Annette Alstadsæter et al., The First Weeks of the Coronavirus Crisis: Who Got Hit, When and Why? Evidence from Norway (2020)
Summary
NBER Working Paper No. 27131, The First Weeks of the Coronavirus Crisis: Who Got Hit, When and Why? Evidence from Norway, by Annette Alstadsæter, Bernt Bratsberg, Gaute Eielsen, Wojciech Kopczuk, Simen Markussen, Oddbjorn Raaum and Knut Røed, dated May 2020. Using real-time register data, it documents the size, dynamics and socio-economic characteristics of temporary and permanent layoffs in Norway after social distancing measures were announced on March 12th, when 360,000 people, approximately 12% of the labor force, signed up for unemployment benefits within weeks. It finds that close to 90% of layoffs were temporary and that the shock hit a financially vulnerable population and parents with younger children hardest. It concludes that the rise in unemployment likely overstates the associated loss of output by about a third.
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
Cite this document
“Annette Alstadsæter et al., The First Weeks of the Coronavirus Crisis: Who Got Hit, When and Why? Evidence from Norway (2020).”
Original at www.nber.org. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/annette-alstads-ter-et-al-the-first-weeks-of-the-coronavirus-crisis-wh-9fa1e5d8eb/
, SHA-256 978b881c31432131f642425f15ed8024acdcaf07adcf690194f4a89d3870e6df.Full text
NBER WORKING PAPER SERIES
THE FIRST WEEKS OF THE CORONAVIRUS CRISIS:
WHO GOT HIT, WHEN AND WHY? EVIDENCE FROM NORWAY
Annette Alstadsæter
Bernt Bratsberg
Gaute Eielsen
Wojciech Kopczuk
Simen Markussen
Oddbjorn Raaum
Knut Røed
Working Paper 27131
http://www.nber.org/papers/w27131
NATIONAL BUREAU OF ECONOMIC RESEARCH
1050 Massachusetts Avenue
Cambridge, MA 02138
May 2020
This article is part of the report from the research project "Temporary and permanent layoffs
under Covid-19". Thanks to the Norwegian Directorate of Labour and Welfare and Statistics
Norway for supplying the registry data, to Trond Vigtel for assistance with the data, and to Maria
Hoen for assistance with the linking of data from the O*NET database to Norwegian occupation
codes. The views expressed herein are those of the authors and do not necessarily reflect the
views of the National Bureau of Economic Research.
NBER working papers are circulated for discussion and comment purposes. They have not been
peer-reviewed or been subject to the review by the NBER Board of Directors that accompanies
official NBER publications.
© 2020 by Annette Alstadsæter, Bernt Bratsberg, Gaute Eielsen, Wojciech Kopczuk, Simen
Markussen, Oddbjorn Raaum, and Knut Røed. All rights reserved. Short sections of text, not to
exceed two paragraphs, may be quoted without explicit permission provided that full credit,
including © notice, is given to the source.
The First Weeks of the Coronavirus Crisis: Who Got Hit, When and Why? Evidence from
Norway
Annette Alstadsæter, Bernt Bratsberg, Gaute Eielsen, Wojciech Kopczuk, Simen Markussen,
Oddbjorn Raaum, and Knut Røed
NBER Working Paper No. 27131
May 2020
JEL No. E24,J11,J4,J6
ABSTRACT
Using real-time register data we document the magnitude, dynamics and socio-economic
characteristics of the crisis-induced temporary and permanent layoffs in Norway. We find
evidence that the effects of social distancing measures quickly spread to industries that were not
directly affected by policy. Close to 90% of layoffs are temporary, although this classification
may change as the crisis progresses. Still, there is suggestive evidence of immediate stress on a
subset of firms that manifests itself in permanent rather than temporary layoffs. We find that the
shock had a strong socio-economic gradient, hit a financially vulnerable population, and parents
with younger children, and was driven by layoffs in smaller, less productive, and financially
weaker firms. Consequently though, the rise in unemployment likely overstates the loss of output
associated with the layoffs by about a third.
Annette Alstadsæter Simen Markussen
School of Economics and Business Ragnar Frisch Centre for Economic Research
Norwegian University of Life Sciences Gaustadalléen 21, 0349 Oslo
P.O. Box 5003, NO-1432 Ås Norway
Norway simen.markussen@frisch.uio.no
annette.alstadsater@nmbu.no
Oddbjorn Raaum
Bernt Bratsberg Ragnar Frisch Centre for Economic Research
Ragnar Frisch Centre for Economic Research Gaustadalleen 21, 0349 Oslo
Gaustadalléen 21, 0349 Oslo Norway
Norway oddbjorn.raaum@frisch.uio.no
bernt.bratsberg@frisch.uio.no
Knut Røed
Gaute Eielsen Ragnar Frisch Centre for Economic Research
Department of Research and Analysis Gaustadalléen 21, 0349 Oslo
Norwegian Directorate of Labor and Welfare Norway
Norway knut.roed@frisch.uio.no
gaute.eielsen@nav.no
Wojciech Kopczuk
Columbia University
420 West 118th Street,
Rm. 1022 IAB MC 3323
New York, NY 10027
and NBER
wk2110@columbia.edu
The coronavirus crisis hit fast all around the world. Its unprecedented nature and the response
to it through social distancing led to massive reduction in economic activity. In Norway, where our
data comes from, the measures were announced on March 12th and during the following few weeks
360,000 people (approximately 12% of the labor force) signed up for unemployment benefits, as
Figure 1 demonstrates.
Figure 1: Unemployment benefits applications in Norway 3/1-4/19/2020
A. New claims per day (1,000) B. Total new claims since March 1st (1,000)
60 350
50 300
250
40
200
30
150
20
100
10
50
0 0
3/2 2/9 3/16 3/23 3/30 4/6 4/13 4/20 3/2 2/9 3/16 3/23 3/30 4/6 4/13 4/20
Temporary layoffs Permanent layoffs
Note: Temporary and permanent layoffs refer to persons who have applied for unemployment benefits under the temporary or
regular schemes of the UI program. Regular claims may include, in addition to dismissals, persons who have completed a fixed
term employment contract. For each applicant we classify the person as temporarily or permanently laid off based on the latest
filed application.
