Has the Willingness to Work Fallen during the COVID Pandemic? — IZA Discussion Paper No. 15086

Summary

An IZA Institute of Labor Economics discussion paper, IZA DP No. 15086, titled Has the Willingness to Work Fallen during the COVID Pandemic?, by R. Jason Faberman, Andreas I. Mueller and Ayşegül Şahin, with a suggested citation dated 2022. Using the Job Search Supplement of the Survey of Consumer Expectations for 2013-2021, the authors construct an Aggregate Hours Gap measure of labor market underutilization. The abstract reports that the pandemic rise in that gap essentially disappeared by the end of 2021, while desired work hours fell by 4.6 percent against a 2.3 percent decline in labor force participation. The authors state that ignoring the drop in desired hours overstates underutilization by 2.5 percentage points and that real reservation wages rose 6.2 percent on average. The paper closes with an appendix table of predicted desired hours by labor force status and demographic.

Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used

Full text

Faberman, Jason; Mueller, Andreas I.; Sahin, Aysegül



Working Paper
Has the Willingness to Work Fallen during the COVID
Pandemic?


IZA Discussion Papers, No. 15086


Provided in Cooperation with:
IZA Network @ LISER, Luxembourg Institute of Socio-Economic Research (LISER)




Suggested Citation: Faberman, Jason; Mueller, Andreas I.; Sahin, Aysegül (2022) : Has the Willingness
to Work Fallen during the COVID Pandemic?, IZA Discussion Papers, No. 15086, Institute of Labor
Economics (IZA), Bonn




This Version is available at:
https://hdl.handle.net/10419/252210

Standard-Nutzungsbedingungen:                                          Terms of use:

Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen        Documents in EconStor may be saved and copied for your personal
Zwecken und zum Privatgebrauch gespeichert und kopiert werden.         and scholarly purposes.

Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle       You are not to copy documents for public or commercial purposes, to
Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich   exhibit the documents publicly, to make them publicly available on the
machen, vertreiben oder anderweitig nutzen.                            internet, or to distribute or otherwise use the documents in public.

Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen         If the documents have been made available under an Open Content
(insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten,       Licence (especially Creative Commons Licences), you may exercise
gelten abweichend von diesen Nutzungsbedingungen die in der dort       further usage rights as specified in the indicated licence.
genannten Lizenz gewährten Nutzungsrechte.
DISCUSSION PAPER SERIES




IZA DP No. 15086

Has the Willingness to Work Fallen
during the COVID Pandemic?


R. Jason Faberman
Andreas I. Mueller
Ayşegül Şahin




FEBRUARY
DISCUSSION PAPER SERIES




IZA DP No. 15086

Has the Willingness to Work Fallen
during the COVID Pandemic?


R. Jason Faberman
Federal Reserve Bank of Chicago
Andreas I. Mueller
UT Austin, CEPR, IZA, and NBER
Ayşegül Şahin
UT Austin and NBER



FEBRUARY




Any opinions expressed in this paper are those of the author(s) and not those of IZA. Research published in this series may
include views on policy, but IZA takes no institutional policy positions. The IZA research network is committed to the IZA
Guiding Principles of Research Integrity.
The IZA Institute of Labor Economics is an independent economic research institute that conducts research in labor economics
and offers evidence-based policy advice on labor market issues. Supported by the Deutsche Post Foundation, IZA runs the
world’s largest network of economists, whose research aims to provide answers to the global labor market challenges of our
time. Our key objective is to build bridges between academic research, policymakers and society.
IZA Discussion Papers often represent preliminary work and are circulated to encourage discussion. Citation of such a paper
should account for its provisional character. A revised version may be available directly from the author.




                                           IZA – Institute of Labor Economics
Schaumburg-Lippe-Straße 5–9                      Phone: +49-228-3894-0
53113 Bonn, Germany                             Email: publications@iza.org                                    www.iza.org
IZA DP No. 15086                                                                                                                FEBRUARY




            ABSTRACT
            Has the Willingness to Work Fallen
            during the COVID Pandemic?*
            We examine the effect of the Covid pandemic on willingness to work along both the
            extensive and intensive margins of labor supply. Special survey questions in the Job Search
            Supplement of the Survey of Consumer Expectations (SCE) allow us to elicit information
            about individuals’ desired work hours for the 2013-2021 period. Using these questions,
            along with workers’ actual labor market participation, we construct a labor market
            underutilization measure, the Aggregate Hours Gap (AHG), following Faberman et al.
            (2020). The AHG captures changes in labor market underutilization for the full population
            along both the extensive and intensive margins using data on desired work hours as a
            measure of their potential labor supply. We find that the sharp increase in the AHG during
            the Covid pandemic essentially disappeared by the end of 2021. We also document a
            sharp decline in desired work hours during the pandemic that persists through the end
            of 2021 and is roughly double the drop in the labor force participation rate. Ignoring
            the decline in desired hours overstates the degree of underutilization by 2.5 percentage
            points (12.5%). Our findings suggest that, as of 2021Q4, the labor market is tighter than
            suggested by the unemployment rate and the adverse labor supply effect of the pandemic
            is more pronounced than implied by the labor force participation rate. These discrepancies
            underscore the importance of taking into account the intensive margin for both labor
            market underutilization and potential labor supply.

            JEL Classification:              E24, J21, J60
            Keywords:                        labor market slack, COVID-19, desired work hours, potential
                                             labor supply



            Corresponding author:
            R. Jason Faberman
            Senior Economist and Research Advisor
            Federal Reserve Bank of Chicago
            230 S. LaSalle Dr.
            Chicago, IL 60604
            USA
            E-mail: jfaberman@frbchi.org




            * We thank Jonas Maibom and various conference and seminar participants for their helpful comments and
            Elizabeth Kepner and Levi Bognar for excellent research assistance. The views expressed in this paper are our own
            and do not necessarily reflect the views of the Federal Reserve Bank of Chicago or the Federal Reserve System.
1. Introduction

The Covid pandemic has had a devastating effect on labor markets in the U.S. and
throughout the world. Payroll employment in the U.S. fell by as much as 16 percent and remained
over 2 percent below its February 2020 level at the end of 2021. The unemployment rate more
than quadrupled, rising from 3.5 percent in February 2020 to a peak of 14.7 percent, and was at
3.9 percent at the end of 2021. Several studies, as well as the Bureau of Labor Statistics (BLS, who
publish the unemployment rate), have pointed out that even this large spike in the unemployment
rate likely understates the degree of labor market slack present during this period.! Moreover,
despite a rapid decline in the unemployment rate and relatively strong employment growth since
the initial aggregate shock, the labor force participation rate remains persistently low. Economists
and the popular press have suggested a broad range of potential causes. These include lagging
cyclical adjustment of labor force participation, added child care burdens among women, generous
unemployment benefits and other fiscal stimulus programs, fear of returning to work while the
virus persists, and structural changes in how individuals view work.

The key to our analysis is the Job Search Supplement of the Survey of Consumer
Expectations (SCE). We designed this survey in 2013 and have administered it through the Federal
Reserve Bank of New York annually each October since then (see Faberman, Mueller, Sahin, and
Topa, forthcoming, for details). The supplement asks a wide range of questions on an individual’s
employment situation, work preferences, and job search behavior. We focus on questions in the

supplement that ask respondents about the number of hours they desire to work and their

' See, for example, Barnichon and Yee (2020), Faberman and Rajan (2020), Forsythe et al. (2020), Bick and Blandin
(2021), as well as the documentation on (mis)measuring unemployment by the BLS at

https://www.bls.gov/covid1 9/effects-of-covid-19-pandemic-and-response-on-the-employment-situation-news-
release. htm.

reservation wage. These measures allow us to identify changes in willingness to work over time.
Furthermore, the 2020 and 2021 SCE Job Search Supplements include new questions on how the
pandemic has affected individual job search and labor market participation decisions. Thus, using
the SCE for our analysis provides two major advantages. First, unlike traditional surveys, it
explicitly asks respondents about their potential labor supply and reservation wages; and second,
unlike newer surveys developed during the pandemic, it provides a benchmark period that allows
a comparison with the pre-pandemic labor market.

Desired labor supply combined with actual hours worked also provides an intuitive
measure of labor market underutilization. We refer to this broader underutilization measure as the
Aggregate Hours Gap (AHG) following Faberman, Mueller, Sahin, and Topa (2020, henceforth
FMST). The AHG measures underutilization through the difference between individuals’ desired
work hours and actual hours worked, regardless of labor force status. FMST (2020) show that the
SCE measure of desired hours is strongly related to job search effort and is a good indicator of
potential labor supply. They also show that the AHG captures aspects of labor market slack missed
by traditional measures, like the unemployment rate, and generally has a stronger relationship with
nominal wage growth. Given that it encapsulates a broad concept of labor market underutilization,
the AHG is well-suited to deal with the unique labor market circumstances of the Covid pandemic.
This includes the misclassification of those on furlough or temporary layoff and any reduction in
labor supply and work hours for reasons specific to the Covid pandemic and its related lockdown
orders.

In this paper, we use the SCE data along with the AHG measure to evaluate and corroborate
the impact of the pandemic on labor market underutilization and aggregate labor supply. We find

a diverging pattern between the AHG and the unemployment rate. The AHG suggests that the
labor market is tighter than what the unemployment rate suggests during the Covid pandemic, with

the AHG already below its February 2020 level by the end of 2021. We find that this is driven

primarily by individuals out of the labor force (and partly by part-time workers). During the

pandemic, these individuals report lower desired work hours, and consequently a lower willingness

to participate in the labor market even marginally. Quantitatively, we find that the decline in

desired work hours across all individuals is 4.6 percent, compared to a decline in the labor force

participation rate of 2.3 percent, over this period. The decline occurs with a sharp drop at the onset

of the pandemic that remains persistently low thereafter. Further analysis of the decline in the AHG

and desired hours shows that a higher share of the employed are essentially on their labor supply

curve (in the sense that their actual hours equal their desired hours) and a higher share of those out

of the labor force that are unwilling to work at all drive the observed aggregate patterns. Following

the Great Recession, the pattern was different. The AHG showed a more sluggish labor market

recovery, and therefore a higher degree of labor market underutilization, driven primarily by

individuals out of the labor force who were willing to work a small amount of hours but were slow

to find work.

