ȱ
ȱ¢ȱDZȱȱ ȱ
ȱ ǯȱ ȱ
ȱȱȱȱ¢ȱ
ȱȱ
ȱȱȱ¢ȱ
¢ȱřŗǰȱŘŖŖŝȱ
ȱȱȱ
ŝȬśŝŖŖȱ
ǯǯȱ
řŗŚşŞȱ
ȱȱȱ
Pr
epared for Members and Committees of Congress
ȱ¢ȱDZȱȱ ȱ
ȱ
¢ȱ
Social Security reform has been a perennial item on the congressional agenda. While Social
Security originated as a Depression-era program aimed at alleviating the economic circumstances
of the elderly, social insurance also corrects market failures in the annuity market (adverse
selection), prevents free-riders (requires workers to provide for their retirement), spreads risk, and
may correct for failure to optimize by shortsighted individuals. The system imposes costs on
society as well, through distortions in savings and labor supply, and political risk.
The need for reform arises from projected financial shortfalls of the current system, a largely pay-
as-you-go (PAYGO) transfer system. Trust fund assets cannot sustain the system. The assets were
arguably not generated through real government saving (in light of the history of persistent
budget deficits). The problem is not merely a blip that occurs as the baby boom retires. The
worker/recipient ratio is projected to fall permanently and ultimately either taxes must be
increased by about 50% or benefits must be cut by one third. Inaction would likely lead to
significant tax increases in the future since it is difficult to cut the benefits of existing recipients.
Reform now would allow future recipients to adapt to benefit changes and the economy to expand
through saving, making the long run problem less burdensome.
Proposed reforms involve revisions to the present system, the introduction of individual accounts,
or a combination of both. All else equal, raising taxes or cutting benefits increases government
saving; most taxes would not significantly affect the private saving rate (especially wage and
consumption taxes). Evidence suggests that some benefit reductions (e.g., raising both the early
and full retirement age) are more likely to increase labor supply and the tax base than others.
Investing trust fund balances in equities cannot, however, provide higher aggregate returns unless
financed through higher taxes or lower government spending. Otherwise, gains from investment
earnings will be offset by lower returns to private portfolios and higher government interest rates.
This analysis suggests that increased government borrowing costs could largely negate the
perceived gains to the government.
Individual accounts are often touted for their higher returns. However, these comparisons are not
accurate: they do not account for general equilibrium effects on interest rates, transition costs of
paying off existing liabilities, the increased risk for individuals, and—in some cases—the full
cost of the current system’s social functions (e.g. disability, transfers). If debt financed, individual
accounts systems would magnify the crisis because transition costs would increase. Moreover, if a
pure individual account system is to be successful in addressing market failures such as adverse
selection and free riders, it must be made mandatory in participation, annuitization, and prudent
investment. Two problems emerge even if these rules are followed. Individual accounts would
redistribute away from the poor because of their shorter lifespan and would eliminate the explicit
redistribution of the current system. Individual accounts would also expose cohorts of retirees to
significant variation in benefits due to the vagaries of the stock market. Individual accounts,
however, could reduce tax distortions and political risk and facilitate budgetary discipline. This
report will not be updated.
ȱȱȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
ȱ
The Economic Rationale for Social Security .................................................................................. 1
Adverse Selection ............................................................................................................... 1
Moral Hazard ...................................................................................................................... 2
Incomplete Private Insurance Markets................................................................................ 2
Failure to Optimize ............................................................................................................. 2
Economic Costs of Social Security..................................................................................... 2
What Is the Problem? ...................................................................................................................... 3
How Big Is the Problem and What Role Do Social Security Surpluses Play?.......................... 6
Economic Effects of the Current System: Savings and Labor Supply .......................................... 10
Social Security and Private Savings........................................................................................ 10
Labor Supply........................................................................................................................... 13
The Solutions: Potential Social Security Reforms within the System........................................... 15
Inaction and the Default Solution............................................................................................ 16
Raising Taxes .......................................................................................................................... 17
Cutting Benefits ...................................................................................................................... 18
Choosing Between Tax Increases and Benefit Cuts.......................................................... 19
Investing the Trust Fund in Higher Yielding Private Assets ................................................... 20
Numerical Simulation ....................................................................................................... 21
Individual Accounts....................................................................................................................... 23
Rate of Return Comparisons ................................................................................................... 23
Transition Costs ................................................................................................................ 24
Social Security’s Social Functions.................................................................................... 29
Administrative Costs......................................................................................................... 32
Can Individual Accounts Fulfill the Objectives of Social Insurance?..................................... 32
Restrictions on Choice ...................................................................................................... 32
Minimum Benefit Guarantees and Moral Hazard............................................................. 33
Should Annuitization Be Mandatory?............................................................................... 34
Should Bequests Be Allowed? .......................................................................................... 35
Risk: Collective Social Insurance vs. Individual Accounts .............................................. 35
Conclusion..................................................................................................................................... 39
ȱ
Figure 1. Worker-Recipient Ration, 1970-2080 .............................................................................. 5
Figure 2. Projected Revenues and Outlays of the Social Security System, 2006-2080 .................. 8
Figure 3. Change in Social Security’s Finances From Introducing 2% Individual Accounts ....... 26
Figure 4. Change in the Unified Budget Deficit From Debt-Financing 2% Individual
Accounts..................................................................................................................................... 27
Figure 5. Change in Budget Deficit From Debt-Financing Option 2 ............................................ 28
Figure 6. Change in Budget Deficit From Debt-Financing Option 3 ............................................ 28
ȱȱȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
ȱ
Table 1. Effects on Earnings, Federal Interest Costs and Rates of Return from Investing
Trust Fund Assets in Equities, Simulation ................................................................................. 22
Table 2. Percentage Increase in Benefit Due to Social Security’s Tax Treatment Relative
to Taxable Private Savings ......................................................................................................... 30
Table 3. The Historical Performance of Hypothetical Individual Accounts.................................. 37
¡ȱ
Appendix. Portfolio Model............................................................................................................ 41
ȱ
Author Contact Information .......................................................................................................... 42
Acknowledgment........................................................................................................................... 42
ȱȱȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
omprehensive Social Security reform has been an issue of debate in Congress for some
time, but no major action has occurred. Reform proposals have been driven in part by a
C recognition that in the future the program is financially unsustainable under current policy.
This report tackles the issue from an economic perspective that focuses not merely on reform that
achieves programmatic sustainability (sustainability within the trust fund), but reform that
achieves sustainability for the government and economy as a whole. Moreover, it stresses the
importance of understanding the economic consequences of program changes if one is to
understand who pays to achieve this sustainability and what type of changes might be consistent
with the economic rationale for social insurance.
ȱȱȱȱȱ¢ȱ
Before turning to the assessment of the current problem and the proposed solutions to that
problem, it is worth taking a moment to consider why we have a Social Security system at all.
Historically, of course, Social Security arose as a Depression-era program to alleviate the
economic circumstances of the elderly. For that reason (and perhaps others) it began as a transfer
rather than a fully-funded retirement system and the susceptibility of a tax and transfer system to
demographic changes is the reason for the current problem. Today, Social Security is more than a
retirement system; it provides disability payments (which account for about 15% of benefits) as
well as benefits to survivors and dependents. (There is a companion program, Medicare, that
provides for health benefits in retirement.) Social Security has also been a way to provide income
redistribution that has lifted many elderly out of poverty without resort to explicit (and to some,
demeaning) welfare programs. To some, these other functions of Social Security represent social
goals that should be separated from the provision of retirement income. For others, they are an
integral part of Social Security’s role of insuring against the risks associated with old age.
Why not simply allow individuals to cope with their own retirement and other risks (such as
disability) through private saving and insurance? Economic theory suggests several reasons that
are discussed below.
ȱȱ
One reason for an aggregated mandatory social insurance program is a problem called “adverse
selection.” Since individuals know more about their own circumstances than insurance firms,
individuals who are more likely to benefit from insurance would tend to purchase it. For example,
using one’s life savings to purchase an annuity that pays a fixed amount per month over the rest
of one’s life would be most attractive to individuals who expect to live a long time (and
unattractive to a person with a terminal disease). So the insurance industry will only offer to pay a
smaller annuity than that justified by average life expectancy because the companies know that
their customers will more likely be those who expect a long life. These less advantageous terms
lead annuities to be attractive to an even smaller group of individuals. The result is that markets
do not provide attractive annuities to the average person. The problem can be avoided by the
government mandating the purchase of annuities or by directly providing them through a tax and
benefit system.
ȱȱȱ
ŗȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
ȱ
£ȱ
A second problem is one known as “moral hazard,” which arises from behavioral responses to
incentives. In this case, moral hazard arises if society has a safety net for only the poor.
Individuals know that if they do not save for retirement, society will not allow them to be
destitute, so there is an incentive to “free ride” on the program—to undersave, to take too many
risks in saving choices, and to spend accumulated savings too quickly. By requiring individuals to
participate in (and pay taxes into) a minimum retirement system that pays a lifetime annuity upon
retirement that is not risky, this moral hazard problem can be avoided.
ȱȱ ȱȱ
A social insurance system also allows for the reduction of individual risk in saving for retirement,
by spreading investment and other risks across individuals within a generation as well as across
generations. Even if private insurance markets are complete within generations, they are unlikely
to be complete between generations, implying that a social insurance system can lead to
efficiency gains for society as a whole.1 A prolonged slump in the stock market, for example, can
be especially damaging to the generation reaching retirement age during that period.
ȱȱ£ȱ
There are other reasons for a mandatory social insurance system. Individuals may not always do
what is in their own best interest, insurance and annuity purchases are complicated, and the
optimal lifetime savings plan is one susceptible to mistakes or myopia. Economists would call
this problem a failure of individual optimization. The fact that so many individuals retire at their
earliest age of eligibility is suggestive that such failures might occur. If the failure to optimize is
the dominant reason for Social Security, then economic models and theories may have very little
predictive power regarding the effects of reform.2
ȱȱȱȱ¢ȱ
Social insurance programs also impose costs on society. If taxes are not directly tied to benefits,
they can cause distortions in labor supply and other economic behavior. Since the current Social
Security system is not fully funded, it may reduce national saving from its efficient level. It may
be politically difficult to maintain an accumulated asset fund that is not viewed as permitting a
greater degree of deficit financing in the remaining part of the budget. And even social insurance
programs that are funded may not reflect the optimal choices in asset holdings.
Since the program is legislated, it can also expose participants to the political risk of arbitrary
transfers among individuals. Note, however, that the potential future tax and benefit changes
necessary to address the current solvency problem would not disappear in a system of private
saving: political risk should be thought of as shifting costs and benefits among individuals and is
1 See Robert Schiller, “Social Insurance and Institutions for Intergenerational, Intragenerational, and International Risk
Sharing,” Carnegie-Rochester Public Policy Conference Series, Apr. 1998.
2 Other economic explanations for Social Security have been advanced. See Casey Mulligan and Xavier Sala-i-Martin,
Social Security in Theory and Practice (II): Efficiency Theories, Narrative Theories, and Implications for Reform,
National Bureau of Economic Research, Working Paper 7119, May 1999.
ȱȱȱ
Řȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
not responsible for real risks such as the demographic changes that are creating the current
financing problem.
These economic rationales for social insurance, and the costs associated with them, provide an
important framework for analyzing alternative approaches to reform and will be referred to
frequently in this report. To illustrate their importance, consider briefly what they imply about
individual accounts, which have been proposed as a full or partial substitute for Social Security
by some and which have the advantage of reducing economic distortions. Individual accounts (as
a complete substitute for Social Security) would not address the problems of adverse selection
and moral hazard unless they were mandatory (both with respect to contributions and to
conversion to a life annuity) and prudently invested. If these restrictions were not introduced, then
there is no economic advantage to government involvement in retirement decisions and such
decisions could be left to the private market. They also do not permit explicit redistribution in
favor of lower income individuals as the current system does. Indeed, given the higher mortality
rates of lower income individuals, higher income individuals would actually enjoy higher benefits
on average. (The longer expected life span of high income individuals give them an advantage in
the current system as well, but explicit redistribution toward lower-paid workers offsets this
effect.)
Therefore, individual accounts are caught on the horns of a dilemma: they cannot simultaneously
satisfy the objectives of eliminating adverse selection and moral hazard and the distributional
objectives of the current system. To maintain all of these objectives, individual accounts would
need to constitute only a part of the Social Security system or the accounts would need to be
subsidized for low income individuals.
ȱ ȱȱǵȱ
Social Security was, and largely remains, a pay-as-you-go (PAYGO) system in which current
workers pay the retirement benefits of current retirees through a payroll tax.3 There is no saving
component to this type of system; from an economic perspective, it is a generational transfer
system, not a pension or investment system. If benefits equal tax payments, as a pure PAYGO
system implies, then total benefits can only grow at the rate that payroll tax revenues grow—with
a stable population growth and age distribution, that will be the growth rate of the economy. Thus,
for an average retiree who has spent his entire life contributing to a PAYGO, the “rate of return”
on his payroll taxes would be about equal to the growth rate of the economy. But for a retiree in
the early years of the system who lived part of his working life before the system’s inception, the
“rate of return” on his payroll taxes will be much higher because he receives the same benefits
without making a lifetime of contributions.4 According to one estimate, past and present
generations will receive $9.1 trillion more in benefits than they paid in contributions in present
value terms, which under current policy will be borne by future generations.5
3 The economic incidence of the payroll tax is borne by the worker regardless of who has the legal obligation to pay.
Thus, the incidence of the one-half of taxes paid by employers is no different from taxes paid by employees.
4 For more information, see John Geanakopolis, Olivia Mitchell, and Stephen Zeldes, Social Security Money’s Worth,
NBER, Working Paper 6722, Sept. 1998.
5 Jagdeesh Gokhale and Kent Smetters, Fiscal and Generational Imbalances (Washington, DC: AEI Press, 2003), p.
26.
ȱȱȱ
řȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
When the number of retirees grows faster than payroll revenues, benefits per average retiree need
to fall for the system to remain solvent, and the return on contributions would fall (and could
become negative). Trouble arises in the future because the retirement of the “baby boom”
generation, along with increasing life spans, cause the growth rate of retirees to greatly outstrip
the growth rate of workers in the system. A pure PAYGO system could rectify this imbalance by
increasing taxes in order to maintain the historical growth rate of benefits or by decreasing the
growth rate of benefits to maintain taxes. Neglecting economic effects for the moment, the choice
is simply a transfer of income: increasing taxes in the future constitutes a transfer from future
workers to the baby boomers, while lowering benefits constitutes a transfer from the baby
boomers to future workers.
