Explaining Foreigners’ Political Rights in the Context of Direct Democracy: A Fuzzy-Set QCA of Swiss Cantonal Popular Votes

In the direct democratic arena, the consensus of voters is required to deliberate policies; without that consent policies are blocked. When bills that support cultural diversity or foreigners’ integration are put into referendums, voters may or may not exert their veto power over the proposed policies. In order to determine under which circumstances these types of bills are successful in referendum, I have undertaken a fuzzy-set qualitative comparative analysis of 39 referendums about the extension of political rights to foreigners at the Swiss cantonal level. My analysis identified a total of five theoreticallyinformed conditions that explain citizenship liberalization and the success of popular votes. I then located these conditions within two configurational hypotheses which postulate how referendum proponents might overcome direct democratic hurdles. The analysis of the success of referendums reveals that the only sufficient path that leads to the popular vote’s success is to insert the sensitive issues into a multi-faceted bill. As demonstrated by a more in-depth case analysis, the sensitive object is successful because it is hidden from voters during the referendum campaign or because other priority objects inside the bill reduce its salience.


Introduction
Christian Joppke (2010) asserted that in Western democracies it is possible to witness a trend toward what he called "citizenship light" (CL) a form of citizenship characterized by the following attributes: easy access to formal citizenship status; more rights than obligations; the lack of a sharp distinction between citizens and certain aliens; and capped by thin identities. Joppke (2007Joppke ( , p. 38, 2010 suggested that citizenship is becoming less exclusive toward aliens, given that aliens and citizens have similar or identical rights, and more inclusive toward minorities, due to the attitude of promoting cultural pluralism.
However, while the literature on CL contributes to the understanding of the citizenship liberalization, it has two important shortcomings. First, CL literature lacks a comprehensive understanding of the CL trend in other democratic decision-making contexts than the representative parliamentary democracy. Second, it lacks a systematic comparative perspective which allows one to validate findings across a medium-large number of cases.
This article aims to close the resultant gap in CL literature regarding the CL phenomenon in the context of direct democracy and identify under which conditions or configuration of conditions CL is successful or unsuccessful in the direct democratic arena by using a systematic comparative analysis on a medium number of cases. From Joppke's suggestions it is possible to ascertain that the representative democratic arena is particularly open to various forms of CL. Nevertheless, this may not be the same case for the direct democratic arena where the characteristics of the political debate and political actors involved are different. The CL process differs from other democratic decision-making processes because in the direct democratic procedure, the consent of the population plays a central role in deliberating policies. Therefore, the population itself constitutes an additional veto player, which could make significant policy changes more difficult to achieve than in the representative democratic context.
An implication of this is that the direct democratic arena may be less prone to CL than the representative democratic arena is. This article individuates the most appropriate formulas that indicate which strategies are successful and which conditions lead to the CL failure. The results, other than stimulating the normative debate on the decision-making procedure of the direct democratic system, offer the opportunity to CL promoters to comprehend which strategies to adopt to achieve their political goals by popular vote.
I decided to focus the analysis on the most homogeneous subfield in terms of the extension of citizenship rights in the direct democratic context. Specifically, I analyzed Swiss popular votes related to the rights of foreigners to vote and/or be elected. The extension of political right to foreigners is a subfield of CL because when political rights are given to non-citizens, political participation shifts from an exclusive form to an inclusive one that is not based on nationhood but rather on a universal dimension of citizenship rights.
The Swiss context plays an important role from a theoretical perspective, given that these types of CL policies in Switzerland have always been subject to a popular vote and have never been implemented through the representative democratic process. In this respect, the Swiss context can be considered as an ideal-typical example of CL in the context of the direct democratic arena.
In order to address this context, I identified a total of five theoretically-informed causal conditions that explain citizenship liberalization and popular vote outcomes. I then located these five conditions in two configurational hypotheses. Finally, I used fuzzy-set qualitative comparative analysis (fsQCA) in order to test conditions and hypotheses on 39 CL popular votes which involved the extension of political rights to non-citizens in the Swiss cantons.