In the Norwegian context, workers filing for unemployment benefits may be laid off temporarily
or permanently. Approximately 90 percent of the layoffs during the period covered here are tempo-
rary. That means that most of those applying for benefits retain employment relationships. There
is, of course, a risk that temporary layoffs may turn into permanent layoffs later on. Berg et al.
(2015), for example, shows that during the recession of 1993-95, about 15% of temporary layoffs
did not return to the previous job. Previous studies also show that those exposed to unemployment
during a temporary economic crisis have higher risk to be out of work in the longer term; see e.g.
Yagan (2019).
On March 16, the Norwegian parliament agreed to change the rules for layoffs with immediate
effect. Laid off workers will be paid full wages (up to an income limit of approximately NOK
600,000 — about USD 56,000) for the first 20 days. Afterwards, the benefits will be equal to 80%
of income under NOK 300,000 and 62.4% for income between NOK 300,000 and 600,000. There is
no compensation for lost income above NOK 600,000. Additionally, employers’ responsibility for
the first period of full pay was reduced from 15 to 2 days.
These changes in layoff regulations have been followed by additional crisis responses aimed at
preventing bankruptcies, maintaining activity in municipalities and increased support for temporary
training of laid off and unemployed.
We use almost real-time unemployment claims information linked to administrative registry data
on past characteristics of individuals and firms to characterize which individuals and businesses were
1
most affected, how the effect unfolded over the first few weeks, and the role that policy played in
the process.
In this article, we take a closer look at who was most affected by the crisis during the first
few weeks, both by loss of work through temporary or permanent layoff, and by being exposed
to infection risk through work with socially critical functions. The article delves into a research
literature that has studied heterogeneous effects of economic crises; see, for example, Hoynes et al.
(2012), who find that the financial crisis in the United States has particularly affected minority
groups, youth, and those with low education. Studies from Germany, Britain and Norway have
also shown that immigrants are particularly vulnerable to economic fluctuations (Dustmann et al,
2010; Bratsberg et al, 2010; 2018). However, the crisis we are facing now is unique, both in terms
of how quickly it has occurred, how many are affected, and not least in its underlying origins. It
is thus far from obvious that studies of previous economic fluctuations provide a good basis for
assessing the distributional effects of the crisis we are currently experiencing.
This paper also contributes to emerging literature on the impact of the crisis on the labor
market and it does so with real-time administrative data rather than relying on survey-based or
partial industry-based information (see for example Adams-Prassl et al, 2020, Bartik et al 2020,
and Coibion, Gorodnichenko and Weber, 2020).
We have six main findings. First, layoffs started in sectors of the economy directly affected
by the policy measures but then quickly spilled over to the rest of the economy so that after 4
weeks 2/3 of layoffs are accounted for by businesses that were not directly targeted. Second, close
to 90% of layoffs are temporary rather than permanent and while this classification may change
as the crisis progresses, that is one glimmer of hope in the data. Third, while permanent layoffs
are a minority, they still correspond to a 1.5 percentage point increase in unemployment — an
unprecedented monthly change. Fourth, the layoffs have a strong socio-economic gradient and
hit financially vulnerable populations. Fifth, there are hints of the important role of childcare
— within firms, layoffs appear to be skewed toward workers with younger children, in particular
toward women. Finally, layoffs are more common in less productive and financially weaker firms so
that the employment loss may be overstating total output loss (although, the potential unobserved
offsetting consideration is the possibility of underemployment of those that remained on the job).
1 The data
The primary database for this article is all individual unemployment benefit applications in Norway
during the period from March 1st to April 19th 2020, with information on whether the applications
concerned ordinary unemployment or temporary layoffs. This means that in practice the analysis
will include wage earners, entitled to unemployment benefits. As the annual income limit for en-
titlement to unemployment benefit was lowered from approximately NOK 125,000 to NOK 75,000
in connection with the current crisis, some applicants are very low income. We do not have infor-
mation on financial problems among self-employed persons and persons whose main income comes
from contract work.
2
The population includes all persons residing in Norway as of January 1st 2019, and provides in-
formation about the employer (including industry), occupation, salary, previous income, education,
age, sex, and country of birth, with links to spouse, children, and parents.
The employer has been identified on the basis of the so-called a-form, filed by all employers to the
tax authority, in December 2019. The a-form includes information on occupation, current wages,
and hours worked. Information on education is updated until October 2018, while information
on previous income from all sources covers the period 1967-2017. The latter means that we can
calculate total professional income so far in life quite precisely for the entire population.
As part of characterizing employees’ occupations, we use information from O*NET (see www.
onetonline.org), which, based on US data, provides a detailed description of the characteristics
of various occupations. This information is then linked to the Norwegian standard for occupational
classification (Hoen, 2016). In this article, one particular characteristic is of interest, namely the
degree to which the job requires physical proximity to other people. Professions are described on
a scale from 0 (do not work in the vicinity of others) to 100 (profession involves direct physical
contact with others). This variable is standardized (in the US data), so that the average is equal
to 0 and standard deviation is equal to 1.
Since our information on occupation and hourly wage is based on information somewhat back
in time, some of the employed will be “misplaced.” This will happen for people who have switched
jobs after December 2019. There are also some unemployment benefit applicants who cannot be
matched to other data. This may be because they were not resident on January 1, 2019, or did not
have a “normal” contract with an employer during 2019. Table 1 gives a more detailed description
of the data. The analysis sample, described in column 1, consists of all wage recipients registered in
2019. Our data links enable us to capture a total of 330,492 of the 362,539 unemployment insurance
applicants (91.2%) in the period in question; see columns (2) and (3). Among the approx. 32,000
applicants that we do not match, one half are not residents in the country as of January 1, 2019
(column 5). The rest fall out of our analysis because we cannot connect them to an employer in
2019 (column 4).
