Ui Generosity And Job Acceptance Effects Of The 2020 Cares Act
Summary
Federal Reserve Bank of San Francisco Working Paper 2021-13, UI Generosity and Job Acceptance: Effects of the 2020 CARES Act, by Nicolas Petrosky-Nadeau and Robert G. Valletta, dated May 2024. The paper assesses the labor market effects of the CARES Act $600 weekly unemployment insurance supplement using monthly CPS microdata and imputed UI benefits in a difference-in-differences framework. It reports moderate disincentive effects, with estimates implying a 22 percent reduction in job-finding rates for the typical UI recipient. A dynamic model of job acceptance derives a reservation benefit level and suggests that approximately 22 percent of UI recipients would refuse an offer to return to work at their previous pay. The paper closes with a table of reservation benefits, replacement rates and offer rejection rates by age, education and occupation under a stronger recovery scenario.
Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used
Full text
FEDERAL RESERVE BANK OF SAN FRANCISCO
WORKING PAPER SERIES
UI Generosity and Job Acceptance: Effects of the 2020 CARES Act
Nicolas Petrosky-Nadeau and Robert G. Valletta
Federal Reserve Bank of San Francisco
May 2024
Working Paper 2021-13
https://doi.org/10.24148/wp2021-13
Suggested citation:
Petrosky-Nadeau, Nicolas and Robert G. Valletta. 2024. “UI Generosity and Job
Acceptance: Effects of the 2020 CARES Act,” Federal Reserve Bank of San Francisco
Working Paper 2021-13. https://doi.org/10.24148/wp2021-13
The views in this paper are solely the responsibility of the authors and should not be interpreted
as reflecting the views of the Federal Reserve Bank of San Francisco or the Board of Governors
of the Federal Reserve System.
UI Generosity and Job Acceptance:
Effects of the 2020 CARES Act
Nicolas Petrosky-Nadeau Robert G. Valletta
Federal Reserve Bank of San Francisco Federal Reserve Bank of San Francisco
April 12, 2024*
Abstract
We assess labor market effects of the CARES Act $600 weekly UI supplement. We analyze
labor force transitions using monthly CPS microdata and imputed UI benefits. The results
show moderate disincentive effects of the supplement on job finding. We rationalize this result
in a dynamic model of job acceptance decisions that yields a reservation level of UI benefits at
which a recipient is indifferent between unemployment and employment at their prior wage.
Calculations based on the model are consistent with the empirical analysis in regard to the
moderate fraction of UI recipients who were likely to reject job offers.
JEL Classification: J64, J65.
Keywords: Unemployment, unemployment insurance, job acceptance, COVID-19, CARES
Act.
* Petrosky-Nadeau: FRB San Francisco, 101 Market Street, San Francisco CA 94105; e-mail: Nicolas.Petrosky-
Nadeau@sf.frb.org. Valletta: FRB San Francisco, 101 Market Street, San Francisco CA 94105; e-mail:
rob.valletta@sf.frb.org. Olivia Lofton and Mary Yilma provided excellent research assistance. This paper includes a
significant expansion of content released in an earlier working paper by Petrosky-Nadeau (FRBSF Working Paper 2020-
28, August 2020), “Reservation Benefits: Assessing Job Acceptance Impacts of Increased UI Payments." We thank the
editor at JPE Macro, Joseph Vavra, and two anonymous referees for helpful comments. The views expressed are those of
the authors and do not necessarily reflect the views of the Federal Reserve Bank of San Francisco or the Federal Reserve
System.
1 Introduction
The Coronavirus Aid, Relief, and Economic Security (CARES) Act, through the Pandemic Unem-
ployment Compensation (PUC) provision, provided an additional $600 per week to supplement
regular unemployment insurance (UI) benefits. The supplement was available during the initial
outbreak of COVID-19 from late March though the end of July 2020. This historically unprece-
dented increase in the level of UI benefit payments meant that most UI recipients received more
weekly income via UI payments than they earned on their prior jobs–i.e., their UI replacement
rates exceeded 100 percent (Ganong, Noel and Vavra 2020). The enhanced benefits prompted con-
cerns that the labor market recovery from the pandemic would be delayed as many UI recipients
rejected offers to return to work, reflecting the standard moral hazard effect of UI benefits on job
search (Feldstein 1976, Baily 1978, Chetty 2008).
We assess the disincentive effects of expanded pandemic-era UI benefits on job search and
acceptance via two complementary approaches: (i) empirical analyses of observed labor force
transitions; (ii) quantitative assessment of a dynamic model of the job acceptance decision. Despite
the large increase in benefits, results from both approaches suggest only moderate disincentive
effects of the $600 weekly UI supplement on job search and acceptance decisions.
We first conduct direct empirical analyses of labor force transitions using monthly data from
the Bureau of Labor Statistics’ (BLS) Current Population Survey (CPS) combined with imputed UI
benefits. Our direct empirical tests are based on a difference-in-differences regression framework.
We use it to assess whether the change in job-finding rates and other labor market transitions
between the pre-CARES and CARES periods is larger for individuals who have higher UI replace-
ment rates as a result of the supplemental payments. Our value added relative to prior analyses
of the potential disincentive effects of the CARES Act supplemental payments arises from two
specific features of our analyses: (i) we exploit individual variation in UI replacement rates rather
than geographic or solely temporal variation; (ii) we directly assess the individual labor market
transitions, in particular job-finding rates (exits from unemployment to employment), that may
be affected by the moral hazard effect of UI benefit generosity.
Our regression analyses rely on labor market transition data formed using data on individuals
matched across consecutive monthly CPS files. We use data for early- to mid-2020 only, to focus
on the impact of the extra $600/week of UI payments specified by the CARES Act. We estimate UI
replacement rates for individuals in our sample by applying the calculator developed by Ganong,
Noel and Vavra (2020) to annual earnings data from the CPS Annual Social and Economic Supple-
ment (ASEC).
Our results show moderate disincentive effects on job finding from this very large increase
in UI replacement rates: for the typical UI recipient, our estimates imply a 22 percent reduction
in job-finding rates due to the $600/week supplement. As we discuss in more detail in Section
2.3.3, the resulting estimated elasticities of search duration with respect to UI replacement rates
are toward the low end of the range based on earlier research using U.S. data (Schmieder and von
Wachter 2016).
1
Our results also are broadly comparable to the findings from two other recent papers that
examine the effects of the pandemic-period expansions using administrative micro-data from fi-
nancial services companies (Coombs et al. 2022, Ganong et al. 2022); using similar identification
strategies, our estimated elasticity is essentially identical to that reported by Ganong et al. (2022).
By contrast, some other recent studies found little or no effect of the pandemic UI enhancements
on labor market outcomes. Altonji et al. (2020), Bartik et al. (2020), and Finamor and Scott (2021)
found that states with more generous UI systems did not experience weaker labor market re-
bounds during the initial phase of economic recovery from the pandemic. As we discuss in Sec-
tion 2.3.3, it is likely that our somewhat larger estimates of the UI benefit effects on job search
arise because we focus on search responses at the individual level, commonly referred to as micro
effects. By contrast, the studies that find smaller effects likely combine micro responses with off-
setting macro effects on aggregate outcomes. The latter include the aggregate stimulus effects of
UI payments, which help sustain labor demand and hence may offset the job search disincentives
for individual UI recipients (Boone et al. 2021, Kekre 2023). We discuss these issues further in
Section 2.3.3.
To complement and reinforce the regression estimates, our second approach relies on the de-
velopment of a dynamic model of job acceptance decisions that includes heterogeneity in the
underlying wage distribution for job seekers. We use this model to derive the level of benefits
necessary for workers to be indifferent between accepting a job offer at their previous wage and
rejecting it to remain unemployed, taking into account the remaining number of weeks of unem-
ployment benefits available to them. We call this the reservation benefit: a job offer at the previous
wage is accepted if the current level of benefits is below this level. For a given job offer, the level
of the reservation benefit is determined by: (i) the expected duration of the employment spell for
an accepted job – longer lasting jobs have a greater value and are rejected only for commensu-
rately more generous unemployment insurance payments; (ii) the rate of arrival of new job offers
– in a depressed labor market, when job offers are few and far between, any job offer is costly
to refuse, raising the reservation benefit amount, and; (iii) the duration of benefits remaining –
an additional week of benefits raises the opportunity cost of accepting an offer and lowers the
reservation benefit level. In the limit of unbounded UI duration the reservation benefit converges
to the prior wage. Conversely, with one week remaining of UI payments, the reservation benefit
is always above the prior wage, implying that many UI recipients will accept a job even if their
benefits exceed the offered wage.
We apply the reservation benefit concept to the period covered by the provisions in the CARES
Act, including the extension of benefit payments for up to 52 weeks with the Pandemic Emergency
Unemployment Compensation (PEUC) and state emergency extensions. We derive the probability
that individuals within a group of job seekers would reject an offer to return to work. As in
the regression analysis, we use data from the CPS ASEC to impute reservation benefit levels for
workers in different skill (education) groups and occupations. Our quantitative analysis suggests
that a moderate fraction of UI recipients, approximately 22 percent in our preferred specification,
2
would refuse an offer to return to work at their previous pay. This proportion of rejected job offers
is essentially identical to the magnitude of the disincentive effects on job finding rates found in
the analysis of individuals’ transition rates out of unemployment (the 22 percent reduction in
job-finding rates noted above).
These findings imply that the value of a sustained job, especially in a depressed labor market,
usually outweighs the value of the temporary additional UI income. By way of example, a typical
(median) worker earning about $717 per week in their previous job received $959 per week in
UI payments under the CARES act. We calculate that an offer received with 8 weeks of CARES
UI payments remaining would be accepted as long as this worker’s current benefit payment was
below $1419, a reservation benefit level that is around twice the previous wage. Further down
the earnings distribution, however, the value of UI benefits under the CARES Act can dominate
the value of a job at the prior wage, thus exceeding the reservation benefit level for these job
seekers. While our model suggests that only a moderate share of the overall population of job
seekers receiving the $600 UI supplement would reject an offer to return to work during this
period, rejection rates vary substantially across levels of education and broad occupations.
Our findings are broadly consistent with prior research on the effects of UI enhancements dur-
ing past recessions. Most notably, our finding of limited disincentive effects of enhanced UI gen-
erosity during the pandemic recession is consistent with other work that finds substantial cycli-
cality in such effects, with little to no impact when labor market conditions are weak (Kroft and
Notowidigdo, 2016). Similarly, analyses of the impact of the historically large increase in potential
UI benefit duration during the Great Recession found small effects on unemployment exit rates,
with the main impact instead being an increase in labor force attachment. (Rothstein 2011, Farber
and Valletta 2015, Chodorow-Reich, Coglianese and Karabarbounis 2019).1
The framework used to derive the reservation benefit statistic is similar to Mortensen (1977)
in incorporating the realistic feature of UI benefits that are limited in duration. It is also broadly
related to the concept of the after-tax reservation wage in Shimer and Werning (2007), which rep-
resents the take home pay required to make a worker indifferent between working and remaining
unemployed.2 Finally, Boar and Mongey (2020) derive a quantitative framework along similar
lines to our reservation benefits analysis and also find a likely limited impact of temporarily in-
creased UI payments on job acceptance decisions during the pandemic.3 Our analysis does not
address the optimality or welfare effects of the supplemental income under the CARES Act. This
1 See Moffitt (1985) for an early empirical study of the effect of UI benefits on unemployment durations. Lalive,
Landais and Zweimüller (2015) use Austrian data and find that search disincentive effects of UI benefit extensions
are offset somewhat by improved search outcomes for individuals who are not eligible for the extensions. A related
question not addressed here is the impact of UI provisions on the joint behavior of workers and firms, and in particular
on the duration of employment spells (see, for instance, Feldstein 1976 and Baker and Rea 1998).
