Inequality in the Impact of the Coronavirus Shock: Evidence from Real Time Surveys

Full text

Inequality in the Impact of the Coronavirus Shock:
Evidence from Real Time Surveys
Abi Adams-Prassl, Teodora Boneva, Marta Golin, and Christopher Rauh∗
April 24, 2020
Abstract
We present real time survey evidence from the UK, US and Germany showing
that the labor market impacts of COVID-19 differ considerably across countries.
Employees in Germany, which has a well-established short-time work scheme, are
substantially less likely to be affected by the crisis. Within countries, the impacts
are highly unequal and exacerbate existing inequalities. Workers in alternative
work arrangements and in occupations in which only a small share of tasks can be
done from home are more likely to have reduced their hours, lost their jobs and
suffered falls in earnings. Less educated workers and women are more affected by
the crisis.
JEL: J21, J22, J24, J33, J63
Keywords: Recessions, Inequality, Labor market, Unemployment, Coronavirus
∗Adams-Prassl: University of Oxford (email: abi.adams@economics.ox.ac.uk). Boneva: Uni-
versity of Zurich (email: teodora.boneva@econ.uzh.ch). Golin: University of Oxford (email:
marta.golin@nuffield.ox.ac.uk). Rauh: University of Cambridge, Trinity College Cambridge (email:
cr542@cam.ac.uk). Ethics approval was obtained from the Central University Research Ethics Com-
mittee (CUREC) of the University of Oxford: ECONCIA20-21-09. We thank Toke Aidt and Hamish
Low for valuable feedback. We are grateful to the Economic and Social Research Council, the Univer-
sity of Oxford, the University of Zurich, and the Cambridge INET for generous financial support, and
Marlis Schneider for excellent research assistance.
1

1 Motivation
In recent weeks, the COVID-19 outbreak has caused severe disruptions to labor supply
in many countries around the world, bringing whole economies grinding to a halt. As
a result of measures that limit people’s ability to do their jobs, individuals are already
suffering large and immediate losses in terms of income and employment. Obtaining a
better understanding of the distribution of impacts of the COVID-19 crisis is crucial for
designing policy responses that target those individuals who have been most affected by
the crisis. In this paper, we provide evidence from real time surveys conducted in the
US, the UK and Germany in March and April 2020, with a total of 20,910 respondents.
We examine which workers were most likely to reduce their hours, lose their jobs and
experience a decrease in their earnings. Our focus lies on documenting cross-country
differences as well as understanding which job characteristics allow individuals to buffer
the shock of the crisis.
The impacts of the COVID-19 crisis are large and unequal within and across coun-
tries. There are several key results that emerge from our study. First, we find staggering
cross-country differences in the labor market impacts of the COVID-19 epidemic. In
early April, 18% and 15% of individuals in our sample report having lost their jobs
within the last four weeks due to the coronavirus outbreak in the US and the UK,
respectively, compared to only 5% in Germany.1Germany has a well-established short-
time work (STW) scheme and we find that 35% of employees have been asked to reduce
their hours to benefit from this scheme. Furloughing has been relatively prevalent in
the UK but not as prevalent in the US; 43% and 31% of employees in the UK and
US respectively report having being furloughed in their main job. Though it might
be too early to claim that the “German economic miracle” witnessed during the Great
Recession (Rinne and Zimmermann, 2012) is repeating itself, we find that the shock
has been much smaller for German workers thus far.
Second, there are striking differences in the impacts within countries depending on
job and worker characteristics. Workers who report that they can do a high share of
their tasks from home are substantially less likely to report to have lost their jobs due
to the coronavirus outbreak. This relationship has become steeper as the crisis has
unfolded. Second, there are large differences in job loss probabilities between employed
and self-employed workers, as well as between employees in different work arrangements.
1We note that our aggregate figures for the US are comparable to recent results from other studies,
e.g. Bick and Blandin (2020) who find that 16.5% of workers in the US lost their jobs.
2

Employees in permanent contracts and in salaried jobs were significantly less likely to
lose their jobs compared to employees in other alternative work arrangements. Third,
there are large differences in job loss probabilities across different occupations, mostly
owing to the fact that the average percentage of tasks workers report being able to do
from home varies substantially across occupations. Interestingly, even within occupa-
tions the percentage of tasks workers can do from home is a significant predictor of job
loss, over and above what can be explained by occupation or other job characteristics.
Turning to individual differences in job loss probabilities, in the US and the UK there
are marked differences between men and women and between people with and without
university education. Women and workers without a college degree are significantly
more likely to have lost their jobs. Remarkably, while occupation fixed effects and
the percentage of tasks one can do from home can account for all of the gap in job
loss between college-educated workers and workers without a college degree, this is
not the case for the gender gap. The gender gap persists even once we control for
these job characteristics, indicating that other factors play a role. This does not only
contrast with usual recessions in which men tend to be more likely to lose their jobs.2
It also stands in contrast with the results from Germany, where neither gender nor
having a college degree significantly predict job loss. Turning to time use data, we note
that amongst the population working from home, women spend significantly more time
homeschooling and caring for children.
Individual outlooks on the future are bleak. The average perceived probability of
losing one’s job within the next months is 37% in the US and 32% in the UK for workers
who are still employed. Even in Germany, where the share of workers who have lost their
job already is much smaller than in the anglophone countries, the average perceived
probability of losing one’s job before August 2020 is 25%. Individuals are worried about
being able to pay their usual bills and expenses. 46% in the US, 38% in the UK, and
32% in Germany already have struggled to pay their usual bills. Overall, the results
suggest that immediate action is required and that policies that aim to mitigate the
shocks of the crisis should take into account the inequality in labor market impacts.
Our paper contributes to several strands of the literature. First, it contributes to
the large literature on the impact of economic downturns on labor market outcomes
(see, e.g., Hoynes, Miller and Schalle 2012; Christiano, Eichenbaum and Trabandt 2015)
and the importance of short-time work schemes to buffer economic shocks (see, e.g.,
2See, for instance, Bredemeier, Juessen and Winkler (2017).
3

Giupponi and Landais 2018; Cahuc, Kramarz and Nevoux 2018; Kopp and Siegenthaler
2018). Second, it closely relates to the literature on alternative work arrangements and
the role of firms in providing workers insurance against shocks to labor demand (Mas
and Pallais 2020; Koustas 2018; Malcomson 1999). We show that firms are sheltering
permanent workers more than those on temporary contracts. More surprisingly, we find
that even amongst those on permanent contracts, workers on flexible hours contracts
or who are paid by the hour have been hit hardest. Third, our paper contributes to
the small but exponentially growing economic literature on the effect of the COVID-19
pandemic. Recent research using real time data has studied the relationship between
the outbreak and stock returns and volatility, subjective uncertainty in business ex-
pectations surveys, business closures, worries regarding the aggregate economy, and
household spending (Alfaro et al. 2020; Baker et al. 2020a,b; Bartik et al. 2020; Fetzer
et al. 2020; Carvalho et al. 2020). Other research using data collected before the crisis
has discussed channels through which the current crisis may affect workers differently
depending on their gender and occupation (Alon et al. 2020; Dingel and Neiman 2020;
Mongey and Weinberg 2020).3Looking at job ads, Kahn, Lange and Wiczer (2020)
find that in the US demand for labor has decreased drastically. We provide real time
evidence on the effect of the pandemic on the supply-side of labor market outcomes.
This paper is structured as follows. Section 2 describes the institutional background,
the characteristics of our sample and the survey design. Sections 3 and 4 present
the inequality in impacts by job characteristics, while Section 5 shows the inequality
in impacts by individual characteristics. Section 6 presents our evidence regarding
expectations for the future, while Section 7 concludes.
2 Institutional Background and Data
2.1 Institutional Background
There are many institutional differences between the US, UK and German labor mar-
kets. In this section we briefly highlight some cross-country differences in labor market
policies that may buffer the negative impacts of the COVID-19 crisis. One prominent
countercyclical policy tool is short-time work (STW) or ‘furloughing’. STW allows firms
3Recent work on COVID-19 has also investigated partisan differences in social distancing (Allcott
et al. 2020), differences in testing and infection rates among different groups in the population (Borjas
2020), or differences in access to high speed internet across regions (Chiou and Tucker 2020).
4

affected by temporary shocks to reduce their employees’ hours instead of laying them
off. Government subsidies pay short-time compensation to employees who reduce their
hours, proportional to the reduction in hours (up to a cap). STW is aimed at correcting
the inefficiencies which arise if liquidity-constrained firms must first fire and then re-hire
and re-train new workers. Separation is costly as match-specific human capital is lost.
STW allows firms to preserve or ‘freeze’ existing matches, thereby contributing to a
swift recovery of the economy in the aftermath of the pandemic. Recent evidence on
the effectiveness of STW schemes suggests that short-time work can have sizeable im-
pacts on employment and firm survival (see, e.g., Giupponi and Landais 2018; Cahuc,
Kramarz and Nevoux 2018; Kopp and Siegenthaler 2018). Furloughing is similar to
short-time work only that working hours typically need to be reduced to zero, i.e. the
employee is not allowed to take up any work for their employer while being furloughed.
Germany has one of the oldest and most comprehensive, well-established STW pro-
grams in the world.4The German Kurzarbeit scheme allows firms to reduce their
employees’ hours for up to 12 months. While a reduction of working hours to zero is
possible, the Kurzarbeit scheme provides a considerable degree of flexibility. Different
employees within the same firm can work 0-100% of their usual working hours. The
rate at which forgone net monthly earnings are replaced (up to a cap) is 60% (or 67%
for employees with children). This wage subsidy is referred to as the Kurzarbeitergeld
and it is claimed by the employer from the Federal Employment Agency. On March 13,
2020, in response to the COVID-19 crisis, the German Bundestag and Bundesrat passed
a law making the eligibility criteria for STW less stringent, allowing more firms and
workers to benefit from the scheme.
In the United Kingdom, the government announced a new scheme to protect jobs
on March 20, 2020. The newly established Coronavirus Job Retention Scheme allows
firms to furlough workers for up to three months, starting March 1, 2020. Through this
scheme, the government replaces 80% of employees’ wages, up to a maximum of £2,500
per month. Employers are responsible for claiming through the Job Retention Scheme
on behalf of their employees. In contrast to the German Kurzarbeit, furloughed workers
cannot undertake any work for their employer. This rigidity may create inefficiencies as
a minimum number of hours may be necessary to sustain critical business operations.
It may also make it more attractive for firms to lay off and re-hire workers rather than
retain them, if workers are not allowed to do any work while being furloughed. Another
4Short-time work dates back to 1910 when it was first used in the mining industry.
5

difference between the UK and German schemes is that the UK scheme is currently
only open for three months. While the government did announce the possibility of an
extension, there is considerable uncertainty about the length of the scheme.
The United States has a similar furloughing scheme in place as the United Kingdom.
The Coronavirus Aid, Relief, and Economic Security (CARES) Act was signed into law
on March 27, 2020. The CARES Act includes provisions to expand unemployment
benefits to include people furloughed, gig workers, and freelancers, with unemployment
benefits increased by $600 per week for a period of four months, as well as direct
payments to families of $1,200 per adult and $500 per child for households making up
to $75,000.5
Germany and the United Kingdom have also made provisions for the self-employed.
To support small businesses, freelancers and the solo self-employed, the German federal
government put together an emergency assistance program which was approved on
March 27, 2020. Businesses with up to five (full-time equivalent) employees can apply
for a one-off payment of up to 9,000 euros for a period of three months. Businesses with
up to ten employees can receive up to 15,000 euros. Federal states have put additional
assistance programs in place, the generosity of which varies considerably across states.
The UK Self-employment Income Support Scheme allows self-employed individuals to
claim a taxable grant worth 80% of their trading profits up to a cap of £2,500 per
month for up to three month. This scheme was announced on March 26, 2020.
2.2 Data Collection and Samples
To study the labor market impacts of the coronavirus shock, we collected primary survey
data on large geographically representative samples of individuals in the United States,
the United Kingdom and Germany. In the US and the UK, we collected two waves of
survey data, while in Germany we collected one wave of data. The data were collected
by a professional survey company.6In the US, the first wave of data (N= 4,003)
was collected on March 24-25, 2020, while the second wave of data (N= 4,000) was
collected on April 9-11, 2020. In the UK, the first wave (N= 3,974) was collected
on March 25-26, 2020, while the second wave (N= 4,931) was collected on April 9-
5Some US states also have short-time compensation (STC) schemes. STC programs are imple-
mented at the sate level and there are differences among state programs.
6All participants were part of the company’s online panel and participated in the survey online. The
survey was scripted in the online survey software Qualtrics. Participants received modest incentives
for completing the survey.
6

