Risk Aspects of Investment-Based Social Security Reform

The University of Chicago Press

Copyright © 2001 National Bureau of Economic Research
All right reserved.

ISBN: 978-0-226-09255-3

Chapter One

Asset Allocation and Risk Allocation

Can Social Security Improve Its Future Solvency Problem by Investing in Private Securities?

Thomas E. MaCurdy and John B. Shoven

Policy makers widely accept that social security faces a long-run solvency problem. The social security trustees publish a seventy-five-year forecast of the OASDI system's finances every year. Since the 1983 amendments, the forecast has gotten consistently more ominous. Figure 1.1 shows the intermediate-assumptions forecast for the trust fund balances in 1983, 1985, and 1998.

The trustees' 1998 report has the OASDI trust fund completely depleted in 2032 (when the youngest baby boomer, born in 1964, will be sixty-eight) rather than 2063 (when the same individual will be ninety-nine years old), the forecast that immediately followed the 1983 social security amendments. The report has as the intermediate or "best-guess" forecast that the system's income (not counting net interest income on the special issue bonds held in the trust fund) will fall short of its costs beginning in 2013. In 1999, the surplus was projected to be 1.52 percent of covered payroll, but this turns negative in 2013, and, by the end of the trustees' forecasting period (2075), the annual shortfall under the intermediate set of assumptions is forecast to be 6.43 percent of covered payroll. That means that, if we operated the system strictly on a pay-as-you-go basis, the payroll tax rate would have to be increased by 6.43 percentage points to achieve balance in 2075. The 2075 OASDI deficit (with the intermediate-assumptions forecast) amounts to 4.26 percent of GDP. One way to look at the financial solvency problem of social security is that the intermediate forecast is for fourteen years of modest and declining surpluses, followed by ever-growing deficits as far as the eye can see. Against that we have a relatively small trust fund (currently $760 billion) that generates roughly $40 billion of interest income but that is now forecast to be completely depleted by 2032, as shown in figure 1.1. Naturally, the financial situation for OASDI is much worse under the trustees' high-cost set of assumptions. With them, the date of OASDI trust fund exhaustion is 2022. The date when payroll-tax receipts first fall short of benefit payments is 2006, and the eventual shortfall of income to costs in 2075 is 16.04 percent of taxable payroll.

The social security trustees summarize the seventy-five year outlook for OASDI by computing the long-run actuarial balance. In 1998, the seventy-five year actuarial balance was in deficit by 2.19 percent of covered payroll with the trustees' intermediate assumptions. What that means is that, had the payroll tax been immediately increased in 1998 by 2.19 percentage points and the increase maintained for the next seventy-five years, the life of the trust fund would have been extended to the seventy-five-year horizon under the intermediate set of economic and demographic assumptions. Even under this hypothetical scenario, the payroll-tax proceeds would be less than benefit payments under the currently legislated benefit structure beginning in roughly 2020. Further, the seventy-five-year actuarial balance would last for exactly one year; with every passing year, one fewer surplus year would be in the seventy-five-year window, and one more deficit year would be included. The immediate payroll-tax increase needed to fix the solvency problem of the current OASDI system permanently would be significantly greater than 2.19 percent. Steven Goss (1999), of the Office of the Actuary of the Social Security Administration, has estimated that the permanent or open-ended actuarial deficit is 4.7 percent of covered payroll under the intermediate set of assumptions.

All this illustrates the dimensions of the problem of returning the existing OASDI system to long-run solvency. Operating within the existing structure of social security, there are only two obvious paths: raise taxes or cut benefits, or both. If we take Goss's 4.7 percent of covered payroll figure for the perpetual actuarial deficit, permanently fixing the finances of the system (under the intermediate set of economic and demographic assumptions) will require immediate tax increases or benefit cuts totaling 38 percent. If the implementation of these steps is delayed, as it almost certainly will be, then the benefit cuts and tax increases will need to be larger. Neither tax increases nor benefit cuts of this magnitude are economically or politically attractive. Naturally, people are looking for a more palatable way out of the social security solvency problem.

One natural place to look is the investment returns earned on the trust fund. Currently, the trust fund is exclusively invested in special nonmarketable U.S. government bonds. When these bonds are issued, their interest rate is set at the average interest rate of marketable Treasury bonds with a maturity of four years or more. The bonds have one special feature, which offsets their nonmarketability: they are redeemable at par at any time. This feature is not generally available on publicly held bonds and notes, with the exception of U.S. government savings bonds. The special issue bonds are certainly safe, with no price risk and with the principal and interest fully backed by the U.S. government, but Treasury interest rates are somewhat less than what is offered on AAA corporate bonds and trail the average total return earned on common stocks by a wide margin.

