Generous to Workers ≠ Generous to All: Implications of European Unemployment Benefit Systems for the Social Protection of Immigrants

Record-high levels of international migration both toward and across Europe have recently given rise to a new body of research on the social protection of immigrants. A recurring argument in this literature maintains that migrants are generally more likely to gain access to social benefits in generous welfare states. The article offers a critical review of this hypothesis with a focus on unemployment benefit provision. The tides of European welfare politics have produced a set of systems in the past which are today highly stratified on the basis of employment. This mechanism generates a considerable benefit gap in reference to migration, especially for those who arrived to their country of residency only recently. Empirical analyses with micro-level data for 14 Western European countries provide supporting evidence for this argument. The findings indicate a negative relationship between generosity and social protection which has not been accounted for in previous research.


Introduction
Several key developments have, over the past 20 years, propelled Europe into what de Haas et al. (2020) refer to as the "Age of Migration." The European Union (EU) nearly doubled its number of Member States during this time, a development which opened up the right to free movement for many citizens living in Central and Eastern Europe. This expansion unfolded in parallel to the Great Recession (2007)(2008)(2009) as well as rising tensions in the Middle East and North Africa, both of which generated an unprecedented increase in political unrest and economic hardship. All these factors combined led to a sharp rise in cross-border migration both toward and across the continent. The total number of foreign-born immigrants 1 increased by 40% from an estimated 56 million in 2000 to roughly 78 million about two decades later (UN, 2017).
The described surge in migration raises crucial questions about the inclusiveness of Europe's comprehensive social protection systems. Immigrants might enter a country for various reasons (e.g., family, work, or refuge), but they all have in common that they arrive to some degree as outsiders to the social, political, and economic system. This implies that migrants are more exposed to social risks such as unemployment and poverty which are typically addressed through public benefits and services in European welfare states (Hooijer & Picot, 2015;Kogan, 2006). Migration thus emerges as a new dimension of social stratification which has long been neglected in comparative welfare state research (Esping-Andersen, 1990;Korpi & Palme, 1998;Lewis, 1992;Marshall, 1950).
Scholars have only recently begun to address this gap in the literature with studies that investigate systematically whether and why migrants' access to social benefits and services varies across Europe (Carmel et al., 2011;Lafleur & Vintila, 2020;Sainsbury & Morissens, 2012). Findings of systematic comparative analyses indicate that immigrants tend to have better access to social rights and a lower poverty risk if they live in a country with a comprehensive welfare state, thus suggesting a positive relationship between benefit generosity and migrants' access to social protection (Corrigan, 2014;Eugster, 2018;Römer, 2017;Schmitt & Teney, 2018).
The article contributes to this emerging body of comparative research with a critical review and fresh set of empirical analyses for the proposed "benefit generosity" hypothesis. It concentrates on the crucial case of unemployment benefit provision for newly arrived immigrants. This choice is informed by the wealth of literature documenting that labor market participation rates are found to be particularly low among newcomers across all forms of entry in countries with developed economies and comprehensive welfare states (Borjas, 1985;Chiswick et al., 2005;Fleischmann & Dronkers, 2010;Kogan, 2006).
Building on the conceptual work of Sainsbury (2012) it is argued that to study the role of generosity in this case we need to differentiate between formal social rights and actual benefit receipt. This becomes clear once the "politics behind the policies" (Sainsbury, 2012, p. 5) are taken into account. Today's benefit systems emerged from the collective action of manufacturing workers in the coordinated market economies of Central Europe (Carroll, 1999;Hall & Soskice, 2001;Sjöberg et al., 2010). They are designed to favor those who were already fully inserted into a wage relationship prior to job loss. Newly arrived immigrants, in contrast, are labor market outsiders upon arrival. This implies that they can be expected to receive a far lower level of social protection than native-born residents in more generous systems, even if they are formally granted equivalent social rights. An empirical assessment of this hypothesis with income data from 14 European welfare states generates supporting evidence. The results highlight a new aspect of the relationship between benefit generosity and migrants' access to social protection which has been unaccounted for in previous research.