14, 2020. In Germany, the data (N= 4,002) was collected on April 9-12, 2020. We
deliberately chose to survey new participants in the second survey wave for the US and
the UK, i.e. there are no participants who participated in the survey twice.
Given the speed at which events and policy responses unravelled, it is important
to situate the moment our surveys were launched. At the time we collected the first
wave of data (in the US and the UK), there were more than 55,000 confirmed cases
of coronavirus and fewer than 1,000 reported deaths in the US. About half of the US
population was already under stay-at-home orders. In the UK, there were still fewer
than 10,000 confirmed cases and 500 reported deaths. The lockdown had already been
in place for a few days, but Prime Minister Boris Johnson had not yet announced the
Self-employment Income Support Scheme. All three countries had some lockdown or
social distancing measures in place at the time we collected data in early April.
To be eligible to participate in the study, participants had to be resident in the US,
UK or Germany, be at least 18 years old, and report having engaged in any paid work
during the previous 12 months, either as an employee or self-employed.7Within each
country, the samples were selected to be representative in terms of region. Appendix
Tables A.1 to A.3 show the distribution of respondents across regions and the compar-
ison to the national distribution of individuals across the different regions, separately
for the three countries in our sample and for each survey wave. As can be seen from
the tables, for all countries and survey waves the two distributions are very similar.
We compare the characteristics of the respondents in our sample to nationally rep-
resentative samples of the working population in each respective country. Appendix
Table A.4 shows the demographic characteristics of a nationally representative sam-
ple and our samples.8While there are some differences between our samples and the
nationally representative samples, all our results are robust to re-weighing our sample
using survey weights.9We present unweighted results throughout the text and weighted
results in the Appendix.
Because we are interested in the recent labor market impact of the COVID-19
pandemic, in all subsequent analysis we focus on respondents who are either still in
work at the time of the survey or lost their job less than a month before the data
7We asked participants to think about all the paid work they engaged in other than completing
surveys.
8For the US, we use the February 2020 monthly CPS data, for the UK the 2019 Labour Force
Survey data, and for Germany the 2017 SOEP data as a benchmark.
9We re-weight our sample to ensure that the joint density of gender, education, and age in our
samples matches that of the economically active population in each respective country.
7

collection due to the coronavirus outbreak. More detail on how we elicit this information
is provided below.
2.3 Survey Design
Information on Employment In all countries and survey waves, we collect detailed
information on respondents’ current work arrangements. We ask respondents to report
how many jobs they have been working in over the past 7 days, either as employees
or as self-employed.10 Respondents who report having at least one job are asked to
provide details on their main job as well as on their second job if they have one. We
also ask all respondents how many hours they worked in the previous week and how
many hours they worked in a typical week in February.
For each job, we collect detailed information on different job characteristics, includ-
ing occupation and industry.11 We further ask respondents to state whether they are
employed or self-employed in this job. Importantly, we ask all respondents what per-
centage of their tasks they could do from home. Answers were recorded using a slider
ranging from 0-100%. To ease comprehension of this question, we provided partici-
pants with some examples. ‘E.g. Andy is a waiter and cannot do any of his work from
home (0%). Beth is a website designer and can do all her work from home (100%)’.
If a respondent reports being employed in any of their jobs, they are further asked
to report whether they are on a permanent or temporary contract, whether their work
schedule is fixed or flexible, and whether they are salaried or non-salaried, i.e. paid in
a different way for their work (e.g. by the hour).
Individuals who report not having a job are asked similar questions about their last
main job. In addition, they are asked to provide information on when they lost their
last job and whether they think they lost their job because of the coronavirus crisis.
Answers to the latter were recorded on a 5-point Likert scale. We classify individuals
as having lost their job due to coronavirus if (i) they lost their job in the four weeks
before data collection, and (ii) if they answer ‘definitely yes’ or ‘probably yes’ to the
question on how likely it was that their job loss could be attributed to the coronavirus
outbreak.
10In the early April wave, in which we also asked about furloughing, we made it explicit that
individuals should count all jobs, including the ones in which they have been furloughed.
11For the US and the UK, we use the Standard Occupations Classification 2018 major groups for our
occupation grouping and the Standard Industry Classification for our industry grouping. For Germany,
we use the main categories from the ISCO-08 classification for the occupation grouping.
8

Information on STW/Furloughing To obtain a better understanding of the use of
furloughing and STW schemes, in the early April survey wave, we included questions on
furloughing and STW. In the US and the UK, if respondents reported being employed
in any of their jobs we asked them to report whether they have been furloughed, and,
if yes, whether they have still been asked by their employer to do any work. In the
UK, respondents provided us with additional information on whether their employer
is topping up the government wage support, and whether they lost any annual leave
entitlements. In the US, we additionally asked whether employees lost their health
insurance coverage. In Germany, we asked employees whether they were on the STW
scheme. We further asked respondents to state the official share of their usual hours
that they are asked to work, and for the share of hours that they actually work.
Monthly Earnings To obtain a clearer picture of the impacts of the crisis and the
earnings lost, we ask all individuals in the early April survey wave to report their net
monthly earnings from all sources for the months of January, February, and March.
Throughout the paper, we define ‘earnings loss’ as a binary variable that takes a value
of one if a respondent earned less in March 2020 compared to his / her average earnings
over the months of January and February 2020. We also ask respondents to state
whether they have already struggled to pay their usual bills or expenses.
Time Use In the early April survey wave, we asked respondents directly about their
time use on a typical working day over the past week. For individuals with children
living in the household, we asked about the number of hours and minutes spent on
active childcare and on homeschooling.
Expectations for the Future To obtain a better sense of how individuals think
about their future labor market outcomes, we ask respondents how likely they think it
is that certain events will occur before August 1st, 2020, on a 0-100% chance scale. Most
notably, those include whether respondents think they will lose their job or shut their
business (if self-employed), and have trouble paying their usual bills and expenses. To
understand how long individuals think the crisis will last, we also asked all individuals in
the second wave how likely they think it is that some form of social distancing measures
will still be in place on August 1st, 2020, using a 0-100% scale.
9

3 Impacts by Job Characteristics
The COVID-19 crisis has had large and unequal impacts on workers in all three coun-
tries. In late March, 11% and 8% of respondents report having lost their jobs within
the last four weeks due to the coronavirus outbreak in the US and UK, respectively. In
early April, those figures rose to 18% (US) and 15% (UK). These figures stand in stark
contrast to the figures from Germany, where only 5% of respondents report having lost
their jobs in early April.
While there are staggering cross-country differences in the percentage of workers who
lost their jobs, there are certain notable similarities in terms of who was most affected
by the crisis. The outbreak has caused significant disruptions to the economy but the
impact has been unequal across different types of jobs. An important characteristic of
a job is the % of tasks individuals can do from home. Figure 1 displays the percentage
of people who lost their job due to the coronavirus outbreak by the percentage of tasks
individuals report being able to do from home (summarized into quintiles). In all three
countries, there is a clear monotonic relationship between the percentage of tasks one
can do from home and job loss. In the US and the UK, this relationship has become
even steeper as the crisis has unfolded. The most salient cross-country differences in
job loss can be observed in the bottom quintile of the distribution. While 40.1% of
workers in the bottom quintile lost their jobs in the US, the corresponding figure is
7.6% in Germany.
Figure 2 displays the probability of job loss across different occupations in the
US (left), UK (center) and Germany (right). Appendix Figure B.2 gives the results by
industry. The percentage of people having lost their jobs varies substantially across the
different occupations and industries. We see that both in the US and the UK people
working in “food preparation and serving” and “personal care and service” are very
likely to have lost their job due to the pandemic. On the other side of the spectrum,
people working in “computer and mathematical” occupations or “architecture and engi-
neering” have been most likely to keep their job. Similarly in Germany, people working
in “auxiliary” and “mechanical” occupations had the highest likelihood of losing their
job, while “technicians” and people working in “office and administration” had among
the lowest.
Turning to differences in job loss for employed workers by work arrangements, Fig-
ure 3 shows the differences in job loss probabilities depending on whether the individ-
ual was employed (i) on a temporary or permanent contract, (ii) had a non-salaried or
10

Figure 1: Job loss probability due to Covid-19 by % tasks that can be done from home
0
.1
.2
.3
.4
Share of workers that lost job due to due Corona
1st
2nd
3rd
4th
5th
Work from home - Quintiles
US
0
.1
.2
.3
.4
Share of workers that lost job due to due Corona
1st
2nd
3rd
4th
5th
Work from home - Quintiles
UK
0
.1
.2
.3
.4
Share of workers that lost job due to due Corona
1st
2nd
3rd
4th
5th
Work from home - Quintiles
Germany
Late March Early April
Notes: The quintiles on the x-axis are defined by the share of tasks the respondents report that they
can do from home. The thin black bars represent the 95% confidence intervals. The figure shows the
share of individuals who were in paid work four weeks before data collection that lost their job due to
Covid-19.
salaried job, and (iii) had varying or fixed hours. We observe the same pattern in all
three countries. Workers with permanent, salaried, fixed hour contracts were less likely
to be affected compared to workers who were on temporary contracts, non-salaried and
whose hours varied.
The share of tasks that can be done from home within occupation and industry is
a powerful predictor of the share of workers that lost their jobs. It alone can explain
69%, 54% and 58% of the variation in job loss due to Covid-19 across occupations in
the US, the UK and Germany, respectively (Figure B.3). As can be seen in Figure B.3
in occupations in which a larger share of tasks can be done from home (x-axis) the
job loss probability due to Covid-19 (y-axis) has been much lower. We find a similar
11

