What Drives the Immigration-Welfare Policy Link? Comparing Germany, France and the United Kingdom

Western European states have increasingly linked immigration and welfare policy. This trend has important implications for European welfare-state trajectories, but accounts of the policy reasoning behind it have diverged. Are policymakers attempting to delimit social citizenship to secure welfare-state legitimacy? Pursuing new, market-oriented welfare-state goals? Symbolically communicating immigration control intentions to voters? Or attempting to instrumentally steer immigration flows? These accounts have rarely been tested empirically against each other. Redressing this, we employ 83 elite interviews in a comparative process-tracing study of policies linking welfare provision and immigration status in Germany, France, and the UK during the 1990s. We find little evidence suggesting welfare-guided policy reasonings. Rather, this policy linkage appears “immigration-guided:” meant to control “unwanted” immigration or resonate symbolically in immigration politics. Differences in exclusions from welfare support for migrants grew from existing national differences in welfare-state design and politicizations of immigration, not from policy intentions, which were largely shared.


Introduction
The linking of welfare-state and immigration policy domains-for instance, by making immigration status a central criterion for social provision, or requiring information exchange between immigration and welfare bureaucracies-has been a prominent trend in Western Europe amid new migration movements following the end of the Cold War. The United Kingdom's "hostile environment" for immigrants (Hiam et al., 2018), restrictions on welfare provision for asylum seekers in countries like Sweden (Scarpa & Schierup, 2018), and the incomplete inclusion of European Union migrants in social protection schemes in countries like Germany, Austria, and Belgium (Heindlmaier & Blauberger, 2017;Lafleur & Mescoli, 2018) demonstrate how robustly this trend endures. A large literature analyzes the effects of increasing migration on political support for welfare states (Nannestad, 2007;Soroka et al., 2016), aiming to discern how this policy intersection will affect social provision in Western democracies.
While the welfare-immigration policy linkage is clear and consistently captures political and academic interest, it is less clear whether there is a shared political logic behind it across states. Instead, past approaches to this topic have produced plausible-but distinct and contrasting-explanations for it. We identify four such explanations in past scholarship that serve as broad hypotheses regarding these welfare-immigration policy links: First, that such policies seek to delimit social citizenship, to address perceived challenges to the legitimacy of the scope of welfare provision; second, that they aim to further new welfare-state policy goals amidst retrenchment or market-oriented reform; third, that they are symbolic immigration politics, communicating hostility to immigration to a domestic political audience, without primary concern for policies' effects; and fourth, that they are instrumental immigration policies, intended to steer migration dynamics.
Systematic analysis of these four accounts is rare. This is despite the differing implications these explanations have for growing discussions about how European welfare states will change amidst increased transnational mobility (Kymlicka, 2015;Sciortino & Finotelli, 2015), and how migrants will be affected (Dwyer et al., 2019). Which of those four political rationalities are empirically most crucial in driving officials to adopt the immigrationwelfare linkage, in countries that differ in their historical politicizations of migration and welfare-state designs? To redress this gap, we conduct a qualitative comparative study, using process-tracing methods, of the adoption of policies that incorporated immigration status as a criterion for social provision in Germany, France, and the United Kingdom in the 1990s, a key period when this trend consolidated in all three countries. We present three narrative case studies, reconstructing the deliberations leading to these policies, drawing on semi-structured interviews with 83 welfare-state and immigration policy actors. We then compare the cases in a section that tests them against those four accounts.
Overall, our findings strongly support the latter two, "immigrationguided" accounts. In all three cases, these linking policies were largely interior-ministry driven efforts to control "unwanted," spontaneous forms of immigration, or to communicate symbolically about immigration with a domestic political audience. Moreover, where welfare logics pertained, they were not exclusionary, against the major previous "welfare-guided" accounts. The ways in which migrants were excluded from welfare differed in each country, despite largely shared policy intentions. These shared "immigrationguided" logics, yet diverging welfare effects, carry implications for how scholars analyze the trajectory and consequences of this policy trend.

