The Process of Labor Market Integration of Male Syrian Nationals

1 Introduction

In May 2018, the statistics of Germany’s ‘basic security benefits scheme’ (Grundsicherung für Arbeitssuchende), recorded a peak of approximately 602,000 Syrian nationals being eligible for full benefits^[1] (Statistik der Bundesagentur für Arbeit [BA Statistik] n.d.– a). This number mirrors that, shortly after their arrival, the majority of refugees holding a Syrian nationality status cannot earn a living (cf., Bähr et al. 2017). However, this does not necessarily have to be the case in the medium turn. Several significant obstacles to joining the labor force, such as language barriers or lack of domestic certificates that precipitate the need for welfare support, might persist only temporarily if educational programs are successfully passed (Bähr et al. 2019; Dietrich et al. 2019; Damelang & Kosyakova 2020).

For Germany, survey data reveals that the employment rate of Syrian born refugees aged 18 to 64 reaches 61 % seven years after immigration (73 % for men) while it reaches 6 % when looking at those living in Germany for a year at most (8 % for men) (Brücker et al. 2024). Evidence from the US (Cortes 2004) suggests that refugees may even outpace economic immigrants in wages and working hours in the medium turn and Brell et al. (2020) report a general pattern of humanitarian immigrants assimilating toward the higher employment share of other migrant groups in their international comparison of industrial host countries – the upraise is often particularly strong in the first years after arrival, which highlights the relevance of this early phase (ibid.; Cortes 2004).

These numbers indicate that temporal development plays a detrimental role. Starting from a similar state of job seeking and welfare receipt, paths diverge over time and lead to heterogeneous labor market outcomes. Based on the administrative dataset Sample of Integrated Welfare Benefit Biographies (SIWBB), this paper describes early labor market activities of male refugees holding Syrian nationality status for around three years after entering the basic security benefits scheme at ages 18 to 60. It tries to shed light on the process of labor market integration and how it is linked to individual and contextual characteristics. To do so, relevant states (education, full-time employment, part-time employment, mini job, unavailability in welfare, activity in welfare, unemployment, residual) are identified for each of 36 consecutive months. At the person level, the monthly states are chronologically merged to form individual sequences of activities. Based on their similarity, the sequences are then grouped into clusters. Hence, the overall picture of how different states develop over time is split up to reveal the typical careers that constitute it. Finally, multinomial regression is applied to estimate the effect of characteristics that, given theoretical arguments, are assumed to be responsible factors for shaping careers.

Since existing data on Syrian-born refugees reveal fundamental differences between women and men with regard to their employment situation (Brücker et al. 2024), a joint analysis of the two groups that does not pay tribute to such throughout the complete analysis (theory-making, model building, interpretation of the results) would undermine meaningful interactions in this context. In fact, an analysis of female refugees entering the basic security benefits scheme shows that the number of women who take up employment in around the next three years is very low. The restriction of the sample to males hence serves as a way of reducing complexity alongside avoiding specific data-related problems (see Appendix A1 for more details as well as empirical findings for women).

The study mainly contributes to the literature in three ways. First, most previous studies on recent refugees’ integration (e.g., Brücker et al. 2020) focus on developing the share of a relevant outcome, e.g., the employment rate, over time. Others concentrate on specific transitions (e.g., transition to the first job or first language class, Kosyakova & Brenzel 2020). In contrast, the current analysis simultaneously refers to the entirety of experienced states, their temporal order, and durations (Scherer & Brüderl 2010; Kogan & Weißmann 2013). While much more information is exploited, the clustering of sequences into distinct types of generic labor market careers simultaneously provides a comprehensive picture of the dominant patterns in early integration. In addition, by evaluating how individuals are distributed across the clusters, it is also possible to assess the quantitative importance of each of the typical careers (Bruckmeier et al. 2020b).

Second, the focus on labor market trajectories also acknowledges that it may not only be of interest which state a person achieved at a specific point in time but also what route led to that state (Karhula et al. 2019). By studying typical labor market careers, it is hence possible to assess in what way trajectories that lead to more successful labor market outcomes in the longer run (e.g., unemployment vs. full-time employment) also systematically differ at earlier phases (e.g., earlier participation in language classes). This deepens the understanding of the process of labor market integration by allowing for a more nuanced description how refugees’ careers develop over time (cf., ibid.; Kogan & Weißmann 2013).

