Appendix I: Empirical Evaluation of the Impact of Social Security Wealth
In one of the first major attempts to evaluate the impact of social security wealth on individual consumption behavior, Feldstein concluded that in 1971 the social security system had halved personal saving. However, the manner in which the social security wealth variable was calculated turned out to be flawed and when that variable was recalculated the effect was found to be insignificant. 61/ Feldstein’s contribution, however, spawned a large literature in which a number of issues were raised.
One specific issue concerns how best to measure an individual’s expected benefits from social security. The traditional approach has been to use the actuarial value of benefits. It has been argued, however, that given the (no bequest) life-cycle perspective the simple discounted value of future benefits (ignoring the possibility of death) would be a more appropriate measure to employ. 62/ A more general theoretical point concerning the linkage between social security and the underlying life-cycle hypothesis is that changes in the parameters of social security affect all the consumption function coefficients as well as the size of social security wealth. 63/
These conceptual difficulties are compounded by problems of econometric identification. 64/ Specifically, social security wealth is typically defined as a deterministic function of other variables such as age, earnings, history, etc. Many of these variables should be included independently in the behavioral equation implying that social security wealth is a function of the other right hand side variables. As a result, in the absence of a detailed knowledge of the functional form of the equation, identification will be extremely difficult and results will be very sensitive to the specification selected.
An alternative strategy for quantifying individuals’ expected social security benefits more directly is to rely on survey data. Bernheim and Levin, for example, use survey data drawn from the Retirement History Survey which tracked a sample of retirement-aged households for a period of ten years, beginning in 1969. 65/ However, the results using survey data also appear mixed, which may in part be due to the ambiguous quality of the data.
APPENDIX II: Structure of Growth Model
The illustrative simulations presented in the text are based on a simple growth model with accompanying data bank that was developed at the Brookings Institution to consider the implications of prospective demographic developments on the social security system. 66/ The assumptions underlying the calculations are based on the mid-range economic assumptions used by the actuaries of the Social Security Administration. 67/
The model consists of five sectors: nonfarm business, agriculture, government, nonprofit institutions, and private households. Most attention is devoted to articulating behavior in the nonfarm business sector, with capital accumulation in that sector being derived from assumptions concerning private saving and government saving. Specifically, the capital stock of the economy is defined as
Kt = (1-δ) Kt-1 + It
where K refers to the capital stock, I to domestic investment and δ to the constant geometric rate of depreciation. On the assumption that net foreign investment is zero 3/, domestic investment is financed by national savings where
S = Sp + Sss + Sgf
where Sp refers to private savings, Sss to the social security surplus/ deficit, and Sgf to net saving in the federal government’s general fund accounts. In the simple framework under consideration, it is further assumed that private saving in the baseline are governed by
Sp = 0.18 GNP
where this value, which is held constant throughout the simulation period, is based on observed behavior in the postwar period.
The social security surplus Sss in the baseline is based on the social security scenario II-B projections (intermediate pessimistic) for OASDI. The baseline for the non-OASDI federal government balance Sgf follows a path consistent with the GRH until 1993—in that year the sum of Sgf and Sss balance to zero—and subsequently it is assumed to be the negative of the actuaries projections for the social security balances under II-B. This implies balance on a unified budget basis after 1993.
Labor and capital are allocated across sectors in light of observed trends or ratios over the postwar period. Output in the nonfarm business sector Y is generated by using a Cobb Douglas production function
where the exponents on K and L are based on the capital income share in the II-B projections. Compensation rates outside the nonfarm business sector are assumed to maintain their historical relationship to those within the nonfarm sector. Hence real wages in the economy are determined within the nonfarm business sector. A(t), total factor productivity, is set so that the baseline projection matches the GNP forecast under the II-B projections.
The baseline assumes that there is no link between investment and the rate of technical advance. In some of the subsequent simulations such a link is assumed. Specifically, to capture the sense of embodied technical progress, the growth rate of total factor productivity was regressed on the level of real gross investment scaled by real GNP over the sample period 1952 to 1988. The estimated equation is:
where the t-ratio is in parentheses.
Finally, the Brookings model has a fully articulated social security sector. This was not employed in the simulations in this appendix given the focus on trends in national savings rather than in the finances of the social security system per se. It should be noted that making the social security system endogenous would not necessarily change the comparative static results appreciably since changes in macroeconomic variables such as growth, for example, will tend to have a similar impact in present value terms on both social security revenues and social security taxes.
Aaron, H.J., P.B. Bosworth, and G. Burtless, 1989, Can America Afford to Grow Old? Paying for Social Security (The Brookings Institution: Washington, D.C.).
Bernheim, B.D., 1987, “The Economic Effects of Social Security: Toward a Reconciliation of Theory and Measurement,” Journal of Public Economics, Vol. 33, No. 3, pp. 273–305.