The Benefit Generosity Hypothesis
The social protection of foreign-born immigrants is still a rather new subject in comparative research on European welfare states. Seminal contributions tend to neglect migration as a dimension of social stratification (Esping-Andersen, 1990;Korpi & Palme, 1998;Lewis, 1992;Marshall, 1950), partly because immigrants where long considered to be incorporated "denizens" who receive access to public welfare even before obtaining full citizenship in their country of residency (Guiraudon, 2000;Hammar, 1985;Soysal, 1994). Only recently, in light of debates about an emerging retreat from multiculturalism in Europe (Banting & Kymlicka, 2013;Joppke, 2007) and the general proliferation of large-scale comparative datasets (Helbling et al., 2017;Huddleston et al., 2015;Koopmans, 2013), have scholars in the field begun to study this topic more explicitly (Corrigan, 2014;Eugster, 2018;Römer, 2017;Sainsbury, 2012;Schmitt & Teney, 2018).
The emerging body of scholarship on migrants' social protection in Europe points to a general pattern. Both Römer (2017) and Schmitt and Teney (2018) find that immigrants, asylum seekers, and permanent workers alike, are more likely to be granted access to social assistance, social security, and housing benefits in more generous welfare states, such as those found in Central and Northern Europe. This result aligns with the findings of Boräng (2018), which indicate that the same countries are also more open to immigration, particularly to those seeking international protection. Large-scale comparative analysis on welfare outcomes lend further support to this "benefit generosity" hypothesis. Immigrants are generally found to have a lower poverty risk in countries with more extensive social protection systems (Corrigan, 2014;Eugster, 2018;Sainsbury & Morissens, 2012).
Evidence in support of the benefit generosity hypothesis is mounting in the literature on migrants' social protection. However, few studies actually disentangle how individual benefit programmes address specific social risks that might occur disproportionately among immigrants. One of the few exceptions is the work of Sainsbury and Morissens (2012). In a descriptive comparison between Denmark, Germany, France, Sweden, the United States, and the United Kingdom the authors find, in line with previous studies, that the comprehensive welfare states of the two Nordic countries perform best at lifting households of immigrants and ethnic minorities out of poverty (p. 121). However, they also point out that the picture becomes more complicated on the level of individual social risks. Households with children and dependent adults receive family benefits and pensions at about the same rate across all subgroups and countries. In contrast, Sweden and Denmark deviate from the pattern when it comes to individuals who are out of work. Immigrants in this group are far are less likely to receive unemployment benefits on a par with the native-born population (p. 126).
It is difficult to draw general conclusions from the findings of Sainsbury and Morissens (2012). Large differences in the data collection process impede on a more substantive interpretation (p. 125). However, the analyses indicate that access within individual benefit programmes seems to be more stratified than the benefit generosity might suggest. This caveat is particularly important for the very programme that sticks out in this case. Previous research on economic integration shows that immigrants, newcomers in particular, face a higher risk of unemployment than native-born residents in the European context (Fleischmann & Dronkers, 2010;Kogan, 2006). Systematic comparative studies on their access to unemployment benefits, however, is scarce. Research on this topic is limited and tends to be confined to individual country cases (Bratsberg et al., 2017;Hansen & Lofstrom, 2011;Wunder & Riphahn, 2014).
Reviewing the nascent literature on migrants' social protection appears to unveil a growing consensus about the benefit generosity hypothesis. However, the findings of Sainsbury and Morissens (2012) show that the argument needs to be revisited with more fine-grained theoretical and empirical analyses. The study addresses this gap in the literature with a particular focus on unemployment benefit provision. A theoretical framework is developed in the next section with a combination of insights from welfare state, labor market, and migration research.

