Dynamic Trade-offs and Labor Supply Under the CARES Act — NBER Working Paper 27727

Summary

NBER Working Paper No. 27727, Dynamic Trade-offs and Labor Supply Under the CARES Act, by Corina Boar and Simon Mongey, dated August 2020. The abstract states that the CARES Act resulted in many unemployed workers receiving benefits that exceeded wages at their previous job and asks whether an unemployed worker would reject an offer to return to the former job at the same wage. It states that the authors provide a simple dynamic model with four reasons the answer could be no, and that under empirically relevant parameters they find it unlikely a worker would reject such an offer. The introduction states that without the CARES Act benefits are 50 percent of previous wage, up to a cap of $458 per week, and that with the CARES Act benefits are increased by $600 per week.

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: Report
Case: Technical Report W27727 National Bureau Of Economic Research

Cite this document

“Dynamic Trade-offs and Labor Supply Under the CARES Act — NBER Working Paper 27727.” Original at www.nber.org. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/technical-report-w27727-national-bureau-of-economic-research/, SHA-256 30d7dad87e404f57f58e3c5d78da017377ea163d3439a9f011c88dc6cf5a33c9.

Full text

NBER WORKING PAPER SERIES

DYNAMIC TRADE-OFFS AND LABOR SUPPLY UNDER THE CARES ACT

Corina Boar
Simon Mongey

Working Paper 27727
http://www.nber.org/papers/w27727

NATIONAL BUREAU OF ECONOMIC RESEARCH
1050 Massachusetts Avenue
Cambridge, MA 02138
August 2020

Thanks to Alexander Weinberg for excellent research assistance. 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 They have not been
peer-reviewed or been subject to the review by the NBER Board of Directors that accompanies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.

© 2020 by Corina Boar and Simon Mongey. 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.
Dynamic Trade-offs and Labor Supply Under the CARES Act
Corina Boar and Simon Mongey
NBER Working Paper No. 27727
August 2020
JEL No. E24,J64,J68

ABSTRACT

The CARES Act resulted in many unemployed workers receiving benefits that exceeded wages at
their previous job. Given this, would an unemployed worker reject an offer to return to their
former job at the same wage? Qualitatively, we provide a very simple dynamic model that
incorporates four reasons the answer could be ‘no’: (i) the temporary nature of the CARES Act,
(ii) uncertainty that their return-to-work offer might expire, (iii) search frictions, and (iv) wage
losses out of unemployment in a recession. Quantitatively, when evaluated under empirically
relevant parameters, we find it unlikely a worker would reject an offer to return to work at the
same wage. We show special cases where this is not true and relate these to anecdotal evidence.

Corina Boar
Department of Economics
New York University
19 West 4th Street
New York, NY 10003
and NBER
corina.boar@gmail.com

Simon Mongey
Kenneth C. Griffin Department of Economics
University of Chicago
1126 E. 59th Street
Chicago, IL 60637
and NBER
mongey@uchicago.edu
Figure 1: Unemployment benefits with and without the CARES Act 2020.
Notes: Without the CARES Act benefits are 50 percent of previous wage, up to a cap on benefits of $458 per week. With the
CARES Act, benefits are increased by $600 per week.

1 Introduction

The CARES Act increased the level of unemployment benefits to many workers in the U.S. economy.
Empirical studies to date have found little effect on employment, as we discuss below. However, there
exists anecdotal evidence of workers turning down return-to-work offers in favor of unemployment.
In this paper we provide a simple theory of labor supply in this environment, and the role of different
features of the worker’s job, the labor market and policy. Taken to data, we show that outside of extreme
cases, it is difficult to rationalize workers not accepting an offer to return to work, hence providing
quantitative support for the empirical results thus far.

CARES Act 2020. How did the CARES Act affect replacement rates?1 Ganong, Noel, and Vavra (2020)
provide a useful calculator for computing the level of weekly benefits given previous weekly wages.
Figure 1 plots the level of benefits and the replacement rate both with and without the CARES Act.
Without the CARES Act, unemployment benefits are 50 percent of the previous wage, up to a cap of
$458 per week. With the CARES Act, all individuals receive an additional $600 per week, regardless of
the level of the previous wage. As reported by Ganong, Noel, and Vavra (2020) this leads to a median
replacement rate of 134 percent. They find that around 68 percent of workers have a replacement rate
greater than 100 percent. If a worker faced with a return-to-work offer were to only make a static trade-off
between UI benefits and weekly earnings, 68 percent of workers would not return to work.
1 For the text of the CARES Act see https://www.congress.gov/116/bills/hr748/BILLS-116hr748enr.pdf.

