Gendered integration? How recently arrived male and female refugees fare on the German labour market

Objective: This paper examines gender differences in the labour market integration of newly arrived refugees in Germany. In particular, we focus on the heterogeneity in employment rates among female refugees. Background: Previous research has demonstrated that refugee women are disadvantaged on the labour market not only compared to their male counterparts, but also compared to other immigrant women. So far, however, little is known about the mechanisms that underlie the specific disadvantages of refugee women. Method: Using data from the IAB-BAMF-SOEP Survey of Refugees, we analyse the labour market participation of refugees who migrated to Germany between 2013 and 2017. To test our theoretical assumptions, we apply logistic regressions. Results: Our results show that compositional differences in qualifications, family structure, institutional support, and networks can partly explain the gender gap in labour force participation that disadvantage refugee women. We find substantial variation in the importance of different determinants for the labour market outcomes of men and women. Conclusion: As the gender gaps in labour supply could be fully attributed to the theoretical mechanisms offered in the literature, further research is needed to disentangle female refugees’ employment behaviour.


Introduction
The influx of refugees who reached Germany in 2015 and 2016 stands out in comparison to the numbers entering other European countries, but also with respect to Germany's history since the end of World War II (Brücker et al. 2020b: 28). The entry of around 1.2 million refugees in these two years has sparked conflicting political and societal discourses in Germany. While some observers interpreted the arrival of these refugees as an economic opportunity given the (until then) growing German economy and shortage of skilled labour, others labelled the influx sceptically as a "refugee crisis" (cf. Karakayali 2018;Brücker et al. 2020b). Refugees' successful integration into the German labour market became one of the main hallmarks for both perspectives by making use of the refugee's human capital and handling the 'crisis' successfully. Moreover, in recent years, Germany has invested considerable resources into the integration of refugees by, for example, accelerating the asylum procedure (Kosyakova & Brenzel 2020); abolishing several politically mandated bureaucratic barriers that had previously prevented refugees from entering the labour market (Etzold 2017); and by instituting new forms of targeted training initiatives, like public language courses and labour market programs (Brücker et al. 2020b: 42). From the perspective of the refugees, participation in the labour market is essential not only to secure their livelihoods, but also, more recently, to improve their prospects of being granted permanent residence.
The discussions on this topic rarely pursue the gender-relevant aspects of the integration process, even though there is considerable evidence that there are gender gaps on the German labour market, not only for immigrants seeking to integrate (Salikutluk et al. 2016), but for men and women overall Buchholz et al. 2015;. The results of both quantitative and qualitative research have shown that there is a gendered process of integration for male and female refugees that disadvantages women (Brücker et al. 2020a, Krämer & Scherschel 2019. For example, it has been reported that female refugees are less likely than male refugees to participate in education and vocational training (Brücker et al. 2020a), and that highly skilled female refugees remain invisible as a special labour administration target group (Krämer & Scherschel 2019: 186). During the five-year period after they entered Germany, less than one-third of female refugees (29 per cent), compared to 58 per cent of male refugees, managed to find employment (Brücker et al. 2020a). While female refugees are less likely than male refugees to participate in the labour market in other OECD countries as well, the international literature has stressed that refugee women face multiple forms of discrimination and disadvantages throughout the asylum and integration processes (Liebig & Tronstad 2018;Pittaway & Bartolomei 2001;Weichselbaumer 2020;Koyama 2015;Jubany 2007). For Germany, little is known about the mechanisms driving the specific disadvantages of refugee women in light of the heterogeneity of this group. Thus, our aim in this paper is to discuss and analyse potential explanations for the gender gap in labour market integration, as measured by labour market activity among recently arrived refugees in Germany. We do so by taking into account the different socio-demographic characteristics of refugees, like their education, work experience, and care responsibilities, as well as whether they wear a hijab as a religious symbol. The latest quantitative literature has shown repeatedly that the participation of refugee women in the labour market is often hindered -or is at least complicated -by their care responsibilities (Brücker et al. 2020a;Brücker et al. 2020b;Fendel 2019). While some aspects of the gendered process of integration are indeed attributable to refugee women spending their time on unpaid care work within the family, others remain unexplained. Thus, an analysis that differentiates between female refugees, and that focuses on mechanisms other than care responsibilities, is still missing. This article addresses this research gap by summarising the findings of existing qualitative and quantitative research on this topic, and by discussing the theoretical arguments for why female refugees are less likely than male refugees to participate in the labour market. We investigate these theoretical explanations using representative quantitative data on refugees provided by the IAB-BAMF-SOEP Survey of Refugees. These data include information specific to the refugees' situations before and after migration. Furthermore, the 2017 sample enables us to examine the question of whether refugee women who wear a hijab differ in their labour marketrelevant characteristics and labour force participation from women who do not wear a hijab.
In the next section, we provide a brief description of the latest political decisions connecting asylum to labour market policies within the German welfare state, which determine the options asylum seekers in Germany have for accessing the labour market. Second, we present the current state of quantitative and qualitative literature regarding the labour market participation of female and male refugees in order to point out existing research gaps, and to formulate our hypotheses. In the subsequent section, we introduce the data and methods we use, and present our results. We then investigate the labour market activity with logistic regressions. The article concludes with a discussion of our main findings.

