The impact of partisan politics on migration policies: the case of healthcare provision for refugees by German states

Comparative migration policy research has increasingly dealt with the question of whether partisan differences in government can explain differences between migration policies. The empirical findings, however, for both, European states and states of the USA, are inconsistent as governing parties’ positions on integration policies do not in all cases explain the differences in migration legislation. This article investigates the conditions under which German states opt for a permissive model of healthcare provision for asylum seekers as an alternative to the existing restrictive bureaucratic model. Using a fuzzy-set qualitative comparative analysis (fsQCA), we find that a left-of-center state government is a necessary condition for the introduction of the alternative model. Full cost reimbursement by the state is the sufficient condition, i.e. the permissive alternative model is fully implemented only in those states with a left-of-center state government and where municipalities do not bear the risk of potentially higher costs for healthcare provision. With this unambiguous partisan effect, this article represents an important contribution to the international debate.


Introduction
Comparative migration policy research examines whether and to what extent migration policies vary between states, whether patterns emerge over time, such as alignment in a particular direction, and which factors account for existing policies. On the one hand, the policies of the OECD states, for example, have converged over time (e.g. Helbling & Kalkum, 2018). On the other, there are still areas with considerable and even increasing differences between sovereign states and between states in a federal structure (e.g. Gulasekaram & Ramakrishnan, 2015;Reich, 2019;Paquet, 2019).
Migration policy research in the 1990s highlighted structural factors for explaining the convergence of migration policies: Freeman (1995) argued that migration policy in liberal democracies has become increasingly expansive and liberal because promigration interests are easier to organize. This is because the benefits of migration (e.g. a labor force for companies) are concentrated whereas the costs remain diffuse.
The literature from the 1990s also assumes that the political elites of liberal states share an 'anti-populist norm' according to which migration policy should not become politicized. Joppke (1998) sees human rights obligations and constitutional principles as additional reasons why liberal democracies also accept unwanted immigration (e.g. family reunification). The assumed 'anti-populist norm' seems to vanish given the rise of right-wing populist movements, making migration a highly contested issue. Recent scholarship investigates whether right-wing parties directly or indirectly affect migration policies (e.g. Rooduijn et al., 2014;Lutz, 2019). The overall findings are nuanced. On one side, immigration policies (measures regulating the entry and stay of migrants) became overall less restrictive over time (de Haas et al., 2018;Helbling & Kalkum, 2018;Lutz, 2019). On the other, integration policies (immigrants' rights and obligations within a country) are affected by anti-migrant mobilization (Lutz, 2019).
In the light of contested migration policy, research has intensified the systematic study of differences and has become more interested in political factors. Hence, the partisan composition of governments (and Parliaments) has become a focus of attention. According to partisan theory, one would assume that left-of-center governments pursue a more permissive policy whereas right-of-center parties pursue a more restrictive policy concerning migration. However, it is a matter of dispute whether this applies in reality. Empirical studies on the migration policies of European and North American states (De Haas & Natter, 2015, p. 17) and the migration policies of US states (Zingher, they can consult the doctor directly. Although the substantial restriction on benefits remains unchanged, this model leads to improved healthcare provision in procedural terms. We, therefore, consider the new model of health care provision a permissive, integration facilitating policy. The city-state of Hamburg introduced this model in 2012. During the so-called "refugee crisis", when in 2015 and 2016 around 1.2 million people came to Germany, the federal government facilitated the introduction of this model by amending legislation. However, only half of the remaining 14 states introduced this model, while the other half rejected it. The contribution of our case to the debate on migration policy is twofold. First, health care policies for refugees represent a powerful indicator for the status of migrants' broader integration into society (Reiter & Töller, 2019;Günther et al., 2020;Bozorgmehr et al., 2020). Despite the highly institutionalized health care systems in industrialized countries, access to health care and provisions for asylum seekers often remains restrictive (Norredam et al., 2006;Biddle et al., 2020). Exploring explanatory factors of policy variance adds to insights which factors may foster the integration of migrants into societies more generally. What is more, studies which analyze the impact of populist right-wing parties on health policies suggest that the rise of right-wing populism tends to politicize the access of migrants to health care (Falkenbach & Greer, 2018). Additionally, given the rise of right-wing populist parties, the question is whether mainstream parties respond to anti-migrant ressentiments with restrictive or permissive policies (Rooduijn et al., 2014;Schumacher & van Kersbergen, 2014;Falkenbach & Greer, 2018;Cammaerts, 2018).
Second, investigating subnational policy variance within federal states is a promising endeavor for research on migration policies. While we know of the variance of migration and integration policies across national states (Helbling & Kalkum, 2018;De Haas & Natter, 2015;de Haas et al., 2018), differences on subnational levels often remain undetected. Controlling for the constant institutional setting the federal state provides, intra-state comparisons allow investigating specific causal explanations for this variance. 1 Additionally, in many federal systems, the subnational level has the legal competency to implement or even shape integration policies, which makes subnational comparisons the plausible level of analysis. Arguably, it is substantively more appropriate to study immigration issues at the sub-national than the national level because this is the place where immigrants integrate and interact with natives (Helbling & Traunmüller, 2016, p. 403) and where integration does or does not occur. Hence, intra-state comparisons are a common research strategy in studies on migration policies of states with federal structures like the USA and Canada (Paquet, 2019;Reich, 2019;Butz & Kehrberg, 2019;Commins & Wills, 2017;Ybarra et al., 2016;Gulasekaram & Ramakrishnan, 2015;Zingher, 2014).