The only other possibility would be to cover the difference through the issuance of government
debt to the public. But this option would simply transfer the burden further into the future because
the budget deficits would decrease the national saving rate and with it the future size of the
economy; in the long run such an approach could not be sustained. In any case, the imbalance is
large enough that debt issuance could not be a primary solution to the problem even in an
intermediate horizon. CBO forecasts suggest that as the baby boomers retire, debt would quickly
exceed 200% of GDP, primarily because of higher health care spending. It would be difficult for
the government to convince its citizens to purchase debt at these levels.
The problem of the baby boomers’ retirement is well appreciated. What may be less well-
understood is that the financing problem is a permanent problem; it is not a decade-long blip
while the baby boomers retire. A key to financing a PAYGO system is the worker to recipient
ratio, and it is the level of the worker-recipient ratio that determines the revenues of a PAYGO
system, not the change in the ratio. As seen in Figure 1, this ratio deteriorates markedly while the
baby boomers retire, from 3.4 workers per recipient today to about 2.0 by 2040.6 If the ratio then
recovered as the baby boomers died, the financing problem would be a temporary one whose cost
could be spread by borrowing to generations before and after the blip. But the problem is
permanent: the ratio is not projected to recover. Part of the reason for this effect is the increase in
life expectancy after-9 retirement, which reflects assumptions about both longevity and earlier
retirement.
6 It is always important to remember that forecasting is not a science, and any future estimates are only the most likely
outcome of an uncertain event. As a projection moves further into the future, uncertainty increases. In their report, the
OASDI Trustees address this issue by offering alternative high and low cost scenarios. Under the high cost scenario,
the worker-recipient ratio would reach 1.8 by 2040 and under the low cost scenario, the ratio would reach 2.4 in 2040.
ȱȱȱ
Śȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
Figure 1. Worker-Recipient Ration, 1970-2080
Source: Trustees of OASDI, Annual Report, 2006, Table IV.B2.
Note: Solid line based on Trustees’ intermediate assumptions. Dashed lines represent worker-recipient ratio
under optimistic and pessimistic scenarios.
One can contrast the past and future financial state of the Social Security system with the state
that would have been achieved if a fully funded system had originally been instated instead of a
pay-as-you-go system. In a fully funded system, current workers pay for their own future benefits
(through either a collective account or individual accounts) instead of the benefits of current
retirees. Until they retire, the proceeds are saved in real financial securities, allowing a fully
funded system to achieve a “market” rate of return. How would outcomes differ under this type of
system as opposed to a PAYGO system? Benefit levels and the “rate of return” on workers’ taxes
would no longer be directly related to the worker-recipient ratio (it would only change as the
capital stock per worker is affected). Thus, the retirement of the baby boomers would not require
reductions in benefits or tax increases because their benefits would come from their own
contributions and the earnings on them.7 Where would the money come from to make this
possible? It would come from the fact that early retirees who spent only a few working years
under the system would have received little or no benefits. Instead, the money they would have
received under a PAYGO system would have been saved and invested for later retirees. Because
money has already been transferred to the early retirees in the current system, there is no reform
to that system to achieve solvency that can avoid lowering someone’s “rate of return.”
The Social Security system has operated differently from a pure pay-as-you-go system retirement
system historically in two important ways. First, unlike a pure PAYGO, the Social Security
system has not always paid out as much as it has taken in. Social Security has always maintained
a small reserve against short-term fluctuations. More importantly, changes to the system in 1983
7 Increasing longevity, however, would result in smaller annual annuities for a fixed amount of contributions under a
funded system; this effect cannot be avoided, but it does not affect the typical rate of return.
ȱȱȱ
śȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
led to Social Security surpluses that have grown to be 1.4% of GDP in 2006. From an accounting
perspective, these surpluses have introduced a partial funding aspect to the system, and the Social
Security trust fund has grown to $1,687 billion as a result. Unfortunately, the budget deficits of
the government in most of these years have arguably prevented the surpluses from partially
funding the system in an economic sense; indeed most years since the 1980s debt held by the
public as a percentage of GDP increased.
Second, the average single retiree will not enjoy the PAYGO rate of return equal to the growth
rate of the economy because a portion of his contributions are diverted to benefits unrelated to
retirement. The current system also provides benefits to the disabled, survivors, and dependents;
in 2003, these uses accounted for more than one-third of total benefits paid (18% for survivors,
15% for disability, and 5% for dependents). In addition, it explicitly redistributes benefits by
earnings level, apparently further lowering the system’s rate of return for higher paid workers. For
example, a single worker retiring at age 65 in 2003 earning low wages would recoup the
retirement portion of his and his employer’s payroll taxes plus interest in 10.4 years, a worker
earning an average wage would recoup them in 14.9 years, and a worker earning the maximum
eligible wage would recoup them in 21.4 years.8 (However, comparisons of the benefit structure
overstate the actual degree of redistribution because of other factors, such as greater life
expectancy for high income individuals).9 It is important to factor in the resources devoted to
these social functions when calculating the return the system can earn for retirees. Even a funded
system could not earn the return of a private saving account if it were to maintain these social
functions.
ȱȱ ȱȱȱȱȱȱȱȱ¢ȱ
ȱ¢ǵȱ
This report defines reform as a measure that returns the system to long-term solvency. The first
question to ask is: what is solvency? The narrow definition, used by the Social Security trustees,
of preventing the Social Security trust fund from being depleted over 75 years seems
unsatisfactory from an economic perspective. After all, the trust fund can be replenished in any
number of ways that would not make it any easier for the government to meet its financing needs
and would have no effect on the economy. For example, income tax revenues could simply be
diverted to the trust fund anytime benefits exceed payroll tax revenues. But if the government
wished to maintain its other spending without raising taxes, it would then have to borrow to
finance spending that would otherwise be financed through income tax revenues. The government
cannot borrow sufficiently to maintain current policy through the retirement of the baby boomers
(largely because of the rapid projected growth of Medicare spending) because debt cannot grow
faster than GDP without limit as current policy implies.
8 CRS Report RL31034, Social Security: The Relationship Between Taxes and Benefits for Past, Present, and Future
Retirees, by Geoffrey Kollmann.
9 See Karen Smith, Eric Toder, and Howard Iams, “Lifetime Distributional Effects of Social Security Retirement
Benefits”; and Lee Cohen, C. Eugene Steuerle, and Adam Carasso, “Social Security Redistribtuion by Education, Race,
and Income: How Much and Why?” Both papers were presented at the third Annual Joint Conference for the
Retirement Research Consortium, “Making Hard Choices About Retirement,” May 17-18, 2001, Washington, DC. See
also Alan l. Gustman and Thomas L. Steinmeier, “How Effective Is Redistribution Under the Social Security Benefit
Formula?” Presented at Second Annual Joint Conference for the Retirement Research Consortium, May 17-18,
Washington, DC. Julia Lynn Coronado, Don Fullerton, and Thomas Glass, The Progressivity of Social Security,
National Bureau of Economic Research, Working Paper 7520, Feb. 2000.
ȱȱȱ
Ŝȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
This analysis raises a further question: does the system become unsustainable when benefits
exceed revenues, as is forecast to occur in 2017, or is it sustainable as long as there are assets in
the trust fund to draw down, as will be the case until 2040? The latter option seems unsatisfactory
when one considers that there are not any “real” assets in the trust fund at all. Since the Social
Security trust fund holds U.S. Treasuries, the government owes money to itself. It is analogous to
saying that one can afford to make a purchase because one’s left pocket has an IOU from one’s
right pocket, even though one’s right pocket is empty. What is in the left or right pocket by itself
is irrelevant, it is total wealth that determines whether the purchase can be made. Similarly, it is
the government’s overall ability to meet the needs of retirees that determines the sustainability of
the Social Security system when the baby boomers retire. The government’s overall ability to pay
benefits comes down to what benefits have been promised, how much overall tax revenue is
being raised, and what is the size of the economy from which taxes are being raised. The answer
to these three questions determines the underlying solvency of Social Security.
Because the decline in the worker to recipient ratio is projected to be permanent, the financing
problem is permanent. This makes the trustees’ estimate of the 75-year actuarial deficit of 2.02%
of payroll a misleadingly low estimate of the system’s financing problems—a trust fund that is
solvent for exactly 75 years merely shifts the permanent tax increases to year 76 and offers no
indication of how the government will finance the redemption of trust fund assets.10 The problem
with current policy is that, mostly due to the projected rise in Medicare spending, the overall gap
between revenues and outlays would place an unsustainable strain on the unified budget deficit.11
For this reason, although the trust funds are assets to the Social Security system, it is doubtful that
there will be general revenues available to honor those assets. Thus, the best measure to gauge the
size of the problem is the difference between taxes and benefits paid over time under current
policy, as shown in Figure 2. For the time being, focusing on this measure sets aside the issue of
trust fund balances, which are not meaningful in an economic sense without considering total
government assets or liabilities.
10 The measure of Social Security’s financing shortfall most commonly cited is the actuarial deficit. The actuarial
deficit is the size of the tax increase or benefit reduction, as a percentage of payroll, that would be required for the trust
fund to exhaust its assets in 75 years. It does not measure the tax increase or benefit reduction needed to prevent cash
deficits within the system over 75 years—the actuarial deficit assumes that cash deficits in any given year can be closed
by drawing down trust fund assets.
11 The indefinite maintenance of current policy implies that the national debt as a percentage of GDP would be
projected to asymptotically approach infinity.
ȱȱȱ
ŝȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
Figure 2. Projected Revenues and Outlays of the Social Security System, 2006-2080
Source: Congressional Budget Office
Note: Lines with markers represents projections under CBO’s intermediate assumptions. Dashed lines
represent projections under optimistic and pessimistic assumptions.
As Figure 2 demonstrates, in the long run, paying benefits under current policy would require
permanent tax increases, so that total revenues equaled 6.2% of GDP in 2030 and 6.9% of GDP
by 2080 (the end of the projection period), ignoring for the time being economic effects.12
Compared with current law, this percentage of payroll translates into a tax increase of roughly
50%. It would also be about 3½ times as large the Trustees’ actuarial deficit of 2.02% of payroll.
(Although this report addresses only Social Security, it should be stressed that Medicare and
Medicaid face projected shortfalls that are even larger, and would require even larger tax
increases to fund.)
The projections are extremely sensitive to underlying assumptions and should be treated with
caution. Under the low-cost alternative scenario, the shortfall would reach 0.5% of GDP by 2030
and 0.9% of GDP by 2080. Under the high-cost scenario, the shortfall would reach 1.9% of GDP
by 2030 and 3.9% of GDP in 2080. Thus, while the extent of needed changes is highly uncertain,
even under the low-cost scenario, some tax increase or benefit reduction would be necessary.
If a tax increase or benefit reduction of nearly 1% of GDP is necessary by 2030 and 2% of GDP
by 2080, but the system is currently in surplus, is there any reason to reform the system before it
12 Congressional Budget Office, Updated Long-Term Projections for Social Security, June 2006.
ȱȱȱ
Şȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
becomes insolvent? The worker to recipient ratio is nearly fixed, barring massive immigration or
higher birth or death rates. Only the future size of the economy and the future state of government
finances can be influenced today. Thus, reform can ease the future financing burden only if reform
causes the economy to grow faster and improves the government’s finances.13 These are
macroeconomic problems that cannot be posed in terms of the Social Security system’s trust fund.
In these terms, according to standard economic theory, the crucial factor in determining whether
the Social Security surpluses increase economic growth and improve the government’s finances
depends on whether they are saved. Saving by the government has two salutary effects. First, it
increases national saving, which increases private capital accumulation, the future size of the
economy, and with it the future size of tax receipts. In this way, saving budget surpluses is
analogous to transferring resources from present to future generations. Second, saving frees up
future government resources (by reducing interest payments and increasing tax revenues in
absolute terms) that make financing future imbalances easier. By contrast, when the Social
Security surplus is used to finance other government activities or tax cuts rather than being saved,
the Social Security trust fund is increased but nothing is done to alleviate the government’s future
fiscal imbalance.
Of course, it is difficult to actually determine whether the Social Security surplus was saved
because the budget that would have been in the absence of surpluses cannot be observed. But it is
known that the government had deficits every year from 1983 to 1997 (and again beginning in
2002), when surpluses were appearing in the Social Security system, and that debt as a percentage
of output increased during that period. If the Social Security surpluses were used to finance other
spending, then running surpluses in the accounts has done nothing to address the crisis in a real
sense. (The counter-argument is that the Social Security surpluses increased national saving
because the historical budget deficits would have been even bigger in their absence.)14
From this discussion it should be clear that saving a surplus from any source will increase the
future size of the economy and improve the government’s future financial position, even though
non-Social Security surpluses do not increase the Social Security trust fund. However, merely
saving all of the projected Social Security surpluses, which are determined by budget accounting
rules rather than the scope of the problem, is grossly insufficient to close the long-term financing
gap.
A word of caution is in order. Simply increasing the size of the economy alone will not convert a
permanently unsustainable system to a sustainable one. A higher wage base will increase tax
receipts but because under current law benefits are linked directly to wages, benefits will increase
as well. (Because it would not raise the benefits of those already retired, higher economic growth
would improve the system’s finances—but only temporarily.) What increased saving will do is
make other changes (such as higher taxes or reductions in benefits relative to wages) easier
because everyone will be better off. For example, if income is higher, a relatively lower tax rate
will be needed to raise a given amount of revenue. Plans that aim to improve sustainability
through economic growth would, therefore, be more effective if the link between wage growth
and benefit growth were weakened.
13 Reform in advance is also essential for approaches that involve benefit reductions so that workers have enough of
their working careers left to adjust their personal saving behavior to the reductions.
14 For alternative perspectives on the Social Security trust fund, see Kent Smetters, Is the Social Security Trust Fund
Worth Anything?, The Wharton School, University of Pennsylvania, Working Paper, May 13, 2002.
ȱȱȱ
şȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
ȱȱȱȱȱ¢DZȱȱ
ȱȱ¢ȱ
Before examining alternative reform proposals, this section begins with two important behavioral
effects of the current system: effects on savings and the supply of labor. The Social Security
system, like all tax and transfer systems, has potential effects on the size of the economy through
effects on labor supply and capital accumulation. First, the tax and transfer system could alter
savings decisions because it reduces income in the present and increases it in the future.
Secondly, the payroll taxes themselves could alter labor supply as would be the case for any tax
on labor income, with the effects depending on the linkage between taxes and benefits. In Social
Security, there is a link between benefits and taxes, however, so payroll taxes are not taxes in a
strict sense; workers may view them more as contributions that will later be recouped. Higher
earnings during the working years result in higher benefits, albeit earning a much smaller return
than would savings and perhaps reflecting forced savings.15 Thus, the behavioral effects of a tax
and transfer system occupy a middle ground between an ordinary tax and expenditures system
and a pension system. Moreover, individuals may perceive the tax as different from other taxes;
indeed, the payroll tax is a tax that has generally not been as unpopular as income taxes.