       To quantify the role of the reduction in labor supply on labor market underutilization, we

perform a counterfactual exercise that holds desired work hours constant at their pre-pandemic

average within detailed labor force and demographic groups and recalculates the AHG and

associated potential work hours from March 2020 forward. The exercise suggests that the decline

in desired work hours reduced the AHG by 2.5 percentage points (12.5 percent) relative to its

estimated value at the end of 2021. It also suggests that essentially all of the discrepancy between

the fall in potential work hours and the labor force participation rate is accounted for by the decline

in desired hours during the pandemic (as opposed to changes in composition). Again, changes in



                                                  4
the desired hours of those out of the labor force and part-time workers account for most of the
differences identified in the counterfactual exercise. Notably, we find only mild differences by
gender in their declines in desired hours—i.e., the result is not driven by a contraction in labor
supply among women which resonates with the findings of Hobijn and Sahin (2021). Instead, we
find the drop in desired hours is pervasive across most demographic groups, with those with less
than a college degree having a notably larger contribution.

Finally, we examine a range of suggestive evidence on the extent that the Covid pandemic
had a direct role on the observed reduction on labor supply. First, we group individuals based on
the degree of social contact required of their current or most recent job, with a higher degree of
social contact implying a higher potential exposure to Covid. We find that individuals in jobs with
at least a moderate degree of social contact had sizable reductions in their desired work hours
during the pandemic, while those in jobs with a low degree of social contact actually increased
their desired work hours. Second, we find that real reservation wages increased during the
pandemic for nearly all labor force and demographic groups, consistent with a decline in
willingness to work. Real reservation wages increased 6 log points (6.2 percent), on average,
across all individuals. Third, using the responses to special Covid-specific questions in the CPS
and SCE, we find that the pandemic caused a sizeable reduction in job search behavior for most of
2020, but that its effect on job search had mostly subsided by the end of 2021. In fact, we find that
job search effort (in terms of job applications sent and the incidence of on-the-job search)
rebounded strongly in 2021.

Our study follows an expansive set of empirical studies on the effects of the Covid
pandemic on the labor market. Many of these studies focus on the initial shock to the labor market

at the onset of the pandemic, while others evaluate the effects of government policies, such as the
expansion of unemployment insurance, on labor supply and employment. 2 Most relevant to our

research are several studies that find labor market tightness was greater than what standard

measures implied because of a reduction in job search effort in the early portion of the pandemic.

These include studies by Forsythe et al. (2020), Marinescu, Skandalis, and Zhao (2020), Brinca,

Duarte, and Faria-e-Castro (2021), and Hensvik, Le Barbanchon, and Rathelot (2021). These

studies identify a reduction in job search effort, either directly or indirectly, as a contributor to

labor market tightness across multiple countries. Our findings are consistent with recent work,

such as Domash and Summers (2022) and Crump et al. (2022), that argue that the U.S. labor market

at the end of 2021 is tighter than in 2019 using alternative measures of labor market tightness. Our

study is also related to research on a declining trend in the willingness to work that predates the

pandemic (Barnichon and Figura, 2016). Finally, our study dovetails with academic research (e.g.,

Bick and Blandin, 2020) and government initiatives, such as the U.S. Household Pulse Survey,

that developed special surveys and survey instruments to study the Covid pandemic.

        The next section describes the conceptual underpinnings of the AHG. Section 3 describes

our data and methodology for estimating the AHG and its components. Section 4 presents the

aggregate time-series behavior of the AHG and its components. Section 5 quantifies the role of

labor supply for the reduction in the AHG and provides supporting evidence on the role of the

Covid pandemic for labor supply and job search. Section 6 concludes.


2. The Aggregate Hours Gap as a Measure of Underutilization

        Labor market underutilization is broadly defined as the ratio of the gap between actual and

potential labor and potential labor supply:


2
 A sampling of these studies include Bartik et al. (2020), Cajner et al. (2020), Ganong, Noel, and Vavra (2020),
Goolsbee and Syverson (2020), and ùahin, Tasci, and Yan (2021).

                                                         6
                                                  ‫݌ܽܩ‬௧
                                                            .                                                    (1)
                                              ‫ݕ݈݌݌ݑܵ ݎ݋ܾܽܮ‬௧

For example, consider the official (U3) measure of the U.S. unemployment rate, produced by the

Bureau of Labor Statistics (BLS). It is equal to the number of unemployed individuals divided by

the total number of individuals in the labor force (employed plus unemployed). The gap is the

number of people who want a job and have actively looked (and are available) for one, or are on

temporary layoff. 3 Total labor supply is measured as the number of people who either have a job

or want a job. 4 Both measures implicitly give all of these individuals a weight of one when

calculating their contribution to labor market underutilization and ignore the variation along the

intensive margin of hours.

         Given the importance of hours as a measure of total labor input in production, one can

define a broader measure, the Aggregate Hours Gap (AHG). FMST (2020) define the measure as:

                                                     σ(‫ܮ‬௜௧ െ ݄௜௧ )
                                          ‫ܩܪܣ‬௧ =                   .                                             (2)
                                                        σ ‫ܮ‬௜௧

         Its denominator is a measure of potential labor supply that aggregates the total amount of

desired work hours at time t across all individuals ݅, σ ‫ܮ‬௜௧ , regardless of their labor force status. Its

numerator is the desired hours gap, σ(‫ܮ‬௜௧ െ ݄௜௧ ), which captures the difference between desired

work hours and actual work hours, ݄௜௧ , and sums this difference across all individuals, regardless

of their labor force status. Individuals who are nonemployed but wish to work add to labor market

underutilization based on the amount of hours they prefer to supply. Those who are employed but

prefer more work hours add to underutilization based on the difference between their current and



3
  Temporary layoffs fit this notion of a gap, because individuals would work their usual hours if they were not on
layoff. This is a particularly important distinction during the Covid pandemic.
4
  This notion of underutilization also holds for the BLS “U6” measure of underutilization. For the U6 measure, the
gap includes all unemployed plus all those who are “marginally attached” to the labor force but not actually a part of
it and those who report that they are part-time rather than full-time for economic reasons.

                                                          7
preferred hours. 5 Consequently, the AHG depends critically on the measures of desired hours and

work hours used in its estimation. FMST (2020) show that a self-reported measure of desired work

hours is a viable measure of labor supply. There is a strong positive relationship between an

individual’s desired work hours and their realized job search effort, and there are intuitive

relationships between individuals’ demographic characteristics, their labor market transitions, and

their desired hours. Thus, the AHG has a clear interpretation as a measure of labor market

underutilization, and the empirical evidence suggests that a direct measure of desired hours is a

valid measure of labor supply. Individual work hours are the obvious measure for ݄௜௧ , but as we

discuss below, the use of usual versus actual hours worked, and how we address measurement

concerns in work hours, are critical for generating consistent estimates of the AHG during the

Covid pandemic. In normal times, the use of usual hours smooths out idiosyncratic hours changes

due to illness, vacations, labor stoppages, and the like. During the pandemic, reported actual hours

worked, despite including such changes for idiosyncratic reasons, also include changes for reasons

specific to the pandemic and therefore capture important measurement issues and labor market

behavior that we want our AHG estimates to address.


3. Data and Measurement

3.1. Data Sources

    We build on FMST (2020) to generate our estimates of the AHG, with several notable

deviations. We rely on two data sources for our analysis. The first is the Current Population Survey

(CPS). The CPS is the survey used to calculate the official U.S. unemployment rate and related


5
  While we mostly abstract from their contribution in this paper, it is worth noting that FMST (2020) show that
individuals who work more than their desired work hours also exert substantial job search effort, suggesting that
their (negative) desired hours gaps are likely also a form of labor underutilization.

                                                          8
labor force statistics. We use the monthly data from January 1994 through December 2021, though

we focus much of our analysis on the most recent years for obvious reasons. 6 We estimate the

share of the total population within detailed labor force states and measure desired hours gaps

within each labor force state using the monthly CPS data.

    Our second data source is the Job Search Supplement to the Survey of Consumer Expectations

(SCE) administered by the Federal Reserve Bank of New York. FMST (Forthcoming) developed

this supplement in earlier work and the supplement has been administered annually each October

since 2013. The supplement asks a broad range of questions on one’s current employment state,

job search activity, employment history, and work preferences (e.g., reservation wage, desired

work hours). These include many questions that are comparable to those in the CPS, allowing us

to directly measure variables related to labor force status, hours worked, and other characteristics

important for our analysis identically across both data sets. Our SCE sample spans 2013 through

2021. We focus on reported differences in desired work hours between the 2013-19 period and the

2020-21 period. Throughout our analysis, it is important to remember that the SCE data are for

October of each year, so the 2020 survey elicits responses about six months after the initial spike

in Covid cases and lockdown period, while the 2021 survey elicits responses between the major

waves of the delta and omicron Covid variants and during a period of relatively strong labor market

growth.

    The 2020 and 2021 SCE Job Search Supplements have additional survey questions that focus

on issues specific to the Covid pandemic. These include follow-ups to its questions on reasons for

not looking for work and reasons for part-time search that elicit whether the Covid pandemic

accounted for these reasons and in what way (e.g., child care issues, fear of contracting the virus,


6
 We only go back to 1994 because it is difficult to produce a consistent measure of our detailed labor force
categories, particularly for those out of the labor force, prior to the 1994 CPS redesign.

                                                          9
caring for someone who was sick, etc.). The new questions also ask about schooling, online

learning, and other aspects of the household that may affect the respondent’s labor supply decision.

We relate this evidence to the behavior of the AHG at the end of our analysis.

