Review of Radical Political Economics

http://rrp.sagepub.com/ Gambling with Retirement: Market Risk Implications for Social Security Privatization
Christian E. Weller Review of Radical Political Economics 2006 38: 334 DOI: 10.1177/0486613406290898 The online version of this article can be found at: http://rrp.sagepub.com/content/38/3/334

Published by:
http://www.sagepublications.com

On behalf of:

Union for Radical Political Economics

Additional services and information for Review of Radical Political Economics can be found at: Email Alerts: http://rrp.sagepub.com/cgi/alerts Subscriptions: http://rrp.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.com/journalsPermissions.nav Citations: http://rrp.sagepub.com/content/38/3/334.refs.html

>> Version of Record - Aug 7, 2006 What is This?

Downloaded from rrp.sagepub.com by Daniel Silva on November 1, 2013

Gambling with Retirement: Market Risk Implications for Social Security Privatization
CHRISTIAN E. WELLER
Senior Economist, Center for American Progress, 1333 H Street NW, 10th Floor, Washington, D.C. 20005; e-mail: cweller@americanprogress.org

Abstract
Under Social Security privatization, workers would be allowed to divert some of the money that currently goes to Social Security into private accounts. This would expose them to market risk, that is, the risk of a substantial drop in equity prices or of a prolonged bear market. This could result in generations of workers with less money than they thought they would have for retirement. A privatized system could require the government to intervene, for example, by expanding social programs. The primary alternative to a government bailout of the Social Security system, older workers working longer, would create enormous labor market pressures. Other alternatives, such as working longer or diversification, also encounter obstacles. Many middle-class families affected by market risk already save too little for retirement, and optimal diversification may prove too costly for many low- and moderate-income households. JEL classification: E62; G12; H55; J26 Keywords: Social Security; private accounts; stock market risk; public programs

1. Introduction Under Social Security privatization, workers would be allowed to divert a large share of the money that currently goes to Social Security into private accounts. Hence, workers may face the chance of prolonged bear markets, so-called market risk. A workers birth date could determine the size of his or her retirement account. The difference from worker to worker could vary widely. This could result in generations of workers with less money than they thought they would have for retirement and considerably less than they would have under the current Social Security system. Market risk is severe. Depending on a workers birth date, if the privatization approach proposed by President George W. Bushs Commission to Strengthen Social Security (CSSS) had been enacted at the start of the Social Security program, retirement benefits after thirty-five years would have ranged from less than 20 percent to almost 40 percent of preretirement earnings. In this scenario, a privatized Social Security system could have cost the government more than $1 trillion in 2004 dollars between 1974 and
Review of Radical Political Economics, Volume 38, No. 3, Summer 2006, 334-344 DOI: 10.1177/0486613406290898 2006 Union for Radical Political Economics