Specific Cases that Will Be Analysed
The population of cases was built using information gathered from the database of the Centre for Research on Direct Democracy (http://www.c2d.ch). I have focused the study on both a specific homogeneous geographical area and jurisdiction level: Switzerland and the cantonal level, respectively (in Appendix A the referendums list).
The Swiss cantonal cases differ from the degree of political rights extension object. These differences are due to the specific Swiss socio-cultural context and to the legal characteristic of Swiss referendums.
The differences in the object's content are due to specific circumstantial elements which are intrinsic to the Swiss context, characterized by different types of cleavages that separate cantons. According to the cleavage approach (Bartolini & Mair, 1990, pp. 213-220), the population is separated by certain ascriptive sociocultural elements. These ascriptive elements are sources of normative values that add a sense of collective identity and determine the attitudinal characteristics of a group. In Switzerland, scholars individuated empirical evidence of the existence of linguistic cleavage between French-speaking and German-speaking Switzerland (Kriesi, 1998a) and religious cleavage between Protestant and Catholic areas (Caramani, 2004, p. 273;Goldberg & Sciarini, 2014); cleavage was also found between rural and urban zones (Kriesi, 1998b), in class differences between blue-and white-collar workers (Kriesi, 1998b), and in new classes difference between the so called losers and winners of globalization (Giugni & Sciarini, 2008;Goldberg & Sciarini, 2014). These cleavages are sources of normative common values and beliefs that contribute to building identity within social groups and around which both the elite and the electorate organize their action. The difference in the object might be related to the diverse social sensibility of the referendums' promoters toward immigration and national identity topics (Eugster & Strijbis, 2011) and the diverse perception of the direct democratic instruments of popular votes. Such diversity is due to differences in democratic participatory models (Bühlmann, Vatter, Dlabac, & Schaub, 2014) and explains why the reforms are generally more radical in certain cantons, such as Neuchâtel or Geneva, than in others, such as Appenzell Ausserrhoden or Graubünden.
In addition to the cleavage perspective, differences in object content are also related to the typology of the referendum. The Swiss popular vote system includes four typologies of direct democratic instruments: 1) The mandatory referendum, which is a required, binding, and non-popular initiated vote; these referendums are usually held on important matters such as a constitutional amendment; 2) The facultative referendum, which is binding, popular initiated and not a required vote; these referendums aim to oppose a law proposed and voted on by the parliament; 3) The popular initiative, which is binding, popular initiated and not a required vote; these referendums aim to introduce a new constitutional amendment; 4) The counterproposal to the popular initiative, which is binding, non-popular initiated and a nonrequired referendum on a similar object of the popular initiative.
The popular votes under my investigation all refer to mandatory referendums, popular initiated referendums, and counterproposals. The most radical forms of political rights extension are proposed during popular initiated referendums, as this type of referendum does not need the consensus of the majority of the political actors to be held but rather can be requested by a small group of citizens.

Research Question
This article seeks to answer the following question: "under which combinations of conditions are attempts to extend political rights to foreigners by popular vote successful or unsuccessful?" The focus is of this study is to individuate the causal frames that lead to the referendums' success or failure. I presume that there is more than one path that leads to the outcome. This implies a complex reality in which causal conditions are not separate entities but often works in conjunction with set theoretic manner.
Essentially, the hypothetical framework to which I refer, presumes a non-linear causal ontology which relies on the concepts of sufficiency and necessity. This type of causal ontology has a holistic understanding of causation which implies the existence of a deterministic reality. Nevertheless, it is important to point out that the meaning of deterministic reality here merely indicates the presence of necessary and sufficient conditions, and not to the presence of certainty or absence of error in the results.

Conditions Selection
The conditions selection process is linked to two specific elements strictly related to the type of study conducted in this article.
The first element refers to the nature of the object at stake which can be analysed by using conditions linked to the CL or direct democratic literature. In relation to this I have decided to consider both literatures in order to have a wider understanding of the phenomenon under analysis.
The second element refers to the typology of condition that should have a well-defined nature in order to answer the research question. This means that they must have a qualitative threshold below which the effect does not occur, and according to previous studies, they should be determinative in terms of sufficiency and necessity for the outcome. The implication of this second element is to exclude conditions which consider the probability of occurrence of an outcome to be dependent on the degree of the condition and conditions whose causal effect are already covered by other conditions.

Government Ideology
The literature has often claimed that the enfranchisement of resident aliens is affected by party ideology. Governments composed predominantly of leftist parties tend to provide more inclusive and generous citizenship policies than rightist governments (Bird, Saalfeld, & Wüst, 2010;Green, 2005;Janoski, 2010). Meanwhile, strong rightist governments tend to restrict citizenship rights (Joppke, 2003) in order to accommodate xenophobic movements (Joppke, 2008, p. 166). This condition implies that rightist governments alone are not sufficient for CL opposition.

Far-Right Populist Parties
According to Howard (2010), anti-immigrant and xenophobic sentiments pre-exist within the population, and populist parties are the catalyst which mobilize these sentiments. Strong right-wing populist parties are a necessary and sufficient condition that prevents citizenship liberalization (Howard, 2010, pp. 735-751). Howard describes a non-symmetrical relationship, in which the absence of the strong right-wing populism is a necessary but not sufficient condition for CL (Howard, 2010, p. 747).