    We focus on a sample of individuals aged 18 to 79 with nonmissing data on labor force status

and broad demographics (age, gender, race, education, marital status) since these are the

individuals we can observe in both the SCE and CPS. The CPS is a fairly large sample of about

60,000 households per month. The SCE, however, is much smaller. The Job Search Supplement

averages just under 1,200 respondents per year. 7 We use a sample that pools individuals across all

survey years to generate most estimates of the labor market measures described below, but in our

estimation of the AHG, we split out our desired hours estimates between the 2013-19 and 2020-

21 periods. This requires us to make some additional adjustments to deal with small sample cell

issues.


3.2. Estimating Desired Hours

    Our desired work hours measure comes from the SCE Job Search Supplement. Specifically,

the survey question asks,

    “Assuming you could find suitable/additional work, how many hours PER WEEK would you

    prefer to work on this new job?”

The survey only asks this question to individuals who responded that they actively looked for work

or stated that they would or “might” take a job if offered to them. For those who consequently do

not have a response, we assign them their total usual hours worked if they are employed and zero

desired hours if they are out of the labor force. We do this on the assumption that, for each group,


7
 Complicating our analysis is a relatively low response rate in 2020 due to the pandemic, when the survey has only
965 respondents.

                                                        10
their current hours equal their desired hours since their behavior reflects zero desire to change their

current work situation. These adjustments generally impute a zero desired hours gap to these

individuals. 8

       We match our SCE estimates of desired work hours to individuals in the CPS using their

demographics and detailed labor force status. We divide individuals into one of nine labor force

states. Four of these represent the employed, which we distinguish by whether they are part-time

or full-time, and within each of these categories, whether or not they are a multiple jobholder. We

identify the unemployed based on the standard CPS definition (those who want work, have actively

searched and are available for work, plus those on temporary layoff), and distinguish them by

whether they are short-duration job seekers (looking for 6 months or less) or long-duration job

seekers (looking for more than 6 months). Finally, we distinguish those out of the labor force by

whether they state wanting work (but otherwise fail to meet the criteria for unemployment), are

retired, or are out of the labor force for some other reason (disabled, attending school, or otherwise

not working or not wanting work).

       Within each labor force state, we categorize individuals based on their gender, age, and

education, splitting them into one of three age groups (18 to 24, 25 to 54, and 55 or older) and two

education groups (less than a college degree, and a college degree or more). This approach

potentially creates up to 108 cells of desired hours estimates, but many of these cells are too small

in the SCE data. We therefore follow the approach of FMST (2020) where we aggregate these into

an unbalanced panel of 39 labor force status × demographics cells based on their cell size and

similarity of reported desired hours. All labor force states are at least disaggregated by gender, and

larger categories are disaggregated further by age group and education as the data allow. The most



8
    The exceptions are individuals whose actual hours deviate from their usual hours, which we deal with separately.

                                                           11
disaggregated category is the full-time employed with a single job (10 out of 12 demographic

categories), and the least disaggregated categories are the part-time employed with multiple jobs,

the short-term and long-term unemployed, and those who are out of the labor force but want work

(each only disaggregated by gender).

    We estimate these cells separately for the 2013-19 and 2020-21 periods. This creates within-

cell estimates of desired work hours that vary before and during the pandemic, but are otherwise

time-invariant. Aggregate estimates of desired work hours will additionally vary each month due

to changes in the population shares across labor force states and demographic groups. 9 We report

the desired hours estimates for each specific cell and period in Appendix Table A.1.

        We must also deal with the issue that the 2020 and 2021 surveys, while pooled together,

have relatively small sample cells for several of the labor force status × demographics categories.

We deal with this by generating predicted estimates of desired hours for each individual using a

regression of their reported desired hours on fixed effects for their demographics, nine-state labor

force status, and interactions of their demographics and labor force status with each other and a

dummy variable for the 2020-21 period. We then calculate the mean desired hours for each of our

39 labor force status × demographics categories separately for the 2013-19 and the 2020-21 periods

by calculating the (sample-weighted) mean of the predicted values from this regression. This

approach reduces the effects of outliers and sampling error on the SCE estimates that we match to

the CPS data.

        Formally, we pool ݅ individuals together across the ‫ ݐ‬years of the SCE survey and regress

their reported desired hours, ‫ܮ‬௜௝௧ , on a set of dummy variables for their detailed labor force state


9
  FMST (2020) provide a more thorough examination of the demographic disaggregation of these categories, and
explore the potential effects on the AHG of using time-varying estimates of desired hours that they derive from the
relation of desired hours to aggregate labor market conditions. They find time-varying estimates of desired hours
amplify the cyclicality of the AHG but otherwise preserve its qualitative time-series behavior.

                                                        12
J, their gender, age (using the three categories above), education (high school or less, some college,
college or more), race (White, Black, Hispanic, all other), marital status (married or not), an
interaction between gender and marital status, and indicator for year > 2020, and interactions with
this year indicator with the other dummy variables.'° We also include interactions between each
individual’s broader labor force state (employed, unemployed, out of the labor force) and their
gender, age, education, race, and marital status, but do not further interact these with the year
indicator to avoid overfitting given our relatively small sample size.

Denote the regression’s predicted values for each individual as L; jt- The estimated desired
hours for our set of 39 labor force status x demographics groups is Lg jt» Which is the sample-
weighted mean of L; jt for individuals within that group, and where d represents the demographic

group defined by gender, age, and education.

3.3. Estimating Hours Gaps

We define the desired hours gap as the difference between desired work hours and actual work
hours. We estimate this gap after matching our estimates of desired hours, Lg jt» to individuals in
the CPS by year, detailed labor force status, and demographics (gender, age, and education) for all
individuals in our CPS sample (1.e., all age 18 to 79 between January 1994 and December 2021).
The desired hours gap is then the difference between this estimate and the CPS respondent’s actual

work hours, hjj, which we sum across all jobs reported by individual i.

'0 We topcode our individual-level measures of desired hours in the SCE at 80 hours per week to avoid adverse
effects of any outliers. When we match our estimates to the CPS data, we do the same for total actual hours worked.

13
         We adjust the desired hours at the individual level to impose the constraint of a zero

minimum hours gap. 11 Specifically, adjusted desired hours are

         x   ‫ܮ‬ௗ(௜)௝௧ = ‫ܮ‬෨ௗ௝௧                 for all non-employed, and

         x   ‫ܮ‬ௗ(௜)௝௧ = max൛‫ܮ‬෨ௗ௝௧ , ݄௜௝௧ ൟ for all employed.

     The resulting desired hours gap for each individual in the CPS in month ‫ ݐ‬is ‫ܮ‬ௗ(௜)௝௧ െ ݄௜௝௧ .

Note that our use of actual hours worked deviates from the approach of FMST (2020), who use

usual hours worked. The actual hours measure captures many transitory changes that are unique

to the pandemic and not captured by usual hours. As others, including the BLS, have pointed out,

there was also considerable labor force misclassification during the pandemic. Many furloughed

individuals reported themselves as employed and on leave rather than on temporary layoff (and

therefore unemployed).

     By using actual work hours in the AHG estimation, we are able to address the misclassification

issues that plagued the CPS during the Covid pandemic. Unfortunately, we may also incorrectly

attribute hours reductions due to vacations, illness, or other types of leave to labor market slack.

We deal with this by further adjusting the desired hours gap estimate for any individual who reports

being employed but on leave (regardless of whether the leave was paid or unpaid). If an individual

in the CPS reports that they were on leave due to vacation, maternity or paternity leave, schooling

or training, civic or military duty, a labor dispute, or weather, we keep their desired hours equal to

‫ܮ‬ௗ(௜)௝௧ , but use their usual hours worked as their measure of ݄௜௝௧ . We do this because these reasons

for leave generally involve idiosyncratic events that are in the control of both the worker and the

firm (e.g., through the bargained labor contract or lack thereof) or neither the worker and the firm,


11
  FMST (2020) show that individuals with negative hours gaps tend to exert significant search effort, consistent
with the behavior of those with positive hours gaps. This suggests that, if anything, these individuals should not
reduce the amount of aggregate labor market slack.

                                                         14
and generally do not reflect a reduction in labor supply alone. If an individual reports they are on

leave because of child care issues, family obligations, or their own temporary illness, we set their

desired hours to their reported actual hours, which equals zero in the absence of work at an

additional job. This implicitly sets the hours gap to zero as well. We do this because these instances

reflect a reduction in labor supply, though potentially involuntary, for reasons outside of what the

worker and firm could contract over. These adjustments affect about 3.5 percent of employment,

and has a modest effect on the hours gap of the employed, though the adjustment is particularly

important during the Covid pandemic since there is a large spike in individuals who report

themselves on leave during this period. The BLS provides evidence that those on furlough who

misreport themselves as employed identify themselves as on leave for “other reasons,” which is a

category that we specifically do not adjust for this reason. 12


3.4. Deriving the Aggregate Hours Gap

           We use our micro-level estimates of desired work hours and the resulting hours gaps to

derive our monthly estimates of the Aggregate Hours Gap (AHG). Recall from Section 2 that one

can define a typical measure of labor market underutilization as the ratio of some gap to a measure

of potential labor supply. The AHG uses population share estimates for each of the nine detailed

labor force states described above and weights them using a measure of the average desired hours

gap within each category.

           Define the share of the population in labor force state ݆ in month ‫ ݐ‬as ߱௝௧ , with σ௝ ߱௝௧ = 1.

The gap is the sum of these population shares in each state ݆ weighted by its average desired hours

gap, ‫ܮ‬௝௧ െ ݄௝௧ . Similarly, our measure of potential (desired) work hours is the sum of these



12
     The mismeasurement issue is also examined in detail by Faberman and Rajan (2020) and Forsythe et al. (2020).

                                                         15
population shares weighted by their desired hours alone, ‫ܮ‬௝௧ . We interpret the potential hours

measure as an estimate of potential labor supply.