A new intersection of asylum and labour market policies -the case of Germany
Integration refers to "the process of settlement of newcomers in a given society, to the interaction of these newcomers with the host society, and to the social change that follows immigration" (Penninx 2019: 5). This definition is associated with a two-way process shaped by the newcomers and the receiving society. Recent studies have pointed out that in recent decades, the outcomes of efforts to integrate migrants have varied due to changes in national labour markets and governmental policies and institutions (among others). These authors have called for a stronger distinction to be made between analysing integration processes and analysing integration policies (Penninx 2019;Gisselquist 2020).
In line with theoretical arguments that (flight) migration should not be detached from the sociological mainstream of migration and integration research, and should instead be reembedded into a more general understanding of contemporary society (Castles 2010;Kalter & Kogan 2020), we link our analysis of the integration of refugees to more general mechanisms within the German welfare state, and thus show that the current integration policies in Germany have been shaped by the intersection of asylum and labour market policies.
benefits are excluded from active employment promotion services due to the enforcement of family subsidiarity principles (Betzelt 2015) and discrimination against immigrant women (Jaehrling & Knuth 2010). The activation policies of the German welfare state also apply to refugees. Economic and social integration into society is generally equated with participation in the labour market (Bakoben et al. 2020). To secure the legal status of "refugee" under international and national rules, a person needs to demonstrate that s/he needs protection. Nonetheless, labour market performance and participation in paid employment as guiding principles for granting the right to remain have found their way into refugee policies in Germany, and have established a new intersection of asylum and labour market policy that has been called a "meritocratic turn" (Schammann 2017: 741;Maroufi 2017). For example, new forms of legal status have been established for refugees who are able to enrol in vocational training. The so-called "Ausbildungsduldung" under paragraph 60c of the German Residence Act grants refugees enrolled in vocational training a limited legal status for three years, and for an additional two years if they are able to find employment after successfully completing the vocational training. Thus, refugees' labour market access and their options for prolonging their existing legal status are subjected to intersectional inequalities that tend to discriminate against women (Menke 2020;Krämer & Scherschel 2019;Scherschel 2016), refugees with low levels of education, and refugees with few economic resources (Maroufi 2017;Farolfi 2016). These forms of discrimination also apply to the family members of recognised refugees (mostly women) who came to Germany via the right of family reunification (for more on the role of marriage migration in women's entry into the German labour market, see Samper and Kreyenfeld in this volume).
The intersection of asylum and labour market policies outlined above suggests that female refugees are likely to face specific gendered and migration-related barriers when trying to participate in the German labour market. In the following section, we provide evidence supporting this assumption by exploring the current state of the literature on the labour market outcomes of female refugees, and by offering theoretical explanations for the gender gap in labour market outcomes among refugees.

Gendered integration of refugees in Germany: What do we know so far?
Across the OECD countries, refugee women experience specific disadvantages. Compared to their male counterparts and other immigrant women, they have the lowest labour force participation rate, are at higher risk of being unemployed, are more likely to work parttime, and are more likely to be overqualified for their job (Dumont et al. 2016;Liebig & Tronstad 2018;Salikutluk et al. 2016). Studies on refugees' efforts to access the labour market have underlined the importance of refugees having personal pro-active agency, making use of social capital, and acquiring post-migration human capital (Verwiebe et al. 2019(Verwiebe et al. : 1406. However, compared to their male counterparts, female refugees in Germany are less likely to enrol in educational institutions like general or vocational schools or universities (Brücker et al. 2020a), and tend to have fewer social contacts with Germans and with individuals from their country of origin (Fendel & Yildiz 2020). At first sight, these findings seem to confirm the narratives about female refugees that are prevalent in the receiving countries: namely, that they have a traditional gender role orientation, and thus prefer to focus on performing care work within the family, rather than on participating in the labour market or obtaining qualifications that are relevant for the labour market (Lokot 2018). The empirical picture is, however, in stark contrast to the stereotypes about refugee women, as they show a high propensity to engage in paid work.
In the IAB-BAMF-SOEP Survey of Refugees, 86 per cent of women indicate that they are "likely" or "certain" to take up employment (Fendel 2019). The high employment inclination of female refugees in Germany is confirmed not only in the quantitative literature, but also in the qualitative literature (Author forthcoming; Krämer & Scherschel 2019;Pallmann et al. 2019;Mußinghoff 2017). In the following sections, we will discuss potential explanations for the discrepancy between female refugees' aspiration to engage in paid work and their poor labour market access by considering individual as well as structural factors.

