Report | Retirement

The Feldstein-Samwick ‘Two Percent Solution’—New Social Security Proposal Is Both Inadequate and Risky

 Download PDF

Share this page:

November 24, 1998 | Issue Brief #128

The Feldstein-Samwick ‘Two Percent Solution’

New Social Security proposal is both inadequate and risky

by Edith Rasell

Among the proposals for strengthening Social Security is a recent plan by economists Martin Feldstein and Andrew Samwick that, they argue, would eliminate the long-term shortfall in the trust fund with no changes in benefits or taxes. Their proposal, however, is based on flawed assumptions that would translate into inadequate trust fund reserves or increased costs to the federal government.

Under the Feldstein-Samwick plan, workers would have the option of contributing 2% of their earnings (below the cap on income subject to the Social Security payroll tax, currently $68,400) to a personal retirement account (PRA). Their contributions would be offset, dollar for dollar, with a refundable personal income tax credit. Consequently, the worker would have no reduction in income as a result of his or her PRA contribution. The tax credits, which in the first 20 years of the plan would total about 0.8% of gross domestic product each year, would be paid out of the projected federal budget surplus. If the surplus were inadequate, a tax increase, cuts in other spending, or deficit spending would be necessary.

When a worker reached age 65, his PRA would be converted into an annuity. For each dollar of retirement income benefits received from the annuity, the retiree’s Social Security benefits would be reduced by 75 cents. Feldstein and Samwick calculate that the reduction in outlays from the Social Security Trust Fund would extend its life beyond the 75-year planning horizon.

Flaws in the Feldstein-Samwick plan
The Feldstein-Samwick plan has three fundamental problems:

1. It overestimates the rate of return that would be earned on the private accounts. A more realistic return would require new, additional payroll contributions as high as 3.5% of earnings, not 2%. The cost of the tax credit offsetting these contributions would require tax increases, cuts in other federal spending, or deficit spending.

2. Administering a system of PRAs is much more expensive than investing in the stock market directly through the Social Security Trust Fund. Annual administrative fees of 1% to 2% of the value of each account would significantly reduce the money available to be paid out in benefits, and converting the PRA into an annuity would cut the value another 15% to 20%.

3. It increases the possibility of fraud and abuse, thus requiring more costly government regulation.

These serious shortcomings are discussed below.

Overstated rate of return
Feldstein and Samwick estimate that a PRA (with an assumed investment mix of 60% stocks and 40% bonds) will earn an inflation-adjusted average annual return of 5.5%, “the average return on [such a portfolio] during the postwar period through 1994″ (p. 5). But because of the slowdown in economic growth projected for the future – which is, after all, a major cause of the Social Security shortfall – returns on the portfolio described by Feldstein and Samwick will average about 3.8%, not 5.5%.

This lower return occurs because earnings on stocks will fall as U.S. economic growth slows, a trend projected by the Social Security trustees. (Specifically, the trustees project real economic growth to average 1.5% annually between 1998 and 2075, compared to 3.0% per year over the past 75 years.) Since the growth in stock prices mirrors the growth in corporate earnings, and because corporate earnings growth tracks economic growth over the long term, stock price growth in the future will probably average about 1.5% annually, 1.5 percentage points less than its historic rate. Adding in dividends paid to stock owners, currently about 2.5% of average stock prices, total returns on stocks over the next 75 years will average approximately 4.0% annually (the sum of 1.5% and 2.5%). (For a more detailed discussion of future stock returns, see Baker 1997.) Returns on corporate bonds will average about 3.5% annually – this includes the typical 0.75% premium over the rate on treasury securities, estimated by the Social Security trustees to average 2.8% annually over the next 75 years. Thus, real returns on a portfolio divided 60-40 among stocks and bonds will average 3.8% per year.

The difference between a return of 5.5% and 3.8% is large over a worker’s lifetime. At 5.5%, an investment of $1,000 will grow to $6,514 after 35 years, compared to just $3,689 (43% less) at a 3.8% annual rate. To reach the same level of savings one would accumulate from an initial investment of 5.5% held over 35 years, a worker with a portfolio earning 3.8% would need to contribute 3.5% of earnings, rather than 2%. Consequently, the cost to the federal government of the Feldstein-Samwick tax credits would rise as well, necessitating either increased taxes, reductions in other federal spending, or expanded budget deficits.