Figure 2: Job loss probability due to Covid-19 by occupation
0
.1
.2
.3
.4
.5
Farming, Fishing, and Forestry
Computer and Mathematical
Architecture and Engineering
Business and Financial Operations
Healthcare Support
Legal
Community and Social Service
Management
Installation, Maintenance, and Repair
Office and Administrative Support
Life, Physical, and Social Science
Building and Grounds Cleaning and Maintenance
Construction and Extraction
Educational Instruction and Library
Healthcare Practitioners and Technical occ.
Sales and Related Occupations
Arts, Design, Entertainment, Sports, and Media
Personal Care and Service
Production
Transportation and Material Moving
Food Preparation and Serving
US - Early April
0
.1
.2
.3
.4
.5
Computer and Mathematical
Architecture and Engineering
Life, Physical, and Social Science
Protective Service
Community and Social Service
Business and Financial Operations
Healthcare Support
Management
Legal
Office and Administrative Support
Educational Instruction and Library
Healthcare Practitioners and Technical occ.
Production
Transportation and Material Moving
Arts, Design, Entertainment, Sports, and Media
Sales and Related Occupations
Installation, Maintenance, and Repair
Construction and Extraction
Food Preparation and Serving
Personal Care and Service
Building and Grounds Cleaning and Maintenance
UK - Early April
0
.1
.2
.3
.4
.5
Military
Technician and comparable non-technical
Office and administration
Management
Craftsmen and women
Service and retail
Academic
Farming, fishing, and forestry
Mechanical
Auxiliary
Germany - Early April
Notes: The thin black bars represent the 95% confidence intervals. The figure shows the share of
individuals who were in paid work four weeks before data collection that lost their job due to Covid-19
by occupation.
pattern when we investigate the relationship between the share of tasks that can be
done from home and job loss across industries (Figures B.4).
Appendix Figure B.6 shows the average share of tasks that can be done from home
by occupation (y-axis) and industry (x-axis), while Appendix Figure B.7 shows the
share of jobs lost due to Covid-19. Occupations in industries in which less tasks can be
done from home have seen more jobs being lost due to the pandemic. Whether or not
the share of tasks one can do from home predicts job loss over and above what can be
predicted by occupation and industry is a question we explore in Section 4.12
12In Appendix Figures B.1 and B.5 we see that even within occupations and industries the share of
tasks that can be done from home varies substantially.
12

Figure 3: Job loss probability due to Covid-19 by work arrangements
0
.1
.2
.3
.4
Job lost due to due Corona
US - Early April
0
.1
.2
.3
.4
Job lost due to due Corona
UK - Early April
0
.1
.2
.3
.4
Job lost due to due Corona
Germany - Early April
Temporary Permanent Not Salaried Salaried Vary Hours Fixed Hours
Notes: The thin black bars represent the 95% confidence intervals. The figure shows the share of
individuals who were employees four weeks before data collection that lost their job due to Covid-19.
13

Furloughing and STW Another response to the coronavirus crisis has been the
introduction and increased use of furloughing and STW schemes. In early April,
31% (US), 43% (UK) and 35% (Germany) of employees report being furloughed or
in STW. Figure 4 shows the percentage of employees who are still employed without
being furloughed or in STW (blue), as well as the percentage of employees who have
been furloughed or in STW (yellow) or laid off (red) by the percentage of tasks in-
dividuals report being able to do from home (again summarized into quintiles).13 In
all three countries, the percentage of employees being furloughed or in STW is sub-
stantial and increases somewhat with the percentage of tasks one can do from home.
Figures B.11 and B.12 in the Appendix show the same breakdown by occupation and
industry, respectively. There is substantive variation in the extent to which employees
were furloughed across industries. In the UK, for example, 68% of employees working
in the “mining and quarrying” industry were furloughed, against a figure of around 5%
for “public administration and defence”. Similarly, furloughing and STW schemes are
differentially used across occupations. Within countries, we also see significant vari-
ation in the terms of furloughing. In the UK for example, employers can choose to
top up the wage of their furloughed employees and 70% of our respondents who were
furloughed report that their employer offered to do so. However, 50% of employees
in the UK were also asked to take annual leave and 15% of them were asked to work
while on furlough. In the US, 53% of employees who were furloughed also lost their
health insurance coverage. Remarkably in Germany, we find no difference between the
percentage of hours that employees were officially asked to work while on STW (49%
on average) and the hours that they actually work (50% on average).
Impact on Hours Worked Conditional on Working Job loss is only one aspect
of the labor market shock. Workers who have kept their job might now be working
different hours. Adjustment on the intensive margin could be driven by changes in
the level and distribution of aggregate economic activity or by changes in labor supply
arising from health restrictions or other responsibilities such as child care. Among those
who still had a paid job in early April, we observe a stark decline in the number of
hours worked. The average change in hours worked (compared to a typical week in
February) was 5 hours (US), 7 hours (UK) and 4 hours (Germany). Figure 5 shows the
average change in hours worked by occupation amongst workers still working. The x-
13Note that in the figures, “Furloughing” should be interpreted as the STW scheme in Germany.
14

Figure 4: Employment status by % of tasks that can be done from home
0
20
40
60
80
100
1st Quintile
2nd Quintile
3rd Quintile
4th Quintile
5th Quintile
US
0
20
40
60
80
100
1st Quintile
2nd Quintile
3rd Quintile
4th Quintile
5th Quintile
UK
0
20
40
60
80
100
1st Quintile
2nd Quintile
3rd Quintile
4th Quintile
5th Quintile
Germany
Lost job Furloughed Employed
Notes: The figure shows the share of individuals who are employed, furloughed or lost their job due
to the COVID-19 crisis, by the percentage of tasks workers report being able to do from home. The
sample is restricted to employees (in their main or last job) only.
axis displays the difference between the two. We see that, across all occupations, those
in paid work are working fewer hours on average. However, there is large variation
across occupations and sectors. For instance, in the UK workers in “computer and
mathematical” occupations on average saw hardly any change in hours worked , while
for those working in “educational instruction and library” the average drop in hours
worked was about 12 hours in the given week. In Appendix Figure B.9, we see that
occupations that saw the largest drop in hours also saw the largest share of workers
laid off. Note that this is not a mechanical effect as the reduction in hours worked is
amongst those that are still working so the change only reflects the intensive margin.
In Appendix Figures B.8 and B.10 we see that the same patterns hold within industry.
15

Figure 5: Change in hours worked by occupation
-15
-10
-5
0
Food Preparation and Serving
Educational Instruction and Library
Production
Transportation and Material Moving
Sales and Related Occupations
Arts, Design, Entertainment, Sports, and Media
Building and Grounds Cleaning and Maintenance
Personal Care and Service
Office and Administrative Support
Installation, Maintenance, and Repair
Construction and Extraction
Legal
Healthcare Practitioners and Technical occ.
Community and Social Service
Management
Business and Financial Operations
Computer and Mathematical
Healthcare Support
Architecture and Engineering
Life, Physical, and Social Science
US - Early April
-15
-10
-5
0
Educational Instruction and Library
Personal Care and Service
Construction and Extraction
Food Preparation and Serving
Installation, Maintenance, and Repair
Arts, Design, Entertainment, Sports, and Media
Production
Sales and Related Occupations
Transportation and Material Moving
Office and Administrative Support
Healthcare Practitioners and Technical occ.
Management
Protective Service
Building and Grounds Cleaning and Maintenance
Business and Financial Operations
Healthcare Support
Architecture and Engineering
Legal
Community and Social Service
Life, Physical, and Social Science
Computer and Mathematical
UK - Early April
-15
-10
-5
0
5
Mechanical
Service and retail
Management
Academic
Office and administration
Craftsmen and women
Technician and comparable non-technical
Auxiliary
Military
Farming, fishing, and forestry
Germany - Early April
Notes: The thin black bars represent the 95% confidence intervals. The figure shows the average change
in hours worked between a usual week in February and the last week by occupation, for individuals
that were in paid work at the time of data collection.
16

4 Predictors of Job and Earnings Loss
We now move on to analyzing the predictors of job and earnings loss in a regression
framework, where we estimate linear probability models focusing on data from the early
April wave. Columns (1) - (3) of Table 1 show regressions where the dependent variable
is a binary variable for having lost one’s job in the last month because of coronavirus.
All specifications control for region, occupation, and industry fixed effects. In all three
countries, workers’ ability to perform more of their tasks from home is associated with
a lower likelihood of them losing their job. Interestingly, this relationship survives even
when we control for occupation and industry fixed effects, suggesting that the variation
in the percentage of tasks one can do from home within an occupation also plays an
important role in explaining differences in job loss probabilities.
Table 1 also speaks to the importance of contractual arrangements in sheltering
workers from the economic downturn that the COVID-19 outbreak induced. Controlling
for workers’ ability to work from home and the occupation and industry they work in,
we find that employees in less secure work arrangements are more likely to have lost
their jobs following the coronavirus outbreak. In the UK, employees with a permanent
contract are 17 percentage points less likely to have lost their job relative to employees
on temporary contracts. In the US and Germany, permanent employees are 7 and
5 percentage points less likely to now be out of work. Salaried employees in the US
(Germany) were 6 (2) percentage points less likely to lose their jobs relative to non-
salaried employees.14
Among the respondents in our sample who still have a paid job in early April,
35% (US), 30% (UK) and 20% (Germany) report having had lower earnings in March
(compared to Jan-Feb). We now investigate which job characteristics predict whether
individuals experienced a drop in earnings. As can be seen in Columns 4 and 5 of
Table 1, the probability of a fall in labor earnings is larger for workers in the US and
the UK who can perform fewer of their tasks from home. For Germany we do not find
a similar association. In the US (UK), individuals who can perform all of their tasks
from home are 25 (15) percentage points less likely to have suffered a fall in earnings
compared to individuals who cannot work from home.
As for job loss, the likelihood of earnings loss significantly varies with work ar-
14In Appendix Table B.5 we pool the first and second survey wave for the US and the UK and
additionally control for a dummy variable indicating whether respondents were part of the second
survey wave. Individuals in the second wave were significantly more likely to report having lost their
job. All other results are robust to using both survey waves.
17

