The Impact of the Federal Pandemic Unemployment Compensation on Job Search and Vacancy Creation
Summary
NBER Working Paper No. 28567, The Impact of the Federal Pandemic Unemployment Compensation on Job Search and Vacancy Creation, by Ioana Marinescu, Daphne Skandalis, and Daniel Zhao, dated March 2021. The abstract states that the Federal Pandemic Unemployment Compensation (FPUC) increased US unemployment benefits by $600 a week and that the authors estimate its effect on job applications and vacancy creation week by week from March to July 2020 using data from Glassdoor. It states that a 10% increase in unemployment benefits caused a 3.6% decline in applications but did not decrease vacancy creation, so that FPUC increased tightness. It also states that tightness was unusually depressed during the FPUC period and that the results help explain prior findings that FPUC did not decrease employment.
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
- Issuer
- Source documents
- Document type
- The Impact Of The Federal Pandemic Unemployment Compensation On Job Search And Vacancy Cre
- Case
- The Impact Of The Federal Pandemic Unemployment Compensation On Job Search And Vacancy Creation
Cite this document
“The Impact of the Federal Pandemic Unemployment Compensation on Job Search and Vacancy Creation.”
Original at doi.org. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/the-impact-of-the-federal-pandemic-unemployment-compensation-on-job-search-and-vacancy-creation/
, SHA-256 823c79af19a7421962d470e723757f0743f807f1bad144a12c475d77287d3f31.Full text
NBER WORKING PAPER SERIES
THE IMPACT OF THE FEDERAL PANDEMIC UNEMPLOYMENT COMPENSATION
ON JOB SEARCH AND VACANCY CREATION
Ioana Marinescu
Daphne Skandalis
Daniel Zhao
Working Paper 28567
http://www.nber.org/papers/w28567
NATIONAL BUREAU OF ECONOMIC RESEARCH
1050 Massachusetts Avenue
Cambridge, MA 02138
March 2021
We would like to thank Hyeri Choi and Jie Guan for excellent research assistance. We would like
to thank for their very helpful comments Steffen Altmann, Antoine Bertheau, Sydnee Caldwell, Nikolaj
Harmon, Robert Mahlstedt, Pascal Noel, as well as the participants of the seminars in UCD Dublin,
Harvard, Essex/Royal Holloway/Bristol. The views expressed herein are those of the authors and do
not necessarily reflect the views of the National Bureau of Economic Research.
NBER working papers are circulated for discussion and comment purposes. They have not been peer-
reviewed or been subject to the review by the NBER Board of Directors that accompanies official
NBER publications.
© 2021 by Ioana Marinescu, Daphne Skandalis, and Daniel Zhao. All rights reserved. Short sections
of text, not to exceed two paragraphs, may be quoted without explicit permission provided that full
credit, including © notice, is given to the source.
The Impact of the Federal Pandemic Unemployment Compensation on Job Search and Vacancy
Creation
Ioana Marinescu, Daphne Skandalis, and Daniel Zhao
NBER Working Paper No. 28567
March 2021
JEL No. J63,J64,J65
ABSTRACT
During the COVID-19 pandemic, the Federal Pandemic Unemployment Compensation (FPUC)
increased US unemployment benefits by $600 a week. Theory predicts that FPUC will decrease
job applications, and could decrease vacancy creation. We estimate the effect of FPUC on job
applications and vacancy creation week by week, from March to July 2020, using granular data
from the online jobs platform Glassdoor. We exploit variation in the proportional increase in
benefits across local labor markets. To isolate the effect of FPUC, we flexibly allow for different
trends in local labor markets differentially exposed to the COVID-19 crisis. We verify that trends
in outcomes prior to the FPUC do not correlate with future increases in benefits, which supports
our identification assumption. First, we find that a 10% increase in unemployment benefits
caused a 3.6% decline in applications, but did not decrease vacancy creation; hence, FPUC
increased tightness (vacancies/applications). Second, we document that tightness was unusually
depressed during the FPUC period. Altogether, our results imply that the positive effect of FPUC
on tightness was welfare improving: FPUC decreased competition among applicants at a time
when jobs were unusually scarce. Our results also help explain prior findings that FPUC did not
decrease employment.
Ioana Marinescu Daniel Zhao
University of Pennsylvania Glassdoor, Inc.
School of Social Policy & Practice 100 Shoreline Highway, Suite 100 A
3701 Locust Walk Mills Valley, CA 94941
Philadelphia PA, 19104-6214 daniel.zhao@glassdoor.com
and IZA
and also NBER
ioma@upenn.edu
Daphne Skandalis
University of Copenhagen
dskandalis@gmail.com
1 Introduction
In March 2020, the coronavirus COVID-19 led to a dramatic surge in business closures
and job losses. To address this crisis, the Federal Pandemic Unemployment Compensation
(FPUC) was voted on March 27 as part of the CARES Act. FPUC provided unemployed
workers with an additional $600 a week in unemployment benefits, until the end of July
2020. This represented an unprecedented increase in unemployment insurance generosity:
76% unemployed workers had a replacement rate above 100% with FPUC, i.e. collected
more in unemployment benefits than in their prior wage (Ganong et al., 2020). Economic
theory predicts that more generous benefits decrease job search effort and could decrease
vacancy creation (Landais et al., 2018b). Therefore, FPUC might have increased unem-
ployment and dampened economic activity, in particular during the summer of 2020, when
many lockdown measures were lifted. Yet, the early evidence suggests that there were at
most very limited effects on employment (Bartik et al. (2020); Altonji et al. (2020), Dube
(2020)). Why? We show that the FPUC did indeed have a negative effect on search ef-
fort. However, we provide two pieces of evidence that can explain why FPUC did not
decrease employment. First, FPUC did not decrease vacancy creation—not even during
the re-opening phase in June-July 2020. Second, labor market tightness was unusually low
during FPUC: since jobs were already receiving an unusually high number of applications,
a decrease in applications likely had limited effects on the number of workers hired.
Identifying the effect of FPUC is not straighforward. In principle, one can estimate the
effect of FPUC on local labor markets by exploiting the large disparities in the propor-
tional increase in unemployment benefits: for workers who would have been eligible to lower
weekly benefits levels before the CARES Act, the additional $600 per week represents a
larger relative increase. However, the passage of the FPUC coincided with exceptionally
large and brutal changes in the labor market: just before the implementation of FPUC, the
number of job listings collapsed (Forsythe et al., 2020a), while the number of unemployed
workers skyrocketed. What is more, workers in low-paying jobs were most likely to lose
their jobs, and these workers are precisely the ones who experienced the largest increase
in benefits with the FPUC. These shocks are potential confounding factors for the identi-
fication of the effect of the FPUC on the labor market. To credibly neutralize them in our
empirical analysis, we track labor market outcomes at a very granular level.
We collect detailed data on job applications and job postings from Glassdoor.com, for
January 2018 to July 2020. We measure search effort and vacancy creation by counting
the number of applications and new job vacancy postings each week, in each local labor
market —defined as the interaction of state, occupations and industries and wage deciles.
We compute potential replacement rates using the calculator in Ganong et al. (2020). To
strengthen the external validity of our analysis, we re-weigh observations such that each
State×industry×occupation reflects the proportion of the US labor force in that labor
1
market in the Current Population Survey (CPS) for 2018-2020.
In the first part of the paper, we estimate the impact of FPUC on search effort, job
vacancies and labor market tightness in March-July 2020. We exploit differences in propor-
tional increases in benefits across local labor markets. We include several control variables
to flexibly account for changes in the local labor market that were unrelated to the FPUC.
We allow for market-specific seasonal variation, using week-of-year by local labor market
fixed effects. We include state by week fixed effects to control for differential COVID-19
related policies across states, and other factors that change at the state level. We control
for industry by week fixed effects, which accounts for the differential impact of the crisis
by industry. Additionally, we allow for different trends in local labor markets that were
differentially exposed to unemployment during the COVID-19 crisis prior to the CARES
Act. Our identifying assumption is that, in the absence of the policy change, the outcome
variables would have evolved similarly in local labor markets with different increases in
UI, conditional on controls. We show that future increases in FPUC do not correlate with
trends in job applications, job vacancies or labor market tightness in the weeks prior to
the enactment of the FPUC. This supports the validity of our identification strategy. We
also confirm in robustness checks that our main results hold in the construction sector,
where job loss before FPUC was not systematically correlated with earnings (and thus the
increase in the replacement rate), and hence identification is less challenging.
We find that a 10% increase in the benefit replacement rate due to the FPUC leads to
a 3.6% decline in job applications. At the same time, FPUC had no effect on job vacancies.
The absence of an effect on vacancy creation indicates that employers’ recruiting was not
limited by workers’ lower search effort or potentially higher wage demands, but rather by
other factors, consistent with job rationing models (Michaillat, 2012). Taking the effect on
applications and vacancies together, a 10% in the benefit replacement rate increased labor
market tightness by 3.3%. We then analyze the timing of the effect. One could have expected
to see a larger impact of FPUC during the reopening phase when most lockdowns were
lifted (June-July 2020), since economic activity was picking up. However, FPUC still had
no effect on vacancy creation. Further, the effect of FPUC on job applications decreased
from May until the end of FPUC in July 2020, suggesting that job seekers might have
increased their search effort in anticipation of FPUC expiration.
Our results are similar when we exclude periods of lockdown, or when we focus on
teleworkable occupations, which are less affected by social distancing measures. In markets
where workers are more likely to be on temporary layoffs, the effects of FPUC on search
effort are significantly smaller. This is consistent with the idea that workers are less likely
to search when they expect to be recalled, and suggests that the prevalence of temporary
layoffs during the COVID-19 crisis might have contributed to attenuating the effect of
unemployment insurance on job search.
