Comparing Proposals for Social Security Reform

Policy Reports | Social Security

No. 227
Wednesday, September 01, 1999
by Liqun Liu and Andrew J. Rettenmaier


  1. Report of the 1994-1996 Advisory Council on Social Security, Advisory Council on Social Security, Washington, D.C., 1997.
  2. 1999 Annual Report of the Board of Trustees of the Federal Old-Age and Survivors Insurance and Disability Insurance Trust Funds, Washington, D.C., March 1999.
  3. Laurence Kotlikoff, "Privatizing Social Security," NCPA Policy Report No. 217, July 1998.
  4. Kevin M. Murphy and Finis Welch, "Real Wage Growth and OASDI Trust Fund Solvency," working paper, 1996.
  5. How good a deal one can get from one's Social Security contributions can be measured either by the (internal) rate of return or by the money's worth ratio, defined as the ratio of discounted benefits to discounted taxes paid by both employees and employers, with the discount rate being the real rate of return. If the internal rate of return is larger than the real rate of return, or equivalently, the money's worth ratio is larger than one, then Social Security participation is a better deal than private savings. Otherwise, it is a worse deal.
  6. However, this does not mean that national saving will rise by simply replacing the unfunded system with mandatory individual savings accounts. The debt management policy can be such that the additional private saving in the form of the mandatory savings accounts is fully absorbed by an increase in the government debt with no increase in capital stock. This can occur if the newly added explicit debt after privatization has exactly the same schedule as the old unfunded liabilities. National saving would increase only if fiscal policy were changed to make the debt shrink over time relative to the old liability.
  7. For a thorough analysis of the deadweight loss due to reduced saving and work efforts, as well as other negative incentive effects caused by the current pay-as-you-go system, see Martin Feldstein, "The Missing Piece in Policy Analysis: Social Security Reform," American Economic Review, May 1996, Vol. 86, pp. 1-4.
  8. Martin Feldstein and Andrew Samwick, "The Economics of Prefunding Social Security and Medicare Benefits," National Bureau of Economic Research, NBER Working Paper No. 6055, June 1997. The authors estimate that the reduced deadweight loss associated with the lower cost of financing is 3.75 percent of wages at the end of their simulation period. This amount exceeds the 2.23 percent of wages needed to finance retirement benefits.
  9. For example, in Kotlikoff, the switching of tax bases is achieved by eliminating the payroll tax and financing the transition with a consumption tax. See Laurence J. Kotlikoff, "Privatizing Social Security at Home and Abroad," American Economic Review, Vol. 86, No. 2, pp. 368-72; and Kotlikoff, "Privatization of Social Security: How It Works and Why It Matters," in James Poterba, ed., Tax Policy and the Economy, Vol. 10, Cambridge, Mass.: MIT Press, 1996).
  10. Many economists argue that the efficiency gains from changing tax bases can be realized without privatization, by simply changing the financing tax (say from the payroll tax to a consumption tax) within the pay-as-you-go system.
  11. Phil Gramm, "Investment-Based Social Security," NCPA Policy Report No. 220, December 1998.
  12. Sens. Gramm and Domenici have proposed a 3 percent carve-out. However, under the balanced budget restriction imposed in this paper and under assumptions made about capital market returns, only a 2.5 percent carve-out can be allowed initially. See the analysis below.
  13. While recognizing that even a slight change to Social Security can have a dramatic effect throughout the rest of the economy, we analyze the proposals using an actuarial model similar to that of the Social Security Administration. In this approach, some economic effects, such as labor supply responses, are ignored. We have two reasons for taking this approach. First, such models are well accepted among policy makers and are ultimately used to evaluate any reform proposal. Second, the proposals themselves were developed within the context of such a model. A description of the model and further explanation of some of the assumptions used can be found in Appendix A.
  14. James M. Poterba estimates that the pretax rate on nonfinancial corporate sector capital averaged 8.5 percent between 1959 and 1996. See James M. Poterba, "The Rate of Return to Corporate Capital and Factor Shares: New Estimates Using Revised National Income and Product Accounts and Capital Stock Data," NBER Working Paper 6263, National Bureau of Economic Research, Cambridge, Mass., April 1999.
  15. We ignore the Medicare program on the theory that it is a separate program with similar problems that also require a long-term solution.
  16. Specifically, we assume that the federal government can capture the additional revenues collected by state and local governments by reducing match grants and other payments to the states by a like amount.
