The Swiss case

Switzerland has a fairly long tradition of granting asylum to political refugees, with the significant exception of during World War II. In the 1970s, when the composition of asylum seekers changed and their number grew, asylum policy became a major public issue. A host of revisions to the first asylum law from 1981 to date has consisted essentially in the constant increase of the threshold for granting asylum recognition (Lavenex Reference Lavenex2007, 632–635). Although asylum proceedings are regulated at the federal level, the pending asylum seekers are allocated to the cantons, proportionate to population size. The cantons organise housing, allocate social assistance, foster integration and, if the asylum decision is negative, are responsible for expulsion. “The tightening of the Asylum Law has led to a shift of the financial burden of the asylum system towards the cantons, which in turn has resulted in the increased politicization of the topic. The cantons are thus responsible on the one hand for the often protracted implementation of the removal of rejected asylum seekers, and, on the other hand, for the additional costs caused by rejected applicants (…)”(Lavenex Reference Lavenex2007, 635).

Until 2004, the federal government bore the costs of social assistance in the asylum system. In 1995, flat-rate contributions were introduced. To make the cantons reduce their spending on social assistance for asylum seekers, the revised Federal Decree on measures to decrease spending on the asylum system introduced severe cutbacks in the flat rates in 1999. This instrument was in force until 2004, when a federal austerity programme again tightened the law. Since then, rejected asylum seekers are no longer entitled to statutory social assistance except for “emergency assistance” (Nothilfe).

We study the time between 2000 and 2003, when the federal government covered limited per diems for asylum seekers in need of social benefits. In this particularly politicised policy field, the cantons had two options for reducing social spending on refugees: the expulsion of asylum seekers with a negative decision, or enabling the integration of pending asylum seekers into the labour market (Schenkel 2005). The asylum law allowed for the gainful employment of pending asylum applicants after a period of three months. However, the cantons were not obliged to apply this rule. As high expulsion rates do not indicate restrictions of asylum seekers’ labour market integration and vice versa, we focus only on the integration measure.Footnote ¹ Labour market integration denotes the openness of cantonal regulation to the integration of asylum seekers into the labour market. It is measured using Piguet and Misteli’s (Reference Piguet and Misteli1996) labour market restriction index as applied in the study by Spörndli et al. (Reference Spörndli, Holzer and Schneider1998). The index encompasses the following:

1. a. the cantonal regulations regarding the length of the working prohibition,

2. b. the principle of preferring natives for open jobs,

3. c. the obligation for employers to actively look for native job candidates,

4. d. working permit limitation for certain industries and

5. e. the possibility for asylum seekers to switch between different industries.

The data were newly collected for this study on the basis of a written survey of all responsible cantonal authorities in 2004 (return rate 100%) (Schenkel 2005). The original survey in German is indicated in the online appendix. Piguet and Misteli (Reference Piguet and Misteli1996) define three degrees of restriction – namely, low restriction (1–5.5), medium restriction (6–8.5) and high restriction (9–15). Low restrictiveness indicates that cantonal policies enable the labour market integration of asylum seekers (INT); high restrictiveness indicates that cantonal policies impose restrictions on labour market integration (int). We consider cantons with a labour market restriction index of less than 7.25 – that is, exactly in the middle of the “medium restriction” category – as cases that enable integration.

In comparing how Swiss cantons implement the same policy, we employ a “most similar systems” design that holds most institutional and macroeconomic contextual conditions constant (Rihoux and Ragin 2009). This enables us to focus on selected possible explanations for the different degrees to which some cantons chose to give asylum seekers access to the job market, while others did not (Figure 1).

Figure 1

Labour market restriction index, 2000–2003. Note: Restriction index=sum of cantonal scores in length of working prohibition according to art. 21 AsylG (six months=3, three months=0), the principle of preferring natives for open jobs (allover=3, partly=1.5, none=0), the obligation of employers to actively look for native job candidates (general obligation=3, obligation for certain branches=1.5, no obligation=0), working permit limitation to certain branches (existent=3, inexistent=0) and the possibility for asylum seekers to switch between different branches (not possible=3, possible after 24 months=1.5, possible after 12 months=1, no restriction=0). AG=Aargau; AI=Appenzell Innerrhodes; AR=Appenzell Outerrhodes; BE=Berne; BL=Basel Land; BS=Basel City; FR=Fribourg; GE=Geneva; GL=Glarus; GR=Grisons; JU=Jura; LU=Lucerne; NE=Neuchâtel; NW=Nidwald; OW=Obwald; SG=St Gall; SO=Solothurn; SZ=Schwyz; TI=Ticino; UR=Uri; VD=Vaud; VS=Valais; ZG=Zug; ZH=Zurich.

