Social Security: Program's Role in Helping Ensure Income Adequacy
(30-NOV-01, GAO-02-62). 					 
								 
Before Social Security was enacted in 1935, at least half of	 
those 65 and older in the United States were financially	 
dependent upon others, including family members and public	 
assistance. Today, dependency on public assistance has dropped to
a fraction of its depression-era levels, and poverty rates among 
the elderly are now lower than for the population as a whole. At 
the same time, Social Security has become the single largest	 
source of retirement income for over 90 percent of these 65 and  
older. Reductions in promised benefits and increases in program  
revenues will be needed to restore the long-term solvency and	 
sustainability of the program. Within the program's current	 
structure, possible benefit changes might include changes to the 
benefit formula or reductions in cost-of-living increases. Among 
other options,	revenue increases might include increases in	 
payroll taxes or transfers from the Treasury's general fund. From
Social Security's inception, one of its fundamental objectives	 
has been to reduce the extent of dependency on public assistance 
programs, and the program's evolution has reflected that	 
objective. In 1972, automatic adjustments were introduced to	 
reflect increases in the cost of living. Other program changes	 
expanded the number of people receiving Social Security benefits,
such as gradually increasing coverage to larger portions of the  
workforce and extending eligibility to family members and	 
disabled workers. Other benefits programs such as Supplemental	 
Security Income (SSI), Medicare, and Medicaid, were instituted	 
that were also intended to help cover costs of living for the	 
aged and disabled. With regard to measuring income adequacy,	 
various measures help examines different aspects of this concept,
but no single measure can provide a complete picture. For various
subgroups of beneficiaries that have lower lifetime earnings,	 
poverty rates have also declined. While the Social Security	 
benefit formula favors lower lifetime earners, their lower	 
earnings and work histories can leave them with incomes below	 
poverty levels when they retire or become disabled. The outlook  
for future Social Security benefit levels and income adequacy	 
depend on how the program's long-term financing imbalance is	 
addressed, as well as on the measures used.			 
-------------------------Indexing Terms------------------------- 
REPORTNUM:   GAO-02-62						        
    ACCNO:   A02517						        
  TITLE:     Social Security: Program's Role in Helping Ensure Income 
Adequacy							 
     DATE:   11/30/2001 
  SUBJECT:   Income maintenance programs			 
	     Income statistics					 
	     Social security benefits				 
	     Projections					 
	     Medicare Program					 
	     Medicaid Program					 
	     Old Age Survivors and Disability			 
	     Insurance Programs 				 
								 
	     Social Security Program				 
	     Supplemental Security Income Program		 
	     Medicare Hospital Insurance Program		 
	     Old Age Assistance Program 			 
	     Old Age Insurance Program				 

******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO Testimony.                                               **
**                                                              **
** 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.                                         **
**                                                              **
******************************************************************
GAO-02-62
     
Report to the Chairman, Subcommittee on Social Security, Committee on Ways
and Means, House of Representatives

United States General Accounting Office

GAO

November 2001 SOCIAL SECURITY Program?s Role in Helping Ensure Income
Adequacy

GAO- 02- 62

Page i GAO- 02- 62 Social Security and Income Adequacy Letter 1

Results in Brief 3 Background 6 Social Security Has Focused on Reducing
Dependency From its

Inception 7 No Single Measure Gives Complete Picture of Income Adequacy 11
Adequacy of Beneficiaries? Total Income Has Generally Increased

Over Time 14 Income Adequacy in the Future Will Depend on Extent and Nature

of Program Changes 24 Concluding Observations 36 Agency Comments 39

Appendix I Descriptions of Adequacy Measurement Methods 40 Current Poverty
Thresholds 46 Experimental Poverty Thresholds 48 Family Budgets 50 Family
Expenditures 54 Material Hardship 55 Median Family Income 57 One- Half
Median Family Income 57 Per Capita Personal Income 58 Public Assistance 60
Public Opinion 62 Earnings Replacement Rates 64

Appendix II Characteristics of Adequacy Measures Used in Our Analyses 67

Current Poverty Thresholds 67 Median Family Income and 50 Percent of Median
Family Income 70 Public Assistance 71 Earnings Replacement Rates 72

Appendix III Defining Benchmarks for Analysis of Future Benefits 73 Criteria
73 Tax- Increase- Only Benchmark Policies 74 Benefit- Reduction- Only
Benchmark Policies 74 Contents

Page ii GAO- 02- 62 Social Security and Income Adequacy Appendix IV
Description of Approaches Used to Forecast Future

Benefits 80 MINT 80 GEMINI 81 Brief Comparison of the Models 83 Illustrative
Workers 84

Appendix V GAO Contacts and Staff Acknowledgments 87 GAO Contacts 87
Acknowledgments 87

Tables

Table 1: Summary of Benchmark Policy Scenarios 26 Table 2: Overview of the
Purpose, Construction, and Other

Dimensions of 11 Methods To Measure Income Adequacy 42 Table 3: Summary of
Benchmark Policy Scenario Parameters 79 Table 4: Comparison of GEMINI Data
to Published Social Security

Data 83 Table 5: Median Annual Social Security Income for Individuals in

2020 for the 1955 Cohort Only by Marital Status: A Comparison of MINT and
GEMINI 84 Table 6: Distribution of Actual Workers Retiring in 1999 by Level,

Relative to PIA Levels for Hypothetical Steady Earnings Cases Retiring in
1999 86

Figures

Figure 1: Dependence of Elderly on Federal Means- Tested Cash Benefits Has
Declined as More Receive Social Security 15 Figure 2: Poverty Rates for
Elderly Have Declined Faster Than for

Other Groups 16 Figure 3: Median Income of the Elderly Has Grown Faster Than
the

Poverty Threshold But Not Faster Than the One- Half Median Income Threshold
(Single Person Households Aged 65 and Older) 17 Figure 4: Historical Social
Security Replacement Rates for

Illustrative Workers 19 Figure 5: Share of Elderly Receiving Other Sources
of Retirement

Income Has Also Grown 21

Page iii GAO- 02- 62 Social Security and Income Adequacy

Figure 6: Certain Subgroups of the Elderly More Likely to be in Poverty
(1999) 23 Figure 7: Increasing Retirement Age Diminishes Replacement

Rates Under Current Law 28 Figure 8: Adequacy Outlook Depends on Adequacy
Standard Used

(Retired Workers, All Retiring at Age 62, Under TaxIncrease Only Benchmark)
30 Figure 9: Adequacy of Income in Future Depends on Benefit

Reduction Approach (Retired Workers, All Retiring at Age 62) 32 Figure 10:
Benefit Reduction Approaches Have Different Effects on

Different Benefit Levels (Retired Workers Born in 1985, All Retire at Age
62) 34 Figure 11: Effects of Benefit Reductions on Various Groups Reflect

Varying Effects by Income Levels 35

Abbreviations

AIME Average Indexed Monthly Earnings BLS Bureau of Labor Statistics CEX
Consumer Expenditure Survey CPS Current Population Survey GAO General
Accounting Office HEW Health, Education, and Welfare MBR Master Beneficiary
Records MINT Model of Income in the Near- Term MTR Maintain Tax Rates NAS
National Academy of Science NIPA national income and product accounts OAA
Old- Age Assistance OAI Old- Age Insurance OASDI Old- Age, Survivors, and
Disability Insurance PIA Primary Insurance Amount PENSIM Pension Policy
Simulation Model PSG Policy Simulation Group SSA Social Security
Administration SSI Supplemental Security Income

Page 1 GAO- 02- 62 Social Security and Income Adequacy

November 30, 2001 The Honorable E. Clay Shaw, Jr. Chairman Subcommittee on
Social Security Committee on Ways and Means House of Representatives

Dear Mr. Chairman: Before Social Security was enacted in 1935, at least half
of those 65 and older in the United States were financially dependent upon
others, including family members and public assistance. 1 Today, dependency
on public assistance has dropped to a fraction of its depression- era
levels, and poverty rates among the elderly 2 are now lower than for the
population as a whole. At the same time, Social Security has become the
single largest source of retirement income and provides income for over 90
percent of those 65 and older. However, in the future, Social Security?s
role could change. Reductions in promised benefits and/ or increases in
program revenues will be needed to restore the long- term solvency and
sustainability of the program. Within the program?s current structure,
possible benefit changes might include changes to the benefit formula or
reductions in cost- of- living increases, among other options; revenue
increases might include increases in payroll taxes or transfers from the
Treasury?s general fund. 3 The various approaches to restoring solvency and
sustainability will have different effects on benefit levels.

To gain a better understanding of how Social Security benefits help ensure
adequate income, you asked us to address: (1) how concern over income
adequacy has shaped the Social Security program; (2) what measures help to
examine income adequacy; (3) how income adequacy has changed over

1 S. Rep. No. 74- 628 at 4 (1935). 2 In this report, we use ?elderly? to
refer to persons aged 65 and older. 3 Also, some proposals would change the
structure of the program to incorporate a system of individual accounts.
Many such proposals would reduce benefits under the current system and make
up for those reductions to some degree with income from the individual
accounts. Individual account proposals also try to increase revenues, in
effect, by providing the potential for higher rates of return on account
investments than reserve funds would earn under the current system.

United States General Accounting Office Washington, DC 20548

Page 2 GAO- 02- 62 Social Security and Income Adequacy

time, especially for different groups of beneficiaries; and (4) under the
current system, how Social Security benefit levels and adequacy might change
in the future.

To assess how concern over income adequacy has shaped Social Security, we
examined the legislative history, reviewed the applicable literature, and
interviewed key individuals familiar with the program?s history. To identify
measures that help examine income adequacy, we reviewed literature and
interviewed experts. (See app. I and II.) To examine historical adequacy
trends, we analyzed published government data on historical income and
benefits, examined benefits for illustrative workers, and examined detailed
data for the current population. To illustrate the range of potential future
benefit levels under the current system, we developed and used alternative
benchmark policies that achieve 75- year solvency either by only increasing
payroll taxes or by only reducing promised benefits. Our tax- increase- only
benchmark uses a one- time permanent increase in payroll taxes beginning in
2002, though any policy that achieved solvency only by increasing taxes
would have identical effects on adequacy. Our benefit- reduction- only
benchmarks gradually phase in reductions from 2005 to 2035; the reductions
are accomplished by changing the parameters of the benefit formula in
various ways to achieve either proportional or progressive reductions. (See
app. III.) As we agreed, we did not examine any reforms that would
restructure the current system, for example by creating a new system of
individual accounts. Also, we did not report on any measures of individual
equity, which measure the relationship between contributions and benefits,
such as rates of return. Any of our benchmarks would have effects on
individual equity as well as income adequacy. With our benchmarks, we used
two models to simulate benefit levels for future populations of
beneficiaries, using the 2001 Social Security Trustees? intermediate
assumptions. 4 At our request, the Social Security Administration?s (SSA)
Office of Policy applied our benchmark policies to its Model of Income in
the Near- Term (MINT); and under contract to us, the Policy Simulation Group
applied our benchmark policies to its GEMINI model. Both models are designed
to examine a representative sample of future beneficiaries. (See app. IV.)
We conducted

4 For the annual report of the Board of Trustees for the Social Security
Trust Funds, SSA actuaries project future revenues and benefits. For these
projections, they use alternative assumptions regarding economic and
demographic trends, including average earnings, mortality, fertility, and
immigration. The intermediate assumptions represent the board?s best
estimate of future trends.

Page 3 GAO- 02- 62 Social Security and Income Adequacy

our review from March through November 2001 in accordance with generally
accepted government auditing standards.

From Social Security?s inception, one of its fundamental objectives has been
to reduce the extent of dependency on public assistance programs, and the
program?s design has reflected that objective as it has evolved. Over time,
public discussion has included the role that Social Security play in as
helping to ensure adequate incomes for the beneficiary population. According
to our analysis of the program?s history, ?adequacy? was never explicitly
defined, although the Congress expected that benefits would eventually
provide more than a minimal subsistence in retirement for fulltime, full-
career workers. Benefits were set in a way that focused especially on
replacing some portion of pre- retirement earnings. In 1950, the first
benefit increases were enacted since benefits had first been paid;
government reports had called for increases, partly out of concern that
inflation had eroded the purchasing power of benefits. Then, for more than
20 years, benefits were increased periodically but on an ad hoc basis
reflecting judgments of how high they should be and taking into account
economic and fiscal conditions. In 1972, automatic adjustments were
introduced to reflect increases in the cost of living. Other program changes
expanded the number of people receiving Social Security benefits, such as
gradually increasing coverage to larger portions of the workforce and
extending eligibility to family members and disabled workers. Other benefit
programs such as Supplemental Security Income (SSI), Medicare, and Medicaid,
were instituted that were also intended to help cover costs of living for
the aged and disabled. Also, as House and Senate reports from 1939 noted,
individual savings and other resources were also expected to play a
significant role. 5

With regard to measuring income adequacy, various measures help examine
different aspects of this concept, but no single measure can provide a
complete picture. Three examples illustrate the variety of approaches.

 Dependency rates measure what proportion of the population depends on
others for income support, or more specifically on government income support
programs such as SSI. Such rates reflect Social Security?s goal of reducing
dependency on public assistance.

5 H. R. Rep. No. 76- 728 at 5 (1939); S. Rep. No. 76- 734 at 5 (1939).
Results in Brief

Page 4 GAO- 02- 62 Social Security and Income Adequacy

 Poverty rates measure what proportion of the population have incomes below
the official poverty threshold, which is just one of many adequacy standards
used in similar rate calculations. The poverty threshold provides a minimal
standard of adequacy, but moderate standards have also been developed over
the years for similar rate calculations, such as moderate family budgets for
elderly couples. Each standard reflects a different outlook on what adequacy
means.

 Earnings replacement rates measure the extent to which retirement income
replaces pre- retirement income for particular individuals and thereby helps
them maintain a pre- retirement standard of living. This measure reflects
the way Social Security?s benefit formula is designed to replace earnings,
and financial advisors commonly use it for individual retirement planning.
Still, a very low earner could have a high replacement rate and still have
very low income, while a high earner could have a low replacement rate and
live quite comfortably.

For any of these measures, the meaning of a given value of the measure is
not clear. For example, what value of a dependency or poverty rate is
considered low enough and what replacement rate is considered high enough
are quite subjective. Moreover, all of these types of measures depend
significantly on what types of income are counted, such as beforeor after-
tax income or noncash benefits such as Medicare and Medicaid. As a result,
the measures are most useful for making comparisons, whether over time,
across different subpopulations, or across different policy scenarios.

The adequacy of income for the elderly has increased substantially by
several measures. For example, since 1940, the percentage of the elderly
receiving public income support has fallen from 22 to 9 percent, and since
1959, poverty rates for the elderly have fallen from 35 to 10 percent.
However, the greatest changes occurred in the first 20 to 40 years of the
program. In the last 20 years, the extent of any change depends on the
measure used; change has slowed or even stopped according to some measures.
Coinciding with these changes have been changes in the various sources of
retirement income. Most notably, Social Security is now the largest single
source of retirement income, providing more than 40 percent of aggregate
income to more than 90 percent of elderly households. Also, coverage under
employer pensions has increased, retirement savings have increased, work
patterns of the elderly have changed, and major noncash benefits have become
widely available, such as Medicare. For various subgroups of beneficiaries
that tend to have lower lifetime earnings, poverty rates have also declined
considerably. However, high proportions of such beneficiary groups continue
to have incomes below the poverty

Page 5 GAO- 02- 62 Social Security and Income Adequacy

threshold. For example, poverty among blacks aged 65 and older has fallen
from over 50 percent to just over 20 percent, but at that level, it remains
more than twice as high as for all the elderly. While the Social Security
benefit formula favors lower lifetime earners, their lower earnings and work
histories nevertheless can leave them with incomes below poverty when they
retire or become disabled.

As our analysis illustrates, the outlook for future Social Security benefit
levels and income adequacy generally will depend on how the program?s long-
term financing imbalance is addressed, as well as on the measures used. Even
without reductions in currently promised benefits, monthly benefit levels
could decrease as the program?s full retirement age increases under current
law, depending on the retirement decisions of future retirees. Still, in
that same case, the outlook varies for different adequacy measures. For
example, using adequacy measures that reflect wage increases, such as median
family incomes, adequacy should not change markedly from today because the
current benefit formula is also designed to reflect wage increases. 6 In
contrast, using measures that reflect only price increases such as the
official poverty threshold, adequacy would appear to improve considerably if
wages increase faster than prices on average, as the Social Security
trustees? report assumes they will. In that case, benefits would in turn
increase faster on average than the threshold. 7 Moreover, the outlook would
depend on how any benefit reductions might be made. As our analysis
illustrates, benefit reductions that do more to protect lower earners would
increase the risk of poverty less than ones that do not. In turn, such
progressive benefit reduction approaches would increase the risk of poverty
less for beneficiary groups that tend to have low earnings, such as
minorities and women. Finally, the adequacy of total incomes will depend on
what happens to other sources of income, such as employer- sponsored
pensions, individual savings, employment earnings, SSI, and noncash benefit
programs such as Medicare.

6 Some reform proposals would revise the benefit formula to reflect price
increases instead of wage increases. 7 The outlook would also depend on how
workers respond to policy changes. For example, faced with increased Social
Security taxes, workers might save less and have less retirement income from
saved assets as a result.

Page 6 GAO- 02- 62 Social Security and Income Adequacy

Title II of the Social Security Act, as amended, establishes the Old- Age,
Survivors, and Disability Insurance (OASDI) program, which is generally
known as Social Security. It provides cash benefits to retired and disabled
workers and their dependents and survivors. Workers become eligible when
they have enough years of earnings covered under Social Security; they and
their employers pay payroll taxes on those covered earnings. In 1999, about
96 percent of all U. S. jobs are covered , and over 40 million people
received $386 billion in benefits, which averaged about $800 per month or
$9600 per year. The benefit formula takes into account the lifetime history
of earnings and replaces a higher percentage of earnings for lower earners
than for higher earners.

In contrast, the Supplemental Security Income (SSI) program provides income
support to eligible aged and disabled persons regardless of their earnings
history. Funds for SSI benefits come from general revenues, not payroll
taxes. Persons with income or assets that exceed certain thresholds are not
eligible for SSI. In 2001, the maximum federal SSI monthly benefit is $531
for an individual and $796 for a couple and is reduced to reflect receipt of
other income, including OASDI benefits. In December 1999, over 6.5 million
people received federally- administered SSI benefits; of these about 6.3
million received a federal benefit and about 2.4 million received an SSI
state supplemental benefit. In December 1999, the average monthly federal
benefit was $342; the average monthly federally- administered state
supplement was $111.

Medicare?s Hospital Insurance benefits are generally provided automatically
and free of premiums to persons aged 65 or older who are eligible for Social
Security or Railroad Retirement benefits. Similarly, individuals who have
been entitled to Social Security or Railroad Retirement disability benefits
for at least 24 months are entitled to such benefits. In addition,
Supplementary Medical Insurance benefits are available on a voluntary basis
with a monthly premium to cover doctors? services, tests, and a variety of
over medical services. In 1999, Medicare paid a total of $210 billion in
benefits and covered nearly 40 million enrollees. According to current
estimates, the Hospital Insurance trust fund will be exhausted in 2029.

Medicare beneficiaries and others who have low incomes and limited resources
may also receive help from the Medicaid program. In 1998, Medicaid made $142
billion in payments for medical services for 41 million recipients, of which
about 4 million were aged 65 or older and 6.6 million were disabled. Average
payments were about $10,200 for the aged and $9,100 for the disabled.
Roughly $32 billion was paid for nursing facilities. Background

Page 7 GAO- 02- 62 Social Security and Income Adequacy

According to the OASDI Trustees? 2001 intermediate, or best- estimate,
assumptions, Social Security?s cash flow is expected to turn negative in
2016. In addition, all of the accumulated Treasury obligations held by the
trust funds are expected to be exhausted by 2038. Social Security?s longterm
financing shortfall stems primarily from the fact that people are living
longer while having fewer children. As a result, the ratio of workers paying
into the system to beneficiaries has been falling and is projected to
decline from 3.3 today to about 2 by 2030. 8

To address the program?s long- term financing shortfall, a variety of
proposals have been offered. In choosing among proposals, we have suggested
that policymakers should consider three basic criteria:

 the extent to which the proposal achieves sustainable solvency and how the
proposal would affect the economy and the federal budget;

 the balance struck between the twin goals of individual equity 9 (rates of
return on individual contributions) and income adequacy (level and certainty
of benefits); and

 how readily such changes could be implemented, administered, and explained
to the public.