In Germany, the access of asylum seekers to social benefits (accommodation, subsistence allowance, cash benefits or benefits-in-kind, healthcare provision) has been the focus of the migration policy debatenot just since the refugee crisis in 2015. Yet, studies that compare the implementation of those policies at the subnational level in 1 In our case, for example, conditions on the federal level remained constant during the period under investigation: the partisan composition of the federal government, a "Grand Coalition" of Christian-Democrats and Social-Democrats, and an overall prospering economic situation. Thus, during the period under investigation, 2015-2016, "timing" or "sequencing" does not pose a problem.
We, therefore, ask for explanatory factors for the puzzle why German federal states introducedor refrained from introducinga permissive model of healthcare provision. Based on the explanatory approaches of comparative public policies and comparative migration policy research, we investigate the effect of the partisan composition of the relevant state government, specific institutional constellations, the proportion of foreigners, and the socio-economic situation in the respective state. We subsequently test the factors resulting from these explanatory approaches with the help of fuzzy-set qualitative comparative analysis (fsQCA) to determine whether they alone or in combinations of the factors are a necessary or sufficient condition for the commissioning of health insurance companies by a state. We proceed as follows: in the next section, we explain our research design and in particular the choice of our method. We then present the phenomenon to be explained (the outcome in QCA terminology). Subsequently, we present the conditions whose potential influence we have derived from the literature and our case studies, 2 and formulate hypotheses. In a next step, we perform the analysis and present our results. Finally, we sum up and conclude.

Research design and method
This paper systematically examines the partisan effect as well as other factors that may influence the commissioning or non-commissioning of health insurance companies to provide healthcare for asylum seekers in states. Multivariate regression analysis based on large numbers of cases, that is common e.g. for analyzing the migration policies in the 50 US states (e.g. Reich, 2019), is of limited use in the context of German states. With generally 16 states in the German federal system, and only 14 states in our case, our N is too large to determine causal interrelations by carrying out a qualitative comparative case study but too small for regression analysis. Therefore, we apply qualitative comparative analysis (QCA) which is settled "in the middle" between large-N and small-N comparative analyses in terms of the number of manageable cases but also in terms of causal logic. It goes back to Charles Ragin (1987) and has established itself as a methodology in policy studies generally (Rihoux et al., 2011) and in research into the policies of German states in particular (e.g. Stoiber & Töller, 2016;Hörisch, 2018). The method combines the assumptions of set theory with the logic of Boolean algebra (Schneider & Wagemann, 2010, p. 404). In contrast to quantitative methods, we do not speak of dependent and independent variables, but of an outcome and conditions for the occurrence or non-occurrence of the outcome.
This method allows for identifying necessary and sufficient conditions, or combinations of conditions that explain the occurrence or non-occurrence of the outcome for as many cases as possible. As a set-theoretic method, QCA distinguishes cases in terms of the presence or absence of a social phenomenon (Thomann & Maggetti, 2017, p. 5). Moreover, by taking into account the concept of equifinality, 3 this method offers 2 Data for the QCA were derived from qualitative case studies analyzing sources taken from the political process, four semi-structured interviews with the responsible ministries in Brandenburg, Schleswig-Holstein, Thuringia, and Mecklenburg-Western Pomerania, as well as few secondary analyses (in particular Wächter-Raquet, 2016). 3 Or in QCA logic, different conditions which in combination are sufficient or necessary.
explaining the (non-)occurrence of an outcome through different paths. A further argument for the chosen method is its case sensitivity, which in contrast to quantitative methods, requires a more in-depth knowledge of cases (Thomann & Maggetti, 2017, p. 4).
In addition to the original form of the QCA with crisp sets, there are further variants of the method (e.g. fuzzy-set QCA, multi-value QCA, and temporal QCA). QCA's classic form, the crisp-set variant, uses dichotomized data. In our case, the advantage of fsQCA lies in the representation of the differentiated expression of the relevant outcome: while some states were never interested in commissioning the health insurance companies, others sought to do so but failed to negotiate a framework agreement; furthermore, there are also differing degrees of the involvement of health insurance companies within a state. In some states, only a small or a medium proportion of municipalities entered into a framework agreement whereas in other states this occurred on a large scale. As a dichotomous operationalization of the outcome would obfuscate these variances of the phenomenon, we chose to carry out a fuzzy-set QCA. It allows the introduction of so-called "fuzzy sets" which, in addition to distinguishing between membership (1) and non-membership (0), allows for distinguishing partial memberships, thus enabling subtle differentiations (Ragin, 2008, pp. 71-73.). At the same time, the value 0.5 forms a qualitative threshold value between membership and nonmembership-accordingly, this value is not allocated during calibration (Schneider & Wagemann, 2012, pp. 32-35.). The use of this QCA variant enables us to depict the described qualitative differences of the outcome by assigning fuzzy values.
To perform the fsQCA, we created a data set 4 that includes qualitative as well as quantitative data. When calibrating qualitative data (the outcome, the partisan composition of state government, and the cost reimbursement regulation) we allocate fuzzy values based on theoretical considerations that reflect empirical reality as precisely as possible. To do so, we use a calibration with four values between the values 0 and 1 consisting of the gradations 0.33 and 0.67 (see the calibration tables in the Online Appendix 2). To calibrate the metric and interval-scaled data-in our case the proportion of foreigners, the unemployment rate, the per capita gross domestic product, and the per capita debt-we summarized the data first into four clusters with the help of hierarchical cluster analysis in the SPSS software using the quadratic Euclidean distance. The cluster analysis aims at forming groups whose members have a high degree of homogeneity concerning the characteristic value. In a second step, the "significance" of the anchor points 0 and 1 were determined (e.g. GDP 0 = low GDP, GDP 1 = high GDP). The two clusters lying the furthest away from each other form the outer boundaries. We allocated the values 0.33 and 0.67 to the clusters closest to the outer boundaries.