Third, this paper aims to test theoretical claims on the responsible factors shaping specific career trajectories. As shown by Bruckmeier et al. (2020b) who analyze pathways of basic security benefit recipients, specific subgroups are disproportionately associated with distinct trajectory clusters (e.g., low educated people are disproportionately prevalent in the cluster representing long-time dependency). At the same time, certain groups may need to be addressed differently by policymakers to support them optimally (e.g., older unemployed individuals versus parents in part-time jobs). Hence, the present study may help to understand which characteristics are associated with a particularly high risk for a disadvantageous or advantageous development (ibid.).

This paper proceeds as follows: first, the theoretical framework and previous empirical findings are presented and the hypotheses to be tested are formulated. Thereafter, information on the dataset and the research methodology is provided. Empirical findings are described and the regression results are interpreted. The paper concludes by summarizing and critically discussing the results generated.

2 Theoretical Framework and Prior Empirical Findings

Following previous work (FitzGerald & Arar 2018; Kogan & Kalter 2020), the present analysis frames the explanation of recent refugees’ integration patterns as “a special case of general mechanisms” (Kogan & Kalter 2020). This assessment implies that the processes relevant to the labor market integration of the group under study are assumed to follow the same regularities as observed for other newly arrived immigrant populations. However, it is also acknowledged that there are several parameters that, on the one hand, have shown to be of importance for early integration and, on the other hand, are, on average, pronounced differently among recent refugees compared to other foreign-borns (ibid.).

For example, human capital theory suggests “that the migrant’s goal is to maximize utility by choosing the location that offers the highest net return to human capital” (Bodvarsson et al. 2015: 9). However, in the case of refugees, immigrants can be assumed to face considerably more restrictions selecting their destination country. They have fewer possibilities to prepare, e.g., by learning the relevant foreign language or evaluating the fit of their personal qualifications to the requirements of a host country’s labor market (Brell et al. 2020). As a consequence, recent refugees under study “are less likely to have immediately transferable skills” (Kogan & Kalter 2020: 9).

This impedes initial labor market entries. Empirically, only 8 % of male refugees born in Syria aged 18 to 64 are in paid employment or self-employed when evaluating their labor market integration in the first twelve months after their arrival. However, this proportion increases to 25 % in year two, to 42 % in year three and to 48 % in year four (Brücker et al. 2024). These numbers suggest that taking up employment often requires preceding phases of investments, particularly in language skills which highlights the presence of path dependencies (cf., Bernhard & Röhrer 2020).

These considerations motivate the hypothesis that shortly after entering the welfare scheme, employment only plays a minor role (H1).

Survey results collected in 2017 reveal that 95 % of the Jobcenter staff supervising refugees find that a lack of language skills are a barrier to taking up employment (Dietz et al. 2018). In fact, almost all refugees of the cohort under study share the attribute of having entered Germany with no command of German, a characteristic that distinguishes them from other groups of immigrants (Kosyakova et al. 2021). For many Syrian refugees, language acquisition is synchronized with being a client at the Jobcenter, the organization administering basic security benefits (cf., BA Statistik 2020; Bernhard & Röhrer 2020; Knuth 2024).^[2] Most Syrian applicants for asylum of the cohort under study receive a protection status^[3] and hence fulfill an important criterion required to be eligible for the receipt of basic security benefits (Bähr et al. 2017; Weiser 2017). The scheme aims at covering individuals’ costs for basic needs as well as enabling them to cover the costs on their own by reintegrating them to the labor market. Recipients are confronted with a policy containing elements of ‘support’ (Fördern) but also elements of ‘demand’ (Fordern) (Eichhorst et al. 2010). “On the one hand, welfare recipients can be assisted to take up employment (e.g., through active labor market policies [ALMPs], job placement services). On the other hand, benefit recipients have to make certain efforts to end (or reduce) their level of need (e.g., participate in ALMPs, actively search for a job or to accept any job)” (Dummert et al. 2024, parentheses in the original). This also holds for participation in the BAMF integration course, an important measure to support language acquisition as also indicated by the number of 160 000 first-time participants holding a Syrian nationality status in 2016 alone (Bundesamt für Migration und Flüchtlinge [BAMF] 2017; Tissot & Zimmer 2021).