1
Approach. We write down a very simple model that describes the labor supply decision of a worker.
The worker earned a wage w in a previous job and is faced with the decision of whether to return to
work at their previous employer and again earn w or be unemployed under the CARES Act. The return-
to-work offer is not assured, and expires with some probability that we view as capturing how easily
replaceable the worker is. Similarly, the additional benefits of the CARES Act have only a limited du-
ration. If the worker’s return-to-work offer expires, they search for a new job which may take some to
find, and is possibly associated with wage losses, which are common when workers experience unem-
ployment in a recession.

Results. Our main analysis consists of plotting what we call the reservation replacement rate, as a function
of the workers’ previous wage. We define the reservation replacement rate as the replacement rate under
the CARES Act that would make the worker indifferent between accepting the return-to-work offer at
wage w and remaining unemployed. We denote this ρbC (w). We compare this to the actual replacement
rate under the CARES Act ρC (w), which we plotted in Figure 1B. If the reservation replacement rate
exceeds the replacement rate under the CARES Act, then the worker would prefer to return to work.
First, we show two special cases where the reservation replacement rate is equal to one, such that
68 percent of workers would prefer to remain unemployed: (i) when there are no search frictions and
job loss in a recession is associated with no wage losses, and (ii) when the worker is never at risk of
losing their offer to return to work at their previous employer. Second, we show that empirically reason-
able deviations from these special cases lead to a reservation replacement rate that exceeds ρC (w). This
rationalizes the empirical results we discuss below. In other words, workers would only choose unem-
ployment over a return-to-work offer, if the additional unemployment benefits under the CARES Act
were much larger than $600 per week. The reservation replacement rate increases in (i) the probability
that the CARES Act expires, (ii) the probability that the return-to-work offer is lost, (iii) the expected
amount of time until a new job offer arrives if they turn down their return-to-work offer, and (iv) wage
losses relative to their former employment.

Empirical literature. These results are consistent with the empirical literature that has studied the ef-
fects of the CARES Act. Using data from online vacancy postings Marinescu, Skandalis, and Zhao (2020)
find that employers did not struggle to find applicants while the CARES Act was in place. Using data
on plant payroll from Homebase, Altonji, Contractor, Finamor, Haygood, Lindenlaub, Meghir, O’Dea,
Scott, Wang, and Washington (2020) find that workers who faced larger expansions in UI benefits re-
turned to their previous jobs at similar rates to those with smaller expansions in UI benefits. Using the
same data, Bartik, Bertrand, Lin, Rothstein, and Unrath (2020) come to a similar conclusion. Using the

2
high-frequency Census Household Pulse Survey, Dube (2020) finds that across-state variation in replace-
ment rates did not affect employment across states.

Quantitative literature. Our short paper complements recent papers that integrate the labor supply
decisions of workers in frictional labor markets with models of disease transmission. A key theme of
this work is that increased unemployment benefits dampen the infection externality associated with
workers’ labor supply decisions. Birinci, Karahan, Mercan, and See (2020) find that the optimal pol-
icy during a pandemic includes both additional UI benefits and payroll subsidies. Gregory, Menzio,
and Wiczer (2020) find that short-term lockdowns have long-lasting negative effects on unemployment.
Tertilt, Brotherhood, Kircher, and Santos (2020) study various policies in a model with incomplete infor-
mation and heterogenous agents. Kapička and Rupert (2020) find that more generous unemployment
benefits can reduce overall infection by 17 percentage points. In this short paper we do not consider
the health consequences of employment, and provide a very simple framework for thinking through the
non-health related incentives toward and against individual labor supply under the CARES Act.
The most related papers are Petrosky-Nadeau (2020) and Mitman and Rabinovich (2020). The former
computes the reservation benefit level required to convince workers in different occupations to turn
down a new job offer and remain in unemployment under the CARES Act. The latter computes the
optimal UI policy in a model where search effort responds to benefits. A key result in Mitman and
Rabinovich (2020) is that tieing benefits to the unemployment rate puts too much history dependence
into UI policy. They find that the optimal replacement rate in the reopening phase is 70 percent. We
consider a slightly a setting in which a worker holds a job offer from their previous employer which has
some risk of expiring. We also consider the possibility of wage losses following unemployment in the
recession, which occurs if the worker’s return-to-work offer expires. We find that the reservation level of
benefits is highly sensitive to these wage losses and that an empirically reasonable degree of wage loss
leads to reservation levels of benefits that exceed what is offered under the CARES Act.

