United States General Accounting Office
GAO Report to the Chairman, Committee on
Appropriations, U.S. Senate
August 1997
FEDERAL PENSIONS
Relationship Between
Retiree Pensions and
Final Salaries
GAO/GGD-97-156
United States
GAO General Accounting Office
Washington, D.C. 20548
General Government Division
B-270600
August 11, 1997
The Honorable Ted Stevens
Chairman, Committee on Appropriations
United States Senate
Dear Mr. Chairman:
Federal spending on pensions for retired civilian employees of the federal
government represents a significant share of the budget. In fiscal year
1996, excluding interest on the public debt, civilian employee pension
benefits (i.e., civil service retirement and disability) was the seventh
largest mandatory spending program, with nearly $40 billion in payments
to 2.3 million retirees and survivor annuitants. Although current
employees finance a portion of these benefits through the contributions
they make, the federal government pays most pension costs, as do states
and localities and private sector employers. Thus, it is important for
policymakers to understand how key features of federal retirement
policy—set in statute—affect pension costs.
At your request, we are responding to a series of questions about federal
and nonfederal retirement programs. This report addresses the part of
your request that concerns pension costs and retirement policy. As agreed
with your office, our objectives were to (1) determine the number of
federal retirees, if any, whose pensions have come to exceed the final
salaries that they earned while working; (2) explain why these retirees’
pensions came to exceed their final salaries; and (3) determine the
difference, if any, in these retirees’ pension amounts if current
cost-of-living-adjustment (COLA) policy—that is, the COLA policy enacted in
1984, which established the formula and schedule used today by the Office
of Personnel Management (OPM)—had been in effect without interruption
since 1962, and also determine any difference in the number of retirees
whose pensions would have exceeded their final salaries.1
We collected data for the Civil Service Retirement System (CSRS) and the
Federal Employees Retirement System (FERS) general employees, as well
as for all former Members of Congress who were retired and still living as
of October 1, 1995, using a computerized personnel database and case file
information maintained by OPM. Although the preliminary results for
Members appear to be about the same as the results for general
employees, as agreed with your office, we are reporting the results for
general employees in this letter, and we will report on Members
1
The COLA policies we refer to in this report were set by various federal statutes.
Page 1 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
B-270600
separately. We used a number of different approaches to meet our
objectives, including simulation and statistical analyses of a randomly
selected, projectable sample of CSRS retirees. The sample and techniques
that we used are described in greater detail in the Scope and Methodology
section of this report.
CSRS and FERS are the two largest retirement programs for federal civilian
Background employees. At the beginning of fiscal year 1995, these programs covered
about 2.8 million federal employees, or 90 percent of the current civilian
workforce. OPM administers CSRS and FERS. CSRS and FERS pension benefits
are financed partly by federal agency and employee contributions and
partly by other government payments to the Civil Service Retirement and
Disability Fund.2
Although CSRS and FERS both provide pensions, the programs are designed
differently. CSRS was established in 1920 and predates the Social Security
system by 15 years. When the Social Security system was established,
Congress decided that employees in CSRS would not be covered by Social
Security through their federal employment. CSRS is a stand-alone pension
program that provides an annuity determined by a formula as well as
disability and survivor benefits.3 The program was closed to new entrants
after December 31, 1983, and, according to OPM actuaries, is estimated to
end in about 2070, when all covered employees and survivor annuitants
are expected to have died. FERS was implemented in 1987 and generally
covers those employees who first entered federal service after 1983 as well
as those who transferred from CSRS to FERS. The primary impetus for the
new program was the Social Security Amendments of 1983, which required
that all federal employees hired after December 1983 be covered by Social
Security.4 FERS is a three-tiered retirement program that includes Social
2
The Department of the Treasury also makes annual payments that are to cover interest on unfunded
liabilities, payments for spouse equity, as well as amortization payments to finance supplemental
liabilities for FERS.
3
If a survivor annuity benefit is chosen, pensions may be reduced by as much as 10 percent. Pensions
are reduced to provide for spousal benefits or insurable interest benefits (i.e., a person designated by
the retiree as expecting to receive some financial benefit from the continuance of the life of the
retiree), but not for children’s benefits. Children’s benefits are provided by law and do not need to be
elected by an employee or retiree. If a spousal survivor annuity is chosen and the spouse predeceases
the retiree, the annuity reduction is eliminated upon notification to OPM. At the time of retirement,
CSRS pensions may also be reduced for other reasons, including reductions for age and unpaid
deposits. FERS pensions may be reduced for age.
4
After December 31, 1983, certain rehires participating in CSRS before 1984 could elect to either stay in
that plan under special rules that integrate CSRS and Social Security or transfer to FERS. For a more
detailed discussion of the transition from CSRS to FERS, see Federal Retirement: Federal and Private
Sector Retirement Program Benefits Vary (GAO/GGD-97-40, Apr. 7, 1997).
Page 2 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
B-270600
Security and a Thrift Savings Plan —in addition to a basic pension. Like
CSRS, FERS provides disability and survivor benefits.
A distinctive feature of CSRS and FERS pensions is the annual COLAs they are
to provide. COLAs are post-retirement increases in pension amounts that
generally are given on either an ad hoc or automatic basis to offset
increases in living costs due to inflation. Congress enacted the first
automatic COLA for CSRS annuitants in 1962 (effective January 1963). At that
time, the automatic adjustment was viewed as a way of controlling
pension costs, because prior ad hoc adjustments had been criticized as
being unrelated to price increases and subject to political manipulation.
Although COLAs generally have been provided on an automatic basis since
1962, COLA policies have been modified numerous times over the years. As
shown in table 1, the changes made during the 1960s and 1970s were
intended to enhance pension purchasing power with respect to inflation as
measured by the consumer price index (CPI), but some of the changes
made during the 1980s had the effect of reducing purchasing power.5 Table
1 is based on information in the Congressional Research Service (CRS)
Report for Congress, 94-834 EPW, updated March 13, 1996.