2 The Dynamics of the Crisis
Even though the crisis hit widely, it did not hit at random. In this section we look more closely at
which employees had to bear the negative consequences in the first few weeks. We will first and
foremost focus on who lost the job, through temporary or permanent layoffs. But we will also look
at who might bear the brunt by continuing to work in jobs with the risk of incurring infection.
This latter group consists of people in “socially critical” professions with a great deal of physical
proximity to other people, such as health personnel, nursing and care staff, and staff in grocery
∗
stores.
Figure 2 shows the composition of layoffs by different types of policies. The initial measures
requiring social distancing were announced on Thursday, March 12, and the layoffs responded
immediately, with a big spike in claims immediately after the weekend on Monday March 16 when
3
Table 1: Summary statistics
UI benefit applicants 3/1 - 4/19/2020
Analysis Analysis sample Not included in
sample analysis sample
Temporary Permanent Resident Non
layoff layoff resident
(1) (2) (3) (4) (5)
Age 40.7 38.6 35.6 39.8 37.5
Female(%) 48.5 45.7 46.6 44.9 22.8
Norwegian-born(%) 82.7 75.7 65.6 74.8 7.6
Western Europe(%) 3.6 4.5 4.2 4.4 28.4
New EU country(%) 5.0 7.8 12.3 6.2 59.4
Other countries(%) 8.7 12.0 17.9 14.5 4.6
Education (years) 13.8 12.9 12.7 12.7 12.3
Income rank 56.1 52.6 41.4
Hourly wage(NOK) 305.1 287.6 261.7
Physical proximity index 0.307 0.275 0.474
Observations 2,672,044 295,848 34,644 16,005 16,042
more generous rules (from both employee and employer’s point of view) were announced.
We decompose this increase into four different categories. Group 1 includes individuals working
in industries that were subject to the direct ban of activity (e.g., hairdressers, tattoo salons, bars).
Group 2 includes workers of businesses that were subject to an implicit ban, typically from dis-
tancing restrictions (dentists, restaurants, etc), while Group 3 are workers of businesses that were
subject to restrictions that prevented consumers from using those services, such as for hotels and
airlines. Just 1% of workers were in Group 1, 5.7% in Group 2, 4.4% in Group 3, and 88.8% of
workers were not subject to any direct regulation.
4
Figure 2: Layoffs by the type of restrictions
A. Cumulative layoffs since March 1st B. Percent of group
80
300
60
Number of layoffs (1000s)
200
Percent laid off
40
100
20
0 0
3/2 3/9 3/16 3/23 3/30 4/6 4/13 4/20 3/2 3/9 3/16 3/23 3/30 4/6 4/13 4/20
regtype 1 regtype 2 regtype 3 no regulation regtype 1 regtype 2 regtype 3 no regulation overall
As a share of the group, layoffs in Group 1 were the largest, with close to 70% of the group
laid off by the end of the period; the majority of them in the first few days. Groups 2 and 3
behaved similarly — a sizable immediate increase, a slower increase afterwards, reaching over 40%
by the end of the period. These responses are not particularly surprising — they correspond to
explicit or implicit restrictions on the activity. All three of these groups add up to a bit over 10%
of employment but account for close to 1/3 of layoffs.
The remaining 2/3 of layoffs are accounted for by industries that did not face direct restrictions
on their activity. Layoffs in that group were more spread out over time — they lagged layoffs in
directly affected industries, but started building up a bit more gradually (although still with more
than half of the layoffs over the whole period happening during the first week).
Figure 3 shows decomposition of layoffs into temporary and permanent ones. They do not
follow the same time pattern — permanent layoffs build up more slowly over time. Of course,
the size of the two groups is dramatically different. Permanent layoffs can be taken as a lower
bound of employment ties that were severed permanently. In particular, the Norwegian system
does not discourage temporary layoffs, so it is unlikely that permanent layoffs would revert to an
employment relationship with the same employer at a later date. Viewed in that way, 1.5% of
employment relationships in the whole economy ended permanently in the course of four weeks
after March 12 — this would be a very large increase in unemployment at any time. In particular,
a 1.5 percentage point increase is larger than any single month change in unemployment during the
financial crisis.
5
Figure 3: Temporary and permanent layoffs by the type of restrictions
A. Temporary layoffs B. Permanent layoffs
80 4
60 3
Percent laid off
40 2
20 1
0 0
3/2 3/9 3/16 3/23 3/30 4/6 4/13 4/20 3/2 3/9 3/16 3/23 3/30 4/6 4/13 4/20
regtype 1 regtype 2 regtype 3 no regulation overall
Speculating somewhat, permanent layoffs are likely to be revealing businesses in severe distress
that are unlikely to survive the shock. One group of workers that may be permanently laid off
are those on fixed term contracts that expire, but it is unlikely that they constitute the majority
of the short-term effect. Otherwise, a permanent layoff in Norway requires the employer to pay
severance, so that firms are expected to prefer temporary layoffs unless they are going bankrupt
and this should be especially true at the time of a liquidity shock.
6
Figure 4: Layoffs by firm size and the type of restrictions
A. All layoffs, small firms B. Temporary layoffs, small firms C. Permanent layoffs, small firms
80 80 4
60 60 3
40 40 2
20 20 1
0 0 0
Percent
D. All layoffs, large firms E. Temporary layoffs, large firms F. Permanent layoffs, large firms
80 80 4
60 60 3
40 40 2
20 20 1
0 0 0
3/2 3/9 3/16 3/23 3/30 4/6 4/13 4/20 3/2 3/9 3/16 3/23 3/30 4/6 4/13 4/20 3/2 3/9 3/16 3/23 3/30 4/6 4/13 4/20
regtype 1 regtype 2 regtype 3 no regulation overall
Figure 4 shows that the behavior of small and large firms (below/above the median of 63
employees) in industries affected directly by restrictions is quite similar, with about the same level
of layoffs and their composition into temporary and permanent ones. The main difference between
larger and smaller firms comes from the “no regulation” group in which small firms reacted much
more strongly.