Qualifications acquired abroad and in the country of residence
Most authors have attached great importance to human capital theories in seeking to explain immigrants' disadvantages on the labour markets of their host countries (e.g., Chiswick et al. 2005;Granato & Kalter 2001). According to these scholars, immigrants who have invested more in their education (e.g., a long period of schooling, high formal qualifications, specific training) have higher probabilities of becoming employed (Becker 1964). Later on, the (rather deterministic) argument that a person's labour market opportunities are primarily (if not exclusively) defined by her/his accumulated human capital has been linked to aspects of the gendered division of labour (Becker 1985). Additionally, human capital is not easily transferable across countries. Migration is accompanied by an initial devaluation of human capital (Chiswick et al. 2005;Friedberg 2000). Segmentation theories divide labour markets into at least two (or more) segments, and take into consideration different working conditions in different sectors that are related to each other. For example, while Piore (1979) argued that precarious jobs stabilise standardised, insurable employment; Castles & Kosack (1997) countered that the former constitute a constant threat to the latter's conditions and wages. Feminist revisions of these theories have pointed to their inherent blind spot: i.e., that they ignore feminised reproductive labour, whether in the form of unpaid labour or service sector work (Neuhauser 2019). Compared to male refugees, female refugees do, on average, have less schooling, and they are more likely to have no schooling at all. However, when women with no degree are excluded from such comparisons, they generally have similar qualification levels in terms of completed vocational training and tertiary education (Brücker et al. 2020a). Notable gender differences in work experience before their arrival in the host country can also be observed: i.e., 75% of male refugees, compared to 37% of the female refugees, report having worked before migration (Brücker et al. 2020a). Strikingly, no statistically significant relationship between women's work experience before fleeing their country of origin and their employment probabilities in Germany has been found (Brücker et al. 2020a). Qualitative studies indicate that university-educated women are the most likely to participate in the German labour market, but often work in jobs that are below their qualification levels. These findings point to the high degree of mediation into low-skilled work (Krämer & Scherschel 2019;Pallmann et al. 2019), in line with segmentation theories. We test how differences in qualifications are linked to labour market participation. Our measure of labour market participation differentiates between individuals who are either in paid work or are looking for a job, and those who are not willing to take up paid work. We assume that female refugees' lower average levels of education and of labour market experience relative to those of male refugees are negatively associated with their labour market outcomes. Hence, we assume that refugee women's lower levels of labour force participation are partly due to their lower human capital endowments (H1a).
The literature clearly shows the importance of language proficiency for the successful integration of immigrants into the receiving country (e.g., Dustmann & Fabbri 2003;Dustmann & Van Soest 2002). Being fluent in the receiving country's language can help refugees employ their existing labour market experience and qualifications. As we described above, several measures have been introduced that encourage refugees to enrol in language and integration courses. Attending an integration, language, or vocational preparation course is often the first opportunity a refugee has to acquire language skills, and is thus highly relevant for the individual's transition into the labour market. Female refugees in particular tend to take advantage of such institutional support in Germany. After attending such a course, their probability of becoming employed increases significantly (from a low level) (Brücker et al. 2020a: 12). At the same time, women are less likely than men to participate in these courses, especially if they have children. The lower participation rate of women is usually explained by their care responsibilities, and the lack of child care services that would allow them to attend courses (Brücker et al. 2020a;Croisier et al. 2019;Brücker et al. 2020b;Kosyakova & Brenzel 2017). As women are the main providers of child care, men have more capacity to attend language and integration courses. Moreover, men may attach greater importance to learning the host country's language quickly to improve their chances of entering the labour market (van Tubergen 2010). As a consequence, female refugees with and without children generally report having lower (self-estimated) German language skills than male refugees (de Paiva Lareiro et al. 2020). Although attending a language course can delay their access to the labour market, completing a language course can positively affect their probability of being employed (Kosyakova & Brenzel 2020). In light of previous research that showed the positive effects of finishing a language course and of language proficiency on immigrants' labour market integration, we assume that part of the gender gap in labour force participation among refugees can be explained by gender differences in language course participation and language proficiency (H1b).

Gender roles
How gender shapes migration and integration processes is discussed broadly within the migration literature (for an overview, see Schrover & Moloney 2013), and underlines the paradoxical effects in between the ascription of gender-specific norms and behaviour and female emancipation and empowerment (Schrover 2014;Morokvasic 1993, Morokvasis 2009). How the role of gender transforms and continues to transform throughout migration and integration processes depends not only on individual behaviour, but on social policies, labour markets, and normative understandings of gender and of a migrant's country of origin within the receiving country (Calavita 2006). Research on female refugees in the United States seems to confirm these assumptions (Gowayed 2019; Koyama 2015).
Independent of their actual care responsibilities, having a traditional gender role orientation can affect women's attitudes towards paid work, and how much time they invest in obtaining the relevant qualifications and in searching for a job. A study by Khoudja & Fleischmann (2015) showed that the more traditional gender role orientations of female immigrants can explain part of the gap in labour force participation between immigrant and native women in the Netherlands. Moreover, recent literature on refugees in Germany has found that female refugees provide the majority of unpaid care work, and therefore have only limited time to participate in education and paid work (Bujard et al. 2019). However, the gender role orientations of recently arrived female refugees in Germany have hardly been explored. Hence, based on these arguments and to fill this gap, we derive two hypotheses: We assume that refugee women's lower labour force participation is partly driven by their child care responsibilities and by living with a partner (Hypothesis 2a). In addition to their actual family arrangements, we expect that having traditional gender role attitudes decreases female refugees' labour force participation (Hypothesis 2b).