The question that we address in this paper is whether a significant fraction of the whole solvency problem could reliably be solved by having the Social Security Administration invest the OASDI trust fund in higher-yielding private securities. The analysis in MaCurdy and Shoven (1992) suggests that such a strategy might yield an improvement in social security's finances. Of course, there is a more fundamental question regarding whether society as a whole would benefit from this new asset allocation. Another question that we address is how risky such a change in the trust fund asset allocation would be and who would bear that risk. Related to this question, we discuss the feasibility of social security delivering a true defined-benefit pension program to its participants.

1.1 The Proposals

We focus on two of the reform proposals that rely on investing the central trust fund in higher-yielding private securities in order to maintain the general benefit structure of social security. However, our analysis is applicable to any plan that attempts to make progress on the solvency of the system by simply reallocating portfolios toward higher-yielding securities, including the plan outlined by President Clinton in his 1999 State of the Union Address. The two plans of this type that we describe in detail are the "maintain-benefits" (MB) proposal of the 1994-96 Advisory Council (often referred to as the Bob Ball plan) and the plan offered recently by Henry Aaron and Robert Reischauer (1998). Among the measures that both plans advocate, the asset reallocation is credited with the largest effect on reducing or eliminating the seventy-five-year actuarial deficit.

The details of the two plans are shown in table 1.1. There are many similarities between them. The numbers in the tables come directly from Aaron and Reischauer (1998, table 6-1) and from volume 1 of the Report of the 1994-96 Advisory Council on Social Security (SSA 1997, table 1, app. II). The figures are based on estimates of the Office of the Actuary of the Social Security Administration, assuming the intermediate alternative II economic and demographic conditions. The numbers for the same action may differ slightly for the two plans since the estimates for the Aaron-Reischauer plan were made roughly two years later than those for the MB plan. Further, the plans differ slightly in how they implement each measure. Even with that said, both proposals have the same goal-to deal with the seventy-five-year actuarial deficit within the context of the current structure of benefits. Both assume that the measurement of the CPI will be changed in ways that lower official inflation and cause the program's costs to grow more slowly. Both advocate covering all newly hired state and local workers, increasing the number of years included in the AIME (average indexed monthly earnings) formula from thirty-five to thirty-eight, and taxing all social security benefits received over and above the employees' own actual contributions. Both proposals characterize their changes in the personal income taxation of social security benefits as putting them on the same basis as private pension income. However, their plans involve significantly higher taxes than the Schieber-Shoven PSA-2000 plan, which also claims to put social security and private pension benefits on the same tax footing (Schieber and Shoven 1999). The PSA-2000 plan taxes half of social security benefits on the argument that half of social security contributions were made with before-tax money (the employer contribution) and half with after-tax dollars (the employee contribution). Essentially, Schieber and Shoven's plan would have half the contributions treated like a Roth IRA (after-tax contributions with tax-free withdrawals) and half like a regular IRA (before-tax contributions with taxable withdrawals). In contrast, the Aaron-Reischauer and Ball plans tax 85 percent or more of social security benefits.

The MB plan contains two features that Aaron and Reischauer did not choose to copy. One is a 1.6 percentage point payroll-tax increase in 2045, and the other is a redirection of money from Medicare to OASDI. The latter feature of the MB plan seems ill advised as the long-run finances of Medicare are in worse shape than are those of OASDI. Similarly, the Aaron-Reischauer plan contains several features not in the MB plan. These include accelerating and extending the increase in the age of normal retirement, advancing the age of early retirement, and readjusting the spousal and surviving spouse benefits. The Office of the Actuary of Social Security estimates that, both plans if enacted immediately and completely, would more than eliminate the seventy-five-year actuarial deficit of the system.

The common feature of the two plans shown in table 1.1 is that the biggest contributor to eliminating the long-run solvency problem is investing part of the OASDI trust fund in private securities. The two plans differ in the details of how they would do that. In fact, the MB plan in the 1994-96 Advisory Council report suggests that further study be given to the idea before implementation. Nonetheless, when the Advisory Council scored the plan to see whether it achieved the goal of eliminating the seventy-five-year actuarial deficit, it included the provision that the trust fund would begin investing in stocks in the year 2000 and that the proportion of trust fund assets in stocks would gradually rise until it reached 40 percent in 2015. The assumed real rate of return on the stock portion of the portfolio in the MB plan is 7.0 percent. The Aaron-Reischauer plan is to have the trust fund balances that exceed 150 percent of one year's benefits gradually invested in common stocks and corporate bonds. Since the OASDI trust fund balances currently exceed the 150 percent of annual payout criterion, the switch to private securities would begin immediately under the Aaron-Reischauer program. Remarkably, Aaron and Reischauer estimate that 55 percent of the whole actuarial deficit of 2.19 percent of payroll would be eliminated by this asset reallocation alone. They also estimate that the combination of the asset reallocation and the effect of the changes that the Bureau of Labor Statistics is making in the construction of the CPI eliminates over 75 percent of social security's projected long-term deficit. Any 75 percent cure that is this painless deserves careful scrutiny.

