Social Security: Long-Term Financing and Reform

views updated

SOCIAL SECURITY: LONG-TERM FINANCING AND REFORM

The Social Security Act and its transformation

The provisions of the Social Security Act (signed into law by President Franklin D. Roosevelt on 14 August 1935) were tailored to model a private insurance system and avoid any hint of socialism. Some degree of individual equity would be maintained, with those who paid in more receiving greater benefits upon retirement. OASI (Old Age and Survivors Insurance) would not be means-tested, reflecting American preferences against welfare. These provisions helped to make the Social Security Act the most successful and enduring of the New Deal programs— indeed, many analysts claim that Social Security is the most popular federal government program ever adopted. In truth, however, Social Security has never closely approximated a private insurance plan, and changes to the Social Security Act over the years have moved it ever farther from that model. Thus, while Americans still tend to think of Social Security as a system of ‘‘retirement insurance,’’ in fact almost 28 percent of the beneficiaries of OASI (which excludes disability benefits covered under the full program of Old Age, Survivors and Disability Insurance, OASDI) in 1997 were spouses or survivors of covered workers. Though the discussion that follows will focus on the aspects of Social Security that relate most closely to issues surrounding aging and retirement, it must be kept in mind that Social Security is a much bigger program with broader coverage, and any reforms must take account of the large percent of beneficiaries without normal work histories.

U.S. workers began to pay payroll taxes for Social Security in 1937, and the first benefits were paid in 1940. Over the years tax rates were increased, the percent of the workforce covered grew, and benefits were expanded. A substantial change was made in 1972, when automatic cost-of-living adjustments were added to benefits. The next few years experienced high inflation and slower real economic growth—together these raised program benefit payments and lowered tax receipts, generating the first crisis for Social Security when it was feared that receipts might fall below expenditures. This led to the first significant cutbacks in the program’s history, with inflation adjustments delayed, tax rates increased, and some benefit cuts for civil servants. By the early 1980s it was realized that these changes had not been sufficient. A commission was appointed (with future Federal Reserve Board Chairman Alan Greenspan at its head) to study the long-term financial situation of Social Security. One significant issue that had arisen was the aging of the U.S. population. This was compounded by the baby boom bulge created in the early postwar period when fertility rates (number of children per woman) rose and remained high until the baby boom bust of the 1970s, when fertility rates fell. Combined with rising longevity, this ensures that the baby boomers will create a relatively large number of elderly Social Security beneficiaries and relatively fewer workers to support them between 2015 and 2035. While the problems created by the baby boom receive most of the press, falling fertility rates and rising life expectancy are common experiences in all the major developed nations, leading to the social challenges that result from an aging population. Indeed, on current projections, Social Security will experience its greatest financial problems after all the baby boomers have died.

The Greenspan Commission published findings that resulted in the most significant changes made to Social Security since the addition of Medicare in 1965. These included higher tax rates, imposition of taxes on Social Security benefits, and phased increases in the normal retirement age. Most important, the 1983 revisions changed Social Security from pay-as-you-go to advance funding. In a pay-as-you-go (or paygo) system, current-year revenues are balanced against current-year expenditures. However, as the Greenspan Commission recognized, the aging of America created a special financing problem. During the late twentieth and early twenty-first centuries, OASDI was projected to run large annual surpluses, but sometime during the second decade of the twenty-first century, the program would begin to run annual deficits. Indeed, it was feared that the deficits would eventually become so large that it might be politically infeasible to raise taxes or cut benefits by the amount required to return the program to balance. For this reason, the paygo system was abandoned in favor of advance funding. In an advance funded system, near-term annual surpluses are accumulated in a trust fund that purchases special U.S. Treasury securities to earn interest. When program spending rises above tax revenues, interest earnings supplement revenues to maintain balance. At some point in the future, taxes plus interest earnings will fall below annual benefit payments; then the trust fund can sell its Treasury securities to make up the difference. In this way surpluses over the first three or four decades following the Greenspan Commission’s changes could be used to offset projected annual deficits during the final decades, ensuring long-term financial solvency.