Moreover, as we have said, reform proposals should be evaluated as packages
that strike a balance among individual reform elements and important
interactive effects. Overall evaluation of each proposal depends on the
weight individual policymakers place on each criterion. 10

From its inception, Social Security was intended to help reduce the extent
of dependency on public assistance programs. As it has evolved, the
program?s design has reflected that objective. Over time, that objective has
come to be stated more broadly as helping ensure adequate incomes. While the
Congress has never explicitly defined what constitutes an adequate level of
benefits, it stated as early as 1939 that its objective was

8 See Retirement Income: Implications of Demographic Trends for Social
Security and Pension Reform (GAO/ HEHS- 97- 81, July 11, 1997). 9 For a
discussion of individual equity, see Social Security: Issues in Comparing
Rates of Return with Market Investments (GAO/ HEHS- 99- 110, Aug. 5, 1999).
10 See Social Security: Criteria for Evaluating Reform Proposals (GAO/ T-
HEHS- 99- 94, Mar. 25, 1999) and Social Security: Evaluating Reform
Proposals

(GAO/ AIMD/ HEHS- 00- 29, Nov. 4, 1999). Social Security Has

Focused on Reducing Dependency From its Inception

Page 8 GAO- 02- 62 Social Security and Income Adequacy

to ?afford more adequate protection.? However, individual savings and other
resources were also expected to play a significant role.

In response to the grave economic problems of the Great Depression,
President Franklin Roosevelt created the Committee on Economic Security in
1934 to study the economic insecurity that individuals faced and to make
recommendations on how to address it. The committee?s recommendations became
the basis of the Social Security Act of 1935, which created several programs
to meet the needs of different population groups, including the aged. Two
programs specifically addressed the aged population- Title I?s Old- Age
Assistance (OAA) program and Title II?s OldAge Insurance (OAI) program. 11
OAA benefits, administered by the states with both state and federal funds,
were intended to provide immediate cash income for millions of elderly
persons without sufficient income for a decent subsistence. OAI benefits,
administered by the federal government and funded by equal contributions
from both employees and employers, were designed for younger workers to
build up their rights to annuities in old age gradually. In effect, the
contributions would purchase insurance to protect workers against lost wages
when they became too old to work.

In debating the creation of OAI, proponents made a variety of arguments in
its favor and mentioned several objectives that it would serve. Of these,
helping reduce dependency on public assistance was arguably the most
fundamental. The Congress was clearly concerned that an increasing number of
people were becoming dependent upon the public for their well- being; Social
Security would eventually provide benefits that workers and their employers
would pay for. Other objectives that were discussed in the debate included
stimulating the economy by providing cash income that people would spend and
opening up jobs for younger workers by freeing older workers to retire.

Implicitly, the Congress designed Social Security benefits with a focus on
replacing lost wages. The original formula computed benefits as a percentage
of lifetime wages covered under the program in a way that favored lower
earners, reflecting a special concern for their benefit levels. 12 Social
Security?s framers had targets in mind for benefit levels, but

11 In the act, Title I was named ?Grants to States for Old- Age Assistance,?
and Title II was named ?Federal Old- Age Benefits.? 12 When the formula was
changed in the 1939 amendments to replace monthly wages instead of lifetime
wages, the new formula was also progressive. Social Security Intended to

Insure Against Lost Wages

Page 9 GAO- 02- 62 Social Security and Income Adequacy

these targets did not appear to be based on any type of scientific research
or data analysis. While the Congress made no assertions concerning whether
the resulting benefits would be adequate, Senate and House reports stated
respectively that under Social Security it would be possible to provide
?more than reasonable subsistence? and ?not merely subsistence but some of
the comforts of life.? The House report also noted that the ?benefits
provided for workers who have been employed during substantially all their
working life will probably be considerably larger than any Federal- aided
State pensions could be.?

As time passed, the Social Security program grew and evolved. Even before
the first monthly benefits were paid in 1940, the Congress enacted
amendments in 1939 to ?afford more adequate protection to more of our
people,? as House and Senate committee reports put it. Changes to benefit
levels, coverage of earnings, and eligibility are especially relevant to the
program?s adequacy goals. In addition, the introduction of new programs
addressed specific needs, such as covering health care costs, promoting
retirement saving, and promoting and protecting employer- sponsored
pensions.

Changes to monthly benefit levels came in different forms at different
times. From 1939 until 1950, there were no changes to the benefit formula,
and benefit levels, after adjusting for inflation, fell as a result. 13 The
1948 Trustees? Report expressed concern that inflation was diminishing the
adequacy of Social Security benefits and presented a chart showing the
decline in inflation- adjusted benefit levels. The 1950 amendments to the
Social Security Act increased benefit levels substantially. Then, until
1972, periodic amendments made various ad hoc adjustments to benefit levels.
Economic prosperity, along with actuarial methods that often left the Trust
Funds with substantial surpluses, facilitated gradual growth of Social
Security benefit levels through these ad hoc adjustments.

In light of the steady growth of benefit levels, the 1972 amendments
instituted automatic adjustments to constrain the growth of benefits as well
as to ensure that they kept pace with inflation. Parameters of the

13 Each year?s group of new retirees would generally have higher average
wages than the previous years? groups (even though the maximum covered
earnings level did not change). Initial benefit levels increased somewhat as
a result, but with the tilt in the benefit formula, a larger and larger
portion of those wages were replaced at a lower marginal rate, and on net,
inflation eroded the purchasing power of those initial benefit levels.
Various Program Changes

Have Been Related to Income Adequacy

Page 10 GAO- 02- 62 Social Security and Income Adequacy

benefit formula were automatically adjusted to reflect inflation, and the
adjustments affected levels of benefits for both existing and new
beneficiaries. However, wages grew more slowly and prices grew more quickly
in the 1970s than they had historically. As a result, initial benefit levels
grew faster than intended. 14 The program?s first benefit reductions in 1977
attempted to correct for those unintended consequences of the 1972
amendments, and the resulting pattern of increasing and then declining
benefit levels has become known as the ?notch.? 15 In the process, the
benefit formula was redesigned so that initial benefits would generally
increase with wages for each new group of beneficiaries. As individuals
aged, annual cost- of- living adjustments would then increase benefits to
keep pace with inflation. In effect, the new formula?s design would
generally replace pre- retirement wages for similar individuals at a
consistent rate across age groups. Implicitly, this episode illustrates the
focus of the Congress on replacing wages and also identifies benefit levels
that the Congress considered higher than they intended. The only other
significant benefit reductions came in 1983 when the Congress delayed cost-
of- living adjustments primarily to address short- term financing problems
and gradually increased the retirement age to address long- term financing
problems. 16

In addition, a variety of other types of program changes had effects on the
extent to which the program helped ensure income adequacy. As amendments
extended Social Security coverage to more jobs, more workers would
eventually receive benefits. Initially, Social Security only covered the
roughly 60 percent of workers in ?commerce and industry? whose wages could
most easily be taxed and tracked. As the program matured, coverage was
gradually extended to new groups of workers, such as farm workers, domestic
workers, self- employed workers, and

14 This unintended benefit growth occurred because larger and larger
portions of slowergrowing average monthly wages were replaced at a higher
marginal rate, which was intended to benefit lower earners but increased
with faster- growing prices.

15 See Social Security: GAO?s Analysis of the Notch Issue (GAO/ T- HEHS- 94-
236, Sept. 16, 1994). 16 Increasing the retirement age might not be
considered a benefit reduction because individuals are living longer and
therefore collecting benefits for more years. However, for a person retiring
at a given age, benefit levels are reduced because the actuarial reduction
for early retirement is larger. See Social Security Reform: Implications of
Raising the Retirement Age (HEHS- 99- 112, Aug. 27, 1999). Also, there have
been changes that tightened eligibility standards for disability benefits,
but for those who are eligible, benefit levels were not affected.

Page 11 GAO- 02- 62 Social Security and Income Adequacy

some federal and state government workers. Today, Social Security covers
about 96 percent of all U. S. jobs .

Moreover, various amendments extended eligibility to more types of
beneficiaries. Under the 1935 act, only some retired workers were to receive
benefits. The 1939 amendments extended benefit eligibility to wives, widows,
children, and dependent parents age 65 and older. The 1956 amendments
extended eligibility to disabled workers, and the 1958 amendments extended
eligibility to their dependents. 17 In addition, the 1956 and 1961
amendments extended eligibility to women and men, respectively, at age 62
for retired workers, spouses, and widow( er) s, though worker and spouse
benefits taken before the full retirement age were reduced to take account
of the longer period over which they would be paid.

Outside Social Security, other legislation also addressed income adequacy in
various ways. Other benefit programs were created and changed to help ensure
adequate incomes. In 1965, Medicare and Medicaid were created to alleviate
the historically increasing strains on incomes from paying for health care.
In 1972, Title XVI?s Supplemental Security Income replaced Title I?s Old-
Age Assistance.

Moreover, as both House and Senate reports noted in 1939, ?individual

savings and other resources must continue to be the chief reliance for
security.? Over the years, the Congress has enacted legislation to promote
employer- sponsored pensions and make them more secure. The Congress has
also enacted legislation to promote individual retirement savings and
encourage greater work- force participation by the aged and disabled.

Various measures have been developed to examine different aspects of income
adequacy, but no single measure offers a complete picture. A universally
accepted definition of ?income adequacy? does not exist; focusing on a
single measure would implicitly endorse the concept of adequacy it measures
while dismissing other concepts. Several examples of three broad types of
measures illustrate the range of relevant measures. Each measure has
characteristics that reflect different outlooks on the issue, including how
it is calculated, how it accounts for different types of

17 Other types of eligibility expansions included dependent parents,
disabled widows, and divorced spouses. No Single Measure

Gives Complete Picture of Income Adequacy

Page 12 GAO- 02- 62 Social Security and Income Adequacy

households, how it accounts for geographic variations, and how it is updated
over time. In addition, for any type of measurement, what types of income
are counted presents a key issue.

The first type of measure includes variations of dependency rates.
Dependency rates speak to Social Security?s fundamental objective of
reducing dependence on public assistance programs, such as SSI or state and
local general assistance programs. Some sources have reported dependency
rates over the years that reflect a wide variety of sources of income
support while other sources report rates that only reflect federal income
support programs. For example, as cited by congressional reports, the
dependency rate of over 50 percent of the elderly in the 1930s reflected
dependence on family members and private charities as well as public
assistance. Moreover, public assistance includes a variety of federal,
state, and local programs in addition to OAA and SSI. As a result of the
extensive effort required to identify all sources of support, the most
readily available annual dependency rate data reflects only dependency on
OAA and SSI. Accounting for different types of households, geographic
variations, and changes over time are not critical concerns in calculating
the rates because the rates simply measure whether individuals or households
receive public assistance, wherever they are, and whatever the eligibility
criteria happen to be. However, the issues of geography and eligibility do
raise questions about how to interpret the rates because benefit standards
and eligibility provisions for public assistance programs have varied
considerably by location and over time.

The second type of measure includes rates that express the percentage of the
population that has incomes below a given adequacy standard. For example,
the poverty rate shows the percentage of individuals whose household income
falls below the official poverty thresholds, which attempt to specify an
income that would afford a minimal standard of living. Different thresholds
apply for different types and sizes of households but are the same for every
location in the country. The official poverty thresholds were originally
developed in 1963 and were built upon a government family food plan.
Initially, the thresholds were updated to reflect the change in the cost of
the food plan, but since 1969, they have been updated annually to reflect
changes in the Consumer Price Index (CPI). In 1969, the Bureau of the Budget
established the thresholds as the official definition of poverty for
statistical use in all executive departments.

The poverty threshold is only one of many adequacy standards that have been
developed over the years. Moreover, various government programs

Page 13 GAO- 02- 62 Social Security and Income Adequacy

and descriptive statistics use different percentages of the poverty
threshold, for example, 125 or 150 percent of poverty in determining
benefits or eligibility. 18 Some standards focus on determining the income
level needed for a moderate subsistence, not merely a minimal one. The bases
of the various standards include government- developed family budgets,
expenditure data, income data, and even public opinion polls. The various
adequacy standards have also used different approaches to capture household
and geographic variations and to reflect changes over time. A variety of
studies have evaluated the poverty threshold and explored possible changes
to it. (See app. II.)

The third type of measure, the replacement rate, speaks to Social Security?s
objective of replacing lost wages, which is implicit in the program?s
benefit formula. In contrast to other types of measures, it focuses on
whether retirement income is sufficient to maintain the standard of living a
given household enjoyed before retirement, not just meet some socially
defined standard of adequacy. Generally, it is calculated as the ratio of
retirement income in the first year of retirement to household income in the
year immediately preceding retirement. 19 However, the actual experience of
a given household could easily involve phased- in retirement or situations
where one spouse retires while the other continues to work. Such
irregularities present problems in interpreting replacement rates for actual
households. Still, these rates can be useful for demonstrating the effects
of program changes by focusing on illustrative workers with standardized
work experiences. With replacement rates, geographic variations and updating
the measure over time are not relevant issues because the household?s own
experience is the basis for the measure regardless of location or year.

All of these types of measures depend significantly on what types of income
are counted. Some dependency rates look only at specific sources of public
assistance, while others attempt to reflect all types of public assistance
and some even try to reflect dependency of private charities and family
members. In the case of poverty rates, one criticism has been

18 See Means- Tested Programs: Determining Financial Eligibility Is
Cumbersome and Can Be Simplified (GAO- 02- 58, Nov. 2, 2001). 19 SSA
typically calculates replacement rates as the ratio of initial Social
Security benefits to pre- retirement covered earnings. However, pension
analysts typically calculate replacement rates as the ratio of total
retirement income to total pre- retirement income. Total retirement income
would include sources such as employer- sponsored pensions and income from
saved assets.

Page 14 GAO- 02- 62 Social Security and Income Adequacy

that before- tax income is compared with thresholds based on after- tax
income. In the case of replacement rates, researchers have noted that the
measures of retirement and pre- retirement incomes should be consistent,
especially with respect to before- or after- tax status. Finally, a wide
range of noncash benefit programs, notably Medicare and Medicaid, also
support the standards of living of their beneficiaries though such benefits
are not always reflected in measures of income adequacy. For example,
replacement rates typically only consider cash income before and after
retirement. Also, noncash benefits are not included as income in determining
poverty status, and the living costs they support are not explicitly
reflected in the poverty threshold against which income is compared. In
particular, considerable debate surrounds how to treat medical care needs
and resources in measuring adequacy.

The adequacy of income for the elderly has generally increased since the
1930s, according to various measures. For example, dependence on public
assistance has fallen, as have poverty rates for the elderly. The largest
changes occurred in the first few decades of the program?s history;
improvements in the past 20 years have slowed or even stopped, depending on
the measure used. At the same time, Social Security has become the most
important source of income for the elderly and disabled. Savings and other
assets, employer- sponsored pensions, and earnings have also increased as
sources of income. Still, relatively high poverty rates remain for subgroups
that typically have low life- time earnings, whether for old- age or
disabled beneficiaries.

The dependency rate for the elderly has fallen from almost 22 percent in
1940 to about 6 percent in 1999, using a rate that only reflects OAA or SSI
benefits 20 and does not include dependency on relatives and friends.
Meanwhile, receipt of Social Security benefits among the elderly has grown
significantly from less than 1 percent to over 90 percent. (See fig. 1.) A
1938 Social Security Bulletin reported a dependency rate of 65 percent,
which included assistance to those who were totally or partially dependent
on friends and relatives. Among the elderly, OASDI beneficiaries outnumbered
OAA beneficiaries for the first time between 1950 and 1955 and, by 1960, a
majority received Social Security benefits.

20 In 1974, the federally administered Supplemental Security Income (SSI)
program replaced the state run OAA program. Adequacy of

Beneficiaries? Total Income Has Generally Increased Over Time

Two Measures Show Increased Income Adequacy for the Elderly

Page 15 GAO- 02- 62 Social Security and Income Adequacy

Since 1980, roughly 90 percent of the elderly have received benefits. The
rapid increase in the percentage receiving benefits and the eventual
leveling off illustrates the natural maturing of the Social Security system.
When monthly benefits were first paid in 1940, only those just turning 65
received benefits; older individuals were not eligible. As each year passed,
one additional age group was added to the beneficiary rolls, and more
individuals from the earlier, ineligible age groups died.

Figure 1: Dependence of Elderly on Federal Means- Tested Cash Benefits Has
Declined as More Receive Social Security

Note: Effective in 1974, SSI replaced OAA. SSI is a federal- state income
support program for the needy aged, blind, and disabled, as was OAA for the
aged. State and local governments also provide other forms of public
assistance.

Source: Annual Statistical Supplement to the Social Security Bulletin, 2000,
SSA.

Poverty rates for the elderly have also declined, from 35 percent in 1959 to
about 10 percent in 1999. (See fig. 2.) Since 1959, the elderly population
has experienced the greatest reduction in poverty rates, compared with
children 18 years and younger and adults aged 18 to 64. 21

21 Near poverty rates, which measure the percentage of the elderly whose
income falls between 100 and 125 percent of the poverty threshold, show
another important aspect of well- being. In 1999, 6 percent of the elderly
were near- poor.

9.3 21.7

90.8 0.7 5.8 0 10 20

30 40

50 60

70 80

90 100

1940 1950 1960 1970 1980 1990 1996 1998 Social Security

OAA/SSI with or without OASDI OAA/SSI Only Percent of elderly population
with income from source

Page 16 GAO- 02- 62 Social Security and Income Adequacy

Figure 2: Poverty Rates for Elderly Have Declined Faster Than for Other
Groups

Note: Data for years indicated by dashed lines were not available but are
available for 1959. Source: U. S. Bureau of the Census.

Examination of dependency and poverty rates for the elderly reveal that much
of the improvements occurred during the early decades of the program. (See
figs. 1 and 2.) The dependency rate declined at a much faster rate in the
early years until about 1965 when declines slowed to a more level trend.
Declines in the poverty rate for the elderly were most dramatic from 1959 to
1974 (more than 1 percent per year on average) and have continued since
then, but at a slower rate.

Using a different adequacy standard than the official poverty threshold,
adequacy appears to have changed little in recent years, using the
difference between median incomes and two alternative thresholds to examine
adequacy. For example, figure 3 shows the median total income and median
Social Security income for single persons aged 65 and older. It also shows
the official poverty threshold for such persons and one- half median income
for all single person households, which some researchers have used as an
alternative adequacy standard. The figure shows that the gap between median
total income and the poverty threshold consistently widens over the years
shown and widens more than the gap relative to the one- half median income
threshold. Moreover, median Social Security income is lower than the poverty
threshold in 1978 and higher by 1998 Extent of More Recent

Changes Depends on Measure Used

0 5

10 15

20 25

30 35

40 1959 1964 1969 1974 1979 1984 1989 1994 1999

Adults 65+ Children < 18 Adults 18-64 Percent of population below poverty

Page 17 GAO- 02- 62 Social Security and Income Adequacy

while it is consistently lower than the one- half median income threshold.
22 The primary difference between the two adequacy standards is that the
official poverty threshold is updated annually to reflect inflation while
median income grows at a rate similar to wage growth because wages are the
largest component of family income. Initial Social Security benefits, and
other sources of retirement income as well, also tend to grow at a rate
closer to wages than to prices. Historically, wages have generally grown
faster than prices, especially when their respective rates of change are
averaged over long periods of time.

Figure 3: Median Income of the Elderly Has Grown Faster Than the Poverty
Threshold But Not Faster Than the One- Half Median Income Threshold (Single
Person Households Aged 65 and Older)

Note: The official poverty threshold increases each year to reflect
inflation. Median income increases roughly by the rate of wage growth, which
has generally been faster than inflation.

Source: Income of the Population 55 and Older, 1998, 2000 Annual Statistical
Supplement, 1999 Money Income in the United States.

22 We use a single person household for illustration because most elderly
households are single person households. A similar pattern applies for two-
person households.

0 2,000

4,000 6,000

8,000 10,000

12,000 14,000

1978 1980

1982 1984

1986 1988

1990 1992

1994 1996

1998 Median total

income One-half median income threshold

Median Social Security income

Poverty threshold 2001 dollars

Page 18 GAO- 02- 62 Social Security and Income Adequacy

Over the same period that the income adequacy has increased for the elderly,
Social Security has become the single largest source of retirement income.
As discussed below, program changes have increased the real value of
benefits, and more and more elderly have received benefits as the program
has matured. Other sources of retirement income have also grown. Periods of
economic prosperity have contributed to the growth of all sources of
retirement income.

Social Security?s benefit levels have generally increased over the years.
Replacement rates for illustrative workers with steady lifetime earnings
histories show how changes in the benefit formula have affected benefit
levels because using such workers holds other factors equal that might also
have an effect. 23 (See fig. 4.) For example, using illustrative workers
filters out the effects of changes in the covered population or changes in
work and retirement patterns. The declining replacement rates during the
early years reflect that no benefit increases were enacted until 1950; fig.
4 also shows a sharp increase in replacement rates that coincides with the
1950 amendments. From 1950 until the early 1970s, replacement rates
fluctuated noticeably more from year to year than over other periods; this
pattern reflects the ad hoc nature of benefit increases over that period.
The rapid increases in the 1970s and the rapid decline in the early 1980s
reflects the effects of the notch and efforts to correct it. 24 The smoother
pattern that appears since that time reflects the automatic indexing of
benefits as enacted in 1977. 25 While there have been many changes in the
program for many reasons at different points in time, the replacement rates
experienced by today?s new retirees are notably consistent with the

23 We used the Social Security Administration?s low, average, and high
steady earners as our illustrative workers. Steady earners have earnings
equal to a constant percentage of Social Security?s Average Wage Index in
every year of their careers. Those percentages are 45, 100, and 160
respectively for low, average, and high earners. See appendix IV for further
discussion.

