Different Patterns of Labor Market Integration by Migration Motivation in Europe: The Role of Host Country Human Capital

Transcription

1 Original Article Different Patterns of Labor Market Integration by Migration Motivation in Europe: The Role of Host Country Human Capital International Migration Review 1-31 ª The Author(s) 2018 Reprints and permission: sagepub.com/journalspermissions.nav DOI: / journals.sagepub.com/home/mrx Wouter Zwysen University of Essex Abstract We study whether the acquisition of host country human capital, such as obtaining equivalent qualifications, good language skills, or naturalization, explains differences in labor market integration between migrants depending on their initial motivation. We use cross-national European data from the 2008 ad hoc module of the Labour Force Survey to analyze migrant gaps in labor market participation, employment, occupational status, and precarious employment. We find that different rates of and returns to host country human capital explain a substantial part of the improvements in labor market outcomes with years of residence, particularly for noneconomic migrants who experience faster growth on average. Introduction Migrants labor market outcomes are generally found to improve over time as they acquire better knowledge of the host country and relevant human capital (Duleep and Regets 1999; Chiswick 2009). Little is known about whether this pattern holds for all migrants, however. Migrants differ from each other in their resources and the Corresponding Author: Wouter Zwysen, Department of Sociology, University of Essex, Wivenhoe Park, CO4 3SQ Colchester, Essex, UK. wzwyse@essex.ac.uk

2 2 International Migration Review XX(X) extent of governmental support they receive, their motivation and match with the host country, and their prospects of return migration. This affects their choices made in the country of residence, which in turn affects labor market outcomes (Duleep and Regets 1999; Cortes 2004). The main question of this article was the role played by migrants host country human capital, such as language skills, obtaining equivalent and recognized qualifications, or having acquired host country nationality, on labor market integration and the extent to which these factors can explain differences between groups of migrants depending on their main motivation for migration. Understanding whether these factors operate similarly over groups of migrants is an important step toward improving labor market integration for all migrants. This is particularly important with regard to the current refugee crisis in Europe, which highlights the variation in migrant motivations and the worries about integrating large groups of primarily noneconomic migrants. We use detailed cross-national data from the 2008 EU Labour Force Survey (LFS) to estimate the differences in labor market integration by self-reported reason for migration. We measure labor market integration as the difference in terms of participation, employment, and job quality of migrants compared to similar natives. There are very few studies using cross-national European data in studying integration patterns by several categories of migration motivation (Damas de Matos and Liebig 2014; Cangiano 2015; Dumont et al. 2016; Dustmann et al. 2016), and none of them study the determinants of these differences in patterns. In the remainder of this article, we present our conceptual framework and form hypotheses regarding the integration patterns with years of residence for different groups of migrants. We then present the data in more detail, discuss our measures of acquired host country human capital, and explain our methodology to test the drivers of integration patterns. The results section shows that host country human capital drives a substantial part of the improvement in labor market integration over time, but this differs between groups of migrants. We test the robustness of these findings by studying differences in the effects of host country human capital by country of residence and studying different subgroups of migrants. Finally, we draw our conclusions. First, we discuss some of the previous literature on the labor market outcomes of migrants depending on their primary reason for migration, however. Conceptual Framework Heterogeneity among Migrants by Motivation There are several different ways to categorize migrants with the choice often depending on available data. Some studies are able to differentiate by categories of admission into the country, as, for example, in Sweden (Lundborg 2013; Bevelander and Pendakur 2014; Luik, Emilsson, and Bevelander 2016), Norway (Bratsberg, Raaum, and Røed 2017), the United Kingdom (Ruiz and Vargas-Silva 2017),

