Policy Research Working Paper                           9873




                      Occupational Hazards
  Why Migrants Faced Greater Economic and Health Risks
           during the COVID-19 Pandemic

                                 Laurent Bossavie
                               Daniel Garrote Sanchez
                                 Mattia Makovec
                                   Caglar Ozden




Social Protection and Jobs Global Practice
  &
Development Research Group
December 2021
Policy Research Working Paper 9873


  Abstract
 This paper investigates the economic and health risks arising                      pandemic. This has led to an increase in both their income
 from the COVID-19 pandemic for migrant workers in the                              and employment risks. Immigrants from regions outside
 European Union. It assesses migrants’ economic and health                          Europe were more vulnerable than those from within
 vulnerabilities using ex ante measures based on both supply                        Europe or native-born workers. The paper shows that indi-
 and demand shocks. The analysis finds that immigrants                              vidual characteristics, such as educational attainment, age,
 were more vulnerable than native-born workers to both                              and geographical location, fail to explain the native-migrant
 income- and health-related risks, and that this greater expo-                      gap in exposure to economic and health risks posed by the
 sure stems from the occupations in which migrant workers                           pandemic. Limited language ability, the concentration of
 are concentrated. Migrants work to a greater degree than                           migrants in jobs with labor shortages among native-born
 native-born citizens in occupations that are less amena-                           workers, and a reliance on immigrant networks to find jobs
 ble to teleworking arrangements, and in economic sectors                           all appear to play significant roles in migrants’ exposure to
 that experienced greater reductions in demand during the                           pandemic-related risks.




 This paper is a product of the Social Protection and Jobs Global Practice and the Development Research Group, Development
 Economics. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to
 development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://
 www.worldbank.org/prwp. The authors may be contacted at at lbossavie@worldbank.org, dgarrotesanchez@worldbank.org,
 mmakovec@worldbank.org, and cozden@worldbank.org.




         The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development
         issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the
         names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those
         of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and
         its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.


                                                       Produced by the Research Support Team
                               Occupational Hazards:
   Why Migrants Faced Greater Economic and Health
       Risks during the COVID-19 Pandemic

                                       Laurent Bossavie
                                     Daniel Garrote Sanchez
                                        Mattia Makovec
                                         Caglar Ozden




JELcodes: J17, J21, J28, J61
Keywords; migrant workers, COVID-19 pandemic, labor market vulnerability.
1. Introduction


COVID-19 exerted unforeseen and sudden pressures on the labor markets across the world, leading
unemployment rates to rise and wages to fall. Even in high-income, Western European countries
with extensive social safety nets, large swaths of the labor force were negatively affected. On the
supply side, workers in many different occupations and sectors faced significant disruptions as
governments imposed various lockdown measures, restricting mobility, and closing non-essential
businesses to slow down the spread of the epidemic. On the demand side, the sudden decline in
overall economic activity set off a chain reaction of earnings losses and downsizing in
transportation, tourism, direct retail, and related sectors both upstream and downstream. Relatively
large government stimulus packages and increases in demand in other sectors, such as health care
and internet retail, failed to compensate for these losses. Because workers sort into different types
of jobs according to their comparative advantages, educational backgrounds, and social networks
(Patel and Vella, 2013; World Bank, 2018), the upheaval of the pandemic rendered some workers
more vulnerable to these various shocks simply by virtue of the work they do.
This paper focuses on the labor market and health ramifications of the pandemic for a specific
group of workers: immigrants in the European Union (EU). We show that immigrant workers were
uniquely vulnerable to pandemic-related shocks, and that their heightened vulnerability is the
result of the greater concentration in occupations that, by the nature of the work involved, proved
to pose greater health and income risks. We show that the risks were greater among immigrants
from certain countries, and that certain factors – such as limited language skills, a reliance on
immigrant networks to find employment, and a tendency for immigrants to fill jobs that have
shortages of native-born workers –likely played a role in this increased concentration in higher-
risk work.
Income losses have welfare repercussions for workers, and the ramifications can be especially
severe for migrant workers. Many migrants lack access to social protection programs in destination
countries (Avato, Koettl, and Sabates-Wheeler 2010); access to such support often depends on
maintaining a job, or on having a permanent rather than temporary immigration status. Indeed, the
loss of a job can threaten a migrant’s ability to remain in the destination country. Income losses
for migrants can have large ripple effects across the world, especially in poorer origin countries
where remittances from workers abroad form a large share of the GDP. Migrants’ remittances to
their families in their home countries provide vital income support to many households, especially
the poorest; receipt of such remittances has been associated with better health and education
outcomes and lower poverty rates (Amuedo-Dorantes and Pozo 2010; World Bank, 2018).
Eurostat data show that employment of migrants in EU28 countries fell by 6.1 percent in the one-
year period between the second quarters of 2019 and 2020; this compares to a 2 percent drop for
natives (Eurostat, 2021). Studies have also found a higher incidence of COVID-19 cases among
the migrant populations (Diaz et al., 2021). Occupation-specific and individual-level analyses of
the drivers of employment trends across different groups have not been possible because of the
lack of the microdata needed. Thus, in this paper we use ex ante measures of exposure to COVID-

                                                 2
19-specific risks to assess the vulnerabilities of native and migrant workers in Europe. We show
that that these ex-ante measures correlate with ex-post employment trends, thus validating the
methodology.
On the labor supply side, we construct three different metrics: (i) a measure of the share of jobs
that are not amenable to telework; (ii) a measure of jobs that were not deemed essential by
European governments; and (iii) a measure of income risk that combines these two measures. On
the labor demand side, the paper measures economic exposure by using estimates of changes in
consumption of goods and services across sectors. The analysis of supply and demand shocks
provides insights on the extent that a worker is in jeopardy to losing a job or facing a reduction in
hours or wages, depending on his or her age, occupation, or level of education; it also examines
the degree to which certain workers are in jobs that entailed higher health risks both because they
could not carry out their work from home, and because their work required extensive face-to-face
interactions. This is the case for those who work in many essential services, such as health care
and education.
The paper finds that migrants in the EU were particularly vulnerable to both income- and health-
related risks. The greater exposure stems from the characteristics of the occupations in which
migrant workers tend to be employed. Prior to the pandemic, migrants tended to work to a greater
degree in occupations that were less likely to be possible to do via teleworking arrangements, and
more likely to be exposed to demand shocks. The concentration of migrants in these occupations
led them to face greater income risks when the pandemic hit. Furthermore, migrants were
employed in occupations that were more likely to require face-to-face interactions, increasing their
exposure to health risks too.
We also show, however, that exposure to these pandemic-related risks varies widely among
migrants themselves, depending on which part of the world they emigrate from. Migrants from
other high-income countries that are members of the Organization for Economic Co-operation and
Development (OECD) (in Western Europe, North America or East Asia) are less vulnerable than
the native workers; by contrast, migrants from other specific locations – relatively new EU member
states (such as Poland and Hungary), Eastern Europe, the Middle East and Sub-Saharan Africa –
are significantly more vulnerable than native workers.
Migrants tend to fill specific labor-market gaps according to their age, skills, and education levels;
however, these differences in demographic profiles explain only a small part of the diverging
occupational choices between native-born and migrant workers – and, hence, between the different
levels of risk exposure that we document. We examine three additional factors that may contribute
to the patterns we find. First, migrants’ skills and human capital are transferred in imperfect ways
when they seek work in destination countries’ labor markets. Migrant workers face barriers that
stem from their own language skills, and from the difficulty that destination countries have in
recognizing foreign credentials. As a result, migrants tend to select into different occupations,
which differ vertically (the occupations require lower levels of education) or horizontally (i.e., the
occupations require the same education levels but entail different types of tasks). These jobs tend
to be less amenable to work-from-home options and more likely to be strongly hit by the pandemic-
related supply restrictions (Mongey, Pilossoph and Weinberg, 2020). Second, migrants rely on

                                                  3
networks of preceding waves of friends, families, and diasporas to find jobs when they arrive. This
process creates a path dependency in the selection into specific occupations and perpetuates the
occupational divide. Finally, migrants are more likely to be employed in dynamic occupations with
higher employment turnover which leads to higher vulnerability.
Our paper contributes to several strands of literature in economics and other disciplines. First, it
adds to the growing literature on the employment and labor market consequences of the COVID-
19 pandemic. Most papers thus far have explored the overall labor-market impacts in high-income
countries (Adams-Pressel et al., 2020a; Adams-Pressel et al., 2020b; Bartik et al., 2020; Cortes
and Forsythe 2020; Rojas et al., 2020). Related studies have explored the impact on different
groups within the labor force (Albanesi and Kim, 2021; Mongey et al., 2020; Alon et al., 2021).1
The emerging pattern is that the groups of workers who were already vulnerable to income and
employment risks prior to the pandemic are also the ones most negatively impacted from COVID-
19-related shocks. The literature thus far shows that employment risks associated with COVID-19
have disproportionately affected low-wage and less-educated workers (Rio Chanona et al., 2020;
Mongey, Pilossoph and Weinberg, 2020). Similarly, Garrote-Sanchez et al. (2020) find that the
pandemic exacerbated inequality in the labor market, especially in richer countries where better-
paid and better-educated workers were almost fully insulated from the shock.
We focus on migrant laborers, who represent an under-examined category of workers within the
literature that has analyzed the labor-markets impacts of the pandemic thus far. A handful of
studies have looked at this group, finding that immigrant workers tended to be more exposed to
the negative labor market consequences of the pandemic (Boeri et al. 2020; Borjas and Casidi,
2020; Fassani and Mazza, 2020a; Fasani and Mazza, 2020b; Yasenov, 2020). However, the drivers
of increased exposure among migrant workers have yet to be analyzed in detail. In contrast to
previous contributions, this paper provides a detailed descriptive analysis of the potential drivers
of the higher economic and health risks faced by migrants during the pandemic. We focus on
migrant workers in the EU, one of the main destinations for immigrants; the region hosts some 60
million of the world’s estimated 272 million immigrant workers. By taking advantage of the
harmonized labor-force data available for multiple destination countries in Europe, we examine
potential drivers at a granular level, taking into account migrants’ personal characteristics,
occupations, and origin countries.
The paper also contributes to a literature examining and comparing the occupational choices of
immigrant and native workers, and the risk levels for the two groups that stem from these
occupational choices. The literature has primarily focused on occupational hazards and has shown
that immigrant workers are over-represented in occupations with higher rates of injuries and
fatalities. This has been evidenced in the US context (Orrenius and Zavodny 2009; Hersch and
Viscusi, 2010; Dávila, Mora and González 2011; Zavodny, 2015), but also in Spain (Díaz-Serrano,
2010; Solé, Díaz-Serrano and Rodríguez, 2010), Germany (Giuntella et al., 2015), Canada (Premji
et al., 2010; Smith, Kosny and Mustard, 2009) and Australia (Corvalan, Driscoll and Harrison,


1
 For evidence on the labor market effects of COVID-19 in developing economies, see for example Kamis et al. (2021)
and Kugler et al. (2021).

                                                        4
1994).2 The economics literature has shown that immigrant workers are concentrated in
occupations that are more sensitive to business cycle fluctuations, exposing them to economic
downturns and shocks (Dustmann, Glitz and Vogel, 2010; Orrenius and Zavodny, 2010). In this
paper, we show that the greater exposure of immigrant workers to employment-related risks also
manifested itself in the specific context of the COVID-19 pandemic, and we provide descriptive
evidence of the drivers behind this greater risk exposure.3
The remainder of the paper is organized as follows: Section 2 describes our measurements of
occupational vulnerability to COVID-19. Section 3 describes the datasets used in the paper.
Section 4 provides descriptive statistics on occupational exposure to COVID-19 in the EU member
countries. Section 5 discusses the empirical strategy. Section 6 explores the main contributors of
the greater occupational exposure of immigrant workers to the COVID-19 shock. It investigates
whether systematic differences in observable characteristics between native and migrant workers
can explain differences in exposure, and it explores additional factors, such as the effects of
networks and the imperfect transferability of skills. Section 7 explores the role of differences in
responsiveness between migrants and natives to occupation-specific demand dynamics. Section 8
concludes.



2. Measurement of Occupational Exposure to COVID-19

2.1 Different Measures of Risk Exposure


Risk exposure is an ex-ante measure; it needs to be assessed before the harmful events take place
and people react to them. Otherwise, people’s endogenous responses to the events and resulting
changes in their behavior will confound attempts to measure the risk exposure. Therefore, we use
information on workers’ occupations and labor market outcomes prior to the emergence of the
pandemic to measure their economic risk exposure and health vulnerability to COVID-19. This is
also the general approach adopted in the rapidly expanding literature (see Dingel and Neiman,
2020; Garrote Sanchez et. al. 2020; Fasani and Mazza, 2020a; Fasani and Mazza, 2020b).
We use five different metrics of occupational vulnerability and risk throughout this paper. The first
metric categorizes whether jobs are or are not essential (i.e., whether the jobs are exempt from the
mobility restrictions and lockdowns imposed by governments to limit the spread of the pandemic).
2
  Orrenius and Zavodny (2013) also summarize the literature on the higher concentration of immigrant workers in
risky occupations.
3
  Limited host country language proficiency has been shown to be associated with higher-risk occupations in the
United States (Orrenius and Zavodny, 2009; Marvasti, 2010; Dávila, Mora and González, 2011) and Australia
(Corvalan, Driscoll and Harrison, 1994). A negative relationship between occupational risk and the number of years
since migration in Taiwan, China (Wu et al., 1997) and the United States (Hao, 2008); this relationship could be due
in part to increased proficiency in the host country language over time. It has also been argued that immigrant workers
may have different risk preferences than natives; by virtue of selection, immigrants may be less risk averse, as
suggested by a willingness to move overseas (Bonin et al., 2006; 2009).


