Published March 3, 1999
Testimony provided to U.S. House of Representatives
Summary of A New Study on Winners and Losers from “Privatizing” Social Security
Testimony submitted to the Social Security Subcommittee of the Committee on Ways and Means U.S. House of Representatives
by John Mueller
Senior Vice President and Chief Economist, Lehrman Bell Mueller Cannon, Inc.
March 3, 1999
Thank you, Mr. Chairman. I’d like to describe for you a new study on “Winners and Losers from ‘Privatizing’ Social Security” I’ve just completed. The report is sponsored by the National Committee to Preserve Social Security and Medicare. This is my fourth paper on Social Security reform sponsored by the National Committee.
For the past decade I’ve been a principal of a financial-markets forecasting firm. I first became involved in the issue of Social Security reform in the 1980s, when I served as Economic Counsel to the House Republican Caucus under Jack Kemp. Like many Reagan Republican conservatives, I began with the conviction that pay-as-you-go Social Security ought to be “privatized.” But analyzing the facts and sifting the arguments turned me around. I was surprised to discover that Social Security is one of those cases, like national defense, in which the government is necessary to perform a role that the private markets alone cannot — in this case, providing the “foundation layer” of retirement security. Social Security was never intended to grow so large as to “crowd out” investment in the private capital markets, or the even more important investment in raising and educating future citizens. But the current study clearly illustrates why moving in the opposite direction — “privatizing” Social Security — would be a big mistake.
The current project was undertaken in co-operation with the Employee Benefit Research Institute (EBRI) and Policy Simulation Group. Under Dallas Salisbury, EBRI has performed a valuable service to the nation’s debate over Social Security. As part of its Social Security Project, EBRI has helped to develop a Social Security policy simulation model called “SSASIM.” SSASIM has been developed by Martin Holmer of Policy Simulation Group, at first under contract with the Social Security Administration, then most intensively in partnership with EBRI. The National Committee recently joined EBRI in its effort effort to develop SSASIM, by funding two important features that had not yet been incorporated: the model’s ability to calculate stock-market returns consistent with long-term economic projections, and to include Social Security benefits for spouses and survivors of covered workers. SSASIM is now arguably the most advanced Social Security policy simulation model in existence. Martha McSteen of the National Committee asked me to use the model to undertake the current study of winners and losers under “privatization.”
To anyone who has closely followed the debate over Social Security, it has become increasingly obvious that the quality of analysis has lagged far behind the importance of the subject. Congress is being asked to make informed judgments about proposals for sweeping changes that would affect the retirement security of American families for at least a century. Yet most studies about who would win and who would lose from “privatizing” Social Security have been flawed by at least three serious errors.
The first error is that projections for returns on stocks and bonds are inconsistent with projections for Social Security. Projections for Social Security are typically based on the “intermediate” assumptions of the Social Security Actuaries, which envision a sharp slowing of economic growth over the next 75 years — partly on the assumption that the United States will reach “zero population growth.” However, projections for the stock and bond markets are usually based on the past performance of those markets, during a period when the economy and the population were growing almost three times as fast. Putting the two together leads to absurd results. Right now Wall Street is wondering how long the stock market can maintain its record level of about 30 times annual earnings. Under the projections adopted by the 1994-96 Social Security Advisory Council, the Standard and Poor’s 500-stock Index would surpass 500 years worth of earnings in the next 75 years. By the time some of the people in this study retire, each share of common stock would be assumed to be selling for over 1,000 years’ worth of earnings. (See Graph 1.) SSASIM calculates that to be consistent with the Actuaries’ intermediate economic projections, rates of return on common stocks would have to be about 2 percentage points lower than the Advisory Council projected: 4.7% a year instead of 6.7% a year above inflation. In the study I point out a similar inconsistency in the projections for bond yields.
The second kind of error concerns the earnings of American households. Nearly all examples used in the debate over Social Security assume that the average worker — man or woman — has earnings at every age equal to something called the “Average Wage Index.” But Census data show that this is not the case. Earnings vary widely by age, sex, marital status, and education. Moreover, the average man can expect about 1 in 5 zero-earnings years, and the average woman about 1 in 3 — time spent outside the labor market due to unemployment, illness, or family responsibilities. Taking these factors into account, we find that the average man does have lifetime average earnings slightly more than the Average Wage Index — but the average women has lifetime earnings equal to about 53% of the Average Wage Index (Graph 2). This means that most studies have overstated average annual earnings of women by almost 100%. This makes a huge difference in calculating benefits under Social Security and individual accounts. Yet this erroneous method has been widely used even by the Social Security Administration.