Table 1: Job and earnings loss probability
Job loss Earnings loss
US UK DE US UK DE
(1) (2) (3) (4) (5) (6)
Tasks from Home -0.2617∗∗∗ -0.1917∗∗∗ -0.0397∗∗∗ -0.1328∗∗∗ -0.0737∗∗∗ -0.0202
(0.0216) (0.0195) (0.0128) (0.0303) (0.0267) (0.0233)
Self-Employed -0.0996∗∗∗ -0.0463∗0.0051 0.0224 0.0945∗∗ 0.0615∗
(0.0228) (0.0257) (0.0174) (0.0320) (0.0373) (0.0322)
Permanent -0.0659∗∗∗ -0.1711∗∗∗ -0.0546∗∗∗ -0.0116 -0.0224 0.0030
(0.0165) (0.0205) (0.0114) (0.0233) (0.0302) (0.0210)
Salaried -0.0632∗∗∗ 0.0110 -0.0193∗-0.0911∗∗∗ -0.0455∗∗ -0.0629∗∗∗
(0.0181) (0.0154) (0.0108) (0.0248) (0.0207) (0.0197)
Fixed Hours 0.0022 -0.0094 0.0035 -0.0714∗∗∗ -0.1108∗∗∗ -0.0927∗∗∗
(0.0164) (0.0151) (0.0097) (0.0232) (0.0203) (0.0175)
Constant 0.4475∗∗∗ 0.2720∗∗∗ 0.1288∗∗∗ 0.3757∗∗∗ 0.3765∗∗∗ 0.2933∗∗∗
(0.0875) (0.0667) (0.0355) (0.1208) (0.0886) (0.0645)
Observations 2995 3760 3354 2396 3111 3165
R20.1600 0.1138 0.0654 0.1057 0.0890 0.0671
Region F.E. yes yes yes yes yes yes
Occupation F.E. yes yes yes yes yes yes
Industry F.E. yes yes yes yes yes yes
Notes: OLS regressions. The dependent variable in Columns 1 - 3 is a binary variable for whether a respondent lost their
job within the past month and attributed the job loss to the coronavirus outbreak. The dependent variable in Columns 4
- 6 is a binary variable for whether a respondent earned less in March 2020 than the average earnings over January and
February 2020. In Columns 4 - 6 the sample is restricted to those who were in work at the time of data collection. Tasks
from Home is the fraction of tasks respondents could do from home in their main or last job. Self-employed is a binary
variable for being self-employed in the main or last job. Permanent, salaried and fixed hours take value 1 for employees
with permanent contracts, who are salaried and whose work hours are fixed, respectively. Region fixed effects refer to state
fixed effects for the US and Germany, and fixed effects for regions as reported in Table A.1 for the UK.
rangements in all three countries. Amongst those who have kept their job, salaried
employees and those with fixed work schedules have been relatively sheltered from the
shock. We find that salaried employees are between 5 and 9 percentage points less likely
to have seen their earnings fall between January-February and March 2020, compared
to non-salaried employees. Similarly, employees with fixed hour contracts have a 7-11
percentage point lower likelihood of losing any of their earnings compared to workers
whose work hours vary.
18

5 Impacts by Individual Characteristics
An important question that emerges is whether the impact of the COVID-19 outbreak
varies across individuals with different background characteristics. Table 2 shows the
results from linear probability models in which the dependent variable is job loss. The
results in Columns (1) and (3) suggest that in the US and UK women were significantly
more likely to lose their jobs, while people with a university degree were significantly
less likely to experience job loss. The magnitudes of the effects are large. Women in the
US (UK) are 7 (5) percentage points more likely to lose their jobs (compared to men),
while workers with a college degree in the US (UK) were 8 (6) percentage points less
likely to lose their jobs (compared to workers without a college degree). In Germany we
find that neither gender nor a university degree predict job loss significantly. However,
in Germany we do find that those under the age of 30 were more likely to lose their job.
In the US and UK, we find a large gender gap in respondents’ ability to work
from home: in the US (UK), women on average report they can do 42% (41%) of
their tasks from home, compared to 53% (46%) for men. In contrast, in Germany we
find no significant difference: men report that 41% of their tasks can be done from
home and women report 39%. Further, previous literature shows that men and women,
as well as workers with different levels of educational attainment, sort into different
occupations. In order to take these differences into account, in Columns (2), (4) and
(6) we additionally control for the percentage of tasks that can be done from home as
well as occupation and industry fixed effects. The coefficient on university education is
no longer significant in these specifications and estimated to be close to zero, indicating
that the percentage of tasks one can do from home and occupation dummies can explain
most if not all of the variation in job loss across the two education groups. In contrast,
the gender coefficient is still positive and significant in the US and UK, albeit reduced
in size, suggesting that other factors we are not capturing in this regression play a role
in driving the gender gaps.
One potential reason for these gender differences is that women are spending more
time homeschooling and caring for children. Figure B.14 presents the average number of
hours that men and women who are working from home reported spending on different
activities during a typical work day. As can be seen from this figure, women spend a
lot more time on childcare than men. In Appendix Table B.6, we show that restricting
the sample to those that are spending some time working from home and controlling
for a range of individual, job, and geographic characteristics, we still find that women
19

Table 2: Job loss probability - Individual characteristics
United States United Kingdom Germany
(1) (2) (3) (4) (5) (6)
Female 0.0652∗∗∗ 0.0321∗∗ 0.0483∗∗∗ 0.0242∗0.0014 -0.0002
(0.0151) (0.0157) (0.0124) (0.0129) (0.0077) (0.0084)
University degree -0.0789∗∗∗ -0.0050 -0.0629∗∗∗ -0.0070 -0.0116 0.0071
(0.0151) (0.0161) (0.0123) (0.0131) (0.0083) (0.0098)
30-39 -0.0325 -0.0043 0.0222 0.0304∗-0.0436∗∗∗ -0.0188∗
(0.0201) (0.0195) (0.0156) (0.0156) (0.0097) (0.0103)
40-49 -0.0286 -0.0087 0.0259 0.0229 -0.0343∗∗∗ -0.0143
(0.0214) (0.0209) (0.0171) (0.0173) (0.0115) (0.0124)
50-59 0.0005 0.0171 0.0036 -0.0074 -0.0338∗∗∗ -0.0207
(0.0247) (0.0241) (0.0215) (0.0216) (0.0120) (0.0127)
60+ 0.0135 0.0111 0.0256 0.0111 0.0318 0.0289
(0.0257) (0.0253) (0.0366) (0.0359) (0.0201) (0.0207)
Tasks from home -0.2574∗∗∗ -0.1913∗∗∗ -0.0406∗∗∗
(0.0219) (0.0197) (0.0132)
Self-Employed -0.1003∗∗∗ -0.0477∗0.0059
(0.0230) (0.0260) (0.0176)
Permanent -0.0639∗∗∗ -0.1720∗∗∗ -0.0511∗∗∗
(0.0166) (0.0206) (0.0116)
Salaried -0.0592∗∗∗ 0.0112 -0.0193∗
(0.0185) (0.0156) (0.0109)
Fixed Hours 0.0018 -0.0123 0.0057
(0.0165) (0.0152) (0.0097)
Constant 0.2371∗∗∗ 0.4311∗∗∗ 0.1191∗∗∗ 0.2454∗∗∗ 0.0857∗∗∗ 0.1317∗∗∗
(0.0689) (0.0888) (0.0253) (0.0678) (0.0132) (0.0358)
Observations 3025 2995 3816 3760 3584 3354
R20.0448 0.1618 0.0169 0.1161 0.0170 0.0679
Region F.E. yes yes yes yes yes yes
Occupation F.E. no yes no yes no yes
Industry F.E. no yes no yes no yes
Notes: OLS regressions. The dependent variable is a binary variable for whether a respondent lost their job within the
past month and attributed the job loss to the coronavirus outbreak. Tasks from home is the fraction of tasks respondents
could do from home in their main or last job. Self-employed is a binary variable for being self-employed in the main or
last job. Permanent, salaried and fixed hours take value 1 for employees with permanent contracts, who are salaried and
whose work hours are fixed, respectively. Region fixed effects refer to state fixed effects for the US and Germany, and fixed
effects for regions as reported in Table A.1 for the UK.
20

spent about one hour more on childcare and home schooling. However, in Germany
we find a sizeable gender gap in active time spent on children but no such gap in the
probability of job loss.
Appendix Figure B.13 presents the coefficients of the occupation fixed effects from
the regressions in Columns (2), (4) and (6) in Table 2. We see that qualitatively the
estimated coefficients resemble the unconditional patterns presented in Figure 2. “Food
preparation and serving”, for instance, is associated with a 13 (12) percentage point
higher job loss probability in the US (UK).15
Table B.4 presents results from linear probability models in which the dependent
variable is whether or not the individual has experienced an earnings loss between
January-February and March 2020, and where the sample is restricted to respondents
who report still being in work in April 2020. In all three countries, women who did
not lose their job were no more likely to experience a fall in their income compared to
men. In the US and the UK, college-educated workers still in work were less likely to
experience a fall in their earnings compared to workers without a college degree. We
do not find a similar pattern in Germany.
15The industry fixed effects are less precisely estimated (not presented) suggesting that occupation
might be the dimension which is better at explaining job loss.
21

6 Expectations for the Future
Focusing on workers in the second survey wave who still report having a job, we find
that individual outlooks on the future are bleak. On average, those still in work report
a perceived likelihood of losing their job within the next few months of 37% and 32%
in the US and UK. In Germany, where job loss has been much less prevalent, still 25%
fear losing their job over the next months. Table 3 shows the results of least square
regressions in which we show what characteristics predict individual perceptions of the
likelihood of job loss. We find that older workers and employees on more secure work
arrangements perceive a lower chance of job loss, with the exception of workers on
permanent contracts in the US. Interestingly, women and those who report being able
to do fewer tasks from home are more optimistic about their chance of keeping their job
in the US and UK. This stands in contrast to the realized experience of these groups
so far.
We also analyze whether individual beliefs about the likelihood of social distancing
measures still being in force in August 2020 are associated with their job loss percep-
tions. Individuals believe it is likely that some form of social distancing measures will
be in place at the end of the summer; the average response to this question was 58% in
the US, 62% in the UK, and 53% in Germany. Those who believe that social distancing
measures will persist into the summer perceive the chance that they will lose their job
as significantly higher.
All respondents irrespective of their current employment status were further asked
about their perceived likelihood of struggling to pay their usual bills and expenses in the
future. The average response to this question was 53% in the US, 46% in the UK, and
33% in Germany, indicating that many individuals think they will struggle financially.
Indeed, 46%, 38%, and 32% of individuals in the US, UK, and Germany report that
they have already had more difficulties meeting their usual bills and expenses compared
to normal. Providing timely assistance to those most affected should be a high priority.
22