3 Zwysen 3 and North America (Cortes 2004; Aydemir 2009; Connor 2010). This captures the legal framework along with barriers and support systems under which migrants enter, but may not always reflect the actual motivation insofar as the entry category is heavily shaped by the legal constraints of the receiving country (Campbell 2014). Other studies approximate the reason for migration through the country of origin (Kausar and Drinkwater 2010; Sarvimäki 2017). This can be problematic when migrants from the same country have different reasons for migration. In this article, we distinguish migrants based on their self-reported primary reason for migration. As this is asked in an anonymized survey, people may be more likely to report truthfully. Migrant motivations are often complex, however, and may not be easily captured by one category, for instance in the case of people fleeing persecution as well as entering a country as a dependent, which is relatively common for female migrants (Campbell 2014; Dumont et al. 2016). Cangiano (2015, 426) discusses the appropriateness of using these self-reported motivations as a proxy for admission categories. They maintain that the broad trends with regard to countryspecific flows of types of immigrants are consistent with other data sources on visa admissions and on prior expectations. Regardless of the measurement, noneconomic migrants are generally found to do substantially worse in terms of employment and quality of work than economic migrants on the labor market, and family migrants tend to do slightly better than refugees among noneconomic migrants (e.g., Aydemir 2009; Connor 2010; Bevelander 2011; Lundborg 2013; Bevelander and Pendakur 2014; Campbell 2014; Cangiano 2015; Luik, Emilsson, and Bevelander 2016; Bratsberg, Raaum, and Røed 2017; Sarvimäki 2017). The EU LFS data, which we use here, have previously been used to study country differences in the composition, reception, and relative outcomes of migrants, particularly those arriving for humanitarian reasons (Cangiano 2015; Dumont et al. 2016; Dustmann et al. 2016). This allows for country comparisons, whereas most papers focus on only one country. Differences in labor market outcomes between migrant categories appear to be lower in countries that prioritize labor migration such as Spain and Italy, while it is more stark in Germany, France, and Sweden (Cangiano 2015; Dumont et al. 2016). There are several possible reasons for differences in outcomes between migrant categories. First, noneconomic migrants are thought to be less positively selected in terms of human capital and motivation than economic migrants (van Tubergen, Maas, and Flap 2004; Lundborg 2013). Second, migrants differ in the resources and networks available to them early on in the country of origin which can lead to better outcomes. Migrants arriving for family reasons often have existing networks to build on (Aydemir 2009), while student migrants build up networks and connections as well as obtaining qualifications in the country of residence. Humanitarian migrants often receive more governmental support, for example, through integration courses and assistance, but they also face more uncertainty regarding their status initially and may have suffered trauma impeding their labor market performance (Connor 2010; Hainmueller, Hangartner, and Lawrence 2016). Third, migrants may also differ in

4 4 International Migration Review XX(X) their evaluation of what makes a successful migration experience, which can affect the reservation wage. Luthra, Platt, and Salamońska (2016) show that among recently arrived Eastern European migrants, unemployment had a far worse effect on people s satisfaction and happiness for economic migrants than for noneconomic, who did not attribute as much of their feeling of success to finding any type of work. This different evaluation resulted in economic migrants being much more likely to take any job, including lower quality work. Host Country Human Capital as a Driver of Different Trajectories This article addresses whether the way labor market outcomes change with years of residence differs depending on migrant motivation. Migration itself is often associated with a decrease in labor market outcomes as migrants human capital is not completely transferable. Over time, labor market outcomes improve as migrants acquire host country human capital such as language skills and knowledge of the labor market (Duleep and Regets 1999; Chiswick 2009; Duleep, Liu, and Regets 2014). In their Immigrant Human-capital Investment model, Duleep and Regets (1999) postulate a heterogeneity in this process, as minorities that face the worst initial disadvantage also face lower opportunity costs to make further investments and are then most likely to invest in further human capital. This is directly relevant for the study of migrant motivations as there are large differences in their initial disadvantage, which then affect the opportunity cost of further investments. Relatedly, migrants also differ in their future plans and prospects, which can affect their integration patterns in the country of residence. Those who are less likely or less able to return to their country of origin, such as family and humanitarian migrants, could be expected to make more investments in the country of residence, such as learning the language or taking up extra training, as they expect more returns from this human capital. Cortes (2004) tests this in a comparison of refugee and economic migrants in the United States between 1980 and 1990 and finds that the earnings of refugees rise faster than those of economic migrants over time. He finds that this is largely due to a more rapid increase in English language skills. Previous cross-sectional studies generally found convergence between groups over time with more initially disadvantaged groups such as family and refugee migrants experiencing faster growth in their probability of employment and earnings. Some studies also find indications that this is related to an increase in, for example, language skills. Cangiano (2015) uses the LFS 2008 and finds partial convergence over time as differences between migrant groups decline substantially after 10 years in terms of employment and access to qualified jobs, although even in the long run employment gaps with natives remain for noneconomic migrants and for almost all groups with regard to job quality. Dustmann et al. (2016) use the same data to show that humanitarian migrants have far larger initial employment gaps than non-eu migrants, but these differences disappear after almost 20 years of residence. A recent report by the

