The Government Accountability Office (GAO) recently released a study on how participants
fare in cash balance plan conversions. In a sharp departure from earlier
studies, the GAO report concludes that such conversions significantly reduce most
participants’ retirement benefits. According to the new study, only 13 percent of workers
aged 30 at conversion, 1.8 percent of those aged 40 and 0.8 percent of those aged 50
would receive better benefits under the cash balance plan than they would have under the
traditional plan it replaced. Fueled by the report, congressional critics of cash balance and
other hybrid plans released a meeting notification on Capitol Hill proclaiming: “New GAO
Report Makes Clear That Cash Balance Conversions Hurt All Workers.”
The GAO’s report is both mistaken and misleading. The analysis is based on an ill-defined
methodology, and its universe of plan designs does not reflect the real-world experience of
sponsors or participants. The report’s conclusions are simply not supportable.
The analysis is presented in three sections: a literature review of previous research, a
description of “typical” traditional and cash balance plans, and an analysis of a conversion’s
effects on different employees. The methodology, assumptions and analysis are flawed at
every one of the three stages.
GAO Literature Review
The opening sentences of the literature review note that “current pension and economic
literature provides little conclusive evidence about the effects of CB plan conversions on
benefits. In many cases, data and other methodological issues…limit the generalization
The report essentially dismisses all research undertaken by business, nonprofit and academic
researchers. The GAO apparently considers all research conducted by private-sector
analysts to be biased, and the report lists but generally ignores research done by analysts at
the Federal Reserve Board and the Urban Institute.
As for Watson Wyatt’s research, our analysts met at length with GAO staff to discuss our
2000 study, The Unfolding of a Predictable Surprise: A Comprehensive Analysis of the Shift from
Traditional to Hybrid Plans. We based that study on 78 plan conversions before 2000, and a
subsequent analysis studied another 55 plan conversions from 2000 through 2004. Our
analysts were able to access detailed conversion data, including the prior plan formula, the
new formula, conversion factors and dates, transition provisions, and changes to other
retirement plans adopted at the same time.
Yet the GAO report specifically cites this work as insufficiently documented, concluding
that “because some specifics of the simulations presented in some studies do not include
sufficient detail, it is difficult to evaluate the quality of the estimates in some cases.”
The report does not acknowledge the detailed discussions of the research by Watson Wyatt
researchers and GAO staff. Nor does it mention Watson Wyatt’s offer to allow GAO staff to
review all the conversions, as long as the GAO agreed not to identify specific companies.
The report also criticizes Watson Wyatt’s study of employers’ reasons for shifting to a CB
plan, because the study used “survey data in an attempt to determine the reasons why
employers initiate CB plan conversions.” According to the GAO, these studies “contain
methodological limitations and base their conclusions on employers’ self perceptions along
with additional biases, and cannot be extended beyond the small samples of firms
studied.” In other words, the GAO determined that asking employers why they switched
to a cash balance design was an inappropriate methodology and possible evidence of bias.
Finally, the GAO report concludes the review of prior literature with the assertion that
“Current research provides limited evidence as to...why sponsors convert to CB plans.”
While it is undeniable that all such research contains some limits — it is impossible
to uncover all motivating factors for every hybrid plan conversion — the GAO’s report
fails to provide any additional evidence or insights into why employers convert to cash
GAO Plan Sample Analysis
The GAO’s description of its methodology is extremely vague, failing to explain the data
source or collection methods. The report indicates that the “sample design for this study
was a stratified random sample of CB plans, the 45 largest plans comprising the first
stratum, and an additional 160 plans selected from the remaining plans, producing a total
sample of 205 plans.” Of these, the report states that 31 of the large plans and 102 of the
smaller plans met the conversion criteria, for a total of 133 plans — the same number of
conversions in Watson Wyatt’s prior two studies, which the GAO dismissed as not being
“a reliable guide to gauging the impact of a typical cash balance conversion on workers.”
Based on the sample size, the report claims a 95 percent confidence level that the results
are within ± 9 percentage points. So the estimate that 47 percent of plan converters offered
some level of grandfathering suggests a real-life range between 38 percent and 56 percent.
Analysts gathered this evidence by reviewing plan sponsors’ Form 5500s and answering
51 questions about the plan conversions, including characteristics of the traditional plan,
conversion dates, affected employees, transition provisions and ongoing features of the
CB plan such as pay and interest credits.
While the GAO study criticizes the methodologies used by other researchers, it does not
reveal exactly what data were collected from the Form 5500s or the completion rates for
the information elements. The only hint of the robustness of the underlying data is the
report’s indication that “23 of 39 plans with data available used conversion interest rates
within 1% of the prior month’s 30-year Treasury rate,” suggesting that the determination of
the interest rate used to set initial cash balances was based on 39 — rather than 205 or
even 133 — conversions.
As noted above, it is not clear where the GAO found its information. Plan sponsors do not
report the interest rates used to calculate opening balances in cash balance conversions or
the number of grandfathered participants on their Form 5500s. Based on the testing
methodology described in the report, a 95 percent confidence level of results within
±9 percentage points seems unwarranted.
GAO Simulation of Plan Conversions
The GAO’s study was based on conversions of a traditional final average pay (FAP) plan
design to a cash balance plan. But the GAO’s “typical FAP plan” bears little resemblance to
typical real-life pension plans. As for the cash balance plan, the report doesn’t provide
enough information to determine its benefit formula.
The “typical” FAP plan in the GAO study is an excess integrated plan. Its base accrual rate
is 1.5 percent per year of service for the average of the highest five consecutive years’ pay,
and the rate for pay over a bend point based on Social Security limits is 1.95 percent.