Administrative costs
The PRAs proposed by Feldstein and Samwick would be expensive to administer. Fees would take a 1-2% bite out of an account’s value each year (this is the range charged on private accounts in Britain’s privatized system; see Diamond 1997). The conversion to an annuity would eat up another 15-20% of the balance in the account (Mitchell, Poterba, and Warshawsky 1998), reducing the annuity payouts by 15% to 20% and reducing the savings to the Social Security Trust Fund by the same amount.

Fraud and abuse
PRAs and other types of self-directed, individual accounts that permit account owners to place their money in a wide range of investment vehicles raise concerns about fraud and abuse. Writing in the Washington Post (November 16, 1998), Arthur Levitt, chairman of the Securities and Exchange Commission, noted that “[a] system of self-directed individual accounts would require an unprecedented level of broad-scale policing of the equity markets. Otherwise, fraud and sales practice abuses could be perpetrated against society’s most vulnerable investors.” Levitt goes on to cite the experience in England, where reform of the retirement system and creation of individual accounts led to “billions of dollars in losses for investors” through fraud and other abuses.

A better way to use the surplus
Returns on stock are higher, as a rule, than on bonds, but so are the risks. If the nation decides to take the risk in order to capture these higher returns for Social Security, then part of the Social Security Trust Fund could simply be invested in the stock market, a step that would also avoid the high administrative costs associated with individual accounts. While there are legitimate concerns about placing public money in this type of investment, investing through the trust fund rather than through individual accounts would generate more money for Social Security beneficiaries and be less prone to fraud and abuse.

An alternative proposal could work in the following way. Keeping the Feldstein-Samwick suggestion, assume an additional 2% of taxable payroll, funded by the federal budget surplus, were earmarked for Social Security. But rather than have individuals deposit this money in PRAs, the funds could go directly into the Social Security Trust Fund. A portion of this money could then be invested in the stock market, as some policy makers and members of the Gramlich Commission have recommended, thereby boosting the rate of return over that received on the fund’s safer, but lower yield, U.S. treasury securities.

In this scenario, fees and expenses would be much lower than in the Feldstein-Samwick plan. New administrative costs would be negligible because there would be no individual accounts to manage, and there would be no need to create annuities and, thus, no conversion
fees. All the accumulated returns could be paid out in benefits.

Since stocks are inherently more risky than treasury bonds, risk would also be increased compared to the current system. But in the trust fund, as opposed to a PRA, risk would be spread widely across time and individuals.

Investing a portion of the trust fund in stocks creates the risk that political considerations could affect investment decisions. To prevent this possibility, the trust fund could be restricted to purchases in index funds only. Because it might also be tempting to use public policy to ensure a steady, uphill course for the stock market, adequate fire walls and other safeguards would have to be constructed to prevent interference in the market.

Personal retirement accounts that allow workers to manage their Social Security savings may seem to be a politically attractive, easy way to solve the problem of a projected shortfall in the Social Security Trust Fund. However, decisions about the future of Social Security must not be based on overestimates of future returns nor on a less-than-complete understanding of the added costs and risks associated with these accounts. If the country decides that it is appropriate that workers’ core retirement money – Social Security – be invested in the stock market, then the most efficient way to accomplish this goal would be through the trust fund.


Baker, Dean. 1997. Saving Social Security With Stocks: The Promises Don’t Add Up. New York: Twentieth Century Fund and the Economic Policy Institute.

Diamond, Peter. 1997. “Macroeconomic Aspects of Social Security Reform.” Brookings Papers on Economic Activity, Vol. 2, pp. 1-89.

Feldstein, Martin, and Andrew Samwick. 1998. “Two Percent Personal
Retirement Accounts: Their Potential Effects on Social Security Tax Rates and National Saving.” National Bureau of Economic Research, Working Paper No. 6540. Cambridge, Mass.: NBER.

Mitchell, Olivia S., J.M. Poterba, and M.J. Warshawsky. 1998. “New Evidence on the Money’s Worth of Individual Annuities.” National Bureau of Economic Research, Working Paper No. 6002. Cambridge, Mass.: NBER.

See related work on Retirement

See more work by Edith Rasell