                                                          5
The second metric assesses whether jobs are or are not amenable to work-from-home
arrangements, which proved to be crucial when mobility restrictions were in place. The third
measure is a combination of the previous two, and it serves as a proxy for the overall vulnerability
to income risks for a given job as measured from the labor-supply side. The fourth metric focuses
on the extent to which a given job involves intensive face-to-face interactions and, thus, results in
more exposure to COVID-19 health risks. Finally, the fifth metric is based on exposure to demand
shocks; it measures how much consumption of goods and services produced by that sector are
likely to be affected by COVID-19.


Non-essential Jobs
When categorizing jobs as non-essential, we start with Fasani and Mazza (2020a), who use the
European Commission’s guidelines concerning the exercise of the free movement of workers
during the COVID-19 outbreak4 and then, supplement it with the Dutch government’s definition
of “key workers.” After the onset of the pandemic, many governments created lists of occupations
that were considered essential and thus were exempt from mobility restrictions and closures that
were imposed during lockdowns. While the criteria for essential occupations vary from country to
country, there is significant overlap. Every list, for example, includes core sectors that deliver
critical healthcare (doctors, nurses and other medical professionals) and those that provide basic
goods and services (food delivery, utilities, transport, security, ICT, and research and science).
Lockdowns were initially imposed, progressively lifted during the summer of 2020, and then
reintroduced by some governments over the course of the pandemic; thus, these restrictions had
significant consequences on the ability to go to work. Based on these criteria, we identify essential
workers at the three-digit ISCO-08 classification codes (which includes 182 occupations).


“Non-teleworkable” Jobs
The second measure we use is related to the degree to which the tasks of a given occupation can
be performed from home. Recent literature on the impact of COVID-19 on labor markets
highlights the feasibility of home-based work (telework) as the key to overcome governments’
mobility restrictions and associated income/unemployment risks, and as a way to limit exposure
to COVID-19 (Dingel and Neiman, 2020; Garrote Sanchez et al., 2020). In constructing our
measure of the “teleworkability” of a job, we follow Dingel and Neiman (2020), who calculate the
amenability of different occupations for telework in the United States by using surveys from the
Occupational Information Network (O*NET) on work context and generalized work activities.
Jobs that are not considered to be amenable to telework arrangements are defined to include those
with the following attributes: requiring workers to spend most time outdoors; requiring workers to
perform physical activity, or to handle moving objects, such as vehicles or machines; requiring
workers to repair large mechanical or electronic equipment; entailing worker exposure to diseases
or infections; involving work in which email is very seldom used. This measure, however, is based

4
    https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:52020XC0330(03)

                                                       6
on six-digit-level disaggregation of the Standard Occupational Classification (SOC) system. Given
that the EU Labour Force Survey has a coarser disaggregation of occupations, we aggregate our
telework measure to three-digit SOC occupations using employment weights based on the Current
Population Survey of the US Bureau of Labor Statistics. Therefore, the six-digit-level variable
used by Dingel and Neiman (2020) becomes a continuous variable between zero and one in our
data; this variable reflects the share of jobs within each broad occupation category that is amenable
to telework. We then transform this classification to the International Standard Classification of
Occupations (ISCO08) used in the EU labor force survey.


Jobs with Greater Income Risks
Using these criteria, we categorize a job as vulnerable to income risks if it is neither essential nor
able to be undertaken via telework. Workers in these jobs are exposed to higher risks of dismissal
and loss of income given that mobility restrictions impact their ability to perform their duties at
their workplaces.


Jobs with Greater Health Risks
When assessing job-specific exposure to health risks, as our fourth measure, we first quantify the
extent of face-to-face interactions required by occupations. We follow Avdiu and Nayyar (2020)
who create an index based on O*NET surveys. In this case, we assume that a job requires high
face-to-face interactions if it involves tasks such as establishing and maintaining personal
relationships, assisting and caring for others, performing for or working directly with the public,
or selling to or influencing others through extensive personal interaction (Garrote-Sanchez et al.,
2020). We categorize jobs as exposed to health risks, if they require above average face-to-face
interaction level, are deemed essential and cannot be performed from home.5 These jobs, such as
those in healthcare or public transportation, are generally shielded from economic risks as they are
considered essential, but they are highly exposed to the COVID-19 contagion given their high
personal interaction and lack of amenability to work from home.


Jobs That Are More Vulnerable to Demand Shocks
COVID-19 affected not only the labor supply (by restricting mobility) but also labor demand by
changing consumption patterns (Rio Chanona et al., 2020; Baker et al., 2020, Brinca et al., 2020).
While certain sectors, such as the health care sector, experienced an increase in demand, overall
economic activity fell sharply as sectoral GDP estimates have shown. Sectors such as tourism,
hospitality and recreational activities were particularly hit hard as consumers reduced their
exposure to the virus. We observe that jobs that have greater income risks are more likely to have

5
  Non-essential jobs that require a high degree of face-to-face interactions suffered more from income risks rather than
from health risks; this is because mobility restrictions in many situations prevented workers from performing their
duties.

                                                           7
experienced a reduction in demand too; so, there is a strong positive correlation between the supply
and demand shocks. However, the demand shock may have continued for after government
lockdown measures ended and other supply restrictions were lifted as the result of lower household
spending capacity, and higher levels of uncertainty about the economy.
To capture the uneven impact of the pandemic experience on consumer spending across sectors,
we follow Rio Chanona et al. (2020) who base their analysis on an earlier model developed by the
US Congressional Budget Office (CBO, 2006) to generate scenarios to examine the possible
impacts that could occur in the event of an influenza pandemic. The most-severe scenario, modeled
after the SARS episode in east Asia, estimated that an 11.5 percent reduction in consumer spending
would occur (CBO, 2006). For comparison, private consumption in 2020 in the Euro Area fell by
8.4% in real terms compared to 2019, and by 12.8% in the United Kingdom.6 The CBO provides
estimates of the size of the change in demand across two-digit categories of the North American
Industry Classification System (NAICS); we convert these to two-digit categories used in the
statistical classification system of the EU (NACE). Under a scenario that assumed an outbreak
similar to the one that occurred during the Spanish flu pandemic of 1918-1919, the US
Congressional Budget Office estimated that consumption in the hospitality and in arts,
entertainment and recreation sectors would plummet by 80 percent (CBO, 2006). Under this
scenario, it estimated that demand for healthcare services would rise by 15 percent, and that no
major consumption changes would occur in other sectors, such as utilities, professional activities,
financial and ICT services, public administration and education.


2.2 Predictive Power of the Different Vulnerability Measures on Recent Employment Trends


Before proceeding to formal analysis of migrant workers’ risk exposure in the labor market due to
COVID-19 related supply and demand shocks, we assess the overall validity of our five different
ex-ante measures of job vulnerability. We examine the correlations between these measures and
the actual employment growth rate between the second quarter of 2019 and the second quarter of
2020 for every ISCO one-digit occupation, every NACE one-digit sector, and each country in the
EU, using recent employment statistics from Eurostat. We regress the employment growth rate on
the risk measure; we include country fixed-effects to control for country-specific factors that would
impact labor markets.
As shown in Table 1, our measures of exposure to supply and demand shocks are strongly
correlated with the job losses that occurred in the quarter when the pandemic related closures began
to take place throughout Europe. Jobs that we classify has having greater income risks show a drop
in employment of around 8 percent between 2019Q2 and 2020Q2 (column 3), even after
controlling for country fixed effects. As expected, occupations categorized as having health risks
were not correlated with changes in employment, as the vulnerability is not related to dismissals
but rather exposure to the COVID-19 virus itself (column 4). With respect to the demand side, a 1
percent decline in the estimated sectoral demand in a given occupation was associated with a 0.15

6
    OECD (2021).

                                                 8
percent drop in employment (column 5). To eliminate potential pre-pandemic trends, we undertake
a similar regression analysis (shown in Table 2) for two periods: growth between 2018Q2 and
2019Q2 (pre-pandemic), and growth between 2019Q2 and 2020Q2 (post-pandemic). In general,
the different measures of vulnerability do not explain changes in employment before the pandemic
emerged (aside from in telework jobs), while each measure of supply- and demand-side income
vulnerability is highly significant for the latest period (2019Q2-2020Q2), as shown by the
significance of the interaction terms with a dummy variable for the 2020Q2 period. These results
provide comfort on the validity of our metrics to measure risk exposure to the pandemic-generated
crisis in the absence of fully available and homogenous employment microdata in the EU.



3. Data Sources

The primary data source used throughout the paper is the European Labor Force Survey (EU-LFS),
which is constructed by harmonizing nationally representative labor force surveys in all of the 27
member countries of the European Union, in addition to Norway, Switzerland and the United
Kingdom. The total sample size ranges between 4.5 million and 5.2 million observations,
depending on the year. The dataset annually compiles detailed information on labor-market
outcomes of the working-age population and their demographic characteristics, such as gender,
age, and educational attainment. We focus on the survey results from 2018, the latest available
year prior to the COVID-19 pandemic.
The dataset provides information on whether the individuals were born in the country in which
they currently reside. We do not observe the specific country of birth or citizenship status of the
migrant workers, but we see nine world regions. These are (i) the high-income EU-15 countries
(including the UK),7 (ii) high-income European Free Trade Association (EFTA) countries,8 (iii)
the new member states of the EU (referred to as NMS13)9, (iv) non-EU European countries10, (v)
Latin America, (vi) Asia, (vii) the Middle East and North Africa, (viii) Sub-Saharan Africa, and
(ix) other high-income OECD countries that are not in Europe (mainly, the US, Australia, Canada
and New Zealand).
We have detailed information on various labor-market outcomes of the individuals in the dataset.
In terms of labor market status, people are labeled as (i) out of the labor force, (ii) seeking
employment, or (iii) currently employed. For the employed workers, there is detailed information
on the characteristics of their jobs. These variables include the sector of activity at the one-digit-

7
  At the time of the EU-LFS surveys, the United Kingdom was a member of the European Union, but formally left on
January 31, 2020. As such, in most of our analysis, the United Kingdom is included with the high-income EU15
countries.
8
  The EFTA Member States are Iceland, Liechtenstein, Norway, and Switzerland.
9
  These are the 13 countries that joined the EU as part of 2004 enlargement and subsequently. The group consists of
Bulgaria, Croatia, Cyprus, the Czech Republic, Estonia, Hungary, Latvia, Lithuania, Malta, Poland, Romania,
Slovenia, and the Slovak Republic.
10
   These countries are Albania, Armenia, Azerbaijan, Belarus, Bosnia-Herzegovina, Georgia, Republic of Moldova,
Montenegro, the Russian Federation, Serbia, Turkey and Ukraine.

                                                        9
disaggregation level (using NACE classifications) and detailed occupational categories at the
three-digit level (using ISCO-08 classifications). This information enables us to categorize jobs
according to the degree of exposure to COVID-19.
The 2014 survey of the EU-LFS includes a special migration module that contains further
information on immigrants. These variables include the reason for migration, the level of language
fluency of the migrant, the main barriers to labor-market participation (legal documents,
recognition of foreign qualifications, language fluency, ethnic or religious background), and the
extent to which the migrant is overqualified for the job. These data allow us to better understand
the skill profiles of migrants in comparison to those of native-born workers.
To assess the task content and different characteristics of occupations, we use two surveys
administered by O*NET, the program sponsored by the US Department of Labor to improve the
understanding of the nature of work. The O*NET database contains a vast array of standardized,
occupation-specific characteristics for almost 1,000 occupations.



4. Exposure of Immigrant and Native-Born Workers to Pandemic-Related Risks

Measure 1: Non-essential Occupations
Our first comparison between immigrants and native-born workers focuses on the share of workers
in non-essential occupations to assess their relative economic exposure to measures that were put
in place to contain the spread of COVID-19. We find that migrants in the EU are less likely than
native-born citizens to work in occupations that are defined as non-essential. As shown in Panel
A of Figure 1, 58 percent of immigrants and 63 percent of native-born workers were employed in
non-essential occupations in 2018.11 Although the share of non-essential occupations in the labor
force varies across all destination countries, migrants are under-represented in these roles in almost
every host country in the EU15. This is especially the case with migrants from non-EU countries
or new member states (NMS13), as depicted by the last two columns of Panel A; however,
migrants from within the EU15 countries (the high-income Western and Northern European
countries that were already EU members before the 2004 enlargement) are represented in these
jobs to an almost identical degree as native workers.
The second important observation is about the age distribution of workers in non-essential jobs.
Among all workers, older workers are less likely to work in non-essential occupations. Their share
in these jobs declines with age. The relationship between employment in non-essential occupations
and age is stronger among immigrants than among natives (Figure 1, Panel B). For example, 60
percent of native-born workers and 70 percent of immigrant workers ages 50 to 64 are in non-

11
  This estimate is slightly higher than that of Fasani and Mazza (2020a), who estimate that 33 percent of workers in
the EU work in essential sectors. The disparities emerge from two sources. First, we include the occupations related
to “Agriculture, Forestry and Fishery Laborers” (ISCO -08 code 921) based on government regulations. Second, we
use an approximation of essential occupations at the two-digit level for Bulgaria, Poland and Slovenia because the
EU-LFS does not include three-digit occupations for those countries.

