SSA Disability: SGA Levels Appear to Affect the Work Behavior of 
Relatively Few Beneficiaries, but More Data Needed (16-JAN-02,	 
GAO-02-224).							 
                                                                 
The Social Security Administration's (SSA) Disability Insurance  
(DI) program paid $50 billion in cash benefits to more than five 
million disabled workers in 2000. Eligibility for DI benefits is 
based on whether a person with a severe physical or mental	 
impairment has earnings that exceed the Substantial Gainful	 
Activity (SGA) level. SSA terminates monthly cash benefit	 
payments for beneficiaries who return to work and have earnings  
that exceed the SGA level--$1,300 per month for blind		 
beneficiaries and $780 per month for all other beneficiaries. GAO
found that the SGA level affects the work patterns of only a	 
small proportion of DI beneficiaries. However, GAO also found	 
that the SGA may affect the earnings of some beneficiaries. About
13 percent of those beneficiaries with earnings near the SGA	 
level in 1985 still had earnings near the SGA level in 1995, even
though the level was increased during that period. The absence of
key information identifying the monthly earnings of		 
beneficiaries, their trial work period status, and whether they  
are blind limited GAO's ability to definitively identify a	 
relationship between SGA levels and beneficiaries' work patterns.
Data limitations also make the effect of the SGA on DI program	 
entry and exit rates difficult to isolate. Although the rate of  
program entry increased in the years immediately following a 1990
increase in the SGA level, it then gradually declined to a level 
below the pre-1990 entry rates. Since 1990, DI exit rates	 
continue to be driven largely by beneficiary death and conversion
to retirement benefits. However, the percentage of all exits	 
caused by improvements in medical conditions or a return to work 
increased slowly, from 1.9 percent in 1985 to 9.2 percent in	 
1996, and then rose dramatically to 19.9 percent in 1997. A	 
substantial increase in the number of continuing disability	 
reviews done by SSA may account, in part, for this 1997 upturn,  
but data limitations preclude GAO from obtaining a full 	 
understanding of the link between the SGA and exit behavior.	 
-------------------------Indexing Terms------------------------- 
REPORTNUM:   GAO-02-224 					        
    ACCNO:   A02641						        
  TITLE:     SSA Disability: SGA Levels Appear to Affect the Work     
Behavior of Relatively Few Beneficiaries, but More Data Needed	 
     DATE:   01/16/2002 
  SUBJECT:   Beneficiaries					 
	     Data integrity					 
	     Disability benefits				 
	     Disability insurance				 
	     Eligibility criteria				 
	     Employees with disabilities			 
	     Federal social security programs			 
	     Income statistics					 
	     Social Security Disability Insurance		 
	     Program						 
                                                                 
	     Substantial Gainful Activity			 

******************************************************************
** 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-02-224
     
United States General Accounting Office

GAO Report to the Committee on Finance, U.S. Senate and the Committee on
Ways and Means, House of Representatives

January 2002

SSA DISABILITY

SGA Levels Appear to Affect the Work Behavior of Relatively Few
Beneficiaries, but More Data Needed

GAO-02-224

Contents

Letter

Results in Brief
Background
The SGA Level Appears to Affect the Work Effort of Relatively Few

Beneficiaries Effects Of SGA on Program Entry and Exit Rates Difficult to

Isolate Conclusions Recommendations for Executive Action Agency Comments and
Our Response

                                     1

                                    2 4

                                     8

17 21 22 23

  Appendix I                           Scope and Methodology                          26
                                            Our Sample                                26
                   Analysis of the Effect of the SGA on Beneficiaries' Earnings       27
               Analysis Of the Effect of the SGA on Program Entry and Exit            28
                                    Limitations in Our Analysis                       29
 Appendix II                        Comments From the Social Security Administration  32

Tables

Table 1: DI Beneficiaries with Earnings Between 75-100 Percent of the
Annualized SGA Level Table 2: Number and Average Annual Earnings of DI
Beneficiaries Who Worked

         Table 3: Subsequent Earnings of Beneficiaries Who Had 1985
                  Earnings Exceeding the SGA Level, 1986-97
         Table 4: Subsequent Earnings of Beneficiaries Who Had 1985

Earnings Between  75-100 Percent  of the SGA,  1986-95 Table  5: Reasons for
Exiting DI

                                     10

                                     11

                                     14

                                   15 21

Figures

Figure 1: Distribution of DI Beneficiaries' Annual Earnings in 1985 and 1997
13 Figure 2: Entry and Exit Rates for the DI Program 18

Abbreviations

AWI average wage index
CWHS Continuous Work History Sample
DI Disability Insurance
SGA Substantial Gainful Activity
SSA Social Security Administration
SSI Supplemental Security Income

United States General Accounting Office Washington, DC 20548

January 16, 2002

The Honorable Max Baucus
Chairman
The Honorable Charles E. Grassley
Ranking Minority Member
Committee on Finance
United States Senate

The Honorable William M. Thomas
Chairman
The Honorable Charles B. Rangel
Ranking Minority Member
Committee on Ways and Means
House of Representatives

The Social Security Administration's (SSA) Disability Insurance program is
the primary federal income program for workers with disabilities, paying
about $50 billion in cash benefits to over 5 million disabled workers in
2000. Eligibility for Disability Insurance benefits is based on whether a
person with a severe physical or mental impairment has earnings that
exceed the Substantial Gainful Activity (SGA) level, which represents
SSA's principal standard for determining whether a disabled individual is
able to work. SSA terminates monthly cash benefit payments for
beneficiaries who return to work (after completing a trial work period)
and have earnings that exceed the SGA level, set at $1,300 per month for
blind beneficiaries and at $780 per month for all other beneficiaries in
2002.

Some researchers and disability advocacy groups believe that the SGA
level serves as a significant work disincentive for Disability Insurance
beneficiaries, with many working beneficiaries "parking," or earning
amounts that are close to, but never exceeding, the SGA level. According
to this view, a large increase in, or elimination of, the SGA level would
result in increased work by this population. However, others believe that
while the SGA may serve as a work disincentive for some beneficiaries,
this disincentive effect is likely to be small in comparison to the various
other work limitations faced by beneficiaries. Rather than emphasizing the
effect of the SGA on the work behavior of those already receiving
Disability Insurance benefits, some of these observers point out that
increasing or eliminating the SGA level could result in a significant
increase in the number of disabled workers entering the Disability

Results in Brief

Insurance program and a reduction in the number exiting the program, placing
additional fiscal stress on the program.

The Ticket to Work and Work Incentives Improvement Act of 1999 requires that
we assess the effects of changes in the SGA level on the Disability
Insurance program. Specifically, we examined (1) the effects of the SGA on
the work patterns of Disability Insurance beneficiaries and (2) the effects
of the SGA on Disability Insurance program entry and exit rates. To assess
these effects, we reviewed the economic and disability literature related to
the effects of the SGA. We also analyzed SSA's Disability Insurance program
data, including the Continuous Work History Sample (CWHS) over the period
1985 through 1997. In addition, we interviewed various SSA policy officials,
academic experts, and representatives from disability advocacy groups. We
performed our work in accordance with generally accepted government auditing
standards from December 2000 to December 2001. (See app. I for a more
detailed discussion of our scope and methodology.)

Our work found that the SGA level affects the work patterns of only a small
proportion of Disability Insurance beneficiaries. On average, about 32,000,
or 7.4 percent, of those Disability Insurance beneficiaries who worked in
any given year during the period 1985 through 1997 had earnings between 75
and 100 percent of the annualized SGA level.1 These beneficiaries comprised
about 1 percent of all Disability Insurance beneficiaries. The proportion of
Disability Insurance workers with earnings in this range of the SGA remained
relatively small even as the total number of Disability Insurance
beneficiaries who worked grew by almost 80 percent from 1985 to 1997. Even
among those beneficiaries who have earnings near the SGA level in any given
year, most experience a substantial decline in earnings over time. For
example, almost half of those with earnings near the SGA level in 1985 had
no earnings by 1989. However, our work also found evidence that the SGA may
affect the earnings of some beneficiaries. About 13 percent of those
beneficiaries with earnings near the SGA level in 1985 still had earnings
near the SGA level in 1995, even though the level was increased during that
period.