Conditions for integration

We consider the Federal Decree on measures to decrease spending on the asylum system from 1999 as opening an implementation window (Ridde Reference Ridde2009; Zahariadis and Exadaktylos Reference Zahariadis and Exadaktylos2016) within which the cantons had to choose their implementation strategies. The MSA allows us to model different constellations of actors, problem perceptions and policy contexts to understand the respective decisions. This framework conceives of policy decisions as the result of a simultaneous conjunction of three different streams at a given time: problems, policies and politics are independent, exogenous streams flowing through a system (Zahariadis Reference Zahariadis2014, 31). Policy decisions are not necessarily a rational reaction aimed at resolving functional problems (Sager and Rielle Reference Sager and Rielle2013, 6). “More often solutions search for problems. People work on problems only when a particular combination of problem, solutions, and participants in a choice situation make it possible” (Kingdon Reference Kingdon1984, 91). Along with Spohr (Reference Spohr2016), this article argues that the MSA analysis should integrate inherited policy paths, which distribute decision power and coin actors’ identities and interpretations of situations (Olsen Reference Olsen2001; Boräng Reference Boräng2015). Although the implementation window was present in all cantons between 1999 and 2004, what differs between the cantons is the interplay of actor constellations (politics stream), problems and the policy inheritance (policy stream).

To address the “many variables, few cases” issue, we analyse six causal conditions, which are composed, overall, of 10 indicators. If two indicators represent the same theoretical reasoning, then we consider them as functionally equivalent secondary-level dimensions of one basic-level condition. Such factors are combined using the logical operator “OR”, denoted by +. When several indicators need to be observed in combination, this is expressed with the logical operator “AND”, denoted by * (Goertz Reference Goertz2006, 40, 41). Adopting QCA notation, we use capital letters if a feature is present and lowercase letters if the feature is absent.

Problem stream

The problem stream refers to conditions that policymakers, interest groups and other policy actors believe warrant attention (Ackrill et al. Reference Ackrill, Kay and Zahariadis2013, 873). However, what decisionmakers conceive to be the relevant problem that requires a solution is not exogenously given: problems do not exist but have to be defined by someone (Knaggård Reference Knaggård2015, 452). Although the MSA has always highlighted this (Kingdon Reference Kingdon1984; Herweg et al. Reference Herweg, Huß and Zohlnhöfer2015, 436), Knaggård (Reference Knaggård2015, 451) also emphasises that “studies using the MS[A] hide the power present in defining public problems. Public problems are here seen as those understood as being in need of political action”.

Party political actors seek to comply with voters’ preferences because of the need to be re-elected (Sager Reference Sager2007a, 435). “Parties are interested in promoting issues they own, and care about how much their proposals impact their chances of re-election” (Herweg et al. Reference Herweg, Huß and Zohlnhöfer2015, 441). Voters’ preferences are influenced by “the cultural characterisations or popular images of the persons or groups whose behaviour and well-being are affected by public policy” (Schneider and Ingram Reference Schneider and Ingram1993, 334). Therefore, the problem perceptions of responsive politicians are to a large extent influenced by the social construction of the relevant target groups, which “indicate whether the problems of the target population are legitimate ones for government attention, what kind of game politics is (…), and who usually wins” (Schneider and Ingram Reference Schneider and Ingram1993, 340). Hence, the population defines the problem by social construction, which in turn influences public judgements of problem saliency and of the preferred policy approaches (Steinacker Reference Steinacker2006, 460). This resonates with Herweg et al. (Reference Herweg, Huß and Zohlnhöfer2015, 438) who also argue that the relevance of problems in the problem stream is related to public opinion: “policy makers will start to perceive a condition as a relevant problem if they believe that its persistence threatens their re-election”. Thus, we expect the social construction of the target group “asylum seekers” to be the relevant problem definition for the implementation decision of political parties in need of re-election.