24 As discussed earlier, the 1972 amendments created automatic adjustments
to the Social Security benefit formula, but initial benefit levels grew
faster than intended. The 1977 amendments created a new formula and reduced
benefits to correct for those unintended consequences, and the resulting
pattern of increasing and then declining benefit levels has become known as
the ?notch.?

25 While this replacement rate analysis captures the effects of many program
changes, some of the patterns result from irregularities in the growth rates
of wages, which help determine the earnings histories of the illustrative
workers. Such irregularities are primarily responsible for the dip in the
late 1990s, which reflects the unusually high growth in wages during that
period. Social Security and Other

Types of Retirement Income Have Grown Over Time

Page 19 GAO- 02- 62 Social Security and Income Adequacy

levels that Social Security?s designers envisioned for a fully mature system
over 60 years ago.

Figure 4: Historical Social Security Replacement Rates for Illustrative
Workers

Notes: Replacement rates are the annual retired worker benefits at age 65
divided by the earnings in the previous year. Steady earners have earnings
equal to a constant percentage of Social Security?s Average Wage Index in
every year of their careers. Those percentages are 45, 100, and 160,
respectively for low, average, and high earners. Some fluctuations in the
graph do not result from program changes but rather from fluctuations in the
growth rate of wages, which helps determine the earnings histories of the
illustrative steady earners.

Source: SSA Office of the Chief Actuary.

At the same time that benefit levels have increased, so has the share of
elderly receiving benefits. This is also true of employer- sponsored
pensions, earnings, and income from saved assets. Like figure 1, figure 5
shows that the percentage of the elderly receiving Social Security benefits
has increased as dependence on public assistance has declined. Figure 5 also
shows that Social Security provides income to more elderly households than
any other source of retirement income, although other sources have also
increased in importance. The percentage of the elderly who receive income
from employer pensions increased from 5 percent in

0 10

20 30

40 50

60 70

80 1940 1950 1960 1970 1980 1990 2000

Low Steady Earner

Average Steady Earner

High Steady Earner Replacement Rate

Page 20 GAO- 02- 62 Social Security and Income Adequacy

1937 to 43 percent in 1998. 26 The percentage receiving income from saved
assets increased from about 15 percent to over 60 percent. The percentage
receiving earned income increased from 1937 to 1962 but dropped from 1962 to
1998. In addition to sources of cash income, noncash benefit programs that
did not exist in the 1930s now play a major role in supporting the standards
of living of Social Security beneficiaries. For example, Medicare is
available to all Social Security beneficiaries aged 65 and older and all
disabled beneficiaries after 24 months, among others.

26 While there have been significant improvements in pension receipt for
retired workers, pensions have failed to reach a significant portion of the
elderly. In 1998, 48 percent of retired persons reported that they had no
pension income of their own or from a spouse. Retired persons without
pension income were more likely to be single, female, less educated,
minority, and poor. About 21 percent of retired persons without pension
income had incomes below the federal poverty threshold, compared with 3
percent with pension income. They were also less likely than retired persons
with pension income to have income from other sources, except SSI and other
public assistance programs. For example, persons without pension income were
less likely to have income from assets (51 percent) than those with pension
income (78 percent). See Pension Plans: Characteristics of Persons in the
Labor Force Without Pension Coverage (GAO/ HEHS- 00- 131, Aug. 22, 2000).

Page 21 GAO- 02- 62 Social Security and Income Adequacy

Figure 5: Share of Elderly Receiving Other Sources of Retirement Income Has
Also Grown

Note: 1962 and 1998 data indicated above represent aged unit income that is
either the income of nonmarried persons or the sum of income within a
married couple.

Source: Income of the Aged ChartBook, 1998, ?Economic Status of the Aged.?
Social Security Bulletin, March 1938.

In addition to providing some income to nearly all elderly persons, Social
Security is the largest source of income for most. In 1998, Social Security
provided more than 50 percent of total income for 63 percent of aged
beneficiaries, and it was the only source of income for about 18 percent of
aged beneficiaries. Still, other sources of retirement income largely
determine who will have the highest retirement incomes. Elderly households
with the highest levels of income tend to have substantial income from
employer pensions, earnings from employment, and saved assets, while those
with the lowest incomes do not. For example, in 1996, 18 percent of all aged
beneficiary units without earnings from employment were poor as compared
with only 2 percent who received earnings. 27

27 See Retirement Income: Implications of Demographic Trends for Social
Security and Pension Reform (GAO/ HEHS- 97- 81, July 11, 1997).

0 69

90 15

54 63

5 18

43 14

36 21 65

14 5 0 10

20 30

40 50

60 70

80 90

100 1937 1962 1998 Social Security Asset income Pensions Earnings Public
assistance Percent of elderly population

Page 22 GAO- 02- 62 Social Security and Income Adequacy

Income adequacy has also improved substantially for specific subgroups of
beneficiaries, such as the very old (85+ years of age), minorities, women,
singles, widows, and the disabled. However, even with those improvements,
significant levels of poverty remain. This fact largely reflects that
lifetime earnings and access to other sources of retirement income tend to
be lower among such groups. Social Security is a major component of
retirement income for these sub- populations. For example, in 1998, when we
exclude Social Security income from total income, 67 percent of unmarried
women aged 85 and over have income that falls below the poverty line.

As figure 6 shows, poverty rates are higher than average for older age
groups, for women, for minorities, and those living alone. Those individuals
in older age groups are less likely to have pension benefits or income from
saved assets. Women also experience high rates of poverty as compared to
men. Of the 3. 2 million aged persons who were poor in 1999, 2.2 million
were women. Minorities such as Hispanics and blacks experience higher levels
of poverty than their white counterparts, as do unmarried women and women
living alone. Subgroups With Lower

Lifetime Earnings Remain Vulnerable

Page 23 GAO- 02- 62 Social Security and Income Adequacy

Figure 6: Certain Subgroups of the Elderly More Likely to be in Poverty
(1999)

Note: In 1999, the poverty rate for all those aged 65 and older was 10
percent. Source: Current Population Survey, 2000.

Poverty rates also vary by living situation. In 1999, elderly persons living
alone were more likely to be poor (14 percent of men and 20 percent of
women) than married couple families (6 percent). Of the 1.8 million elderly
poor who lived alone in 1999, about 1.5 million were women. Aged African-
Americans and Hispanics females living alone are most at risk for living in
poverty. In 1999, almost 58 percent of aged Hispanic females living alone
were in poverty, while 44 percent of aged African- American females were in
poverty.

Individuals who fall into more than one group with higher poverty rates are
especially at risk of poverty. For example, in 1998, 56 percent of unmarried
black females aged 85 and older were poor. Over 60 percent of unmarried
Hispanic females aged 75 to 84 were poor. In contrast, 21 percent of white
females aged 65 to 74 were poor, and poverty rates for the male counterparts
for each category were either less or there were too few cases available to
make an assessment.

6 21

16 20

14 23

8 20

14 10

9 0 5 10 15 20 25 Married couples

Unmarried women Unmarried men

Women living alone Men living alone

Blacks Whites

Hispanic Aged 85+

Aged 75-84 Aged 65-74

Percent of elderly households with income below poverty

Page 24 GAO- 02- 62 Social Security and Income Adequacy

Social Security provides an important source of income for the disabled. In
1999, disabled workers made up 11 percent of all OASDI beneficiaries. As
with the elderly, Social Security is a major component (38 percent) of
family income for disabled worker families. Also, 48 percent of disabled
worker families get half of their income or more from Social Security, while
6 percent have no other income. Unlike the elderly, however, earnings are an
equally large source of family income (38 percent) for disabled worker
families. 28

At 19 percent, poverty rates are nearly twice as high for the disabled as
for the elderly. Still, like the elderly, poverty rates for disabled workers
are higher for women, minorities, unmarried persons, and those living alone.
Of all disabled beneficiaries, 23 percent of females were poor compared with
15 percent of men. Fifteen percent of disabled beneficiaries were white, 31
percent were black, and 26 percent were Hispanic. Only 12 percent of the
disabled who lived with relatives lived in poverty, compared with 35 percent
who did not. Ten percent of disabled workers who were married lived in
poverty, compared with 27 percent who were not. Disabled workers who were
widowed, never married or divorced experienced poverty rates of 30, 25, and
24 percent, respectively.

The outlook for future Social Security benefit levels and thus their effect
on income adequacy generally will depend on how the program?s longterm
financing imbalance is addressed, as well as on the measures used. To
illustrate the range of possible outcomes, we developed benchmark policy
scenarios that either only increase taxes or only reduce benefits. Even
without new benefit reductions, our analysis shows that replacement rates
could decrease as the program?s full retirement age gradually continues to
increase under current law, depending on the retirement decisions of future
retirees. However, even with those reductions, our analysis shows that the
adequacy of retirement income would improve markedly using one adequacy
standard but change very little using another. Future benefit levels will
also depend on the extent and nature of any benefit reductions. More
progressive approaches to benefit reductions would result in greater
adequacy for lower- earning beneficiaries. In turn, adequacy for various
subgroups of beneficiaries would depend in turn on the earnings levels
typical of those subgroups. Moreover, the adequacy of

28 As a source of income, pensions make up only 7 percent of income for
disabled worker families. Income Adequacy in

the Future Will Depend on Extent and Nature of Program Changes

Page 25 GAO- 02- 62 Social Security and Income Adequacy

total incomes will depend on how individuals adjust their retirement
planning in reaction to any program changes 29 and on what happens to other
sources of cash and noncash income. In particular, Medicare also faces
serious long- term financing problems. However, our analysis does not
reflect interactions with other income sources but focuses on the effects of
changes in Social Security benefits, holding all else equal.

To illustrate a full range of outcomes that might result from alternative
approaches to restoring long- term solvency, we developed hypothetical
benchmark policy scenarios that would restore solvency over the next 75
years either by only increasing payroll taxes or by only reducing benefits.
Our tax- increase- only benchmark simulates ?promised benefits,? or those
benefits defined under current law, while our benefit- reduction- only
benchmarks simulate ?funded benefits,? or those benefits for which currently
scheduled revenues are projected to be sufficient. These benchmarks used the
program?s current benefit structure and the 2001 OASDI Trustees?
intermediate, or best- estimate, assumptions. The benefit reductions are
phased in between 2005 and 2035 to strike a balance between the size of the
incremental reductions each year and the size of the ultimate reduction. At
our request, SSA actuaries scored our benchmark policies and determined the
parameters for each that would achieve 75- year solvency. Table 1 summarizes
our benchmark policy scenarios. For our benefit reduction scenarios, the
actuaries determined these parameters assuming that disabled and survivor
benefits would be reduced on the same basis as retired worker and dependent
benefits. If disabled and survivor benefits were not reduced at all,
reductions in other benefits would be deeper than shown in this analysis.
(See app. III for more on our benchmark policy scenarios.)

29 However, it is difficult to anticipate how individual behavior would
change in response to policy changes, and none of our modeling approaches
were equipped to do so. For example, in response to benefit reductions,
workers could increase their retirement savings, depending on their capacity
to do so. On the other hand, in response to payroll tax increases, workers
might decrease retirement saving in the face of lower disposable incomes.
Still, those retirees with the lowest retirement incomes already receive
virtually no income from saved assets. It should also be noted that tax
increases would have effects on both the standard of living enjoyed during
working years and on the individual equity of benefits, or how benefits
relate to contributions. Benchmark Policy

Scenarios Illustrate Range of Possible Outcomes

Page 26 GAO- 02- 62 Social Security and Income Adequacy

Table 1: Summary of Benchmark Policy Scenarios Ultimate new benefit
reductions a (percent) Benchmark policy scenario Description Phase- in
period Minimum Maximum

Tax- increase- only Increases payroll taxes in 2002 by amount necessary to
achieve 75- year solvency (0.95 percent of payroll each for employees and
employers) b

Immediate 0 0 Proportional benefit reduction Reduces benefit formula factors
proportionally

across all earnings levels 2005- 2035 24 24 Progressive benefit reduction
Reduces benefit formula factors by smaller

proportion for lower earners 2005- 2035 11 33 Limited proportional benefit
reduction Does not reduce benefit formula factor at all for

lowest earners (below first formula bendpoint) and reduces them
proportionally for earnings above that level

2005- 2035; with smaller reduction in

first 10 years 0 34

a These benefit reduction amounts do not reflect the implicit reductions
resulting from the gradual increase in the full retirement age that has
already been enacted. b Any policy scenario that achieves 75- years solvency
only by increasing revenues would have the same effect on the adequacy of
future benefits in that promised benefits would not be reduced. Source:
GAO?s analysis.

We then modeled future benefit levels with these benchmarks and calculated a
variety of measures to look at income adequacy. However, we did not examine
any measures of individual equity, such as rates of return, which any of our
benchmark policies would also affect. We examined adequacy measures for
illustrative workers with different steady lifetime earnings histories, for
the entire beneficiary population, and also for different subgroups. To look
at representative samples for the beneficiary population and subgroups, we
used both SSA?s MINT model and the Policy Simulation Group?s GEMINI model.
The MINT model allows us to look at total retirement income in 2020 across
different age groups and races while the GEMINI model allows us to focus on
specific birth cohorts reaching age 62 in various years, which we selected
to look at long- term trends. As with any such simulation models, these
models simulate income using a combination of historical data from small
samples of the population and a variety of assumptions about future trends.
30 At their best, such models can only provide very rough estimates of
future incomes. Still, they can provide valuable comparisons over time and

30 While these models use sample data, our report, like others using these
models, does not address the issue of sampling errors. The results of the
analysis reflect outcomes for individuals in the simulated populations and
do not attempt to estimate outcomes for an actual population.

Page 27 GAO- 02- 62 Social Security and Income Adequacy

across alternative policy scenarios, holding all else equal. Thus, any
analysis should focus on such comparisons rather than on the literal values
of the estimates. (See app. IV for more on our modeling analyses.)

Our tax- increase- only benchmark illustrates that monthly benefit levels
could already decrease as the program?s full retirement age increases under
current law, depending on the retirement decisions of future retirees. In
turn, replacement rates would decrease by the same proportion because they
are defined as the annual benefit amount divided by the last year of
earnings. Figure 7 shows future replacement rates under our taxincrease-
only benchmark for a range of illustrative retired workers. The full
retirement age is the age at which full benefits are paid and historically
has been age 65. Under current law, the full retirement age is gradually
increasing, beginning with retirees born in 1938, and will reach 67 for
those born in 1960 or later. For workers who retire at a given age, an
increase in the full retirement age reduces monthly benefits because the
actuarial reduction for early retirement increases. For example, for workers
who will face a full retirement age of 67 and retire early at 65, monthly
benefits will be reduced actuarially by 13.3 percent while their benefits
would not have been reduced at all if the full retirement age had been kept
at 65. 31 Moreover, the 13.3 percent reduction applies to such workers
equally at all earnings levels. As a result, increasing the full retirement
age from 65 to 67 implies that replacement rates for illustrative low
earners would decline from 57 to 49 percent while for illustrative high
earners they would decline from 35 to 30 percent. 32 Therefore, under such a
proportional reduction, lower earners face a larger percentage- point
reduction than higher earners. Still, the effect of such reductions would be

31 Compared with earlier retirees, beneficiaries affected by the higher full
retirement age are expected to live longer on average. Therefore, on
average, they will collect benefits longer, and the reduction in their
lifetime benefits will not be as high as the reduction in their monthly
benefit amount. Still, the adequacy of income depends on how much people
have to spend on a monthly or annual basis, not on a lifetime basis, because
the longer individuals live, the longer they will need income to cover the
costs of living. Nevertheless, workers will have the ability to avoid the
monthly benefit reductions by waiting until their full retirement age to
retire. Such behavioral responses to program changes make it especially
difficult to anticipate future income adequacy, which such responses would
affect directly.

32 We use the illustrative steady earners defined by SSA?s Office of the
Chief Actuary; these steady earners have earnings equal to a constant
percentage of average wages in every year of their careers. Those
percentages are 45, 100, and 160, respectively, for low, average, and high
earners. See appendix IV for more on our illustrative steady earner
analysis. Monthly Benefits Could Be

Lower Even Without Benefit Reductions

28 GAO- 02- 62 Social Security and Income Adequacy

diminished to the extent that workers choose to retire later than today?s
workers do.

Figure 7: Increasing Retirement Age Diminishes Replacement Rates Under
Current Law

Notes: Replacement rates are the annual retired worker benefits at age 65
divided by the earnings in the previous year. Steady earners have earnings
equal to a constant percentage of Social Security?s Average Wage Index in
every year of their careers. Those percentages are 45, 100, and 160,
respectively, for low, average, and high earners. Taxable Maximum Earners
have earnings equal to the maximum earnings taxable under OASDI in each
year. Annual benefits at age 65, and replacement rates in turn, decline as
the full retirement age increases because actuarial reductions for early
retirement become higher as the gap between the actual retirement age and
the full retirement age increases. As shown in figure 7, the slight
increases in replacement rates from 2000 to 2005 do not result from any
program changes but rather from short- term fluctuations in the growth rate
of wages which helps determine the earnings histories of the illustrative
steady earners.

Source: GAO?s analysis using SSA?s ANYPIA program.

While replacement rate analysis suggests that income adequacy will decline
in the future, other ways of assessing adequacy suggest that it will change
little or even improve dramatically. The GEMINI model allows us to
illustrate this point best by showing changes over long periods of time.
Using our tax- increase- only benchmark policy, we calculated the percentage
of retired workers with Social Security benefits that fall below two
different adequacy standards- the official poverty threshold and onePage
Different Measures

Suggest Different Outlooks

0% 10%

20% 30%

40% 50%

60% 2000 2005 2010 2015 2020 2025 2030 2035 2040 2045 2050

Year earner attains age 65 Low steady

earner Average steady earner High steady earner

Taxable maximum earner Replacement rate at age 65

Page 29 GAO- 02- 62 Social Security and Income Adequacy

half median income. 33 The official poverty threshold is adjusted each year
to reflect inflation. In contrast, our simulation assumes that the one- half
median income threshold will grow at the same rate as Social Security?s
Average Wage Index, since wages are the largest component of family income.
Figure 8 shows that the percentage of retired workers 34 with benefits below
the poverty threshold drops dramatically over time while the percentage with
benefits below one- half median income changes very little. The difference
in these percentage measures simply reflects differences in the assumptions
underlying each adequacy standard. Since initial Social Security benefits
are designed to increase with wages, and wages are assumed to grow faster
than prices, benefit levels will grow faster than an adequacy standard that
grows only by prices. In contrast, benefits will grow at roughly the same
rate as a standard that grows by wages. In a fashion similar to poverty
rates, dependency rates would also decline relatively rapidly because they
focus on SSI benefit standards that increase with prices, not wages.

33 Using the poverty threshold with the GEMINI model does not allow us to
calculate a simulated ?poverty rate? because a poverty rate would reflect
total income, not just Social Security benefits, and our GEMINI analysis
only simulated future Social Security benefits and not total retirement
income. In contrast to GEMINI, the MINT model does simulate total income,
but only in the near term. However, the sensitivity of adequacy measures to
the standard used is greater over longer periods of time. Given the various
limitations of both models, the focus should not be on the specific
estimates of adequacy measures but rather the differences between them
across types of measures, across beneficiary groups, and across policy
scenarios.

34 Here, we used retired workers for simplicity. Our focus is on
illustrating the effect of using alternative adequacy standards holding
everything else equal; our focus is not on the type of beneficiary. We use
age 62 as the retirement age because most retired worker beneficiaries
retire at that age, and our GEMINI analysis required that we use a single
retirement age. Since spouses receive a benefit equal to 50 percent of the
retired workers benefit (unless they are entitled to a higher retired worker
benefit on their own earnings record), the percentage of spouse
beneficiaries below each adequacy standard would be much higher than for
retired workers. Moreover, the value of an adequacy standard differs for
different household types, so ideally benefit levels for retired workers,
their spouses, and all other types of beneficiaries would be compared to the
standard appropriate for their household type. However, this more complex
analysis is not necessary to illustrate the point about how the choice of
the adequacy standard itself drives the results. The complexities concerning
beneficiary and household type are yet another reason that our estimates
should not be interpreted as poverty rates.