2 Berg (1990) extends Mortensen’s analysis to a non-stationary environment to study the dynamic evolution of a
worker’s reservation wage as economic conditions evolve (exogenously). Contrary to Mortensen’s reservation wage
our reservation benefit and the reservation wage in Shimer and Werning (2007) are not choice variables affecting the
arrival rate of job offers.
3 Marinescu and Skandalis (2021), using French administrative data, find evidence of declining reservation wages
(measured as a desired target wage) as exhaustion of UI benefit payments nears.
3
is the focus of Mitman and Rabinovich (2021), who argue the $600 supplemental income approx-
imated an optimal UI benefit given the large and transitory nature of the COVID-19 shock to the
labor market.
The rest of this paper is organized as follows. Section 2 describes the empirical design and
provides the results from regression analyses of labor market transitions using matched monthly
CPS data. Section 3 develops a modeling framework that rationalizes our empirical results. Sec-
tion 3.1 describes the decision problem and derives a reservation benefit as a function of the state
of the labor market, the wage offer, and the number of weeks of UI payments remaining. Section
3.2 adapts the reservation benefit statistic to the details of the CARES Act and uses CPS data to
calculate benefit amounts for different categories of workers. Section 4 concludes.
2 CARES Act UI expansion and labor market transitions
We begin with a direct empirical assessment of the effects of the increase in UI benefit payments
during the pandemic on job-finding rates and other labor market flows. We rely on a before/after
regression framework to assess whether the change in job-finding rates and other labor market
transitions between the pre-CARES and CARES periods is larger for individuals who receive the
largest UI replacement rates due to the supplemental payments. As described in Section 1, our
value added relative to prior analyses of the pandemic UI supplements arises from our reliance
on individual variation in UI replacement rates and direct measurement of job-finding rates.4
Our regression analyses rely on labor market transition data formed using data on individuals
matched across consecutive monthly CPS files. We use data for early- to mid-2020 only, to focus
on the impact of the extra $600/week of UI payments specified by the CARES Act and available
from late March through the end of July. We combine the monthly CPS data with estimated UI
replacement rates formed using the calculator developed by Ganong, Noel and Vavra (2020). Our
specific calculations rely on annual earnings data from the CPSASEC for the individuals observed
in our matched monthly CPS data. We discuss these steps in detail in the next two sub-sections,
including a discussion of the distribution of replacement rates in our sample.
4 By comparison, Bartik et al. (2020) rely on state-level variation in median replacement rates and employ-
ment/hours, and Altonji et al. (2020) and Finamor and Scott (2021) examine labor market status but not flows between
labor market states. These papers reported little or no disincentive effects of the enhanced UI payment generosity on
employment status. In research conducted in parallel with ours, Coombs et al. (2022) and Ganong et al. (2022) used
administrative data from financial services companies and found effects of the CARES Act UI benefit increases on job
finding that are similar to or smaller than our estimates. We discuss the comparison of our results to these studies in
more detail in Section 2.3.3. In addition, Marinescu, Skandalis and Zhao (2021) examined job applications in local labor
markets during the pandemic and found moderate reductions in application rates in areas with greater increases in UI
benefit amounts.
4
2.1 Matched CPS data on labor market flows
We use matched monthly data on individual labor force participants from the CPS (age 16 and
over).5 Because our empirical strategy requires linking monthly CPS files to annual earnings data
from the CPS ASEC (see the next sub-section), our matched observations are limited to the months
of February through July of 2020. This timeframe is narrow but enables us to focus on the period
when the $600 supplement was in place (and the preceding two months of 2020, which are used
as a pre-treatment comparison period).6
Due to the rotating sampling scheme used for the CPS, surveyed households and individuals
are in the sample for two separate periods of 4 consecutive months (with an intervening 8-month
period spent out of the sample). This enables consecutive month-to-month matching for about
70 percent of the sample.7 The monthly match is based on household identifiers, which we vali-
date by ensuring that the reported data on age, education, race, and gender do not conflict across
matched observations. We identify labor market transitions by comparing an individual’s labor
force status in consecutive months. We focus primarily on transitions out of unemployment (U),
to employment (E) or out of the labor force (N), denoting them as UE and UN transitions respec-
tively. Given relaxed job search requirements under the CARES Act UI expansions, the behavior
of individuals not actively searching for work may have been affected by the UI supplements. We
therefore also examine transitions from out of the labor force to employment (NE).
A well-known concern regarding matched CPS data is the likelihood of spurious transitions in
labor force status arising from inconsistent or error-ridden survey responses rather than meaning-
ful changes (Abowd and Zellner 1985, Poterba and Summers 1986, 1995). Such spurious transi-
tions could impart a downward bias to the estimated effects of UI payments on labor force transi-
tions and reduce the precision of the estimates. We therefore follow past research by adjusting the
data to minimize the incidence of spurious transitions (Rothstein 2011, Valletta 2014, Farber and
Valletta 2015, Farber, Rothstein and Valletta 2015). In particular, for individuals identified as leav-
ing unemployment one month, either through job finding or labor force exit, and then returning
to unemployment the next month, we recode their records to show no transition (and retain the
newly created observations). We refer to these as “two-month matches,” although the resulting
transitions are still measured on a consecutive monthly basis. This adjustment requires restriction
of the final analysis sample to individuals who are observed to be in their first or second month
of a consecutive four-month span in the sample, thereby reducing the matched sample count by
approximately one-third and eliminating July 2020 observations from our analyses.8
5 See Valletta (2014) for more details on construction of a similar sample for an earlier timeframe (in particular, Table
2 and the associated discussion in that paper). We exclude individuals who identify as serving in the armed forces.
6 The ASEC is administered primarily in March, although some CPS respondents receive the supplement in other
months. With the 4-month rotation in the monthly CPS, this enables us to use observations with ASEC information for
the months of January through July of 2020. Forward matching yields observations on labor force transitions starting
in February 2020.
7 Most of the non-matched observations are from the “outgoing rotation groups” that are exiting the sample for eight
months or permanently (one quarter of each monthly sample). In addition, a modest fraction of observations is lost
because respondent households that move to different geographic locations are not followed.
8 The adjustment reduces the monthly incidence of transitions out of unemployment by about 5 percentage points
5
The results for unemployment exits reported in subsequent sections generally are based on
these adjusted transitions, although we also provide some comparison to specifications that do
not make this adjustment. We do not apply this adjustment to our analysis of transitions from
out of labor force to employed (NE), because the measurement distortion generally applies to
transitions in and out of unemployment. As we will see in the results below, the adjustment
for unemployment exits makes a substantial difference for the key results. This likely reflects
the turbulence in labor market transitions and measurement during the time period used for our
estimates, which corresponds to the early phase of the pandemic. Adjusting for short-term or
spurious transitions is especially important in such circumstances.
2.2 UI replacement rates from CPS ASEC data
Our analysis relies on UI replacement rates calculated at the individual level, defined as the ra-
tio of weekly UI payments to weekly earnings prior to the job loss that resulted in the UI claim.
As discussed in Ganong, Noel and Vavra (2020), median UI replacement rates across all eligible
workers typically are slightly below 0.5 in the United States (50 percent of prior earnings), absent
benefit supplements. They estimated that the $600 CARES Act UI supplement raised the typical
replacement rate substantially, to a median value of 1.34, implying that the majority of UI recip-
ients were eligible for UI payments that exceeded their prior weekly earnings. As part of their
research, Ganong et al. constructed a calculator for replacement rates based on individuals’ recent
prior earnings history, which they have made publicly available.9
We use the Ganong et al. calculator to form estimated UI replacement rates for the individuals
in our data. Precise measurement requires individual employment and earnings data from prior
quarters. We therefore restrict our matched monthly CPS sample to individuals who are included
in the 2020 CPS ASEC sample. As noted above, this limits the sample to the months of January
through July 2020. The ASEC includes information on weeks worked, hours, and earnings in the
prior calendar year (2019 in this case, which largely contains the qualifying earnings period for
potential UI recipients in our sample from early 2020).10 Because no information is provided on
the timing of employment and earnings across the four quarters of the year, we spread them out
evenly across all four quarters for the purposes of applying the UI benefits calculator.11
One notable feature of the distribution of UI replacement rates is that because normal UI pay-
ments generally are determined as a fraction of prior earnings, the uniform $600 supplement in-
creased replacement rates more for individuals with low versus high prior earnings. To illustrate
this, we divided our sample of unemployed individuals during the months when the $600 supple-
ment was available (April-July) into quintiles based on weekly earnings. The median replacement
on average (Valletta 2014).
9 https://github.com/PSLmodels/ui_calculator
10 Ganong et al. used pre-pandemic labor market data and 2018 as their base earnings year. Our use of observations
on actual unemployed individuals in early 2020 combined with their 2019 earnings should yield relatively accurate
measurement of UI replacement rates in our sample.
11 The rules specifying which prior earnings quarters are used to determine UI eligibility and weekly payments vary
across states.
6
rate ranged from about 2.5 in the lowest quintile down to about 0.6 in the highest quintile. This
will also be reflected in the distribution of replacement rates across industries and occupations,
given variation in typical skill levels and hence earnings across sectors.
This variation in UI replacement rates due to the $600 supplement raises potential concern
about the identifying information that we use to estimate their effects on job-search behavior. In
particular, job losses early in the pandemic were heavily weighted toward low-wage sectors and
workers, notably individuals in high-contact services jobs concentrated in the retail, leisure and
hospitality, and personal services sectors. Activity in these sectors remained disrupted well into
the pandemic, curtailing job prospects for individuals laid off from these sectors. This correspon-
dence between UI replacement rates and sectoral disruptions during the early pandemic period
raises potential concern that our estimates of UI effects may be contaminated by sector-specific
labor market conditions.
In our subsequent empirical analysis, we address this concern about possibly confounding
effects from the sectoral pattern of labor market disruptions during the early pandemic period.
To preview, Figure 1 illustrates the distribution of replacement rates across our complete sample
of unemployed individuals and also within major industries and occupations (calculated for the
months of April-July 2020, including the $600 supplement). The median replacement rate for the
full sample is 1.40, slightly higher than the estimate of 1.34 from Ganong, Noel and Vavra (2020).12
The figures show that median and mean replacement rates vary notably across industries and
occupations, with higher replacement rates evident in low-wage sectors such as the leisure and
hospitality industry and also for services occupations. However, replacement rates vary more
within than between sectors, as reflected in standard deviation spreads within sectors that gen-
erally extend to or beyond the range of means and medians across sectors. This suggests that
variation in replacement rates is not closely related to sector-specific effects of the pandemic. We
explore this issue further when we present our regression results in the next section.
2.3 Regression specification and results
We begin our analyses with a conventional before/after regression design, essentially a differences-
in-differences analysis that yields an estimate of the average treatment effect of variation in UI
replacement rates. We include imputed UI replacement rates using the procedure described in the
preceding section. Our specific regressions take the form:
Pr (Yit = 1) = δRi + (π × Ri × ( Apr − July)) + γt + ϕs + βXi,t−1 + λZs,t−1 (1)
In this equation, the dependent variable Yit is an indicator for whether an individual i tran-
12 Our higher estimated replacement rates are as expected: Ganong, Noel and Vavra (2020) noted that replacement
rates estimated from individuals unemployed during the early pandemic period are likely to exceed their estimates
based on pre-pandemic data, given the tilt toward low-wage individuals among job losers during this period. Our
estimates are very similar between the full sample of unemployed individuals and our restricted two-month match
sample.