                                                        10
essential jobs. By contrast, the share of young (between the ages of 20 to 24) native and immigrant
workers who are in non-essential occupations is almost identical (around 70 percent).
Essential workers tend to be concentrated at the two extremes of the occupational skill/education
spectrum, both in higher-skilled/education careers (such as those in healthcare) and in lower-
skilled/education occupations (such as those in food delivery and transportation). This pattern
holds for both the immigrant and native-born workers (Figure 1, Panel C). While the share of non-
essential jobs held by immigrant workers is lower across all education categories, the gap between
the share of immigrant and native workers in groups with low levels of education is wider.
A natural consequence of essential workers clustering in these groups with lowest and highest
levels of education is that they are also overrepresented in the highest and lowest income groups.
And as it is the case with education categories, a significantly lower share of the non-essential
immigrant workers is in the lowest income deciles relative to the native-born essential workers
(Figure 1, Panel D). The gap between immigrant and native-born workers is negligible in the fifth
through the ninth income deciles. Native workers surpass the immigrant workers only in the
highest decile.
The concentration of immigrants in lower-skill (and therefore lower-income) jobs within essential
occupations shows some variation across destination countries. Figure 2 presents the share of
migrant workers in six of the essential occupations with larger shares of immigrant employment
for the five largest destination countries. This list includes high-skilled occupations such as
medical doctors, nurses, and software developers; middle-skilled occupations such as personal
care workers; and low-skilled occupations such as drivers, cleaners, and helpers. Except for the
United Kingdom, where the share of migrants is higher for medical doctors, migrant workers in
the main EU countries are concentrated to a greater degree in lower-skilled occupations within the
essential occupations, particularly as cleaners and helpers.


Measure 2: Occupations That Are Not Amenable to Telework
Our second risk measure is the likelihood that a worker is in an occupation that cannot be
performed from home (“non-teleworkable”). We find that immigrants are significantly more likely
to work in jobs that are not amenable to telework (Figure 3, Panel A). While 64 percent of natives
have jobs that do not make teleworking possible, the ratio rises up to 72 percent for migrants,
increasing their exposure to the pandemic-related risks posed by this type of work. In addition to
this overall gap, we again find significant heterogeneity across different immigrant groups.
Migrants from new member states (NMS13) are particularly at high risk given that 79 percent of
them work in occupations that involve tasks that cannot be performed from home. By contrast,
migrants from higher-income EU15 countries are even less likely to have non-teleworkable jobs
than natives (60 percent vs. 64 percent).
Across age groups, prime-age (25-39) workers are more likely to telework, while very young
workers (15-24) who recently entered the labor force are the least likely to do so (Figure 3, Panel
B). The native-immigrant teleworking gap is larger for older populations. In terms of education

                                                11
and income groups, jobs amenable to telework are more concentrated among highly educated and
high-income workers, regardless of whether they are native-born or immigrants. This teleworking
gap between migrants and natives is quite stable across education levels, at slightly less than 10
percent (Panel C). The gap, however, declines as income levels increase (Panel D). While the
highest-income natives and migrants hold almost identical shares of jobs that are amenable to
teleworking, the gap is substantial for the lower deciles. For example, within the bottom four
income groups, around 85 percent of migrant workers have non-teleworkable jobs, compared to
70 to 75 percent of native workers.


Measure 3: Occupational Exposure to Income Risks
Occupations that face the highest income and unemployment risks are those that are categorized
as non-essential and, at the same time, as not amenable to telework. When an occupation is both
non-essential and non-amenable to telework, social-distancing policies and mobility restrictions
become truly binding, and lead to significant unemployment and income losses. Combining these
two metrics, we find that migrants, in particular those from NMS13 countries, are more likely to
work in occupations exposed to income risks (Figure 4, Panel A). On average, jobs defined as
exposed to income risks (both non-essential and non-teleworkable) account for 38 percent of all
employment in the EU. The share of migrants employed in essential jobs is higher, and the share
employed in occupations that allow for teleworking is lower. Thus, on net, migrants are 5 percent
more exposed than natives to the income risks that arose during the pandemic. Overall, 40 percent
of migrants fit into this income risk category. By region of origin, migrants from the NMS13
countries are particularly vulnerable to these risks, with 44 percent of them working in such
occupations.
Migrants and natives exposed to the same economic risks have similar demographic patterns, with
younger groups significantly more likely to work in jobs that are both non-essential and non-
teleworkable (Figure 4, Panel B). By education level, both native and migrant workers who have
secondary education or less schooling are twice as likely as their peers who completed a university
degree to have jobs exposed to income risks (Figure 4, Panel C). Overall, migrants in the poorest-
earning deciles are less exposed to income risks than low-income natives (Figure 4 Panel D). This
is due to the fact that low-income migrant workers are more likely to be in essential jobs. However,
migrants in middle-income groups are the most vulnerable to income losses as they tend to work
more in jobs that are non-essential and non-teleworkable compared to natives with similar earning
levels.


Measure 4: Occupational Exposure to Health Risks
We measure health risks by the extent of face-to-face interactions required by an occupation by
using a methodology similar that used in Avdiu and Nayyar (2020). Migrants tend to work in
occupations that are more exposed to COVID-19-related health risks (Figure 5, Panel A). While
only 15 percent of native workers in the EU28 have jobs in sectors that have high health risks, 23
percent of migrants do so. Exposure to COVID-19-related health risks again varies across

                                                12
immigrant groups. By region of origin, migrants from NMS13 and non-EU28 countries show the
greatest health vulnerability (23 percent and 24 percent), while migrants from other EU15
countries and natives have similar exposure levels (16 percent). Migrant workers (who account for
13.5 percent of the working population in the EU) are overrepresented in essential occupations
that have above-average health risks, accounting for more than 20 percent of medical doctors,
nearly 25 percent of personal care workers in health services, and almost 40 percent of cleaners
and helpers (Figure 5, Panel B). This picture contrasts with the lower prevalence of migrants in
essential sectors that do not require frequent in-person interactions, and/or do not offer the prospect
of telework; some 15 percent of migrant workers and 22 percent of native-born workers are
employed in these types of jobs.
Older and less- educated workers are more likely to work in jobs exposed to greater health risks;
this is particularly true for immigrants (Figure 6, panels A and B). Low-income workers tend to
work in occupations that are more exposed to COVID-19 health risks. This is especially true for
low-income migrants from developing countries (Figure 6, Panel C).


Measure 5: Occupational Exposure to Demand Shocks
The previous measures examined labor supply-side risks (arising from government restrictions, an
inability to work from home, and greater face-to-face interactions). But immigrant workers also
face labor demand-side related risks that stem from their disproportionate share of employment in
certain occupations. During the pandemic these risks arose due to rapidly declining consumer
demand. The sharp decline of employment in travel-related sectors is a prominent example.
Our data reveal a positive correlation between the supply and demand shocks; that is, sectors that
have higher labor-supply disruptions are also likely to have a larger drop in demand (Figure 7,
Panel A). Given the different greater presence of migrant workers in vulnerable sectors, negative
demand shocks associated with COVID-19 entailed larger labor disruptions for migrant than for
natives. While natives work in sectors that have an estimated demand reduction of 11 percent, the
estimated average fall in consumption in the sectors in which migrants work appears to have been
much greater. Our figures show a decline of about 15.5 percent in those sectors which represents
a 29 percent larger reduction than in sectors in which native-born workers concentrate (Figure 7,
Panel B). Demand shocks in the short and medium terms might thus have posed a higher threat to
the sustainability of migrants’ jobs than those of native workers. The effects are also different for
migrants from different regions of origin. Comparisons of demand for the average sectoral
occupations of native-born workers with demand for the average sectoral occupations of migrant
workers shows that the fall in demand was 2.3 percentage points greater for EU15 migrants, 5
percentage point greater for NMS13 migrants, and 5.5 percentage points greater for NMS13
migrants.
For both native and migrant workers, younger cohorts and those with lower levels of education
suffered higher demand shocks in their sectors of employment (Figure 8). Across income
categories, workers from the lowest earning deciles faced a larger drop in sectoral consumption
and thus likely faced more pressure to retain jobs. The demand shock had a similar impact on high-

                                                  13
income natives and migrants. However, the migrant-native gap is significantly larger for waged-
workers in the medium-income deciles. While the demand shock increases in an almost linear
fashion the lower the income decile of workers, the largest negative impact among migrants
appears to have been among the lower-middle-income deciles. Migrants in the poorest deciles were
slightly more shielded from a drop in demand because certain low-skilled occupations like
personal care, or cleaning experienced demand reductions to a lesser degree.



5. Methodology

The data and figures in the previous section present the patterns of exposure to various types of
risks for native and immigrant workers. Since the risk measures are constructed at the occupational
level, the differences between the natives and the immigrant workers are mainly due to their
distributions across occupations. In addition, immigrants differ from native workers in terms of
their age and education levels, which further impact workers’ risk exposure as the data indicate.
This section aims to see if the correlations between risk exposure and immigration status (native-
born vs. immigrant) hold once we control for differences in observable personal characteristics,
such as age and education.
To examine these issues, we first estimate a multinomial logit model at the worker level, where
the dependent variable is a categorical variable for the labor-market status. We construct three
possible outcomes for working-age adults in the sample for this variable:12
         (a) Employed in “safe jobs:” These are workers with jobs in occupations that are classified
             as safe from health or income risks. That is, these are the occupations that are income
             safe (they can be performed from home, or they are deemed essential for the economy),
             and they do not require extensive face-to-face interactions. An example would be a
             computer programmer who can work from home.
         (b) Employed in occupations with health risks: These are workers with jobs in essential
             sectors that cannot telework. They require above-average numbers of face-to-face
             interactions. An example would be a surgeon or a bus driver.
         (c) Employed in occupations with income risks: These are workers in occupations that are
             exposed to supply shocks because they are not considered essential and because they
             are not able to telework; thus, they are more at risk of being dismissed. An example
             would be the members of the kitchen staff at a restaurant.




12
   For robustness checks, we also run another specification that includes a fourth category indicating if a worker is not
employed (in addition to the three employment categories of safe jobs, jobs with health risks, and jobs with income
risks). To minimize the number of adults that are not employed due to continuing education, the sample is restricted
to adults ages 25 to 64. The main results are qualitatively similar to the results we obtain when using the three
categories. Results are available upon request.

                                                           14
These outcomes are mutually exclusive and capture all possible occupational categories. Figure 9
displays several Venn diagrams of the different categorization of jobs for workers from diverse
regions of origin. Each circle represents whether a job is non-essential, non-teleworkable, or
requires intense face-to-face interactions. (Jobs outside all three circles are thus essential,
teleworkable, or does not require intensive face-to-face interactions). Jobs with a high degree of
income risk because they are both non-essential and non-teleworkable are shown in red. Panel A
shows that 37.6 percent of jobs held by native workers fall into that category; this is a lower share
than the 39.6 percent of such jobs held by migrants (Panel B).
Jobs that entail high health risks because they are essential, non-teleworkable, and require intense
face-to-face interaction are shown in green. Again, we observe that migrants are overrepresented
in this category; 22.5 percent of the jobs held by migrants fall into this category, compared to 15.5
percent of the jobs held by native-born workers.
Finally, safe jobs are all other jobs. These are jobs that are essential, with low numbers of face-to-
face interactions; these jobs can be either teleworkable or non-teleworkable. These are shown in
the gray areas in the figures, which show that 46.9 percent of jobs held by native-born workers fall
into this category, compared to 37.9 percent of the jobs held by migrants.
When we look at regional distribution of the migrants, those from western Europe (EU15) have
almost the same occupational distribution as native workers (Panel C) across the three categories.
By contrast, migrants from new member states of the EU (Panel D) and outside the EU (Panel E)
have higher shares in jobs exposed to health risks (green area) and income risks (red area).
With the labor-market status defined as above, we estimate the following equation:
Y iok = α + Ωo + β Xi + λ Li + γ Πi + Δ1FiNi + Δ2 (1-Fi)*Ni + φk + ε iok                                   (1)
where Y iok is the categorical variable for one of the three occupational statuses of individual i
currently living in region of residence k in the EU13 and originally from origin o. Ωo is a vector of
fixed effects for the world regions of origin o for the worker i. The regions of origin are defined
as those countries that we put in the following categories: the EU15; the NMS13; the European
Free Trade Association (EFTA); other parts of Europe; Asia; Latin America; Sub-Saharan Africa;
the Middle East and Northern Africa (MENA); and non-European OECD countries.14 Xi is a vector
of individual-level observable socioeconomic characteristics (age, age-squared, gender, and years
of education). Li is a vector of dummy variables that capture the different levels of fluency of an
individual in the language of the host country. L1 indicates a beginner level of command of the
new host-country language; L2 indicates an intermediate level; L3 indicates an advanced level. The
baseline indicates having language skills only in one’s native language, not in the language used
in the host country. Although questions concerning language ability in the EU-LFS 2014 ad hoc

13
  Region is defined at the second level of the EU Nomenclature of Territorial Units for Statistics (NUTS2)
14
  EU15 is composed of Western EU countries. NMS13 are the Eastern European countries that joined the EU since
2004. EFTA countries include Iceland, Lichtenstein, Norway, and Switzerland; these are countries that are also part
of the Schengen Area allowing free mobility within the EU. Other European countries are those that are not in the EU
or in the EFTA; these are mostly countries in the Balkans and eastern Europe. Non-European OECD countries are
Australia, Canada, New Zealand, and the United States. Please see earlier description above.

                                                        15
migration module are posed only to migrants, our model assumes that native-born workers have
perfect language fluency. Πi is a dummy variable that takes the value of one if immigrant
respondents report that certifying their foreign education certificates was a major barrier in the
host country’s labor market. Ni is a dummy variable that takes value of one if a worker found
his/her current job through friends or relatives, or zero if the job was found through other channels
(to capture the role of networks). We interact this variable with the dummy variable denoting
whether a person is a migrant or a native to allow for differential impacts of networks on the
outcomes of interest for migrants and natives. φk is a vector of dummy variables for each NUTS2
region of residence K in the EU.