Table 3: Perceived probability of job loss
United States United Kingdom Germany
(1) (2) (3) (4) (5) (6)
Female -0.0998∗∗∗ -0.0590∗∗∗ -0.0533∗∗∗ -0.0100 0.0222∗∗ 0.0447∗∗∗
(0.0138) (0.0141) (0.0106) (0.0107) (0.0102) (0.0095)
University degree 0.0198 0.0167 0.0136 0.0092 0.0656∗∗∗ 0.0327∗∗∗
(0.0140) (0.0146) (0.0106) (0.0108) (0.0112) (0.0112)
30-39 0.0129 0.0075 -0.0491∗∗∗ -0.0243∗-0.0001 0.0152
(0.0185) (0.0176) (0.0135) (0.0129) (0.0128) (0.0117)
40-49 0.0084 0.0022 -0.1407∗∗∗ -0.0873∗∗∗ -0.0909∗∗∗ -0.0216
(0.0195) (0.0189) (0.0147) (0.0144) (0.0153) (0.0140)
50-59 -0.1269∗∗∗ -0.0849∗∗∗ -0.2361∗∗∗ -0.1571∗∗∗ -0.1465∗∗∗ -0.0609∗∗∗
(0.0229) (0.0220) (0.0183) (0.0177) (0.0156) (0.0143)
60+ -0.2102∗∗∗ -0.1505∗∗∗ -0.2514∗∗∗ -0.2087∗∗∗ -0.1858∗∗∗ -0.1124∗∗∗
(0.0239) (0.0232) (0.0317) (0.0299) (0.0270) (0.0241)
Tasks from home 0.1105∗∗∗ 0.1180∗∗∗ 0.1385∗∗∗
(0.0200) (0.0166) (0.0152)
Self-Employed 0.0059 -0.1077∗∗∗ -0.0932∗∗∗
(0.0206) (0.0231) (0.0205)
Permanent 0.0443∗∗∗ -0.0778∗∗∗ 0.0023
(0.0152) (0.0186) (0.0135)
Salaried -0.0244 -0.0297∗∗ -0.1086∗∗∗
(0.0163) (0.0129) (0.0125)
Fixed Hours -0.0368∗∗ -0.0587∗∗∗ -0.0297∗∗∗
(0.0150) (0.0125) (0.0111)
Measures still in August 0.3562∗∗∗ 0.2170∗∗∗ 0.2147∗∗∗
(0.0238) (0.0203) (0.0164)
Constant 0.3804∗∗∗ 0.1608∗∗ 0.4165∗∗∗ 0.3478∗∗∗ 0.3368∗∗∗ 0.3363∗∗∗
(0.0639) (0.0801) (0.0214) (0.0563) (0.0182) (0.0420)
Observations 2402 2382 3115 3094 3179 3116
R20.1320 0.2713 0.0831 0.2333 0.0792 0.3085
Region F.E. yes yes yes yes yes yes
Occupation F.E. no yes no yes no yes
Industry F.E. no yes no yes no yes
Notes: OLS regressions. The dependent variable is a binary variable for whether a respondent lost their job within the
past month and attributed the job loss to the coronavirus outbreak. Tasks from home is the fraction of tasks respondents
could do from home in their main or last job. Self-employed is a binary variable for being self-employed in the main or
last job. Permanent, salaried and fixed hours take value 1 for employees with permanent contracts, who are salaried and
whose work hours are fixed, respectively. ‘Measures still in August’ refers to the perceived probability of some social dis-
tancing measures being in place in August. Region fixed effects refer to state fixed effects for the US and Germany, and
fixed effects for regions as reported in Table A.1 for the UK.
23

7 Conclusion
The COVID-19 crisis has had large impacts on the economy. The results from our
study suggest that the impacts are highly unequal. The percentage of tasks workers
can do from home is highly predictive of job loss and so are individual work arrange-
ments. Firms have played some role in smoothing the shock for permanent and salaried
employees, and for those who usually work on fixed schedules.
In the US and UK women and workers without a college degree are significantly
more likely to already have lost their jobs, while younger individuals are significantly
more likely to experience a fall in their earnings. The outlook on the future is bleak with
many workers expecting to lose their jobs over the next months. The results highlight
the need for immediate policy responses that target those groups in the population that
are most affected by the crisis.
Finally, we find large differences in the magnitude of the shock between the an-
glophone countries, the US and the UK, versus Germany. The anglophone countries
have seen much more employment ties cut. This might not only increase the share
of population suffering hardship at the moment, but could also prove important for
recovery as well due to the need for matching between workers and firms and the loss in
employer-employee specific human capital. Further research into understanding which
institutional factors are driving these differences is of high policy importance.
24

References
Alfaro, Laura, Anusha Chari, Andrew Greenland, and Peter Schott. 2020.
“Aggregate and firm-level stock returns during pandemics in real time.” National
Bureau of Economic Research 26950.
Allcott, Hunt, Levi Boxell, Jacob C Conway, Matthew Gentzkow, Michael
Thaler, and David Y Yang. 2020. “Polarization and public health: Partisan dif-
ferences in social distancing during the Coronavirus pandemic.” National Bureau of
Economic Research 26946.
Alon, Titan, Matthias Doepke, Jane Olmstead-Rumsey, and Michele Tertilt.
2020. “The impact of COVID-19 on gender equality.” National Bureau of Economic
Research 26947.
Baker, Scott, Nicholas Bloom, Steven Davis, Kyle Kost, Marco Sammon,
and Tasaneeya Viratyosin. 2020a. “The unprecedented stock market impact of
COVID-19.” National Bureau of Economic Research 26945.
Baker, Scott R., R.A. Farrokhnia, Steffen Meyer, Michaela Pagel, and Con-
stantine Yannelis. 2020b. “How does household spending respond to an epidemic?
Consumption during the 2020 COVID-19 pandemic.” National Bureau of Economic
Research 26949.
Bartik, Alexander W., Marianne Bertrand, Zoe B. Cullen, Edward L.
Glaeser, Michael Luca, and Christopher T. Stanton. 2020. “How are small
businesses adjusting to COVID-19? Early evidence from a survey.” National Bureau
of Economic Research 26989.
Bick, Alexander, and Adam Blandin. 2020. “Real Time Labor Market Estimates-
During the 2020 Coronavirus Outbreak.” Mimeo.
Borjas, George J. 2020. “Demographic determinants of testing incidence and COVID-
19 infections in New York City neighborhoods.” National Bureau of Economic Re-
search 26952.
Bredemeier, Christian, Falko Juessen, and Roland Winkler. 2017. “Man-
cessions, fiscal policy, and the gender composition of employment.” Economics Let-
ters, 158: 73–76.
Cahuc, Pierre, Francis Kramarz, and Sandra Nevoux. 2018. “When short-time
work works.” CEPR Discussion Paper 4654-1530893226.
Carvalho, Vasco. M., Juan R. Garcia, Stephen Hansen, Álvaro Ortiz,
Tomasa Rodrigo, José V. Rodríguez Mora, and José Ruiz. 2020. “Track-
25

ing the COVID-19 Crisis with High-Resolution Transaction Data.” Cambridge-INET
Working Paper Series No: 2020/16.
Chiou, Lesley, and Catherine E. Tucker. 2020. “Social distancing, internet access
and inequality.” National Bureau of Economic Research 26982.
Christiano, Lawrence J, Martin S Eichenbaum, and Mathias Trabandt. 2015.
“Understanding the great recession.” American Economic Journal: Macroeconomics,
7(1): 110–67.
Dingel, Jonathan, and Brent Neiman. 2020. “How many jobs can be done at
home?” National Bureau of Economic Research 26948.
Fetzer, Thiemo, Lukas Hensel, Johannes Hermle, and Chris Roth. 2020.
“Coronavirus Perceptions and Economic Anxiety.” arXiv:2003.03848.
Giupponi, Giulia, and Camille Landais. 2018. “Subsidizing labor hoarding in re-
cessions: The employment & welfare effects of short time work.”
Hoynes, Hilary, Douglas L. Miller, and Jessamyn Schalle. 2012. “Who suffers
during recessions?” Journal of Economic Perspectives, 26(3): 27–48.
Kahn, Lisa B., Fabian Lange, and David Wiczer. 2020. “Labor Demand in the
time of COVID-19: Evidence from vacancy postings and UI claims.” Mimeo.
Kopp, Daniel, and Michael Siegenthaler. 2018. “Does short-time work prevent
unemployment?” KOF Swiss Economic Institute, ETH Zurich 106.
Koustas, Dmitri. 2018. “Consumption insurance and multiple jobs: Evidence from
rideshare drivers.” Unpublished working paper.
Malcomson, James M. 1999. “Individual employment contracts.” Handbook of labor
economics, 3: 2291–2372.
Mas, Alexandre, and Amanda Pallais. 2020. “Alternative work arrangements.”
National Bureau of Economic Research.
Mongey, Simon, and Alex Weinberg. 2020. “Characteristics of workers in low
work-from-home and high personal-proximity occupations.” Working Paper.
Office for National Statistics. 2019. “Estimates of the population for
the UK, England and Wales, Scotland and Northern Ireland.” Data re-
trieved from https://www.ons.gov.uk/peoplepopulationandcommunity/
populationandmigration/populationestimates/datasets/
populationestimatesforukenglandandwalesscotlandandnorthernireland.
Rinne, Ulf, and Klaus F Zimmermann. 2012. “Another economic miracle? The
German labor market and the Great Recession.” IZA journal of labor policy, 1(1): 3.
Statistische Ämter des Bundes und der Länder. 2018. “Fläche und
26

Bevölkerung nach Ländern.” Data retrieved from https://www.statistikportal.
de/de/bevoelkerung/flaeche-und-bevoelkerung.
U.S. Census Bureau, Population Division. 2019. “Estimates of the Total
Resident Population and Resident Population Age 18 Years and Older for the
United States, States, and Puerto Rico: July 1, 2019 (SCPRC-EST2019-18+POP-
RES).” Data retrieved from https://www.census.gov/newsroom/press-kits/
2019/national-state-estimates.html.
27

A Online Appendix A: Data Description
Table A.1: Distribution of respondents across regions - UK
Region National Late March Early April
Scotland 8.42 8.48 8.54
Northern Ireland 2.76 2.57 2.80
Wales 4.79 4.83 4.87
North East 4.06 4.08 4.12
North West 11.00 11.02 11.11
Yorkshire and the Humber 8.24 8.28 8.34
West Midlands 8.80 8.86 8.92
East Midlands 7.27 7.32 7.38
South West 8.59 8.63 8.70
South East 13.70 13.79 13.87
East of England 9.29 8.91 8.03
Greater London 13.15 13.24 13.32
Observations 3974 4931
Notes: National figures refer to the latest available estimates for the population of residents
aged 18 or above and come from the Office for National Statistics. Data source: Office for
National Statistics (2019).
Table A.2: Distribution of respondents across area codes - US
Region National Late March Early April
Area code 0 7.40 7.39 7.40
Area code 1 10.33 10.32 10.32
Area code 2 10.04 10.04 10.05
Area code 3 14.41 14.41 14.40
Area code 4 10.02 10.02 10.03
Area code 5 5.25 5.25 5.25
Area code 6 7.17 7.17 7.18
Area code 7 11.94 11.94 11.95
Area code 8 7.13 7.12 7.13
Area code 9 16.30 16.34 16.30
Observations 4003 4000
Notes: National figures refer to the latest available estimates for the popu-
lation of residents aged 18 or above and come from the United States Cen-
sus Bureau. Data source: U.S. Census Bureau, Population Division (2019).
28

Table A.3: Distribution of respondents across states - Germany
Region National Early April
Baden-Württemberg 13.33 13.29
Bayern 15.75 15.74
Berlin 4.39 4.40
Brandenburg 3.03 3.02
Bremen 0.82 0.82
Hamburg 2.22 2.22
Hessen 7.55 7.55
Mecklenburg-Vorpommern 1.94 1.97
Niedersachsen 9.62 9.62
Nordrhein-Westfalen 21.60 21.59
Rheinland-Pfalz 4.92 4.92
Saarland 1.19 1.20
Sachsen 4.91 4.90
Sachsen-Anhalt 2.66 2.65
Schleswig-Holstein 3.49 3.50
Thüringen 2.58 2.60
Observations 4002
Notes: National figures refer to the latest available estimates for the pop-
ulation of residents and come from the Statistische Ämter des Bundes und
der Länder. Data source: Statistische Ämter des Bundes und der Länder
(2018).
Table A.4: Demographic Variables in the Population & Surveys
US UK DE
CPS March April LFS March April SOEP April
Female 0.472 0.621 0.581 0.47 0.532 0.552 0.481 0.475
University 0.395 0.440 0.494 0.357 0.422 0.488 0.255 0.323
<30 0.231 0.322 0.255 0.232 0.295 0.281 0.168 0.398
30-39 0.224 0.262 0.264 0.230 0.272 0.333 0.205 0.284
40-49 0.203 0.179 0.215 0.217 0.203 0.238 0.209 0.146
50-59 0.198 0.130 0.136 0.217 0.151 0.114 0.251 0.132
60+ 0.144 0.107 0.130 0.104 0.079 0.033 0.166 0.040
Notes: The table shows the mean demographic characteristics of economically active individuals in each
respective country. These were calculated using the frequency weights provides in the CPS for the US,
LFS for the UK, and SOEP for Germany. The unweighted averages of these demographic variables in our
survey waves are also reported.
29