ȱ¢ȱȱȱȱ
To an individual, participation in the Social Security system appears to be a form of saving: it
reduces disposable income in working years and increases it in the future. In certain models of
household behavior over time (e.g., the life cycle model), such an income shift should reduce
personal saving. A natural response is to rebalance consumption over time by increasing
consumption today and reducing funds available in the future that are less needed. In a simple
model where Social Security earns a normal return and borrowing is available, savings should fall
by exactly the amount of Social Security taxes. This expected effect has been the basis for much
criticism of the unfunded Social Security system. That is, the Social Security system can be seen
as a system that provides effects similar to saving in a pension plan, except that no real
accumulation of capital necessarily occurs. (The system has generated surpluses but this effect is
not very meaningful for two reasons: first, most of the system is still pay-as-you-go and the
surpluses are small in relation to the future size of the unfunded liabilities, and secondly, these
surpluses may have permitted larger deficits in programs outside of Social Security, so that it is
not clear that real savings has occurred).
However, this effect on savings is not certain for several reasons. First, to the extent that a Social
Security system substitutes for pre-existing intergenerational transfers, no savings effects would
be expected. For example, if, in the absence of Social Security, individuals expect to support their
parents and expect, in turn, to be supported by their children, and that obligation is reduced or
eliminated by Social Security, the transfer system is acting as a substitute for private transfers and
would not necessarily reduce private saving. Similarly, if parents had expected to leave bequests
to their children, the parents would have increased their bequests to offset the payroll tax, and
children would wish to save that bequest to relieve their own children of the tax. (Most families
do not leave bequests of any significance, however, so it is the former rather than the latter
15 In the Medicare program, there is no link between the level of earnings and the level of benefits.
ȱȱȱ
ŗŖȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
phenomenon that is more likely). At the extreme, it can be argued that if all individuals are
connected through intergenerational transfers, anything the government does is offset by what
individuals do (this phenomenon is referred to as Ricardian equivalence). Although such a model
is obviously not entirely realistic (some individuals have no children, for example), some
elements of it may be present in the U.S. economy.
In either of these models, outcomes may also be affected to the extent that some individuals are
liquidity constrained; that is, they would like to consume more now but cannot. Liquidity-
constrained individuals are affected only by cash flow effects; that is, taxes lower consumption
and future income such as Social Security has no effect on current consumption. There is
evidence that a significant fraction of individuals falls into this category. People may also save for
precautionary reasons because of future uncertainty and the potential for misfortune. Social
Security may lower private saving by reducing uncertainty and the need for precautionary saving.
(Then again, the family safety net may make precautionary saving largely unnecessary, so that the
introduction of Social Security has little effect on precautionary saving.) Finally, this analysis of
savings effects assumes that individuals are making optimizing choices. However, making the
appropriate lifetime savings decision is not only complicated, it is done only once so people
cannot learn from mistakes. Thus, the possibility for a significant error must be considered.
All of this discussion suggests that turning to empirical evidence to assess the effects on savings
is necessary. Much of this approach has been based on time series studies, which examine the
change in savings rates over time with changes in Social Security wealth. Unfortunately, the
econometric evidence has been mixed.16 An early influential study by Feldstein (1974) found
Social Security significantly reduced private savings (actually increasing consumption). However,
that study was found to have a programming error by Leimer and Lesnoy (1982), which caused
the results to become smaller and generally statistically insignificant. Feldstein (1982, 1996) still
found significant results with additional data: his findings suggested that Social Security reduced
the saving rate by half in 1992. These results were questioned by Meguire (1998), who found a
result about a tenth as large and by Coates and Humphreys (1999) who found results that tended
to be smaller and, in some cases, in the opposite direction. A part of the problem with these
studies is that it is very difficult to measure Social Security wealth; another is that it is difficult to
control for other effects; and finally, the results tend to be sensitive to functional form. In general,
studies have found some savings offset, but typically not enough for Social Security to offset
private wealth dollar for dollar.
Some studies have used cross section data, and in most cases found some effect on saving, but
considerably less than the full savings reduction implied by the life cycle modeling approach.
Most of the effects were not statistically significant, however. These data are, in any case, quite
questionable because Social Security rules are generally applied uniformly; thus, any variation is
16 See Martin Feldstein, “Social Security, Induced Retirement and Aggregate Capital Accumulation,” Journal of
Political Economy, 82 (Sep./Oct. 1974), pp. 905-926; Dean R. Leimer and Selig D. Lesnoy, “Social Security and
Private Saving: New Time-Series Evidence.” Journal of Political Economy 90 (June 1982), pp. 606-629; Martin
Feldstein, “Social Security and Private Saving: Reply,” Journal of Political Economy, 90 (June 1982), pp. 630-642;
Martin Feldstein, “Social Security and Savings: New Time Series Evidence,” National Tax Journal 49 (June 1996), pp.
151-163; Philip Meguire, “Social Security and Private Savings,” National Tax Journal, 51 (June 1998), pp. 339-358;
Dennis Coates and Brad R. Humphreys, “Social Security and Savings: A Comment,” National Tax Journal 51 (June
1999), pp. 261-268.
ȱȱȱ
ŗŗȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
correlated with other characteristics (income, family status, age) that could independently affect
savings.17
A final approach is to compare savings rates in countries with different Social Security systems.
These results tend to not be very conclusive, finding both positive and negative effects, usually
without statistical significance. But perhaps it is not surprising that it is difficult to detect results
given the possible influence of cultural mores on saving.18 In general, therefore, the evidence
suggests a savings effect but one that is by no means large enough to be consistent with the life
cycle model.
At this point, a word of caution is also in order, for predictions of the effects of Social Security
reform that rest on simulation models. An example is the recent study that uses a dynamic model
to estimate the effects of Social Security revisions.19 This model is a life cycle model (although it
allows for bequests, the bequest motive is not the type that leads to no effect of Social Security on
savings). This study tends to find privatization an attractive option, in part because of its effect on
saving. But this result is driven in part by the fundamental nature of the model. A different model
would not have led to the same predicted effect.
It should also be recognized that if individuals are well-informed, rational, and optimizing, the
projected insolvency of the Social Security system should already have an effect on the national
saving rate. If people working today know that the system will be capable of paying only, say,
two-thirds of their benefits as promised under current law, then they should be saving more than
they would if the system faced no crisis. On the other hand, if individuals currently believe they
will pay the taxes and receive the benefits promised under current law, reform could induce more
private saving. This could happen because individuals are not well informed (e.g., they base their
expected benefit on account statements from the Social Security Administration instead of the
financial position of the system) or because they think that others will bear the burden of solving
the problem (e.g., they believe taxes will be raised after they have retired).
The theoretical and empirical evidence suggests that Social Security has had a less clear cut effect
on saving than is often assumed. If Social Security has not caused individuals to save less, then
the PAYGO approach may not be necessarily faulted as discouraging savings. In any case, it is
one thing to argue that had we originally introduced a fully funded system rather than a PAYGO
system, we would have achieved higher national saving rates. But it is another to argue that now
that we have had a PAYGO system for over 60 years, shifting to a funded system will raise the
national saving rate. As discussed below, moving from a PAYGO to fully funded system involves
a significant transition cost, and if this transition cost is financed through debt issuance, national
saving will not rise. Thus, the move to a fully funded system is only likely to increase national
saving if real resources are devoted to financing the transition in the form of higher taxes or lower
benefits.
17 See the review in Social Security and Private Saving: A Review of the Empirical Evidence, Congressional Budget
Office, July 1998.
18 Ibid.
19 Laurence J. Kotlikoff, Kent Smetters, and Jan Walliser, Finding a Way Out of America’s Demographic Dilemma,
National Bureau of Economic Research, Working Paper 8258, April 2001.
ȱȱȱ
ŗŘȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
ȱ¢ȱ
Social Security systems can also affect labor supply. Economic theory suggests that wage taxes
have the potential to reduce or increase work effort depending on offsetting income and
substitution effects; these effects may be reflected in hours worked or the workforce participation
rate. If the tax reduces hours worked or participation, raising payroll taxes themselves could have
a negative effect on the size of the work force relative to the retired population and the size of the
economy. Most evidence, however, suggests that labor supply is not very responsive to the wage
rate, because of offsetting income and substitution effects. Indeed, for men the evidence suggests
that an increase in net wages would slightly reduce work, while for married women the effect is a
slight increase; both of these effects are more likely to reflect participation choices rather than
choices of hours worked. Based on empirical studies that reflected taxes, the Congressional
Budget Office puts the total response at an elasticity of between 0 and 0.3, meaning that a 10%
reduction in wage might reduce work effort by up to 3%, reflecting a response of between -0.1 to
0.2 for men and 0.3 to 0.7 for married women.20
Social Security has some particular attributes for the effect on work that might modify these
results in various ways. First, there is a link between payroll taxes and the size of Social Security
at retirement that makes the payroll tax only partially a tax. Because of redistribution in the
system and the PAYGO nature, this relationship is not perfect (i.e., the return earned is too low
and variation in taxes does not lead to the same degree of variation in benefits), but the payroll
tax is more directly linked to benefits than are other taxes. Secondly, there is a ceiling on covered
wages, so that higher income workers do not experience the marginal tax effects that tend to lead
to reduced work. Indeed, for such individuals (which would include higher income men), there is,
if anything, an inducement to increase work from payroll taxes via income effects.
For married women, who have generally been viewed as most responsive to taxes in their labor
supply, the link between taxes and benefits is lessened because they may receive spousal benefits
even if they do not work. This latter link suggests that the benefits to contributing are smaller than
would be the case for other workers, and the presence of the spousal benefit could reduce labor
force participation. Benefits for a non-working spouse are essentially a transfer, and some have
argued that such benefits should not be larger for non-working wives of higher income men;
rather the payment should be a flat payment. A smaller flat benefit would increase the tie between
Social Security contributions and benefits for married women who work, and should increase
their labor supply. However, empirically estimating the labor supply responsiveness of women to
wages is always difficult because potential wages of non-workers are not observed. Moreover, the
responsiveness to Social Security benefits also depends on perceiving and expecting that working
will result in losing the non-working spouse transfer; younger women particularly may be
unaware of how spousal benefits are formulated. Empirical evidence that focuses particularly on
Social Security is relatively limited. One study found a significant effect on labor force
participation of the implicit benefit differential due to being eligible for a spousal benefit for older
married women, but not for younger ones; another study found that the benefit differential led to
earlier retirement.21
20 U.S. Congressional Budget Office, Labor Supply and Taxes, Memorandum, Jan. 1996.
21 See Therese A. McCarty, “The Effect of Social Security on Married Women’s Labor Force Participation,” National
Tax Journal, vol. 43 (Mar. 1990), pp. 95-110; Jessica Primoff Vistnes, “An Empirical Analysis of Married Women’s
Retirement Decisions,” National Tax Journal, vol. 47 (Mar. 1994), pp. 135-156.
ȱȱȱ
ŗřȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
Perhaps the clearest effect of Social Security on work effort has been the increase in retirement of
older men since the introduction of the system. Between 1950 and 1989, labor force participation
decreased from 46% to 17% for those 65 and over, and from 87% to 67% for those 55 to 64; the
median age of retirement for both men and women fell from around 67 years to about 62 years.22
Labor force participation of men aged 62-64 fell dramatically after the introduction of early
retirement at 62 in the early 1960s, even though there was an actuarial correction to prevent early
retirees from drawing larger lifetime benefits. In 1963, 76% of these men worked; by 1985, only
46% participated (the number has roughly stabilized at that rate).23 The fall in retirement age
appears to have slowed down, and perhaps stabilized from the 1970s to the 1990s, however.
There are a number of factors that might affect this decline, but some of it is estimated to be due
to Social Security, in part because Social Security makes retirement feasible, and in part because
Social Security has an earning test that discourages work. In 2000, Congress repealed the test for
recipients 65 and older; it is probably too early to measure the potential effects of this change. For
those expecting not to live very long, retirement at 62 would be attractive despite the aggregate
actuarial correction. Moreover, by setting a minimum age for benefit eligibility, Social Security
may have contributed to setting some social norm.
Statistical studies have generally found an effect, although in some cases not a large effect, of
Social Security on retirement age. Some researchers attribute a significant portion of the
reduction in labor force participation to mandatory retirement policies adopted by businesses and
only recently made illegal, and to features of private pension systems which encouraged
retirement at a specific age.24 As in the case of saving, studies that rely on cross sections within
the United States may be questionable because Social Security wealth is correlated with other
variables that may affect retirement. Moreover, the observation of a falling retirement age
following the introduction of earlier retirement age eligibility and the dramatic fall in labor
participation has also occurred in other countries that did not experience major changes in private
pensions; older men’s labor force participation has been declining in these countries as well.
These studies suggest that a statutory early retirement age significantly influences the average age
of retirement.25
If these studies are correct, an important method of raising the worker to recipient ratio may be to
raise the age of early retirement (the “normal” retirement age is already being increased). Critics
argue that early retirement reduces the pressure on disability insurance and that many individuals
cannot easily work into older age. However, most occupations have probably become physically
less demanding. They also argue that allowing earlier retirement prevents people from being
thrust into poverty. The evidence suggests that about a quarter of men receiving early retirement
had a self-reported work-limiting disability. Of all early retirees, fewer than one in ten were both
disabled and had non-Social Security income below the poverty level.26 Another study estimated
22 See Andrew Samwick, “New Evidence on Pensions, Social Security and the Timing of Retirement,” Journal of
Public Economics, vol. 70 (Nov. 1998): 207-36; Murray Gendell, “Retirement Age Declines Again in 1990s,” Monthly
Labor Review, Oct. 2001.
23 William J. Wiatrowski, “Changing Retirement Age: Ups and Downs,” Monthly Labor Review, Apr. 2001.
24 Ibid; also see Catherine Philips Montalto, Yoonkyng Yuh, and Sherman Hanna, “Determinants of Planned
Retirement Age,” Financial Services Review 9 (2000); 1-15, and Congressional Budget Office, Raising the Earliest
Eligibility Age for Social Security Benefits, Jan. 1999.
25 See Jonathan Gruber and David Wise, Social Security Programs and Retirement Around the World, National Bureau
of Economic Research, Working Paper 6134, Aug. 1997.