Outline. This paper has four remaining sections. Section 2 describes the model. Section 3 describes our
main results. Section 4 shows how different workers may make different labor supply decisions. Section
5 concludes with caveats and possible avenues for future research.

3
2 Model

2.1 Environment

We consider the problem of a single worker, that earned a wage w in a previous job and is faced with the
decision of whether to go back to that job and earn a wage w or be unemployed under the CARES Act.
Workers discount the future at rate β < 1, and live hand-to-mouth such that consumption c is equal to
their wage w or unemployment benefits b each period, for which they get utility u(c).2
When unemployed under the CARES Act, the worker receives benefits bC (w) that are equal to a
replacement rate ρC (w) times their previous wage, w. The CARES Act expires each period with proba-
bility χ, which we choose such that the expected duration of the Act is 4 months, as in the data. After
the CARES Act expires, benefits are b(w) and are determined by a replacement rate ρ(w).
The return-to-work offer consists of being paid the previous wage w, however with probability δ
the offer permanently expires. If the return-to-work offer expires without being accepted, the worker
searches for a new job. A new job arrives with probability φ ≤ 1, and has a wage that is (1 − ψ) ≤ 1
times the previous wage w. If the worker accepts this job offer, or accepts the return-to-work offer, they
hold their job forever.

2.2 Bellman equations

We are interested in the decision of an unemployed worker choosing between returning to work, or
remaining unemployed under the CARES Act. The worker compares two values:

n o
max V (w) , U1C (w) ,

where V (w) is the presented discounted value of accepting the return-to-work offer, and U1C (w) is the
value of unemployment under the CARES Act. Since we assume that an individual holds a job forever,
then
u (w)
V (w) = u (w) + βV (w) =⇒ V (w) = .
1−β
Let the present discounted value of lifetime utility for an individual that is unemployed under the
CARES Act with a return-to-work offer be given by U1C (w). Then U1C (w) satisfies the following Bell-

2 As shown in Figure 1, replacement rates are highest for low wage workers. Evidence that low wage workers’ consumption

responses to benefits have been consistent with hand-to-mouth behavior is provided by Chetty, Friedman, Hendren, Stepner,
and Team (2020); Cox, Ganong, Noel, Vavra, Wong, Farrell, and Greig (2020); and Coibion, Gorodnichenko, and Weber (2020).

4
man equation:

Uf(w) =u (pC(w)w) +B

(1 — x) (1-6) max {V (w) ,UF(w) } (1)

+ (1—x)6 [pmax {V ((1— p) w) ,UF(w) } + (1-9) UF (w)|
+x (1 = 6) max {V (w) , Un(w)}

+x8 | pmax {V ((1—p) w) ,Ud(w)} + (1-9) unio)

With probability (1 — x7) (1 — 6), nothing changes—the CARES Act remains in place, and the return-to-
work offer is still available—so the worker faces the same decision again tomorrow. With probability
(1 — x) 6 the CARES Act continues but the return-to-work offer expires. The worker finds a new job with
probability ¢, but only receives a wage of (1 — #) w at the new job. The value of unemployment Uf (w)
will reflect the loss of the return-to-work option, and is written below. With probability x the CARES Act
expires, which changes the value of unemployment. The last line combines the expiration of the offer
and the CARES Act.

The remaining unemployment values in (1) are comprised of special cases:

No CARES/No offer: Up(w) = u(p(w)w)+ 6 G max {v (1—yp)w), Up(w) } +(1-@) Up(w)|

No CARES/Offer: U;(w) = u(o(w)w) +B (1 — 6) max {v (w) ,U, (w)}

+6 [pmax {V (1 — p) w) ,Uo(w)} + (1 — $) Uo(w)]

CARES/No offer: US(w) = u(p°(w)w) + Bi (1 —x) |~max {v ((1-y)w), us (w)} +(1-9) Us (w)|

+a [max {V (1 =) w) , Uo(w)} + (1 = $) Uo(w)}

2.3. Reservation replacement rate

We define the reservation replacement rate 0° (w) as the replacement rate under the CARES Act that would
make a worker with previous wage w indifferent between accepting a return-to-work offer and remain-
ing in unemployment:

V(w) = UF (w; °(w)). (2)
The value of unemployment under the CARES Act is increasing in the replacement rate. Therefore, if
the actual replacement rate in the CARES Act ρC,Data (w)—plotted in Figure 1B—is less than ρbC (w), then
the individual would find it optimal to return to work.