6. Determinants of Exposure to COVID-19 Related Risks

6.1. Demographic Characteristics of Migrants and Their Exposure to Income and Health Risks
Due to the Pandemic


Tables 3 through 5 present the results for the multinomial logit estimation of Equation 1. Table 3
reports the results for the “safe jobs.” Table 4 shows results for the “jobs with health risks.” Table
5 shows results for the “jobs with income risks.” Each reported coefficient in the tables is the
marginal effect (in percentage points) of that explanatory variable on the likelihood of being in the
respective labor market category (safe, health-risk, or income-risk jobs). For example, each
coefficient for a world region (such as EU15, NMS13 or MENA) in Table 3 is the difference in
percentage points in the likelihood of a migrant from that region being in a safe job vis-à-vis native
worker (the baseline group).
Each table has seven columns corresponding to different specifications of the explanatory
variables. Column 1 includes only a dummy variable of being foreign born. Column 2 includes
separate dummy variables for each of the world regions of origin (the baseline being native born).
In Column 3, we add personal characteristics such as age (and age-squared to capture non-
linearities), gender (via a dummy variable that is equal to one if the person is male) and years of
education. Column 4 adds dummy variables for each one of the 159 NUTS2 regions of residence
in the EU. Column 5 includes the three dummies that capture the different levels of language
fluency (compared to the baseline that is having fluency in the local language only). Column 6
includes a dummy variable that captures the certification barriers faced by the migrants. Column
7 adds two variables assessing the impact of job referrals through networks for both migrants and
natives.
As shown in Column 1 of Table 3, immigrants are 10 percent less likely to have safe jobs than
natives. Immigrants tend to be over-represented in jobs exposed to both health and income risks
(Column 1 of tables 4 and 5). Even though this is true for most migrants, there are critical
differences across regions of origin (Column 2 of tables 3-5). Migrants from higher-income
countries (EFTA, EU15, or other OECD countries) have similar labor outcomes. For example,


                                                 16
migrants from EU15 countries are 4 percent more likely than the natives to be in a safe job and,
similarly, those from non-European, high-income OECD countries 21.5 percent more likely to
work in such safe jobs (Table 3, Column 2). Migrants from the rest of the world, by contrast, are
significantly less likely than native-born workers to be in safe jobs. The risks for these migrants
are 18.9 percent greater for those from NMS13 countries, 20.7 percent for those from the rest of
Europe, and 8.6 percent greater for those from MENA. Naturally the reverse is true for jobs with
income or health risks. Migrants from Sub-Saharan Africa and Latin America are highly likely to
be found in jobs with health risks (Table 5) while migrants from NMS13 and non-EU Eastern
European countries have the largest share of jobs with income risks.
Controlling for individual characteristics such as age, gender or education levels of individuals
and their country of residence does not significantly change these results. The coefficients in
columns 3 and 4 of tables 3 to 5 are not much different from those in Column 2. For example, the
probability of migrants from NMS13 countries to be in income-risk jobs relative to natives goes
declines by only 2.9 percentage points (5.5 percent compared to 8.4 percent). The likelihood that
Sub-Saharan African or Latin American migrants are in jobs with high health risks essentially
stays the same. Thus, the large gaps in income and health risks faced by migrants from lower-
income regions of the world (relative to the native-born workers) are mostly due to factors other
than personal characteristics, such as age and education.


6.2. Language Barriers and Recognition of Foreign Credentials


The imperfect transferability of skills and human capital from home to the host country is among
the most-cited factors behind the disparities observed between migrants and native-born residents
in terms of their labor-market outcomes. Language barriers and difficulties in having credentials
recognized are two main obstacles to skill transferability; these are among the possible reasons
why migrants are sorted into lower-wage and less-appealing jobs at the destination.
The literature on the importance of linguistic skills has identified several key patterns. Different
levels in local language skills lead to different sectoral and occupational allocations even if
migrants have similar education levels. Low-skilled workers tend to have lower language fluency,
which tends to limit them to a greater degree to jobs that involve manual labor, rather than
interactive tasks, as Peri and Sparber (2009) show. As a result, migrants concentrate in manual
labor, while natives have a comparative advantage in finding occupations that involve more
complex communication tasks (D’Amuri and Peri, 2014; Foged and Peri, 2016). Similarly, Peri
and Sparber (2011) find occupational differences between migrants and native-born workers even
if they are highly educated. Among workers with tertiary education, immigrants are concentrated
in occupations that involve more quantitative and analytical tasks; by contrast, native-born workers
tend to be in jobs that require more intense communication skills.
Thus, we want to examine whether language ability shapes the sorting of migrants into occupations
with different skill requirements in the EU – and, as a result, the exposure to the different risks
associated with the COVID-19 pandemic. We begin by using the O*NET survey to identify the

                                                17
intensity of the host-country language requirement per occupation. In particular, the knowledge
questionnaire includes the question “What level of English language knowledge is needed to
perform your current job?” with answers ranging from one to seven. We transform this measure
available for each occupation in the US to ISCO three-digit occupations and create a standardized
index of intensity of the host-country language requirement with mean of zero and standard
deviation of one.
Our analysis shows that immigrant workers in Western European countries are disproportionately
concentrated in occupations that require lower levels of host-country language fluency (Figure 10,
Panel A). Across the EU labor markets, immigrants work in occupations in which language skill
requirements are 0.2 standard deviations below the mean (Figure 10, Panel B). This difference is
statistically significant at the 1 percent level. However, the difference in occupational choices
between migrants and natives varies with migrants’ fluency in the host-country language. Migrants
with advanced or above-average levels of fluency have occupations with similar levels of language
requirements to those required for native workers (just 0.03 standard deviations lower than those
of the native-born workers), while migrants with beginner or intermediate levels of language
ability are employed in jobs in which the language requirements are, on average, 0.64 standard
deviations lower. OLS regressions show that, even after controlling for other individual
characteristics (age, age squared, gender, years of education, as well as fixed effects for both
regions-of-origin, and countries-of-destination), language fluency is significantly positively
associated with the language intensity required in the occupation of workers.
Occupations with higher language requirements are associated with lower levels of job
vulnerability. In particular, occupations that require a higher command of the native language, as
well as higher communication skills, are significantly more likely to be amenable to telework,
which shields them from pandemic-related income risks (Figure 11, Panel A). Such occupations
typically require higher levels of education, but this finding holds even when controlling for
differences in workers’ average education by occupation. The correlation between language ability
required per occupation and the exposure of occupations to health risks is less clear (Figure 10,
Panel B).
Language fluency is a strong predictor of the type of jobs in which migrant workers are employed
and the associated level of vulnerability that surfaces in the multinomial logit regression (Column
5, tables 3-5). Compared to workers for whom the language used in their place of residence is also
their mother tongue, those with beginner or intermediate levels of language fluency are about 10
percentage points more likely to be employed in a safe job (Column 5, Table 3). This effect is very
large, given that less than half of the employed population has jobs with neither income nor health
risks. Moreover, language ability explains a significant gap in the share of “safe jobs” for several
migrant groups. For example, the gap between migrants from NMS13 countries and natives’
likelihood of having held a “safe” job during the pandemic falls from 16 percentage points to 9
percentage points when language skills are taken into account; the gap falls from 19 percentage
points to 13 percentage points for migrants from other European countries, from 13 percentage
points to 7 percentage points for Asians, from 10 percentage points to 6 percentage points for those


                                                18
from Sub-Saharan Africa, and from 9 percentage points to 3 percentage points for those from
MENA countries.
Barriers to recognition of foreign credentials are also significantly correlated with the exposure to
the labor-market risks of the pandemic. Migrants who report a lack of recognition of their
credentials as a main barrier to integrate in the host-country labor market are 16 percentage points
less likely to be employed in “safe jobs” compared to similar migrants who did not face difficulties
validating their degrees (Column 6, Table 3). Although skill validation is an important factor, it
does not explain a large part of the overall higher vulnerability of migrants in the labor market; as
captured in the EU-LFS, it represents a main barrier for only 6 percent of working-age migrants.
Moreover, when we compare the coefficients for different world regions (tables 3-5, columns 5-
6), we do not see much of a change either. This indicates that barriers for certifying foreign
credentials, although relevant for some migrants, are not one of the main factors that would explain
the higher economic risk exposure of migrants from lower-income countries to the EU15.


6.3. Immigrant Network Effects


Even after we control for personal characteristics such as education level and fluency in the local
language, there are still unexplained differences between migrants from different regions of the
world, in terms of their occupations and the resulting income and health risks to which they are
exposed. Results (tables 3-5, columns 5-6) show that migrants from NMS13 countries are still 9.5
percentage points less likely than native workers to be in safe jobs. The gap is 13.5 percent for
migrants from other eastern European countries, 6.0 percent for those from Sub-Saharan Africa,
6.6 percent for those from Asia, and 10.5 percent for those from Latin America (Table 3, Column
5).
Another pattern found in the literature for the clustering of migrants into specific occupations is
the role of migrant networks. Existing social and cultural networks provide numerous services,
especially to newly arriving migrants. In addition to helping them with housing, schools and
deciphering the cultural norms of the destination country, migrant networks help the new migrants
find jobs. These jobs tend to be in sectors in which previous cohorts of migrants are employed.
This process leads migrants from the same communities to cluster in specific occupations and
sectors. Beaman (2012) shows immigrant enclaves and networks in different localities of residence
influence the occupational choices of different generations of immigrants. Similarly, Patel and
Vella (2013) find that immigrants from specific countries cluster in particular occupations across
different regions in the United States. This fact is not due to immigrants’ skills or other
characteristics but rather to referrals from previous cohorts of immigrants and the informational
advantages they offer. While such networks help immigrants integrate in the host labor market and
provide invaluable help in finding a job, they can at the same time restrict the occupational and
spatial mobility of their members (Munshi and Rosenzweig, 2006). Therefore, the historical
clustering of past waves of migrants in specific occupations and regions can have long-lasting
repercussions through path dependence.


                                                 19
Although patterns of occupational clustering have been examined in different regional contexts,
the role of immigrant networks in Europe has received less attention. We use two different
migration waves to assess the path dependence in occupational choices of migrants in the EU.
Figure 12 presents the relationship between the percentage of migrants in an occupation in 2005
(x-axis) and the percentage of recently arrived migrants in those occupations in 2018. To show the
main trend line, we plot the fitted line of a kernel-weighted local polynomial regression as well as
the confidence interval.15 There is a strong and positive correlation between the relative share of
migrant employment in an occupation and destination region in 2005 (representing an older wave
of migration) and the share of recently arrived migrants in 2018 in each occupation and region.16
These results remain robust when differentiating migrants by region of origin and controlling for
occupation, region of origin, and region-of-destination (NUTS2 level) fixed effects, or when
excluding occupations in which there were no migrants in 2005 (see Appendix 1).
Network effects as drivers of occupational choices can, in turn, have an impact on the vulnerability
of migrants to pandemic-related shocks. Past studies analyze the role of networks using different
measures, but in general, either using the share of co-ethnic migrants in the same region of
residence (as in Beaman, 2012), or the share of co-ethnic migrants in the region of residence
working in the same occupation as the migrant (Patel and Vella, 2012). To assess the role of
networks, we cannot use these measures in the multinomial logit from the previous section given
that the dependent variable is based on occupations; thus, any explanatory factor that is also based
on ISCO occupations would create perfect multicollinearity, while region-level measures would
be captured by the fixed effects.
We proceed by using a measure of networks at the individual level. The EU-LFS ad hoc migration
module of 2014 includes a question on the method the individual used to find a current job. Options
include advertisements; public or private employment agencies; training institutions; direct contact
with employers; and contacts through relatives, friends, or acquaintances. As equation (1) shows,
Ni is a dummy variable that takes the value one if a worker used networks of relatives, friends, or
acquaintances, and zero if s/he used any other method. The raw averages show that migrants are
significantly more likely than native workers to use networks to find a job; on average, 40 percent
of migrants used this network, compared to 30 percent of native-born workers. To see how this
fact translates into occupational risk exposure, we use the multinomial logit model with the three
types of occupations (safe, at health-risk and at income-risk) displayed in equation (1). We add
two variables based on the networks for each of the migration categories (migrants and natives) so
we can assess whether the impact of networks on job safety is different between those two groups.
We show that both natives and migrants who used networks to find a job are about 6 percent less
likely to have a safe job and that they are more likely to have jobs with either higher income risks
or higher health risks (Tables 3-5, Column 7). The use of such networks is mostly associated with
higher health risks among migrant workers; by contrast, among native workers, the use of these
networks increases the probability of having a job with greater income risks. Furthermore, the

15
   The confidence interval expands as the share of migrants increases (x-axis) since there are fewer number of such
occupations.
16
   We restrict the sample of migrants in the 2018 EU-LFS to those that had been in the country for at most 10 years,
so the newer cohort of migrants do not overlap with the older one in 2005.

                                                         20
average migrant-native gap in the probability of having a safe job is reduced in half for migrants
from NMS13 and Latin America, and becomes insignificant for migrants from Sub-Saharan
Africa, Asia or MENA. Therefore, while networks (in general of co-nationals) are a key source for
finding employment for many migrants, the use of such networks can translate into an occupational
concentration in certain jobs; the jobs in which migrants who used these networks tended to
concentrate proved to be more vulnerable to both health and income risks related to the COVID-
19 pandemic.