26 Congressional Budget Office, Raising the Earliest Eligibility Age for Social Security Benefits, Jan. 1999.
ȱȱȱ
ŗŚȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
that cutting early Social Security benefits would increase the probability of normal retirement by
twice as much as the probability of disability retirement.27
ȱDZȱȱȱ¢ȱȱ
ȱȱ¢ȱ
Approaches to the Social Security’s financing imbalance fall into the following five major
categories:
• doing nothing at this time
• raising current and/or future taxes
• decreasing current and/or future benefits
• investing the trust fund in higher yielding assets
• the adoption of individual accounts
Whichever approach is ultimately chosen, changes will have to be made to the system in one
fashion or another. While this report does not provide a detailed quantitative analysis of the
options, they are each discussed briefly in light of the objectives of social insurance, the
distributional effects (both across generations and across income levels), and implications for
economic behavior. The efficacy of any reform will depend on whether it limits the size of
eventual tax increases or benefit reductions implied in current policy.
In some cases, the effects of specific revisions are not as clear as they are often portrayed. An
illustration is a proposal to invest trust fund assets in the stock market rather than in government
securities, thereby earning a higher rate of return. However, in the absence of basic increases in
the supply of capital and labor, this change cannot be a free lunch: somewhere, someone will pay
for these apparent gains. Another example is the potential effect of moving to voluntary
individual accounts, which can result in higher interest payments and, in the presence of adverse
selection and moral hazard, higher welfare costs in the rest of the budget. One of the most
important effects of this proposal is that individual accounts are supposed to shift from pay-as-
you-go finance to funded finance. To simultaneously solve the problem of financing promised
benefits in the Social Security system and shift to a permanently pre-funded (even if only in part)
system would require a great deal of new resources during the transition period.
Commentators tend to divide reform proposals into “sacrifice” approaches, such as benefit
reductions or tax increases, or “investment” approaches such as equity investment by the trust
fund or the introduction of individual accounts. This report suggests that policymakers cannot
simply choose one approach or the other; they must be linked to other changes to succeed. Tax
increases or benefit reductions at present are only useful if they result in a higher government
saving rate; this effect can be accomplished through debt reduction or an investment approach. If
changes lead to higher government spending outside of Social Security or to tax cuts, they will
have increased the size of the trust fund without improving the government’s ability to honor trust
27 See Olivia S. Mitchell and John W. R. Phillips, Retirement Responses to Early Social Security Benefit Reductions,
National Bureau of Economic Research, Working Paper 7963, Oct. 2000.
ȱȱȱ
ŗśȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
fund promises in the future. Likewise, investment approaches are only useful if they lead to a
higher government saving rate, and a higher government saving rate is only possible through tax
increases or benefit reductions. Otherwise, investment in the trust fund or individual accounts will
be “debt financed” and have no positive effect on national saving.
To make the system sustainable or solvent after reform, Social Security should have roughly no
effect on the unified budget balance every year. To achieve this if Social Security is to remain in
its current (PAYGO) form would require that its cost rate be approximately equivalent to the
income that has been designated to the program. It is also possible for some of the system’s
income to come from investment earnings, if the trust fund or individual accounts were invested,
or for the system to be redesigned to receive revenue from on-budget sources, such as the
proceeds of an increase in income tax revenues.
This definition of solvency may seem overly strict since there is nothing preventing the system
from running deficits as long as the trust fund possesses assets and nothing preventing the
government from supplementing the system with general revenues if it desires. Nevertheless, the
definition used in this report is conceptually useful because it is the only definition that requires
an explicit description of how the financing gap would be closed in each year. Reform options
that include proposals such as the use of unidentified general revenue transfers, the redemption of
trust fund assets, or the issuance of debt—on a temporary or permanent basis—implicitly depend
on further tax increases, benefit reductions, or reductions in other government spending, but do
not explain how they will occur. Again, it should be stressed that deficit financing does not
impose a smaller burden than tax increases or benefit reductions; it simply shifts the burden
forward to future generations.
Of course, reforms can only return the system to projected solvency, and the margin of
uncertainty on projections is large. If the projections are too pessimistic, then reforms could turn
out to be overly harsh. If the projections are too optimistic, then further tax increases or benefit
reductions would become necessary in the future.
ȱȱȱȱȱ
Inaction itself does have some consequences. As stressed at the beginning of this report, the
Social Security system is not sustainable in its current form, and inaction shifts the burden of
fixing it to the future. Without action now, however, these future changes cannot be softened by
preparation or warning. As noted above, given the permanent gap between income and outflow,
eventually payroll taxes will have to increase by about one half or benefits be cut by about one
third in order to match income and outflow. On the other hand, if Social Security cannot be more
successfully insulated from the rest of the budget and gains from reform are used to finance other
government spending or tax cuts as has happened in the past, inaction may be preferable.
A second reason to take action soon is that if benefits were to be reduced, it would be appropriate
to determine those reductions in advance so that workers could plan for them. Policy makers are
reluctant, for obvious reasons, to reduce benefits of those currently retired, and taking no action
today may simply result in increased taxes in the future as a default condition.
ȱȱȱ
ŗŜȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
ȱ¡ȱ
Raising taxes now and saving the proceeds by improving the unified budget balance would raise
the national saving rate and improve the government’s finances. (The beneficial effect would not
occur if taxes were raised and the proceeds were not saved.) Raising taxes now would impose
part of the adjustment burden on current workers, including the baby boom generation, if this
accumulation could occur without causing deficit finance in the rest of the budget. Increases in
the capital stock result in a true improvement in sustainability, as wage income would rise.
However, the improvement in solvency would be partly offset unless there were some reduction
in benefits relative to wages. That is, one would need to keep benefits on the path they were
already on which would effectively mean a cut in Social Security benefit relative to the larger
economy. Raising taxes now and keeping them permanently higher as a method of achieving
sustainability could convert the system to a partially funded system.
While tax increases are never popular, it should be noted that in this case—if one’s goal is to
maintain the current size of the Social Security system—raising taxes now is more economically
efficient than raising them in the future. Since the deadweight loss of taxation grows
exponentially as taxes increase, there is a smaller efficiency loss from raising taxes now than
maintaining them for the time being and raising them to a greater extent in the future.
If the current payroll tax base were raised, there would probably be relatively little adverse effect
on private savings or labor supply. Generally wage taxes are unlikely to have an effect on the
savings rate (or private capital accumulation) in the economy through price effects. Its only effect
may be that because any tax reduces net income, it reduces the saving of those who save at a
fixed rate. That is, some small part of any tax increase may come out of savings because
individuals wish to spread the reduction in consumption over their lifetimes.28 There are also
concerns about the regressivity of Social Security taxes, which might argue for increases in the
earnings base as well (currently there is a ceiling on wages subject to the payroll tax).
Alternatively, income taxes could be increased to help reduce the system’s liabilities. These taxes
are relatively progressive and fall on the elderly as well as workers, but they may have somewhat
larger distorting effects on savings and even labor supply. Income taxes have mixed effects: they
still have some effect on income in retirement, but they also have the potential for discouraging
savings through effects on the rates of return. These offsetting effects may be the reason that
substituting wage for income taxes typically has little effect on the economy in life cycle
models.29
Some studies have proposed using a consumption based tax, such as a value-added tax, to
supplement payroll taxes.30 One argument for using such a tax is that it is less likely to adversely
28 There are economic models where such an effect would not occur, including infinite-horizon models with fixed labor
supply, where taxes have no effects (only spending). However, in the context of these models, the government as well
as the economy must be dynamically stable and one could view the taxes simply as a consequence of spending
decisions.
29 See Eric Engen, Jane Gravelle, and Kent Smetters, “Dynamic Tax Models: Why They Do the Things They Do,”
National Tax Journal, vol. 50 (Sep. 1997), pp. 657-682.
30 See Laurence J. Kotlikoff, Kent Smetters, and Jan Walliser, “Finding a Way Out of America’s Demographic
Dilemma,” National Bureau of Economic Research, Working Paper 8258, Apr. 2001. Note that their study shows very
beneficial effects from VAT finance as opposed to finance via payroll or income taxes because in a life-cycle model
saving tends to respond very powerfully to such a tax regime.
ȱȱȱ
ŗŝȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
affect savings and labor supply. While payroll taxes fall solely on workers, consumption taxes
also impose a lump sum tax (i.e., a tax that does not distort economic behavior) on older
individuals because they fall on assets as well as on wages. In general, Social Security benefits
are not affected since they are inflation-indexed, so that elderly individuals without income from
wealth would not bear this burden. Moreover, because it falls on existing retired workers it can be
imposed at a lower rate than a wage tax.
Another approach to raising tax revenue would be to change the tax treatment of Social Security
benefits for purposes of the income tax to make it consistent with the taxation of private pensions,
as discussed below. Currently, about two-thirds of Social Security recipients pay no income tax
on their benefits. This approach could also be thought of as an effective benefit reduction. It
would fall relatively lightly on low-income individuals who would be less likely to pay income
tax because of standard deductions and personal exemptions in the tax system. Its impact would
also be limited for higher income individuals who already pay tax on a larger fraction of their
benefits.
ȱȱ
There are many ways to cut future benefits, which this report defines as reducing benefits relative
to current law rather than in absolute terms. Most proposals would make benefit cuts prospective,
to provide time for adjustment, so little of the effect would be felt by the current elderly. A
proportional, across-the-board, phased-in decrease could be used, or a slowing of indexation to
prices and wages could be used. The advantage of the latter approach is that the savings to the
government would grow over time, offsetting the growth in funding shortfall (although if carried
out indefinitely, it would cause Social Security benefits to become extremely small relative to
pre-retirement income). CBO estimates that one proposal along these lines could reduce outlays
to 4.1% of GDP from 6.2% of GDP (under current law) by 2050.31
Two particular approaches to restoring solvency that could also induce economically beneficial
behavioral effects are reducing benefits for dependents of high wage earners and increasing the
retirement age. As noted earlier, one of the transfers in the Social Security system is a spousal
benefit guarantee that rises with the payment to the primary worker. For a secondary worker, the
existence of these benefits means that contributions paid into Social Security earn a lower return,
and thus may have adverse behavioral effects more in the nature of pure taxes. Lowering the
expected benefit to women who do not work increases the return to working and should, in
theory, result in more labor force participation. Evidence (cited earlier) suggests that this
treatment particularly has an effect on the labor force participation of older married women.
A second approach is to increase the normal (or full benefit) retirement age. Proponents justify
this approach on the grounds that rising life expectancy means that retirees are collecting benefits
for longer than previous retirees. The full benefit retirement age is already scheduled to increase
to 67. Raising the full benefit retirement age would result in direct savings in terms of benefit
reductions, either because of fewer delayed retirement credits or because of reductions in
payments received in early retirement. CBO estimated that a proposal to raise the retirement age
and reduce early retirement benefits would reduce outlays in 2050 from 6.2% of GDP to 5% of
GDP.32 Another study found that raising the projected retirement age by an additional three years
31 Congressional Budget Office, The Long Term Budget Outlook, Dec. 2003, p. 22.
32 Congressional Budget Office, The Long Term Budget Outlook, Dec. 2003, p. 24.
ȱȱȱ
ŗŞȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
would reduce the projected increase in spending on Social Security (as a percentage of GDP) by
about 40%.33 To the extent that workers delayed retirement because of these changes, output
would be higher and more taxes would also be collected.
An alternative that might have a greater behavioral effect is to raise the age of early and full
benefit retirement or to reduce benefits for early retirees. Such an approach might have a larger
effect on delaying retirement, which would also boost output and tax revenue. Raising the early
retirement age should also reduce the amount of adverse selection in the system (the tendency of
shorter-lived people to choose early retirement); reducing benefits to early retirees permits
adverse selection but does not reward it as much. However, if early retirement decisions are not
primarily driven by adverse selection, the behavioral effect of lower early retirement benefits may
be limited. These proposals have a cost, in that the burden may fall disproportionately on
demographic groups with lower life expectancy (including lower income individuals, men, and
minorities) or those who work in arduous occupations, and it may increase claims for retirement
based on disability.
ȱ ȱ¡ȱ ȱȱȱȱ
Given that tax increases or benefit cuts appear necessary to restore the system to solvency, does
analysis provide any guidelines as to what mixture of the two would be desirable? The answer to
this question depends on the economically optimal size of the Social Security system, which in
turn depends on the tradeoff between the system’s costs and benefits.34 The smaller the system,
the less it is able to alleviate the market failures it was designed to cure—moral hazard, adverse
selection, incomplete insurance markets, and failure of optimization. The larger the system, the
more it distorts labor and saving decisions. Higher taxes have negative substitution effects, while
benefit reductions have none. Thus, in economic models, benefit cuts would typically lead to
better macroeconomic outcomes than higher taxes because there would be no negative effect on
saving and labor supply.
Because evaluation of the costs and benefits of Social Security is so deeply embedded in value
judgements, the optimal size of the system must ultimately be determined through the political
process, and cannot be evaluated in this report. If the current size is too large, a reform should be
directed more to a benefit cut, with private saving making up the difference; if the current size is
too small, a reform should be more directed to a tax increase.
If the current system is assumed to already be the optimal size, however, at least some rationale
exists for providing a mixture of relatively small benefit cuts and large tax increases. If
individuals want to smooth the effects of reform over their lifetimes after reform is completed and
adjusted to, they might prefer a roughly proportional effect on their standard of living. Since
Social Security benefits are a larger fraction of retirement income than Social Security taxes are
of workers’ income, it could be argued that much of the adjustment might be made in tax
increases. As an illustration, consider a case with a 10% contribution during a working period of
45 years, to finance an annuity for a retirement span of 10 years. Assume a 6% rate of return and
33 Ibid. After 75 years, costs (including Medicare) were projected to rise to about 0.264 from 0.137; instead they would
rise to 0.209. These projections included some adverse labor supply and savings responses.
34 It can be argued that the solution to the system’s funding crisis is not primarily an equity argument. If higher taxes or
lower benefits affect the same people (at different points in life), then there is no equity rationale for favoring one
approach over the other.
ȱȱȱ
ŗşȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
a 2% growth in wages. If the retirement span doubled to 20 years, one could either increase the
contribution by 55% or decrease the annual annuity by 35%. Suppose, however, one desired a
proportional decrease in income for all years. To accomplish that would require a tax increase of
about 47% and an annuity decrease of 4.7%—most of the adjustment (85%) would come on the
tax side. The share allocated to taxes would still be significant if the Social Security annuity
represented only part of retirement income. For example, the average share of retirement income
from Social Security is 51% for singles and 37% for married couples.35 With these shares, the tax
adjustment would be between about two-thirds and about three-fourths of the total adjustment,
respectively (rather than 85%). This relative shift away from tax increases to benefit cuts occurs
because a change in benefits now constitutes a smaller portion of overall retirement income.
