that 13 percent of white males and 22 percent of African-American males will die between the ages of 50 and 65. Another 8 percent of all white males and 11 percent of all African-American males will die between the ages of 65 and 70. Thus, one in three African-American males and one in five white males will die between ages 50 and 70.


Stanford University economist Daniel Garrett drew on such data and calculated the variation in returns from Social Security for a single cohort of individuals with the same average life expectancy and income. These variations are shown in Chart 5. For this set of workers, the lifetime net present value of participation in Social Security ranges from -$92,259 for the worst-performing percentile to $85,993 for the best-performing percentile, in terms of 1988 dollars in 1990 present values.22

  • Social Security's Political Risk. The political risk in Social Security arises because workers and families do not enjoy secure property rights, which are enforceable in court, over their future Social Security benefits. The U.S. Supreme Court has ruled in Fleming v. Nestor that a worker's claim to Social Security benefits is "non-contractual and cannot be soundly analogized to that of a holder of an annuity, whose right to benefits are [sic] bottomed [based] on his contractual premium payments.... To engraft upon the Social Security system a concept of accrued property rights would deprive it of the flexibility and boldness in adjustment to ever-changing conditions which it demands."23

In other words, the future benefits of retirees are completely dependent upon future voters and politicians. Given the tax burden needed to fund promised benefits under the current system, it seems appropriate to assign a considerable degree of political risk to future Social Security benefits.

On Figuring the Private Rates of Return

Criticism
The rates of return on private investments assumed in the Heritage study are too high. This exaggerates the benefits of a privately held individual account.

Response
We used very cautious assumptions regarding the rates of return paid on private investments. For the years up to 1997, we used the actual annual historical rates of return on bonds and equities. For 1998 and future years, the real rate of return on equities was assumed to be 5.7 percent, and the real rate of return on bonds was projected to be 2.8 percent.

The 5.7 percent real rate of return on equities lies well below the long-term rates found in the professional literature. For example, the Social Security Administration's own 1994-1996 Advisory Council used a projected return of 7 percent on equities after considering a wide range of expert testimony.24 During the 1926 to 1997 period, large company stock returns averaged 7.7 percent after inflation, while small company stocks yielded an average post-inflation return of 9.3 percent.25 Heritage reduced even these returns on equities and used a return of 5.7 percent.

The 2.8 percent return on U.S. government bonds is the same as the long-term rate used by the Social Security Administration in the 1998 Report of the Trustees of the Federal Old-Age and Survivors' Insurance and Disability Insurance Trust Funds.

However, even if ultra-pessimistic predictions regarding the returns on stocks are adopted, the major conclusions of the Heritage study would be unaffected. One critic of the study cited a report by Dean Baker of the Economic Policy Institute26 in which the claim was made that economic growth, as projected in the Social Security Trustees' Report (whose assumptions were used as the basis for the Heritage study), was consistent with a real rate of return on stocks of only 4.5 percentage points. Citing this rate of return on equities does not, however, indict the Heritage analysis: Our assumed rate of return is even lower, at a very cautious 4.25 percent. In the great majority of cases, returns from a private account exceeded returns from Social Security, even where taxes were invested wholly in ultra-low-risk U.S. government bonds.

In our study, we assumed that individuals were extremely risk-averse in their investment strategies and would concentrate their investments among low-yield, ultra-secure investments. The riskiest portfolio we used was one in which half of all investments were made in long-term government bonds and the remainder in a broad market equity index. The projected future rate of return on this portfolio is 4.25 percent, with the bond component returning only 2.8 percent annually.

On Administrative Costs and Private Rates of Return

Criticism
Administrative costs would eat up 1.5 percent to 2 percent of all private funds annually. This would remove much or all of the gains from privatization for most workers.