7. Migrants’ Responsiveness to Labor Demand

Prior studies show that the effects of local labor-demand conditions on immigrants’ labor market
outcomes are stronger than those for natives (Azlor et al., 2020). Importantly, immigrants are more
responsive to changes in local labor-demand conditions, choosing locations with more available
jobs and filling labor shortages more rapidly (Cadena and Kovak, 2016). The responsiveness of
migrants to labor demand occurs not only geographically, but also occupationally, where migrants
are found to be significantly more mobile than natives in changing jobs (Green, 1999).
Occupations with labor shortages or openings that attract migrants are typically also characterized
by poorer and riskier working conditions (Orrenius and Zavodny, 2009), more manual work (Peri
and Sparber, 2009), and lower skill requirements (Dustmann and Frattini, 2013). All of these
factors make those jobs less attractive to natives who may prefer to stay unemployed, depending
on the welfare system (Constant, 2014). These more precarious occupations could, in turn, be more
exposed to the negative shocks that surfaced during the pandemic.
To assess the role of labor demand by occupation, we follow a recent study by the European
Commission (2015); we proxy for labor demand by using the share of new hires in total
employment in each ISCO-08 occupation.17 The EU-LFS provides information on whether a
person has been employed in the current job for less than a year. We therefore define new hires as
those who obtained their job within the previous 12 months.
The evidence for western Europe shows a strong and positive correlation between the share of new
hires in an occupation (which we refer to as “turnover”) and the share of foreign workers (Figure
13, Panel A). That is, migrants tend to work in occupations with a more dynamic structure and
with turnover that might be unmet by natives’ supply of labor, resulting in labor shortages. The
elasticity of the share of migrants to new employment is about 0.1, meaning that a 1 percent
increase in the share of new hires is associated with a 0.1 percent increase in the share of migrants
in a given occupation of a NUTS2-level division of destination (see Appendix 2). The correlation
between turnover and the prevalence of immigrant workers per occupation is stronger for non-EU


17
   We acknowledge that this measure captures not only changes in demand in an occupation (because increases in
demand are associated with larger share of new hires), but also other characteristics such as differences in turnover
and job tenure. However, better-suited proxies, such as job vacancies in the EU, are not available at finer levels of
disaggregation of occupations and sectors.

                                                         21
migrants and those coming from NMS13 countries, while migrants from high-income EU15
countries are not likely to be in high-turnover occupations (Figure 13, panels b, c and d).
As a final step, we show that the higher responsiveness of migrants to labor-demand conditions
and job openings results in higher exposure of this group to income shocks derived from supply
and demand changes that surfaced during the pandemic. Figure 14 shows that ISCO-08 three-digit
occupations in the EU with a higher share of new hires in 2018 also have higher ex ante exposure
to income risks. In conclusion, there is a strong, positive relationship between job openings, the
share of immigrants (especially from NMS13 countries and non-EU countries) and high-income
risks at the occupational level in high-income EU countries. Thus, negative demand shocks that
reduced jobs in certain sectors during the pandemic are likely to have had more impact in
occupations in which immigrants are overrepresented.



8. Conclusion

This paper provides evidence that migrant workers in the EU were more exposed to both economic
and health shocks associated with the COVID-19 pandemic. The source of this greater risk
exposure of migrant workers to COVID-19 stems from the systematic differences in the
occupations in which they are concentrated to a greater degree than native-born workers. The
additional risks stem from three factors. First, migrants work to a greater degree in occupations
that cannot be undertaken at home. Second, migrants work to a greater degree in occupations that
were more severely hit by COVID-19-related demand shocks. Both of these exposed migrants to
income risks to a greater degree than natives. Third, migrants who are employed in essential jobs
are working in positions that require face-to-face interactions; thus, during the pandemic, they have
been exposed to greater risks to their health, in addition to the risks to their economic security.
Our findings reveal notable differences in risk faced by migrant groups coming from different
regions of the world. While immigrants from other EU countries were not more exposed to
pandemic-related risks than native-born workers, immigrant workers from other regions of the
world faced higher economic and health risks associated with COVID-19.
We find that differences in personal characteristics, such as age and education levels, explain only
a small fraction of the risk exposure gap between native and immigrant workers. By contrast, we
show that several other factors lead to the overrepresentation of migrant workers in occupations
that proved to be high-risk jobs during the pandemic. One factor is the imperfect transferability of
human capital across countries. A limited command of the host-country language leads migrants
to concentrate occupations that require lower skill and have higher levels of COVID-19 exposure
than is the case for native-born workers with similar levels of education. Second, migrant workers
also tend to work in occupations in which there are larger labor shortages; by contrast, native
workers are reluctant to take up these jobs. Third, the occupational sorting of migrant workers is
reinforced by the effects of migrant networks between origin and destination countries.



                                                 22
In addition, migrant workers often have weaker social and employment protection in destination
countries than those available to native-born citizens.
Our findings call for policy interventions targeted at migrant workers to mitigate the adverse
effects of the pandemic experience on this particularly vulnerable population. In particular,
programs created to respond to the COVID-19 outbreak could be made more migrant-sensitive to
take into account the unique challenges faced by migrant workers (Moroz et al., 2020). For
example, emergency social safety net programs created to respond to the pandemic could include
migrant workers and their families; other non-pandemic-specific measures could expand migrants’
access to existing welfare programs; new schemes could be devised that specifically target foreign
workers. Needed measures go beyond such safety-net programs to provide migrants with access
to health services, which are particularly urgent given migrant workers’ greater exposure to
COVID-19 itself and related health risks. In addition, job losses for migrant workers can in some
cases threaten their legal status in destination countries; such a situation raises the need for
interventions – even if temporary – to relax visa or residency-status restrictions to protect migrant
workers.
In the longer term, migrants’ integration in the labor market can also be measured, not just by
wages and employment rates, but also by gaps in the types of employment (e.g., the share of
teleworkable jobs). Such gaps can predict vulnerability to certain employment shocks, as
underlined in the pandemic. To reduce those gaps, interventions should address the main drivers
of occupational disparities, in particular the limited language fluency and networks that migrants
have in the host country.




                                                 23
References


Adams, Richard H., and John Page. 2005. “Do International Migration and Remittances Reduce
Poverty in Developing Countries?” World Development, 33 (10): 1645–69.
Adams-Prassel, Abi, Teodora. Boneva, Marta Golin and Christopher Rauh. 2020a. “Inequality in
the Impact of the Coronavirus Shock: Evidence from Real Time Surveys.” Journal of Public
Economics, 189: 104245.
———. 2020b. “Work Tasks That Can Be Done from Home: Evidence of the Variation Within
and Across Occupations and Industries.” Cambridge Working Papers in Economics 2040, Faculty
of Economics, University of Cambridge.
Ahonen, Emily Q., and Fernando G. Benavides. 2006. “Risk of Fatal and Non-fatal Occupational
Injury in Foreign Workers in Spain.” Journal of Epidemiology and Community Health, 60: 424-
426.
Albanesi, Stefania, and Jiyeon Kim. 2021. “The gendered impact of the COVID-19 recession on
the US labor market.” No. w28505. National Bureau of Economic Research.
Aleksynska, Mariya and Ahmed Tritah. 2013. “Occupation–education mismatch of immigrant
workers in Europe: Context and policies,” Economics of Education Review, 36: 229-244.
Alon, Titan, Matthias Doepke, Jane Olmstead-Rumsey and Michele Tertilt. 2020. “The impact of
covid-19 on gender equality,” National Bureau of Economic Research, Working Paper 26947
doi:10.3386/w26947.
Amuedo-Dorantes, Catalina, and Susan Pozo. 2010. “Accounting for Remittance and Migration
Effects on Children’s Schooling.” World Development, 38 (12): 1747–59.
Autor, David H., and Michael J. Handel. 2013. “Putting Tasks to the Test: Human Capital, Job
Tasks, and Wages,” Journal of Labor Economics, 31 (S1): 59-96.
Autor, David H., Frank Levy and Richard J. Murnane. 2003. “The Skill Content of Recent
Technological Change: An Empirical Exploration,” Quarterly Journal of Economics, 118(4):
1279-1334.
Avato, Johanna, Johannes Koettl and Rachel Sabates-Wheeler. 2010. “Social Security Regimes,
Global Estimates, and Good Practices: The Status of Social Protection for International Migrants, ”
World Development, 38 (4): 455-466.
Avdiu, Besart, and Gaurav Nayyar. 2020. “When face-to-face interactions become an occupational
hazard: Jobs in the time of COVID-19,” World Bank Policy Research Working Paper 9240.
Azlor, Luz, Anna Piil Damm and Marie Louise Schultz-Nielsenc. 2020. “Local labour demand
and immigrant employment,” Labour Economics, 63, 101808.



                                                24
Baker, Scott R., R.A. Farrokhnia, Steffen Meyer, Michaela Pagel, and Constantine Yannelis. 2020.
“How does household spending respond to an epidemic? consumption during the covid-19
pandemic,” National Bureau of Economic Research Working Paper 26949.
Barrett, Alan, Seamus McGuiness, Martin O'Brien and Philip O'Connell. 2013. “Immigrants and
Employer‐Provided Training,” Journal of Labor Research, 34 (1): 52–78.
Bartik, Alexander W., Marianne Bertrand, Feng Lin, Jesse Rothstein, and Matt Unrath.
2020. “Measuring the labor market at the onset of the COVID-19 crisis.” National Bureau of
Economic Research, Working Paper 27613.
Bayer, Patrick, Stephen L. Ross, and Giorgio Topa. 2008. “Place of Work and Place of Residence:
Informal Hiring Networks and Labor Market Outcomes,” Journal of Political Economy, 116 (6):
1150-1196.
Beaman, Lori A. 2012. “Social Networks and the Dynamics of Labor Market Outcomes: Evidence
from Refugees Resettled in the U.S.” Review of Economic Studies, 79 (1): 128-161.
Betcherman, Gordon, Nicholas Giannakopoulos, Ioannis Laliotis, Ioanna Pantelaiou, Mauro
Testaverde, and Giannis Tzimas. 2020. “Reacting quickly and protecting jobs: The short-term
impacts of the COVID-19 lockdown on the Greek labor market.” IZA Discussion Paper No. 13516.
Bonin, Holger, Amelie Constant, Konstantinos Tatsiramos, and Klaus F. Zimmermann. 2006.
“Ethnic Persistence, Assimilation and Risk Proclivity.” IZA Discussion Paper No. 2537.
Bonin, Holger, Amelie Constant, Konstantinos Tatsiramos, and Klaus F. Zimmermann. 2009.
“Native-Migrant Differences in Risk Attitudes.” Applied Economics Letters, 16: 1581-1586.
Borjas, George J. 1985. “Assimilation, Changes in Cohort Quality, and the Earnings of
Immigrants,” Journal of Labor Economics, 3 (October): 463–89.
______ 2015. “The Slowdown in the Economic Assimilation of Immigrants: Aging and Cohort
Effects Revisited Again,” Journal of Human Capital, 9 (4): 483-517.
Borjas, George J. and Hugh Cassidy. 2020. “The Adverse Effect of the COVID-19 Labor Market
Shock on Immigrant Employment,” National Bureau of Economic Research Working Paper
27243.
Brinca, Pedro, Joao B. Duarte, and Miguel Faria-e-Castro. 2020. “Measuring Sectoral Supply and
Demand Shocks during COVID-19,” Working Paper 2020-011B, Federal Reserve Bank of St.
Louis.
Cadena, Brian C. and Brian K. Kovak. 2016. “Immigrants Equilibrate Local Labor Markets:
Evidence from the Great Recession,” American Economic Journal: Applied Economics, 8(1): 257–
290.
Carlsson, Magnus, and Dan-Olof Rooth. 2007. “Evidence of ethnic discrimination in the Swedish
labor market using experimental data,” Labour Economics, 14 (4): 716-729.


                                              25
Cattaneo, Cristina, Carlo V. Fiorio and Giovanni Peri. 2015. “What Happens to the Careers of
European Workers When Immigrants “Take Their Jobs”?,” Journal of Human Resources, 50 (3)
655-693.
Chiswick, Barry R. 1980. “An Analysis of the Economic Progress and Impact of Immigrants,”
National Technical Information Service, no. PB80-200454. Washington, DC: U.S. Dept. Labor.
Chiswick, Barry R., Yew Liang Lee and Paul W. Miller. 2005. “A Longitudinal Analysts of
Immigrant Occupational Mobility: A Test of the Immigrant Assimilation Hypothesis ,”
International Migration Review, 39(2), 332–353.
Chiswick, Barry R. and Paul W. Miller. 2009. “The international transferability of immigrants’
human capital skills,” Economics of Education Review, 28(2), 162-169.
Congressional Budget Office. 2006. “Potential influenza pandemic: Possible macroeconomic
effects and policy issues”. https://www.cbo.gov/sites/default/files/109th-congress-2005-
2006/reports/12-08-birdflu.pdf
Constant, Amelie F. 2014. “Do migrants take the jobs of native workers?” IZA World of Labor: 10
doi: 10.15185/izawol.10.
Corvalan, Carlos F., Timothy R. Driscoll and James E. Harrison. 1994. “Role of migrant factors in
work-related fatalities in Australia.” Scandinavian Journal of Work, Environment & Health 20(5):
364-70.
Cortes, Matias and Eliza Forsythe. 2020. “Impacts of the covid-19 pandemic and the cares act on
earnings and inequality”, IZA Discussion Paper No. 13643.
Damm, Anna Piil. 2009. “Ethnic enclaves and immigrant labor market outcomes: quasi-
Experimental evidence,” Journal of Labor Economics, 27 (2): 81-314.
______ 2014. “Neighbourhood quality and labour market outcomes: evidence from quasi-random
neighbourhood assignment of immigrants,” Journal of Urban Economics, 79: 139-166.
D'Amuri, Francesco and Giovanni Peri. 2014. “Immigration, Jobs, and Employment Protection:
Evidence from Before and During the Great Recession,” Journal of the European Economic
Association, 12 (2): 432-464.
Dávila, Alberto, Marie T. Mora, and Rebecca González. 2011. “English-Language Proficiency and
Occupational Risk Among Hispanic Immigrant Men in the United States.” Industrial Relations,
50 (April): 263-296.
Diaz, Esperanza, Svenn-Eirik Mamelund, Jarle Eid, Henriette S. Aasen, Oddvar Martin Kaarbøe,
Rebecca J. C. Brokstad, Siri Gloppen, Anders Beyer, and Bernadette N. Kumar. 2021. “Learning
from the COVID-19 pandemic among migrants: An innovative, system-level, interdisciplinary
approach is needed to improve public health.” Scandinavian Journal of Public Health
14034948211019795.