       We aggregate these gaps within each labor force state as

                                                  ߱௜௝௧
                              ‫ܮ‬௝௧ െ ݄௝௧ = ෍            ൫‫ܮ‬ௗ(௜)௝௧ െ ݄௜௝௧ ൯.
                                              ௜‫א‬௝ ߱௝௧


That is, the mean desired hours gap for labor force state ݆ in month ‫ ݐ‬is the population-weighted

mean calculated across all individuals in ݆, where ߱௜௝௧ is the share of the population made up by

individual ݅ (i.e., the respondent’s sample weight) and ߱௝௧ is the share of the population in labor

force state ݆ in month ‫ݐ‬. We calculate both using the monthly CPS data. Note that the gap measure

within labor force state ݆ will vary over time due to changes in the demographic composition of

those in state ݆, and among the employed, changes in actual hours worked. It will also change due

to changes in desired hours within each category over time, though in our implementation ‫ܮ‬ௗ(௜)௝௧

will only differ before and during the Covid pandemic. Similarly, potential hours is

                                                   ߱௜௝௧
                                     ‫ܮ‬௝௧ = ෍            ‫ܮ‬ௗ(௜)௝௧ .
                                               ௜‫א‬௝ ߱௝௧


Potential hours within labor force state ݆ will also vary over time due to changes in the

demographic composition of those in state ݆ and due to changes in desired work hours within each

category before and during the pandemic. Plugging the measures for the desired hours gap and

potential work hours into equation (1) implies that the AHG measure is

                                          σ௝ ߱௝௧ ൫‫ܮ‬௝௧ െ ݄௝௧ ൯
                                 ‫ܩܪܣ‬௧ =                       .                                (3)
                                              σ௝ ߱௝௧ ‫ܮ‬௝௧

The numerator of equation (3), the gap, will vary over time due to the demographic, work hours,

and desired hours variation noted above, as well as variations in the population share of each labor

force state ݆. The denominator of equation (3), potential work hours, will vary over time due to the


                                                 16
demographic and desired hours changes noted above, and changes in the population share of each

labor force state ݆.


4. Evidence on the Aggregate Hours Gap and Desired Hours

4.1. The Aggregate Hours Gap over the Business Cycle

        We start by presenting the aggregate time-series evidence for the Aggregate Hours Gap

(AHG) and potential work hours (i.e., potential labor supply). Figure 1 presents the time-series of

the AHG and compares it to the cyclical behavior of the unemployment rate. The AHG and the

unemployment rate track each other closely through the Great Recession, but diverge thereafter,

with the AHG implying a persistently higher level of underutilization following the Great

Recession. This divergence is notably absent during the Covid pandemic, a point we return to

shortly.

        Figure 2 compares the cyclical behavior of potential (desired) work hours and the labor

force participation rate. Again, the two series track each other closely through the Great Recession.

There is a brief divergence following the Great Recession, with potential work hours falling faster

than the participation rate, but the two series converge by 2014. The two series both exhibit large

persistent drops during the Covid pandemic. Their declines are quite different quantitatively,

however, with potential work hours falling about twice as much as the participation rate. By the

end of 2021, the labor force participation rate is about 2.3 percent lower than its February 2020

level while potential work hours are about 4.6 percent lower than their February 2020 level. The

impact of the sharp decline in desired work hours will be a recurring theme throughout our

analysis.




                                                 17
       Figure 3 examines the differential behavior of the AHG and the unemployment rate in more

detail. The top panel normalizes each measure to its 2007 average and plots its behavior during

and after the Great Recession. They key feature is the persistently higher level of labor market

underutilization implied by the AHG relative to the unemployment rate. By the end of 2015, the

unemployment rate is nearly back to its pre-recession level, while the AHG is still 2.1 percentage

points above its pre-recession level.

       The bottom panel of Figure 3 normalizes the two measures to their 2019 averages and

shows that this pattern is practically reversed during the Covid pandemic. Both measures spike

sharply at the onset of the pandemic and fall sharply thereafter. By the end of 2021, however, the

AHG is slightly below its pre-recession level while the unemployment rate remains somewhat

above its pre-recession level. As we show in the next section, the key driver of the divergent

behavior in both cases is the movement in desired work hours over time.

       Figure 4 shows the changes in the AHG over time by its detailed labor force components,

with employment components in the first panel, unemployment components in the second panel,

and out of the labor force components in the last panel. Each component is equal to its population

share multiplied by its gap contribution, all divided by mean potential work hours across all

individuals. This ensures that the sum of the components across all labor force states equals the

AHG estimate depicted in Figure 1. The first panel of Figure 4 shows that multiple jobholders

contribute a negligible amount to movements in the AHG. Those with a single full-time or part-

time job exhibit a notable hours gap throughout the sample period, though their cyclical

movements during the Great Recession were relatively modest. In contrast, the full-time employed

exhibit large and uneven spikes in their hours gaps during the Covid pandemic. These spikes

primarily reflect workers who are furloughed or have had their hours otherwise cut due to business



                                               18
closures and cutbacks from multiple waves of Covid cases. 13 These individuals remain employed

(as defined in the CPS survey) and prefer to work at their usual hours, but have their actual hours

cut, leading to an increase in their hours gaps. Part-time workers exhibit a notable decline in their

hours gap during the pandemic. The decline persists through the end of 2021 and contrasts with

the slight rise in the hours gap for part-time workers during the Great Recession. We show in the

next section that a fall in desired work hours (rather than a rise in hours worked) among part-time

workers is a key driver of the decline.

        The second panel of Figure 4 shows the hours gaps of the short-term unemployed (those

unemployed 6 months or less) and long-term unemployed (those unemployed more than 6

months). The two series closely parallel the behavior of the short-term and long-term

unemployment rates in the CPS. This is because both groups report desiring close to full-time work

hours, and this preference changes little during the pandemic. Consequently, the unemployment

components of the AHG contribute significantly to its cyclical movements over time, but explain

almost none of the AHG’s divergence with the (overall) unemployment rate.

        Finally, the last panel of Figure 4 reports the hours gaps of our three out of the labor force

categories: those who want work, the retired, and all others. Three facts stand out. First, those who

are out of the labor force for other reasons exhibit a large and persistent drop in their hours gap

during the pandemic. As we show in the next section, this reflects a large decline in their desired

work hours. Second, those who are out of the labor force but report wanting work exhibit a large

and persistent spike in their hours gap during the pandemic. This is driven primarily by movements

in the share of individuals who are out of the labor force but want work during this period, and less

so by changes in their desired hours. The contribution of this group to the AHG rose during the


13
   Goolsbee and Syverson (2021) were among the first to show that changes in economic activity followed changes
in local Covid case rates more than changes in government shutdown order.

                                                      19
Great Recession as well, but not nearly to the sharp and large degree that it does during the

pandemic. Lastly, the contribution of the retired follows a steadily rising trend throughout the

sample period. This trend is practically undisturbed during the pandemic. 14


4.2. The Behavior of Desired Hours

         We next show how desired hours in the SCE data have changed during the pandemic for

various subgroups on the data. Table 1 reports the mean estimates of desired work hours for 2013-

17, 2018-19, and 2020-21, along with the difference in desired hours between 2018-19 and 2020-

2021, by detailed labor force state and selected demographics. 15 The table shows that, prior to

2020, those who work full-time generally prefer full-time hours and those working part-time

generally prefer fewer hours. The unemployed, regardless of their duration, generally prefer full-

time hours, on average. Those out of the labor force that report “wanting work” (but either did not

actively search or were not available for work) prefer close to full-time hours (about 30 hours per

week), but fewer hours than the unemployed, on average. Those who are retired or out of the labor

force for other reasons prefer a small but nontrivial amount of work hours, generally between one

and two days per week of work. Table 1 also shows that desired hours fell, on average, during the

Covid pandemic across nearly all labor force states and for all reported demographic groups. 16

Within labor force states, the declines were largest among part-time workers (3.1 hours, or 11.5

percent) and those out of the labor force who either want work (2.5 hours, or 8.4 percent), or were



14
   Several studies, such as Atkinson et al. (2021), and Nie and Yan (2021) highlight a strong potential role for
retirements in explaining the persistently low labor force participation rates. While we find a strong role for those
out of the labor force in explaining the low rates, we find that the main drivers are those out of the labor force for
other reasons, particularly when one accounts for changes in desired hours among individuals.
15
   We impose that the desired hours of the employed do not imply a negative hours gap, so that the estimates are as
comparable to the adjustments we make when estimating the AHG as the SCE data allow.
16
   Note that, in unreported results, we find that desired hours are similar for both 2020 and 2021, despite a strong
rebound in the labor market during 2021. Consequently, our results are little changed if we examine changes for
these years separately, but statistical precision declines considerably because of our small sample sizes in the SCE.

                                                          20
otherwise out of the labor force for reasons besides retirement (3.3 hours, or 22.8 percent). The
full-time employed and the unemployed exhibit almost no change in their desired work hours.
Returning to the patterns we observe in Figure 4, these results imply that movements in the AHG
for full-time workers reflect changes in the share of individuals working full-time as well as
changes in their actual work hours—i.e., changes in their desired work hours have little effect on
their behavior. Movements in the AHG for the unemployed only reflect changes in the share of
individuals who are unemployed because their desired hours are relatively constant as well. The
results also shed light on why the AHG for part-time workers falls during the pandemic—those
who remain part-time employed prefer fewer work hours during the pandemic. The AHG for those
who are out of the labor force but “want work” spikes up and remains elevated throughout the
pandemic. This is because the increase in the population share for this group more than offsets the
fact that individuals in this group prefer fewer work hours, which would otherwise give them a
smaller hours gap. Finally, those who are out of the labor force for other reasons (besides
retirement), exhibit a sharp drop in their AHG contribution almost entirely because of a major
decline in their desired work hours. The results for those out of the labor force are consistent with
the cyclical behavior of participation documented by Hobijn and Sahin (2021). We complement
their findings by highlighting the importance of intensive hours adjustments that accompany
extensive-margin participation movements.

The results by demographics show that desired work hours fell for all groups save for the
college educated. Men had slightly smaller reductions in desired hours compared to women, likely
because of added child-rearing burdens falling disproportionately upon women during the Covid

pandemic.'” In special questions fielded in the SCE Job Search Supplement in 2020 and 2021, we

'7 Alon et al. (2020) discuss the disproportionate impact the initial shock of the Covid pandemic had on women’s
employment and discuss its potential implications in the medium-run.