Four Accounts of the Immigration-Welfare State Policy Linkage
Why have Western European states increasingly linked welfare-state and immigration policy? This trend is strikingly prevalent, even though European countries share neither generic welfare-state designs nor historic politicizations of migration. Indeed, these links are not indisputably natural ones. The closed nature of immigration and citizenship (Bommes, 2012) and the inclusive tendencies of welfare-state bureaucracies (Guiraudon, 2000) suggest that the two sides of such policy links might naturally repel each other. We identify four main explanations in the literature on the intersection of immigration and welfare for why states elaborate such links. These accounts are not necessarily mutually exclusive, but they suggest four quite distinct political rationalities that may drive this process. We also identify several likely indicators for the presence of each form of policy reasoning in establishing these links, reflected in the key actors involved in the measures; the policies' mode of adoption; the problem conceptions policymakers articulate surrounding these policies; and policies' implementation. 1 The question, "How does immigration affect welfare states?" has become a popular starting point in immigration policy research (Boswell & D'Amato, 2012, p. 12). Mirroring this question's focus on the effects experienced by welfare states, the first two accounts we term "welfare-guided," since they read this link primarily through the lens of welfare states' politics and goals. The first account is one where these policies are intended to delimit social-citizenship boundaries, responding to perceived challenges to the social legitimacy of the scope of welfare provision. While welfare states are supposed to provide for human needs, they also rely on notions of social membership (Ferrera, 2005) which migration seems to complicate (Alesina & Glaeser, 2004;Bommes, 2012;Freeman, 1986). Where welfare provision to people perceived as "outsiders" becomes politicized, the social legitimacy of the scope of welfare provision may erode (Mau & Burkhardt, 2009;Reeskens & van Oorschot, 2012). By bringing the scope of social provision into greater alignment with such social-membership images, governments are effectively inscribing "welfare chauvinist" approaches (Banting, 2010, p. 798), preserving provision for natives while enacting relative exclusion of immigrants. In terms of key policy actors, we would expect welfare-state actors with strong interests in their policies' perceived social legitimacy-and especially welfare ministers and political advisers who prioritize public attitudes (Dahlström, 2011, p. 307)-to play key roles in linking welfare and immigration policy. In terms of problem conceptions, we would expect interview data to reflect that in policy deliberations, questions of social legitimacy were perceived as at stake for the welfare state. We would also expect, in terms of implementation, for ministries to evaluate these policies based on whether new policies for provision to non-nationals seemed to satisfy widely held images of social citizenship.
The second account focuses on how immigration has burgeoned as a political issue in Western democracies amidst an already-extant period of welfare state retrenchment (Soroka et al., 2016), questioning welfare states' fiscal viability, policy goals, and generosity, as immigration's challenges have simultaneously grown (Schierup et al., 2006). Here, more exclusionary stances toward immigrants aid the pursuit of new welfare-state policy goals, for instance reducing state spending (Huber & Espenshade, 1997;Taylor-Gooby, 2016), or moving away from supporting vulnerable groups toward incentivizing labor market participation (Lafleur & Mescoli, 2018). While such policies may affect migrants adversely (Heindlmaier & Blauberger, 2017), the policies' main purpose is neither controlling immigration per se, nor affirming particular social-membership definitions. Rather, these policies serve as a means to achieve new welfare-state policy goals where immigrant groups are ripe targets for realizing them. As indicators of this logic, in terms of key actors behind linking policies, we would expect leading bureaucrats in social-service agencies or treasuries to play substantial roles, as they are key in managing new welfare-state goals and leading on the design of policies to achieve them (Genieys & Hassenteufel, 2015). Regarding implementation, we would expect these policies' success to be evaluated based on welfare retrenchment outcomes. We would expect policymakers to articulate that scarcity of welfare-state resources was a key problem. In terms of the mode of policy adoption, the packaging of these policies within larger welfare-state reform proposals would likely indicate this logic.
In contrast, the next two accounts we identify foreground the other side of the link: they are "immigration-guided," rather than fundamentally welfareoriented. Substantial unwanted migration is liable to be seen by publics as a crisis, amid endemic difficulties in achieving immigration-policy goals (Cornelius & Tsuda, 2004). This dynamic underlies a third account of immigration-welfare links: that they are symbolic immigration politics, intended primarily to convey signals about immigration control to a domestic political audience, rather than to instrumentally achieve policy goals. In this view, excluding immigrants from welfare becomes one manifestation of a symbolic political discourse that avoids acknowledging the inevitability of permanently settled migrant populations (Faist, 1994). For political expediency, policymakers may adapt policies to resonate with strongly embedded public or political accounts of sensible immigration control, without expecting immigration control to improve (Slaven & Boswell, 2019). Tying in welfare provision "sends a message" (Calavita, 1996) about immigration control intentions, or makes a statement about immigration in the run-up to an election (Han, 2013). In contrast to the first account, this political rationality is distinguished by the intention to symbolize immigration control specifically, rather than any main intention to address welfare-state legitimacy issues. While, again, these policies may affect immigrants adversely, their main purpose is to demonstrate commitment to controlling immigration rather than to effectuate it. As indicators of this logic, we might expect implementation to be half-hearted even by bureaucracies that led in policy development, as rhetoric is "decoupled" from practice (Boswell, 2008). In terms of the mode of adoption, policymakers seeking public political settings above administrative ones might indicate this logic, suggesting policymakers prioritized public visibility for these measures. We would expect key policymakers to reflect that citizens were not seeing enough action being taken on immigration (Newton, 2015) as an important problem conception. In terms of key actors, we would expect political actors and ministers, who prioritize politically effective interventions, to take on relatively greater roles. While a leading role for social-service bureaucracies would likely indicate welfare-guided logics, conversely, policies aimed at "immigration-guided" symbolic political goals would likely be driven by immigration-control bureaucracies.
While the political symbolism of cracking down on unwanted immigration is clear, a fourth account reads these links as instrumental immigration policy-not mainly expressive, but primarily intended to achieve immigration control goals. As migrants navigate the "social space created by the differentiation of society" (Bommes & Sciortino, 2011, p. 224), the state attempts to close this space, bringing other social systems into policing migrations (Engbersen & Broeders, 2011). States see welfare as key to the livelihoods of many "unwanted" migrant populations, so they use welfare to try to control migration (Ataç & Rosenberger, 2019;Sales, 2002). The destitution and limited integration caused by such policies keep immigrants deportable (De Genova & Peutz, 2010;Geddes, 2000) or aim to deter them punitively (Squire, 2009). As indicators of this logic, in terms of key actors, we would again expect linkages to be driven by immigration-control bureaucracies. In terms of the mode of adoption, we would expect linking policies to be packaged within immigration-control reforms. As regards implementation, we would expect these policies' success to be evaluated in light of immigration-control data, and for there to be thorough interior-ministry attempts at implementation, but possibly dissension by social-service actors (Marrow, 2009). Policymakers expressing that the key problem was that immigration per se needed to be controlled-articulating discourses of deterrence, strengthened enforcement, or incentivizing departure-would also indicate this logic.
These explanations are not mutually exclusive; policies can be adopted for multiple reasons. However, they illustrate that the prevalence of immigration-welfare state policy links could be guided by political rationalities differing substantially in their premises and aims, creating uncertainty about which policymaking considerations drive this linkage. Despite this uncertainty, there has been little effort to characterize systematically which logics are most empirically important in the development of these links in Western Europe, or to consider how these rationalities might interact. With few exceptions (Izambert, 2018), studies of the intersection of welfare and immigration have not employed methods that expose policy reasonings, often focusing on exclusionary effects on migrants without researching state intentions. Where studies have tried to illuminate these intentions, they have often taken two other approaches: a functionalist one, where policies' effects on target populations are read as "shed[ding] considerable light on the broader intentions. . . of government policy" (Zetter & Pearl, 1999, p. 236); or broad systems-theoretical or political-economic reasoning about state objectives. Therefore, we know relatively little about why policymakers across national contexts have so frequently linked immigration and welfare policy.