1SSA collects annual, rather than monthly, earnings data. However, the SGA
represents a monthly earnings limit. To permit comparison of the monthly
limit to the annual data, we multiplied the monthly SGA amount by 12 to
develop an annualized SGA. For example, the SGA level in 1995 was $500 per
month and the annualized SGA level was $6,000 ($500 X 12). See app. 1 for
further details on our methodology.

However, the absence of key information identifying the monthly earnings of
beneficiaries, their trial work period2 status, and whether they are blind
limit our ability to definitively identify a relationship between SGA levels
and beneficiaries' work patterns.

Data limitations also make the effect of the SGA on Disability Insurance
program entry and exit rates difficult to isolate. While the rate of program
entry increased in the years immediately following a 1990 increase in the
SGA level, it then gradually declined to a level below the pre-1990 entry
rates. Although some researchers and policy makers believe that an increase
in the SGA could encourage more people who are capable of working to enter
the rolls, our analysis indicates that most new Disability Insurance
beneficiaries were either not able or not inclined to increase their
earnings or work at all. However, due to data limitations and the wide range
of other possible factors affecting program entry--such as labor force
responses to the 1990-91 recession and subsequent economic expansion--the
link between the increase in the SGA level and these trends in entry is
unclear. CWHS data indicate that, since 1990, Disability Insurance exit
rates continue to be driven largely by beneficiary death and conversion to
retirement benefits. However, the percentage of all exits caused by
improvements in medical conditions or a return to work increased slowly,
from 1.9 percent in 1985 to 9.2 percent in 1996, and then rose dramatically
to 19.9 percent in 1997. While a substantial increase in the number of
continuing disability reviews3 conducted by SSA may account, in part, for
this 1997 upturn, data limitations preclude us from obtaining a full
understanding of the link between the SGA and exit behavior.

2The trial work period allows a beneficiary to earn any amount for 9 months
(which need not be consecutive) within a 60-month period and still receive
full cash and medical benefits. At the end of the trial work period, if a
beneficiary's earnings exceed the SGA level, cash benefits continue for an
additional 3-month grace period and then stop. For 36 months after the trial
work period ends, referred to as the extended period of eligibility, cash
benefits will be reinstated for any month in which the person does not have
countable earnings above the SGA level.

3SSA periodically conducts continuing disability reviews to verify that an
individual on the rolls still has a disability that prevents that person
from engaging in substantial gainful activity. Continuing disability reviews
may be conducted when (1) substantial earnings are posted to a beneficiary's
employment record, (2) a report of medical improvement is received from a
vocational rehabilitation agency, (3) the beneficiary provides a voluntary
report indicating medical improvement or return to work, or (4) a medical
reexamination is scheduled based on an expectation that a beneficiary's
impairment will improve.

Background

This report contains a recommendation to the Commissioner of SSA concerning
the types of data SSA needs to collect in order to assess the effects of the
SGA on Disability Insurance program beneficiaries. In its written comments,
SSA agreed that it needed to improve its collection of data on Disability
Insurance program beneficiaries' earnings and employment and also provided a
number of technical comments.

From its origin in 1956, the Disability Insurance (DI) program has provided
compensation for the reduced earnings of individuals who, having worked long
enough and recently enough to become insured,4 have lost their ability to
work due to a severe, long-term disability. The program is administered by
SSA and is funded through payroll deductions paid into a trust fund by
employers and workers. In addition to cash assistance, DI beneficiaries
receive Medicare coverage after they have received cash benefits for 24
months. In 2000, about 5 million disabled workers received DI cash benefits
totaling about $50 billion, with average monthly cash benefits amounting to
$787 per person.5

To qualify for benefits, an individual must have a medically determinable
physical or mental impairment that (1) has lasted or is expected to last at
least 1 year or result in death and (2) prevents an individual from engaging
in substantial gainful activity. Individuals are considered to be engaged in
substantial gainful activity if they have countable earnings at or above a
certain dollar level.6 In addition to determining initial eligibility, the
SGA standard also applies to the determination of continuing eligibility for
benefits. Beyond a 9-month trial work period and an additional 3-month grace
period during which beneficiaries are allowed to have any level of earnings
without losing benefits, benefit payments are terminated once SSA determines
that a beneficiary's countable earnings exceed the SGA

4To be eligible for disability benefits, workers must be fully insured and,
except for those who are disabled due to blindness, must also meet a test of
substantial recent covered work. Under this test, workers aged 31 and older
must have credit for work in covered employment for at least 20 quarters of
the 40 calendar quarters ending with the quarter the disability began.
Workers disabled before age 31 may qualify for benefits under a special
insured status requirement.

5In the same year, the DI program also paid about $5 billion in cash
benefits to about 1.6 million spouses and children of disabled workers.

6To calculate countable earnings, SSA deducts from gross earnings the cost
of items that, because of the impairment, a person needs to work (for
example, attendant care services performed in a work setting, wheelchairs,
or Braille devices).

level. DI benefits are also terminated when a beneficiary (1) dies, (2)
reaches age 65, upon which DI benefits are automatically converted to Social
Security retirement benefits, or (3) medically improves, as determined by
SSA through periodic continuing disability reviews.

Under the Social Security Act, the Commissioner of Social Security has the
authority to set the SGA level for individuals who have disabilities other
than blindness. SSA has increased the SGA several times over the past
decade, to $500 per month in 1990 and to $700 per month in July 1999. In
December 2000, SSA finalized a rule calling for the annual indexing of the
nonblind SGA level to the average wage index (AWI)7 and recently increased
the level to $780 on the basis of this indexing. The SGA level for
individuals who are blind is set by statute and indexed to the AWI.8
Currently, the SGA for blind individuals is $1,300 of countable earnings.9

Despite considerable disagreement and uncertainty among researchers, policy
makers, and disability advocates over the employment effects of the SGA on
DI beneficiaries, there is a theoretical basis for believing that the SGA
acts as a work disincentive. That is, to maximize income, maintain health
insurance coverage, or achieve a desirable labor-leisure tradeoff,
beneficiaries may be inclined to limit their work effort to remain eligible
for program benefits. This economic rationale is supported by anecdotal
evidence from some beneficiaries who have reported that, although they would
prefer to work or have greater earnings, they are fearful of doing so
because of the severe financial consequences of exceeding the SGA- losing
cash benefits and, eventually, Medicare benefits. In addition, some workers
with disabilities whose current earnings are above the SGA level,

7The AWI is a measure of average wages of all employees in the United
States.

8The Social Security Act did not initially distinguish between the SGA
levels for blind and nonblind DI beneficiaries. This was changed in 1977
when the Social Security Financing Amendments (P.L. 95-216) set the SGA
level for individuals who are blind equal to the monthly earnings limit set
for Social Security retirees aged 65-69 (a dollar level higher than the SGA
for nonblind beneficiaries). The Senior Citizens' Right to Work Act of 1996
(P.L. 104-121) removed the link between the retirement earnings limit and
the SGA level for the blind. However, the act retained the higher SGA level
for the blind that was in place at that time and allowed for continued
annual indexing to the AWI.

9Blind and nonblind beneficiaries are also treated differently under several
other DI provisions. For example, blind beneficiaries age 55 or older whose
earnings exceed the SGA level are evaluated differently than nonblind
beneficiaries. If the work performed requires a lower level of skill and
ability than work done prior to age 55, benefits are suspended rather than
terminated and will be reinstated in any month that earnings fall below SGA.

making them ineligible for the DI program, may reduce their earnings to
become eligible for DI benefits.