Response
Heritage's first rate of return study did not consider administrative costs explicitly. Instead, these costs were taken into account implicitly through an assumption of extremely low rates of return on private assets. However, both a Social Security Administration study and empirical data show that administrative fees will be much lower than the critics' 1.5 percent to 2 percent projection. A study by the Actuary's Office for the 1994-1996 Social Security Advisory Council estimated that administrative costs for the Personal Security Accounts (PSA) plan, which would privatize a substantial part of Social Security, would be only 1.0 percent of fund assets.27

In actual practice, costs are even lower. A 1996 U.S. Department of Labor study showed that the administrative costs for private-sector, multi-employer defined contribution plans were only 0.82 percent of assets. The mean administrative cost for Standard & Poor's 500 Index mutual funds was lower still--0.39 percent, according to Lipper Analytical Services.28 And the Thrift Savings Plan, a privatized retirement plan run by the federal government for its employees, has costs for its three funds that range from 0.08 percent to 0.10 percent.

These lower estimates are supported by data from Australia's privatized social security system, in which annual administrative costs average 0.8 percent of fund assets.29 The structure of the plan is also important. Limiting investment options and creating larger investment pools will hold costs down. These are features of most privatization plans. Also, costs decline rapidly after the plan starts. For instance, administrative costs for the Thrift Savings Plan are 76 percent lower than they were when the plan began operations in 1988.30

One low-cost option would be to allow individuals to invest their Social Security taxes in the new 30-year Series I Savings Bonds, which currently pay a return of 3.3 percent over the inflation rate. These bonds can be obtained at virtually no cost, and they pay a substantially higher rate of return than does the current Social Security system.

On the Employer's Share of Payroll Taxes

Criticism
The Heritage study included not only the employee's share of taxes, but also those paid by the employer. This overestimates the costs of the program to workers.

Response
Glen Lane, district manager of the Social Security Field Office in Cedar Rapids, Iowa, was among those who criticized our inclusion of the employer's share of the Social Security tax burden in our study.31 However, the "employer's share" of Social Security taxes is part of the total amount an employer expends on employee compensation, which includes the worker's wages and employer-provided benefits. The ascription of the term "employer's share" is an accounting label, rather than a meaningful distinction. In the absence of Social Security taxes, this money from the employee's paycheck would be available for the worker to invest in a private account or to use as an addition to take-home pay. As Dean Leimer, chief author of the Social Security Administration's own calculations of its rate of return, has noted:

In any event ignoring the employer share of the tax is clearly inappropriate, because it results in the comparison of benefits with taxes that are insufficient to fund those benefits; as a consequence, Social Security appears to be a much better deal than it actually is when all taxes required to fund the program are considered.32

On Judging Social Security's Effectiveness by Its Rate of Return

Criticism
The rate of return is not a proper measure of the effectiveness of the Social Security program. Rather, the system should be judged on social criteria, such as its success in reducing the poverty rate among the elderly.

When the rate of return from Social Security for lower-income workers is below the rate available from alternative investments, the program actually may add to poverty--or at least slow wealth accumulation--by reducing the resources available to a family over their lifetime.
Response
To be judged effective, a retirement social insurance program not only must protect all workers from the threat of poverty when they are elderly, but also must provide an efficient level of retirement income for the taxes paid. The rate of return measures the difference between the money that Social Security takes from a family and the money that the family receives from Social Security. A low or negative rate of return means that individual families are foregoing higher retirement income because Social Security is returning less to them than they could have accumulated had they been able to invest their payroll taxes in private accounts. When the rate of return from Social Security for lower-income workers is below the rate available from alternative investments, the program actually may add to poverty--or at least slow wealth accumulation--by reducing the resources available to a family over their lifetime.