21
find evidence consistent with added child care burdens for women during the pandemic.

Specifically, we find that in October 2020, about 83 percent of households report having their

children’s schooling at least partly affected by Covid, and that women report spending an average

of 8.4 hours per week on their children’s schooling, compared to 5.7 hours reported by men. In

October 2021, 36 percent of households report schooling at least partly affected by Covid, with

women spending 3.0 hours and men spending 2.6 hours per week on their children’s schooling.

Among the other demographic groups, only those with less than a college degree exhibit a notably

larger decline in their desired hours than the other groups.

       Figure 5 reports the distributions of desired work hours by broad labor force status, as well

as the implied hours gap for the employed (desired work hours and the hours gap are equivalent

for the nonemployed, since they work zero hours). We report the distributions of the unadjusted

desired hours from the SCE. Consequently, the employed can have a negative hours gap if their

desired work hours are less than their actual work hours. Our method for deriving the AHG

essentially sets these negative gaps to zero, as noted earlier.

       Figure 5 reports the distributions of desired hours and hours gaps for the pre-pandemic

SCE data (2013-19) and the pandemic-period data (2020-21). The distributions of hours gaps for

the employed become more kurtotic during the pandemic. That is, there is a higher share of full-

time and part-time workers who report working exactly the hours they prefer and lower shares of

workers reporting either a positive or negative hours gap. For full-time workers, this arises through

a slightly higher share of them preferring full-time work of 35 hours or more. For part-time

workers, this arises through a lower share of them preferring full-time work despite working part-

time. The unemployed have essentially no change in the distribution of their desired work hours

during the pandemic. Those out of the labor force, however, are much more likely to prefer no



                                                  22
work at all during the pandemic, and much less likely to prefer some part-time work. The fraction

of those out of the labor force who prefer zero work hours rises from 41 percent to 50 percent

during the pandemic.

       Thus, the Covid pandemic is characterized by a notably large contraction in labor supply.

The contraction is concentrated among part-time workers and those out of the labor force who

would normally prefer at least some part-time work. We characterize these changes as being

concentrated along the intensive margins of labor force participation. These are individuals who

are normally only marginally attached to the labor force, in the sense that they work infrequently,

and when they do, they prefer less than full-time work. During the pandemic, we find that many

of these individuals chose to remain out of the labor force and not work at all. We find little

variation across demographic groups, save for that the decline in desired work hours appears

concentrated among the less educated.


5. The Covid Pandemic and Labor Supply

       In the remainder of the paper, we quantify the effect of the decline in labor supply on the

AHG and provide supporting evidence for the specific role the Covid pandemic plays in the

observed reduction in desired work hours.


5.1. Quantifying the Role of Lower Labor Supply

       We start with a counterfactual analysis that quantifies the effect of the reduction in desired

work hours on the AHG. When we construct our AHG measure, we allow for essentially a discrete

change in (predicted) desired work hours within our 39 labor force x demographic estimation cells.

Our counterfactual exercise simply holds these cell estimates constant at their 2013-19 values and

then recalculates the AHG from March 2020 forward. Under this approach, the counterfactual

                                                23
AHG estimates over this period show what the level of labor market underutilization would be had

there been no change in desired work hours. The difference between our baseline and

counterfactual AHG estimates gives the quantitative effect of labor supply changes during the

pandemic, while the difference between baseline and counterfactual potential work hours gives the

change in labor supply independent of compositional changes in the labor market during the

pandemic.

       Figure 6 shows the results of our counterfactual exercise. The top panel shows our baseline

and counterfactual estimates of the AHG compared to the unemployment rate, while the bottom

panel shows our baseline and counterfactual estimates of potential hours compared to the labor

force participation rate. The figure shows that ignoring the fall in desired work hours during the

pandemic implies a higher AHG estimate, and therefore a higher level of underutilization. By the

end of 2021, the level of underutilization implied by the counterfactual AHG is also substantially

higher than the unemployment rate. Quantitatively, the counterfactual AHG implies a degree of

labor market underutilization that is 2.5 percentage points (12.5 percent) higher than the level

implied by the baseline AHG in December 2021.

       The bottom panel of Figure 6 shows that ignoring the fall in desired work hours accounts

for essentially all of the difference in declines between the baseline estimate of potential work

hours and the labor force participation rate. Our counterfactual potential hours series essentially

lies on top of the labor force participation rate before and during the pandemic. Together with our

baseline estimates of potential hours, the results imply that labor supply fell by about twice as

much as the labor force participation rate during the pandemic (and remains depressed), and all of

the discrepancy between the two measures reflects a reduction in desired work hours throughout

the labor market.



                                                24
       Figure 7 and Table 2 highlight the sources of the differences between our baseline and

counterfactual AHG estimates by labor force status. Figure 7 plots the contributions of the

employed, unemployed, and those out of the labor force to the baseline and counterfactual AHG

measures. The figure shows that most of the difference between the two estimates (and most of

the decline in desired work hours) occurs among those who are out of the labor force. There is also

a smaller but notable contribution by the employed. The unemployed contribute essentially nothing

to the difference between our baseline and counterfactual estimates, reflecting the consistency of

their desired work hours we observe in Table 1.

       Table 2 breaks out the differences in what the baseline and counterfactual AHG estimates

imply about underutilization during the pandemic. It reports the (baseline) AHG and the

contribution of detailed labor force states in February 2020, just before the start of the pandemic,

and the change of each between then and December 2021 under the baseline and counterfactual

estimates. Overall, the baseline AHG is nearly 0.5 percentage points below its February 2020 level,

while the counterfactual AHG is 2.0 percentage points above the February 2020 level. These

combine to produce the 2.5 percentage point difference highlighted earlier. Table 2 shows that the

bulk of this difference is accounted for by part-time workers and those out of the labor force but

do not want work. Part-time workers contribute 0.5 percentage points to the overall difference; the

retired contribute 0.4 percentage points, and those out of the labor force for other reasons

contribute 1.7 percentage points. Full-time workers actually have a slightly lower hours gap in the

counterfactual case and therefore reduce the gap between the baseline and counterfactual AHG by

0.2 percentage points. All of these contributions are consistent with the main drivers of the decline

in desired work hours reported in Table 1.




                                                 25
         Finally, Table 3 reports the differences between the baseline and counterfactual AHG and

potential work hours estimates by gender, age, and education. The table shows that, under the

baseline estimates, most demographic groups are near or at their February 2020 levels of

underutilization, with men and the college-educated considerably below their February 2020

levels, by December 2021. Older individuals and those with less than a college degree have the

largest differences between their baseline and counterfactual AHG estimates.


5.2. Evidence on the Role of Covid

         Next, we examine evidence on the extent that Covid is directly related to the observed

decline in desired work hours. We start by estimating the change in desired work hours by the type

of job individuals currently or most recently worked at. We group job types by the degree of social

contact required of the job, since higher degrees of social contact imply a higher potential exposure

to Covid. Specifically, we calculate an effective social proximity index for all two-digit

occupations and major industry sectors. Our index is the social proximity index for each of these

occupations and industries developed by Leibovici, Santacreu, and Famiglietti (2020) multiplied

by one minus the share of individuals who can work from home estimated by Dingel and Neiman

(2020). The index derived by Leibovici et al. assigns occupations an index value based on the tasks

required of the job in their O*NET job description. For industries, the index is based on the

occupational mix of that industry. We group occupations and industries into (roughly employment-

weighted) thirds based on their effective proximity index value. 18



18
   The occupations with the highest effective social proximity include healthcare professionals, food preparation
workers, and those in production, construction, transportation, and personal care services. The occupations with the
lowest effective social proximity include managers, technical professionals, legal professionals and those in education.
The industries with the highest effective social proximity are health services, leisure and hospitality, retail,
construction, resources, and transportation and warehousing. The industries with the lowest effective social proximity
are professional and business services, financial activities, government, information, wholesale, and education.

                                                          26
       Table 4 presents our results. We find that individuals who currently or recently worked in

occupations and industries with at least a relatively moderate degree of social contact desired to

work fewer hours during the Covid pandemic. Those in the middle and high social proximity

groups exhibited a decline in desired work hours between 0.8 and 4.0 hours (3 and 12 percent). In

contrast, those in occupations and industries with the lowest degree of social contact actually

desired to work more hours during the pandemic, reporting increases between 0.3 and 1.3 hours

(1 and 4 percent). Thus, the evidence based on potential Covid exposure at one’s job is consistent

with Covid playing a role in the observed decline in labor supply.

       Table 5 presents a sort of consistency check on the notion that the decline in desired work

hours reflects a reduction in labor supply. It reports the (log) level and change over time of real

hourly reservation wages reported by individuals in the SCE Job Search Supplement. In general,

a rise in the average reservation wage across individuals should accompany any contraction in

labor supply. Issues related to the pandemic should also cause an increase in reservation wages.

These include the higher health risk of potentially contracting Covid and higher opportunity costs

of work because of increased family responsibilities at home. Table 5 shows that this is indeed

supported by the data. Reservation wages across all individuals rose a statistically significant 6 log

points (6.2 percent) during the pandemic. The employed and those out of the labor force each had

an increase over 7 log points, and all reported demographic groups show an increase in their

reservation wages as well. The largest increases were for men, prime-age workers, and less-

educated workers. Only the unemployed report a decline in the reservation wage.

       In general, a reduction in job search effort may accompany a contraction in labor supply.

Both the CPS and the SCE Job Search Supplement allow us to go a step further and examine the

direct role of Covid in reducing search effort. Starting in May 2020, the CPS began asking



                                                 27
respondents who were out of the labor force if they did not look for work because of Covid. In its

2020 and 2021 surveys, the SCE supplement follows up with preexisting questions on why

individuals (regardless of labor force status) did not search or only searched for part-time work to

see if their search behavior was due to the Covid pandemic.

