Personal Reemployment Accounts: Results from Bonus Experiments
Prepared for Members and Committees of Congress
The purpose of Personal Reemployment Accounts (PRAs) is to provide persons likely to exhaust
their Unemployment Insurance (UI) benefits a choice in the type and source of reemployment
services and to induce UI claimants and exhaustees to speed their reemployment by providing a
bonus equal to the balance in their Personal Reemployment Accounts when they obtain new jobs.
PRA recipients would have to pay from their accounts for some reemployment services now
available free at one-stop centers, if they chose to utilize those services (e.g., training). The Bush
Administration initially proposed the accounts in its 2003 economic stimulus package and
included a PRA demonstration in its FY2005 budget request.
In March 2005, the House passed H.R. 27, a bill that would reauthorize the Workforce Investment th
Act (WIA) and authorize a PRA demonstration. (During the 108 Congress, the House passed
H.R. 444, which included PRA provisions identical to those in H.R. 27.) The U.S. Department of
Labor announced in October 2004 that seven states had been selected to participate in a PRA
PY2004 demonstration. Funding for the demonstration was not included in FY2005
appropriations, and the Administration did not request PRA funds for FY2006. In September
2005, it proposed recovery accounts for workers displaced from jobs as a result of the Gulf Coast
hurricanes that appear to be much like PRAs. H.R. 3976, introduced in October 2005, similarly
would create worker recovery accounts for those who lost jobs due to Hurricanes Katrina and
Proponents of the concept point to the results of experiments conducted during the 1980s as
support for the efficacy of a reemployment bonus. The experiments produced mixed results: only
one of six bonus designs in Washington significantly reduced the length of unemployment and the
amount of UI benefits, while all but one of the designs in Pennsylvania significantly sped
reemployment; in both states, the costs of the bonus experiments exceeded their benefits to the UI
system. Based upon data from the two experiments, it was subsequently estimated that targeting
bonus offers to likely UI exhaustees yielded somewhat better outcomes than untargeted bonuses
and that varying the value of the bonus could affect the number of weeks UI benefits must be paid
as well as the bonus receipt rate. Another PRA simulation involving likely UI exhaustees in
Georgia estimated that 59% would not have qualified for a bonus; it suggested that PRA
recipients might reduce their use of fee-based reemployment services to maximize the bonus
The design of the experiments differs from the PRA demonstrations in H.R. 27 and the PY2004
project in ways that could affect the applicability of the experiments’ results. For example, those
eligible for the demonstrations would be more disadvantaged than experiment participants. The
experiments’ bonus amount generally was a fixed amount, while the demonstration participants
could affect its value based on their use of fee-based services. Further, experiment participants
had to remain employed for a given time before qualifying for a bonus; demonstration
participants would receive a portion of their PRA balances immediately upon reemployment. This
report will be updated as warranted.
Personal Reemployment Accounts and the Workforce Investment Act..........................................1
A Comparison of the Design of Bonus Experiments and of the Current Bonus Proposal..............3
Results of Selected State Bonus Experiments.................................................................................6
The Washington Reemployment Bonus Experiment................................................................6
The Pennsylvania Reemployment Bonus Experiment..............................................................8
Reanalysis of the Washington and Pennsylvania Experiments...............................................10
A Targeted Bonus..............................................................................................................10
What Happens When You Change Program Characteristics?...........................................10
Illinois Claimant Bonus Experiment........................................................................................11
Table 1. Treatments in the Washington Experiment........................................................................6
Table 2. Treatments in the Pennsylvania Experiment......................................................................9
Author Contact Information..........................................................................................................12
s part of its 2003 economic stimulus package, the Bush Administration proposed Personal
Reemployment Accounts (PRAs). The Secretary of Labor testified in February 2003
before the House Committee on Education and the Workforce that the accounts are A
intended not only to speed reemployment into stable jobs through provision of a cash bonus to
Unemployment Insurance (UI) claimants that states determine are likely to exhaust their benefits,
but also to give the targeted individuals an opportunity to choose the type of services (e.g.,
retraining, child care, and transportation) and the source of services (i.e., one-stop delivery system
or other service providers) they believe will be most useful to them in their pursuit of new jobs.
Proponents of the accounts point to the results of experiments conducted in four states during the
mid- to late 1980s as support for the efficacy of offering a new-jobs bonus to UI claimants. In her
testimony, the Secretary stated that experiments in Washington, Pennsylvania, and New Jersey
showed that a reemployment bonus of $300 to $1,000 motivated the recipients to become
reemployed, reduced the duration of UI by almost one week, and resulted in new jobs
comparable in earnings to those obtained by workers who were not eligible for the bonus and 1
remained unemployed longer.