7
(a) UI replacement rate, by major industry
(b) UI replacement rate, by major occupation
Figure 1: UI replacement rates with CARES Act $600 supplement
Notes: Calculated from authors’ CPS monthly-ASEC match, using the UI benefits calculator from Ganong,
Noel and Vavra (2020). Vertical bars extend one standard deviation above and below the mean.
8
sitions between the specified labor market states across consecutive months (observed in month
t, based on status in months t and t − 1). We focus primarily on job finding rates from unem-
ployment (UE transitions) but also examine transitions that involve labor force exit or entry (UN
and NE transitions), as discussed more below. Our preferred estimates rely on our two-month
matched CPS data for the reasons described in Section 2.1, although we also examine results from
the single-month matched sample. The underlying sample contains observations for transitions
observed in the months of February through July of 2020, although the two-month match estima-
tion samples end in June.
The key explanatory variables are the individual’s imputed UI replacement rate (Ri ) under the
CARES Act and its interaction with an indicator for observations corresponding to the months of
April through July, when the CARES Act $600 supplement was available (with estimated coeffi-
cients δ and π). The replacement rate with the $600 supplement included varies across individuals
but not over time and hence is not the key source of variation in this equation. Instead, the treat-
ment effect of the $600 CARES supplement is captured by the impact of the replacement rate after
the CARES Act was implemented and the supplemental payments were available. This period
began in late March 2020, between the March and April CPS reference periods. These effects are
estimated by the coefficient on the interaction between the replacement rate Ri and an indicator
for observations in the months of April through July (or June for the two-month matched sample).
This represents a conventional before/after estimation approach with regression controls (with
the months of February and March combined used as the baseline control period).
We also estimate an expanded equation that allows the effects of the CARES Act supplement
to vary across the months during which it was active:
Pr (Yit = 1) = δ1 Ri + (δ2 × Ri × Mar ) + (π1,2,3,4 × Ri × ( Apr, May, Jun, Jul ))
+γt + ϕs + βXi,t−1 + λZs,t−1 (2)
In this expanded equation, we separately identify each month in the sample, which enables
us to examine the time pattern in the effects of the $600 supplement. The months of February
and March are once again used as baseline control periods. The February control period effect is
identified as the omitted category in the regression (the first term on the right-hand side), and the
March control period is separately identified (via the second term on the right hand side). The
effects of the UI replacement rate (Ri ) for each month when the supplement was available are
represented by the coefficients π1,2,3,4 . As above, the sample period ends in June rather than July
when we use our two-month matched sample.
The regression specifications also include indicators for calendar months (γt ) and state of resi-
dence (ϕs ). In addition, the vector Xi,t−1 consists of individual-level controls observed in the base
(pre-transition) month: age (eight categories), education (five categories), race/ethnicity (five cat-
egories), gender by marital status, broad occupation (10 categories) and industry (13 categories)
of prior employment, and duration to date of the individual’s unemployment spell (10 categories,
9
with the final category indicating duration of longer than one year).13 The model also includes
several state/month labor market controls (Zst ): cubics in the state unemployment rate and the
three-month employment growth rate.
Estimation is via a logit model, with reported parameter estimates converted into average
marginal effects. All estimates are weighted by the longitudinal weights that adjust the sample
for the characteristics of the sequentially matched observations.14 The analysis is restricted to
individuals with non-zero estimated UI replacement rates under the CARES Act—i.e., individuals
who are identified as eligible to receive UI payments based on their prior earnings history. This is
a direct implication of our before/after design, since individuals who are not eligible to receive UI
payments do not contribute any identifying variation to the estimation.
2.3.1 Main results
The results for the before/after regression specification for unemployment exits and other labor
force transitions, based on equation 1 in the preceding section, are shown in Table 1 (estimated
coefficients, with robust standard errors in parentheses below them).
Results from the preferred specification for unemployment exits are reported in the first col-
umn; this specification relies on the two-month match that corrects for temporary exits from unem-
ployment. The estimated effect of the UI replacement rate measure during the combined months
of April through June when the $600 supplement was available is negative and very precise in the
first column, attaining significance at better than the 1 percent level.
The second column of Table 1 shows results from the alternative specification for job-finding
rates (UE transitions), with the two-month match restriction removed: all consecutive monthly
transitions are included and no correction is made for reported temporary exits from unemploy-
ment. This enables use of observed transitions through July. As expected, the prevailing labor
market turbulence during our sample frame appears to introduce noise in the measurement of
monthly transition rates: the estimated interaction coefficient is reduced somewhat in size, al-
though it remains statistically significant at nearly the 1 percent level.
The estimated effects of UI replacement rates during the CARES Act period in columns 1 and 2
of Table 1 are economically meaningful, implying moderate disincentive effects of the enhanced UI
payments on job acceptance decisions. In particular, because they are stated as average marginal
effects, the interaction coefficients imply a reduction in job-finding rates ranging from 7.2 per-
centage points (column 1) down to 4.9 percentage points (column 2) when UI replacement rates
increase from 0.5 to 1.5 (for example). We discuss interpretation of these magnitudes in more detail
in in Section 2.3.3 below, where we compare our estimates to findings from the existing literature
on UI benefit levels and job search.15
13 For regressions in which the initial state is out of the labor force, the unemployment duration, industry, and occu-
pation variables are excluded.
14 The regression results are highly robust to use of different survey weights.
15 The disincentive effects of UI generosity on job-finding might vary depending on the duration of unemployment.
Given the massive job loss early in the pandemic, the typical duration of an in-progress unemployment spell in our
10
Table 1: Regression results: UI replacement rates and labor force transitions, before/after
design
(1) (2) (3) (4)
UE (2-month UE (1-month UN (2-month NE (1-month
match) match) match) match)
UI rep rate 0.050** 0.026 0.004 -0.001
(0.022) (0.019) (0.016) (0.007)
UI rep rate*CARES months -0.073*** -0.049** 0.013 -0.013
(0.027) (0.022) (0.016) (0.010)
Observations 2860 5449 2768 7124
*** p<0.01 ** p<0.05, * p<0.1 Notes: Logit regression model results (average marginal effects, with
robust standard errors in parentheses) from matched CPS micro-data, Feb.-Jul. 2020, combined with
2020 CPS ASEC data to form individual UI replacment rates (including the $600 supplement from
the CARES Act). Regression controls include: age (eight categories), education (five categories),
race/ethnicity (five categories), gender by marital status, broad occupation (10 categories) and in-
dustry (13 categories) of prior employment, and duration to date of the individual’s unemployment
spell (10 categories, with the final category indicating duration of longer than one year); state/month
economic conditions (cubics in the unemployment rate and log 3-month employment growth); and
complete vectors of calendar month and state dummies. The duration, occupation, and industry con-
trols are excluded from column 4.
One notable element of the results is that the pre-CARES (February and March) exit rates
are higher for individuals with the highest UI replacement rates under the CARES Act enhance-
ments, with a statistically significant estimate evident for this baseline effect in the first column.
This higher unemployment exit rate for individuals who will later receive high UI replacement
rates suggests possible violation of the conventional parallel trends assumption for the validity
of difference-in-differences estimates. While we cannot reject this interpretation, we view this
baseline difference as reflecting systematic unobserved differences between individuals with high
and low replacement rates–e.g., it is likely that individuals with high replacement rates under the
CARES Act supplement were employed prior to the pandemic in low-wage labor markets with
high turnover and job-finding rates. Moreover, this baseline difference is not evident in the second
column, which is based on the full set of monthly matches.
We also examined whether UI generosity affects exits from unemployment to out of labor force
(UN), with the results displayed in column 3 of the table.16 This follows earlier empirical results
suggesting that UI benefits may increase labor force attachment, because active job search gen-
erally is a requirement for UI eligibility in the United States (e.g., Farber, Rothstein and Valletta
2015, Card, Chetty and Weber 2007). The results in column 3 show no meaningful effect of UI
sample is quite short: the mean and median duration are about 8.7 and 5 weeks, and only about 10 percent of the
sample has been unemployed for more than 6 months (27 weeks or more). This short duration distribution precludes
reliable estimation of our model for individuals with prolonged spells of unemployment. Estimates with the sample
divided between those above and below the mean duration suggests somewhat larger and more consistent effects for
those with shorter durations.
16 Relative to the column 2 sample, a small number of observations are lost due to exact collinearity in the column 3
regression.
11
Figure 2: Job finding rates (from unemployment), by UI replacement rates (pre/post CARES Act)
Notes: Calculated from logit regression results (column 1 of Table 2).
replacement rates on reported labor force exits from unemployment. This contrasts with the ear-
lier empirical findings of enhanced labor force attachment due to extended UI durations, likely in
part because the job search requirements for UI eligibility were relaxed during the initial phase of
the COVID-19 pandemic in the first half of 2020. Finally, column 4 presents results for job-finding
rates from out of the labor force (NE). We include this analysis because the relaxation of job-search
requirements implies that the $600 supplement may have altered the job search and acceptance
decisions of individuals who self-report as not actively searching and hence out of the labor force.
However, the results provide no evidence that these transition rates were affected by the increase
in UI generosity due to the CARES Act.
The regression results based on equation 2 in section 2.3, which allows the effects of the UI re-
placement rates to vary across months, are displayed in Table 2. In this specification, the omitted
month is February, so the UI replacement rate variable and its interaction with the month indica-
tor for March both reflect the pre-CARES baseline comparison period. The interaction effects for
subsequent months represent the impact of the higher replacement rates generated by the extra
$600/week UI benefits available through the CARES Act. The table is otherwise structured iden-
tically to Table 1, with the results for our preferred two-month match specification reported in the
first column.
The results show that the estimated negative effect of UI benefit generosity from column 1
of table 1 vary somewhat across the months when the $600 supplement was available. The es-
timates are the same size and attain similar statistical significance for April and May, while the
estimate for June is slightly smaller and much less precise. In other words, the estimated effects
are largest early in the implementation period and then decline somewhat over time, although
the monthly estimates are too imprecise for the differences between them to be statistically signif-
icant. Declining effects of the $600 supplemental payments as their expiration date approaches is
one implication of our model of reservation benefits discussed subsequently, in Section 3.
Figure 2 shows the time pattern of UI generosity effects on job-finding rates based on the
12
Table 2: Regression results: UI replacement rates and labor force transitions, monthly effects
(1) (2) (3) (4)
UE (2-month UE (1-month UN (2-month NE (1-month
match) match) match) match)
UI rep rate 0.057* 0.022 0.016 0.002
(0.033) (0.026) (0.015) (0.010)
UI rep*Mar -0.013 0.007 -0.029 -0.005
(0.043) (0.037) (0.034) (0.013)
UI rep*Apr -0.080* -0.025 0.003 -0.011
(0.043) (0.033) (0.018) (0.017)
UI rep*May -0.080** -0.046 -0.001 -0.021
(0.038) (0.030) (0.017) (0.013)
UI rep*June -0.069 -0.064** 0.011 -0.020
(0.046) (0.032) (0.028) (0.019)
UI rep*July – -0.004 – 0.008
(0.046) (0.037)
Observations 2860 5449 2768 7124
*** p<0.01 ** p<0.05, * p<0.1 Notes: See Table 1 notes. Pre/post-CARES period replaced by individual
month indicators.
column 1 results from Table 2, comparing exit rates for individuals at the pre- and post-CARES
Act average levels of replacement rates (holding both groups characteristics at the full sample
averages). A drop in relative job-finding rates for those with higher replacement rates is evident
in April. In subsequent months, job-finding rates increase for both groups, but the job-finding
rates for those with higher post-CARES replacement rates remain somewhat lower than for those
with lower replacement rates.