       Figure 8 shows the time series behavior of the share of those out of the labor force in the

CPS that report not looking for work because of Covid. It reports the share for all of those out of

the labor force and for the subset that report “wanting work.” The figure shows that over 9 percent

of those out of the labor force, and 56 percent of those who “want work” reported Covid as the

reason they were not looking for work in May of 2020. The shares fall to about 4 percent and 25

percent, respectively, by October 2020, and are much lower, at 1 percent 7 percent, respectively,

by the end of 2021. The estimates line up well to the responses from the SCE supplement, which

shows that 9 percent of the nonemployed did not search due to Covid in October 2020, with the

share falling to just over 1 percent in October 2021. The SCE supplement provides additional

statistics that show a broader, but generally consistent view of the effect of Covid on search

behavior. Across all individuals, 4 percent report not looking for work because of Covid in October

2020 and this share falls to 1 percent in October 2021. Fear of contracting the virus and a perceived

lack of job opportunities are the main reasons respondents cite in 2020, while fear of catching the

virus remains as the most notable reason for not looking in 2021. Covid is much less likely to

reduce on-the-job search among the employed. Across demographic categories, Covid is more

likely to affect the incidence of job search for women, older workers, and those with less than a

college degree. Covid had quantitatively similar effects on the incidence of searching only for part-

time work (conditional on actively looking for work). Across all respondents, 9 percent of job

seekers report looking for only part-time work because of Covid in October 2020. This falls to 2



                                                 28
percent in October 2021. The share is highest in 2020 for women, prime-age workers, and those

with less than a college degree. Notably, it is similar for both the employed and nonemployed, and

remains relatively elevated for the nonemployed and for women into 2021.

        Finally, Table 7 reports measures of job search effort over time from the SCE Job Search

Supplement. We report the 2020 and 2021 estimates separately from average estimates for 2013-

17 and 2018-19. In general, our measures of search effort show a decline in search in 2020

followed by a rebound in 2021. Keep in mind that the October 2020 survey occurs six months after

the worst of the Covid pandemic’s effects on the labor market, so the decline in search effort may

have been much larger early on. 19 The number of applications sent in the prior four weeks falls

then rebounds across all labor force states (employed, unemployed, and out of the labor force).

The incidence of on-the-job search falls and then recovers as well.

        Overall, the evidence on job search behavior paints a mixed picture of the effects of Covid

on job search effort. The results show that Covid clearly had a large impact on search effort, and

ultimately, the willingness to work, in the first months of the pandemic. In 2021, however, the

share of individuals citing Covid as a reason for not looking for work declined precipitously.

Despite this, we observe no recovery in desired work hours or the labor force participation rate.

Several reasons may account for the divergence of labor supply and job search toward the end of

2021. The tightening of the labor market likely improved the returns to job search while lingering

concerns about health and issues related to child and dependent care may have kept labor supply

depressed. For example, we find that 36 percent of household children still had their schooling

affected by Covid in 2021, which may affect the work decisions of their parents. The Covid

pandemic may have also permanently changed the attitudes toward work. Some individuals may


19
  Forsythe et al. (2020) and Marinescu, Skandalis, and Zhou (2021) present richer evidence on search behavior
during the early months of the pandemic.

                                                       29
have become accustomed to a greater degree of work flexibility, including working from home,
while others may now prefer a greater degree of work-life balance.”° Others still may now prefer
a change in career towards something that provides different wages, hours, and benefits than what
they previously had. From this perspective, it will be of great interest to follow the evolution of

desired work hours in the SCE in the coming years.

6. Conclusions

In this paper, we analyze labor market underutilization and labor supply behavior of
workers during the Covid pandemic, building on the Aggregate Hours Gap developed by
Faberman, Mueller, Sahin, and Topa (2020). We find diverging patterns between the AHG and the
unemployment rate, though the nature of the divergence is not the same during the Covid pandemic
as it was following the Great Recession. The Covid pandemic is characterized by a relatively tight
labor market, with the AHG already below its February 2020 level by the end of 2021. We find
that this is driven by individuals out of the labor force and part-time workers whose desired work
hours dropped substantially during the Covid pandemic. The decline in desired work hours is more
than double the decline in the labor force participation rate and is just as persistent throughout the
pandemic. We perform a counterfactual exercise that holds desired work hours constant at their
pre-pandemic average within detailed labor force and demographic groups and recalculates the
AHG and associated potential work hours from March 2020 forward. We find that the decline in
desired work hours reduced the AHG by 2.5 percentage points (12.5 percent) relative to its
estimated value at the end of 2021, and it accounts for essentially all of the discrepancy between

potential work hours and the labor force participation rate. Our evidence does not support the

2° Barreo, Bloom, and Davis (2021) argue that an increase in the share of individuals working from home is likely a
permanent structural change in the labor market.

30
notion that the contraction in labor supply is driven mostly by women responding to child care

demands. Instead, the drop in desired hours is pervasive across most demographic groups, with

somewhat larger declines among those with less than a college degree. The decline is also

concentrated among what we refer to as the intensive margin of labor force participation. This

margin represents individuals who prefer to work infrequently, and when they do, they generally

prefer part-time work hours.

       Finally, we find a range of suggestive evidence that shows that the Covid pandemic likely

plays a considerable role in the observed reduction in desired work hours—including lower desired

hours among those in jobs with a higher potential exposure to Covid—but that its effects on job

search behavior mostly dissipate by the end of 2021. We also find that most individuals have

persistently higher reservation wages throughout the pandemic, consistent with a reassessment of

their labor supply decisions. Our findings suggest that an overall lower willingness to work has

led to a contraction in labor supply that persists throughout the Covid pandemic. This decline

contributed to the rapid tightening of the labor market following the onset of the pandemic.

       Our results also demonstrate the value of collecting survey data on desired work hours that

are consistently fielded over time. Adopting such a question in household surveys such as the

Current Population Survey would allow researchers and policymakers to assess the role of the

intensive margin of aggregate labor supply in future downturns using larger samples and with

greater detail than currently possible in the SCE.




                                                31
References

Alon, Titan, Matthias Doepke, Jane Olmstead-Rumsey, and Michele Tertilt, 2020. “This Time It's
Different: The Role of Women's Employment in a Pandemic Recession,” NBER Working Paper
No. w27660.

Atkinson, Tyler, Jim Dolmas, Marc Giannoni, Robert S. Kaplan, and Karel Mertens, 2021. "The
Labor Market May Be Tighter than the Level of Employment Suggests." Dallas Fed Economics
blog.

Barnichon, Regis, and Andrew Figura, 2016. "Declining Desire to Work and Downward Trends
in Unemployment and Participation." NBER Macroeconomics Annual 30(1): 449-494.

Barnichon, Regis, and Winnie Yee, 2020. “Adjusting the Unemployment Thermometer.” Federal
Reserve Bank of San Francisco Economic Letter, 2020-27.

Barrero, Jose Maria, Nicholas Bloom, and Steven J. Davis, 2021. “Why Working from Home Will
Stick,” NBER Working Paper No. w28731.

Bartik, Alexander W., Marianne Bertrand, Feng Lin, Jesse Rothstein, and Matthew Unrath, 2020.
"Measuring the Labor Market at the Onset of the COVID-19 Crisis." Brookings Papers on
Economic Activity, No. 2: 239-268.

Bick, Alexander, and Andrew Blandin, 2021. “Real-Time Labor Market Estimates During the
2020 Coronavirus Outbreak.” Arizona State University, mimeo.

Cajner, Tomaz, Leland D. Crane, Ryan A. Decker, John Grigsby, Adrian Hamins-Puertolas, Erik
Hurst, Christopher Kurz, and Ahu Yildirmaz, 2020. "The US Labor Market during the Beginning
of the Pandemic Recession." Brookings Papers on Economic Activity: 3-34.

Crump, Richard K., Stefano Euseppi, Mare Giannoni, and Aysegiil Sahin, 2022. “The
Unemployment-Inflation Trade-off Revisited: The Phillips Curve in COVID Times,” mimeo.

Dingel, Jonathan I. and Brent Neiman, 2020. “How Many Jobs can be done at Home?” Journal of
Public Economics 189: 104325.

Domash, Alex, and Lawrence H. Summers, 2022. “How Tight are U.S. Labor Markets?” NBER
Working Paper No. 29739.

Faberman, R. Jason, Andreas I. Mueller, Aysegtil Sahin and Giorgio Topa. “Job Search Behavior
among the Employed and Non-Employed,” forthcoming, Econometrica.

Faberman, R. Jason, Andreas I. Mueller, Aysegiil Sahin and Giorgio Topa, 2020. “The Shadow
Margins of Labor Market Slack,” Journal of Money, Credit, and Banking, 52(S2), 355-391.

Faberman J. and A. Rajan, 2020. “Is the Unemployment Rate a Good Measure of People
Currently Out of Work?" Chicago Fed Insights blog.

32
Forsythe, Eliza, Lisa B. Kahn, Fabian Lange, and David Wiczer, 2020. “Searching, Recalls, and
Tightness: An Interim Report on the Covid Labor Market,” NBER Working Paper No. 28083.

Ganong, Peter, Pascal Noel, and Joseph Vavra, 2020. "US Unemployment Insurance Replacement
Rates during the Pandemic." Journal of Public Economics 191: 104273.

Goolsbee, Austan, and Chad Syverson, 2021. “Fear, Lockdown, and Diversion: Comparing
Drivers of Pandemic Economic Decline 2020,” forthcoming, Journal of Public Economics.

Hobijn, Bart, and Aysegiil Sahin, 2021. “Maximum Employment and the Participation Cycle,”
NBER Working Paper No. w29222.

Leibovici, Fernando, Ana Maria Santacreu, and Matthew Famiglietti, 2020. “Social Distancing
and Contact-Intensive Occupations,” On the Economy Blog, Federal Reserve Bank of St. Louis,
March 24.

Marinescu, Ioana, Daphne Skandalis, and Daniel Zhao, 2021. "The Impact of the Federal
Pandemic Unemployment Compensation on Job Search and Vacancy Creation." Journal of Public
Economics 200: 104471.

Mongey, Simon, Laura Pilossoph, and Alex Weinberg, 2021. “Which Workers Bear the Burden
of Social Distancing Policies?” Journal of Economic Inequality.