Theoretical Considerations
To review the benefit generosity hypothesis with a focus on migration and unemployment we first need to map out how European welfare states address the social risk of unemployment to begin with. A common characteristic of all systems is that the most generous transfers are paid through insurance funds on a membership basis (Esser et al., 2013). Tax-funded government programs for minimum income protection only play a complementary role in the European context (Bahle & Hubl, 2011). This "insurance model" has particular implications for the social protection of immigrants which are thoroughly discussed in this section. The central hypothesis of the study is derived at the end by applying the theoretical considerations to the particular case of immigrants who arrived to their country of residency only recently.

The Politics of Unemployment Insurance
The dominance of insurances for the provision of unemployment benefits in Europe can, at a basic level, be explained with help of the so-called Varieties of Captialism (VoC) approach (Hall & Soskice, 2001). European labor markets are marked by a high demand for industry-and firm-specific skills. Social protection for the unemployed is crucial in this context as it ensures that members of the workforce will be willing to invest in these skills despite the high risk of becoming dependent on them in the long term (Estévez-Abe et al., 2001).
The VoC approach offers a compelling explanation for the relative strength and omnipresence of social protection for the unemployed in Europe. However, it is less equipped to explain why income security is primarily provided though insurances, even in a liberal market economy with less specific skill demands like the United Kingdom (Esser et al., 2013). This gap can be addressed with help of Power Resource (PR) theory which concentrates more on the "politics behind the policies" (Sainsbury, 2012, p. 5), particularly on the strength of organized labor (Esping-Andersen, 1990;Korpi & Palme, 2003;Stephens, 1979). The very first unemployment insurance funds were created among labor unions in the Belgian city of Ghent during the Industrial Revolution (Vandaele, 2006). The city provided subsidies to keep membership costs low, especially for workers who were most at risk of losing their job. This "Ghent system" caught attention abroad and led to the introduction of the first nationally legislated unemployment benefit system in France in 1905 (Sjöberg et al., 2010, p. 428).
It is worth noting at this point that the strength of organized labor tends to be toned down in studies on unemployment insurance, even among scholars who follow PR theory (Carroll, 1999;Sjöberg et al., 2010). This is primarily because later developments deviated somewhat from the original Ghent model. The United Kingdom was the first country to introduce a nation-wide unemployment insurance system in 1911 (Sjöberg et al., 2010, p. 422). The difference was that it made insurance membership mandatory. Other countries followed shortly thereafter leading to an expansion of compulsory unemployment insurance systems first among Western European countries and then across the entire continent after the end of the Cold War (Vodopivec et al., 2005).
The general trend toward statutory mandatory instead of voluntary state subsidized insurance meant that power over benefit provision shifted gradually from labor unions to national governments. However, workers maintained influence through the political system. Denmark, Finland, and Sweden, three countries with dominant social-democratic parties, still operate a version of the "Ghent system" today (Vandaele, 2006). Moreover, Korpi and Palme (2003) found that left-leaning cabinets were much more cautious to cut benefits during phases of welfare state retrenchment (1980s and 1990s). The social insurance model is thus kept firmly in place by the political power of organized labor in European welfare politics.

Unemployment Insurance in the Age of Migration
A central implication of the described insurance model and its roots in the workers' movement is that today's systems rely heavily on the definition of employment rather than that of unemployment. All systems in Europe follow a "logic of reciprocity" (Leitner & Lessenich, 2003). Transfers tend to be most generous for labor market insiders, particularly middle-income earners with stable employment contracts (Jahn, 2018). Eligibility is based on prior contributions to unemployment insurance and/or employment records (worktesting). Those who do not qualify are referred to lower second-tier systems such as social assistance, unemployment assistance, or housing benefits instead (means-testing) (Esser et al., 2013, p. 12).
Taking into account how European welfare states structure social protection for the unemployed around the concept of reciprocity is particularly relevant when considering that the systems are otherwise rather neutral in regard to migration. In a recent comprehensive analysis of Europe's welfare states Lafleur and Vintila (2020) find that none of the region's unemployment insurance systems "[. . .] imposes specific migration-related conditions that could directly obstruct immigrants' access to welfare" (p. 26). In other words, any legal resident of a European country is formally treated equally as an incorporated "denizen" (Hammar, 1985) in reference to unemployment insurance. Stratification can instead be expected to emerge in the actual distribution of benefits, or what Sainsbury (2012) refers to as "substantive" access to social protection: only foreign-born immigrants with a regular employment contract are de facto incorporated into the system.