5
The CPI is compiled by the Bureau of Labor Statistics and is intended to measure the average change
in the prices paid by urban consumers for a fixed market basket of goods and services. It is calculated
monthly for two population groups, one consisting only of wage earners and clerical workers and the
other consisting of all urban families. The wage earner index—CPI-W—is the index used for federal
COLA purposes. Because it is a national average, it affects retirees differently, depending on whether
they live in areas where the CPI-W differs from the national average. Also, because the CPI is a
statistical average, it may not reflect an individual’s experience, particularly an individual whose
expenditures differ greatly from the “average” consumer’s. Moreover, whether the CPI accurately
estimates inflation is currently being debated. In a 1996 report, the Advisory Commission to Study the
Consumer Price Index concluded that the CPI overstates inflation. The Commission recommended
that the market basket on which the CPI depends be updated more frequently than is currently done
and that adjustments be made to correct any bias in the estimates.
Page 3 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
B-270600
Table 1: Major Changes Made to COLA
Policy Since Automatic Adjustments Year Public law Description
Began 1962 P.L. 87-793 Provided the first automatic adjustments whenever the
CPI in a given year exceeded the CPI for the year of the
last adjustment by 3 percent or more. This was later
modified to provide for adjustments whenever the CPI
rose 3 percentage points or more above the CPI in the
month of the last adjustment and remained at or above
this level for 3 consecutive months.
1969 P.L. 91-93 Added an extra 1 percent to the adjustment—known as a
kicker—to offset the erosion in pension benefits due to the
time lag between increases in living costs and benefit
adjustments.
1976 P.L. 94-440 Repealed the kicker because it had been found to
overcompensate for inflation. However, Congress
replaced the kicker with semiannual COLAs as another
way to address the time lag.
1981 P.L. 97-35 Replaced semiannual COLAs with annual COLAs based
on the change in the CPI from December to December
and payable in March of the following year, thereby
saving money by having benefits held constant for longer
periods.
1982 P.L. 97-253 Added a restriction in certain cases to ensure that
pensions would not exceed the current maximum pay for
a General Schedule (GS) 15 federal employee.
1983 P.L. 98-270 Established the formula upon which COLAs currently are
(enacted in based and made COLAs effective in December of the
1984) current year and payable in January of the following year.a
1984 P.L. 98-369 Specified that COLAs were to be payable in checks
issued the first business day of the month following the
month for which they are scheduled or effective.
1985 P.L. 99-177 Suspended COLAs for fiscal year 1986 and for all
subsequent years in which specified deficit reduction
targets would not otherwise be met.
1986 P.L. 99-509 Reinstated COLAs for programs that had been subject to
the suspension under P.L. 99-177 for calendar years
1987-1991.b
1993 P.L. 103-66 Changed the effective dates for COLAs from December
to March for fiscal years 1994 through 1996.c
a
This formula and schedule are the same as those used for Social Security COLAs, which were
established for that program in P.L. 98-21. This law also eliminated the COLAs scheduled for May
1984 and June 1985. Instead, COLAs were scheduled for December 1984, payable in January 1,
1985, checks.
b
The Balanced Budget and Emergency Deficit Control Reaffirmation Act of 1987 (P.L.
100-119) permanently exempted federal pension COLAs from suspension under P.L. 99-177.
c
The COLAs were in checks payable the first business day of April rather than January. This law
did not change the CPI measuring period.
Source: CRS.
Page 4 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
B-270600
One of these changes provides especially relevant background for
considering the relationship between current pensions and final salaries
and requires a more complete discussion. As noted in table 1, P.L. 97-253
(the Omnibus Budget Reconciliation Act of 1982) restricted COLAs in
relation to final salaries in certain cases. Under this restriction, a pension
may not be increased by a COLA to an amount that exceeds the greater of
the current maximum pay for a GS-15 federal employee or the final pay of
the employee (or high-3 average pay, if greater), increased by the overall
annual average percentage adjustments (compounded) in rates of pay of
the general schedule for the period beginning on the retiree’s annuity
starting date and ending on the effective date of the adjustment. In effect,
the statute requires that a retiree’s pension is to be capped at an amount
not to exceed the maximum pay of a general schedule employee (i.e.,
GS-15) or an amount that represents the value of the retiree’s final or
average pay, adjusted for the general schedule pay adjustments that had
been provided since the annuitant retired. According to OPM’s policy
handbook, because the cap applies to COLA increases to pensions, in no
instance would a pension already exceeding the cap be reduced.6
As noted earlier, under current policy—enacted in 1984—COLAs for CSRS
and FERS retirees are based on increases in living costs as measured by the
CPI-W between the third quarter (July through September) of the current
calendar year and the third quarter of the previous year. Although the COLA
formula and schedule are the same for FERS and CSRS, FERS COLAs are
limited if inflation is over 2 percent. If inflation is between 2.0 and
3.0 percent, the FERS COLA is 2.0 percent; if inflation is 3.0 percent or more,
the COLA is the CPI minus 1 percent. If, however, inflation is less than
2 percent, FERS COLAs are to be fully adjusted for inflation. Also, CSRS
benefits are to be fully indexed from the time of retirement, and FERS
pensions are to be indexed beginning at age 62 for regular retirees.7
An estimated 459,000 (or about 27 percent) of the 1.7 million retirees who
Results in Brief were on the federal pension rolls as of October 1, 1995, were receiving
pensions that had come to exceed their final salaries when these salaries
were not adjusted for inflation. However, when their salaries were
adjusted for inflation—i.e., expressed in constant dollars—no retiree was
6
Under CSRS, initial annuities are also capped. As described in greater detail later in this report, with
certain exceptions, the maximum initial annuity that a retiree can receive under CSRS is 80 percent of
his or her high-3 average salary.
7
The first FERS COLA was effective in December 1988 and payable in January 1989. FERS participants
of any age who retired on disability are to receive COLAs after their first year of disability.
Page 5 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
B-270600
receiving a pension that was larger than his or her final salary. As a general
rule, using constant dollars provides a more meaningful way to compare
monetary values across time, because the use of constant dollars corrects
for the effects of inflation or deflation.