This motivates the hypothesis that shortly after entering the welfare scheme, (language) training in the welfare system increasingly becomes the dominating activity (H2).

Often starting from a similar position as an unemployed asylum seeker with no command of German, differences in personal resources and individual opportunity structure may cause career paths to diverge over time (cf., Kogan & Kalter 2020). In this context, the mere local availability of spots in language courses may play a role. With regard to access to language training, differences in the local supply of language classes may represent disparate opportunity structures affecting individual integration trajectories. The local density of public language classes offered varies enormously by region, with more than 4,500 courses starting in North Rhine-Westphalia in 2017 compared to less than 200 in Mecklenburg-Western Pomerania (BAMF 2020; Kanas & Kosyakova 2023). A great distance to the nearest course provider can impede participation and may become even more relevant with regard to less frequently offered course types, such as alphabetization classes (Tissot & Zimmer 2021; Rösch et al. 2020). Similarly, Kanas and Kosyakova (2023) hint at local disparities following their evaluation of the number of course starts and the number of vouchers offered for these courses: despite the fact that between 2015 and 2016 “the increase in vouchers issued more than doubled in 165 counties, the supply of language courses doubled in less than half of these counties” (ibid.: 10).

Based on the considerations on the role of language learning in public courses shortly after entering the welfare scheme, the question arises how this phase may be related to labor market outcomes in the medium and long-term. Generally, Kanas and Kosyakova (2023) find a positive effect of local provision on the likelihood of becoming employed. It also “significantly affects further intermediary outcomes, such as refugees’ language proficiency (though the effect is modest), language course completion and certification, but only in the initial periods since arrival” (ibid.: 26, parenthesis in the original). Similarly, a recent paper of Marbach et al. (2025) studying refugees aged 18 to 35 at arrival who have a Syrian, Eritrean, Iraqi and Iranian nationality status shows that participation in the public BAMF integration course increased employment by between 4.4 and 4.8 percentage points one year after course start depending on the estimation method (cf., ibid.).

Overall, it is hence assumed that a low local supply of BAMF integration course spots is negatively associated with trajectories that are characterized by early (language) training and subsequent transitions into full-time employment (H3).

From a theoretical perspective, a lack of formal qualification can correspond to a lack of efficiency in language acquisition (Kristen & Seuring 2021). For example, a poorly educated person who first needs to become proficient in reading the Latin alphabet may represent a very different type of learner compared to a Syrian academic speaking multiple languages. For foreign-borns lacking formal education, studying a foreign language may be associated with a limited prospect of success (cf., Bernhard & Röhrer 2020). This highlights that formal education is an important individual resource. Empirical data reveals that the group of well-educated refugees is substantial. In their study, Kristen et al. (2020) report that 26 % possess degrees corresponding to tertiary education. Empirical data gained in qualitative research also suggests that educational attainment affects individual pathways: while poorly educated Syrian refugees would tend to remain passive when supervised by a Jobcenter caseworker, educated Syrian refugees would try to actively push their careers, e.g., by vying for a spot in their preferred language class (Bernhard & Röhrer 2020).

However, the role of education goes beyond its impact on language acquisition. While a large share of the refugee cohort under study have acquired vocational experience abroad, only a minority holds a corresponding formal certificate (Liebau & Salikutluk 2016). This lack is penalized in Germany, where access to many skilled forms of employment is restricted to those who can provide proof of competence, usually acquired by completing a domestic dual apprenticeship or ensuing qualifications (Jacobsen 2021; Kogan 2011; Liebau & Salikutluk 2016). It can hence be mandatory to successfully complete a procedure of credential recognition to be allowed to work in a certain job even for those holding a degree or a certified vocational qualification (cf., Jacobsen 2021).