5 Zwysen 5 European Commission and the OECD (Dumont et al. 2016) goes some way to explore the drivers of this convergence using the 2014 LFS to show that humanitarian migrants are more likely to naturalize and experience a stronger improvement in language skills with time than economic migrants. These factors seem to boost employment probability somewhat more for humanitarian than economic migrants. They do not address the extent to which this is driven by host country human capital and how it varies by other groups than refugees and economic migrants, however. Some country-specific studies have also addressed the trajectories over time between different types of migrants. Campbell (2014) uses the UK LFS to show different trajectories by original motivation with student migrants and refugees increasing their employment probability and wages over time in the country, while female work migrants start at high employment levels, but these decrease over time. He finds that differences in language skills account for a small part of the earnings penalty of refugees relative to work immigrants and also shows that migrants who migrated for family or protection reasons are more likely to take up the British nationality than student or economic migrants. Bevelander (2011) documents a convergence in the employment penalties with natives between refugees, family, and labor migrants in Sweden after 15 to 20 years of residence, although it does not close. These studies investigate labor market integration over time using cross-sectional data, meaning the outcomes of different cohorts of migrants are compared at one point in time. A few studies have been able to use longitudinal data where the same migrants are followed over time, which does not make the assumption that all cohorts of migrants of a specific type have similar integration trajectories. Ruiz and Vargas-Silva (2017) use longitudinal data to compare outcomes of recently accepted refugees in the United Kingdom to all other immigrants and found that refugees were 7 percentage points (p.p.) more likely to be unemployed eight months after receiving protection, but 13 months later, there was no longer a statistically significant gap. Lundborg (2013) also shows that refugee gaps compared to natives diminish substantially over time, which is also found by Sarvimaki (2017) in Finland. Bratsberg, Raaum, and Røed (2017) use longitudinal register data in Norway and find evidence of divergence, however, as the employment gaps of refugees compared to natives are at their lowest after five to nine years of residence and then increase substantially again. They also find this for female family migrants, while for male family migrants the employment probability decreases linearly over time. Learning the language, taking up host country nationality, or obtaining qualifications or training are forms of human capital that can be specific to the country of residence. Acquiring these often requires a deliberate decision; costs time, effort, and sometimes money; and brings about an expected return on the labor market (Cortes 2004; Chiswick 2009; De Vroome and van Tubergen 2010). This is not to say that they necessarily follow from a free choice as migrants can face substantial legal and financial constraints in acquiring receiving society-specific human capital such as language skills, qualifications, job training, or citizenship (Barry 2001).

6 6 International Migration Review XX(X) Based on the discussion above, we first of all expect that there are differences in labor market integration trajectories between groups of migrants and that part of this is due to differences in the acquisition and returns to host country human capital (Duleep and Regets 1999; Cortes 2004). We then expect less difference in labor market integration over time once accounting for these levels of human capital [H1]. We expect host country human capital to have larger positive returns on the labor market for humanitarian migrants and family migrants than it does for economic migrants [H2]. This follows the idea that, because of their initially discounted human capital, the opportunity costs of further investment are lower and the returns to host country human capital larger for migrants that are more disadvantaged initially. The strong signal given off by acquired host country human capital can be particularly important in counteracting possible statistical discrimination as well as providing opportunities for more disadvantaged migrants (Duleep and Regets 1999; Corluy, Marx, and Verbist 2011; Duleep, Liu, and Regets 2014; Bratsberg, Raaum, and Røed 2017). Finally, we expect host country human capital to mainly increase the opportunities to work on high-quality jobs, rather than finding work in general (Jarreau 2015; Ruiz and Vargas-Silva 2017). Labor market participation and employment are expected to be less affected by host country human capital as many migrants generally find work in lower-skilled sectors where high human capital is not required (e.g., Reyneri and Fullin 2011). Higher host country human capital may even affect employment negatively if the reservation wage rises. As economic migrants arrived to work, they generally have lower reservation wages than noneconomic migrants (Luthra, Platt, and Salamońska 2016), meaning an increase in reservation wages would affect their employment probability more. We therefore expect that host country human capital particularly improves the quality of work, rather than activity or employment [H3]. Data and Methods European Labor Force Survey We test these hypotheses using the 2008 ad hoc module of the European LFS. This is a large-scale harmonized survey of European countries. While it does not primarily target migrants, the LFS has been shown to produce reliable estimates in terms of stocks of non-nationals (Martí and Ródenas 2007). One limitation is that this dataset does not reach irregular migrants who make up large fractions of the immigrant population in some countries (Boeri et al. 2015). The 2008 ad hoc module contains harmonized questions on the main reason for migration, perceptions of language skills, services used, and whether migrants naturalized. The sample for the main analyses is restricted to first-generation migrants who were not nationals at birth, aged 25 to 65. After listwise deletion of observations with

7 Zwysen 7 missing values, our sample includes 20,198 migrants in 14 countries. 1 We restrict the analyses to migrants who have been in the country for longer than five years. We do this for two main reasons. First, return migration is substantially more likely in the first five years of residence and in this article we want to estimate the labor market integration of long-term migrants (Lundborg 2013; Campbell 2014; Ruiz and Vargas-Silva 2017). Second, after five years most migrants would have the right to stay and work in most countries under study, meaning we limit the heterogeneity in legal treatment between migrants (Campbell 2014; Cangiano 2015). Labor Market Integration To measure being integrated on the labor market, we consider participation on the labor market and employment when active, but also the occupational status of jobs measured through the ISEI scale (Ganzeboom and Treiman 1996) 2 and an indicator for working on an involuntary nonstandard contract (Buchholz et al. 2009). This bad job dummy indicates that someone works part-time but wants to work more hours or works on a temporary contract because a permanent position was not available. While employment of migrants likely matters most to public opinion, the quality of work is also crucial for full integration. We measure labor market integration as the difference in labor market outcomes of migrants and similar natives (Algan et al. 2010). We use coarsened exact matching to estimate the difference in all four outcomes between migrants (treated) and similar natives (controls) (Iacus, King, and Porro 2012). A total of 483,707 natives in the LFS are used to form counterfactuals, divided over 6,322 strata made up of combinations of region at the NUTS level, 3 age, 4 gender, being married, and the highest obtained qualification, 5 in which at least five natives were present. The difference between the observed labor force participation, employment status, occupational status, and working on an involuntary nonstandard contract of each migrant with the counterfactual within each stratum is then used as our measure of labor market integration. Through equalizing the distribution of covariates between 1 Austria, Belgium, Germany, Spain, France, Greece, Ireland, Italy, Luxembourg, the Netherlands, Norway, Portugal, Sweden, and the United Kingdom. 2 The ISEI score was developed to measure occupational status and is computed in such a way that the direct effect of education on income is minimized. The measure ranges between 16 and 90, and a higher score indicates higher socioeconomic status. Ganzeboom and Treiman (1996) provide tables to match this score to three-digit isco88 occupations. 3 One-digit NUTS level: three in Austria; three in Belgium; 16 in Germany; seven in Spain; eight in France; four in Greece; two in Ireland; five in Italy; one in Luxembourg, one in the Netherlands; seven in Norway; five in Portugal; three in Sweden; 11 in the United Kingdom. 4 Age is measured in five-year intervals from 25 to Seven ISCED codes ranging from preprimary/no degree to upper tertiary.