Further, she noted that in an experiment that took place in Illinois, “a reemployment bonus of
$500 reduced the duration of unemployment by more than a week and did not lead to lower
earnings at the worker’s next job.”
This report begins by briefly summarizing the relevant provisions contained in legislation th
pending before the 109 Congress. It then compares the design elements of the state bonus
experiments with those of the bonus demonstrations included in legislation to reauthorize the
Workforce Investment Act and in a Program Year (PY) 2004 project of the U.S. Department of
Labor (DOL). The report closes with a review of the results of selected state bonus experiments.
On January, 4, 2005, two bills were introduced in the House that would authorize a PRA
demonstration under the Workforce Investment Act (WIA): H.R. 26, a stand-alone bill, and H.R. 2
27, a broader bill to reauthorize WIA. (WIA, 29 U.S.C. §2811 et seq.). Both bills contain
identical language about the demonstration project under Section 171 of WIA to test and evaluate 3
PRAs. On March 2, 2005, the House passed H.R. 27.
The principle features of the PRA demonstration are as follows:
1 U.S. Congress, House Committee on Education and the Workforce, 108th Congress, 1st sess., Back to Work: The
Administration’s Plan for Economic Recovery and the Workforce Investment Act, Feb. 12, 2003, Serial no. 108-1, pp.
59-60 (statement of Elaine Chao, Secretary of Labor).
2 For more information on WIA, see CRS Report 97-536, Job Training Under the Workforce Investment Act: An
Overview, by Ann Lordeman.
3 The PRA provisions of H.R. 26 and H.R. 27 are identical to the provisions contained in H.R. 444, passed by the
House in the 108th Congress.
• States and other eligible entities would provide accounts of up to $3,000 per
eligible individual. The initial account balances must be the same for participants
within a local area or within a state.
• Eligible individuals primarily would be UI claimants identified through worker
profiling as likely to exhaust their benefits and in need of job search assistance to 4
obtain new employment. Eligible entities could, at their option, extend accounts
to individuals who had exhausted their benefit entitlement and (a) are in training
for which completion requires additional support or (b) were laid off from an
industry or occupation with declining employment or which no longer provides
jobs in a local area.
• Individuals could use the account funds, at their own discretion, to purchase a
variety of employment-related services (i.e., intensive services, training, and
support services and assistance with buying or leasing a car) on a fee-for-service
basis from the one-stop delivery system or other service providers. Recipients
could only receive training, intensive, and support services on a fee-for-service
basis during the one-year period from the date of establishment of their accounts.
Core services provided by one-stop centers would remain free to account
recipients (e.g., access to job listings and assistance with writing résumés).
• Individuals with PRAs who obtain full-time jobs within a 13-week qualification
period (dating from initial receipt of UI benefits for those still eligible for
benefits and from the account’s establishment for those who had exhausted their
benefits) would receive any funds remaining in their accounts as bonuses. The
bonuses would be dispensed in two installments: 60% upon reemployment and
These features are similar to an $8 million PRA demonstration for which the U.S. Department of 5
Labor (DOL) announced awards to seven states on October 29, 2004. For FY20005, the
President had requested $50 million under the authority of WIA Section 171 for a PRA
demonstration. Funding for the demonstration was not included in FY2005 appropriations (P.L.
As part of the Administration’s response to Hurricanes Katrina and Rita, it proposed recovery
accounts—seemingly like PRAs—for workers displaced from their jobs as a result of the natural
disaster. H.R. 3976, introduced in October 2005, similarly would create worker recovery accounts
for workers who lost jobs due to the Gulf Coast hurricanes and who are likely to exhaust the
benefits they are receiving under the Disaster Unemployment Assistance or Unemployment
Insurance programs. No action has been taken on this approach to aiding workers displaced as a
result of the hurricanes.
4 P.L. 103-152, which amended the Social Security Act, required states to develop systems of profiling new UI
claimants in order to identify those likely to exhaust their benefits. After states screen out initial claimants on recall
status (including in some cases persons laid off from seasonal industries) and those who exclusively use union hiring
halls, the states use statistical models or characteristic screens to identify potential UI exhaustees. Measures taken into
account include previous occupation, industry, wages, and job tenure; educational attainment and other claimant
characteristics; local economic indicators; and the individual’s UI weekly benefit amount. States are not permitted to
include such equal opportunity characteristics as age, race, gender, and disability status.
