Social Security: Issues in Comparing Rates of Return With Market
Investments (Chapter Report, 08/05/1999, GAO/HEHS-99-110).

Pursuant to a congressional request, GAO: (1) examined the estimates of
social security's implicit rates of return for different birth years,
earning levels, household configurations, and other demographic
groupings; (2) examined rates of return available on private market
investments; and (3) discussed the issues that arise from comparing
social security and market investment returns.

GAO noted that: (1) implicit rates of return that workers receive on
their social security contributions vary significantly across a number
of dimensions; (2) the variations mostly reflect several types of income
transfers that the program is designed to provide as part of its social
insurance function; (3) implicit returns vary by birth year, reflecting
the program's income transfers to the first generations of retirees from
subsequent generations; (4) implicit returns that workers receive also
vary on average by their earnings level, by the number of dependents and
survivors, and by their life expectancies; (5) these characteristics
vary by race and gender and therefore rates of return do also; (6) rates
of return on private market assets vary substantially, depending on the
investment risks associated with those assets, particularly the risk of
asset price volatility and the risk of firms defaulting on obligations;
(7) the choice of assets in a portfolio and the timing of investment
decisions ultimately help determine the returns individuals receive and
the risk they bear; (8) a simple comparison between the Social Security
program and market investments would not reflect all the costs
associated with a new system with individual accounts; (9) costs for
both managing and annuitizing the new accounts would reduce actual
retirement incomes and therefore the effective rates of return workers
enjoyed; (10) future rates of return for either market investments or
social security as it is currently structured could differ from their
historic averages; (11) risks differ between the Social Security program
and market investments; (12) instead of making simple comparisons
between social security and historical market returns, one should make
any rate of return comparisons among comprehensive return estimates for
specific reform proposals that include both the individual accounts and
the social security components of the resulting system; (13) such return
estimates would accurately measure the relationship between all the
contributions and benefits implied in each proposal, including both the
social security and individual account components; (14) in particular,
they would reflect the effect of measures taken to ensure the
sustainable solvency of the system; and (15) however, such rate of
return comparisons among reform proposals have some limitations of their
own and address only one of several criteria on which to compare
proposals.

--------------------------- Indexing Terms -----------------------------

 REPORTNUM:  HEHS-99-110
     TITLE:  Social Security: Issues in Comparing Rates of Return With
	     Market Investments
      DATE:  08/05/1999
   SUBJECT:  Social security benefits
	     Projections
	     Comparative analysis
	     Federal social security programs
	     Retirement benefits
	     Investments
IDENTIFIER:  Social Security Program
	     SSA Individual Account Program

******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO report.  Delineations within the text indicating chapter **
** titles, headings, and bullets are preserved.  Major          **
** divisions and subdivisions of the text, such as Chapters,    **
** Sections, and Appendixes, are identified by double and       **
** single lines.  The numbers on the right end of these lines   **
** indicate the position of each of the subsections in the      **
** document outline.  These numbers do NOT correspond with the  **
** page numbers of the printed product.                         **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
** A printed copy of this report may be obtained from the GAO   **
** Document Distribution Center.  For further details, please   **
** send an e-mail message to:                                   **
**                                                              **
**                                            **
**                                                              **
** with the message 'info' in the body.                         **
******************************************************************

Cover
================================================================ COVER

Report to the Chairman, Special Committee on Aging, and to the
Honorable Richard C.  Shelby, U.S.  Senate

August 1999

SOCIAL SECURITY - ISSUES IN
COMPARING RATES OF RETURN WITH
MARKET INVESTMENTS

GAO/HEHS-99-110

Social Security Rates of Return

(207449)

Abbreviations
=============================================================== ABBREV

  AIME - average indexed monthly earnings
  IA - individual accounts
  MB - maintain benefits
  PIA - primary insurance amount
  PSA - personal security accounts
  SSA - Social Security Administration
  TSP - Thrift Savings Plan

Letter
=============================================================== LETTER

B-281321

August 5, 1999

The Honorable Charles E.  Grassley
Chairman
Special Committee on Aging
United States Senate

The Honorable Richard C.  Shelby
United States Senate

As the Congress and the nation have examined how best to restore the
long-term solvency of the Social Security system, many proposals to
restructure the system to include individual accounts have been
offered.  Many who favor individual account proposals point to the
low rates of return that workers can expect from the current system
and the opportunity that individual accounts would offer for
improving rates of return on retirement contributions.  Opponents of
individual accounts have taken exception to the usefulness and
validity of focusing on rates of return.  This report, entitled
Social Security:  Issues in Comparing Rates of Return With Market
Investments, provides a discussion of the key issues to consider in
comparing Social Security and private market rates of return. 

We are sending this report to the Commissioner of Social Security and
relevant congressional committees and subcommittees.  The report will
be available to others on request. 

This report was prepared under my direction.  Please contact Charles
A.  Jeszeck, Assistant Director, at (202) 512-7036 if you have
questions. 

Barbara D.  Bovbjerg
Associate Director, Education, Workforce,
 and Income Security Issues

EXECUTIVE SUMMARY
============================================================ Chapter 0

   PURPOSE
---------------------------------------------------------- Chapter 0:1

Social Security forms the foundation for our retirement income
system, providing crucial benefits to millions of Americans. 
However, the program faces a significant long-term financing
shortage, according to government projections.  In the debate about
how to address this problem, some proposals would restructure Social
Security to include individual retirement savings accounts that would
either supplement or partially replace the current program's
benefits.  According to proponents, such accounts would substantially
improve the rates of return individuals could receive on their
retirement contributions relative to the current system.  The
proponents assert that rates of return under the current system will
be near zero and even negative for many future retirees.  According
to others, however, a new system of individual accounts is not the
only way to raise average rates of return for individuals; investing
some portion of the Social Security trust funds in the stock market
could also help do that.  Moreover, opponents of individual accounts
assert that the rate of return concept should not be applied to
Social Security because it is a social insurance program and should
not be viewed strictly as an investment program.  Still, if rates of
return are considered in weighing Social Security reforms, doing so
raises numerous issues that should be kept in careful perspective. 

In recognition of the role that rate of return comparisons are
playing in the current reform debate, the Senate Special Committee on
Aging and Senator Richard Shelby asked GAO to (1) examine estimates
of Social Security's implicit rates of return for different birth
years, earnings levels, household configurations, and other
demographic groupings; (2) examine rates of return available on
private market investments; and (3) discuss the issues that arise
from comparing Social Security and market investment returns. 

   BACKGROUND
---------------------------------------------------------- Chapter 0:2

In the midst of the Great Depression, Social Security was enacted to
help ensure that the elderly would have adequate retirement incomes
and would not have to depend on welfare.  It would provide benefits
that workers had earned to some degree because of their contributions
and those of their employers, and these benefits would be related to
the earnings on which contributions would be based.  Today, less than
11 percent of the elderly have incomes below the poverty line,
compared with 35 percent in 1959; for about half of the elderly,
incomes excluding Social Security benefits are below the poverty
line.  However, Social Security does not only provide benefits to
retired workers.  In 1939, coverage was extended to their dependents
and survivors, and, in 1956, the Disability Insurance program was
added. 

The current Social Security program is not designed to pay interest
on workers' contributions the way banks pay interest on a savings
account; it is not a system of individual savings accounts.  Rather,
Social Security is financed largely on a "pay-as-you-go" basis, in
which each year's revenues are primarily used to pay that year's
benefits.  Contributions are not deposited in interest-bearing
accounts for individual workers but are instead credited to the
Social Security trust funds.  Under current law, the trust funds must
invest any surplus funds in interest-bearing federal government
securities.  However, the benefit payments to any given individual
are derived from a formula that does not use interest rates or the
amount of contributions but, rather, uses average lifetime
earnings.\1

Still, the benefits workers eventually receive reflect an "implicit"
rate of return they receive on their contributions.  This implicit
rate of return provides one measure of the relationship between
contributions and benefits.  It equals the average interest rate
workers would hypothetically have to earn on their contributions in
order to pay for all the benefits they and their families will
receive from Social Security.  Note that this implicit rate of return
that individuals receive on their contributions is not the same as
the rate of return (or interest rate) that the Social Security trust
funds earn on their assets.  Implicit rates of return for individuals
depend on the relationship between lifetime benefits and
contributions, while the interest earned by the trust funds reflects
the prevailing rate of interest in the market.  In part, implicit
rates of return for individuals depend on the interest earned by the
trust funds but only because it reduces the contribution rates
required to fund benefits.  In addition to depending on trust fund
interest earnings, implicit returns depend on long-term demographic
and economic trends that affect the program's flows of contributions
and benefits. 

To be accurate and consistent, rate of return estimates must reflect
all the contributions and other revenues associated with the benefits
that will eventually be received.  For example, they should reflect
the employers' payroll taxes as well as the employees' taxes.  Also,
given current law and actuarial projections, total revenues will not
be sufficient to fund all the benefits anticipated by 2034.  Rate of
return estimates are misleading if they reflect a long-term imbalance
between revenues and benefits.  In addition, if rate of return
estimates include contributions for survivors and dependents or for
disability benefits, they should also include these benefits.  While
disagreement exists concerning the merits of including nonretirement
benefits in rate of return calculations, estimates should treat
benefits and contributions consistently.  This report only presents
estimates that satisfy these and similar standards of analytical
rigor.  Moreover, actual rates of return vary tremendously by
individual, particularly because life spans vary; some die early and
receive virtually no benefit payments while others live long past the
average life expectancy.  Therefore, rate of return estimates are
used more appropriately for group averages than for individuals. 

--------------------
\1 In technical terms, Social Security provides a defined-benefit
pension, not a defined-contribution pension.  A defined-benefit
pension provides a benefit based on a specific formula generally
linked to each worker's earnings and years of employment.  In
contrast, a defined-contribution pension resembles an individual
savings account; retirement income from this type of pension depends
on the total amount of contributions to the account and any
investment earnings.  As an example, 401(k) accounts are a type of
defined-contribution pension. 

   RESULTS IN BRIEF
---------------------------------------------------------- Chapter 0:3

Implicit rates of return that workers receive on their Social
Security contributions vary significantly across a number of
dimensions.  The variations mostly reflect several types of income
transfers that the program is designed to provide as part of its
social insurance function.  Implicit returns vary by birth year,
reflecting the program's income transfers to the first generations of
retirees from subsequent generations.  For example, the
inflation-adjusted (or "real") implicit rate of return averaged more
than 25 percent annually for the earliest retirees covered by Social
Security and is projected to average roughly 2 percent for baby
boomers, according to a Social Security Administration (SSA) study. 
Implicit returns that workers receive also vary on average by their
earnings level, by the number of their dependents and survivors, and
by their life expectancies.  These characteristics vary by race and
gender and therefore rates of return do also. 

Rates of return on private market assets vary substantially,
depending on the investment risks associated with those assets,
particularly the risk of asset price volatility and the risk of firms
defaulting on obligations.  For example, historical
inflation-adjusted returns on stock market investments, which have
relatively high investment risk, have averaged roughly 7 to 8 percent
over the past 60 to 70 years, compared with roughly 2 to 3 percent
for long-term corporate bonds and roughly 0 to 2 percent for
government securities, which have very low investment risk.  The
choice of assets in a portfolio and the timing of investment
decisions ultimately help determine the returns individuals receive
and the risk they bear. 

A simple comparison between the rates of return for the current
Social Security program and private market investments would be
misleading because of several key issues that such comparisons raise. 
First, a simple comparison between the current Social Security
program and market investments would not reflect all the costs
associated with a new system with individual accounts.  In
particular, the returns individuals would effectively enjoy under a
new system would depend on how the unfunded liabilities of the
current system would be paid off.  Also, costs for both managing and
annuitizing the new accounts would reduce actual retirement incomes
and therefore the effective rates of return workers enjoyed.  Second,
future rates of return for either market investments or Social
Security as it is currently structured could differ from their
historic averages.  Third, risks differ between the current Social
Security program and market investments. 

Instead of making simple comparisons between Social Security and
historical market returns, one should make any rate of return
comparisons among comprehensive return estimates for specific reform
proposals that include both the individual accounts and the Social
Security components of the resulting system.  Such return estimates
would accurately measure the relationship between all the
contributions and benefits implied in each proposal, including both
the Social Security and individual account components.  In
particular, they would reflect the effect of measures taken to ensure
the sustainable solvency of the system.  However, such rate of return
comparisons among reform proposals have some limitations of their own
and address only one of several criteria on which to compare
proposals.  Other criteria include the adequacy and predictability of
benefits, the extent of solvency improvement, and the effect on the
federal budget and national saving. 

   PRINCIPAL FINDINGS
---------------------------------------------------------- Chapter 0:4

      IMPLICIT RATES OF RETURN
      VARY BECAUSE OF SOCIAL
      SECURITY'S INCOME TRANSFERS
-------------------------------------------------------- Chapter 0:4.1

Social Security's implicit rates of return vary significantly by
birth year, earnings level, household composition, and other
demographic characteristics.  Social Security insures workers against
the uncertainties associated with various life events and low
lifetime earnings.  Its income transfers help ensure that
beneficiaries have adequate incomes, and the program has proven
effective in reducing poverty.  For example, Social Security
transfers income to persons who live longer--and therefore need
income longer--from those who do not.  Those who receive such
transfers get higher rates of return than those who do not. 

In the case of variation by birth year, Social Security's average
implicit rates of return have fallen continuously since the beginning
of the program.  According to an SSA study, inflation-adjusted
returns averaged more than 25 percent annually for Social Security's
first retirees in the 1940s and are estimated to average roughly 4
percent for today's retirees, roughly 2 percent for baby boomers, and
1 percent for those who will be born 40 years from now.  (See fig. 
1.) These estimates do not include Social Security disability
contributions and benefits but do reflect tax rates that would
maintain actuarial balance on a pay-as-you-go basis. 

   Figure 1:  Social Security's
   Implicit Rates of Return Are
   Higher for Earlier
   Beneficiaries

   (See figure in printed
   edition.)

Note:  Inflation-adjusted rates, average for all workers in each
birth year.  These estimates do not include Social Security
disability contributions and benefits.  They do reflect tax rates
that would maintain actuarial balance on a pay-as-you-go basis.  They
also reflect employer as well as employee contributions.  This is the
most complete set of estimates by birth year and one of very few that
compute average rates of return for all workers born in a given year. 

Source:  Dean R.  Leimer, Cohort-Specific Measures of Lifetime Net
Social Security Transfers, working paper 59 (Washington, D.C.:  SSA,
Office of Research and Statistics, Feb.  1994). 

This decline in rates of return is primarily a natural and
anticipated consequence of the maturing of a pay-as-you-go system. 
Although both Social Security benefits and contributions have always
been based on earnings, early beneficiaries made contributions over a
smaller portion of their careers.  Also, from 1937 to 1949, Social
Security's tax rates were relatively low at 1 percent of payroll each
for employees and employers, compared with 6.2 percent today.  Higher
rates were not necessary because only a small share of the elderly
had contributed enough to the program to qualify for benefits.  Early
beneficiaries as a group received benefits that were large relative
to their contributions, and therefore the implicit rates of return
they enjoyed were very high.  As the system matured--that is, as each
year passed and another group of people reaching retirement age
qualified for benefits--benefit costs increased.  Tax rates
eventually increased accordingly, and benefits were smaller relative
to contributions.  In effect, the start-up phase provided large
transfers of income to the first generations of retirees from
subsequent generations. 

Now that the system is essentially mature, the lower rates of return
for more recent and future retirees reflect an underlying
relationship between a mature pay-as-you-go system's long-term
average implicit rates and national trends in total wages covered by
the system.  While the declines have been dramatic, future declines
should be small because the returns are now fundamentally tied to the
growth of total wages because both contributions and benefits are
based strictly on earnings. 

In the case of variation by earnings level, Social Security's
implicit rates of return are higher on average for workers with low
lifetime earnings than for those with high earnings.  For example,
for single women born in 1973, SSA projects that inflation-adjusted
implicit rates of return will range from 2.8 percent annually for
women with low earnings to 0.4 percent for those with the maximum
earnings on which Social Security taxes are paid.  This pattern
reflects the way the benefit formula transfers income from high to
low earners. 