Cases and Methods
In addressing this gap, we argue a comparative, process-tracing approach is particularly valuable. The analytical puzzle we have identified is a surplus of plausible accounts of the political rationalities underlying immigration-welfare state policy links. These require confirmatory or disconfirmatory evidence, toward building general theoretical propositions about the causal mechanisms driving this policy link. To provide this, we engage in purposeful sampling of "typical case" studies-ones that appear representative of purported cross-case relationships between immigration and welfare policy-as the most suitable case-selection logic for addressing this research question (Seawright & Gerring, 2008, pp. 299-300). This approach allows us to analyze qualitative within-case data and match it to previously hypothesized mechanisms (Campbell, 1975).
A comparative study is particularly valuable here. Literatures on the migration-welfare intersection have often presented quite broad-brush postulates, owing to systems-theory or political-economy starting points, discussing the intersections of these policy domains, and the imperatives of Western states, writ large (Bommes, 2012;Freeman, 1986). However, the Western European cases where these relationships purportedly hold display some potentially significant differences in welfare-state design and immigration politics (Sainsbury, 2006). In light of such differences, single-case studies themselves are unlikely to offer compelling evidence for or against those accounts in general (Rueschemeyer, 2003, pp. 310-311). This makes it important to complement the strengths of qualitative research at withincase analysis with cross-case comparison (Mahoney, 2007, pp. 134-136). The need to gather detailed, qualitative within-case data points to a small-N study as the most robust viable option.
We employ Germany, France, and the UK as "typical" case studies of countries with advanced welfare states, which have received increased immigration, and then displayed the outcome to be explained: strengthening links between immigration status and welfare provision. Established by the 1990s as the three major Western European countries of immigration, all increasingly incorporated immigration status as a criterion for various welfare-state services in this period (Owers, 1994). However, among possibly significant differences these cases display, they contrasted in welfare-state design, with the UK a typical "liberal" welfare state and France and Germany "conservative," "Bismarckian" ones (Esping-Andersen, 1990). They also differed in the inclusion of migrants in welfare regimes prior to the 1990s. In the UK, a discretionary power to exclude some immigrants from income and housing support (but not health care) existed, but was applied only ad hoc to some migrants with temporary status like students and fiancé(e)s; migrants with recourse to welfare, including asylum seekers, were included in general social provision. In France, from the 1980s, irregular migrants and asylum seekers were excluded from social housing and family benefits, but not health care. In Germany, lacking the right to work, asylum seekers and irregular migrants were excluded from participating in the general contributory system but they could in principle access basic social benefits. Comparing these countries allows us to test whether previous accounts apply across cases where they would purportedly pertain, despite some possibly significant differences.
Methodologically, process-tracing is an essential approach for investigating this kind of puzzle, as it aims to identify particular causal mechanisms operating between structural independent variables and policy outcomes (George & Bennett, 2005;Trampusch & Palier, 2016). In exposing the reasoning of actors in the policy process, it is particularly valuable for testing inferences made by previous comparative research (Bennett, 2010). We engage in a "theory-building" process-tracing approach that comparatively tests previous accounts by making theoretically informed inferences about what would evidence them (Beach & Pedersen, 2013, p. 16). Process-tracing requires gathering substantial within-case data, principally relying on contemporaneous documents and elite interviews (Checkel, 2008). To this point, we focus on policies that incorporated immigration status into welfare eligibility in our three national cases in the 1990s, for two reasons. First, previous research has identified the 1990s as a critical juncture for sense-making in immigration-related policy across Western Europe, yielding problem conceptions that have proven very durable (Geddes, 2018). The 1990s have also been identified as a period of substantial reform pressure upon Western European welfare states (Palier, 2010;Taylor-Gooby, 2004). Focusing on the 1990s allows us to investigate this period of "puzzling" (Hoppe, 2011), before such measures may have become a bureaucratically internalized or politically routine policy lever, providing a greater likelihood of richer data about the basic policy deliberations underlying these links. Second, investigating a relatively recent period where key policymakers are likely no longer to be active alleviates some issues regarding access and candor that may complicate interviewing active policymakers (Ball, 1994). Gathering such rich within-case data opens up the possibility of developing new theoretical propositions about this policy link (Falleti, 2016, p. 457;Rueschemeyer, 2003). 2 This article emerges from a major comparative project funded by the Economic and Social Research Council (UK). 3 Drawing from that project's general survey of immigrant-related policy changes in these countries in this period, this analysis identified significant policy interventions that linked immigration and welfare in general. As policy elites relevant to particular decisions are often identifiable from public sources (Knoke, 1993), we sampled elites using a combination of "decisional" sampling based on sources like civil service yearbooks, as well as snowball sampling. Because of their awareness of both the political and bureaucratic imperatives in relevant policy areas, and their integral roles in implementation and in policymaking at times of lesser political salience (Alesina & Tabellini, 2008), we focused our interview sample on high-ranking civil servants, though we also sampled ministers and political advisers, and external members of policy networks. To achieve a productive basis for comparative analysis, we developed a shared interview schedule to guide interviews across our cases-though, as semistructured interviews, they featured substantial variation in discussion within main shared themes. Likewise, interview transcripts were coded according to a shared coding schedule. This study is based on interviews with 83 policy actors from this period-35 German, 28 French, and 20 British. To triangulate interview data, we also consulted contemporaneous government and media documents, including white papers and press coverage. The article presents a "theoretically explicit narrative" (Falleti, 2016) of each case study individually, describing the dominant welfare-regime model, the incorporation of immigrants into welfare states, and the politicization of immigration entering the 1990s, before presenting sequential accounts of causal processes underlying relevant linking policies. A discussion section then assesses the cases comparatively against the four accounts we identified.

Germany
Germany possesses a paradigmatic Bismarckian or conservative welfare state model, in which relatively extensive social insurance systems are funded primarily through employee and employer contributions. In the early 1990s, asylum seekers represented a majority of immigrants in Germany, and, lacking a right to employment, they were excluded from this contributory system. They were, however, entitled to the basic provision of welfare benefits (Sozialhilfe) that, in cases of emergency, is accessible to Germans and foreigners-including, in principle, irregular migrants (Schönwälder et al., 2004, p. 41). Immigration governance in Germany (Ausländerrecht) is characterized by a robust monitoring system which includes one of the first and largest state databases on migrants in Europe (Badenhoop, 2020). Given Germany's federal structure, competences in the policy area of migration control are shared between the federal and the Länder (states') interior administrations. The Federal Interior Ministry holds law-making competence, while the Länder Interior Ministries and their local Foreigners Authorities hold responsibility for law enforcement. To investigate both policy design and implementation, we analyzed federal authorities and selected Länder: Bavaria representing a more restrictive political approach to immigration, and Hamburg a less restrictive one.
Following the end of the Cold War, Germany received more migrants than other Western European countries, registering one million ethnic German immigrants between 1989 and 1992 and 1.8 million asylum applications between 1987 and 1995-including 80 percent of all asylum applications in Western Europe in the peak year of 1992 (Geddes, 2003, p. 85). However, only five percent of asylum applications in Germany succeeded (Bade, 2002, p. 389). The majority of (rejected) asylum seekers was therefore considered to be irregular: known to authorities, but not authorized, rather, merely "tolerated" (Geduldete) and "obliged to leave the country" (Ausreisepflichtige). In the early 1990s, hostile political and media discourse represented asylum seekers as "economic migrants" burdening the welfare state (Herbert, 2001, p. 299). Germany was also shaken by a series of deadly racist attacks on migrants, and electoral successes by far-right parties (Herbert, 2001, pp. 302-319).
In this context, the coalition government of Christian Democrats and Free Democrats adopted a series of restrictive links between immigration and welfare. Two legislative acts aimed to significantly tighten immigration control, introducing public reporting duties in 1990 and excluding asylum seekers from basic social benefits in 1993. The predominant logic in the German case was instrumental, aiming at deterrence and departure and signaling a difficult reception to potential migrants. The interior ministries driving these reforms were also aware of lax policy implementation and counted on the symbolic effectiveness of these policies. Protecting the contributory welfare system served as accompanying rhetoric justifying exclusion, not as a main policy rationale.