334

Weller / Gambling with Retirement

335

2004 if the government decided to help out those who accumulated less than expected for retirement. 2. Market Risks and Retirement Savings With private accounts, workers face risks that are not part of Social Security. An important risk here is that financial markets stay below their historical averages for long periods of time, so-called market risk. This risk is real. Although the real rate of return of the stock market has averaged 6.6 percent during the past 100 years, its average rate of return across thirty-five-year periods has fluctuated between 3 percent and 10 percent (Weller 2005). Privatization proposals call for part of the payroll tax to go into private accounts. The rest would continue to support Social Security benefits, albeit at lower levels. Many analyses consider option II of President Bushs CSSS as a sample plan most likely to mirror the presidents proposal (CSSS 2001; Diamond and Orszag 2002). Under this plan, workers could invest 4 percent of payroll in private accounts up to $1,000 per year and 2 percent thereafter. The contribution is voluntary, but it is subsidized,1 making it likely that everybody would invest (Diamond and Orszag 2002). For illustrative purposes, these contributions are invested in a balanced portfolio, half stocks and half corporate bonds. This also reflects the limits a privatized Social Security system would place on investment choices (CSSS 2001; Diamond and Orszag 2002). Additional contributions are credited to the account at the end of each year. Each year, workers are charged an administrative cost of 0.7 percent of assets.2 When workers retire, their savings are converted into inflation-adjusted lifetime annuities, based on the real bond rate at the time of retirement. At that time, workers are charged 5 percent of their account for converting their savings into annuities. Also, benefits for new retirees would grow only at the rate of inflation and no longer at the rate of wage increases. Thus, the living standard that Social Security benefits would afford workers would be frozen at the level of the year the privatization begins. Furthermore, Social Security benefits are reduced by the amount contributed to the private account and assessed with a real interest rate of 2 percent per year. Thus, workers would receive a loan from Social Security for the money they contribute to their private accounts, which they would have to repay on retiring (Diamond and Orszag 2002). Finally, option II includes a minimum benefit of 120 of poverty (inflation indexed) to a thirty-year minimum wage worker (Favreault et al. 2004).
1. This subsidy is rather complex. As employees contribute to a private account, a shadow liability account is established. This liability account is credited with the same amount that is contributed to the private account. Each year, the liability account increases by an interest rate that is 2 percent higher than inflation. On retirement, the savings in the private account are offset by the liability account balance. In essence, employees receive a loan from Social Security to invest in their private account, and this loan is due with interest when they retire. The interest rate on this loan is 2 percent higher than the inflation rate. As long as earnings in the private account are higher than 2 percent plus inflation, employees will actually have some additional savings. Because the interest rate on the loan is lower than the average long-term rate for bonds, employees could expect to generate even some additional savings without investing in stock. 2. Experience in U.S. financial markets and with other countries privatization efforts suggests that administrative costs will be substantially higher than the low cost estimate of 0.3 percent (Favreault et al. 2004).

336

Review of Radical Political Economics / Summer 2006

45% 40% 35% 30% Percent 25% 20% 15% 10% 5% 0% 1974 1979 1984 1989 Year real replacement rates 1994 1999

Figure 1. Real Replacement Rates after Thirty-five Years, CSSS Option II Source: Authors calculations based on Bureau of Labor Statistics (2004a), Favreault et al. (2004), and Shiller (2000).

If this plan had been in place since 1940, when Social Security began paying benefits, and if we start with a replacement rate of 42.5 percentequal to Social Securitys replacement rate of average earnings in 2004 (Trustees of the Social Security Administration 2004)the ultimate replacement rates would have ranged from 18 percent in 1978 to 39 percent in 1999 (figure 1). Interestingly, the experience of the past few years was significantly better than what should typically be expected based on historic trends. To calculate the distribution of replacement rates that retirees could expect, 1,000 hypothetical scenarios are created. In each case, earnings, interest rates, and stock market rates of return are created randomly for thirty-five years;3 average life expectancy at age sixty-five is held constant at 16.4 years; and inflation of 4 percent is assumed. Under these assumptions, the possibility of having a replacement rate of less than 20 percent has a 9 percent chance, and the possibility of having a replacement rate of less than 30 percent is slightly greater than 50 percent (figure 2). The probability of having a replacement rate of more than 40 percent of preretirement income is only 8 percent. Thus, the past few years represented extraordinary circumstances.
3. The average nominal earnings growth is 5.4 percent with a standard deviation of 3.9 percent, the average stock market rate of return is 10.8 percent with a standard deviation of 16.8 percent, and the average interest rate is 6.2 percent with a standard deviation of 3.0 percent. Rates cannot deviate more than one standard deviation from the average. In each instance, five-year averages are used.

Weller / Gambling with Retirement

337

60% 50.9% 50% 40% 30% 20% 10% 0% probability of <20% probability of <30% probability of >40% probability of >50% 8.6% 7.9% 0.3%