Institutional support and social network
Recently arrived male and female refugees are offered a variety of institutional supports to facilitate their language acquisition and labour market integration. As well as having lower participation rates in language and integration courses than men, refugee women in Germany and other European countries tend to make less use of institutional labour market integration supports for a number of reasons. For example, if a woman who was in a relationship before she fled her country of origin arrived later than her partner in the receiving country, she may have a different legal status that prevents her from taking part in labour market programs. Moreover, she may be mainly responsible for providing child care within the family (Liebig & Tronstad 2018). We expect to find that the female refugees in our sample are less likely than male refugees to receive institutional support provided by the Employment Agency. We assume that this lack of support is negatively associated with their labour force participation (Hypothesis 3a).
Besides having institutional support, having social contacts in the receiving country plays an essential role for both recently arrived immigrants and refugees, as these contacts can provide important information that the newcomers need to settle into their new surroundings (Fendel & Yildiz 2020;Verwiebe et al. 2019). For obtaining information about the receiving country and practical aid, social contacts are beneficial, since the refugees' countries of arrival are often chosen coincidentally, or may change while they are on the move (Bitterwolf et al. 2016). Having inter-ethnic contacts in particular is typically seen as an indicator of social cohesion (van der Meer & Tolsma 2014), and is regarded as helpful for the labour market integration for recently arrived immigrants (Kalter & Kogan 2014). However, the refugees' labour market outcomes are likely to differ depending on the different types of social capital these contacts can provide; and the different forms of support the refugees receive during the labour market integration process can lead to either adequate employment, or low-skilled work and underemployment (Gericke et al. 2018). Migrants with (mainly) co-ethnic contacts tend to find employment quickly, but these jobs are often of low quality (Kogan 2016). However, quantitative analyses have shown that although most female refugees in Germany had high levels of social network support prior to their arrival (Brücker et al. 2020b), they tend to have fewer "weak ties" after their arrival. This applies to contacts with Germans, as well as with individuals from the country of origin or from other countries (Fendel & Yildiz 2020). In a recent study, Hartmann and Steinmann (2021) examined whether the traditional gender role attitudes of refugee women and their partners hamper women in establishing friendships and acquaintances. Although they found that female refugees have fewer social contacts than male refugees overall, they also reported that female refugees' language proficiency, rather than their traditional gender role attitudes, have more negative effects on their levels of contact with Germans. We assume that the gender gap in levels of contact with Germans is negatively associated with refugee women's labour market participation (Hypothesis 3b).

Disadvantages based on religion
Discussions in Germany and elsewhere in Europe about refugees have been coloured by many stereotypes and prejudices, especially about Muslim females wearing headscarves (Helbling 2014). Veiling practices have been equated with an unwillingness to integrate or to participate in paid work, or with being a threat to Western culture, particularly in terms of gender equality (e.g., Berghahn & Rostock 2015). The scientific literature regarding Muslim women and how they fare on the labour markets of Western countries has been two-sided. On the one hand, there is a strong focus on female Muslims' lack of qualifications, such as their low levels of education and language proficiency, and on their tendency to adhere to traditional gender roles by staying home to care for children instead of being active on the labour market (e.g., Koopmans 2016; Khoudja & Fleischmann 2015;Inglehart & Norris 2003). On the other hand, recent correspondence studies have detected significant discrimination against Muslim applicants. For instance, various field studies carried out in Germany have shown that there are lower call-back rates for women who have a Turkish name and who wear a hijab in the photo in the application documents than for a female applicant with a "native" name (Weichselbaumer 2016; for the US, see: Ghumman & Ryan 2013). Based on a large-scale correspondence study, Koopmans et al. (2018) concluded that employers make strong assumptions about cultural differences and incompatible values among Muslims, and thus tend to reject Muslim applicants. Additionally, qualitative research has uncovered anti-Muslim racism in the context of labour market activation programs for female Muslim refugees, including among volunteers providing refugee assistance (Menke et al. forthcoming). The expectation of Muslim women who wear a hijab that they will experience religious discrimination on the labour market leads them to have lower expectations of receiving a job offer than Muslim women who do not wear a headscarf (Ghumman & Jackson 2010).
Contrary to many studies based on survey data, the IAB-BAMF-SOEP Survey of Refugees from 2017 contains the information on whether female refugees wear a hijab. Although we are not able to test whether discrimination is one of the main drivers of the gender gap in labour market integration among refugees, we are able to explore two hypotheses for a sub-sample: Female refugees who wear a hijab are less likely to participate in the labour market than female refugees who do not wear a hijab and refugee men (Hypothesis 4a). We assume that a part of these differences can be explained by their lower levels of qualifications, resources, and contacts (Hypothesis 4b).