The third kind of error concerns treatment of the “transition tax” involved in any move away from pay-as-you-go Social Security. “Pay-as-you-go” means that each generation of workers pays for the benefits of its parents. That’s why the first generations covered by Social Security received rates of return far above the long-term average. It also means that if pay-as-you-go Social Security were ended, the last couple of generations would have to pay twice: they would keep paying for their parents’ Social Security benefits, while receiving drastically reduced benefits, or no benefits, themselves; at the same time, they would have to save for their own retirement through individual accounts (Graph 3). If pay-as-you-go Social Security were phased out over one lifetime, this “transition tax” would fall partly on the Baby Boom, but especially on “Generation X” and the children of the Baby Boom.
This “transition tax” far exceeds any changes in payroll taxes and/or Social Security benefits that would be necessary to balance the existing pay-as-you-go system. The reason is simple: paying once for retirement is always cheaper than paying twice. Those who favor ending pay-as-you-go Social Security have resorted to various tricks of creative accounting to try to disguise the “transition tax.” These techniques generally involve income tax credits, or borrowing against general revenues, or some combination. But creative accounting can only try to disguise the cost of retirement benefits; it cannot change that cost by a single penny. Likewise, income-tax credits only shuffle the cost around without changing it.
SSASIM is especially suited to a study of this kind, because it permits us to sidestep the errors I’ve just described. First, the model makes it possible to project financial market returns that are consistent with projections for the economy. The model can also realistically account for transaction costs involved in investing in stocks and bonds, and in purchasing the private annuities that would have to replace Social Security. Second, SSASIM also makes it possible to simulate the impact of Social Security reform on American households with a high degree of realism, surpassing the “flat unisex earnings” assumptions of the Social Security Administration. Finally, SSASIM has the great virtue of requiring you to say exactly how the cost of benefits will be paid: no creative accounting tricks are possible. I think you’ll find that the results of a rigorous analysis of Social Security reforms are eye-opening.
The study is divided into two parts (along with appendixes examining some of the important issues raised). The first part illustrates the basic choices for Social Security reform in their effect upon the average benefits and rates of return for couples representing four different birth-years (1955, 1975, 1990 and 2025). Those born in 1955 are the middle of the Baby Boom; those born in 1975 are the smallest cohort in “Generation X”; those born in 1990 are the largest cohort in the “Echo of the Baby Boom”; and those born in 2025 represent the longer-term effects of various kinds of Social Security reform. The second part of the study looks at winners and losers among households within each of those birth-years or “cohorts” — unmarried or married; two-earner or one-earner couples; those with average, above-average or below-average earnings; and also selected African-American households.
There are essentially two possible approaches to Social Security reform: either balance the current pay-as-you-go retirement program, or else replace it with a system of individual retirement accounts. Every reform plan does one or the other, or else some combination. The current study examines the whole range of options, by comparing results under three different plans to reform the current OASI (Old Age and Survivors Insurance) program.
The first plan would completely “privatize” Social Security, by replacing it over the course of one lifetime with a system of individual accounts. Initial benefits for new retirees would be phased out evenly over 45 years, which means that the payroll tax would disappear after about 80 years. At first, the total contribution rate would be increased by 2 percentage points, though it would later fall back to equal the current payroll tax rate.
The couple born in 1955, whom I call John and Debra, would therefore live under a partially privatized system — putting about 3 percentage points into individual accounts and about 9 percentage points into Social Security. Their Social Security retirement benefits would be just under two-thirds of current law. This approximates the various plans to “carve” out part of workers’ contributions for individual accounts to supplement reduced Social Security benefits.
The couple born in 1975, Carl and Amy, would live under a mostly privatized retirement system. By the time they retired, the payroll tax would have fallen, and account contributions risen, by about 3 percentage points. Social Security retirement benefits would be reduced by about four-fifths. Plans like the Personal Security Accounts (PSAs) favored by some of the Social Security Advisory Council members, or the more recent Feldstein plan (which would combine investment in individual retirement accounts with Social Security benefits reduced from current law by 75 percent), would fall somewhere between the experience of John and Debra and the one for Carl and Amy.