5 The seven states are Florida, Idaho, Minnesota, Mississippi, Montana, Texas and West Virginia. For more
information, see http://www.dol.gov/opa/media/press/opa/OPA20042248.htm.
Bonus experiments were conducted in Washington, Pennsylvania, New Jersey, and Illinois. This
report omits the New Jersey experiment because it had a very limited eligible population (i.e.,
dislocated workers, who were defined as persons having worked for the same employer for three
years before filing for UI benefits). As a result, the findings from the experiment are not
generalizeable to a bonus program having a broader eligible population such as the one currently 6
proposed. New Jersey’s bonus experiment had a mandatory job-search component as well.
The experiments in Washington, Pennsylvania, and Illinois had a number of elements in common.
Differences between the design of the experiments and the current bonus proposal will affect how
well the results from the former can be applied to the latter.
In the experiments, UI claimants were randomly assigned to a control (i.e., comparison) group
that received existing services and to one or more treatment groups that received, in addition to
existing services, time-limited offers of bonuses which they were awarded after a specified period
of reemployment. Accordingly, treatment group members did not differ from typical UI recipients
in any particular way (e.g., they were not more likely to exhaust their benefits). In addition,
participants in the experiments excluded persons who already had exhausted their UI benefits. On
both counts, then, individuals in the bonus experiments were likely to have better reemployment
prospects than the eligible population of current legislation and DOL’s PY2004 demonstration.
Even among persons profiled as most likely to exhaust UI benefits, a PRA simulation utilizing
employment service records of Georgia UI claimants found that 59% would have been ineligible
for a bonus given their reemployment histories. About two-fifths found jobs within 13 weeks and
consequently would have qualified for an initial (60%) bonus payment. A little more than one-7
fourth remained employed long enough to have received both bonus payments.
6 The experiment also does not provide “much guidance for policy, because ... bonus offers were made only after seven
weeks of insured unemployment, and the pending offer was unknown to the selected participants prior to that time.
Such a situation could not be replicated in a real program, as knowledge of the pending offer would be available to all
claimants from the start of their benefit year (and probably prior to that ... ). This knowledge can be expected to
critically affect job-search behavior during the first seven weeks of the benefit year, as well as during the period in
which the bonus was available.” Walter A. Corson and Robert G. Spiegelman, “Introduction and Background of the
Reemployment Bonus Experiments,” in Philip K. Robins and Robert G. Spiegelman, eds., Reemployment Bonuses in
the Unemployment Insurance System (Kalamazoo, MI: W.E. Upjohn Institute for Employment Research, 2001), p. 14.
(Hereafter cited as Robins and Spiegelman, Reemployment Bonuses in the Unemployment Insurance System.)
7 Christopher J. O’Leary and Randall Eberts, “Personal Reemployment Accounts,” Employment Research, vol. 11, no.
1 (Jan. 2004). (Hereafter cited as O’Leary and Eberts, Personal Reemployment Accounts.) See also Christopher J.
O’Leary and Randall Eberts, Personal Reemployment Accounts: Simulations for Planning Implementation, W.E.
Upjohn Institute Staff Working Paper No. 04-110, May 2004.
In most treatments, the value of the bonus offer was a multiple of an individual claimant’s weekly
benefit amount (WBA). The dollar value of bonuses therefore differed from one claimant to
another because of the claimants’ varying benefit entitlements. The bonus offers were the same,
however, in terms of the opportunity cost of unemployment to claimants (i.e., a reduction in one
or more weeks of unemployment cost each person one or more weeks of UI benefits).
As described in current proposals and the PY2004 demonstration, there is no link between the
value of a bonus and the value of an individual’s UI benefit. Instead, variability within a state in
participants’ bonuses (i.e., in their PRA balances) would depend upon their differing utilization of
fee-based reemployment services. While persons in the experiments could not control the size of
their bonuses, individuals in the proposed program could do so through their decisions about
expenditures from the accounts. Some might try to maximize their potential bonus by relying
only on the free services that still would be available to them at one-stop centers. Those in areas
with relatively low unemployment rates or with signs of an improving economy could be more
likely to think they need few, if any, fee-based reemployment services.
Some information on the possible utilization of different reemployment services can be gleaned
from the PRA simulation of likely UI exhaustees in Georgia. The study found that relatively few
persons drew upon training and supportive services during their first 13 weeks of benefit receipt.