Data
To analyse gender differences in the labour market integration of refugees, we use data from the IAB-BAMF-SOEP Survey of Refugees (IAB-BAMF-SOEP Survey of Refugees 2020; Kühne et al., 2019). Based on the data of the German Central Register of Foreigners, refugees who arrived in Germany between January 2013 and December 2016 were identified as the target population. More concretely, the sample includes immigrants to Germany who applied for asylum, who received some type of protection (recognised refugee, subsidiary protection), or whose deportation was deferred. Details on the data collection and sampling procedures can be found in Kroh et al. (2017). The data capture information on various topics, including on the refugees' flight experiences and views on asylum procedures and integration-related measures. The data also include standard socio-economic variables that refer to the refugees' status before and after migration.
To increase the number of cases with complete information on the relevant variables and measures, we pool data from 2017 and 2018. As we are interested in labour market integration, we exclude respondents who are younger than 18 years or are older than 64 years at the date of the interview. We also exclude respondents from our analyses who have lived less than one year (N=17) in Germany and those who have lived more than six years in Germany (N=30). After applying multiple imputation to impute missing information on the variables we use in our analyses (Rubin 1987), our sample contains 5,171 cases. 2 We provide additional analyses in which we differentiate between women depending on whether they wear a hijab. As this information was collected in 2017 only, our analysis sample is reduced to 1,204 cases.

Dependent variable
Our dependent variable is labour force participation (for additional calculations based on employment status, see Appendix Table 4). Following the ILO definition, we distinguish between active (employed and unemployed) and non-active (in non-vocational education, on maternity leave, non-working) respondents. We count respondent who were in vocational education or marginally employed at the time of the interview as being active in the labour market.

Qualifications
Our analyses take into account the different types of qualifications the refugees obtained either before or after migration. Based on the ISCED-2011 categories provided in the SOEP, we capture respondents' highest educational degree on four levels: (1) primary education (ISCED 1), (2) lower secondary education (ISCED 2), (3) upper secondary education (ISCED 3), and (4) tertiary education (ISCED 4 and higher). Furthermore, we use proxy indicators for full-time and part-time labour market experience acquired before migrating to Germany. These indicators are rather rough, and do not consider interruptions, or in which country other than Germany the labour market experience was gained. However, it is plausible to assume that most of the respondents mainly took part in paid work in the country of origin.
Regarding qualifications that were acquired in Germany, we take into account the respondents' full-time and part-time labour market experience gained in Germany, their language proficiency, and whether they finished any language or integration courses. For the latter, we generated a variable that specifies whether the respondents have completed or are taking part in any language training or integration courses, such as courses offered by the Federal Office for Migration and Refugees. We differentiate between respondents (1) who have not participated in a course, (2) who are still enrolled in a course, and (3) who have already finished a course. Each year, respondents are asked to assess their skills in speaking, writing, and reading German on a five-point Likert scale. The sum score we calculated reflects higher proficiency with increasing scores.

Family structure
We combine information on whether the respondents have a partner and the partner's labour force status, which leads to three categories: no partner, the partner is employed, or the partner is not employed. To account for different care requirements, we use two indicators for the number of children by counting the number of children who are three years old or younger, and the number of children who are between ages four and 16.

Gender attitudes and gender equality
In their first interview, a set of statements were presented to the respondents to capture their attitudes related to gender and gender equality. They were asked to state how strongly they agree or disagree with the following: "Having a job is the best way for a woman to be independent"; "Even a married woman should have a paid job so that she can be financially independent"; "If a woman earns more money than her partner, this inevitably leads to problems"; "For parents, vocational training or higher education for their sons should be more important than vocational training or higher education for their daughters"; and "At home, the husband should have the final say". Using Principal Component Factor Analysis, we predict factor scores for a component of gender attitudes that explains 65 per cent of the variation. The lower the values, the less traditional the respondents' gender attitudes are. As these items were measured only one time, we use the score for both waves. Additionally, we use the respondents' answers to the question on gender equality in their home country ("To what extent has the equal treatment of men and women been achieved in your country of origin?") based on a scale from zero (very low) to 10 (very high).

Support and social network
Our analyses include two indicators for measuring respondents' contact with natives, and whether they have received institutional support for accessing the labour market. They were asked whether they are informed about the consultation opportunities on labour market issues by the German Federal Employment Agency. The response options were: "Yes, I have already used this service"; "Yes, but I haven't made use of it yet"; and "No, I do not know it". Additionally, respondents reported how often they have contact with Germans in their friendship network and their neighbourhood. The frequency was measured by the categories "every day", "several times per week", "every week", "every month", "less often", and "never". We combined the answers given to both questions and calculated the average contact frequency with Germans. Note that it is not possible to determine the causal direction in the relationship between the two variables and the labour force participation, as those who are motivated to work should also be more motivated to search for consultation options or for contacts with Germans in order to get information about the German labour market.