From a theoretical perspective, “direct indicators of workers’ skills, such as education and training, […] according to the human capital model, represent investments that increase productivity” (Kogan 2007: 10). At the same time, the “employer, […] seeks to recruit the most productive and the least costly applicant” (ibid.: 9) – education hence improves migrants’ labor market prospects (ibid.). From a similar point of view, vocational certificates can be understood as signals that provide information on the level of refugees’ vocational skillset (Damelang & Abraham 2016). By supporting employers’ “search process that is characterized by incomplete and asymmetric information” (ibid.: 93), formal degrees may hence improve labor market prospects. While a local degree may be superior to a foreign one, it may still represent an advantageous resource compared to having work experience only (ibid.; Damelang et al. 2020; Tibajev & Hellgren 2019).

These considerations describing the interplay of education, educational certificates and language acquisition lead to the hypothesis that having a degree is positively associated with trajectories that are characterized by early language training and succeeding transitions into full-time employment (H4).

Analogously, the assumed role of high education for language acquisition and labor market entry raises the question of what the lack of such implies for individual careers. Kristen et al. (2020) report that around a quarter of Syrian refugees had completed primary education at most. Hence, the low-qualified also constitute a substantive subgroup. In fact, at the time of their arrival between 2013 and 2016, only a third could read and write using the Latin alphabet, as discovered by Scheible (2018). However, the general BAMF integration courses, which consist of 600 lessons of language acquisition, are designed for individuals who have already obtained these abilities. Hence, around 50 % of refugees literate in another alphabet, and especially the illiterate 15 %, are confronted with additional hurdles. Originally, these two groups were supposed to attend the same literacy class as part of the BAMF integration course portfolio consisting of 1200 units. Alphabetized individuals could then transfer to the regular BAMF integration course once they made enough progress. Following the identification of problematic waiting times implied by this course structure, the portfolio was enlarged by a separate class for those literate in another alphabet in February 2017 (BAMF 2018; Scheible 2018). Bernhard and Röhrer (2020) highlight that Syrian refugees’ phases of formal language acquisition – and hence their early integration trajectories – are commonly linked to interruptions and long waiting times, which usually result from a mismatch of individual needs, abilities, and available course spots.^[4] They also highlight that long phases of language training do not necessarily imply that people aim at reaching high levels of German skills.

Given the considerations on the role of education with regard to learning efficacy outlined above, it is hence assumed that low education is positively associated with careers characterized by alternating phases of training and unemployment in the welfare system and only seldom transitions to employment (H5).

However, next to individual characteristics, transitions into employment can also be affected by macro conditions such as the local supply of and demand for employees. A factorial survey experiment by Damelang and Abraham (2016) identifies a negative effect of holding specific foreign nationalities net of differences in human capital on the likelihood of being invited to a job interview. For the group under study, not only anti-foreigner but also anti-refugee and anti-Muslim sentiments can lead to such disadvantages (cf., Kosyakova & Kogan 2022). Such sentiments may become relevant when assuming that “the employer presumably ranks job candidates in a queue, while an individual’s position in the queue is determined […] by his/her rank in relation to other candidates according to characteristics perceived as relevant” (Kogan 2007: 9) and consequently, it becomes more likely that refugees need to go to the end of the line. A theoretical framework of this approach is offered by Thurow’s ideas about employers’ applicant selection that highlights their ranking and thus the role of job competition (Herwig 2017; Thurow 1978). Resulting from this, the regional context may affect refugees’ individual outcomes: “[a]s a higher unemployment rate indicates […] more competition for vacant positions, it is directly related to the likelihood of finding a job. […] [S]uch a high local labor supply may lead to refugees being considered last for hiring” (Tsolak & Bürmann 2023: 3). On the other hand, low competition and hence an increased demand for employees should increase refugees’ job prospects.

This leads to the hypothesis that the lower the unemployment rate, the higher the likelihood of showing a career characterized by transitions into full-time employment (H6) and the lower the likelihood of showing a career consisting of alternating phases of (language) training and unemployment in the welfare system (H7).

Finally, to increase job prospects in the longer run, one may also strive for completing a local degree. For many, this goal implies not only longer phases of language acquisition, but further phases of qualification (cf., Bernhard & Röhrer 2020). This situation raises the question of who is able and willing to further invest in training and education and who decides in favor of a relatively quicker labor market entry. Those who would profit the most from higher future returns in the long run are the youngest (cf., Damelang & Kosyakova 2020). In this context, the opportunity costs of a decision, i.e., the “forgone income or other losses when a specific alternative is chosen” (ibid.: 12) need to be considered. These are less likely to be compensated for, the shorter the remaining time to achieve income from labor market activities are (cf., ibid.).