In contrast, about two-fifths utilized two intensive services, namely, customer service plans and
counseling. The researchers estimated that the intensive services cost $356 and $712,
respectively—much less than the little-used options of training and supportive services ($1,424
and $1,068, respectively). The implementation of a PRA program might affect this usage,
however, as participants would be faced with having to pay for some services. The analysts
further estimated that if individuals spent the entire PRA amount of $3,000 to buy services
because they believed doing so would speed reemployment or help get higher paying jobs, the
unemployed workers “must expect either earnings to be nearly 14 percent higher or that
employment will occur at least 6 weeks sooner.” As they noted that earnings and employment
gains of this magnitude are not supported by prior research, “PRA recipients might therefore 8
reduce use of services in hopes of receiving larger reemployment bonuses.”
Another difference between the experiments and the currently proposed design with regard to the
value of the bonus is the amounts available to treatment group members were below the $3,000
maximum value. In 1989 dollars, the year when the experiments mostly took place, $3,000 was
equivalent to $2,040—considerably more than the $300-$1,000 range of bonuses in the 9
Of course, eligible entities would be free under the PRA provisions in H.R. 27 and in the PY2004
demonstration to set a bonus amount below $3,000, which would, in turn, allow them to make
offers to a larger share of eligible UI recipients. However, while $3,000 might appear to be a
generous bonus amount, it might not seem so in terms of covering the cost of reemployment
services for which account funds could be used. The adequacy of the PRA amount would vary by
8 O’Leary and Eberts, Personal Reemployment Accounts, pp. 3-4.
9 Paul T. Decker and Irma L. Perez-Johnson, What Can We Expect Under Personal Reemployment Accounts?
Predictions and Procedures, Mathematica Policy Research, Inc., Jan. 23, 2004. (Hereafter cited as Decker and Perez-
Johnson, What Can We Expect Under Personal Reemployment Accounts?.)
geographic area as well, in terms of both its component uses. For example, the value of the bonus
would be greater to someone living in an area with a lower cost of living, and reemployment
services presumably would be less costly to someone residing in an area with a community 10
Treatment group members who found new full-time jobs within a given period could apply for a 11
bonus. It was thought that the optimal “qualification period” should not be so long that many UI
claimants could obtain bonuses without altering their behavior (i.e., receive bonuses without
increasing their job search effort in order to become reemployed sooner than they otherwise
would have). Alternatively, it was thought that the qualification period should not be so brief that
it discouraged claimants from attempting to intensify their job-seeking activities.
Treatment group members generally received the same bonus amount (e.g., six times their WBA)
regardless of when within the qualification period they obtained new jobs. It was estimated that
the one experimental treatment in Pennsylvania that provided a declining bonus amount over the
qualification period did not significantly hasten an individual’s reemployment. As currently
proposed, bonus amounts would decline over the 13-week qualification period if individuals draw
upon their PRAs. Individuals effectively would control the rate at which their bonuses decreased.
It thus appears that, under the currently proposed design, participants could themselves determine
whether they faced a constant or a declining bonus offer.
UI recipients in the experiments had to remain employed for a prescribed length of time before
receiving bonuses. Its length was chosen to avoid providing a bonus for unstable seasonal work
and to reduce an individual’s inclination to take any job just to qualify for a bonus.
As currently proposed, individuals would get 60% of their bonus amounts immediately upon
reemployment. Those in the state experiments got none of their bonuses until completion of the
reemployment period. The reemployment period was about four months in each experiment,
which is two months shorter than the reemployment period required in H.R. 27 and the PY2004
demonstration to obtain the remaining 40% of the bonus. Immediate receipt of a large share of a
substantial bonus might be a greater inducement to speedy reemployment and acceptance of a
low-paying job compared to the experiments’ designs. The longer reemployment period for
receipt of the remaining portion might be comparatively less of an inducement for remaining in
the new job, particularly if the value of the bonus were fairly small. In other words, the actual
amount of the bonus could influence whether the timing of its receipt makes any difference to the
individual’s job search and retention behavior. The researchers who conducted the PRA
simulation of Georgia UI beneficiaries noted that, while “the reemployment earnings of those
11 Jobless persons on recall status generally could not obtain bonuses if they subsequently got back their previous jobs
at the employers who laid them off. This also typically was the case for jobless persons who obtained employment
through union hiring halls because, in both instances, the bonus offer did not cause them to change their behavior.
Instead of the bonus offer prompting them to accelerate their job search in order to obtain employment sooner than they
otherwise would have, they continued to rely on the actions of others (i.e., former employers and referral unions) to
determine the timing of their reemployment.
offered bonuses [in the experiments] were at least a high as the control groups, [t]he timing of 12
bonus payments under the proposed PRAs might yield a different impact on wages.”