The 1990 Foreigner Act (Bundesausländergesetz)
Our interview data suggest that the rationality behind the adoption of this policy was to instrumentally affect immigration movements through deterrence and strengthening enforcement. This act placed a reporting duty on all public authorities to immediately inform the Foreigners Authorities once they gained knowledge of an individual migrant's irregular status, for example overstaying after a rejected asylum application (AuslG paragr. 76 (2)). This was done knowing that the provisions would apply to welfare providers including hospitals and schools, and would therefore disrupt the provision of these services to unauthorized immigrants. As a senior civil servant from the Federal Work Ministry recounted, there was a fear the German welfare system was a "pull factor" for irregular migration, and that allowing irregular immigrants to participate in the contributory welfare system was sending out inviting signals that "anyone can come to Germany and access the healthcare system as if they were a fully legal worker paying social insurance contribitions, regardless of whether they hold a residence permit or not" (Interview G3).
While the German case showed some elements of welfare chauvinist rhetoric, the policy intention behind the reporting duties was to instrumentally steer immigration inflows and encourage departure. The senior federal civil servant in charge of drafting the 1990 reform explained that reporting duties were introduced because "we somehow thought they would work better" in making the living conditions of unauthorized migrants more difficult, inducing them to leave (Interview G6). Policymakers believed irregular migrants' uptake of social services would decline with the new reporting requirements. These duties were designed to signal a more difficult reception to potential immigrants and aid in enforcement, while simultaneously signaling a tough stance on immigration to voters.
However, the reporting duties increasingly manifested symbolic immigration politics during their implementation because, despite their instrumental intentions, policymakers who helped to design these measures were aware of, and often even tolerated, their limited enforcement (cf. Boswell & Badenhoop, 2020). First, this was because humanitarian concerns interfered with immigration-control logics. As one policymaker conceded, for instance, "A woman who does not have a residence permit should be able to give birth to her child in absolute safety without any risk of deportation" (Interview G3). Even the Bavarian interior minister, known for a tough stance on immigration, did not want to be seen as intruding into classrooms, and instructed that only heads of schools, not ordinary teachers, were obliged to report (Interview G2). Federal policymakers perceived that such an attenuated interpretation of the reporting duties rendered them completely ineffective. As the senior civil servant in charge of foreigner laws reflected, "I always had the feeling that these rules were undermined, for example in the education sector. Of course, there were irregular pupils who should have been reported to the Foreigners Authorities by the school management but this was often omitted for humanitarian reasons" (Interview G6). Indeed, immigration officials in Munich (one of the largest municipal Foreigners Authorities in Germany) recalled never receiving a report from a school (Interview G1).
Second, the visible tool of reporting duties-explicit provisions laid down in a public act-were met by vocal resistance from welfare professionals typically bound by non-disclosure agreements, such as teachers, medics and social authorities (Interview G13). All German interviewees agreed that the 1990 Act's reporting duties were largely not implemented. The federal foreigners commissioner from this time explained that they did not lead to any actual detections and deportations of irregular migrants (Interview G4), and activists confirmed this impression (Interview G13). From the perspective of the federal interior administration, the issue was a lack of enforcement ("many regulations disappeared into space" (Interview G6)). However, activists argued that the reporting duties were, as a legal opinion commissioned by the Catholic Church found, not legally watertight (and thus plainly open to challenge)-and instead represented "what you call symbolic politics!" (Interview G13).
Later, during the 2000s, a forum of activists and the Catholic Church successfully argued to policymakers that the reporting duties produced unintended consequences. By forcing undocumented migrants to exclude themselves from vital services such as education and health care, these duties created new social problems rather than facilitating detection and deportation. Even proponents of reporting duties supported the explicit exemption of schools (Interview G3), which took effect in 2011. While the instrumental aims of the reporting duties produced sectoral tension that diminished their efficacy, these provisions still clearly symbolized a hard line, and were kept in law for all public authorities except education providers.

Social Welfare Act for Asylum Seekers (Asylbewerberleistungsgesetz)
Similar to the reporting duties discussed above, the 1993 Act was part of an immigration reform package motivated by the rationality of deterrence. This act was justified by a rhetoric of protecting an exclusive, contributory welfare state from non-contributing immigrants. Following the "asylum compromise," which effectively abolished the German constitutional right to asylum by rejecting asylum seekers from "safe third countries," the 1993 Act excluded asylum seekers from the existing basic social benefits system (Sozialhilfe) and created a separate, lowered provision of financial benefits, housing, and health care. This act applied to a range of regular and irregular migrants, including asylum seekers whose applications were being processed or who had successfully gained a humanitarian residence permit, as well as rejected asylum applicants who were "tolerated" as long as they could not be deported (Geduldete), and those who were "obliged to leave" Germany (vollziehbar zur Ausreise Verpflichtete). The 1993 Act thus aligned legally resident asylum seekers who were awaiting their decision with unauthorized migrant residents, fueling discourses linking asylum and illegality.
According to the then-Bavarian interior minister, the "central issue" during the late 1980s and early 1990s was that the majority of rejected asylum seekers were not allowed to work, and received social benefits (Interview G2). For him, excluding asylum seekers from the labor market while including them in the welfare system presented a political dilemma: It was seen as hardly justifiable to German nationals "who have been working hard for 30, 35 or 40 years" that they receive the "same money as someone who came here to apply for asylum, was rejected but simply refuses to leave;" yet, "on the other hand, once you allow someone to work, it becomes practically impossible to remove him from the country" (Interview G2). The 1993 Act addressed this dilemma, as it significantly lowered the level of welfare provision for asylum seekers to deter newcomers, while continuing the structural dependence on welfare of those who were already resident. That way, asylum seekers continued to be known to the authorities, excluded from the labor market, and easier to deport. This act contained further initiatives to "reduce" Germany's "attractiveness" by introducing other "incentives" such as material benefits (Sachleistungsprinzip). Bavaria added cash or vouchers for construction materials to rebuild houses in Bosnia for those who volunteered to leave, though these "voluntary return transports" proved ineffective (Interview G2).
The 1993 Act further deteriorated living conditions for irregular migrants in Germany. In combination with earlier reporting duties, the 1993 Act effectively prevented undocumented migrants from accessing medical treatment. While under the 1993 Act, rejected asylum seekers (vollziehbar Ausreisepflichtige) were entitled to health care, if they exercised this right, they risked detection and deportation: the 1993 Act required migrants to contact the Social Authorities first to obtain a sick certificate (Krankenschein), but if they did that, then the Social Authorities would be obliged to report them under the 1990 Act discussed above. That way, "accessing health care the regular way under the 1993 Act is impossible without taking the risk of detection" (Interview G7). To redress this severe social exclusion, some Länder and local authorities in cooperation with NGOs and medical professional organizations later developed alternative models enabling undocumented migrants to access health care without risking detection. Anonymous healthcare funds adopted in Munich and Hamburg cover emergency treatment (Interviews G28, G23), while anonymous sick certificates (anonymer Krankenschein) were first introduced in Lower Saxony in 2014 (Niedersächsischer Landtag, 2014), followed by Thuringia (MDR, 2018) and Berlin (Memarnia, 2019). However, these models were strongly opposed by the Federal Work Ministry and Social Authorities in some Länder, who feared abuse and wanted to protect contributory health insurance from those who do not officially pay into the system (Interviews G3, G19).
While these municipal and Länder initiatives mitigate some of the exclusionary effects created by federal policies, the reporting duties and the 1993 Act are still in force, even though the low level of social security offered by the 1993 Act was declared unconstitutional by the Federal Constitutional Court in 2012 (Pro Asyl, 2017). In Germany, linking policies developed in the early 1990s targeted unknown undocumented migrants and known rejected or potentially unsuccessful asylum seekers alike, making their access to welfare services such as health care more difficult. Both policies were interior-ministry driven and motivated by instrumental logics, with the 1990 reporting duties aggravating exclusions which the 1993 Act created in an attempt to deter asylum-seeking.