Implications for Newcomers
Paid work as a precondition for social protection can be expected to generate particularly strong barriers for immigrants who just arrived to their country of destination. After all, newcomers first need to establish themselves on the labor market in order to join the risk pool of work-tested benefit programs. This mechanism forces us to rethink what "generosity" actually means in this context. Unemployment benefit systems are generally considered to be generous if they provide a high level of payments for a long period of time on the basis of low qualifying conditions (Esping-Andersen, 1990;Esser et al., 2013;Scruggs & Allan, 2006). However, even the most comprehensive system are still based on work-testing. More generous systems thus only raise the level of substantive social protection for the core workforce, not for all residents. Those who are born in their country of residency grow up in preparation for the local labor market, which gives them an edge at entering this risk pool relative to all newcomers. Social protection among the unemployed is, as a result, expected to be more stratified on the basis of country of birth in more generous systems.
The expected relationship between generosity and social protection is arguably further exacerbated by the nexus between generosity and labor market dualization. Unemployment insurances are designed to be "passive" or "decommodifying" (Esping-Andersen, 1990) labor market interventions. They provide citizens with an income sources that makes them less dependent on paid work. That in turn, however, raises the reservation wage for which job seekers are willing to work (Feldstein & Poterba, 1984). Generous unemployment insurance systems thus contribute to the dualization of labor markets between well paid stable employment at the core and temporary low paid jobs at the fringes (Lindbeck & Snower, 1988). This makes it even more difficult for newly arrived immigrants to amass the required amount of contributions and derive social protection from a work-tested benefit system.
We can generally expect the discussed stratification to be strongly patterned by country of origin and form of entry. After all, both are found to be reliant predictors of economic integration in the long term (Borjas, 1985;Rendall et al., 2010). Benefit gaps with the native-born population are, however, expected to be most pronounced across all groups during the first couple of years after arrival, due to the entanglement between work-testing and labor market dualization. This overarching expectation can be expressed with the following hypothesis: Newly arrived immigrants are less likely to receive work-tested unemployment benefits on a par with the native-born population in more generous benefit system.
The proposed hypothesis might appear to offer a counter argument to the benefit generosity hypothesis at first glance. After all, the social protection of immigrants is expected to be lower in countries with more generous benefit systems. However, it should be noted that it does not explicitly rule out that immigrants are more likely to receive formal social rights if benefits are generous. What it instead points to is that substantive access to social protection can be highly stratified even in the most inclusive system. Data sources for an empirical test of the proposed hypothesis are introduced in the next section.