3 Production loss
How large is the magnitude of the shock in terms of lost output? While direct information on
this is not yet available, we can provide a guess based on employment loss and measures of worker
productivity. Worker productivity is measured as value added (profits plus total wage bill) per
man-year. Figure 5 shows the production loss based on the assumption that all workers are equally
productive, such that the percentage loss in production corresponds to that of employment (yellow
line). This is likely an overestimate of the production loss because industries that are affected
are likely less productive. Reweighting employment loss by the average industry-level productivity
implies output loss of about 15% by the end of the period. Using a firm-specific measure of produc-
tivity allows for accounting for selection within industries. As expected (because less productive
firms are more likely to lay off employees), it attenuates the effect somewhat further.
7
Figure 5: Production loss based on employment loss
0
-5
Reduction, %
-10
-15
-20
March 2 March 9 March 16 March 23 March 30 April 6 April 13 April 20
Production loss, private sector
Production loss, private sector, from industry means
Number of employees, private sector
Is change in labor employed a good proxy for production loss? Of course, this is not a prediction
of the loss of output in general. It does not account for the fact that even workers who remain
employed may be less productive than they would have been otherwise. Going in the other direction,
in Norway workers may be laid off for a fraction of their time and some temporarily laid off workers
may have returned to work by April 19. At this point, our data does not allow for observing that.
4 The “social gradient” of the crisis
Figure 6 shows the incidence of socially critical risky work among all wage earners in Norway,
based on their position in the age- and gender-specific distribution of total employment income
throughout the professional career. Socially critical risky work is then defined as having a job
of critical importance, in combination with a profession that requires physical closeness beyond
the average in the whole economy. We divide the wage earners into ten equal groups by gender,
according to their position in the age- and gender-specific income distribution.
The likelihood of holding an essential job with a high degree of physical proximity to others falls
markedly with income levels, especially for men. There is also a remarkably large gender difference:
Women are far more than men exposed to the combination of critical tasks and physical closeness
8
to others. As many as 24.5 percent of the female workers have such jobs, which is about twice the
corresponding proportion for men.
Figure 6: Share of employees in socially critical risk work, according to rank in age- and gender-
specific distribution of lifetime income
30
25
20
Percent
15
10
5
0 20 40 60 80 100
Percentile
Men Women
Note: Critical function with a high degree of physical proximity is based on the coding of the ministries’ list of personnel groups
/ positions covered by the exception to the main rule that children should be kept home from childcare and school to prevent
the spread of infection, interacted with an indicator that the profession involves more physical closeness with others than the
average. We have removed from this category 14,995 people classified as having an essential job, who were actually laid off or
terminated in our data window. 63 percent of them come from one of the three professions salesperson in grocery / pharmacy
/ gas station, “other care worker,” or bus and tram driver. The proportion with essential work with close physical proximity
to others is 8.7 percent for men and 24.5 percent for women.
Figure 7 shows the extent of permanent and temporary layoffs for men and women by their
position in the age and gender specific income distribution. We see that the likelihood of being
hit by layoffs or termination falls sharply with income rankings. This social gradient is even more
marked for permanent than for temporary layoffs. The probability of having permanently lost work
during the first weeks of the coronavirus crisis is more than six times higher for people in the lower
income deciles than at the top. The likelihood of being laid off is a little bit higher for men than
for women.
9
Figure 7: Percentage of temporarily and permanently laid-off workers according to rank in age-
and gender-specific distribution of lifetime income
A. Temporary layoff B. Permanent layoff
16
3
14 2.5
2
Percent
12
1.5
10
1
8 .5
0 20 40 60 80 100 0 20 40 60 80 100
Percentile Percentile
Men Women
Note: The percentage of temporary layoffs in our data is 11.7 percent for men and 10.4 percent for women. Corresponding
shares of permanent layoff are 1.3 percent for men and 1.2 percent for women. The income rank is based on lifetime income. It
is calculated by summing up all income earned in the period 1967-2017 for all residents in Norway. Subsequently, all employees
are divided into ten equal groups by ranking in the income distribution. This is done within each birth cohort and separately
for men and women.
We will now take a closer look at who, in the period from March 1st to April 19th , lost the
job on a temporary or permanent basis, and who applied for unemployment benefit. We will not
separate between temporary and permanent unemployment benefits in what follows unless stated
otherwise. The distinction between these two outcomes over the longer haul is not clear, as some
of the temporary unemployed may end up being permanently dismissed
Figure 8 gives a broader idea of the socioeconomic composition of all the new claims during
the first stage of the coronavirus crisis. We see that the social gradient is clear no matter how we
capture socio-economic background. There is an overrepresentation among new unemployed people
of low-income, low-education, and low hourly wages.
10
Figure 8: Layoffs by socioeconomic status
A. Education B. Parental earnings rank
20
14
15 13
12
10
11
5
10
0 9
10 12 14 16 18 20 0 20 40 60 80 100
Years Percentile
Percent
C. Earnings rank D. Hourly wage
18 20
16
14 15
12
10
10
8 5
0 20 40 60 80 100 200 400 600 800
Percentile NOK
Men Women
Note: Panel A shows the number of school years associated with the highest completed education on the horizontal axis. In
panel B, the scale is the parent’s income rank, and the location of the data points is determined by a division into 10 equal
cells. Parents’ income is measured at age 52-58 for all employees born before 1983, then the measurement window is shifted
one year forward for each year the person is born after this. In Panel C, the income rank is based on their own “so far-in-life”
professional income (up to 2017) and each birth cohort and gender are ranked separately. Both in panels B and C, the rankings
were made within the entire population residing in Norway, while the division into 10 equal cells was made within the group of
workers included in our analysis.