1.2 The Net Transaction

The net transaction involved in having the central OASDI trust fund invest a portion of its assets in common stocks is an asset swap. Social security or the federal government sells additional government bonds to the public and uses the proceeds to acquire common stock from the public. In the social security context, the system is selling one set of assets (U.S. government bonds) in order to acquire another set of assets (a diversified portfolio of common stocks) of equal value. Of course, this transaction can be examined without reference to social security's financial problems at all. The real issue is whether the government can improve the welfare of taxpayers (or social security participants) by issuing and selling bonds and using the proceeds to buy common stocks.

There are at least two reasons to be skeptical about the advantages of the net transaction being discussed. First, the total capital stock and wealth in the economy (at least to a first approximation) are unaffected by the asset swap. Therefore, the level of GDP and national income is unchanged. If social security or the government can systematically improve its financial position by making this exchange, the private sector, which presumably is on the other side of the transaction, is systematically losing. It is hard to imagine that the politicians and bureaucrats running social security are systematically getting the better of the pension fund managers and institutional investors who are buying the bonds from the government in exchange for common stocks. It is possible, of course, that the exchange could be played out on international capital markets so that Americans as a whole could conceivably end up as net winners (or losers) in the transaction. Second, there is the matter of risk. While stocks have a much higher expected return than government bonds, they also involve much higher risks (particularly given the fact that the current special issue bonds have the feature that they can be redeemed at par at any time). The two plans discussed above are very vague as to how much additional risk the system would be assuming and how that risk would be shared among taxpayers and social security participants. As we discuss in section 1.5 below, there is a real question about whether it is feasible to maintain a universal coverage defined-benefit pension plan funded with risky securities.

Having the government exchange bonds for stocks in the hope of relieving the solvency problems of social security is a form of financial engineering. The recent history of such schemes is not promising. Savings and loans sell short-term liabilities (certificates of deposit and demand deposits) and acquire higher-yielding long-term mortgages. The savings-and-loan crisis of the 1980s was brought about when short-term interest rates peaked in 1980 in the mid-teens when the savings-and-loan mortgage portfolio yielded 6-8 percentage points less. The usual yield relation did not hold, and the resulting bankruptcies and bailouts cost taxpayers massive sums of money. The Orange County debacle resulted from a similar failed attempt to exploit the shape of the yield curve. The recent massive losses of hedge funds resulted from asset swaps that failed to perform as expected. Before the U.S. government engages in exchanging bonds for stocks, careful analysis is clearly warranted.

1.3 Analyzing the Stocks-for-Bonds Swap

To gauge the riskiness of the net transaction of selling government bonds and buying corporate stocks, we use market returns from 1954 to 1997 and simulate what would have happened had the government completed the transaction. We separately look at what would have happened with the government selling ten-year and twenty-year bonds. We assume that market returns would have been unaffected by the government transaction. This almost certainly favors the exchange strategy that in reality might very well drive up interest rates and depress equity returns. The counterfactual simulations have the government selling bonds in the past, buying the S&P500 stock portfolio with the proceeds, and then paying all the bond payments (interest and principal) from the resulting stock portfolio. The strategy is deemed successful if all the payments on the bonds can be made with the stock funds with money left over. The strategy is deemed to have been a failure if the stock portfolio is unable to generate sufficient cash to make the required bond payments. In the discussion presented below, we initially explore how the strategy would have worked with the actual time series of returns generated by the stock market (and the actual historical interest rates on government bonds). The interest rates are those published in the Economic Report of the President (1998) and computed by the Board of Governors of the Federal Reserve System. The stock market returns are the "large company stocks" (i.e., S&P500) total return series in Ibbotson Associates (1998). The problem with this historical approach is that there are not many independent ten- or twenty-year periods in our data set (1954-97). In fact, there are only four completely independent ten-year runs of data and two of twenty years. We could add more data by examining pre-1954 information, and we do so in some of the analyses in this section. There is a serious question about whether adding data before 1954 helps or hurts when assessing the viability of a proposed strategy for the twenty-first century. The problem is that the additional data are likely to be drawn from a different regime and, therefore, may do more harm than good to the analysis. After examining the actual performance of the stock market in the period 1954-97, we report on extensive bootstrap simulations of what could have happened using the same annual return data, but now with the order of the returns randomly scrambled according to a bootstrap statistical approach.


Excerpted from Risk Aspects of Investment-Based Social Security Reform Copyright © 2001 by National Bureau of Economic Research. Excerpted by permission.
All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.