Although no retiree’s pension exceeded his or her final salary in constant
dollar terms, our analysis confirmed that three factors played an important
role in explaining why the retirees’ pensions came to exceed their
unadjusted final salaries—the number and size of COLAs that retirees
received, the number of years that they had been retired, and the number
of years of their federal service. The first two factors in combination
reflect retirement policies that are intended to maintain most or all of a
pension’s purchasing power. Although the COLAs that the sample retirees
received caused their pensions to increase at rates that generally were to
equal inflation during retirement, their unadjusted final salaries remained
the same. Thus, the longer the annuitants had been retired, the more COLAs
they would have received and the more likely their pensions would have
come to exceed their unadjusted final salaries. Also, because COLAs were
to be automatic and inflation continued throughout the period we
reviewed, the number of COLAs that the sample retirees would have
received was highly correlated with the number of years that they had
been retired. The third factor—a retiree’s years of federal service—also
contributed, because years of service is a major component in determining
the amount of a retiree’s initial pension. Specifically, the sample retirees
with many years of service would have received initial pensions that came
closer to the amounts of their final salaries than the retirees with fewer
years of service, other factors being equal. Smaller beginning differences
between initial pensions and final salaries, in turn, would have caused the
pensions of the first group of retirees to have exceeded their unadjusted
salaries sooner than the second group’s pensions.
Our analysis of the effects that COLA policies have had on retiree pensions
suggests that the policies have played an important role in maintaining the
purchasing power of retiree pensions since automatic COLAs began. It also
suggests that the effects COLA policies actually have had on retiree pension
amounts cannot be summarized easily because of the numerous changes
that have been made in COLA policies over the past 35 years. COLA policy
changes have affected individual retirees differently, depending on when
their retirements began. For example, because the effects of COLAs and
COLA policy changes compound over time, the COLA policies of the late
1960s and 1970s, which overcompensated for inflation, will continue to
affect the pensions of those retirees who receive them as long as they are
Page 6 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
B-270600
alive, just as the suspensions of some COLAs in the 1980s will continue to
be reflected in the pensions of anyone who retired before the suspensions
occurred.
If current COLA policy—that is, the policy that was enacted in 1984—had
been in effect without interruption since automatic COLAs began in 1962,
the pensions of some of the sample retirees would have been smaller than
the pensions that they actually received, and the pensions of other retirees
would have been larger. Our comparison of the effects of current and
historical COLA policy (as shown in table 1) on pension amounts suggests
that, other factors being equal, a majority of those who retired before 1970
would have received smaller pensions had current COLA policy been
continuously in effect during their retirement, and about 90 percent of
those who retired after 1970 would have received larger pensions. The
changes that would have occurred in the sample retirees’ pension amounts
under current policy were enough to cause about a three percentage point
(3.0) increase in the number of retirees whose pensions would have come
to exceed their unadjusted final salaries.
To respond to your request, we used a computerized personnel database
Scope and of CSRS and FERS retirees and case file information maintained by OPM. At
Methodology the time of our analysis, the latest available data were for living CSRS and
FERS annuitants who were retired as of October 1, 1995. The database and
case files provided much of the information that we needed for our
analysis, including the retirees’ initial and 1995 pensions, retirement dates,
high-3 average salaries, service histories, survivor benefits, and other
retirement-related information. However, the database did not have
information on retirees’ final salaries, which we needed in order to
compare their final salaries to their 1995 annuities. The database did have
information on “high-3” average salaries, which are used in calculating
initial pensions. Thus, we compared the retirees’ high-3 average salaries to
their 1995 pensions to identify a set of retirees whose pensions were most
likely to have exceeded their final salaries. From this group, we selected a
random sample of 400 from among the 524,435 CSRS retired general
employees whose annuities exceeded their high-3 average salaries and all
105 FERS retired general employees for whom the database reported
annuities exceeding their high-3 average salaries.8 We reviewed the
selected retirees’ case files to verify that those we had selected had 1995
pensions that, in fact, exceeded their unadjusted final salaries.
8
We did not sample from the 66 CSRS annuitants whose high-3 average salaries were listed as zero in
the database.
Page 7 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
B-270600
From our review of the sample of 400 CSRS annuitants, we identified 348
whose 1995 pensions exceeded their final salaries. We identified and
removed from our sample 50 with pensions below their final salaries, 1
whose case file did not have the data we needed for our analysis, and
another whose case file was not available for our review. From our case
file review of the 105 FERS annuitants, we identified and removed 104 that
did not match our criterion (i.e., did not have a 1995 annuity that exceeded
the retiree’s final salary). The remaining case had a pension that exceeded
the final salary. However, the pension combined both FERS and CSRS
benefits. This retiree had transferred from CSRS to FERS and thus was
receiving benefits that were neither wholly FERS nor wholly CSRS.
Consequently, we included this individual in our estimates of the number
of retirees who had annuities that exceed their final salaries, but excluded
this individual from our regression analysis.
We weighted the CSRS sample results to estimate the number of retired
general employees in the population whose pensions had come to exceed
both their final salaries and high-3 average salaries. In making these
estimates, we assumed that the small number of FERS and CSRS cases for
which data were not available were similar to the cases that we had
reviewed. The sample results thus estimate the total number of general
employees whose pensions exceed both their final salaries and their high-3
average salaries. As the final salary is generally included in the three
highest salaries that are averaged, these employees are described as
having pensions that exceed their “final salaries” in the remainder of the
report. We also adjusted the retirees’ final salaries for inflation, using the
1995 CPI-W, and made a second estimate of the number of retirees whose
1995 pensions exceeded their final salaries, expressed in constant dollar
terms.