real replacement rates

Figure 2. Probabilities of Specific Replacement Rates, CSSS, Option II, 2038

3. Market Risks and Potential Fiscal Burden Less than expected retirement income is a result of underperforming markets affecting entire generations. This problem could become too large to ignore. Future governments may have to cover the shortfalls, for example by instituting new social programs to support the elderly. To calculate the costs of a bailout, a minimum level of retirement income needs to be established. A reasonable threshold for the privatized Social Security system would be the replacement rate for average workers that the current Social Security system pays to workers retiring at age sixty-five. Hence, the threshold would be 42.5 percent of an average workers last earnings before retirement. Based on the distributional assumptions discussed below, a threshold for each quintile in each year is calculated. Under CSSS option II, benefits would be a reduced Social Security benefit plus private account savings. The replacement rates from Social Security vary with income. To model the distribution of income on retirement, it is assumed that the income distribution in the year of retirement reflects the distribution of Social Security benefits of new retirees, which in turn is assumed to mirror the wage distribution of the population as a whole.4 Using the current Social Security formula, an unadjusted replacement rate from Social Security benefits for each quintile is calculated. This replacement rate is adjusted for the benefit cut resulting from the change from wage to price indexation by discounting it each year by the difference between wage and price growth. For each quintile of the earnings distribution, the Social Security benefit is compared to the minimum benefit
4. Average earnings are calculated for each quintile. Annual earnings are calculated as fifty times average weekly earnings using the Current Population Survey for the years from 1979 to 2003 (Center for Economic and Policy Research 2004). For earlier years, it is assumed that average earnings grew at the rate of average incomes (U.S. Census Bureau 2004).

338

Review of Radical Political Economics / Summer 2006

70 60 50 Billions $ 40 30 20 10 0 1974 1979 1984 1989 Year 1994 1999

Figure 3. Cost of Government Bailouts, CSSS Option II, Constant 2004 Dollars Note: All figures in billions of 2004 dollars. CPI is used as deflator.

under CSSS option II, and retirees are given the larger of the two. The benefits from private accounts, which are offset by the liability account, are added to the Social Security benefit. The costs of bailouts are calculated as follows. They are the difference between the threshold and the actual replacement rate times the average income times the number of people retiring in each quintile.5 The bailout pays new retirees this additional benefit for their entire retirement on an inflation-adjusted basis, just like Social Security would. Hence, the net present value of inflation-adjusted future benefits financed by the bailout is calculated for each cohort. From 1974 to 2003, the government would have had to bail out private accounts every single year (figure 3). For a number of years, the bailouts would have exceeded $50 billion (in 2004 dollars) annually. At their lowest point, the bailouts would still have cost $5.1 billion in 2004 dollars. The sum of all bailouts during the past thirty years would have totaled $1.1 trillion in 2004 dollars. It seems reasonable to assume that bailouts would occur for several years in a row. For one, it would be hard for the government to bail out one generation of retirees but not another. Moreover, even if the government would notice that one generation of retirees is likely to have less retirement income than they expected, there is little else that can be done to boost the replacement rate because the government has no direct influence over the stock market. Third, requiring higher tax rates would leave less money for other forms of saving and may not necessarily increase overall retirement wealth. The results depend on the choice of the threshold replacement rate. There are two possible objections to using thresholds for each quintile that are tied to an average replacement rate of 42.5 percent. First, one could argue that the government may want to bail out
5. The number of people retiring is one-tenth of the number of employees in the age group fifty-five to sixty-four.

Weller / Gambling with Retirement

339

only lifetime low-income earners. This would, however, create a heavy bureaucratic burden. It would also be politically difficult to decide where the income cutoff for bailout beneficiaries should be. After all, everybody paid into the system; that is, Social Security is an earned benefit and not a welfare program. Second, the initial replacement rate should not be the threshold for the system because benefits are being reduced annually. This, however, contradicts the rhetoric of those who favor privatization. Workers are promised that they could at least make up for the reduction in Social Security benefits with the savings in private accounts. Consequently, workers should expect that their replacement rate will on average at least remain constant. Furthermore, even if no specific threshold is replaced, the starting replacement rate is likely a good target because it is the benefit level that the first generation under the system was actually given. Demands by subsequent generations to see at least the same level of benefits, relative to their preretirement earnings, could likely only be muted if future retirees would build up additional savings elsewhere. This is not, however, part of the Social Security privatization debate. In addition, economic evidence indicates that workers will not compensate for the loss of guaranteed Social Security benefits with additional savings (Bernheim and Levin 1989; Wolff 1988). Last, the mere possibility of future bailouts reduces peoples incentives to save. Future governments cannot be committed to do nothing if retirees have saved too little. Thus, workers know that if the shortfalls become large enough in the aggregate, future governments will intervene. As a result, their incentive to save more outside of Social Security is reduced. 4. Market Risk and Labor Market Workers may want to work longer if they received less retirement income than they expected.6 Working longer is feasible only if employers are hiring older workers at prevailing wages. Historically, it has been the case that would-be retirees would have ended up with less than they expected when unemployment rates were already high. The statistical results show that labor market pressures should be expected to increase exactly when the labor market is already weak. Consequently, workers who accumulated less than expected savings in their retirement accounts would have to either swell the ranks of the unemployed or put downward pressure on wages. In this regard, the results suggest that lower earnings precede higher unemployment rates (table 1). That is, when workers stay in the labor force due to below average rates of return, the unemployment rate is more likely to be high than low, and wages are likely to hold steady, after their growth has already slowed due to labor market weaknesses. For purely illustrative purposes, assume that the labor supply increases but that wage rigidities maintain current wages. The unemployment rate would have risen in every year between 1974 to 2003 (figure 4).7 In extreme cases, the unemployment rate could have
6. The other adjustment is diversification. The examples discussed earlier are, however, already based on diversified portfolios, showing that even with diversification, risks remain. 7. All would-be retirees stay in the labor force until the combination of aging, additional savings, and more earnings on assets generates a replacement rate greater than the threshold. It is assumed that the basic Social Security replacement rate is not reduced after age sixty-five; that is, delayed retirement credits offset the cuts to initial benefits resulting from price instead of wage indexation.