Bernheim, B.D. and L. Levin, 1988, “Social Security and Personal Saving: An Analysis of Expectations,” American Economic Review, Papers and Proceedings, Vol. 79, No. 2, pp. 97–102.
Boskin, M.J., ed., 1977, The Crisis in Social Security: Problems and Prospects, (San Francisco: Institute for Contemporary Studies).
Boskin, M.J., L.J. Kotlikoff, D.J. Puffert, and J.B. Shoven, 1981, “Social Security: A Financial Appraisal Across and Within Generations,” National Tax Journal, Vol. 40, pp. 19–34.
Diamond, P.A. and J.A. Mirrlees, 1978, “A Model of Social Insurance with Variable Retirement,” Journal of Public Economics, Vol. 10, No. 3, pp. 295–336.
Ebrill, L.P. and O. Evans, 1988, “Ricardian Equivalence and National Saving in the United States, International Monetary Fund, WP/88/96.
Ebrill, L.P., 1989, “Some Microeconomics of Fiscal Deficit Reductions: The Case of Tax Expenditures,” International Monetary Fund, WP/89/14.
Eisner, R., 1988, “Extended Accounts for National Income and Product,” Journal of Economic Literature, Vol. 26, No. 4, pp. 1611–1684.
Feldstein, M.S., 1974, “Social Security, Induced Retirement, and Aggregate Capital Accumulation,” Journal of Political Economy, Vol. 82, No. 5, pp. 905–26.
Heller, P.S., R. Hemming, and P. Kohnert, 1986, Aging and Social Expenditure in the Major Industrial Countries, IMF Occasional paper No. 47 (International Monetary Fund: Washington, D.C.).
Heller, P.S., 1989, “Aging, Savings, and Pensions in the Group of Seven Countries: 1980-2025,” International Monetary Fund, WP/89/13.
Inman, R.P., 1985, “The Funding Status of Teachers’ Pensions: An Econometric Approach,” National Bureau of Economic Research, Inc., Working Paper No. 1727.
Kotlikoff, L.J., A. Spivak, and L.H. Summers, 1982, “The Adequacy of Savings,” American Economic Review, Vol. 72, No. 5, pp. 1056–69.
Kotlikoff, J.J., 1984, “Taxation and Savings: A Neoclassical Perspective,” Journal of Economic Literature, Vol. 22, No. 4, pp. 1576–1629.
Leimer, D.R. and S.D. Lesnoy, 1982, “Social Security and Private Saving: The Time-Series Evidence,” Journal of Political Economy, Vol. 90, No. 3, pp. 606–29.
Pechman, J.A., H.J. Aaron, and M.K. Taussig, 1968, Social Security: Perspective for Reform, (Brookings Institution: Washington, D.C.).
Weaver, C.L., 1989, “Social Security’s Looming Surpluses: Controlling the Risks Posed by Advance Funding,” AEI Conference Paper, March 30.
Williamson, S.H. and W.L. Jones, 1983, “Computing the Impact of Social Security Using the Life-Cycle Consumption Function,” American Economic Review, Vol. 73, No. 5, pp. 1036–1052.
The author is grateful to Yusuke Horiguchi, Charles Adams, Lans Bovenberg, Owen Evans, and Steven Fries for their helpful advice. The author is also grateful to Mr. J. Sabelhaus of the Congressional Budget Office for his assistance in preparing scenarios.
Net saving is defined as gross saving less capital consumption allowances.
The nature of the social security trust funds and their relationship to budgetary accounting is detailed later.
Budget of the United States Government, FY 1990, Special Analysis.
Congressional Budget Office, The Economic and Budget Outlook: Fiscal Years 1990-1994, January 1989.
The HI and SMI programs are discussed in Ebrill (1989b).
Source: Budget of the United States Government, FY 1990. The data for FY 1990 and beyond are budget projections.
Interestingly in light of the current debate, the social security financing schedule adopted at the outset envisaged the accumulation of reserves so that by 1980 interest earnings would cover about one third of the outlays. This policy, which would have implied significant funding of the program, was abandoned in 1939. See Thompson (1983).
Solvency is not an absolute concept and the criteria for ensuring the solvency of the system have changed over time. The current practice in the case of OASDI is to calculate an actuarial balance over a 75-year time horizon where this balance is based on the present value of future income, outgo, and taxable payroll. In the case of HI, the projection period used to be 35 years on the grounds that the revenues and expenses of this program are more difficult to project than in the case of OASDI. However, the actuarial balance in this program has also recently been extended to 75 years, though the balance is calculated as the simple arithmetic average over that period of the difference between the annual cost and annual contribution rates as a percentage of taxable payroll. See 1988 Annual Report of the Board of Trustees of the Federal OASDI trust funds, pp. 29 et seq. and 1988 Annual Report of the Board of Trust ees of the Federal HI Trust Fund, pp. 38 et seq.