The goals of the experiments were two-fold. One was to reduce the amount of unemployment,
and the other was to reduce the cost of the states’ Unemployment Trust Funds. The eligible
population for participation in the experiments therefore was restricted to UI beneficiaries. The
key outcome measures were the difference between the treatment and control groups in the
duration of insured unemployment and in the amount of UI benefits received.
UI claimants were randomly assigned to one of six treatment groups or to a control group. The
treatments involved variations on the size of the bonus offered and the length of the qualification
period. Although the bonuses shown in Table 1 were calculated as a multiple of an individual’s
UI weekly benefit amount (WBA), the actual offers made to persons assigned to the treatment
groups were expressed as dollar amounts. Qualification period aside, the lowest bonus amount
averaged about $300; the middle bonus amount, about $600; and the highest bonus amount, about
$900. The reemployment period was set at four months.
The evaluators determined that only the highest bonus level of six times a claimant’s WBA
produced a large and statistically significant effect on job search behavior by reducing both the
length of unemployment and the amount of benefits. In 2003, the average WBA in Washington
was about $324. The highest bonus level in current dollars thus would be equal to $1,944 on
average, which is almost two-thirds of the proposed $3,000 maximum for PRAs that are meant to
go toward both reemployment services and bonuses.
The highest bonus level and the long qualification period had the largest impact, namely, a 0.75-
week reduction in unemployment and a $140 reduction in UI benefits. The average response
across all treatment groups was a 0.41-week reduction in the duration of UI payments and a $65
reduction in benefits. The most cost-effective of the approaches also was estimated to be
Treatment 6 (i.e., shifting from lower bonus offers and the short qualification period to the
maximum bonus and qualification period had a statistically significant impact on both outcome
Table 1. Treatments in the Washington Experiment
Bonus amount Shortb Longc
Two times a claimant’s weekly benefit amount (WBA) Treatment group 1 Treatment group 4
12 O’Leary and Eberts, Personal Reemployment Accounts, p. 4.
13 This section is derived from U.S. Department of Labor, The Washington Reemployment Bonus Experiment Final
Report, Unemployment Insurance Occasional Paper 92-6, 1992.
Bonus amount Shortb Longc
Four times a claimant’s WBA Treatment group 2 Treatment group 5
Six times a claimant’s WBA Treatment group 3 Treatment group 6
a. Assuming that individuals’ job-search efforts are associated with their duration of UI benefit entitlement, a
qualification period that is a specified share of duration (as opposed to a specified number of weeks) treats
all claimants the same in a state with varying lengths of benefit entitlement. Claimants’ benefit entitlement in
Washington ranged from 10 weeks to 30 weeks.
b. Twenty percent of a claimant’s benefit entitlement plus the one-week waiting period. The short qualification
period ranged from three weeks to seven weeks; the average was almost six weeks.
c. Forty percent of a claimant’s benefit entitlement plus the one-week waiting period. The long qualification
period ranged from five weeks to 13 weeks; the average was 11 weeks.
Individuals assigned to the treatment groups were found to have exited the unemployment rolls
sooner than control group members during the qualification period. The quicker reemployment of
treatment group members was not associated with their accepting poorer quality jobs as measured
by lower hourly wages or fewer hours worked.
There were limited differences in the job search responses of subgroups within the treatment
groups. For example, older workers (i.e., age 45 or older) with high earnings (i.e., in the top third
of the earnings distribution or at least $17,366) experienced among the largest effects in terms of
shortened unemployment and reduced UI benefits. The evaluators speculated that this might have
been the case because older, high-earning workers could have become discouraged from looking
for work after they had been let go from jobs they had held for many years. These individuals
thus had room to improve upon their job search efforts. For all dislocated workers, however, only
those with the broadest definition used by the evaluators (i.e., individuals continuously employed
for three years prior to filing for UI) and who were offered the largest bonus had outcome
measures that differed substantially from others in the treatment groups.
Similarly, the effects of the experiment on unemployment duration and UI payments differed
according to economic conditions. The evaluators found stronger responses to the bonus offer in
areas within the state with low unemployment rates (i.e., below 5%). Workers in those areas
might have thought that increasing their job search efforts could well elicit more job offers;
workers in areas with higher unemployment rates might have thought that employer demand was
too soft to generate additional opportunities regardless of their job search intensity. This finding
suggests that a bonus offer could have limited impact during a cyclical downturn of the economy
or in an area experiencing long-term decline.