Of course, to the extent that the cause of the crisis is greater longevity, individuals might also
prefer a delay in retirement age (which is characterized above as a retirement benefit reduction,
but which also increases the revenues available for benefits). In the illustration above, retirement
would need to be delayed by about four years to maintain benefits. If increased longevity were
accompanied by increased health during the foregone retirement years, then an increased
workspan might also be an optimal solution. Nevertheless, it would be problematic for those
whose health does not permit a longer working period.36
In sum, what this life cycle discussion suggests is that if Social Security were initially set at
optimal levels, only a relatively small portion of an adjustment to reflect longevity should be in
the form of a general benefit cut, and most of the adjustment should take place in either higher
taxes or higher retirement age.
ȱȱȱȱȱ
ȱȱȱȱ
One proposal intended to alleviate the Social Security problem is to invest funds in higher
yielding assets. Although most often associated with the development of individual accounts,
other proposals have suggested that such investments be made by the trust fund itself instead. In
either case, the results would be the same.
The key to whether this strategy provides any net national benefit rests on whether it is debt
financed. If the investments were funded by raising taxes or lowering government spending, they
will result in higher rates of capital accumulation, which would increase the future size of the
economy and increase future resources. But if the investments were financed through the issuance
of government debt or the use of existing surpluses, the exercise amounts to a mere shifting of
assets into different portfolios. Since this shift results in no increase in net national savings, there
would be no increase in the total capital stock, and the exercise would not create more resources
for the nation as a whole. As will be shown in the simulation below, diverting funds out of non-
marketable Treasury bonds and into equities would decrease returns to other investments and
raise the government borrowing rate. Much of the improvement in the system’s finances would be
offset by higher interest payments in the overall government from the debt issued to finance the
trust fund’s investments. Moreover, increases in the cost of government debt would further
35 Social Security Administration, Office of Policy, Income of the Population 55 and Older, Feb. 2002.
36 One solution would be to shift more workers with poor health near normal retirement age into the disability program.
Another solution would be to increase the normal retirement age, continue to allow early retirement with a reduced
benefit on average, but provide subsidies to permit maintaining a minimum benefit so that these individuals are not
thrust into poverty.
ȱȱȱ
ŘŖȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
increase the budget deficit. To the extent that the system’s equities earn a higher rate of return
than the government must pay on its additional debt, the government’s overall finances would
have improved at the expense of a burden placed on anyone who holds private equities, including
the retired.37
Assuming the fund remained a defined benefit plan, any risk of lower return would also need to
be made up, presumably by higher taxes. Estimates that find that investment in equities improves
trust fund balances are based on expected returns. The variance of financial markets implies that
in reality returns on assets could be higher or lower than their expected return. For example, one
study found that if equity markets continued to perform as they have historically, there is a greater
than 25% chance that the value of a trust fund invested in equities would be lower after 10 years
than if it were invested in government bonds, as it is now. Over longer periods of time, however,
the probability diminishes: after 75 years there is a less than 1% chance that the equity-invested
trust fund would have a lower value.38
ȱȱ
A simulation will illustrate how investment by the trust fund affects financial markets and the
government budget. In a simple model with only one type of debt and equity, the effects of
shifting from holding debt to equity is clear, and the gain in earnings will be divided into an
offsetting rise in government interest payments, and a fall in the return on equities. (This fall in
the return on equities also reduces the return on trust fund assets, although this is a second order
effect for a small change.) The same results would occur if the government introduced individual
accounts with debt financing, and the accounts were invested in the stock market.
Moreover, one can estimate the basic magnitude of the effects (as discussed briefly in the
appendix, using 2000 asset shares), which depend on the willingness of business to substitute debt
and equity and the willingness of individuals to substitute in their portfolios. (The degree of
willingness is called an elasticity; it is the percentage change in the ratio of assets divided by the
percentage change in the ratio of asset returns.) The smaller these elasticities are, the larger the
effect on government borrowing. For example, if both elasticities are 0.3 (and there is some
evidence to support small elasticities),39 then 45% of equity earnings will show up as an offsetting
increased cost of borrowing, so the government as a whole gains only about half of the amount
projected. The public, in its role as a private investor, loses the remaining 55%. The amount that
shows up as higher interest costs falls as elasticities rise, becoming 12% at elasticities of one.
37 An extensive literature describes the presence of what economists call an equity premium in financial markets. That
is, the spread between the returns on equities and the returns on bonds is larger than can be explained by models which
attempt to weigh the tradeoff between risk and return. If the equity premium does (still) exist and it is caused by a
failure to optimize, investment by the Social Security system (centrally or through individual accounts) has the
potential to lead to an improvement in social welfare. By contrast, if the equity premium is the result of other factors
(e.g., transaction costs) or mis-specification by economists, then there will be no increase in social welfare. Even if
social welfare was increased, the reduction in the equity premium would still, by definition, lead to the same relative
reduction in rates of return.
38 Amy Harris, Noah Meyerson, and Joel Smith, “Social Insecurity? The Effects of Equity Investments on Social
Security Finance,” paper presented at National Tax Association Symposium, April 2001.
39 See Jane G. Gravelle, The Economic Effects of Taxing Capital Income (Cambridge MA: MIT Press, 1994), p. 84 for
a summary.
ȱȱȱ
Řŗȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
This model is highly simplified, and one of the most important simplifications is that it assumes
only two assets. The model would be richer if assets were divided into riskless government debt,
risky private debt, and equities. The results for a three-asset model are presented in Table 1. As
before, the outcome depends on assumptions about elasticities of substitution: the degree to which
firms are willing to substitute debt for equity in their demands for capital (σ), the degree to which
individuals are willing to substitute debt for equity in their portfolios (ε) , and the degree to which
individuals are willing to substitute government for private debt (S). The limited evidence we
have suggests that debt and equity are not perceived as close substitutes and we set those
elasticities relatively low; we assume, however, that private and public debt is a much closer
substitute and set those elasticities relatively high.
Table 1. Effects on Earnings, Federal Interest Costs and Rates of Return from
Investing Trust Fund Assets in Equities, Simulation
Values of σ, ε,
Increased
Increased
Percent Increase in Private Interest
and S
Earnings of
Federal
Rate (I
Trust Fund (or
Interest, as % of
p), Equity Return (E),
Government Interest Rate (I
Individual
Col. 2
g) and Total
Private Return (R)
Accounts)
(% of GDP)
Ip
E
Ig
R
0.3, 0.5, 2.0
0.2136
59.0
-5
a
56 -1.1
0.6, 0.5, 2.0
0.2150
55.0
-7
a 53 -1.1
0.3, 0.8, 2.0
0.2155
59.0
-5
a 56 -1.1
0.3, 0.5, 1.5
0.2136
89.0
-5
a 85 -0.3
0.3, 0.5, 2.5
0.2136
43.0
-5
a 41 -1.5
0.3, 0.5, 1.01
0.2135
163.0
-5
a 156 1.8
Source: Author’s Calculations.
Notes: σ, ε, and S refer respectively to the elasticity of substitution between debt and equity by firms, the
elasticity of substitution between equity and a composite of government and private debt in individual portfolios
and the substitution between private and government debt in individual portfolios. Results based on 2000 asset
shares. See Appendix for model details.
a. Less than 1%.
The results suggest that the substitutability between government and private debt is an important
factor in determining whether the benefit of higher returns with a shift to equity primarily results
in a decline in private returns or an increase in government interest payments. In the simulation,
the effect on the return to equities themselves, is relatively small. With that elasticity set at 2.5,
interest costs rise by 43% of the gain in the return to the trust fund, while the remaining 57%
appears in a decline in private returns, mostly through a decline in private interest rates. With an
elasticity of 1.01, interest costs rise by more than the gain in trust fund earnings and private
returns rise.
This model is not definitive, but rather illustrative. It suggests that a potentially significant
amount of the gain in earnings would be offset by an increased burden on taxpayers through
higher interest rates. The results are the same whether assets are accumulated centrally by the
government or in individual accounts, although in the latter case the investment returns would
flow to account owners while the debt payments would be borne by the government.
ȱȱȱ
ŘŘȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
ȱȱ
Perhaps the most sweeping change is a proposal that Social Security move away from its
collective, defined benefit format towards a system of defined-contribution individual accounts.
Were these accounts to completely replace the current system, then the PAYGO system would be
replaced by a fully funded system.
Other advocates have argued that individual accounts should augment rather than replace the
current system. One of the key differences among competing individual accounts proposals
concerns whether they would be “added on to” or “carved out of” the current Social Security
system’s tax and benefit structure. But pursuing either of these approaches alone would not
reduce the system’s unfunded liabilities. Add-on accounts that use additional revenues (in the
form of higher taxes, general revenues, or voluntary contributions) to finance individual accounts
but leave the current benefit structure of the system intact would not alter the system’s future
financing imbalance. Likewise, proposals for carve-out accounts that divert revenues from the
payroll tax into individual accounts, and then lower benefits paid by the system to offset the
diverted revenues only reduce the system’s unfunded liabilities to the limited extent that the
benefit offset is greater than the diverted revenues, as explained below. The unfunded liabilities
can only be significantly reduced by raising (“adding on”) taxes while holding benefits constant
or cutting (“carving out”) benefits while holding taxes constant.40
The objective of the following discussion of individual accounts is to analyze the claimed merits
of individual accounts, as well as explaining some of their disadvantages. For analytical clarity,
consider the economic effects of a hypothetical proposal to move completely from the current
PAYGO system to a fully funded system of pure individual accounts. (The report’s conclusions
apply to specific proposals only where explicitly noted.) Once the costs and benefits of a pure
system have been identified, one can evaluate whether specific costs and benefits of a mixed
system are attributable to its individual accounts or its government-provided PAYGO portion. A
mixed system could address some of the market failures and social goals that a pure system of
individual accounts could not, but would also reduce the benefits stemming from individual
accounts. First, we begin by discussing a claim commonly made for individual accounts: that
rates of return are larger than the return in the Social Security system and that these higher returns
can help achieve solvency.
ȱȱȱȱ
There is no doubt that expected returns are higher on private assets than benefits under the current
system. Yields for current workers are expected to be small or even negative because of the fall in
the worker-retiree ratio upon the retirement of the baby boomers. But even setting this
demographic problem aside, and considering a sustainable steady state, if the economy is growing
at a real rate of 2% per year, a dollar paid into a PAYGO system over, say a 25-year period, would
permit a benefit payment of $1.64. A dollar invested in an individual account that yielded an
40 From an economic perspective, carve-out accounts would be unlikely to significantly affect public or private saving.
Add-on accounts, however, could lead to some increase in national saving, and hence economic growth, but only if
they did not supplant private saving. This would be the case for individuals who under-save. For individuals who
planned their saving through optimization and were not liquidity constrained, add-on accounts would strictly supplant
private saving unless they were subsidized, in which case they would instead reduce public saving.
ȱȱȱ
Řřȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
expected 8% return would be expected to permit a benefit payment of $6.85. Such is the power of
compound interest.
There are several important problems with this argument. The one noted in the previous
simulation holds for individual accounts as well: without an increase in the capital stock, there
can be no net gain to society. Gains in individual accounts will be offset by losses to other
investors and to the rest of the government budget. By our calculations, it is certainly possible for
half or more of the apparent gain from investing in equities to appear as an added interest cost in
the general budget. While an individual account may appear to outperform Social Security,
society as a whole (including individual account holders) cannot escape the unfunded liabilities of
the current system.
There are other reasons that Social Security cannot earn a market rate of return. First, over one-
third of contributions finance benefits for the disabled, survivors, and dependents. Contributions
also redistribute income to lower income recipients. Second, individual accounts are likely to
have higher administrative costs than Social Security. Finally, the risky returns offered by
individual accounts cannot be meaningfully compared to the safer returns offered by Social
Security since people willingly accept lower returns to avoid risks.
The economy-wide benefit to individual accounts would not come from the rate of the return that
they earned. It would come from the stronger link between benefits and contributions that should
reduce or eliminate the distortions arising from social insurance, such as effects on labor supply
and savings. These distortions may not be completely eliminated as long as there are mandatory
components to the system and some individuals save more than they would otherwise prefer. But
in general they should reduce distortions in private decisions to work and save.
ȱȱ
Furthermore, a reform plan based on individual assets will have to deal with the cost of paying off
currently accrued obligations and making the transition to a funded system. This effect is far from
trivial. Suppose all contributions were immediately shifted into individual accounts but the
system was committed to make no one who had already participated in the system worse off.
Existing retirees would still have to receive benefits, and working individuals would have to pay
close to the existing 12.4% payroll tax, as well as set aside their new contributions, to finance
those benefits. These payroll tax dollars would earn no return, even of principal, for their
contributors.
What if a mixed system were introduced that diverted some of the payroll tax, say 2%, into
individual accounts and used the rest to pay current benefits? Would this allow the future funding
crisis to be diverted without tax increases or benefit reductions? To make this evaluation more
concrete, we analyze the budgetary effects of the proposals of the President’s Commission to
Strengthen Social Security, based on the official estimates of the Social Security Administration
(SSA) actuaries. The estimates make it clear that introducing 2% individual accounts without
introducing a new stream of revenue to finance them would not prevent a funding crisis.41
41 See also Peter Diamond and Peter Orszag, “Reducing Benefits and Subsidizing Individual Accounts: An Analysis of
the Plans Proposed by the President’s Commission to Strengthen Social Security,” Center for Budget and Policy
Priorities and the Century Foundation, June 18, 2002.
ȱȱȱ
ŘŚȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
The President’s Commission on Social Security proposed three such plans. In Option 1, workers
under 55 could elect to deposit 2% of their OASDI taxable earnings in an individual account,
with an offsetting reduction in benefits based on diverted amounts compounded at a real rate of
return 3.5%. It is the only option that relies solely on individual accounts. In the second, the
amounts diverted were 4%, up to $1,000 per year, of taxable earnings with benefits offset by
compounding at a real rate of return of 2%. This plan also included a benefit cut: to slow the
growth of benefits by indexing to prices rather than wages. Option 3 allowed the diversion of
2.5% of earnings not to exceed $1,000 plus a required additional 1% of taxable earnings for
which a refundable tax credit would be received. The diverted amounts would reduce benefits,
compounded at a real rate of return of 2.5%. According to the trustees, Options 2 and 3 lead to a
significant long-term improvement (but short-term deterioration) in the system’s finances, while
Option 1 does not.42 None of the reforms achieve solvency according to this report’s definition,
however.