Data
Data for this study stems primarily from the European Statistics on Income and Living Conditions (EU-SILC). National statistics offices from over 30 European countries provide yearly samples of household-and individuallevel data to this system, using a combination of surveys and administrative records (Iacovou et al., 2012). Many studies in the literature on economic integration use this dataset due to its strong harmonization of key variables (Bárcena-Martín & Pérez-Moreno, 2016;Corrigan, 2014;Eugster, 2018;Hooijer & Picot, 2015).
The dataset is first pooled across all currently available waves (2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017) and then restricted to the sub-population of individuals in working age (15-64 years) who reported at least 1 month of unemployment during the reference period (past 12 months). Benefit receipt can be operationalized with help of a binary variable indicating transfers from the country's unemployment benefit system. Such transfers cover both long-and short-term benefits in EU-SILC. However, all payments have in common that they are explicitly work-tested. 2 Information on migration is largely standardized in EU-SILC, but also somewhat limited. Length of residency is unfortunately only available in 5-year intervals and data on country of birth is collapsed into three categories ("Local," "European," and "Non-European"). The differentiation between newly arrived and more established immigrants is thus strongly predetermined as is the possibility for a breakdown by origin or entry category. However, the dataset still covers essential information for research on migration in this context. Intra-European immigrants can access the labor market of their destination country without particular restrictions, due to a range of multilateral agreements and institutions, including the European Union, the European Economic Area, and the Schengen Area. Employment rights for non-European migrants, in contrast, depend on bilateral agreements which are more restrictive in most cases (Geddes, 2018). Comparative analyses show that it takes non-European migrants roughly 5 years on average to legally obtain either permanent residency status or citizenship in the European context (MIPEX, 2015, p. 50). The coding of EU-SILC thus entails rather broad, but very crucial distinctions for the purpose of this study.
Benefit generosity is measured with help of Scruggs and Allan's (2006) Benefit Generosity Index (BGI). This operationalization is used extensively in comparative welfare state research and particularly in the case of migrants' social protection (see Boräng, 2018;Corrigan, 2014;Römer, 2017;Schmitt & Teney, 2018). The index for unemployment benefit generosity is comprised of five indicators: coverage within the labor force, length of the qualifying period for benefit access, number of waiting days before first payments, replacement rate relative to prior earnings, and benefit duration. All five measures refer to work-tested benefits only (Scruggs et al., 2017).
Each of the BGI sub-indicators is expected to be positively associated with a person's probability of benefit receipt. Less restrictive eligibility rules increase the share of recipients by definition. Higher benefit levels and longer payments can be expected to make the system more attractive, leading to a higher number of claimants. Hence, an index that systematically combines all of these indicators is preferred over individual proxies such as those provided by the OECD or Eurostat. The original BGI will be included in its entirety for the main analyses. Additional robustness checks with individual sub-indicators as independent variables can be found in the Supplemental Appendix.
Data with sufficient coverage and quality could be retrieved for 14 Western European countries: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Italy, Ireland, the Netherlands, Portugal, Spain, Sweden, and the United Kingdom. Information on time since arrival is unfortunately only available from 2010 onward. This is particularly problematic when considering that coverage for the BGI currently ends in 2011 (Scruggs et al., 2017). What helps addressing the described problem is the finding in previous studies that cross-country rankings of payment levels and qualifying conditions have been rather stable over the past 20 years, even in light of smaller temporary changes throughout the Great Recession (Helgason, 2019;Jahn, 2018;Vis et al., 2011). This can be confirmed when looking at trends of the BGI across countries (Table 1). The only two cases that deviate from the  general pattern are Sweden and Denmark. Both countries show substantial drops in generosity levels which occurred primarily due to a decline in replacement rates (Jahn, 2018). However, changes in both countries took place before 2008 and thus prior to the main observation period (2011)(2012)(2013)(2014)(2015)(2016)(2017). The problem of missing values on the BGI could therefore be addressed by carrying forward each country's last observation in 2011.
It is unfortunately not possible to further extend the presented list of macro-level units due data limitations. However, the given selection has its advantages, too. All countries are relatively similar in many regards. They are all located in the same region and have all had rather stable democracies and economies for the past 40 years. Migration patterns for the selected cases are shaped by similar forces through their spatial proximity as well as their membership in the European Union. Their welfare systems, on the other hand, are still fairly diverse, despite growing convergence over the past decades (Arts & Gelissen, 2010). A study with this selection of countries thus fits well with the "most similar systems" framework of comparative research (Lijphart, 1971, p. 687).