Overall, this suggests that there is a subgroup of male refugees who prioritize obtaining German degrees over instant labor market entry and that this option is particularly chosen by younger people. Hence, it is assumed that there is a separate typical career characterized by transitions into educational programs (e.g., apprenticeships or university education). The younger the person, the higher the likelihood of showing this type of career (H8).

3 Research Methodology

4 Results

4.1 Distribution of States over Time

Figure 1 depicts the distribution of states over time. With regard to the graphical description of the cluster solution, no regular ‘Relative Frequency Sequence Plots’ (Fasang & Liao 2014) or index plots can be shown due to rules relating to data protection. For the same reason, an additional state anonymous is introduced. The reported and graphically displayed numbers can therefore slightly deviate from the original distribution.

In the first month of the period under review (the third month after the appearance of records in the Unemployment Benefit II Recipient History), the two states ‘unemployment’ and ‘activity in welfare’ are assigned to almost 50 % and 40 % of the sample, respectively. While the share of the state ‘activity in welfare’ is strongly increasing for around six months, the opposite is true for ‘unemployment’ whose proportion is decreasing to less than a quarter in the same time period. Until the end of the period under review, the share in ‘unemployment’ remains comparably constant with some up and downs. On the other hand, after its peak at almost 60 %, the proportion of ‘activity in welfare’ decreases to less than 20 % in the 36th month.

States relating to employment show a different development over time. Full and part-time employment play practically no role in the first month but rise similarly to around 4 % and 3 % in the 12th month. Full-time employment then shows a stronger increase that yields a proportion of approximately a fifth in the 36th month which equals around double the share of part-time employment. Mini jobs play a comparably larger role in around the first 18 months but the share then stops increasing and instead stagnates at around 10 %. These differences in the temporal dynamics between different forms of employment are neglected when looking at the development of the overall employment rate.

Furthermore, almost no individuals are assigned ‘education’ during the first year. Then, the rate starts rising reaching around 6 % at the end of the period under review. While the development of the share in the residual category mirrors that of ‘part-time employment’, the proportion of ‘unavailability in welfare’ declines from around 6 % to around 4 % over time.

Overall, these results indicate that while only few individuals are (self-)employed at the beginning of the period under review, around 40 % are at its end. Considering that dual apprentices are assigned to the state ‘education’ (that makes up almost 6 %) and that self-employed individuals outside the basic security benefits schemes are not reflected in the numbers, these shares add to a similar employment rate as the 42 % and 48 % attributed to Syrian-born male refugees in the third and fourth year after arrival based on survey data (Brücker et al. 2024). This similarity supports the validity of the information on employment of both data sources given some definitional differences between the two approaches.

Finally, H1 and H2 can be confirmed: shortly after entering the welfare scheme, employment plays only a minor role. At the same time, (language) training in the welfare system is increasingly becoming the dominating activity.

Figure 1:

Distribution of States Over Time

4.2 Cluster Solution

Based on the clustering solutions and the evaluation of indicators (see Studer 2013), a typology of four clusters is chosen (ASW = 0.265).

• The largest cluster Alternating (in)activities in welfare is assigned to around 60 % of the studied population (n = 4887, ASW = 0.299). As shown in Figure 2, ‘activity in welfare’ and ‘unemployment’ are the dominating states. In the first month, almost 40 % are assigned ‘activity’ and in the last month, almost 30 % do so. In the second half of the first year, a peak around 65 % can be observed. On the other hand, more than half of refugees are in ‘unemployment’ at the beginning of the period under review while around 35 % do so in the last month. A minimum of around 25 % is likewise observed in the second half of the first year. An increasing dynamic with regard to transitions into employment can be observed in the third year: the share who are employed in the last month makes up almost a quarter. Figure 3 shows the five most common orders of states occurring within the cluster’s sequences. All typical orders consist of alternations of the two dominating states and start by ‘unemployment-activity-unemployment’. While the share of individuals in unemployment never falls below a quarter, only a small minority are assigned unemployment over the whole period under review. Median durations of 19 months (‘activity’) and 11 months (‘unemployment’) show that, despite not being employed, activities to acquire human capital, e.g., participation in language courses, play a dominant role. Still, time is not spent most effectively with regard to improving individual labor market prospects given the durations in unemployment. The pattern also aligns with research highlighting the role of involuntary waiting times and interruptions (Bernhard & Röhrer 2020; Homrighausen & Salwan 2021) related to language course participation.