France
The French welfare state in the 1990s, like Germany's, followed the conservative or Bismarckian model, with a core contributory healthcare system funded by worker and employer contributions, and managed by the Caisses Primaires d'assurance Maladie (hereafter, the "insurance system"). Also like in Germany, a tax-funded safety net provided for those who did not contribute enough to access the general contributory system (hereafter, the "assistance system"). Entering the 1990s, irregular immigrants were in principle banned from working, and therefore were excluded indirectly, rather than specifically, from contributing to (and accessing) the insurance system. However, enforcement loopholes and street-level bureaucratic practices meant many such immigrants did work and thus contribute. As a high-ranking civil servant who evaluated these issues in the mid-1990s recalled, "We were very surprised to observe that there were visibly several thousands of people . . . who were 'declared' clandestine workers. . . . These people didn't hold work permits, but they were regularly registered to the [insurance system], and they enjoyed all the social rights attached to it, and the employers were paying the social contributions absolutely regularly" (Interview F2). The assistance system, on principle, did not exclude undocumented foreigners; access was provided on the basis of residency, so those rejected by the insurance system would turn there. The assistance system was mainly managed by local government entities, the départements, and was designed to include recipients on the basis of their address. However, département interpretations of this requirement varied.
Unlike in Germany and the UK, immigration politics in France entering the 1990s was not especially centered upon new post-Cold War migrations such as refugee movements from the former Yugoslavia or new sources outside Europe. Asylum applications had risen in the late 1980s, but policymakers considered the issue solved after reforms and investments in French asylum agency and the exclusion of asylum seekers from the labor market in 1991 (and, consequently, from the insurance system). Political attention in France was more directed to long-term irregular labor or asylum migration from former colonies, as reflected in the sans-papiers movement (Siméant, 1998).
The two main reforms examined here were both introduced after parliamentary elections and changes in government: the exclusion of undocumented migrants in 1993 from the insurance system, after the election of a right-wing government; and the introduction of a specific health care scheme for undocumented immigrants by a new left coalition government following the 1997 elections. While in France the prevalent logic was symbolic immigration politics in the adoption of the 1993 law, at the end of the 1990s, a new welfare-guided reform included undocumented migrants in a specific scheme. This reform, though not revisiting earlier exclusions of undocumented workers, was not driven by exclusionary aims.

The 1993 Pasqua Act
The Pasqua Act in 1993 is often referred to as a turning point. This legislation, packaged among other measures aimed at adopting a tougher stance on immigration control, explicitly excluded undocumented immigrants from the insurance system. Driven by interior ministry actors, this reform was, according to our data, never entirely instrumental, but rather punitive, expressing a principle that people who should not be in the territory should not have access to health care. The symbolic dimension of this measure was underlined by the extent to which interior ministry actors prioritized publicly displaying a tougher stance on immigration above attaining concrete policy changes.
Welfare policy actors' concern with migrant access to the French welfare state in this period was focused on "healthcare tourism:" people who came to France for medical attention, which would result in unpaid hospital bills, detrimentally affecting healthcare budgeting. Such concerns were not about undocumented immigrants who intended long-term residence. As a former immigration official recalled, "the main concern was not so much irregular foreign workers, but irregular use [of social rights]. For example, families who come to France for holidays and use their cousin's social security card" (Interview F25). While "the consensus was on the fraud" (Interview F25), lack of immigration status was seen as a key characteristic of healthcare tourists. In this view, access to the insurance system for the undocumented grew from loopholes in immigration enforcement, and closing these loopholes enjoyed consensus among left-and right-wing political elites, from the interior as well as the social affairs ministries. Despite the lack of a clear power to do so, some local caisses already discriminated on the basis of immigration status before 1993 (Izambert, 2018). These exclusionary practices, sometimes dictated by the welfare-state hierarchy, were contested in administrative courts and frequently condemned (Isidro, 2015).
However, the first version of the 1993 legislation went much further than this developing consensus: It aimed to exclude undocumented migrants not just from the insurance system, but also the assistance system. This proposal had two notable qualities. First, in proposing legislation, Charles Pasqua, the interior minister, was elevating the issue into visible public political arenas, where it had previously been dealt with less visibly through regulatory processes. A former high-ranking interior-ministry civil servant summarized that, "to be honest, I think [discrimination] already existed, but [the Act] corresponded to a political demand to bring it to light in legislation, which in my opinion should not be stated in a law" but rather in administrative orders (Interview F11). Following the rationale of symbolic immigration politics, legislating escalated the measure's political visibility, requiring debate in Parliament. Second, the legislation's proposed double ban (on insurance and assistance) appeared to put aside developing interdepartmental consensus in favor of a more sweeping exclusion of "undeserving" irregular immigrants from the tax-funded backstop system. This escalation rattled social-affairs decision-makers up to the highest levels: A former caisses agent recalled a phone call he received from the social affairs minister, expressing a concern the proposal "was going very far" (Interview F22). This reluctance from key social affairs actors suggests a lack of welfare-guided intentions.
Eventually, the final version of the law excluded undocumented foreigners only from the insurance system, limiting the prospect of any instrumental impacts. Its limited scope was mainly due to the intervention of a right-wing MP, Claude Malhuret, a doctor and former president of Médecins Sans Frontières, who argued that doctors had a duty to care for everyone, and that emergency health care in hospital would become much more expensive. Here, containing costs was invoked, as part of an argument to include the undocumented in the assistance system, rather than exclude them. This underscores how the original proposal was not integrated with efforts to cut state spending.
The 1993 law aimed to connect welfare-state and interior ministry files: one of its key measures was an obligation for administrators in the caisses who registered beneficiaries to check the validity of residence permits, establishing that these agents could access the foreigner files held by prefectures ("AGDREF") to make these verifications. In line with instrumental immigration-control logics, the interior ministry pursued implementation. However, attempts by the interior ministry to connect these files in order to exclude immigrants from the insurance system were met with often lenient practices from caisses administrators. As one recalled, "I would not speak about resistance, but there was no endorsement. And [enforcement] wasn't in the habit, in the intellectual mindset, of health insurance" (Interview F22). This lack of enthusiasm was not lost on the interior ministry: as a senior interior ministry civil servant recalled, caisses administrators' "state of mind was, 'we protect; we don't want to be cops' auxiliaries'" (Interview F10). Social affairs officials did not publish the circular clarifying the types of documents required to access the insurance system until May 1995. Interior ministry effort to contact various agencies within the social affairs ministry to implement these measures "was confronted with a total inertia" (Interview F10). Informationsharing practices were left highly dependent on caisses administrators' discretion, as "systematic automatized access" (Interview F10) to AGDREF files was only enacted in 2012. Social-affairs and interior-ministry differences on these measures stretched from the level of street-level bureaucrats up to major policy decision-makers. Consistent interior-ministry efforts to implement these measures suggest they had more than symbolic intentions, though they often went unfulfilled.
Ultimately, the 1993 law amounted to the formalization and post-hoc legitimation of some caisses' existing practices excluding undocumented migrants from health care, while patchy implementation left in place more lenient practices where they had existed. Still, the law had effects beyond signaling within domestic politics. With the legal duty of caisses administrators to check the regularity of foreign workers publicly highlighted, even if not fully carried out, undocumented migrants began turning to hospitals and NGOs (Interview F22).