By the end of 1982, the net assets position of the OASI trust fund was negative, and money was borrowed from the DI and HI trust funds to meet the December 1982 benefit payments. See Thompson (1983), p. 1432.
Emphasis should be placed on “in effect.” At the time of the reform, the emergence of surpluses on the scale now projected was not envisaged. However, the expansion of the U.S. economy since 1983 greatly exceeded expectations and fueled a dramatic increase in the size of projected trust fund cash-flow surpluses.
Under Alternative II-B (Intermediate Pessimistic). This scenario is the alternative most commonly used in the public debate concerning the social security system. Further, Aaron, Bosworth, and Burtless (1989, page 41) conclude that on balance the assumptions underlying the scenario (Alternative II-B) are the most plausible; they note that while some of the assumptions appear overly optimistic—labor productivity is assumed to grow at 1.7 percent annually after 2010—others may be overly pessimistic—real interest rates are projected to be 2 percent after 1998. Any evaluation of the plausibility of this scenario should, however, recognize that the projections are very sensitive to the behavior of underlying variables which are themselves uncertain.
Including interest on existing stock of assets.
The combined OASDI trust funds are estimated to be exhausted in 2048 under Alternative II-B. See 1988 Annual Report of Trustees of the Federal OASDI Trust Funds.
While the hospital insurance (HI) trust fund is currently generating significant cashflow surpluses, its financial future is more precarious than that of the OASDI trust funds. Under Alternative II-B, the HI trust fund is projected to be exhausted by 2005 and the average annual deficit over the next 75 years is projected to be 2.37 percent of taxable payroll. See 1988 HI Annual Report.
In this connection, it is interesting to note that more than half of U.S. taxpayers pay more in social security taxes than in federal income taxes. See Weaver (1989).
This strand is of course related to the earlier strand since pensions can also be viewed as insurance schemes.
In other words, the relevant version of the life-cycle model assumes that there is no bequest motive. This rules out the possibility that the intergenerational altruism which underlies the Ricardian Equivalence proposition would lead to the current generation being so concerned about the welfare of future generations as to reverse through bequest behavior any governmental action that might have the effect of redistributing resources across generations. See Ebrill and Evans (1988) for further discussion of the Ricardian Equivalence Proposition. Also, Barro (1974).
The fact that social security income takes the form of an annuity in a world where the market for annuities is incomplete might have the additional impact of curbing the need for precautionary saving against the eventuality of living longer than expected.
The extensive empirical literature is reviewed in Annex I.
The issue here concerns the realtive efficiency implications of smooth versus variable tax rate paths and is discussed further in Ebrill (1989).
What this implies in operational terms is discussed later.
In the absence of the reform, there is a presumption that the taxes paid by the next generation would have to be increased to finance the current generation’s retirement consumption. This of course is not a foregone conclusion, because social security benefits could then be pared; taxes (e.g., consumption taxes) could be pitched toward the elderly; and/or there could be resort to increased borrowing.
Note that even if the surpluses are used to finance investments there could still be a potential fiscal problem due to the efficiency implications of having to use higher tax rates.
To the extent that the economy is open, this effect will be ameliorated.
The aging of the population is in this connection demographically more significant than the fact that as the population ages there will tend to be relatively fewer children.
An alternative approach, of course, would be to effect an increase in the labor supply. That raises the question of immigration policy.
The presumption is that the economy’s wealth accumulation is proceeding at a pace below that corresponding to the golden rule, a plausible presumption in the context of the U.S. economy.
It is certainly an imperfect proxy. The fact that the social security system is not actuarially balanced if the time horizon is extended beyond 75 years indicates that the 1983 reform was insufficient to ensure that the system is balanced from a pay-as-you-go perspective over the very long run. Moreover, taking a broader perspective on the intergenerational implications of baby boom demographics, there is also the prospect of large HI deficits. See Ebrill (1990).
In this respect, the impact on national savings of the social security surpluses associated with the demographic shock mirror the transitional effects discussed earlier which were associated with the introduction of the social security system.
Using the demographic and interest rate assumptions of the social security administration together with supplementary assumptions on other exogenous variables such as the private savings rate meant that total factor productivity had to behave residually to project a path of the economy which would generate social security trust fund flows of the same magnitude as those projected by the social security actuaries. The net effect is a growth rate in total factor productivity which increases from 0.9 percent a year in the late 1990s to 1.24 percent per annum in the late 2020s, a large increase. However, the scenario results presented in this appendix are in terms of deviations from the baseline, permitting one to abstract from many of the issues associated with interpreting the absolute levels of variables. The Aaron, Bosworth, and Burtless baseline incorporates judgmental changes to the social security system that would ensure the actuarial soundness of the system over the whole projection period.