Analytical Strategy
The proposed hypothesis of this study suggests that unemployment benefit receipt can be modeled as a function of welfare state generosity (macro-level) and country of birth (micro-level) in a comparative design. An empirical analysis is therefore best realized by incorporating both levels into one single framework. Multi-level modeling is particularly suited for this purpose. Estimates could in principle be realized with single-level analyses as well. However, ignoring the hierarchical structure of cross-country data is rather problematic. Every single object is treated as an independent case rather than as a member of a higher-level cluster. Standard errors tend to be underestimated in this case, leading to overly confident interpretations of study results (Hox et al., 2018, p. 4).
Multi-level modeling takes the hierarchical structure into account and enables unbiased estimates with corrected standard errors even for computations with few upper-level units (Elff et al., 2021). It is therefore rather unsurprising that the method is widely adopted in comparative, including the literature on migrants' social protection (Bárcena-Martín & Pérez-Moreno, 2016;Corrigan, 2014;Eugster, 2018). The study employs a specific type, logistic random-intercept multi-level modeling with maximum-likelihood estimations, for all regression analyses due to the non-linear nature of the dependent variable (Hox et al., 2018, p. 103).
Fixed effects are only included for the year of data collection, leaving room for country-specific random intercepts and cross-level interactions. Estimates for the relationship between time since arrival and benefit generosity are presented both as output tables and plotted average margins to avoid false-positives (see Ai & Norton, 2003). Results of analyses with ordinary least squares and separately clustered standard errors on country-and yearlevel instead of joined logit estimates can be found in the Supplemental Appendix, along with additional robustness checks. All continuous variables are z-standardized in order to facilitate interpretation.
Pooling data for 14 countries over 8 years (2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017) generates 112 country-year clusters with full data coverage. Sample sizes per country vary between around 4,000 (Denmark) and 37,000 observations (Spain), summing up to a total number of about 200,000. The number of observations and highlevel clusters for this dataset is substantial for the purpose of multi-level modeling (Elff et al., 2021). However, it is still debatable whether this is enough for a detailed analysis of the proposed hypotheses. Restricting the dataset to roughly two dozen Western European countries means that each individual case exerts a considerable weight in the analysis (Bryan & Jenkins, 2016). Robustness checks were performed in order to estimate the degree to which this influence distorts the overall analysis. Results can be found in the Supplemental Appendix. Replication materials for all analyses are compiled by Gschwind (2020).

Control Variables
The central hypothesis of the study builds on the argument that substantive social protection is stratified in reference to country of birth due to a systematic division between (unemployed) insiders and outsiders on the labor market. More generous unemployment benefit systems are expected to be associated with larger benefit gaps both through their internal mechanism of work-testing and external impact on labor market dualization. This relationship can only be assessed empirically under the prerequisite that newcomers encounter similar conditions for employment across countries, apart from the variation in unemployment benefit systems. A range of both macro-level variables are included in the analyses for this reason. Controls on the country-level are added to account for structural employment barriers that migrants might encounter upon arrival. Information on economic downturns is included using a lagged variable for GDP growth 3 to control for the fluctuation of labor demand. Educational requirements for employment are modeled with an estimate for the share of workers in highskill occupations. 4 Targeted migrant-specific restrictions to labor market mobility are, lastly, controlled for with an indicator for the inclusiveness of employment rights retrieved from the Migrant Integration Policy Index (MIPEX) 5 .
It is unfortunately not possible to cover a broader set of macro-level variables due to the limited number of country cases. A viable alternative in such a situation is the use of fixed effects for clusters of structurally and geographically similar countries (see Kogan, 2006). This would clearly help to account for other factors, such as the volume and source of cross-border movements as well as differences in migration policies and welfare state characteristics. A general problem with this approach, however, is that it will always consume much of the relevant variation in unemployment benefit generosity between clusters. Supplementary analyses where nevertheless performed to explore this alternative. Their results are briefly discussed in the main text. Details can be found in the Supplemental Appendix.
Micro-level variables are, finally, included to account for heterogeneity between newly arrived immigrants and native-born residents that is not directly related to migration as such. The study includes a measure for the length of unemployment for this reason, as well as standard socio-demographic indicators for gender, age, education, and household type. Benefit receipt probability is expected to by highest for men at the average age of the sample population who live in a one-person household and do not possess a higher education degree. This expectation is based on prior findings in gender analyses of the welfare state which indicate that unemployment insurance systems gravitate toward protecting the "average production worker" (Lewis, 1992;Sainsbury, 1996). The longitudinal component of EU-SILC can unfortunately not be used in this context as it does not include information on migration. Any interpretation of the results thus needs to take into account that a person's labor market status prior to unemployment is only approximated through the socio-demographic characteristics that were listed above.