Although asylum seekers are per definition powerless, the political attitudes of the population indicate their social image as being deserving or undeserving (Lenschow et al. Reference Lenschow, Liefferink and Veenman2005; Lavenex Reference Lavenex2007). The main indicator for political attitudes in Switzerland is language. German-speaking regions (G) have a less liberal attitude towards asylum seekers than many cantons in the Latin cultural settings, and urbanised regions (U) are generally more liberal than rural regions (Christin et al. Reference Christin, Hug and Sciarini2002). We further consider the cantonal results in the public vote on the SPP’s popular initiative “Against asylum abuse” on 24 November 2002, which called for a further tightening of the asylum law. An attitude that is unfavourable towards asylum seekers (A) is observed in two instances: first, if a canton is both German-speaking (G) AND rural (u), OR, second, if a high share of the population voted for the “Against asylum abuse” initiative (V). The canton Ticino, for instance, is not German speaking, but still displays an unfavourable attitude because of the high share of anti-immigration votes:

$${\rm A}\,{\equals}\,{{\rm G^{\asterisk}u}} {\plus}{\rm V}$$

According to Steinacker (Reference Steinacker2006), if asylum seekers are perceived as deserving (a; dependents), then the choice of policy is likely to be beneficial, in our case, enabling integration (INT). Conversely, if asylum seekers are constructed negatively (A; deviants), then party politicians seek to claim credit for being tough on the unworthy group (int).

Conversely, we posit that cantonal bureaucracies define the problem functionally in view of the labour market situation: “public administration (…) will seek the technically best policy design, that is, the solution most adequate to the problem under discussion” (Sager Reference Sager2007a, 436). If the labour market is already highly saturated, then it would constitute a contradiction to integrate a fresh workforce. High cantonal unemployment rates (UR) may indicate high saturation. Yet, most asylum seekers with work permits concentrate only on the hotel and restaurant industry, where a varying share of seasonal workers obtain their permits from the cantons (SW).Footnote ² The labour market is too saturated to facilitate the integration of asylum seekers (S) if either the rate of unemployment is high (UR) OR there are already many seasonal workers (SW):

$${\rm S\,{\equals}\,UR{\plus}SW}$$

Policy stream

In his conceptualisation of the policy stream, Spohr (Reference Spohr2016, 252) employs historical institutionalism by claiming that “ideas have to fulfil the criteria of technical feasibility and value-acceptability to win acceptance in policy-networks. Technical feasibility means that policies have to be implementable to be chosen, which involves compatibility with existing institutional or organisational structures (…). Thus, the institutional setting is important for the criterion of technical feasibility”. Accounting for existing policies means rejecting “the traditional postulate that the same operative forces will generate the same results everywhere in favour of the view that the effect of such forces will be mediated by the contextual features of a given situation often inherited from the past” (Hall and Taylor Reference Hall and Taylor1996, 941; see also Thomann and Manatschal Reference Thomann and Manatschal2015).

The need to make an implementation decision – in this case, the choice between enabling or restricting integration – was present in all cantons between 1999 and 2004. What differs between the cantons is the policy inheritance within which this decision was made. Cantonal policy paths characterise the implementation of unemployment law (Battaglini and Giraud Reference Battaglini and Giraud2003). A reintegration-oriented implementation strategy is directed towards improving placement capability, filling gaps in the qualifications of the unemployed, and diminishing the negative consequences of unemployment (Battaglini and Giraud Reference Battaglini and Giraud2003, 288).

Inherited policy paths can be expected to play an important role in the MSA framework, because the problem perceptions of policymakers may partly be based on a sense of identity that is known, accepted and anticipated (Olsen Reference Olsen2001). Especially in migration policy, existing welfare state institutions impact the boundaries of social solidarity, enhance generalised trust, and influence citizens’ views of how the state can and should protect individuals (Boräng Reference Boräng2015). If this is true, inherited policy paths defining the policy stream can “rule out” and replace the prevailing problem perceptions (attitudes or labour market saturation). Existing, integration-oriented welfare institutions are hence functionally equivalent to an integration-prone problem stream; it takes either of the two for policies that grant asylum seekers the right to access labour markets. For both political party actors and bureaucratic actors, a reintegration-oriented policy path (P) can equally produce a tendency towards enabling integration (INT).