In the second part of the paper, we describe the context in which FPUC was imple-
2
mented, as it is crucial to assess the impact of FPUC on unemployment and welfare. We
document the evolution of applications, vacancies, and labor market tightness. During
FPUC, seasonally adjusted aggregate applications slightly increased by 4.4% relative to
January-February 2020, despite the negative effect of FPUC on search effort. This prob-
ably reflects in part the drastic increase in the number of unemployed workers. At the
same time, job vacancies declined by 26%. As a result, labor market tightness decreased
by 31%. This implies that on average employers got more applicants for their vacancies,
and recruiting was easier. Even in the markets most affected by FPUC (4th quarter of the
increase in replacement rate due to FPUC), tightness during the FPUC was 19% lower
than in January-February 2020. During the re-opening period, tightness was only back to
its pre-pandemic level, even though many unemployed workers were receiving FPUC as
new job openings picked up.
We finally discuss how our results contribute to understanding the welfare impact of the
FPUC during the COVID-19 crisis. Landais et al. (2018b) formalize that increases in unem-
ployment insurance improve welfare when unemployment insurance increases tightness—
note that the effect on tightness depends on the effect on vacancy creation, and is a priori
ambiguous—and tightness is inefficiently low. We demonstrate that this is the case for
FPUC. First, we show that FPUC did not affect vacancy creation and hence increased
tightness. Second, we document that labor market tightness was particularly low during
FPUC, suggesting that tightness in the absence of FPUC would have been inefficiently low.
Importantly, these results hold for the reopening phase (June-July 2020). Taken together,
our results hence indicate that the effect of FPUC on labor market tightness was welfare
improving, during the whole period of FPUC. Our paper focuses on welfare considerations
that are related to labor market tightness. For a comprehensive welfare analysis of FPUC,
one should also take into account the welfare gains associated with limiting the transmis-
sion of the virus by avoiding work (Fang et al., 2020), and welfare gains associated with
consumption smoothing for UI recipients and the stimulus effect on the economy (Ganong
et al., 2021). These additional welfare considerations would likely reinforce our conclusion
regarding the desirability of high unemployment insurance in the context of FPUC.
Our paper contributes to the body of work analyzing the effects the CARES Act on
employment. Several studies show no association between FPUC and employment, either
using the Homebase dataset of hourly employees in small businesses (Bartik et al., 2020;
Altonji et al., 2020; Finamor and Scott, 2021) or the Census Household Pulse Survey (Dube,
2020). Relative to this literature, we provide two major contributions. First, our data allow
us to improve the identification strategy: measuring search effort and vacancy creation
instead of employment, and at a very granular level helps us control for changes that are
due to the COVID-19 crisis rather than FPUC. Second, we provide a potential explanation
for the limited effect of FPUC on employment: FPUC did not affect vacancy creation,
and its large negative effect on search effort happened at a time when returns to search
3
were particularly low. This complements other explanations for the limited employment
effect in the literature. In their models, Boar and Mongey (2020); Petrosky-Nadeau (2020);
Mitman and Rabinovich (2021) emphasize that the effect of FPUC on search behavior
should be attenuated because unemployed workers anticipated its expiration. Although we
find a substantial effect of FPUC on search effort, our finding that it decreased between
May and July is consistent with the idea that workers reacted less when they anticipated
the expiration. Ganong et al. (2021) show that through a stimulus effect on consumption,
FPUC also had a positive effect on employment, which could offset the potential negative
effect mediated through the labor market.
Second, this article contributes to the literature on the impact of unemployment ben-
efit levels on job search, and vacancy creation. The increase in the level of benefits with
FPUC was exceptionally large (Schmieder and von Wachter, 2016). Yet, our findings are
consistent with the estimated effect of unemployment insurance on search effort in prior
literature (Krueger and Mueller, 2010; Fradkin and Baker, 2017; Lichter and Schiprowski,
2020). They also confirm the findings of dynamic effects in Marinescu and Skandalis (2021)
and DellaVigna et al. (2020). Moreover, our results add to the limited direct evidence on
the impact of unemployment benefits on vacancy creation: we find that the number of
vacancies was unaffected by the increase in unemployment insurance generosity, similar to
what happened during the Great Recession (Marinescu, 2017). This is consistent with the
evidence of spillover effects of unemployment benefits extensions (Lalive et al., 2015).
Last, our analysis contributes more generally to the study of the labor market during
COVID-19 (Bartik et al., 2020; Gupta et al., 2020; Cheng et al., 2020; Fairlie et al., 2020;
Montenovo et al., 2020). We add to work investigating trends in the applications and job
postings during COVID in Sweden (Hensvik et al., 2020), and job postings in the US
(Forsythe et al., 2020a; Campello et al., 2020; Forsythe et al., 2020b,c).
2 Data
We combine several datasets from Glassdoor, one of the world’s largest jobs and career sites.
We use job vacancy listings and job applications from Glassdoor’s online jobs platform in
April 2018 to July 2020 for the U.S. We also exploit information on self-reported users’
salaries to measure replacement rates.
2.1 Data on job listings and job applications
Glassdoor contains job openings that companies post directly to the site, or that are col-
lected on company career sites or third-party job boards. Glassdoor then ties job listings
data to specific industries and canonical occupations that can be mapped into O*NET-
SOC codes. We focus on new job listings posted each week rather than the total stock of
4
active job listings. The flow is more responsive to sharp changes in policy and labor mar-
ket conditions, and captures a disproportionate share of application activity. To measure
search effort, we take the count of applications, defined as a user starting an application
by clicking on the “Apply Now” button1 . Applications provide a proxy of search effort,
under the assumption that the probability of applying on Glassdoor’s website rather than
using other search channels is relatively stable over time. We allocate applications into an
industry and occupation category based on jobs characteristics, and into a state based on
applicants’ address. Applications on Glassdoor don’t come exclusively from unemployment
benefits recipients, therefore our estimates at the local labor market level also capture the
search effort of job seekers who do not receive UI. This is the right measure for the pur-
pose of investigating overall labor market tightness. As job seekers who do not receive UI
should react less to changes in the generosity unemployment insurance, these market level
estimates could be smaller than estimates of the microeconomic impact on the search of
UI recipients.
Glassdoor’s job listings capture a large portion of online U.S. job listings, which them-
selves comprise a substantial percentage of all job openings.2 Job listings at very small
businesses or managed by unions may be underrepresented (Chamberlain and Zhao, 2019).
As a result, compared to the Current Population Survey (CPS) labor force by industry
in 2018-2020, some sectors like Construction are underrepresented, while other sectors like
Information Technology are overrepresented (Table A.1). We hence use weights in all our
analysis, based on the average number workers in the labor force in the industry, occupa-
tion and state from the CPS for January 2018- July 2020. We confirm that weighted job
listings evolve very similarly—the correlation is 0.6 for 2019-2020—to vacancies in the rep-
resentative Job Openings and Labor Turnover Survey (JOLTS) (Figure A.2). In particular,
the magnitude of the decrease in job listings in April and May 2020 is the same in the two
data sources.
2.2 Data on earnings and unemployment benefits
To study the role of UI generosity, we need to determine the “potential replacement rates”
for job seekers in our sample, i.e. the replacement rates that those eligible for unemploy-
ment insurance would receive at each point in time. The replacement rate level in normal
times depends on rules specific to each state. We use the benefits calculator created by
Ganong et al. (2020), which computes unemployment benefits based on individuals’ state
of residence and pre-displacement quarterly earnings based on UI guidance provided by
1
The data is a subset of total applications on Glassdoor. To enhance data quality, Glassdoor’s data
processing retains likely people (not bots) who apply after organically searching for jobs on Glassdoor
(rather than being redirected to the job posting from elsewhere, e.g. from Glassdoor emails or from paid
advertisements on other sites).
2
The count of job openings reported in Glassdoor’s Job Market Report represents 81% of the count job
openings reported in the Job Openings and Labor Turnover Survey.
5
the Department of Labor. We add $600 to standard weekly benefits levels to compute the
replacement rate with the FPUC.
We need two key pieces of information to compute the standard weekly benefits level:
the average weekly earnings before unemployment and the number of days worked for each
quarter in the year before unemployment. For wages, we use earnings self-reported by Glass-
door users. Although it is not required to search for jobs on the platform, some Glassdoor
users report information on their wage, including base salary and additional compensation
like tips and cash bonuses3 . We take the average prior earnings for each state-occupation,
using a very disaggregated occupation definition (O*Net 6-digit codes, 416 categories).4
The second piece of information we need to infer workers’ potential replacement rate is the
number of days worked in each quarter of the year before unemployment. We assume that
individuals worked continuously during the four quarters preceding unemployment.5
2.3 Panel at the week and local labor market level
We arrange our data as a panel, at the calendar week and local labor market level. To be
able to add granular fixed effects to our regressions, we define a local labor market as the
interaction of state, 2-digit occupations (76 categories in our sample), 2-digit industries (12
categories in our sample) and wage deciles. We keep local labor markets with at least 1000
applications in 2018-2020, such that we can detect variation in weekly applications count
(our results are robust to using alternative thresholds). We present descriptive statistics
for the main variables in Table A.2.