While the study does not explain the derivation of this formula, it is clearly not typical
of traditional FAP plans. According to the GAO’s own supporting material, less than
26 percent of FAP plans use an excess integrated formula, and almost 63 percent of FAP
plans are not integrated with Social Security at all.
The same supporting material states that the mean base benefit percentage for excess integrated
plans is 1.07 percent — significantly less than the 1.5 percent in the study’s FAP
plan. Although the supporting material provides mean base benefit percentages for different
industries, the average base rate never comes close to 1.5 percent.
The GAO’s typical FAP plan would pay a benefit of 45 percent of final pay to a
30-year-career worker with earnings below the bend point. At a 1.07 percent accrual
rate, the accumulated benefit would be 30.2 percent of average earnings. According to
Watson Wyatt’s 2003-2004 COMPARISON™ database of pension plans, at the 50th percentile
of plans, a worker with 30 years of service and a final average salary below the
formula bend point would receive a benefit of 34.3 percent — much closer to the 1.07
formula identified as typical in the study’s supporting material than the plan formula used
by the study. In short, the GAO’s FAP plan would provide 50 percent more benefits than
an average real-world plan would provide to comparable participants.
The GAO’s “typical” cash balance plan provides an increasing pay credit based on the participant’s
age and service, starting with a minimum of 3 percent and increasing to a maximum
of 9 percent. The report does not state why or when pay credits increase, making it
impossible to calculate benefits or to assess whether the plan is truly typical.
According to the report’s supporting material, however, 28 percent of cash balance plans
have a flat contribution rate unrelated to age or service, and the average pay credit is
estimated at 4.4 percent per year. Assuming that pay credits increase in quarter percentage
increments along the natural growth path from the 3 percent minimum to the 9 percent
maximum, the GAO’s typical cash balance plan appears to provide lower benefits to
younger workers than the average plan. For a worker who starts a job at age 25, a flat
pay credit rate of 4.4 percent would exceed the GAO’s simulated benefit for the first
20 years of service. In other words, many cash balance plans with flat rates would provide
significantly more generous benefits to workers with less than 20 years of service than the
plan cited as typical by the GAO.
Moreover, the report’s supporting material indicates that almost one-third of cash balance
plans provide increasing pay credits based on a combination of age and service, similar
to the GAO’s typical plan. For those plans, the average minimum contribution rate is
3.4 percent per year of service and the maximum rate is 11.8 percent, so both the average
minimum and the average maximum fall outside the range used by the GAO. In other
words, most of these plans would provide more generous benefits than the GAO’s typical
cash balance plan.
For the remaining 40 percent of cash balance plans, about half have a flat- or fixed-pay
crediting rate and base additional pay credits on service. The others increase pay credits
according to service alone. According to the report’s supporting material, the average base
contribution rate is 4 percent — 33 percent more generous than the base rate in the GAO
plan. The average maximum rate is 8.2 percent — about 9 percent below the maximum
rate in the GAO plan. It is impossible to project the effects of alternative accrual
patterns without looking at specific schedules. But higher pay credits earlier in the career
boost eventual benefit amounts considerably due to interest compounding. The GAO’s
typical cash balance plan provides lower pay credits early in participants’ careers than the
average plans described in its own supporting material.
The GAO’s analysis seems to systematically overestimate the benefits paid by typical
traditional pension plans and systematically underestimate the benefits provided by
typical cash balance plans. Participants would naturally be worse off after a cash balance
conversion if the traditional plan were much richer than average and the cash balance
plan much poorer.
The report’s comparison of a typical FAP plan converted to a “constant cost” cash balance
plan is also confusing, since it concluded that most workers would be better off under
a traditional plan than in an equally generous cash balance plan. In estimating
post-conversion costs and benefits, it is generally assumed that 80 percent of a vested
workforce will no longer be working for their current employer at age 55, with some variance
according to industry, age makeup of the workforce and so forth. It would be very
unusual for most, or even half, of an employer’s vested workers to still be working for the
same firm at age 55. Shifting from a traditional FAP plan to a cash balance or other hybrid
plan would redistribute benefits among participants, increasing benefits for employees
with shorter tenures and those who leave the employer before becoming eligible for retirement.
Given employee turnover rates, the only way that most of a vested workforce would
receive greater benefits from a traditional FAP plan would be if the FAP plan were more generous
overall than the cash balance plan.
One point made in the GAO’s report is noteworthy: Shifting to a cash balance plan
consistently delivers greater benefits to affected participants than simply terminating
the old plan. Given the number of defined benefit plans that have terminated in recent
years — leaving workers to fund their own retirement — this point should be emphasized.
For many employees, a cash balance plan would be an enormous step up in
Using the pension simulation model referenced in the supporting materials should have
enabled the GAO to perform a fair and reasonable analysis of cash balance conversions.
The pension simulation model, called PENSIM, was funded by the federal government to
gain a detailed understanding of the current range of plans and their prevalence. Rather
than comparing the effect of converting an atypical FAP plan to an atypical cash balance
plan, GAO analysts could have used PENSIM to model what happens to workers shifting
out of and into a range of truly typical plans. Such an analysis could have contributed
valuable data and new insights to the public debate on retirement.
The defined benefit system is in trouble. Many employers are giving up on the system
altogether, leaving their employees to fend for themselves. Cash balance and other hybrid
plan sponsors are voluntarily choosing to stay in the system so they can continue providing
valuable and secure retirement benefits to millions of U.S. workers.
Converting a traditional defined benefit plan to a hybrid plan design may redistribute
future benefits, potentially disrupting some participants’ retirement plans. To minimize the
disruption, most employers provide transition benefits and protections to older workers.
The GAO’s analysis did not study typical plans and so did not ascertain typical effects on
workers. Any study that purports to analyze an important retirement trend but fails to
consider workplace and economic realities does a disservice to the current public debate
and the retirement future of American workers.