The founders of Social Security recognized the importance of the program's rate of return. Arthur J. Altameyer, chairman of the Social Security Board from 1937 to 1946 and the first Commissioner of the Social Security Administration, argued against policies that would lead to the evolution of a social security system that robbed workers of the chance of higher lifetime incomes or a more elaborate safety net by subjecting them to rates of return below those available from private markets. As Altameyer stated in 1945,

Therefore, the indefinite continuation of the current contribution rate will eventually necessitate raising employees' contributions later to a point where future beneficiaries will be obliged to pay more for their benefits than if they had obtained this insurance from a private insurance company.... I say it is inequitable to compel them to pay more under this system than they would have to pay to a private insurance company, and I think that Congress would be confronted with that embarrassing situation.33

On Payroll Tax Assumptions

Criticism
Heritage inappropriately assumes that if Social Security is not partially privatized, it will be restored to balance entirely by raising payroll taxes and that this tax increase will begin in 2015, a decade earlier than the Social Security actuaries project would be necessary.34

Response
There are several ways to balance the Social Security system within its current framework. In addition to increases in payroll taxes, Congress could cut benefits, increase the retirement age, and require all state and local government workers to participate. Each of these proposals would have a different impact on workers of different ages and income levels. For example, extending Social Security coverage to all state and local government workers would create a massive unfunded liability among existing state and local employee retirement funds that would have to be corrected either by cuts in payments to retired state and local employees or by increased taxes.35

In their calculations of the rate of return to the current system, Social Security's own actuaries used two assumptions to reflect the financial imbalance in the system. The first of these assumes that the system is balanced through across-the-board cuts in Social Security benefits. The second assumes that balance is achieved by increases in payroll tax rates. Dean Leimer, who authored SSA's rate of return calculations, found that the rate of return from Social Security for workers born between 1932 and 1975 is higher under a regime of payroll tax increases than in a scenario where benefit cuts are used to balance the system.36 This higher return occurs because current workers bear the full costs of benefit cuts while bearing only a partial share of future tax increases.

We used one of the two assumptions adopted by Social Security in its examination of the current system, and the assumption that we selected for Social Security's rate of return was the one that yielded the higher rate of return. Had we chosen the assumption of reduced future benefits, the rate of return would have been even lower.

The Social Security trust funds are composed entirely of U.S. government bonds, which means they are a set of IOUs that one part of the federal government (the U.S. Treasury Department) has written to another branch of the federal government (the Social Security Administration). When the Social Security system starts taking in less money than it needs to pay its promised benefits (as it is scheduled to do in 2013),37 then the federal government as a whole will have to meet the shortfall. It can do this either by redeeming the IOUs in the Social Security trust fund (which would mean raising non-Social Security taxes or cutting non-Social Security spending) or by cutting promised Social Security benefits or raising payroll taxes.

In each case, Social Security participants will have to bear the burden of this shortfall through increased federal non-Social Security taxes, reduced federal non-Social Security spending, Social Security benefit cuts, or Social Security tax hikes. In making their projections, Social Security's actuaries merely assume that the IOUs in the trust fund are redeemed, and do not take into account the non-Social Security tax hikes and spending cuts that the rest of the federal government will have to implement should it repay these IOUs. The day of financial reckoning is easily within the lifetime of the baby boomers and their children. Unless Congress raises taxes or cuts benefits and other spending, the Social Security Trustees will begin calling in their loans to the U.S. Treasury by about 2012. By about 2030, the Trustees will have been paid back all of their loans and will have to begin making sharp reductions in Social Security's basic programs.

A LACK OF COMPETING ANALYSES BY OUR CRITICS

The criticisms leveled at Heritage's rate of return analysis have not succeeded in altering our finding: Social Security offers a very low rate of return for most Americans, including minorities and low-income families. Not only does a low rate of return reduce a family's potential retirement income, but it also diminishes the ability of families to pass wealth on to children.

That Heritage's major finding remains unrefuted is perhaps best underscored by the failure of any of its critics to publish their own estimates of Social Security's rate of return. In advancing their criticisms, neither the Center on Budget and Policy Priorities, nor the American Association of Retired Persons, nor Robert Myers, nor the Institute for America's Future has produced their own estimates of the rate of return for Social Security or the degree to which our estimate is affected by the alleged errors in its analysis.

However, one major question remains: Why has the Social Security Administration itself not published calculations of the impact of the current program (or any of the major reform alternatives) on minorities, especially in light of the fact that it readily answers rate of return questions based on age and income? This stunning silence is puzzling, given that Social Security constitutes the federal government's largest domestic program, that the mortality and income data required to complete such a study are readily available to federal researchers,38 and that the impact on minorities of almost every other federal program has been subjected to extensive analysis.