                                               26
Díaz-Serrano, Luis. 2010. “Do Legal Immigrants and Natives Compete in the Labour Market?
Evidence from Catalonia.” IZA Discussion Paper No. 4693.
del Rio-Chanona, R. Maria, Penny Mealy, Anton Pichler, Francois Lafond and Doyne Farmer.
2020. “Supply and demand shocks in the COVID-19 pandemic: An industry and occupation
perspective,” Oxford Review of Economic Policy, Volume 36, Issue Supplement_1, 2020, Pages
S94–S137, https://doi.org/10.1093/oxrep/graa033.
Dingel, Jonathan I, and Brent Neiman. 2020. “How many jobs can be done at home?” Journal of
Public Economics, Volume 189, September 2020, 104235.
Duleep, Harriet O., and Daniel J. Dowhan. 2002. “Insights from Longitudinal Data on the Earnings
Growth of U.S. Foreign-Born Men.” Demography 39 (August): 485–506.
Duleep, Harriet O., and Mark C. Regets. 1999. “Immigrants and human-capital investment,”
American Economic Review, 89(2): 186–191.
Dustmann, Christian, Albert Glitz, A. and Thorsten Vogel. 2010. “Employment, wages, and the
economic cycle: Differences between immigrants and natives.” European Economic Review,
54(1): 1–17.
Dustmann, Christian and Tommaso Frattini. 2013. “Immigration: The European experience.” In
D. Card and S. Raphael (Eds.), Immigration, Poverty, and Socioeconomic Inequality: Russell Sage
Foundation.
Dustmann, Christian, Albrecht Glitz, Uta Schönberg and Herbert Brücker. 2016. “Referral-based
Job Search Networks, Review of Economic Studies, 83 (2): 514–546.
Edin, Per-Anders, Peter Fredriksson and Olof Åslund. 2003. “Ethnic Enclaves and the Economic
Success of Immigrants—Evidence from a Natural Experiment,” The Quarterly Journal of
Economics, 118 (1): 329–357.
European Commission. 2015. “Labor Market Shortages                in the European Union,”
IP/A/EMP/ST/2013-06.
European Commission. 2020. “Communication from the Commission Guidelines concerning the
exercise of the free movement of workers during COVID-19 outbreak,” 2020/C 102 I/03.
https://eur-lex.europa.eu/legal-
content/EN/TXT/PDF/?uri=CELEX:52020XC0330(03)&from=EN
Eurostat. 2021. “Labor Market Database”, Eurostat, European Commission, reference period
2019q2/2020q2, Employment by sex, age and country of birth (1 000) [lfsq_egacob],
https://appsso.eurostat.ec.europa.eu/nui/submitViewTableAction.do (accessed on 02 November
2021).
Fasani, Francesco, and Jacopo Mazza. 2020a. “A Vulnerable Workforce: Migrant Workers in the
COVID-19 Pandemic,” European Commission, JRC - EUR 30225



                                              27
Fasani, F and J Mazza. 2020b., “Being on the Frontline? Immigrant Workers in Europe and the
COVID-19 Pandemic,” IZA Discussion Papers 13963.
Friedberg, Rachel M. 2000. “You can’t take it with you? Immigrant assimilation and the portability
of human capital,” Journal of Labor Economics, 18(2): 221–51.
Foged, Mette and Giovanni Peri. 2016. “Immigrants' Effect on Native Workers: New Analysis on
Longitudinal Data,” American Economic Journal: Applied Economics, 8 (2): 1-34.
Garrote Sanchez, Daniel, Nicolas Gomez Parra, Caglar Ozden, and Bob Rijkers. 2020. “Which
jobs are most vulnerable to COVID-19? Analysis of the European Union,” World Bank Research
& Policy Briefs No. 34. Washington, D.C.: World Bank Group.
Giuntella, Osea, and Fabrizio Mazzonna. 2015. “Do immigrants improve the health of natives?”
Journal of Health Economics 43: 140–53.
Goel, Deepti, and Kevin Lang. 2017. “Social Ties and the Job Search of Recent Immigrants,” ILR
Review.
Green, David A. 1999. “Immigrant Occupational Attainment: Assimilation and Mobility over
Time,” Journal of Labor Economics 17 (1): 49-79.
Hao, Lingxin. 2008. “Workplace Nonfatal Injuries Among Immigrants to the U.S.” Mimeo, Johns
Hopkins University Department of Sociology. Available online at
http://paa2008.princeton.edu/download.aspx?submissionId=80386.
Hersch, Joni, and W. Kip Viscusi. 2010. “Immigrant Status and the Value of Statistical Life.”
Journal of Human Resources 45 (Summer): 749-771.
Hildebrandt, Nicole, and David J. McKenzie. 2005. “The Effects of Migration on Child Health in
Mexico.” Economia 6 (1): 257–89.
Juranek, Steffen, Jörg Paetzold, Hannes Winner, and Floris Zoutman. 2020. “Labor market effects
of COVID-19 in Sweden and its neighbors: Evidence from novel administrative data.” NHH Dept.
of Business and Management Science Discussion Paper 2020/8.
Khamis, Melanie, Daniel Prinz, David Newhouse, Amparo Palacios-Lopez, Utz Pape, Michael
Weber. 2021. “The Early Labor Market Impacts of COVID-19 in Developing Countries: Evidence
from High-Frequency Phone Surveys.” Policy Research Working Paper No. 9510. World Bank,
Washington, DC. © World Bank.
Kramarz, Francis and Oskar Nordström Skans. 2014. “When Strong Ties are Strong: Networks
and Youth Labour Market Entry,” Review of Economic Studies, 81 (3): 1164–1200.
Kugler, Maurice; Mariana Viollaz, Daniel Duque, Isis Gaddis, David Newhouse, Amparo
Palacios-Lopez, Michael Weber. 2021. “How Did the COVID-19 Crisis Affect Different Types of
Workers in the Developing World?” Jobs Working Paper No. 60. World Bank, Washington, DC.
© World Bank.


                                               28
Lubotsky, Darren. 2007. “Chutes or Ladders? A Longitudinal Analysis of Immigrant Earnings,”
Journal of Political Economy, 115:5, 820-867.
Martin, Philip L. 1993. “The Missing Bridge: How Immigrant Networks Keep Americans Out of
Dirty Jobs.” Population and Environment 14 (July): 539-565.
Marvasti, Akbar. 2010. “Occupational Safety and English Language Proficiency.” Journal of
Labor Research 31: 332-347.
Mongey, Simon, Laura Pilossoph, and Alex Weinberg. 2020. “Which workers bear the burden of
social distancing policies?’, National Bureau of Economic Research, Working Paper No 27085.
doi:10.3386/w27085.
Montgomery, James D. 1991. “Social Networks and Labor-Market Outcomes: Toward an
Economic Analysis,” American Economic Review, 81 (5): 1408-1418.
Munshi, Kaivan. 2003. “Networks in the Modern Economy: Mexican Migrants in the U.S. Labor
Market”, Quarterly Journal of Economics, 118: 549–597.
Munshi, Kaivan and Mark Rosenzweig. 2006. “Traditional Institutions Meet the Modern World:
Caste, Gender, and Schooling Choice in a Globalizing Economy,” American Economic Review,
96(4): 1225-1252.
OECD. 2021. “Main Economic Indicators - complete database”, Main Economic Indicators
(database), https://doi.org/10.1787/data-00052-en (accessed on 02 November 2021).
Oreopoulos, Philip. 2011. “Why Do Skilled Immigrants Struggle in the Labor Market? A Field
Experiment with Thirteen Thousand Resumes.” American Economic Journal: Economic Policy
3(4): 148-171.
Orrenius, Pia M, and Madeline Zavodny. 2009. “Do Immigrants Work In Riskier Jobs?”
Demography. 46(3): 535–551.
Orrenius, Pia M. and Madeline Zavodny, M. 2010. “Mexican immigrant employment outcomes
over the business cycle.” American Economic Review 100(2): 316–20.
Orrenius, Pia M., and Madeline Zavodny. 2013. “Immigrants in risky occupations.” In Amelie F.
Constant and Klaus F. Zimmermann (Eds.), International Handbook on the Economics of
Migration, pp. 214–26. Cheltenham, UK: Edward Elgar.
Ottaviano, Gianmarco I. P., Giovanni Peri, and Greg C. Wright. 2013. “Immigration, Offshoring,
and American Jobs.” American Economic Review, 103 (5): 1925-59.
Patel, Krishna, and Francis Vella. 2013. “Immigrant Networks and Their Implications for
Occupational Choice and Wages,” Review of Economics and Statistics, 95 (4): 1249-1277.
Peri, Giovanni, and Chad Sparber. 2009. “Task Specialization, Immigration, and Wages,”
American Economic Journal: Applied Economics, 1(3): 135–169.



                                             29
Peri, Giovanni, and Chad Sparber. 2011. “Highly Educated Immigrants and Native Occupational
Choice,” Industrial Relations: a journal of economy and society, 50 (3): 385-411.
Piracha, Matloob and Florin Vadean. 2013. “Migrant educational mismatch and the labour
market.” In A. F. Constant & K. F. Zimmermann (Eds.), International Handbook on Economics
of Migration (pp. 176-192). Edward Elgar Publishing Ltd.: Cheltenham, UK.
Premji, Stephanie, Patrice Duguay, Karen Messing and Katherine Lippel. 2010. “Are immigrants,
ethnic and linguistic minorities over‐represented in jobs with a high level of compensated risk?
Results from a Montréal, Canada study using census and workers' compensation data.” American
journal of industrial medicine, 53(9): 875-885.
Rojas, Felipe Lozano, Xuan Jiang, Laura Montenovo, Kosali I. Simon, Bruce A. Weinberg, and
Coady Wing. 2020. “Is the cure worse than the problem itself? Immediate labor market effects of
COVID-19 case rates and school closures in the US.” National Bureau of Economic Research
Working Paper No.27127.
Rich, Judith. 2014. “What Do Field Experiments of Discrimination in Markets Tell Us? A Meta
Analysis of Studies Conducted since 2000.” Institute for the Study of Labor (IZA) Discussion
Paper 8584.
Sanromá, Esteve, Raúl Ramos and Hipólito Simón. 2015. “Portability of Human Capital and
Immigrant Overeducation in Spain,” Population Research and Policy Review, 34 (2): 223-241.
Smith, Peter M., Agnieszka A. Kosny, and Cameron A. Mustard. 2009. “Differences in access to
wage replacement benefits for absences due to work‐related injury or illness in Canada.” American
Journal of Industrial Medicine, 52(4), pp.341-349.
Solé, Meritxell, Luis Díaz-Serrano, and Marisol Rodríguez. 2010. “Work, Risk and Health:
Differences between Immigrants and Natives in Spain.” IZA Discussion Paper No. 5338.
World Bank. 2018. “Moving for Prosperity: Global Migration and Labor Markets,” Policy
Research Report. Washington, DC, The World Bank.
Wu, Trong-Neng, Saou-Hsing Liou, Chao-Chun Hsu, Show-Lin Chao, Shu-Fen Liou, Kquei-Nu
Ko, Wen-Yu Yeh, and Po-Ya Chang. 1997. “Epidemiologic Study of Occupational Injuries
Among Foreign and Native Workers in Taiwan.” American Journal of Industrial Medicine 31:
623-630.
Yasenov, Vasil. 2020. “Who Can Work from Home?” IZA Discussion Paper No. 13197.
Zavodny, Madeline. 2015. “Do immigrants work in worse jobs than U.S. natives? Evidence from
California.” Industrial Relations 54(2): 276–93.




                                               30
Figure 1: Share of non-essential jobs in the EU by country of residence, age group, income decile, education
level and nativity

Panel a. By region of birth                                                   Panel b. By age group

                    64%
                    63%
 % of non-essential jobs




                    62%
                    61%
                    60%
                    59%
                    58%
                    57%
                    56%
                    55%
                    54%
                           Natives   Foreign    EU15   NMS13 Non-EU28
                                      born     migrant migrant migrant


Panel c. By education levels                                                  Panel d. By income decile




Source: Own calculation based on EU-LFS 2018 data, following EC directive (2020) and Fasani and Mazza (2020a).




                                                                         31
Figure 2: Share of foreign workers in some essential occupations by main country of residence




Source: Own calculation based on EU-LFS 2018 data.




                                                     32
Figure 3: Share of non-teleworkable jobs in the EU by country of residence, age group, income decile,
education level and nativity

Panel a. By region of birth                                                 Panel b. By age group

                      80%

                      75%
 % of non-telework jobs




                      70%

                      65%

                      60%

                      55%

                      50%
                            Natives   Foreign    EU15     NMS13 Non-EU28
                                       born     migrant   migrant migrant


Panel c. By education level                                                 Panel d. By income decile




Source: Own elaboration based on data from the 2018 EU-LFS and following Dingel and Neiman (2020)
methodology.
Note: 95% confidence intervals are included for comparison.




                                                                     33
Figure 4: Share of jobs exposed to income risks (non-essential and non-teleworkable)

Panel a. By place of birth                                   Panel b. By age group

 50%
 48%
 46%
 44%
 42%
 40%
 38%
 36%
 34%
 32%
 30%
        Natives   Foreign    EU15     NMS13 Non-EU28
                   born     migrant   migrant migrant


Panel c. By education level                                  Panel d. By income decile




Source: Own elaboration based on data from the 2018 EU-LFS and combining Dingel and Neiman (2020), Rio-
Chanona et al (2020) and Avdiu and Nayyar (2020) methodologies.
Note: 95% confidence intervals are included for comparison.