There is also a systematic pattern by age and country of birth. This is illustrated in Figure
9. There is a clear pattern that young people have been hit harder than the elderly, and that
immigrants are more vulnerable than Norwegian-born.
11
Figure 9: Layoffs by gender, age and country of birth
A. Men B. Women
30 30
25 25
20 20
Percent
15 15
10 10
5 5
20 30 40 50 60 20 30 40 50 60
Age
All Native West Europe New EU Other
Note: Native includes children born in Norway by immigrant parents. In the group “Western Europe etc.,” Sweden, Germany
and Denmark are the three largest country groups in the analysis sample. “New EU countries” includes countries with mem-
bership since 2004; Poland and Lithuania are the two largest country groups in the analysis sample. In the “Other” group, the
Philippines, Thailand, Eritrea, Somalia and Iraq are the five largest country groups.
Within our data period, few wage earners have lost income during the crisis, since the laid off
are secured full pay for the first 20 days up to an annual income of approximately NOK 600,000
(USD 56,000). However, if the crisis persists, there is a risk of significant loss of income for the
laid-off and their families. The ability of families to manage a period of reduced (or lost) income
will depend on how much income they used to have and access to liquid financial reserves. Figure 10
shows clear indications that the likelihood of being laid off or dismissed is higher for less financially
secure households. This pattern is most clearly seen in panel A (household income) and B (the
size of bank deposits), but there is also a clear systematic pattern when we look at the household’s
interest burden relative to income. Those with the greatest interest burden are at the highest risk
of being laid off or terminated. The U-shape is due to the fact that many people with particularly
low financial resources also do not have debt, primarily because a large proportion of them do not
own their own housing.
12
Figure 10: Layoffs by household income and total bank deposits
A. Household income B. Household bank deposits C. Household interest burden
18
16
16
16
14
14
14
Percent
12
12
12
10
10
8 8 10
0 500 1000 1500 0 500 1000 1500 0 5 10 15 20
1000 NOK 1000 NOK Percent of income
Men Women
Note: Household income is adjusted for family composition (household income divided by the square root of the number of
family members) .
Having children is another important dimension of socioeconomic differences between families,
both because of the implications for the behavior of parents and because of the potential of adverse
outcomes for children. Figure 11 illustrates that families with children experienced a little bit
higher rate of layoffs at the bottom of the distribution and a bit lower at the top. Figure 11
focuses on children under 13, because this is a group for which childcare considerations are very
important. The interaction of the crisis with childcare is complex — on one hand, school closures
have an obvious consequence of creating a childcare crisis as well. The crisis package in Norway
adjusted the care benefit parents normally receive to look after their sick children to fit the current
circumstances. The days were doubled for each parent, there was no requirement of the child being
sick, and it allowed parents to transfer the days to the other parent in cases where one parent
had socially critical work. This alleviated this problem somewhat. However, laid off workers by
definition are not the ones that used the care benefit. At the same time, workers that can work
from home have a (costly, but still) ability to take care of their own children while continuing to
work. These considerations may be contributing to a different direction of the effect for lower and
higher income individuals.
13
Figure 11: Layoffs depending on the presence of a child under 13 at home. By position in the age-
and gender-specific lifetime earnings distribution
A. Men B. Women
20 20
15 15
Percent
10 10
5 5
0 20 40 60 80 100 0 20 40 60 80 100
Percentile
Children below 13 No children
The discussion so far makes it clear that layoffs are strongly associated with a number of
socioeconomic characteristics and result in a “social gradient” of the crisis. The social gradient
mainly arises as a result of the underlying social gradient in the professional structure in Norway.
The first phase of the coronavirus crisis affected people in occupations involving close physical
contact with other people, but without essential jobs. This pattern follows, to some extent, directly
from the authorities’ decision to close or restrict some types of businesses that involve the risk of
spreading the virus, such as hairdressing salons and restaurants. The result of this is shown in Figure
12, where we show the proportion of laid-off workers by occupation, with occupations placed on the
horizontal axis according to the extent to which they involve physical proximity with other people.
The size of each data point is proportional to the size of the occupational group in our data, and
for some of the largest occupational groups we have applied a professional designation in the figure.
Figure 12 illustrates a clear positive correlation between physical proximity in the profession and
the proportion of layoffs among both women and men.
14
Figure 12: Layoffs by occupation and the profession’s tendency to involve physical proximity with
others
A. Men B. Women
Hairdressers
70 70
60 60
Cooks
50 50
Waiters
Receptionist Waiters
Percent
40 40
Kitchen help
Car, taxi an
30 General mana
30
Tech sales Shop salespe
Tech sales Cleaners
20 Auto mech Shop salespe 20 Clerical off
Cleaners Stock clerks Electricians
Directors an
10 Heavy truck Carpenters a 10 Child-care w
Computer sys
0 0
Low High Low High
Degree of physical proximity to others
Note: Physical proximity at work is based on information provided by O*NET; see further description in section on data. The
data is limited to occupations in the private sector; occupation - gender combinations with fewer than 2,500 individuals are
excluded from the figure. The figures also show the regression fit of industry-gender-specific unemployment claims on physical
proximity, weighted by the number of workers in the occupation. The regression coefficient is 3.73 [0.69] for men and 4.97 [0.83]
for women.
Do industry and occupation specific impacts fully account for the presence of the gradient? We
investigate this question by regressing the likelihood of layoffs on individual characteristics and a
set of progressively more detailed fixed effects in Table 2. The sample is restricted to private sector
employees only. Without accounting for any firm or occupational characteristics, the multivariate
analysis confirms the relevance of education and hourly wage. It also indicates that younger workers
and those with shorter tenure were more strongly affected. There is an indication here that women
were more affected (this is in contrast to Figure 7 that includes both public and private sector
employees), and so were parents. Adding occupational or industry fixed effects attenuates gender,
education, tenure and wage effects, but they remain statistically significant. Accounting for industry
or occupation characteristics flips the sign of the age effect, suggesting that within industries or
occupations it may be older workers that are more strongly affected, for a given tenure. Firm and
combined firm and occupation effects attenuate most of these further somewhat.