The second column of Table 2 show results for the alternative specification for job-finding rates,
similar to the same column in Table 1, with the two-month match restriction removed. As in Table
1, this reduces the magnitude and precision of the estimated coefficients. Only the June estimate
is statistically significant at the conventional 5 percent level (despite the reduced standard errors
afforded by the larger sample size compared with the first column).
Finally, results for the UN and NE transitions in columns (3) and (4) confirm no effect of UI
replacement rates on these transitions during any of the months of our sample.
2.3.2 Sensitivity to industry and occupation
As noted in Section 2.2, the concentration of early pandemic job losses among low-wage services
sectors raises potential concern that our estimates of UI benefit effects may be contaminated by
sector-specific labor market conditions. More precisely, rather than solely reflecting effects of ex-
panded UI generosity, our estimates may instead reflect reduced job prospects for workers pre-
viously employed in sectors in which activity was directly constrained by pandemic effects. As
already illustrated in that earlier section, the wide distribution of UI replacement rates across and
13
within occupations and industries in our sample partly alleviates this concern. This concern is also
partly addressed by the inclusion of broad occupation and industry dummies in the regressions
reported in the preceding section.
We further explore the sensitivity of our results to sector-specific effects by restricting the sam-
ple used for our regressions based on occupation and industry. The small sample sizes for indi-
vidual industry and occupation groups preclude reliable estimation of UI replacement rate effects
within those groups.17 However, we can explore whether our main results are driven by key
sectors by sequentially excluding each occupation and industry from the estimation sample.
The results are shown in panels A and B of Tables 3. For straightforward interpretation, we
once again rely on the simplified “before/after” design, as in Table 1 above. The specification
is identical to that from column 1 of Table 1, but with the indicated occupations and industries
sequentially excluded from the full sample. Across all columns of both panels, the estimated
effect of the UI replacement rate during the CARES Act $600 supplement period (April-July) is
tightly distributed around the full sample estimate of -0.072 (from column 1 of Table 1) and in
all cases is precisely estimated. One notable and perhaps surprising exception is the leisure and
hospitality industry (column 11 of Table 3, panel A). When this industry is excluded, the estimated
effect of UI replacement rates under the CARES Act rises substantially, suggesting that the effect is
small in this sector. Overall, these results bolster the case for interpreting our findings as reflecting
variation in the UI replacement rates associated with the $600 supplement rather than sector-
specific differences in job prospects for individuals who lost jobs early in the pandemic.
17 When estimated for subsets of grouped occupations and industries, the estimated effects of the UI replacement rate
in our before/after design varied widely across sectors and generally were statistically imprecise.
14
Table 3: Regression results: UI replacement rates and job finding (UE rates, two-month match), before/after design, occupations and industries
eliminated
Panel A: occupations eliminated
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Farm/ Const/ Inst/ Trans/
Manag Prof/Tech Services Sales Admin Prod.
Fish Extract Maint/Rep Materials
UI rep rate 0.057** 0.040 0.049* 0.045* 0.062** 0.050** 0.045* 0.054** 0.050** 0.048**
(0.024) (0.025) (0.026) (0.024) (0.026) (0.023) (0.024) (0.023) (0.022) (0.023)
UI rep* -0.074*** -0.074** -0.073** -0.070** -0.076*** -0.072*** -0.062** -0.074*** -0.068** -0.074***
CARES months (0.028) (0.030) (0.029) (0.029) (0.029) (0.027) (0.027) (0.027) (0.026) (0.028)
Observations 2568 2409 2191 2547 2611 2833 2582 2774 2645 2580
Panel B: industries eliminated
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
15 Trans/ Prof/ Educ/
Agric Mining Const Manuf Whl/Ret Info Financial
Util Bus Health
UI rep rate 0.050** 0.051** 0.043* 0.059** 0.059** 0.054** 0.051** 0.045** 0.051** 0.048*
(0.023) (0.023) (0.023) (0.024) (0.024) (0.023) (0.023) (0.023) (0.025) (0.025)
UI rep* -0.072*** -0.071*** -0.058** -0.076*** -0.079*** -0.076*** -0.073*** -0.067** -0.067** -0.078**
CARES months (0.027) (0.027) (0.027) (0.028) (0.029) (0.027) (0.027) (0.027) (0.029) (0.031)
Observations 2825 2828 2547 2548 2487 2730 2817 2746 2567 2362
(11) (12) (13)
Leis/Hosp Oth Serv Gov
UI rep rate 0.046* 0.048** 0.051**
(0.026) (0.022) (0.023)
UI rep* -0.091*** -0.062** -0.072***
CARES months (0.029) (0.026) (0.027)
Observations 2336 2699 2811
Notes: *** p<0.01 ** p<0.05, * p<0.1 Notes: Occupations and industries excluded as indicated in column titles. Specification is from column 1 of Table 2.
2.3.3 Assessing the magnitude of the UI effect on job finding
We assess the magnitude of the estimated impact of the CARES supplement based on the results
from the first column of Table 1, which applies the before-after design to our two-month match
specification. We focus on this specification because the two-month match correction for spuri-
ous transitions bolsters the precision of the results, and the before/after framework provides a
straightforward averaging of the UI replacement rate effects across the months when the $600
supplement was available.
Given the wide span of post-CARES UI replacement rates observed in our data, various met-
rics could be used to interpret the size of the estimated effect.18 Interpretation of the coefficients
is straightforward, however: the replacement rate is measured relative to a value of 1.0 (UI pay-
ments equal to prior earnings), and the coefficients are average marginal effects that represent the
effect of an increase in the UI replacement rate of 1 (100 percentage points) on the probability of
observing the relevant transition.
We conduct straightforward calculations based on these considerations. The $600/week addi-
tional payments raised the median replacement rate from 0.5 to 1.39 in our two-month matched
sample of unemployed individuals.19 This represents an increase in the typical replacement rate of
0.89. Based on the coefficient on the UI replacement rate interaction with the CARES Act months
in column 1 of Table 1, this implies that the job-finding rate for the typical recipient of enhanced
UI benefits during those months was reduced by about 6.5 percentage points, or 0.065.20 This is of
moderate size relative to job-finding rates averaging about 0.29 during the months when the $600
supplement was available. Specifically, this represents a 22 percent reduction in job-finding rates
due to the availability of the $600 weekly UI benefit supplement.
We can also translate our estimate into an elasticity of job-finding with respect to variation in
UI replacement rates. This enables us to put our findings further into context and compare them
directly with results from other research. We use mean rather than median changes for consistency
with conventional elasticity calculations. The $600 supplement raised average UI replacement
rates from 0.44 to 1.53 in our sample from column 1 of Table 1 (a 245 percent increase). This
increase of 1.09 in the average replacement rate lowers job-finding rates by 7.8 percentage points
(1.09*(-0.073)=-0.080), or 28 percent relative to a base rate of about 0.29. The calculated job-finding
elasticity is -0.11, with a standard error of 0.042 and a 95 percent confidence interval of -0.045 to
-0.18.21
18 As noted earlier, the $600 supplement substantially raised the typical replacement rates. It also widened the dis-
persion substantially, with the standard deviation of replacement rates across UI-eligible individuals rising by nearly a
factor of seven.
19 We limit this calculation to individuals observed during the period when the supplement was available, from April
through July. The median replacement rate of 1.39 in this two-month matched sample is very close to the value of 1.40
for the full sample of unemployed individuals discussed in section 2.2.
20 The specific calculation is 0.89;(-0.073)=-0.065.
21 The exact calculation is is [(-0.078)/(0.286)]/[(1.090)/(0.444)]=-0.111. The standard error and confidence intervals
for this point estimate are obtained via a bootstrap, using 500 iterations of random re-sampling with replacement from
the regression sample. Our estimated elasticity range is somewhat lower if we instead use the interaction coefficient
from the sample of one-month transitions in column 2 of Table 1 (point estimate of -0.077, with a standard error of 0.031
16
This estimated elasticity of -0.11 and corresponding 95 percent confidence interval are at the
low end of the range of elasticity estimates using U.S. data summarized in Schmieder and von
Wachter (2016). In particular, the low end of the range summarized in their Table 2 is -0.10 to
-0.15, with about half of the reported estimates exceeding -0.5.22
Our estimated elasticity is close to two other careful estimates of pandemic UI effects, both
using very different data than ours (administrative microdata from financial services companies,
from two separate sources). Ganong et al. (2022) rely on two identification strategies, including
one that is very similar to ours, based on a difference-in-differences framework using individual
variation in UI replacement rates. With this approach they obtain an estimate of the elasticity of
the UI benefit effect on job finding that is essentially identical to ours, -0.11.23 Coombs et al. (2022)
obtain a somewhat higher estimated elasticity range of -0.13 to -0.22; this is largely within the
confidence interval for our estimate and like ours is toward the lower end of the range from past
estimates (Schmieder and von Wachter 2016).
As noted in Section 1, other studies found much smaller labor market effects of the $600/week
UI benefit enhancement. In particular, Bartik et al. (2020), Altonji et al. (2020), and Finamor and
Scott (2021) found that states that experienced the largest increases in UI payments due to the
$600/week supplement did not experience weaker labor market rebounds during the initial phase
of economic recovery from the pandemic.
It is likely that our somewhat larger estimates of the pandemic UI benefit effects on job finding,
along with those in Ganong et al. (2022) and Coombs et al. (2022), arise because we focus on
search responses at the individual level, commonly referred to as micro effects. Our regressions
rely on variation in UI replacement rates across individuals before and during the periods when
the CARES Act $600 supplement was available, with no direct channel for broader macro effects
to influence the estimates. Reinforcing this point, although we include measures of state labor
market conditions in our regressions, our results for the UI replacement rates are largely invariant
to their inclusion or exclusion.
By contrast, the studies that find smaller effects likely combine responses along the micro or
individual margin with more general macro effects on aggregate outcomes. These macro effects
can take various forms. The most important one in regard to the pandemic UI payments is their
aggregate stimulus effects via consumption spending. Recent research has found these stimulus
effects to be substantial during the pandemic, with a large marginal propensity to consume out of
the UI benefits paid (Ganong et al. 2022). This added household spending likely helped sustain
labor demand and hence offset the job search disincentives for individual UI recipients (Boone
et al. 2021, Kekre 2023). Analyses that rely on outcomes aggregated above the individual level
(i.e., state and local labor markets) are likely to capture a combination of the micro responses at
and 95 percent confidence interval of -0.015 to -0.117)
22 Schmieder and von Wachter (2016) focus on the elasticity of search duration, which takes on positive values, in
their literature summary. Under the assumption of constant exit rates from unemployment, the elasticities of search
duration and job-finding probabilities are nearly identical in absolute value.
23 Ganong et al. (2022) report their estimated elasticities in duration rather than transition terms, hence with opposite
sign to ours.