Social Security's average implicit rates of return also differ
considerably for workers if their family situations differ.  For
example, for workers with average earnings born in 1973, SSA projects
that inflation-adjusted implicit rates of return will range from 3.7
percent for one-earner couples to 1.3 percent for single men. 
Workers' earnings may generate Social Security benefits for their
spouses and dependents as well as themselves, both while they are
receiving benefits and after they have died.  Because workers do not
make any additional contributions to receive any of these auxiliary
benefits, workers with families that get them receive a higher
implicit rate of return than workers without such families. 

Social Security's average implicit rates of return also vary by
demographic characteristics, such as race and gender, even though
Social Security's benefit and contribution provisions are
structurally neutral with respect to these characteristics.  These
variations in implicit returns arise because different demographic
groups have different average earnings levels, life expectancies, and
household configurations.  Social Security's income transfers are
designed to help ensure adequate incomes for beneficiaries and are
not intended to mitigate any inequalities among various demographic
groups in income or longevity that exist in our society. 

      PRIVATE MARKET RATES OF
      RETURN VARY BY RISK AND
      PORTFOLIO COMPOSITION
-------------------------------------------------------- Chapter 0:4.2

The private market offers a wide variety of investment opportunities
with widely varying rates of return that reflect variations in the
riskiness of those investments.  Portfolio composition and the
performance of the market ultimately determine the return individuals
receive and the risks they bear.  Over the long term, riskier
investments offer higher average rates of return.  The risk from
volatile asset prices can be managed both by holding riskier
investments over longer periods and by managing portfolios so that
such risks tend to offset one another.  While managing a portfolio's
composition in this manner generally requires sophisticated data
analysis and expertise, individual investors can take advantage of
such expertise to some degree by investing in widely diversified
mutual funds. 

Social Security reform proposals that create individual accounts vary
in the degree of latitude that workers would have in managing their
investments.  Some proposals would have the government centrally
manage the accounts and limit the range of investments workers could
choose.  Others would have workers manage their own accounts and
place few restrictions on their investment options.  Such features
would significantly determine the range of returns and risks workers
would face with their market investments and also the costs of
administering their accounts. 

      ISSUES IN COMPARING RATES OF
      RETURN SUGGEST THAT
      COMPARISONS SHOULD BE MADE
      ACROSS REFORM PROPOSALS
-------------------------------------------------------- Chapter 0:4.3

A simple comparison between the rates of return for the current
Social Security program and private market investments would be
misleading because of several key issues that such comparisons raise. 
First, a simple comparison of rates of return for the current Social
Security system and private market investments would be misleading
because it would not capture all the relevant costs that a new system
would imply.  Most significantly, the transition to a new system
would entail costs to pay off unfunded liabilities of the current
system.  The amount necessary to pay the benefits already accrued by
current workers and current beneficiaries is roughly $9 trillion,
according to SSA.  In a pay-as-you-go system, an unfunded liability
always exists and will be covered by future program revenues or
reduced by benefit cuts or both.  However, during the transition to a
restructured system, financing these costs would significantly reduce
returns, and the transition could last for a generation or longer,
depending on how its costs were paid.  In addition, costs for both
managing and annuitizing the new accounts would reduce actual
retirement incomes and, therefore, the effective rates of return
workers enjoyed. 

Second, future average rates of return on either market investments
or Social Security as it is currently structured could differ
significantly from their historical averages, and the gap between
these rates could narrow.  Trends in rates of return on market
investments and Social Security are difficult to predict for many
reasons.  Still, economic growth fundamentally drives rates of return
for both, and projections for either are misleading if they are not
consistent with economic growth projections.  Current SSA projections
suggest that economic growth will be slower in the future than in the
past.  They also suggest that labor will become relatively more
scarce.  In addition, capital may become more plentiful.  Combined,
these trends suggest that market investment returns may be smaller
and that Social Security returns may be relatively larger than they
would be without these trends. 

Third, both the level and type of risk differ between the current
Social Security program and private market investments.  Some of the
risks are market and economic risks that affect rates of return on
either investments or Social Security or both.  Such risks include
the volatility of investment returns and the potential for broad
economic downturns.  Other risks are political, relating to
uncertainties about what changes the Congress might make to either
the current system or a new one.  Estimates of average rates of
return do not measure risk by themselves, and predicting the
statistical variability of those estimates is difficult.  In
particular, rates of return do not measure whether retirement incomes
will be adequate, which is a primary risk that Social Security is
designed to help address.  In addition, they do not measure the
certainty or predictability of retirement incomes. 

In contrast to simple comparisons between the current Social Security
program and market investments, comprehensive rate of return
comparisons among specific reform proposals, with all their
components, address many of the various issues that arise.  Such
comparisons among reform proposals reveal that transition costs would
reduce rates of return to the extent that many participants would not
get significantly higher rates of return than they would under the
current system.  However, such comparisons also show that once the
transition costs are paid off, participants could potentially enjoy
significantly higher returns, depending on market performance and
economic trends.  Comprehensive rate of return comparisons among
reform proposals also capture the effects of administrative and
annuity costs, which would depend to a large extent on the specific
design of the proposals.  Such comparisons can reflect the ways that
portfolio choices and current economic projections might affect
investment earnings. 

Still, comparing rate of return estimates among specific proposals
has some limitations.  Some reform provisions are not easily
incorporated into rate of return estimates.  For example, some
proposals would tap general revenues to help finance the system in
addition to using payroll taxes, but how return estimates could
incorporate such nonpayroll tax revenues is not clear.  Moreover,
average rate of return estimates do not by themselves reveal the
different levels of risk that individuals would face under
alternative reform proposals.  Examining how total retirement incomes
might vary under alternative proposals can suggest to a limited
extent how much risk individuals might face in terms of the adequacy
and predictability of their incomes.  In addition, just as a
trade-off exists between risk and return in market investments, the
same trade-off exists among alternative approaches to Social Security
reform.  Some proposals might offer higher rates of return on Social
Security contributions but might also increase the risk of inadequate
retirement incomes.  Alternatively, provisions that attempt to
mitigate the risk of market investments, such as guarantees, might
create incentives for individuals to take excessive investment risks. 
Such individuals would enjoy any gains from such excessive risk while
the government would incur any losses insured by the guarantees. 
However, any additional costs resulting from such guarantees would
ultimately lower participants' rates of return. 

   AGENCY COMMENTS
---------------------------------------------------------- Chapter 0:5

GAO obtained comments on a draft of this report from SSA.  SSA
generally agreed with GAO's treatment of the issues and offered a
number of technical comments, which were incorporated where
appropriate. 

INTRODUCTION
============================================================ Chapter 1

Social Security forms the foundation for our retirement income
system, providing crucial benefits to millions of Americans. 
However, the program faces a long-term financing shortage, according
to government projections.  In the current Social Security reform
debate, the rates of return workers implicitly receive on their
Social Security contributions have received considerable attention. 

Some proponents of reform assert that for many future retirees, the
inflation-adjusted rates of return on Social Security contributions
will be near zero and even negative for some people.  However, others
believe that the rate of return concept should not be applied to
Social Security because it is a social insurance program and should
not be viewed strictly as an investment program.  Still others view
Social Security as a tax-transfer program, in which taxes should not
be associated with future benefits but simply considered to be
transfers to current beneficiaries, replacing to some degree
transfers workers would have otherwise made--for example, to their
own parents--in the absence of the program.  Nevertheless, if rates
of return are considered in weighing Social Security reforms, they
should be kept in careful perspective. 

When applied to Social Security, the rate of return concept
fundamentally measures the relationship between benefits and
contributions, just as other so-called money's-worth measures do. 
Providing a fair return on contributions is just one of Social
Security's objectives.  In particular, this objective competes to
some degree with the objective of helping ensure adequate incomes,
which Social Security's various income transfers try to achieve. 

Estimating rates of return involves complex actuarial computations
and requires accounting for all contributions and benefits in a
correct and consistent manner.  Many of the relevant factors are
subject to considerable uncertainty, so estimates ideally incorporate
the statistical probabilities associated with the uncertainties of
those factors.  As a result of these uncertainties, actual rates of
return for individuals vary tremendously; hence, rate of return
estimates are used more appropriately for group averages than for
individuals. 

   THE CURRENT SOCIAL SECURITY
   PROGRAM AND THE REFORM DEBATE
---------------------------------------------------------- Chapter 1:1

In the midst of the Great Depression, Social Security was enacted to
help ensure that the elderly would have adequate retirement incomes
and would not have to depend on welfare.  It would provide benefits
that workers had earned to some degree because of their contributions
and those of their employers, and these benefits would be related to
the earnings on which contributions would be based.  However, Social
Security does not only provide benefits to retired workers.  In 1939,
coverage was extended to their dependents and survivors.  In 1956,
the Disability Insurance program was added. 

Profound demographic trends are contributing to Social Security's
long-term financing shortfall.  While 3.3 workers support each Social
Security beneficiary today, only 2 workers are expected to be
supporting each beneficiary by 2030.  This trend reflects increasing
longevity and declining fertility for all future workers, not just
the baby boom generation.  Restoring Social Security's long-term
solvency will require increased revenues, reduced expenditures, or
some combination of both. 

A variety of options are available within the current structure of
the program.\2 However, some proposals would go beyond restoring
long-term solvency and restructure the program to include individual
retirement savings accounts to either supplement or partially replace
the current program's benefits.  In effect, nontax revenues could be
added to the program if the retirement funds could earn a higher rate
of return than Social Security's current funds do.  According to
proponents, a new system of individual accounts would substantially
improve the rates of return individuals can receive on their
retirement contributions.  However, others point out that reforms
within the current structure could also improve rates of return.  For
example, increasing the build-up of the Social Security trust funds
and having the government invest some of those funds in the stock
market would also draw nontax revenues into the program and raise
rates of return.\3

Improving rates of return is just one of many criteria by which to
evaluate alternative reform proposals.\4 It is also important to
examine the effect of reforms on the adequacy of retirement incomes,
in terms of both the level and the certainty of those incomes.  In
addition, reforms should restore solvency in a way that is likely to
be sustained over time.  Moreover, reforms will have effects on the
federal budget and the prospects for economic growth.  Reforms should
also be evaluated for how readily they can be implemented,
administered, and explained to the public.  Finally, reform proposals
should be evaluated as entire packages, weighing all their many
effects together. 

--------------------
\2 See Social Security:  Different Approaches for Addressing Program
Solvency (GAO/HEHS-98-33, July 22, 1998). 

\3 John Geanakoplos, Olivia S.  Mitchell, and Stephen P.  Zeldes,
Would a Privatized Social Security System Really Pay a Higher Rate
of Return? in R.  Douglas Arnold, Michael Graetz, and Alicia H. 
Munnell, eds.  Framing the Social Security Debate (Washington, D.C.: 
Brookings Institution, 1998), pp.  137-56.  This paper makes the
distinction between three distinct types of reform:  privatization,
prefunding, and diversification.  Creating a new system of individual
accounts would achieve all three, but only the last two would be
necessary to improve rates of return.  Privatization would transfer
retirement funds from the government to individuals.  Prefunding
would build up retirement funds in advance, in contrast to the
current system's pay-as-you-go financing structure.  Diversification
would invest those funds in a wider range of market investments than
just government bonds. 

\4 See Social Security:  Criteria for Evaluating Social Security
Reform Proposals (GAO/T-HEHS-99-94, Mar.  25, 1999). 

   IMPLICIT RATES OF RETURN RELATE
   BENEFITS TO CONTRIBUTIONS
---------------------------------------------------------- Chapter 1:2

By design, Social Security contributions are not deposited in
interest-bearing accounts for individual workers but are credited to
the Social Security trust funds, which are primarily used to pay
current benefits.\5 The trust funds are invested in interest-bearing
federal government securities.  However, the benefit payments to any
given individual are derived from a formula that does not use
interest rates or the amount of contributions but rather uses average
lifetime earnings.\6

Even though workers do not earn interest on their contributions as
they would on a savings account, the benefits they receive do reflect
a rate of return they implicitly receive on their contributions. 
This implicit rate equals the interest rate that workers would
hypothetically have to earn on their contributions in order to pay
exactly for all the benefits they and their families will receive
over the course of their lives.\7 This implicit rate of return
provides one measure of the relationship between contributions and
benefits.  It is important to recognize that this implicit rate of
return that individuals receive on their contributions is not the
same as the interest that the Social Security trust funds earn on
their assets.  Implicit rates of return for individuals depend on the
relationship between lifetime benefits and contributions, while the
interest earned by the trust funds reflects the prevailing rate of
interest in the market.  In part, implicit rates of return for
individuals depend on the interest earned by the trust funds but only
because it reduces the contribution rates required to fund benefits. 
In addition to depending on trust fund interest earnings, implicit
returns depend on long-term demographic and economic trends that
affect the program's flows of contributions and benefits.\8

--------------------
\5 The Social Security trust funds are not trust funds in the sense
used in the private sector.  They are primarily used to keep track of
amounts earmarked for a specific purpose.  The Department of the
Treasury has permanent authority to make Social Security benefit
payments as long as there is a fund balance.  As a result, benefit
payments do not require annual appropriations from the Congress.  The
trust funds also provide a contingency reserve to help ensure that
short-term economic downturns do not result in funding shortfalls. 
Currently, the trust fund balances equal about 194 percent of annual
benefit payments. 

\6 In technical terms, Social Security provides a defined-benefit
pension, not a defined-contribution pension.  A defined-benefit
pension provides a benefit based on a specific formula generally
linked to each worker's earnings and years of employment.  In
contrast, a defined-contribution pension resembles an individual
savings account; retirement income from this type of pension depends
on the total amount of contributions to the account and any
investment earnings.  As an example, 401(k) accounts are a type of
defined-contribution pension. 

\7 A more technically precise definition of the rate of return for
Social Security contributions would be the constant discount rate
that equates the present discounted value of contributions with the
present discounted value of benefits. 

\8 Rates of return are most useful when they are adjusted for
inflation to reveal how much the purchasing power of an invested sum
of money has increased.  For example, the yield on a 3-month Treasury
bill in 1981, not adjusted for inflation, was 14.0 percent, but
inflation was 10.3 percent.  Adjusted for inflation, the yield was
3.4 percent.  In contrast, the same yield in 1986, not adjusted, was
much lower at 6.0 percent although inflation was only 1.9 percent. 
Adjusted for inflation, the yield in 1986 was 4.0 percent, higher
than in 1981. 

      IMPLICIT RATES OF RETURN ARE
      ONE TYPE OF "MONEY'S-WORTH"
      MEASURE
-------------------------------------------------------- Chapter 1:2.1

Implicit rates of return are one type of so-called money's-worth
measure and measure the "individual equity" of the program--that is,
how benefits compare with contributions.  Such measures also reflect
how well benefits compare with the income workers would have if they
could keep their contributions and invest them elsewhere. 

Other measures of Social Security's money's-worth include payback
periods, lifetime benefit/tax ratios, and the dollar value of net
lifetime transfers.  Such measures begin with an interest rate
workers could earn on their contributions.  The payback period is how
long it takes their benefits to pay back their contributions plus
interest.  Benefit/tax ratios, or money's-worth ratios, compare the
interest-adjusted value of lifetime benefits with lifetime
contributions.  In general, these alternative measures yield
conclusions similar to the rate of return, but none gives a complete
picture.  For example, early Social Security beneficiaries enjoyed
very high rates of return.  However, those returns were on very small
contributions, so the absolute dollar value of the income transfer
they received was relatively small. 

Money's-worth calculations measure only individual equity, which is
just one of Social Security's objectives.  The program's insurance
features inherently place greater emphasis on helping ensure that
beneficiaries have adequate income; without its built-in income
transfers across and within cohorts, Social Security would provide
identical rates of return on contributions.  In contrast, measures of
"income adequacy" include how total retirement income, including
benefits, compares with the poverty line.  Today, less than 11
percent of the elderly have incomes below the poverty line, compared
with 35 percent in 1959.  For about half of the elderly, incomes
excluding Social Security benefits are below the poverty line.  Also,
"replacement rates," which equal the initial annual benefit amount
divided by the earnings in the worker's last year of work, show how
well benefits compare with or "replace" preretirement income.  For
example, workers who retired in 1999 at age 65 with a history of
average earnings had a replacement rate of 40 percent and an annual
benefit of $11,454. 