Other researchers and policy makers believe that although the SGA level may
serve as a work disincentive for some beneficiaries, this disincentive
effect is likely to be very limited for several reasons. First, because
severe long-term disability is a central criterion for DI eligibility, many
DI beneficiaries may be unable to perform any substantial work. Even if they
are willing and able to work, beneficiaries may face employment barriers,
such as high costs for supportive services and equipment or discrimination.
In addition, we reported previously that many beneficiaries are unaware of
DI program provisions affecting work,10 and several researchers we spoke
with said that some beneficiaries may not even know how much they are
allowed to earn. In terms of the SGA's effect on those not currently on the
DI rolls, disability advocates have stated that workers turn to the DI
program only as a last resort and are not inclined to reduce income for the
sole purpose of qualifying for benefits. Also, some studies indicate that
the difficulty of qualifying for DI benefits-having to limit or cease work
for at least 5 months before receiving benefits and undergoing a stringent
review to certify one's condition as severely disabled-may itself be a
factor discouraging workers with disabilities from applying for these
benefits.11

Few empirical studies have examined the effects of the SGA on the work
patterns of disabled beneficiaries and nonbeneficiaries. Two studies
conducted in the late 1970s by SSA researchers found that the SGA level does
not have a substantial effect on the work behavior of beneficiaries.12 These
studies examined past increases in the SGA level to assess whether

10See Social Security Disability Insurance: Multiple Factors Affect
Beneficiaries' Ability to Return to Work (GAO/HEHS-98-39, Jan. 1998) and SSA
Disability: Program Redesign Necessary to Encourage Return to Work
(GAO/HEHS-96-62, Apr. 1996).

11Jonathan Gruber and Jeffrey D. Kubik, "Disability Insurance Rejection
Rates and the Labor Supply of Older Workers," Journal of Public Economics,
Vol. 64, Issue 1, 1997, pp. 1-23; Brent Krieder, "Social Security Disability
Insurance: Applications, Awards, and Lifetime Income Flows," Journal of
Labor Economics, Vol. 17, No. 4, Pt. 1, 1999, pp. 784-827.

12Paula A. Franklin, "Impact of the Substantial Gainful Activity Level on
Disabled Beneficiary Work Patterns," Social Security Bulletin, Vol. 39, No.
8, 1976, pp. 20-29; Paula A. Franklin and John C. Hennessey, "Effect of the
Substantial Gainful Activity Level on Disabled Beneficiary Work Patterns,"
Social Security Bulletin, Vol. 42, No. 3, 1979, pp. 3-17.

these increases led to greater labor force participation on the part of DI
beneficiaries. Neither study identified any clear change in beneficiary
earnings as the SGA level increased. However, a study conducted by the
Office of Inspector General (OIG) at the Department of Health and Human
Services (HHS) found that some beneficiaries who had completed a trial work
period subsequently reduced their earnings below the SGA level so they could
continue to receive DI benefits.13 Out of the 100 cases sampled, 18
beneficiaries who were capable of working had quit work or reduced their
earnings to maintain DI benefits. In addition, an internal study conducted
by SSA researchers examined how the earnings patterns of DI beneficiaries
age 55 or older changed after they converted to retirement benefits at age
65.14 This study found that beneficiaries were more likely to return to work
after converting to retirement benefits, which were subject to a more
generous earnings limit. This evidence suggests that the SGA standard leads
some beneficiaries to work less than they could.

Despite the difficulties inherent in comparisons of different programs,
studies of earnings limits in other programs may also provide some insights
on the effect of the SGA. For example, studies of the retirement earnings
test15 indicate that this limit probably caused some retirees to restrain
their earnings in order to avoid having their benefits reduced. However,
this "parking" effect appeared to be limited to only a relatively small
proportion of the retiree population. For example, one study found that only
about 2 percent of insured workers aged 65-69 had earnings at or near the
retirement earnings limit.16

13HHS/OIG, Audit of the Effectiveness of Title II Disability Work
Incentives, A-13-92-00223 (Washington, D.C., 1993).

14John C. Hennessey and L. Scott Muller, "The Search for Evidence of Labor
Supply Response to a Benefit Offset," unpublished manuscript, Nov. 1999.

15The retirement earnings test has undergone a number of changes over the
years. For most of the 1980s, this test resulted in a $1 reduction in Social
Security benefits for every $2 in earnings above an exempt amount for
recipients aged 65-69. In the 1990s, the reduction in benefits was changed
to $1 for every $3 in earnings above the exempt amount for beneficiaries
aged 65-69. The exempt amount-which was automatically adjusted based on
increases in the national average wage index-increased from $6,600 a year in
1983 to $15,500 in 1999. The Senior Citizen's Freedom to Work Act of 2000
(P.L. 106-182) eliminated the retirement earnings test for beneficiaries age
65 and older, effective for taxable years after December 31, 1999.

16Michael V. Leonesio, "Social Security and Older Workers," Social Security
Bulletin, Vol. 56, No. 2, 1993, pp. 47-57. This study examined 1988 earnings
data.

The SGA Level Appears to Affect the Work Effort of Relatively Few
Beneficiaries

A study of the Supplemental Security Income (SSI)17 program's 1619(b)
provision18 also indicates that an earnings limit can result in
beneficiaries limiting their work effort.19 As the 1619(b) earnings
threshold was increased, some SSI beneficiaries increased their earnings in
line with this threshold, which is consistent with the idea that
beneficiaries restrain earnings in order to maintain program (in this case,
Medicaid) eligibility. However, this "parking" behavior was limited to only
those beneficiaries who had significant earnings-a group comprising about 2
percent of all adult, disabled SSI beneficiaries.

Our analysis of SSA data indicates that the work patterns of most DI
beneficiaries are unlikely to be affected by the SGA level. For example,
from 1985 through 1997, on average, about 7.4 percent of DI beneficiaries
who worked had annual earnings between 75 and 100 percent of the SGA level.
These beneficiaries comprised only about 1 percent of the total DI caseload.
The proportion of beneficiaries with earnings in this range of the SGA
remained relatively small even though the number and proportion of DI
beneficiaries who work rose dramatically during this period, increasing by
almost 80 percent. Although almost one-fourth of working beneficiaries had
earnings above the SGA level, most had very low earnings, well below the
annualized SGA level. Even among those beneficiaries with earnings near the
SGA level in a given year, most experience an eventual reduction in earnings
in subsequent years.

Nevertheless, some beneficiaries may change their work effort in response to
the SGA level. For example, we found that about 13 percent of working
beneficiaries who had earnings between 75 and 100 percent of the

17SSI is a means-tested income assistance program for disabled, blind, or
aged individuals who have low income and limited resources. Unlike the DI
program, SSI has no prior work requirement. Eligible SSI applicants
generally begin receiving cash benefits immediately upon entitlement and, in
most cases, receipt of cash benefits makes them eligible for Medicaid
benefits.

18Section 1619 of the Social Security Act allows SSI beneficiaries to keep
Medicaid coverage even when earnings exceed the SGA level. SSI beneficiaries
may keep their Medicaid coverage until earnings increase to a point -
referred to as the threshold amount --that SSA considers high enough to
replace the equivalent of SSI cash and Medicaid benefits.

19The Lewin Group, Inc., "Exploratory Study of Health Care Coverage and
Employment of People with Disabilities: Final Report," prepared for the
Department of Health and Human Services, Office of the Assistant Secretary
for Planning and Evaluation, July 1998.

annualized SGA level in 1985 still had earnings near the SGA level in 1995,
even though the SGA had increased from $300 to $500 a month during this
period. In addition, about 7 percent of beneficiaries who did not have any
earnings in the years immediately preceding their retirement earned income
in the one or more years following retirement, when the SGA earnings limit
no longer applied. However, while these findings are suggestive of a
possible effect on work effort, our analysis could not definitively link
beneficiary work patterns to the SGA level due in part to various
limitations in SSA data, such as the lack of monthly earnings data.

About 1 Percent of All DI Beneficiaries Have Annual Earnings Near the SGA
Level

From 1985 through 1997, on average, about 7.4 percent of DI beneficiaries
who worked -comprising about 1 percent of the total DI caseload - had annual
earnings between 75 and 100 percent of the SGA level (see table 1).20 On an
annual basis, the number of beneficiaries with incomes clustering at or just
below the SGA level increased almost fourfold in absolute terms from 15,800
in 1985 to almost 60,000 in 1997. However, the annual percentage of working
beneficiaries with earnings between 75 and 100 percent of the SGA level
fluctuated from 8.5 percent in 1988 to 5.1 percent in 1990 to 8.9 percent in
1997.