3 Institutional background and empirical strategy
3.1 Institutional background
The CARES Act was signed into law on March 27 2020. As part of the CARES Act, the
Federal Pandemic Unemployment Compensation (FPUC) was an additional $600 a week
of unemployment benefits for all workers eligible to unemployment insurance. The FPUC
was also available to workers who are not normally eligible for benefits, but were eligi-
ble for Pandemic Emergency Unemployment Compensation (PEUC) benefits or Pandemic
Unemployment Assistance (PUA) benefits. PEUC provided an additional 13 weeks of ben-
efits for those whose regular unemployment benefits have been exhausted. PUA provided
3
Karabarbounis and Pinto (2018) show that the average and variance of the distribution of salaries
within geographical region and within industry are fairly representative when compared against the Quar-
terly Census of Employment and Wages and Panel Study of Income Dynamics.
4
We do not use industry for prior earnings because the cells become very small and most of the variation
in wages is explained by our narrow occupation categories in our data.
5
This is the case for most individuals eligible for unemployment insurance, given the average number
of days worked reported in the 2019 Annual Social and Economic Supplement (ASEC) of the CPS.
6
payment to workers not traditionally eligible for unemployment benefits (self-employed,
independent contractors, workers with limited work history, and others) who were unable
to work as a direct result of the coronavirus public health emergency. The $600 FPUC
amount was chosen to raise the UI replacement rate to 100% for the average U.S. worker.
Because of technical issues in state unemployment insurance systems, it was not possible
to tailor FPUC to each worker’s prior earnings. Once the CARES Act was voted, on March
27, unemployed workers could anticipate their upcoming replacement rate, and potentially
adjust their behavior accordingly. But it took a few weeks, until the end of April, for all
states to start paying out benefits (Bartik et al., 2020). The last FPUC payment was the
week of July 20, 2020.
3.2 Empirical strategy
Variation in the increase in replacement rate For identification, we exploit the
large disparities in the magnitude of the increase in replacement rate. They mostly come
differences in pre-displacement earnings: with a fixed $600 increase in benefits, unemployed
workers with low prior wages experienced a larger increase in replacement rate.6 We focus
on the estimation of impact of the relative increase in replacement rate, in order to calculate
elasticities. We leverage differences in the relative increase in replacement rate across local
labor markets. Because our local labor markets are relatively narrow, they capture a lot
of the variation in pre-displacement earnings. In particular, we capture the variation in
earnings across occupations, and even within occupation due to seniority and specialization
(Marinescu and Wolthoff, 2019).
Identification strategy We can identify the effect of the increase in replacement rate
under the assumption that, in the absence of a change in unemployment insurance, the out-
come variable would have evolved similarly in local labor markets with different increases in
UI (conditional on controls). The identification of the effect of FPUC on the outcome vari-
ables is threatened by any factor that affects the outcomes around the date of the CARES
Act and correlates with the increase in UI generosity. The enactment of FPUC coincided
with an exceptionally large and abrupt negative labor market shock due to COVID-19.
In March, employment decreased precipitously and unemployment skyrocketed (appendix
Figure A.1). This shock disproportionately hit low-paying jobs, and hence workers who ex-
perienced the largest increase in their potential replacement rate due to FPUC. Therefore,
failing to account for this shock would likely bias estimates of the causal effect of FPUC on
aggregate labor market outcomes. We address this identification challenge in several ways.
First, we control for various observable variables to absorb the effect of the COVID-19
crisis, or other potential confounders. We include week-of-year×local labor market fixed
6
This is true both in absolute terms (i.e. the additional benefit over their prior wage is larger) and in
relative terms (i.e. the additional benefit over their unemployment benefit without FPUC is larger).
7
effects, to flexibly allow for seasonal variation in each local labor market. As the COVID-19
crisis hit different states and industries in different ways, we include time×state fixed ef-
fects, and time×industry fixed effects. We also take a different approach, and try to directly
quantify the exposure to the COVID-19 crisis in each local labor market. We estimate the
risk of unemployment immediately prior to the CARES Act. Using the monthly CPS for
January to March 2020 matched with the Outgoing Rotation Group 2019 (ORG) (Flood
et al., 2020), we predict the probability of being unemployed for individuals in the labor
force, in a logit model including industry, occupation, wage decile and state fixed effects.
We use the estimated coefficients to predict the risk of unemployment in our sample. We
allow for different trends in markets differentially exposed to the COVID-19 crisis in our
estimation model, by including exposure decile×industry fixed effects.7 . Figure A.3 shows
that substantial variation in the increase in replacement rate remains even once we take
out the portion explained by state, industry, and by exposure decile.
Second, we test the plausibility of our identifying assumption by analyzing the correla-
tion between the evolution of outcomes before the CARES Act, and the magnitude of the
benefits increase induced by FPUC. If our identification strategy is valid, our pre-CARES
Act estimates should not be significant. This is equivalent to the test of the parallel trend
assumption in a differences-in-differences setup.
Third, we conduct a robustness check where we focus on a segment of the economy where
identification problems appear less severe. Specifically, we analyze the correlation or prior
earnings with the probability of becoming unemployed at the onset of the COVID-19 crisis,
by sector, using the CPS for January-March 2020, matched with ORG 2019. Consistent
with descriptive statistics presented in Figure A.1, there is a strong negative relationship
between prior earnings and the risk of unemployment (Table A.3), but this correlation is
small and insignificant in the Construction sector (in the Mining sector as well, but there
is virtually no applications for this sector on Glassdoor). We hence re-estimate the effect
of FPUC for Construction, using the same empirical model without including week by
state, week by industry and week by exposure fixed effects.8 The estimates give the causal
effect of FPUC, without imposing any functional form to the confounding effect of the
COVID-19 crisis on outcomes, under the alternative identifying assumption that there is
no substantial confounding factor in this sector.
Estimation models We estimate the following model:
7
We use exposure deciles for more flexibility, but our results are completely similar if we instead control
linearly for exposure×week.
8
There is substantial variation in earnings across frequent occupations in the construction sector in CPS-
ORG: for instance, they range around $1410 for community association managers ; $1020 for electricians
; $960 for construction laborers.
8
RepF P U C
βτ · 1[t = F P U C + τ ] · ln(
X
ln(Yt,s,o,i,w ) = )s,o,i,w +
Rep
τ 6=−1 (1)
λweek(t),s,o,i,w + ρs,t + φi,t + ψt,Exposure(s,o,i,w) + εt,s,o,i,w
ln( RepRep
F P UC
)s,o,i,w gives the relative increase in the replacement rate associated with
FPUC in each local labor market, defined as the intersection of a state s, an occupation
o, an industry i and a wage decile w. ln(Yt,s,o,i,w ) denotes the logged outcome variable in
calendar week t (to avoid missing values, we add 1 to each outcome before taking the log).
1[t = F P U C + τ ] indicates that the calendar week t is τ weeks before the start of FPUC
(for τ < 0), or after (for τ ≥ 0). We take as reference period τ = −1, i.e. the week before the
start of FPUC (we consider that FPUC starts with the CARES Act, the week of March 23,
2020). The βτ are the coefficients of interest. We include week-of-year×local labor market
fixed effect, λweek(t),s,o,i,w ; we let time t fixed effects vary by state (ρs,t ), industry (φi,t ) and
Exposure decile (ψt,Exposure(s,o,i,w) ). εt,s,o,i,w is the error term. We cluster standard errors at
the state level, and observations are weighted by the average number workers in the labor
force in the industry, occupation and state from the CPS for January 2018- July 2020.
4 The impact of FPUC on the labor market
4.1 Main results
Pre-FPUC trends Figure 1 presents the correlation between labor market outcomes and
the increase in UI generosity induced by FPUC, week by week. We present our preferred
specification, including the full set of controls: time×state fixed effects, time×industry
fixed effects and time×Exposure decile fixed effects. There is no significant correlation be-
tween applications and future benefits increases in weeks prior to the adoption of FPUC.
This supports the credibility of our identifying assumption that, in the absence of FPUC,
the evolution of outcomes would be similar in markets that experienced different benefits
increases, conditional on the control variables included. Under this assumption, the coeffi-
cients for the period after the adoption of FPUC can be interpreted as the causal effect of
FPUC.
9
Figure 1: The impact of FPUC on applications, vacancies and tightness, week by week
(1) Applications (2) Vacancies (3) Labor market tightness
10 Notes: This Figure reports the estimates from regressions of the logged count of applications, the logged count of vacancies, and the logged labor
market tightness (i.e. ratio of vacancies over applications) on the interaction of each week around the enactment of the FPUC and the potential
increase in UI generosity ln( RepRep
F P UC
). The coefficients before the enactment of the FPUC help test our identification assumption. The coefficients
after represent the elasticity of the outcome with respect to benefits levels. We use a panel at the calendar week and local labor market level—where
local labor markets are defined as as the interaction of state, 2-digit occupations and 2-digit industries and wage deciles. We include calendar week
and local labor market fixed effects. We control for seasonality using week-of-year×local labor market fixed effects. This Figure presents our preferred
specification, with the full set of controls for potential confounders: time×state fixed effects, time×industry fixed effects and time×Exposure deciles
fixed effects. We use weights to reflect the proportion of the labor force in each State×industry×occupation in the CPS. Thin lines denote 95%
confidence intervals, based on robust SE clustered at the state level.
The effect of FPUC The increase in unemployment insurance generosity leads to a
decrease in applications immediately after the CARES Act, which remains visible until
FPUC expires (Figure 1, (1)). Quantitatively, Table 1 shows that the corresponding elas-
ticity of applications with respect to the benefit replacement rate is -0.36, (col. (1)): a 10%
increase in the replacement rate leads to a 3.6% decline in job applications, consistent with
the theoretical prediction that a higher benefit replacement rate decreases search effort. In
contrast, FPUC had no clear effect on job vacancies (Figure 1 (2)). While there is a decrease
in the number of vacancies immediately after FPUC was voted and until mid-April, it is
only transitory and imprecisely estimated. Since FPUC decreased applications and had no
effect on vacancies, it increased labor market tightness (Figure 1, (3)). Consistently, we see
in column (2) of Table 1 that the elasticity of vacancies with respect to the replacement
rate is close to zero and statistically insignificant. Column (3) shows that the elasticity
of tightness (vacancies/applications) is therefore close to the opposite of the elasticity of
applications: a 10% increase in the benefit replacement rate leads to a 3.34% increase in
labor market tightness.