The couples born in 1990 (Patrick and Hilary) and 2025 (Abraham and Dorothy) would receive benefits that depended entirely on their individual accounts — similar to plans favored by the Cato Institute. The main difference is that the Patrick and Hilary, born in 1990, would still have paid substantial payroll taxes to fund benefits to earlier retirees, but payroll tax rates would almost have disappeared by the time the Abraham and Dorothy retired nine decades from now.
The other two plans are “traditional” reforms to balance the current pay-as-you-go Social Security system.
One “traditional” plan would raise the Normal Retirement Age faster and higher than under current law: this was recommended by a majority of the 1994-96 Social Security Advisory Council. The OASI payroll tax would be maintained at the current rate of 10.6% until the trust fund fell to one year’s reserve; then the payroll tax would be adjusted as necessary to keep the system in balance.
The other “traditional” plan has two features. First, the system would be restored to a pure pay-as-you-go basis by cutting the payroll tax rate 20% (just over 2 percentage points) immediately; at the same time, benefit formulas for new retirees would be scaled back over 45 years until they reached 80% of current law. Inflation-adjusted Social Security retirement benefits would therefore rise, but not as fast as under current law. If the trust fund reserve ever fell to the minimum one-year reserve, payroll tax rates would be adjusted to keep income and outgo in balance.
Of course, any actual reform plan would be more complicated; these were chosen to illustrate the major issues.
The study examines four different sets of assumptions, from the most extreme to the most realistic. The most important conclusion from the first part of the study is this: for everyone now alive, it doesn’t greatly matter what assumptions you use about the stock and bond markets. (See Graph 4.) For the Baby Boom, Generation X and the children of the Baby Boom, not even unrealistically high stock and bond market returns can offset the “transition tax.” For everyone now alive, both average benefits and rates of return are much higher under even a scaled-back Social Security system than could be had from a partly or fully privatized retirement system.
The assumptions do make a difference for people who are not born yet. The couple born in 2025 would come out ahead under a “privatized” retirement system, if you assumed that the stock market’s price/earnings ratio will in fact surpass 1,000 years worth of earnings. But with reasonable financial asset returns, the average couple born in 2025 would be better off with even a scaled-back pay-as-you-go Social Security system than with a system of individual retirement accounts.
The second part of the study goes into much deeper detail about the impact of the various plans upon a wide variety of American households. A sample population of 32 individuals was created for each of the four birth-years, using Census data about earnings and employment by age, sex, marital status, and education level. The individuals are chosen to reflect differences in marital status (unmarried individuals or married couples); labor market behavior (two-earner couples with wives working full-time or part-time, as well as one-earner couples), and education levels (average education [some college], high-school graduates and college graduates). Individuals with below-average education and earnings are assumed to have higher-than-average mortality (that is, they don’t live as long), and vice versa.
Three of the 16 pairs of men and women in each birth-year are intended to represent selected African-American households which may differ from the general population: one unmarried man and woman each, and a two-earner married couple (all with average education and earnings for African-American men and women); as well as a couple intended to represent the most economically and socially disadvantaged African-Americans: a single mother with children and her separated partner, both of whom are high-school dropouts. All the African-American individuals are assumed to have significantly higher mortality (shorter lives) than the general population with the same education and earnings. There are no separate examples for college-educated African-Americans, on the assumption that such households have socioeconomic characteristics very similar to those for other college graduates.
The results of the second part of the study are richly detailed, and bear examining in some detail. However, in summarizing the results, I will concentrate on a few overriding themes, and focus on the most realistic set of assumptions. (See Graph 5.)
The first important finding is that the largest group of losers from “privatizing” Social Security would be women. This is true for women in all birth-years, all kinds of marital status, all kinds of labor-market behavior, and all income levels. The main reason is that Social Security was specifically designed to protect women in three ways, all of which would be eliminated by “privatization.” First, Social Security benefits are progressive, and at all education levels, women have lower average earnings than men. Second, Social Security provides the same annual benefits to men and women with equal earnings, but women live longer and so collect more benefits. Third, Social Security provides benefits for spouses of retired workers, as well as survivors benefits, which are far more advantageous to women than the private market can provide. The study shows that even unmarried women with high earnings, and women with high incomes in two-earner families, lose from privatization. However, the losses are even greater for women who work part-time, intermittently or not at all in the labor market.