20There are no clear criteria for identifying the cutoff point at which a
beneficiary can be said to be earning "near" the SGA level. Therefore, we
examined beneficiaries' earnings at several increments between 75 and 95
percent of the SGA level. The increment reported here represents the
broadest range (75 percent) that we examined. See app. II for more
information on our methodology.

Table  1:  DI Beneficiaries  with  Earnings  Between 75-100  Percent of  the
Annualized SGA Level

Year

Number with earnings between 75-100 percent of the SGA Percentage of working
                        DI beneficiaries Percentage of all DI beneficiaries

Note: The
annualized SGA
level was $3,600
from 1985-1989
and $6,000 from
1990-1997.
Sampling errors
for the number of
beneficiaries
with earnings
between 75-100
percent of the
SGA from 1985 to
1993 do not
exceed 17 percent
of the value of
those estimates.

Source: GAO analysis of CWHS data.

The proportion of beneficiaries with earnings at or just below the SGA level
remained small even though the proportion of DI beneficiaries who worked
rose dramatically, increasing by almost 80 percent between 1985 and 1997
(see table 2).21 The number of beneficiaries who worked increased from about
220,000 in 1985 to over 675,000 in 1997 and increased as a percent of all DI
beneficiaries in every year, including during the 1990-91 recession.

21We considered a beneficiary to be working in a given year if SSA records
for that individual indicated annual earnings greater than zero. See app. I
for further information on SSA's earnings data and its limitations.

 Table 2: Number and Average Annual Earnings of DI Beneficiaries Who Worked

Year

                                      Number of DI beneficiaries who worked

Percentage of all DI beneficiaries

Mean earnings

Median earnings

Note: Working is
defined as having
posted earnings
greater than
zero. Earnings
are in constant
1997 dollars. The
95-percent
confidence
interval for the
1985 estimate of
median earnings
ranged from 1,812
to 2,246.

Source: GAO
analysis of CWHS
data.

Throughout the period, most working DI beneficiaries had very low earnings.
For example, in 1995, the median annual earnings of working beneficiaries
were about $2,15722 and the majority of working beneficiaries-about 58
percent-earned no more than 50 percent of the annualized SGA level.23
Although median earnings of working DI beneficiaries were about 15 percent
higher in 1997 than they had been in 1985, they remained well below the
annualized SGA level. While mean earnings for this group fluctuated between
a high of $5,851 in 1985 and a low of $4,697 in 1993, figure 1 indicates
that even with the 67 percent

221997 dollars. Figures were adjusted based on the Bureau of Labor
Statistics' Consumer Price Index for All Urban Consumers.

23About 23 percent of beneficiaries who worked in 1995 had earnings above
the annualized SGA level.

increase in the SGA level in 1990, the earnings distribution of DI
beneficiaries did not change considerably from 1985 to 1997.24

24We found that, from 1990 to 1997, only 1 to 2 percent of beneficiaries who
had earnings less than the SGA level prior to 1990 had increased their
earnings to an amount between 75 and 100 percent of the new SGA level. For
beneficiaries who had no earnings prior to 1990, less than half of 1 percent
had earnings between 75 and 100 percent of the SGA level from 1990 to 1997.

Figure 1: Distribution of DI Beneficiaries' Annual Earnings in 1985 and 1997

200,000 Number of Beneficiaries 180,000

160,000

140,000

120,000

100,000

80,000

60,000

40,000

20,000 0 0

901 - 12001 - 300 301 - 600601 - 900 1201 - 15001501 - 18001801 - 21002101 -
24002401 - 27002701 - 30003001 - 33003301 - 36003601 - 39003901 - 42004201 -
              45004501 - 48004801 - 51005101 - 54005401 - 5700

ver 60005701 - 6000oAnnual Earnings (Constant 1997 Dollars)

Note: For purposes of illustration, the bars indicating the number of
beneficiaries are truncated at 200,000. The actual numbers estimated from
the data for beneficiaries with no earnings are 2,306,400 in 1985 and
3,665,500 in 1997.

Source: GAO analysis of CWHS data.

We also examined beneficiaries who had earnings above the SGA level to see
if, over time, they tended to reduce their earnings to an amount less than
but close to the SGA level in order to maintain eligibility for DI benefits.
We found that the majority of beneficiaries in 1985 who had earnings
exceeding the SGA level eventually experienced a reduction to no earnings or
to an amount less than 75 percent of the SGA (see table 3).

     Table 3: Subsequent Earnings of Beneficiaries Who Had 1985 Earnings
      Exceeding the SGA Level, 1986-97 Percentage of 1985 cohort with:

Year

Earnings above the SGA

Earnings between 75-100 percent of the SGA

Earnings between 1-74 percent of the SGA

No earnings

         1986                   55.8                  7.4                 14.2
         1987                   45.7                  6.3                 12.6
         1988                   40.0                  4.7                 11.6
         1989                   41.1                  2.1                  8.4
         1990                   31.6                  5.8                 17.4
         1991                   32.6                  4.2                 10.5
         1992                   32.1                  1.6                 10.0
         1993                   29.0                  4.2                 12.6
         1994                   29.5                  1.1                 13.2
         1995                   28.4                  3.7                 11.1
         1996                   30.0                  1.1                 14.2
         1997                   31.6                  2.1                  9.5

Note: For this analysis, we examined all 190 cases in the CWHS where DI
beneficiaries had earnings above the SGA level in 1985 and remained on the
rolls through 1997. We estimate that, in 1985, about 57,400 DI beneficiaries
(comprising about 2 percent of all DI beneficiaries and about 26 percent of
beneficiaries who worked) had earnings that exceeded the SGA level. We also
estimate that about 19,000 of these beneficiaries remained on the DI rolls
through 1997.

Source: GAO analysis of CWHS data.

By 1989, 48 percent of these individuals had no earnings and only 2 percent
had earnings between 75 to 100 percent of the annualized SGA level. This
indicates that most beneficiaries who at some point have earnings above the
SGA level do not subsequently engage in "parking" to remain on the DI rolls.
Nevertheless, the large shift that we observed from earnings above the SGA
to no or very low earnings does suggest decreasing ability or motivation to
work.

However, as late as 1997, about 32 percent of these beneficiaries had
earnings exceeding the SGA level, indicating that some beneficiaries
maintain their ability to achieve relatively substantial earnings. It is
unclear why these individuals are able to consistently earn above the SGA

level while retaining eligibility for DI benefits. Although beneficiaries in
a trial work period or an extended period of eligibility may have earnings
that exceed the SGA level, these work incentive periods are time-limited.
Only beneficiaries who are blind are permitted, on a continuing basis, to
earn above the SGA level that applies to nonblind individuals. However, we
could not determine the status of individuals who had earnings exceeding the
SGA level because SSA's principal program data do not reliably identify
whether a beneficiary is in a trial work period or extended period of
eligibility and do not contain an indicator denoting whether a beneficiary
is blind.

Most Beneficiaries with Among beneficiaries who have earnings at or near,
but not exceeding, the Earnings Near the SGA SGA level in a given year, most
experience a reduction in earnings in Level Do Not Maintain that subsequent
years. For example, of beneficiaries in 1985 who earned

between 75 to 100 percent of the annualized SGA level, 47 percent had
noLevel of Earnings earnings by 1989, while the earnings of another 26
percent had fallen to between 1 and 74 percent of the annualized SGA level
(see table 4).