William W. Beach is John M. Olin Senior Fellow in Economics and Director of The Center for Data Analysis at The Heritage Foundation. Gareth G. Davis is a former Policy Analyst in The Center for Data Analysis at The Heritage Foundation.

Endnotes

1. William W. Beach and Gareth G. Davis, "Social Security's Rate of Return," Heritage Foundation Center for Data Analysis Report No. CDA98-01, January 15, 1998.

2. Remarks by President Bill Clinton before the National Forum on Social Security, Kansas City, April 7, 1998.

3. William W. Beach and Gareth G. Davis, "Social Security's Rate of Return for Hispanic Americans," Heritage Foundation Center for Data Analysis Report No. CDA98-02, March 27, 1998.

4. Members of the 1996 Social Security Advisory Council asked the Office of the Chief Actuary to calculate several rates of return based on several factors, including life expectancy, adjusted for income. The Social Security Administration refused their request. See Sylvester Schieber, Rates of Return on Social Security Contributions: Good Deal, Bad Deal, or Do We Even Care? testimony before the Committee on the Budget, U.S. Senate, January 21, 1998.

5. William W. Beach and Gareth G. Davis, "Social Security's Rate of Return," Heritage Foundation Center for Data Analysis Report No. CDA98-01, January 15, 1998; "Social Security's Rate of Return for Hispanic Americans," Center for Data Analysis Report No. 98-02, March 27, 1998; and "Social Security's Rate of Return for Union Households," Center for Data Analysis Report No. 98-06, September 7 1998. See also William W. Beach, Gareth G. Davis, and Sarah E. Youssef, "A State-by-State Analysis of the Returns from Social Security," Center for Data Analysis Report No. 98-05, July 30, 1998.

6. Mark Weisbrot, Flawed Assumptions, Fatal Errors: An Analysis of the Recent Heritage Foundation Report on Social Security's Rate of Return (Washington, D.C.: Institute for America's Future, undated), p. 2.

7. Social Security Advisory Council, "Findings and Recommendations," Report of the 1994-1996 Advisory Council on Social Security, Vol. I, January 1997, Washington, D.C., p. 51.

8. Steve Goss, Deputy Chief Actuary, Social Security Administration, memorandum, "Problems with `Social Security's Rate of Return: A Report of the Heritage Center for Data Analysis,'" February 4, 1998.

9. Kilolo Kijakazi, African Americans, Hispanic Americans and Social Security: The Shortcomings of the Heritage Report (Washington, D.C.: Center on Budget and Policy Priorities, October 5, 1998).

10. National Center for Health Statistics, Vital Statistics of the United States, 1992 Life Tables, Vol. II, Section 6, April 1998. It should be noted that this life table is based only on conditions prevailing in 1992. It does not reflect changes in life expectancy that may occur in subsequent years. The original Heritage analysis uses a life table that was adjusted to take into account changes in longevity. The 1992 Life Table cited here is also the one quoted by Steve Goss in his "Problems" memorandum and is used for the purposes of allowing direct comparison with his examples.

11. These are the ratios of median-wage, full-time-employed white males and black male workers in the final quarter of 1997. See U.S. Department of Labor, Bureau of Labor Statistics Release, "Usual Weekly Earnings of Wage and Salary Workers, Fourth Quarter, 1997," January 22, 1998.

12. In calculating this rate of return, Heritage analysts made a number of assumptions in order to keep the calculation as close as possible to the example contained in the Goss memorandum. It is assumed that current law taxes and benefits continue in effect, even though the Social Security Trustees project that Trust Fund outgo will exceed income from 2013 onwards. The calculations were based entirely on the mortality conditions contained in the National Center for Health Statistics' 1992 Life Tables of the United States, the source used by Steve Goss in his analysis of the life expectancies of the two workers contained in his memorandum. Because mortality rates for 1992 are available only up to age 85, post-age 85 mortality rates in 1992 are assumed to be the same ratio of the death rate at age 85 as they were reported to be in the National Center for Health Statistics' 1989-91 Life Tables of the United States. Only Old-Age and Survivors' Insurance and tax benefits are contained in these calculations.