Table 1: Elasticity of search with respect to benefits levels
Applications Vacancies Tightness
(ln) (ln) (ln)
(1) (2) (3)
ln( RepRep
F P UC
)×FPUC period -0.361*** -0.027 0.334***
(0.071) (0.069) (0.106)
Exposure decile×Time FE X X X
Industry×Time FE X X X
State×Time FE X X X
Labor market×Week of year FE X X X
No. of Obs 1,523,252 1,523,252 1,523,252
Notes: This Table reports the estimated elasticity of job applications, vacancy creation and labor
market tightness (vacancies/applications) with respect to unemployment benefits levels. The es-
timates are obtained from the regression of logged labor market outcomes on the logged increase
in benefits interacted with a dummy for the period between the vote of FPUC and its expiration,
in a panel at the calendar week and local labor market level—where local labor markets are
defined as as the interaction of state, 2-digit occupations and 2-digit industries and wage deciles.
We use weights to reflect the proportion of the labor force in each State×industry×occupation
in the CPS. Robust standard errors clustered at the state level are in parenthesis (* p< 0.10, **
p< 0.05, *** p< 0.010).
4.2 Additional results
We first analyze the timing of the effect of FPUC on search effort. In Figure 1 (panel (1)),
we see that the negative effect on search strongly increased in the first 3 weeks, possibly due
to delays in initial benefit receipt. The effect then remained roughly constant until June,
11
Table 2: Additional results on the impact of FPUC on applications
Applications (ln)
(1) (2) (3) (4)
ln( RepRep
F P UC
)×FPUC period, March -0.144*
(0.079)
ln( RepRep
F P UC
)×FPUC period, April -0.385***
(0.075)
ln( RepRep
F P UC
)×FPUC period, May -0.436***
(0.065)
ln( RepRep
F P UC
)×FPUC period, June -0.395***
(0.080)
ln( RepRep
F P UC
)×FPUC period, July -0.326***
(0.085)
ln( RepRep
F P UC
)×FPUC period×Lockdown -0.356***
(0.070)
ln( RepRep
F P UC
)×FPUC period×Open -0.368***
(0.075)
ln( RepRep
F P UC
)×FPUC period, Telework -0.382***
(0.100)
ln( RepRep
F P UC
)×FPUC period, Not telework -0.311***
(0.086)
FPUC period × Telework 0.131
(0.141)
ln( RepRep
F P UC
)×FPUC period, High temp layoff -0.173**
(0.080)
ln( RepRep
F P UC
)×FPUC period, Low temp layoff -0.414***
(0.075)
FPUC period × High temp layoff -0.361***
(0.107)
P-Value Test May = July .013
P-Value Test Lockdown = Open .702
P-Value Test Telework = Not telework .554
P-Value Test High temp layoff = Low temp layoff .009
Exposure decile×Time FE X X X X
Industry×Time FE X X X X
State×Time FE X X X X
Labor market×Week of year FE X X X X
No. of Obs 1,523,252 1,523,252 1,496,988 1,523,252
Notes: This Table reports the estimates from the regression of logged search effort on various
independent variables, in a panel at the calendar week and local labor market level—where
local labor markets are defined as as the interaction of state, 2-digit occupations and 2-digit
industries and wage deciles. We use weights to reflect the proportion of the labor force in each
State×industry×occupation in the CPS. Robust standard errors clustered at the state level are
in parenthesis (* p< 0.10, ** p< 0.05, *** p< 0.010).
and decreased until the end of FPUC. The estimated elasticity is -0.44 in May, at its largest;
in July, as the extra benefits are about to expire, the elasticity is significantly smaller (in
absolute value) at -0.33 (Table 2, col. 1).9 Our results are consistent with theory predicting
9
We present the estimates from the same specifications for vacancies and tightness in A.4.
12
that job seekers increase their search effort as benefits expiration approaches (Marinescu
and Skandalis, 2021), and therefore that increases in unemployment insurance have smaller
effects on job search when workers anticipate that they are temporary (Boar and Mongey,
2020; Petrosky-Nadeau, 2020; Mitman and Rabinovich, 2021).
Over time, the economic environment also changed with many states adopting lockdown
measures in March that were later relaxed. As these measures could hinder economic
activity, one could expect the impact of unemployment insurance on the labor market
to be attenuated during lockdowns. The elasticity of search with respect to unemployment
insurance was slightly smaller when states implemented a lockdown, but the difference is
small and insignificant (Table 2 col. (2)). This suggests that the difference in the effect of
unemployment insurance on search over time is not explained by lockdown measures, but
rather by unemployed workers’ anticipation of FPUC expiration. In Appendix A.4 (col.
(2)) we also show that if anything, there is more of a negative effect on vacancy creation
during lockdowns (although the estimate is insignificant).
One could expect that the reaction of job seekers to the increase in unemployment insur-
ance was smaller than in normal times, because the risk of contamination with COVID-19
imposed constraints on search and on work. Yet, we find little evidence for this: in Table 2
(col. 3), we show that the effect of FPUC on search effort is similar in “teleworkable” and
non-“teleworkable” occupations (defined as in Dingel and Neiman (2020)).
The situation of unemployed workers during the COVID-19 crisis was also atypical in
that a large fraction of them were on temporary layoff (Birinci et al., 2021). We estimate
the chance to be on temporary layoff in the CPS for January to July 2020 conditional on
being unemployed, in a logit model with two-digit occupation, two-digit industry, state and
wage decile fixed effects. We predict the probability of temporary layoff among unemployed
workers in each local labor market based on the estimated coefficients. In Table 2 (col. (4)),
we show that the effect of FPUC on search effort is significantly smaller in markets with
a high (above median) chance of temporary layoff (elasticity of -0.17) than in the others
(elasticity of -0.41). It suggests that the prevalence of temporary layoffs has contributed to
attenuating the effect of FPUC on job search, since workers who expect to be recalled are
search less irrespective of their unemployment insurance.
4.3 Robustness checks
To illustrate the role of the controls we include in our favorite specification, we present
estimates when we add them one by one in Figure A.4, Figure A.5 and Figure A.6. All
control variables contribute to making the pre-trends disappear, in line with the economic
intuitions behind our estimation strategy. For instance, when we only control for seasonal-
ity, the coefficients in panel (2) of Figure A.4 exhibit a steep increase in the weeks before
the CARES Act, which might reflect the abrupt surge in unemployment among low-wage
13
workers. Although the different controls seem to improve our identification of the effect of
FPUC, it is reassuring to note that our qualitative conclusions do not crucially depend on
one particular specification. The estimates obtained in all the specifications in A.4 are sug-
gestive of a negative effect of FPUC on job search. None of the specifications in Figure A.5
provide any indication of a negative effect on vacancy creation. Finally, all specifications
in Figure A.6 are suggestive of a positive effect of FPUC on tightness.
We then analyze more flexibly the relationship between labor market outcomes and the
increase in unemployment benefits induced by FPUC in Table A.5. We look separately at
the evolution of outcomes in local labor markets at different quartiles of the distribution
of relative increase in benefits in columns (1)-(3). As expected, the largest increase in
unemployment benefits was associated with the largest decrease in applications. The effect
for the third and fourth quartile is essentially the same, suggesting there is no additional
decrease in applications for very high increases in the replacement rate. In columns (4)-
(6) we estimate the effect of the absolute increase in replacement rates. The results are
qualitatively similar.
Finally, to confirm our main results, we focus on the construction sector, where job
loss before FPUC was not systematically correlated with earnings. While the estimates are
naturally less precise than in the full sample, we see similar qualitative patterns: FPUC
decreased applications, had no effect on vacancies, and therefore ultimately increased labor
market tightness (Figure A.7). These results are reassuring: our key results hold in a seg-
ment of the economy where there are no substantial factors confounding the identification
of the effect of FPUC.
5 Labor market tightness during the FPUC
In order to fully understand the impact of FPUC, we now document the labor market
situation when it was introduced.
5.1 The evolution of labor market tightness
We present the evolution of seasonally adjusted applications, new vacancy postings, and
labor market tightness relative to their baseline levels in Jan-Feb 2020 in Figure 2. The
number of vacancies strongly declined both before and right after FPUC. In contrast, the
number of job applications exhibited limited change. This suggests that the various factors
that could influence search in that period (the increase in unemployment, FPUC, the
infection risk, uncertainty, etc.) largely compensated each other.10 Therefore the decline in
vacancies entirely drove the evolution of labor market tightness.
10
In the non seasonally adjusted series (Figure A.8, left panel), applications decline a bit after FPUC,
but this pattern occurs every year (appendix Figure A.9).
14
Figure 2: Changes in applications, vacancies, and labor market tightness relative to
Jan-Feb 2020 (seasonally adjusted)
Notes: The Figure presents the seasonally adjusted changes in the total weekly count of applica-
tions, the weekly count of new posted vacancies, and labor market tightness (ratio of vacancies
over total applications), relative to their baseline levels in Jan-Feb 2020. The Figure is obtained
by regressing the logged variables on the calendar week coefficients and week-of-year fixed effects
to control for seasonal variation, and then subtracting to each calendar week coefficient the aver-
age of estimates for Jan-Feb 2020. We use weights to reflect the proportion of the labor force in
each State×industry×occupation in the CPS.