Figure 3 illustrates the effect of introducing individual accounts on the income and costs of the
Social Security system as a percentage of taxable payroll.43 The estimates are based on the
introduction of individual accounts under the Option 1 plan of the President’s Commission on
Strengthening Social Security. Option 1 makes the effects of individual accounts clearest because
it is the only option that does not include benefit cuts. Roughly 40 years after the introduction of
individual accounts, when enough retirees had spent their entire career making contributions, the
accounts would indeed allow a larger reduction in Social Security benefits than the revenue lost
from a 2% marginal payroll tax cut.44 Over time, the savings to the government would get larger
and larger and eventually become quite significant. Unfortunately, for the first 40 years after the
introduction of individual accounts, the accounts would generate less revenue than was being
diverted from beneficiaries.
42 For the sake of brevity, many of the details of the commission’s proposals are omitted here. For a detailed description
of the commission and its proposals, see CRS Report RS21095, Social Security: Report of the President’s Commission
to Strengthen Social Security, by Dawn Nuschler.
43 This graph does not illustrate the overall income and cost rates under a system of individual accounts, but rather the
change in income and cost rates caused by the introduction of individual accounts. Thus, the future financing crisis is
not averted if the proposals generate additional revenues. It is only averted if the proposals generate enough additional
revenues to cover the entire financing gap.
44 This graph does not illustrate the revenue generated to the individuals who own the individual accounts. It illustrates
the reduction in benefits that the government is obligated to pay because Social Security benefits are offset by an
amount equal to the diverted payroll tax compounded at a 3.5% real rate per year.
ȱȱȱ
Řśȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
Figure 3. Change in Social Security’s Finances From Introducing 2% Individual
Accounts
Source: Chief Actuaries of Social Security Administration, Memorandum to President’s Commission to Strengthen
Social Security, December 2001, p. 53.
Notes: Estimates based on Commission’s Option 1 reform plan of an individual account with contributions
equal 2% of payroll and a 66.7% participation rate. It assumes accounts are introduced in 2004.
Thus, individual accounts would exacerbate the financing problems of the Social Security system
for roughly 40 years, but improve the system’s finances from that point on. Whether this is
desirable can only be determined by judging their effect on the economy and government as a
whole. And such accounts improve the government’s overall finances and national saving only if
the revenue shortfalls caused by them in the first 40 years of their existence are financed through
tax increases or benefit reductions. In this case, the improvement in the government’s finances
could be calculated by compounding the rising line in Figure 3 (reduction in cost rate). By
contrast, Figure 4 illustrates the effects the accounts have on the unified budget balance if the
accounts are debt financed (i.e., introduced without an accompanying source of financing). As
explained in the introduction, the effect on the unified budget balance is the most meaningful
measure because it is the only one that clearly identifies the cost of reform to the government as a
whole. It may surprise some readers to see that individual accounts, if debt financed, still generate
less revenue than the current system for the entire 75-year projection window. In other words,
debt-financed individual accounts worsen Social Security’s financing crisis for the next 75 years.
ȱȱȱ
ŘŜȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
Figure 4. Change in the Unified Budget Deficit From Debt-Financing 2% Individual
Accounts
2%
1%
0%
% of GDP
-1%
-2%
2004
2014
2024
2034
2044
2054
2064
2074
Year
66.7% Participation
100% Participation Rate
Source: Chief Actuaries of Social Security Administration, Memorandum to President’s Commission to Strengthen
Social Security, December 2001, pp. 53-54.
Notes: Estimates based on Commission’s Option 1 reform plan of an individual account with contributions
equal to 2% of payroll and a 66.7% participation rate. It assumes accounts are introduced in 2004.
Since the private assets in individual accounts are growing through the accumulation of interest,
these results may seem puzzling. Although the assets in individual accounts are growing through
the accumulation of interest, the debt issued by the government to finance the accounts is also
growing because of interest accumulation. The individual accounts offset government benefits at
a higher rate of return than the interest payments on government debt, so eventually (outside the
75-year window) the individual accounts save the government money. But because so much debt
must be issued immediately, it takes more than 75 years before the accounts pay for themselves.
In fact, the actuaries of the SSA make two sets of assumptions, both of which are illustrated in
Figure 4. Since the accounts are voluntary, they estimated the costs under the assumptions (1)
that 66.7% of eligible individuals elect to participate, and (2) that 100% elect to participate.
Comparing the results under the two assumptions leads to a surprising result: the best way to limit
the costliness of debt-financed individual accounts is to reduce the participation rate.
ȱȱȱ
Řŝȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
The President’s Commission on Social
Security made it clear that Option 1would not
Figure 5. Change in Budget Deficit From
avert a future financing crisis. For that reason,
Debt-Financing Option 2
it also proposed two further options that
5%
coupled the introduction of individual
accounts with benefit reductions. These two
4%
options, it asserted, would make Social
3%
Security solvent. In both cases, the actuaries
P
D 2%
of the SSA believe that a 66.7% participation
f G
rate is most likely, and the figures below are
% o 1%
based on this assumption. The reforms do not
0%
make the system solvent, however, if
solvency is defined as the avoidance of cash
-1%
2004
2014
2024
2034
2044
2054
2064
2074
deficits throughout the 75-year forecast
Year
window; both proposals would greatly reduce
Benefit Cuts Only
Benefit Cuts and Individual Accounts
the size of the deficits in the second half of the projection window, however. Under Option 2, the
system would generate cash deficits from 2010 to 2059. Under Option 3, it would generate cash
deficits from 2014 to 2072. In other words, additional unidentified funding would need to be
found to pay benefits promised under reform, and this funding would ultimately come from
further tax increases or reductions in benefits or other government spending. However, these
options do achieve solvency if solvency is defined as achieving a cash surplus at the end of the
forecast window. This means that—unlike reform proposals that achieve 75-year trust fund
balance—they do not result in renewed crisis in 76 years.
By examining Figures 5 and 6, which
Figure 6. Change in Budget Deficit From
illustrate the effect these options would have
Debt-Financing Option 3
on the unified budget balance, it can been
seen that if the individual accounts are debt
4%
financed, only the benefit reductions are
3%
responsible for the improvement in the
system’s finances.45 Indeed, a much larger
P2%
D
improvement in the system’s finances is
G fo
possible if the individual accounts were
1%
%
excluded from Option 2 and 3 and only
0%
benefit reductions were made (assuming the
proceeds were saved). In other words, the
-1%
2004
2014
2024
2034
2044
2054
2064
2074
system’s finances could be improved to the
Year
same degree with smaller benefit reductions
Benefit Cuts Only
Benefit Cuts and Individual Accounts
than recommended in Option 2 and 3 if individual accounts were not implemented.
Source: Chief Actuaries of Social Security Administration, Memorandum to President’s Commission to Strengthen
Social Security, December 2001, pp. 57-58.
Notes: Estimates based on Commission’s Option 2 and Option 3 reform plan with 66.7% participation rate and
contribution rates equal to 2.39% and 1.97% of taxable earnings, respectively. It assumes individual accounts are
introduced in 2004 and are debt financed.
45 Although the benefit reductions in Options 2 and 3 grow significantly over time, for the first half of the projection
window they would generate insufficient budgetary savings to finance the individual accounts. For that reason, the
creation of individual accounts would result in an increasing debt for roughly the first half of the projection period
unless taxes were increased or other government spending were reduced.
ȱȱȱ
ŘŞȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
Throughout this discussion, the report has focused on the effect of individual accounts on the
government’s finances. This may seem at odds with many other evaluations of individual
accounts which center on whether there are gains to the worker. But the effect on the
government’s finances is quite separate from the effect on the worker. Government finances are
affected because the benefits the government must pay are reduced by the presence of individual
accounts (e.g., in Option 1, benefits are reduced by the amount of tax redirected to the account
compounded at a real rate of 3.5%). The potential gain to the individual comes from the fact that
the accounts are expected to be worth more than the amount the individual has lost in reduced
benefits (e.g., under the assumptions of the SSA actuaries, the accounts would earn a real rate of
return of 4.6% whereas benefits would be reduced by 3.5% in Option 1).46 Debt financed
accounts would only result in an (eventual) cash-flow improvement in government finances if the
compounding factor used in the benefit offset is greater than the interest rate on government debt.
This is not the case in Option 2 or Option 3, which call for offsets with real interest compounding
factors of 2.0% and 2.5%, respectively, compared with a 3% rate of return on U.S. Treasuries.
The reason this report focuses on the effect on government finances is that the system’s crisis
centers on the fact that the government does not have enough money to pay promised benefits. In
preventing that crisis, estimating the gain to individuals is largely irrelevant. Individual accounts
could earn rates of return of 100% a year, but if they did not offset government-provided benefits,
then the funding shortfall would be the same, and the same unidentified financial sacrifice would
need to be made in the form of higher taxes or lower benefits. Furthermore, even if a reform
proposal reduced the system’s unfunded liabilities, if it did not eliminate them, then further
unspecified sacrifices would have to be made at some point. Since those unspecified sacrifices
would need to come from the same workers possessing the individual accounts, calculating the
rate of return on the accounts gives only a partial picture of the proposal’s overall effect on the
individual.
ȱ¢Ȃȱȱȱ
As discussed above, only 62% of the current system’s income is paid out to retired workers. The
other 38% is paid to disabled workers (13%), survivors (19%), and dependents (6%). These
payments for “social functions” directly lower the “rate of return” available to healthy single
retired workers. One, therefore, should make sure that, in comparisons, these social functions are
being included in any comparisons of the return on social security and individual accounts. For
example, if one contracted to invest and purchase an old age annuity, one would not be covered
for disability or insurance payments for early death; the purchase of such insurance would use
some of the investment that could grow in the annuity.
In addition, the “rate of return” available to wealthier retirees is lowered further because of the
system’s progressive benefit formula. The degree of this redistribution is not certain, because of
offsetting factors. Higher income individuals tend to have larger ratios of retirement years to
working years, largely because they live longer. They also tend to enter the workforce later
because of more extensive schooling. Moreover, the distributional effect depends on whether one
examines families or individuals. Higher income men who marry wives who do not work outside
46 Note that the SSA estimates do not take into account the negative effects on rates of return that portfolio shifts may
cause, as suggested in the numerical simulation above. Thus, individual accounts may have a lower return than the SSA
estimates suggest for any given portfolio.
ȱȱȱ
Řşȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
the home (a typical event in the past) receive benefits for their wives that are proportional to their
own benefit levels, and if one includes these benefits, high income families have a higher return.
Another regressive feature of the current Social Security system is the tax treatment of benefits
and payroll contributions. If Social Security is thought of as a retirement plan, contributions are
(like other retirement plans) subsidized through the tax system compared to other investments.
There is a subsidy because contributions made by employers are not subject to individual income
tax at the time of contribution and because “earnings” (benefits in excess of contributions) are
taxed only when received, and then taxed only partially. Social Security benefits are taxed
differently from pensions, however. A fraction of Social Security benefits is taxed when income
exceeds certain levels, with individuals with relatively high incomes paying tax on 85% of
benefits, low income individuals paying no tax, and some individuals falling in between.
Currently about two-thirds of all individuals do not pay a tax but the number covered will grow
absent legislative change because the exclusions and phase-outs are not indexed. Pensions are
taxed in a way that explicitly permits previously taxed benefits to be excluded (and many pension
systems allow full exclusion of contributions from the income tax base).
Even though there are no aggregate assets in a PAYGO transfer system, there is a return from the
perspective of the individual since benefits typically exceed contributions.47 These tax benefits
favor higher income individuals, and the subsidy per dollar of payroll tax is more pronounced,
other things equal, the higher the individual tax rate, the longer the holding period and the higher
the rate of return. Benefits are positive at any reasonable rate of return, and are also larger in
absolute value for higher income individuals because they have larger contributions. Table 2
shows the ratio of end-of-period value from an investment receiving Social Security’s tax
treatment compared to an investment subject to a normal tax for a variety of returns (including
holding periods), depending on whether benefits are taxed. The current system offsets these
effects through both direct redistribution and exclusions from benefit taxation, the latter of which
favors middle income individuals but not the poor, whose ratio is one (i.e. who would pay no
income tax in any case because of the income tax system’s regular exclusions).
Table 2. Percentage Increase in Benefit Due to Social Security’s Tax Treatment
Relative to Taxable Private Savings
Percentage Increase in Benefit for:
Income Tax Rate
Asset Held for 35 Years
Asset Held for 20 Years
0% Yield 5% Yield 10% Yield 0% Yield 5% Yield 10% Yield
15%
No Benefit Taxa 8.8 39.9 76.0 8.8 25.6 43.2
85% Taxedb -5.1
22.2
53.5
-5.1
9.6
25.0
47 As explained above, the rate of return on contributions to a PAYGO system would average the economy’s growth
rate in the steady state.
ȱȱȱ
řŖȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
Percentage Increase in Benefit for:
Income Tax Rate
Asset Held for 35 Years
Asset Held for 20 Years
0% Yield 5% Yield 10% Yield 0% Yield 5% Yield 10% Yield
25%
No Benefit Taxa 16.7
77.4
160.8
16.7
48.2 84.8
85% Taxedb -8.2
39.7
105.4
-8.2
16.7
45.5
Source: Author’s calculations.
Note: Calculated as [0.5/(1-t) + 0.5][1-0.85at]{(1+r)/[1+r(1-t)]}T - 1 where t is the tax rate, r is the yield, T is
the holding period and a is equal to zero with no benefit tax and one if 85% of benefit is taxed.
a. Assumes no part of Social Security benefits are taxed.
b. Assumes 85% of Social Security benefits are taxed.
A pure system of individual accounts would, therefore, tend to favor higher income individuals if
the current Social Security system’s tax treatment were applied to those accounts. The relative
subsidy to higher income individuals would increase in the post-transition period as rates of
return rise and the benefit exclusions are eroded through inflation. This tax subsidy differs from
that of IRAs, which have fixed dollar rather than percentage-of-wage limits (constraining the
benefit for high income individuals); the treatment is more like that provided for private pensions.
One could offset that effect by withdrawing favorable tax treatment (taxing employer
contributions and taxing account earnings as accrued) or by introducing offsetting subsidies, such
as higher contribution levels or government matching contributions for low income individuals.
When all of these factors are taken into account, the magnitude of the distributional effects of the
current system are not entirely clear.48 Nevertheless, these observations suggest that individual
accounts would place lower income individuals at a disadvantage compared to the current system
because they would lose the explicit redistribution in the current system. It is also possible that
they would place low income individuals at a further disadvantage if their tax treatment were
similar to the current system, unless the accounts were subsidized by earnings. At the same time,
because the poor do not live as long, individual accounts would maintain one of the
redistributional disadvantages of the current system if beneficiaries were forced to annuitize (for
the reasons discussed below).