To understand why retiree pensions could come to exceed unadjusted
final salaries as much as they did, we used regression analysis to model
the relationship between key retirement policy variables and the extent to
which the pensions of the sample retirees exceeded their unadjusted final
salaries. Regression is a statistical technique that can be used to measure
the relationship between a dependent variable and a set of independent
(i.e., explanatory) variables and isolate their independent effects. This
analysis was based on the subsample of 348 CSRS employees whose 1995
pensions exceeded their final salaries. This subsample did not include the
single FERS annuitant whose pension exceeded the final salary, the two
sampled cases with missing information, nor the 50 sampled cases whose
Page 8 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
B-270600
1995 pensions did not exceed their final salaries.9 We used the percentage
by which the retirees’ pensions exceeded final salaries as the dependent
variable in the model, because our sample did not include retirees whose
pensions were below their high-3 average salaries.10 We selected
retirement variables to use as independent variables because they were
(1) required to be used for computing pension benefits (e.g., years of
service); or (2) known to affect pension amounts for some or all retirees
(e.g., COLAs and the selection of spousal survivor benefits).11 Although
variables representing changes in a retiree’s personal circumstances (e.g.,
marriage, death of a spouse, or divorce) that would have changed his or
her pension over the period of retirement were not included in the final
regression model, we reviewed the retirees’ case files to determine what
effects these changes may have had on individual sample retirees. We
found that these changes in personal circumstances could cause an
individual retiree’s pension to fluctuate (e.g., increase and/or decrease)
during his or her retirement depending on whether survivor’s benefits
were being deducted.
To compare the effects of current and historical COLA policy on retirees’
pensions, we reviewed federal retirement-related documents and
identified the historical changes in COLA policy since the inception of
automatic COLAs in 1962.12 Using this information, we calculated the
pensions that the sample of 398 retirees would have received each year
from 1962 through 1995 had current COLA policy been in effect without
interruption. We compared these results to the pensions that they would
have received under actual COLA policy, absent other changes that might
have affected their pensions (e.g., adjustments due to death of a spouse
when survivor benefits had been chosen). We then compared the resulting
numbers to assess the probability that the change, if any, in the number of
retirees whose 1995 pensions had exceeded their unadjusted final salaries
was statistically significant, that is, unlikely to be due to sampling error.
9
Our regression estimates are not applicable to the larger population of all retirees, because no FERS
participants and no retirees with 1995 pensions lower than their high-3 average salaries were included
in the analysis.
10
More than two-thirds of all annuitants retired in 1995 received pensions that were below their high-3
average salaries.
11
It is important to note that the model’s parameter estimates of the effects of the retirement policy
variables are for those retirees whose 1995 pensions had come to exceed their final salaries. Had all
retirees been used, the parameter estimates could have been different because the analysis would have
examined instances in which retirees’ 1995 pensions had not come to exceed their final salaries.
12
Payment of COLAs specified by the current COLA policy enacted in 1984 has been interrupted
several times since then, as shown in table 1. Our simulations of current COLA policy did not include
these interruptions.
Page 9 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
B-270600
To illustrate the effects that the different COLA policies could have had on
pensions during the sample annuitants’ retirements, we simulated the
effects of current and actual policy on pension amounts for three different
retirement periods. To simplify the analysis, our simulation of the impacts
of current COLA policy implemented without interruption since 1984 was
not adjusted to reflect the actual effective dates of COLAs, the actual pay
dates, “lookback” payments or adjustments, or prorated to reflect the
month an employee retired.13 We selected 1961 to 1995, 1968 to 1995, and
1981 to 1995 to show the cumulative effects that the COLAs of the 1960s and
1970s, which overcompensated for inflation, and the suspensions of COLAs
in the 1980s could have had for different periods of retirement. We used
the average initial pension for the sample annuitants who had retired in
the first year of each of the three periods for our starting pension amounts
(e.g., the average initial pension of those annuitants who retired in 1961).
Our analysis had several limitations. As agreed with your office, we did not
independently verify the accuracy of OPM’s database. However, we did
verify the accuracy of the data for the cases used in our analysis. Also, the
number of retirees whose pensions had come to exceed their final
unadjusted salaries could be somewhat higher than we estimated for two
reasons. As noted, we used high-3 average salary to identify a population
that we believed would be most likely to have pensions that had come to
exceed final salaries, because OPM’s computerized database did not include
final salary information. Thus, our estimates do not include those retirees
whose pensions were lower than their high-3 salaries but whose pensions
were higher than their final salaries. Also, the annuity amounts contained
in the case files already had survivor benefit reductions, if any, taken.
Thus, retirees who selected survivor benefits would have had higher initial
pensions than the pensions reported in OPM’s files. However, we could not
take this reduction into account, because the automated data file did not
identify those retirees who had selected this benefit. On the basis of our
examination of the data and our knowledge of the key retirement policy
variables used in our analysis, we believe that any such underestimate
would have been small.
We requested comments on a draft of this report from the Director of OPM,
and those comments are discussed at the end of this letter. We did our
13
The lookback adjustment, or comparative annuity computation, was established by P.L. 93-136 and
applied to retirees whose immediate annuities commenced on or after July 2, 1973, and before
January 20, 1981. Under this COLA provision, a retiree was assured that his or her annuity would be no
less than it would have been if the annuity had commenced on the effective date of the COLA and had
included the increase payable on that date. P.L. 96-499 eliminated the lookback adjustment and,
instead, provided for the proration of a retiree’s initial COLA increase.
Page 10 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
B-270600
review from December 1995 to July 1997 in Washington, D.C., according to
generally accepted government auditing standards.
As of 1995, 1.7 million retirees who were covered by the CSRS and/or FERS
Some Retirees’ pension plans were on the federal retirement rolls.14 Our estimate of the
Pensions Exceeded number of these retirees whose 1995 pensions exceeded their final salaries
Their Unadjusted differed, depending on whether we adjusted the retirees’ final salaries for
inflation. When we did not adjust the salaries for inflation, about 459,000,
Final Salaries or 27 percent, of the total general employee retirees received pensions that
in nominal dollars exceeded their final salaries. However, when we
adjusted the final salaries for inflation, no retiree received a pension that
exceeded his or her final salary.
As a general rule, using constant—rather than nominal—dollars is more
meaningful for examining dollar values across time, because constant
dollars correct for the effects of inflation or deflation. Constant dollars are
especially appropriate for comparing current pensions and final salaries,
because the number of years that the annuitants in our sample had been
retired averaged 22 years and ranged from 8 to 42 years. Table 2 compares
the 1995 pensions and the nominal and inflation-adjusted final salaries for
three illustrative retirees in our sample. The illustrative pensions shown in
the table are the average amounts received by those sample annuitants
who had retired in the years 1961, 1968, or 1981.