15
Quantitatively, seasonally adjusted labor market tightness declined by 30.6% during the
FPUC period relative to January-February 2020 (Table 3, col. 3, upper panel), reflecting a
labor market more favorable to recruiting firms, despite the large increase in unemployment
benefits. During the whole FPUC period, applications slightly increased by 4.4% (col. 1,
upper panel), while vacancies declined by 26.2% (col. 2). During the reopening phase, in
June-July 2020, seasonally adjusted applications, vacancies, and labor market tightness
were similar to their level in Jan.-Feb. 2020 (col. 1-3, lower panel). Overall, the evidence is
not consistent with significant hiring difficulties during the reopening phase.11
Table 3: Changes in labor market tightness during the period of the FPUC, relative to
Jan-Feb 2020
Potential increase in UI
Sample: All Q1 Q2 Q3 Q4
Outcome: Applications Vacancies Tightness Tightness Tightness Tightness Tightness
(1) (2) (3) (4) (5) (6) (7)
FPUC period 0.044** -0.262*** -0.306*** -0.413*** -0.334*** -0.287*** -0.190***
(0.018) (0.022) (0.021) (0.023) (0.024) (0.046) (0.032)
Week of year FE X X X X X X X
No. of Obs 1,540,598 1,540,598 1,540,598 613,586 412,050 297,346 217,616
FPUC, Mar-May 0.076*** -0.448*** -0.524*** -0.547*** -0.558*** -0.547*** -0.444***
(0.021) (0.023) (0.021) (0.023) (0.021) (0.048) (0.035)
FPUC, Jun-Jul 0.004 -0.030 -0.034 -0.246*** -0.055 0.039 0.126***
(0.019) (0.022) (0.025) (0.028) (0.040) (0.054) (0.037)
Week of year FE X X X X X X X
No. of Obs 1,540,598 1,540,598 1,540,598 613,586 412,050 297,346 217,616
Notes: This Table reports changes in the count of job applications, the count of new posted
vacancies, and labor market tightness during the period of the FPUC, relative to their baseline
levels in Jan-Feb 2020. In the upper panel, we present estimates for the total period of FPUC,
while we divide the FPUC period in two sub-periods in the lower panel. The estimates are
obtained by regressing each logged outcome on dummy variables for the FPUC period and for
all calendar weeks outside of the baseline period of January-February 2020, in a panel at the
calendar week and local labor market level. We include week-of-year fixed effects to control for
seasonal variation. The estimates are obtained in the full sample in col. (1)-(3), and separately
for local labor markets in each quartile of the distribution of potential UI increase in col. (4)-(7).
We use weights to reflect the proportion of the labor force in each State×industry×occupation
in the CPS. Robust standard errors clustered at the state level are in parenthesis (* p< 0.10, **
p< 0.05, *** p< 0.010).
11
In JOLTS, the hiring rate was above its pre-COVID levels in June-July 2020, in line with our findings.
16
5.2 Heterogeneity
One could be concerned that recruitment difficulties increased in the local labor markets
most affected by FPUC. We present the evolution of tightness by quartile of the increase in
replacement rate due to FPUC, in columns 4-7 of Table 3 and graphically in Figure A.10.
We show that even in the markets most affected by FPUC, seasonally adjusted tightness
was 19% lower during FPUC than at baseline (Table 3, col. 7). During the re-opening
phase, tightness for the fourth quartile was slightly higher (12.6%) than at baseline. For the
three lower quartiles (col. 4-6), tightness during reopening remained below or at baseline.
Finally, Figure A.11 and Figure A.12 show that the evolution of applications, vacancies
and labor market tightness are qualitatively similar across sectors. The magnitudes of the
fluctuations differ by sector in expected ways: for example, during the lockdown, vacancies
in sectors that were directly affected like leisure & hospitality and transport & logistics
understandably decreased more than in other sectors. In leisure & hospitality, applications
also declined during the FPUC period, suggesting that workers might have decreased their
search due to infection risk. In construction, which we analyzed above, the evolution of
tightness is similar to that in other sectors.
6 Discussion
6.1 Model of the labor market
We discuss how our results help understand the functioning of the labor market during the
start of the COVID-19 crisis. Although most models predict that more generous unem-
ployment insurance should decrease search, one important question in general equilibrium
models is to which extent unemployment benefits affect vacancy creation, and hence la-
bor market tightness (Landais et al., 2018b). In standard models, higher unemployment
benefits diminish job search and exert upward pressure on wages by raising the outside
option of unemployed workers, thereby discouraging vacancy creation (Pissarides, 2000).
In “job rationing” models, vacancy creation primarily depends on the marginal product of
labor, and is hence unaffected by unemployment insurance (Michaillat, 2012). Our find-
ing that FPUC did not affect vacancy creation is consistent with job rationing. What is
the economic intuition? In the context of the COVID-19 crisis, the marginal product of
a worker might have been pushed down by social distancing measures. Other factors less
specific to the COVID-19 crisis might have also contributed to the job rationing, since we
detect no effect on vacancy creation even after lockdown measures were lifted, and even
for teleworkable occupations. Evidence of job rationing were also found in other recessions
(Lalive et al., 2015; Marinescu, 2017; Landais et al., 2018a).
In job rationing models, unemployment insurance has more limited effects on employ-
ment (Landais et al., 2018b): because UI increases labor market tightness, it decreases
17
the competition among job seekers and raises the returns to each unit of search effort.
Additionally, labor market tightness was particularly low when FPUC was implemented,
such that decreasing search effort likely had limited effects on unemployment. Our find-
ings hence help explain why prior studies have found no effect of FPUC on unemployment
(Bartik et al., 2020; Altonji et al., 2020; Finamor and Scott, 2021; Dube, 2020).
6.2 The impact of FPUC on welfare
Our results speak to aspects of the impact of FPUC on welfare that have to do with labor
market tightness. Landais et al. (2018b) formalize that the welfare impact of unemploy-
ment insurance depends on three elements: first, its impact on labor market tightness;
second, the impact of tightness on welfare, and third, the trade-off between consumption
smoothing and moral hazard. Our paper provides evidence on the first two. We show that
FPUC increased tightness, and that tightness was particularly low at that time. Our results
together suggest that FPUC increased welfare by bringing labor market tightness back to-
wards a more efficient level. Intuitively, it strongly decreased the cost of unemployment
for job seekers without a commensurate increase in recruitment difficulties for employers.
Importantly, this conclusion also holds for the reopening phase (June-July 2020). In this
period, one could have expected a clearer negative impact of FPUC on vacancy creation,
since economic activity was no longer hindered by lockdown measures. However, we find
no evidence of this. One could also have feared labor market tightness to be particularly
high during this period, since firms’ recruiting was picking up while many unemployed
workers were receiving FPUC. We show that labor market tightness was just back to
its pre-pandemic level, despite the positive effect of FPUC on tightness. In the absence
of FPUC, the counterfactual level of tightness would likely have remained lower than in
baseline.
While our analysis focuses on labor market tightness, one should also consider other
important elements to determine the welfare impact of FPUC. During the COVID-19
crisis, the sanitary situation made unemployment more desirable than in normal times, to
the extent that unemployed workers were less likely to get infected with COVID-19 and
infect others (Fang et al., 2020). Moreover, workers appeared to rely heavily on FPUC to
sustain their consumption, and FPUC likely boosted employment through a stimulus effect
(Ganong et al., 2021).12 Overall, these elements all point towards the desirability of high
unemployment insurance benefits during this crisis.
12
We don’t capture stimulus effects with our identification strategy, as they don’t necessarily affect jobs
where workers experience the largest increase in potential replacement rates.
18
7 Conclusion
During the COVID-19 crisis, a 10% increase in unemployment benefits due to the Federal
Unemployment Pandemic Assistance (FPUC) led to a 3.6% decline in Glassdoor job ap-
plications. Since the FPUC had no effect on vacancies, its effect on applications led to a
commensurate 3.3% increase in labor market tightness (vacancies/applications). However,
despite the negative effect of FPUC on applications, seasonally adjusted aggregate job ap-
plications remained fairly stable during the FPUC period. With a 4% increase in aggregate
applications and a steep 26% decline in job vacancies relative to January-February 2020,
seasonally adjusted labor market tightness declined by 31% during the FPUC period. Even
in the top 25% of labor markets with the highest increase in the level of unemployment
benefits, tightness still declined by 19% during the FPUC period. By decreasing applica-
tions, FPUC has thus contributed to lifting up a depressed labor market tightness, likely
increasing welfare. In a context of excessive competition for jobs among workers, more
generous unemployment insurance reduces wasteful applications and has little effect on
employment. Our results can help explain why higher unemployment benefits had no effect
on employment after the CARES Act (Bartik et al., 2020; Altonji et al., 2020; Dube, 2020).
19
Bibliography
Altonji J, Contractor Z, Finamor L, Haygood R, Lindenlaub I, Meghir C, O’Dea C, Scott
D, Wang L, Washington E. 2020. Employment Effects of Unemployment Insurance
Generosity During the Pandemic : 24.
Bartik A, Bertrand M, Lin F, Rothstein J, Unrath M. 2020. Measuring the Labor Market
at the Onset of the COVID-19 Crisis. SSRN Electronic Journal ISSN 1556-5068.
URL https://www.ssrn.com/abstract=3633053
Birinci S, Karahan F, Mercan Y, See K. 2021. Labor market policies during an epidemic.
Journal of Public Economics 194: 104348. ISSN 0047-2727.
URL https://www.sciencedirect.com/science/article/pii/S0047272720302127
Boar C, Mongey S. 2020. Dynamic Trade-offs and Labor Supply Under the CARES Act.