Table 1 reports the six conditions that result from these 10 indicators and their expected impact on both enabling and restricting integration. The MSA implies that theoretical expectations must be formulated on the combined, instead of isolated, effects of these factors. In terms of method, we take this causal complexity into account by employing fsQCA. In terms of theory, we now derive theoretical expectations on the interplay of politics, problems and institutional path dependencies.

Table 1

Conditions for labour market integration and directional expectations

Note: SPP=Swiss Peoples Party.

Hypotheses

We have outlined four assumptions:

1. 1. Policy choices are made only when a particular combination of actor constellations (politics), problems and solutions in a choice situation makes it possible (Sager and Rielle Reference Sager and Rielle2013, 6). Politics, problems and policies are a necessary part of different conjunctions, which only jointly are sufficient for (→) enabling or restricting integration.

2. 2. Different social constructions of asylum seekers provoke different problem assessments of political party actors seeking re-election.

3. 3. The cantonal bureaucracies act according to the salience of the technical (functional) problem of labour market saturation.

4. 4. Both political and bureaucratic actors refer to cantonal policy paths, which can replace the problem stream in the MSA.

We can hence hypothesise two alternative explanations for policies enabling integration and two additional restrictions on integration.

First, a strong political left will favour enabling integration if the population’s attitude is favourable towards asylum seekers or the canton has an integration-friendly policy path. Generally, party political actors are only clearly dominant when the cantonal bureaucracy is weak. Formally, we can write this as follows:

$${\rm H1\colon\, L{\asterisk}b{\asterisk}}\left( {{\rm a{\plus}P}} \right)\to{\rm INT}$$

Second, however, if the cantonal bureaucracy is strong, then it favours enabling integration if either the labour market is not too saturated or the canton has an integration-friendly policy path. For the cantonal bureaucracy to dominantly favour enabling integration, the political right must be weak:

$${\rm H2\colon\, B{\asterisk}r{\asterisk}}\left( {{\rm s{\plus}P}} \right)\to{\rm INT}$$

Third, in the absence of a strong bureaucracy, a strong political right achieves restrictions of integration if the attitude of the population is unfavourable towards asylum seekers and the canton has an integration-averse policy path:Footnote ³

$${\rm H3\colon\, R{\asterisk}b{\asterisk}A{\asterisk}p}\to{\rm int}$$ Finally, fourth, if no dominant political left pushes for integration, then a strong bureaucracy will favour restrictions of integration when the labour market is already saturated and when the canton has an integration-averse policy path:

$${\rm H4\colon\,B{\asterisk}l{\asterisk}S{\asterisk}p}\to{\rm int}$$ The combined effects of these factors are evaluated using the research strategy currently presented.

Data and methods

To account for the interaction of the politics stream, the problem stream and inherited policy paths, we employed fsQCA (Ragin Reference Ragin2000; Rihoux and Ragin Reference Rihoux and Ragin2008). On the basis of Boolean algebra and its fuzzy extension, the core interest of the QCA method lies in identifying the necessary or sufficient conditions for a certain outcome.Footnote ⁴ QCA allows for the causal complexity behind a given phenomenon. This entails, first, the assumption of conjunctural causality, meaning that the effect of a single condition unfolds in combination with other conditions (subsequently called configurations or paths), as the MSA assumes. The notion of equifinality then captures that integration may have several, mutually nonexclusive, context-dependent explanations. The assumption of causal asymmetry entails that enabling integration can be explained differently than restrictions of integration (Schneider and Wagemann Reference Schneider and Wagemann2012, 78).