An intriguing finding is the effect of having young children that remains almost unaffected once
firm and job fixed effects are controlled for. Furthermore, while the gender difference for those
without children disappears, the gender interaction with children becomes stronger. Once firm and
occupational differences are accounted for, the gender effect is only associated with the presence of
young children: women with young children are more likely to be laid off and this is a within-firm
and within-occupation effect. This potentially points to the employers accounting for childcare
in making layoff decisions, either unilaterally or in cooperation with employees who may prefer
15
temporary layoff in the presence of childcare obligations. It may also reflect employers´ belief that
the productivity of female workers would be most strongly affected by child care obligations due to
school and kindergarten closures, although it remains puzzling why temporary layoffs rather than
the adjusted care benefit would be used.
Table 2: Socio-economic characteristics and layoffs. Private sector.
(1) (2) (3) (4) (5)
Fixed Effects
OLS Occupation Industry Firm Firm and
occupation
Female 5.246*** 1.264*** 0.518*** 0.348*** 0.158
(0.345) (0.209) (0.156) (0.125) (0.107)
Child under 13 1.012*** 1.016*** 1.102*** 0.908*** 0.844***
(0.134) (0.108) (0.106) (0.0901) (0.0874)
Female and 0.103 0.800*** 1.346*** 1.256*** 1.242***
child under 13 (0.232) (0.178) (0.163) (0.159) (0.154)
Age -0.0265** 0.0317*** 0.0644*** 0.0495*** 0.0490***
(0.0108) (0.00755) (0.00660) (0.00590) (0.00608)
Education -1.106*** -0.343*** -0.357*** -0.122*** -0.0882***
(years) (0.0559) (0.0255) (0.0249) (0.0180) (0.0174)
Tenure -0.185*** -0.143*** -0.124*** -0.103*** -0.0967***
(0.0246) (0.0160) (0.0130) (0.00994) (0.00966)
Log hourly -4.649*** -3.114*** -2.366*** -1.147*** -1.038***
wage (0.275) (0.182) (0.165) (0.122) (0.119)
N 1723833 1723833 1723833 1723833 1723833
# Fixed effects 349 571 149963 149963+349
R2 0.020 0.129 0.184 0.434 0.439
Note: The sample is restricted to private sector workers. Dependent variable is an indicator variable set to 100 if laid off
between 3/1 and 4/19. The sample mean of the dependent variable is 18.1%. Standard errors are clustered within firms and
reported in parentheses. Significance level of 1% is denoted by ***, 5% by ** and 10% by *.
Figure 13 illustrates that this gender and parental effect is not an artifact of the choice of
controls. It shows the deviation of the likelihood of layoffs from firm-specific means (residuals from
regression of the layoff indicator on firm fixed effects) for men and women, separately for those
with and without children under 13. The effect of having a child under 13 is stronger for women
than for men and in each case it is driven by workers at the bottom of the distribution.
16
Figure 13: Within-firm relationship between having children under 13 and layoffs. By position in
the age- and gender-specific lifetime earnings distribution
A. Men B. Women
Deviation from group mean (percentage points)
3 3
2 2
1 1
0 0
-1 -1
-2 -2
0 20 40 60 80 100 0 20 40 60 80 100
Percentile
Children below 13 No children
Note: The Figure shows the average residuals from regression of layoff indicator on firm fixed effects and should be interpreted
as a deviation from firm-specific layoff rate.
5 The driving forces behind lay off decisions
The relevance of physical proximity for layoff decisions is natural, but it masks the dynamics that
these considerations played. As Figure 14 demonstrates, this association applies to the immediate
effect of the lockdown measures. The top panels show the relationship during the first 9 days after
lockdown measures were introduced, the bottom ones show it in the weeks that followed. The
association is much stronger in the initial weeks. A one standard deviation increase in physical
proximity requirements of an occupation is associated with a very significant 2.75 percentage point
increase in layoffs for men (t-stat of over 5) and 4.82 percentage point increase for women (t-stat
of over 6). The effect weakens afterwards to 0.92 for men (still significant with t-stat of 4) and a
small and insignificant one for women. While physical proximity plays a large role in the initial
impact, it becomes less of a factor over time, suggesting the presence of spillovers to other parts of
the economy.
17
Figure 14: Layoffs by occupation and the profession’s tendency to involve physical proximity with
others, over time
A. Men, layoffs 3/12-3/20 B. Women, layoffs 3/12-3/20
70 70 Hairdressers
60 60
50 50
40 Cooks 40
Waiters Receptionist Waiters
30 30
Kitchen help
20 Car, taxi an 20 Shop salespe
Shop salespe Tech sales
Cleaners
10 Tech sales
Stock clerks 10 Clerical off
Auto
Cleaners mech Electricians Child-care w
Computer sys Directors Heavy
an
truck
Carpenters a
0 0
Percent
C. Men, layoffs 3/21-4/19 D. Women, layoffs 3/21-4/19
70 70
60 60
50 50
40 40
30 30
20 20
Tech Auto
sales mechShop Waiters
salespe Tech sales Shop salespe
10 Stock
Cleaners clerks
Electricians 10 Clerical off Waiters
Cleaners
Computer sys Directors Heavy truck Carpenters a Child-care w
0 an 0
Low High Low High
Degree of physical proximity to others
Note: See Figure 12. The regression coefficients are 2.75 [0.50] and 0.92 [0.23] for men and 4.82 [0.67] and 0.15 [0.23] for women.