      CONSIDERATIONS RELATING TO
      WHICH BENEFITS AND
      CONTRIBUTIONS ARE INCLUDED
      IN RATE OF RETURN
      CALCULATIONS
-------------------------------------------------------- Chapter 1:2.2

Because rates of return show the relationship between benefits and
contributions, rate of return calculations depend critically on which
benefits and contributions are included.  To be consistent,
calculations must carefully include all the benefits associated with
any of the contributions that are included, and vice versa.  In
particular, considerations in properly accounting for all benefits
and contributions include the treatment of (1) inflation adjustment
of benefits, (2) employers' contributions, (3) any actuarial
imbalance in the system, and (4) nonretirement benefits. 

         INFLATION ADJUSTMENT OF
         BENEFITS
------------------------------------------------------ Chapter 1:2.2.1

Rate of return estimates should reflect the automatic annual
inflation adjustment of Social Security benefits, which is a
significant part of the benefit package that the payroll tax
finances. 

         EMPLOYERS' CONTRIBUTIONS
------------------------------------------------------ Chapter 1:2.2.2

Including the employers' share of the payroll tax has a significant
effect on rate of return calculations since it is half of all payroll
taxes.  Currently, both the individual and the employer pay a
6.2-percent tax on covered earnings for retirement, survivors, and
disability benefits combined.\9 Although a few studies use only the
workers' contributions to calculate rates of return, most studies use
both the employers' and employees' contributions.  Most analysts
agree that employees ultimately pay the employers' share because
employers pay lower wages than they would if the employers'
contribution did not exist.  Furthermore, estimates that leave out
employers' contributions reflect the full benefits but not the full
costs of providing those benefits. 

--------------------
\9 Self-employed workers pay a contribution rate of 12.4 percent,
half of which is tax deductible as a business expense. 

         REFLECTING AN ACTUARIALLY
         BALANCED SYSTEM
------------------------------------------------------ Chapter 1:2.2.3

Rate of return calculations that include only contributions and
benefits as defined under current law are misleading, because the
system is not in actuarial balance.  The returns that workers
actually receive will be different from any returns estimated using
current contribution and benefit levels, depending on how the
financing shortfall is addressed.  One approach to resolving this is
to use contribution levels that would restore actuarial balance on a
pay-as-you-go basis--that is, raising tax rates as the funds are
needed to pay benefits.  Another is to use reduced benefit levels
that would require no tax increases.  According to 1996 estimates by
the Social Security actuaries, for example, two-earner couples born
in 1973 with average earnings would receive an inflation-adjusted
return of 1.9 percent under the tax-increase approach but would
receive a return of 1.7 percent under the benefit-cut approach.  In
contrast, their estimated rate of return would be 2.1 percent using
contributions and benefits from the current, imbalanced system.\10

--------------------
\10 Advisory Council on Social Security, Report of the 1994-1996
Advisory Council on Social Security, Vol.  1 (Washington, D.C.:  Jan. 
1997), p.  222. 

         NONRETIREMENT BENEFITS
------------------------------------------------------ Chapter 1:2.2.4

Rate of return calculations must be clear and consistent about
whether benefits and contributions are included for survivors,
dependents, and disabled workers as well as for retired workers.\11
If rate of return calculations included the full range of benefits
provided by the Social Security program rather than retirement
benefits alone, the calculations would also need to include the full
range of contributions made for those benefits.  Conversely, if the
calculations included only the retirement portion of the benefits,
then the contributions would need to be reduced accordingly. 

Although disagreement exists about whether return estimates should
include survivors and disability benefits, either approach can
theoretically produce reasonable estimates as long as the
contributions and benefits used are comparable.  Analysts who prefer
to exclude survivors and disability benefits from the computations
believe that these benefits are distinct and separable from
retirement benefits and that they are more like true insurance. 
Death and disability can strike at any time, but workers can plan and
save for retirement over a known period.  So providing for retirement
is an issue more of saving than insurance, according to this view. 
In contrast, analysts preferring return estimates for the whole
program point out that returns can vary significantly across
retirement, survivors, and disability benefits for various groups of
beneficiaries.  For example, focusing only on retired worker rates of
return for African-Americans, who have shorter life expectancies than
whites, ignores that African-Americans are considerably more likely
to receive disability and survivors benefits.\12

--------------------
\11 In 1996, retired workers accounted for 61 percent of all Social
Security beneficiaries, and they received 68 percent of the benefits. 

\12 In addition, the appropriate contribution rate to attribute to
Disability Insurance is not as clear-cut as it may seem.  Even though
a distinct contribution rate exists for Disability Insurance, the
Congress has occasionally adjusted the rates to manage the financial
balances of the separate funds.  The problems are more complicated
for survivors and dependents benefits under the Old-Age and Survivors
Insurance program, which does not have a separate contribution rate
for each type of benefit. 

      SEVERAL FACTORS ARE SUBJECT
      TO UNCERTAINTY
-------------------------------------------------------- Chapter 1:2.3

Several factors that affect rate of return calculations are subject
to uncertainty, which makes projections complex and subject to error. 
Contributions depend on each worker's earnings level and the tax
rate.  Workers' benefits depend on various uncertain life events,
such as when they retire, become disabled, or die; whether they have
spouses or dependents who are eligible for benefits; and how long
benefits are paid.  Their benefits also depend on their lifetime
earnings histories.  Further, their benefits depend on national
trends in wage and price levels.  Both benefits and contributions
depend on any changes in the law that the Congress may make.  Rate of
return estimates can reflect averages relating to these uncertainties
across large groups, such as all average-income workers born in a
given year.  However, any projections for individual workers would
prove to be misleading for many of them because their actual
experience can vary so much from the average.\13

To account for various uncertainties, accurate rate of return
estimates require complex actuarial calculations.  The most rigorous
calculations produce an estimate of what workers can expect to
receive from the time they start paying taxes.  "Expected values"
describe the average return for all possible outcomes, weighted for
the probability of each outcome.  In the perfect case, this would
involve projecting statistical probabilities for each life event,
including disability or death at each age, age at retirement,
earnings in each year, marital status, number of children, and so on. 
The return calculations would then use these probabilities in the
weighted average of all the benefits received under each of the many
different possible scenarios.  In contrast, less rigorous
calculations estimate a rate of return for a small number of specific
illustrative outcomes, such as having a low, average, or high level
of lifetime earnings and being single or in a one-earner or
two-earner couple.  Such "hypothetical worker" calculations are by
far the most common type of rate of return estimate.  However, while
they can be accurate for such specific cases and can be useful for
making comparisons across types of individuals, they do not and
cannot represent what all workers in a particular group can expect to
receive on average.  Still, even these simpler, hypothetical worker
calculations require proper actuarial methods. 

--------------------
\13 See SSA Benefit Estimate Statement:  Adding Rate of Return
Information May Not Be Appropriate (GAO/HEHS-98-228, Sept.  2, 1998). 

   OBJECTIVES, SCOPE, AND
   METHODOLOGY
---------------------------------------------------------- Chapter 1:3

In recognition of the role that rate of return comparisons are
playing in the current reform debate, the Senate Special Committee on
Aging and Senator Richard Shelby asked us to (1) examine estimates of
Social Security's implicit rates of return for different birth years,
earnings levels, household configurations, and other demographic
groupings; (2) examine rates of return available on private market
investments; and (3) discuss the issues that arise from comparing
Social Security and market investment returns.  To answer these
questions, we conducted an extensive review of the growing literature
on the subject and interviewed experts familiar with the estimates
available.  Many estimates of Social Security's rate of return have
been made.  Analysts generally agree on which approaches for
calculating returns are the most rigorous and which are flawed.  We
examined estimates from several studies and found that the most
rigorous ones produced generally consistent estimates.  For example,
estimates made by Social Security actuaries for the Report of the
1994-1996 Advisory Council on Social Security were among the most
rigorous.  In this report, we present only estimates that meet a
rigorous standard, and we note any limitations or qualifications.  In
particular, we present only estimates that include employers' as well
as employees' contributions and that reflect an actuarially balanced
system.  We conducted our work between January 1998 and June 1999 in
accordance with generally accepted government auditing standards. 

IMPLICIT RATES OF RETURN VARY
BECAUSE OF SOCIAL SECURITY'S
INCOME TRANSFERS
============================================================ Chapter 2

Social Security's implicit rates of return vary significantly by
birth year, earnings level, household composition, and other
demographic characteristics.  These variations reflect several types
of income transfers that the program provides as part of its social
insurance function.  Social Security insures workers against the
uncertainties associated with various life events and low lifetime
earnings.  In effect, any type of insurance transfers income to
persons who incur losses from those who do not.  Similarly, Social
Security transfers income, for example, to persons who live
longer--and therefore need income longer--from those who do not. 
Persons who receive such transfers get higher rates of return than
those who do not.\14

--------------------
\14 This chapter summarizes only the basic dimensions by which rates
of return vary:  birth year, earnings level, and household
composition.  In addition, interactions exist among these dimensions
that present a more complicated picture.  For example, a high-earning
one-earner couple earns lower returns than average by virtue of its
earnings level but higher returns by virtue of its household
composition.  Rate of return estimates can reveal the net effect for
each particular combination of characteristics.  However, as noted in
chapter 1, rates of return by themselves do not provide a complete
picture even then.  The dollar value of the income transfer can be
relatively low even when the rate of return is relatively high.  For
a more complete set of rates of return and money's-worth measures,
see Advisory Council on Social Security, Report of the 1994-1996
Advisory Council on Social Security, Vol.  1, pp.  165-230. 

   VARIATION BY BIRTH YEAR
   REFLECTS MATURING OF SYSTEM AND
   WAGE GROWTH
---------------------------------------------------------- Chapter 2:1

Social Security's implicit rates of return have fallen continuously
since the beginning of the program.  This decline is primarily a
natural and anticipated consequence of the maturing of a
pay-as-you-go system, in which each year's revenues are primarily
used to pay that year's benefits.  When a pay-as-you-go system is
started, rates of return are high for earlier retirees because they
receive large transfers of income from subsequent generations.  While
the declines were dramatic initially, they have been much smaller as
the system has approached maturity.  (See fig.  2.1.)

   Figure 2.1:  Social Security's
   Implicit Rates of Return Are
   Higher for Earlier
   Beneficiaries

   (See figure in printed
   edition.)

Note:  Inflation-adjusted rates, average for all workers in each
birth year.  These estimates do not include Social Security
disability contributions and benefits.  They do reflect tax rates
that would keep the system in actuarial balance on a pay-as-you-go
basis.  They use the intermediate assumptions of the 1991 Social
Security Trustees' Report.  This is the most complete set of
estimates by birth year and one of very few that compute average
rates of return for all workers born in a given year. 

Source:  Dean R.  Leimer, Cohort-Specific Measures of Lifetime Net
Social Security Transfers, working paper 59 (Washington, D.C.:  SSA,
Office of Research and Statistics, Feb.  1994). 

Figure 2.1 illustrates inflation-adjusted average rates of return for
all workers born in given years--that is, for "birth groups." These
estimates include all Social Security benefits and contributions
except disability, and they assume that payroll tax rates will
increase on a pay-as-you-go basis to keep the system actuarially
balanced.  Inflation-adjusted rate of return estimates were more than
25 percent per year for birth groups born in 1880 or earlier. 
However, these returns were on relatively small contributions, so the
dollar value of the income transfer they received was relatively
small.  Rate of return estimates were more than 10 percent for birth
groups born before 1905.  They fell below 6 percent for those born in
1920, below 3 percent for those born in about 1940, and below 2
percent for those born in about 1960.  They will reach 1 percent for
those who will be born in about 2040.\15

Implicit rates of return declined for successive groups of workers
during the maturing phase of Social Security's history.  From 1937 to
1949, Social Security's tax rates were a relatively low 1 percent of
payroll each for employees and employers, compared with 6.2 percent
today.  Higher rates were not necessary because only a small share of
the elderly had contributed enough to the program to qualify for
benefits.\16 In addition, early beneficiaries made contributions over
fewer years in covered employment than later beneficiaries.  As a
result, the benefits they received were very high relative to these
smaller contributions, and the implicit rates of return they enjoyed
were very high.  As the system matured--that is, as each year passed
and another group of people reaching retirement age qualified for
benefits--benefit costs increased.  Tax rates eventually increased
accordingly, newer beneficiaries had made contributions over more
years, and benefits became smaller relative to those
contributions.\17

In designing Social Security, the Congress chose a pay-as-you-go
system rather than an "advance-funded" one in which tax levels are
high enough to finance future benefit promises.\18 In effect, the
Congress provided large income transfers to early generations of
beneficiaries.  This decision reflected concern that the government
might amass huge reserve funds and the prospect that this could
weaken the economy.  It also reflected a concern about helping
improve retirement incomes much sooner than an advance-funded system
would have done.  While these early beneficiaries may have received a
substantial income transfer within the Social Security system, as a
group they contributed substantial amounts outside the system to the
retirement incomes of their parents' generation, which did not
qualify for Social Security benefits.  Such contributions included
not only income support that some provided to their own parents but
also taxes and charitable contributions that paid for other forms of
support. 

In a fully mature pay-as-you-go system, long-term average implicit
returns roughly equal the growth of total wages covered by the system
because both contributions and benefits are based directly on covered
wages.\19 In turn, total wage growth depends significantly on the
growth of labor productivity and the growth of the labor force. 
Since both of these growth rates have slowed in recent years and are
projected to remain low, implicit Social Security returns have been
declining even though the system is now essentially mature.  However,
as long as total wage growth remains positive, long-term average
returns on Social Security for birth groups will also generally
remain positive.\20 This remains true over the long-term even with
increasing longevity and a declining ratio of workers to
beneficiaries.  The estimates in figure 2.1 take these projected
demographic changes into account and also reflect tax rates that
would keep the system in actuarial balance.  Under this scenario, tax
rates would increase but so would lifetime benefits as people live
longer. 

--------------------
\15 Dean R.  Leimer, Cohort-Specific Measures of Lifetime Net Social
Security Transfers, working paper 59 (Washington, D.C.:  SSA, Office
of Research and Statistics, Feb.  1994). 

\16 In addition, the maximum annual earnings subject to the payroll
tax were only $3,000 in 1937.  However, in 1937, 97 percent of all
covered workers had total earnings below $3,000, while today about 94
percent have total earnings below the taxable maximum.  Still, the
percentage of workers with total wages under this ceiling was much
lower from about 1950 to 1978, when this percentage ranged between 64
and 85 percent.  So this pattern of relatively lower contributions
also contributed to higher rates of return for those who paid taxes
during this period. 

\17 Technically speaking, more than one generation of retirees
benefited from the transfers that resulted from starting a new
system.  The Congress increased Social Security benefit levels many
times over several years and expanded coverage to new sets of
workers.  Each time benefits are added or coverage is expanded, those
incremental changes begin a new maturing process of their own, which
extends the maturing process for the system as a whole. 
Nevertheless, the current system can now be considered to be
essentially mature since any remaining part of that process is
relatively small. 

\18 Social Security actually began in 1935 as a partially funded
pension plan; however, the 1939 amendments modified it to more of a
pay-as-you-go pension plan. 

\19 While this fundamental relationship between Social Security's
rate of return and wage growth may not be immediately obvious, the
academic literature has shown it to be true.  In short, if
demographic and economic conditions and program provisions were all
constant in a mature pay-as-you-go system, then benefits for one
generation of retirees would equal the contributions paid by its
children's generation.  Those contributions would equal the retirees'
contributions plus wage growth, since contributions are based on
wages.  If any of the constants were to change, program provisions
would have to change to restore balance.  Once balance were restored
and all factors became constant again, this relationship between
contributions and benefits would be restored.  Of course, in the real
world, returns vary within these long-term averages because
contribution and benefit patterns can vary somewhat and still reflect
long-term actuarial balance.  For example, the Congress can increase
taxes or cut benefits to achieve actuarial balance, but such policy
changes can affect those born earlier more than those born later or
vice versa and still achieve the same level of long-term balance. 

\20 This relationship between returns and wage growth helps explain
why Social Security is not a "Ponzi" or pyramid scheme.  It is
mathematically impossible for a pyramid scheme to continue
indefinitely.  As layers are added at the bottom of the pyramid, the
number of participants required grows exponentially and eventually
there would never be enough people to complete a full layer. 
However, under Social Security, positive rates of return on average
can exist indefinitely as long as total wage growth remains positive. 