340

Review of Radical Political Economics / Summer 2006

Table 1 Granger Causality Tests for Stock Market Returns, Unemployment, and Earnings Total Real Rate of Return (35-Year Average) Unit root test (January 1948 to December 2004) Unit root test (January 1974 to December 2004) Determining variable Total real rate of return (35-year average) 3 lags 4 lags 6 lags Unemployment rate 3 lags 4 lags 6 lags Year-to-year average real hourly earnings growth 3 lags 4 lags 6 lags Determining variable Total real rate of return (35-year average) 3 lags 4 lags 6 lags Unemployment rate 3 lags 4 lags 6 lags Year-to-year average real hourly earnings growth 3 lags 4 lags 6 lags Determining variable Total real rate of return (35-year average) Unemployment rate Year-to-year average real hourly earnings growth 0.00 0.03 2.88** (0.04) 2.62* (0.08) Year-to-Year Average Real Hourly Earnings Growth 3.36** (0.01) 3.36** (0.01)

Unemployment Rate 3.34** (0.01) 2.67* (0.09)

Test statistics for Granger causality tests (1948 to 2004)

4.47*** 3.90*** 3.37***

0.68 0.73 0.81 1.91 1.68 1.53

0.57 0.45 1.42

2.08 1.73 1.31

3.83** 2.83** 2.11**

Test statistics for Granger causality tests (1974 to 2004)

3.99*** 2.76** 2.08**

0.52 0.33 0.45 2.13* 1.41 0.84

3.94*** 3.68*** 2.49**

2.89** 2.15** 1.50

6.27*** 3.89*** 2.30** Sum of coefficients for 4 lags (1948 to 2004) 0.79*** 0.07**

0.02 0.03* (continued)

Weller / Gambling with Retirement Table 1 (continued) Total Real Rate of Return (35-Year Average)

341

Unemployment Rate

Year-to-Year Average Real Hourly Earnings Growth

Sum of coefficients for 4 lags (1974 to 2004) Total real rate of return (35-year average) Unemployment rate Year-to-year average real hourly earnings growth 0.02*** 0.01 0.03** 0.18*** 0.00 0.00

Note: Unit root test statistics are augmented Dickey-Fuller statistics, and values in parentheses are MacKinnon one-sided p values. * Significant at 10% level. ** Significant at 5% level. *** Significant at 1% level.