Main Results
A first set of bivariate analyses with mean values for the years 2010 to 2017 is presented in Figure 1. The plot in the upper left corner shows that the share of benefit recipients among immigrants falls below that of the native-born in most countries. A comparison by length or residency indicates that this benefit gap increase sharply with the level of generosity for those who arrived less than 5 years prior to data collection. Migrant groups with longer periods of residency appear to be less affected.
An overview of the the main findings for the multivariate analyses is presented in Table 2. Detailed statistics can be found in the Supplemental Appendix. Estimating a random-intercept model without macro-or microlevel control variables returns a significant negative estimate for country of birth (Model 1). The main coefficient increases further once control variables for length of unemployment, gender, age, education, and household type are included (Model 2). Hence, migrants receive benefits with a lower probability when compared to natives, even under control for other predictors of benefit receipt. The remaining unexplained macro-level variance decreases notably once unemployment benefit generosity is included (Model 3). The estimate for this variable is significant and positive which shows that the   unemployed are generally more likely to receive benefits in countries with more generous systems.
Interacting benefit generosity with country of birth returns a significant and negative estimate (Model 4), indicating that benefit gaps are larger in more generous systems. The main relationship between generosity and benefit receipt decreases somewhat once all other controls are interacted (Model 5). 6 Adding control variables for GDP growth, inclusiveness of employment rights, and the share of high-skilled workers does not shake up this result substantively (Model 6). The final estimate subdivides the foreign-born by length or residency. It shows that both the main estimate and the interaction effect are strongest for newly arrived immigrants.
Results of the logistic multi-level analysis can be illustrated graphically by plotting the average marginal effects (AME) for each migrant group against the benefit generosity index (Figure 2). The results show that the unemployed are generally less likely to receive work-tested benefits if they are foreignborn. This gap widens with the level of benefit generosity. However, it does so with varying intensity across the different migrant groups. Those who just arrived to their country of residency are most affected. Their probability to receive benefits drops from −10 to −30 percentage points between the highest and the lowest level of the scale. Established migrants are more likely to receive payments on a par with the native-born, even in the most generous benefit systems.

Robustness Checks and Supplementary Analyses
The results presented before largely support the central hypothesis of this study. Robustness checks and additional analyses can be found in the Supplemental Appendix. They show, firstly, that the main interaction effects remain negative and significant even if ordinary least squares (OLS) regression with clustered standard errors is used instead of logistic multi-level modeling. This is also the case if the Benefit Generosity Index is replaced with either of its sub-indicators. 7 Analyses with clusters of structurally and geographically similar countries return a somewhat weaker, but still robust effect. Extending and subdividing the dataset by year of data collection further shows that the effect remains stable throughout the Great Recession (2007)(2008)(2009)) and the latest peak in refugee migration (2015). The findings, lastly, hold as well if individual country cases are dropped to gauge their influence on the overall results. Supplementary analyses were performed to further investigate underlying mechanisms of the relationship between generosity and benefit receipt. Breaking down the sample by region of origin shows that the results are similar for European and non-European migrants, albeit somewhat stronger for the latter. Similar results can be found when subdividing the sample by age group. The effect of generosity is negative and significant even if newly arrived immigrants are compared to labor market newcomers among the native-born, namely those below the age of 30. These results further support the argument that generous work-tested insurance systems contribute to a structuring of the labor market that stratifies social protection between nativeborn residents and newly arrived migrants even among those with equivalent economic rights and prerequisites on the labor market.
What, finally, remains open for investigation is the question whether those without transfers from the unemployment benefit system receive income from other benefit programs instead. EU-SILC allows for some additional analyses to investigate this question (see Supplemental Appendix). Replacing unemployment benefits with either family or housing allowances as dependent variables returns estimates similar to those of the main analyses. In contrast, newly arrived immigrants do seem to receive means-tested income protection at a higher rate than native-born residents in countries with generous unemployment benefit systems. The coding in EU-SILC lumps together a whole range of programs for this indicator, including both social assistance and introductory benefits for asylum seekers. Yet, all of have in common that they provide substantially benefits lower than payments from the unemployment insurance (Esser et al., 2013, p. 14). It is thus unlikely that access to such benefits would keep migrants from claiming work-tested benefits for which they are eligible.