• The second largest cluster Job seeking to full labor market entry (n = 1663, ASW = 0.182) is characterized by transitions into full-time employment. The related share mainly increases in the second year while toward the end of the period under review, the proportion is stagnating and then even slightly decreasing. With regard to potential path dependencies, it may worth noting that the cluster with the strongest labor market attachment shows the smallest share of initial inactivity of the three main clusters (36.5 %). The five most common orders show phases of activity and unemployment (sometimes recurring) that are followed by full-time employment.

• Transitions from Job seeking to partial labor market entry forms a separate typical career in male refugees’ early integration pathways (n = 1084, ASW = 0.175). Almost a tenth already has a mini job in the first month. The fraction increases to almost 60 % around two years after the start of the process and then shrinks to around 40 % in the last observed month. Similar to the two larger clusters, the states ‘activity in welfare’ and ‘unemployment’ dominate in early phases. However, in contrast to the prior cluster, transitions within different forms of employment, from mini jobs to part-time employment, appear in the top 5 orders. The proportion of people in mini jobs is even shrinking toward the end, in favor of a rise in full-time and particularly part-time employment. The results hence suggest that labor market integration can also be a step-wise process within employment. At the same time, it is less clear whether this partial entry occurs next to an actual main activity, e.g., education that cannot be identified in the data, or whether employment in mini or part-time jobs is chosen as an alternative because no job in full-time employment is found.

• Finally, the smallest cluster Residual (n = 459, ASW = 0.416) serves, as suggested by its name, as a residual category and is strongly characterized by capturing gaps in the data. Therefore, the interpretation, also in the upcoming section, focuses on the other clusters.

The results have also shown that no separate cluster characterized by transitions into educational programs such as apprenticeships or university programs appears. H8 is not confirmed. However, note the results of a robustness analysis based on a broader definition of educational activities that are further described in Appendix A2.

Figure 2:

Distribution of States Over Time by Cluster

Figure 3:

Five Most Common Orders of States by Cluster

4.3 Sample Description and Regression Results

Table 2:

Sample Description – Means and Standard Deviations/ Percentages

Variable	Mean (std) before z-standardization / %	N
Age	30.3 (9.7)	8,083
Monthly indicator	February 2016 (7.4)	8,083
Unemployment rate	7.0 (3.1)	8,083
Course ratio  Medium/high supply  Low supply	73.5  26.6	5,937 2,146
Degree  No degree  Degree  Missing	51.8 16.2 32.0	4,190 1,310 2,583
Household context and marital status  No partner, single, no child  No partner, (formerly) married, no child  Partner/child	55.0 21.4 23.6	4,441 1,732 1,910
Settlement structure  Big cities  Urban  Rural  Very rural	31.0 35.8 16.4 16.8	2503 2897 1328 1355
East/West  West  East	82.5  17.5	6,668 1,415

Figure 4:

AMEs on Cluster Membership With 95 % CI

Descriptive statistics can be found in Table 2.

The independent variables are regressed on the cluster membership to test the theoretically expected associations. Controlling for the effects of the other variables, on average, holding a degree increases the probability of being assigned to ‘job seeking to full labor market entry’ by 5.2 percentage points (p < 0.001) compared to possessing no degree. While a negative and significant association is found as far as ‘job seeking to partial labor market entry’ (- 2.7, p < 0.05) is concerned, the point estimate regarding ‘alternating (in)activities in welfare’ (- 2.5, p ≥ 0.05) is likewise negative but fails common significance levels.

Moreover, the local unemployment rate is negatively associated with the likelihood of being assigned to ‘job seeking to full labor market entry’. On average, an increase in the rate by one standard deviation is associated with a decrease in the likelihood of being assigned to ‘job seeking to full labor market entry’ by 5.3 percentage points (p < 0.001). At the same time, the likelihood of being assigned to ‘alternating (in)activity in welfare’ would be raised by 4.5 percentage points (p < 0.001). With regard to ‘job seeking to partial labor market entry’, the coefficient stands close to zero.