   VARIATION BY EARNINGS LEVEL
   REFLECTS INCOME REDISTRIBUTION
---------------------------------------------------------- Chapter 2:2

Social Security's implicit rates of return are higher on average for
workers with low lifetime earnings than for those with high earnings. 
(See fig.  2.2.) This pattern reflects the way the benefit formula
transfers income from high to low earners.\21 To help ensure that
beneficiaries with low lifetime earnings have adequate incomes, the
benefit formula was designed to be progressive and replace a higher
percentage of average lifetime earnings for low earners than for high
earners.\22

   Figure 2.2:  Social Security's
   Implicit Rates of Return Are
   Higher for Low Than for High
   Earners

   (See figure in printed
   edition.)

Note:  Inflation-adjusted rates, single women born in 1973.  These
estimates include all Social Security contributions and benefits,
including disability, and reflect tax rates that would keep the
system in actuarial balance on a pay-as-you-go basis.  These
estimates do not reflect the fact that life expectancy is lower for
lower earners.  These estimates are for hypothetical workers to
illustrate differences across earnings levels.  Each earnings level
estimated represents one earnings amount in each year; the estimates
do not represent ranges of earnings.  The average earnings level
equals the average Social Security covered earnings in each year, the
low earnings level equals 45 percent of the average, and the high
level equals 160 percent of the average.  The maximum taxable
earnings level reflects an earnings history in which the workers'
earnings equaled the maximum taxable level in each year.  In 1998,
the average earnings level was about $29,000, implying a low earnings
level of roughly $13,000 and a high level of roughly $46,000.  The
maximum taxable earnings level was $68,400.  Returns for single men
were roughly 0.5 percentage points lower at each earnings level. 

Source:  SSA. 

However, some analysts have noted that lower earners have lower life
expectancies on average, which reduces their rates of return.\23
Various sets of estimates have attempted to demonstrate the size of
the effect on implicit returns from life expectancy differences
across income groups.  While some of these estimates have been
flawed, rigorous and reasonably accurate estimates have shown that
the life expectancy differences between income groups do lower rates
of return for low earners and increase them for high earners. 
However, these effects are not large enough to reverse the overall
progressivity of the benefit structure.  For example, for workers
born between 1917 and 1922, one study estimated that adjusting life
expectancy for income differences would decrease the average annual,
inflation-adjusted, implicit returns for low-wage men from 6.23 to
6.17 percent and would increase such returns for high-wage men from
4.99 to 5.04 percent.\24

These estimates did not include disability benefits or contributions. 

--------------------
\21 The estimates for figures 2.2 and 2.3 are for illustrative,
hypothetical workers.  However, a recent study raises questions about
whether the low and average earnings levels reflect earnings that
are truly low and average.  As a result, rates of return for truly
low and average earnings levels would actually be somewhat higher
than these estimates suggest.  Earnings records for hypothetical
workers are assumed to follow a steady, smooth lifetime earnings
pattern.  In reality, earnings patterns vary considerably, and many
workers have some years of zero earnings.  Those zero earnings in
particular years are not reflected in the average earnings level used
for the hypothetical worker cases, but they can affect the Social
Security benefit calculation.  As a result, the study found that the
hypothetical low earnings level of $13,000 actually falls between
the low and average earnings level.  Similarly, the hypothetical
average earnings level of $29,000 actually falls between the
average and high levels.  Because rates of return are lower for
higher earnings, return estimates for these hypothetical earnings
levels may be misleadingly low.  Nevertheless, they do illustrate the
general pattern by earnings level.  See Gary Burtless, Barry
Bosworth, and C.  Eugene Steuerle.  Changing Patterns of Lifetime
Earnings:  What Do They Tell Us About Winners and Losers From
Privatization? Paper presented at the First Annual Joint Conference
for the Retirement Research Consortium, New Developments in
Retirement Research, Boston College Center for Retirement Research
and Michigan Retirement Research Center, Washington, D.C., May 20-21,
1999. 

\22 Specifically, the primary insurance amount (PIA) is the full
monthly benefit payable to retired workers at age 65 or to disabled
workers when first eligible.  Retired workers are first eligible for
benefits at age 62 but the monthly benefit is reduced for each month
they receive benefits before age 65.  For those first eligible for
benefits in 1998, the PIA equaled (1) 90 percent of the first $477 of
average indexed monthly earnings (AIME) plus (2) 32 percent of the
next $2,398 of AIME plus (3) 15 percent of AIME over $2,875.  The
bend points in this formula (dollar amounts of AIME defining each
bracket) are indexed to increases in average national earnings. 

\23 Also, lower earners tend to enter the workforce earlier than
higher earners, who tend to have more years in school.  Therefore,
lower earners are likely to have more years of nonzero earnings,
which diminishes the benefit formula's progressivity.  However, lower
earners more commonly have interrupted work histories or work outside
of covered employment, which strengthens progressivity. 

\24 James E.  Duggan, Robert Gillingham, and John S.  Greenlees,
Progressive Returns to Social Security?  An Answer from Social
Security Records, research paper 9501 (Washington, D.C.:  U.S. 
Treasury Department, Nov.  1995), p.  14. 

   VARIATION BY HOUSEHOLD TYPE
   REFLECTS THE ROLE OF
   DEPENDENTS' BENEFITS
---------------------------------------------------------- Chapter 2:3

Social Security's implicit rates of return also vary considerably for
workers if their family situations differ.  Workers' earnings
generate Social Security benefits for themselves and may also
generate benefits for their spouses and dependents, both while they
are receiving benefits and after they have died.\25 Because workers
do not make any additional contributions for any of these auxiliary
benefits, workers with families that get them receive a higher
implicit rate of return than workers without such families.  Also,
one-earner and two-earner couples both receive some combination of
retired worker and spouse benefits, but the two-earner couples make
contributions based on two earnings records instead of one.\26 As a
result, one-earner couples receive significantly higher implicit
rates of return than two-earner couples or single earners.  (See fig. 
2.3.) For these estimates, the hypothetical one-earner couples are
those in which one spouse works steadily until retirement while the
other does not work at all.  In reality, a couple could have the
second spouse work and make Social Security contributions for some
number of years; if that spouse's average lifetime earnings were low
enough, the couple might still receive the same benefit as the
hypothetical one-earner couple.  Such a couple would have a lower
implicit rate of return than the one-earner case in which the second
spouse makes no contributions.  The hypothetical one-earner case
illustrates only one relatively extreme scenario.\27

   Figure 2.3:  Social Security's
   Implicit Rates of Return Are
   Higher for One-Earner Couples

   (See figure in printed
   edition.)

Note:  Inflation-adjusted rates, average earners born in 1973.  These
estimates include all Social Security contributions and benefits,
including disability, and reflect tax rates that would keep the
system in actuarial balance on a pay-as-you-go basis.  These
estimates are for hypothetical workers with earnings equal to the
national average each year; for the one-earner couple, one spouse
does not work at all.  In 1998, the average earnings level was about
$29,000.  The estimates illustrate differences across household types
but they are not averages for all workers in each type.  In addition,
they do not reflect any differences in average income that may exist
across these groups. 

Source:  SSA. 

These patterns reflect that Social Security is designed to provide
income transfers from families without dependents to those with them. 
In particular, Social Security's benefit provisions for spouses have
the effect of subsidizing or in some way recognizing the work efforts
of spouses who do not work and earn income outside the home. 
However, women are increasingly participating in the labor force for
greater proportions of their working years, so the role of spousal
benefits may be declining in importance for such women but would
still be significant for those who do not work outside the home. 

--------------------
\25 Social Security also pays benefits to divorced spouses.  However,
most divorced women do not qualify for divorced spouse benefits
because most marriages that end in divorce last less than 10 years,
the minimum marriage duration needed to qualify for such benefits. 
In addition, many divorced women who were married at least 10 years
do not receive divorced spouse benefits because they either
subsequently remarry or have retired worker benefits that exceed
their benefit as a divorced spouse. 

\26 The spouses with the lower earnings are eligible to receive
spouse benefits based on their spouse's earnings record as well as
retired worker benefits based on their own earnings, but they cannot
receive both full benefits simultaneously.  Essentially, these
beneficiaries, who are called dually entitled, receive their own
retired worker benefit and the difference between that and the spouse
benefit if it is higher.  Spouse benefits equal 50 percent of the
worker's benefit, which may be higher than the spouse's own retired
worker benefit if the difference in their average lifetime earnings
is large enough. 

\27 In addition, the hypothetical couple does not capture the effect
of the age difference between spouses; it assumes that spouses are
the same age and have two children born when the spouses are in their
mid-20s.  Couples with large age differences may get higher rates of
return than those with no age difference because, on average, they
may receive benefits for longer periods. 

   VARIATION BY DEMOGRAPHIC GROUP
   REFLECTS THE PROGRAM'S SOCIAL
   INSURANCE ROLE
---------------------------------------------------------- Chapter 2:4

Social Security's implicit rates of return also vary by demographic
characteristics, such as race and gender, even though Social
Security's benefit and contribution provisions are structurally
neutral with respect to these characteristics.  Rather, these
variations in implicit returns arise because such demographic groups
have different average earnings levels, life expectancies, and
household configurations.  These factors significantly affect rates
of return as a result of Social Security's insurance role and income
transfers.  Its income transfers are designed to help ensure adequate
incomes for beneficiaries and are not intended to mitigate any
inequalities in income or longevity that exist in our society among
various racial, ethnic, or gender groups. 

For example, figure 2.3 illustrates the difference in returns for
hypothetical single men and women both with the same earnings equal
to the national average earnings in each year.  The difference in
implicit returns between single men and women reflects the greater
life expectancies of women.  At age 65, women today have a life
expectancy of about 19 additional years, compared with 16 years for
men.  However, note that women have lower incomes on average than
men, which the estimates in figure 2.3 do not reflect; these
estimates are for illustrative households in which all workers have
equal earnings.  Estimates of average implicit returns for all
workers born in the same year would show that the difference between
single men and women would be even greater because of the difference
in average income. 

With respect to race differences, nonwhites tend to have lower
incomes than whites, which tends to increase the implicit returns of
nonwhites.  However, African-Americans tend to have shorter life
expectancies than whites, which tends to decrease their implicit
returns.  Still, African-Americans are relatively more likely to be
disabled, die before retirement, and have dependents than whites.\28
As a result, implicit rates of return are probably higher for
African-Americans if the full range of Social Security benefits is
included than if only retirement benefits are included.  However,
none of the currently available rate of return studies that examine
race differences have included disability benefits.  Still, rigorous
and accurate return estimates that do not include disability benefits
generally show that both African-Americans and other nonwhites have
higher average implicit rates of return from Social Security than
whites.\29

--------------------
\28 For example, while African-Americans make up 12 percent of the
nation's population, they make up only 8 percent of Social Security
retirement beneficiaries.  However, they make up 18 percent of
disabled beneficiaries and 23 percent of child beneficiaries.  Also,
nearly half of all African-American beneficiaries receive disability
or survivor benefits compared with 28 percent of white beneficiaries. 

\29 James E.  Duggan, Robert Gillingham, and John S.  Greenlees,
"Returns Paid to Early Social Security Cohorts," Contemporary Policy
Issues, Vol.  11 (Oct.  1993), pp.  1-13; Charles Meyer and Nancy
Wolff, "Intercohort and Intracohort Redistribution under Old Age
Insurance," Public Finance Quarterly, Vol.  15, No.  3 (July 1987),
pp.  259-81.  For a review of the literature on this point, see Dean
R.  Leimer, "Guide to Social Security Money's Worth Issues," Social
Security Bulletin, Vol.  58, No.  2 (summer 1995), p.  13. 

PRIVATE MARKET RATES OF RETURN
VARY BY RISK AND PORTFOLIO
COMPOSITION
============================================================ Chapter 3

The private market offers a variety of investment vehicles with
widely varying rates of return, reflecting differences in the degree
of risk associated with those investments.  Portfolio composition and
the performance of the market ultimately determine the returns
individuals receive.  Social Security reform proposals that would
create individual accounts vary in the degree of latitude that
workers would have in choosing their investments, and the returns
they would potentially enjoy would depend on such provisions. 

   RISKIER INVESTMENTS GENERALLY
   YIELD HIGHER LONG-TERM AVERAGE
   RETURNS
---------------------------------------------------------- Chapter 3:1

Over long periods of time, riskier investments generally yield higher
average rates of return.  Over the past 60 to 70 years, returns on
low-risk government securities have been lower over the long term
than private securities, with a compound annual average return of
roughly 0 to 2 percent per year on an inflation-adjusted basis before
personal income taxes.\30 In contrast, compound annual returns on
stocks in Standard & Poor's composite stock index have averaged
roughly 7 to 8 percent per year on an inflation-adjusted basis.  On
long-term corporate bonds, inflation-adjusted annual returns have
averaged roughly 2 to 3 percent. 

Two specific types of risk are particularly relevant to returns on
market investments as they might relate to individual accounts. 
"Default" or "credit" risk is the risk of borrowers defaulting on
their obligations, such as bonds.  Bond-rating firms grade borrowers
on the risk of default.  Highly graded bonds--that is, bonds with low
default risk--have consistently been sold at lower interest rates. 

In contrast, "market" risk relates to the volatility of the price of
broad groups of assets, such as stocks, bonds, and other types of
investments.  The volatility of asset prices is reflected in the
volatility of the rates of return on those assets.  For example,
annual returns on a broad portfolio of stock investments are more
volatile than returns on government bonds.  On a long-term average
basis, the market compensates for this greater market risk by
offering higher average returns on riskier investments.  For example,
the year-to-year variation in rates of return is much greater for
stocks than for government securities, and their long-term compound
average annual rate of return is higher--roughly 7 to 8 percent per
year compared with roughly 0 to 2 percent per year on government
securities. 

--------------------
\30 Compound average annual rates of return reflect the total return
on an investment over a number of years, figured on a constant annual
basis; this is not the same as the arithmetic average of rates for
each year. 

   PORTFOLIO STRATEGIES CAN MANAGE
   RISK
---------------------------------------------------------- Chapter 3:2

Investors can manage the riskiness of their portfolios by both how
long they hold specific investments and how they compose their
portfolios.  Historical data suggest that over long periods of time,
riskier investments have quite reliably offered higher average rates
of return than less risky investments.  For example, figure 3.1 shows
that over any 20-year holding period since 1940, compound average
annual returns for the Standard & Poor's composite stock index have
been higher than for U.S.  Treasury bills, which have both less
default risk and less market risk.  However, figure 3.1 also shows
that returns can still vary significantly across 20-year holding
periods.  For example, from 1953 to 1972, inflation-adjusted returns
on Standard & Poor's index averaged 9.1 percent.  However, for the
20-year holding period starting just 2 years later in 1955, returns
averaged less than half that rate at 4.2 percent.  This illustrates a
related type of risk--"liquidity risk," or the risk of having to
liquidate investments when market prices are not favorable.  In
addition, figure 3.1 shows that even conservative investments face
the risk of being eroded by inflation.  For example, Treasury bills
provided negative inflation-adjusted returns for several 20-year
holding periods. 

   Figure 3.1:  Holding Risky
   Investments for Long Periods
   Diminishes Risk

   (See figure in printed
   edition.)

Note:  Inflation-adjusted compound annual average rates of return
over rolling 20-year holding periods. 

Source:  GAO analysis using data from Robert J.  Shiller, Market
Volatility (Cambridge, Mass.:  MIT Press, 1989), available at
www.econ.yale.edu/~shiller/chapt26.html; Council of Economic
Advisers, Economic Report of the President, 1999 (Washington, D.C.: 
U.S.  Government Printing Office, Feb.  1999). 