B Online Appendix B: Additional Tables and Fig-
ures
Figure B.1: Share of tasks that can be done from home by occupation
-50
0
50
100
150
Management
Business and Financial Operation
Computer and Mathematical
Architecture and Engineering
Life, Physical, and Social Scien
Community and Social Service
Legal
Educational Instruction and Libr
Arts, Design, Entertainment, Spo
Healthcare Practitioners and Tec
Healthcare Support
Protective Service
Food Preparation and Serving
Building and Grounds Cleaning an
Personal Care and Service
Sales and Related Occupations
Office and Administrative Suppor
Farming, Fishing, and Forestry
Construction and Extraction
Installation, Maintenance, and R
Production
Transportation and Material Movi
Military Specific Occupations
US
0
50
100
Management
Business and Financial Operation
Computer and Mathematical
Architecture and Engineering
Life, Physical, and Social Scien
Community and Social Service
Legal
Educational Instruction and Libr
Arts, Design, Entertainment, Spo
Healthcare Practitioners and Tec
Healthcare Support
Protective Service
Food Preparation and Serving
Building and Grounds Cleaning an
Personal Care and Service
Sales and Related Occupations
Office and Administrative Suppor
Farming, Fishing, and Forestry
Construction and Extraction
Installation, Maintenance, and R
Production
Transportation and Material Movi
Military Specific Occupations
UK
0
50
100
Management
Academic
Technician and comparable non-te
Office and administration
Service and retail
Farming, fishing, and forestry
Craftsmen and women
Mechanical
Auxiliary
Military
Germany
30

Figure B.2: Job loss probability due to Covid-19 by industry
0
.2
.4
.6
Public Administration and Defence
Water Supply etc.
Real Estate Activities
Electricity, Gas, Steam etc.
Information and Communication
Finacial and Insurance Activities
Agriculture Forestry and Fishing
Mining and Quarrying
Professional Activities
Administrative and Support Services
Human Health and Social Work
Other
Construction
Manufacturing
Other Service Activities
Education
Transportation and Storage
Wholesale and Retail Trade
Arts, Entertainment and Recreation
Accommodation and Food Service Activities
Activities of Households as Employers
US - Early April
0
.1
.2
.3
.4
.5
Mining and Quarrying
Electricity, Gas, Steam etc.
Water Supply etc.
Real Estate Activities
Administrative and Support Services
Information and Communication
Public Administration and Defence
Finacial and Insurance Activities
Human Health and Social Work
Transportation and Storage
Professional Activities
Agriculture Forestry and Fishing
Education
Manufacturing
Construction
Wholesale and Retail Trade
Other
Arts, Entertainment and Recreation
Other Service Activities
Accommodation and Food Service Activities
Activities of Households as Employers
UK - Early April
0
.05
.1
.15
.2
.25
Information and Communication
Finacial and Insurance Activities
Administrative and Support Services
Water Supply etc.
Real Estate Activities
Construction
Public Administration and Defence
Agriculture Forestry and Fishing
Human Health and Social Work
Manufacturing
Professional Activities
Mining and Quarrying
Transportation and Storage
Other Service Activities
Electricity, Gas, Steam etc.
Other
Wholesale and Retail Trade
Education
Activities of Households as Employers
Arts, Entertainment and Recreation
Accommodation and Food Service Activities
Germany - Early April
Notes: The thin black bars represent the 95% confidence intervals. The figure shows the share of
individuals who were in paid work four weeks before data collection that lost their job due to Covid-19
by occupation.
31

Figure B.3: Share of tasks that can be done from home versus job loss probability due
to Covid-19 by occupation
Slope -.44, R-squared .69
0
.1
.2
.3
.4
Share that lost job due to Coronavirus
0.2 .4 .6 .8
Tasks possible from home
US - Early April
Slope -.32, R-squared .54
0
.1
.2
.3
.4
Share that lost job due to Coronavirus
0.2 .4 .6 .8
Tasks possible from home
UK - Early April
Slope -.16, R-squared .58
0
.05
.1
.15
Share that lost job due to Coronavirus
0.2 .4 .6 .8
Tasks possible from home
Germany - Early April
Notes: Each bubble represents an occupation and the size is proportional to the number of observations
we have for that occupation.
32

Figure B.4: Share of tasks that can be done from home versus job loss probability due
to Covid-19 by industry
Slope -.39, R-squared .61
0
.1
.2
.3
.4
.5
Share that lost job due to Coronavirus
0.2 .4 .6 .8
Tasks possible from home
US - Early April
Slope -.27, R-squared .37
0
.1
.2
.3
Share that lost job due to Coronavirus
0.2 .4 .6 .8
Tasks possible from home
UK - Early April
Slope -.13, R-squared .22
0
.05
.1
.15
.2
Share that lost job due to Coronavirus
0.2 .4 .6 .8
Tasks possible from home
Germany - Early April
Notes: Each bubble represents an industry and the size is proportional to the number of observations
we have for that industry.
33

Figure B.5: Share of tasks that can be done from home by industry
-50
0
50
100
150
Agriculture Forestry and Fishing
Mining and Quarrying
Manufacturing
Electricity, Gas, Steam etc.
Water Supply etc.
Construction
Wholesale and Retail Trade
Transportation and Storage
Accommodation and Food Service A
Information and Communication
Finacial and Insurance Activitie
Real Estate Activities
Professional Activities
Administrative and Support Servi
Public Administration and Defenc
Education
Human Health and Social Work
Arts, Entertainment and Recreati
Other Service Activities
Activities of Households as Empl
Other
US
-50
0
50
100
150
Agriculture Forestry and Fishing
Mining and Quarrying
Manufacturing
Electricity, Gas, Steam etc.
Water Supply etc.
Construction
Wholesale and Retail Trade
Transportation and Storage
Accommodation and Food Service A
Information and Communication
Finacial and Insurance Activitie
Real Estate Activities
Professional Activities
Administrative and Support Servi
Public Administration and Defenc
Education
Human Health and Social Work
Arts, Entertainment and Recreati
Other Service Activities
Activities of Households as Empl
Other
UK
-50
0
50
100
150
Agriculture Forestry and Fishing
Mining and Quarrying
Manufacturing
Electricity, Gas, Steam etc.
Water Supply etc.
Construction
Wholesale and Retail Trade
Transportation and Storage
Accommodation and Food Service A
Information and Communication
Finacial and Insurance Activitie
Real Estate Activities
Professional Activities
Administrative and Support Servi
Public Administration and Defenc
Education
Human Health and Social Work
Arts, Entertainment and Recreati
Other Service Activities
Activities of Households as Empl
Other
Germany
34

Figure B.6: Share of tasks that can be done from home by occupation and industry
Notes: Joint data for US and UK from wave 2 of the surveys. Cells with less than 10 observations are
dropped.
35

Figure B.7: Jobs lost due to Coronavirus by occupation and industry
Notes: Joint data for US and UK from wave 2 of the surveys. Cells with less than 10 observations are
dropped.
36

Figure B.8: Change in hours worked by industry
-15
-10
-5
0
5
Education
Arts, Entertainment and Recreation
Transportation and Storage
Accommodation and Food Service Activities
Other Service Activities
Real Estate Activities
Other
Wholesale and Retail Trade
Administrative and Support Services
Manufacturing
Professional Activities
Construction
Information and Communication
Electricity, Gas, Steam etc.
Human Health and Social Work
Finacial and Insurance Activities
Agriculture Forestry and Fishing
Water Supply etc.
US - Early April
-15
-10
-5
0
5
Arts, Entertainment and Recreation
Accommodation and Food Service Activities
Education
Real Estate Activities
Other
Wholesale and Retail Trade
Transportation and Storage
Other Service Activities
Professional Activities
Construction
Manufacturing
Mining and Quarrying
Human Health and Social Work
Information and Communication
Administrative and Support Services
Finacial and Insurance Activities
Agriculture Forestry and Fishing
Public Administration and Defence
Electricity, Gas, Steam etc.
Water Supply etc.
UK - Early April
-15
-10
-5
0
5
Education
Accommodation and Food Service Activities
Arts, Entertainment and Recreation
Agriculture Forestry and Fishing
Wholesale and Retail Trade
Other Service Activities
Manufacturing
Professional Activities
Other
Administrative and Support Services
Human Health and Social Work
Transportation and Storage
Public Administration and Defence
Water Supply etc.
Finacial and Insurance Activities
Real Estate Activities
Construction
Mining and Quarrying
Electricity, Gas, Steam etc.
Information and Communication
Germany - Early April
Notes: The thin black bars represent the 95% confidence intervals. The figure shows the average
change in hours between a usual and the last week by industry.
37

Figure B.9: Change in hours worked (conditional on working) vs jobs lost due to
Coronavirus by occupation
0
.1
.2
.3
.4
Share that lost job due to Coronavirus
-15 -10 -5 0 5
Change in hours worked
US - Early April
-.1
0
.1
.2
.3
Share that lost job due to Coronavirus
-15 -10 -5 0 5
Change in hours worked
UK - Early April
0
.05
.1
.15
Share that lost job due to Coronavirus
-15 -10 -5 0 5
Change in hours worked
Germany - Early April
Notes: Each bubble represents an occupation and the size is proportional to the number of observations
we have for that occupation. The figure shows the average change in hours between a usual and the
last week by occupation on the x-axis and the share of workers that their jobs due to Coronavirus on
the y-axis.
38

Figure B.10: Change in hours worked (conditional on working) vs jobs lost due to
Coronavirus by industry
0
.1
.2
.3
.4
Share that lost job due to Coronavirus
-15 -10 -5 0 5
Change in hours worked
US - Early April
-.1
0
.1
.2
.3
Share that lost job due to Coronavirus
-15 -10 -5 0 5
Change in hours worked
UK - Early April
0
.05
.1
.15
Share that lost job due to Coronavirus
-15 -10 -5 0 5
Change in hours worked
Germany - Early April
Notes: Each bubble represents an industry and the size is proportional to the number of observations
we have for that industry. The figure shows the average change in hours between a usual and the last
week by industry on the x-axis and the share of workers that their jobs due to Coronavirus on the
y-axis.
39