Page 30 GAO- 02- 62 Social Security and Income Adequacy

Figure 8: Adequacy Outlook Depends on Adequacy Standard Used (Retired
Workers, All Retiring at Age 62, Under Tax- Increase Only Benchmark)

Notes: The rates shown in the graph are not poverty rates because they
reflect only Social Security income, not total income. Also, if we had
assumed that all workers retired at a higher age, the reduction for early
retirement would be lower, their Social Security benefits would be higher,
and each of the percentages in this chart would unambiguously be lower.
Still, the focus should not be on the specific values from the simulation
but on the difference in the pattern over time between the two thresholds,
which would not be notably different with other retirement age assumptions.
The official poverty threshold increases each year to reflect inflation. We
projected future poverty lines using OASDI Trustees? intermediate
assumptions for inflation. One- half median family income has been used as
an alternative adequacy threshold. It increases each year roughly by the
rate of wage growth. We projected future values of this threshold using the
OASDI Trustees? intermediate assumptions for the Average Wage Index.

Source: GAO?s analysis using the GEMINI model.

Future benefit levels and income adequacy will also depend considerably on
how any benefit reductions are made. Figure 9 shows that the percentage of
retired workers 35 with Social Security benefits below the official poverty
threshold would be greater under a proportional benefit reduction approach
than under a progressive benefit reduction approach.

35 Here again, we used retired workers for simplicity. In this case, our
focus is on the difference between benefit reduction approaches, holding
everything else equal. See footnote 26. Effects of Different Benefit

Reduction Approaches Would Vary By Earnings Levels

51 22

17 11 62

54 57 57 0 10

20 30

40 50

60 70

1935 (1997) 1955 (2017) 1970 (2032) 1985 (2047)

Birth year (age 62 year)

Poverty threshold One-half median income threshold Percent of cohort with
Social Security benefits below respective threshold

Page 31 GAO- 02- 62 Social Security and Income Adequacy

The difference between the two approaches grows slightly over time. 36 The
proportional benefit- reduction- only benchmark would reduce benefits by the
same proportion for all beneficiaries born in the same year. The progressive
benefit- reduction- only benchmark would reduce benefits by a smaller
proportion for lower earners and a higher proportion for higher earners.
Both benefit reductions benchmark policies would be phased in gradually from
2005 to 2035. The tax- increase- only (no benefit reduction) benchmark
estimates are shown for reference. Also, the figure shows that percentage of
workers with benefits below the poverty threshold would be slightly higher
in our simulations for those retiring in 2032 rather than 2017. This
reflects primarily that the benefit reductions in our benchmarks are more
fully phased in for the 2032 group. The declines in the percentages from the
2032 to 2047 retirement years largely reflects the effects of the disparity
between growth in wages and prices, as illustrated earlier; since the
benefit reductions are fully phased in by 2035, the last two age groups
experience nearly the same benefit reductions.

36 In contrast to using the poverty threshold, for our 1970 and 1985 birth
cohorts, Social Security benefits fall below 50 percent of median family
incomes for nearly all retired workers under our proportional benefit-
reduction benchmark and for all of them under our more progressive benefit-
reduction benchmarks.

Page 32 GAO- 02- 62 Social Security and Income Adequacy

Figure 9: Adequacy of Income in Future Depends on Benefit Reduction Approach
(Retired Workers, All Retiring at Age 62)

Notes: The rates shown in the graph are not poverty rates because they
reflect only Social Security income, not total income. For the 1935 birth
cohort, which reached age 62 in 1997, 51 percent of our simulated sample had
Social Security benefits below poverty, as shown in figure 8. Also, if we
had assumed that all workers retired at a higher age, the reduction for
early retirement would be lower, their Social Security benefits would be
higher, and each of the percentages in this chart would unambiguously be
lower. Still, the focus should not be on the specific values from the
simulation but on the difference in the pattern among the benchmarks, which
would not be notably different with other retirement age assumptions. Under
the proportional reduction approach, all beneficiaries in a given birth year
are subject to a benefit reduction that is a constant proportion of their
benefits. Under the progressive reduction, beneficiaries with lower benefits
receive a smaller proportional reduction than those with higher benefits.
See appendix III for more details on the alternative benefit reduction
benchmarks.

Source: GAO?s analysis using the GEMINI model.

The differences in adequacy estimates across benefit- reduction scenarios
reflect how different benefit reduction approaches will have different
effects on workers with different earnings. Lower earners have benefits that
are closer to the poverty threshold than higher earners, so a progressive
approach to reducing benefits would decrease the chances that lower earners?
benefits fall below that threshold. Figure 10 illustrates how different
benefit reduction approaches would produce benefit

29.7 31.8 24.6 28.5 29.5

21.8 22.4 17.5

11.5 0 5

10 15

20 25

30 35

1955 (2017) 1970 (2032) 1985 (2047)

Birth year (age 62 year)

Proportional reduction Progressive reduction Tax-increase only Percent of
cohort with Social Security benefits below poverty

Page 33 GAO- 02- 62 Social Security and Income Adequacy

reductions that would vary by benefit levels. 37 The proportional
benefitreduction benchmark results in identical percentage benefit
reductions, while two alternative, progressive benefit- reduction benchmarks
would result in smaller reductions for lower earners and larger reductions
for higher earners. The so- called ?limited- proportional? benefit-
reduction benchmark would be even more progressive than the progressive
benefitreduction benchmark because a portion of benefits below a certain
level are protected from any reductions while reductions above that level
are proportional. The 1985 birth cohort will be subject to the largest
benefit reductions of the four cohorts we simulated; therefore, it best
illustrates the potential disparity in benefit reductions by benefit level.

37 In this case, we used retired workers not only for simplicity but also
because we explicitly wanted to examine the distribution of benefit
reductions across earnings levels. Spousal benefits equal 50 percent of the
retired worker benefit, therefore including them would bias the
distribution. In any event, the percentage reduction would be the same for a
spouse as for his or her retired worker spouse.

Page 34 GAO- 02- 62 Social Security and Income Adequacy

Figure 10: Benefit Reduction Approaches Have Different Effects on Different
Benefit Levels (Retired Workers Born in 1985, All Retire at Age 62)

Notes: Quintiles are by benefit levels. Percentage reductions are calculated
for beneficiaries closest to the median of each quintile that appear in each
scenario?s sample. For the 1985 birth cohort, all benefit reductions are
fully phased- in under our benchmark scenarios. Under the proportional
reduction approach, all beneficiaries in a given birth year are subject to a
benefit reduction that is a constant proportion of their benefits. Under the
progressive reduction, beneficiaries with lower benefits receive a smaller
proportional reduction than those with higher benefits. Under the limited
proportional reduction, a portion of benefits below a certain level are
protected from any reductions while reductions above that level are
proportional. See appendix III for more details on the alternative benefit
reduction benchmarks.

Source: GAO?s analysis using the GEMINI model.

The different benefit reduction approaches would have different effects on
various subgroups of beneficiaries because of the differences in the
lifetime earnings levels that are typical of those groups. Women,
minorities, and never married individuals all tend to have lower lifetime
earnings than men, whites, and married beneficiaries, respectively. Figure
11 shows how future poverty rates mirror these patterns. Moreover, it
illustrates again how more progressive benefit- reduction approaches would
result in lower poverty rates for these groups in particular. In this case,
we present our analysis using SSA?s MINT model because it allows us to
examine different races. However, these estimates for the year 2020 reflect
benefit reductions that are not fully phased in as well as benefits for
beneficiaries from many birth cohorts who will be subject to various Effects
on Different

Subgroups Reflect Varying Effects by Earnings Level

24 16

11 24

21 23 24 23 28

24 25 31

24 30

32 0 5

10 15

20 25

30 35

Proportional reduction Progressive reduction Limited proportional reduction

Lowest quintile Second Third Fourth Highest quintile Percentage reduction in
Social Security benefits

Page 35 GAO- 02- 62 Social Security and Income Adequacy

levels of the phased- in benefit reductions. 38 For later beneficiaries with
fully phased- in benefit reductions, poverty rates could be higher.

Figure 11: Effects of Benefit Reductions on Various Groups Reflect Varying
Effects by Income Levels

Notes: We looked at data for 2020 because it provides the most complete data
for the elderly population, given the cohorts represented in the MINT. In
the MINT model, total incomes do not include means- tested benefits, while
the official poverty rate does reflect such benefits. In addition, MINT uses
income data from the Survey of Income and Program Participation (SIPP) while
official poverty rates use income data from the Current Population Survey
(CPS). According to CPS data for 1999, which accounts for means- tested
benefits, 10 percent of the elderly had incomes below the poverty threshold.
According to SSA analysts, poverty rates calculated using SIPP data are
generally lower than those using CPS data. Groups shown are not all the
groups for marital status or race; groups for one characteristic overlap
groups for another. Under the proportional reduction approach, all
beneficiaries in a given birth year are subject to a benefit reduction that
is a constant proportion of their benefits. Under the progressive reduction,
beneficiaries with lower benefits receive a smaller proportional reduction
than those with higher benefits. Under the limited proportional reduction, a
portion of benefits below a certain level are protected from any reductions.
See Appendix III for more detail on the alternative benefit reduction
benchmarks.

Source: GAO?s analysis using SSA?s MINT model.

38 We looked at data for 2020 because it provides the most complete data for
the elderly population, given the cohorts represented in the MINT.

1.2 17.1

3.2 12.3

10.1 2.9

6.0 1.2

16.5 3.1

11.8 9.7

2.8 5.8

1.1 15.8

3.0 11.1

9.3 2.6

5.6 1.0

15.3 2.7

10.7 8.9

2.5 5.3

0 5

10 15

20 Married Never married White Non-

Hispanic Black Non- Hispanic Hispanic Male Female

Proportional reduction Progressive reduction Limited proportional reduction
Tax-increase-only Percent of beneficiaries with total incomes below poverty
in 2020

Percent of all beneficiaries with total incomes below poverty (4.1%)

Sex Race/Ethnicity Marital status

Page 36 GAO- 02- 62 Social Security and Income Adequacy

Very old beneficiaries are another subgroup that has tended to be at higher
than average risk of poverty. Several factors relating to multiple sources
of income have contributed to this risk, and many of these factors can be
expected to have similar effects in the future. As people get older, they
may spend down their retirement savings, especially as health and longterm-
care costs mount up, and they are less likely to work. Also, they are more
likely to be widowed. For a couple receiving one retired worker benefit and
one spouse benefit, the household?s Social Security benefits would fall by
33 percent when either is widowed. For a couple in which both spouses
receive retired worker benefits on their own earnings records, household
benefits could fall by as much as 50 percent when either is widowed. In
addition, widows might lose employer- sponsored pension benefits, which
would happen if their spouse elected a self- only annuity instead of a
joint- and- survivor annuity. 39 Also, while Social Security benefits
increase each year to reflect inflation, not all employersponsored pension
benefits do. Of these various factors, all could affect future retirees,
though employer pensions have been changing in design. 40

Based on our review, reducing dependency on public assistance appears to
have been the primary objective of the Social Security program. While many
have noted the importance that Social Security plays in helping ensure
adequate incomes for its beneficiaries, the Congress has never explicitly
defined the term ?adequacy.? In the end, setting benefit levels to address
the adequacy issue will always be, as it has always been, a policy decision
for the Congress. Still, income adequacy is only one of several criteria to
consider in an overall evaluation of comprehensive Social

39 The 1984 Retirement Equity Act made the joint- and- survivor benefit the
default option for married workers retiring on an employer- sponsored
pension. Under the joint- and- survivor option, a surviving spouse will
continue to receive pension benefits once the pension beneficiary dies. To
waive this option, both the retiring worker and his or her spouse must sign
a waiver form. Prior to this law, waivers from spouses were not required,
and few retiring workers selected the joint- and- survivor option. As a
result, many spouses were left without pension benefits when the retiree
died first.

40 Neither our analysis with the MINT model nor with the GEMINI model were
well suited to analyze the outlook for the very old. The very old in the
MINT model are beneficiaries who have already begun their retirements and
will not be subject to any of the benefit reductions in our benchmark policy
scenarios. Our GEMINI analysis was not able to examine other sources of
retirement income, which relate to many of the most significant poverty
risks for the very old. Nevertheless, with GEMINI we were able to examine
patterns in Social Security benefits for people in given cohorts as they age
through retirement. See appendix IV. Concluding

Observations

Page 37 GAO- 02- 62 Social Security and Income Adequacy

Security reform proposals. 41 Specifically, income adequacy should be
balanced against individual equity, or the extent to which benefits are
proportional to contributions. Other criteria include the extent to which
proposals achieve sustainable solvency, how they would affect the economy
and the federal budget, and how readily changes could be implemented,
administered, and explained to the public.

Current demographic trends confront us with a reality that cannot be
ignored. If people will be living longer, then maintaining today?s levels of
monthly benefits for all beneficiaries would require either more revenues,
from whatever sources, or would require that workers wait longer to collect
them. The other alternative of reducing monthly benefits would tend to
diminish income adequacy for beneficiaries. However, our analysis shows that
more progressive approaches to reducing monthly benefits would have a
smaller effect on poverty rates, for example, than less progressive
approaches. Also, reductions that protect benefits for survivors, disabled
workers, and the very old would help minimize reductions to income adequacy,
though they would place other beneficiaries at greater risk of poverty.

More broadly, the choices the Congress will make to restore Social
Security?s long- term solvency and sustainability will critically determine
the distributional effects of the program, both within and across
generations. In turn, those distributional effects will determine how well
Social Security continues to help ensure income adequacy across the
population.

As our analysis has also shown, the effects of some reform options parallel
those of benefit reductions made through the benefit formula, and those
parallels provide insights into the distributional effects of those reform
options. For example, if workers were to retire at a given age, an increase
in Social Security?s full retirement age results in a reduction in monthly
benefits; moreover, that benefit reduction would be a proportional, not a
progressive reduction. Another example would be indexing the benefit formula
to prices instead of wages. Such a revision would also be a proportional
reduction, in effect, because all earnings levels would be treated the same
under such an approach. In addition, holding all else

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

(GAO/ AIMD/ HEHS- 00- 29, Nov. 4, 1999).

Page 38 GAO- 02- 62 Social Security and Income Adequacy

equal, such an approach would implicitly result in future poverty rates that
would be close to today?s rates instead of falling as they would with the
current benefit formula.

Therefore, in finding ways to restore Social Security?s long- term solvency
and sustainability, the Congress will address a key question, whether
explicitly or implicitly:

What purpose does it want Social Security to serve in the future?

 to minimize the need for means- tested public assistance programs;

 to minimize poverty; using what standard of poverty;

 to replace pre- retirement earnings;

 to maintain a certain standard of living; or

 to preserve purchasing power? The answer to this question will help
identify which measures of income adequacy are most relevant to examine. It
will also help focus how options for reform should be shaped and evaluated.
Our analysis has illustrated how the future outlook depends on both the
measures used and the shape of reform. While the Congress must ultimately
define Social Security?s purpose, our analysis provides tools that inform
its deliberations.

Still, changes to benefit levels would typically only be part of a larger
reform package, and Social Security is only one part of a much larger
picture. As we have said in the past, reform proposals should be evaluated
as packages that strike a balance among their component parts. Furthermore,
Social Security is only one source of income and only one of several
programs that help support the standard of living of our retired and
disabled populations. All sources of income and all of these programs should
be considered together in confronting the demographic challenges we face.
For example, changes to Social Security could potentially affect SSI
benefits, employer- sponsored pensions, retirement savings, and the work and
retirement patterns of older workers. Such interactions should actively be
considered. Moreover, several programs provide noncash benefits that also
play a major role in sustaining standards of living for their beneficiaries.

Importantly, examining the adequacy of cash income alone would ignore the
major role of noncash benefits and the needs they help support. This is
especially critical in the case of Medicare beneficiaries. Considering these
important noncash benefits in any adequacy analysis could have a very

Page 39 GAO- 02- 62 Social Security and Income Adequacy

material effect on both the absolute and relative positions of senior
citizens as compared to other groups of Americans.

We provided a draft of this report to SSA. SSA provided a number of
technical comments, which we have incorporated where appropriate.

We are sending copies of this report to the Commissioner of the Social
Security Administration and other interested parties. We will also make
copies available to others on request. If you or your staff have any
questions concerning this report, please call me on (202) 512- 7215. Key
contributors are listed in appendix V.

Sincerely yours, Barbara D. Bovbjerg, Director Education, Workforce, and
Income Security Issues Agency Comments

Appendix I: Descriptions of Adequacy Measurement Methods

Page 40 GAO- 02- 62 Social Security and Income Adequacy

Several methods have been used to measure the level of adequate income- what
it costs to live. We identified 11 methods that have been used to develop
measures against which income from Social Security benefits might be
compared for determining adequacy. 1 These methods include the current
poverty thresholds, experimental poverty thresholds, family budgets, family
expenditures, material hardship, median family income, one- half median
family income, per capita personal income, public assistance, public
opinion, and earnings replacement rates.

These methods vary along a number of dimensions. These include their
purpose, features of their construction, years for which they measure
adequacy, and frequency of publication. In some instances where the method
has been used to develop more than one measure, we selected one of the
measures as an example of the method and used it for the description of the
method.

The methods also vary in whether they are absolute or relative. Absolute
measures are derived from a fixed bundle of goods and services that does not
vary in mix, quantity, or quality regardless of when or where it is applied.
For example, an absolute measure would be one based on a list of goods and
services that are judged to be necessary for a family to meet its basic
needs. The list of goods and services would need to be changed periodically
to reflect changes in living standards over time. In contrast, relative
measures change with current income or consumption. Measurement experts who
have served on various panels to study the issue have not agreed on which is
more appropriate to determine how much it costs to live. Table 2 provides an
overview of the 11 methods with regard to several dimensions.

1 Two of these measures are nonmonetary and therefore cannot be used in
direct comparison with a monetary amount. Appendix I: Descriptions of
Adequacy

Measurement Methods

Appendix I: Descriptions of Adequacy Measurement Methods

Page 41 GAO- 02- 62 Social Security and Income Adequacy

Table 2 is followed by a fuller description of each method, with particular
attention to how each is constructed, its uses, and issues that panels and
experts have raised regarding the measure.

Appendix I: Descriptions of Adequacy Measurement Methods

Page 42 GAO- 02- 62 Social Security and Income Adequacy

Table 2: Overview of the Purpose, Construction, and Other Dimensions of 11
Methods To Measure Income Adequacy

Measurement method Purpose of measure Absolute or

relative measure a

Current poverty thresholds Determine whether persons living in a family are
officially poor. Absolute

Experimental poverty thresholds Provide reasonable thresholds to derive
poverty statistics. Relative

Family budgets Estimate what it costs a working family of four to live.
Absolute

Family expenditures Describe consumer spending and determine cost- of-
living indexes. Relative

Material hardship Identifies individuals who do not consume minimal levels
of goods and services.

b Median family income Estimates the income of the family at

the middle of the income distribution. Relative

Appendix I: Descriptions of Adequacy Measurement Methods

Page 43 GAO- 02- 62 Social Security and Income Adequacy

Construction of measure Basis of measurement Family unit Geographic

unit Method of updating

Observation about the measure Years of measure Frequency of

publication

1964 cost of a food plan for the family multiplied by three

By age of family head, family size, and number of children under 18 years

b Increased annually by the Consumer Price Index (CPI)

(See app. II) 1959 through present Annually

Median expenditures in specific categories adjusted for other necessities,
family size, and cost of housing

By family size and number of children under 18 years c

41 geographic areas based on area and population size

Median expenditures in the specific categories are re- estimated

Lacks scientific basis to support the use of median level of expenditures.

1990 to 1997 Initially in 1995; Census published additional years of
thresholds in 1999.

The total costs of a list of items associated with a specific level of
living

Separate budgets specified for a family of four and a retired couple d

Urban United States, 40 metropolitan areas, and 4 nonmetropolitan regions

In addition to revising the lists of goods and service, (1) the lists were
periodically repriced; or (2) previous costs increased annually by the CPI

Expert panel recommended methodology be changed; found subjective judgment
used in developing lists of items.

1946 to 1951, 1959, and 1966 to 1981 Periodically;

no longer published

Consumer expenditures as reported in a national survey

By various family characteristics

By type of area (urban and rural) and 4 regions

Data are collected quarterly; prior to 1980, data collected about every 10
years

Averages may not represent expenditures made by families with specified
characteristic.

1980 through present Annually

Self- assessment on specific events, such as not having enough to eat

By various family characteristics

b b Measure is not in monetary terms. 1992 and 1995;

Chicago residents in the mid- 1980s

Census published reports for 1992 and 1995. Money income as reported in a
national survey

By various family characteristics

4 regions and by type of residence- inside or outside metropolitan areas

Money income data are collected annually

(See app. II) 1947 through present Annually

Appendix I: Descriptions of Adequacy Measurement Methods

Page 44 GAO- 02- 62 Social Security and Income Adequacy

Measurement method Purpose of measure Absolute or

relative measure a

One- half median family income Provides a means for comparative analysis of
poverty status. Relative

Per capita personal income Presents the nation?s personal income on a per
person basis. Relative

Public assistance Determines whether a person is financially dependent.

b Public opinion Tracks the size of groups of people at

different standards of living. b

Earnings replacement rates Determine extent that individuals can maintain
pre- retirement standard of living.