Hijab
As we mentioned above, in 2017, female refugees were asked whether they wear a hijab in public for religious reasons. Women who wear the hijab always, most of the time, or sometimes, are grouped as one category. In additional analyses based on the 2017 sample, we compare them with women who wear no hijab, and with refugee men.

Controls
In all analyses, we control for the respondents' age, age-squared, years since migration, and the country of origin. Based on our analysis sample, we can separate respondents from Syria (N=2,984), Afghanistan (N= 634), Iraq (N=719), and Eritrea (N=152). Refugees from other countries are grouped into a single category, as the numbers of cases are not sufficient to further differentiate them by their home countries (N=692). Respondents were asked about their current legal status in Germany. We combined the answers into two categories: (1) temporary resident permit ("Aufenthaltsgestattung"), suspension of deportation ("Duldung"), probationary permit ("Fiktionsbescheinigung"), and no residence status; and (2) residence permit as a recognised refugee ("anerkannter Flüchtling"), refugee status ("Flüchtlingseigenschaft"), subsidiary protection, and permanent residence permit ("Niederlassungserlaubnis"). Furthermore, we include a subjective measure of the respondents' health status. The respondents were asked indicate the state of their current health on a five-point scale: the higher the score on the scale, the lower their self-reported health status. Finally, to roughly account for regional variations in institutional settings and the labour market, we control for the federal state in which the respondents live.

Analyse strategy
First, we run one logistic regression model on labour force participation that includes all respondents. In the second step, we estimate separate models for women and men. Finally, with the subsample of 2017, we further differentiate between women who do and do not wear a hijab. Due to the low number of cases in this sub-sample, we do not estimate separate models for men, women wearing no hijab, and women wearing a hijab.
The set-up of our analyses is as follows: We start by showing the gross gender gap in labour force participation. Subsequently, we add the explanatory variables into the models. All results are weighted and controlled for the year of the survey. We use cluster robust standard errors and present our results as average marginal effects.

Descriptive differences
We start by presenting a selected descriptive overview of gender differences in the variables and the measures that we use for our analyses (Table 1). Not surprisingly, larger shares of refugee men than of refugee women are shown to be active in the labour market. We find a gender difference in labour supply of 26 percentage points. The partner variable reflects this gender pattern: only four per cent of men have an employed partner. The results further indicate that 80 per cent of refugee men, compared with 53 per cent of refugee women, have a non-working partner. Around one-third (30 per cent) of the women in our analysis sample have no partner at all.
When we look at the refugees' educational qualifications, we see that a larger share of women (45 per cent) than of men (40 per cent) did not progress beyond primary education. However, while there are men than women in the middle education categories, there is only a marginal difference in the shares of women and men with tertiary education. Women also have less working experience than men, on average. The gap is most pronounced for full-time working experience before migration, although women also reported having slightly less experience in part-time work in their home country and in Germany in general.
In line with previous literature, we find that women are less likely than men to participate in language training or take part in integration courses (28 per cent compared to nine per cent). These results reflect the shares of men and women in our analysis sample who were enrolled in a course at the time of the survey. In addition, more men than women reported that they had already completed a course. Presumably, this explains why refugee women have, on average, lower German language skills than men.
We find an advantage for male refugees in terms of both institutional support and contact with Germans. The results also show that women (42 per cent) are less informed than men (34 per cent) about consultation opportunities offered by the Employment Agency, and that women have less frequent contact with Germans in their neighbourhood and through friendships than men. Although the latter difference appears to be small, it is statistically significant.
We also want to highlight some differences found between women who do and do not wear a hijab. First, when we look at their labour market outcomes, we see that the share of hijab-wearing women who are active in the labour market is one percentage point lower than that of non-hijab-wearing women. We observe differences in the explanatory variables as well. Compared to non-hijab-wearing women, women who wear a hijab are, on average, higher educated, but have accrued less labour market experience before and after migration. Hijab-wearing women tend to have more children, on average, and more traditional gender attitudes. By contrast, non-hijab-wearing women seem to be better equipped for the labour market than hijab-wearing women: i.e., they are more likely to participate in a language or integration course, have better German skills, and have more contacts with Germans.
Overall, female refugees have less beneficial (starting) conditions than male refugees. In the next sections, we will examine how this disadvantage contributes to the abovedescribed gender differences in the labour force supply in favour of men.  Table 2 presents a set of logistic regressions on the respondents' labour force status. The first two models show the gross and the net gender gap in labour market participation. By estimating separate models for men and women (Models 3 and 4), we can examine whether the indicators operate for both genders in the same way. The first model shows that women are 29 percentage points less likely to be active in the labour market than men when we control only for the year of the survey. Model 2 demonstrates how this gap changes when we add our explanatory variables: i.e., the gender difference in the dependent variable is reduced to 21 percentage points. Hence, a large share of the gender gap seems to be produced by compositional differences in qualifications and other characteristics that are relevant for the labour market. When we test Hypothesis 1a, we find that the average marginal effects for qualifications acquired abroad are small and statistically not significant. Furthermore, refugees' labour market experience before migration is shown to have either a small and negative effect (full-time working experience) or no effect at all (part-time working experience). By contrast, in line with Hypothesis 1b, we observe that the qualifications and experience refugees acquire in Germany are important for their labour market participation: i.e., having better German language skills, being enrolled in any type of language or integration course, and having full-time working experience in Germany are found to increase the probability of labour market participation. While all of these effects point in the same direction for men and women, not all of these effects are statistically significant for women (Models 3 and 4).