Some proponents of the current system base their opposition to individual accounts on the fact
that these social functions could not be incorporated into individual accounts—“ownership” of
assets is incompatible with horizontal and vertical redistribution. While that is true, these social
functions could be carried out through new government programs that were financed through
general revenues.49 What should be realized is that a comparison between the rate of return
48 See Karen Smith, Eric Toder, and Howard Iams, “Lifetime Distributional Effects of Social Security Retirement
Benefits,” presented at the third Annual Joint Conference for the Retirement Research Consortium, “Making Hard
Choices About Retirement,” May 17-18, 2001, Washington, DC., who find that the current system favors the lowest
income group and projects redistribution to increase in the future under current policy. However, their paper does not
consider the distributional effects of tax treatment.
49 Survivor and disability benefits could be partially incorporated in a system of individual accounts through bequests
and annuities. Maintaining the level of benefits available under the current system would require government
subsidization, however. For example, personal accounts could not replace current survivor benefits when death comes
before retirement age.
ȱȱȱ
řŗȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
offered under the current system, which fulfills several social functions in addition to “paying
back” workers’ “investments” (payroll taxes), and the rate of return under an individual account
system that does not finance such social functions is invalid. Stated differently, dropping Social
Security’s social functions would be one way to reduce its unfunded liabilities, but the option is
not predicated on moving to a system of individual accounts.
ȱȱ
Another factor reducing the rate of return that could be achieved by a system of individual
accounts is the administrative costs that nearly all analysts agree would be higher than under the
current system. Administering the system would be significantly more complex than the current
system—individual account balances would need to be tracked, funds would need to be
transferred between investment portfolios upon request, assets would need to be bought and sold,
and so on. There is both an explicit cost to consider, in terms of the fees charged by the system’s
administrator (or funds appropriated if the system were administered by the government) and the
implicit cost to employers and employees in terms of more complicated paperwork and higher
compliance costs. The more choice that was allowed in the system, the higher the administrative
costs would be.
The administrative costs of the current system equaled 1% of benefits paid in 2000. MIT
economist Peter Diamond believes that the administrative costs of individual accounts could run
four to five times as high as the current Social Security system.50 The Thrift Savings Plan’s (TSP)
administrative costs are fairly low. As the retirement system for the country’s largest employer,
however, the centrally-administered TSP may enjoy economies of scale that would be unavailable
to private sector firms if individual accounts were administered competitively. The TSP also
offers relatively little investor choice (participants choose among five passive index funds). Thus,
administrative costs could be higher for private firms than the TSP, and this additional cost would
directly reduce the benefits paid to retirees.
ȱ ȱȱȱȱȱȱȱ ǵȱ
As explained above, the economic rationale for Social Security rests in part on the presence of
adverse selection, moral hazard, incomplete generational insurance markets, and non-
optimization in the market for retirement planning. This section explores the restrictions that
would need to be placed on individual accounts to correct for these factors. Without these
restrictions, there would be no economic advantage to the government provision of individual
accounts and retirement choices could be better left to the market. In the latter case, the
government could pay off the outstanding obligations of the existing system and leave individuals
on their own.
ȱȱȱ
Many of the arguments for individual accounts and issues in their design revolve around how
much choice individuals should have over their individual accounts. Many see an expansion of
choice as a compelling moral argument in favor of individual accounts, compared with what they
50 Peter Diamond, “Administrative Costs and Equilibrium Charges with Individual Accounts,” Working Paper, Mar.
2000.
ȱȱȱ
řŘȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
judge to be the paternalistic nature of Social Security. They argue that individuals should be able
to choose how to deal with their own contributions: whether to participate, what to invest in, and
whether to annuitize or withdraw their assets as they desire.
On the other hand, if the purpose of Social Security is to correct for the adverse selection, moral
hazard, and non-optimization, then these restrictions on choice are fundamental to the system.
They include not only mandatory participation, but also mandatory annuitization, restrictions on
account withdrawals, and restrictions to ensure that investments are prudent. The problem is not
only the lack of actuarially fair annuities, but also the possibility of free riders who do not save
and rely on welfare to support them in old age. In a sense, choice is in fundamental conflict with
the basic economic reason for social insurance.
Partial restrictions, such as requiring individuals to save but not annuitize, would also run afoul of
both adverse selection and moral hazard issues. Adverse selection would lead to smaller annuities
for those who take them, and still allow individuals to spend down their assets and then rely on
welfare. As a simple example, in an economy where half of the individuals live five years and
half live 20 years after retirement (assume all individuals know their own life span), an average
monthly actuarially fair mandatory annuity would pay $970 per $100,000 of assets. If annuities
were made voluntary, the individuals with five-year life spans would not purchase an annuity and
the annuity would fall to $454 per $100,000 of assets. But the need for mandatory annuities to
avoid moral hazard and adverse selection also means that it would be impossible to rely
completely on individual accounts without higher income individuals and women enjoying higher
rates of return, on average, than lower income individuals and men since the latter have shorter
life spans.
ȱȱ ȱȱȱ
£ȱ
To reproduce Social Security’s role as a dependable, safe portion of an individual’s retirement
portfolio (or for others, the only significant source of retirement income), some individual
account advocates have suggested that the government should “guarantee” a minimum benefit to
retirees in case their accounts suffer investment losses. It can be argued that such provisions, in
effect, would defeat the purpose of individual accounts, which is to allow individuals to shoulder
more risk if they desire and offer them more control over their retirement saving. Furthermore,
many proponents of individual accounts state that individual accounts could help fund the
system’s unfunded liabilities by achieving higher rates of return. The introduction of a minimum
benefit guarantee would make stand alone rate-of-return comparisons misleading because these
guarantees pass on the benefits of higher returns to the retiree while leaving the government with
the same liabilities to retirees as the current system.
In fact, the guarantees would increase the potential liabilities of the current system because of
moral hazard. If individuals received all of the benefits of holding high risk/return assets, but
assumed few of the risks (because they were guaranteed a minimum benefit), then they will shift
their portfolio towards riskier assets than they would otherwise desire. This choice increases the
expected value of liabilities to the government, since the presence of moral hazard makes it more
likely that the government will have to pay minimum guarantees.
The moral hazard problem can be reduced—but not eliminated—by placing restrictions on
investment decisions (e.g., limiting equities to, say, 60% of the portfolio) for a portion of the
portfolio, although this approach would run contrary to the goal of promoting investment
freedom. And the restrictions, by forcing investors to hold lower risk/return assets, would reduce
ȱȱȱ
řřȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
the expected return on individual accounts and make their funding advantage over the current
system less favorable. Individual accounts that entirely eliminate risk without creating moral
hazard would have to be invested in much lower yielding assets than proponents typically use in
rate of return comparisons.
Even without an explicit minimum guarantee, the government might feel compelled to use other
government programs, such as Supplemental Security Income, Medicaid, or the food stamp
program, to aid retirees whose life savings were lost. Thus, the moral hazard problem may not be
avoidable even in the absence of an explicit guarantee because individuals may act in the belief
that losses they suffer would be implicitly guaranteed for political reasons. If avoiding moral
hazard is the motivation for individual accounts, however, why not maintain Social Security as a
PAYGO system of a smaller size (that covers only a minimum benefit) and leave individuals free
to make other investments if and when they choose? And while higher spending on other
programs is not a “cost” to the retirement program, it is still a cost to the government that lowers
the funding advantage of individual accounts.
ȱ£ȱȱ¢ǵȱ
Aside from the volatility risk associated with private investments, retirees would face another risk
with individual accounts that they do not face with the current Social Security system: the risk
that they will outlive their assets, known as “longevity risk.” Because one’s time of death is
uncertain, there is the risk that retirees will draw down their assets too quickly and be left
impoverished in their last days. This risk can be avoided through annuitization, that is, through
the exchange with a financial intermediary of one’s assets for a promised stream of income.
Unfortunately, annuities can currently be purchased in private markets only on terms unfavorable
to retirees—on average, the expected value of the annuity benefit can be 15%-25% lower than life
expectancy would suggest.51 Since annuities would presumably be more desirable in the absence
of the current Social Security system, the purchase cost might go down as the market grew,
particularly if it reduced adverse selection. The only way to eliminate adverse selection
completely, however, would be for the government to make annuitization mandatory. This would
not prevent companies from attempting to “cherry pick” unhealthy retirees from each other,
however—either government provision or regulation would be needed to prevent it.
The problem of moral hazard also argues for making annuitization mandatory. Here again, the
government would need to weigh the benefits of allowing individuals greater freedom of
decision-making (e.g., the ability to spend a large amount of wealth suddenly in response to a
personal calamity) against the social desire to prevent people from becoming destitute because
they ran down their assets too quickly. Even if it were in a person’s best interest to buy an
annuity, he or she may not do so if given the choice. And again, if annuitization were not made
mandatory, the government might feel obliged to use other government resources to prevent
individuals who had run down their assets from becoming destitute.
Yet mandatory annuitization introduces an element of risk that is not present in the current
system: the problem of market fluctuations close to retirement undermining the value of an
individual account. If the market were to fall dramatically before retirement, mandatory
annuitization—particularly of accounts invested in equities—would “lock in” that low market
value, whereas individuals might feel that they could regain some of that lost income as the
51 Congressional Budget Office, Social Security Privatization and the Annuities Market, Feb. 1998.
ȱȱȱ
řŚȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
market turned around if they were allowed to hold on to their portfolio into retirement. (Of
course, their misfortune could be compounded if markets continued to fall.)
ȱȱȱ ǵȱ
With proponents’ stress on the “ownership” qualities of individual accounts, a logical next
question is whether individuals should be able to bequeath the balance of their account to their
heirs upon death. Doing so would raise some problems. Since closing the current system’s
unfunded liabilities is one of the major justifications proponents raise for switching to a system of
individual accounts, it is important to note that bequests would increase the unfunded liabilities of
the system. That is because they would represent a new set of benefits that do not exist under the
current system. The current system does pay survivor benefits to spouses and dependents, and to
the extent that bequests partially replaced those benefits, they would not represent a new liability.
But the ability to transfer benefits to a broader circle of heirs upon death—particularly in the case
of a premature death—is not possible under the current system and would represent a new
liability.
In a pure system of individual accounts, allowing bequests would reduce annuities. For example,
returning to our case where half of the individuals live for five years and half live for 20 years
after retirement, guaranteeing a minimum 10-year stream of payments (with the heir to receive
the payments in the case of premature death) would reduce the monthly annuity from $970 to
$807 per $100,000 of assets. However, giving individuals a choice between a larger annuity with
no bequest and a smaller annuity with a bequest would re-introduce the adverse selection
problem, because healthy individuals would likely choose the former and sick individuals would
choose the latter. This feature would lead to smaller annuities for the healthy than in the absence
of choice.
DZȱȱȱ ȱǯȱ ȱȱ
Individual accounts differ from an aggregated system in the allocation of risk. There are different
types of risk that face individuals in planning for retirement: some are unavoidable in any system,
while some can be mitigated, or worsened, by collective provision. Before examining specific
types of risks, it is useful to ask how risk affects individuals’ preparation for retirement, both
privately and through Social Security.
What Is the Role of Risk in Social Insurance? Generally in investment markets, higher rates of
return can be enjoyed only by taking on greater risk. The fact that an aggregate social insurance
system offers a relatively low return is not necessarily an undesirable feature of the system; nor in
a funded aggregate system would a portfolio invested in low return/low risk securities (such as
government securities) necessarily be undesirable. For most individuals willing to set aside
adequate investments for retirement, Social Security is only one portion of their investment
portfolio and most individuals would like to have some part of their retirement kept relatively
safe. They are free to invest their private assets in more risky ventures. Other individuals (who
fail to optimize) do not set aside adequate retirement savings, for whatever reason. Because they
are relying primarily on Social Security in their old age, a strong case can be made that it should
remain risk-free.52
52 Note that when using the term risk free, we are speaking of avoiding investment risk; demographic risks that are
(continued...)
ȱȱȱ
řśȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
How do the higher rates of return offered by individual accounts affect these two types of
individuals? For those individuals who save adequately, is the current system keeping too much
of their retirement income safe (and earning low returns)? Or if given the choice in an individual
account, would they place an equal proportion of their overall portfolio in riskless assets? If that
were the case, then the rate of return on their individual account would be no higher than their
overall portfolio is at the present, after accounting for transition costs. For those individuals who
do not save for retirement outside of Social Security, should society allow them to expose their
individual accounts to greater risk (and return), knowing that if they are unlucky they will have
no other private source of income to fall back on and that it will fall to government transfer
programs financed by all taxpayers to provide some minimum level of subsistence?
Investment Risk. A move away from the current PAYGO system towards individual accounts or
trust fund investments in private assets introduces a new type of risk into Social Security: the risk
of investment market volatility. The two approaches could theoretically distribute market risk
very differently, however. Some market risks can be reduced relatively easily, for example
through portfolio diversification (although it is not clear that all individuals reduce these risks
optimally in reality). There are other market risks that are more difficult to pool, however, such as
risk across time.
As explained above, a dollar invested in private assets by the Social Security system would have
the same rate of return as a dollar invested in the same asset through an individual account.
Obviously, the same amount of risk is associated with that investment in either case as well;
losses associated with trust fund investments would ultimately result in lower benefits or higher
taxes. But there is a significant difference in who bears the risk (and enjoys the profit) associated
with any given investment between the two systems. In a system of individual accounts, the entire
risk is placed on the individual who chose to make that investment. In a system with a
collectively invested “trust fund” and benefits only indirectly tied to the system’s income, those
risks can be spread both among the members of a generation (because the system has “defined
benefits” rather than “defined contributions”) and between generations if the system is allowed to
build up and draw down its assets over time.
Is the difference in risk-bearing between the two systems significant? Table 2 demonstrates that,
if the stock market performs in the future as it has in the past, the historical difference in
outcomes would have been extremely large.53 Historically, the stock market has frequently gone
through long phases of doing extremely well (e.g., the 1990s) or extremely poorly (e.g., the
1970s), and this would be fully reflected in the value of accounts accrued during those times. The
first column of the table demonstrates that if a system of individual accounts had been in place
between 1927 and 2001, the value of the account invested entirely in the Standard and Poor’s
composite stock index would have varied from 22.8% (for a retiree in 1974) to 97.8% (for a
retiree in 1999) of the value of Social Security benefits, depending solely upon the year in which
the worker retired.54 Of course, individuals could reduce risk by choosing a portfolio that mixed
(...continued)
currently affecting the return to Social Security, such as increased life span, would affect any retirement plan, whether
private or public.
53 For more information on the results in Table 2 and results under alternative assumptions, see CRS Report RL31324,
Social Security Reform: The Effect of Economic Variability on Individual Accounts and Their Annuities, by Geoffrey
Kollmann, Dawn Nuschler, and Patrick Purcell.
54 The table is based on workers who have typical work patterns that produce the Social Security benefit of someone
who always earned an average wage and who contribute 2% of pay for 41 years to individual accounts. In the first
(continued...)