8 8 International Migration Review XX(X) migrants and natives, we do not have to rely on extrapolation through modeling assumptions and potential nonlinearity is taken into account. Migrant Categories Migrants are divided into six categories based on their main reason for migrating to the current country of residence. We include two types of economic migrants: those who arrive as intracorporate transfers or who arrived with a contract, and those who arrive to look for work. Noneconomic migrants are separated into three groups: migrants who arrived as students; those seeking international protection; and migrants arriving for family formation or reunification. Finally, we include a rest category of migrants with other motives. This category likely also captures migrants who could not pick one main motivation. To measure labor market integration over time we include years of residence. This is reported as exact years until 10 and afterward as five-year intervals. We impute the midpoint of these intervals and treat the variable as continuous. Country of birth is only provided in highly aggregated groups. 6 As these groups are associated with migrant motivation and not directly meaningful, we do not include them, but use relevant characteristics from the sending region, calculated as the average of the values for the relevant countries, weighted by the share of migrants of a specific country of origin within the region of origin in a given country of destination. The shares are obtained by the Database on Immigrants in OECD Countries from The share of Christians in the sending country, obtained through the world religion dataset for the year 2005 (Maoz and Henderson 2013), is used to indicate religious closeness to the receiving country. We also include female labor force participation and the share of graduates (van Tubergen, Maas, and Flap 2004; Fleischmann and Dronkers 2010). Following van Tubergen, Maas, and Flap (2004), we include two ratio variables (in logarithmic form) to capture selection of migrants based on perceived earnings distributions and level of development: the ratio of the Gini coefficient in the sending region to that of the receiving country and the ratio of GDP per capita (ppp). The share of graduates, female labor force participation, Gini coefficient, and GDP per capita (in current US dollars) are all obtained from the World Bank. 7 6 Categories are as follows: own country, EU15, NMS10, NMS3, EFTA, Other Europe, North Africa, Other Africa, Near Middle East, East Asia, South and South-East Asia, North America, Central America, South America, Australia/ Oceania, Latin America with some countries having further aggregations of all new member states; EFTA and other Europe; North Africa and the Near Middle East; South and East Asia; North America and Australia. 7 Values from 2008 are used (and average of if 2008 was not available).

9 Zwysen 9 We include a dummy variable for being born in a country that was an EU member state at the time of the survey as EU migrants are more free in their migration and face far fewer disadvantages on the labor market (e.g., Damas de Matos and Liebig 2014). As a sensitivity test, we also carry out all analyses restricting the sample to non-eu migrants finding no real difference in the relative effects of motivation. Table 1 presents the weighted averages of all variables in our sample by migrant motivation for women and men. Family is the most important reason for women to migrate (48%), followed by looking for work (20%). Around 6 percent of female migrants arrive for study and protection. Men primarily migrate for economic reasons with 35 percent arriving without a contract and 17 percent with a contract. A further 22 percent arrive as family migrants, and 10 percent of male migrants seek protection. Almost all migrants are more likely to be active on the labor market than similar natives, with the exception of male protection migrants. Family and protection migrants are most disadvantaged in terms of employment probability, but not much more so in terms of job quality when working than the majority of economic migrants. In terms of job quality, student and economic migrants with a contract upon arrival perform best on average. Student migrants are substantially younger than the others on average. The majority of migrants, especially family migrants, are married. There are large differences in qualifications as around 60 percent of student migrants have tertiary qualifications, compared to 40 to 45 percent of economic migrants with a contract, but only 26 to 28 percent of family and protection migrants and 12 to 19 percent of economic migrants without a contract. Economic migrants without a contract have spent least time in the country of residence on average (11 13 years), likely reflecting their higher probability of return migration. The characteristics of the sending region also differ between groups. Almost half of the migrants who arrived for economic reasons with a contract ready or for other reasons originate from a current EU member state, as do female student migrants, while only 17 percent of male student migrants are born in the current European Union. Only around one tenth of protection migrants were born in EU member states. They originate from regions with fewer Christians on average and lower female labor force participation rates. This cultural distance is also seen in the share of female labor force participation in the region of origin which is lowest on average for protection migrants. Male student migrants and all economic migrants without a contract tend to originate from regions with low shares of graduates. While all migrants were on average born in less economically equal and less wealthy regions than their current residence country, this difference is largest for protection migrants. On average, migrants seeking protection come from more culturally distant and less economically developed regions which could impede their integration. Male student migrants have similar patterns, while this is very different for female students who are more likely to come from more developed countries.