fsQCA (Ragin Reference Ragin2000) translates variables into sets – for example, the set of “cantons enabling labor market integration”. During the “calibration” process, the cases’ set membership is determined on the basis of substantive and theoretical knowledge. Set membership may vary from the thresholds for full and nonmembership with a crossover point (neither in nor out). For example, a canton may fully enable integration (1) or fully restrict integration (0). The decisive crossover point above which integration policies are more enabling than restricting (INT) and below which integration is more restricted than enabled (int) is 0.5. As the numerical distances between the scores do not continuously have the same qualitative meaning (see Figure 1), we calibrate labour market integration using the indirect method of calibration, which involves an initial grouping of cases into previously defined set-membership scores (Schneider and Wagemann Reference Schneider and Wagemann2012, 35). Owing to limited space, the measurement, descriptive statistics and calibration of the sets are discussed in depth in online appendix A. Lacking information on policy paths led to the dropout of the cantons Schaffhausen and Thurgau (Battaglini and Giraud Reference Battaglini and Giraud2003).

After calibration, all logically possible combinations of conditions are listed in a “truth table”. Each case belongs to a configuration (row). The following logical minimisation process identifies the shortest possible causal expression for those configurations, implying the outcome (the solution term). The basic idea is that if an outcome, for example, INT, is found in a canton displaying conditions L, B and P, as well as in another case displaying L and B, but not P, then the presence or absence of P obviously does not make a difference for the occurrence of INT. Formally, we can write this as L*B*P+L*B*p→INT and minimise it accordingly to L*B→INT.

To evaluate the results, consistency indicates the degree to which the empirical evidence at hand is in line with the statement of sufficiency or necessity. Furthermore, the proportional reduction in inconsistency (PRI) indicates the degree to which a given causal configuration is not simultaneously a subset of both the occurrence and the nonoccurrence of the outcome. Coverage indicates how much of the empirical information has been explained (Schneider and Wagemann Reference Schneider and Wagemann2012, 128, 139, 235–239).

The choice of appropriate levels for these measures is research specific. Consistency sufficiency should not fall below 0.75. For necessary conditions, a lower consistency threshold of 0.9 applies; coverage expresses their empirical relevance in terms of the condition set not being much larger than the outcome set, and the relevance of necessity (RoN), in terms of the condition being close to a constant (Schneider and Wagemann Reference Schneider and Wagemann2012, 143, 147). When deciding upon the raw consistency threshold for sufficient truth table rows, we considered “gaps” in the consistency values and low PRI scores. A further criterion was then the absence of deviant cases consistency in kind with qualitatively different membership scores in the condition and outcome sets (Schneider and Rohlfing Reference Schneider and Rohlfing2013). The prioritisation of consistency and differing calibration methods produce low coverage values that do not adequately reflect the solution’s acceptable explanatory power: approximately two-thirds of all cases are explained.

We tested the robustness of our results against different calibration thresholds; see online appendix B (Skaaning Reference Skaaning2011; Schneider and Wagemann Reference Schneider and Wagemann2012, 285–293). Applying the enhanced standard analysis procedure, we resort to the intermediate solution, make theoretically informed directional expectations (Table 1) and ensure that the results of necessity and sufficiency are coherent with each other (software: R packages QCA and SetMethods; Quaranta Reference Quaranta2013; Duşa and Thiem Reference Duşa and Thiem2014). The raw data, prime implicants, alternative solution terms, truth tables, indications on limited diversity and untenable and simplifying assumptions are reported in online appendix C (Schneider and Wagemann Reference Schneider and Wagemann2012, 167ff, 200ff).

Theory evaluation

We now discuss our hypotheses following Ragin’s principles of theory evaluation, as extended by Schneider and Wagemann (Reference Schneider and Wagemann2012, 295–305). The scenarios expected and those not expected in the hypotheses are compared with the scenarios that were empirically (not) observed to answer three questions: first, which parts of the theory are supported by the findings? These are the areas shaded white in Tables 4 and 5. Second, in which direction should theory be expanded (grey areas)? Third, which parts of the theory need to be dropped (dark areas)? We computed the intersections of the theoretical expectations with the conservative solutions and their logical complements. The formal details are presented in online appendix D.

Table 4

Theory evaluation for enabling integration (INT)

Note: Bold: hypothesised combinations.

Hypotheses: L*b*a+L*b*P+B*r*s+B*r*P → INT.