3 Results

3.1 Special cases

If the standard replacement rate in absence of the CARES Act, ρ(w), is less than one, then there are two
special cases under which ρbC (w) = 1. The case of ρ(w) < 1 is the empirically relevant case, as shown
in Figure 1B, where the replacement rate without the CARES Act is 50 percent or less. These cases are
independent of all other parameters of the model.

1. No frictions, no wage loss - If φ = 1 and ψ = 0, then even if the worker’s return-to-work offer
expires, they can transition immediately to another job with no wage loss. Therefore the level
of benefits that would make the worker indifferent between taking the return-to-work offer or
remaining unemployed under the CARES Act is equal to their previous wage.

2. Can always return-to-work - If δ = 0, then the worker’s return-to-work offer is always available.
Since ρ(w) < 1, when the CARES Act expires they will always return to work at wage w. Returning
to (1), the value of unemployment under the CARES Act and the value of employment are

h n oi
U1C (w) = u(ρC (w)w) + β χV (w) + (1 − χ) max V (w), U1C (w) ,

V (w) = u(w) + βV (w).

This makes clear that to be indifferent between returning to work and unemployment before the
expiration of the CARES Act, their benefits must equal their previous wage.

Under these special cases, all workers that receive a replacement rate under the CARES Act of
ρC,Data (w) > 1, will choose to remain unemployed as long as the CARES Act has not expired. As noted
by Ganong, Noel, and Vavra (2020), this is around 68 percent of workers. These are the same outcomes as
a static decision. In the first case, no frictions and no wage loss make the future the same as today, which
makes the decision essentially static. In the second case, no loss of the return-to-work offer preserves
today’s outside option into the future, which similarly makes the decision static.

6
3.2 Parameters

We find it instructive to move away from these special cases in steps. We fix some parameters, and then
move {φ, ψ, δ, χ} to empirically reasonable values one by one.

Fixed parameters. A model period is one week, and we set β such that it implies a 5 percent risk-free
rate. The expiry rate of the CARES Act, χ, is set to 1/16 in our benchmark, giving an expected duration
of four months: April-July, 2020. We use log utility: u(c) = log c. We set the replacement rate absent the
CARES Act to replicate Figure 1B: ρ(w) = min { 0.50w , 458 }.

Variable parameters. We consider ranges of the parameters {φ, ψ, δ, χ} as follows. The average
monthly job finding rate at the trough of post-war recessions is around 0.40 (Shimer, 2012). We con-
sider values of φ that deliver steadily worse monthly job finding rates: {0.80, 0.60, 0.40}. The param-
eter ψ governs the loss in lifetime earnings associated with a transition through unemployment in a
recession. The headline result of Davis and von Wachter (2011), is that this value is around 15 per-
cent (see also Huckfeldt, 2016). We consider values of ψ that give steadily worse lifetime wage losses:
ψ ∈ {0.01, 0.02, 0.03}. We turn to the empirically relevant case of ψ = 0.15 at the end. With little data
on recall rates in the current recession, we consider values of δ that give steadily lower recall prospects.
We consider values of δ that deliver probabilities of the return-to-work offer no longer being available
after 4 months of the CARES Act of 0.05, 0.25 and 0.50. We use π to denote this probability.3 Lastly, we
consider values of χ that imply durations of the CARES Act benefits of {4, 8, 12} months.

3.3 Comparative statics

In this subsection we qualitatively compare reservation replacement rates under these different configu-
rations of parameters. In the following section we quantitatively assess the empirically relevant case.

1. Labor market frictions - φ. Figure 2A considers the case where labor market frictions may lead to
protracted unemployment if the return-to-work offer expires before it is accepted. The pink downward
sloping line plots the replacement rate under the CARES Act as a function of the previous wage (cf.
Figure 1B). The remaining lines plot the reservation replacement rate from the model under different values
of φ, which deliver different monthly job finding rates. As φ decreases and the job finding rate falls, the
reservation replacement rate increases from the orange to the blue dashed line.