17
the individual level and the offsetting stimulus effects at the macro level. Our estimates instead
focus on the narrow micro or individual responses.”4
3. Ulincome and job acceptance decisions: a reservation benefits frame-
work
Our direct assessment of individual job-finding rates in the preceding section uncovered mod-
erate disincentive effects of the $600 weekly UI supplement. In order to further understand this
result, this section develops a dynamic framework of job acceptance decisions and applies it to
the specific features of the CARES Act UI benefits expansion. Importantly, our modeling frame-
work incorporates heterogeneity in the distribution of prior earnings within groups of job seekers
(defined primarily by educational attainment or prior occupation). The model sheds light on why
many UI recipients will accept job offers even when their benefits exceed the offered wage. As in
the preceding section, we use data from the CPS ASEC to infer the underlying wage distribution
for job seekers and CARES Act replacement rates. Our quantitative analysis suggests that a mod-
erate fraction of UI recipients would have refused an offer to return to work at their previous pay
during the period when the $600 weekly supplement was available.
3.1 Reservation benefits
We study the problem of a risk neutral insured job seeker considering a job offer to return to work
at the previous wage, wj;, characterizing the level of UI benefits that leave a job seeker indifferent
between accepting and rejecting the offer.? The framework highlights the circumstances under
which a job seeker may accept a job offer for a wage below the value of their current weekly UI
payments. Finally, note that the reservation benefit statistic developed here does not take into
account risk aversion, which would increase the value of a long stream of earned income on the
job compared with temporary UI payments.
Consider a worker comparing the present value of the job, We(w;) to that of remaining unem-
ployed with UI benefits b;, which depend on prior earnings w;, and ¢ remaining weeks of eligibil-
ity, Wy(b;,t). The decision takes into account the likely duration of the job and that of finding an
alternative offer — through the probabilities of losing and finding a job s and f, respectively — and
the discounting of time at rate r:
We(w;) = Wer [(1 _ S) We (w)) +sWy (bj, T)| (3)
1
1+r
24Other channels for macro effects that may offset the individual search response include job rationing and spillovers
to individuals not covered by the UI enhancements (Michaillat 2012, Lalive, Landais and Zweimiiller 2015, Landais,
Michaillat and Saez 2018). These channels may have affected recent assessments of expanded UI effects, although
the unusually widespread availability of enhanced UI benefits during the pandemic likely limited the spillovers to
ineligible individuals.
25 Although there is little evidence of significant wage cuts during the recession triggered by the COVID-19 pandemic,
the approach developed here is straightforward to adapt to any wage offer.
18
Wu (byt) = b+ [Af Wulbjet 1) + fmax (Welw), Wulbjt DI] @)
forl<t<T
Wu (Be) = bj + [CA f) WO) + fF max [We(w}), Wu (0)]] 6)
Wu(0) = 0+ [(1 =f) Wu(0) + f max [We(w;), Wu(0)]] 6)
where T is the maximum duration of UI, Wy(0) is the value of unemployment after exhaustion of
unemployment benefits, Wi(b;, T) is the value of unemployment at the start of a new unemploy-
ment spell following a job loss and, for a positive wage, max |We(w;), Wu(0)| = We(w;).7°
If employment if preferred to remaining unemployed at a date tf + 1 then, from the value
functions above, the value of unemployment up to the maximum duration of UI of T weeks can
be re-expressed as:
Wu (bjt) = By(t)+ (4) We(w;) forl <t <T (7)
which highlights that unemployment is valued for the discounted present value of expected UI
payments with ¢ weeks of eligibility remaining, Bj(t) = Liz bj (=) \ and the discounted value
of finding a job and moving into employment.
Since the value of unemployment in (7) is increasing in the weekly benefit amount, there exists
a reservation benefit Di(t, w;) to be paid out for the remaining weeks of eligibility t such that an
individual is indifferent between remaining unemployed and receiving that amount or accepting
a job offering pay w;. That is, a job offer with pay w; will be turned down if the current level of
weekly benefit payments b; is greater than this reservation level bi(t, w;). Formally:
Proposition 1. The reservation benefit for an unemployed individual with t weeks of UI eligibility remain-
ing and considering a job offer at wage w; solves:
Wu (bj (t,~)),£) = We(w)) (8)
Given the value functions for employment and unemployment (3) and (7) the reservation benefit is
- for 0O<t<T (9)
where
26We assume that employment immediately affords eligibility to full UI whereas state UI systems have different work
and earnings requirements to establish UI eligibility. Moreover, we make the simplifying assumption of considering
job offers at the previous wage OF and an arrival rate of job offers that is independent of prior earnings. Detailed
derivations for all results are provided in the appendix.
19
Job seekers will accept an offer to return to work at their previous wage if weekly income from
UI benefits is lower than their reservation level of benefits with t weeks of payments remaining,
b j < br (t, w j ).
For a given wage offered, the level of reservation benefit leading to a job offer being rejected is
determined by: (i) the duration of benefits remaining (t); (ii) the expected duration of the employ-
ment spell (≈ 1/s), and; (iii) the rate of arrival of new job offers ( f ). With an unbounded duration
of UI payments (T → ∞) the reservation benefit is equal to the wage, brj (∞) = w j . In this limit,
a replacement rate above 100 percent will induce workers to reject a job offer at their previous
wage rate. With one week remaining, the reservation benefit br (1, w j ) is the annuity value of the
present discounted value of the job offered. It is always the case that, with a week remaining, the
reservation benefit is greater than the wage offer (br (1, w j ) > w j ). In other words, replacement
ratios above 100 percent do not necessarily lower job offer acceptance rates. More generally, for
UI benefit payments of finite duration, the reservation benefit br (t, w j ) is declining with weeks
remaining of UI benefits, trading off an additional week of benefits at the reservation level against
the forgone employment value.
The level of the reservation benefit depends crucially on the expected duration of the employ-
ment spell and the rate of arrival of new job offers, over and above the considerations from the
duration of remaining weeks of UI eligibility. Longer lasting employment spells (lower s) are
of greater value and rejected only for commensurately generous unemployment insurance pay-
ments. In a depressed labor market, when job offers are few and far between (low f ), any job
offer is costly to refuse as new offers are hard to find. This can be seen in the discounting terms in
equations (9) and (10).
3.2 Reservation benefits during the pandemic
This section provides estimates of reservation benefits for different categories of workers during
the COVID-19 recession. We adapt the general problem to reflect institutional details from the
CARES Act and then use micro data from the CPS to obtain the relevant moments entering the
definition of a reservation benefit level. The main set of results are based on the experience dur-
ing the recovery out of the Great Recession of 2007-09, especially with respect to the expected
hazard rates out of unemployment. Additional results, obtained by varying the assumptions on
the expected durations of unemployment and employment spells, are provided and are meant
to capture the ranges of reservation benefits and offer rejection rates under alternative worker
assessments of the state of the labor market.
3.2.1 CARES Act specific formulation
The temporary nature of the supplemental PUC income relative to the duration of payments of
baseline UI requires a small modification to the unemployment Bellman equations above. Let tc
denote the weeks of expanded UI eligibility, and t p the weeks of supplemental UI income under
20
the PUC remaining for a given unemployment spell. For simplicity it is assumed that tp < te
for all unemployed. In addition, let b; denote baseline UI payments and the additional income
provided through the PUC by by. Note the baseline payments b; are conditional on past earnings
while the PUC payments by are not. The value of unemployment under the CARES Act is:
_ _ 1 _
Wu (bjrterbprtp) = bj + bp + a= [(L— f) Wu (bj, te — 1, bp, tp — 1)
+f max [We(w;),Wu(bj, tc — 1, bp, tp —1)]] forte, tp > 1 (11)
1 -
We (w)) = Wir T4r [(1 —s) We (w)) +sWy (bj, Tc) | (12)
Following similar steps as in the previous section, the value of unemployment under the CARES
Act with tf, weeks of regular UI payments and t, weeks of PUC payments may be expressed as:
Wu (Bj, te-Bprty) = Bilte) + Bp(ty) + 2 We(w)
+f
where Bj(t) = ar bj (55) is the present discounted value of expected baseline UI payments
and B,(t) = ar bp (4) the present discounted value of expected supplemental UI payments.
We focus on the the level of supplemental UI payments leading to indifference to job offers at the
previous wage w;, denoted bf jktr t-, ;). The reservation benefit for a job seeker during the period
of the CARES Act is the sum of regular benefit payments b; and this supplemental reservation
benefit payments: bi(t, te, Wj) = b; + Dt Mt, t-,w;). This level of reservation for the supplemental
benefit depends on the number of weeks of regular benefit payments remaining, t,, and, for 1 or t
weeks remaining in PUC payments, is given by:
by j(Irtert)) = Wel) ~ By(te) (13)
bY (1, te, Wj)
pipes J
Bi i(trtert)) = PE eT (14)
t-1 (1-f\!'
Li-o (3)
3.2.2 Calculating reservation benefits during the pandemic
As previously mentioned, the reservation benefits during the pandemic calculated below is the
sum of regular and supplemental reservation benefit payments for an individual considering a
of job offer at their previous wage w;, bi(t, te, Wj) = bj + bi lt, t-, w;). We specify the baseline UI
program as a weekly payment b; = min [tT x wj,bcap| for a maximum duration of T = 26 weeks,
where 7 € (0,1) is a replacement rate set to 50 percent and bap a cap on weekly payments of
$500.2” The PEUC extended the duration of UI payments an additional 13 weeks for a total of 39
27 This assumption for regular UI compensation is slightly more generous than the typical U.S. state program. See
Department of Labor (2019) for a review of the heterogeneity in eligibility requirements and benefit levels and duration
across US states. Note also the discount rate r is set to an annualized rate of 5 percent.
21
weeks, but in some states emergency extensions provide an additional 13 weeks for a maximum
of 52 weeks. We set T c to 52 weeks. The additional income provided through the PUC is denoted
by b p = $600 per week. Payments first began the week ending April 4, 2020 and the last went out
the week ending July 25, 2020, for a total of Tp = 17 weeks. Finally, the CARES Act provision of
additional UI income is assumed to no longer be available at the end of the employment spell of
any job offer under consideration.28
The remaining moments for weekly earnings, and job finding and job separation rates required
to calculate reservation benefits are obtained from the monthly CPS. These are reported for the
overall population, prime aged workers, by level of education, and by occupation of prior job in
Table 4. As in the regression analysis, individual weekly earnings from 2019 are obtained from the
2020 CPS ASEC file, for individuals recorded as unemployed in the months of January through
July 2020; this information captures the likely distribution of earnings for workers considering job
offers during the period of the CARES Act provisions. More precisely, we assume that the distri-
butions of weekly earnings for job seekers within a group are drawn randomly from a log-normal
distribution G (·), with mean µw and standard deviation σw ; these parameters are estimated di-
rectly from the CPS ASEC individual earnings data just referenced. For example, the distribution
of log weekly earnings for our overall sample of unemployed individuals has a mean of 6.58 and
a standard deviation of 0.73, while the distribution of log weekly earnings for unemployed indi-
viduals who were in managerial occupations in their prior job has a mean of 7.01 and a standard
deviation of 0.69.
In our baseline calculations, the measure of expected durations of unemployment and employ-
ment spells draws on the experience from early recovery phase following the Great Recession (the
full calendar year 2010). This period is chosen as a reasonable reference point for a job seeker’s
expectation of job offer arrival rates coming out of the initial phase of the COVID-19 recession.