We carried out the fuzzy-set QCA using the fsQCA 3.0 software developed by Ragin and Davey (2016). As threshold values for consistency, 5 the values commonly used in the literature are 0.8 for sufficient conditions and 0.95 for necessary conditions. The practical implementation followed the procedure described by Ragin (2010). First, we carried out a test to determine the conditions necessary for the outcome. Next, a test 4 See Online Appendix 1 "Data matrix". 5 Consistency indicates the extent to which a coincidence of a condition and the outcome correspond to the pattern of a necessary or sufficient condition whereas coverage indicates how many cases from the data set can actually be explained with the relevant condition (Schneider & Wagemann, 2012, pp. 119-148).
for sufficient conditions was carried out in which a so-called truth table is produced, which should then lead to a (parsimonious) solution to explain the outcome.
The outcome: the introduction or non-introduction of the alternative model As mentioned above, healthcare provision for asylum seekers in the first 18 months of their stay in Germany is regulated in the German Asylum Seekers' Benefits Act (Asylbewerberleistungsgesetz, AsylbLG), which is a federal law. In section 10 AsylbLG, the federal legislator delegated the implementation to the states, which in most instances delegated these tasks fully or in part to the municipalities. Under section 4 (1) AsylbLG, benefits are limited to necessary treatment for acute illness and pain. Unless otherwise provided for, the standard operating procedure for the implementation provides that an asylum seeker has to visit the municipal social security office when he or she needs medical assistance. 6 In this case, refugees must apply for a certificate of entitlement to treatment (valid for 3 months). This entitles them to make an appointment with a doctor in practice. However, before this can come about, they must also possess a treatment certificate, which the doctor often requests from the authorities while the patient is sitting in the waiting room (Schammann, 2015, p. 175). If the authorities find that the requirements of the AsylbLG are met, they fax the treatment certificate to the practice. If doubts arise, the local public health department is consulted. Prescriptions for medication require further approval by the authorities. 7 This bureaucratic effort combined with the lack of medical skills in the social service authorities can make effective healthcare provision more difficult. What is more, this model leads to the stigmatization of asylum seekers in doctors' practices (e.g. Lindner, 2015). Likewise, these processes were challenged for being ineffective and inefficient in terms of administrative procedures, as for instance, the regional audit office of Hamburg criticized (Burmester, 2015, p. 195).
The alternative model, which was first introduced in the city-state of Bremen in 2005 and then in the city-state of Hamburg in 2012, provides that municipalities commission health insurance companies with healthcare provision for asylum seekers. Asylum seekers-just like people with statutory health insurance-obtain an electronic healthcare card with which they can visit the doctor's practice. The asylum seeker's limited benefit entitlement is stored on this card for use in the practice. In this model, it is the doctor who decides whether an illness is acute and treatment necessary. This model simplifies the procedure of healthcare provision, even though there is still a substantial limitation to benefits. To introduce this model, the respective state Ministry of Health must conclude a framework agreement with health insurance associations. This agreement regulates the cover of the incurred treatment costs and provides for appropriate reimbursement of the administrative expenses of the health insurance companies (Wächter-Raquet, 2016, p. 16). The municipalities carry out the concrete commissioning of the health insurance companies. In 2015, the federal legislator passed an amendment to section 264 (1) of the German Social Insurance Code (Sozialgesetzbuch) Book V in order to facilitate the implementation of this model. Accordingly, healthcare insurance companies are now obliged to be commissioned with providing healthcare to asylum seekers by the state government (German Bundestag paper [BT-Drs.] 18/6185, 60).
As to whether, how, and when the remaining 14 states switched or considered switching from the standard to the alternative model between 2015 and 2016, 8 we identified four different groups: Group 1: States in this group did not consider switching from the standard model to the alternative model. This applies to Bavaria, Saxony, and Baden-Wuerttemberg since March 2016. Group 2: These states considered switching from the standard model to the alternative model and also entered into negotiations with health insurance companies (and in part with municipalities), but failed to reach an agreement; therefore, the states continued using the standard model. This applies to Mecklenburg-Western Pomerania, Saarland, Hesse, Saxony-Anhalt, and Baden-Wuerttemberg until March 2016. Group 3: These states considered switching from the standard model to the alternative model and concluded a corresponding framework agreement with health insurance companies. However, the-required-implementation in the municipalities remained incomplete. This applies to Lower Saxony, where only one out of 45 municipalities implemented this model, and to Rhineland-Palatinate where only three out of 36 municipalities implemented it. We also include North Rhine-Westphalia in this category where 22 out of 53 municipalities have commissioned health insurance companies with healthcare provision for asylum seekers, but this has only come into effect for 26% of the respective persons in North Rhine-Westphalia. Group 4: These states implemented the alternative model on a large scale, i.e. they have concluded a framework agreement at state level and have subsequently commissioned health insurance companies in all municipalities. This is the case in Berlin, Schleswig-Holstein, Brandenburg, 9 and Thuringia.