Figure B.11: Employment status by occupation
0
20
40
60
80
100
Management
Business and Financial Operation
Computer and Mathematical
Architecture and Engineering
Life, Physical, and Social Scien
Community and Social Service
Legal
Educational Instruction and Libr
Arts, Design, Entertainment, Spo
Healthcare Practitioners and Tec
Healthcare Support
Protective Service
Food Preparation and Serving
Building and Grounds Cleaning an
Personal Care and Service
Sales and Related Occupations
Office and Administrative Suppor
Construction and Extraction
Installation, Maintenance, and R
Production
Transportation and Material Movi
US
0
20
40
60
80
100
Management
Business and Financial Operation
Computer and Mathematical
Architecture and Engineering
Life, Physical, and Social Scien
Community and Social Service
Legal
Educational Instruction and Libr
Arts, Design, Entertainment, Spo
Healthcare Practitioners and Tec
Healthcare Support
Protective Service
Food Preparation and Serving
Building and Grounds Cleaning an
Personal Care and Service
Sales and Related Occupations
Office and Administrative Suppor
Construction and Extraction
Installation, Maintenance, and R
Production
Transportation and Material Movi
UK
0
20
40
60
80
100
Management
Academic
Technician and comparable non-te
Office and administration
Service and retail
Farming, fishing, and forestry
Craftsmen and women
Mechanical
Auxiliary
Military
Germany
Lost job Furloughed Employed
Notes: The figure shows the share of individuals who are employed, furloughed or lost their job due
to the COVID-19 crisis, by occupation.
40

Figure B.12: Employment status by industry
0
20
40
60
80
100
Agriculture Forestry and Fishing
Manufacturing
Electricity, Gas, Steam etc.
Construction
Wholesale and Retail Trade
Transportation and Storage
Accommodation and Food Service A
Information and Communication
Finacial and Insurance Activitie
Real Estate Activities
Professional Activities
Administrative and Support Servi
Education
Human Health and Social Work
Arts, Entertainment and Recreati
Other Service Activities
Other
US
0
20
40
60
80
100
Agriculture Forestry and Fishing
Mining and Quarrying
Manufacturing
Electricity, Gas, Steam etc.
Water Supply etc.
Construction
Wholesale and Retail Trade
Transportation and Storage
Accommodation and Food Service A
Information and Communication
Finacial and Insurance Activitie
Professional Activities
Administrative and Support Servi
Public Administration and Defenc
Education
Human Health and Social Work
Arts, Entertainment and Recreati
Other Service Activities
Other
UK
0
20
40
60
80
100
Agriculture Forestry and Fishing
Mining and Quarrying
Manufacturing
Electricity, Gas, Steam etc.
Water Supply etc.
Construction
Wholesale and Retail Trade
Transportation and Storage
Accommodation and Food Service A
Information and Communication
Finacial and Insurance Activitie
Professional Activities
Administrative and Support Servi
Public Administration and Defenc
Education
Human Health and Social Work
Arts, Entertainment and Recreati
Other Service Activities
Other
Germany
Lost job Furloughed Employed
Notes: The figure shows the share of individuals who are employed, furloughed or lost their job due
to the COVID-19 crisis, by industry.
41

Figure B.13: Occupation fixed effect for job loss
-.2
-.1
0
.1
.2
Healthcare Support
Installation, Maintenance, and Repair
Architecture and Engineering
Educational Instruction and Library
Healthcare Practitioners and Technical occ.
Office and Administrative Support
Business and Financial Operations
Life, Physical, and Social Science
Computer and Mathematical
Construction and Extraction
Legal
Community and Social Service
Building and Grounds Cleaning and Maintenance
Management
Personal Care and Service
Sales and Related Occupations
Production
Arts, Design, Entertainment, Sports, and Media
Transportation and Material Moving
Food Preparation and Serving
US - Early April
-.2
-.1
0
.1
.2
Protective Service
Computer and Mathematical
Life, Physical, and Social Science
Educational Instruction and Library
Architecture and Engineering
Community and Social Service
Healthcare Support
Legal
Business and Financial Operations
Transportation and Material Moving
Production
Management
Office and Administrative Support
Healthcare Practitioners and Technical occ.
Arts, Design, Entertainment, Sports, and Media
Sales and Related Occupations
Installation, Maintenance, and Repair
Construction and Extraction
Personal Care and Service
Food Preparation and Serving
Building and Grounds Cleaning and Maintenance
UK - Early April
-.15
-.1
-.05
0
.05
.1
Military
Technician and comparable non-technical
Office and administration
Service and retail
Management
Craftsmen and women
Academic
Farming, fishing, and forestry
Auxiliary
Mechanical
Germany - Early April
Notes: The thin black bars represent the 95% confidence intervals. The bars represent coefficients for
occupation fixed effects from the regressions in Table 2 columns (2), (4), and (6) for the US and UK,
respectively. Management is the baseline occupation.
42

Figure B.14: Hours spent on a “typical” work day during the past week on active
childcare and home schooling
0
1
2
3
4
Hours
Childcare Home schooling
US
0
1
2
3
4
Hours
Childcare Home schooling
UK
0
1
2
3
4
Hours
Childcare Home schooling
Germany
Men Women
Notes: Data from wave 2 of the surveys. The thin black bars represent the 95% confidence intervals.
The figure shows average number of hours that men and women reported spending on childcare and
homeschooling. We restrict the sample to individuals with children who report working from home,
and whose answers to the time use questions combined do not exceed 24 hours.
43

Table B.1: Job and earnings loss probability (weighted)
Job loss Earnings loss
US UK DE US UK DE
Tasks from home -0.2522∗∗∗ -0.1996∗∗∗ -0.0619∗∗∗ -0.1404∗∗∗ -0.0756∗∗∗ -0.0322
(0.0218) (0.0196) (0.0127) (0.0299) (0.0264) (0.0224)
Self-Employed -0.0887∗∗∗ -0.0429∗0.0119 0.0271 0.0673∗0.0773∗∗
(0.0227) (0.0259) (0.0191) (0.0314) (0.0374) (0.0348)
Permanent -0.0616∗∗∗ -0.2011∗∗∗ -0.0965∗∗∗ -0.0006 -0.0527∗-0.0032
(0.0169) (0.0213) (0.0128) (0.0234) (0.0310) (0.0233)
Salaried -0.0732∗∗∗ 0.0290∗-0.0049 -0.1005∗∗∗ -0.0172 -0.1145∗∗∗
(0.0187) (0.0153) (0.0111) (0.0251) (0.0203) (0.0195)
Fixed Hours 0.0088 -0.0079 0.0024 -0.1049∗∗∗ -0.1473∗∗∗ -0.0756∗∗∗
(0.0168) (0.0152) (0.0097) (0.0232) (0.0200) (0.0169)
Constant 0.5098∗∗∗ 0.2916∗∗∗ 0.1546∗∗∗ 0.4225∗∗∗ 0.2951∗∗∗ 0.2918∗∗∗
(0.0865) (0.0651) (0.0414) (0.1200) (0.0857) (0.0725)
Observations 2995 3760 3354 2396 3111 3165
R20.1630 0.1244 0.0909 0.1229 0.1029 0.0926
Region F.E. yes yes yes yes yes yes
Occupation F.E. yes yes yes yes yes yes
Industry F.E. yes yes yes yes yes yes
Notes: OLS regressions. The dependent variable in Columns 1 - 3 is a binary variable for whether a respondent lost their
job within the past month and attributed the job loss to the coronavirus outbreak. The dependent variable in Columns 4
- 6 is a binary variable for whether a respondent earned less in March 2020 than the average earnings over January and
February 2020. In Columns 4 - 6 the sample is restricted to those who were in work at the time of data collection. Tasks
from home is the fraction of tasks respondents could do from home in their main or last job. Self-employed is a binary
variable for being self-employed in the main or last job. Permanent, salaried and fixed hours take value 1 for employees
with permanent contracts, who are salaried and whose work hours are fixed, respectively. Region fixed effects refer to
state fixed effects for the US and Germany, and fixed effects for regions as reported in Table A.1 for the UK.
44

Table B.2: Job loss probability - Individual characteristics (weighted)
United States United Kingdom Germany
(1) (2) (3) (4) (5) (6)
Female 0.0480∗∗∗ 0.0259∗0.0479∗∗∗ 0.0385∗∗∗ -0.0051 0.0034
(0.0150) (0.0156) (0.0124) (0.0130) (0.0077) (0.0084)
University degree -0.0898∗∗∗ -0.0135 -0.0611∗∗∗ -0.0065 -0.0232∗∗∗ -0.0132
(0.0153) (0.0164) (0.0130) (0.0137) (0.0088) (0.0104)
30-39 -0.0243 -0.0030 0.0302∗0.0371∗∗ -0.0405∗∗∗ -0.0099
(0.0223) (0.0215) (0.0180) (0.0178) (0.0129) (0.0133)
40-49 -0.0186 -0.0115 0.0283 0.0239 -0.0383∗∗∗ -0.0142
(0.0229) (0.0223) (0.0183) (0.0185) (0.0127) (0.0133)
50-59 0.0228 0.0244 0.0135 0.0062 -0.0334∗∗∗ -0.0129
(0.0232) (0.0230) (0.0184) (0.0188) (0.0123) (0.0129)
60+ 0.0267 0.0165 0.0161 0.0099 0.0340∗∗ 0.0349∗∗
(0.0251) (0.0248) (0.0239) (0.0239) (0.0137) (0.0143)
Tasks from home -0.2467∗∗∗ -0.1996∗∗∗ -0.0557∗∗∗
(0.0220) (0.0198) (0.0130)
Self-Employed -0.0912∗∗∗ -0.0443∗0.0083
(0.0229) (0.0263) (0.0194)
Permanent -0.0596∗∗∗ -0.2021∗∗∗ -0.0957∗∗∗
(0.0170) (0.0214) (0.0130)
Salaried -0.0700∗∗∗ 0.0277∗-0.0031
(0.0190) (0.0155) (0.0112)
Fixed Hours 0.0087 -0.0106 0.0032
(0.0169) (0.0152) (0.0097)
Constant 0.3303∗∗∗ 0.4963∗∗∗ 0.1274∗∗∗ 0.2601∗∗∗ 0.1017∗∗∗ 0.1623∗∗∗
(0.0669) (0.0879) (0.0253) (0.0661) (0.0155) (0.0418)
Observations 3025 2995 3816 3760 3584 3354
R20.0481 0.1648 0.0152 0.1277 0.0289 0.0963
Region F.E. yes yes yes yes yes yes
Occupation F.E. no yes no yes no yes
Industry F.E. no yes no yes no yes
Notes: OLS regressions. The dependent variable is a binary variable for whether a respondent lost their job within the
past month and attributed the job loss to the coronavirus outbreak. Work from Home is the fraction of tasks respondents
could do from home in their main or last job. Self-employed is a binary variable for being self-employed in the main or
last job. Permanent, salaried and fixed hours take value 1 for employees with permanent contracts, who are salaried and
whose work hours are fixed, respectively. Region fixed effects refer to state fixed effects for the US and Germany, and fixed
effects for regions as reported in Table A.1 for the UK.
45