Technical Report w27727, National Bureau of Economic Research.
URL https://www.nber.org/papers/w27727
Campello M, Kankanhalli G, Muthukrishnan P. 2020. Corporate Hiring under COVID-
19: Labor Market Concentration, Downskilling, and Income Inequality. SSRN Scholarly
Paper ID 3596655, Social Science Research Network, Rochester, NY.
URL https://papers.ssrn.com/abstract=3596655
Chamberlain A, Zhao D. 2019. Glassdoor Job Market Report Methodology. Technical
report, Glassdoor.
URL https://www.glassdoor.com/research/glassdoor-job-market-report-methodology/
Cheng W, Carlin P, Carroll J, Gupta S, Rojas FL, Montenovo L, Nguyen TD, Schmutte IM,
Scrivner O, Simon KI, Wing C, Weinberg B. 2020. Back to Business and (Re)employing
Workers? Labor Market Activity During State COVID-19 Reopenings. Working Paper
27419, National Bureau of Economic Research.
URL http://www.nber.org/papers/w27419
Chetty R, Friedman J, Hendren N, Stepner M, Team TOI. 2020. The Economic Impacts
of COVID-19: Evidence from a New Public Database Built Using Private Sector Data.
Technical Report w27431, National Bureau of Economic Research, Cambridge, MA.
URL http://www.nber.org/papers/w27431.pdf
DellaVigna S, Heining J, Schmieder JF, Trenkle S. 2020. Evidence on Job Search Models
from a Survey of Unemployed Workers in Germany : 49.
Dingel JI, Neiman B. 2020. How many jobs can be done at home? Journal of Public
Economics 189: 104235. ISSN 0047-2727.
URL https://www.sciencedirect.com/science/article/pii/S0047272720300992
Dube A. 2020. The Impact of the Federal Pandemic Unemployment Compensation on
Employment: Evidence from the Household Pulse Survey : 7.
Fairlie RW, Couch K, Xu H. 2020. The Impacts of COVID-19 on Minority Unemployment:
First Evidence from April 2020 CPS Microdata. Working Paper 27246, National Bureau
of Economic Research.
URL http://www.nber.org/papers/w27246
20
Fang L, Nie J, Xie Z. 2020. Unemployment Insurance During a Pandemic. SSRN Scholarly
Paper ID 3666545, Social Science Research Network, Rochester, NY.
URL https://papers.ssrn.com/abstract=3666545
Finamor L, Scott D. 2021. Labor market trends and unemployment insurance generosity
during the pandemic. Economics Letters 199: 109722. ISSN 0165-1765.
URL http://www.sciencedirect.com/science/article/pii/S0165176520304821
Flood S, Warren RJ, King M, Rodgers R. 2020. Integrated Public Use Microdata Series,
Current Population Survey: Version 8.0 [dataset].
URL https://doi.org/10.18128/D030.V8.0
Forsythe E, Kahn LB, Lange F, Wiczer D. 2020a. Labor demand in the time of COVID-
19: Evidence from vacancy postings and UI claims. Journal of Public Economics 189:
104238. ISSN 0047-2727.
URL https://www.sciencedirect.com/science/article/pii/S004727272030102X
Forsythe E, Kahn LB, Lange F, Wiczer D. 2020b. Labor demand in the time of COVID-
19: Evidence from vacancy postings and UI claims. Journal of Public Economics 189:
104238. ISSN 0047-2727.
URL https://www.sciencedirect.com/science/article/pii/S004727272030102X
Forsythe E, Kahn LB, Lange F, Wiczer DG. 2020c. Searching, Recalls, and Tightness:
An Interim Report on the COVID Labor Market. Technical Report w28083, National
Bureau of Economic Research.
URL https://www.nber.org/papers/w28083
Fradkin A, Baker S. 2017. The impact of unemployment insurance on job search: Evidence
from google search data. Review of Economics and Statistics 99: 756–768. Tex.date-
added: 2018-06-07 15:57:53 +0000 tex.date-modified: 2018-06-07 15:59:49 +0000.
Ganong P, Greig F, Liebeskind M, Noel P, Sullivan DM, Vavra J. 2021. Spending and Job
Search Impacts of Expanded Unemployment Benefits: Evidence from Administrative
Micro Data : 72.
Ganong P, Noel P, Vavra J. 2020. US Unemployment Insurance Replacement Rates During
the Pandemic. SSRN Scholarly Paper ID 3601492, Social Science Research Network,
Rochester, NY.
URL https://papers.ssrn.com/abstract=3601492
Gupta S, Montenovo L, Nguyen TD, Rojas FL, Schmutte IM, Simon KI, Weinberg BA,
Wing C. 2020. Effects of Social Distancing Policy on Labor Market Outcomes. Working
Paper 27280, National Bureau of Economic Research.
URL http://www.nber.org/papers/w27280
Hensvik L, Le Barbanchon T, Rathelot R. 2020. Job Search during the COVID-19 Crisis.
SSRN Scholarly Paper ID 3598126, Social Science Research Network, Rochester, NY.
URL https://papers.ssrn.com/abstract=3598126
Karabarbounis M, Pinto S. 2018. What Can We Learn from Online Wage Postings? Ev-
idence from Glassdoor. SSRN Scholarly Paper ID 3322205, Social Science Research
Network, Rochester, NY.
URL https://papers.ssrn.com/abstract=3322205
21
Krueger AB, Mueller A. 2010. Job search and unemployment insurance: New evidence
from time use data. Journal of Public Economics 94: 298–307.
URL http://ideas.repec.org/a/eee/pubeco/v94y2010i3-4p298-307.html
Lalive R, Landais C, Zweimueller J. 2015. Market Externalities of Large Unemployment
Insurance Extension Programs. American Economic Review 105(12): 3564–3596.
Landais C, Michaillat P, Saez E. 2018a. A Macroeconomic Approach to Optimal Unem-
ployment Insurance: Applications. American Economic Journal: Economic Policy 10:
182–216. ISSN 1945-7731, 1945-774X.
URL https://pubs.aeaweb.org/doi/10.1257/pol.20160462
Landais C, Michaillat P, Saez E. 2018b. A Macroeconomic Approach to Optimal Unemploy-
ment Insurance: Theory. American Economic Journal: Economic Policy 10: 152–181.
ISSN 1945-7731.
URL https://www.aeaweb.org/articles?id=10.1257/pol.20150088
Lichter A, Schiprowski A. 2020. Benefit duration, job search behavior and re-employment.
CESifo Working Papers Tex.date-added: 2017-12-12 13:36:52 +0000 tex.date-modified:
2020-05-14 14:57:32 +0000.
Marinescu I. 2017. The general equilibrium impacts of unemployment insurance: Evidence
from a large online job board. Journal of Public Economics 150: 14–29. ISSN 0047-2727.
URL http://www.sciencedirect.com/science/article/pii/S0047272717300348
Marinescu I, Skandalis D. 2021. UNEMPLOYMENT INSURANCE AND JOB SEARCH
BEHAVIOR. THE QUARTERLY JOURNAL OF ECONOMICS : 45.
Marinescu IE, Wolthoff R. 2019. Opening the Black Box of the Matching Function: the
Power of Words. Journal of Labor Economics ISSN 0734-306X.
URL https://www.journals.uchicago.edu/doi/abs/10.1086/705903
Michaillat P. 2012. Do Matching Frictions Explain Unemployment? Not in Bad Times.
American Economic Review 102: 1721–50.
Mitman K, Rabinovich S. 2021. Whether, When and How to Extend Unemployment
Benefits: Theory and Application to COVID-19 : 41.
Montenovo L, Jiang X, Rojas FL, Schmutte I, Simon K, Weinberg B, Wing C. 2020.
Determinants of Disparities in Covid-19 Job Losses. Technical Report w27132, National
Bureau of Economic Research, Cambridge, MA.
URL http://www.nber.org/papers/w27132.pdf
Petrosky-Nadeau N. 2020. Reservation Benefits: Assessing Job Acceptance Impacts of In-
creased UI Payments. Technical Report 2020-28, Federal Reserve Bank of San Francisco.
URL https://www.frbsf.org/economic-research/publications/working-papers/
2020/28/
Pissarides C. 2000. Equilibrium Unemployment Theory. Cambridge: MIT Press., 2nd ed.
edition.
Schmieder JF, von Wachter T. 2016. The Effects of Unemployment Insurance Benefits:
New Evidence and Interpretation. Annual Review of Economics 8: 547–581.
URL https://doi.org/10.1146/annurev-economics-080614-115758
22
Appendix
Figure A.1: Descriptive statistics on the state of the labor market when FPUC was voted
Notes: This Figure shows the evolution of the number of UI claims (new and continuing) in the
upper panel. In the lower panel, it shows the evolution of employment rates among workers at
different parts of the wage distribution (low wage, middle wage or high wage), relative to their
levels in January 2020. The data for the lower panel are described in Chetty et al. (2020), and
made publicly available on the website https : //tracktherecovery.org.
1
Figure A.2: Comparison of the measure of new vacancies from JOLTS and Glassdoor
Notes:The Figure presents the variation in the count of new job listing at the monthly level,
relative to the average level in January-February 2020. We compare the measure of new vacancies
from JOLTS, to the measure from Glassdoor data—which we use in the rest of the paper. For
the measure from Glassdoor, we present the weighted sum of job listings, using weights reflecting
the size of each State×industry×occupation in the labor force in the CPS.