ȱȱȱ
řŜȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
equities and bonds, but this would also reduce the rate of return they could enjoy, as demonstrated
in the second column. In fact, even though risk would be reduced, more than 25% of the time the
mixed portfolios of stocks and bonds would have underperformed the worst performing year for
the portfolio of 100% equities.55 This means that even if individual accounts matched the benefits
paid by Social Security on average, absent government guarantee subsidies, some beneficiaries
would do much better than under Social Security while others would do much worse.
Table 3. The Historical Performance of Hypothetical Individual Accounts
Statistical Measure
Percentage of Social Security benefit
replaced for account invested in:
100% Equities
60% Equities,
40% Bonds
Mean (Average)
49.1%
26.2%
Minimum 22.8
15.3
25th Percentile
41.0
20.6
50th Percentile
47.9
26.7
75th Percentile
55.6
31.3
Maximum 97.8
40.7
Standard Deviation
15.7
6.1
Source: CRS Report RL31324, Tables 4 and 12.
Demographic Risk. Another type of risk that is already present in the current system is
demographic risk, which cannot be avoided since mortality will always be unknown. In Social
Security or individual accounts with annuitization, demographic risk comes from dying before
one has received the value of one’s contributions. In individual accounts without annuitization,
the demographic risk comes from outliving one’s assets. (Demographic risk could be avoided
through annuitization with bequests, but would result in smaller annuity payments.)
A demographic risk that is part of the cause for the current insolvency of the system is the
increase in life spans that is not anticipated in advance, and will leave individuals with smaller
retirement accumulations and potentially smaller retirement incomes. Just as an individual facing
such an increase must either save more or accept a smaller retirement income (or delay
retirement), a social insurance system must increase taxes or reduce benefits. Thus potential
variability in taxes or benefits (which is sometimes characterized as part of the “political risk”) is
not a risk that can be avoided with an individual account system. And the large increase in taxes
or decrease in benefits needed to restore the current system to solvency is still present in any shift
to individual accounts. A pure system of individual accounts would still have to alter contribution
(...continued)
column, it is assumed that the accounts earn the same historical rate of return as the S&P 500 minus a 1%
administrative fee; in the second column, it is assumed that 60% of the portfolio earns the same rate of return as the
S&P 500 and 40% earns the historical return on long-term U.S. government bonds. The table assumes that the amount
accumulated in the account is converted to a fixed life annuity based on the prevailing rate of interest on long-term U.S.
bonds. It should be noted that volatility is higher if accounts are accumulated over shorter careers.
55 Similar results are confirmed by Gary Burtless, Social Security Privatization and Financial Market Risk, Center on
Social and Economic Dynamics, Working Paper 10, Feb. 2000.
ȱȱȱ
řŝȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
levels or benefits in the face of increasing life spans. Another type of risk, changes in fertility,
would not affect private accounts or a fully-funded social insurance system, but such changes
might still affect private retirement incomes by altering the availability of private
intergenerational transfers. Thus, these risks would not be avoided, but rather would be addressed
in different ways.
Political Risk. A final type of risk that many argue would be diminished by individual accounts is
political risk. Because there is no “ownership” of benefits in the current system, retirees and
workers planning for retirement are always at the mercy of legislative change. Individual
accounts would remove assets, income, and benefit flows from the government budget and create
individual “ownership” of retirement benefits that would be less subject to manipulation through
legislative changes. As a result, it might be more difficult to use surpluses in the retirement
system to finance other government spending. It is important, however, to distinguish this
political risk from actual risks such as demographic or investment risks discussed above. Only
political risk, which involves shifting benefits and costs among individuals, is zero-sum for
society as a whole.
Political uncertainty is a particularly compelling issue when one considers that the current system
is unsustainable because of anticipated life span and fertility changes. At some point, either
workers will see taxes raised or retirees will see benefits reduced from their current level. Since
much of the political risk revolves around future uncertainty, the transition to a fully funded, pure
individual account system would also have the benefit of forcing the question of who will bear
the costs of reforming the system to be answered immediately. In other words, a pure system of
individual accounts would bring the future funding crisis forward to the present by making
implicit liabilities explicit. Then individuals would no longer be left uncertain of how they will be
affected by reform in the future arising from information already available today. This argument
does not apply to most current proposals, however, because the proposals do not call for pure
individual account systems. They call for mixed systems, and many of these mixed systems do
not increase the system’s prefunding. Thus, proposals that do not address the issue of the system’s
current unfunded liabilities do not really reduce political risk, even if they have created individual
ownership, because they make further, undefined reform in the future necessary.
The risk of credit distortion is another potential political risk present if a centralized trust fund
were to make private investments, but avoidable in a system of individual accounts. If markets
allocate capital among different types of investments efficiently on their own, then, from an
economic perspective, any move towards investment by the government retirement system should
preserve that market allocation or it would lower economic efficiency. A system of individual
accounts would largely preserve the market allocation of capital because private individuals
would continue to choose where they wished to invest their money. By contrast, if the
government were to make private investments centrally through the trust fund, there would be the
possibility that investments would be made on the basis of political criteria rather than the profit
motive. This potential risk seems relatively easy to avoid, however, by stipulating that the trust
fund be invested on the basis of a passive index, similarly to some mutual funds and the TSP.
Of course, political risks are present in any sort of government intervention into the provision of a
fully-funded retirement system. For example, ideally the government may be able to offer
individual account holders actuarially fair annuities on more favorable terms than the private
sector. But were the government to do so, there also would be the possibility that it would offer
actuarially unfair annuities instead for political reasons (e.g., making annuities overly generous to
win political support among the elderly). Such an outcome could create new unfunded liabilities
ȱȱȱ
řŞȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
that could undermine government finances in the long term. Similarly, in theory the government
could equitably spread the risks of a centrally-funded trust fund over time, but it could also distort
risk-spreading formulas in ways favorable to current generations which store up problems for
future generations.
ȱ
The analysis in this report points to two major conclusions. First, it reemphasizes the old saying:
“There is no free lunch.” Someone has to pay to address Social Security’s funding imbalance, and
while the burden can be distributed differentially across individuals and across generations, it
cannot disappear.
Inaction continues uncertainty as to who will pay and when. Tax increases and benefit cuts
provide at least a general picture of who might bear the burden. Some approaches, such as
investing the trust fund (or individual account) savings in higher yielding assets other than
government bonds, have the appearance of being costless, but the benefits to be gained from these
higher yields appear as costs in other parts of the economy: higher interest rates on government
debt or lower returns on private investment (which may include direct reductions in the returns to
private securities in the accounts or trust fund), and changes in risk. Analyses of such approaches
need to account for any declines in private returns to capital, and also address what taxes are to be
increased, or government services to be cut, to pay for the increased government borrowing cost.
None of the current proposals being discussed has provided such detail. Of course, one way to
avoid an increase in interest costs would be to increase national saving but, again, such saving
would mostly likely only be accomplished by some forced government saving plan, and the
policies to accomplish it would need to be spelled out to clarify who pays.
Trying to judge the merits of alternative approaches based on which approach poses a smaller
burden is not really meaningful: all plans face the same problem, that income does not cover cost.
The only difference between plans is who bears the burden. Rather, they may be framed in terms
of losses and gains in economic efficiency. If one rules out options that spread the costs to other
parts of the economy, the choice between higher taxes and lower benefits primarily depends on
the optimal size of the Social Security system. The smaller the system, the less it is able to
alleviate the market failures it was designed to cure—moral hazard, adverse selection, incomplete
insurance markets, and failure of optimization. The larger the system, the more it distorts labor
and saving decisions. The form of tax increases or benefit cuts can also have implications for
economic behavior and distribution. Besides economic efficiency, the decision depends on social
goals and distributional effects that may affect social welfare.
Funding the system can occur under either a centralized system or individual accounts since it is
national saving, not saving in the accounts, that matters. The principal distinction between
maintaining an aggregated system such as we have currently versus a set of individual accounts
involves efficiency gains, and not differential rates of return. Many of the potential criticisms of
individual accounts are mirror images of the rationales for social insurance. If individual accounts
are to fulfill the objectives of social insurance, then they must be very much like an aggregate
system in that they must involve mandatory contributions, mandatory annuitization, and
requirements for prudent investment. But such requirements are in direct contradiction with the
notions of choice and of the complete elimination of distortions induced by taxes; indeed, choice
is itself in conflict with the rationale for social insurance.
ȱȱȱ
řşȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
From this analysis, it follows that if individual accounts meet the minimum requirements for the
objectives of social insurance (that contributions and annuitization are to be mandatory and
investment choice restricted), there are still two limits to individual accounts. First, they could
not, by themselves, fulfill the current system’s social objectives, which include redistribution and
the provision of benefits to survivors, dependents, and the disabled. They could reverse the
distributional objective of the current system, shifting income from the poor to the rich on
average because of the latter’s longer expected life span, and potentially through the tax treatment
of the accounts. This effect could be addressed by subsidizing the poor and taxing the rich (e.g.,
government matching rates for low income individuals), but that revision would move away from
a pure individual account system and its economic benefits. Second, they would introduce
elements of risk, including risk across cohorts reflecting variations in returns and in stock market
valuations. If the goal of reform were to move towards greater pre-funding, an aggregate system
could be more successful at spreading risk among and across generations.
Nevertheless, it is important to also understand the advantages of individual accounts. Individual
accounts do maintain a connection between marginal contribution and marginal benefit that can
reduce distortions in private work and saving behavior (although the mandatory nature of the
contributions may undermine this gain to some extent). Individual accounts also eliminate
political risk, at least within the program itself, which makes the retirement plan itself less risky
in some ways. Finally, individual accounts may impose more budgetary discipline by explicitly
removing any Social Security surplus from government control, so that in reality, they may be
more likely to lead to an increase in national savings than would be the case of the aggregate trust
fund. But for this to occur, individual accounts must be financed through tax increases or benefit
cuts.
Many proposals actually include a mixture of individual accounts and some smaller version of the
current system. Such a mixed system might be used to address some of the problems with pure
individual accounts discussed above, such as their tendency to redistribute from the poor to the
rich and the problem of moral hazard. Such a shift dilutes both the costs and benefits of individual
accounts. In particular, it reduces the return to investment and it leaves a significant portion of the
system as a PAYGO system. But it is the interaction between PAYGO financing and demographic
change that is responsible for the current system’s impending insolvency. There is a fundamental
tension between the objectives of social insurance (to deal with adverse selection and failure of
optimization as well as moral hazard and risk spreading) and the merits of individual accounts
(choice and reduction of other behavioral distortions); any mixed system is giving up some of the
benefits of the first to attain the benefits of the second. Evaluating a plan necessarily involves
assessing those costs and benefits.
It is important to keep this trade-off in mind. For example, proponents of a mixed plan may argue
that their plan does not suffer from moral hazard or undesirable redistribution; nevertheless,
correcting these elements must reduce the rate of return for most individuals, constrain the degree
of choice, and limit reduction in economic distortions that are the rationale for considering
individual accounts in the first place. Thus, mixed plans do not address these problems without a
cost.
ȱȱȱ
ŚŖȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
¡ǯ ȱȱ
The portfolio model is based on a portfolio function for individuals, and a debt-equity trade-off
for firms. Individuals maximize, acting as price takers:
(1) P = (a{ [bB
(1+1/ε)
p (1+1/S)+(1-b)D (1+1/S)](1/(1+1/S)) }(1+1/ε) + (1-a)Ep
)(1/(1+1/ε))
subject to rW = rbBp + rdD + reEp
where Bp is government bonds held by the public, D is corporate debt, and Ep is corporate equities
held by the public, W is a fixed amount of wealth, r is the average return to wealth, rb is the
interest rate on government bonds, rd is the interest rate on corporate debt, and re is the return to
corporate equity. S is the substitution elasticity between bonds and debt, and ε is the substitution
elasticity between equities and the composite of interest bearing assets. In the case of the simple
two asset model, the value of S is set to infinity and the initial interest rates are equated. However,
for a three asset model, the return on government bonds is lower than the return on private bonds.
First order conditions with respect to Bp, D, and Ep lead to the following ratios
(2) Bp/D = [(1-b)/b] S (rb/rd)S
(3) D/Ep = (rd/re)ε [(1-a)/a] ε (1-b)-[(ε+1)S/(S+1)] { [(1+b)/b]S(rb/rd)(S+1)+1} [(ε-S)/(S+1)]
Firms also choose sources of finance (debt versus equity) depending on their cost of debt and
equity:
(4) F = [λD (1-1/σ)+(1-λ)E (1-1/σ)](1/(1-1/σ))
Subject to C = rdD +reE
First order conditions with respect to D and E yield:
(5) (D/E) = (rd/re)-σ [λ/(1-λ)] σ
To complete the model we have the following constraints and identities:
Total equities are:
(6) E = Eg + Ep
where the p subscript refers to equities held by the public and the g subscript to equities held by
the Social Security trust fund. Eg is exogenously determined, and is zero under current rules, but
equal to the total assets in the trust fund in the new equilibrium.
Total government bonds B equal:
(7) B = Bp +Bg
where the p subscript refers to equities held by the public and the g subscript to equities held by
the Social Security trust fund. All are exogenously determined, with Bg equal to the assets in the
Social Security trust fund under current rules, but equal to zero in the new equilibrium.
ȱȱȱ
Śŗȱ
ȱ¢ȱDZȱȱ ȱ
ȱ
(8) K = D + E
The private capital stock, K, is fixed, and is the sum of debt and equity finance, which are
determined endogenously.
Since the private capital stock is fixed, so is the marginal product of capital:
(9) F’(K) K = rdD + reE
Equations (2), (3), (5), (6), (8), and (9) are six equations in six unknowns: rd, rb, re, D, Ep and E.
For purposes of illustration, the model was calibrated for FY2000. Based on data at that time, and
norming output to 1, the private capital stock (K) was 3.35, with 1/3 debt and 2/3 equity, the
government debt held by the public was 0.34 and assets held in the Social Security trust fund
were 0.10. Real interest rates on private debt, equity, and government debt were set at 0.048,
0.088 and 0.017 respectively.
For the two asset model, the model was differentiated and the interest rate set to approximately
0.04, a weighted average of government and private bonds.
ȱȱ ȱ
Jane G. Gravelle
Marc Labonte
Senior Specialist in Economic Policy
Specialist in Macroeconomic Policy
jgravelle@crs.loc.gov, 7-7829
mlabonte@crs.loc.gov, 7-0640
ȱ
The authors gratefully acknowledge the help and comments provided by Geoffrey Kollmann and others in
the development of this report.
ȱȱȱ
ŚŘȱ