Relative

Appendix I: Descriptions of Adequacy Measurement Methods

Page 45 GAO- 02- 62 Social Security and Income Adequacy

Construction of measure Basis of measurement Family unit Geographic

unit Method of updating

Observation about the measure Years of measure Frequency of

publication

50 percent of median family income as reported in a national survey

By various family characteristics

b Money income data are collected annually

No standard method for which family type or proportion of income to use as
adequacy measure.

b b Total of various sources of personal income in the nation divided by the
population of the nation

b By nation, state, region, county, and metropolitan areas

Comprehensive revisions about every 5 years;

Includes income of nonprofit institutions and pension plans.

1929 through present Monthly at the

national level; quarterly at the state level; and annually at the county and
metropolitan area Participation in a welfare program Programs

vary, e. g., Supplemental Security Income (SSI) benefit amounts vary by
family composition

Programs vary, e. g., number of SSI recipients published nationally and by
state

Programs vary, e. g., SSI benefit levels are increased annually by the CPI

Measure is not in monetary terms (see app. II).

Programs vary, e. g., 1974 through present for SSI

Programs vary, e. g., number of SSI recipients published quarterly and
annually

Subjective estimates of what it costs to get along in the community

b bb Wording may be ambiguous to respondent.

Selected years from 1946 through 1992 Periodically;

Gallop measure no longer collected

Retirement income relative to pre- retirement income

May be computed for individuals or households but most meaningful for
households

b b Measure calculated for specific individuals

Measure is not a standard of adequacy but a benchmarking tool.

b Measure determinable for any year. Estimates for illustrative workers
available from 1940

b

a An absolute measure is a fixed bundle of goods and services that does not
vary in mix, quantity, or quality regardless of when or where it is applied.
A relative measure changes directly with current income or consumption. b
None or not applicable.

c One of the experts who recommended these thresholds proposed another
adjustment for families headed by a single adult. d In 1960 and 1968, the
Bureau of Labor Statistics (BLS) published equivalence scales that could be

used to adjust the costs by family composition. Source: GAO?s analysis of
the relevant literature.

Appendix I: Descriptions of Adequacy Measurement Methods

Page 46 GAO- 02- 62 Social Security and Income Adequacy

The poverty thresholds are a measure that attempts to specify the minimum
money income that could support an average family of a given composition at
the lowest level of living consistent with a country?s prevailing standards
of living.

The poverty thresholds are an absolute measure whose initial purpose was to
measure year- to- year changes in the number and characteristics of poor
people.

The poverty thresholds, as originally published by the Social Security
Administration (SSA) in 1963, represent a minimal amount of funds a family
needed to rear its children, what the author termed ?crude indexes? of
poverty. Later the crude indexes were extended to families without children.
If a family?s total money income is less than the poverty threshold for that
family?s composition, which is based on family size, age of the family?s
head, and number of children under 18 years old, then that family, and every
individual in it, is considered poor.

In 1965, the Office of Economic Opportunity adopted the thresholds for
statistical and program planning purposes. The Bureau of the Budget
established the thresholds as the official definition of poverty for
statistical use in all executive departments in 1969. This definition was
reconfirmed in Statistical Policy Directive No. 14, after the bureau became
the Office of Management and Budget.

Poverty thresholds are used mainly for statistical purposes, such as
estimating the number of Americans in poverty each year. 2 This official
measure of poverty is used to measure the nation?s progress in reducing the
extent of poverty and is used to allocate funds and to identify target
populations for various public assistance programs. Policymakers use trends
in poverty rates- the proportion of persons whose family income is below the
poverty threshold- over time and across population groups to make judgments
about particular policies. Poverty statistics are also used to evaluate
government programs for low- income persons and the effects of policies on
the distribution of income.

2 Poverty guidelines, a simplification of the poverty thresholds, are used
for administrative purposes, such as determining financial eligibility for a
federal program. Such means- tested programs include food stamps, low-
income home energy assistance, and legal services for the poor. See Means-
Tested Programs: Determining Financial Eligibility Is Cumbersome and Can Be
Simplified (GAO- 02- 58, Nov. 2, 2001). Current Poverty

Thresholds Purpose and Uses

Appendix I: Descriptions of Adequacy Measurement Methods

Page 47 GAO- 02- 62 Social Security and Income Adequacy

SSA?s 1963 publication based the poverty thresholds on information from a
1955 food consumption survey and the 1964 costs of a food plan. The author
determined from U. S. Department of Agriculture?s (USDA) 1955 Household Food
Consumption Survey that families of three or more people spent approximately
one- third of their after- tax money income on food. The author then tripled
the 1964 costs of USDA?s economy food plan for various compositions of
families. Different procedures were used to calculate poverty thresholds for
two- person families and single individuals. 3 Separate thresholds were
estimated for single individuals and 2- person families headed by an
individual 65 years and over, as well as an individual under 65 years old. 4
There were separate sets of thresholds for farm and nonfarm families, as
well as thresholds by sex of the head of the family. The thresholds that
were based on the sex of the family?s head and by farm residence were
eliminated in 1981.

There is no geographic variation of the poverty thresholds. Although there
were regional costs for the USDA food plan, they were not used to account
for regional variation when the poverty thresholds were developed. 5

Two methods have been used to update the original poverty thresholds.
Initially, the change in the cost of USDA?s economy food plan was used to
annually update the poverty thresholds. In 1969, the method of updating the
thresholds was changed to price changes of all items in the Consumer Price
Index (CPI). 6 The poverty thresholds are increased each year by the

3 The author used a different multiplier of the food costs for 2- person
families and then calculated the single- person thresholds as 80 percent of
the 2- person thresholds. 4 The term ?equivalence scales? is used in
reference to an index that is calculated from the resulting family size and
age of head distinctions embedded in the 1963 methodology. In one such
scale, the individual is the base and other family compositions are in
relation to it. For example, a 4- person family?s equivalence scale is 2.
01, indicating that it is equivalent to 2 individuals.

5 Only the poverty guidelines differentiate by geographic area. Separate
guideline amounts are published for the 48 contiguous states and the
District of Columbia, Alaska, and Hawaii. 6 Poverty thresholds for 1959 to
1967 were recalculated on this basis. Construction of the

Measure

Appendix I: Descriptions of Adequacy Measurement Methods

Page 48 GAO- 02- 62 Social Security and Income Adequacy

same percentage as the annual average CPI for all Urban Consumers (CPI- U).
7 The Census Bureau annually updates and publishes the poverty thresholds. 8

Numerous alternative poverty thresholds have been proposed since the
official adoption of the measure developed in 1963. One such alternative is
the experimental thresholds recommended by a Committee on National
Statistics of the National Academy of Science (NAS) study panel in 1995. 9

The NAS poverty threshold is a relative measure whose stated purpose was ?to
lead to an initial threshold that is reasonable for purposes of deriving
poverty statistics.?

The NAS poverty thresholds have been solely used for research. Census
published a report in 1999 to provide information for evaluating the
implications of many of the NAS panel?s recommendations for a new poverty
measure. To do so, Census reported how estimated levels of poverty for 1990
through 1997 differed from official levels as specific recommendations of
the NAS panel are implemented individually and how estimated trends differed
when many recommendations are implemented simultaneously.

The NAS poverty thresholds represent a dollar amount for basic goods and
services- food, clothing, shelter (including utilities)- and a small
additional amount to allow for other common, everyday needs (e. g.,
household supplies, personal care, and nonwork- related transportation). 10

7 BLS publishes the CPI- U and the CPI for Urban Wage Earners and Clerical
Workers (CPIW). According to BLS, the CPI- U represents about 87 percent of
the U. S. population, and the CPI- W represents about 32 percent of the U.
S. population. BLS began publishing the CPI- U in 1978. Until then, it
published only the CPI- W.

8 For example, see Poverty in the United States: 2000, U. S. Census Bureau,
Current Population Reports: Consumer Income, P60- 214, (Washington: 2001). 9
Constance F. Citro and Robert T. Michael, eds., Measuring Poverty: A New
Approach

(Washington: National Academy Press, 1995). 10 To determine a family?s
poverty status, the NAS panel recommended that family resources- the measure
that is compared to the thresholds- include money and nearmoney income,
minus expenses, such as child care, that divert money away from the basic
goods and services. Experimental Poverty

Thresholds Purpose and Uses

Construction of the Measure

Appendix I: Descriptions of Adequacy Measurement Methods

Page 49 GAO- 02- 62 Social Security and Income Adequacy

First, to develop a threshold for a reference family, a specified percentage
of median annual expenditures from the Consumer Expenditure Survey (CEX)
data is used to determine an amount of food, clothing, shelter expenditures.
11 The reference family consists of two adults and two children. The median
annual expenditure amount is next increased by a modest additional amount to
allow for other necessities. An equivalence scale is then applied to the
reference family threshold to adjust for families of different sizes and
composition.

Further adjustments are made to account for geographic differences in the
cost of housing. The NAS panel developed an index of 41 geographic areas
that is presented by area and population size. These index values are
applied to the thresholds to adjust for differences in the cost of housing.

The NAS panel also recommended a method for updating the initial threshold
that would reflect changes in nominal growth in food, clothing, housing, and
shelter expenditures. To do so, 3 years of the most recent data from the CEX
would be used to determine the threshold for the reference family. The CPI-
U would be used to update these expenditure data to the current period.
Then, the procedures as outlined above are followed to estimate thresholds
for families of other sizes by geographic areas. 12

The NAS panel said that its method of updating the thresholds represented a
middle ground between an absolute approach of simply updating the thresholds
for price changes, which ignores changes in living standards over time, and
a relative approach of updating the thresholds for changes in total
consumption.

One of the NAS panel members dissented from the panel because the major
recommendations and conclusions for changing the measurement of poverty were
the ?outcome of highly subjective judgments? and were not based on
scientific evidence. In his dissent, the member said that there was no
scientific basis to support the use of food, clothing, and shelter

11 Actually, the panel did not use a single percentage of the median. It
concluded that the reference family threshold be from 78 to 83 percent of
the median expenditures of the basic goods and services.

12 The NAS panel recommended that the index used to make housing costs
adjustments be updated every 10 years as new decennial census data become
available. Observations About the

Measure

Appendix I: Descriptions of Adequacy Measurement Methods

Page 50 GAO- 02- 62 Social Security and Income Adequacy

expenditures upon which to develop the thresholds. He also objected to using
the median level of expenditures of these items rather than the CPI to
update the poverty thresholds; he said to do so would change the measure
from an absolute to a relative measure. He had two other objections in that
the NAS panel did not treat medical care as basic service and that the panel
suggested that the poverty line fell within a range of values, of which he
stated did not have the scientific community?s consensus.

Family budgets are an income adequacy measure that dates back to the 19th
century. The measure described in this appendix is for the city worker?s
family budget, 13 whose origins closely relate to the budgets that the Works
Progress Administration constructed in 1935 for a urban family of four.

The city worker?s family budget represents the estimated cost of a list of
goods and services that the 4- person family would need to live at a
designated level of well- being. 14 The level designated in the city
worker?s family budget for 1946 was intended to represent a modest but
adequate standard of living. 15 The same level of well- being was used in
the interim city worker?s family budget with 1959 costs. In the mid- 1960s,
two levels of well- being were added- lower and higher- and the name of the
modest but adequate level was changed to intermediate. Also, the name of the
city worker?s family budget was changed to family budgets.

The city worker?s family budget was an absolute measure that was used to
determine the adequacy of income- what it costs to live- for a city worker?s
family who was defined as a husband, aged 38 and employed full

13 See City Worker?s Family Budget for a Moderate Living Standad, Bureau of
Labor Statistics, Bulletin 1570- 1 (Washington: 1967). BLS? first
publication of a budget was in 1911 for the measurement of the adequacy of
family incomes of textile workers.

14 The elderly couple?s budget was developed by SSA to parallel the city
worker?s family budget. 15 The level designated in SSA?s elderly couple?s
budget was intended to represent a modest but adequate mode of living, which
allowed normal participation in the life of the community in accordance with
current American standards. The level was above subsistence level in that it
provided for more than physical needs or what would be necessary to carry
families through a limited period of stringency. Family Budgets

Purpose and Uses

Appendix I: Descriptions of Adequacy Measurement Methods

Page 51 GAO- 02- 62 Social Security and Income Adequacy

time; a wife who did not work outside the home; a boy aged 13; and a girl
aged 8. 16

The city worker?s family budgets were used as benchmarks in determining
individual family needs, establishing interarea differences in living costs,
and documenting changes in living standards over time. The budget cost
levels were used by federal, state, and local governments as thresholds for
eligibility in administrative programs. The city worker?s family budgets
were widely used in employment compensation determinations, such as wage
negotiations and geographic wage adjustments. Since the costs of the budgets
were city specific, the budgets were also used to construct indexes of
living costs. These indexes showed interarea variations in living costs and
individuals and financial planners used them to examine interarea cost- of-
living differences. The budgets were also used in private and public legal
actions. Researchers continue to construct family budgets to examine the
adequacy of Social Security benefits, 17 as well as the adequacy of wages
paid to single parents. 18

In a number of countries, budget standards are used as reference points in
devising or monitoring income maintenance programs. For example, the
Commonwealth Department of Social Security commissioned the development of a
set of budget standards for Australia. Published in 1998, the budget
standards are expected to inform future Australian governments in relation
to adequacy standards. 19

16 The purpose of SSA?s elderly couple?s budget was to measure the adequacy
of income for a retired couple who was defined as a husband about 65 years
of age or over and the wife a few years, if at all, younger.

17 For example, Joseph White in False Alarm: Why the Greatest Threat to
Social Security and Medicare Is the Campaign to ?Save? Them (Baltimore:
Johns Hopkins University Press, 2001), used an abbreviated method to
calculate 1996 costs of an elderly couple?s budget in 14 nonrandom counties.
From the budget amounts, he then estimated the amounts that retired couples
would need in Social Security benefits to meet the costs of the rudimentary
budgets that he constructed.

18 Family budgets have been calculated for a single- parent with two
children. Wages of the single parent are then compared to the budget amount
to determine if families are provided adequate compensation to meet their
basic needs. For example, see Jared Bernstein, Chauna Brocht, and Maggie
Spade- Aguilar, How Much Is Enough? Basic Family Budgets for Working
Families (Washington: Economic Policy Institute, 2000).

19 Peter Saunders, et al., Development of Indicative Budget Standards for
Australia, Social Policy Research Centre, University of New South Wales,
Policy Research Paper No. 74 (Mar. 1998).

Appendix I: Descriptions of Adequacy Measurement Methods

Page 52 GAO- 02- 62 Social Security and Income Adequacy

The measure involved the formulation of a budget, listing the items and
their quantities that comprised the level of well- being chosen, the pricing
of these items, and computing the aggregate annual cost of the budget. A
group of experts developed a list of goods and services using scientific
standards of requirements, such as the recommendations of the Committee on
Nutrition of the National Research Council for the food segment of the
budget. 20 Where standards had not been developed for the various segments
of the budget, records of family expenditures by 4- person families were
used. These data were studied to determine the level of purchases in
expenditure categories where the families began to purchase higher quality
items in the same expenditure category of items or started to save their
income.

BLS published the costs of the city worker?s family budget for 34 cities for
1946, 1947, 1949, 1950, and 1951. The cost of the interim city worker?s
family budget was published for 20 cities for 1959. The family budgets at
three cost levels were published for 1967 through 1981 (the cost of the
intermediate level was also published for 1966) for urban United States, 40
individual metropolitan areas, and 4 nonmetropolitan regions. 21

The early city worker?s family budgets could not be used for families other
than those consisting of a husband, wife, and two young children. In 1960,
BLS published equivalence scales that could be used to adjust the costs by
family composition. BLS updated the equivalence scales in 1968.

The city work?s family budgets are no longer published. With the release of
the 1981 budget costs, BLS terminated the family budgets program because
funding was not available for a revision.

In addition to re- specifying the lists of items in a revision, two methods
were used to update the city worker?s family budget costs. The first method
recollected price data for the individual items on the budget list and then
aggregated those costs for an annual amount. The other method, which was
used to estimate the 1949 through 1951 and the 1969 through

20 Scientific standards for housing were also used. Provisions of the
Medicare program were used for the medical segment of the retired couple?s
budgets. 21 The costs of the elderly couple?s budget were published
initially for 8 cities and later for 13 cities. The 1950 costs were
published for 34 cities. The 1959 interim retired couple?s budget was
published for 20 cities. The retired couple?s budget costs were published
for urban United States, metropolitan and nonmetropolitan areas, 40
individual metropolitan areas, and 4 nonmetropolitan regions. Construction
of the

Measure

Appendix I: Descriptions of Adequacy Measurement Methods

Page 53 GAO- 02- 62 Social Security and Income Adequacy

1981 costs, was to use the CPI?s component index numbers to update the costs
for the segments of the budgets. 22 Revisions of the budgets occurred in
1959 and 1966 when the lists of goods and services were re- specified by
experts to account for changes in the modest but adequate standard of
living.

In response to a congressional mandate and in recognition that the family
budgets needed to be improved, in the 1970s, BLS contracted with the
Wisconsin Institute for Research on Poverty to recommend revisions in the
Family Budgets program. In 1980, the Expert Committee on Family Budget
Revisions recommended that the methodology be changed and that scientific
standards no longer be used. 23 The committee asserted that a scientific
basis does not exist by which to develop commodity- based lists for the
budgets.

One of the reasons the Expert Committee on Family Budget Revisions
recommended a change in methodology was that it found that large elements of
relativity and subjective judgment entered into the development of the lists
of goods and services, including those for which scientific standards were
used. The committee recommended that actual overall levels of expenditures
be used to measure adequacy. Specifically, it recommended that median
expenditure of two- parent families with two children be used to develop the
?prevailing family standard? budget and that three other standard budgets be
developed as proportions of the prevailing family standard budget amount. 24
In a dissent, a committee member said that a measure of well- being that
uses an average (or median) of total family expenditures, which is obtained
from a consumer expenditure survey, does not take into consideration the
specifics of what that amount will buy or whether the actual quantities of
goods and services available within the amount are enough to supply what is
needed.

22 The two methods were compared for differences between 1946 and June 1947
elderly couple?s budget costs. In this comparison the actual costs of the
budgets rose less rapidly than when the budgets were updated with the CPI.
However, this finding could not be applied to periods outside these 2 years.

23 Expert Committee on Family Budget Revisions, New American Family Budget
Standards, Working Paper, Institute for Research on Poverty, University of
WisconsinMadison (May 1980).

24 BLS researchers recently implemented the committee?s recommendations with
consumer expenditure data for 1989, 1994, and 1998. See David S. Johnson,
John M. Rogers, and Lucilla Tan, ?A Century of Family Budgets in the United
States,? Monthly Labor Review,

Vol. 124, No. 5 (May 2001) pp. 28- 45. Observations About the

Measure

Appendix I: Descriptions of Adequacy Measurement Methods

Page 54 GAO- 02- 62 Social Security and Income Adequacy

Family expenditures are the averages of consumer purchases that are recorded
in survey data arrayed by family characteristics, such as age of reference
person.

Family expenditures is a relative measure whose purpose is to describe
consumer spending and to determine cost- of- living indexes. The basic
premise is that the living standards of society can be measured with current
consumption expenditure levels and patterns. The early family expenditure
surveys, which were conducted in the late 19th century, were concerned with
the cost of living of the ?working man? and his family, that is the amount
of dollars a family needed to live.

Family expenditure data are used by government and private agencies to study
the welfare of particular segments of the population. The data are used by
economic policymakers interested in the effects of policy changes on various
groups.

CEX data are used to estimate aggregate family expenditures. There are three
basic methods to measure family expenditures: current consumption, used in
the CEX before 1980; total expenditures, used in the CEX since 1980; and
current outlays, an alternative measure used to approximate out- of- pocket
expenditures, which is also used in the CEX since 1980.

Current consumption expenditures method includes the transaction costs of
goods and services, excise and sales taxes, the price of durables (e. g.,
vehicles) at the time when the purchased, and home mortgage interest
payments. It excludes the payment of principal on loans, gifts to persons
outside the family, personal insurance, and retirement and pension payments.

The total expenditures method is the same as the current consumption
expenditures method, except it includes gifts, personal insurance, and
retirement and pension payments.