The 1999 creation of Aide Médicale d'Etat
In 1999, the new left-wing government of Lionel Jospin introduced Couverture Maladie Universelle (CMU, "universal healthcare coverage"). This measure expanded the insurance system to include people who did not previously contribute enough to access it, funding their participation through taxes. This policy change entailed the transfer of all assistance-system beneficiaries to the caisses, with one exception: undocumented migrants. However, faced with this constraint, the new government created a specific mechanism linked to the assistance system to deal with undocumented migrants' health care: Aide Médicale d'Etat (AME). The government explicitly linked undocumented status to welfare provision, but adopted a largely inclusive measure, against the main theorizations of welfare-guided linking logics we have identified.
Indeed, the major incentive in this reform related to healthcare budgeting, since it clarified whose fiscal responsibility healthcare provision to irregular migrants was. This issue had seen constant confrontation between local and central governments. Within the assistance system, the départments had been in charge of funding health care for their residents, while the central state provided for individuals who could not prove their residency in the départment. Local authorities started to deny access to the assistance system to undocumented migrants, claiming irregularity was proof that they were not residing there. Unpaid hospital bills piled up. The 1993 Act had somewhat clarified responsibilities for funding health care for undocumented migrants: the départments were responsible for all individuals who could prove residency of more than three years, whether regular or irregular, but the départements still contested this arrangement. The creation of AME further clarified these responsibilities: the central state assumed responsibility for the assistance system. From this date, hospital administrations created social work units to provide information and support to undocumented migrants so that they could access AME (Interview F22). AME effectively shifted many of these patients from hospital emergency departments to less costly general practitioners. Thus while there was a cost-cutting intention, this did not correspond to excluding undocumented migrants, but rather, to their inclusion in a specific tax-funded scheme. While AME has been criticized for providing a relatively lower level of provision that can be difficult to access (Izambert, 2010), in contrast to the UK and German cases, this highly visible scheme is aimed specifically at undocumented migrants, and continues to provide health care for hundreds of thousands of individuals.
In France, immigration status was consolidated as a criterion for healthcare access in the 1990s, in interior ministry-led reforms intending mainly to symbolically communicate tightening in immigration policy. Though interior-ministry officials seemed invested in the implementation of scaled-down measures, welfare-state actors focused on resolving funding responsibilities for undocumented immigrants' care in the tax-funded assistance system. That system has been very visibly retained solely for undocumented immigrants since 2000-suggesting the importance of a welfare logic, though not one of the exclusionary ones previous theorizations have suggested.

The United Kingdom
Britain's welfare regime has been seen as an exemplar of the liberal model, with extensive means-testing and limited universal provision . However, the UK welfare state includes one highly notable example of tax-funded universal provision: the National Health Service, which, unlike in France or Germany, is a direct provider of universal health care, rather than an insurance system. The UK entered the 1990s with only some "ad hoc" (Cohen, 2001, p. 20) measures excluding immigrants from welfare systems. A discretionary power to exclude immigrants from income and housing support-but not health care-existed in law, but had been applied only to migrants with specific temporary statuses, like students or fiancé(e)s (Gordon, 1986, pp. 25-26). Asylum seekers, for instance, were included in general social-support systems. While non-resident patients were supposed to pay for NHS care, this did not involve systematic checks of immigration status. The UK was considered to lack the ability to carry out these kinds of checks, as it had not developed extensive population monitoring or identity-fixing capabilities such as national ID cards commonplace in continental Europe (Agar, 2001). Accordingly, the Home Office, Britain's interior ministry, worked from the premise that as an island nation, the UK possessed an ability to control its borders that countries like France and Germany lacked (Morris, 1998, p. 954), and therefore ran an immigration control system that was distinctly ports-based.
The post-Cold War increase in asylum seeking rattled these basic presumptions. Though dwarfed numerically by Germany, the rise of asylum applications in the UK-from 3,998 in 19883,998 in to 44,840 in 19913,998 in (Fiddick, 1999. 7)-shook officials' notion of an effective ports-based control system, since cases had to be adjudicated internally. At the ports, a leading operational official recounted, "The essence of the decision-making had been taken away" (Interview UK8). The significant proportion of asylum claims made in-country, following enforcement actions, was interpreted as irregular immigrants circumventing enforcement through asylum. With increasing applications, the asylum system was beset by long delays. These factors precipitated a sense of dysfunction and crisis among policymakers, which persisted for years as these issues grew in political salience.
This pressure triggered successive reforms that linked immigration and welfare with the instrumental goal of immigration control through deterrence. Four new immigration acts were adopted sequentially, after the previous measures were evaluated as not having reduced asylum seeking sufficiently. Similar to Germany, while the dominant logic was instrumental, policymakers were also aware of these measures' symbolic value in politically signaling immigration control, even when they did not achieve intended outcomes. Over time, the key deterrent logic shifted from excluding asylum seekers from general systems of support, to substituting much less generous systems intended to deter applications by additionally rendering asylum seekers more controllable.

The 1993 and 1996 Acts
By 1993, earlier measures to externalize border control were deemed to have not sufficiently limited asylum seeking. Home Office officials turned to the question: "Why are people coming to the UK?" (Interview UK3). UK policymakers, similar to Germany, aimed to reduce perceived "pull factors," following an instrumental immigration-control logic. In an ongoing crisis atmosphere with pressure to act, "nobody knew in a scientific way what all the different millions of decisions were that went into determining how many asylum claims the UK got in a year. . . [but] the number of claims was potentially sensitive to the signals that the system as a whole sent out" to migrants (Interview UK7).
The Asylum and Immigration Appeals Act 1993, passed under a Conservative Party government, was the first UK legislation to try to steer migration dynamics through reducing such "pull factors." Policymakers at this time saw the lengthy asylum adjudication process as the most critical such factor. But the 1993 Act also restricted public housing provision to asylum seekers, making them ineligible for social housing if they had any form of accommodation, even temporary, and even then, limiting eligibility to only temporary accommodation. Social housing in the UK was affected by scarcity, while concentrations of asylum-seekers meant disproportionate pressure on a few localities. Still, as a housing policymaker involved in drafting the 1993 Act reflected, "The issue around asylum seekers was, relatively speaking, a secondary concern, because they were never going to have . . . that kind of acute political and operational importance in [housing] ministers' minds that the domestic pressure on temporary accommodation was going to have" (Interview UK17). While "a sudden rise in a particular group which will have a lot of political resonance" (Interview UK17) meant an opportunity to address a contributing factor to social-housing scarcity in what was primarily a Home Office bill, from the Home Office perspective, including housing was relevant because, simply, asylum seekers "had to find somewhere to live" (Interview UK2), and making this more difficult would deter them.
However, the 1993 legislation did not effectuate a lasting reduction in asylum claims, the key metric by which it was assessed. This prompted policy changes in 1996 that applied this instrumental deterrence logic more expansively. Restriction of asylum seekers' eligibility for a broad array of benefits, including ending their eligibility for all forms of public housing, was packaged within the Asylum and Immigration Act 1996, which additionally further reformed the administration of the asylum system and introduced employer sanctions for hiring unauthorized workers. Further welfare restrictions not requiring legislation were promulgated administratively by the social security minister. While the Home Office led on this legislation, social security ministers echoed that these steps primarily aimed to "discourage unfounded claims from people who are actually economic migrants," while also decrying such claims' fiscal costs (HC Deb 11 January 1996, col. 331). By this time, such tropes were featuring frequently in the popular press (Kaye, 1999), which had "conflated benefits scroungers. . . single mothers, and the drain on the state with immigration" (Interview UK14), though polls at this time still showed few voters considered asylum or immigration as among the UK's most important issues.
While officials clearly saw the resonance of such policies in domestic politics, they were by no means purely symbolic: Home Office officials did intend-and cautiously expect-these measures to have an instrumental effect through deterrence. While evidence of what was attracting asylum seekers was anecdotal, "if you haven't got anywhere to live, you can't earn any money, you're not getting any benefits, surely that must be a dis-attraction?" (Interview UK13). At the same time, Home Office interest in restricting healthcare provision encountered firm opposition from the NHS, which saw erecting barriers to care as anathema. As a Home Office official recounted, "it has been a tenet of the Health Service for a long time that they just treat people because they need treating. . . . [w]hat we came up against when we tried to implement this or tried to see our way around this policy, [is] that it is an article of faith" (Interview UK9).