Nie, Jun, and Shu-Kuei X. Yang, 2021. "What Has Driven the Recent Increase in Retirements?"
Federal Reserve Bank of Kansas City, Economic Bulletin, August 11.

Sahin, Aysegiil, Murat Tasci, and Jin Yan, 2021. “Unemployment in the Time of COVID-19: A

Flow-Based Approach to Real-Time Unemployment Projections,” NBER Working Paper No.
w28445.

33
Table 1. Desired Work Hours by Labor Force Status and Demographics
                                                                                                   Difference,
                         October of…           2013-17          2018-19          2020-21         ‘20-21 – ‘18-19
 Labor Force Status
                                                 41.54            41.69            41.10               -0.59
 Employed
                                                 (0.19)           (0.31)           (0.34)              (0.47)
                                                 44.88            45.11            45.11               -0.00
   Full-time
                                                 (0.16)           (0.25)           (0.26)              (0.45)
                                                 28.68            27.10            23.97               -3.13
   Part-time
                                                 (0.49)           (0.82)           (0.72)              (0.93)
                                                 35.62            36.07            35.80               -0.27
 Unemployed
                                                 (0.70)           (1.53)           (1.01)              (2.30)
                                                 13.06            12.01            11.06               -0.94
 Out of the Labor Force
                                                 (0.33)           (0.51)           (0.55)              (0.70)
                                                 29.91            30.00            27.47               -2.53
   Want work
                                                 (1.86)           (2.74)           (2.09)              (3.14)
                                                 10.99            10.25             9.68               -0.57
   Retired
                                                 (0.36)           (0.57)           (0.57)              (0.71)
                                                 15.45            14.46            11.17               -3.29
   Others out of the labor force
                                                 (0.65)           (1.02)           (1.22)              (1.02)
 Gender
                                                 34.42            34.16            33.42               -0.74
 Men
                                                 (0.34)           (0.56)           (0.58)              (0.79)
                                                 30.98            30.94            29.94               -1.00
 Women
                                                 (0.33)           (0.54)           (0.59)              (0.81)
 Age
                                                 39.70            40.98            40.06               -0.91
 Prime Age (25-54)
                                                 (0.26)           (0.39)           (0.40)              (0.67)
                                                 23.44            22.22            21.46               -0.76
 Older (55+)
                                                 (0.36)           (0.59)           (0.65)              (0.74)
 Education
                                                 30.87            31.00            29.36               -1.63
 Some college or less
                                                 (0.35)           (0.60)           (0.67)              (0.70)
                                                 36.47            35.75            35.53               -0.22
 College degree or more
                                                 (0.31)           (0.49)           (0.49)              (0.94)
Notes: Table reports mean desired work hours for each labor force state or demographic group. Estimates are for
respondents in each category pooled across SCE surveys within each listed time period, and desired hours are adjusted
to impose a zero-minimum desired hours gap among the employed. The last column reports the difference between
the 2018-19 mean and the 2020-21 mean for each category. Standard errors are in parentheses.




                                                          34
Table 2. Baseline vs. Counterfactual Changes in the AHG by Labor Force Status

February Change, Dec. 2021 — Feb. 2020
2020 Value Baseline Counterfactual
All Individuals, Age 18 to 79
Aggregate Hours Gap 20.40 -0.45 2.00
Potential Work Hours 31.81 -1.47 -0.52
AHG Components by Labor Force Status
Employed
Full-time 1.59 0.73 0.53
Part-time 1.78 -0.58 -0.08
Unemployed
Short-term (< 6 mos.) 2.27 0.17 0.13
Long-term (> 6 mos.) 0.53 0.43 0.37
Out of the Labor Force
Want work 1.62 0.28 0.39
Retired 5.18 0.04 0.46
Other reason 7.43 -1.51 0.21

Notes: Estimates are from authors’ calculations using CPS and SCE data for all individuals age 18 to 79. Baseline and
counterfactual estimates of the AHG and potential work hours replicate those from Figure 6, with the AHG broken
out by detailed labor force state. Counterfactual estimates recalculate each series holding desired hours constant within
each labor force status x demographic estimation cell from March 2020 forward. See text for details.

Table 3. Baseline vs. Counterfactual Changes in the AHG by Demographics

February Change, Dec. 2021 — Feb. 2020

2020 Value Baseline Counterfactual
Aggregate Hours Gap
Gender
Men 17.33 -0.57 1.75
Women 23.89 -0.17 2.28
Age
Prime Age (25-54) 13.46 0.31 1.95
Older (55+) 30.95 -0.12 2.59
Education
Some college or less 24.36 0.03 2.59
College or more 13.84 -0.91 1.24

Notes: Estimates are from authors’ calculations using CPS and SCE data for all individuals age 18 to 79. Baseline and
counterfactual estimates of the AHG replicate those from Figure 6, with the AHG broken out by demographic group.
Counterfactual estimates recalculate each series holding desired hours constant within each labor force status x
demographic estimation cell from March 2020 forward. See text for details.

35

Table 4. Desired Work Hours by Social Proximity Required of Occupation or Industry
                                                                         Difference,
                      October of… 2013-17       2018-19     2020-21    ‘20-21 – ‘18-19
 Prior Month’s Occupation
                                      36.04      35.05       36.32           1.27
 Low Social Proximity
                                                     (0.37)          (0.62)           (0.62)               (1.01)
                                                    31.90            33.56            29.51                -4.04
 Medium Social Proximity
                                                     (0.41)          (0.69)           (0.76)               (1.08)
                                                    32.38            32.19            31.36                -0.83
 High Social Proximity
                                                     (0.49)          (0.80)           (0.89)               (1.03)
 Prior Month’s Industry
                                                    31.95            31.16            31.43                 0.28
 Low Social Proximity
                                                     (0.38)          (0.61)           (0.60)               (0.91)
                                                    33.85            34.52            32.20                -2.32
 Medium Social Proximity
                                                     (0.45)          (0.73)           (0.85)               (1.13)
                                                    33.34            33.37            32.50                -0.86
 High Social Proximity
                                                     (0.41)          (0.71)           (0.75)               (0.96)
Notes: Table reports mean desired work hours for each occupation or industry group. Estimates are for respondents in
each category pooled across SCE surveys within each listed time period, and desired hours are adjusted to impose a
zero-minimum desired hours gap among the employed. The last column reports the difference between the 2018-19
mean and the 2020-21 mean for each category. Social proximity is defined by the degree of interpersonal contact and
the ability to work from home for the respondent’s current or most recent job. See text for details of the identification.
Standard errors are in parentheses.




                                                           36
Table 5. Real Reservation Wages by Labor Force Status, Demographics, and Job Type
                                                                          Difference,
                      October of… 2013-17          2018-19  2020-21     ‘20-21 – ‘18-19
 Real Reservation Wage (log hourly wage, 2019 dollars)
                                       3.034         3.092   3.153           0.060
 All Individuals
                                                 (0.010)         (0.017)         (0.019)            (0.025)
 Labor Force Status
                                                  3.156           3.210           3.281             0.071
 Employed
                                                 (0.012)         (0.019)         (0.021)            (0.028)
                                                  2.634           2.840           2.759             -0.081
 Unemployed
                                                 (0.032)         (0.106)         (0.078)            (0.110)
                                                  2.773           2.778           2.854             0.076
 Out of the labor force
                                                 (0.019)         (0.028)         (0.037)            (0.049)
 Gender
                                                  3.226           3.231           3.307             0.076
 Men
                                                 (0.015)         (0.025)         (0.027)            (0.035)
                                                  2.868           2.951           3.007             0.055
 Women
                                                 (0.013)         (0.021)         (0.025)            (0.034)
 Age
                                                  3.116           3.174           3.248             0.075
 Prime Age (25-54)
                                                 (0.013)         (0.022)         (0.023)            (0.031)
                                                  2.918           2.982           2.998             0.016
 Older (55+)
                                                 (0.015)         (0.025)         (0.031)            (0.042)
 Education
                                                  2.842           2.873           2.927             0.053
 Some college or less
                                                 (0.013)         (0.022)         (0.026)            (0.029)
                                                  3.424           3.471           3.489             0.018
 College or more
                                                 (0.014)         (0.022)         (0.023)            (0.036)
Notes: Table reports estimates of the log real hourly reservation wage by selected labor force and demographic
characteristics. Estimates are for respondents age 18 to 79 in each category pooled across SCE surveys within each
listed time period. Standard errors are in parentheses.




                                                       37
Table 6. Effects of the Covid Pandemic on Search Behavior
                                 Pct. Only Searched Part-Time                         Pct. Did Not Search
                                 due to Covid, Cond. on Search                            Due to Covid
                 October of…         2020             2021                            2020           2021
                                      9.0              2.0                             4.1            1.0
 All Individuals
                                              (2.4)                (1.0)               (0.6)             (0.3)
  Child care/family reasons                    9.0                  2.0                0.7                0.2
  No work in area                              ---                  ---                1.4                0.1
  Fear of contracting Covid                    ---                  ---                2.0                0.6
 Labor Force Status
                                               9.7                  0.4                1.6                0.9
 Employed last month
                                              (2.9)                (0.5)               (0.5)             (0.3)
                                               7.9                  4.7                8.9                1.2
 Nonemployed last month
                                              (4.3)                (2.9)               (1.7)             (0.6)
 Gender
                                               2.2                  0.0                3.6                0.5
 Men
                                              (2.0)                (0.0)               (0.9)             (0.3)
                                              12.5                  3.3                4.5                1.5
 Women
                                              (3.5)                (1.7)               (0.9)             (0.5)
 Age
                                              11.4                  2.1                2.9                0.8
 Prime age (25-54)
                                              (3.0)                (1.2)               (0.7)             (0.4)
                                               3.4                  2.1                4.5                1.2
 Older (55+)
                                              (0.3)                (2.0)               (1.1)             (0.5)
 Education
                                              14.3                  2.1                5.0                0.9
 Some college or less
                                              (5.0)                (1.6)               (1.1)             (0.4)
                                               2.2                  1.8                2.6                1.1
 College or more
                                              (1.5)                (1.2)               (0.7)             (0.4)
Note: The table reports the percentage of individuals who actively looked for work that only searched part-time
because of the Covid pandemic, and the percentage of all individuals who did not search for any work because of the
Covid pandemic, for each listed group. Estimates come from authors’ tabulations from the 2020 and 2021 waves of
the SCE Job Search Supplement, for all individuals aged 18-79. Standard errors are in parentheses.