The total outlays method differs from total expenditures in that payments of
principal for home mortgages and financed vehicles are included and the
purchase price of vehicles is excluded. Family Expenditures

Purpose and Uses Construction of the Measure

Appendix I: Descriptions of Adequacy Measurement Methods

Page 55 GAO- 02- 62 Social Security and Income Adequacy

Data from the continuing CEX have been collected quarterly on an ongoing
basis since 1980. Prior to the continuing CEX, the survey was conducted
periodically about once every 10 or so years. BLS annually publishes average
annual expenditures from the continuing CEX for consumer units. 25
Expenditure data are published by type of area (urban and rural) and for
four regions of residence. 26

According to BLS, the published expenditure amounts are averages for
consumer units with specified characteristics, regardless of whether or not
a particular consumer unit purchased an item in the expenditure category
during data collection. Therefore, the average expenditure for an item may
be considerably lower than the average for those who actually purchased the
item. Also, the average may differ from those who purchased the item as a
result of frequency of purchase or the characteristics of the consumer units
that purchased the item. For example, since all consumer units do not
purchase a new vehicle every year, the average expenditure for new vehicles
will be lower than the average for those who actually purchased a new
vehicle because the average expenditure includes those who did not purchase
a new vehicle that year. Even among those who purchase the item, consumer
units may have dissimilar demographic characteristics.

Material hardship measures identify individuals who do not consume minimal
levels of goods and services, such as food, housing, clothing, and medical
care. The material hardship measure presented here is one developed in the
1980s by Susan Mayer and Christopher Jencks in their study of Chicago
residents. 27 This material hardship measure focused on the following
hardships: hunger, cut off of utilities to the home, living in crowded or
dilapidated housing, eviction, inadequate health care, and unmet needs for
dental care.

25 A consumer unit is the members of a household related by blood, marriage,
adoption, or other legal arrangement; a single person living alone or
sharing a household with others but is financially independent; or two or
more persons living together who share some major expenses.

26 For example, see Consumer Expenditures in 1999, BLS Report 949,
(Washington: 2001). 27 Susan E. Mayer and Christopher Jenks, ?Poverty and
the Distribution of Material Hardship,? The Journal of Human Resources, Vol.
XXIV, (Winter 1989), pp. 88- 113. Observations About the

Measure Material Hardship

Appendix I: Descriptions of Adequacy Measurement Methods

Page 56 GAO- 02- 62 Social Security and Income Adequacy

Material hardship is a measure whose purpose is to provide a means for
policymakers to measure the goal of reducing specific forms of material
hardship. 28

Researchers have used material hardship measures to supplement traditional
measures of poverty, such as to provide a nonmonetary perspective of those
who are experiencing economic difficulties. The measures are used by
researchers to create point- in- time estimates of hardship, describe trends
in hardship, identify predictors of hardship, and develop hardship
indicators to evaluate welfare reform.

Respondents are asked to make self- assessments of specific events in their
lives. For example, they are asked if there was a time in the previous year
when they needed food but could not afford to buy it or could not get out of
the home to get food. Generally asked in a yes/ no format, these indicators
are reported individually but are then summed into a composite deprivation
index. In some instances, respondents are asked to report the hardship on
the basis of a scale. For example, respondents might be asked to categorize
the food eaten in their household as (1) having enough of the kinds of food
they want, (2) enough but not always the kinds they want, (3) sometimes not
enough to eat, or (4) often not having enough to eat. Other than
periodically conducting the surveys, there is no method to update the
material hardship measure.

Until Census began collecting data from a nationally representative sample,
data had been collected of single mothers in Chicago, Illinois, and of
selective populations in other cities. 29

28 Material hardship measures are neither absolute nor relative. They
examine the extent or lack thereof the consumption of a specific good or
service. They do not change with current income or consumption.

29 Material hardship data are collected in selected nationwide surveys. The
Survey of Income and Program Participation (SIPP) and the National Survey of
America?s Families (NSAF) collect data for multiple indicators of material
hardship. Census conducts the SIPP and the Urban Institute in partnership
with Child Trends administers NSAF. Census published extended measures of
well- being from the SIPP for 1992 and 1995. Other data sources are
available for specific material hardship measures. For example, food
insecurity data were collected in the Third National Health and Nutrition
Survey, supplements to the Current Population Survey, and the Longitudinal
Study of Aging. Purpose and Uses

Construction of the Measure

Appendix I: Descriptions of Adequacy Measurement Methods

Page 57 GAO- 02- 62 Social Security and Income Adequacy

Median income is the amount which divides an income distribution into two
equal groups, half having incomes above the median and half having income
below the median. The concept of using the midpoint of the income
distribution as an adequacy measure is that people are social beings and
that full participation within society requires that they ?fit in? with
others. Individuals are not able to participate fully in society if their
resources are significantly below the resources of their members of society,
even if they are able to eat and physically survive.

Median family income is a relative measure whose purpose is to estimate the
income of the family at the middle of an income distribution. Researchers,
analysts, and policymakers use median family income to follow historical
trends and annual changes in income. A relative measure, such as median
family income, is used to provide a perspective of an adequacy measure that
keeps up to date with overall economic changes in the society.

Current Population Survey (CPS) data are used to calculate median family
income. The measure is updated annually through data collection. The median
is based on money income before taxes and does not include the value of
noncash benefits, such as food stamps, Medicare, Medicaid, public or
subsidized housing, and employment- based fringe benefits. The Census Bureau
has annually published median family income since 1947. 30

Median family income data are published by various family characteristics.
The data are also presented by four regions of residence and by type of
residence- inside or outside metropolitan areas. The metropolitan areas are
further broken down by over or under 1 million in population and by inside
or outside central cities.

One- half of median family income (see the previous method for description
of median family income) is a relative poverty standard.

30 For example, see Money Income in the United States: 2000, U. S. Census
Bureau, Current Population Reports: Consumer Income, P60- 213, (Washington:
2001). Median Family

Income Purpose and Uses Construction of the Measure

One- Half Median Family Income

Appendix I: Descriptions of Adequacy Measurement Methods

Page 58 GAO- 02- 62 Social Security and Income Adequacy

One- half median family income is a relative measure that researchers use to
demonstrate the absolute nature of the official poverty thresholds. 31 One-
half of median income for four- person families is also used in comparative
analyses of poverty across nations.

Researchers use one- half of the value of median family income as the
measure.

No standard method is used to establish the measure of a minimal level of
adequacy with median family income. The most commonly proposed measure used
for poverty determination is 50 percent of the median. The standard could be
implemented in several ways, for example, one- half of the median for each
family size. However, median income by family size is bell shaped with the
peak at the four- person family.

Per capita personal income is the amount of personal income from the U. S.
national income and product accounts (NIPA) that would be available to each
individual if all income received by persons was distributed equally among
all people in the nation.

Per capita personal income is a relative measure whose purpose is to present
a measure of a nation?s personal income on a per person basis.

Government and private decision makers, researchers, and the public at large
who need timely, comprehensive, and reliable estimates use per capita
personal income as a measure of the value of and changes in average income
at the national and regional level. Because per capita personal income is
conceptually and statistically consistent with the official measure of
output (Gross Domestic Product), productivity, and other key economic
indicators, national estimates of per capita personal income are key inputs
to the formulation and monitoring of economic activity by the Federal
Reserve Board and to the preparation of projections of federal receipts by
the Congressional Budget Office.

31 For example, see Denton R. Vaughan, ?Exploring the Use of the Public?s
Views to Set Income Poverty Thresholds and Adjust Them Over Time,? Social
Security Bulletin, Vol. 56, No. 2 (Summer 1993) pp. 22- 46. Purpose and Uses

Construction of the Measure

Observations About the Measure

Per Capita Personal Income

Purpose and Uses

Appendix I: Descriptions of Adequacy Measurement Methods

Page 59 GAO- 02- 62 Social Security and Income Adequacy

Regional level estimates, which are consistent with the national estimates,
also are used by state governments for similar purposes and are used in the
allocation of federal funds for key programs.

Per capita personal income data are used as a measure of the economy?s
capacity to pay. For example, the Medicaid funding formula uses state per
capita personal income to provide higher matching percentages for states
that have more limited resources to finance program benefits and more low-
income people to serve.

Personal income is calculated as the sum of incomes received by persons from
production and from transfer payments from government and business.
?Persons? consists of individuals, nonprofit institutions that primarily
serve individuals, private noninsured welfare funds, and private trust
funds.

Wage and salary disbursements, other labor income, 32 proprietors? income,
rental income, dividend income, interest income, and transfer payments to
persons, less personal contributions for social insurance are summed to
calculate personal income. In most cases, only market transactions are used.
In a few cases, nonmarket transactions are used in personal income. These
transactions include home ownership, financial services furnished without
direct payment, and employer contributions for health and life insurance.
The summation of the personal income components is then divided by the
nation?s population to provide per capita personal income. Population is the
total population of the United States, including military personnel.

Each component of personal income is prepared independently using the most
up- to- date and reliable source data. The Commerce Department?s Bureau of
Economic Analysis prepares the estimates of personal income and calculates
per capita personal income. Per capita personal income estimates are
released monthly at the national level, quarterly at the state level, and
annually at the county and metropolitan area levels. Per capita

32 Other labor income consists primarily of employer payments to private
pension and profit- sharing plans, publicly administered government employee
retirement plans, private group health and life insurance plans, and
privately administered workers? compensation plans. Construction of the

Measure

Appendix I: Descriptions of Adequacy Measurement Methods

Page 60 GAO- 02- 62 Social Security and Income Adequacy

personal income is published at both the national and regional- state,
county, and metropolitan area- levels. 33

The base of per capita personal income, personal income, is updated on a
regularly scheduled basis, where the schedule of updates are timed to
incorporate newly available and revised source data. Comprehensive revisions
are carried out at about 5- year intervals. 34 Population estimates are
revised to reflect the results of the latest decennial census of population.

The definition of personal income, which is based on the NIPA definition, is
not what one usually equates to family or household income. For example, it
includes income of ?persons? as defined for the NIPAs, which includes income
of individuals as well as income of nonprofit institutions serving
individuals and the investment income of pension plans. It excludes realized
capital gains or losses and incomes that reflect transfers from other
individuals, such as alimony or gifts. Although, in general, incomes are
recorded when received, benefit payments from pension plans are not included
when the benefits are actually paid. Instead, employer contributions to
these plans are recorded as income to employees when the contributions are
made and the investment income of the plans is recorded when earned. Also,
although Social Security benefit payments are included in personal income,
total personal income is reduced by personal contributions to Social
Security.

If an individual is dependent upon others for cash assistance, then the
individual has inadequate income. Since data about sources of income
provided by others is difficult to obtain, statistical indicators of such
dependency often resort to administrative data from public assistance
programs. As used in this appendix, the receipt of public assistance is a
measure to denote individuals who meet program eligibility criteria and

33 For example, see table J. 3 Per Capita Personal Income and Disposable
Personal Income and table K. 1 Personal Income and Per Capita Personal
Income by Metropolitan Area in Section D of Survey of Current Business.

34 Comprehensive revisions update the NIPAs to portray more accurately the
evolving U. S. economy, to reflect the introduction of new and improved
methodologies, and to incorporate newly available and revised major source
data. Revisions are carried back in time to maintain a consistent time
series of data. Observations About the

Measure Public Assistance

Appendix I: Descriptions of Adequacy Measurement Methods

Page 61 GAO- 02- 62 Social Security and Income Adequacy

have resources below a level that is specified by a state (or federal
government) for its public assistance program.

The dependency on others appears to be the basis on which President
Roosevelt?s Committee on Economic Security made its recommendations in 1935.
Supporting materials prepared for the committee indicate that it used a
?danger line? amount that was used in some of the states for their old- age
public assistance programs. The danger line was an amount ($ 300 per year)
that placed older persons in a dependent class. As an example of this
adequacy measurement method we use the Supplement Security Income (SSI)
program and its predecessor the old- age assistance program.

The federal SSI program was created to provide a positive assurance that the
nation?s aged, blind, and disabled people would no longer have to subsist on
below poverty- level incomes. SSI was conceived as a guaranteed minimum
income for the aged, blind, and disabled. It was to supplement the Social
Security program and to provide for those who were not covered or minimally
covered under Social Security or who had earned only a minimal entitlement
under the program. In 1972, SSI replaced the federal- state old- age
assistance programs in which state benefit amounts were matched by the
federal government up to a specified monthly amount. Under those programs
the states were able to set benefit amounts and the basis for those amounts
was unclear.

The purpose of a measure that examines the receipt of public assistance is
to determine if the person is dependent upon others for his/ her economic
well- being. 35

In staff reports prepared for the 1934 Committee on Economic Security, the
dependency on others is used as a measure of inadequate income. For example,
Edwin Witte, executive director for the committee, estimated that 2.7
million of the 6.5 million persons 65 and older were supported by others,
including those who obtained public assistance.

The National Resources Planning Board in 1942 used the receipt of public
assistance to determine whether old- age and survivors benefits that were

35 Public assistance measures are neither absolute nor relative. They do not
involve a list of necessary goods and services that has been identified by
experts or scientific standards, nor do they change with current income or
consumption. Purpose and Uses

Appendix I: Descriptions of Adequacy Measurement Methods

Page 62 GAO- 02- 62 Social Security and Income Adequacy

payable in 1940 were adequate for the needs of the recipients. The board
said that a large volume of supplementation of social insurance benefits by
other forms of aid would lead it to conclude that insurance payments were
not adequate for a considerable proportion of qualified workers.

The measure is simply the number of persons who receive public assistance.
SSA administrative data are used to determine the number of persons who
receive federally administered SSI benefits. The number of SSI recipients is
continually updated with administrative data. SSA publishes the data
quarterly and annually. The data are published for the United States and by
state. 36

By the nature of SSI?s benefit structure and eligibility criteria,
administrative data can be used to identify the type of family unit, or lack
of, in which the recipients live. For example, there are different benefit
levels for couples, individuals living alone, recipient living in someone?s
household, or individuals in a Medicaid facility.

Public opinion polls have been used to solicit subjective estimates from
individuals on the amount of income that one needs to live. The concept
underlying a public opinion poll to ascertain a subjective measure of
adequacy is that individuals are able to tell a pollster what the minimum
amount of income (or consumption) is that people need to maintain a
minimally adequate level of living. Subjective measures of adequacy are
grounded in the everyday and necessarily subjective perceptions of typical
individuals as to the material requirements associated with differing levels
of economic well- being. The direct question approach is based on the
assumption that people are the experts on the needs of their families and/
or those living in their communities.

The only relatively consistent series of money amounts corresponding to a
living- standard threshold based on judgment of representative samples of
the public is one developed by the Gallup polling organization. 37 The

36 For example, Annual Statistical Supplement, 1999, to the Social Security
Bulletin,

Social Security Administration. 37 The following study, which used the
Gallop data, is often cited: Denton R. Vaughan,

?Exploring the Use of the Public?s Views to Set Income Poverty Thresholds
and Adjust Them Over Time,? Social Security Bulletin, Vol. 56, No. 2 (Summer
1993) pp. 22- 46. Construction of the

Measure Public Opinion

Appendix I: Descriptions of Adequacy Measurement Methods

Page 63 GAO- 02- 62 Social Security and Income Adequacy

subjective measure presented here is the ?get- along? measure that was
collected by the Gallop Organization.

The purpose of a subjective measure of well- being that has been obtained
through a public opinion poll is to track the size of groups enjoying
different standards of living. 38 To do so, the societal views about the
income levels required to support alternative living levels are compared
with average levels of family economic resources.

The primary use of the subjective measure has been to demonstrate the
absolute nature of the official poverty thresholds. For example, the
Committee on National Statistics of the National Academy of Science study
panel and researchers compared trends in the official poverty threshold,
one- half of family median income, and the get- along amount to document
that the official poverty measure is no longer consistent with the society?s
definition of measures of need. Subjective measures have also been used to
produce subjective minimum income thresholds.

The responses to the following question are used as the subjective measure:
?What is the smallest amount of money a family for four (husband, wife, and
two children) needs each week to get along in this community?? The response
when converted into an annual amount is generally referred to as the ?get
along? amount. The Gallup Organization queried samples of adults about the
get- along amount 38 times from 1946 through 1992. There was no regular
publication of the data.

Although the get- along question was asked in the context of the
respondent?s community, no presentation has been made of geographic
differences among the values reported.

Other than periodically making an inquiry through a poll or survey, there is
no method to update the public opinion measure.

38 Subjective measures are neither absolute nor relative. They are not lists
of goods and services identified by experts or with scientific standards.
Neither do they change with current income or consumption. Purpose and Uses

Construction of the Measure

Appendix I: Descriptions of Adequacy Measurement Methods

Page 64 GAO- 02- 62 Social Security and Income Adequacy

As part of a study of subjective assessments of economic well- being,
researchers at the Bureau of Labor Statistics found that respondents have
definite emotional reactions to their financial situations and are willing
and able to discuss them. 39 They also found that the terms used in
subjective questions were ambiguous. In addition, if the respondent was the
designated bill payer, the person?s responses were found to differ from
those in the family who did not pay the bills.

One common measure of retirement income adequacy is the replacement rate,
which represents the income in retirement for a single worker or household
in relation to a measure of pre- retirement earnings, such as earnings in
the year before retirement.

The purpose of the earnings replacement rate is to compare the level of
retirement income with the level of pre- retirement income to help
illustrate the extent to which pre- retirement standards of living can be
sustained in retirement for particular individuals or households. The
replacement rate is a relative measure in that it is relative to an
individual?s or household?s own income, not to some absolute standard of
adequacy.

The earnings replacement rate has been used both with respect to Social
Security and to employer- sponsored pensions. As noted in this report, the
Social Security benefit formula is defined in a way that focuses on
replacing earnings. When calculating replacement rates, SSA typically uses
the ratio of initial Social Security benefits to pre- retirement covered
earnings. A number of researchers have used replacement rates in analyzing
Social Security benefits for many years. Also, an SSA actuarial note
observes that ?policymakers are interested in replacement ratios: (1) as a
means of communicating to prospective beneficiaries approximately how much
they can expect to receive from Social Security, relative to their earnings;
and (2) as a means of deciding if and how the Social Security program should
be changed to meet the needs and desires of the public?? 40

39 This study was not of the Gallup get- along question. It examined four
self- assessment questions that had been used in other research. 40 Joseph
A. Applebaum and Joseph F. Faber. ?The Concept of Replacement Ratios Under
Social Security.? SSA Actuarial Note No. 96, July 1979. Observations About
the

Measure Earnings Replacement Rates

Purpose and Uses

Appendix I: Descriptions of Adequacy Measurement Methods

Page 65 GAO- 02- 62 Social Security and Income Adequacy

Replacement rates have also been used with respect to employersponsored
pensions and retirement income more broadly, using total income amounts in
the ratio. For example, available data suggest that typical pension
replacement rates for a 30- year career worker have been in the 20- to 40-
percent range across the earnings distribution and that lower earners have
received slightly higher replacement rates than higher earners. 41 More
generally, many benefit professionals currently consider a 70 to 80 percent
replacement rate as adequate to preserve the preretirement living standard.
In contrast, Social Security replacement rates for workers who retired in
2001 at age 65 with a history of average earnings had a replacement rate of
roughly 40 percent.

Construction of replacement rates raises a variety of methodological issues,
most notably, how retirement income is measured, how preretirement income is
measured, how the two are compared, and for whom. How these issues are
addressed depends on the purpose at hand. For example, in measuring
retirement income, some researchers feel that income in the first year of
retirement should be used, rather than trying to reflect changes in
retirement income over time. In measuring preretirement income, some
researchers use income in the year immediately before retirement. In
comparing the two, the two measures should be consistent with one another,
for example, with respect to before- or aftertax status. For whom the
comparison is made might include specific individuals or households for
their own retirement planning purposes, illustrative workers such as the
steady- earners used in figures 4 and 7 of this report, or some sample of
individuals or households in the population. If the purpose of the analysis
is to isolate the effects of certain program changes, then the use of
illustrative steady earners in which all are assumed to retire at a given
age might be appropriate. In contrast, if the purpose is to describe the
experience of a population, then using a sample might be appropriate.

The issues of updating and geographic variation are not especially
applicable to the replacement rate by its nature. It is a ratio that is
relative to the earnings of the individuals or households examined, which
themselves change across cohorts.

41 Olivia S. Mitchell, ?New Trends in Pension Benefit and Retirement
Provisions,? Working Paper No. 7381 (Cambridge, Mass.: National Bureau of
Economic Research, Oct. 1999). Construction and Updating

the Measure

Appendix I: Descriptions of Adequacy Measurement Methods

Page 66 GAO- 02- 62 Social Security and Income Adequacy

While replacement rates can be useful for some purposes, such as
illustrating the effects of program changes over time, the meaning of a
specific value of a replacement rate is not clear. For example, a very low
earner could have a high replacement rate and still have very low income,
while a high earner could have a low replacement rate and live quite
comfortably. Thus, desired or target replacement rates can vary
significantly by income level and other factors. 42 Also, the standard that
pension professionals consider an adequate replacement rate has changed over
the years. While a 50 percent replacement rate might have been considered
adequate in the 1930s, when Social Security was instituted, many benefit
specialists and researchers would apply a higher standard today. 43
Moreover, the actual experience of a given household could easily involve
phased- in retirement or situations where one spouse retires while the other
continues to work. Such irregularities present problems in interpreting
replacement rates for actual households.