Long-term prospects

The changes made in 1983, including the move to an advance funded system, were believed at the time to have resolved the challenges created for Social Security by the aging of the population. Over subsequent years the board of trustees developed a rigorous method of financial accounting for the program that reported detailed projections for the next ten years (its short-range forecast) and a projection for the next seventy-five years (its long-range forecast) to capture the effects of demographic shifts, as well as alternative assumptions regarding economic factors, such as economic growth. By the end of the 1990s, they had settled on use of three alternative scenarios for the long-term forecasts: high-cost (pessimistic), intermediate-cost, and low-cost (optimistic). The long-term financial status of the program is summarized in a calculation of the actuarial balance for the seventy-five-year period. This is the difference between the summarized income ratio (the ratio of the present value of payroll taxes to the present value of taxable payroll) and the summarized cost rate (the ratio of the present value of expenditures to the present value of the taxable payroll) over the valuation period. This is essentially similar to any comparison that discounts future revenues and costs to determine long-term net revenues. When the summarized income rate equals or exceeds the summarized cost rate, the program is said to be in actuarial balance. If the difference is negative, the program is in actuarial imbalance with an actuarial gap, measured as a percent of taxable payroll. For example, if discounted revenues fall short of discounted benefit payments by an amount equal to 1 percent of taxable payroll, this is said to represent an actuarial gap of 1 percent. An immediate increase of payroll taxes by a total of 1 percentage point (half on employers and half on employees) would close the gap.

OASDI began to show a large actuarial imbalance by the late 1990s, equal to more than 2 percent of taxable payroll. As an alternative to tax increases, the gap could be closed by cutting benefits or increasing the rate of return earned on trust fund assets. In fact, many proposals for reforming Social Security include provisions that would simultaneously pursue all three alternatives: tax increases, benefit cuts, and higher earnings. In addition, some proposals would move Social Security operations closer to a private insurance fund model—or, indeed, replace Social Security with a privately operated (and even purely voluntary) retirement system. Before turning to such proposals, the underlying causes of the gap will be examined.

According to the Social Security Administration’s 1999 projections, on intermediate-cost assumptions the program achieved actuarial balance only for the first twenty-five years; over the fifty-year period the actuarial gap was -1.26; and it reached -2.07 for the entire seventy-five-year long-range forecast. However, on low-cost assumptions, the program maintained actuarial balance for the whole period, and on high-cost assumptions the program had an actuarial gap even for the first twenty-five-year period. This shows how critical the assumptions used in the forecasts are. Also, any crises are relatively far in the future; indeed, even on high-cost assumptions the actuarial imbalance is quite small for the first quarter of the twenty-first century. According to the 1999 projections, the Social Security trust funds would reach $2.3 trillion by 2008 on intermediate-cost assumptions, peak at more than $4.4 trillion in 2020, then decline to zero by 2035. Using low-cost assumptions, the trust fund would continue to grow over the entire seventy-five-year period, reaching more than $45 trillion by 2075. On the other hand, the trust fund would reach only $2.6 trillion in 2015 according to high-cost assumptions, and would then be depleted quickly, falling to zero by 2025. In other words, if Social Security is analyzed as if it were a private pension plan, it apparently will experience a crisis in 2025 or 2035, using high-cost or intermediate-cost assumptions. Most analysts focus on the intermediate-cost projections, according to which Social Security revenues would be sufficient to cover only three-fourths of expected expenditures after the mid-2030s.

The main demographic and economic assumptions that underlie the projections are fertility rates, immigration, labor force participation rates, longevity, growth of real wages, and taxable base. Together, fertility rates, immigration rates, longevity, and labor force participation rates determine the size of the pools of workers and retirees. The number of Social Security beneficiaries supported by workers will rise sharply in the early twenty-first century. For example, the number of OASDI beneficiaries per one hundred covered workers was thirty-one in 1975, but this will rise steadily between 2010 and 2075, when it will reach fifty-six, on intermediate-cost projections. To put it another way, while the United States had just over 3.3 workers per beneficiary in 2000, it may have fewer than 1.8 by 2075. Thus, the burden required of future workers to provide for OASDI beneficiaries could increase by almost a factor of two. On the other hand, workers in 2075 are projected to support fewer young people. If the population under age twenty is added to the population age sixty-five and over to obtain a dependent population (most of whom would not be expected to be working), the dependency ratio (the ratio of dependents to workers) actually peaked at 0.95 in 1965, fell to 0.71 by 1995, and will rise only slightly to 0.83 by 2075. In other words, the parents of the baby boomers supported more dependents in the mid-1960s than any generation is likely to support in the future.