Concluding Remarks
The study at hand contributes to research on the social protection of immigrants with a novel theoretical approach and a comprehensive set of empirical analyses. Its focus lies on the implications of European unemployment benefit systems for immigrants who arrived to their country of residency only recently. The core hypothesis sets a contrast to previous studies by arguing that newly arrived immigrants are less likely to receive social protection on a par with the native-born population in countries with more generous unemployment benefit systems. This hypothesis is derived from a theoretical framework that combines insights from comparative research on the welfare politics behind today's unemployment insurance systems with the literature on economic integration in the European context.
Empirical analyses across 14 Western European countries with one of today's most extensive datasets on income and migration have generated supporting evidence for the proposed hypothesis. Foreign-born immigrants are generally under-represented among unemployment benefit recipients, especially in countries with comparatively generous systems. This relationship between migration, generosity, and benefit receipt was, in line with the core argument of the study, found to be particularly strong when native-born residents were compared to immigrants with less than 5 years of residency.
Restrictions of the used dataset inhibit more general conclusions on basis of the reported results. Stratification is only observed cross-sectionally for the particular case of unemployment benefit receipt among newly arrived immigrants in Europe. However, the findings retain important implications for debates about the relationship between migration and public welfare more at large. They, firstly, support Sainsbury's (2012) argument that the social protection of immigrants needs to be conceptualized in terms of both formal and substantive access to public welfare. The study shows, secondly, how crucial it is to account for the politics behind social policies to unpack underlying mechanisms of social stratification in relation to migration. Its connection to research on economic integration, finally, highlights the importance of early labor market integration as a tool for immigrants to achieve both economic self-sufficient and income security.
Comparative research on the social protection of immigrants is in many ways still in its early stages. The study offers a point of departure, in this regard, from which several promising pathways for future research open up. Long-term trajectories of substantive access to social protection could be studied with more fine-grained information on migration by using a combination between comparative cross-sectional and longitudinal single-case analyses. It would further be worth exploring how unemployment insurance providers themselves cope with the increasing diversity of the workforce. A crucial case in this regard are today's "Ghent systems" with voluntary unemployment insurances where labor unions act as non-for-profit organizations (NPOs) for the enactment of migrants' substantive social rights (Bruzelius, 2020). Taking stock of such and other developments is highly relevant both for immigrants themselves and for the workings of modern welfare states at large.

Author Note
This article has greatly benefited from feedback received by the editors and anonymous reviewers of CPS. I am also grateful for all suggestions on earlier versions of the manuscript, particularly by Per Adman, Anton Ahlén, Nazita Lajevardi, Pär Nyman, Linda Moberg, and Joakim Palme as well as participants at the 2018 ECPR General Conference in Hamburg, the 2018 ESPAnet Conference in Vilnius, the 2019 IMISCOE Spring Conference in Liège, and the 2019 ECPR Joint Session on "Migrants' Access to Welfare in Times of Crisis." All analyses in this article employ cross-sectional data from the European Union Statistics on Income and Living Conditions (EU-SILC), years 2004 to 2017. Data is provided by the statistical office of the European Union (Eurostat). The responsibility for all conclusions drawn from the data lies entirely with the author.

Declaration of Conflicting Interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding
The author disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Data access was granted as part of the project "Aging migrants, labor market transitions and the welfare state" which received funding by the Swedish Research Council (grant number 2018-04898).