10 10 International Migration Review XX(X) Table 1. Descriptive Statistics of Migrant (a) Women and (b) Men by Motivation. Economic with contract Economic without contract Study Protection Family Other (a) Women Active Difference active Employed Difference employed ISEI Difference ISEI Bad job Difference bad job Age Married No qualifications Primary qualifications Low secondary qualifications Upper secondary qualifications Post-secondary qualifications Low tertiary qualifications Upper tertiary qualifications EU member Christians (%) Female participation (%) Share graduates (%) Ratio Gini (log) Ratio GDP (log) Years of residence Count 923 2, , Share out of migrants (%) (b) Men Active Difference active Employed Difference employed ISEI Difference ISEI Bad job Difference bad job (continued)

11 Zwysen 11 Table 1. (continued) Economic with contract Economic without contract Study Protection Family Other Age Married No qualifications Primary qualifications Low secondary qualifications Upper secondary qualifications Post-secondary qualifications Low tertiary qualifications Upper tertiary qualifications EU member Christians (%) Female participation (%) Share graduates (%) Ratio Gini (log) Ratio GDP (log) Years of residence Count 1,767 3, ,067 2,268 1,046 Share out of migrants (%) Note: Weighted average from EULFS 2008 ad hoc module for sample. Migrants were aged 25 65, not national at birth, and arrived at least five years prior to survey. Host Country Human Capital Migrants are expected to face initial disadvantage because their human capital is not completely transferable (e.g., Chiswick and Miller 2008; Reyneri and Fullin 2011). Damas de Matos and Liebig (2014) show that qualifications of migrants are substantially discounted in Europe and in particular that family and humanitarian migrants have close to zero returns to having tertiary qualifications in terms of employment. Taking steps to have foreign qualifications recognized leads to better outcomes and increases the returns to qualifications. We include a dummy variable indicating that migrants either obtained their highest qualifications in the country of residence or have taken steps toward equivalizing their qualifications obtained abroad, rather than not having equivalent qualifications for any reason. Second, we consider language skills of migrants. Speaking the language well is crucial for further integration and often a requirement for finding good employment

12 12 International Migration Review XX(X) and making use of public services, bringing substantial returns on the labor market (e.g., Dustmann and Fabbri 2003; Cortes 2004; Cebulla et al. 2010; De Vroome and van Tubergen 2010; Cheung 2013; Campbell 2014; Cheung and Phillimore 2014). The ad hoc module includes a question on whether respondents feel they need to improve their language skills to get an appropriate job which is reversed to indicate good language skills. This question only captures self-reported views on language skills, and the phrasing of the question refers specifically to its use in getting an appropriate job, which people may interpret differently. In an evaluation of the 2014 LFS, it was shown that the proportions of migrants who reported not needing better language skills did correspond to the proportion of migrants who reported their language skills as advanced or mother tongue in 2014, indicating that the question does capture perceived language skills (Eurostat 2015). It is important to note that many migrants to European countries originate from former colonial countries and may already speak the language well prior to arrival. While we cannot account for the way in which good language skills were acquired, we do carry out a sensitivity test to assess whether the effects of language skills are driven by migrants who may have spoken the language of the country of residence in their origin country. Third, we include a dummy for taking up host country nationality after the age of 16. Naturalization presents an investment as conditions, differing by country, have to be met, but it can also affect further labor market integration positively. These positive effects may come about for several reasons. The transaction costs of hiring naturalized migrants are generally lower, as there is no need for checking work permits. Employers may also see it as a strong signal, both in terms of permanence of the candidate and in terms of sociocultural integration. Depending by country, there are also jobs that are restricted to citizens, so gaining nationality may increase the possible job pool. The European literature on this topic tends to find small positive naturalization premiums on labor market outcomes, although the effect is heterogeneous and mainly accrues to the most disadvantaged groups (Corluy, Marx, and Verbist 2011; Helgertz, Bevelander, and Tegunimataka 2014; Jarreau 2015). Jarreau (2015) studies the effect of naturalization on earnings in France and finds a positive effect due to migrants moving to better-matched jobs after naturalization. Previous studies found that less advantaged migrants are more likely to naturalize (Vink, Prokic-Breuer, and Dronkers 2013) and that refugees tend to have higher returns to naturalization in terms of employment (Dumont et al. 2016). As a robustness test to account for country differences in the duration of stay required to naturalize, we restrict the sample to migrants who were in the country longer than 10 years (Dumont et al. 2016). Table 2 presents the description of these host country human capital characteristics over migrant categories. Economic migrants without a contract stand out as having relatively low scores on all types of host country human capital, while migrants who arrived with a contract have generally acquired more host country human capital. Student migrants are very likely to have equivalent qualifications and good language skills. Migrants seeking protection and family migrants have