In any case, as the number of beneficiaries rises relative to the number of workers paying Social Security taxes, the actuarial balance is negatively impacted. This results in part from a falling fertility rate, which reduces the size of the younger population from which workers can be drawn. The fertility rate (children born per woman) stood at just over 2 at the end of the 1990s and was projected to fall to 1.9 under the intermediate assumptions. If the fertility rate were to rise back to 3.7 (where it stood in 1957 during the baby boom), over 90 percent of the actuarial gap would be eliminated. On the other hand, a falling fertility rate can be offset by rising net immigration (since immigrants can add to the worker pool, paying payroll taxes) and by rising labor force participation rates (the number working or seeking work per one hundred population). In the late 1990s net (legal and illegal) immigration reached about 960,000 per year. In their projections, however, the trustees assumed that annual net immigration would fall to 900,000 and remain there throughout the seventy-five-year period. Each additional 100,000 net immigrants above that level would reduce the actuarial gap by about 0.07 percent of taxable payroll. The trustees also project that labor force participation rates for men will fall (from 75.5 percent in 1997 to 74 percent by 2075). In contrast, since World War II, labor force participation rates for women have risen sharply (reaching 60 percent in 1997). The trustees project that this will nearly level off (reaching only 60.6 percent by 2075). If male labor force participation rates did not fall, and if female rates continued to rise, some of the actuarial gap would be eliminated. Finally, the trustees project that death rates will fall by 34 percent over the seventy-five-year period, and each ten percentage point decrease in the death rate increases the long-range actuarial gap by about 0.34 percent of taxable payroll.

While most of the debate over the Social Security program’s solvency focuses on these unfavorable demographic trends, they were mostly known to the Greenspan Commission and the 1983 adjustments should have taken care of them. In fact, the reason for the looming Social Security crisis lies not in the demographics but in the increasingly pessimistic economic assumptions adopted by the trustees in their reports in the 1980s and 1990s. The main economic assumptions that lead to the financing gap are low growth of real wages and a falling taxable base. In 1999 the trustees projected that real wages would grow at only 0.9 percent per year. Real wage growth, in turn, is related to productivity gains. The trustees assume that productivity will grow at just 1.3 percent annually over the long-range period—well below long-term U.S. averages. If real wages were to grow at 2 percent per year, more than half of the actuarial gap would be eliminated. Finally, the trustees have projected that the taxable base will fall from 41 percent of GDP in 1999 to only 35 percent in 2075. This is for two reasons. First, Social Security taxes wages and certain kinds of self-employment income. Other types of income, such as interest income, are exempt. If these rise as a share of national income, the percent of income subject to the Social Security tax will fall. Second, payroll taxes are levied on only a portion of one’s wage income—determined by the contribution and benefit base. In 1999 OASDI taxes were applied only to the first $72,600 of employment income (the contribution and benefit base for that year; this base is increased each year with rising nominal average wages). The trustees have assumed that the taxable base will fall both because a smaller portion of income will be received in the form of wages and because a higher percent of wages will accrue to those with earnings above the contribution and benefit base. Therefore, by 2075, a little over a third of national income will be taxed to support OASDI beneficiaries.

Some analysts have questioned the usefulness of calculating actuarial balance over a seventy-five-year period. Projections of demographic trends and, more important, economic variables over such long periods is inherently difficult, and relatively small changes in assumptions can change the projections significantly. Some analysts have argued that projections are based on rather pessimistic economic assumptions. For example, according to intermediate-cost projections, real GDP and labor productivity will grow at only 1.3 percent per year. In fact, labor productivity has grown at a rate of approximately 2 percent per year since 1870, and at 2.7 percent per year between World War II and 1973, while real GDP grew at an annual rate of 3.7 percent from 1870 to 1973. Even if productivity and GDP grew at only two-thirds of long-term trends, Social Security’s financial problems would be eliminated. A counterargument is that U.S. economic performance since 1973 has generally been worse than long-term averages, and it is more prudent to project weak performance into the future than to presume that high growth might return. On the other hand, it has been noted that these pessimistic assumptions were incongruously adopted in the trustee forecasts during the ‘‘Goldilocks’’ 1990s expansion, when U.S. economic performance did return toward historical averages. Furthermore, the assumptions adopted may not be internally consistent. For example, if labor force growth rates are as low as the trustees have assumed, one might expect that real wage growth should be higher than the assumed 0.9 percent as excess demand for labor pushes up its relative return, and labor productivity growth should be higher as firms substitute capital for scarce labor.