The evaluators looked at whether the experiment conferred net benefits or costs from the
perspective of the UI system, all government, society as a whole, and selected population groups.
It was assumed that in an actual bonus program the bonuses would come from the UI Trust Fund;
therefore, the cost to the UI system was considered to be bonus payments and administrative
expenses while the benefit was considered to be savings in UI payments due to earlier
reemployment. The UI system was estimated to experience net costs because the reduction in 14
benefits was smaller than the bonuses paid. In other words, the experiment did not pay for itself.
14 The administrative expenses of the experiment were quite low at $3 per eligible claimant.
If a regular bonus program were to be implemented, this finding suggests that a source other than
the UI Trust Fund would have to be found to fund it (e.g., WIA).
The experiment was slightly more beneficial to all government than to the UI system because
benefits, in addition to reduced UI payments, included the income tax on earnings that
reemployed workers owed the government. From a societal perspective, the reemployment bonus
was estimated to produce large net benefits because the increase in earnings of those who
responded to the bonus offer far outweighed the small increase in administrative costs associated 15
with a bonus program. The net benefits to society estimated for the experiment could be
overstated if the bonus receipt rate were higher in an established program. (Up to one-third of the 16
individuals in the experiment who were entitled to bonuses did not collect them.) In terms of a
bonus offer to subgroups within the population of UI recipients, the evaluators found that a
program directed at older high-earners had large net benefits for the UI system, government, and
UI claimants were randomly assigned to one of six treatment groups or to a control group. The
treatments involved variations on the size of the bonus offered and the duration of the 18
qualification period, as shown in Table 2. As in the Washington experiment, the bonus offers
made to persons assigned to the treatment groups were expressed as a dollar amount although the
bonuses were calculated as a multiple of each claimant’s WBA. Qualification period aside, the
lower bonus amount averaged $500; the higher, $1,000. The reemployment period was set at 16
Unlike in Washington, all treatments in Pennsylvania included the option of attending a job
search workshop. Treatment 6 did not include access to the workshop in order to isolate its
impact. However, participation in the workshop was so low (2.6%) that it prevented estimation of
its impact and resulted in there actually being five rather than six approaches. The evaluators
speculated that participants lacked interest in attending the workshop because of the state’s low
unemployment rate and to the workshop being offered early in an individual’s jobless spell.
15 Transfer payments (i.e., UI benefits and bonus amounts) cancel each other out when computing costs and benefits to
society as a whole because they represent benefits to some (i.e., UI recipients) and costs to others (i.e., taxpayers).
16 The net benefits of the experiment also could differ from the long-run results of an ongoing program for reasons
other than an increased take-up rate. For example, persons offered bonuses to speed their reemployment might take
jobs that other jobseekers would have taken. Because the former’s reemployment would come at the expense of longer
unemployment among the latter, the net benefits of a bonus program would be lower. In addition, because a bonus offer
could increase the value of filing a claim for UI benefits, it might prompt more UI-eligible workers to register with the
system and thereby increase the number of persons receiving UI benefits and potentially receiving bonus payments.
17 Unless otherwise indicated, this section is derived from U.S. Department of Labor, The Pennsylvania Reemployment
Bonus Experiment Final Report, Unemployment Insurance Occasional Paper 92-1, 1992.
18 Unlike Washington’s variable lengths of UI benefit entitlement, virtually all claimants in Pennsylvania were entitled
to 26 weeks of payments. Thus, fixed qualification periods of six weeks and 12 weeks achieved the homogenous
treatment of persons offered bonuses. (For more information see Walter A. Corson and Robert G. Spiegelman, “Design
of Three Field Experiments,” in Robins and Spiegelman, Reemployment Bonuses in the Unemployment Insurance
Also unlike in Washington, Pennsylvania included a treatment (5) with a declining bonus amount
over the qualification period. It was hypothesized that structuring the bonus in this way would
encourage claimants to become reemployed as early in the longer qualification period as possible.
Table 2. Treatments in the Pennsylvania Experiment
Bonus amount 6 weeks 12 weeks
Three times a claimant’s Weekly Benefit Amount (WBA) Treatment group 1 Treatment group 2
Six times a claimant’s WBA Treatment group 3 Treatment group 4a
Initially six times a claimant’s WBA and then declining — Treatment group 5
Six times a claimant’s WBA and no optional job-search workshop — Treatment group 6b
a. The only difference between Treatment groups 4 and 6 was that the latter did not include the offer of
attending a job-search workshop, which very few participants in the experiment attended.
b. All treatment groups except six were offered the option of attending a job-search workshop.