  17. The simulation results for each proposal are presented in Appendix B. All of the proposals are treated in the same way. The Gramm-Domenici plan identifies the degree to which recaptured corporate taxes impact the system's future financial status. The Archer-Shaw and bipartisan plans do not, but for consistency across evaluations, recaptured corporate taxes are included in each table. It is appropriate to count these revenues, as long as the contributions to PRAs reflect additions to the capital stock; that is, if they are either implicitly funded by a budget surplus, mandatory contributions, or a transitional tax. While the proposals differ in how they would fund contributions to the PRAs, we subtract the contributions from the payroll tax. Doing so recognizes the contributions within the summary tables.
  18. The revenues and expenditures depicted in the graph are based on our simulation model. See Appendix A for a description.
  19. Although the Gramm-Domenici plan does not include a provision for repeal of the earnings penalty, Sen. Gramm has introduced other legislation that does provide for repeal.
  20. Under a 1996 change in the law, the earnings limit for 65-to-69-year-olds will increase by a set amount each year until it reaches $30,000 in 2002. The limit for 62-to-64-year-olds is automatically adjusted each year for increases in the national average wage index.
  21. Stephen C. Goss, "Long-Range OASDI Financial Effects of the Social Security Guarantee Plan," memorandum to Harry C. Ballantyne, Chief Actuary, Social Security Administration, April 29, 1999.
  22. See Leora Friedberg, "The Labor Supply Effects of the Social Security Earnings Test," NBER Working Paper No. W7200, National Bureau of Economic Research, Cambridge, Mass., June 1999.
  23. The earnings in each year between now and retirement for each education category are based on our forecasts of participation in the work force, annual hours of work and hourly wages. The Social Security benefits are calculated using the projected rules in place in 2048. The replacement rates are calculated assuming a constant 5.3 percent real rate of return.
  24. The simulation results in the Gramm-Domenici proposal are based on an adaptation of a model developed by Feldstein and Samwick in "The Economics of Prefunding Social Security and Medicare Benefits." The Feldstein-Samwick work has been widely cited and is one of the leading Social Security privatization plans. As a result of our independent estimates of earnings and benefits, our model will produce results that differ in magnitude from those produced by Gramm-Domenici's adaptation of the Feldstein and Samwick model. However, our estimates in large part confirm those suggested by Gramm-Domenici.
  25. The term PRA, for personal retirement account, is used here; Gramm termed the accounts Social Security Individual Investment Accounts or SI accounts.
  26. The Gramm-Domenici plan is detailed in Phil Gramm, "Investment-Based Social Security: Saving Social Security for Our Parents, for Our Children, Forever," National Center for Policy Analysis, NCPA Policy Report No. 220, December 1998.
  27. Our discussion of the transition cost differs from Gramm and Domenici's in the way a budget surplus is counted. We define the transition cost as the difference between net revenues and expenses and thus the surplus must be identified net of the Social Security system, whereas with Gramm and Domenici's treatment the full budget surplus is counted against the transition cost because they present aggregate costs. The difference is a matter of presentation only.
  28. The Gramm-Domenici proposal calculates corporate tax recapture thus: First, annual aggregate PRA balances net of annuity payments are determined. Gramm and Domenici assume that 80 percent of the aggregate PRA balances are net additions to the capital stock, given that the PRAs may displace private saving. They further assume that 15 percent of the investments will not be subject to corporate taxes because they are held in foreign securities of sub-chapter S corporations. Based on these assumptions, they adjust the corporate tax rate downward from 35 percent to 23.8 percent (.35 x .8 x .85). The recaptured corporate taxes are then the aggregate PRA balances times the 8.5 percent rate of return times the adjusted corporate tax rate.
  29. Although the Archer-Shaw proposal would reduce federal tax collections by an amount equal to 2 percent of taxable payroll, we identify the effect in the simulation as a reduction in the payroll tax. Doing so results in payroll taxes net of PRA contributions equal to 8.6 percent in all years. Identifying the effect in this way allows us to more conveniently compare this proposal with the other proposals.
  30. The Social Security Administration used a 5.35 percent rate of return in its evaluation of the Archer-Shaw proposal. There are three primary differences between the Social Security Administration's approach and ours. First, we limit the analysis to the OASI program; they include the DI program. Second, we count the recaptured corporate taxes as revenue; they do not. Third, they count interest on the Trust Fund balance as a revenue source, but we do not, for reasons explained in the text. Thus our findings differ from those of the Social Security Administration, although both show that the Archer-Shaw plan is workable.