More important, it is not clear that a national, public retirement system ought to operate as if it were a private pension fund, building reserves today that earn interest and can be depleted in future years. In, say, 2035 when the trust fund needs to sell securities to the Treasury, the Treasury will have to raise taxes, cut other spending, or sell securities to cover retirement of debt held by Social Security. Surprisingly, this is exactly what the government would have to do even if Social Security had no trust fund at all! Suppose Social Security were operated as a paygo system, with each year’s receipts equal to spending. When revenues began to fall below benefit expenditures, the government would have to increase taxes, reduce other spending, or issue securities to cover the difference. Some, including Milton Friedman, have concluded that the trust fund is nothing more than an ‘‘accounting gimmick,’’ because when Social Security begins to run deficits, existence of a trust fund cannot really provide for financing of its spending. Further, unlike a private firm, the U.S. government’s revenues are not market-determined. If necessary, the government can raise tax rates or can deficit-spend to ensure that it meets its Social Security obligations—things that private firms cannot do. Hence, while government cannot really build up a trust fund, it also does not need to do so.

What really matters is whether the economy will be able to produce a sufficient quantity of real goods and services to provide for both workers and dependents in, say, the year 2035. If it cannot, then regardless of Social Security’s finances, the real living standards of Americans in 2035 will have to be lower than they are today. Those who take this approach argue that any reforms to Social Security made today should focus on increasing the economy’s capacity to produce real goods and services, rather than on ensuring positive actuarial balances. For example, policies that might encourage public and private infrastructure investment will ease the future burden of providing for growing numbers of retirees. In a sense, this would be a real reform rather than a financial reform, although it is possible that financial reforms might encourage greater investment. Indeed, some advocates of privatization argue this will encourage more investment by entrepreneurs. However, even with the trustees’ rather pessimistic economic and demographic assumptions, real living standards are projected to rise substantially for both workers and retirees throughout the seventy-five-year, long-range period. Accordingly, Social Security does not face a real crisis even though it may face a financial crisis. Still, this may not be sufficiently comforting to future workers, because although they will enjoy a growing real economic pie, the share of the pie going to retirees will grow. In fact, the share of GDP going to OASDI will grow from about 5 percent in 2000 to 7 percent for the period between 2030 and 2075. On the one hand, this is a significant increase, but on the other hand, similar shifts have occurred in the past without generating an economic crisis.

Reform proposals

Of the most important reform proposals, perhaps the most extreme would be to abolish Social Security altogether, leaving it up to individuals to decide how to provide for their retirement. (However, it should be recalled that Social Security is not simply an old age security program; privatization could leave widows, dependents, and disabled persons to their own devices.) Some reformers recognize that individuals generally underestimate the future costs of retirement, so some sort of mandatory minimum contribution levels should be maintained even if the program is privatized. Another variation would maintain basic coverage in a mandatory, government-run program, but would allow individual control over supplemental investments in privately run pension funds. Finally, some proposals would retain most features of the current system but would direct the trustees to invest a specified portion of the trust funds in private equities.

Some reformers advocate privatization simply as a matter of principle—for example, Milton Friedman has long argued that there is no justification for mandatory participation in a public system. Others emphasize that, as currently designed, Social Security has a strong redistributional element—both within and across generations. They typically use a money’s worth estimate to calculate a return on one’s contributions. Those with high earnings, and thus high contributions, receive low returns when they eventually collect benefits, while those with low earnings receive higher returns. Many beneficiaries never actually contribute; thus, contributions of others are redistributed to them. Returns also vary greatly by generation—early participants in Social Security received very good returns on their contributions, but returns for later generations are much lower. Thus, some reformers emphasize that reform should be geared toward ensuring better money’s worth outcomes for contributors, both within and across generations. It should be noted, however, that while some attention was paid to such equity concerns in the original Social Security Act, money’s worth was not a high priority in the beginning and most amendments since then have moved the program ever farther from this consideration.