Conditions for the introduction of the healthcare card
As mentioned, comparative migration policy research is debating whether, and if so, under which conditions, partisan composition has an effect on migration policies. Alternatively, studies examine the impact of the economic situation (economic strength, unemployment), the extent of immigration, and the role of political and institutional factors (federalism, proportional representation) (e.g. De Haas & Natter, 2015). In US immigration federalism, studies use the degree of professionalization of the federal legislature (Reich, 2019;Zingher, 2014;Marquez & Schraufnagel, 2013;), crime rates (Creek & Yoder, 2012), the influence of lobbying groups (Butz & Kehrberg, 2019;Commins & Wills, 2017), the proportion of trade union members in the population, and the level of education in a state (Marquez & Schraufnagel, 2013) as alternative explanations. However, our method limits the number of explanatory factors; according to the Boolean logic, which the QCA is built on, the combination of possible conditions increases exponentially with the introduction of every additional condition, so that from a certain point on an analytical reduction is no longer possible (Marx et al., 2014). 8 Baden-Wurttemberg is a special case as the government initially considered negotiations but failed to conclude on a framework agreement. It ruled out further initiatives following the change of government in 2016. We therefore assigned Baden-Wurttemberg to Group 2 in the first phase and Group 1 in the second phase. 9 As a city-state, Berlin is both a state and a municipality in one. As 17 out of 18 municipalities in Brandenburg participate in the model, we have allocated this state to Group 4.
We, therefore, limit ourselves to the factors, which can be considered as explanatory factors according to case studies and research done on Germany.

Partisan theory
Partisan theory is a central explanatory approach in comparative policy studies. Established by Hibbs (1977), it has since been expanded theoretically in many ways and empirically comprehensively tested in many policy areas. Accordingly, political parties' policy positions can be systematically distinguished along definable cleavages. This is because parties have different ideological roots and they court specific voters with specific expectations. When parties are in government, they try to translate these positions into policy. 10 Three mechanisms were identified for this: vote-seeking (i.e. implementing what voters expect), office-seeking (striving for office to implement policies), and policy-seeking (implementing what the party ideology and political objectives suggest) (Müller & Strøm, 1999). Consequently, policies should systematically differ according to which parties are in government and a change of government should correlate with policy changes.
In an international comparison, De Haas and Natter (2015) find hardly any effect of the government's political hue on the migration policies of 21 European and North American states. Significant positive correlations only exist between governments with dominant left-of-center parties and liberal border control regulations and between coalition governments comprising left-of-center and right-of-center parties and liberal integration measures (De Haas & Natter, 2015, p. 17). The partisan effects on the various migration policy measures of the 50 US states are also ambiguous. Zingher (2014) found Republican-controlled legislators increase the likelihood that a state will tighten control over its residency laws. Another study showed that a government dominated by Democrats correlates with a lower number of restrictive laws but not with an increase in permissive laws (Marquez & Schraufnagel, 2013). In Reich's study, a change in government to a Republican-controlled government is accompanied by a more restrictive prosecution of migrants, but also by a positive effect on the extension of healthcare provision for migrants (Reich, 2019, pp. 562-566). Other US studies do not find any evidence of partisan effects (Butz & Kehrberg, 2019;Commins & Wills, 2017;Ybarra et al., 2016).
In terms of German migration policy, Henkes (2008) showed that German conservative state governments clearly have a negative effect on naturalization rates. In Wolf's study of hardship commissions on the right of residence, the establishment of commissions by left-of-center governments but also by social democratic interior ministers has a negative effect on the success rate of these procedures whereas their establishment by a grand coalition government has a positive effect (Wolf, 2013, pp. 141-145). In Hörisch's study (Hörisch, 2018), a left-of-center state government proves to be "part of a sufficient explanation path for high expenditure per refugee" (p. 798). Also, conservative governments alone are sufficient for deportations to Afghanistan (Hörisch, 2018, p. 798). Hörisch and Heiken identify left-of-center state governments as a sufficient condition for a high proportion of decentralized accommodation for asylum seekers (Hörisch & Heiken, 2018, p. 222). Recently, Meyer et al. find that a liberal party program of the government is a necessary condition for permissive migration policies in German states (Meyer et al., 2021).
As a plausible initial hypothesis, we assume that left-of-center parties are more likely to pursue permissive migration policies whereas right-of-center parties pursue a rather restrictive policy, from which the following hypothesis H1 can be derived: H1: In states with left-of-center governments, health insurance companies are commissioned with providing healthcare for refugees. This does not occur in states with right-of-center governments.
The usual operationalization of the partisan composition of the state governments occurs through coalition compositions. In German state parliamentary systems, the partisan composition of the government reflects the government majority in Parliament. We distinguish between left-of-center and right-of-center governments as well as partial membership in cases where coalition governments exist across political blocks. Thus, our calibration distinguishes between state governments that are a) solely composed of conservative parties (this is only the case in Bavaria: the CSU), b) composed of coalition governments led by a conservative party (CDU) with the participation of leftof-center parties as junior partners (SPD, Greens), c) coalition governments led by leftof-center parties (SPD, Greens) with the participation of conservative parties (CDU) as junior partners, and d) state governments composed solely of left-of-center parties (SPD, Greens, and Die Linke [The Left]). We assign the fuzzy values of 0, 0.33, 0.67, and 1 to these four compositions.

Institutional factors: financial relations
From the perspective of rational choice institutionalism, institutions (understood as structures and rules) represent a framework within which rational, benefit-maximizing actors select their strategies (e.g. Peters, 2012, pp. 47-69). Nevertheless, these assumptions are quite general and completely different institutions will play a role depending on the research field (e.g. Stoiber & Töller, 2016). In our case of health care provision within Germany's federal system, the-institutionally anchored, completely different-financial relations between states and their municipalities are of particular relevance as an institutional factor in terms of benefits under the Asylum Seekers' Benefits Act (AsylbLG).