Table 4: Subsequent Earnings of Beneficiaries Who Had 1985 Earnings Between
75-100 Percent of the SGA, 1986-95

                      Percentage of 1985 cohort with:

Year

Earnings above the SGA

Earnings between 75-100 percent of the SGA

Earnings between 1-74 percent of the SGA

No earnings

         1986                  15.7                  27.1                  34.3
         1987                  12.9                  14.3                  35.7
         1988                  20.0                  12.9                  25.7
         1989                  15.7                  11.4                  25.7
         1990                  4.3                    8.6                  40.0
         1991                  2.9                   11.4                  34.3
         1992                  4.3                    8.6                  32.9
         1993                  5.7                   10.0                  31.4
         1994                  5.7                    7.1                  32.9
         1995                  2.9                   12.9                  31.4

Note: For this analysis, we examined all 70 cases from the CWHS where DI
beneficiaries had earnings between 75-100 percent of the SGA level in 1985
and remained on the rolls through 1995. We estimate that, in 1985, there
were 15,800 DI beneficiaries (comprising less than 1 percent of all DI
beneficiaries and about 7 percent of beneficiaries who worked) who had
earnings between 75 to 100 percent of the annualized SGA level. We also
estimate that about 7,000 of these beneficiaries remained on the DI rolls
through 1995. Percentages in this table have sampling errors not exceeding
12.4 percentage points.

Source: GAO analysis of CWHS data.

Nevertheless, about 11 percent of these beneficiaries still had earnings in
1989 between 75 to 100 percent of the annualized SGA level, suggesting

that at least some beneficiaries may be attempting to stay close to the SGA
without exceeding it. Even after the SGA level was increased in 1990, a
small proportion of these beneficiaries continued to have earnings between
75 to 100 percent of the new annualized SGA level. For example, in 1995
about 13 percent of beneficiaries who had earnings between 75 to 100 percent
of the annualized SGA level in 1985 still had earnings within this range of
the higher annualized SGA level.

Small Number of Nonworking Beneficiaries Begin Employment After Initial
Receipt of SSA Retirement Benefits

Our review of the earnings of former DI beneficiaries who were converted to
retirement benefits at age 65 also indicates that the work patterns of only
a small proportion of beneficiaries are affected by the SGA. For example, we
looked at DI beneficiaries who converted to retirement benefits at age 65
between 1987 and 1993. Of those in this group who had no earnings in the 3
years preceding retirement, about 7 percent did have earnings in 1 or more
years following retirement (between ages 66 - 68) when the SGA earnings
limit no longer applied.25 While small, the proportion of beneficiaries
returning to work after retirement is greater than the proportion of older
beneficiaries who return to work while still on the DI rolls. For example,
we found that of beneficiaries who had no earnings at ages 55-57, about 3
percent had earnings at ages 58-60. These data suggest that, at least for a
limited number of beneficiaries, the SGA may serve as a disincentive to
work.

Data Limitations Suggest Caution in Ascertaining SGA's Effects

For each analysis, the absence of key data elements made it difficult for us
to determine the effects of the SGA level. For example, because SSA collects
annual rather than monthly earnings data, we could not observe earnings
relative to the SGA level on a monthly basis. However, many workers with
disabilities may engage in only intermittent work throughout the year. The
annual earnings data did not allow us to observe those individuals who only
work several months out of the year and, in order to ensure receipt of
benefits, "park" at the SGA level in those months.

Another data limitation is the difficulty in identifying whether a DI
beneficiary is in a trial work period. Without reliable information on the

25For this analysis, we examined beneficiaries who had entered the DI rolls
prior to age 62, were converted to retirement benefits at age 65 between
1987 and 1993 (the years from which we identified our cohort for this
analysis), and survived to age 68. These beneficiaries comprised, on
average, about 9 percent of all DI beneficiaries. About 93 percent of these
beneficiaries had no earnings in the 3 years preceding retirement.

Effects Of SGA on Program Entry and Exit Rates Difficult to Isolate

trial work period status of beneficiaries, we could not determine the full
range of work incentives and disincentives potentially affecting the
earnings of DI beneficiaries. In addition, neither the CWHS nor SSA's
principal administrative file for the DI program (the Master Beneficiary
Record) contain data that identify whether a beneficiary is blind.26 Such a
distinction is important to analyses relating to the SGA because blind
beneficiaries are subject to a higher SGA limit than nonblind beneficiaries
are. Distinguishing blind and nonblind beneficiaries may help explain why a
substantial proportion of beneficiaries continue to earn above the nonblind
SGA level while retaining DI eligibility.

Data and methodological limitations make it difficult to ascertain the
effect of the SGA on DI program entry and exit rates.27 After 1990, the rate
of program entry initially increased and then gradually declined. Although
some researchers and policy makers believe that an increase in the SGA could
encourage more people who are capable of working to enter the rolls, our
analysis indicates that most new entrants were either not able or not
inclined to increase their earnings or work at all. However, because of data
limitations and the wide range of other possible factors affecting program
entry, the link between the increase in the SGA level and these trends in
entry is unclear. The analysis of program exits indicated that although the
number of beneficiaries exiting the program rose over the 7 years after the
1990 increase in the SGA level, the annual rate of exit generally declined.
While beneficiary deaths and conversions to retirement benefits accounted
for most program exits, the percentage of exits caused by medical
improvement or a return to work increased gradually, from 1.9 percent in
1985 to 9.2 percent in 1996, and then rose sharply to 19.9 percent in 1997.
However, the aggregation of medical improvement and return-to-work data
prevent us from obtaining a full understanding of the link between the SGA
and DI program exit behavior.

26SSA does maintain data identifying whether a beneficiary is blind in
another data set-the 831 Disability File. The 831 file contains data from
initial medical determinations for individuals applying for DI or SSI
benefits. However, we obtained information from SSA indicating that the data
fields in this file that identify blindness may not be reliable in about 20%
of the cases.

27The number of beneficiaries entering the DI program exceeded the number
exiting the program in every year of our analysis. Therefore, the total
number of DI beneficiaries increased from about 2.7 million to 4.5 million
from 1985 to 1997. However, our discussion in this section deals separately
with patterns of program entry and exit, rather than overall program growth.

Annual Program Entry Our analysis showed that the rate of program entry
varied between 1990 Rates Varied from 1990 and 1997, reaching a high of 19.3
percent in 1991 and then gradually through 1997, but declining, except for a
slight upward movement in 1996, to a low of 10.3 Limitations Prevent percent
in 1997 (see figure 2). In 1990, there was a discernible jump in the

rate of program entry, which continued into 1991. The 1990 and 1991
ratesAssessment of Alternative were higher than the rates in any of the
pre-1990 years we analyzed. Explanations

Figure 2: Entry and Exit Rates for the DI Program

Note: The entry rate was calculated by dividing the number of new
beneficiaries in a given year by the total number of DI beneficiaries on the
rolls at the end of the previous year. The exit rate was calculated by
dividing the number of exiting beneficiaries in a given year by the total
number of DI beneficiaries on the rolls at the end of the previous year.

Source: GAO analysis of CWHS data.

The 1990 increase in the SGA level could have encouraged additional program
entry to the extent that individuals with disabilities whose earnings were
between the pre-1990 SGA level and the 1990 SGA level

could then qualify for benefits.28 Also, some individuals could have reduced
their earnings in order to qualify for DI benefits and then increased their
earnings once they became eligible. However, the data we examined indicate
that most DI beneficiaries who entered the program between 1990 and 1994
were either not able or not inclined to increase their earnings or work at
all after receiving benefits. Relatively few of these new DI
beneficiaries-between 2 to 5 percent-increased their earnings above the SGA
level within the first 3 years after their initial year in the program and
most new beneficiaries had no earnings during these first several years on
the rolls.29

There are a number of factors other than the increase in the SGA level that
likely affected the post-1990 DI program entry rates. For example, given
that entry rates began to increase in 1988, prior to the 1990 SGA increase,
the growth in program entry in 1990 and 1991 may simply represent a
continuation of this earlier trend. In our prior work, we described several
program factors, such as changes in the criteria for evaluating mental
impairment disabilities, that appear to have contributed to this trend.30 In
addition, a general labor force response to the 1990-91 recession might also
explain the increase in entry. The recession could have resulted in layoffs
of individuals with disabilities, as well as other workers. In response,
some of these individuals might have sought entry to the DI program, rather
than continuing a job search, even though they were previously able to work
and earn above the SGA level. From the data, we cannot differentiate the
reason for entry by a beneficiary, and so have no way of determining whether
the increase in entry was related to the increase in the SGA level or some
other factor. Likewise, the ensuing economic expansion may have helped to
ensure continuing work and significant earnings for some disabled workers,
thereby reducing the number of workers seeking and receiving DI benefits. In
addition, advances in medicine and medical care, along with advances in and
increased use of assistive devices and equipment (for example, adapted

28In assessing the effects of increasing the SGA from $500 to $700 in 1999,
SSA estimated that by fiscal year 2004, an additional 27,000 individuals
whose earnings exceeded the prior SGA level but were less than the new level
would receive DI benefits as a result of this increase.