In addition, below we present results under an alternative assumption regarding job offer arrival
rates, to assess model properties and robustness. The job offer arrival rate f t = UEt /Ut−1 is the
sum of transitions from unemployment to employment over the previous period’s stock of unem-
ployed individuals. The separation rate out of employment st = ( EUt + ENt ) /Et−1 is the sum of
transitions out of employment into either unemployment or non-employment over the preceding
period’s stock of employed individuals. It is worth noting that durations of unemployment spells
based on outflow rates are significantly shorter than the average durations reported by CPS re-
spondents. In other words, our chosen measure of job offer arrival rates will imply lower levels of
reservation benefits compared to using self-reported durations of unemployment spells.29
Transition rates into employment in a specific occupation are not easily defined due to the
28 Allowing for the additional UI income to be available upon reemployment, at least partially, would increase the
value of a job offer. The levels of the reservation benefit would be somewhat higher due to strong discounting over the
duration of a typical employment spell.
29 Table A1 provides durations of unemployment spells as self-reported in the CPS for comparison to the durations
implied by the finding rate f . In particular, it reports the average duration of the unemployment spell preceding a
transition into employment, which can be compared to the imputed finding rate based on durations by occupation.
Table A4 of the appendix reports the equivalent moments for 2015. See also the discussion in Farber and Valletta (2015).
22
ambiguity of identifying the pool of potential job seekers within each occupation. Our solution to
this measurement challenge follows Hall and Schulhofer-Wohl (2018). We estimate a logit on the
outcome of a potential transition from unemployment into employment into a specific occupation,
f = exp( β f X )/ 1 + exp( β f X ) , based on a set of demographic characteristics in the vector X that
includes age, education, race/ethnicity, sex, and marital status. The regressions, using all months
of 2010, are then used to predict the average transition rate by occupation (see appendix B for
further details).
3.2.3 Results: Overall, by education, and by occupation
The PUC benefit expired July 31st 2020. As such, we focus on reservation benefit levels, and the
corresponding UI benefit amounts, for individuals considering an offer to return to work at the
previous wage between the end of April and the end of June 2020. This is the core period of con-
cern over the disincentive effects of the supplemental UI payments, during which job seekers were
considering leaving behind 4 to 12 weeks of increased UI payments by returning to work. This
timeframe largely overlaps with that in the analysis of individual transition rates in the preceding
section. The second to last column of Table 4 reports the value of reservation benefits at a midpoint
in this period, the end of May, for the median worker within each group.
The final column of Table 4 displays our main results. It reports the proportion of individ-
uals within each group who will reject a job offer at their prior wage based on the comparison
between their imputed UI benefits and their calculated reservation benefit level. That is, let orj (t)
be an indicator function equal to one if an individual with prior wage w j would reject an offer to
return to work with t weeks of PUC payments remaining: orj (t) = 1 if brj (t) ≤ b j + b p . The share
of individuals rejecting an offer to return to work with t weeks remaining is then calculated as
´ r
o j (t)dG (w j ). For example, for our overall sample, the results in the first row, column 7, show
that the fraction of individuals who would reject job offers based on their UI benefits is about 22
percent; we discuss this result in more detail below.
The findings in column (7) of Table 4 are best understood by reviewing the components in
the prior columns that contribute to the calculation. Focusing on the overall sample in the first
row, the table shows that a typical (median) unemployed worker earning about $717 per week in
their previous job received $959 per week in UI payments under the CARES Act (columns 1 and
5). Considering an offer at the previous wage takes into account that the proposed employment
spell is expected to last just under two years and, if rejected, unemployment can be expected to
last 22 weeks (column 2). We calculate that an offer received with 8 weeks of CARES UI pay-
ments remaining would be accepted as long as this worker’s current benefit payment was below
brj (8, w j = 717) of around $1419 (column 6), a reservation benefit level that is about twice the pre-
vious wage.30 This is well above weekly UI payments under the CARES Act, implying that the
30 The following example illustrates the calculation of the reservation benefit level with 8 weeks of PUC payments
remaining for the median earner facing a weekly job arrival rate of f = 0.047, separation rate s = 0.012. The first step is
to calculate supplemental benefit level with 1 week of PUC and tc = Tc − Tp + 8 weeks of regular payments remaining
23
Table 4: Reservation benefits, replacement rates, and offer rejection rates
(1) (2) (3) (4) (5) (6) (7)
Rejection
Earnings Duration of: Weekly UI comp.
rate (%)
w (wkly) U (wks) E (yrs) b̄ bC br (8) Averagea
Overall 717 22 1.7 359 959 1419 21.5
Age 25 to 54 years 792 21 2.3 396 996 1598 15.6
Education:
Less then HS 514 23 0.74 257 857 908 43.1
High School 650 22 1.6 325 925 1290 22.0
College and above 955 19 2.8 477 1077 1853 15.2
Occupation:
Management 1103 20 2.3 500 1100 2196 8.4
Prof. and tech. 850 20 2.3 425 1025 1639 19.0
Services 541 22 1.2 271 871 1026 34.6
Sales 637 21 1.4 318 918 1206 29.5
Admin. 608 22 1.6 304 904 1215 26.4
Farm., fish. 608 20 1.0 304 904 1084 30.5
Constr., extract. 929 20 1.5 465 1065 1717 13.1
Inst., maint., rep. 831 20 1.7 415 1015 1572 15.3
Production 798 21 1.5 399 999 1545 12.6
Transp., materials 688 21 1.4 344 644 1314 24.2
Notes: Earnings data calculated using the 2019 CPS ASEC for indidivuals unemployed in any of the months
of March through July 2020. Durations of unemployment and employment in columns 2 and 3 are calculated
using the Dec. 2009 to Dec. 2010 CPS. w: median weekly earnings. Weekly job finding and separation rates
entering the resevation benefits are obtained by converting the monthly flow rates to a weekly frequency
(see appendix for details); b̄: median regular weekly unempmloyment benefits; bC : median weekly benefits
under the CARES Act, b̄ + 600$; br (t p ) median reservation benefit level with t p weeks left under the CARES
Act. a: Average rate of rejecting an offer arriving between 12 and 4 weeks of PUC payments remaining.
24
median worker would not reject an offer to return to work at the prior wage. However, there is
significant dispersion in weekly earnings such that many workers at the lower end of the distri-
bution might be dissuaded from accepting a job offer at their prior wage. As noted above, the
final column reports the fraction of workers whose weekly UI payments under the CARES Act
exceed their reservation benefit level, prompting them to reject job offers at their prior wage. We
find that about 22 percent would reject an offer to return to work, calculated as an average for
the period from the end of April through the end of June 2020. This proportion is essentially
identical to the magnitude of the disincentive effects on job finding rates found in the previous
section’s analysis.31 These model-based calculations are not exactly comparable to the regression
results, which are based on an empirical sample and a before/after research design. However, the
close correspondence between the estimated effect of the $600 UI supplement in the two settings
is reassuring, given the independent design of the two approaches.
As shown in the second row of Table 4, the results are similar when the analysis is restricted to
the prime age workforce, aged 25 to 54 years old, though the share who reject job offers is some-
what lower than for the full sample (16 percent versus 22 percent). The next three rows of Table 4
present the results for workers in three educational attainment groups (less than high school, high
school, and college and above). College educated workers were less likely than those with lower
educational attainment to turn down a job offer at the previous wage during the period of sup-
plemental payments under the CARES Act: their earnings are higher relative to UI benefits and
their employment spells are longer compared with individuals with less education. Conversely,
high school educated unemployed workers, with median earnings of $650 per week and expected
durations of employment and job-finding rates close to the overall average, were moderately de-
terred by a high replacement rate under the CARES Act.
The subsequent rows of Table 4 present reservation benefit calculations for workers within 10
major occupations. Median weekly earnings among the unemployed within these occupations
range from just under $541 per week (Services) to over $1100 per week (Managers), with average
durations of employment spells from under a year to about two and a half years. The median
reservation benefits levels with 8 weeks remaining in PUC payments for each occupation are re-
ported in the second to last set of columns in Table 4, while Figure 3 plots an occupation’s median
weekly earnings against the median reservation benefit within that occupation with 8 weeks of
PUC payments remaining. A 100 percent replacement rate (black line) separates the graph in two
regions, shaded in blue for replacement rates below 100 percent. Regular UI payment rates are
represented by the bottom line (red), increasing at a rate of 50 percent of the prior wage until hit-
ting a cap at $1000 in weekly earnings for a maximum benefit payment of $500 per week. The
as in equation (13), resulting in brp,j (1, tc , w j = 717) where we have used the definition of the value for employment in
(12) as detailed in appendix section A.2in equation (A.5). The supplemental benefit level with 8 weeks reamaining is
1− f i
then obtained as brp,j (8, tc , w j = 717) = brp,j (1, tc , w j = 717)/ ∑8i=−01 1+r = 1060, to which we add regular weekly UI
payments of b̄ j = 359, yielding the value reported in the first row of Table 4.
31 As noted in Section 2.3.3: “Specifically, this represents a 22 percent reduction in job-finding rates due to the avail-
ability of the $600 weekly UI benefit supplement.”
25
UI payment schedule under the CARES Act is shifted up by $600 (green line), and any individ-
ual with earnings below $1100 per week receives more on UI with the PUC payments than on
the previous job. Each occupation’s median weekly earnings and reservation benefit level with 8
weeks of PUC supplemental payments remaining are plotted as blue dots. Only median earners
employed in Services have UI payments under the CARES Act close to their respective median
reservation benefit levels.32 Individuals in these occupations were more likely than others to reject
a job offer at the previous wage during the period analyzed.
Figure 3: Regular, CARES Act and reservation level UI benefit payments
for median earner by occupation
Notes: Each dot corresponds to the reservation benefit for a with median earnings within
each occupation calculated according to equation (14) with 8 weeks of PUC payments
remaining.
32 These results by occupation are related to the robustness exercise in the earlier section evaluating whether specific
occupations or industries drive the regression results. However, our sample sizes for the regression analysis are too
small to enable regressions for each occupation.
26
In order to provide a sense of how reservation benefits and rejections of job offers are affected
by varying expectations for labor market conditions, we performed the same calculations as sum-
marized in Table 4 under an alternative assumption for job offer arrival rates and durations of
employment spells. This alternative uses data from 2015 rather than 2010 to obtain transition
rates, thereby incorporating stronger labor conditions and better job-finding prospects than in our
initial calculations. Under this scenario, the arrival rate of job offers is raised by nearly 50 percent.
The results are reported in the appendix, in Table A4, which is structured identically to Table 4.
The increased arrival rate for job offers lowers the level of reservation benefits in all occupations.
For the full sample, about 34 percent would reject an offer to return to work at their previous
wage, notably higher than the 22 percent obtained in our initial scenario and also in the regres-
sion analysis. However, this scenario is meant purely to illustrate the sensitivity of our results to
labor market conditions. We regard the results from our primary scenario displayed in Table 4 as
a better reflection of the prevailing labor conditions and job acceptance decision by recipients of
the $600 UI supplement in early 2020.33
4 Conclusion
This paper examines the impacts of the $600 weekly UI benefits supplement provided by the
CARES Act on job search and job acceptance decisions in early 2020. We first conduct direct empir-
ical analyses of labor force transitions using matched CPS data, linked to annual earning records
from the CPS income supplement to form UI replacement rates. The results show moderate disin-
centive effects of the UI supplemental payments on job finding rates. Our estimated elasticities of
job-finding rates with respect to UI replacement rates are at the low end of the range of prior esti-
mates using U.S. data (Schmieder and von Wachter 2016). Our estimates are broadly comparable
to results from two other recent papers that use administrative micro-data from financial services
companies to assess the impacts of the pandemic UI supplements (Coombs et al. 2022, Ganong
et al. 2022). In fact, our point estimate of the UI benefit elasticity for job finding is essentially iden-
tical to a similarly obtained estimate from different data in Ganong et al. (2022). Our estimates are
aimed at identifying the micro effects of the UI benefit expansions on job search, whereas other
recent analyses of the pandemic UI expansions likely combine micro effects at the individual level
with more general macro effects that alter aggregate labor market conditions (Altonji et al. 2020,
Bartik et al. 2020, Finamor and Scott 2021).