Nevertheless, most reformers in recent years have pushed privatization to resolve the long-range financial imbalance of Social Security, with money’s worth calculations playing a smaller role. It is argued that under current arrangements, the trustees can hold only government bonds with relatively low interest rates. Allowing investments in the stock market and other private assets could increase returns on the trust fund. Such arguments were given a tremendous boost by the spectacular performance of U.S. stock markets from the mid-1980s through 2000. Many proponents argue that equity markets should earn real returns of about 7.5 percent per year—as they have averaged since the 1920s— probably more than double the real return on government debt. Over the long-range, seventy-five-year period, these higher returns could resolve Social Security’s financial problems.

However, opponents have raised several objections. First, there will be a costly transition period as the system becomes privatized, during which current workers must finance the cost of the existing Social Security system (paying taxes to finance the benefits going to current retirees), plus the costs of building up their own retirement funds. Thus, a fairly large, immediate tax increase will be required, and must remain in place for several decades until all those covered under the old system have died. Second, it is probably inconsistent to argue that real GDP growth will slow (to 1.3 percent per year, little over one-third its long-term average) without affecting growth of equity prices. Critics have shown that this implies either that the share of distribution of national income going to profits must grow to implausibly high levels (indeed, Dean Baker has calculated that wages must be negative by 2070), or that price-earnings ratios (already at all-time highs in 2000) must rise to truly astronomical levels (on Baker’s calculation, to 485-to-1 by 2070). On the other hand, privatizers believe that the influx of money into stocks would generate more investment, and thus higher economic growth. Third, privatization might place unacceptable levels of risk on workers. Even if the stock market were to grow at an average rate of over 7 percent per year, there could be relatively long periods of below-normal growth. In the past, equity prices have been fairly flat for decades at a time. Unlucky workers whose retirement happened to come after such a period could face deprivation during retirement. Furthermore, if workers are given individual control over their retirement accounts, they might do poorly even if the markets as a whole are doing well.

Fourth, numerous small retirement accounts can be very costly to administer and supervise. Current costs of administering Social Security are exceedingly small—well under 1 percent of revenues. Privatizers often point to the Chilean example as evidence that a private system can produce high returns for contributors; however, overhead costs in Chile are above 10 percent. While private fund managers were initially supportive of the move to privatize, they have become less enthusiastic as they have come to realize the logistics of managing many small accounts for lower income workers. Fifth, some critics have argued that because women typically have lower incomes, spend more time out of the labor force, and more often work part-time, most privatization reforms would adversely affect benefits paid to women. Others have noted that because African Americans and Hispanics typically have lower income and lower life expectancies, privatization reforms as well as raising the normal retirement age would have a disproportionately negative impact on those groups. Finally, critics note that privatization schemes do not, and probably cannot, offer the same kinds of coverage currently offered by Social Security. For example, private pension plans do not offer inflation indexing, as Social Security does. As discussed, Social Security also offers coverage for many individuals without significant work histories. If the program is privatized, a new social safety net would have to be created to cover individuals who could not purchase private insurance.

With the large turnaround of the U.S. federal government budget in the late 1990s (from chronic deficits to record surpluses), President Bill Clinton and many others proposed that budget surpluses could be set aside to resolve Social Security’s financial problems. Essentially, President Clinton would have increased the size of the trust fund by an amount equal to just under two-thirds of annual budget surpluses. The larger trust fund would then earn more interest and would have more Treasury securities to sell when program revenues fall below expenditures. However, as noted above, when the trust fund sells securities, the Treasury will have to cut other spending, raise taxes, or sell securities to the general public to cover the payments made to the trust fund. Furthermore, like most financial fixes, this reform will not necessarily increase future productive capacity. Indeed, it is not necessary for the federal government to run surpluses for it to credit the trust fund with more securities—the Treasury can add securities worth any amount to the trust fund at any time (in principle, it can add an amount equal to the entire Social Security shortfall today, thereby resolving any financial difficulties). Alternatively, the Treasury could simply agree to pay a higher interest rate on trust fund assets—paying whatever interest rate would eliminate the actuarial gap. Though somewhat ludicrous, these alternatives emphasize that accumulating a trust fund of Treasury securities really cannot resolve future annual deficits in the Social Security program.