13 Zwysen 13 Table 2. Acquired Host Country Human Capital by Migrant Motivation. Economic with contract Economic without contract Study Protection Family Other Women Equivalent qualifications Good language skills Naturalised Men Equivalent qualifications Good language skills Naturalised Note: Weighted average from EULFS 2008 ad-hoc module for sample. Migrants were aged 25 65, not national at birth and arrived at least five years prior to survey. relatively poor language skills, but score high on other types of acquired host country human capital with particularly refugees very likely to take up host country nationality. Country Differences We study labor market integration patterns by motivation for all migrants to European countries. It is important to note that there are substantial differences between countries in their integration policies, which affect the opportunities and incentives for acquiring host country human capital, for instance through the offering or even requiring of integration and language courses for migrants, as in Belgium, France, Germany, and Sweden, or in the availability of assistance in job search (Huddleston, Tánczos, and Wolffhardt 2016). With regard specifically to humanitarian migrants, there are substantial differences in the restrictions to work faced by migrants and the insecurity in the initial period before asylum is granted, which can have long-lasting scarring effects (Aiyar et al. 2016; Dustmann et al. 2016; Hainmueller, Hangartner, and Lawrence 2016). Besides policy differences, the labor market conditions can also shape different integration patterns as they influence the ease of integration on the labor market (Reyneri and Fullin 2011; Barslund et al. 2016; Konle-Seidl and Bolits 2016). The probability of migrants knowing the language of the country of residence differs strongly over the different countries, as do the incentives to learn the language. In some countries and occupations English may be generally used, while migrants may also be more inclined to learn a language such as English that can pay off in other countries as well (Dustmann et al. 2016; Luik, Emilsson, and Bevelander 2016). The returns to equivalent qualifications likely differ depending on the country

14 14 International Migration Review XX(X) of residence, as certificates for instance matter more on the German than on the UK labor market (Dustmann et al. 2016). Finally, naturalization is likely to have a different impact over countries, depending on the regulations allowing it and the possible benefits from it. European countries strongly differ in the conditionality of naturalizations such as allowing dual citizenship or not, tying it to financial independence and different tests (Vink, Prokic-Breuer, and Dronkers 2013). Helgertz, Bevelander, and Tegunimataka (2014) compare the naturalization premium on earnings in Denmark, which has a rather strict conditionality, and Sweden, which is much more open in terms of naturalization. Interestingly, they find very little difference as there is a similar premium to otherwise disadvantaged groups in both countries, consistent with the small positive effect for disadvantaged origin groups found by Corluy, Marx, and Verbist (2011) in their study on Belgium. A detailed discussion of the country differences is beyond the scope of this article, but we carry out a sensitivity analysis to test country differences in the returns to host country human capital. The work by Cangiano (2015) provides more information on policy differences between countries and their effect on different categories of migrants. Estimating Labor Market Integration over Time The descriptive statistics show substantial differences in the levels of labor market integration and host country human capital between migrants with different motivations. To test the extent to which host country human capital explains labor market integration patterns, we model the estimated difference between migrants and similar natives in linear regression models. Standard errors are clustered at the regional level to account for the higher measurement error introduced using an estimated dependent variable (Lewis and Linzer 2005). Analyses are carried out separately by gender as men and women differ in their labor market integration patterns and may also differ in their reporting of main migrant motivation (Dumont et al. 2016). All analyses are weighted using the weights provided by the EU LFS. DðY is ^Y s Þ¼aþb 1 REAS þ b 2 X þ b 3 C þ b 4 YEARS þ b 5 YEARS YEARS þ g 1 REAS YEARS þ g 2 REAS YEARS YEARS þ e ð1þ Equation 1 shows the model to test labor market integration over years of residence by motivation groups. The outcome variable is the difference between the observed outcome for individual migrant i and the estimated counterfactual among natives in stratum s. The main variable of interest is the reason for migration (REAS). X is a vector of controls including marital status, education, and age (squared) as well as controls for the sending region, and C is a vector of dummies for country of destination. We include years of residence (YEARS) and its square. This is interacted with main motivation for migration (REAS) through interaction