The evaluators determined that all of the treatments—with the exception of the declining bonus
offer—significantly reduced both the duration of unemployment and the amount of UI benefits.
The highest bonus level and the long qualification period (Treatment 4) had the largest impact,
namely, a 0.8-week reduction in unemployment and a $130 reduction in UI benefits. These are
almost the same magnitudes produced by a very similar approach, Treatment 6, in the Washington
experiment. (The other Pennsylvania treatments yielded a reduction in the duration of UI
payments of 0.5 weeks on average and a reduction in benefits of $80 on average.) Pennsylvania’s
Treatment 4 also was the only design that significantly lowered the share of claimants who
exhausted their UI benefits.
In 2003, the average WBA in Pennsylvania was about $292. Treatment 4, at six times a claimant’s
WBA, would produce a bonus in current dollars equal to $1,752 on average, which is almost 60%
of the proposed $3,000 maximum for PRAs that are intended to be used for both reemployment
services and bonuses.
Individuals assigned to the treatment groups were found to have exited the unemployment rolls
sooner than control group members during the qualification period. As the differential in UI exit
rates tended to be widest right before or at the end of the qualification period, many treatment
group members appear to have escalated their job search activities just before their bonus
eligibility expired. There was no evidence that the comparatively faster reemployment of
treatment group members meant they accepted less desirable jobs in order to get a bonus.
Very little difference was found in the job search response to a bonus offer among subgroups
within the treatment groups. The impact of the treatments on UI receipt was substantial and
significantly greater among claimants from manufacturing industries than from other industries
(by more than 1.5 weeks), however.
The evaluators looked at whether the experiment conferred net benefits or costs from a variety of
perspectives. The findings from the Pennsylvania experiment were similar to those from the
Washington experiment. Despite the bonus offer producing a significant decrease in UI receipt,
the experiment was not cost-effective because the reduction in benefits did not outweigh the cost
to the UI system of paying bonuses. The outcome for the government was somewhat better, with
some treatments producing net benefits as a result of the income taxes on increased earnings from
workers who qualified for the bonus. Society as a whole was estimated to experience net benefits
because the workers’ increased earnings exceeded the program’s low administrative costs.
Researchers employed data from the Washington and Pennsylvania experiments to estimate the
effects of a reemployment bonus focused on UI recipients with high likelihoods of benefit
exhaustion. Models of the probability of benefit exhaustion were developed for the two states.
They were similar to the Worker Profiling and Reemployment Services (WPRS) systems that
states use to identify UI claimants expected to experience long spells of unemployment.
The researchers examined the effects of limiting bonus offers to individuals with varying
probabilities of benefit exhaustion. They found that very narrowly focusing bonuses on those
most likely to exhaust benefits (e.g., claimants in the top 10% of the exhaustion probability
distribution) was not necessarily the most effective approach. Neither the Washington nor
Pennsylvania experiments produced greater reductions in weeks of UI receipt when they were
targeted at the top 50% rather than the top 25% of persons most likely to exhaust their benefits.
The savings in UI benefits were larger for the top 25% or the top 50% of potential exhaustees
than for all bonus experiment participants. Conversely, bonus and administrative costs were
higher in the targeted design than in the non-targeted designs.
From the perspective of the UI system, net benefits of the combined treatments were greater (or 20
net costs smaller) for the targeted design than for the non-targeted designs. Among the various
bonus treatments in Washington and Pennsylvania, the researchers estimated that a low bonus
amount with a long qualification period (e.g., three times the WBA and a 12-week qualification
period) that is focused on the 50% of UI claimants most likely to exhaust benefits was the most
cost-effective treatment. They concluded that while “the original findings from the [non-targeted]
experiments did not generate overwhelming support for reemployment bonuses,” some degree of 21
targeting based on worker profiling “improves the appeal of the reemployment bonus program.”
Another study also utilized data from the Washington and Pennsylvania experiments to, among
other things, estimate the bonus receipt rate and UI impact if the PRA amount is changed. The
researchers found that if UI recipients were offered a $3,000 bonus, 33% would qualify for and
19 This section was derived from Christopher J. O’Leary, Paul T. Decker, and Stephen A. Wandner, “Cost-
Effectiveness of Targeted Reemployment Bonuses,” Journal of Human Resources, winter 2005, vol. 40, no. 1.
(Hereafter cited as O’Leary, Decker, and Wandner, Cost-Effectiveness of Targeted Reemployment Bonuses.)