The second important finding is that, to the degree Social Security is “privatized,” the current progressivity of benefits would be eliminated. The actual progressivity of Social Security is rather mild: this is because the progressivity of combined taxes and benefits is partly offset by the fact that individuals with less education and lower earnings tend to have shorter lifespans, so they collect benefits for fewer years. Individuals with more education and higher earnings tend to live longer and so collect benefits for more years. However, Social Security is still progressive, and this feature would be eliminated by “privatization.” The study finds that individuals with lower education and earnings will tend to lose more, and individuals with higher education and earnings will tend to lose less, from privatization.
The third important finding is related to the first and second; it concerns the relative treatment of unmarried individuals and married couples. (Of course, we need to bear in mind that fewer than 5% of those who reach retirement age have never been married, and perhaps 10% have never had children.) “Privatizers” have claimed that Social Security is biased against unmarried individuals. But the study shows that this claim is the result of the faulty “unisex wage” assumption. The facts are more complicated, and much more interesting. Unmarried women actually receive a higher rate of return from Social Security than most married couples. Unmarried men receive a significantly lower rate of return than unmarried women, because the men have higher average earnings and don’t live as long. If we considered husbands and wives separately, we would find that married men receive a lower rate of return than unmarried men, because unmarried men spend less time in the labor market, and so have lower average earnings and benefit more from Social Security’s progressivity. Viewed as individuals, the highest rates of return are received by married women — the lower the earnings, the higher the rate of return.
The final important finding concerns how African-American families would fare if Social Security were “privatized.” A recent study sponsored by the Heritage Foundation gained notoriety by claiming, against earlier research, that African-Americans are big losers under Social Security and would gain tremendously if Social Security were replaced by individual retirement accounts. However, the Heritage Study contained all the errors of method outlined earlier — inconsistent projections, unrealistic earnings, entirely ignoring the “transition tax,” and a few more besides. The current study squarely contradicts the Heritage findings. Of African-Americans now alive, the average household would lose about half the value of its retirement savings if Social Security were “privatized.” (See Graph 6.) For African-Americans who are not born yet, the pattern is like that for the general population. The only African-Americans born in 2025 who would not lose under “privatization” would be unmarried men without children: they would receive approximately the same rate of return from Social Security or a private retirement account. However, under realistic assumptions all other future African-American households studied would lose from “privatization” — including typical two-earner married couples, unmarried women, and those with substantially below-average education and earnings.
What does the study imply for the future of Social Security reform?
In the first place, the study is a wake-up call to policy analysts. Policymakers and the public require much better information than they are now getting — about the financial markets, about earnings of American households, and about the funding of Social Security reform. That information is available — but so far, it has not been used.
Second, the study strongly indicates that policymakers should be focusing on fixing the Social Security system, instead of getting rid of it. The “transition tax” under privatization dwarfs any possible cost of balancing Social Security. The low future returns predicted for Social Security are not due to its pay-as-you-go nature, but simply to the projections about the future economy. If the United States is in fact about to enter an Economic Ice Age, then the rates of return on everything — Social Security, stocks and bonds alike — are going to be substantially lower than in the past. It must be said that in the past few years the economy has not been behaving that way. So far, inflation, unemployment and interest rates have all been signficantly lower than the intermediate projections. The economy has so far most closely resembled the Social Security Actuaries’ “Low-Cost” economic assumptions. It would not be surprising if, in the next annual report, the Actuaries’ intermediate assumptions were modified. However, the projections in the current study are based on the 1998 intermediate projections.
But this raises an important issue for policymakers. How is it possible to set retirement policy when the future is so uncertain? As we have seen, even in the long run, the case for “privatizing” Social Security depends entirely on being right about a long string of highly specific — and in part, highly unlikely — forecasts.
Most “traditional” reform plans have the great advantage of not depending on a particular forecast. Unlike “privatization,” they are reversible. If the economy performs better than expected, the announced changes in future benefits might never have to be enacted; or else, under the “Low-Cost” assumptions, payroll tax rates might actually have to be reduced below current law (Graph 7). But if the economy does perform as poorly as projected, a “traditional” reform plan would provide a reasonable way to balance the current system. American families would be prepared long in advance, and any surprises would be positive ones.
Over the years, Social Security has in fact been adjusted several times to allow for changing circumstances. Precisely because it has been able to cope with such uncertainties, Social Security has proven itself to be a “plan for all seasons.”