As the economic crisis spreads both internationally and through the economy through falling
demand for goods and services from the first affected businesses, we expect jobs with low physical
proximity and contagion risk to be affected as well. This is also what we find in the data. In Figure
15 we visualize how the composition of the new unemployment insurance applicants has changed
day by day after the layoff wave took off in earnest on Friday 13 March. While physical proximity
was a driving factor in the early days, the importance of this was greatly diminished in just a few
days. As the crisis spread through the economy, we also see that the average age is increasing and
the proportion of women is falling. In particular, while in the first few days of the crisis layoffs
were skewed toward females, this initial effect reversed quickly and by the end of the period more
men than women were laid off altogether.
When it comes to indicators of the social gradient, the picture is more mixed, and the variations
from day to day are rather moderate (note the scales on the vertical axes). We still see signs that
the average hourly wage among the laid-off increased somewhat during the first week. For all
socio-economic indicators, the affected individuals fall below the mean of the characteristic in the
analysis sample.
18
Figure 15: Characteristics of laid-off workers - day by day 3/3-4/19
A. Physical proximity B. Age C. Women
.8 42 60
55
.6 40
50
Degree Years Percent
.4 38
45
.2 36
40
0 34 35
3/9 3/16 3/23 3/30 4/6 4/13 4/20 3/9 3/16 3/23 3/30 4/6 4/13 4/20 3/9 3/16 3/23 3/30 4/6 4/13 4/20
D. Native born E. Education F. Hourly wage
85 14
300
80
13.5
75 280
Percent Years NOK
70
13
260
65
12.5
60 240
3/9 3/16 3/23 3/30 4/6 4/13 4/20 3/9 3/16 3/23 3/30 4/6 4/13 4/20 3/9 3/16 3/23 3/30 4/6 4/13 4/20
Note: The size of the circles is proportional to the number of layoffs on each day. The definitions of “physical proximity at
work” are described in more detail in the section on databases. See also note to Figure 8. The figures also show the mean value
of each characteristic as a dashed line.
Which firms responded most strongly? Figure 16 shows the relationship between layoffs and a
set of firm characteristics: productivity, equity and employment. In each case there is a very strong
negative relationship between these proxies for the strength of a firm and layoffs.
19
Figure 16: Characteristics of firms and layoffs
A. ln(productivity) B. ln(equity) C. ln(employment)
30 30
25
25
25
20
20
Percent
20
15
15
15
10
10
10
5 5
12.5 13 13.5 14 14.5 6 8 10 12 14 16 0 2 4 6 8
Note: Each point corresponds to a decile of the distribution. Productivity is measured by value added. Employment represents
the number of full-time equivalent workers.
Table 3 shows that this association is strong and statistically significant and it persists even
within industries.
Table 3: Characteristics of firms and layoffs
(1) (2) (3) (4)
OLS Industry FE OLS Industry FE
Log productivity -10.13*** -3.86*** -6.90*** -3.00***
(0.54) (0.39) (0.61) (0.38)
Log equity -1.17*** -0.74***
per worker (0.20) (0.17)
Log employment -2.13*** -1.52***
(No. FTE) (0.21) (0.14)
N 1,415,131 1,415,131 1,343,602 1,343,602
R-sq 0.023 0.169 0.040 0.176
#Industry FE 536 532
Note: The sample is restricted to private sector workers working for firms that can be linked to the 2017 accounting data.
Dependent variable is an indicator set to 100 for workers laid off between March 1st and April 19th . Productivity is calculated
as operating profit plus labor costs divided by the number of man-years, where man-years are calculated on the basis of a-forms
filed in 2019 and accounting figures are based on the annual accounts for 2017. Employment is measured as the number of man-
years (FTE) and calculated fromf a-forms filed in 2019. Standard errors are clustered within firms and reported in parentheses.
Significance level of 1% is denoted by ***, 5% by ** and 10% by *.
Figure 17 shows a timeline focusing on the characteristics of the companies that are laying off
employees. In Panel A we see that the problems affect companies with higher and higher labor
productivity (measured by total wages and profits per full-time equivalent), in Panel B that they
affect larger and larger companies, and in Panels C and D that fewer and fewer of the laid-off
20
workers come from companies with low equity and high debt.
Figure 17: Characteristics of private-sector enterprises laying off workers - day by day in the period
3/3-4/19
A. Productivity B. Size
9 800
8
100,000 NOK
600
Man-years
7
400
6
5 200
3/9 3/16 3/23 3/30 4/6 4/13 4/20 3/9 3/16 3/23 3/30 4/6 4/13 4/20
C. Low equity D. High debt
35
35
30
30
Percent Percent
25 25
20 20
15 15
3/9 3/16 3/23 3/30 4/6 4/13 4/20 3/9 3/16 3/23 3/30 4/6 4/13 4/20
Note: The size of the circles is proportional to the number of layoffs on each day. In Panel A, productivity is calculated as
operating profit plus labor costs divided by the number of man-years, where man-years are calculated on the basis of a-forms in
2019 and accounting figures are based on the annual accounts for 2017. In panel B, the number of man-years is calculated from
a-forms filed in 2019. In Panel C, low equity is defined as equity below NOK 40,000 per full-time equivalent. In Panel D, high
debt is high debt defined as debt that exceeds eight times equity. The figures also show the mean value of each characteristic
as a dashed line.
Figure 18 shows the extent of layoffs in selected industries, affected by the crisis in very different
ways. Hotels and restaurants were among the first industries directly affected, and during the first
days they accounted for 30-35% of the new unemployment benefit applications. In just over a week,
this percentage dropped to about 10%. Soon after, layoffs in the retail sector followed. They were
accompanied by layoffs among wholesalers that then continued in the following weeks. Layoffs in
construction were not as important initially, but then continued at a level about proportional to
the industry share. It appears that the petroleum industry has been more indirectly affected by
a dramatic fall in oil prices than the domestic response to the crisis, as indicated by the absence
of extensive layoffs during the first weeks of the crisis. The sector accounted for only 0.7% of new
unemployment claims during the initial nine days of the crisis, and 1.8% of all layoffs between
March 21 and April 19.