Source: Based on Schneider and Wagemann (Reference Schneider and Wagemann2012, 301).

Table 5

Theory evaluation for restricting integration (int)

Note: Bold: hypothesised combinations. Hypotheses: R*b*A*p+B*l*S*p → int.

Source: Based on Schneider and Wagemann (Reference Schneider and Wagemann2012, 301).

The upper left quadrant of Table 4 provides no empirical support for our first hypothesis on the role of a dominant political left in combination with an integration-prone problem stream. In contrast, our second hypothesis is, in fact, partly reflected in the results. If the politics stream encompasses no strong political right-wing force pursuing an opposite policy (r), then a strong bureaucracy (B) opts for enabling the integration of asylum seekers into the labour market, sticking to the integration-friendly policy path (P). However, H2 was too bold in two respects: first, the bureaucracy also refers to the problem stream in terms of a positive construction of the target group in this situation (a). Hence, second, our results do not support the assertion that the bureaucracy is the actor primarily guided by technical problems. We observed H2 both in cantons with a highly saturated labour market (S) and in cantons with a strong left (L). These two factors only become irrelevant if we introduce counterfactual reasoning (Table 2).

The lower left quadrant of Table 4 turns our attention to alternative, unexpected explanations of why cantons grant asylum seekers access to the labour market. Two cases reveal that even if none of the three types of actors dominate the politics stream (l*b*r), this can imply enabling integration if the problem stream (a*s) and integration-friendly policy paths (P) unambiguously point to integration. Another case suggests that even when the political right is the dominant actor (R*l*b) and the problem stream points to low integration (A*S), asylum seekers are given the right to work if the policy path is integration friendly (P).

The right half of Table 4 indicates a low empirical relevance to delimit theory and confirms that we overlooked possible explanations for enabling integration (Schneider and Wagemann Reference Schneider and Wagemann2012, 302–303). As we argued above, such factors include a more differentiated conception of the politics stream – namely, the personalities of policy entrepreneurs and short-term political considerations in electoral campaigns.

Table 5 neither shows evidence for hypothesis three nor any relevance of a strong political right for restricting access to the labour market in general. In fact, the upper right quadrant reveals that contrary to all of our expectations, the actor constellation of a dominant right with a weak bureaucracy (R*b), a negatively constructed target group (A) and an integration-averse policy path (p) can still imply enabling integration.Footnote ⁵ The case of Aargau, where this happened, is indeed a Sonderfall. The population is unusually split, and there is a strong group of activists who actively support the integration of asylum seekers (Schmidt Reference Schmidt2014).

Conversely, the upper left quadrant of Table 5 provides some support for our fourth hypothesis – namely, that a dominant bureaucracy (B*l) will opt for restricting asylum seekers’ access to the labour market if the latter is too saturated (S) and the policy path is integration averse (p). H4 needs refinement as, again, the assumption that the negative construction of the target group (A) is irrelevant for the bureaucracy was misleading.

Similar to the results for enabling integration, we failed to hypothesise that a politics stream with no dominant actor can imply restricting integration if asylum seekers are constructed as undeserving (l*r*b*A) (lower left quadrant of Table 5). Interestingly, in this situation, the attitudes of the population seem to dominate the problem stream and to somewhat “rule out” the fact that either the labour market is still receptive (s) or the policy path is integration friendly in principle (P).

In summary, we have found no empirical support for our two hypotheses, which stated that an interplay of party political actors seeking re-election with the social constructions of the target group “asylum seekers” in the population and/or policy inheritance would matter for integration. This is somewhat surprising, given the high politicisation of this policy issue. Yet, party politics did matter. First, the absence of a dominant political left favouring integration was a necessary condition for policies restricting the labour market integration of asylum seekers. Second, the issue was instrumentalised in electoral campaigns. Conversely, we did find empirical support that a strong bureaucracy acts according to inherited policy paths. The results highlight the usefulness of complementing the MSA with policy inheritance. Existing welfare institutions sometimes “replaced” the problem stream in the MSA. However, the public administration referred less to technical problems than was expected; the social construction of target groups does matter for the bureaucracy. We conclude by discussing the implications of these findings.