14% 12% 10% Percent 8% 6% 4% 2% 0% 1974 1979 1984 1989 Year actual adjusted for savings fluctuations 1994 1999

Figure 4. Actual and Adjusted Unemployment Rates, CSSS Option II Note: Adjusted unemployment rate adds the number of additional workers to the number of unemployed and the labor force. Source: Bureau of Labor Statistics (2004b) and authors calculations.

risen close to 13 percent. Even if only half of workers had decided to stay in the labor force, the unemployment rate would still have risen to more than 11 percent in this rather simple example. Despite its simplifications, this hypothetical scenario demonstrates two important points. First, financial market outcomes can create large labor market pressures,

342

Review of Radical Political Economics / Summer 2006

and, second, these pressures are likely to be much larger when unemployment is high than when it is low. 5. Additional Savings and Diversification as Potential Alternatives Workers could also save more outside of Social Security. Workers would have to increase their savings in reaction to less than expected wealth accumulation. The literature on the so-called wealth effect does suggest that workers tend to change their consumption and savings patterns in reaction to changes in their wealth. A good approximation appears to be that for each dollar wealth increases, workers increase their consumption by three cents (Poterba 2000). The wealth effect, however, only describes changes in savings behavior around a longer term trend of individual savings. Yet, the longer term trends suggest too few retirement savings for many households. For instance, Weller and Wolff (2005) concluded that among households nearing retirement, between the ages of fifty-six and sixty-four, 44.1 percent did not have enough retirement savings to replace at least 75 percent of their preretirement income in 2001. The shortfalls in retirement savings are especially pronounced among those who are disproportionately likely to rely on Social Security as a source of retirement income. In particular, minorities and single women are likely to see larger shortfalls in retirement savings. Another possibility to reduce the chance of government bailouts could be the reduction of financial market risks through diversification. The financial economics literature has, however, long recognized that workers can reduce market risks only in an incomplete manner due to high transaction costs and liquidity constraints. The relevant argument focuses on the fact that workers receive their labor income predominantly from one source. Workers can experience unanticipated business cycle shocks to labor income. Importantly, labor income is a nontradable implicit asset. To achieve optimal allocation of household portfolios, the fact that labor income is nontradable needs to be balanced with other explicit assets, such as homes, stocks, and bonds (Campbell et al. 1999; Storesletten, Telmer, and Yaron 1998; Viceira 1999; Bodie, Merton, and Samuelson 1991). Researchers have still found, however, that asset holdings tend to be lower than expected (Haliassos and Michaelides 2000; Gomes and Michaelides 2003). This is typically taken as a sign that workers face two obstacles to optimal diversification. For one, households may not be able to borrow money at the riskless interest rate to purchase the optimal mix of assets (Constantinides, Donaldson, and Mehra 1998; Bertaut and Haliassos 1997), and, second, there may be prohibitively high costs (Vissing-Jorgensen 2002; Yaron and Zhang 2000; Abel 1998). The latter may be especially important for low-income households, which often do not hold any equities in their portfolios (Vissing-Jorgensen 2002; Abel 2000; Haliassos and Michaelides 2000; Campbell et al. 1999). A corollary of this argument implies that greater volatility in labor income requires more frequent portfolio adjustments, raising costs and reducing stock holdings (Vissing-Jorgensen 2002). Without efficient diversification, short-term income fluctuations will mean that households will be less likely to accumulate savings when it is financially most opportune, that is, when asset prices are comparatively low. Because households are typically unable to achieve optimal portfolio allocation to reduce financial and labor market risks at the same time, the chances of government bailouts will remain.