                                                        38
Table 7. Search Behavior by Labor Force Status
                          October of…   2013-17                          2018-19            2020              2021
 Applications Sent in Last 4 Weeks
                                          1.12                             0.85              0.76             1.25
 All Individuals
                                                         (0.07)            (0.09)           (0.11)            (0.21)
                                                          1.03             0.76              0.69             1.11
   Employed
                                                         (0.07)            (0.10)           (0.13)            (0.26)
                                                          9.28             8.09              6.14            10.20
   Unemployed
                                                         (1.06)            (1.49)           (1.03)            (2.52)
                                                          0.27             0.29              0.13             0.29
   Out of the labor force
                                                         (0.04)            (0.08)           (0.04)            (0.08)
 Percent of Employed Looking in Last 4 Weeks
                                          21.3                             17.5              15.9             17.8
 All Employed
                                                          (0.6)            (1.0)             (1.4)            (1.4)
Notes: Table reports estimates of search effort by labor force status. Estimates are for respondents age 18 to 79 in each
labor force state pooled across SCE surveys within each listed time period. Standard errors are in parentheses.




                                                          39
Figure 1. The Aggregate Hours Gap vs. the Unemployment Rate




Notes: The Aggregate Hours Gap (AHG) estimates come from authors’ calculations using CPS and SCE data for all
individuals age 18 to 79. The unemployment rate uses CPS data. See text for details.

Figure 2. Potential Work Hours vs. the Labor Force Participation Rate




Notes: Potential work hours estimates are from authors’ calculations using CPS and SCE data for all individuals age
18 to 79. The labor force participation rate uses CPS data. See text for details.



                                                        40
Figure 3. The Aggregate Hours Gap vs. the Unemployment Rate, Selected Periods
                                 (a) The Great Recession




                                             (b) The Covid Pandemic




Notes: The Aggregate Hours Gap (AHG) estimates come from authors’ calculations using CPS and SCE data for all
individuals age 18 to 79. The unemployment rate uses CPS data. The top panel reports each series percentage point
deviation from its 2007 average, while the bottom panel reports each series percentage point deviation from its 2019
average. See text for details.


                                                        41
Figure 4. Components of the Aggregate Hours Gap Over Time
                          (a) Employment Components, Finer Detail




                         (b) Unemployment Components, Finer Detail




                                          42
                            (c) Out of Labor Force Components, Finer Detail




Notes: The Aggregate Hours Gap (AHG) estimates come from authors’ calculations using CPS and SCE data for all
individuals age 18 to 79. The figure reports the individual AHG components by detailed labor force status. All
components are reported as their contribution to the total AHG, as a percentage of mean potential work hours.




                                                     43
Figure 5. Distribution of Desired Hours and Hours Gaps by Labor Force Status
       (a) Hours gap, full-time employed             (b) Hours gap, part-time employed




         (c) Desired hours, full-time employed                    (d) Desired hours, part-time employed




          (e) Desired hours, unemployed                           (f) Desired hours, out of the labor force




Notes: The top two panels report the distribution of the desired hours gap (desired minus actual work hours) across
full-time and part-time workers, respectively. The bottom four panels report the distribution of desired work hours by
labor force status. Estimates are from authors calculations using respondents age 18 to 79 pooled over the 2013-19
surveys (dark blue bars) or 2020-21 surveys (orange bars).




                                                         44
Figure 6. Baseline vs. Counterfactual Movements in the AHG and Potential Work Hours
                              (a) Baseline vs. Counterfactual AHG




                              (b) Baseline vs. Counterfactual Potential Hours




Notes: Estimates are from authors’ calculations using CPS and SCE data for all individuals age 18 to 79. Baseline
estimates of the AHG and potential work hours replicate the estimates from Figures 1 and 2, respectively.
Counterfactual estimates recalculate each series holding desired hours constant within each labor force status ×
demographic estimation cell from March 2020 forward. The unemployment and labor force participation rates use
CPS data. See text for details.


                                                       45
Figure 7. Baseline vs. Counterfactual Movements in the AHG, Additional Detail




Notes: Estimates are from authors’ calculations using CPS and SCE data for all individuals age 18 to 79. Baseline and
counterfactual estimates of the AHG and potential work hours replicate those from Figure 6, with the AHG broken
out by labor force state. Counterfactual estimates recalculate each series holding desired hours constant within each
labor force status × demographic estimation cell from March 2020 forward. See text for details.




                                                        46
Figure 8. Fraction of Those Out of the Labor Force Not Looking Because of Covid




Notes: Estimates are from authors’ calculations for all individuals age 18 to 79 in the CPS, based on a special survey
question implemented starting in May 2020.




                                                         47
Appendix: Desired Hours by Demographics and Labor Force
Status, 2013-19 vs. 2020-21 Samples
       Table A.1 reports the mean desired hours and (adjusted) hours gaps for the 39

demographics × labor force status categories that we use in the estimation of our measure of labor

market underutilization for the 2013-19 and 2020-21 survey years pooled together. The estimates

are for nine different labor force states (as noted in the main text), three age groups (less than 25,

25 to 54, 55 or older), gender, and two education groups (less than a college degree or a college

degree or more). This creates a potential for up to 108 group estimates of desired work hours.

Unfortunately, sparse sample cells limit our ability to generate reliable estimate for all 108 groups.

Consequently, we aggregate individuals into 39 broader groups. These groups are an unbalanced

panel of demographic subgroups across the nine labor force states. These categories represent the

finest level of disaggregation we feel we can use given the sample size constraints. The estimates

are the sample-weighted means of desired work hours predicted by labor force status,

demographics, and their interaction with a fixed effect for the pandemic period as described in the

main text. The table shows considerable heterogeneity in desired hours across the groups, along

with notable heterogeneity in the change in desired hours between the 2013-19 and 2020-21

periods. See the Table 1 and its discussion in the main text for more details.




                                                 48
Table A.1. Desired Hours by Demographics & Labor Force Groups Used in Estimation

Labor Force 2013-19 SCE 2020-21 SCE
Status Description N Mean N Mean
Male, 18-24, all education 31 1a 3 (095)
Male, 25-54, < College 532 003) 124 15
Male, 25-54, > College 1039 (0.02) 323 (0.06)
Male, 55+, < College 242 004) 43 (0.29)
Employed FT, Male, 55+, > College 299 003) 58 ol)
single job Female, 18-24, all education 47 O15) 14 0.29)
Female, 25-54, < College 484 (0.04 90 0.16
Female, 25-54, > College 787 (0.03) 245 oO 08)
Female, 55+, < College 181 o 07 42 15
Female, 55+, > College 162 0.07 58 o 17
Male, 18-54, < College 99 (0.06) 22 (0.20)
Male, 18-54, > College 168 (0.05) 46 028)
Employed FT, Male, 55+, all education 91 (0.05) 19 (05)
multiple jobs Male, 18-54, < College 138 009) 28 O33)
Male, 18-54, > College 202 007 57 O21
Male, 55+, all education 73 ol) 13 o %)
Male, 18-54, all education 100 oon 26 a 48
Male, 55+, < College 103 o 05) 16 O34)
Male, 55+, > College 138 (003) 34 (0.25)
inwle job PT, Female, 18-54, < College 119 (0.09) 32 oa)
Female, 18-54, > College 123 a 08) 33 (024)
Female, 55+, < College 94 (0.09) 36 (0.23)
Female, 55+, > College 92 (0.09) 20 (0.28)

(continued on next page)

49
Table A.1. (continued)

Labor Force 2013-19 SCE 2020-21 SCE
Status Description N Mean N Mean
. 30.25 27.30
Employed PT, Male, all ages, all education 118 (0.12) 15 (0.52)
multiple jobs Female, all ages, all education 164 (0.09) 34 rn 3
. 38.90 37.92
Unemployed Male, all ages, all education 54 (0.31) 18 (0.51)
<6 months Female, all ages, all education 91 o a 30 o m1)
. 37.68 38.30
Unemployed Male, all ages, all education 39 (0.42) 10 (0.29)
> 6 months . 33.89 36.09
Female, all ages, all education 40 (0.51) 13 (1.02)
. 30.37 29.09
OLF, Want Male, all ages, all education 26 (0.59) 13 (1.10)
Work Female, all ages, all education 32 06) 13 orn
11.59 10.52
Male, all ages, < College 425 (0.08) 92 (0.13)
Male, all ages, > College 492 10.95 128 8.28
. (0.07) (0.09)
OLF, Retired
Female, all ages, < College 334 10.08 88 10.07
» all ages, 8 (0.10) (0.17)
9.78 8.10
Female, all ages, > College 240 (0.12) 69 (0.18)
Male, 18-54, all education 102 18) 27 a 7)
Male, 55+ all education 106 13.33 18 9.22
(0.20) (0.40)
OLF, Other 16.57 13.15
Female, 18-54, all education 282 (0.13) 54 (0.31)
Female, 55+ all education 132 (0.18) 42 (030)

Notes: Sample is all individuals in the SCE Job Search supplement aged 18-79 pooled across its 2013-19
and 2020-21 surveys. Estimates represent the sample-weighted mean predicted desired hours, where the
predicted estimates are from the regression of actual desired hours on fixed effects for labor force status,
demographics, and time period. See the main text for details. Standard errors are in parentheses.

50

File and source

File

has-the-willingness-to-work-fallen-during-the-covid-pandemic-2.pdf

Size

1,430,355 bytes

SHA-256

f604b13172ba0b4b3230bbc1b3e31c0c21bddd2b27f7fdc1362d8a4678f1141b

Our copy

has-the-willingness-to-work-fallen-during-the-covid-pandemic-2.pdf

Original

doi.org