42 See Bruce A. Palmer, ?Retirement Income Replacement Ratios: An Update,?
Benefits Quarterly, 2nd Quarter, 1994. 43 See Sylvester J. Schieber, ?The
Employee Retirement Income Security Act: Motivations, Provisions, and
Implications for Retirement Income Security,? paper presented at ?ERISA

After 25 Years: A Framework for Evaluating Pension Reform,? Washington, D.
C., Sept. 17, 1999, pp. 8- 9. Observations About the

Measure

Appendix II: Characteristics of Adequacy Measures Used in Our Analyses

Page 67 GAO- 02- 62 Social Security and Income Adequacy

We examined the characteristics of the 11 measures, which are described in
appendix I, that might help examine income adequacy. Through this
examination, we determined that each had limitations that precluded using
any single measure by itself for our analyses. Given these limitations, we
selected four measures that would, as a group, be more appropriate measures
for our analyses. These are the current poverty thresholds, median family
income, public assistance, and earnings replacement rates. Public assistance
and earnings replacement rates reflect the concern that the framers of the
Social Security Act had about dependency on others and a means to support
people who no longer worked. The current poverty thresholds and median
family income, respectively, provide a lower and upper bounds of the
congressional expectation for Social Security to provide more than a minimal
subsistence level, which is at a level above that estimated by the current
poverty thresholds.

We decided not to use three measures- family budgets, material hardship, and
per capita personal income- because they were outdated or because they did
not allow us to make the comparisons our analyses required. We elected not
to use the family budgets measure because the database on which it was
constructed was 40 years old and because it was no longer officially
published. We elected not to use the material hardship measure because it
produced a nonmonetary value that could not be compared to Social Security
benefit amounts or income dollar amounts. We chose not to use per capita
personal income because by definition it includes income other than that
held by people, specifically, money income held by nonprofit institutions
and pension plans.

In examining the four measures we used, we determined that each had
limitations that precluded using any single measure by itself. Below, we
document the recognized limitations of each for use in our analyses.

Several limitations have been identified regarding the use of current
poverty thresholds for estimating the number of people who live in poverty
each year. Some of these limitations were identified as a result of two
federally sponsored studies in the 1970s and 1990s. Although these studies
did not assess the thresholds as an adequacy measure, the limitations they
identified shed light on the thresholds? ability to identify those whom do
not have the resources to meet subsistence or minimal needs. We also include
concerns expressed by the developer of the current poverty thresholds.
Appendix II: Characteristics of Adequacy

Measures Used in Our Analyses Current Poverty Thresholds

Appendix II: Characteristics of Adequacy Measures Used in Our Analyses

Page 68 GAO- 02- 62 Social Security and Income Adequacy

A 1976 Department of Health, Education, and Welfare (HEW) mandated study of
poverty measures noted that several limitations stemmed from the fact that
the current thresholds were based on one needs standard- food- and its costs
in relation to other nonfood expenditures. The HEW study stated that other
than food there were no other commonly accepted standards of need. In
addition, it noted that the amount of money a family spends on food was only
an approximation of a family?s food needs. The report also stated that the
multiplier that was applied to the food costs was a rough measure of nonfood
requirements.

According to two federally sponsored studies, some of the limitations of the
current poverty thresholds relate directly to their inability to reflect
changes in living standards. The poverty thresholds are an absolute measure
in which the mix of goods and services the thresholds represent has not been
changed for nearly 40 years and, therefore, are not consistent with
prevailing American standards of living. Although the current poverty
thresholds are updated by price changes as reflected in the Consumer Price
Index (CPI), as indicated in these two studies, the items that are updated
reflect a mid- 20th century mix in terms of quality and quantity of goods
and services.

The current poverty thresholds do not reflect how the proportion of income
dedicated to food has changed with rising living standards, according to a
1995 study panel of Committee on National Statistics of the National Academy
of Science (NAS). A research study illustrates how living standards based on
food rise over time- as the population becomes more prosperous, on average,
it devotes a smaller proportion to food expenditures and larger proportions
to nonfood expenditures. The study recalculated the poverty thresholds using
USDA?s 1965 Household Survey to determine the portion of family income
dedicated to food purchases and USDA?s 1975 Thrifty Food Plan to approximate
the cost of food. The thresholds re- estimated for 1977 were about 40
percent higher for the 1person households and about 20 percent higher for 4-
person families. A recent study estimated that the poverty thresholds for 4-
person families would have been 68 percent higher in 1987 if they had been
recalculated with methodology similar to that used to develop the current
poverty thresholds.

The 1976 HEW study and the 1995 NAS study panel noted that, although the
current poverty thresholds are updated by changes in prices paid by
consumers, they do not change with the standard of living. The 1995 NAS
study panel said that the thresholds do not incorporate changes in total
consumption that include spending on luxuries, as well as necessities, or

Appendix II: Characteristics of Adequacy Measures Used in Our Analyses

Page 69 GAO- 02- 62 Social Security and Income Adequacy

declines in the standard of living. The 1976 HEW study noted, however, that
the current poverty thresholds were updated using a relative means- changes
in prices- using the CPI. However, the developer of the current poverty
thresholds voiced concern about updating the thresholds with the CPI. She
noted the uncertainty about the appropriateness of the CPI as a measure of
price changes for the poor. She doubted that one price index could capture
how families at different income levels adjust their spending to accommodate
to price changes. For example, poor families may react to a 10 percent
increase in the price of utilities by reducing expenses in other essential
consumption areas; whereas wealthy families would have more options and
could address the increase in a different manner.

Another limitation of the current poverty thresholds identified by the 1995
NAS study panel is that the thresholds do not account for the fact that
working families pay taxes on their earnings and families on public
assistance do not pay taxes on the cash assistance they receive. According
to the NAS panel, this occurs because the determination of whether or not a
family is poor is based on a comparison of before- tax income with
thresholds based on after- tax income. This comparison ignores the fact that
payment of taxes lowers disposable income. As a result, the comparison of
before- tax income with the current poverty thresholds can make it appear as
if low- income working families are better off than poor families receiving
public assistance. The NAS study panel indicated that this limitation might
affect the manner in which policymakers view the poverty population. For
example, because of the comparison of before- tax income with an after- tax
poverty measure, the adverse effects of tax policy changes for low- income
working families are not captured in the resulting poverty statistics.

The NAS panel identified another limitation of the current poverty
thresholds in that the value of noncash benefits, such as housing subsidies,
are not included as income in the determination of poverty status. According
to the panel, the extent of poverty among the recipients of such benefits is
overstated and the efficacy of government income- support measures is
understated because the current poverty thresholds do not take into account
the receipt of noncash public benefits.

According to the developer of the current poverty thresholds, the thresholds
are inappropriately applied to all types of families. The developer stated
that a major limitation of the thresholds was the failure to differentiate
between a social minimum appropriate for a worker and his family and a more
stringent standard appropriate for a family dependent

Appendix II: Characteristics of Adequacy Measures Used in Our Analyses

Page 70 GAO- 02- 62 Social Security and Income Adequacy

on public assistance. She indicated that the same standard was
inappropriately applied to both types of families.

Furthermore, the developer and the NAS study panel said the current poverty
thresholds inability to address needs that are specific to families with
different living situations was a limitation. The NAS panel stated that the
thresholds do not accurately portray the relative poverty status of working
families with childcare expenses and those without such expenses. The
developer also voiced concern about the tradeoffs that families make and
cited the limitation of the thresholds to address, for example, how higher
expenditures in health care affect other areas of family living. The NAS
panel also said that the thresholds do not distinguish among the health care
needs of different kinds of families or reflect the role of insurance
coverage in reducing families? medical care expenditures.

According to the studies, the current poverty thresholds have limitations in
the manner in which they differentiate for family size and do not account
for geographic differences in the cost of living. The NAS panel questioned
the equivalence scale adjustments for family size- especially thresholds for
single persons and those for aged individuals and couples- because the
composition of families and households has changed since the 1960s. Both the
1976 HEW study and the 1995 NAS report state that the thresholds are limited
in that they do not adjust for interarea price differences and therefore do
not incorporate geographic differences in the cost of living.

Median family income has not been used in any official capacity. Therefore,
only general observations have been documented about its limitations as an
adequacy measure. Limitations are generally expressed in terms of using 50
percent of family median income as a measure of poverty status.

According to one researcher in the field, one limitation concerns the
public?s ability to understand the measure?s income base when it is
accustomed to a measure based on basic needs. The researcher noted that an
income- based measure was less closely linked to the basic concept of
minimum adequacy than an absolute measure. In other words, the public would
have difficulty grasping how it could be a measure of adequacy if it was not
linked to one?s basic needs for food, clothing, and shelter. Median Family

Income and 50 Percent of Median Family Income

Appendix II: Characteristics of Adequacy Measures Used in Our Analyses

Page 71 GAO- 02- 62 Social Security and Income Adequacy

According to the NAS study panel, another limitation is that median family
income changes directly with aggregate income and is difficult for people to
understand its movement when the economy changes. One researcher said that a
relative measure like median family income would fall in real terms during a
recession and that this was less than ideal because the needs of the poor do
not fall similarly. The 1995 NAS study panel also noted the behavior the
measure would demonstrate during recessions and economic upturns and said it
would be hard to explain and justify changes in the measure that are not
simply a reflection of price changes. The researcher noted that opponents of
a relative adequacy measure, such as median family income, say it presents
too much of a moving target for policy assessment purposes and that it is
unreasonable to judge the effectiveness of antipoverty efforts against such
a measure.

Limitations also revolve around how to implement median family income as an
adequacy measure, according to the NAS panel. It noted the problems in
selecting the median family income for a particular family size. The panel
discussed several approaches that have been used to develop an adequacy
measure and limitations of these approaches. For example, it noted that one
approach is to apply an equivalence scale to the income amounts in order to
develop a per capita equivalent income for the reference family. The panel
noted that this approach was sensitive to the particular equivalence scale
that was used. In this report, we used median family incomes by family size
as published by the Census Bureau. For single individuals we used one-
person household median income; for two persons we used two- person family
median income. As noted in Ruggles, this approach also has its limitations
in that median family income has a bell- shaped distribution peaking at the
four- person family size.

Another limitation the NAS panel identified concerned the definition and
sources of income that are used to produce median family income. The NAS
panel noted conceptual problems in using median income as an adequacy
measure because it does not reflect disposable income in the way it handles
taxes, childcare expenses, and other work- related expenses. The NAS panel
also said that median family income does not include noncash benefits, such
as food stamps, but said that is not much of a problem since families at the
median do not generally receive such benefits.

The receipt of public assistance has not been recently reviewed by a group
of experts as an adequacy measure. Therefore, the limitations identified for
this measure are those applicable to the federal- state old- age Public
Assistance

Appendix II: Characteristics of Adequacy Measures Used in Our Analyses

Page 72 GAO- 02- 62 Social Security and Income Adequacy

assistance program. The National Resources Planning Board said, in 1941,
that using the receipt of public assistance as an indicator of whether
Social Security beneficiaries had adequate income had several limitations.
It stated that some of the states, in 1940, were providing a level of living
considerably lower than that provided by Social Security. The board also
reported that some states did not have funds to provide for all of their
needy applicants and chose not to supplement those who received Social
Security benefits.

We used administrative data to report the proportion of the elderly who
received old- age assistance or SSI benefits. We note that some Social
Security beneficiaries who may meet all eligibility criteria may not receive
benefits.

The chief limitation of replacement rates is that the meaning of a specific
value of a replacement rate is not clear. A very low earner could have a
high replacement rate and still have very low income, while a high earner
could have a low replacement rate and live quite comfortably. Also, the
standard that pension professionals consider an adequate replacement rate
has changed over the years. Another important limitation arises in trying to
define replacement rates for actual households. For example, the actual
experience of a given household could easily involve phased- in retirement
or situations where one spouse retires while the other continues to work.
Earnings

Replacement Rates

Appendix III: Defining Benchmarks for Analysis of Future Benefits

Page 73 GAO- 02- 62 Social Security and Income Adequacy

According to current projections of the Social Security trustees for the
next 75 years, revenues will not be adequate to pay full benefits as defined
under current law. Therefore, estimating future Social Security benefits
should reflect that actuarial deficit and account for the fact that some
combination of benefit reductions and revenue increases will be necessary to
restore long- term solvency. To illustrate a full range of possible
outcomes, we developed benchmark policy scenarios that would achieve 75-
year solvency either by only increasing payroll taxes or only reducing
benefits. In developing these benchmarks, we identified criteria to use to
guide their design and selection. We also identified key parameters that
could be used to describe and calibrate the policies to achieve 75- year
solvency. We asked SSA?s Office of the Actuary to score the policies and
determine the precise parameter values that would achieve 75- year solvency
in each case. Once we defined and fully specified our benchmark policies, we
used them to estimate the range of potential future benefit levels using two
representative sample microsimulation models as well as an SSA benefit
calculator for illustrative workers. (See app. IV.)

According to our analysis, appropriate benchmark policies should ideally be
evaluated against the following criteria:

1. ?Distributional neutrality?: the benchmark should reflect current law as
closely as possible while still restoring solvency. In particular, it should
try to reflect the goals and effects of current law with respect to
redistribution of income. However, there are many possible ways to interpret
what this means, such as

a) producing a distribution of benefit levels with a shape similar to the
distribution under current law (as measured by coefficients of variation,
skewness, kurtosis, etc.); b) maintaining a proportional level of income
transfers in dollars; c) maintaining proportional replacement rates; and d)
maintaining proportional rates of return.

2. Demarcating upper and lower bounds within which the effects of
alternative proposals would fall. For example, one benchmark would reflect
restoring solvency solely by increasing payroll taxes and therefore
maximizing benefit levels while another would solely reduce benefits and
therefore minimize payroll tax rates.

3. Ability to model: the benchmark should lend itself to being modeled
within the GEMINI and MINT models. Appendix III: Defining Benchmarks for

Analysis of Future Benefits Criteria

Appendix III: Defining Benchmarks for Analysis of Future Benefits

Page 74 GAO- 02- 62 Social Security and Income Adequacy

4. Plausibility: the benchmark should be politically within reason as an
alternative; otherwise, the benchmark could be perceived as a strawman.

5. Transparency: the benchmark should be readily explainable to the reader.

We used only one tax- increase- only benchmark policy scenario because
policies that only increase payroll tax rates have no effect on benefits.
Our tax- increase- only benchmark would raise payroll taxes once and
immediately (in the next calendar year) by the amount of the OASDI actuarial
deficit as a percent of payroll. It results in the smallest ultimate tax
rate of those we considered and spreads the tax burden most evenly across
generations; this is the primary basis for our selection. The later that
taxes are increased, the higher the ultimate tax rate needed to achieve
solvency, and in turn the higher the tax burden on later taxpayers and lower
on earlier taxpayers. We consider this policy to be plausible because it
would involve less than a 1 percentage point increase on employers and
employees each. Still, any policy scenario that achieves 75- years solvency
only by increasing revenues would have the same effect on the adequacy of
future benefits in that promised benefits would not be reduced.
Nevertheless, alternative approaches to increasing revenues could have very
different effects on individual equity.

We developed three benefit- reduction benchmarks for our analysis. For ease
of modeling, all benefit- reduction benchmarks take the form of reductions
in the PIA formula factors; they differ in the relative size of those
reductions across the three factors, which are 90, 32, and 15 percent under
current law. Each benchmark has three dimensions of specification: scope,
phase- in period, and the factor changes themselves.

For our analysis, we want the benefit reductions in our benchmarks to apply
very generally to all types of benefits, including disability and survivors
benefits as well as old- age benefits. Our objective is to find policies
that achieve solvency while reflecting the distributional effects of the
current program as closely as possible. Therefore, it would not be
appropriate to reduce some benefits and not others. If disabled and survivor
benefits were not reduced at all, reductions in other benefits would be
deeper than shown in this analysis. Tax- Increase- Only

Benchmark Policies Benefit- ReductionOnly Benchmark Policies

Scope

Appendix III: Defining Benchmarks for Analysis of Future Benefits

Page 75 GAO- 02- 62 Social Security and Income Adequacy

We selected a phase- in period that begins with those reaching age 62 in
2005 and continues for 30 years. We chose this phase- in period to achieve a
balance between two competing objectives: 1) minimizing the size of the
ultimate benefit reduction and 2) minimizing the size of each year?s
incremental reduction to avoid notches and unduly large incremental
reductions. Since later birth cohorts are generally agreed to experience
lower rates of return on their contributions already under current law,
minimizing the size of the ultimate benefit reduction would minimize further
reductions in later cohorts? rates of return. The smaller each year?s
reduction, the longer it will take for benefit reductions to achieve
solvency and in turn, the deeper the eventual reductions will have to be.
However, the smallest possible ultimate reduction would be achieved by
reducing benefits immediately for all new retirees by over 10 percent; this
would create a huge notch, that is, creating some marked inequities between
beneficiaries close in age to each other.

Our analysis shows that a 30- year phase- in should produce incremental
annual reductions that would be of palatable size and avoid significant
notches. Therefore it would be preferable to longer phase- in periods, which
would require deeper ultimate reductions.

In addition, we feel it is appropriate to delay the first year of the
benefit reductions for a few years because those within a few years of
retirement would not have adequate time to adjust their retirement planning
if the reductions applied immediately. The Maintain Tax Rates (MTR)
benchmark in the 1994- 96 Advisory Council Report also provided for a
similar delay. 1

When workers retire, become disabled, or die, Social Security uses their
lifetime earnings records to determine each worker?s Primary Insurance
Amount (PIA), on which the initial benefit and auxiliary benefits are based.
The PIA is the result of two elements- the Average Indexed Monthly Earnings
(AIME) and the benefit formula. The AIME is determined by taking the
lifetime earnings earnings record, indexing it, and taking the average. To
determine the PIA, the AIME is then applied to a step- like formula, shown
here for 2001.

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

Defining the PIA Formula Factor Reductions

Appendix III: Defining Benchmarks for Analysis of Future Benefits

Page 76 GAO- 02- 62 Social Security and Income Adequacy

PIA = 90% ! (AIME 1 $561) + 32% ! (AIME 2 > $561 and $3381) + 15% ! (AIME 3
> $3381) where AIME i is the applicable portion of AIME. All three of our
benefit- reduction benchmarks are variations of changes in PIA formula
factors and all are special cases of the following generalized form, where F
i represents the 3 PIA formula factors, which are 90, 32, and 15 percent
under current law.

F i t+ 1 = F i

t - (F i

2001 ! x ! weight x ) - y ! weight y

where

t = the year of the factor,

x = constant proportional benefit reduction,

y = constant ?subtractive? benefit reduction, and weight x and weight y
determine the relative effects of x and y and sum to 1. Our three potential
benchmarks can now be described as follows:

Proportional Offset: weight x = 1 and weight y = 0. The value of x is
calculated to achieve 75- year solvency, given the chosen phase- in period
and scope of reductions.

The formula specifies that the proportional reduction is always taken as a
proportion of the base year factor value rather than the prior year. This
maintains a constant rate of benefit reduction from year to year. In
contrast, taking the reduction as a proportion of the prior year?s factor
value implies a decelerating of the benefit reduction over time because the
prior year?s factor gets smaller with each reduction. To achieve the same
level of 75- year solvency, this would require a greater proportional
reduction in earlier years because of the smaller reductions in later years.

The proportional offset hits lower earners especially hard because the
constant x percent of the higher formula factors results in a larger
percentage reduction over that segment of the formula, while the higher

Appendix III: Defining Benchmarks for Analysis of Future Benefits

Page 77 GAO- 02- 62 Social Security and Income Adequacy

formula factors apply to the lower earnings segments of the formula. For
example, in a year when the cumulative size of the proportional reduction
has reached 10 percent, the 90 percent factor would then have been reduced
by 9 percentage points, the 32 percent factor by 3.2 percentage points, and
the 15 percent factor by 1.5 percentage points. As a result, earnings below
the first bendpoint would be replaced at 9 percentage points less than
current law, while earnings above the second bendpoint would be replaced at
only 1.5 percentage points less than current law. Still, the proportional
offset is easily described as a constant percentage reduction of current law
benefits for everyone. In the example, beneficiaries of all earnings levels
would have their benefits reduced by 10 percent.

Progressive Offset: weight x = 0 and weight y = 1. The value of y is
calculated to achieve 75- year solvency, given the chosen phase- in period
and scope of reductions.

This offset results in equal percentage point reductions in the formula
factors, by definition, and subjects earnings across all segments of the PIA
formula to the same reduction. Therefore, it avoids hitting lower earners
especially hard as the proportional offset does.