The municipalities are delegated the tasks resulting from the AsylbLG and initially incur the costs of these mandatory tasks of e.g. accommodation, subsistence benefits, healthcare costs. Following the principle of connectivity (Art. 104a of the German Basic Law [GG]), states will obtain financial compensation from the federal state and likewise make the financial means available to municipalities (Schammann, 2018, p. 79). However, the reimbursement arrangements vary considerably between states. For our study, the financial framework and its reimbursement rules for healthcare benefits pursuant to the AsylbLG are an institution that could influence the perception of problems and the interest municipalities have in commissioning health insurance companies or not. By commissioning health insurance companies, specific decisions on the acuteness of an illness and the necessity of treatments are transferred to doctors and health insurance companies. In doing so, municipal social security offices lose their gatekeeper function and could thus fear cost increases. Whether municipalities perceive this as a financial risk that would potentially lead to a rejection of this model could be due to the specific rules about cost reimbursement. For this reason, the H2 hypothesis is as follows: H2: In states in which the state reimburses municipalities for the expenditure actually incurred for benefits under the AsylbLG, the health insurance companies are commissioned with healthcare provision for refugees. This does not occur in states where municipalities are only reimbursed a flat rate for healthcare costs.
We differentiate between the reimbursement rules that are mainly determined by the relevant Refugee Reception Acts of the states (Hummel & Thöne, 2016), subsequently the extent to which they place the burden of the risk of the new regulation on the municipalities or the state. There are a) reimbursement rules, which place the risk of specific costs fully on municipalities by providing for flat-rate amounts. This is the case in Baden-Wurttemberg and Lower Saxony. We assign a value of 0 to these states. There are b) reimbursement rules, which partly mitigate this risk burden with the municipalities by providing special reimbursement rules for high-cost healthcare provision cases in addition to the flat-rate amounts. This is the case in Hesse, Saxony-Anhalt, Saxony, Rhineland-Palatinate, and North Rhine-Westphalia. These states receive a value of 0.33 because they only cushion the risk in a few particularly expensive cases, but not in a large number of cases. Finally, there are c) reimbursement rules, which provide for full reimbursement of the costs incurred and thus relieve the municipalities of the full risk of the new constellation. Brandenburg fully reimburses healthcare benefits (Brandenburg State Reception Act [LAufnG] section 15). In Thuringia, the state likewise settles the costs of the healthcare benefits directly in full via a peak settlement with the health insurance companies (Thuringia Refugee Reception Act [ThürFlüAG] section 2 (2) No 3). In Bavaria, Mecklenburg-Western Pomerania, and Saarland, the state reimburses all of the costs incurred for healthcare benefits as a peak settlement. In Schleswig-Holstein, the state has been taking on 90% of the municipal costs for benefits under AsylbLG as of 01/01/2016 since the financial relations were amended at the end of 2015 (instead of 70% previously; Ministerium für Inneres und Bundesangelegenheiten Schleswig-Holstein, 2015). Even if this does not represent a peak settlement in the strict sense, the 90% cost coverage justifies the classification as a reimbursement rule that relieves municipalities. In Berlin, there is no municipal administration level so that there is likewise no financial risk here for the municipal level; therefore, we will likewise include Berlin in this category. We assign the value of 1 because in this case, the financial risks to municipalities as regards healthcare treatment costs range from very low to non-existent.

Share of foreigners
Studies that aim to explain the emergence of permissive or restrictive migration policies usually examine the share of foreigners in the population as an explanatory factor (e.g. Riedel & Schneider, 2017;Hörisch, 2018, p. 789). Apart from the fact that this variable itself is strongly influenced by law and practice of naturalization, there are, however, quite different assumptions both about the direction in which the share of foreigners is supposed to act and about the causal mechanism assumed behind it.
The racial threat hypothesis assumes that high numbers of foreigners lead to restrictive measures. More precisely it is the rapid increase in particular population groups that fosters negative attitudes among native citizens and ultimately gives rise to restrictive legislation against migrants (Butz & Kehrberg, 2019;Commins & Wills, 2017;Ybarra et al., 2016;Marquez & Schraufnagel, 2013;Creek & Yoder, 2012). Although this assumption does not seem implausible on a general note, one would have to look more at the increase in size of specific groups than at the share of foreigners generally. More important, it is difficult to argue that there are systematic differences between the German states, which in turn could explain differences in migration policies. This is because in Germany asylum seekers are distributed among the states according to the 'Königsteiner Schlüssel', based on population size and tax revenue. Therefore, states cannot be affected by a disproportionate influx of asylum seekers that could explain the adoption of different policies.
There are, in turn, two ways in which high shares of foreigners could be associated with permissive migration policies whereas low shares lead to restrictive measures.
First, the contact hypothesis suggests that negative role schemata are dismantled more easily in societies with a high proportion of foreigners (Ellison et al., 2011), at least if there are direct or indirect contacts between individuals, which should lead to a higher acceptance of permissive migration policies. Indeed, data on the segregation of foreigners in German cities and regions indicate that migrants living in states with a higher proportion of foreigners are less segregated from the native population than those living in states with a lower proportion of foreigners (see Baldewein & Keller, 2020, pp. 15-18). These findings support the contact hypothesis implying that frequent contacts allow for changing normative and cultural attitudes, which again foster support for permissive policies. Since the share of foreign population varies considerably between the states, this could indeed account for differences in policies.