The 1999 and 2002 Acts
With UK asylum applications in 1998 reaching new highs, the recently elected Labour Party government legislated further on the issue, with similar intentions as the previous government: to instrumentally steer migration movements, reducing applications and effecting removals. Rather than simply restricting access to benefits, the Home Office was given direct authority over housing and support for asylum seekers through a National Asylum Support Service, packaged within the Immigration and Asylum Act 1999, which also altered appeal procedures, externalized border control, and barred income and housing support to a wider range of mostly family migrants. The act removed asylum seekers from general systems of social support, replaced cash benefits for them with vouchers, and introduced involuntary dispersal for asylum seekers seeking housing assistance.
Our data reflect clearly that the 1999 Act was intended to achieve immigration, and not welfare, policy aims. Relocating welfare functions to the Home Office explicitly decontextualized these steps from broader welfare reforms: "We don't try to revamp the whole of government, we set up our own mini system. We take people out of benefits, we deal with the accommodation, we send them where we think they ought to be, and all of that.
We'll find it easier to keep tabs on them" (Interview UK12). Welfare-state agencies seemed to accept this-with mediation from the powerful Treasury-since dealing with immigration policy-driven changes "was just aggravation for them. . . . [The Home Office] work[ed] out a deal with the Treasury whereby, essentially, we could be given more resources on the basis that we would then be able to deliver a reduction in asylum" (Interview UK13). However, the voucher scheme especially drew opposition from Labour parliamentarians who criticized it as both degrading and more expensive to administer than cash benefits (Bloch, 2000); vouchers were abandoned in 2002.
Immigration and asylum issues were becoming increasingly politicized in the UK, reaching new levels of salience around the turn of the millennium (Jennings, 2009, p. 858). The involvement of ministers-the policy actors most attuned to political pressure-in policies linking immigration and welfare grew around this time. The logic of using such measures to appeal to domestic political audiences therefore seemed to be strengthening, indicating a growing symbolic dimension. As political attention to asylum grew amidst increasing applications notwithstanding the 1999 Act, the government again addressed the issue within the Nationality, Immigration and Asylum Act 2002, which was developed with exceptional input from Tony Blair, the prime minister. The initial bill sought to create a new system of accommodation centers to centrally provide services for asylum seekers deemed likely to become destitute. Here as in 1999, longstanding deterrent aims were now augmented with a complementary emphasis on monitoring asylum applicants "end-to-end" (Interview UK16), to effectuate removals and thus strengthen deterrence. The prime minister's office intervened late in the process to insert a provision barring from support asylum seekers deemed not to have made their claims "as soon as reasonably practical." When plans for the accommodation centers failed due to extensive local opposition, the "as soon as reasonably practical" provision ended up being the 2002 Act's main welfare policy impact.
All of the UK's key linking policies in this period were adopted with instrumental aims, and were assessed based primarily on whether reductions in asylum applications followed. Policymakers subscribed to the notion that in order to steer unwanted immigration inflows downward, the UK had to make its welfare state less of a "pull factor." However, unlike in the other cases, healthcare provision was not successfully targeted. As in Germany, UK policymakers were keenly aware of welfare-chauvinist media tropes. Ministers sought to capitalize on these, suggesting that policymakers saw these measures' symbolic political value as resonating, even as they failed to meet initial goals.

Discussion: Which of the Four Accounts Match the Evidence?
Delimiting Social Citizenship to Respond to Legitimacy Challenges to the Scope of Welfare Provision?
Our three case studies offer little evidence that this was an important logic in the development of policies in these countries that linked immigration and welfare. Rather than being driven by welfare actors aiming to adopt welfarechauvinist images to align with concepts of social membership they believed the public saw as legitimate, policy changes in all three countries were driven by interior ministries, which saw welfare through their typical control-oriented lenses. While welfare policymakers were frequently involved in elaborating policies as secondary actors, none of our three cases presents evidence that welfare policy actors prioritized a problem conception where social legitimacy was at stake for the welfare state or its scope. When key policymakers (often in interior ministries) articulated concepts of deservingness, these were not noticeably connected to any problem concept seen as carrying implications principally for the welfare state. In France, assertive interiorministry attempts in the 1990s to exclude long-term irregular workers not just from the contributory system, but also the basic social-benefits system, incorporated rhetoric about welfare deservingness-but caused great concern among welfare-state officials. Similar rhetoric accompanied pronouncements of UK policies which had clearly immigration-oriented aims. In Germany as well as France, some policymakers expressed fears about "abuse" and wanted to protect the welfare system's contributory logic, but their statements do not suggest that the legitimacy of the scope of welfare provision was seen as a principal matter at stake in reducing provision for some immigrants. Key policymakers driving these policies in our cases focused on other purported problems.

Achieving Welfare-State Policy Goals Amidst Retrenchment or Market-Oriented Reform?
Likewise, the identification of interior ministries, rather than welfare agencies, as the predominant institutional actors in driving immigration-welfare policy links also suggests that achieving new welfare-state policy goalssuch as reducing public spending, or encouraging labor market participation-was not a principal logic in the development of this trend. In France, while cost-cutting was an objective among elite welfare actors, reducing immigrants' welfare access was not seen as a way of achieving this. In Germany, while some policymakers were concerned about "abuse" of welfare provisions, none showed concern about an acute scarcity of welfare-state resources. In the UK, scarcity did seemingly facilitate the early selection of housing as a sector to target, but there is little evidence welfare agencies had a particular zeal for excluding immigrants-though the case of the NHS suggests that strong organizational resistance could be decisive. Later, UK welfare agencies surrendered responsibility for providing for the politically sensitive asylum-seeking population, at the behest of a powerful Treasurybut in order for these monies to be employed to achieve the separate, nonfiscal policy aim of reducing asylum seeking. In France, the question of who would pay for the health care of irregular immigrants was a major concern for policymakers, but this was a matter of clearly attributing responsibility for this group and resolving organizational budget lines. When changes affecting immigrants were packaged in larger welfare reform efforts-as with France's creation of AME in 1999-the measures did not advance exclusion, but sought to resolve issues in the system's multi-level governance. Here, efforts to curb overall state spending actually led to including undocumented immigrants within this specific scheme, rather than in a logic of exclusion. While offering some evidence that welfare-state actors could steer interior ministrydriven exclusion attempts away from or toward certain welfare-state sectors, all three cases also saw the reinstitution of greater levels of provision after previous exclusions presented fiscal or social problems.

Symbolic Immigration Politics?
Our data suggest that this was a major logic in initial developments in France, though the "law-and-order" symbolism of linking immigration and welfarethat concepts of legality in immigration should be reflected in the welfare state, which must not provide to those suspected as corroding immigration control-was also significant in the UK and Germany. In France, the motivation for the 1993 leglsiation appears to have been symbolic to a large extent, with the interior ministry aiming for visible legislative action, even if it ended up reiterating existing practices. In Germany, the symbolic resonance of introducing reporting duties for welfare-state agencies as a measure against unwanted migration was a main reason these duties remained in law, despite their instrumental inefficacy. In the UK, immigration policymakers were clearly conscious of the political symbolism of excluding asylum seekers from welfare, and the interventions of ministers-the policy actors most interested in sending effective domestic-political signals-increasingly sought to solidify welfare-immigration links as immigration rose in political salience. In Germany and the UK, interior-ministry actors found symbolic value in their linking measures after initial proposals had not achieved intended instrumental effects. However, policies adopted publicly for symbolic, domestic-political reasons often had additional signaling effects to immigrants themselves. In France and Germany, immigrants' resulting hesitancy to access health care was later deemed not in the state's interest.