Note that the labor force participation rate only captures the effect of the baby boom generation’s aging on demographics. Since the proportion of the population aged 16 and less might be expected, ceteris paribus, to decrease as the population ages, the behavior of this rate will tend to exaggerate the impact of prospective demographic changes on per capita consumption levels. The effect will nonetheless be large since the dependency ratio under Scenario II-b is projected to rise from about 70 percent in 1990 to about 83 percent in 2065 (OASDI, 1988, op. cit.).
The behavior of real interest rates operates with a lag since it is based on a ten-year moving average of returns on tangible assets. It is important to remember that all changes are expressed in terms of a percentage deviation from baseline values.
The baseline assumes the GRH path in the years up to and including 1993 and zero unified budget balance thereafter, implying that the surpluses are used to finance current government consumption. The alternative of setting the non-OASDI balance equal to zero for the whole period was not adopted since, as already pointed out, the 1983 reforms were insufficient to ensure the very long-run viability of OASDI—this alternative would, therefore, imply exceptionally rapid dissaving in the final years of the simulation.
It should also be remembered that the prospective social security surpluses are only a proxy for the amount of prefunding of the social security system which would be desirable in light of prospective demographic shifts. As already noted, OASDI is not actuarially sound over the longer term suggesting that further changes will be needed and implying that the demographically induced surge in expected social security benefits is considerably larger than the cash-flow surpluses.
Such a policy would also reduce the degree to which increased savings drive up the wages earned by labor in the next generation as a result of domestic capital deepening. While this might not change the share of total labor income it could have adverse welfare implications for labor in the next generation, ceteris paribus.
For example, the CSRS and the MRS were estimated to have net unfunded liabilities of $575 billion and $525 billion as of September 1982. See Leonard (1988).
Federal workers hired prior to January 1984 are eligible to opt for FERS but few have in fact done so.
Broadly, the defined benefits component tops up the employees contributions so that the total of their social security contributions and the topping up under FERS equals employee contributions under CSRS. Concerning the thrift (defined contributions) component, the Federal Government automatically contributes 1 percent of annual salary into a thrift account and up to a further 4 percent in matching contributions. See Kerns (1986).
Upon enactment of legislation in 1986, individuals entering the military after August 1, 1986 and retiring before the age of 62 with less than 30 years service can expect a reduction in their initial retirement annuity. This can be expected to have a significant cash flow impact on the MRS sometime in the next century as the system moves from one class of retirees to the other.
The surpluses are significant. For example, in 1990 the surpluses for military and civilian retirement have been estimated to be $16 billion and $20 billion, respectively. Congressional Budget Office, 1988 Annual Report, February.
By the early 1980s, these involved some 5,000 separate plans covering 12 million employees and holding assets of over $200 billion. See Leonard (1988), page 33.
The result is due to Arnold and is reported in Kotlikoff and Smith (1983), pp. 393-405. However, Inman reported that the unfunded liabilities for state and local teachers’ pensions alone amounted to about $400 billion in 1980. See Inman (1985).
For example, see Appendix XI, SM/88/162, Supplement 2.
The implicit assumption here is that maintaining approximate balance on the nonsocial security component of the budget would be appropriate. In reality, the appropriate nonsocial security fiscal deficit would depend on circumstances.
This includes the fear that if the social security trust funds are not placed truly off-budget—perhaps under the jurisdiction of a separate authority—the temptation to use the surpluses in other ways may prove too great. See, for example, Leonard (1989).
See Bernheim (1987). The essence of the argument is that a life-cycle consumer with positive savings is assumed to purchase a conventional asset for its survival-contingent claim. At the margin, one therefore uses the market price of future resources—effectively the price of survival-contingent resources—to value annuity claims. This amounts to simple discounting of benefit streams. Bernheim concludes that estimates based on actuarial valuation may understate the depressive effect of social security on private savings by a factor of three or more.
In this connection, Williamson and Jones emphasize that in the transition period from no social security to a social security program with extensive coverage—a period from which much of the data for the empirical tests have been drawn—parameters were changing steadily implying that the reported estimates are subject to functional form errors. Williamson and Jones (1983).
See Aaron, Bosworth, and Burtless (1989) for further elaboration of the underlying model. The specific version of the model and the data bank used in this appendix was supplied by the Congressional Budget Office (CBO). In this connection, the cooperation and assistance of Mr. F. Ribe and Mr. J. Sabelhaus, both of CBO, are gratefully acknowledged.
Annual Report of Trustees, op. cit.
The implications of making this assumption are discussed in the main body of the text.