Table 2: a Comparison of the 1995
Pensions and Final Salaries Presented Nominal dollar terms Constant dollar terms
in Nominal and Constant Dollar Terms 1995 pension 1995 pension
for the Average Annuitants Who 1995 Final as a percent of Final as a percent of
Retired in 3 Selected Years. Retirement year pension salary final salary salary final salary
1961 $21,102 $ 7,290 289 $36,291 58
1968 $22,211 $10,175 218 $43,580 51
1981 $24,064 $21,594 111 $35,372 68
Source: GAO analysis of OPM data.
14
About 97 percent were CSRS retirees. Of the approximately 12,000 FERS annuitants added to the
retirement rolls in fiscal year 1995, about 30 percent had prior CSRS service.
Page 11 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
B-270600
Three factors help to explain why some retirees’ pensions came to exceed
Three Factors Help their final salaries when their salaries were not adjusted for the effects of
Explain Why Pensions inflation—the number and size of COLAs that retirees received, the number
Can Come to Exceed of years that they had been retired, and their number of years of federal
service. Two factors—the number and size of the COLAs that the retirees
Unadjusted Final had received and the number of years that they had been
Salaries retired—contributed because they helped to cause the retirees’ pension
amounts to increase over time. The third factor—years of federal
service—contributed because years of service was used in computing the
retirees’ initial pensions. Our regression model showed that the value of
the COLAs that the sample retirees received, as determined by the number
and size of COLAs and the length of employees’ retirement, together with
their years of federal service, explained about 82 percent of the variation
in the percentage by which the retirees’ pensions exceeded their
unadjusted final salaries. The important role that COLAs and length of
service played is a predictable consequence of pension policies that are
designed to reward employee service and maintain the purchasing power
of pensions.
During retirement, the retirees’ pensions increased because the COLAs that
the retirees were to receive increased in number. The amount of the
increase each year fluctuated according to changes in the CPI-W. In
contrast, unadjusted final salaries remained unchanged. Thus, the longer
the annuitants had been retired, the more COLAs they received and the
more likely it was that their pensions exceeded their unadjusted final
salaries. In fact, the average annuitant in our sample had been retired
about 22 years and had received 26 COLAs. The 4 percent who had retired
before 1963 had received 36 COLAs.
Generally, the likelihood that a retiree’s pension exceeded his or her
unadjusted final salary increased when the annuitant had been retired
during periods of high inflation, because larger COLAs were given during
these periods.15 Our model showed that, on average, a 1 percentage point
increase in the total value of the COLAs that a retiree had received would
result in a 0.5 percentage point increase in the amount by which the
retiree’s pension exceeded his or her final salary, other factors being
equal.16 In particular, more than 90 percent of the retirees in our sample
15
As noted, although CSRS and FERS COLA policies differ from each other and from COLA policies of
the past, these differences do not affect whether a pension would come to exceed an unadjusted final
salary, but rather, when.
16
In considering these and the other regression results in this report, it is important to recognize that
the results can be applied only to those retirees whose 1995 pensions had come to exceed their
unadjusted final salaries.
Page 12 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
B-270600
had been retired during all or part of the 1969 through 1980 period when
the most frequent and largest COLAs were given. Over this 12-year period,
pensions increased by 166 percent in nominal terms. Appendix I provides
a summary of COLA history since automatic COLAs were enacted in 1962.
The number of years of federal service also contributed to the explanation
of why some retirees’ pensions exceeded their unadjusted final salaries,
because years of service is included in determining the percentage of
high-3 average salary that a retiree ultimately will receive as his or her
initial pension. For example, under CSRS, an employee who had 41 years,
11 months of service at retirement would have been entitled to receive
80 percent of his or her high-3 average salary—the maximum percentage
allowed—while an employee who had worked 30 years would have been
entitled to receive 56.25 percent.17 As a result, the longer a retiree had
worked for the federal government, the closer the retiree’s initial pension
would have been to his or her unadjusted final salary. Nineteen (5 percent)
of the retirees in our sample had worked 40 years or more for the federal
government, and another 288 (83 percent) had worked 20 to 39 years. The
remaining 41 (12 percent) worked 5 to 19 years.18 Our model showed that
on average, a 1-year increase in a retiree’s federal service time would
result in about a 3.7 percentage point increase in the percentage by which
the retiree’s pension had exceeded his or her final salary, other factors
being equal.
A final factor—whether a retiree had chosen a survivor’s annuity
benefit—helped to explain why some retirees’ pensions had come to
exceed their unadjusted final salaries as much as they did. As noted in the
background section of this report, an employee who chooses a survivor
annuity benefit can have his or her basic annuity reduced by as much as
10 percent. As a consequence, if two retirees retired in the same year and
had the same final salaries and years of service, but only one had chosen a
survivor annuity benefit, the retiree who elected not to take the benefit
would have had a pension that exceeded his or her unadjusted final salary
sooner than the retiree who had chosen the survivor benefit.19 An
17
CSRS retirees may receive additional service credit for unused sick leave, which would allow them to
exceed the 80-percent rule. In contrast, FERS does not have a maximum percentage base. The formula
used to calculate initial annuities under FERS provides a lower annuity than the one used under CSRS.
Thus, it is unlikely that someone who has government service solely under the FERS pension plan
would receive as much as the maximum percentage base allowed under CSRS.
18
The vast majority (76 percent) of these annuitants retired under disability.
19
Also, retirees who had chosen a survivor’s annuity benefit and who became divorced or whose
spouses died during their retirement would have exceeded their final salaries sooner than they
otherwise would have because their pensions were increased due to a change of marital status.
Page 13 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
B-270600
employee who chose a survivor annuity benefit would have reduced the
initial pension and thus increased the gap between the initial annuity and
the final salary. Of the CSRS retirees in our sample, 48 percent were not
having survivor benefits deducted from their pensions.