As it happens, this offset produces exactly the same effect as the offset we
used in our 1990 analysis of a partial privatization proposals. 2 In that
analysis, we were charged with finding a benefit reduction that would leave
the redistributive effects of the program unchanged while allowing a
diversion of 2 percentage points of contributions into individual accounts.
We calculated these benefit reductions by computing the Social Security
annuity value of the balance of a hypothetical account that earned interest
on the diverted contributions at the rate of return for each individual?s
cohort as a whole. We demonstrated the distributional neutrality of this
benefit reduction by showing that if all individuals earned exactly the
cohort rate of return on their individual accounts, then their income under
the proposal from Social Security and the new accounts would be exactly the
same as under current law.

The hypothetical account approach to reducing benefits translates into our
PIA factor changes because such a reduction is proportional to the AIME,

2 Social Security: Analysis of a Proposal to Privatize Trust Fund Reserves.

GAO/ HRD- 91- 22, Dec. 12, 1990.

Appendix III: Defining Benchmarks for Analysis of Future Benefits

Page 78 GAO- 02- 62 Social Security and Income Adequacy

not to the PIA. The contributions to a hypothetical account are proportional
to earnings. Therefore, a benefit reduction based on such an account would
also be proportional to earnings; that is

Benefit reduction = y !AIME

Therefore, the new PIA would be PIA new =90% ! AIME 1 + 32% ! AIME 2 + 15% !
AIME 3 - y ! AIME T Where AIME i is the applicable portion of AIME and AIME
T is the total AIME. In turn,

PIA new =( 90% - y) ! AIME 1 + (32% - y) ! AIME 2 + (15% - y) ! AIME 3 Thus,
the reduction from a hypothetical account can be translated into a change in
the PIA formula factors.

Because this offset can be described as subtracting a constant amount from
each PIA formula factor, it is reasonably transparent, especially in
comparison to describing it as a hypothetical account offset.

Limited Proportional Offset: Other analyses have addressed the concern about
the effect of the proportional offset on low earners by modifying that
offset to apply only to the 32 and 15 percent formula factors. The MTR
policy in the 1994 to 1996 Advisory Council Report used this approach, which
in turn was based on the Individual Account (IA) proposal in that report.
However, the MTR policy also reflected other changes in addition to PIA
formula changes. Our recent report on disability and Social Security reform
also used this ?limited proportional? approach but using PIA formula changes
alone to achieve solvency. 3 In addition, both the Advisory Council and our
analysis favored those closer to retirement by having a smaller percentage
reduction for the first several years and a reduction that was 1 percentage
point higher after those several years and until the end of the phase in.
For simplicity, we apply the first percentage reduction for the first 10
years of the phase- in.

3 Advisory Council on Social Security. Report of the 1994- 1996 Advisory
Council on Social Security, Vols. 1 and 2. Washington, D. C.: Jan. 1997.
Social Security Reform: Potential Effects on SSA?s Disability Programs and
Beneficiaries (GAO- 01- 35, Jan. 24, 2001).

Appendix III: Defining Benchmarks for Analysis of Future Benefits

Page 79 GAO- 02- 62 Social Security and Income Adequacy

Using the generalized form above, this can be expressed as weight x = 1,
weight y = 0

F 90

t = F 90 2001 for all t

F i t+ 1 = F i

t - (F i

2001 ! x p ! weight x ) - y ! weight y

for i = {32,15}, where x p differs for the first 10 and second 20 years of
the phase- in period and is 1 percentage point higher in the second part
than in the first.

Table 3 summarizes the features of our four benchmarks.

Table 3: Summary of Benchmark Policy Scenario Parameters Annual PIA factor
reduction

(percentage point) Ultimate PIA factor (2035) (percent)

Benchmark policy scenario Phase- in

period 90 percent factor 32 percent

factor 15 percent factor 90 percent

factor 32 percent factor 15 percent

factor

Tax- increase- only 2002 0.00 0.00 0.00 90.00 32.00 15.00 Proportional
benefit reduction 2005- 2035 0.71 0.25 0.12 68.10 24.21 11.35 Progressive
benefit reduction 2005- 2035 0.32 0.32 0.32 80.11 22.11 5.11

2005- 2014 0.00 0.27 0.13 Limited proportional benefit reduction 2015- 2035
0.00 0.59 0.28 90.00 16.95 7.94

Source: GAO?s analysis as scored by SSA actuaries.

Appendix IV: Description of Approaches Used to Forecast Future Benefits

Page 80 GAO- 02- 62 Social Security and Income Adequacy

For our analysis of future Social Security benefits, we used two alternative
policy microsimulation models and illustrative worker analysis. We used the
MINT (Modeling Income in the Near Term) model, developed and used by the
Social Security Administration?s Office of Policy, and the GEMINI model,
developed by the Policy Simulation Group. For both models the developers
produced multiple output data sets based on the PIA formula changes
specified by the policy benchmarks. 1 This appendix will briefly describe
the two models and the illustrative worker analysis and illustrate their
different characteristics for analysis of future benefits. While all methods
of analysis were carefully chosen for their unique qualities, presenting
results using three different analytical tools allows for different
perspectives on the uncertainty of the future.

MINT is a detailed microsimulation model developed and used by the Social
Security Administration?s Office of Policy. The MINT model projects
demographic changes, retirement income, and Social Security benefits for the
cohort of persons born between 1926 and 1965. It provides SSA with the
capability to assess the distributional impact of changes to the Social
Security program. The base data sets used in the model are 1990- 93 panels
of the Census Bureau?s Survey of Income and Program Participation (SIPP),
matched to Social Security Earnings Records (SER) and Master Beneficiary
Records (MBR). The SERs give earnings histories for the years 1951 to 1998.
MINT uses data on the matched files for individuals in the 1931 to 1960
birth cohorts to project their incomes. In some cases, additional data are
used to project income and demographic characteristics. 2

MINT estimates earnings histories for persons in the sample who have not yet
completed their careers. MINT also projects the year of initial receipt of
Social Security benefits and benefit amounts, in addition to the other
sources of retirement income including pensions, asset income, and earnings
of working Social Security beneficiaries. MINT does not estimate any
interactions between changes in Social Security and other income

1 See appendix III for more information on the policy benchmark results. 2
For a complete description of the MINT model and projections see The Urban
Institute, Final Report: Modeling Income in the Near Term - Projections of
Retirement Income Through 2020 for the 1931- 1960 Birth Cohorts, 1999, and
RAND, Final Report: Near Term Model Development Part II, 1999 (http:// www.
ssa. gov/ policy/ policyareas/ evaluation/ MINT/ UI/). Appendix IV:
Description of Approaches

Used to Forecast Future Benefits MINT

Appendix IV: Description of Approaches Used to Forecast Future Benefits

Page 81 GAO- 02- 62 Social Security and Income Adequacy

sources. For example, assuming no change in consumption during working
years, our tax increase benchmark may overestimate total retirement income
because no provision is made to decrease income from saved assets that might
diminish as higher payroll taxes reduce disposable income before retirement.

The MINT model has not been well validated against other micro or
macroeconomic projection models. However, SSA analysts note that there are
not many models against which to validate MINT. Moreover, they note a panel
of demographers, economists, and outside experts oversaw the development of
MINT. Additionally, the 1990 to 1993 SIPP data are the most recent available
SIPP data for most income sources other than earnings. In short, more recent
nonearnings income data would be ideal. Nevertheless, the intention of this
report is to present comparisons of distributions between policy benchmarks.
Thus, income- related MINT point estimates should not be considered as
literally as differences between the policy benchmarks.

Methodologically we chose MINT for this report for

 its capability to project total income and therefore permit analysis of
the adequacy of total income;

 its ability to prospectively assess and model various Social Security
programmatic alternatives;

 its ability to examine a large portion (those age 60 to 89) of the Social
Security population at a point in time;

 its ability to examine various subgroups, notably by race and ethnicity;
and

 its use as a policy tool already employed by SSA. GEMINI 3 is a policy
microsimulation model developed by the Policy Simulation Group (PSG). For
our report, PSG produced simulated samples, sometimes called synthetic
samples, of lifetime histories, including earnings, marriage, disability,
death, and Social Security benefits, for the cohorts born in 1935, 1955,
1970, and 1985. Key descriptive statistics for each of the four birth
cohorts are identified through a variety of sources. These statistics
describe life expectancy, educational attainment, employment patterns, and
marital status at age 60. Where possible these targets are set to be
consistent with the 2001 Trustees? Report or generally available
methodologies from the SSA?s

3 For more information on GEMINI go to http:// www. polsim. com/ GEMINI.
html. GEMINI

Appendix IV: Description of Approaches Used to Forecast Future Benefits

Page 82 GAO- 02- 62 Social Security and Income Adequacy

Office of the Chief Actuary. After the calibration targets are determined,
complete life histories for each birth cohort are produced that match the
targets. These life histories are produced by the Pension Policy Simulation
Model (PENSIM), a complementary PSG model integrated with GEMINI.

Once the cohort samples have been generated, each sample is input into
GEMINI, a microsimulation model that has the same Social Security benefit
calculation capabilities as the microsimulation model of SSASIM, which past
GAO reports have used to analyze Social Security reforms. Each sample is run
twice through each of the our benchmark policies and produces output files
that contain detailed information on each member of the sample, including
Social Security benefits for sample individuals and their spouses. Because
GEMINI cannot yet stochastically determine the age at which a member of the
sample applies for benefits, one output file assumes that the all workers
retire at age 62 and the other assumes that they retire at age 65.

Table 4 shows results for GEMINI compared to the 1998 Annual Statistical
Supplement to the Social Security Bulletin. Average benefits are high by
only 0.9 percent for men and high by only 1.6 percent for women. However
this comparison may suffer from a selectivity problem caused by the fact
that, in the actual data, not everyone eligible to apply for retired worker
benefits does so at age 62. If the propensity to retire early at age 62
varies by lifetime earnings level, then the fact that only about sixty
percent actually apply at 62 will complicate the comparison with statistics
from the GEMINI simulation, which assumes everyone applies at age 62. After
adjusting for the selectivity problem we find that benefits are low by 0.4
percent for men and are low by 4.6 percent for women.

Appendix IV: Description of Approaches Used to Forecast Future Benefits

Page 83 GAO- 02- 62 Social Security and Income Adequacy

Table 4: Comparison of GEMINI Data to Published Social Security Data 1935
Birth cohort statistic

GEMINI value (retiring at

age 62) Social

Security Bulletin

(1998)

Gender composition of awardees in percent:

Men 53.4 53.7 Women 46.6 46.3

Average initial monthly benefit of awardees in dollars:

Men 802 795 (805) Women 517 509 (541) Note: The GEMINI column represents an
output run which assumes everyone born in 1935 retires in 1997 at age 62,
for retired- worker and spouse beneficiaries who are dually entitled. The
Social Security Bulletin Annual Statistical Supplement column represents
actual data on initial retired- worker awards for those born in 1935
retiring at age 62 in 1997 from Tables 6.B1 and 6.B2 in the Social Security
Bulletin?s 1998 Annual Statistical Supplement. The number in parenthesis
represents the selectivity- adjusted amount. This amount adjusts the amount
to reflect the fact that persons retiring at age 62 have lower PIAs than the
average of the age 62 to 65 population by sex.

Methodologically, we chose GEMINI for this report for

 its ability to examine the effect of Social Security programmatic changes
on a cohort population and

 its ability to project cohorts and examine policy effects well out into
the 75 year actuarial period (the year 2050).

The MINT takes real people and projects their behavior out into the future
while GEMINI develops a synthetic sample and validates it to recent data.
While these models were developed separately and take somewhat different
modeling approaches, we can see that actual results compare somewhat
favorably. Table 5 compares median annual Social Security benefit income for
the 1955 cohort by marital status for both models. For married and divorced
individuals, the results compare very favorably as the MINT results fall
within the same range as the GEMINI results. The results for never married
and widowed individuals do not align as nicely, though they are within 8.9
percent and 9.4 percent, respectively, of the lower bound of GEMINI
benefits. However, the intent of the report is not to focus on actual values
produced by the models, but how values change across benchmark scenarios.
Brief Comparison of

the Models

Appendix IV: Description of Approaches Used to Forecast Future Benefits

Page 84 GAO- 02- 62 Social Security and Income Adequacy

Table 5: Median Annual Social Security Income for Individuals in 2020 for
the 1955 Cohort Only by Marital Status: A Comparison of MINT and GEMINI

Median annual Social Security income GEMINI Marital Status MINT Retiring at
age 62 Retiring at age 65

Married a $20,456 $19,281 $20,514 Widowed $13,803 $15,242 $16,948 Never
married $11,745 $12,891 $15,950 Divorced $12,597 $12,030 $14,849 a Benefits
for married persons in the MINT model are equivalence- adjusted to
facilitate comparisons between nonmarried persons and married persons, whose
household income includes income from both spouses that can vary
significantly between them. An equivalence scale is used to adjust married
individuals? income. MINT?s equivalence scale is the ratio of the 1998
poverty threshold for households with two persons aged 65 and older divided
by the threshold for households with one person aged 65 and older. The
benefits for the couple are then divided by this ratio (1.26) to arrive at a
constructed individual benefit amount. To compare like amounts among married
individuals across models, we also equivalence- adjusted GEMINI median
benefits in the same manner.

Note: In the MINT model, workers retire at different ages, while in the
GEMINI model, all workers retire at a fixed age. For this analysis, the
marital status definitions are constructed to be consistent across models.
The distribution of marital status (rounded percentages in parenthesis) for
the MINT sample is married (61. 3), widowed (18.0), never married (5. 0),
and divorced (15.8). The distribution of marital status for the GEMINI
samples are married (64.0), widowed (11.2), never married (9.6), and
divorced (15.2).

Source: GAO?s analysis of the MINT and GEMINI models.

For analysis of future replacement rates, we use four illustrative workers.
These illustrative workers are constructed according to the methodology
employed for steady workers by SSA?s Office of the Chief Actuary.
Additionally, our analysis of future steady workers assumes that the average
wage increases according to Alternative II assumptions of the 2001 Trustees
Report.

As defined by SSA?s Office of the Chief Actuary, the steady earnings pattern
assumes that the worker is a steady full- time employee with no interruption
in employment. The steady worker begins working in covered employment at age
22, and the worker?s earnings increase each year at the same rate as Social
Security?s Average Wage Index. For our analysis, workers are continuously
employed between the ages of 22 and 62 (i. e., they do not experience a
period of disability or die). For the steady earnings pattern, the following
four levels of earnings are used: low (annual earnings equal to 45 percent
of the average wage), average (annual earnings equal to the average wage),
high (annual earnings equal to Illustrative Workers

Appendix IV: Description of Approaches Used to Forecast Future Benefits

Page 85 GAO- 02- 62 Social Security and Income Adequacy

160 percent of the average wage 4 ), and maximum (annual earnings equal to
the OASDI Contribution and Benefit Base). To calculate the worker?s monthly
Social Security benefit, we used SSA?s Office of the Chief Actuary?s ANYPIA
program. 5 Finally, to calculate replacement rates, we annualized the
monthly benefit and divided the result by the worker?s age 64 earnings.

In actuality, the year- to- year earnings of most workers do not follow
steady earnings patterns. However, illustrative steady workers offer the
advantage of showing programmatic variation by utilizing a consistent worker
profile. More realistic lifetime earnings profiles would be more significant
if timing of payroll contributions are important to the worker, such as a
policy of contributing a portion of payroll taxes to individual accounts.
The most important metric of adequacy for a life time earner is the workers?
PIA, which can be arrived from any number of different earnings patterns.
Examination of actual workers PIAs to the illustrative steady worker types
shows that women and men are ?best represented? 6 by different worker types.
Table 6 shows that in 1999 the low earner ?best

represents? female workers as 71.7 percent fall closest to that category and
the high earner best represents male workers as 41.1 percent fall closest to
that category.

4 For certain historical years (1944- 1973), earnings of 160 percent of the
Average Wage Index would exceed the OASDI contribution and benefit base. Any
earnings that exceed this base are thus censored at the base. For all steady
workers retiring in 2017 or later, all earnings for our illustrative workers
are below the OASDI contribution and benefit base.

5 Here again, we used assumptions consistent with the 2001 Trustees? Report.
For more information about the ANYPIA program, see http:// www. ssa. gov/
retire2/ anypia1. htm. 6 For purposes of this discussion, an actual worker
is "best represented" by a particular hypothetical worker if that
hypothetical worker's PIA is closest to the actual worker's PIA. As an
illustration, table 6 shows that 44. 0 percent of actual retirees have a PIA
that is closer to the PIA of the low- earning worker than to any other, and
are thus said to be best represented by the low- earning worker.

Appendix IV: Description of Approaches Used to Forecast Future Benefits

Page 86 GAO- 02- 62 Social Security and Income Adequacy

Table 6: Distribution of Actual Workers Retiring in 1999 by Level, Relative
to PIA Levels for Hypothetical Steady Earnings Cases Retiring in 1999

Percent with PIA a less than PIA for hypothetical case Percent with PIA
closest to

PIA for hypothetical case b Hypothetical case Male Female Total Male Female
Total

Low 10.0 48.4 27.8 20.0 71.7 44.0 Average 33.9 86.5 58.3 28.5 21.7 25.3 High
75.2 98.6 86.0 41.1 6. 2 24.9 Maximum 100.0 100.0 100.0 10.4 0. 3 5.7 a
Primary Insurance Amount. The PIA is the full (i. e., unreduced) monthly
benefit level, which is payable to disabled workers and to retired workers
who become entitled at normal retirement age b May not add to 100 percent
due to rounding.

Source: Orlo R. Nichols, Michael D. Clingman, and Milton P. Glanz. ?Internal
Real Rates of Return under the OASDI Program for Hypothetical Workers.? SSA
Actuarial Note #144, June 2001.

Percentages indicated above reflect the status of workers retiring in 1999.
These percentages would likely be different for workers retiring in earlier
or later years. For instance, the increasing employment rates for women over
the last several decades is expected to result in relatively greater
increases in career- average earnings for women than for men in the future.
Thus, the difference in the distributions of male and female retired workers
by benefit levels is expected to diminish in the future.

Appendix V: GAO Contacts and Staff Acknowledgments

Page 87 GAO- 02- 62 Social Security and Income Adequacy

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

In addition to those named above, Ken Stockbridge, Kimberly Granger, Charles
Ford, Brendan Cushing- Daniels, Nila Garces- Osorio, Kim Reniero, Daniel
Schwimer, and Kathleen Scholl made key contributions to this report.
Appendix V: GAO Contacts and Staff

Acknowledgments GAO Contacts Acknowledgments

(130035)

The General Accounting Office, the investigative arm of Congress, exists to
support Congress in meeting its constitutional responsibilities and to help
improve the performance and accountability of the federal government for the
American people. GAO examines the use of public funds; evaluates federal
programs and policies; and provides analyses, recommendations, and other
assistance to help Congress make informed oversight, policy, and funding
decisions. GAO?s commitment to good government is reflected in its core
values of accountability, integrity, and reliability.

The fastest and easiest way to obtain copies of GAO documents is through the
Internet. GAO?s Web site (www. gao. gov) contains abstracts and full- text
files of current reports and testimony and an expanding archive of older
products. The Web site features a search engine to help you locate documents
using key words and phrases. You can print these documents in their
entirety, including charts and other graphics.

Each day, GAO issues a list of newly released reports, testimony, and
correspondence. GAO posts this list, known as ?Today?s Reports,? on its Web
site daily. The list contains links to the full- text document files. To
have GAO E- mail this list to you every afternoon, go to our home page and
complete the easy- to- use electronic order form found under ?To Order GAO
Products.?

The first copy of each printed report is free. Additional copies are $2
each. A check or money order should be made out to the Superintendent of
Documents. GAO also accepts VISA and Mastercard. Orders for 100 or more
copies mailed to a single address are discounted 25 percent. Orders should
be sent to:

U. S. General Accounting Office P. O. Box 37050 Washington, D. C. 20013

To order by phone: Voice: (202) 512- 6000 TDD: (301) 413- 0006 Fax: (202)
258- 4066

GAO Building Room 1100, 700 4th Street, NW (corner of 4th and G Streets, NW)
Washington, D. C. 20013

Contact: Web site: www. gao. gov/ fraudnet/ fraudnet. htm, E- mail:
fraudnet@ gao. gov, or 1- 800- 424- 5454 (automated answering system).

Jeff Nelligan, Managing Director, NelliganJ@ gao. gov (202) 512- 4800 U. S.
General Accounting Office, 441 G. Street NW, Room 7149, Washington, D. C.
20548 GAO?s Mission

Obtaining Copies of GAO Reports and Testimony

Order by Mail or Phone Visit GAO?s Document Distribution Center

To Report Fraud, Waste, and Abuse in Federal Programs

Public Affairs
*** End of document. ***