Diversifying portfolios can also diminish the risks of investment
while still providing relatively higher returns.  A properly selected
combination of risky assets can have a lower risk than any of its
individual assets, and such portfolios would still provide higher
average returns than an asset with equal risk over the long term. 
For example, in the case of market risk, the risks from different
investments can offset one another if their prices do not fluctuate
in a similar pattern, even though they still individually earn higher
average returns.  However, such techniques are very sophisticated,
require substantial data analysis, and require the help of
professional advisers for the average investor.  Still, investors can
also diversify by investing in mutual funds, which do have
professional managers.  Nevertheless, diversifying a stock portfolio
does not protect investors against the risk of large swings in the
market as a whole; diversifying the portfolio to include other types
of investment assets, such as bonds, commodities, or real estate,
could help manage that risk.\31

Measures of investment risk and risk-adjusted rates of return are
available for helping plan portfolios.  Estimating a return on an
investment without taking into account its riskiness is likely to
overstate the benefit of that investment.  There are different ways
to adjust returns for risk, but there is no clear best way to do
so.\32 Moreover, these measures have key limitations that do not
permit making generalizations about the risk-adjusted rates of return
that individuals can earn on their portfolios as a whole.  For
example, a well-diversified portfolio has a different and often lower
risk than that suggested by the risks of its individual
components.\33 Also, some techniques for calculating risk-adjusted
rates relate only to one type of risk, such as market risk.  In
short, the combinations of risk and return that individual investors
face depend fundamentally on how portfolios are managed.\34

--------------------
\31 For a more complete discussion, see Katerina Simons,Risk
Adjusted Performance of Mutual Funds, New England Economic Review
(Federal Reserve Bank of Boston), Sept.-Oct.  1998, pp.  33-48. 

\32 See Social Security:  Capital Markets and Educational Issues
Associated With Individual Accounts (GAO/GGD-99-115, June 28, 1999). 

\33 Other limitations include (1) they are primarily useful for
investments with normal probability distributions, which means, for
example, that the probability of below-average returns equals the
probability of above-average returns; (2) while many individual
investments have such characteristics, different portfolios may not;
and (3) the measures presume that investors are free to borrow and
use leverage in their investment portfolios. 

\34 Some controversy surrounds the issue of risk adjustment; there is
no one risk-adjusted measure that everyone agrees is the correct one. 
For example, some analysts have suggested that the risk-adjusted rate
of return on all assets simply equals the rate on the least risky
assets.  By holding a particular mix of assets, they argue, investors
demonstrate that they are indifferent to the assets or else they
would change the mix.  However, different portfolios can have
identical risk levels but different expected rates of return because
portfolios can vary by how well risks are managed.  Nevertheless,
such analysts make the point that risk adjustment should reflect
investors' subjective preferences as well as objective, statistical
measures of risk.  See John Geanakoplos, Olivia S.  Mitchell, and
Stephen P.  Zeldes, "Social Security's Moneysworth," in Olivia S. 
Mitchell, Robert J.  Myers, and Howard Young, eds., Prospects for
Social Security Reform (Philadelphia:  University of Pennsylvania
Press, 1999), pp.  79-151. 

   INDIVIDUALS' PORTFOLIO CHOICES
   REFLECT THE EXTENT OF RISK
   AVERSION AND RETIREMENT
   PLANNING
---------------------------------------------------------- Chapter 3:3

Investors have varying degrees of aversion to risk that can vary in
particular by income, education, and gender.  Low-income and
less-educated individuals and women tend to choose less-risky
investments with lower average returns than high-income, highly
educated individuals and men.\35 This may reflect more than a lack of
knowledge of how to manage investment risk.  Those with lower income
and wealth have more to lose in relative terms than wealthier
individuals.  For example, if investors with savings of $5 million
each make a risky investment and lose 20 percent of the savings, they
still have $4 million and can still afford a very generous lifestyle. 
However, if investors with savings of only $500,000 lose 20 percent,
the $100,000 they lose can have a significant effect on their
lifestyle in retirement.  For example, if annuities paid an annual
benefit equal to 7 percent of the purchase price, a retiree with
$500,000 could purchase an annuity that paid $35,000 annually,
compared with the $28,000 that $400,000 would buy. 

In choosing the riskiness of their portfolios, prudent investors also
consider how close they are to retirement.  Those who are 20 years
away from retirement face less risk from investments with higher
average returns than those who are only 5 or 10 years away, as fig. 
3.1 suggests.  Shifting assets gradually to less risky investments as
retirement approaches helps guard against a sudden deterioration in
savings balances just before retiring or purchasing an annuity. 

--------------------
\35 For example, see Social Security Reform:  Implications for
Women's Retirement Income (GAO/HEHS-98-42, Dec.  31, 1997), pp. 
9-10. 

   ADMINISTRATIVE COSTS VARY BY
   INVESTMENT STRATEGY
---------------------------------------------------------- Chapter 3:4

Some investment strategies incur smaller administrative costs than
others.  For example, some investment funds are "passively
managed"--that is, the portfolio is based on a broad market index
such as the Standard & Poor 500, and trading activity automatically
follows a formula that tries to match the performance of that index. 
In contrast, some investment funds are "actively managed" by
professionals who pick stocks in an attempt to beat the averages. 
Such funds are more expensive to manage.  Moreover, some individual
investors use brokers to manage their own portfolios rather than just
buy shares in a large fund.  Such investors incur transaction costs
every time they make a trade. 

   PORTFOLIO MANAGEMENT WOULD
   AFFECT RETURNS ON INDIVIDUAL
   ACCOUNTS UNDER A RESTRUCTURED
   SOCIAL SECURITY PROGRAM
---------------------------------------------------------- Chapter 3:5

Portfolio composition and timing would play a large role in
determining the investment returns on individual accounts and, in
turn, the retirement outcomes under Social Security reform proposals
that would create individual accounts.  However, returns would also
depend substantially on the provisions of the proposal, particularly
how much latitude it gave workers to choose their investments and
annuitize their savings. 

For example, the 1994-96 Advisory Council on Social Security offered
three alternative reform proposals, two of which created a new system
of individual accounts.  The "individual accounts" (IA) proposal
would restrict investments to a limited number of passively managed
index funds, similar to the Thrift Savings Plan (TSP) available to
federal employees.  It also would require that workers purchase an
annuity at retirement with their Social Security retirement accounts. 
The "personal security accounts" (PSA) proposal would not impose such
restrictions. 

To illustrate the potential investment returns on the individual
accounts under alternative proposals, SSA actuaries developed a set
of hypothetical portfolio scenarios for the Advisory Council.  Table
3.1 presents these scenarios and the resulting investment yields. 
The scenarios illustrate how the combined effects of investment
choices, allocation changes with age, and administrative costs would
interact with three sets of assumptions for the returns on stock
investments alone.  The intermediate return assumption uses an
inflation-adjusted stock return of 7 percent per year, which reflects
the historical average for the period 1900-95; the high return
assumption uses a return of 9.3 percent.  In addition to making these
return assumptions, the actuaries analyzed a low-return case in which
the hypothetical worker's stock returns are roughly no better than
the returns on government bonds.  As a result, allocation decisions
do not affect the overall yield, although administrative costs still
differ between the IA and PSA proposals.  The low-return assumption
illustrates conservative or poorly timed investments or generally
poor returns on stocks.  In this case, the PSA proposal has a net
yield of 2.0 percent overall for the portfolio, and the IA proposal
has a net yield of 2.3 percent at all ages.  The estimated net yields
in table 3.1 do not project what investment returns would be on
average but simply illustrate a range of possible returns for
hypothetical workers that fit these particular scenarios.  Moreover,
they illustrate only returns on the individual accounts themselves,
not on all retirement contributions under a new system. 

                                        Table 3.1
                         
                         Returns on Market Investments Depend on
                                   Portfolio Strategies

                 PSA proposal--401(k)                 IA proposal--401(k) annuitized
        ---------------------------------------  ----------------------------------------
                                                                     Annual
                           Annual   Portfolio's                administrati   Portfolio's
        Percent of  administrativ           net                  ve expense           net
           account      e expense    inflation-    Percent of        factor    inflation-
        balance in         factor      adjusted       account   (percent of      adjusted
Age          stock    (percent of  annual yield    balance in          fund  annual yield
group       market  fund balance)     (percent)  stock market      balance)     (percent)
------  ----------  -------------  ------------  ------------  ------------  ------------
Intermediate returns: Stocks earn 7 percent
-----------------------------------------------------------------------------------------
Younge          55           1.00         3.885            55         0.105         4.780
 r
 than
 40
40-49           52           1.00         3.744            50         0.105         4.545
50-59           48           1.00         3.556            40         0.105         4.075
60-69           43           1.00         3.321            20         0.105         3.135

High returns: Stocks earn 9.3 percent
-----------------------------------------------------------------------------------------
Younge          55          0.500         5.650            55         0.105         6.045
 r
 than
 40
40-49           52          0.500         5.440            50         0.105         5.695
50-59           48          0.500         5.160            40         0.105         4.995
60-69           43          0.500         4.810            20         0.105         3.595
-----------------------------------------------------------------------------------------
Note:  Returns are adjusted for inflation.  These estimated
investment returns do not project what returns would be on average
but simply illustrate a range of possible returns for hypothetical
workers who fit these scenarios.  The PSA proposal would have
individually held and managed accounts and would not require that the
funds be annuitized at retirement.  The IA proposal would have the
federal government hold and manage the accounts with a limited number
of passively managed investment funds.  It would also require that
funds be annuitized.  In addition to these scenarios, the actuaries
analyzed a low-return case in which the hypothetical worker's stock
returns were roughly no better than the returns on government bonds. 
This would illustrate conservative or poorly timed investments or
generally poor returns on stocks.  In this case, the PSA proposal has
a net yield of 2.0 percent overall for the portfolio and the IA
proposal has a net yield of 2.3 percent at all ages. 

Source:  Advisory Council on Social Security, Report of the 1994-1996
Advisory Council on Social Security, Vol.  1 (Washington, D.C.:  Jan. 
1997). 

The share of the hypothetical portfolios invested in stocks is based
on 401(k)-plan experience about how workers distribute their 401(k)
funds among types of assets at different ages.  Compared with the PSA
proposal, the IA proposal assumptions have a smaller percentage of
funds invested in the stock market as people approach retirement
because of the annuity requirement. 

With regard to administrative costs for the individual accounts, the
hypothetical scenarios illustrate ranges as discussed in the reform
debate.  Account costs for the IA plan are smaller than for the PSA
plan because accounts and transactions are managed centrally by the
government, similar to the TSP plan; they would not vary by
individual.  In contrast, the Advisory Council assumed that the
account costs for the PSA plan would be larger than for the IA plan
since they would be individually managed.  Moreover, individuals
could manage their accounts very differently with widely ranging
administrative costs; some might modify their portfolios only rarely,
incurring very few transaction costs, while others might trade very
actively.  The actuaries assumed lower administrative costs for the
PSA high-return case than for the intermediate-return case; this
lower cost assumption helps define a more optimistic, illustrative
scenario. 

Note that this table presents one limited set of returns illustrating
one hypothetical worker's investment allocation choices.  In fact, as
some critics have contended, allocation choices could and would vary
significantly, especially by income, because low-wage workers tend to
invest more conservatively than high-wage workers.\36 Still, despite
their limitations, the Advisory Council's portfolio scenarios
represent one of the few efforts to illustrate the interaction
between portfolio management choices and overall stock returns.  Its
scenarios could be interpreted to reflect variations among
individuals as well as variations in market averages. 

--------------------
\36 Gordon P.  Goodfellow and Sylvester J.  Schieber, Simulating
Benefit Levels Under Alternative Social Security Reforms, in
Mitchell, Myers, and Young, eds., Prospects for Social Security
Reform, pp.  152-83. 

SIGNIFICANT ISSUES IN COMPARING
RATES OF RETURN
============================================================ Chapter 4

A simple comparison between the rates of return for the current
Social Security program and private market investments would be
misleading because of several key issues that such comparisons raise. 
First, such comparisons do not capture all the relevant costs that a
new system would imply, such as transition, administrative, and
annuity costs.  Second, future returns on both market investments and
Social Security as it is now structured may not be the same as in the
past, and the gap between those returns may narrow.  Third, risks
differ between the current Social Security program and private market
investments.  In contrast to simply comparing the current Social
Security program with market investments, many of these issues can be
addressed by estimating rates of return for specific reform proposals
and including both the individual account and the Social Security
components in those comprehensive estimates.  Still, even comparisons
of such return estimates among reform proposals have key limitations. 
For example, rates of return by themselves do not measure the risks
workers may face with respect to their retirement incomes. 

   ADDITIONAL COSTS NEED TO BE
   CONSIDERED IN COMPARING SOCIAL
   SECURITY AND MARKET RETURNS
---------------------------------------------------------- Chapter 4:1

Simple comparisons between returns on market investments and the
current Social Security program do not reflect all the costs that
would accompany a new system with individual accounts.  Such costs
include

  -- transition costs:  making the transition to the new system would
     involve the substantial costs of covering the unfunded
     liabilities of the current system;

  -- administrative costs:  administering the individual accounts and
     managing the investment of their funds would incur costs beyond
     the administrative costs of the current system; and

  -- annuity costs:  converting the account balances at retirement
     into annuities would also incur costs beyond the current
     system's administrative costs. 

All these costs would affect either the total contributions or the
total retirement income benefits or both under the new system. 
Moreover, the size of these costs and who pays for them would depend
on the provisions of a particular proposal.  These costs would not
necessarily be paid through the payroll taxes of the new system. 
Whoever pays these costs and how, they should all be reflected in any
rate of return estimates made for the new system.  Calculating valid,
comprehensive rates of return for a new system requires taking into
account all the contributions and benefits of the new system,
including some new types of contributions and benefits that are not
present in the current system.  A simple comparison between the
current program and historical market investments would not capture
all the contributions and benefits implied by a new system. 

      TRANSITION COSTS
-------------------------------------------------------- Chapter 4:1.1

A new system with individual accounts would generally increase the
degree to which retirement benefits are funded in advance.  Today's
pay-as-you-go system largely funds current benefits from current
contributions, but those contributions also entitle workers to future
benefits.  The amount necessary to pay the benefits already accrued
by current workers and current beneficiaries is roughly $9 trillion,
according to SSA.  In a pay-as-you-go system, an unfunded liability
will always exist and will be covered by future revenues or reduced
by benefit cuts or both.\37 However, any changes that would create
individual accounts would require revenues both to deposit in the new
accounts for future benefits and to pay for existing accrued
benefits.  Rate of return estimates for such a system should reflect
all the contributions and benefits implied by the whole reform
package, including the costs of making the transition. 

The effect of transition costs on rates of return depends greatly on
how those transition costs would be paid.  Tax rates could be
increased right away or many years later.  The costs could also be
paid for with benefit cuts, again either sooner or later.  Moreover,
some proposals would increase federal borrowing for some period of
time.  If such debt is repaid very slowly by rolling over the debt,
transition costs could be paid gradually over several generations. 
For some reform proposals, the "contributions" to pay transition
costs would include general revenues, not payroll taxes or account
deposits; general revenues primarily come from individual and
corporate income taxes.  Workers who pay these transition costs,
whoever they are in whichever generation, would receive lower overall
returns than those who do not. 

Some proponents of individual accounts point out that making the
transition to increased advanced funding is critical and has
implications for comparing rates of return.  They observe that rates
of return from the individual accounts in an advance-funded system
fundamentally differ from Social Security's implicit rates of return
because individual accounts would provide a new source of investment
funds and would increase national saving.  This increased pool of
investment would produce real increases in economic activity that
would make society better off.  In contrast, they assert that Social
Security only transfers income from taxpayers to beneficiaries,
detracts from saving and long-term economic growth, and produces no
real economic returns. 

Other analysts contend that workers paying the transition costs must
receive lower returns than they would otherwise in order to improve
returns for future generations.\38

Moreover, some observe that increasing the advance funding of Social
Security would not necessarily increase national saving.  Consumers
might compensate for their increased savings in their individual
accounts by saving less elsewhere or borrowing more.  National saving
also depends on federal budgets and surpluses, which could be
affected by the specific aspects of any changes enacted.  For
example, any federal borrowing that helps pay for transition costs
would offset any corresponding increases in individual account
balances to some degree. 

--------------------
\37 Note that the unfunded liability of $9 trillion is not the same
as the actuarial imbalance, which equals roughly $3 trillion,
according to SSA.  The actuarial imbalance reflects both future
revenues and future benefit accruals.  In contrast, the unfunded
liability reflects neither of these but rather the dollar value of
benefits accrued to date but not yet paid.  Under any reform proposal
that restored long-term solvencythat is, reduced the actuarial
imbalance to zeroadditional future program revenues would cover some
portion of the imbalance while any benefit reductions would eliminate
the remaining portion. 

\38 Geanakoplos, Mitchell, and Zeldes, Would a Privatized Social
Security System Really Pay a Higher Rate of Return?