3 That is, δ = 1 − (1 − π )1/16 , this requires values of δ equal to 0.018, 0.032, and 0.042, respectively.

7
Figure 2: Effects of labor market frictions and persistence of return-to-work offer.
Notes: In both cases wages on job loss are only 1 percent (ψ = 0.01). In Panel A, δ is such that the probability of the return-
to-work offer expires after 4 months is π = 0.05. The parameter φ is varied to deliver different job finding rates, JFR ∈
{0.40, 0.60, 0.80}. In Panel B, φ is such that the monthly job finding rate is 0.80. The parameter δ is varied to deliver different
probabilities of the return-to-work offer expiring after 4 months, π ∈ {0.50, 0.25, 0.05}. The black vertical line gives the median
wage.

Under this parameterization, the worker has a 95 percent chance that their return-to-work offer being
available after four months. Nonetheless, if the offer expires without having been accepted, and the
CARES Act expires, the worker will be faced with the lower replacement rate of the usual unemployment
insurance policy, while they find a new job. As the job finding rate decreases, the expected length of time
on these lower benefits becomes longer, which reduces the value of being unemployed under the CARES
Act. This requires a higher level of benefits under the CARES Act to make the worker indifferent between
unemployment and returning to work.

2. Recall - δ. Figure 2B shows how the reservation replacement rate varies with the probability that
the workers’ return-to-work offer is still available to them in the future. Recall the special case: if δ = 0,
such that the probability the job is always available in four months, then the reservation replacement
rate is equal to one. As δ increases, and thus π increases, the reservation replacement rate increases, as
can be seen from comparing the orange to the blue line. Intuitively, when the return-to-work offer is
only held tenuously, there is a higher probability that turning down the offer will leave the worker in
unemployment when the CARES Act expires. Additionally, with ψ slightly above zero, the possibility of
lower future wages also requires higher benefits under the CARES Act to leave the worker indifferent.
Note that even if ψ were equal to zero, the prospect of lower future benefits after the expiration of the
CARES Act would by itself require a higher reservation replacement rate under the CARES Act.

8
Figure 3: Effect of re-employment wage losses and length of CARES Act
Notes: In both cases φ is such that the monthly job finding rate is 0.80, and δ is such that the probability of the return-to-work
offer expires after 4 months is only 5 percent (π = 0.05). In Panel A, the CARES Act lasts for 4 months, and the parameter
ψ is varied such that reemployment wage losses are {3, 2, 1} percent of the previous wage. In Panel B, reemployment wages
are only 1 percent less than the previous wage (ψ = 0.01), and the parameter χ is varied to change the expected length of the
CARES Act between 4, 8 and 12 months. The black vertical line gives the median wage.

3. Wage losses - ψ. Figure 3A shows how the reservation replacement rate varies with the prospect
of lower future wages in the case that the return-to-work offer is not accepted, and the worker ends up
in a new job after a spell of unemployment. Even with a very high probability that the worker’s job is
still available after four months (π = 0.05), the reservation replacement rate is highly sensitive to re-
employment wages. As the loss in present discounted value of wages, ψ, increases from one percent to
three percent the reservation replacement rate almost doubles. Why such a large increase? The CARES
Act is short in duration, while the scarring effects of job loss in a recession are long-lived. Substantially
higher benefits under the CARES Act are required for a worker to be indifferent between short-term
higher benefits and long-term lower wages.

4. Expiration of CARES Act - χ. Finally, Figure 3B shows how the reservation replacement rate varies
with respect to the expected duration of the CARES Act provisions. If the duration of the CARES Act
were significantly longer, then the reservation replacement rate would also be lower. With a shorter
duration, the reservation replacement rate increases.

9
Figure 4: Reservation replacement rates under empirical values of ψ and φ.
Notes: Here ψ = 0.15, consistent with evidence from Davis and von Wachter (2011) and φ is such that the monthly job finding
rate is 0.40, consistent with evidence from Shimer (2012). The parameter δ is varied to change the probability that the return-
to-work offer expires in 4 months (π ), which is reported in the legend. The black vertical line gives the median wage.