Labour force participation
Turning to the family constellations, we find that for both men and women, having no partner is associated with a lower probability of labour market participation than having a non-employed partner. Similarly, for both men and women, being in a partnership with an employed person increases the likelihood of labour force participation compared to the reference category. While we find no evidence to support Hypothesis 2a, we observe, in line with Hypothesis 2b, a gender difference in the effects of having children. The average marginal effect found in Model 2 indicates that the presence in the household of each child younger than three years old decreases the probability of the parents being active on the labour market by 10 percentage points. Noticeably, this negative effect seems to be larger for women (-14 percentage points per child in Model 3) and smaller for men (-6 percentage points per child in Model 4). To capture gender role orientations apart from lived gender roles, we added the respondents' gender attitudes and assessment of the level of gender equality in their home country. It appears that neither indicator helps to explain the gender gap in labour supply. Hence, we find a statistically significant relationship only between having young children and the parents' labour supply. We find no evidence for any effects exerted by gender role orientation on (female) refugees' likelihood of participating in the labour force.
These descriptive results indicate that male and female refugees differ in the information they have about institutional support opportunities, and in the frequency of their contacts with Germans. In the pooled and both separate models, knowing and already having made use of the advisory service offered by the Employment Agency is found to be positively associated with labour market activity, which provides support for Hypothesis 3a. However, interacting with German neighbours and friends is not shown to be associated with labour force participation. Thus, we have to reject Hypothesis 3b regarding labour supply. Source: IAB-BAMF-SOEP Survey of Refugees, own calculations. Note: Pooled and imputed data for 2017 and 2018; weighted and cluster robust results. Results based on logistic regressions and reported as average marginal effects. Controls include age, age-squared, country of origin, legal status, and federal state; year of survey is controlled in all models (for full models, see: Appendix Table A2). + p<0.1, * p<0.05, ** p<0.01, *** p<0.001. Table 3 displays the results for Model 1, which considers the raw gender gap in labour supply; and for Model 2, which includes qualifications, family constellation, institutional support, networks, and the control variables. Overall, we find mixed evidence for our Hypotheses 4a and 4b. Women wearing a hijab are found to be 32 percentage points less likely than men to be active in the labour market. The results indicate that women who do not wear a hijab do not differ significantly differ from women who wear a headscarf, as they are also less likely than men to participate in the labour market. In the second model, the gaps between refugee men and women decrease. But the model also shows that women who do and do not wear a hijab still have statistically significant lower probabilities of being active in the labour market; i.e., of between 26 and 22 percentage points. We find no statistically significant differences between the two groups of women. The differences between these two groups decrease in the second model when we add our explanatory and control variables. The gap found between men and women who wear a hijab seems to be partly attributable to compositional differences in the explanatory variables. However, these compositional differences cannot close the gender gap in labour supply, regardless of whether women do or do not wear a hijab. Note: Imputed data for 2017; weighted and cluster robust results. Results based on logistic regressions and reported as average marginal effects. In Model 2, we control for education, labour market experience, partner status, number of children younger than three years old, number of children between three and 16 years old, German proficiency, course participation, gender attitudes, gender equality, consultation, contact with Germans, age, age-squared, country of origin, legal status, and federal state; year of survey is controlled in all models (for full models, see: Appendix Table A3). + p<0.1, * p<0.05, ** p<0.01, *** p<0.001.