Weller / Gambling with Retirement

343

6. Conclusion Under Social Security privatization, workers face the risk that financial markets can underperform during their working careers. This could happen for entire cohorts of retirees. Large market fluctuations are unavoidable for individual accounts. These fluctuations can create substantial shortfalls in savings for entire generations of retirees. One way to address this shortfall would be for workers to work longer, which would create enormous labor market pressures, with much higher unemployment rates or lower wage growth as the result. To avoid such large labor market pressures, the federal government would likely have to bail out individual accounts instead. Thus, the government would likely have to implement financial support programs to avoid spikes in old-age poverty. These bailouts could come in a number of forms, such as expanded social programs or direct transfers to account holders. Thus, privatization would essentially amount to a system of insured gambling, in which the government pays the bill if the market underperforms. References
Abel, A. 1998. The aggregate effects of including equities in the Social Security Trust Fund. Wharton School of the University of Pennsylvania, Philadelphia. . 2000. The effects of investing Social Security funds in the stock market when fixed costs prevent some households from holding stocks. NBER Working Paper no. 7739. Cambridge, MA: National Bureau of Economic Research. Bernheim, D. B., and L. Levin. 1989. Social Security and personal saving: An analysis of expectations. American Economic Review 79 (2): 97102. Bertaut, C. C., and M. Haliassos. 1997. Precautionary portfolio behavior from a life-cycle perspective. Journal of Economic Dynamics and Control 21: 151142. Bodie, Z., R. Merton, and W. Samuelson. 1991. Labor supply flexibility and portfolio choice in a life cycle model. Journal of Economic Dynamics and Control 16: 42749. Bureau of Labor Statistics. 2004a. Employment, hours, and earnings from the current employment statistics survey (national). Washington, DC: Bureau of Labor Statistics. . 2004b. Labor force statistics from the current population survey. Washington, DC: Bureau of Labor Statistics. Campbell, J. Y., J. F. Cocco, F. J. Gomes, and P. J. Maenhout. 1999. Investing retirement wealth? A life-cycle model. Harvard Institute of Economic Research Discussion Paper no. 1896. Cambridge, MA: Harvard University. Center for Economic and Policy Research. 2004. CPS ORG uniform data files version 0.9.3. Washington, DC: Center for Economic and Policy Research. Commission to Strengthen Social Security (CSSS). 2001. Strengthening Social Security and creating personal wealth for Americans. Washington, DC: CSSS. Constantinides, G. M., J. B. Donaldson, and R. Mehra. 1998. Junior cant borrow: A new perspective on the equity premium puzzle. NBER Working Paper no. 6617. Cambridge, MA: National Bureau of Economic Research. Diamond, P., and P. Orszag. 2002. Reducing benefits and subsidizing private accounts: An analysis of the plans proposed by the presidents Commission to Strengthen Social Security. Washington, DC: Center on Budget and Policy Priorities. Favreault, M., J. Goldwyn, K. Smith, L. Thompson, C. Uccello, and S. Zedlewski. 2004. Reform model two of the presidents Commission to Strengthen Social Security: Distributional outcomes under different

344

Review of Radical Political Economics / Summer 2006

economic and behavioral assumptions. Urban Institute Report, September 30. Washington, DC: Urban Institute. Gomes, F., and A. Michaelides. 2003. Portfolio choice with internal habit formation: A life-cycle model with uninsurable labor income risk. London Business School. Haliassos, M., and A. Michaelides. 2000. Portfolio choice and liquidity constraints. University of Cyprus, Nicosia. Poterba, J. 2000. Stock market wealth and consumption. Journal of Economic Perspectives 14 (2): 99119. Shiller, R. 2000. Irrational exuberance. Princeton, NJ: Princeton University Press. Storesletten, K., C. Telmer, and A. Yaron. 1998. Asset pricing with idiosyncratic risk and overlapping generations. GSIA Working Paper no. 1998-E226. Pittsburgh, PA: Carnegie Mellon University. Trustees of the Social Security Administration. 2004. Annual report of the board of trustees, Federal Old-Age and Survivors Insurance and Disability trust funds, 2004. Washington, DC: Government Printing Office. U.S. Census Bureau. 2004. Historical income data. Washington, DC: U.S. Census Bureau. Viceira, L. 1999. Optimal portfolio choice for long-horizon investors with nontradable labor income. NBER Working Paper no. 7409. Cambridge, MA: National Bureau of Economic Research. Vissing-Jorgensen, A. 2002. Towards an explanation of household portfolio choice heterogeneity: Nonfinancial income and participation cost structures. NBER Working Paper no. 8884. Cambridge, MA: National Bureau of Economic Research. Weller, C. 2005. Social Security privatization: The retirement savings gamble. CAP Economic Policy Report. Washington, DC: Center for American Progress. Weller, C., and E. Wolff. 2005. Retirement income: The crucial role of Social Security. Washington, DC: Economic Policy Institute. Wolff, E. 1988. Social Security, pensions and the life cycle accumulation of wealth: Some empirical tests. Annales Dconomie et de Statistique (9): 199226. Yaron, A., and H. Zhang. 2000. Fixed costs and asset market participation. Revista de Analisis Economico 15 (1): 89109.

Christian E. Weller is a senior economist at the Center for American Progress in Washington, DC. His work specializes on retirement income security, macroeconomics, labor markets, and international finance. Prior to coming to American Progress, he worked as research economist for the Economic Policy Institute in Washington, DC, where he remains a research associate.