Had current COLA policy—that is, the COLA policy enacted in 1984, which
Some Retirees’ established the formula and schedule used today by OPM—been in effect
Pensions Would Have without interruption since 1962, some sample retirees’ pensions would
Been Smaller, Others have been smaller than the pensions that they actually received, and other
retirees’ pensions would have been larger. Our simulations suggest that
Larger, Had Current other factors being equal, the majority of those who retired before 1970
Policy Been in Effect would have received smaller pensions, while about 90 percent of those
who retired after 1970 would have received larger ones.20 If current policy
Without Interruption had been in effect for all retirees in the sample, the number of retirees
whose pensions would have exceeded their unadjusted final salaries
would have increased by about 3 percentage points.
The Effects of COLA The following examples compare the pensions that retirees would have
Policies Would Have received under current versus actual COLA policy by simulating the effects
Differed, Depending on that changes in COLA policy would have had on pension amounts, other
factors being equal. The examples cover three different periods—1961 to
When Annuitants Retired 1995, 1968 to 1995, and 1981 to 1995—and show how the impacts would
have varied, depending on the period of retirement.21 In considering the
meaning of the figures, it is important to recognize that the trend lines
refer to current versus historical CSRS COLA policy. FERS lines were not
presented because, as stated earlier in this report, none of the FERS retirees
received an annuity that was based solely on his or her FERS participation.
Figure 1 shows the relative effects of current and actual policy for a CSRS
participant who retired in 1961. As the figure shows, if the current policy
had been in effect without interruption, the retiree’s pension would have
been smaller over the period. Our analysis showed that by 1995 the
retiree’s pension would have been 6.3 percent smaller than it was under
the actual COLA policy. However, as the gap shown between the 1995
pension and the unadjusted final salary amount makes clear, such a
20
The margin of error is plus or minus 5 percent with a 95-percent confidence interval.
21
As stated in the scope and methodology section of this report, we used the average initial pension for
the sample annuitants who retired in the first year of each period as the starting pension amount for all
three figures. However, the amount that we used for the beginning pension did not matter because, in
percentage terms, the impacts would have been the same for any beginning annuity that we selected.
Page 14 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
B-270600
reduction would not have been nearly enough to have caused the retiree’s
pension to fall below his or her final unadjusted salary.
Figure 1: Comparison of the Effects of Actual COLA Policy and Current COLA Policy, Had It Been in Effect for the Average
Sampled CSRS Employee Who Retired in 1961
Annuities (dollars in thousands)
22,000
20,000
18,000
16,000
14,000
12,000
10,000
8,000
6,000
4,000
2,000
0
Ap 8 1
Ap 3
'67
Au 71
Ja 8 4
5
'63
A u '74
Ja 6 5
A p 79
J a 87
J a 91
'62
80
'64
Oc 69
A p 78
'92
F e '66
D e '6 9
'7 4
Oc 8
Oc 9
'89
O c '8 0
'68
Se 75
'85
Ja '88
'90
'61
Ju 0
Ap 7
M a '82
Ap 4
3
'86
Au 72
'73
Ap 7 6
O c '77
Ap 5
'9
r. '9
7
r. '7
7
7
r. '9
y '8
'7
t. '
r. '
ly '
t. '
'
'
t. '
c. '
'
'
t. '
t. '
r. '
'
'
r. '
n.
n.
ne
b.
n.
n.
n.
n.
b.
b.
n.
n.
n.
r.
g.
r.
r.
b.
g.
g.
r.
pt.
Ap
Ap
Ja
Fe
Fe
Ja
Ja
Ja
Fe
Ju
Years in retirement
Actual paid policy (CSRS)
Current COLA Policy (CSRS)
Final salary
Years in which semiannual COLAs occurred
Source: GAO analysis of OPM data.
Page 15 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
B-270600
Figure 2 shows similar results for an annuitant who retired in 1968. In this
example, our analysis showed that the retiree’s pension would have been
3.5 percent smaller if current policy had been in effect without
interruption. The reduction in this annuitant’s pension is less
proportionally than the reduction in the pension of the annuitant who had
been retired since 1961 (shown in fig. 1), primarily because of the
difference in the number of the COLAs that were received and, to a lesser
extent, the shorter period of compounding. Again, the reduction would not
have been large enough to cause the retiree’s 1995 pension to fall below
his or her unadjusted final salary.
Page 16 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
B-270600
Figure 2: Comparison of the Effects of Actual COLA Policy and Current COLA Policy, Had It Been in Effect for the Average
Sampled CSRS Employee Who Retired in 1968
Annuities (dollars in thousands)
22,000
20,000
18,000
16,000
14,000
12,000
10,000
8,000
6,000
4,000
2,000
0
5
7
8
9
'75
'6 8
9
4
0
1
2
'8 4
'8 6
1
2
3
4
5
0
5
7
8
9
0
1
2
9
6
0
3
4
3
7
8
9
A pr. '7
A pr. '9
A pr. '7
A pr. '7
Jan . '9
Aug. '7
Aug. '7
July '7
Feb. '7
Apr. '6
Sept. '7
A ug. '7
Feb. '7
Apr. '7
Jan. '8
Jan. '8
Jan. '8
Jan. '8
Jan. '9
Jan. '9
Jan. '9
Dec. '6
Apr. '8
O ct. '8
Apr. '8
Apr. '8
May '8
Apr. '9
O ct. '7
O ct. '7
O ct. '7
S ept.
Years in retirement
Actual paid policy (CSRS)
Current COLA policy (CSRS)
Final salary
Years in which semiannual COLAs occurred
Source: GAO analysis of OPM data.
The third example (fig. 3) shows the results for an annuitant who retired in
1981. The retiree’s pension would have been larger if current policy had
been in effect without interruption. As the figure shows, under actual
policy, the retiree did not receive a COLA in 1984 or 1986, which caused this
retiree’s pension to fall somewhat short of the pension that he or she
would have received had current policy been in effect. Because the effects
Page 17 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
B-270600
of these suspensions continued to be reflected in the pension amounts that
the retiree received in subsequent years, by 1995 the retiree’s pension
would have been 1.4 percent larger under current, compared to historical,
COLA policy.