Second, in line with more general functionalist explanations in policy research (e.g. Wilensky, 2002), it is suggested that higher shares of foreigners produce a higher "functional need" for pro-integration policies (Dobbin et al., 2011) even though the empirical support for this argument is rather weak (Dobbin et al., 2011, p. 387). This would for our study, however, imply not to look at the share of foreigners in general but at the share of asylum seekers in the first phase of their stay who would benefit from facilitated health provision for this group. Looking at our data, it is not fully implausible that this factor played a role. However, we observed that variance occurred rather on the time axis, i.e. politicians gave the issue of health provision to newly arriving asylum seekers high relevance as long as their numbers were high, and less, when they were decreasing. Yet, we do not see variance between the states that could explain different policies, because, as outlined above, the numbers of newly arriving asylum seekers in each state were determined according to the 'Königsteiner Schlüssel'.
Summarizing our discussion, we favor the contact hypothesis over other theoretical assumptions because this is the only option that provides us with variance between the states that could possibly explain the adoption of either permissive or restrictive policies. Accordingly, we formulate hypothesis H3 as follows: H3: In states with a high proportion of foreigners, health insurance companies are commissioned with the healthcare provision for refugees. This does not occur in states with a low proportion of foreigners.
We carried out a cluster analysis for the proportion of foreigners in the overall population of the states in 2015 based on data from the German Federal Statistical Office and four clusters were determined. The first cluster includes the eastern German states and covers a low proportion of foreigners of between 3.11% and 3.33%-a value of 0 is determined for cases in this cluster. The second cluster includes cases with a somewhat higher proportion of foreigners of between 5.8% and 8.8% (Lower Saxony, Rhineland-Palatinate, Saarland, and Schleswig-Holstein)-a value of 0.33 is assigned to this cluster. Whereas these two clusters include states with a low proportion of foreigners, the next two clusters depict states that tend to have a higher proportion of foreigners. The third cluster shows the value range of between 10.97% and 11.1% and includes the states of Bavaria and North Rhine-Westphalia-this cluster is calibrated with a value of 0.67. The last cluster comprises Berlin, Hesse, and Baden-Wuerttemberg, states with the highest proportion of foreigners (of between 13.2% and 14.9%) and is thus calibrated with a value of 1.

Socio-economic factors
Ultimately, previous studies have assumed that a state's economic situation effects a society's willingness to integrate migrants (Riedel & Schneider, 2017, p. 31;Hörisch, 2018, p. 788). Accordingly, wealthy states are more likely to be pro-integration whereas poor states tend to restrict migrants' access to benefits. In the US context, this assumption has been confirmed in terms of the expansion of healthcare provision for migrants, less stringent criminal prosecution measures applicable to migrants and less restrictive immigration policies (Reich, 2019;Commins & Wills, 2017). In a similar vein, Ybarra et al. (2016) argue in their study investigating the impact of the 2007 to 2009 economic crisis on US states' immigration policy that states particularly affected by the economic crisis approved increasingly restrictive immigration measures. By contrast, other studies find that the economic crisis or the unemployment rate have no effect (Butz & Kehrberg, 2019;Zingher, 2014;Creek & Yoder, 2012).
In the German context, Wolf's study on hardship case proceedings confirms the assumption that the GDP has a positive impact on the approval and success rate of hardship case proceedings (Wolf, 2013, p. 141). In Hörisch's study, this has, by contrast, no significant effect on spending for asylum seekers and deportation policy (Hörisch, 2018, 798). Hörisch and Heiken identify higher GDP as a sufficient condition for a higher proportion of decentralized accommodation for asylum seekers (Hörisch & Heiken, 2018, p. 222). Riedel and Schneider correlate a high unemployment rate with lower acceptance rates (Riedel & Schneider, 2017, p. 42). In any case, it seems to make sense to examine the impact of a state's economic wellbeing, which we measure using the unemployment rate and the GDP (e.g. Hörisch, 2018, p. 789). As the willingness of the municipalities, who bear the cost of healthcare benefits under AsylbLG, to cooperate is key in our case, we examine a state's debt level as an additional indicator of its socio-economic situation. Contrary to the theoretical assumption, Riedel and Schneider (2017, p. 42) also find a significant positive correlation between the debt level of a state and a higher acceptance rate. As a proxy for the fiscal situation, we operationalize the debt level as a per capita debt for municipalities or municipal associations and the state. 11 Thus, Hypotheses 4a, 4b, and 4c are as follows: H4a: In states with a low unemployment rate, health insurance companies are commissioned with healthcare provision for refugees. This does not occur in states with a high unemployment rate.
Unemployment rates in states range between 3.6% (Bavaria) and 10.7% (Berlin). Based on a cluster analysis, we assign a value of 0 to states with an unemployment rate of 10.2% to 10.9%. A second cluster includes states with unemployment rates of 7.2% to 8.7% to which we assign a value of 0.33; states with unemployment rates between 5.2% and 6.5% obtain a value of 0.67 and states that have unemployment rates between 3.6% and 3.8% are assigned a value of 1.
H4b: In states with a high GDP, health insurance companies are commissioned with healthcare provision for refugees. This does not occur in states with a low GDP.
Per capita GDP ranges from €25,232 in Mecklenburg-Western Pomerania to €43,365 in Bavaria (Statistisches Landesamt Baden-Wuerttemberg, 2019). Based on a cluster analysis, we assign states with a per capita GDP of between €25,232 and €28,040 a value of 0. The second category includes a per capita GDP of €30,332 to which we assign a value of 0.33. The third cluster contains states that have a per capita GDP of between €33,971 and €36,559; these obtain a value of 0.67. States with a per capita GDP of between €42,654 and €43,365 obtain a value of 1.
H4c: In states with a low debt level, health insurance companies are commissioned with healthcare provision for refugees. This does not occur in states with a high debt level.