3.4 Main results - Empirically relevant case

We now turn to the empirically relevant case. First, in contrast to the above comparative statics, we
incorporate the effect of job loss in a recession on the present discounted value of lifetime earnings as
estimated by Davis and von Wachter (2011). This is substantially larger than the above cases. We set
ψ = 0.15, which leads to a 15 percent loss in lifetime earnings, as opposed to the value of ψ = 0.01 in
the above exercises. Second, in contrast to the above comparative statics, we incorporate the job finding
rate commonly found in recessions. We set φ to match a 40 percent monthly job finding rate, as opposed
to 80 percent in the above exercises. Without empirical estimates we consider values of δ such that the
probability the return-to-work offer expires within four months is 0.05, 0.25, and 0.50.
Figure 4 plots the reservation replacement rate under these three cases and compares it to the replace-
ment rate under the CARES Act. Our main result is that only a worker with a low previous wage and a
nearly certain return-to-work offer, would turn down their old employer and remain unemployed under
the CARES Act. To put this in context, the left panel—in which the worker faces only a 5 percent chance
that their old job is not available after four months—shows that the reservation replacement rate is less
than the CARES Act replacement rate for workers with hourly wages less than $12.50 an hour. If the
probability of the return-to-work offer remaining available and the wage are positively correlated—such
that low wage jobs have more transitory return-to-work offers—then it may be the case that very few
workers are in this region. The right panel shows that for higher probabilities that the return-to-work
offer expires, reservation replacement rates far exceed that which is offered under the CARES Act.

10
Figure 5: Rationalizing the return to work decisions of two types of workers.
Notes: Here φ is set such that the monthly job finding rate is 0.40, consistent with evidence from Shimer (2012). The four panels
correspond to different levels of wages, which imply different replacement rates under the CARES Act, which we note in the
title of each figure. On the y-axis we vary wage losses on re-employment ψ between zero and 20 percent. On the x-axis we
vary the probability that the worker’s return-to-work offer is no longer available in 4 months time, π, between 0 and 1. The
pink region to the lower-left gives values of (π, ψ) such that the worker would choose to remain unemployed. The green region
to the upper-right gives values of (π, ψ) such that the worker would choose to return to work. The mark for a Dental assistant
corresponds to the median hourly wage of $19.27/hr. The mark for a Food preparation / serving worker (hospitality worker)
corresponds to the median hourly wage of $12.01/hr. In both cases wage losses after unemployment are 3 percent: ψ = 0.03.

4 Two types of workers

Despite our previous results, there is anecdotal evidence that some workers may have chosen to remain
unemployed under the CARES Act. As an example, we have both attended dentists in New York City
and Chicago that reported having trouble getting their Dental Assistants to return to work under the
CARES Act. Meanwhile, we have both found that in reopened coffee shops, most of the usual staff have
returned. The model neatly rationalizes this through the probability that the worker’s return-to-work
offer remains valid.
In Figure 5 we compare a hypothetical Dental Assistant to a hypothetical Food Preparation / Serving
Worker, and also plot the return to work decisions of different types of workers by their initial wage,

11
π and ψ. Consistent with Figure 4, choosing to remain unemployed requires a low previous wage, a
low probability that the return-to-work offer stays available, and wage losses out of unemployment in a
recession that are in general less than the empirically relevant case of ψ = 0.15.
We now turn to our two cases in panels A and C. The median weekly wage of a Food Preparation
/ Serving Worker is $480/week, and that of a Dental Assistant is $770/week.4 We assume low wage
losses out of unemployment are for both workers and set ψ = 0.03; the former due to a minimum wage
that provides a floor to wage losses, and the latter due to specialized skills. We assume both workers
face the same labor market frictions. The key difference is in the probability that the return-to-work offer
expires. The Dental Assistant is hard to replace, and has only a 5 percent chance of their job not being
available after four months (π = 0.05). The hospitality worker is more easily replaceable, and has an 80
percent chance of their job not being available after four months (π = 0.80). As shown in the figures,
the median hospitality worker would return to work, while the median dental assistant would choose
to remain unemployed under the CARES Act and return to work when it expires.5

5 Conclusion

In this short paper we use a simple dynamic model to enumerate reasons workers may return to work
despite having access to unemployment benefits in excess of their previous wage. If workers made a
static comparison of their previous wage to the benefits available under the CARES Act, then 68 percent
of laid off workers would not have returned to work if offered their previous jobs over April to July
2020. A dynamic comparison factors in the limited duration of CARES Act benefits, the possibility that the
option to return to their previous job is rescinded, the fact that finding a new job takes time if their old
job is lost, and the prospect of permanently lower wages after the career setback of being unemployed
in a recession. Quantitatively, once these factors are taken into account, we find that very few workers
would not have returned to work. Only workers with a low previous wage, an almost certain return-to-
work offer, and very low wage losses after unemployment would turn down their old job and remain
unemployed under the CARES Act. Nonetheless, we describe jobs that satisfy these characteristics.
We conclude that high unemployment in April to July 2020 is unlikely to be due to low labor supply
directly linked to the CARES Act. The causes of high unemployment are therefore outside of the scope
of our analysis, and thus more likely due to low labor demand, or low labor supply due to health risks.
In this paper we do not attempt to distinguish the two.
4 See BLS Occupationa and Employment Statistics https://www.bls.gov/oes/current/oes319091.htm for Dental Assistant