15 Zwysen 15 terms y 1 and y 2, allowing for different labor market integration patterns by migration motivation. DðY is ^Y s Þ¼aþb 0 1 REAS þ b 0 2 X þ b 0 3 C þ b 0 4 YEARS þ b 0 5 YEARS YEARS þ g 0 1 REAS YEARS þ g 0 2 REAS YEARS YEARS þ d 1 HC þ d 2 REAS HC þ e Equation 2 includes the indicators of host country human capital (HC), equivalent qualifications, naturalization, and language skills. This is interacted with migration motivation to allow for different returns. Comparing the coefficients and estimated effects in equation 1 to those of equation 2 indicates whether part of the differences between migrant groups or in their trajectories over time is mediated by host country human capital. A reduction in the effect of years since migration and the variation between migrant groups would indicate that part of the differences is due to differences in acquired host country human capital (Mackinnon and Dwyer 1993). To facilitate the interpretation of results, we study marginal effects and predicted probabilities. As the LFS is a cross-sectional dataset, we cannot study the accumulation of host country human capital and changing labor market integration for one person over time, but only compare migrants who arrived in different years. By comparing cohorts of migrants within each motivation group, we make the assumption that they have the same integration pattern. This may not hold as initial conditions can vary and lead to different labor market trajectories (Lundborg 2013; Barslund et al. 2016; Hainmueller, Hangartner, and Lawrence 2016). We may also observe diminishing labor market penalties with years of residence due to attrition rather than a catching up over time if migrants who do less well economically are more likely to move away and drop out of the sample. The propensity to leave the country of residence differs between groups and is generally higher for economic and student migrants than humanitarian or family migrants. This means the attrition bias, where labor market outcomes seem to improve over time because those that did less well drop out of the sample, would be worse for these groups (Cortes 2004; Lundborg 2013; Campbell 2014; Bratsberg, Raaum, and Røed 2017). As a sensitivity test, we restrict the sample further to those who have remained in the country for at least 10 years. ð2þ Results Patterns of Labor Market Integration over Time Figure 1A and 1B presents the estimated patterns of labor market integration over time. Overall, we see that labor market outcomes improve with years of residence, although there are variations. We also see that generally the improvement evens out over years of residence. The interactions between reason for migration and years of residence (squared) are jointly statistically significant (at p<0.05) for activity for

16 16 International Migration Review XX(X) A active employed gap gap years of residence years of residence status bad job gap years of residence gap years of residence Economic with contract Study Family Economic without contract Protection Other B active employed gap years of residence gap years of residence status bad job gap gap years of residence years of residence Economic with contract Study Family Economic without contract Protection Other Figure 1. Labor market integration over time for (A) women and (B) men. Note: Figures show predicted migrant-native penalty by main motivation over years of residence, controlling for socio-demographic and sending region characteristics.

17 Zwysen 17 female immigrants and employment and working on an involuntary nonstandard contract for men. Migrants remain more likely to be active on the labor market than natives regardless of years of residence, but with the exception of humanitarian migrants, who are initially similar to natives but become increasingly more likely to be active on the labor market, there is little change with years of residence. We see convergence to natives and between migrant groups with regard to employment and job quality, although for many groups, gaps remain after long periods of stay in the country of residence. Migrants seeking protection generally experience the largest initial gaps, but experience a relatively rapid increase in labor market integration, consistent with the literature (e.g., Cortes 2004; Cangiano 2015). Employment gaps are largest initially for noneconomic migrants, but in terms of occupational status, the most disadvantaged group are economic migrants without a contract who experience large gaps and do not catch up to other groups. Among men, there is no real growth over time in occupational status for economic migrants, but their probability of working on an involuntary nonstandard contract does decrease over time. Migrants who arrived as students are the only group estimated to obtain parity with natives in terms of occupational status over time. Explaining Trajectories over Time Table 3 shows the marginal effect of an extra year of residence estimated at the grand margin at five years of residence, before and after including host country human capital in the regression models as described by equations 1 and 2, respectively, in the Data and Methods section. The percentage reduction in the effect after including host country human capital is also shown. Full coefficients for these models can be found in Supplemental Table S1. Activity gaps decrease by 1 p.p. for an extra year of residence among migrant women who came for economic reasons with a contract, but increase significantly (p < 0.1) for male and female protection migrants and female family migrants. After including information on migrants equivalent qualifications, language skills, and nationality, these effects are no longer statistically significant (p < 0.1) and reduced by a quarter for female family migrants and over half for the migrants seeking protection. This means the increased labor market participation of noneconomic migrants is largely explained by their higher host country human capital. Employment integration also improves most rapidly for migrants who arrived for family reasons, by around half a p.p. per year, and those seeking protection, by 1 to 1.5 percentage points. Accounting for host country human capital renders the convergence for protection migrants statistically insignificant and reduces the improvement over time for family migrants by 14 to 18 percent. All female migrants and male noneconomic migrants work on jobs with occupational status closer to that of their native counterparts as they have been in the country longer. Acquired host country human capital only accounts for a small part