The debt level of municipalities and states per inhabitant ranges from €1279 in Saxony to €17,391 in Saarland. Based on a cluster analysis, the following clusters and assigned values result: states with a per capita debt level between €16,819 and €17,391 obtain a value of 0 while states with a debt level between €8863 and €11,154 obtain a value of 0.33. We assign a value of 0.67 to states in which the per capita debt ranges between €6993 and €7964, and a value of 1 to states in which it ranges between €1279 and €4243.

Analysis
In part one, we present results for the "introduction of the alternative model" outcome where health insurance companies were commissioned with healthcare provision for asylum seekers. Part two presents the results for the complementary outcome, the "non-introduction of the alternative model".
Determining factors for the "introduction of the alternative model" outcome Test for the necessary conditions Table 1 shows that the condition "left-of-center government" is the only one to achieve and exceed the (conservative) consistency value of 95% and is therefore assumed a necessary condition for the outcome. This confirms our hypothesis H1, which reflects the partisan effect. Only states with a left-of-center government commissioned health insurance companies with healthcare provision for asylum seekers. The "introduction of the alternative model" outcome is, therefore, a subset of the "left-of-center state government" condition, from which follows: Left−of −center state governments←commissioning of health insurance companies Furthermore, the test for necessary conditions for the economic conditions of GDP and unemployment shows that states with a high and a low GDP, as well as those with a high and a low unemployment rate, commission health insurance companies with healthcare provision for refugees. The alternative model was also introduced both in states with a high proportion of foreigners (Berlin, North Rhine-Westphalia, and Lower Saxony) as well as in states with a low proportion of foreigners (Brandenburg, Thuringia, and Schleswig-Holstein), which contradicts the H3 contact hypothesis.

Test for sufficient conditions 12
After the previous section identified the necessary condition "left-of-center government", the next test identifies sufficient conditions and combinations of conditions. The intermediate solution (see Online Appendix 3) results in four consistent solution paths, three of which are fully consistent with a value of 1. As the findings of the intermediate solution are ambiguous, especially concerning the conditions GDP, unemployment rate, and the proportion of foreigners, and because each of the solution terms only covers a small portion of the cases, we will study only the parsimonious solution (Table 2) in more detail below. In this parsimonious solution, the four solution paths are reduced to two parsimonious solution terms with the largest possible coverage and consistency. 13 The parsimonious solution comprises the two solution terms left-of-center government in conjunction with a high debt level for the state and the municipalities as well as a left-ofcenter government in conjunction with a cost reimbursement regulation. When studying the respective cases of the two solution terms covered, it can be noted that, apart from Berlin and Schleswig-Holstein, the first solution path only covers states in which a framework agreement was indeed concluded, but in which municipalities did not participate across the board. However, for the cases of Berlin and Schleswig-Holstein, cost reimbursement also exists in addition to the high debt level, which is why the second parsimonious solution term covers these cases. Besides Berlin and Schleswig-Holstein, the second solution term also covers Brandenburg, Thuringia, and Mecklenburg-Western Pomerania. This solution term shows lower consistency than the first, caused by the case of Mecklenburg-Western Pomerania. In this case, the relevant conditions are present but yield no outcome. Mecklenburg-Western Pomerania thus constitutes a deviant case that our fsQCA-model cannot explain.
Overall, the parsimonious solutions show that states with a left-of-center government and a high debt level have indeed concluded a framework agreement with health insurance companies; however, this has only amounted to incomplete participation by municipalities. On the other hand, apart from Mecklenburg-Western Pomerania, all states with a framework agreement, in which the state reimburses municipalities for the healthcare costs actually incurred for benefits under AsylbLG, show widespread implementation by municipalities Berlin,14 Thuringia,and Brandenburg). From this, we infer that in states with option solutions, a high debt level without a cost reimbursement regulation explains why only a small proportion of municipalities have opted in. Evidently,  (Schneider & Wagemann, 2012, p. 248). In order to counter this problem, we also consider the PRI consistency, which shows no abnormalities for the indicated solution terms (it always lies above the threshold of 0.8). Hence, we can rule out the probability of erroneous conclusions resulting from the impact of irrelevant cases in our study. 14 As a city-state, Berlin is a special case because no autonomous municipal level exists.
without reimbursement of the actual costs by the state and because of their financial situation, municipalities had concerns they would be further burdened in the event of increasing healthcare provision costs. It can therefore be assumed that a high debt level in connection with a left-of-center government does in fact lead to the conclusion of a framework agreement at state level. However, only a cost reimbursement for healthcare pursuant to AsylbLG enables widespread participation at municipal level.
Determining factors for the "non-introduction of the alternative model" outcome In addition to the above results, we present the results for the complementary outcome, i.e. non-introduction of the alternative model. Again, we first perform a test to determine the necessary conditions and subsequently identify sufficient conditions and combinations of conditions.

Test for the necessary conditions
The test for the necessary conditions for the complementary "non-commissioning of health insurance companies with healthcare provision for refugees" outcome shows that none of the conditions achieves the consistency for a necessary condition. The highest consistency constitutes the "low debt level" condition, which, at 0.81, is far below the threshold of 0.95 (Table 3).

Test for sufficient conditions 15
The following (parsimonious) solution term was determined as a sufficient condition (Table 4): The test for the sufficient conditions for the outcome results in the intermediate solution in two consistent 16 solution paths that cannot be reduced any further in the parsimonious solution. This means that the first solution path only contains the condition "rightof-center government coalition" and covers the cases of Bavaria, Saarland, Hesse, Saxony-Anhalt, and Saxony. It follows that the condition "right-of-center state government" is a sufficient condition for non-commissioning health insurance companies with healthcare provision for refugees. The second consistent solution path consists of the two conditions of low debt level and flat-rate settlement and covers the cases of Baden-Wuerttemberg (in both its 2015 and its 2016 composition), and Saxony. From this, it follows: Right−of −center state government þ low debt levelÃflat−rate settlement→Non−introduction of the alternative model The result for the complementary outcome again illustrates the strength of the partisan effect and the differences between a left-of-center and right-of-center government in migration policy.