Instrumental Immigration Policy?
Our data show that instrumental immigration policy logics were the principal rationalities behind this policy link in Germany and the UK, while there is evidence that they were also significant in France. British and German policymakers both sought to deter asylum seeking in the post-Cold War context of increasing inflows, while keeping unsuccessful asylum seekers returnable. The UK case data especially reflect how these policies were adopted as a result of broad reasoning about migrants' life courses, with rigorous analysis forestalled by a perpetual sense of crisis and a need to act. In France, where the issue of illegality was not especially tied to post-Cold War asylum-seeking, the initial interior ministry intention behind the 1993 legislation was to bar undocumented immigrants from the tax-funded healthcare assistance system-but political opposition felled this effort. The interior ministry was invested in the implementation of their exclusion from the contributory healthcare system, suggesting some instrumental (rather than solely symbolic) intent, but met with resistance from social-affairs officials. In all three cases, there are clear attempts at "outsourcing," or the involvement of other social sectors in enforcing immigration compliance (Düvell, 2006).

Conclusion
Comparing the cases of Germany, France, and the UK in the 1990s offers the opportunity to test contrasting previous accounts for why policymakers link immigration and welfare policy. This study's clear finding is that policymakers in all three countries adopted this link overwhelmingly for "immigrationguided" reasons-to try to control unwanted immigration or to communicate symbolically about this issue-rather than for previously theorized reasons guided by welfare-state politics or policy goals. Policymakers saw welfare states as key to the life chances of immigrants, and sought to use them as tools of immigration control, or for symbolic political purposes. Immigrationwelfare links were driven by interior ministries, which assessed these policies' successes through their traditional control lenses; these policies' later welfare impacts were more often brought to attention by outside campaigns than by state-driven policy evaluation. Scholars have frequently discussed how increased immigration raises legitimacy issues for the welfare state, but in our cases, it does not appear that any such legitimacy challenge-either to the welfare state generally or, more narrowly, its scope of social provisionprimarily spurred the consolidation of welfare-immigration policy links in this key period. While the policies we analyze all inherently had implications for both welfare states and immigration control, and reflected tensions in these systems' central notions of membership (Bommes, 2012;Sciortino & Finotelli, 2015), decisions to enact these linking policies were not informed by an equal consideration of both sides of this link. Rather, the priorities and goals of immigration-policy actors guided this process.
Given this, it is striking how the formation of the immigration-welfare policy link in this period was conditioned by particular notions of "unwanted" migration. Beyond testing previously proposed accounts of this policy linkage, our data consequently yield some further findings with important implications for a growing literature on migrants' access to social rights, as a growing number of forms of migration within Europe are politicized as "unwanted," including some forms of intra-EU work migration (Heindlmaier & Blauberger, 2017;Lafleur & Mescoli, 2018). The implications are in two main areas: how forms of exclusion take shape, and what populations are affected. Regarding the first, in our cases, differences in enacted forms of exclusion-for instance, which welfare-state sectors were targeted, and how successfully-did not emerge from major differences in policy intentions. Rather, they grew from how these intentions interfaced with institutionalized welfare-state policy designs. For instance, health care was successfully targeted for action within the contributory systems of France and Germany. However, this encountered strong resistance during the 1990s in the UK, where the state provides health care directly and exclusion would have meant direct exclusion from care. 4 While we find welfare-guided logics were not key to the emergence of this policy linkage, welfare bureaucracies often did partly direct or shape these links. Combined with advocates' later successes in each case in partially reinstating social provision for some affected immigrants, these dynamics make it difficult to glean policy intentions from overall effects of policies on target populations. Process-tracing accounts are essential to uncovering such intentions.
Second, policies in this period linking welfare and immigration overwhelmingly targeted groups associated with spontaneous, "unwanted" migrations (Düvell, 2006). The fact that the immigration-welfare policy link emerged largely in relation to groups problematized in this way is an important point within a literature which has often interpreted the inclusion of immigrants in general as what is at issue in welfare-chauvinist challenges (Kitschelt & McGann, 1997, p. 22). Of course, this derives from the citizen/ non-citizen distinction historically key to the European welfare state's emergence (Freeman, 1986). We observe that when this policy link was elaborated during this crucial period in its consolidation, not all foreign nationals were targeted, but only particular categories. Yet, despite targeting groups problematized in similar ways, differing contextual politicizations and national monitoring practices meant differences in which precise migrant populations policies targeted. In France and Germany, with more robust systems of population monitoring, "undocumented" or "illegal" immigrants were targeted. In Germany, failed asylum applicants represented a major portion of this larger category due to an especially high number of asylum applications. In the UK, without robust monitoring practices, exclusions were aimed predominantly at asylum seekers-a population already visible to the state. We observe that the intertwinement of national monitoring practices, welfare-state design, and immigration politicizations seems to have important effects on the particular exclusions that are implemented. Diverging policy effects in our cases did not grow from distinct policy intentions-which were in fact broadly sharedbut rather from how these intentions interfaced with nationally distinct immigration politicizations, monitoring practices, and welfare-state designs.
Our findings of course carry some caveats. While Germany, France and the UK displayed diverse welfare-state designs, none of the national cases follows the social-democratic (Esping-Andersen, 1990) or southern European (Ferrera, 1996) welfare regime models. Making robust comparisons with emblematic examples of either would have entailed analytical difficulty, since these countries are much smaller (as with social-democratic countries like Sweden), were not as established as immigration countries in this period (as with southern European countries like Italy or Spain), or both (as with Denmark or Greece). Perhaps delimiting social citizenship is more important in countries with strongly embedded social-democratic welfare regime models. Furthermore, given that populist right parties are key mobilizers of welfare chauvinism (Afonso, 2015), perhaps their greater electoral success since the 1990s has increased the relevance of logics where policymakers link immigration and welfare in order to align policy with welfare-chauvinist images of social membership. Likewise, even if new welfare-state goals were not important to this linkage in the 1990s, this rationality may have emerged later, as such links became more routine, and amidst EU expansions that brought new forms of work migration to Western European countries.
Nonetheless, this article offers strong evidence that scholars working at the intersection of migration and welfare in Western Europe should understand the linking of these policy domains as an "immigration-guided" phenomenon: one that responds first to immigration politics, and seeks primarily immigration policy goals. While welfare states are inextricably part of any such "linking" policies, their politics and policy goals were clearly secondary in the development of this link. Finally, our findings suggest that national differences in which immigrants were excluded from which welfare benefits grew in great part from the intertwinement of nationally distinctive monitoring practices, welfare-state designs, and immigration politicizations-rather than from major differences in policy intentions. These interconnections therefore stand out as particularly important topics for further research in order to understand the continuing evolution of migrants' access to social rights in Western Europe.