Furthermore, as emphasized above, what really matters is the economy’s capacity to produce real goods and services in the future. Hence, if the amount that can be produced will not be sufficient to provide the level of consumption desired by all generations in the future, it will be necessary to either to boost production or to ration consumption. Extending the normal retirement age (which is essentially a benefit cut) will keep workers in the labor force longer, and will reduce the number of years they must be supported during retirement. Increasing taxes on future workers will leave them with lower purchasing power, ensuring that more of the nation’s output can go to retirees. Cutting future OASDI benefits will do the opposite—allocating more output toward workers and others with incomes that are not dependent on Social Security. Fortunately, given increases in worker productivity that reasonably can be expected to occur, plus increases in production facilities that are likely to take place as a result of public and private investment, it appears quite likely that future workers and future retirees will enjoy higher living standards than do their counterparts today—in spite of the aging of America.

L. Randall Wray

See also Baby Boomers; Population Aging; Social Security, and the U.S. Federal Budget; Social Security, History and Operations.

BIBLIOGRAPHY

Advisory Council on Social Security. Report of the 1994–1996 Advisory Council. Vol. I: Findings and Recommendations. Washington, D.C.: Government Printing Office, 1997a.

Advisory Council on Social Security. Report of the 1994–1996 Advisory Council. Vol. II: Report of the Technical Panel on Trends and Issues in Retirement Savings, Technical Panel on Assumptions and Methods and Presentations to the Council. Washington, D.C.: Government Printing Office, 1997b.

Baker, D. ‘‘Saving Social Security with Stocks, Introduction, Policy in Perspective.’’ The Century Foundation. www.tcf.org

Baker, D. ‘‘Saving Social Security in Three Steps.’’ Briefing Paper. Washington, D.C.: Economic Policy Institute, 1998.

Ball, Robert M.; Fierst, Edith U.; Johnson, Gloria T.; Jones, Thomas W.; Kourpias, George; and Shea, Gerald M. ‘‘Social Security for the 21st Century: A Strategy to Maintain Benefits and Strengthen America’s Family Protection Plan.’’ In Advisory Council on Social Security, Report of the 1994–1996 Advisory Council. Vol. 1: Findings and Recommendations. Washington, D.C.: Government Printing Office, 1997.

Clinton, W. J. State of the Union Address, January 19, 1999. Transcript. CNN Interactive, January 20. www.cnn.com

Congressional Budget Office. Social Security Privatization: Experiences Abroad. Washington, D.C.: Government Printing Office, 1999.

Friedman, M. ‘‘Social Security Chimeras.’’ New York Times, January 11, 1999.

Langer, D. ‘‘Social Security Finances are in Fine Shape.’’ Remarks at the Congressional Conference to Fight Social Security Privatization. Washington, D.C.: January 21, 1999.

Langer, D. ‘‘Social Security vs. Stock.’’ Op-ed. Christian Science Monitor, June 19, 2000. Page 19.

Papadimitriou, D. B., and Wray, L. R. ‘‘Does Social Security Need Saving?’’ Public Policy Brief no. 55. Annandale-on-Hudson, N.Y.: The Jerome Levy Economics Institute, 1999a.

Papadimitriou, D. B., and Wray, L. R. ‘‘How Can We Provide for the Baby Boomers in Their Old Age?’’ Policy Note 1999/5. Annandale-on-Hudson, N.Y.: The Levy Economics Institute, 1999b.

Papadimitriou, D. B., and Wray, L. R. ‘‘More Pain, No Gain: Breaux Plan Slashes Social Security Benefits Unnecessarily.’’ Policy Note 1999/8. Annandale-on-Hudson, N.Y.: The Levy Economics Institute, 1999c.

Skidmore, M. J. Social Security and its Enemies: The Case for America’s Most Efficient Insurance Program. Boulder, Colo.: Westview Press, 1999.

Social Security Administration. Annual Report of the Board of Trustees of the Federal Old-Age and Survivors Insurance and Disability Insurance Trust Funds. Washington, D.C.: Government Printing Office, 1998.

Stein, H. ‘‘How to Solve Almost Everything.’’ New York Times, February 3, 1999.

Wray, L. R. ‘‘Can the Social Security Trust Fund Contribute to Savings?’’ Journal of Post Keynesian Economics 13, no. 2 (1990–1991).

Wray, L. R. ‘‘The Emperor Has No Clothes: President Clinton’s Proposed Social Security Reform.’’ Policy Note 1999/2. Annandale-on-Hudson, N.Y.: The Levy Economics Institute, 1999a.