Controlling for the effects of the other variables, on average, a low supply of integration courses goes along with a decrease in the likelihood of being assigned to ‘job seeking to full labor market entry’ by 2.6 percentage points (p < 0.05) and an increase in the likelihood of being assigned to ‘alternating (in)activity in welfare’ by 2.6 percentage points (p < 0.05).^[9]

With regard to the hypotheses, the results support the assumption that the lower the local unemployment rate, the lower the likelihood of showing a career consisting of alternating phases of (language) training and unemployment in the welfare system (H7). Similarly, having a degree goes along with a higher likelihood of showing a trajectory that is characterized by early language training and succeeding transitions into full-time employment (H4). These hypotheses are hence confirmed.

Also, having no academic degree or vocational qualification is associated with a higher likelihood of showing a pathway characterized by alternating phases of training and unemployment in the welfare system and no transition to employment over the observation period (H5). Moreover, a low local supply of BAMF integration courses is negatively associated with the probability of showing a trajectory characterized by early (language) training and succeeding transitions into full-time employment (H3). Finally, the lower the local unemployment rate, the higher the probability of showing a career characterized by transitions into full-time employment (H6). However, robustness analyses (see Appendix A2) reveal that significance (p < 0.05) can or cannot be achieved depending on the respective model's specification. Interpreting the findings with caution, the hypotheses are not confirmed.

5 Discussion and Summary

The present analysis has explored the development of labor market careers of male Syrian refugees aged 18 to 60 by tracking their (in)activities for three years. To do so, I have analyzed data from a longitudinal administrative dataset called Sample of Integrated Welfare Benefit Biographies that stores information generated during operating procedures of the Federal Employment Agency. This includes the administration of the basic security benefits scheme Grundsicherung für Arbeitssuchende, a scheme that has provided welfare benefits to the majority of the group under study at some point in time (Bähr et al. 2017; Dummert et al. 2022; Oertel & Thomsen 2018).

For the analysis, each person’s state – education, full-time employment, part-time employment, mini job, unavailability in welfare, activity in welfare, unemployment, residual – in each of 36 months is identified. By applying clustering techniques, the sequences are summarized into a typology of four clusters. The largest cluster which covers around 60 % of sequences represents pathways of job seekers that are overwhelmingly spent receiving welfare. Often, phases of unemployment and activity, e.g., participation in language courses, are alternating. As theoretically expected, the likelihood of following this kind of career is positively associated with high local unemployment rates as revealed by the results of multinomial regressions of individual and contextual characteristics on cluster assignment.

A second cluster covering around a fifth of the observations shows a steep rise in full-time employment that mainly takes place in the second year. The probability of showing this type of career is positively associated with having a formal degree. A third cluster that represents 13 % of careers is characterized by an increase of the share having a mini job in the first and second year. Toward the end of the period under review, the share decreases for a rise in part-time and full-time employment. The cluster hence captures careers that show transitions from welfare to partial labor market entry. It is less clear whether this partial entry occurs next to an actual main activity, e.g., education that cannot be identified in the data, or whether employment in mini or part-time jobs is chosen as an alternative because no job in full-time employment is found. This uncertainty is mirrored in less clear regression results. A fourth cluster mainly captures trajectories related to the residual category.

More generally speaking, these results reveal several insights on male refugees’ labor market careers. First, comparing the results with those of general research on pathways after entering the German basic security benefits scheme (e.g., Bruckmeier et al. 2020b, Seibert et al. 2017), a remarkable difference can be identified that relates to the circumstance that no cluster characterized by relatively instant full-time employment occurs. However, the absence of initial employment is a common finding in research on the labor market integration of the refugee cohort under study and is attributed to the specific conditions discussed earlier that impede early labor market entries (cf., Brücker et al. 2024, Kosyakova & Kogan 2022).