In the second part of our analysis, we derive a level of UI benefit payments over the duration
of remaining UI eligibility at which unemployed individuals are indifferent between accepting a
job paying their previous wage and remaining unemployed. With finite UI benefit duration, this
33 All else equal, the frequency of rejected job offers in the model increases with the expected job offer arrival rate.
Intuitively, UI recipients are more willing to continue receiving benefits if they know that additional job offers will
come sooner rather than later. Similarly, if UI recipients expect the availability of benefits to be extended, as they were
in the aftermath of the recession of 2007-09, this will increase the job offer rejection rate in the context of our model. We
interpret the close alignment of our regression results and model calculations in Table 4 as indicating that the latter are
based on a reasonably accurate representation of labor market conditions and job seeker expectations in early 2020.
27
reservation benefit is always above the previous wage. In a depressed labor market with lower job
offer arrival rates, the gap between the previous wage and the reservation benefit widens, leaving
room for replacement ratios above 100 percent without large negative effects on job acceptance
rates and the resulting speed of the labor market recovery. Our model-based calculations using
CPS monthly and ASEC microdata show that only a moderate faction of unemployed individuals
receiving the temporary CARES Act $600 weekly UI supplement would refuse an offer to return to
work at their previous wage. Under our preferred, most likely scenario, the model-based calcula-
tions yield reductions in job acceptance rates that are essentially identical to the estimated impact
on job finding in our preceding regression analysis.
It is worth noting a few considerations that may have a meaningful impact on an individual’s
job acceptance decision in the context of our job-search model. First, our analysis does not incorpo-
rate job-related health risks during the pandemic that would reduce the value of a job offer and the
corresponding value of the reservation benefit. We focus more narrowly on the financial disincen-
tive of the supplemental UI income on the job acceptance decision, and the calculations based on
the microdata suggest that the additional UI income alone was not likely to have deterred a large
nmber of workers from returning to work. Second, our model does not incorporate human capital
depreciation or other factors that would result in a declining job arrival rate over the duration of
the unemployment spell. This consideration would act to increase the reservation benefit level,
especially as individuals experience longer unemployment spells during a protracted slowdown.
Finally, these are partial equilibrium exercises, which do not take into account general equilibrium
effects of expanding UI policies on job offer arrival and separation rates, and are meant to be the
model counterparts of the individual level estimates uncovered in the first part of the analysis.
Consideration of such general equilibrium effects is left to future work.
28
References
Abowd, John, and Arnold Zellner. 1985. “Estimating Gross Labor Flows.” Journal of Business and
Economic Statistics, 3(3): 254–83.
Altonji, Joseph, Zara Contractor, Lucas Finamor, Ryan Haygood, Ilse Lindenlaub, Costas
Meghir, Cormac O’Dea, Dana Scott, Liana Wang, and Ebonya Washington. 2020. “Employ-
ment Effects of Unemployment Insurance Generosity During the Pandemic.” Yale University
Manuscript.
Baily, Martin Neil. 1978. “Some aspects of optimal unemployment insurance.” Journal of Public
Economics, 10(3): 379 – 402.
Baker, Michael, and Samuel A. Rea. 1998. “Employment Spells and Unemployment Insurance
Eligibility Requirements.” The Review of Economics and Statistics, 80(1): 80–94.
Bartik, Alexander W., Marianne Bertrand, Feng Lin, Jesse Rothstein, and Matthew Unrath. 2020.
“Measuring the labor market at the onset of the COVID-19 crisis.” Brookings Papers on Economic
Activity, 51(2): 239–268.
Boar, Corina, and Simon Mongey. 2020. “Dynamic Trade-offs and Labor Supply Under the
CARES Act.” NBER Working Paper 27727.
Boone, Christopher, Arindrajit Dube, Lucas Goodman, and Ethan Kaplan. 2021. “Unemploy-
ment Insurance Generosity and Aggregate Employment.” American Economic Journal: Economic
Policy, 13(2): 58–99.
Card, David, Raj Chetty, and Andrea Weber. 2007. “The Spike at Benefit Exhaustion: Leaving the
Unemployment System or Starting a New Job?” American Economic Review, 97(2): 113–118.
Chetty, Raj. 2008. “Moral Hazard versus Liquidity and Optimal Unemployment Insurance.” Jour-
nal of Political Economy, 116(2): 173–234.
Chodorow-Reich, Gabriel, John Coglianese, and Loukas Karabarbounis. 2019. “The Macro Ef-
fects of Unemployment Benefit Extensions: A Measurement Error Approach.” Quarterly Journal
of Economics, 134(1): 227–279.
Coombs, Kyle, Arindrajit Dube, Calvin Jahnke, Raymond Kluender, Suresh Naidu, and
Michael Stepner. 2022. “Early Withdrawal of Pandemic Unemployment Insurance: Effects on
Employment and Earnings.” AEA Papers and Proceedings, 112: 85–90.
Department of Labor. 2019. “Comparison of State Unemployment Laws 2019.”
Farber, Henry S., and Robert G Valletta. 2015. “Do extended unemployment benefits lengthen
unemployment spells? Evidence from recent cycles in the US labor market.” Journal of Human
Resources, 50(4): 873–909.
29
Farber, Henry S., Jesse Rothstein, and Robert G. Valletta. 2015. “The Effect of Extended Un-
employment Insurance Benefits: Evidence from the 2012-2013 Phase-Out.” American Economic
Review, 105(5): 171–76.
Feldstein, Martin. 1976. “Temporary Layoffs in the Theory of Unemployment.” Journal of Political
Economy, 84(5): 937–57.
Finamor, Lucas, and Dana Scott. 2021. “Labor market trends and unemployment insurance gen-
erosity during the pandemic.” Economics Letters, 199: 109722.
Ganong, Peter, Fiona E Greig, Pascal J Noel, Daniel M Sullivan, and Joseph S Vavra. 2022.
“Spending and Job-Finding Impacts of Expanded Unemployment Benefits: Evidence from Ad-
ministrative Micro Data.” National Bureau of Economic Research Working Paper 30315.
Ganong, Peter, Pascal Noel, and Joseph Vavra. 2020. “US unemployment insurance replacement
rates during the pandemic.” Journal of Public Economics, 191: 104273.
Hall, Robert E., and Sam Schulhofer-Wohl. 2018. “Measuring Job-Finding Rates and Matching Ef-
ficiency with Heterogeneous Job-Seekers.” American Economic Journal: Macroeconomics, 10(1): 1–
32.
Kekre, Rohan. 2023. “Unemployment Insurance in Macroeconomic Stabilization.” Review of Eco-
nomic Studies, 90(5): 2439–2480.
Kroft, Kory, and Matthew J. Notowidigdo. 2016. “Should Unemployment Insurance Vary with
the Unemployment Rate? Theory and Evidence.” The Review of Economic Studies, 83(3): 1092–
1124.
Lalive, Rafael, Camille Landais, and Josef Zweimüller. 2015. “Market Externalities of Large Un-
employment Insurance Extension Programs.” American Economic Review, 105(12): 3564–96.
Landais, Camille, Pascal Michaillat, and Emmanuel Saez. 2018. “A Macroeconomic Approach
to Optimal Unemployment Insurance: Theory.” American Economic Journal: Economic Policy,
10(2): 152–81.
Marinescu, Ioana, and Daphné Skandalis. 2021. “Unemployment Insurance and Job Search Be-
havior.” The Quarterly Journal of Economics, 136(2): 887–931.
Marinescu, Ioana, Daphné Skandalis, and Daniel Zhao. 2021. “The impact of the Federal Pan-
demic Unemployment Compensation on job search and vacancy creation.” Journal of Public Eco-
nomics, 200: 1–21.
Michaillat, Pascal. 2012. “Do Matching Frictions Explain Unemployment? Not in Bad Times.”
American Economic Review, 102(4): 1721–1750.
30
Mitman, Kurt, and Stanislav Rabinovich. 2021. “Whether, when and how to extend unemploy-
ment benefits: Theory and application to COVID-19.” Journal of Public Economics, 200: 104447.
Moffitt, Robert. 1985. “Unemployment insurance and the distribution of unemployment spells.”
Journal of Econometrics, 28(1): 85–101.
Mortensen, Dale T. 1977. “Unemployment Insurance and Job Search Decisions.” Industrial and
Labor Relations Review, 30(4): 505–517.
Poterba, James, and Lawrence Summers. 1995. “Unemployment Benefits and Labor Market Tran-
sitions: A Multinomial Logit Model with Errors in Classification.” The Review of Economics and
Statistics, 77(2): 207–16.
Poterba, James M., and Lawrence H. Summers. 1986. “Reporting Errors and Labor Market Dy-
namics.” Econometrica, 54(6): 1319–1338.
Rothstein, Jesse. 2011. “Unemployment Insurance and Job Search in the Great Recession.” Brook-
ings Papers on Economic Activity, 42(2 (Fall)): 143–213.
Schmieder, Johannes F., and Till von Wachter. 2016. “The Effects of Unemployment Insurance
Benefits: New Evidence and Interpretation.” Annual Review of Economics, 8(1): 547–581.
Shimer, Robert, and Iván Werning. 2007. “Reservation Wages and Unemployment Insurance.”
The Quarterly Journal of Economics, 122(3): 1145–1185.
Valletta, Robert G. 2014. “Recent extensions of U.S. Unemployment benefits: Search responses in
alternative labor market states.” IZA Journal of Labor Policy, , (3): 1–25.