18 Table 3. Effect of Extra Years of Residence Before and After Including Host Country Human Capital, at Five Years of Residence. Active Employed Occupational status Bad job M1 M2 Change (%) M1 M2 Change (%) M1 M2 Change (%) M1 M2 Change (%) Women Economic þ contract Economic no contract 1.00 (0.23)*** 0.99 (0.22)*** (0.22) 0.19 (0.22) (0.20)* 0.31 (0.18)* (0.34) 0.22 (0.32) (0.27) 0.36 (0.29) (0.21) 0.06 (0.22) (0.14)*** 0.35 (0.13)*** (0.29)** 0.53 (0.26)** 9 Study 0.05 (0.38) 0.26 (0.35) (0.38) 0.29 (0.36) (0.28)*** 0.87 (0.27)*** (0.65)** 1.22 (0.60)** 17 Protection 1.52 (0.55)*** 0.68 (0.75) (0.73)** 1.23 (0.76) (0.24)* 0.06 (0.29) (0.70)** 0.92 (0.64) 37 Family 0.32 (0.17)* 0.24 (0.17) (0.13)*** 0.35 (0.13)*** (0.10)*** 0.30 (0.09)*** (0.21)*** 0.53 (0.22)** 14 Other 0.16 (0.29) 0.06 (0.32) (0.38) 0.18 (0.37) (0.19)* 0.14 (0.18) (0.47) 0.29 (0.51) 20 Men Economic þ contract Economic no contract 0.14 (0.19) 0.17 (0.18) (0.19) 0.29 (0.20) (0.12) 0.13 (0.10) (0.27) 0.48 (0.27)* (0.15) 0.05 (0.15) (0.18) 0.13 (0.18) (0.08) 0.01 (0.08) (0.27)** 0.68 (0.29)** 2 Study 0.03 (0.36) 0.04 (0.35) (0.47) 0.71 (0.52) (0.17)** 0.21 (0.22) (0.41) 0.71 (0.49) 97 Protection 0.63 (0.31)** 0.26 (0.39) (0.43)* 0.39 (0.37) (0.23)* 0.29 (0.22) (0.53)** 1.10 (0.52)** 7 Family 0.07 (0.14) 0.17 (0.14) (0.17)*** 0.50 (0.18)*** (0.13)*** 0.40 (0.14)*** (0.21)*** 0.79 (0.19)*** 8 Other 0.20 (0.17) 0.23 (0.18) (0.21) 0.18 (0.20) (0.13)** 0.10 (0.14) (0.36) 0.07 (0.36) 166 Note: Estimated marginal effect of extra year of residence, at five years, from model controlling for sociodemographic characteristics (age, gender, marital status, education), country of residence fixed effects, and sending region characteristics (EU member state, share of Christians, female labor force participation, share of graduates, log of ratio Gini and log of ratio GDP) in M1. M2 adds host country human capital (language, equivalent qualifications, and naturalization); table also shows percentage difference in the point estimate of M2 compared to M1. *p < 0.1. **p < ***p <

19 Zwysen 19 of this effect for economic migrants, but almost all of the change over time for female humanitarian migrants. After including host country human capital, the effects for migrants who sought protection or arrived for other reasons, and male student migrants, are no longer statistically significant (p < 0.1). Migrants are generally more likely to work on involuntary nonstandard contracts, but this penalty decreases over time for all groups with the exception of economic migrants with a contract, those with other motivations, and male student migrants. Including host country human capital accounts for very little of these changes over time, however, with the exception of female refugee migrants where it accounts for over a third of the effect, and there is no significant change over time after including host country human capital. To summarize, we find substantial improvements in labor market outcomes for some groups at five years of residence, particularly for groups that are relatively disadvantaged, namely migrants seeking protection and family migrants. Accounting for three measures of host country human capital explains away substantial parts of these improvements in activity, employment and occupational status, particularly for family and humanitarian migrants. There is less change over time for the more advantaged groups such as economic migrants with a contract and acquired host country human capital accounts for less of it. There are generally fewer differences between groups in the trajectories once accounting for host country human capital, supporting our first hypothesis. This mechanism seems to be particularly relevant with regard to occupational status. We re-estimate the effect of an extra year of residence at 15 rather than five years of residence as the change over time evens out somewhat by then, and find similar results with regard to relative effect sizes and the reduction due to host country human capital (shown in Supplemental Table S2). Effects of Host Country Human Capital The effects of host country human capital on labor market integration are shown in Table 4. Migrants who report their language skills as adequate work on higherquality jobs on average, although this effect is not statistically significant for male protection and female student migrants. Language skills are also important in increasing the labor market participation and the employment rate of more vulnerable groups, such as family and protection migrants. Having obtained qualifications in the country of residence or taken steps to equivalize foreign qualifications is generally associated with working on higherquality jobs, although not statistically significantly so (p <0.1) for male student and family migrants. There is no such positive effect on the probability of working on involuntary nonstandard work, however, and for economic migrants without a contract, it actually increases the probability of these types of contract, although this is only statistically significant for women. Having equivalent qualifications is associated with lower relative employment, although this is only statistically significant for women migrants arriving with a contract. This could mean that migrants with