      ADMINISTRATIVE COSTS
-------------------------------------------------------- Chapter 4:1.2

Market investments entail a variety of transaction and administrative
costs, which reduce the rates of return that investors effectively
earn.\39 These costs are not present in the current Social Security
system, at least not in the same form or to the same degree.  For
example, stock brokers charge commissions for making trades, mutual
fund managers are compensated for managing the funds, and making
deposits into accounts and recordkeeping entail some administrative
costs.  Reflected in such costs are marketing and advertising
expenses, including sales commissions, incurred as money managers and
brokers compete for the investors' business.  In some countries that
have privatized their social security systems, these costs have been
quite high.  In contrast, SSA does not maintain actual accounts for
each individual but simply keeps records of earnings.  Administrative
costs for Social Security's Old-Age and Survivors Insurance program
are less than 1 percent of annual program revenues.\40

In a new Social Security system with individual retirement accounts,
the size and effect of administrative costs would depend
significantly on how the new system is designed.  For example, just
as administrative costs vary between active and passive investment
strategies for individual investors, the range of investment
strategies permitted under the new system would affect its
administrative costs.  A centrally managed approach, such as that
envisioned in the IA proposal noted earlier, could minimize costs
associated with recordkeeping and financial transactions.  Limiting
the range of investment options to a few types of funds available
through the central system could also avoid substantial marketing
costs that might arise if individuals had the freedom to switch from
one money manager to another.  Moreover, the effect of administrative
costs on rates of return could vary across workers, depending on how
those costs are paid.  If individuals were charged a flat fee per
account for administrative costs, accumulations in small accounts
would be affected to a greater extent than if they were charged an
annual percentage.  Therefore, such costs would diminish the
effective rates of return more for low-income workers with smaller
balances than for high-income workers.  Finally, higher
administrative costs could be associated with more customer services,
and some of the additional administrative costs would also provide
other, nonquantifiable benefits, such as investors' freedom of
choice. 

--------------------
\39 See Social Security Reform:  Administrative Costs for Individual
Accounts Depend on System Design (GAO/HEHS-99-131, June 18, 1999). 

\40 In addition to direct administrative costs, various indirect
costs exist under the current system, such as those that Treasury and
employers incur for various processing tasks.  Indirect costs would
also exist in a restructured system, including some new costs
potentially, such as costs for investor education.  Because indirect
costs may or may not have an effect on individuals' specific
retirement contributions and benefits and in many cases are difficult
to measure, it is not clear how or whether to incorporate them into
rate of return estimates.  Such costs may also ultimately be paid in
the form of lower wages to workers or higher prices to consumers,
which further complicates how to treat them in rate of return
estimates. 

      ANNUITY COSTS
-------------------------------------------------------- Chapter 4:1.3

In addition to the costs of managing the accounts before retirement,
the costs of annuitizing the balances at retirement would affect the
retirement incomes individuals actually enjoy and therefore their
effective rates of return.  Like other investments, annuities
purchased in the private market entail a variety of transaction and
administrative costs.\41 However, annuities are also a form of
insurance, and annuity prices in a free market reflect profits that
insurers make.\42 Moreover, annuity costs could vary substantially
from person to person, depending especially on interest rates at the
time of purchase.  Annuity costs greatly depend on interest rates,
with higher interest rates increasing the size of the annuity
benefit.\43

Annuity costs could also vary considerably across groups of people
with different life expectancies, depending on the ability of annuity
providers to charge different prices to different groups, such as
groups defined by gender or health status.  Those groups with longer
life expectancies would receive their annuities longer, and their
annuity providers would therefore incur higher annuity costs for
them.  Reform provisions might prohibit annuity providers from
charging different prices based on race, gender, health status, or
other factors that reflect differences in life expectancy.\44 Such
prohibitions would reduce the variation in annuity costs across
groups.  However, they would also implicitly transfer income from
those groups with shorter life expectancies--such as men or the poor
or African-Americans--to groups with longer life expectancies--such
as women or the wealthy or whites.  Rates of return would vary across
such groups accordingly.  Still, such prohibitions might not prevent
annuity providers from using marketing and advertising to appeal to
retirees with shorter life expectancies. 

Prohibiting annuity providers from charging different prices to
different groups would probably have a limited effect unless
annuities were mandatory.  Otherwise, individuals with shorter life
expectancies might perceive annuity costs to be too high and choose
not to buy them.  In effect, they would "self-annuitize" and face the
risk that they might outlive their retirement savings.  In the
current annuity market, consumers who expect to live a long time
because of health status or family history are much more likely to
purchase annuities than those who do not.  As a result, annuity
purchasers as a group have a longer life expectancy at any given age
than the population at large, so annuity prices are higher than they
would be if everyone purchased an annuity.  This problem, known as
"adverse selection," would not be nearly as significant if annuities
were mandatory because people with lower life expectancies would not
be able to opt out of buying an annuity. 

By helping reduce adverse selection, making annuities mandatory in a
new system could significantly reduce annuity costs for individuals
who would buy annuities anyway while increasing costs for those who
otherwise would not.  It would implicitly transfer income from those
who die earlier to those who die later but help ensure adequate
retirement incomes for those who die later. 

--------------------
\41 See Social Security Reform:  Implications of Annuities for
Individual Accounts (GAO/HEHS-99-160, July 30, 1999). 

\42 In contrast, when the government provides annuities, such as
Social Security benefits, it does not make a profit. 

\43 The more money an annuity fund can earn in interest, the more it
can pay out in benefits. 

\44 Requiring insurers to use unisex annuity rates would be an
example of this.  Unisex annuity rates are currently required for
employer-provided group annuities, but annuities sold to individuals
are usually based on gender-specific life tables.  The current Social
Security program, in effect, also provides unisex annuities, which
results in an income transfer from men to women and higher rates of
return for women. 

   FUTURE AVERAGE RATES COULD
   DIFFER FROM HISTORIC AVERAGES
---------------------------------------------------------- Chapter 4:2

In the future, average rates of return on either market investments
or Social Security as it is currently structured could differ
significantly from their historical averages.  Moreover, the gap
between these rates could narrow.  Fundamentally, economic growth
drives rates of return for both market investments and Social
Security.  Rate of return projections for either are misleading if
they are not consistent with economic growth projections.  More
specifically, capital productivity helps determine market rates of
return while the growth of labor productivity helps determine Social
Security's long-term average rates of return.  Trends in the
productivity of both capital and labor are difficult to predict for
various reasons, and the markets for capital and labor interact with
each other in determining what share of the national income is paid
to each. 

      RETURNS ON MARKET
      INVESTMENTS DEPEND ON
      ECONOMIC GROWTH AND MARKET
      FORCES
-------------------------------------------------------- Chapter 4:2.1

For market investments in capital, long-term average rates of return
ultimately depend on whether those investments produce more income by
producing more goods or services.  However, capital comes in many
forms, such as land, buildings, technology, machinery, supplies, and
product inventory.  Not all these forms of capital are financed
through the stock and bond markets, but they all compete with labor
for their share of the national income, which determines their rates
of return.  These rates of return are related to the productivity of
each factor of production, but many economic forces work through the
markets to determine the rates of return they earn individually. 

Several issues make it hard to predict returns on capital. 
Investment in new technology can result in major breakthroughs that
change the way we all live or it can go down dead-end paths with
little if any result.  Moreover, measuring productivity growth from
advances in information technology, for example, has proven
difficult.  Also, the dynamics of the stock market may be changing. 
For example, the difference, or "spread," between rates of return on
stocks and Treasury securities has been shrinking.  Some economists
have suggested that this trend reflects that the economy appears to
be less susceptible to recessions, making stock investments less
risky.\45

Still, in the aggregate, returns to capital and labor must
fundamentally reflect growth in national income, so projections of
future returns depend on assumptions about economic growth.  In fact,
the growth of the U.S.  economy is expected to slow as the population
ages.  The rate of national saving and the growth in productivity and
wages have slowed notably in the past two decades, and these trends
relate to economic growth.  The Social Security trustees' projections
reflect an assumption that growth will slow as the baby boom
generation retires and relatively fewer young people enter the labor
force.  From 1989 to 1997, the economy grew at an inflation-adjusted
average annual rate of 2.2 percent.  The trustees' intermediate
assumptions use a growth rate of 2.0 percent over the next decade and
1.4 percent by 2020.  One analysis estimated that these growth rate
assumptions imply that future stock market returns could be as low as
4.0 percent.\46 If the rate of economic growth turns out to be higher
than these projections, returns to capital could be higher, but so
could be returns to labor and in turn Social Security's implicit
rates of return and its actuarial balance. 

--------------------
\45 For more discussion of factors that could diminish future stock
returns, see Social Security Financing:  Implications of Government
Stock Investing for the Trust Fund, the Federal Budget, and the
Economy (GAO/AIMD/HEHS-98-74, Apr.  22, 1998), pp.  41-44. 

\46 Dean Baker, Saving Social Security with Stocks:  The Promises
Don't Add Up (New York:  Twentieth Century Fund, 1997). 

      RETURNS ON SOCIAL SECURITY
      DEPEND ON TOTAL WAGE GROWTH
-------------------------------------------------------- Chapter 4:2.2

While rates of return on market investments depend on capital's share
of income from economic growth, the current Social Security program's
long-term average rates of return depend on labor's share.  As
discussed earlier, in a mature pay-as-you-go Social Security system,
long-term average implicit returns depend predominantly on the growth
rate of all wages covered by Social Security.  Growth in total
covered wages reflects both average wage increases and growth of the
labor force.  Wage increases depend on the growth of labor
productivity, the growth of the economy as a whole, and the results
of other market forces in determining labor's share of national
income.  As the baby boom generation retires, labor force growth is
expected to slow dramatically but average wages could be bid up in
response.  As a result, the net effect on total covered wage growth
is unclear. 

      THE GAP BETWEEN MARKET AND
      SOCIAL SECURITY RETURNS MAY
      NARROW
-------------------------------------------------------- Chapter 4:2.3

As a result of anticipated trends, market forces could leave a
smaller gap between the long-term average rates of return on market
investments and Social Security as it is currently structured. 
Capital and labor compete in the market for their shares of the
national income, and they interact with each other.  For example,
capital investment tends to improve the productivity of labor, which
in turn tends to increase wages.  As noted earlier, capital is
expected to be relatively more plentiful, and labor is expected to be
relatively more scarce in the future.  Either of these trends could
decrease returns to capital and increase returns to labor.  In turn,
the rates of return available from a new system with individual
accounts could be smaller than historical returns on market
investments might suggest, and Social Security's implicit rates of
return could be higher than they are expected to be using the current
trustees' assumptions regarding wage growth. 

Since 1956, the growth rate of total inflation-adjusted wages has
averaged roughly 3 percent on a compound annual basis; since 1967, it
has averaged 2.4 percent.  These growth rates are lower than the
long-term average annual rate of return of 7 to 8 percent on stocks. 
However, figure 4.1 illustrates that over 20-year periods on a
compound annual average basis, stock returns actually dipped below
the growth rate of total covered wages during several periods since
the 1950s.  Still, it remains difficult to say just how narrow the
gap will be and how much it may fluctuate. 

   Figure 4.1:  Twenty-Year
   Average Rates of Return on
   Market Investments Compared
   With Growth Rate in Total
   Covered Wages

   (See figure in printed
   edition.)

Note:  Inflation-adjusted compound annual averages over rolling
20-year periods.  In the early years of Social Security's history,
total covered wages increased dramatically in some years as coverage
was extended to more workers or the maximum taxable earnings
increased. 

Source:  From GAO analysis of data from SSA; from Robert J.  Shiller,
Market Volatility (Cambridge, Mass.:  MIT Press, 1989), available at
www.econ.yale.edu/~shiller/chapt26.html; and from Council of Economic
Advisers, Economic Report of the President, 1999 (Washington, D.C.: 
U.S.  Government Printing Office, Feb.  1999). 

   RISKS DIFFER BETWEEN SOCIAL
   SECURITY AND MARKET INVESTMENTS
---------------------------------------------------------- Chapter 4:3

Simple rate of return comparisons between the current Social Security
program and market investments do not take into account the
differences in risk associated with those returns.  Economic
uncertainty affects the risks and returns of private market
investments but also, in a different way, of the Social Security
system.  In addition, political risks exist for both the current and
any restructured Social Security system.  For retirement incomes, a
primary risk is that they may not be adequate.  In addition, risks
make retirement income less predictable, which diminishes the ability
of individuals and the society as a whole to set aside a level of
resources for retirement that is neither too high nor too low.  By
themselves, rate of return estimates reflect only the average of the
possible retirement incomes, not their adequacy or the degree to
which they could vary.  While some approaches do exist for assessing
the risks of alternative reform proposals, none fully capture the
full range of variability in retirement incomes.  Just as a trade-off
largely exists between risk and return in market investments, the
same trade-off exists among alternative approaches to Social Security
reform. 

As discussed earlier, rates of return in the private market vary
considerably according to various types of investment risk, including
market risk, default risk, and the like, and also according to how
well investors manage those risks.  As a result, in a new Social
Security system with individual accounts, average rates of return
would vary both by year of birth and by individual, and this source
of variation is not present in the current system.  Different groups
of retirees born in different years would accumulate savings and
receive investment earnings over different sets of years; the returns
that the private market offers could vary substantially between those
sets of years, as figure 3.1 illustrated earlier.  Moreover, retirees
born within a given year and facing the same investment period could
have very different rates of return, depending on how they allocated
and timed their investments.  Even restrictive individual account
proposals would permit workers to invest all their funds
conservatively and switch back and forth between alternative funds
with different types of assets.  In contrast, Social Security's
current structure results in rates of return that vary relatively
little from year to year because its rate of return depends on
long-term economic trends, not market fluctuations. 

In addition to investment risk, participants face political risk
under either the current system or a new one.  That is, the Congress
could enact changes to the system, such as cutting benefits, raising
taxes, changing the tax treatment of retirement benefits, or
guaranteeing a minimum retirement income, that would affect returns
on retirement contributions.  Rate of return comparisons should
ideally account for differences in both market and political risk,
but political risks are not easily quantified and both require
subjective judgments. 

As discussed earlier, a variety of attempts have been made to measure
rates of return on a risk-adjusted basis.  While some measures have
been developed for individual investments based on the statistical
variation of their rates of return, these measures adjust only for
market risk.  Moreover, risk fundamentally depends on portfolio
choice because portfolios can be designed so that investment risks
offset one another to some degree. 

One recent study analyzed historical investment returns in various
countries and examined how much retirement incomes in an individual
account system would vary from workers who retire in one year to
those who retire in the next.  In the case in which workers with
average earnings invested half their portfolio in stocks and received
a pension equal to 50 percent of average earnings, identical workers
in the next year could expect, on average, to get a pension equal to
anywhere from 46 to 54 percent of average earnings.  In some years,
this variation could be less, in others more.\47

Just as a trade-off exists between risk and return in market
investments, the same trade-off exists among alternative approaches
to Social Security reform.  Any Social Security changes enacted will
implicitly reflect the relative priorities placed on maximizing
returns or minimizing risks for workers and beneficiaries.  For
example, some individual account proposals would guarantee that
workers would have at least as much retirement income as they do
under the current system.  To some degree, such guarantees provide an
incentive to take greater investment risks.  If some workers do
poorly enough that the government must make up the difference,
taxpayers paying the subsidies will have lower rates of return than
they would otherwise.  Thus, efforts to minimize risk could also
reduce returns. 

--------------------
\47 Lawrence H.  Thompson, Predictability of Individual Pensions,
Ageing Working Paper 3.5 (Paris:  Organisation for Economic
Co-operation and Development, 1997),
www.oecd.org/els/pds/socialpolicy/ENG5.PDF. 

   COMPARISONS BETWEEN REFORM
   PROPOSALS HELP CAPTURE RELEVANT
   ISSUES
---------------------------------------------------------- Chapter 4:4

As the preceding discussion demonstrates, the rates of return that
participants would enjoy under a restructured Social Security program
are not equal to the returns they might receive on their market
investment accounts, so a simple rate of return comparison between
the current program and market investments would be misleading in
assessing the advantages of a new system.  All the costs participants
pay and all the benefits they receive under the new system should
enter into the rate of return calculations.  Including both the
individual account and Social Security components in one
comprehensive rate of return estimate provides the best basis for
comparing the individual equity of alternative reform proposals. 
Still, individual equity is only one of many criteria to use in
comparing proposals, and rates of return are only one measure of
individual equity. 