21
Figure 18: Layoffs in different industries over time
A. Retail B. Restaurants C. Wholesale
25 25 10
20 20 8
Percent Percent Percent
15 15 6
10 10 4
5 5
2
3/9 3/16 3/23 3/30 4/6 4/13 4/20 3/9 3/16 3/23 3/30 4/6 4/13 4/20 3/9 3/16 3/23 3/30 4/6 4/13 4/20
D. Construction E. Hotels F. Petroleum
12 20
4
10
15 3
8
Percent Percent Percent
10 2
6
4 5 1
2
0 0
3/9 3/16 3/23 3/30 4/6 4/13 4/20 3/9 3/16 3/23 3/30 4/6 4/13 4/20 3/9 3/16 3/23 3/30 4/6 4/13 4/20
Note: The petroleum industry with 3.6 percent of total private-sector employment comprises extraction of crude oil and natural
gas, drilling and other services related to extraction, construction of platforms and modules, furnishing and installation work
and supply bases. The horizontal line represents the share of employment accounted for by a given industry
6 Summary and conclusion
The coronavirus crisis struck broadly, but not randomly. There are clear systematic patterns in
which workers are most affected. The public sector is so far shielded from layoffs, but there are
many public employees in socially critical and risk-exposed professions, especially in the health and
care sectors. Here, women are far more exposed than men, and people with low incomes are more
exposed than those with high incomes.
Among all employees, we find that the risk of layoffs or dismissals during the first phase of
the crisis is higher the lower the education, income, hourly wage and social class background of
an employee. While initially, women bear the brunt of the layoffs due to their over-representation
in sectors explicitly targeted by social distancing policy measures, over time this effect was more
than offset as the rest of the economy responded. Males are more exposed to layoffs than women,
because more men work in the private sector. Within the private sector, the gender effect is
associated with the presence of young children: women with young children are more likely to be
laid off comparing workers within the same firm and occupation by means of fixed effects. This
22
points to the importance of childcare in understanding the labor market impact of the crisis (see
also Dingel, Patterson and Vavra, 2020).
As the crisis spreads, both internationally and across sectors domestically, we see signs that
unemployment seekers are coming from ever-increasing occupations in the private sector. The
companies that are laying off have also changed over time. While industries such as tourism and
businesses with low productivity and equity dominated the first few days, we see a trend towards
more “average” businesses being affected.
The Norwegian context is also useful in judging the extent to which the crisis is expected to be
permanent. In the short term, about 1/10th of layoffs are classified as “permanent” and correspond
to severing employment relationships. The remaining layoffs are temporary. While the incentives
to lay off workers permanently are not strong even for firms in distress, as temporary layoffs may
be converted to permanent later on and permanent layoffs may trigger severance obligations, this
decomposition gives at least some reasons for hope that the great majority of layoffs in countries
where this choice is not easily available (such as the US) may in fact be reversible.
References
Adams-Prassl, A., Boneva, T., Golin, M., and Rauh, C. (2020) Inequality in the Impact of the
Coronavirus Shock: Evidence from Real Time Surveys, IZA Discussion Paper 13183.
Bartik, A., Bertrand, M., Lin, F., Rothstein, J., and Unrath, M. (2020) Labor Market Impacts of
COVID-19 on Hourly Workers in Small- and Medium-Sized Businesses: Four Facts from Home-
base Data., Online at https://irle.berkeley.edu/labor-market-impacts-of-covid-19-on-
hourly-workers-in-small-and-medium-sized-businesses-four-facts-from-homebase-data-
2/, accessed on 4/30/2020.
Berg, H., Larsen, K. A., Klingenberg, S., og Solheim, Ø. B. (2015) Permitteringer og avgang fra
arbeidslivet. Rapport 2015-007 Proba samfunnsanalyse.
Bratsberg, B., Raaum, O., and Røed, K. (2010) When Minority Labor Migrants Meet the Welfare
State. Journal of Labor Economics, Vol. 28, No. 3, 633-676.
Bratsberg, B., Raaum, O., and Røed, K. (2018) Job Loss and Immigrant Labor Market Perfor-
mance. Economica, Vol. 85, 124–151.
Coibion, O., Gorodnichenko, Y., and Weber, M. (2020) Labor Markets During the COVID-19
Crisis: A Preliminary View, NBER Working Paper No. 27017.
Direktoratet for samfunnssikkerhet og beredskap (2016) Samfunnets kritiske funksjoner. HR 2350.
Dingel, J., Patterson, C., and Vavra, J. (2020) Childcare Obligations Will Constrain Many Workers
When Reopening the US Economy. Becker Friedman Institute for Economics at University of
Chicago, Working Paper 2020-46.
23
Dustmann, C., Glitz, A., and Vogel, T. (2010) Employment, Wages, and the Economic Cycle:
Differences between Immigrants and Natives. European Economic Review, Vol 54, No. 1, 1-17.
Hoen, M. (2016) Occupational crosswalk, data and language requirements. Working Paper 1/2016.
The Ragnar Frisch Centre for Economic Research.
Hoynes, H. Miller, D. L., and Schaller, J. (2012) Who Suffers During Recessions? Journal of
Economic Perspectives, Vol. 26, No. 3, 27-48.
NAV (2020) Statistics on unemployment insurance applications. Updated daily at 10am on http:
//www.nav.no.
Yagan, D. (2019) Employment Hysteresis from the Great Recession. Journal of Political Economy,
Vol. 127, No. 5, 2505-2558.
24
File and source
- File
- alstadster-2020-the-first-weeks-of-the-coronavirus-crisis-who-got-hit-when-and-why-evidenc.pdf
- Size
- 593,448 bytes
- SHA-256
- 978b881c31432131f642425f15ed8024acdcaf07adcf690194f4a89d3870e6df
- Our copy
- alstadster-2020-the-first-weeks-of-the-coronavirus-crisis-who-got-hit-when-and-why-evidenc.pdf
- Original
- www.nber.org