29We tracked the earnings, through 1997, of beneficiaries who entered the DI
rolls between 1990 and 1995. For the years we analyzed, the percentage of
these new beneficiaries who had no subsequent earnings ranged from about 76
to 88 percent.

30See Social Security: Disability Rolls Keep Growing, While Explanations
Remain Elusive (GAO/HEHS-94-34, Feb. 1994).

computers/keyboards),  may  have allowed  some  disabled workers  to  remain
gainfully employed.

Program Exits Since 1990 Driven by Retirement and Death but Data Are Limited

Our analysis of DI program exits indicated that the yearly rate of exit
generally declined over the 1990 to 1997 period31 even though the number of
beneficiaries exiting the program was increasing (see figure 2).

Program exit is largely driven by beneficiaries' death or their conversion
to retirement benefits, which together account for about 95 percent of
aggregate program exits between 1985 and 1997 (see table 5).32 While medical
improvement or return to work gradually increased from 2 to 9 percent of all
exits between 1985 and 1996, there was a dramatic increase in the percentage
of DI beneficiaries exiting the program in 1997 for these reasons.

31However, this trend in exit rates began prior to 1990, and a variety of
factors, such as a decline in the average age of new beneficiaries, may have
contributed to it.

32Our figures on the reasons for program exit, or termination, differ
somewhat from those computed based on data reported by SSA (see Tim Zayatz,
A.S.A., "Social Security Disability Insurance Program Worker Experience,"
Actuarial Study No. 114, July 1999). In particular, SSA data indicate
somewhat higher exit rates due to reasons other than death and conversion to
retirement benefits. These differences are likely attributable, in part, to
the use of different sources of data on program exit. However, despite these
differences, the trends portrayed in our data on exits are generally
consistent with those indicated in the SSA data.

                    Table 5: Reasons for Exiting DI Year

Conversion to retirement benefits Death

Returned to work/medical improvement

Note: In cases
where the data
indicated that a
person who had
been eligible for
DI benefits was
no longer
eligible, we
looked to see
whether a
retirement or
death indicator
was recorded. In
cases where
neither of these
indicators were
shown, we
inferred that the person had left the DI rolls due to either one of two
other possible reasons; they were determined to have medically improved or
they were engaged in substantial gainful activity upon returning to work .

Source: GAO analysis of the CWHS data.

It is unclear what effect, if any, the SGA may have had on these program
exits because, although the data indicate whether the beneficiary reached
retirement age or died, they do not indicate whether the beneficiary
returned to work or whether a continuing disability review determined that
they had medically improved. The large increase in the percentage of
beneficiaries returning to work or medically improving for 1997 may be
related, in part, to an increase in the number of continuing disability
reviews that occurred during 1997.33 However, a strong economy that drew
more DI beneficiaries into the labor force or other factors also may have
played a role.

Conclusions Our analysis of DI beneficiary earnings from the mid-1980s to
the mid-1990s suggests that the SGA level may act as a work disincentive for
only a small proportion of DI beneficiaries. This is generally consistent
with

33Congress authorized about $4.1 billion dollars to fund a 7-year initiative
by SSA to conduct about 8.2 million continuing disability reviews during
fiscal years 1996 through 2002. In fiscal years 1996 and 1997, SSA conducted
1.2 million continuing disability reviews.

studies of the SGA and of earnings limits in related programs, which
indicate that such limits, at most, affect a relatively small proportion of
beneficiaries. However, the limitations in the available data mean that our
findings should be accepted with caution. The lack of data on monthly
earnings; on beneficiaries who are blind or are in a trial work period; and
on beneficiaries who return to work, to name only a few areas, all hampered
our efforts to arrive at more definitive conclusions. In particular, the
lack of data identifying whether a beneficiary is blind precluded us from
analyzing the effect of different SGA levels on blind and nonblind DI
beneficiaries.

We place significance on our finding that the SGA's effect remained small
even as increasing numbers of DI beneficiaries entered the labor force.
While the DI program had grown by almost 72 percent from 1985 to 1997, the
number of employed DI beneficiaries more than tripled. The number of working
DI beneficiaries increased every year, even during the recession of the
early 1990s. Yet it is unclear what has been driving this increase in
employment. Given that most of these new workers have earnings far below the
SGA level and remain at those low levels for many years afterwards, it is
unlikely that this increase was caused by an increase in the SGA level.
Other possible explanations include a buoyant economy throughout most of
this period since 1985, enhanced employment protections for the disabled,
increased availability of assistive technology, and a greater acceptance of
hiring the workers with disabilities by society in general. While this
development has important implications for the DI program, the lack of data
again makes it difficult for program officials, researchers, and policy
makers to gain a better understanding of this phenomenon and reconfigure the
DI program's return-to-work incentives to reinforce this trend.

The DI program, program beneficiaries, policy makers, and the general public
could all greatly benefit from the collection of data that would facilitate
a more comprehensive analysis of critical employment and program policy
issues. Therefore, we recommend that the Commissioner of SSA take action to
identify the full range of data necessary to assess the effects of the SGA
on DI program beneficiaries, develop a strategy for reliably collecting
these data, and implement this strategy in a timely manner, balancing the
importance of collecting such data with considerations of cost, beneficiary
privacy, and effects on program operations. In our study, we noted several
key data elements that would be needed for a comprehensive assessment of the
effects of the SGA level on program beneficiaries. These include data that
identify the monthly

Recommendations for Executive Action

Agency Comments
and Our Response

earnings of beneficiaries and whether a beneficiary is blind, is
participating in a trial work period, or has exited the DI program based on
a return to work. Some of these data, such as information identifying
whether a beneficiary is blind or is participating in a trial work period,
is already collected by SSA but is not reliably recorded and maintained in
SSA's principal DI program data base. Other information, such as monthly
earnings data, may be difficult to collect and involve data issues that
extend beyond the DI program. There may also be additional information,
beyond the data elements we discussed, that SSA may consider necessary for
assessing the effects of the SGA.

In commenting on a draft of this report, SSA agreed with our recommendation.
The agency, while acknowledging that it currently does not have the
capability in place to track the employment and earnings patterns of DI
beneficiaries, noted that it has made a commitment to collecting and
analyzing DI beneficiary data. SSA stated that it is currently reaffirming
that commitment and is developing a strategy to improve its efforts to
collect such data. (SSA's comments appear in app. II.)

We believe that SSA's stated commitment to developing improved data on DI
beneficiaries' earnings and employment represents a positive development.
Such a commitment should include the development and implementation of a
comprehensive strategy that would collect the data required for assessing
the earnings and employment of all DI beneficiaries rather than just a
subset, such as those who participate in particular programs initiated under
the Ticket to Work Act. This strategy should also include additional data
elements that would provide insight into our understanding of DI
beneficiaries' employment, such as data identifying beneficiaries who are
blind or who are participating in a trial work period.

SSA also provided some technical comments. The agency noted that although
our report acknowledges various data limitations that affected our analysis,
including limitations in SSA's earnings data, we did not sufficiently
emphasize the extent to which these earnings data might include income that
is not related to current employment. In addition, SSA stated that our data
on reasons for exit, or termination, from the DI program varied from those
published by SSA's Office of the Chief Actuary. Finally, SSA questioned our
analysis of beneficiaries whose earnings consistently exceed the SGA level.