20 Defined as UI benefit savings plus UI tax contributions on additional earnings minus bonus payment and program
21 O’Leary, Decker, and Wandner, Cost-Effectiveness of Targeted Reemployment Bonuses, p. 278.
22 This section was derived from Decker and Perez-Johnson, What Can We Expect Under Personal Reemployment
receive the first bonus installment; if UI recipients targeted for WPRS were offered a $3,000
bonus, 31% would qualify for and receive the first installment equal to 60% of the account’s
balance. This exceeds the bonus receipt rates of the state experiments not only because of the
PRA’s relatively greater value, but also because reemployment under H.R. 27 triggers immediate
payment of 60% of the account’s balance compared to the experiments’ requirement that
individuals retain their jobs for 16 weeks before they receive bonuses. In contrast, a lower PRA
amount should yield a lower receipt rate because it reduces people’s incentive to try for the
bonus. Thus, a $2,000 bonus could reduce the receipt rate of the first installment to 27% for UI
recipients generally and 30% for those targeted for WPRS. If the bonus were lowered to $1,000—
about the amount (in real terms) offered in the state experiments—the receipt rates (20% and
Similarly, the researchers estimated that the effect on the length of UI receipt would somewhat
exceed the impacts in the state experiments. The latter ranged between -0.26 weeks to -0.82
weeks per person offered bonuses with a qualification period of 11 or 12 weeks. Alternatively, a
$3,000 bonus was found to reduce the length of UI receipt by 1.66 weeks per account recipient;
$2,000, by -1.38 weeks; and $1,000, by -1.09 weeks. The reason for the potentially superior
performance of the PRA vis-a-vis the state experiments is twofold: the PRA amount would be
higher, which should accelerate reemployment and thereby reduce the benefit period; and the
PRA would be focused on those expected to have long durations of UI receipt, which should
increase the average contraction in benefit weeks for persons who find jobs during the bonus
In this earliest experiment, new UI claimants were randomly assigned to one treatment and one
control group. The treatment involved the offer of a $500 bonus for reemployment full-time
within 11 weeks of filing for benefits and after having been in the job for four months. The bonus
amount was equal to about four weeks of UI payments. The 11-week qualification period
represented about 40% of the UI potential benefit entitlement in Illinois’ regular program (i.e., 26
The $500 bonus was estimated to have produced a large and statistically significant effect on job
search behavior by reducing the average length of unemployment by 1.15 weeks. The bonus also
lowered UI benefit receipt significantly: the average benefit of those in the treatment group was
$158 or $194 less than received by those in the control, depending on whether Federal
Supplemental Compensation (FSC) benefits were included.
The FSC program ended during the Illinois experiment, which meant about half the claimants had
26 weeks of UI eligibility under the state’s regular program and about half had 38 weeks because
of the supplemental program. It was determined that the two groups had very different responses
to the bonus offer. The 1.15-week reduction in UI benefit receipt was the product of a 1.78-week
reduction among FSC-eligibles and a 0.71-week reduction among FSC-ineligibles. The response
of the latter group more closely resembles the results from the Washington and Pennsylvania
experiments, which suggests that the Illinois results overstate the reduction in unemployment
23 Unless otherwise indicated, this section was derived from Stephen A. Woodbury and Robert G. Spiegelman,
“Bonuses to Workers and Employers to Reduce Unemployment: Randomized Trials in Illinois,” American Economic
Review, vol. 77, no. 4 (1987).
duration that could be expected from a bonus focused on claimants in states’ regular UI 24
Relatively more individuals assigned to the treatment group than to the control group exited the
UI rolls during the qualification period and relatively fewer treatment than control group
members exhausted their benefits. As in the Pennsylvania and Washington experiments, the
comparatively faster reemployment of treatment group members in Illinois was not accompanied
by lower earnings.
It was estimated that the experiment conferred large net benefits from the perspective of the UI
system. This outcome greatly differs from those of the Washington and Pennsylvania
experiments. The discrepancy was due to the much larger reduction in weeks of UI receipt found
in Illinois. As noted above, however, when Illinois’ treatment group is made more comparable to
the other two states by limiting its claimants to those in the regular UI program, the decrease in
duration of UI receipt more nearly matches the experience in Washington and Pennsylvania.
Linda Levine Ann Lordeman
Specialist in Labor Economics Specialist in Social Policy
email@example.com, 7-7756 firstname.lastname@example.org, 7-2323
24 Paul T. Decker, Christopher J. O’Leary, and Stephen A. Woodbury, “Bonus Impacts on Receipt of Unemployment
Insurance,” in Robins and Spiegelman, Reemployment Bonuses in the Unemployment Insurance System.