                                                        34
Figure 5: Share of jobs exposed to health risks

Panel a. % of jobs exposed to health risks                           Panel b. % of migrants in essential jobs by health exposure

                        30%


                        25%
% of health-risk jobs




                        20%


                        15%


                        10%


                        5%


                        0%
                              Natives   Foreign    EU15   NMS13 Non-EU28
                                         born     migrant migrant migrant


Source: Own elaboration based on data from the 2018 EU-LFS and combining Dingel and Neiman (2020), Rio-
Chanona et al (2020) and Avdiu and Nayyar (2020) methodologies.
Note: 95% confidence intervals are included for comparison. The exposure index is the combination of the share of
jobs not amenable to telework in each 3-digit occupation and the intensity of face-to-face interactions. The size of the
circles represents the share of each occupation in the total employment in the EU.




                                                                            35
Figure 6: Share of jobs with health risks by age group, education and income levels

Panel a. By age group                                    Panel b. By age group




Panel c. By income decile




Source: Own elaboration based on data from the 2018 EU-LFS and combining Dingel and Neiman (2020), Rio-
Chanona et al (2020) and Avdiu and Nayyar (2020) methodologies.




                                                    36
Figure 7: COVID-19 demand shock: comparison with supply shock and uneven impact by region of birth

Panel a. Supply and demand by occupation              Panel b. Expected demand shock given occupation profile
                 (by ISCO 3-digit)                               (CBO 2006, Rio-Chanona et al 2020)

                                                                                   0
                                                                                  -2




                                                              Demand shock (%)
                                                                                  -4
                                                                                  -6
                                                                                  -8
                                                                                 -10
                                                                                 -12
                                                                                 -14
                                                                                 -16
                                                                                 -18




Source: EU-LFS 2018 database, following estimates of Rio-Chanona et al (2020) and CBO (2006).
Note: The sectoral demand estimates by the US Congressional Budget Office (2006) are based on Hong Kong’s
experience with the past influenza pandemic “SARS” between 2002 and 2004. The CBO provides two scenarios and,
following Rio Chanona et al. (2020), we use the severe scenario which describes a pandemic that is similar in size to
the 1918-1919 Spanish flu outbreak. The CBO lists demand side estimates for broad industry categories, which we
mapped to the NACE sectors.




                                                         37
Figure 8: Demand shock across age groups, education levels, income and nativity

Panel a. By age group                                  Panel b. By education level




Panel c. By income decile




Source: EU-LFS (2018).




                                                  38
Figure 9: Share of each type of occupation by migration status

Panel a. Natives                                                                          Panel b. Migrants



      5.1%                                     11.0%                                             5.1%                           9.8%

                                      Non-telework                                                                   Non-telework


                            21.5%                                                                               21.6%
                                                        15.5%                                                                               22.5%
                                        16.1%                                                                              18.0%
       Non-essential                                                                             Non-essential
               11.3%                                   F2F                                               8.5%                             F2F
                                       13.8%                                                                             10.2%
                                                         5.7%                                                                               4.2%




C. Migrants from Western Europe D. Migrants from Eastern Europe E. Migrants not from Europe


    7.4%                       9.0%                        3.6%                         11.7%                     5.0%                      9.4%

                       Non-telework                                          Non-telework                                           Non-telework


                   17.7%                                                 29.3%                                                  20.3%
                                       15.8%                                                     23.3%                                              24.3%
                            17.5%                                                 14.9%                                                  19.2%
    Non-essential                                          Non-essential                                          Non-essential
           13.2%                      F2F                         6.9%                          F2F                      7.6%                      F2F
                           13.9%                                                 7.4%                                                   10.1%
                                        5.5%                                                      3.0%                                               4.2%




Source: Own elaboration based on data from the 2018 EU-LFS and combining Dingel and Neiman (2020), Rio-
Chanona et al (2020) and Avdiu and Nayyar (2020) methodologies.




                                                                                 39
Figure 10: Language requirements and share of migrants per occupation and average by type of migrants

Panel a. Language requirements and share of migrants Panel b. Language requirement by type of migrant

                                                                                                     0.1




                                                                Standardized language requirements
                                                                                                       0
                                                                                                     -0.1




                                                                       in average occupation
                                                                                                     -0.2
                                                                                                     -0.3
                                                                                                     -0.4
                                                                                                     -0.5
                                                                                                     -0.6
                                                                                                     -0.7
                                                                                                            Natives Migrants Migrants Migrants
                                                                                                                               high      low
                                                                                                                             language language
                                                                                                                             command command


Source: Own elaboration based on the EU-LFS 2018 and the ad-hoc migration module of 2014.
Note: The size of the circles represents the share of each occupation in the total employment in the EU. 95% confidence
intervals are included for comparison. Panel b shows language intensity of in the average occupation in standardized
terms, that is, the original language intensity minus the mean, divided by the standard deviation.




                                                          40
Figure 11: Language requirements per occupation and share of at-risk jobs

Panel a. % of jobs with income risks                           Panel b. % of jobs with health risks




Source: Own elaboration based on the EU-LFS 2018 and the ad-hoc migration module of 2014.
Note: The size of the circles represents the share of each occupation in the total employment in the EU.




                                                         41
Figure 12: Correlation between occupational choices of migrants in the EU in 2005 and 2018




Source: Own elaboration based on the EU-LFS 2005 and EU-LFS 2018.
Note: Share of migrant workers in each occupation for each NUTS2 division of residence. For 2018, migrants are
restricted to those recently arrived in the country (less than ten years) so there is no overlap with the migrants in the
country in 2005.




                                                           42
Figure 13: Correlation between turnover (% of new hires) and the share of migrants in each occupation

Panel a. Total migrants                                       Panel b. EU15 migrants




Panel c. NMS migrants                                         Panel d. Non-EU28 migrants




Source: EU-LFS (2018).
Note: Correlation based on the 136 ISCO-08 3-digit occupations that represent more than 0.1% of the total
employment in the EU. The size of the circles represents the share of each occupation in the total employment in the
EU. In grey, 95% confidence intervals are reported for comparison.




                                                        43
Figure 14: Correlation between turnover (% of new hires) and the share of income-risk jobs in each
occupation




Source: EU-LFS (2018).
Note: Correlation based on the 136 ISCO-08 3-digit occupations that represent more than 0.1% of the total
employment in the EU. The size of the circles represents the share of each occupation in the total EU employment.




                                                       44
        Table 1: Correlations between ex-ante measures of job vulnerability and post-COVID annual employment
        growth rates (2020q2-2019q2) in EU countries, at the occupation-sector-country level
                       (1)                (2)                 (3)                (4)                 (5)               (6)
VARIABLES         Employment         Employment          Employment         Employment          Employment        Employment
                 growth 2020q2-     growth 2020q2-      growth 2020q2-     growth 2020q2-      growth 2020q2-    growth 2020q2-
                     2019q2             2019q2              2019q2             2019q2              2019q2            2019q2

Non-telework        -9.298***
                      (1.322)
Non-essential                          -3.104***
                                        (0.626)
Income-risk                                                -8.120***                                               -5.335***
                                                             (1.159)                                                 (1.036)
Health-risk                                                                      -0.276
                                                                                (0.767)
Demand shock                                                                                      -0.151***        -0.130***
                                                                                                    (0.018)          (0.018)
Constant            -7.089***          -4.758***             -0.425           -3.571***           -1.707***           0.116
                      (0.467)           (0.219)             (0.455)             (0.123)             (0.230)          (0.441)

Observations         1,893                1,893               1,893               1,893              1,893           1,893
R-squared            0.076                0.022               0.055               0.015              0.088           0.104
Time periods     1 (2019-20q2)       1 (2019-20q2)        1 (2019-20q2)      1 (2019-20q2)       1 (2019-20q2)   1 (2019-20q2)
Country FE            YES                  YES                 YES                 YES                YES             YES
        Note: Robust standard errors are reported in parentheses; ***: p<0.01, **: p<0.05, *: p<0.1.




                                                                45
Table 2: Correlations between ex-ante measures of job vulnerability and pre- and post-COVID annual
employment growth rates (2020q2-2019q2-2018q2) in EU countries, at the occupation-sector-country level
                               (1)        (2)        (3)        (4)        (5)        (6)
  VARIABLES                  Annual     Annual     Annual     Annual     Annual     Annual
                           employment employment employment employment employment employment
                             growth     growth     growth     growth     growth     growth

  Non-telework               -3.641***                                                              -3.835***
                               (1.144)                                                                (1.137)
  Non-telework*2020q2        -5.647***                                                               -3.549**
                               (1.378)                                                                (1.339)
  Non-essential                               -0.535
                                             (1.026)
  Non-essential*2020q2                      -2.732**
                                             (1.199)
  Income-risk                                               -1.920
                                                           (1.271)
  Income-risk*2020q2                                     -6.239***
                                                           (1.733)
  Health-risk                                                             -0.173
                                                                         (1.094)
  Health-risk*2020q2                                                      -0.050
                                                                         (1.526)
  Demand shock                                                                           -0.003        0.009
                                                                                        (0.012)       (0.012)
  Demand shock*2020q2                                                                 -0.149***     -0.137***
                                                                                        (0.027)       (0.025)
  2020q2                     -8.098***     -7.050***     -3.639***     -5.995***      -4.234***     -5.671***
                               (0.698)       (0.681)       (0.909)       (0.557)        (0.524)       (0.525)
  Constant                    0.878**       2.054***      3.016***      2.246***       2.315***       0.753*
                               (0.392)       (0.419)       (0.630)       (0.314)        (0.302)       (0.405)

  Observations                  3,863          3,863         3,863          3,863         3,863        3,863
  R-squared                     0.091          0.062         0.078          0.058         0.093        0.114
  Time periods                2 (2018-       2 (2018-       2 (2018-       2 (2018-      2 (2018-    2 (2018-
                               19q2 &         19q2 &        19q2 &         19q2 &        19q2 &       19q2 &
                            2019-20q2) 2019-20q2) 2019-20q2) 2019-20q2) 2019-20q2)                  2019-20q2)
  Country FE                    YES            YES            YES            YES           YES         YES
Note: Robust standard errors are reported in parentheses; ***: p<0.01, **: p<0.05, *: p<0.1.
                                                          1




                                                       46
               Table 3: Probability of having a safe job based on marginal effects of the multinomial logit
                                Model 1:     Model 2:     Model 3:                               Model 5:     Model 6:        Model 7:
                                Raw gaps     Raw gaps     Individual            Model 4:         Language     Barriers to     Networks
VARIABLES                                    by origin    characteristics       NUTS2 FE         ability      certification
Foreign born                    -0.101***
                                (0.004)
Origin: EU15                                 0.040***     0.013                 -0.006           0.034***     0.033**         0.028*
                                             (0.009)      (0.009)               (0.009)          (0.095)      (0.010)         (0.016)
Origin: NMS13                                -0.189***    -0.168***             -0.162***        -0.095***    -0.089***       -0.051**
                                             (0.007)      (0.008)               (0.009)          (0.012)      (0.012)         (0.019)
Origin: EFTA                                 -0.022       -0.032                -0.009           0.003        0.002           -0.023
                                             (0.026)      (0.028)               (0.028)          (0.028)      (0.028)         (0.058)
Origin: other Europe                         -0.207***    -0.196***             -0.192***        -0.135***    -0.128***       -0.110***
                                             (0.008)      (0.008)               (0.008)          (0.011)      (0.012)         (0.018)
Origin: MENA                                 -0.086***    -0.068***             -0.085***        -0.027       -0.017          0.022
                                             (0.015)      (0.017)               (0.017)          (0.019)      (0.019)         (0.029)
Origin: Sub-Saharan Africa                   -0.089***    -0.084***             -0.104***        -0.060***    -0.057***       -0.039
                                             (0.014)      (0.015)               (0.015)          (0.016)      (0.016)         (0.025)
Origin: Asia                                 -0.126***    -0.121***             -0.134***        -0.066***    -0.062***       -0.033
                                             (0.010)      (0.011)               (0.011)          (0.014)      (0.014)         (0.022)
Origin: LAC                                  -0.183***    -0.172***             -0.131***        -0.105***    -0.092***       -0.061**
                                             (0.011)      (0.012)               (0.013)          (0.014)      (0.014)         (0.022)
Origin: non-Europe OECD                      0.215***     0.151***              0.130***         0.146***     0.150***        0.163***
                                             (0.020)      (0.023)               (0.023)          (0.023)      (0.023)         (0.033)
Age                                                       0.005***              0.008***         0.008***     0.008***        0.006***
                                                          (0.001)               (0.001)          (0.001)      (0.001)         (0.001)
Age^2                                                     0.000***              -0.000***        -0.000***    -0.000***       -0.000***
                                                          (0.000)               (0.000)          (0.000)      (0.000)         (0.000)
Male                                                      0.032***              0.034***         0.034***     0.034***        0.034***
                                                          (0.003)               (0.003)          (0.003)      (0.003)         (0.003)
Years of Education                                        0.053***              0.054***         0.054***     0.054***        0.055***
                                                          (0.001)               (0.001)          (0.001)      (0.001)         (0.001)
Language – advance                                                                               -0.065***    -0.060***       -0.058***
                                                                                                 (0.011)      (0.011)         (0.016)
Language – intermediate                                                                          -0.114***    -0.103***       -0.091***
                                                                                                 (0.013)      (0.013)         (0.019)
Language – beginner                                                                              -0.093***    -0.086***       -0.079***
                                                                                                 (0.018)      (0.018)         (0.024)
Barriers Certification                                                                                        -0.161***       -0.171***
                                                                                                              (0.021)         (0.028)
Job through networks*Migrant                                                                                                  -0.060***
                                                                                                                              (0.014)
Job through networks*Native                                                                                                   -0.063***
                                                                                                                              (0.005)
Observations                   395,284       395,284       394,291               394,291          393,300     393,300         373,350
R-squared                      0.003         0.006         0.078                 0.088            0.089       0.089           0.088
NUTS2 division FE              No            No            No                    Yes              Yes         Yes             Yes
P(safe jobs)                   0.452         0.451         0.460                 0.465            0.465       0.465           0.472
             Note: Robust standard errors are reported in parentheses; ***: p<0.01, **: p<0.05, *: p<0.1.