Figure 3: Comparison of the Effects of
Actual COLA Policy and Current COLA
Policy, Had It Been in Effect for the Annuities (dollars in thousands)
Average Sampled CSRS Employee
26,000
Who Retired in 1981
24,000
22,000
20,000
18,000
16,000
14,000
12,000
10,000
8,000
6,000
4,000
2,000
0
1981 4/82 5/83 1984 1/85 1986 1/87 1/88 1/89 1/90 1/91 1/92 1/93 4/94 4/95
Years in retirement
Actual paid COLA (CSRS)
Current COLA policy (CSRS)
Final salary
Source: GAO analysis of OPM data.
Page 18 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
B-270600
The Percentage of Retirees The increases in the pensions of some sample retirees, if current policy
Whose Pensions Exceeded had been in effect the entire time, would have been enough to cause an
Their Unadjusted Salaries increase of 3.0 percentage points in the number of retirees whose pensions
exceeded their unadjusted final salaries. When we estimated what the
Would Have Been Higher If sample retirees’ pensions would have been if current policy had been in
Current Policy Had Been in effect without interruption, we found that about 29 percent of retirees
Effect would have had annuities that exceeded their unadjusted final salaries,
compared to about 26 percent under the actual policy simulation.22
Although the difference was quite small, it was statistically significant.23
The two estimates differed by about 3 percentage points in part because
the effects of COLAs on pension amounts are cumulative and compound. In
particular, the suspensions of COLAs during 1980s tended to offset the COLA
policies of the 1960s and 1970s that overcompensated for inflation.
Our analysis of the effects that COLA policies have had on retiree pensions
Observations shows that the policies have played an important role in maintaining the
purchasing power of retiree pensions since automatic COLAs began.
Although COLA policies of the 1960s and 1970s overcompensated for the
effects of inflation as measured by the CPI, COLA policies of the 1980s
sometimes under-compensated. And, although current COLA policy would
have tracked the CPI more closely had it been applied over the period we
reviewed compared with some past COLA policies, the numerous changes
that have been made in COLA policies over the past 35 years did not cause
any retiree’s pension to exceed his or her final salary when the salaries
were adjusted for inflation.
Our analysis also shows that the effects that COLA policies actually have on
retiree pension amounts cannot be summarized easily. Generalization is
difficult, in part because no one COLA policy has ever been implemented for
a sustained period. For example, although the current underlying policy
22
Since legislative changes made after 1984 did not permanently affect the COLA formula or schedule,
we did not include them in our analysis of current COLA policy. However, these changes were
included in our actual COLA policy analysis. Thus, because our simulation of COLA policies used the
initial annuity as the starting point for adding COLAs, our simulation did not include any adjustments
(e.g., loss of survivor’s annuity benefit due to spouse’s death) to annuities subsequent to the
calculation of the initial annuity. When these adjustments are considered by using the annuity received
in 1995, the percentage of those retirees exceeding their final salaries is 27 percent.
23
Of the 398 sample cases for which data were available, 38 had pensions that were below their final
salaries under actual COLA policy but above their final salaries under current COLA policy. None of
the sample retirees whose pensions were above their final salaries under actual COLA policy had
pensions below their final salaries under current COLA policy. The estimate that about 3 percent more
of the pensions would have exceeded final salaries under current COLA policy compared to actual
COLA policy is surrounded by a 95-percent confidence interval that extends from about 2 to 4 percent.
Page 19 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
B-270600
has been in effect since 1984, Congress has modified this policy several
times for limited periods to help reduce the deficit. Also, the effects of
many individual COLAs and COLA policy changes are cumulative and
compound over time. As a consequence, COLA policy changes have affected
individual retirees differently, depending on when they retired. In
particular, the effects of the COLA policies of the 1960s and 1970s that
overcompensated for inflation will continue to have an effect on retiree
pensions for as long as those who received them are alive, just as not
receiving scheduled COLAs in 1984 and the suspension of COLAs in 1986 will
continue to be reflected in the pensions of anyone who retired before
these years.
We received oral comments on a draft of this report from OPM on July 16,
Agency Comments 1997. OPM officials who provided comments included Federal Retirement
and Our Evaluation Benefits Specialists from the Retirement Policy Division and a Program
Analyst from the Retirement and Insurance Service. These officials
generally concurred with the information and conclusions presented in
our report. In particular, they agreed that using constant dollars, rather
than nominal dollars, is a more meaningful way to compare retiree
pensions to final salaries and that the statutory factors that are designed to
maintain pension purchasing power and reward employees with longer
service play a major role in determining whether pensions come to exceed
nominal final salaries. These officials also provided a number of technical
and clarifying comments, which we incorporated into this report where
appropriate.
We are sending copies of this report to the Ranking Minority Member of
your Committee and the Chairmen and Ranking Minority Members of the
Subcommittee on International Security, Proliferation, and Federal
Services, Senate Committee on Governmental Affairs; and to the
Subcommittee on Civil Service, House Committee on Government Reform
and Oversight. Copies of this report are also being sent to the Director of
OPM and other parties interested in federal retirement matters and will be
made available to others upon request.
Page 20 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
B-270600
Major contributors to this report are listed in appendix II. If you have any
questions, please call me at (202) 512-9039.
Sincerely yours,
Michael Brostek
Associate Director, Federal Management
and Workforce Issues
Page 21 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
Contents
Letter 1
Appendix I 24
Summary of COLA
History Since
Automatic COLAS
Were Enacted in 1962
Appendix II 26
Major Contributors to
This Report
Tables Table 1: Major Changes Made to COLA Policy Since Automatic 4
Adjustments Began
Table 2: A Comparison of the 1995 Pensions and Final Salaries 11
Presented in Nominal and Constant Dollar Terms for the Average
Annuitants Who Retired in 3 Selected Years.