Comparing such comprehensive rates of return for reform proposals can
show how transition costs will have different effects on workers born
in different years and, hence, can reveal their effects on
intergenerational equity.  They can show that returns on the entire
package of retirement contributions depend on the proportion that is
deposited into individual accounts.  They can show how returns depend
on different provisions relating to administrative costs, annuities,
and investment restrictions.  They can show how returns depend on the
economic assumptions that drive the rates of return on market
investments and Social Security benefits generally. 

However, such comparisons among reform proposals are limited because
many of these effects are difficult to predict and model.  Moreover,
such comparisons should be made only between proposals that achieve
comparable levels of long-term actuarial balance.  Also, some reform
provisions under consideration, such as the use of general revenues,
are complicated to incorporate in rate of return calculations. 
Finally, rates of return alone do not measure the risks that
individuals would face in terms of the adequacy and predictability of
their retirement incomes. 

      ADVISORY COUNCIL ESTIMATES
      ILLUSTRATE RETURNS FOR
      ALTERNATIVE PROPOSALS
-------------------------------------------------------- Chapter 4:4.1

While many studies have published rate of return estimates for the
current Social Security program, very few have published estimates
for alternative reform proposals.  The Report of the 1994-1996
Advisory Council on Social Security provides an extensive set of rate
of return estimates for reform proposals.\48 These estimates are now
somewhat dated, especially since they are based on projections from
the 1995 Social Security trustees' report.  Since then, the economy
has grown faster than projected, and the long-term actuarial balance
has improved somewhat.  Also, other reform proposals have been
introduced that warrant study.  Still, the Advisory Council's rate of
return estimates are the best available and are sufficient to
illustrate some key points about comparing returns across reform
proposals.  Moreover, the Council's three alternative proposals
provide a broad range of reform approaches that reflect the essence
of key components of more recent proposals. 

The Advisory Council report provides estimates for three reform
proposals and two benchmark cases of particular interest.  The IA and
PSA plans are individual account proposals, described in chapter 3. 
The third proposal, the "maintain benefits" (MB) plan, would make
changes within the current program structure to restore solvency.  In
part, as one alternative, the MB plan would increase revenues by
investing up to 40 percent of the trust funds in the stock market. 
Although it would not create a new system of individual accounts, it
would increase advance funding somewhat.  The report also provides
estimates for two illustrative benchmark cases.  The first, known as
present law-PAYGO, makes no changes except for increasing taxes
sufficient to restore solvency on a pay-as-you-go basis.  The second,
known as maintain tax rates, makes no changes except to cut
benefits enough to restore solvency with the current tax levels. 

Among individual account proposals, the IA and PSA plans represent
two ends of a spectrum along which most individual account proposals
fall.  The IA plan would have deposits to the accounts equal to 1.6
percent of workers' earnings, while the PSA plan would have deposits
of 5 percent.  Several recent proposals currently under discussion
have deposits in the range of 2 to 2.5 percent of earnings, while a
few others have deposits as low as 1 percent and as high as 10
percent.  The IA plan would have the federal government centrally
manage the accounts on workers' behalf, while the PSA plan would have
individuals manage their own accounts.  The IA plan would provide a
limited selection of investment options, while the PSA plan would
place few restrictions on how workers invest their funds.  The IA
proposal would require workers to purchase an annuity at retirement,
while the PSA plan would not.  The IA plan would retain the current
structure of Social Security benefits but would reduce benefits so
that current Social Security payroll tax rates would adequately fund
them.  The PSA plan would replace the current Social Security benefit
with a relatively small flat benefit that would not depend on
lifetime earnings. 

--------------------
\48 One other study has published payback ratios, which are another
type of money's-worth measure, for some stylized, illustrative reform
approaches but only for workers from two different birth years.  See
Kelly A.  Olsen and others, How Do Individual Social Security
Accounts Stack Up?  An Evaluation Using the EBRI-SSASIM2 Policy
Simulation Model, issue brief 195 (Washington, D.C.:  Employee
Benefits Research Institute, Mar.  1998). 

      REFORM PROPOSAL COMPARISONS
      ILLUSTRATE THE EFFECT OF
      TRANSITION COSTS AND
      INTERGENERATIONAL EQUITY
-------------------------------------------------------- Chapter 4:4.2

Social Security reforms will have different effects on different
generations depending on their specific provisions.  One criterion
for evaluating alternative proposals is the "intergenerational
equity" they provide, or whether rates of return are fairly
consistent across generations.  The way proposals would handle the
current long-term financing shortfall and the costs of making a
transition to a new system would have especially significant effects
on intergenerational equity. 

Figure 4.2 provides the rate of return estimates for one illustrative
type of household with average earnings for workers born in different
years, as calculated by SSA actuaries for the Advisory Council.\49
Also, these estimates illustrate only the intermediate return case in
which any stock market investments in a household's portfolio earn an
inflation-adjusted average annual return of 7 percent.  Rates of
return for workers born in earlier years would not vary significantly
among the reform options because none of them would reduce benefits
for those already retired or nearing retirement.  The declining rates
of return for persons born earlier reflect the maturing of the
current system and recent declines in total wage growth, as discussed
earlier. 

   Figure 4.2:  Rate of Return
   Comparisons for Reform
   Proposals Illustrate Effects on
   Intergenerational Equity

   (See figure in printed
   edition.)

Note:  Inflation-adjusted rates, two-earner couples with average
earnings.  All proposals achieve comparable actuarial balance over 75
years.  These estimates include all Social Security contributions and
benefits, including disability.  In 1998, the average earnings level
was about $29,000.The raise taxes only option makes no changes to the
current program except to raise taxes on a pay-as-you-go basis.  The
cut benefits only option cuts benefits sufficiently to maintain the
current tax rate within the current program structure.  The MB
(maintain-benefits) proposal, among other provisions, provides for
investing 40 percent of trust fund assets in stocks.  The last two
proposals establish individual savings accounts with various
provisions, including different provisions about the range of
investment flexibility.  The MB and the PSA and IA intermediate
return cases reflect an annual inflation-adjusted rate of return on
equities equal to 7 percent. 

Source:  Advisory Council on Social Security, Report of the 1994-1996
Advisory Council on Social Security, Vol.  1 (Washington, D.C.:  Jan. 
1997). 

The trough in rates of return for both the IA and PSA intermediate
cases reflects the effect of transition costs, with rates of return
depressed while these costs are paid off.\50

As a result, many participants would not get significantly higher
rates of return than they would under the current system.  However,
rates of return then improve as the transition costs diminish.  This
improvement also reflects that persons born in each successive year
have had more years in which to make individual account deposits. 
Each successive group has a larger proportion of retirement income
coming from these accounts and has more to gain from the new system's
potentially higher investment returns.  In contrast, rates of return
are roughly level for the MB plan from the 1943 birth year on.  The
MB plan offers higher rates of return than either the
raise-taxes-only or cut-benefits-only cases, largely because it draws
new revenue from higher investment returns. 

For the raise-taxes-only case, rates of return decline for the later
birth years because taxes increase only as revenues are needed to pay
benefits in this scenario.  Under current projections, no further tax
increases would be needed until 2034, and further increases would be
required in later years.  Rates of return therefore diminish for
persons working in later years because they pay more in taxes without
any corresponding increases in benefit levels.  In contrast, the
effect on rates of return of the cut-benefits-only approach becomes
more level because the tax rate remains constant from now on.  While
benefit cuts are necessary to sustain solvency in this case, rates of
return remain fairly constant.  In effect, the cuts in the benefit
amounts are compensating for the fact that people are living longer
and collecting benefits longer; but on a total lifetime basis,
benefits are roughly constant, as is the tax rate.\51 This
observation underscores the fact that increasing longevity is one of
the root causes of Social Security's long-term financing problem,
since it contributes substantially to the declining ratio of workers
to beneficiaries.  Current benefit levels cannot be sustained under
any scenario without additional revenues of one sort or another,
which could include higher investment returns. 

--------------------
\49 Other types of households, such as single workers or one-earner
couples, with different earnings levels exhibit somewhat similar
patterns with regard to intergenerational equity.  However, higher
earners have generally lower rates of return, and lower earners have
generally higher rates of return.  Also, one-earner couples have
generally higher rates of return under the MB plan than under either
the PSA or IA plan.  As noted earlier, some interactions exist among
the various characteristics--for example, between household type and
earnings level.  All the various combinations present a more
complicated picture.  For a more extensive set of rates of return and
money's-worth measures, see Advisory Council on Social Security,
Report of the 1994-1996 Advisory Council on Social Security, Vol.  1,
pp.  165-230.  In addition, these estimates are for hypothetical
workers with a steady pattern of lifetime earnings.  As noted in
chapter 2, the hypothetical average earner may have earnings
somewhat higher than the true average.  As a result, workers with
earnings closer to the true average would have higher rates of return
on the Social Security component of their retirement income. 
Moreover, rates of return will vary for the individual account
component by the shape of the earnings history.  For example, for a
given lifetime average earnings level, workers who have higher
earnings earlier in their careers would have higher rates of return
on their individual accounts than those with lower early earnings
since their account deposits would have more years to earn interest. 
See Burtless, Bosworth, and Steuerle, Changing Patterns of Lifetime
Earnings.

\50 Note that the PSA proposal has a higher comprehensive rate of
return than the IA proposal, even though the IA individual account
component has a higher yield.  This reflects the difference in the
size of each proposal's account.  See below for further discussion. 

\51 In theory, a scenario in which annual benefit levels are
maintained could be sustained by taxes that increase gradually in a
way that reflects longevity improvements.  Such an approach could
also result in relatively level rates of return for different birth
years if tax increases were actuarially calculated to reflect
longevity trends.  The raise-taxes-only scenario illustrated here
does not do that because the tax increases reflect the cash flow
demands of the program, not the actuarial cost of each year's newly
accrued benefit promises. 

      REFORM PROPOSAL COMPARISONS
      ILLUSTRATE THE EFFECTS OF
      ACCOUNT SIZE AND COSTS
-------------------------------------------------------- Chapter 4:4.3

For individual account proposals, the comprehensive rates of return
will depend primarily on (1) what proportion of retirement
contributions can be invested in the market for potentially higher
returns and (2) what net returns those market investments actually
earn.  In addition to depending on market outcomes, net returns will
depend on administrative and annuitization costs, the effect of
annuitization requirements on investment strategies, and the range of
permitted investment options.  The Advisory Council's rate of return
estimates reflect these various factors to some degree.  For example,
as discussed earlier and as illustrated in table 3.1, these estimates
reflect the administrative and annuity costs implied under the
proposals. 

Figure 4.3 includes two lines each for the IA and PSA proposals, one
for low and one for high investment returns, as well as the benchmark
case of cutting benefits only.  Rates of return for the low-return
cases do not vary significantly from each other or from the option of
restoring actuarial balance by cutting benefits alone.  This largely
reflects that the low-investment earnings assumption roughly
parallels Social Security's long-term implicit rate of return. 
However, these low-return scenarios also illustrate that any
improvement in rates of return from individual account proposals
depends on actually realizing higher investment returns.  Increasing
the level of advanced funding alone does not improve returns, even
after all transition costs have been paid. 

   Figure 4.3:  Rate of Return
   Comparisons for Reform
   Proposals Illustrate the
   Effects of Account Size and Net
   Returns

   (See figure in printed
   edition.)

Note:  Inflation-adjusted rates, two-earner couples with average
earnings.  All proposals achieve comparable actuarial balance over 75
years.  These estimates include all Social Security contributions and
benefits, including disability.  In 1998, the average earnings level
was about $29,000.The high-return cases reflect an annual
inflation-adjusted rate of return on equities equal to 9.3 percent. 
The low-return cases reflect a 2.3-percent rate of return, which is
comparable to returns earned by the Social Security trust funds. 

Source:  Advisory Council on Social Security, Report of the 1994-1996
Advisory Council on Social Security, Vol.  1 (Washington, D.C.:  Jan. 
1997). 

In the high-return cases, the PSA proposal yields higher
comprehensive rates of return, largely because a larger proportion of
earnings is going into the individual accounts than with the IA
proposal.  Since the accounts are earning a high rate of return, the
larger the account the more it raises the comprehensive rate of
return.  This also explains why the PSA plan provides higher returns
than the IA plan in the intermediate-return case illustrated in
figure 4.2.  In the intermediate case, the IA accounts yield a higher
investment return than the PSA plan in all but the last age range. 
However, when averaged in with the Social Security component in the
comprehensive rate of return, the PSA still yields a higher overall
return because the PSA accounts provide a larger share of retirement
income.  In the high-return cases, the higher returns for the PSA
plan also reflect the assumption that workers under the PSA plan
would have a larger share of their accounts invested in stocks at
later ages, as illustrated in table 3.1.  For both plans, stocks are
assumed to provide an inflation-adjusted return of 9.3 percent
annually in the high-return scenarios.  The three alternative return
assumptions of 2.3, 7, and 9.3 percent are arbitrary illustrative
cases agreed on by the Advisory Council; they do not necessarily
reflect the latest assumptions about economic growth or other market
projections. 

      REFORM PROPOSAL COMPARISONS
      HAVE SOME LIMITATIONS
-------------------------------------------------------- Chapter 4:4.4

Even though comparing rates of return for reform proposals is much
more valid than simply comparing returns for the current system with
those for market investments, limitations and cautions still arise. 
For example, any reform proposals that are compared should achieve
the same degree of long-term solvency.  Also, it may not be possible
to incorporate the effects of some specific provisions of reform
proposals.  Moreover, by themselves, rate of return estimates do not
measure the risks that workers may face in terms of the
predictability or adequacy of their retirement incomes. 

Some reform provisions make it difficult to generalize exactly what
contributions and benefits would be, which complicates rate of return
analysis.\52 For example, some reform proposals would draw on general
revenues of the federal government as well as on Social Security's
own revenues.  General revenues come from a wide variety of sources,
including both personal and corporate income tax.  Shareholders,
employees, suppliers, or consumers ultimately end up paying corporate
income tax in the form of reduced earnings, reduced wages, reduced
supplier prices, or increased consumer prices.  Rate of return
estimates should include all contributions to the new Social Security
system made by all who benefit from it, regardless of how those
contributions are made.  It is not at all clear how to incorporate
contributions from general revenues into return estimates because
general revenues come from many current and future beneficiaries born
in various years with various incomes and household sizes who provide
those revenues in varying proportions.  However, to leave any general
revenue contributions out of return estimates would artificially make
rates of return look better than they would actually be. 

As discussed earlier, any rate of return is associated with some
level of risk, but the return estimate itself does not measure that
risk.  For rates of return under a restructured Social Security
system, two distinct types of risk are of interest.  First, how much
could actual rates of return vary from the average projected rate? 
This variability arises on both the aggregate and the individual
level.  A projection that stock investments will earn 7 percent over
some future period represents an average for a number of possible
aggregate outcomes with different probabilities.  The actual
aggregate outcome could be higher or lower.  However, even if the
aggregate outcome actually turns out to be 7 percent, it would
represent an average across many different investors.  So the first
type of risk, variability, reflects the risk both that the aggregate
projection may be wrong and that an individual's return could vary
from the average.  The second type of risk is the risk for specific
individuals that retirement outcomes are not adequate.  For example,
what is the probability for a given individual of winding up with a
retirement income below the poverty line?  Workers value not only
being able to predict their retirement income but also knowing that
it will be adequate. 

The Advisory Council estimates do not really illustrate the risk of
either variability or inadequacy.  They do suggest a range of
possible outcomes, but no probabilities are associated with those
outcomes.  So these estimates do not reveal the degree to which
actual retirement incomes could vary from one worker to another or
from one birth group to another; they illustrate only that they could
vary using arbitrarily chosen examples. 

Some studies have examined the statistical variation of outcomes from
various reform packages, but this analysis still goes only so far.\53
In particular, they examine two types of outcomes.  Two studies
examine the variation in dollar retirement incomes under alternative
proposals while a third examines the variation in rates of return
that workers experience.  Such studies make an important
contribution, but it is necessary to appreciate their limitations and
how such analysis might be extended.  They do study the aggregate
variability risk--that is, how much outcomes vary because of how much
actual aggregate returns could vary from the projected average. 
Studies examining retirement incomes address adequacy somewhat by
making comparisons with other reforms that do not involve individual
accounts.  However, all these studies assume that all individuals
have identical investments and earn the aggregate rate of return. 
This understates an individual's risk of inadequacy because it does
not reflect individual variation in investment returns.  Also, these
studies do not examine the possible variability among persons born in
different years--for example, if there were a dramatic surge or drop
in the stock market or interest rates from one retirement year to the
next. 