2
Table A.1: Descriptive statistics: Count of applications and job postings on Glassdoor by
industry
Data source Glassdoor data CPS
Variable Application count Job posting count US labor
per week×industry per week×industry force
Mean % Mean % %
Agriculture 15 0.00 7 0.00 1.59
Business Service 81,112 17.78 84,050 13.84 12.42
Construction 4,420 0.97 3,776 0.62 7.30
Consumer Service 6,447 1.41 8,135 1.34 4.82
Finance 49,031 10.75 39,249 6.46 6.80
Government 7,047 1.55 8,421 1.39 4.67
Health, Education 68,506 15.02 164,691 27.12 22.51
Information 105,109 23.05 65,283 10.75 1.80
Leisure, hospitality 30,366 6.66 72,982 12.02 9.34
Manufacturing 46,673 10.23 33,052 5.44 9.88
Retail 43,129 9.46 69,722 11.48 12.92
Transportation, warehousing 14,217 3.12 57,889 9.53 5.66
Notes: This Table presents the mean weekly count of applications and new job postings and the
proportion of applications and new vacancy postings by industry, in our study sample. In the
last column, the Table reports the proportion of US labor force by industry based on the CPS in
2018-2020.
3
Table A.2: Descriptive statistics in our study sample
Mean SD p25 p50 p75
(1) (2) (3) (4) (5)
Count of applications, each week in each local labor market 82.2 204.3 12.0 26.0 70.0
Count of vacancies, each week in each local labor market 189.6 659.1 16.0 44.0 126.0
Tightness (vacancies / applications) 4.7 22.6 0.7 1.4 3.4
Replacement rate without FPUC 43.6 8.4 38.2 45.4 49.2
Replacement rate with FPUC 146.1 47.9 107.3 143.5 173.8
Notes: This Table presents descriptive statistics in our study sample. All variables are computed
for each calendar week from January 2018 to July 2020, in each local labor market. We use weights
reflecting the size of each State×industry×occupation in the labor force in the CPS. Note that
the distribution of potential replacement rates in local labor markets in this Table is not directly
comparable to the distribution of replacement rate in the population of unemployed workers
computed in Ganong et al. (2020): the level of observation is different since one observation
represents one local labor market in our analysis, while one observation represents one worker
who lost their job during the COVID-19 crisis in Ganong et al. (2020). However, we note that
these statistics are relatively close, with the first quartile of replacement rate ranging around
103%, the median around 145%, third quartile around 195% (Ganong et al., 2020).
4
Figure A.3: Variation in the residualized increase in replacement rate in our data
(1) After including (2) After including (3) After including
State FE State and industry FE State, industry, exposure decile FE
5
Notes: This Figure illustrates the variation in the residualized increase in replacement rate in our data, which serves to identify the effect of FPUC.
The Figure shows that increases in replacement rates are negatively correlated with pre-displacement weekly earnings, and that substantial variation
in increase in replacement rate remains even once we take out the portion explained by state, industry, and exposure decile. They are obtained in
the cross section of all the local labor market (state by occupation by industry by wage decile) in our study sample. We use weights to reflect the
proportion of the labor force in each State×industry×occupation in the CPS, like in the rest of the analysis. As an example, we show in red the
observations corresponding to local labor markets for the occupation “salesmen”.
Table A.3: The probability of unemployment in January-March 2020 depending on past
earnings, by industry
Pba Unemployed
(1) (2) (3)
ln(prior earnings) -0.372***
(0.024)
Agriculture × ln(prior earnings) -0.525*** -0.541***
(0.198) (0.208)
Mining × ln(prior earnings) -0.061 -0.047
(0.393) (0.395)
Construction × ln(prior earnings) -0.096 -0.097
(0.088) (0.089)
Manufacturing × ln(prior earnings) -0.531*** -0.553***
(0.097) (0.096)
Retail × ln(prior earnings) -0.433*** -0.449***
(0.081) (0.081)
Transportation × ln(prior earnings) -0.521*** -0.539***
(0.111) (0.110)
Information × ln(prior earnings) -0.248* -0.236*
(0.130) (0.134)
Finance × ln(prior earnings) -0.266*** -0.277***
(0.057) (0.059)
Business services × ln(prior earnings) -0.388*** -0.403***
(0.066) (0.067)
Education & health × ln(prior earnings) -0.360*** -0.370***
(0.043) (0.045)
Leisure & hospitality × ln(prior earnings) -0.262*** -0.264***
(0.087) (0.088)
Consumer services × ln(prior earnings) -0.657*** -0.663***
(0.132) (0.137)
Government × ln(prior earnings) -0.603*** -0.575***
(0.136) (0.137)
Industry FE - X X
State FE - - X
No. of Obs 54,023 54,017 54,017
Notes: This table reports the estimates from the logit regressions of a dummy for unemploy-
ment on local labor market characteristics: prior earnings, two-digit industries and states. The
information on workers’ unemployment status, industry, and state come from the monthly CPS
data for January-March 2020, and information on their prior earnings come from the Outgoing
Rotation Group data for 2019. The two datasets are matched at the individual level. We use indi-
vidual survey weights from the CPS 2020. Note that as we use the information on earnings from
ORG 2019, we could also have computed alternative weights accounting for the composition of
respondents in ORG 2019 and correcting for missing earnings values. We use the standard survey
weights for replicability, but we checked that this does not affect our results. Robust standard
errors in parenthesis (* p< 0.10, ** p< 0.05, *** p< 0.010).
6
Table A.4: Additional results on the impact of FPUC on vacancies and tightness
Vacancies (ln) Tightness (ln)
(1) (2) (3) (4) (5) (6) (7) (8)
ln( RepRep
F P UC
)×FPUC period, March -0.123 0.021
(0.075) (0.084)
ln( RepRep
F P UC
)×FPUC period, April -0.247** 0.138
(0.115) (0.133)
ln( RepRep
F P UC
)×FPUC period, May -0.011 0.425***
(0.069) (0.106)
ln( RepRep
F P UC
)×FPUC period, June 0.026 0.421***
(0.087) (0.133)
ln( RepRep
F P UC
)×FPUC period, July 0.111 0.437***
(0.092) (0.132)
ln( RepRep
F P UC
)×FPUC period×Lockdown -0.109 0.246**
(0.066) (0.094)
ln( RepRep
F P UC
)×FPUC period×Open 0.063 0.430***
(0.084) (0.126)
7 ln( RepRep
F P UC
)×FPUC period × Telework 0.079 0.461***
(0.098) (0.153)
ln( RepRep
F P UC
)×FPUC period × Not telework -0.066 0.245**
(0.077) (0.114)
FPUC period × Telework -0.096 -0.227
(0.083) (0.150)
ln( RepRep
F P UC
)×FPUC period × High temp layoff 0.044 0.217
(0.114) (0.139)
ln( RepRep
F P UC
)×FPUC period × Low temp layoff -0.148 0.267**
(0.090) (0.118)
FPUC period × High temp layoff -0.193 0.168
(0.171) (0.179)
P-Value Test May = July .047 .852
P-Value Test Lockdown = Open .004 .008
P-Value Test Telework = Not Telework .062 .116
P-Value Test High temp layoff = Low temp layoff .178 .741
Exposure decile×Time FE, Industry×Time FE, State×Time FE, X X X X X X X X
Labor market×Week of year FE X X X X X X X X
No. of Obs 1,523,252 1,523,252 1,496,988 1,523,252 1,523,252 1,523,252 1,496,988 1,523,252
Notes: This Table reports the estimates from the regression of logged vacancies and tightness on various independent variables, in a panel at the
calendar week and local labor market level—where local labor markets are defined as as the interaction of state, 2-digit occupations and 2-digit
industries and wage deciles. We use weights to reflect the proportion of the labor force in each State×industry×occupation in the CPS. Robust
standard errors clustered at the state level are in parenthesis (* p< 0.10, ** p< 0.05, *** p< 0.010).
Figure A.4: The impact of FPUC on applications week by week
(1) No additional controls (2) Labor Market ×week-of-year FE
(3) Labor Market ×week-of-year (4) Labor Market ×week-of-year
+ State ×time FE + State ×time + Industry ×time FE
(5) Labor Market ×week-of-year (6) Labor Market ×week-of-year
+ State ×time FE + State ×time + Industry ×time FE
+ Exposure decile ×time FE + Exposure decile ×time FE
Notes: This Figure reports the estimates from regressions of logged count of applications on
the interaction of each week around the enactment of the FPUC and the potential increase in
UI generosity ln( RepRep
F P UC
). We use a panel at the calendar week and local labor market level
—where local labor markets are defined as as the interaction of state, 2-digit occupations and
2-digit industries and wage deciles. All models include calendar week and local labor market fixed
effects. We also add various controls in different panels. We use weights to reflect the proportion of
the labor force in each State×industry×occupation in the CPS. Thin lines denote 95% confidence
intervals, based on robust SE clustered at the state level.
8
Figure A.5: The impact of FPUC on new vacancies week by week
(1) No additional controls (2) Labor Market ×week-of-year FE
(3) Labor Market ×week-of-year (4) Labor Market ×week-of-year
+ State ×time FE + State ×time + Industry ×time FE
(5) Labor Market ×week-of-year (6) Labor Market ×week-of-year
+ State ×time FE + State ×time + Industry ×time FE
+ Exposure decile ×time FE + Exposure decile ×time FE
Notes: This Figure reports the estimates from regressions of logged count of new vacancies on
the interaction of each week around the enactment of the FPUC and the potential increase in
UI generosity ln( RepRep
F P UC
). We use a panel at the calendar week and local labor market level
—where local labor markets are defined as as the interaction of state, 2-digit occupations and
2-digit industries and wage deciles. All models include calendar week and local labor market fixed
effects. We also add various controls in different panels. We use weights to reflect the proportion of
the labor force in each State×industry×occupation in the CPS. Thin lines denote 95% confidence
intervals, based on robust SE clustered at the state level.