and https://www.bls.gov/oes/current/oes319091.htm for Food Preparation / Serving Worker.
5 Not shown in the figure, this also holds for workers in a large part of the wage distribution of each type.

12
References
A LTONJI , J., Z. C ONTRACTOR , L. F INAMOR , R. H AYGOOD , I. L INDENLAUB , C. M EGHIR , C. O’D EA , D. S COTT,
L. WANG , AND E. WASHINGTON (2020): “Employment Effects of Unemployment Insurance Generosity During
the Pandemic,” .

B ARTIK , A. W., M. B ERTRAND , F. L IN , J. R OTHSTEIN , AND M. U NRATH (2020): “Measuring the labor market at
the onset of the COVID-19 crisis,” Working paper, BPEA Conference Drafts.

B IRINCI , S., F. K ARAHAN , Y. M ERCAN , AND K. S EE (2020): “Labor Market Policies During an Epidemic,” .

C HETTY, R., J. N. F RIEDMAN , N. H ENDREN , M. S TEPNER , AND O. I. T EAM (2020): “How Did COVID-19 and
Stabilization Policies Affect Spending and Employment? A New Real-Time Economic Tracker Based on Private
Sector Data,” Working Paper 27431, National Bureau of Economic Research.

C OIBION , O., Y. G ORODNICHENKO , AND M. W EBER (2020): “How Did U.S. Consumers Use Their Stimulus Pay-
ments?,” Working Paper 27693, National Bureau of Economic Research.

C OX , N., P. G ANONG , P. N OEL , J. VAVRA , A. W ONG , D. FARRELL , AND F. G REIG (2020): “Initial impacts of the
pandemic on consumerbehavior: Evidence from linked income, spending,and savings data,” Brookings Papers
on Economic Activity.

D AVIS , S. J., AND T. M. VON WACHTER (2011): “Recessions and the Cost of Job Loss,” Working Paper 17638,
National Bureau of Economic Research.

D UBE , A. (2020): “The Impact of the Federal Pandemic Unemployment Compensationon Employment: Evidence
from the Household Pulse Survey,” .

G ANONG , P., P. N OEL , AND J. S. VAVRA (2020): “US Unemployment Insurance Replacement Rates During the
Pandemic,” Working paper.

G REGORY, V., G. M ENZIO , AND D. W ICZER (2020): “Pandemic Recession: L- or V-Shaped?,” Quarterly Review,
40(01), 1–31.

H UCKFELDT, C. (2016): “Understanding the Scarring Effect of Recessions,” Working paper.

K API ČKA , M., AND P. R UPERT (2020): “Labor Markets during Pandemics,” .

M ARINESCU , I. E., D. S KANDALIS , AND D. Z HAO (2020): “Job Search, Job Posting and Unemployment Insurance
During the COVID-19 Crisis,” .

M ITMAN , K., AND S. R ABINOVICH (2020): “Optimal Unemployment Benefits in the Pandemic,” .

P ETROSKY-N ADEAU , N. (2020): “Reservation Benefits: Assessing Job Acceptance Impacts of Increased UI Pay-
ments,” Discussion paper.

13
S HIMER , R. (2012): “Reassessing the ins and outs of unemployment,” Review of Economic Dynamics, 15(2), 127 –
148.

T ERTILT, M., L. B ROTHERHOOD , P. K IRCHER , AND C. S ANTOS (2020): “An economic model of the Covid-19
epidemic:The importance of testing and age-specific policies,” Discussion paper, CEPR Discussion Paper No.
14695.

File and source

File: technical-report-w27727-national-bureau-of-economic-research.pdf
Size: 968,366 bytes
SHA-256: 30d7dad87e404f57f58e3c5d78da017377ea163d3439a9f011c88dc6cf5a33c9
Our copy: technical-report-w27727-national-bureau-of-economic-research.pdf
Original: www.nber.org