So, while such statistical approaches help describe the minimum
extent of variability, they do not describe the maximum variation
possible.  Moreover, they do not capture how individuals subjectively
assess and respond to risk in their own investment choices.  Some
individuals may be indifferent to receiving a lower rate of return
with less risk and a higher rate with more risk, and such preferences
vary by individual.  Risk analysis based on objective statistical
measures is possible and useful, but ultimately it is limited to some
degree in its ability to address individuals' subjective preferences
regarding risk. 

--------------------
\52 In addition, one recent study points out that these estimates do
not capture the difference in the tax treatment of Social Security
benefits under the alternative proposals.  Under the IA and MB plans,
Social Security benefits would be subject to income tax to the extent
that they exceeded contributions, although the personal account
portion of the IA plan would not be taxable.  Under the PSA proposal,
retirement benefits would not be taxable.  More generally,
incorporating tax effects vastly complicates rate of return analysis. 
Because income tax rates depend on all sources of income, not just
income from Social Security or the individual accounts, two retirees
could have the same retirement benefits from the new system but pay
different tax rates on those benefits.  One retiree may have income
from an employer pension, employment during retirement, or other
saved assets, while another retiree may have none of these.  See
Goodfellow and Schieber, Simulating Benefit Levels.

\53 Goodfellow and Schieber, Simulating Benefit Levels; Olsen and
others, How Do Individual Social Security Accounts Stack Up?; Lee
Cohen, Laurel Beedon, and Carlos Figueiredo, A Critical Look at
Equity Investment in the 1994-1996 Advisory Council on Social
Security Recommendations, issue brief 30 (Washington, D.C.:  American
Association of Retired Persons, Public Policy Institute, Apr.  1998). 

OBSERVATIONS
============================================================ Chapter 5

Comparing rates of return on Social Security and private market
investments has frequently been discussed in evaluating options for
reform.  Social Security's implicit rate of return provides a measure
of individual equity--that is, whether workers get a fair level of
benefits relative to their contributions.  Intuitively, it gives a
sense of whether workers get their money's worth from Social
Security, especially in relation to what they could have earned on
their contributions elsewhere.  However, simply comparing the current
Social Security program's implicit rate of return with historical
returns on market investments reveals little about what workers have
to gain from alternative reform proposals.  Rather, if rates of
return are to be compared, they should ideally be compared among
complete reform proposals to capture all the costs that the proposals
imply and to reflect the latest projections of future economic and
demographic trends. 

Even such rate of return comparisons among reform proposals must be
kept in careful perspective.  Rates of return address individual
equity alone, which is just one of many factors that should be used
in considering Social Security reform alternatives.  One of Social
Security's primary objectives has always been to help ensure adequate
incomes not just for the elderly but also for the disabled and for
dependents and survivors.  The current Social Security system
attempts to strike a balance between the competing goals of income
adequacy and individual equity.  Social Security's income transfers
are a primary means of helping ensure income adequacy but implicitly
diminish individual equity at the same time.  Reforms could alter the
balance between equity and adequacy, but any such change should be a
conscious and informed choice. 

In addition to the adequacy-equity balance, several other
considerations deserve attention in weighing alternatives for reform. 
Potential effects on the federal budget and the national economy are
key factors to examine.  Reforms could have significant implications
for the level of national saving, which fundamentally affects the
prospects for economic growth.  Such growth can substantially ease
the pressures of an aging population in which relatively fewer
workers will support more retirees. 

Addressing Social Security's financing issues is similarly essential. 
Reforms clearly must address the long-term actuarial balance of the
Social Security system and whether that balance is sustainable as
time goes on.  Potentially improving rates of return on workers'
contributions cannot in itself restore Social Security's solvency
without additional changes to the current system. 

Also, proposals should be examined for a number of design and
implementation issues and whether the new system would function
effectively at a reasonable cost.  Finally, the public will need to
be able to understand how a reformed Social Security system will be
financed and how benefits will be determined. 

Restoring Social Security's long-term solvency will require making
difficult choices involving many complex and sometimes conflicting
objectives.  Given the complexity of the program, its financing, and
how it fits in with the rest of the government and the economy as a
whole, the results and implications of any changes cannot be known
with certainty.  Improving rates of return has been one objective
that has received much attention in the Social Security reform
debate.  However, it is also one of the most complex and contentious
issues, and it is fraught with many key subtleties and
qualifications.  Moreover, it is just one of many important
considerations in finding the best approach to restoring Social
Security's long-term solvency.  While rates of return may continue to
receive much attention, they should be kept in careful perspective,
acknowledging both the inherent complexities of rate of return
analysis and the larger context of making trade-offs among several
other important objectives. 

(See figure in printed edition.)Appendix I
COMMENTS FROM THE SOCIAL SECURITY
ADMINISTRATION
============================================================ Chapter 5

(See figure in printed edition.)

GAO CONTACTS AND STAFF
ACKNOWLEDGMENTS
========================================================== Appendix II

GAO CONTACTS

Barbara D.  Bovbjerg, (202) 512-7215
Charles A.  Jeszeck, (202) 512-7036

STAFF ACKNOWLEDGMENTS

In addition to the persons named above, Ken Stockbridge, William
McNaught, and Francis P.  Mulvey made key contributions to this
report. 

BIBLIOGRAPHY
============================================================ Chapter 1

Aaron, Henry J.  "The Social Insurance Paradox." Canadian Journal of
Economics and Political Science, Vol.  32, No.  3 (Aug.  1966), pp. 
371-74. 

Advisory Council on Social Security.  Report of the 1994-1996
Advisory Council on Social Security, Vols.  1 and 2.  Washington,
D.C.:  Jan.  1997. 

Baker, Dean.  Saving Social Security with Stocks:  The Promises Don't
Add Up.  New York:  Twentieth Century Fund, 1997. 

Beach, William W., and Gareth G.  Davis.  Social Security's Rate of
Return.  Washington, D.C.:  Heritage Foundation, Jan.  15, 1998. 

Burtless, Gary, and Barry Bosworth.  Privatizing Social Security: 
The Troubling Tradeoffs, policy brief 14.  Washington, D.C.: 
Brookings Institution, 1997. 

Burtless, Gary, Barry Bosworth, and C.  Eugene Steuerle.  "Changing
Patterns of Lifetime Earnings:  What Do They Tell Us About Winners
and Losers From Privatization?" Paper presented at the First Annual
Joint Conference for the Retirement Research Consortium, "New
Developments in Retirement Research," Boston College Center for
Retirement Research and Michigan Retirement Research Center,
Washington, D.C., May 20-21, 1999. 

Chen, Yung-Ping, and Stephen C.  Goss.  "Are Returns on Payroll Taxes
Fair?" In Social Security in the 21st Century, Eric R.  Kingson and
James H.  Schulz, eds.  New York:  Oxford University Press, 1997. 

Cohen, Lee.  "Moneysworth Measures for Social Security." Paper
presented at the 5th International Conference on Social Justice
Research, Reno, Nevada, June 28, 1995. 

Cohen, Lee, Laurel Beedon, and Carlos Figueiredo.  A Critical Look at
Equity Investment in the 1994-1996 Advisory Council on Social
Security Recommendations, issue brief 30.  Washington, D.C.: 
American Association of Retired Persons, Public Policy Institute,
Apr.  1998. 

Council of Economic Advisers.  Economic Report of the President,
1999.  Washington, D.C.:  U.S.  Government Printing Office, Feb. 
1999. 

Duggan, James E., Robert Gillingham, and John S.  Greenlees. 
Progressive Returns to Social Security?  An Answer from Social
Security Records, research paper 9501.  Washington, D.C.:  U.S. 
Treasury Department, Nov.  1995. 

Duggan, James E., Robert Gillingham, and John S.  Greenlees. 
"Returns Paid to Early Social Security Cohorts." Contemporary Policy
Issues, Vol.  11 (Oct.  1993), pp.  1-13. 

Feldstein, Martin.  "The Case for Privatization." Foreign Affairs,
July-Aug.  1997, pp.  24-38. 

Feldstein, Martin, and Andrew Samwick.  "The Transition Path in
Privatizing Social Security." In Martin Feldstein, ed., Privatizing
Social Security.  Chicago:  University of Chicago Press, 1998.  Pp. 
215-64. 

Ferrara, Peter, and Michael Tanner.  A New Deal for Social Security. 
Washington, D.C.:  CATO Institute, 1998. 

Garrett, Daniel M.  "The Effects of Differential Mortality Rates on
the Progressivity of Social Security." Economic Inquiry, Vol.  33
(July 1995), pp.  457-75. 

Geanakoplos, John, Olivia S.  Mitchell, and Stephen P.  Zeldes. 
"Social Security's Moneysworth." In Olivia S.  Mitchell, Robert J. 
Myers, and Howard Young, eds., Prospects for Social Security Reform. 
Philadelphia:  University of Pennsylvania Press, 1999.  Pp.  79-151. 

Geanakoplos, John, Olivia S.  Mitchell, and Stephen P.  Zeldes. 
"Would a Privatized Social Security System Really Pay a Higher Rate
of Return?" In R.  Douglas Arnold, Michael Graetz, and Alicia H. 
Munnell, eds., Framing the Social Security Debate.  Washington, D.C.: 
Brookings Institution, 1998.  Pp.  137-56. 

Golob, John, and David Bishop.  "What Long Run Returns Can Investors
Expect from the Stock Market?" Economic Review (Federal Reserve Bank
of Kansas City), Vol.  82, No.  3 (third quarter 1997), pp.  5-20. 

Goodfellow, Gordon P., and Sylvester J.  Schieber.  "Simulating
Benefit Levels Under Alternative Social Security Reforms." In Olivia
S.  Mitchell, Robert J.  Myers, and Howard Young, eds., Prospects for
Social Security Reform.  Philadelphia:  University of Pennsylvania
Press, 1999.  Pp.  152-83. 

Harmelink, Philip J., and Janet F.  Speyrer.  "Social Security Rates
of Return and the Fairness of Benefits." CATO Journal, Vol.  14, No. 
1 (spring-summer 1994), pp.  37-54. 

Hieger, Melissa, and William Shipman.  Common Objections to a
Market-Based Social Security System:  A Response, SSP10.  Washington,
D.C.:  The Cato Project on Social Security Privatization, July 22,
1997. 

Hurd, Michael D., and John B.  Shoven.  "The Distributional Impact of
Social Security." In David A.  Wise, ed., Pensions, Labor, and
Individual Choice.  Chicago:  University of Chicago Press, 1985.  Pp. 
193-221. 

Kijakazi, Kilolo.  African Americans, Hispanic Americans, and Social
Security:  The Shortcomings of the Heritage Foundation Reports. 
Washington, D.C.:  Center on Budget and Policy Priorities, Oct.  5,
1998. 

Kollman, Geoffrey.  How Long Does It Take New Retirees to Recover the
Value of Their Social Security Taxes?, report 94-5 EPW.  Washington,
D.C.:  Congressional Research Service, Feb.  1, 1996. 

Leimer, Dean R.  Cohort-Specific Measures of Lifetime Net Social
Security Transfers, working paper 59.  Washington, D.C.:  Social
Security Administration, Office of Research and Statistics, Feb. 
1994. 

Leimer, Dean R.  "Guide to Social Security Money's Worth Issues."
Social Security Bulletin, Vol.  58, No.  2 (summer 1995), pp.  3-20. 

Menchik, Paul.  "Economic Status as a Determinant of Mortality Among
Black and White Older Men:  Does Poverty Kill?" Population Studies,
Vol.  47 (1993), pp.  427-36. 

Meyer, Charles, and Nancy Wolff.  "Intercohort and Intracohort
Redistribution under Old Age Insurance," Public Finance Quarterly,
Vol.  15, No.  3 (July 1987), pp.  259-81. 

Myers, Robert J., and Bruce D.  Schobel.  "An Updated Money's Worth
Analysis of Social Security Retirement Benefits." Transactions
(Society of Actuaries), Vol.  44 (1992), pp.  247-75. 

Olsen, Kelly A., and others.  How Do Individual Social Security
Accounts Stack Up?  An Evaluation Using the EBRI-SSASIM2 Policy
Simulation Model, issue brief 195.  Washington, D.C.:  Employee
Benefits Research Institute, Mar.  1998. 

Orszag, Peter R.  Individual Accounts and Social Security:  Does
Social Security Really Provide a Lower Rate of Return?  Washington,
D.C.:  Center on Budget and Policy Priorities, Mar.  1999. 

Panis, Constantijn W.A., and Lee A.  Lillard.  Socioeconomic
Differentials in the Returns to Social Security, working paper 96-05. 
Santa Monica, Calif.:  RAND, Feb.  1996. 

Shiller, Robert J.  Market Volatility.  Cambridge, Mass.:  MIT Press,
1989. 

Simons, Katerina.  "Risk Adjusted Performance of Mutual Funds." New
England Economic Review (Federal Reserve Bank of Boston), Sept.-Oct. 
1998, pp.  33-48. 

Steuerle, C.  Eugene, and Jon M.  Bakija.  Retooling Social Security
for the 21st Century.  Washington, D.C.:  Urban Institute Press,
1994. 

Thompson, Lawrence H.  Predictability of Individual Pensions, Ageing
Working Paper 3.5.  Paris:  Organisation for Economic Co-operation
and Development, 1997.  www.oecd.org/els/pdfs/SocialPolicy/ENG5.PDF. 

Thompson, Lawrence H.  "The Social Security Reform Debate." Journal
of Economic Literature, Vol.  21 (Dec.  1983), pp.  1425-67. 

U.S.  General Accounting Office.  Retirement Income:  Implications of
Demographic Trends for Social Security and Pension Reform. 
GAO/HEHS-97-81, July 11, 1997. 

U.S.  General Accounting Office.  Social Security:  Analysis of a
Proposal to Privatize Trust Fund Reserves.  GAO/HRD-91-22, Dec.  12,
1990. 

U.S.  General Accounting Office.  Social Security:  Capital Markets
and Educational Issues Associated With Individual Accounts. 
GAO/GGD-99-115, June 28, 1999. 

U.S.  General Accounting Office.  Social Security:  Criteria for
Evaluating Social Security Reform Proposals.  GAO/T-HEHS-99-94, Mar. 
25, 1999. 

U.S.  General Accounting Office.  Social Security:  Different
Approaches for Addressing Program Solvency.  GAO/HEHS-98-33, July 22,
1998. 

U.S.  General Accounting Office.  Social Security:  Individual
Accounts as an Element of Long-Term Financing Reform. 
GAO/T-HEHS-99-86, Mar.  16, 1999. 

U.S.  General Accounting Office.  Social Security:  Restoring
Long-Term Solvency Will Require Difficult Choices.  GAO/T-HEHS-98-95,
Feb.  10, 1998. 

U.S.  General Accounting Office.  Social Security:  The Trust Fund
Reserve Accumulation, the Economy, and the Federal Budget. 
GAO/HRD-89-44, Jan.  19, 1989. 

U.S.  General Accounting Office.  Social Security Financing: 
Implications of Government Stock Investing for the Trust Fund, the
Federal Budget, and the Economy.  GAO/AIMD/HEHS-98-74, Apr.  22,
1998. 

U.S.  General Accounting Office.  Social Security Reform: 
Administrative Costs for Individual Accounts Depend on System Design. 
GAO/HEHS-99-131, June 18, 1999. 

U.S.  General Accounting Office.  Social Security Reform: 
Implications of Annuities for Individual Accounts.  GAO/HEHS-99-160,
July 30, 1999. 

U.S.  General Accounting Office.  Social Security Reform: 
Implications for Women's Retirement Income.  GAO/HEHS-98-42, Dec. 
31, 1997. 

U.S.  General Accounting Office.  SSA Benefit Estimate Statement: 
Adding Rate of Return Information May Not Be Appropriate. 
GAO/HEHS-98-228, Sept.  2, 1998. 

*** End of document. ***