9
Figure A.6: The impact of FPUC on labor market tightness, week by week
(1) No additional controls (2) Labor Market ×week-of-year FE
(3) Labor Market ×week-of-year (4) Labor Market ×week-of-year
+ State ×time FE + State ×time + Industry ×time FE
(5) Labor Market ×week-of-year (6) Labor Market ×week-of-year
+ State ×time FE + State ×time + Industry ×time FE
+ Exposure decile ×time FE + Exposure decile ×time FE
Notes: This Figure reports the estimates from regressions of logged labor market tightness (va-
cancies/applications) on the interaction of each week around the enactment of the FPUC and the
potential increase in UI generosity ln( RepRep
F P UC
). We use a panel at the calendar week and local
labor market level —where local labor markets are defined as as the interaction of state, 2-digit
occupations and 2-digit industries and wage deciles. All models include calendar week and local
labor market fixed effects. We also add various controls in different panels. We use weights to
reflect the proportion of the labor force in each State×industry×occupation in the CPS. Thin
lines denote 95% confidence intervals, based on robust SE clustered at the state level.
10
Table A.5: Robustness checks on the impact of FPUC on applications, vacancies and
tightness
Applications Vacancies Tightness Applications Vacancies Tightness
(ln) (ln) (ln) (ln) (ln) (ln)
(1) (2) (3) (4) (5) (6)
Q2 of ln( RepRep
F P UC
)×FPUC period -0.113*** -0.022 0.091
(0.029) (0.046) (0.058)
Q3 of ln( RepRep
F P UC
)×FPUC period -0.184*** 0.005 0.189***
(0.037) (0.042) (0.066)
Q4 of ln( RepRep
F P UC
)×FPUC period -0.188*** 0.007 0.195***
(0.046) (0.037) (0.065)
(RepF P U C − Rep)×FPUC period -0.218*** -0.013 0.205***
(0.032) (0.036) (0.045)
Exposure decile×Time FE X X X X X X
Industry×Time FE X X X X X X
State×Time FE X X X X X X
Labor market×Week of year FE X X X X X X
No. of Obs 1,523,252 1,523,252 1,523,252 1,523,252 1,523,252 1,523,252
Notes: This Table reports the estimates from the regression of logged outcomes on various in-
dependent variables, in a panel at the calendar week and local labor market level—where local
labor markets are defined as as the interaction of state, 2-digit occupations and 2-digit industries
and wage deciles. Rep is the replacement rate. We use weights to reflect the proportion of the
labor force in each State×industry×occupation in the CPS. Robust standard errors clustered at
the state level are in parenthesis (* p< 0.10, ** p< 0.05, *** p< 0.010).
11
Figure A.7: The impact of FPUC on applications, vacancies and tightness in the construction sector, week by week
Applications Vacancies Labor market tightness
Notes: This Figure reports the estimates from regressions of the logged count of applications, the logged count of vacancies, and the logged labor
12
market tightness (i.e. ratio of vacancies over applications) in the construction sector on the interaction of each week around the enactment of the
FPUC and the potential increase in UI generosity ln( RepRep F P UC
). The coefficients before the enactment of the FPUC help test our identification
assumption. The coefficients after represent the elasticity of the outcome with respect to benefits levels. We use a panel at the calendar week and
local labor market level—where local labor markets are defined as as the interaction of state, 2-digit occupations and 2-digit industries and wage
deciles. We include calendar week and local labor market fixed effects. We control for seasonality using week-of-year×local labor market fixed effects.
We use weights to reflect the proportion of the labor force in each State×industry×occupation in the CPS. Thin lines denote 95% confidence intervals,
based on robust SE clustered at the state level.
Figure A.8: Changes in applications, vacancies, and labor market tightness relative to Jan-Feb 2020, not seasonally adjusted
Without week of year FE With week of year FE
13
Notes:The Figure presents the changes in tightness (ratio of the weekly count of new posted vacancies over weekly count of applications), relative to
their baseline levels in Jan-Feb 2020. We regress the logged tightness on the time coefficients, and then subtract the average of the estimates obtained
for the period Jan-Feb 2020, such that each coefficient represents the relative variation with respect to the average level in Jan-Feb 2020. We use
weights to reflect the proportion of the labor force in each State×industry×occupation in the CPS.
Figure A.9: Weekly changes in applications, vacancies and tightness relative to Jan-Feb, each year
Applications Vacancies Tightness
Notes: The Figure presents the weekly changes in the count of applications, of new posted vacancies and of labor market tightness (ratio of the weekly
14
count of new posted vacancies over weekly count of applications), relative to their baseline levels in Jan.-Feb., such that each coefficient represents
the relative variation with respect to the average level in Jan.-Feb.. We present them separately for the years 2018, 2019 and 2020. We use weights
to reflect the proportion of the labor force in each State×industry×occupation in the CPS.
Table A.6: Changes in labor market tightness during the period of the FPUC relative to
Jan-Feb 2020, without accounting for seasonal variation
Potential increase in UI
Sample: All Q1 Q2 Q3 Q4
Total
Outcome: Applications Vacancies Tightness Tightness Tightness Tightness Tightness
(1) (2) (3) (4) (5) (6) (7)
FPUC period -0.128*** -0.237*** -0.109*** -0.221*** -0.127*** -0.091* 0.004
(0.021) (0.024) (0.024) (0.022) (0.032) (0.046) (0.042)
Week of year FE No No No No No No No
No. of Obs 1,540,598 1,540,598 1,540,598 613,586 412,050 297,346 217,616
FPUC, Mar-May -0.117*** -0.415*** -0.298*** -0.356*** -0.325*** -0.310*** -0.201***
(0.022) (0.023) (0.023) (0.023) (0.024) (0.043) (0.043)
FPUC, Jun-Jul -0.142*** -0.014 0.128*** -0.052** 0.121** 0.183*** 0.260***
(0.022) (0.027) (0.029) (0.026) (0.046) (0.056) (0.045)
Week of year FE No No No No No No No
No. of Obs 1,540,598 1,540,598 1,540,598 613,586 412,050 297,346 217,616
Notes: This Table presents the same estimates as Table 3, except that we do not control for
seasonal variation. The Table reports changes in the count of job applications, the count of new
posted vacancies, and labor market tightness during the period of the FPUC, relative to their
baseline levels in Jan-Feb 2020. In the upper panel, we present estimates for the total period of
FPUC, while we divide the FPUC period in two sub-periods in the lower panel. The estimates
are obtained by regressing each logged outcome on a dummy for the FPUC period, and a dummy
for all calendar weeks outside of the baseline period of January-February 2020 in a panel at the
calendar week and local labor market level. The estimates are obtained in the full sample in col
(1)-(3), and separately for local labor markets in each quartile of the distribution of potential
UI increase in col (4)-(7). We use weights to reflect the proportion of the labor force in each
State×industry×occupation in the CPS. Robust standard errors clustered at the state level are
in parenthesis (* p< 0.10, ** p< 0.05, *** p< 0.010).
15
Figure A.10: Changes in applications, vacancies and labor market tightness relative to
Jan-Feb 2020, by UI increase quartile
First quartile (lowest increase) Second quartile
Third quartile Fourth quartile (largest increase)
Notes: The Figure presents, for each quartile of the increase in the replacement rate of UI, the
seasonally adjusted changes in the total weekly count of applications, the weekly count of new
posted vacancies, and labor market tightness (ratio of vacancies over total applications), relative
to their baseline levels in Jan-Feb 2020. The Figure is obtained by regressing the logged variables
on the calendar week coefficients and week-of-year fixed effects to control for seasonal variation,
and then subtracting to each calendar week coefficient the average of estimates for Jan-Feb 2020.
We use weights to reflect the proportion of the labor force in each State×industry×occupation
in the CPS.
16
Figure A.11: The evolution of job applications & job postings in specific sectors:
Business Services Construction Consumer Services
Finance Government
17
Notes: The Figure presents, fr each industry, the seasonally adjusted changes in the total weekly count of applications, the weekly count of new
posted vacancies, and labor market tightness (ratio of vacancies over total applications), relative to their baseline levels in Jan-Feb 2020. The Figure
is obtained by regressing the logged variables on the calendar week coefficients and week-of-year fixed effects to control for seasonal variation, and
then subtracting to each calendar week coefficient the average of estimates for Jan-Feb 2020. We use weights to reflect the proportion of the labor
force in each State×industry×occupation in the CPS.
Figure A.12: The evolution of job applications & job postings in specific sectors:
Health, education Information Leisure & hospitality
Manufacturing Retail Transport & logistics
18
Notes: The Figure presents, for each industry, the seasonally adjusted changes in the total weekly count of applications, the weekly count of new
posted vacancies, and labor market tightness (ratio of vacancies over total applications), relative to their baseline levels in Jan-Feb 2020. The Figure
is obtained by regressing the logged variables on the calendar week coefficients and week-of-year fixed effects to control for seasonal variation, and
then subtracting to each calendar week coefficient the average of estimates for Jan-Feb 2020. We use weights to reflect the proportion of the labor
force in each State×industry×occupation in the CPS.
File and source
- File
- the-impact-of-the-federal-pandemic-unemployment-compensation-on-job-search-and-vacancy-creation.pdf
- Size
- 1,358,183 bytes
- SHA-256
- 823c79af19a7421962d470e723757f0743f807f1bad144a12c475d77287d3f31
- Our copy
- the-impact-of-the-federal-pandemic-unemployment-compensation-on-job-search-and-vacancy-creation.pdf
- Original
- doi.org