With regard to our discussion of limitations in the earnings data, we agree
with SSA that these limitations are considerable and have noted that

throughout the report. In particular, SSA highlighted the potential for SSA
earnings records to include income that may not be related to current work.
It is unclear whether a substantial portion of the earnings data we analyzed
was unrelated to current work. For example, an SSA study34 stated that the
agency's earnings data may include "certain payments from profit sharing
plans." However, the study also noted that few beneficiaries had actually
participated in such plans. In addition, although this study indicated a
sizeable discrepancy between SSA earnings data and earnings reported by some
beneficiaries in a survey interview, it was unclear whether this discrepancy
was due to limitations in SSA data or to limitations inherent in
self-reported data.

Regarding the differences between our data on the reasons for program exit,
or termination, and the data reported by SSA, we acknowledge in the report
that SSA data indicate somewhat higher exit rates due to reasons other than
death and conversion to retirement benefits. We believe that these
differences are likely attributable to the use of different sources of data
on program exit. We used the CWHS because it was the most appropriate data
set for conducting a longitudinal analysis of beneficiaries' earnings in
relation to the SGA level. Further, although the termination rates we report
do differ from SSA's data, the trends portrayed in our data on exits are, in
fact, generally consistent with those indicated in the SSA data. For
example, where SSA's data indicate a 10.5 percentage point increase in
program exit due to medical recovery or return-to-work from 1996 to 1997
(from 12.3 percent to 22.9 percent), GAO's data similarly indicate a 10.7
percentage point increase (from 9.2 percent to 19.9 percent). Given that our
discussion of program exits focuses primarily on trends rather than absolute
numbers, we believe that our data adequately support our finding.

Finally, regarding the issue of some beneficiaries being able to
consistently earn above the SGA level, we identified in the report several
reasons why some beneficiaries might do so. For example, such beneficiaries
may be blind and thus subject to a higher SGA level than nonblind
beneficiaries. We also note that without better DI program data, including
data identifying whether a beneficiary is blind or in a trial work period,
we could not provide a more definitive explanation of this

34L. Scott Muller, "Comparisons of Self-Reported Work Activity and
Administrative Earnings Reports of Individuals Recently Entitled to Social
Security Disability Insurance Benefits," unpublished manuscript, Apr. 1990.

phenomenon. Examination of individual case folders to determine why
beneficiaries continued to earn above the SGA level-an approach suggested by
SSA--was not a viable option for us on this study given our resources and
timeframes for completing the study.

SSA also made a few other technical comments, which we incorporated where
appropriate.

We are sending copies of this report to the Honorable Jo Anne B. Barnhart,
Commissioner of Social Security; appropriate congressional committees;
and other interested parties. We will make copies available to others on
request. This report is also available on GAO's home page at
http://www.gao.gov.

If you or your staff have any questions concerning this report, please call
me at (202) 512-7215 or Charles A. Jeszeck at (202) 512-7036. Other
individuals making key contributions to this report include Mark Trapani,
Michael J. Collins, and Ann Horvath-Rose.

Barbara D. Bovbjerg
Director, Education, Workforce,

and Income Security

                     Appendix I : Scope and Methodology

Our Sample

To conduct our work, we analyzed data from the Social Security
Administration's (SSA) Continuous Work History Sample (CWHS). The CWHS
consists of records representing a longitudinal 1 percent sample of all
active Social Security accounts. It is designed to provide data on earnings
and employment for the purpose of studying the lifetime working patterns of
individuals. The data, drawn from SSA administrative data sets, contain
information on an individual's Disability Insurance (DI) eligibility,
earnings,1 and demographic characteristics. We did not independently verify
the accuracy of the CWHS data because they were commonly used by researchers
in the past and they are derived from a common source of DI program
information.

From the total sample of 2,955,942 individuals, we selected a subsample of
92,662 individuals with disabilities for the 1984 to 1998 period. To obtain
this sample, we excluded individuals whose Social Security record indicated
a gap in DI entitlement, DI beneficiary status beginning before age 18 or
continuing past age 64, a date of death before their DI beneficiary status,
and those not identified as the primary beneficiary. We could not determine
the exact date of eligibility because the CWHS only provides eligibility
status as of December 31 of each year. Therefore, individuals were included
in our analysis only as of their second year of DI eligibility to assure
that the earnings we observed occurred only while an individual was in
beneficiary status.2 In addition to our main sample, we selected another
subsample of 9,990 DI beneficiaries who reached age 65 during the 1984 to
1990 time period for the purpose of analyzing DI beneficiaries who were
converted to retirement benefits.

1Reports of earnings must be filed annually with SSA by every employer who
is required to withhold income tax from wages and/or who is liable for
Federal Insurance Contributions Act taxes-which are used to finance the Old
Age and Survivors Insurance and Medicare Hospital Insurance programs. While
almost all jobs in the U.S. are covered by these reporting requirements,
there are several excluded categories of workers, such as federal civilian
employees hired before 1984 and certain state and local government employees
who participate in alternative retirement systems. In addition, some
earnings reported to SSA may actually represent income derived from work
activity in a previous year, such as commissions or bonuses from prior work.

2The one exception to this "two-year rule" occurred in our analysis of new
program entrants where we identified DI eligibility as the first year that
such eligibility was indicated in the data set.

                     Appendix I : Scope and Methodology

Sampling Errors

Analysis of the Effect of the SGA on Beneficiaries' Earnings

All samples are subject to sampling error, which is the extent to which the
sample results differ from what would have been obtained if the whole
universe had been observed. Measures of sampling error are defined by two
elements-the width of the confidence interval around the estimate (sometimes
called precision of the estimate) and the confidence level at which the
interval is computed. The confidence interval refers to the fact that
estimates actually encompass a range of possible values, not just a single
point. This interval is often expressed as a point estimate, plus or minus
some value (the precision level). For example, a point estimate of 75
percent plus or minus 5 percentage points means that the true population
value is estimated to lie between 70 percent and 80 percent, at some
specified level of confidence.

The confidence level of the estimate is a measure of the certainty that the
true value lies within the range of the confidence interval. We calculated
the sampling error for each statistical estimate in this report at the
95-percent confidence level. All percentage estimates from the sample have
sampling errors (95 percent confidence intervals) of plus or minus 10
percentage points or less, unless otherwise noted. All numerical estimates
other than percentages have sampling errors of 10 percent or less of the
value of those numerical estimates, unless otherwise noted.

To analyze the effects of the SGA on the earnings of DI beneficiaries, we
attempted to determine whether DI beneficiaries engage in "parking," that
is, whether they limit their earnings to a level at or just below the SGA
limit in order to maintain eligibility for benefits. If beneficiaries do
indeed park, then we would expect to find a clustering of earnings just
below the SGA level. The occurrence of such clustering would provide a
fairly strong indication that beneficiaries are limiting their employment
and earnings to stay in the DI program, thereby reducing program exit. In
addition, to the extent that beneficiaries park or otherwise limit their
earnings due to a work disincentive effect of the SGA, we would expect an
increase in the SGA level to result in a corresponding increase in
beneficiaries' earnings.

To determine if earnings clustered around the SGA level, we examined the
distribution of earnings both before and after the 1990 increase in the SGA
level to see what proportion of beneficiaries had annual earnings at or
within 5 percent, 10 percent, and 25 percent of the annualized SGA level. We
also tracked those beneficiaries who had earnings near the annualized SGA
level in a given year to see if they maintained this level of earnings in
subsequent years. In addition, we tracked those beneficiaries who were on
the rolls and had no earnings or had earnings below the annualized SGA

                     Appendix I : Scope and Methodology

level prior to 1990 to see if they increased their earnings and clustered
around the new annualized SGA level. Finally, we examined beneficiaries who,
in a given year, had earnings above the annualized SGA level to see if, over
time, they tended to reduce their earnings to an amount near, but below, the
SGA to maintain program eligibility.

To further analyze whether DI beneficiaries limit their earnings due to the
SGA, we observed how these individuals behave once they are no longer
subject to the SGA level. We did this by looking at the earnings of DI
beneficiaries who reached age 65 and were converted to the Old Age and
Survivors Insurance (OASI) program. Once DI beneficiaries reach age 65, they
are converted to retired worker status and their benefits are paid from the
OASI trust fund. Likewise, they are no longer subject to the SGA limit.3 If
beneficiaries are limiting their earnings due to the SGA, then we would
expect them to increase their earnings after retirement at age 65.
Therefore, a finding that a significant proportion of former DI
beneficiaries return to work or increase earnings after conversion would
serve as some evidence for the work disincentive effect of the SGA. For DI
beneficiaries who had entered the DI rolls prior to age 62, remained on the
rolls until being converted to retirement benefits at age 65, and survived
to age 68, we examined their earnings between ages 66 - 68 to determine
whether there was an increase in earnings and employment after they left the
DI program.