Summary & discussion
The results of the analysis for the "introduction of the alternative model" outcome show that a left-of-center state government is a necessary condition for the occurrence of the outcome. This confirms our H1 hypothesis: only left-of-center governments or governments led by the left, commission health insurance companies with healthcare provision for refugees (partisan theory). Furthermore, the test for sufficient conditions identified the two condition combinations of a left-of-center government and a high state and municipal debt level (H4b) as well as a left-of-center government and the existence of a cost reimbursement regulation for healthcare benefits under AsylbLG (H2). It demonstrated, particularly in cases in which a cost reimbursement regulation existed, that widespread participation by municipalities can also be determined. We summarize that without a simultaneous cost reimbursement regulation, the debt level of states and municipalities poses a potential obstacle to the willingness of municipalities to join the corresponding framework agreements. This confirms our H2 hypothesis. On the other hand, regarding the H4(a-c) hypotheses, which focus on socio-economic factors, neither GDP (H4c) nor the unemployment rate (H4a) was relevant, and contrary to our hypothesis, a high debt level (H4b) is significant for the introduction. The contact hypothesis, which was operationalized by the proportion of foreigners, could not be confirmed either.
However, the results also identified a deviant case that we could not explain using fsQCA and hence needs to be examined in more detail. In the case of Mecklenburg-Western Pomerania, although the necessary condition (left-of-center state government) and the sufficient condition (left-of-center state government and complete cost reimbursement by the state) were present, health insurance companies were not commissioned. Thus, why did the factors not produce the expected outcome in this deviant case?

Conclusion
Comparative research on migration policy has been particularly interested in the effect of partisan politics on migration policy. However, the assumption resulting from partisan theory that left-of-center parties tend to approve permissive migration policies whereas right-of-center parties tend to pursue restrictive migration policies only partly holds in empirical studies on various jurisdictions. Yet, several studies point to a potential partisan effect on the migration policies of German states. Against the backdrop of this debate, this study investigated the conditions under which German states provide healthcare for asylum seekers. On the backdrop of a generally restricted access of this group to healthcare provision, the states apply either a bureaucratic, restrictive, model or an alternative, permissive model that commissions health insurance companies. Besides partisan theory, we examined the role of the other 'usual suspects' in comparative policy research and migration policy studies, namely the institutional constellations (in this case the financial relations between states and their municipalities), the proportion of foreigners ("contact hypothesis"), and the socio-economic situation in the respective state. We examined the factors resulting from these explanatory approaches using fuzzy-set qualitative comparative analysis to determine whether they alone or in combination of the factors represent a necessary or sufficient condition for the adoption of the alternative model.
Our results are unequivocal: a left-of-center state government was in office in all cases in which a framework agreement was agreed at state level. In most cases, these were left-ofcenter coalition governments, composed solely of left-of-center parties (SPD, Greens, and Die Linke). In one case (Berlin), there was a grand social-democratic/conservative coalition which we included in left-of-center governments (and assigned a value of 0.67) because it was led by the social democrats (SPD). Yet, not all left-of-center governments have implemented this measure and so partisanship does not represent a sufficient condition. Rather, we identified complete cost reimbursement by the state to be a condition, which, in conjunction with a left-of-center government, leads to the outcome as a sufficient condition. In other words, the cost reimbursement rule is, in turn, the decisive factor for whether municipalities in states with a left-of-center government and in which a framework agreement existed with the health insurance companies are also willing to implement the model across the board. 17 This appears rational as many municipalities expressed concerns that with the new model, they would lose control over healthcare costs if social security departments no longer decide which illness is acute and which treatment is necessary.
Such an unequivocal partisan effect on healthcare provision for asylum seekers by German states represents an important contribution to the debates on whether party ideologies actually account for variance of migration policies in cross-national (De Haas & Natter, 2015) or subnational comparisons (Reich, 2019;Butz & Kehrberg, 2019;Zingher, 2014). Whereas in most comparative literature, clear partisan effects remain a rare finding, for Germany, partisan effects on migration policies have been identified in a number of studies (e.g. Henkes, 2008;Wolf, 2013;Hörisch, 2018;Meyer et al., 2021). Our finding is also relevant given the rise of right-wing populist parties which may result in possible contagion effects in the field of migration policies in general and the provision of health care to refugees in particular (Rooduijn et al., 2014;Schumacher & van Kersbergen, 2014;Falkenbach & Greer, 2018;Cammaerts, 2018). While the results of our study so far argue against a contagion effect, we see an urgent need for further research in this area, as populist parties are becoming more established, and there is a possibility that they influence migration policy discourses, party positions, and ultimately policies in their favor.
With regard to other cases, our study allows for two conclusions. First, health care provision for refugees is a likely case for partisan effects, which makes comparative studies that not only compare health care provision to refugees (e.g. Biddle et al., 2020) but also explain the differences an important task. Second, if we want to assure that in spite of the success of populist parties partisan effects prevail, time must be included in such studies. This would, however, require a different methodological approach than the one we applied, due to the limits of the fuzzy-set QCA in terms of time.
Additional file 3: Online Appendix 3. Intermediate and Complex Solution.