13. See James Duggan, Robert Gillingham, and John S. Greenlees, "Returns Paid to Early Social Security Cohorts," Contemporary Policy Issues, Vol. 11, No. 4 (October, 1993), pp. 1-13.

14. Goss, "Problems with `Social Security's Rate of Return.'" The authors are puzzled by Goss's criticism that they did not use these data in their rate of return studies, because the Duggan et al. study is based on data that are not available to non-federal researchers.

15. For information on the widening socioeconomic differentials in mortality, see G. S. Popper, W. Hadden, and G. Fisher, "Increasing Disparity in Mortality Between Socioeconomic Groups in the U.S.," New England Journal of Medicine, July 8, 1998.

16. Duggan et al. did estimate an average for all of the observations in their data. However, because of the lack of data on spouses and family members, these calculations cannot be viewed as unbiased estimates of returns received by the entire white and black populations. For a more extensive discussion, see Daniel Garrett, "The Effects of Differential Mortality Rates on the Progressivity of Social Security," Economic Inquiry, July 1995.

17. See Schieber, Rates of Return on Social Security Contributions: Good Deal, Bad Deal, or Do We Even Care?

18. Ibid., p. 30.

19. Disabled retirees may receive an Old-Age benefit that equals their previous DI payment.

20. National Center for Health Statistics, 1992 Life Tables of the United States.

21. Goss, "Problems with `Social Security's Rate of Return.'"

22. Garrett, "The Effects of Differential Mortality Rates on the Progressivity of Social Security."

23. Fleming v. Nestor, 363 U.S. 603 (1960).

24. Social Security Advisory Council, "Findings and Recommendations," Report of the 1994-1996 Social Security Advisory Council, Vol. I, January 1997, p. 35.

25. Ibbotson Associates, Stocks, Bonds, Bills and Inflation, 1997 Yearbook (Chicago, Ill.: Ibbotson Associates, 1998).

26. "Saving Social Security With Stocks; The Promises Don't Add Up," The Twentieth Century Fund, 1997. Also see Weisbrot, Flawed Assumptions, Fatal Errors.

27. David C. John and Gareth G. Davis, "The Costs of Managing Individual Social Security Accounts," Heritage Foundation Backgrounder No. 1238, December 3, 1998.

28. Lipper Analytical Services, unpublished data, October 1998, available from the authors upon request.

29. Insurance and Superannuation Commission, Bulletin, Australian Government Publishing Services, various issues.

30. See Thrift Savings Plan at http://www.tsp.gov/features/tspcx.html#sub3.

31. Glen Lane, "Don't Distort Benefits Offered by Social Security," Cedar Rapids Gazette, February 5, 1998.

32. Dean Leimer, "A Guide to Social Security Money's Worth Issues," Social Security Administration, Office of Research and Statistics, Working Paper Series No. 67, April 1995.

33. Quoted in Schieber, Rates of Return on Social Security Contributions. Also see I. S. Falk, "Questions and Answers on Financing of Old-Age and Survivors' Insurance," memorandum to O. C. Pogge, Director, Bureau of Old-Age and Survivors' Insurance, February 9, 1945, p. 13.

34. Kijakazi, African Americans, Hispanic Americans and Social Security.

35. See Robert J. Scott, Testimony Before the Social Security Subcommittee of the House Ways and Means Committee Concerning Mandatory Social Security Coverage of Public Employees, March 21, 1998.

36. Leimer, "A Guide to Social Security Money's Worth Issues."

37. Social Security Administration, 1998 Report of the Trustees of the Federal Old-Age and Survivors' Insurance and Disability Insurance Trust Funds.

38. See Gareth G. Davis, "Ethnic and Racial Differentials from Social Security Old-Age and Survivors' Insurance," unpublished essay, November 1998, available upon request from the author.

About the Author