31
Online appendix
A. Detailed derivations
A.1 Main derivations
Recall the Bellman equations:
We = wt ~ [(1 —s) We +sWu (b;,T) | (A.1)
Wu (bt) = b+ i+? [(1 — f) Wu(bj,t-1) + fWe] forT >t>1 (A.2)
Wu (0,1) = b+ ~ [(1 — f) Wu (0) + fWe] (A.3)
Wu(0) = 0+ [= f) Wu(0) + fe] (A4)
From the last line we have Wy (0) = -4;We, then:
Wu (bj),1) = b+ ~ c — f) Ws + fe] = bj + We
1
Wu (bj,2) = bj + aa [CL f) Wa (5,1) + fe]
- 6 +8) (FS f\ 4 ! [a - pt +f] We
1+ 1+r r+f
_ f f
= +b; Git)+ tem
and finally:
Wu (bt) = ro (54) + (=) "
Let bi(t,wj;) denote the value of unemployment benefit with t weeks of eligibility remaining
such that an individual is just indifferent between a job offer and remaining unemployed. With
one week of benefits remaining:
Wu (0/(1,w)),1) = We
bi (1, wj) + We = We
_f_
r+f
bi(1,w;) = (45) We
With two weeks remaining:
Wu (6F(2,0;),2) = We
bj (2, wj) [1+ (74) + toe — Ws
bi (2, w)) (sy) We vy)
+
such that bj (2,wj) < b;(1,w;). More generally: for T > t > 1
bi (1 ei)
r _ Jd
i= =0 Tr
Finally, we can re-express the value of employment as:
i (eS)
rWe = op es (1+ nay +58)(T)]
such that
(1+1r) w; +sB;,(T)
r+s+f
bi(1, w;) =
A.2. Application to the 2020 CARES Act
The value of unemployment under the CARES Act is:
Wu (Bj-ter bp, tp) = bj + bp + Tyr [(1 — f) Wu(bj, te — 1, bp, ty — 1)
+f max [We(w;),Wu(bj, te — 1,bp, tp —1)]] for te,tp > 1
Wu (bj, tc, bp, 1 ) = Bj + bp + [(1 — f) Wu (bj, tc — 1,0,0) + f max [We(w)), Wu(b;, te — 1,0,0)} |
u (By 10,0) = B+ [(1—f) Wu(0) + fmax [We(w;), Wu0)]
Wl) = pWe(w)
We(wj) = w+ = [(1 — s) We(w;) + sWu (bj, Te) |
With one week and tf, weeks of regular UI remaining and exhaustion of PUC benefits:
Wu (bi, 1,0, 0) = b; + ; f ; E(w;)
_ _ t-1 1— i 7
Wu (bi, tc, 0,0) = bj dL € —*) + — we(w)) = Bj(tc) + — we(w))
With t, weeks of regular UI payments and one week of PUC payments:
- - 1 -
Wu (bj, te,bp, 1) = dj + bp + Tay [(1 — f) Wu (bj, te — 1,0,0) + fWe(w;)|
With t, weeks of regular UI payments and t, weeks of PUC payments:
Wu (Bj, te, bp, tp) = Bi(te) + Bo(tp) +
Reservation supplemental benefit with one week of PUC remaining bj (tc, tp = 1, wj):
Wu (bj, te, b,(1),1) = We(wj)
Bj(t.) +5 ,(1, te, ,) + > we(w)) = We(w;)
, r _
Using similar steps as in the previous section, this can be expressed as
14+r)w;+sB,(T- b;
UY (1,te, Wj) = O41) w+ 8Bi(Te) _ I (A.5)
r+s+f ri (Ef)
Reservation supplemental benefit with two weeks of PUC remaining bi (tc, tp = 2, wj):
Wu (bj, te, bi, :(2),2) = We(wj)
Bi(tc) + B,(2) + — we(w)) = We(wj)
b’ .(2,t-) =
p2rte) 1 (5)
i=0 \ I+r
Reservation supplemental benefit with t weeks of PUC remaining bi (te, tp = t, Wj):
Wu (bj, te bp (t,t) = We(w;)
Bi(te) + Bp(t) + 7 We(w)) = We(w;)
r
r4f
bi, (tte) =
B’. Data
Unemployment duration is the inverse of the weekly job finding rate calculated by converting the
monthly flow rate fn = UE;/U;_1, to a weekly frequency as fw = 1— (1 — fin) *; The duration of
an employment spell is the inverse of the weekly job separation rate calculated from the monthly
flow rate S = (EU; + EN;)/E;_1, converted to a weekly rate by solving
5 = Sw {{( ~ fv) + (1—sw)] (25.0fe + (1— fw)? + (1 Sw)’) \.
Table A1: Measures of weekly earnings, unemployment and employment duration
Log weekly earnings dist. Duration of: unemployment* employment?
Reported Flow Flow
Hw Ow mean cond.onU-E 1/fw 1/syp
Overall 6.58 0.73 31.7 20.5 22 1.7
Age 25 to 54 years 6.67 0.68 33.73 22.1 21 2.3
Education:
Less then HS 6.24 0.70 28.6 18.4 23 0.74
High School 6.48 0.61 32.5 21.1 22 1.6
College and above 6.86 0.81 32.8 21.1 19 2.8
Occupation:
Management 7.01 0.69 - 18.9 20 2.3
Prof. and tech. 6.74 0.81 - 16.9 20 2.3
Services 6.29 0.64 - 20.6 22 1.2
Sales 6.46 0.77 - 22.9 21 1.4
Admin. 6.41 0.67 - 23.9 22 1.6
Farm., fish. 6.41 0.53 - 16.7 20 1.0
Constr., extract. 6.84 0.64 - 16.4 20 1.5
Inst., maint., rep. 6.72 0.62 - 21.1 20 1.7
Production 6.68 0.53 - 17.9 21 1.5
Transp., materials 6.53 0.70 - 20.3 21 1.4
Notes: (a) weeks; (b) years. Earnings data calculated using the Jan. to Dec. 2019 CPS. Durations calculated
using Dec. 2009 to Dec. 2010 CPS. w: weekly earnings; Weekly job finding fj» and separation sw rates
calculated by converting the monthly flow rates to a weekly frequency.
Job finding rates by major occupation are obtain from a logit on the outcome of a transition
from unemployment into employment, f = exp(6¢X)/ [1+ exp(6X) |, based on a set of demo-
graphic characteristics in the vector X that includes age, education, race/ethnicity, sex and marital
status. The regression results are reported in Table A2.
Table A2: Predicting Finding and Separation Rates for 2010
UE EU + EN
Age
25-34 0.0128 -0.0539 -0.953 -0.833
(0.0341) (0.0362) (0.0215) (0.0227)
35-44 -0.0316 -0.135 -1.166 -0.976
(0.0356) (0.0408) (0.0222) (0.0257)
45-54 -0.195 -0.310 -1.274 -1.070
(0.0363) (0.0430) (0.0220) (0.0263)
55-64 -0.333 -0.460 -0.970 -0.757
(0.0437) (0.0504) (0.0230) (0.0275)
65-79 -0.468 -0.604 -0.0557 0.159
(0.0759) (0.0812) (0.0268) (0.0315)
Education
H.S. Diploma 0.0721 0.0755 -0.536 -0.529
(0.0336) (0.0336) (0.0211) (0.0211)
Some College 0.149 0.170 -0.672 -0.672
(0.0355) (0.0356) (0.0214) (0.0215)
College Degree & Above 0.287 0.309 -1.020 -1.014
(0.0408) (0.0410) (0.0236) (0.0236)
Race/Ethnicity
Black -0.373 -0.343 0.408 0.356
(0.0353) (0.0357) (0.0221) (0.0224)
Hispanic 0.147 0.137 0.269 0.268
(0.0322) (0.0323) (0.0209) (0.0209)
Asian/Pacific Islander -0.248 -0.260 0.147 0.141
(0.0635) (0.0637) (0.0338) (0.0338)
Other -0.0771 -0.0627 0.291 0.267
(0.0623) (0.0624) (0.0403) (0.0404)
Sex
Female -0.169 0.0984
(0.0238) (0.0141)
Marital Status
Married (Spouse Absent) 0.243 0.221
(0.0866) (0.0572)
Widowed -0.0420 0.109
(0.0962) (0.0465)
Divorced -0.133 0.0810
(0.0393) (0.0254)
Separated 0.00183 0.213
(0.0669) (0.0477)
Never Married -0.185 0.291
(0.0323) (0.0195)
Constant -1.540 -1.314 -1.761 -2.070
(0.0323) (0.0450) (0.0210) (0.0277)
Observations 52442 52442 536849 536849
Note: Groups “16-24”, “Less than H.S. Diploma”, “White”, “Male”, and “Married
(Spouse Present)” are included as reference categories, respectively.
5
C Additional tables
6
Table A3: Reservation benefits, replacement rates, and offer rejection rates
Rejection
Earnings Duration of: Weekly UI compensation
Rate (%)
w (wkly) U (wks) E (yrs) b̄ bC br (12) br (8) br (4) 12 wks 8 wks 4 wks
Overall 717 22 1.7 359 959 1105 1419 2373 38.3 21.7 5.8
Age 25 to 54 years 792 21 2.3 396 996 1245 1598 2668 30.3 15.2 2.4
Education:
Less then HS 514 23 0.74 257 857 708 908 1514 65.8 45.4 14.6
High School 650 22 1.6 325 925 1003 1290 2161 42.0 21.8 3.3
College and above 955 19 2.8 477 1077 1461 1853 3060 27.0 15.1 3.0
Occupation:
Management 1103 20 2.3 500 1100 1712 2196 3672 17.7 7.7 1.0
7 Prof. and tech. 850 20 2.3 425 1025 1287 1639 2702 32.6 19.2 5.0
Services 541 22 1.2 271 871 801 1026 1714 57.8 35.9 8.4
Sales 637 21 1.4 318 918 944 1206 2002 47.8 30.5 8.9
Admin. 608 22 1.6 304 904 942 1215 2038 46.3 26.8 5.7
Farm., fish. 608 20 1.0 304 904 854 1084 1785 56.5 30.9 4.2
Constr., extract. 929 20 1.5 465 1065 1355 1717 2828 26.8 12.4 1.6
Inst., maint., rep. 831 20 1.7 415 1015 1237 1572 2597 30.7 14.8 1.9
Production 798 21 1.5 399 999 1207 1545 2580 28.7 11.0 0.7
Transp., materials 688 21 1.4 344 644 1026 1314 2186 42.7 24.5 5.3
Notes: Earnings data calculated using the 2019 CPS ASEC for indidivuals unemployed in any of the months of March through July 2020.
Durations of unemployment and employment in columns 2 and 3 are calculated using the Dec. 2009 to Dec. 2010 CPS. w: median weekly
earnings. Weekly job finding and separation rates entering the resevation benefits are obtained by converting the monthly flow rates to
a weekly frequency (see appendix for details); b̄: median regular weekly unempmloyment benefits; bC : median weekly benefits under
CARES act, b̄ + 600$; br (t p ) median reservation benefit level with t p weeks left under the CARES act.
Table A4: Reservation benefits, replacement rates, and offer rejection rates - stronger recovery
scenario
(1) (2) (3) (4) (5) (6) (7)
Rejection
Earnings Duration of: Weekly UI compensation
rate (%)
w (wkly) U (wks) E (yrs) b̄ bC br (8) Averagea
Overall 717 22 1.7 359 959 1141 34.2
Age 25 to 54 years 792 21 2.3 396 996 1291 26.9
Education:
Less then HS 514 23 0.74 257 857 780 54.3
High School 650 22 1.6 325 925 1028 37.7
College and above 955 19 2.8 477 1077 1485 25.1
Occupation:
Management 1103 20 2.3 500 1100 1700 18.1
Prof. and tech. 850 20 2.3 425 1025 1262 32.5
Services 541 22 1.2 271 871 811 52.9
Sales 637 21 1.4 318 918 953 44.6
Admin. 608 22 1.6 304 904 941 43.8
Farm., fish. 608 20 1.0 304 904 842 53.8
Constr., extract. 929 20 1.5 465 1065 1354 26.0
Inst., maint., rep. 831 20 1.7 415 1015 1252 28.8
Production 798 21 1.5 399 999 1239 25.8
Transp., materials 688 21 1.4 344 644 1044 39.0
Earnings data calculated using the 2019 CPS ASEC for indidivuals unemployed in any of the months
of March through July 2020. Durations of unemployment and employment in columns 2 and 3 are
calculated using the Dec. 2014 to Dec. 2015 CPS. w: median weekly earnings. Weekly job finding and
separation rates entering the resevation benefits are obtained by converting the monthly flow rates to
a weekly frequency (see appendix for details); b̄: median regular weekly unempmloyment benefits;
bC : median weekly benefits under CARES act, b̄ + 600$; br (t p ) median reservation benefit level with
t p weeks left under the CARES act. a: Average rate of rejecting an offer arriving between 12 and 4
weeks of PUC payments remaining.
8
File and source
- File
- ui-generosity-and-job-acceptance-effects-of-the-2020-cares-act.pdf
- Size
- 605,637 bytes
- SHA-256
- 4f14de87da163acef9600e627f6bb49d09e5aea391f331ae7458d77ebbd6c927
- Original
- doi.org