Figures Figure 1: Comparison of the Effects of Actual COLA Policy and 15
Current COLA Policy, Had It Been in Effect for the Average
Sampled CSRS Employee Who Retired in 1961
Figure 2: Comparison of the Effects of Actual COLA Policy and 17
Current COLA Policy, Had It Been in Effect for the Average
Sampled CSRS Employee Who Retired in 1968
Figure 3: Comparison of the Effects of Actual COLA Policy and 18
Current COLA Policy, Had It Been in Effect for the Average
Sampled CSRS Employee Who Retired in 1981
Abbreviations
COLA cost-of-living adjustment
CPI consumer price index
CPI-W wage earner index
CRS Congressional Research Service
CSRS Civil Service Retirement System
FERS Federal Employees Retirement System
OPM Office of Personnel Management
Page 22 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
Page 23 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
Appendix I
Summary of COLA History Since Automatic
COLAS Were Enacted in 1962
Effective Date CSRS FERS
Measuring period datea paidb COLA COLAc
* 1/63 2/63 5.0 -
** 12/65 1/66 6.1 -
** 1/67 2/67 3.9 -
** 5/68 6/68 3.9 -
** 3/69 4/69 3.9 -
** 11/69 12/69 5.0 -
** 8/70 9/70 5.6 -
** 6/71 7/71 4.5 -
** 7/72 8/72 4.8 -
** 7/73 8/73 6.1 -
** 1/74 2/74 5.5 -
** 7/74 8/74 6.3 -
** 1/75 2/75 7.3 -
** 8/75 9/75 5.1 -
** 3/76 4/76 5.4 -
June-December 1976 3/77 4/77 4.8 -
December-June 1976/77 9/77 10/77 4.3 -
June-December 1977 3/78 4/78 2.4 -
December-June 1977/78 9/78 10/78 4.9 -
June-December 1978 3/79 4/79 3.9 -
December-June 1978/79 9/79 10/79 6.9 -
June-December 1979 3/80 4/80 6.0 -
December-June 1979/80 9/80 10/80 7.7 -
June-December 1980 3/81 4/81 4.4 -
Dec. 1980-Dec. 1981 3/82 4/82 8.7 -
Dec. 1981-Dec. 1982 4/83 5/83 3.9d -
e
3rd qtr. 1984-3rd qtr. 1983 12/84 1/85 3.5 -
3rd qtr. 1985-3rd qtr. 1984 12/85 1/86 0.0 -
3rd qtr. 1986-3rd qtr. 1985 12/86 1/87 1.3 -
3rd qtr. 1987-3rd qtr. 1986 12/87 1/88 4.2 -
3rd qtr. 1988-3rd qtr. 1987 12/88 1/89 4.0 3.0
3rd qtr. 1989-3rd qtr. 1988 12/89 1/90 4.7 3.7
3rd qtr. 1990-3rd qtr. 1989 12/90 1/91 5.4 4.4
3rd qtr. 1991-3rd qtr. 1990 12/91 1/92 3.7 2.7
3rd qtr. 1992-3rd qtr. 1991 12/92 1/93 3.0 2.0
3rd qtr. 1993-3rd qtr. 1992 3/94 4/94 2.6 2.0
(continued)
Page 24 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
Appendix I
Summary of COLA History Since Automatic
COLAS Were Enacted in 1962
Effective Date CSRS FERS
Measuring period datea paidb COLA COLAc
3rd qtr. 1994-3rd qtr. 1993 3/95 4/95 2.8 2.0
3rd qtr. 1995-3rd qtr. 1994 3/96 4/96 2.6 2.0
Legend
* = Adjustments made whenever the CPI in a year exceeded the CPI in the base year by
3 percent or more.
** = Adjustments made whenever the CPI in a month rose by at least 3 percent over the month of
the last adjustment and remained at or above that level for 3 consecutive months.
a
The “effective date” column indicates the month the COLA went into effect.
b
The “date paid” column indicates the month the retiree received the COLA.
c
All disability retirees (and survivors) and nondisability retirees age 62 or over. (The first FERS
COLA was effective in December 1988 and payable in January 1989.)
d
The COLA rate was 3.3 percent for nondisabled retirees under age 62.
e
Due to a change in the adjustment period, no COLA paid in 1984.
Sources: OPM and CRS.
Page 25 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
Appendix II
Major Contributors to This Report
Margaret T. Wrightson, Assistant Director
General Government Gregory H. Wilmoth, Senior Social Science Analyst
Division, Washington,
D.C.
Tyra J. DiPalma, Senior Evaluator
Dallas Field Office Enemencio S. Sanchez, Evaluator
In addition to those named above, Jerry T. Sandau, Social Science Analyst,
Acknowledgement GGD, contributed through his development of the regression analysis
results presented in this report.
(410004) Page 26 GAO/GGD-97-156 Relationship Between Retiree Pensions and Final Salary
Ordering Information
The first copy of each GAO report and testimony is free.
Additional copies are $2 each. Orders should be sent to the
following address, accompanied by a check or money order
made out to the Superintendent of Documents, when
necessary. VISA and MasterCard credit cards are accepted, also.
Orders for 100 or more copies to be mailed to a single address
are discounted 25 percent.
Orders by mail:
U.S. General Accounting Office
P.O. Box 6015
Gaithersburg, MD 20884-6015
or visit:
Room 1100
700 4th St. NW (corner of 4th and G Sts. NW)
U.S. General Accounting Office
Washington, DC
Orders may also be placed by calling (202) 512-6000
or by using fax number (301) 258-4066, or TDD (301) 413-0006.
Each day, GAO issues a list of newly available reports and
testimony. To receive facsimile copies of the daily list or any
list from the past 30 days, please call (202) 512-6000 using a
touchtone phone. A recorded menu will provide information on
how to obtain these lists.
For information on how to access GAO reports on the INTERNET,
send an e-mail message with "info" in the body to:
info@www.gao.gov
or visit GAO’s World Wide Web Home Page at:
http://www.gao.gov
PRINTED ON RECYCLED PAPER
United States Bulk Rate
General Accounting Office Postage & Fees Paid
Washington, D.C. 20548-0001 GAO
Permit No. G100
Official Business
Penalty for Private Use $300
Address Correction Requested
Federal Pensions: Relationship Between Retiree Pensions and Final Salaries
Published by the Government Accountability Office on 1997-08-11.
Below is a raw (and likely hideous) rendition of the original report. (PDF)