Second, prior research also suggests that participating in the BAMF integration course has a positive effect on employment prospects (cf., Marbach et al. 2025). In fact, the cluster ‘job seeking to full labor market entry’ that is most strongly characterized by full-time employment also shows the highest initial share in ‘activity in welfare’ (most likely participation in integration courses) and the lowest share in ‘unemployment’ in the first month of the period under review. While the present study cannot make any causal claims and self-selection into courses may occur (cf., Bernhard & Röhrer 2020), it can provide descriptive evidence of a relationship between an early onset of training in the welfare system and later labor market success.^[10]

Third, splitting this overall development into a typology of careers has allowed for a test of expected associations between individual and contextual characteristics and the likelihood of following a certain type of career. Coefficients relating to ‘alternating (in)activities in welfare’ and ‘job seeking to full labor market entry’ usually show the opposite sign which underlines the diverging development for specific subgroups in the most or least promising direction. Men having a degree who live in counties with low unemployment and a high/medium level of supply of language courses are predicted a much higher likelihood of showing the advantageous career ‘job seeking to full labor market entry’ given the model estimates. The relevance of contextual characteristics, particularly the unemployment rate, also raises questions about the current policy of allocating asylum seekers across Germany that does not consider local economic conditions that potentially facilitate successful integration (Meyer & Winkler 2023; Brücker et al. 2022).

At the same time, it is essential to remember that the associations presented cannot be interpreted causally. For example, internal migration may lead to processes of self-selection that produce the existence or absence of specific correlations between contextual variables and individual outcomes. In this context, it is also worth noting that all independent variables are measured before the start of the process. Hence, if a person moves to another place that is characterized by totally different employment rate during the period under review, the individual is still assigned the covariate value valid before the start of the process.

In addition, variables of interest (e.g., the supply of language courses) may show a strong monthly dynamics while they are only provided on a yearly basis. Also, they can only be associated to the regional level of counties (Kreise), while smaller units may be more appropriate (Meyer & Winkler 2023).

Moreover, periods spent in school-based vocational education or university studies are usually determined based on certain assumptions (in contrast to apprenticeships). Similarly, participation in BAMF integration courses cannot be directly measured. Also, self-employment can only be observed when individuals are currently registered at a Jobcenter. Thus, phases in education and self-employment may be overlooked, particularly when there are no simultaneous entries to the basic security benefits scheme.

In a similar vein, identification of the correct state may depend on the correctness of the entries made by Jobcenter staff. Potentially, such errors disproportionally imply a confusion of ‘unemployment’ with ‘activity in welfare’. This may particularly be the case before spring 2019 when measure to improve data quality were implemented (Bundesrechnungshof 2020).

In addition, the period under review is restricted to three years. Future research based on a longer time horizon may also be able to meaningfully evaluate the overall gains of different types of careers in terms of cumulative earnings or occupational status (cf., Fuller 2015; Kogan & Weißmann 2013).

Another limitation of this study relates to the circumstance that not all processes of theoretical interest can be considered. For example, Dietrich et al. (2023) show that high scores of posttraumatic stress disorder in 2016 is associated with a significant reduction in the probability of being full-time employed in 2021. Next to health, similar limitations concern aspects like discrimination, access to networks or the duration of the asylum application procedures (see Kosyakova & Kogan 2022 for an overview). Also, a higher quality and further differentiation of the variable related to educational attainment appears to be preferable from a theoretical perspective.

Overall, one general conclusion can be drawn: the analysis of administrative datasets of the Federal Employment Agency can provide valuable insights which should be reflected in the provision of an administrative research dataset designed for the application to research on refugees’ labor market integration. With regard to some issues addressed in the discussion of limitations, the relevant information required is already available in datasets originating from the working processes of the Federal Labor Agency. For example, the study of Marbach et al. (2025) merge an administrative dataset similar to the one used for the present analysis with a dataset containing information on participation in the BAMF integration course. They can also work with immigration-specific variables, such as the date of arrival, which would have been important information for this analysis as well. The large number of Ukrainian refugees entering the scheme further underlines the demand for an evidence-based analysis of refugees’ labor market integration and calls for making use of the unique strength of administrative data with regard to sample size and panel stability. Using information that already exists in the databases of the Federal Employment Agency, tailoring them to the needs of research on refugee integration and providing access to external researchers would not only be a benefit to the questions asked in the present analysis but far beyond.