Table B.3: Earnings loss probability - In-work respondents
United States United Kingdom Germany
(1) (2) (3) (4) (5) (6)
Female 0.0126 0.0143 0.0082 0.0273 0.0104 0.0130
(0.0202) (0.0217) (0.0166) (0.0174) (0.0145) (0.0151)
University degree -0.1501∗∗∗ -0.0758∗∗∗ -0.0206 0.0287 -0.0022 0.0325∗
(0.0209) (0.0226) (0.0169) (0.0176) (0.0165) (0.0177)
30-39 -0.0129 -0.0044 -0.0777∗∗∗ -0.0447∗∗ -0.0567∗∗∗ -0.0288
(0.0271) (0.0272) (0.0209) (0.0211) (0.0182) (0.0185)
40-49 -0.0484∗-0.0676∗∗ -0.0686∗∗∗ -0.0219 -0.0302 0.0019
(0.0286) (0.0291) (0.0229) (0.0235) (0.0218) (0.0223)
50-59 -0.0973∗∗∗ -0.1084∗∗∗ -0.0994∗∗∗ -0.0612∗∗ -0.0465∗∗ -0.0121
(0.0335) (0.0339) (0.0285) (0.0290) (0.0222) (0.0228)
60+ -0.1044∗∗∗ -0.1290∗∗∗ -0.1045∗∗ -0.0861∗-0.1176∗∗∗ -0.1072∗∗∗
(0.0349) (0.0356) (0.0491) (0.0485) (0.0382) (0.0382)
Tasks from home -0.1224∗∗∗ -0.1258∗∗∗ -0.0990∗∗∗ -0.0785∗∗∗ -0.0280 -0.0281
(0.0274) (0.0304) (0.0236) (0.0269) (0.0213) (0.0239)
Self-Employed 0.0293 0.1045∗∗∗ 0.0678∗∗
(0.0319) (0.0377) (0.0326)
Permanent -0.0230 -0.0147 0.0078
(0.0234) (0.0303) (0.0214)
Salaried -0.0683∗∗∗ -0.0472∗∗ -0.0641∗∗∗
(0.0252) (0.0210) (0.0198)
Fixed Hours -0.0699∗∗∗ -0.1087∗∗∗ -0.0901∗∗∗
(0.0231) (0.0204) (0.0176)
Constant 0.4013∗∗∗ 0.4164∗∗∗ 0.3640∗∗∗ 0.3751∗∗∗ 0.1789∗∗∗ 0.2812∗∗∗
(0.0939) (0.1225) (0.0347) (0.0901) (0.0272) (0.0650)
Observations 2405 2396 3123 3111 3201 3165
R20.0661 0.1207 0.0214 0.0932 0.0139 0.0712
Region F.E. yes yes yes yes yes yes
Occupation F.E. no yes no yes no yes
Industry F.E. no yes no yes no yes
Notes: OLS regressions. Sample is restricted to those who were in work at the time of the survey. The dependent variable
is a binary variable for whether a respondent earned less in March 2020 than the average earnings over January and Febru-
ary 2020. Tasks from home is the fraction of tasks respondents could do from home in their main or last job. Self-employed
is a binary variable for being self-employed in the main or last job. Permanent, salaried and fixed hours take value 1 for
employees with permanent contracts, who are salaried and whose work hours are fixed, respectively. Region fixed effects
refer to state fixed effects for the US and Germany, and fixed effects for regions as reported in Table A.1 for the UK.
46

Table B.4: Earnings loss probability - In-work respondents (weighted)
United States United Kingdom Germany
(1) (2) (3) (4) (5) (6)
Female 0.0218 0.0235 0.0058 0.0250 0.0042 0.0110
(0.0197) (0.0212) (0.0164) (0.0173) (0.0140) (0.0147)
University degree -0.1429∗∗∗ -0.0720∗∗∗ -0.0127 0.0367∗∗ -0.0099 0.0264
(0.0205) (0.0221) (0.0174) (0.0181) (0.0169) (0.0183)
30-39 -0.0292 -0.0175 -0.0714∗∗∗ -0.0407∗-0.0441∗-0.0095
(0.0292) (0.0291) (0.0238) (0.0236) (0.0234) (0.0233)
40-49 -0.0494 -0.0673∗∗ -0.0603∗∗ -0.0192 -0.0303 0.0133
(0.0301) (0.0304) (0.0242) (0.0247) (0.0231) (0.0233)
50-59 -0.1196∗∗∗ -0.1278∗∗∗ -0.0900∗∗∗ -0.0573∗∗ -0.0406∗0.0022
(0.0310) (0.0315) (0.0243) (0.0251) (0.0222) (0.0227)
60+ -0.1212∗∗∗ -0.1457∗∗∗ -0.0994∗∗∗ -0.0831∗∗∗ -0.1081∗∗∗ -0.0925∗∗∗
(0.0336) (0.0342) (0.0316) (0.0319) (0.0251) (0.0256)
Tasks from home -0.1282∗∗∗ -0.1417∗∗∗ -0.0909∗∗∗ -0.0833∗∗∗ -0.0169 -0.0427∗
(0.0268) (0.0299) (0.0230) (0.0266) (0.0201) (0.0229)
Self-Employed 0.0386 0.0805∗∗ 0.0920∗∗∗
(0.0313) (0.0379) (0.0352)
Permanent -0.0144 -0.0426 0.0045
(0.0235) (0.0312) (0.0237)
Salaried -0.0772∗∗∗ -0.0216 -0.1176∗∗∗
(0.0254) (0.0205) (0.0196)
Fixed Hours -0.1013∗∗∗ -0.1451∗∗∗ -0.0758∗∗∗
(0.0231) (0.0201) (0.0169)
Constant 0.4498∗∗∗ 0.4672∗∗∗ 0.3476∗∗∗ 0.2984∗∗∗ 0.1654∗∗∗ 0.2722∗∗∗
(0.0949) (0.1217) (0.0342) (0.0869) (0.0297) (0.0731)
Observations 2405 2396 3123 3111 3201 3165
R20.0743 0.1400 0.0197 0.1080 0.0191 0.1005
Region F.E. yes yes yes yes yes yes
Occupation F.E. no yes no yes no yes
Industry F.E. no yes no yes no yes
Notes: OLS regressions. Sample is restricted to those who were in work at the time of the survey. The dependent variable
is a binary variable for whether a respondent earned less in March 2020 than the average earnings over January and Febru-
ary 2020. Work from Home is the fraction of tasks respondents could do from home in their main or last job. Self-employed
is a binary variable for being self-employed in the main or last job. Permanent, salaried and fixed hours take value 1 for
employees with permanent contracts, who are salaried and whose work hours are fixed, respectively. Region fixed effects
refer to state fixed effects for the US and Germany, and fixed effects for regions as reported in Table A.1 for the UK.
47

Table B.5: Job loss probability - Waves 1 and 2
United States United Kingdom
(1) (2) (3) (4) (5) (6)
Work from Home -0.2685∗∗∗ -0.2482∗∗∗ -0.1372∗∗∗ -0.1858∗∗∗ -0.1506∗∗∗ -0.1091∗∗∗
(0.0117) (0.0127) (0.0111) (0.0112) (0.0124) (0.0108)
Wave 2 0.0905∗∗∗ 0.0936∗∗∗ 0.1977∗∗∗ 0.0882∗∗∗ 0.0896∗∗∗ 0.1743∗∗∗
(0.0088) (0.0087) (0.0072) (0.0080) (0.0079) (0.0068)
Self-Employed -0.0513∗∗∗ -0.0267∗
(0.0117) (0.0149)
Permanent -0.0327∗∗∗ -0.1051∗∗∗
(0.0086) (0.0122)
Salaried -0.0317∗∗∗ 0.0103
(0.0094) (0.0087)
Fixed Hours 0.0035 -0.0007
(0.0085) (0.0086)
Constant 0.2557∗∗∗ 0.2420∗∗∗ 0.1018∗∗∗ 0.1363∗∗∗ 0.1028∗∗∗ 0.0932∗∗∗
(0.0401) (0.0421) (0.0361) (0.0148) (0.0195) (0.0203)
Observations 6289 6282 5901 7024 7010 6709
R20.1007 0.1226 0.1811 0.0553 0.0783 0.1411
Region F.E. yes yes yes yes yes yes
Occupation F.E no yes yes no yes yes
Notes: OLS regressions. The dependent variable is a binary variable for whether a respondent lost their job within the
past month and attributed the job loss to the coronavirus outbreak, and zero if they did not. Work from Home is the
fraction of tasks respondents could do from home in their main or last job. Self-employed is a binary variable for being
self-employed in the main or last job. Permanent, salaried and fixed hours take value 1 for employees with permanent
contracts, who are salaried and whose work hours are fixed, respectively. Region fixed effects refer to state fixed effects
for the US, and fixed effects for regions as reported in Table A.1 for the UK.
48

Table B.6: Hours spent on a “typical” work day during the past week on active childcare
or home schooling
United States United Kingdom Germany
(1) (2) (3) (4) (5) (6)
Female 1.3178∗∗∗ 1.0663∗∗ 0.9021∗1.3876∗∗∗ 1.2538∗∗∗ 1.2373∗∗∗
(0.4173) (0.4758) (0.4818) (0.2039) (0.2238) (0.2236)
University degree -0.0189 0.1077 0.1043 0.1963 0.1961 0.2005
(0.4423) (0.4910) (0.4902) (0.2148) (0.2302) (0.2301)
Number of kids 0.1462 0.0790 0.0786 0.5518∗∗∗ 0.6184∗∗∗ 0.6249∗∗∗
(0.2120) (0.2359) (0.2356) (0.1251) (0.1288) (0.1286)
Married 0.3577 0.4534 0.4647 0.2971 0.3673 0.3758
(0.5084) (0.5525) (0.5524) (0.2533) (0.2602) (0.2603)
30-39 -0.5580 -0.4830 -0.4904 0.8583∗∗∗ 0.6391∗∗ 0.6397∗∗
(0.5170) (0.5743) (0.5759) (0.2568) (0.2699) (0.2702)
40-49 0.2264 -0.0719 -0.0982 0.2239 -0.0413 -0.0413
(0.5492) (0.6219) (0.6290) (0.2872) (0.3043) (0.3069)
50-59 -1.7315∗-1.6476∗-1.8368∗-1.8240∗∗∗ -2.2041∗∗∗ -2.1552∗∗∗
(0.8833) (0.9919) (1.0013) (0.4224) (0.4440) (0.4457)
60+ -1.6086 -1.6823 -1.7829 -2.8146∗∗∗ -2.9806∗∗∗ -3.0226∗∗∗
(1.0472) (1.1566) (1.1550) (0.9283) (0.9515) (0.9509)
Tasks from home -0.7789 -0.8137 -1.0187∗∗∗ -1.0978∗∗∗
(0.7520) (0.7647) (0.3928) (0.4018)
Hours worked outside home -0.0631 -0.1137∗∗
(0.0814) (0.0472)
Hours worked from home 0.1067 -0.0520
(0.0678) (0.0367)
Constant 1.5196 1.1854 1.2252 3.5933∗∗∗ 2.7605∗∗ 3.0701∗∗∗
(1.8156) (2.3639) (2.3616) (0.4639) (1.1043) (1.1092)
Observations 433 429 429 1310 1273 1273
R20.1665 0.2726 0.2810 0.1094 0.1530 0.1575
Region F.E. yes yes yes yes yes yes
Occupation F.E. no yes yes no yes yes
Industry F.E. no yes yes no yes yes
Notes: OLS regressions. The dependent variable is the number of hours spent on child care or home schooling on a typical day
during the last week. Work from home is the fraction of tasks respondents could do from home in their main or last job. Region
fixed effects refer to state fixed effects for the US and Germany, and fixed effects for regions as reported in Table A.1 for the UK.
49