To examine the effects of the SGA on DI program entry and exit rates, we
looked at the rate of entry and exit both before and after the increase in
the SGA.4 If people respond to the change in the SGA then we might expect
the rate of entry to increase after the increase in the SGA level. With the
higher SGA level, some individuals with disabilities would now qualify for
benefits if their earnings are between the old and new SGA level.5 Likewise,
some individuals with earnings just above the new SGA level may reduce their
earnings in order to qualify and then increase their

3Although they were, until January 2000, subject to the retirement earnings
limit, which was much higher than the nonblind SGA level during our period
of analysis.

4We calculated entry rates by dividing the number of new beneficiaries in a
given year by the total number of DI beneficiaries on the rolls in the
previous year. Similarly, we calculated exit rates by dividing the number of
exiting beneficiaries in a given year by the total number of DI
beneficiaries on the rolls in the previous year.

5Assuming they also meet other DI eligibility criteria.

Analysis Of the Effect of the SGA on Program Entry and Exit

                     Appendix I : Scope and Methodology

Limitations in Our Analysis

earnings after they become eligible. Therefore, we examined the earnings,
through 1997, of new beneficiaries who entered the DI program between 1990
and 1995 to see if they tended to increase their earnings after becoming
eligible for benefits.

In terms of program exit, we might expect exit rates to decrease after an
increase in the SGA level since many working beneficiaries may now be
further from the new level and some may even increase their earnings to an
amount near the new level (but higher than the old level) without having
their benefits terminated. We examined data indicating the reasons that
beneficiaries' exit DI to determine the extent to which program exits
resulted from beneficiaries returning to work or medically improving versus
retirements or deaths.

The absence of key data in the CWHS and in other SSA data sets limited our
ability to draw clear conclusions from our analysis. For example, while the
SGA is a monthly level, the available earnings data are recorded only on a
yearly basis. Therefore, we were not able to analyze DI beneficiaries'
monthly earnings in relation to the actual, monthly SGA limit. Instead, we
examined beneficiary earnings in terms of the annualized SGA level; that is,
we multiplied the monthly SGA amount by 12 to permit comparison of the
monthly limit to the annual data. (For example, the SGA level in 1995 was
$500 per month, so the annualized SGA level was $500 multiplied by 12, or
$6,000.) As a result, we were not able to identify parking that might have
occurred among beneficiaries who, for example, worked for only a few months
during the year but limited their earnings to a level near, but not
exceeding, the SGA level in each of those months.6 Nevertheless, our
analysis did allow us to identify individuals who consistently have earnings
at or near the SGA level. To the extent that beneficiaries are trying to
maximize their income-that is, earn as much as they can within a given year
while maintaining DI eligibility-there may be a significant number of
beneficiaries who have sustained earnings up to the SGA level through much
of the year.

6It is also possible that some beneficiaries who worked only intermittently
during the year had annual earnings at or near the annualized SGA level.
However, while earnings at or near the SGA level may be suggestive of
parking behavior, we could not identify whether a beneficiary with earnings
in this range was truly parking or had limited earnings due to other
reasons.

Appendix I : Scope and Methodology

Another data limitation concerned beneficiaries who are in a trial work
period. The trial work period allows beneficiaries to test their ability to
work without penalty. Therefore, beneficiaries can earn any amount without
being subject to the SGA limit. Neither the CWHS nor other SSA data sets
provide a reliable means for identifying beneficiaries in a trial work
period. As a result, in our parking analysis, we were not able to
distinguish the earnings of beneficiaries who are subject to the SGA limit
from those who are not subject to this limit. Although the trial work period
allows beneficiaries to earn any amount, there is no reason to believe that
all beneficiaries in a trial work period will have earnings greater than the
SGA level. An individual's disability may limit his/her earnings to well
below the SGA level. However, we do not believe that this limitation
affected our analysis to a great extent because it is unlikely that the
earnings of beneficiaries in the TWP would systematically fall at or near
the SGA level and thereby skew our analysis.

The identification of blind and nonblind beneficiaries also created a
limitation in our analysis. The CWHS does not allow us to distinguish
between blind and nonblind DI beneficiaries, which is important since blind
beneficiaries are subject to a higher SGA limit. Some of the beneficiaries
that we observe earning above the nonblind SGA limit may actually be blind
individuals. If a substantial number of blind beneficiaries had earnings
just below the nonblind SGA level, then our analysis could exaggerate the
existence of parking. However, this limitation is not likely to have
substantially impacted our analysis of parking among nonblind beneficiaries
because blind individuals represent only about 2 percent of the DI caseload
and therefore probably comprised a very small portion of our sample. Perhaps
more importantly, the inability to identify blind beneficiaries means that
we could not assess the extent to which they exhibit parking behavior. As a
result, our analysis may be understating the extent of parking in the DI
program.

Finally, the lack of data on impairment-related work expenses (IRWE) also
limited our ability to analyze the effects of the SGA level on employment.
SSA deducts the cost of certain impairment-related expenses needed for work
from earnings when making SGA determinations. The inability to identify IRWE
could exaggerate the effect of the SGA on earnings since some beneficiaries
near or above the SGA level may not have been at this level once IRWE was
subtracted from their earnings. However, the inability to determine IRWE is
not likely to have significantly impacted our analysis because SSA officials
told us that IRWE was applied in only a very limited number of cases during
the years of our analysis

Appendix I : Scope and Methodology

Despite these substantial limitations, the CWHS is the best available data
set for identifying the basic program information needed to conduct our
analysis within acceptable timeframes. The principal alternative data set
within SSA-the Master Beneficiary Record-does not lend itself to easy
analysis because it is designed to fulfill SSA's administrative objectives.
In particular, we did not choose to use this data set because it would not
have provided the longitudinal data that we needed unless it was linked with
other SSA administrative files containing DI program information. Linking
these complex files would have raised many uncertainties regarding the
ultimate quality of the data and would have added substantial time and
complexity to our analysis. In addition, non-SSA data sets, such as the
Census Bureau's Current Population Survey, could not serve our needs
because, among other limitations, we would not be able to adequately
identify DI program participation for most of the years of our analysis.

In addition to data limitations, our analysis was also constrained by the
lack of any quantitative evaluation of other possible factors affecting the
earnings of DI beneficiaries and disabled workers. For example, our analysis
does not control for other factors in the economy such as recessions,
implementation of the Americans With Disabilities Act (ADA), advances in
medicine and medical care, and advances in and increased use of assistive
devices and equipment. A recession may increase entry into the DI program,
but implementation of the ADA and improvements in medical care and assistive
devices and equipment may either decrease entry or increase exit. The
inability to control for these factors limited our ability to make clear
inferences from the data regarding the effects of the SGA.

Appendix II: Comments From the Social Security Administration

Appendix II: Comments From the Social Security Administration

Appendix II: Comments From the Social Security Administration

Appendix II: Comments From the Social Security Administration

GAO's Mission

Obtaining Copies of GAO Reports and Testimony

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 www.gao.gov and select
"Subscribe to daily e-mail alert for newly released products" under the GAO
Reports heading.

Order by Mail or Phone 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: (202) 512-2537 Fax: (202)
512-6061

Visit GAO's Document GAO Building

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

To Report Fraud, Contact: Web site: www.gao.gov/fraudnet/fraudnet.htm,

Waste, and Abuse in E-mail: [email protected], or

Federal Programs 1-800-424-5454 or (202) 512-7470 (automated answering
system).

Jeff   Nelligan,  Managing   Director,   [email protected]   (202)  512-4800

Public Affairs U.S.  General Accounting Office, 441 G. Street NW, Room 7149,
Washington, D.C. 20548
*** End of document. ***