Discussion
The issue of the labour market integration of refugees has been high on the political agenda in Germany in recent years. Connected to these discussions about the integration of refugees are debates about the willingness of female refugees to engage in paid work that are often rooted in stereotypical assumptions about the prevalence of traditional gender roles in immigrant communities that keep women from participating in the labour market.
In this paper, we addressed these gender-related issues using data from the IAB-BAMF-SOEP Survey of Refugees. Specifically, we analysed how the labour market participation of female refugees compared to that of male refugees, and the extent to which the gender gap in labour market participation among refugees can be explained by compositional differences in their qualifications and other resources that are relevant for the German labour market. While most of the current quantitative literature has attributed these gender gaps in participation rates among refugees primarily to women having care responsibilities and different resources than men, this article takes into consideration the heterogeneity of female refugees by, for instance, considering the number of (young) children they have or whether they wear a hijab.
Are female refugees less willing to be active in the labour market? Descriptive findings from a previous study (Brücker et al. 2016) reported that over 80 per cent of the female refugees interviewed for the first sample of the IAB-BAMF-SOEP Survey of Refugees expressed the ambition to take up employment in the future. However, our multiple regressions show that the probability of being active in the labour market is around 30 percentage points lower for women than for men. We assumed that women's lower labour supply is associated with their lower endowments in terms of their educational and vocational qualifications. The results show that for both men and women, qualifications obtained abroad have no effect on their labour force participation in Germany. This could be due to the failure to recognise refugees' formal and informal qualifications. Our findings also indicate, however, that labour market experience in Germany and language course attendance are positively linked with labour force participation. We found that the shares of female refugees who have enrolled in and completed a language course, and who have labour market experience in Germany, are not only smaller than those of male refugees, but that women who have newly acquired qualifications profit from them less than men.
For our hypotheses on gender roles, we found support only for the impact of practised gender roles. Having young children was found to decrease the labour market participation of women. Although traditional gender conceptions are often given as a reason for refugee women's low labour market orientation in the literature (e.g., Lokot 2018), we found no association between their gender-related attitudes and their labour supply.
In our analyses, we also tested the relationship between refugees' levels of institutional support and contact with Germans and their willingness to participate in the labour market. As expected, we found that both men and women seemed to benefit from institutional support provided by the Federal Employment Agency. However, we were unable to find any link between spending time with Germans and women's labour market activity.
Finally, we estimated our models by further differentiating between female refugees based on information about whether women wear a hijab. To summarise the main results, we found that women who wear a hijab do not differ from other female refugees in their labour market activity. When we took their qualifications, resources, and family structure into account, the gap among women, but also between female refugees who wear a hijab and men, decreased, but the gap between men and both groups of women was still statistically significant.
These results suggest that to increase their labour market integration, refugee families need better access to existing family policy instruments and infrastructure. Up to now, refugees have only have limited access to many forms of public support. For example, in most of the federal states in Germany, access to parental leave schemes and to full-day child care facilities depends on the employment status of both parents (Menke & Klammer 2017). As previous research has shown, refugee women have less contact with Germans (Fendel & Yildiz 2020). This lack of social contact can have unfavourable effects on women's labour market integration, even if they aspire to start working in a paid job.
When we examined the validity of common stereotypes that female refugees prefer not to participate in the labour market because they adhere to traditional gender roles, we found no clear evidence in our analyses that this was the case. While having to provide care for young children was found to explain in part why the labour market participation of female refugees in Germany is lower, their gender role orientation was not. We were able to test whether women who wear a hijab are less equipped for and less likely to participate in the labour market than women who do not. The results showed that although these women have, on average, slightly lower levels of educational and vocational qualifications and less experience in paid work, their labour force participation does not differ markedly from that of other refugee women.
Some important limitations should be noted. First, the sample size is quite small, as only a small share of the refugees surveyed were already working at the time of the interview. This issue particularly applies to our sub-sample for the interviews conducted in 2017 (as well as supplementary investigations on employment rates, see Appendix Table A4). To our knowledge, no other quantitative study has investigated whether wearing a hijab affects the labour market outcomes of female refugees. Thus, although our results have to be interpreted with caution, the analyses provide novel insights into female immigrants' levels of integration in Germany depending on their religious practices. Second, with the available data, it was not possible to determine the mechanisms that underlie these effects. For example, we were unable to disentangle whether the rates of labour force participation are lower among refugee women because men and women use different job search strategies, or because employers are less likely to hire them. Third, at the moment, the share of refugee women who are employed is relatively small. Although we estimated models on employment probabilities (which can be found in Appendix A4 3 ), it was not possible to analyse gender differences in job characteristics among refugees. Thus, this issue should be tackled by future studies. Finally, as the results are based on pooled cross-sectional analyses, no causal interpretation can be drawn from this study, and reverse causality might be a problem in some cases. For instance, respondents who are employed might be more likely to meet Germans at work, and are able to improve their German skills by spending leisure time with them.
It is important to relate these quantitative findings to broader societal contexts, such as the interplay of asylum and labour market policies in Germany, as doing so can shed light on the societal and political meaning of participation in paid work. In the past, the stagnation in the integration process for refugee women was an issue that received little research attention. When having access to and success in the labour market become increasingly critical for securing legal status, intersectional inequalities increase. Additionally, a question that has yet to be answered and urgently needs to be addressed in future research is why female refugees benefit less than male refugees from their labour market experience and educational qualifications. These gender differences also apply to other areas of integration, like social integration, which, in turn, greatly affects the structural integration of female refugees. Further qualitative and quantitative research is needed to investigate the specific hurdles refugee women face.  Source: IAB-BAMF-SOEP Survey of Refugees, own calculations. Note: The dependent variable differentiates between refugees who are employed (1) or who are actively searching for a job (0). Pooled and imputed data for 2017 and 2018; weighted and cluster robust results. Results based on logistic regressions and reported as average marginal effects. Year of the survey is controlled in all models. + p<0.1 *, p<0.05, ** p<0.01, *** p<0.001.