                                                                      47
           Table 4: Probability of having an income-risk job based on marginal effects of the multinomial logit
                                        Model 1:     Model 2:     Model 3:     Model 4:      Model 5:    Model 6:        Model 7:
                                        Raw gaps     Raw gaps     Individual   NUTS2         Language    Barriers to     Networks
VARIABLES                                            by origin    character.   FE            ability     certification
Foreign born                            0.013***
                                        (0.004)
Origin EU15                                          -0.055***    -0.032***    -0.001        -0.017      -0.017          -0.019
                                                     (0.008)      (0.009)      (0.010)       (0.010)     (0.010)         (0.015)
Origin NMS13                                         0.084***     0.055***     0.061***      0.044***    0.044***        0.011
                                                     (0.009)      (0.009)      (0.009)       (0.012)     (0.012)         (0.018)
Origin EFTA                                          0.002        0.017        0.015         0.011       0.011           0.052
                                                     (0.025)      (0.027)      (0.027)       (0.027)     (0.027)         (0.053)
Origin other Europe                                  0.103***     0.082***     0.082***      0.069***    0.069***        0.035*
                                                     (0.008)      (0.009)      (0.009)       (0.011)     (0.011)         (0.018)
Origin MENA                                          0.020        -0.026*      0.020         0.002       0.002           -0.028
                                                     (0.015)      (0.016)      (0.017)       (0.018)     (0.018)         (0.026)
Origin Sub-Saharan Africa                            -0.075***    -0.088***    -0.053***     -0.059***   -0.059***       -0.104***
                                                     (0.013)      (0.013)      (0.014)       (0.015)     (0.015)         (0.021)
Origin Asia                                          0.041***     0.015        0.038***      0.022       0.022*          -0.005
                                                     (0.011)      (0.011)      (0.012)       (0.014)     (0.013)         (0.020)
Origin LAC                                           -0.026**     -0.041***    -0.037***     -0.041***   -0.041***       -0.098***
                                                     (0.012)      (0.012)      (0.012)       (0.013)     (0.013)         (0.018)
Origin non-Europe OECD                               -0.156***    -0.098***    -0.076***     -0.084***   -0.084***       -0.092***
                                                     (0.017)      (0.020)      (0.021)       (0.021)     (0.021)         (0.031)
Age                                                               -0.012***    -0.013***     -0.013***   -0.013***       -0.012***
                                                                  (0.001)      (0.001)       (0.001)     (0.001)         (0.001)
Age^2                                                             0.0001***    0.0001***     0.0001***   0.0001***       0.0001***
                                                                  (0.0000)     (0.0000)      (0.0000)    (0.0000)        (0.0000)
Male                                                              0.089***     0.085***      0.085***    0.085***        0.080***
                                                                  (0.003)      (0.003)       (0.003)     (0.003)         (0.003)
Years of Education                                                -0.049***    -0.050***     -0.050***   -0.050***       -0.051***
                                                                  (0.001)      (0.001)       (0.001)     (0.001)         (0.001)
Language – advance (reference
category is mother tongue)                                                                   0.026***    0.023***        0.025
                                                                                             (0.010)     (0.009)         (0.016)
Language – intermediate
(reference category is mother
tongue)                                                                                      0.034***    0.031***        0.017
                                                                                             (0.013)     (0.012)         (0.019)
Language – beginner (reference
category is mother tongue)                                                                   0.038***    0.034***        0.019
                                                                                             (0.017)     (0.016)         (0.023)
Barrier Certification                                                                                    0.095***        0.115***
                                                                                                         (0.022)         (0.031)
Job through networks*Migrant                                                                                             0.023*
                                                                                                                         (0.013)
Job through networks*Natives                                                                                             0.044***
                                                                                                                         (0.005)
Observations                           395,284       395,284      394,291      394,291       393,300     393,300         373,350
R-squared                              0.003         0.006        0.078        0.088         0.089       0.089           0.088
NUTS2 division of residence FE         No            No           No           Yes           Yes         Yes             Yes
P(income-risk jobs)                    0.368         0.369        0.358        0.356         0.356       0.356           0.353
          Note: Robust standard errors are reported in parentheses; ***: p<0.01, **: p<0.05, *: p<0.1.



                                                                   48
              Table 5: Probability of having a health-risk job based on marginal effects of the multinomial logit
                                    Model 1:                                                                              Model 7:
                                    Raw        Model 2:     Model 3:                      Model 5:        Model 6:        Networks
                                    gaps       Raw gaps     Individual    Model 4:        Language        Barriers to
VARIABLES                                      by origin    character.    NUTS2 FE        ability         certification
Foreign born                        0.089***
                                    (0.004)
Origin EU15                                    0.015**      0.019**       0.007           -0.017**        -0.016**        -0.009
                                               (0.007)      (0.007)       (0.007)         (0.007)         (0.007)         (0.011)
Origin NMS13                                   0.105***     0.114***      0.101***        0.065***        0.048***        0.040***
                                               (0.008)      (0.008)       (0.008)         (0.009)         (0.009)         (0.013)
Origin EFTA                                    0.020        0.014         -0.006          -0.014          -0.013          -0.029
                                               (0.023)      (0.023)       (0.021)         (0.020)         (0.020)         (0.037)
Origin other Europe                            0.104***     0.114***      0.109***        0.065***        0.061***        0.076***
                                               (0.008)      (0.008)       (0.008)         (0.009)         (0.009)         (0.014)
Origin MENA                                    0.066***     0.095***      0.065***        0.025*          0.020           0.007
                                               (0.013)      (0.015)       (0.014)         (0.013)         (0.013)         (0.018)
Origin Sub-Saharan Africa                      0.164***     0.171***      0.157***        0.118***        0.117***        0.143***
                                               (0.014)      (0.015)       (0.015)         (0.014)         (0.014)         (0.022)
Origin Asia                                    0.085***     0.106***      0.095***        0.044***        0.042***        0.037**
                                               (0.009)      (0.009)       (0.009)         (0.010)         (0.010)         (0.014)
Origin LAC                                     0.210***     0.213***      0.169***        0.146***        0.139***        0.159***
                                               (0.012)      (0.012)       (0.011)         (0.011)         (0.011)         (0.018)
Origin non-Europe OECD                         -0.058***    -0.054***     -0.054***       -0.062***       -0.064***       -0.071***
                                               (0.014)      (0.015)       (0.015)         (0.014)         (0.014)         (0.019)
Age                                                         0.006***      0.006***        0.006***        0.006***        0.006***
                                                            (0.001)       (0.001)         (0.001)         (0.001)         (0.001)
Age^2                                                       0.000***      0.000***        0.000***        0.000***        0.000***
                                                            (0.000)       (0.000)         (0.000)         (0.000)         (0.000)
Male                                                        -0.122***     -0.119***       -0.120***       -0.120***       -0.114***
                                                            (0.002)       (0.002)         (0.002)         (0.002)         (0.002)
Years of Education                                          -0.004***     -0.004***       -0.004***       -0.004***       0.003***
                                                            (0.000)       (0.000)         (0.000)         (0.000)         (0.000)
Language – advance (reference
category is mother tongue)                                                                0.039***        0.037***        0.033***
                                                                                          (0.008)         (0.003)         (0.011)
Language – intermediate
(reference category is mother
tongue)                                                                                   0.080***        0.072***        0.074***
                                                                                          (0.010)         (0.010)         (0.014)
Language – beginner (reference
category is mother tongue)                                                                0.055***        0.052***        0.060***
                                                                                          (0.013)         (0.013)         (0.016)
Barrier certification                                                                                     0.069***        0.056***
                                                                                                          (0.014)         (0.016)
Job through networks*Migrants                                                                                             0.037***
                                                                                                                          (0.009)
Job through networks*Natives                                                                                              0.018***
                                                                                                                          (0.004)
Observations                        395,284    395,284      394,291       394,291         393,300         393,300         373,350
R-squared                           0.003      0.006        0.078         0.088           0.089           0.089           0.088
NUTS2 division of residence FE      No         No           No            Yes             Yes             Yes             Yes
P(health-risk jobs)                 0.180      0.180        0.182         0.179           0.179           0.179           0.175

                                                                  49
 Appendix 1: The role of networks in explaining migrants’ occupational choices

 To better understand the role of networks of co-national migrants in the occupational choices of
 migrants, we estimate the following regression using two different waves of the EU-LFS:
 Log(pct_migrant_2018) cok = α + β log(pct_migrant_2005)cok + ηc + Ωo + φk + ε cok                  (2)
 where log(pct_migrant_2018)cok is the logarithm of the percentage of migrants in 2018 in each
 ISCO 3-digit occupation c that are from the region of origin o in NUTS2 division of destination k.
 Log(pct_migrant_2005) cok is the logarithm of the share of migrant population from region of
 origin o in each occupation c residing in NUTS2 division of residence k in 2005. Ωo is a vector of
 regions of origin o (native, EU15, NMS13, EFTA, Other European countries, MENA, Sub-
 Saharan Africa, Asia, Latin America and non-Europe OECD countries); ηc is a vector of dummy
 variables for each ISCO 3-digit occupation; and φk is a vector of dummy variables for each NUTS2
 division of residence K. In order to avoid overlapping of migrants in the two periods, we restrict
 the sample of migrants in 2018 to those that arrived within the previous ten years.
 Migrant networks are particularly strong among co-nationals in the same area of residence. We
 run the OLS regression in equation (2) in which the cell of analysis is the ratio of migrants from
 world region of origin o to native in NUTS2 division of residence k working in occupation c. The
 results of this regression analysis show a strong path dependence in occupational choices of
 migrants even after controlling for occupation, region of origin and NUTS2 division of residence
 fixed effects (Table A1). Our preferred regression includes all occupations – even those without a
 single migrant – and origin region, host NUTS2 division and occupation fixed effects. The results
 (Column 2) show that a 1 percent increase in the share of co-national migrants in an occupation in
 2005 is associated with a 0.13 percent increase in the share of new migrants in that occupation in
 2018.
 Table A1: Regression analysis on the role of networks in occupational choices of migrants in the EU

Dependent variable: log (% of migrants in         Model 1            Model 2            Model 3   Model 4
2018) per 3-digit occupation, nuts2 division
of residence and region of origin                  (OLS)              (OLS)              (OLS)     (OLS)

Log(% migrants in 2005)                           0.188***           0.127***         0.0828***   0.0344***
                                                 (0.00539)          (0.00518)         (0.00628)   (0.00611)
Constant                                         -6.983***          -6.529***         -3.830***   -2.944***
                                                  (0.0472)            (0.129)          (0.0421)     (0.256)

Observations                                       92,294             92,294              8,439     8,439
R-squared                                           0.060              0.262              0.041     0.500
NUTS2 division of residence FE                       NO                YES                 NO       YES
Region of origin FE                                  NO                YES                 NO       YES
ISCO 3-digit Occupation FE                           NO                YES                 NO       YES
Including occupations with no migrants              YES                YES                 NO        NO
  Note: Robust standard errors are reported in parentheses; ***: p<0.01, **: p<0.05, *: p<0.1.


                                                            50
Appendix 2: Labor demand and share of migrants per occupation

In order to formally assess the link between labor demand and the prevalence of migration in each
occupation we estimate the following regression:
Log(pct_migrants) ck = α + β log(pct_new_hires)ck + ηc + φk + ε ck                                    (3)
where log(pct_migrants)ck is (the natural logarithm of) the percentage of migrants in 2018 in each
ISCO 3-digit occupation c in UTS2 region of residence k. Log(pct_new_hires) ck is the logarithm
of the share of new hires migrant population from origin region o in each occupation c residing in
NUTS2 division k in 2005; ηc is a vector of dummy variables for each ISCO 3-digit occupation;
and φk is a vector of dummy variables for each NUTS2 division of residence K.
The results of the above regression in Equation 3 are reported in Table A2. The strong relationship
between labor demand and the share of migrants per occupation in the EU holds even when we
control for ISCO-08 3-digit occupation and NUTS2 region of residence fixed effects (Column 3).
In particular, a 1 percent increase in the share of new hires in an occupation in a given NUTS2
division –which we use as a proxy for the dynamism of the labor demand– is associated with a
0.02 percent increase in the share of migrants per occupation. The coefficient is statistically
significant at the 1 percent level.


Table A2: Regression analysis on the role of labor demand in occupational choices of migrants in the EU

  Dependent variable: Log (% of migrants) per             Model 1              Model 2          Model 3
  3-digit occupation and NUTS2 division in 2018           (OLS)                (OLS)            (OLS)

  Log (% new_hires (<1 year))                            0.102***             0. 089***         0.020***
                                                          (0. 003)             (0.003)           (0.003)
  Constant                                               -1.793***            -1.883***        -2. 675***
                                                           (0.012)             (0.112)           (0.678)

 Observations                                               23,803              23,803          23,803
 R-squared                                                  0.066                0.147           0.408
 NUTS2 division of residence FE                              NO                  YES             YES
 ISCO 2-digit Occupation FE                                  NO                   NO             YES
Note: Robust standard errors are reported in parentheses; ***: p<0.01, **: p<0.05, *: p<0.1.




                                                         51