  31. In our simulation, corporate income taxes arising from the additional national savings are used to help finance the transition. The Archer-Shaw proposal does not mention recaptured corporate taxes nor does the evaluation of the Archer-Shaw proposal that was performed by the Social Security Administration. As the transition progresses there is less need for this alternate funding source. Recaptured corporate taxes are identified here for consistency with the evaluations of the other proposals. The taxes can be used to reduce the payroll tax or returned to the Treasury as general revenues. The consequences of both treatments are presented in Appendix B, Table B-IIIa and Table B-IIIb.
  32. The difference between our results and those of the SSA are a consequence of using our own revenue forecasts and benefits forecasts and our focus on the retirement and survivors portion of the program. [See Appendix B, Table B-IV for a comparison of our results and the Social Security Administration's evaluation.] Our forecasts of OASI revenues and expenditures produce estimates that are more conservative than the SSA actuaries' intermediate assumptions. For example, in 2010 we estimate that revenues will exceed expenditures by an amount equal to 0.69 percent of taxable income, while the actuaries' intermediate and high cost assumptions are equal to 1.10 and 0.28 percent, respectively. In 2075, we estimate that expenditures will exceed revenues by an amount equal to 7.49 percent of taxable income. The SSA actuaries' intermediate estimate in that year is 5.79 percent and the high cost estimate is 13.57 percent.
  33. The Bipartisan Social Security Reform Plan is sponsored by members of both houses including Sens. Judd Gregg of New Hampshire, John Breaux of Louisiana, Bob Kerrey of Nebraska, Charles Grassley of Iowa, Charles Robb of Virginia, Fred Thompson of Tennessee and Craig Thomas of Wyoming. Congressmen Jim Kolbe of Arizona and Charles Stenholm of Texas have introduced a similar proposal in the House.
  34. Another provision allows the lower-earning partner in a two-earner couple to drop the five lowest earnings years from his or her benefit calculation. Thus for each couple the benefit calculation for the higher earner will be based on the highest 40 earnings years, and the benefit calculation for the lower earner will continue to be based on the 35 highest earnings years.
  35. Recall that our simulation model uses historical data from the 1964 to 1998 March Supplements to the Current Population Survey as the basis for predicted future wages. To allow for a changing distribution of wages, we predicted earnings within age by sex by education categories. Thus the historical spreading in the distribution of earnings persists in our predictions of future earnings. We estimate that the present value of the system's revenues when the ratio of taxable to total wages is set at .86 is 9 percent higher than the present value of the system's revenues under current law.
  36. The Social Security Administration has evaluated the effects of reform measures that are quite similar to those suggested in the Bipartisan Proposal. The separate effects of each reform are presented in "Distributional Effect of Social Security Options by Gender and Income," memorandum from Keith Fontenot, Assistant Deputy Commissioner for Policy, to Jane Ross, Deputy Commissioner for Policy, May 18, 1999. In that study, the effect of each reform is treated separately. Since we look at the cumulative effects of the bipartisan reforms, the closest comparison between our results and theirs is the first reform we consider: the effect of maintaining the tax base at 86 percent of total wages. They find that maintaining a tax base of 90 percent of total wages reduces the long-term deficit by 29 percent. We find that maintaining a tax base of 86 percent of total wages reduces the unfunded liability by 34 percent. There are two primary reasons for the difference between the results. First, the unfunded liability we calculated is the present value of the difference between future expenditures and tax revenues. Their revenue stream includes Trust Fund interest revenues as well. Second, our result maybe higher than the SSA's because of the degree to which the wage distribution is expected to widen in the future. For more detail on how we forecast future wages, see Rettenmaier and Saving, The Economics of Medicare Reform, forthcoming.
  37. Recall that the tax rate is based on a higher level of taxable earnings under the Bipartisan Proposal than under the other plans.
  38. A 10 percent bonus is used because a 20 percent bonus requires a higher contribution rate, and the aggregate contributions of the young would grow faster than is justified under the 2.5 percent budget constraint.
  39. As mentioned earlier, the redistribution could occur at retirement, as is shown here, or the contributions of low-income workers could be enhanced in such a way that their savings produce an annuity that replaces Social Security benefits. The point is that some tax revenues will be required regardless of how the redistribution is realized.
  40. Part of the reason for the difference is the lower bonus under the NCPA-PERC plan.

Read Article as PDF