Popular Initiated Referendums
The legal characteristics of a popular vote can influence the vote's outcome. As previously indicated, mandatory referendums are instruments that allow citizens to decide on particular policies adopted by the political authorities; meanwhile, popular initiatives are instruments that enable parts of the population to enforce popular votes that may go against the will of parliamentary majorities (Freitag & Vatter, 2006). In our specific context, the popular initiative usually proposed more radical policies on the extension of political rights to foreigners, transversal to the Swiss socio-cultural cleavages. This condition is a prerequisite for popular vote failure. Moreover, in Switzerland, the success rate of popular initiatives is under 10% (Setälä, 2009, p. 49), which is already a quasi-sufficient condition for the failure of popular votes. In addition, Kriesi (2006, pp. 605-606) observes that the government succeeds whenever it opposes a popular initiative. Therefore, in the context of popular initiatives, the opposition of the government is a sufficient and necessary factor for an initiative's failure.

Popular Votes with Single or Multiple Issues
Popular votes with multiple issues (those that gather together several policies into a single referendum question) and those with single issues (those that address only specific policies in the referendum question) can behave in different ways. Popular votes with multiple issues have a better chance of passing because the complexity of the vote can reduce the sensitive object's salience (Ginsburg, 2009, p. 3). In such cases, CL matters can be hidden inside a broad bill and the political elite can better control the propaganda related to the bill. In contrast, a single-issue popular vote on CL policy has a lower chance of passing because the salience of the CL matters would be high (Cooter & Gilbert, 2010, p. 745). In the Swiss context, multiple-issue popular votes only refer to mandatory referendums, while single-issue popular votes can be proposed in mandatory referendums, popular initiatives, or counterproposals.

Political Elite: Split or Cohesive
The degree of polarization of the political elite influences the outcome of a popular vote: the more fragmented the political elite is on the issue put to popular vote, the less likely the citizens are to agree with the government (Kriesi, 2006, pp. 601-602). This hypothesis has a symmetrical nature; when all parties are unified in favor of a proposal, a one-sided information flow exists and the elites will be more persuasive (Ray, 1999, p. 298).

Expected Causal Configurations
The five conditions established above are divided into two groups: i) The conditions government ideology and presence/absence of strong far-right populist parties belong to the group of CL literature's conditions that explains CL restriction or extension in the sole context of representative democracy; ii) The conditions popular initiatives opposed by the government, multiple-issue referendum, and elite cohesive or split belong to the group of general direct democracy literature that explains the popular vote's success or failure.
These two groups of conditions are circumstantial to their specific contexts; none of them fully explain the specific context of CL extension or restriction in a direct democracy. Likewise, the outcome of interest can be explained by a configuration of those conditions related to the CL literature and those related to the direct democratic literature. Therefore, the potential explanations expressed in the conditions contain both elements that have been discovered in the context of representative democracy and those that are important only in the context of direct democracy. The presence of both elements highlights the fact that even when the final decision is made in a direct democratic fashion, the political process involves elements of both representative and direct democracy. As a result, the nature of the phenomenon under our investigation is configurative; the use of such an approach allows us to obtain the set of potential configurations of conditions that can combine the two different groups of conditions outlined above. In relation to this approach, we believe that our outcome of interest would be best understood in terms of configurative causation rather than the average causal effects of variables across cases.

Expected Configurations in CL Success in the Context of Direct Democracy
In the direct democratic system, the consent of the population plays a central role because without it, policies can be blocked or delayed. Therefore, when trying to amend new policies during the popular voting process, both the government and the parliamentary majority must consider the population as a new veto player. CL policies may be easily subject to the voters' veto due to pre-existent negative attitudes toward aliens (Howard, 2010), the presence of difference socio-cultural cleavage or, as pointed out by Sniderman and Hagendoorn (2007, pp. 105-106) and Marcus, Sullivan, Theiss-Morse and Wood (1995), because of the less-liberal attitude toward cultural diversity and policy change of the electorate compared to the more-liberal representative democratic actors. As result, in the representative democratic arena political outcomes can deviate from voters' preferences, which is not the case in the direct democratic arena given that voters themselves are called upon to express their preferences on the outcome.
The hurdles posed by the direct democratic arena and the explanations expressed in the conditions contain elements of both the representative and the direct democratic arenas; such elements are translated into a complex configurative combination of conditions that work together in order to achieve CL success. Therefore, in relation to the outcome "CL success" during referendum, we expect configurations of conditions with at least one element coming from the group of CL conditions and an element coming from the group of direct democratic conditions.
The configurational hypotheses consider the direct democratic hurdles as having a central role. CL promoters must adopt complex political strategies in order to pass a CL bill. I have identified two such strategies: 1) the open strategy, and 2) the hidden strategy.
In the open strategy, CL issues are openly debated during the referendum campaign. Considering the voters' veto role in the direct democratic arena, promoters of CL policies should provide a one-sided influx of information (without strong opposition) in order convince voters about the CL object. In these circumstances, the CL promoters will seek elite cohesion in order to create little to no opposition to CL and positively influence the voters' preferences on the issue; however, the absence of far-right populist parties is also necessary in order to avoid conflicts over the CL reform. In these circumstances, CL promoters can decide to pose the CL issue in the form of a referendum question with little fear that the CL policy will be considered unpopular by voters. Indeed, the strong elite's propaganda in favor of CL and the absence of mobilizing actors that could activate latent xenophobic and anti-immigration sentiments would lead voters to accept the bill.
The open strategy leads to the following configurational hypothesis:

Configurational Hypothesis 1 (CH1): absence of populist parties (∼POP) AND elite cohesion (ELITE) AND non-popular initiative (∼INI). 1
This configurational hypothesis considers that CL can be achieved if a cohesive elite can introduce the issue without being afraid of resistance from a populist rightwing party.
CL promoters might also adopt a hidden strategy of communication in order to pass the bill. In this context, CL's promoters fear the possible mobilization of pre-existent xenophobic sentiments among the population. Basically, left-wing governments, as main promoters of CL policies, are aware that voters might stop the bill because of their conservative attitudes on this type of object. Therefore, these governments might decide to avoid informing voters about the bill's content in order to minimize the risk that the voters would focus on the sensitive issue in the campaign (Besley & Coate, 2008;Ginsburg, 2009). This strategy also reduces conflict among political actors during the referendum debate, given that the issue is not discussed. We define this possible strategy with the term hidden strategy, given that the CL issue is hidden inside a broad bill and people are not aware of the bill's content.
The hidden strategy leads to the following configurational hypothesis: Configurational Hypothesis 2 (CH2): left-wing government (LEFT) AND multiple-issue referendum (MULT) AND non-popular initiated initiative (∼INI). 2 This configurational hypothesis considers that CL can be achieved if a left-wing government includes the issue in a multiple-issue referendum. In this situation, unpopular policies would be kept hidden from the people. As a consequence, voters' preferences would not be determined by the CL object but by other objects included in the referendum's question.
The open strategy and the hidden strategy refer to an optimistic and a pessimistic view, respectively, of voters' attitudes toward aliens. Indeed, the open strategy considers voters' negative sentiments toward aliens as latent and thereby able to be activated only by an agency factor. Therefore, in order for the vote to be successful, it is sufficient to deactivate or avoid activating the action of the agency factor (i.e., populist parties). The hidden strategy considers that voters' negative sentiments toward aliens are already active because they are pre-existent. Therefore, in order for the vote to succeed, it is important to avoid activating this pre-existent sentiment by hiding the issue within a broad bill.

Expected Configurations in CL Failure in the Context of Direct Democracy
In relation to the failure of CL, it is assumed that the many hurdles facing the passage of a bill in a direct democracy are translated into many sufficient conditions which alone are enough to stop CL. The rejection of a CL bill in the context of a direct democracy would be relatively easy due to the hurdles of the direct democratic arena explained above. In these circumstances, conditions that explain referendum failure and CL failure in the context of a representative democratic system are alone sufficient to the failure of the referendum.
Hypothesis 3 (H3): CL bill will fail when it is opposed by a rightist government (∼LEFT) Hypothesis 4 (H4): CL bill will fail when it is supported by a divided government (∼ELITE) Hypothesis 5 (H5): CL bill will fail when it is opposed by a strong far-right populist party (POP) Hypothesis 6 (H6): CL bill will fail when is proposed as popular initiative (INI) Hypothesis 7 (H7): CL bill will fail when is put in a single-issue referendum (∼MULT) 3

Research Design
In order to test these hypotheses, I decided to use the fsQCA (Ragin, 2000(Ragin, , 2008. The choice of this method is primarily linked to configurative ontology and the assumptions of causal asymmetry as set out by the hypothesis above. A configurative ontology presupposes that conditions can work together in a configurational or interactive way within a specific theoretical framework. Meanwhile, causal asymmetry implies that an explanation of the failure of CL reform proposals cannot draw on simple negations of the factors that explain the success of CL reform proposals. The alternative outcome of a social process therefore requires its own explanation.

Condition Operationalization
All variables are transformed into sets of values; this allows one to individuate sufficient and necessary relationships.
Conditions with a defined nature receive a crisp binary score of 0 or 1. In this study, these conditions refer to the outcome of the popular vote (Yes or No), the presence of a multiple or single object referendum and the presence of a popular initiated referendum.
Conditions with a non-defined nature receive a fuzzy set score which is a continuous score between 0 and 1 with a qualitative anchor point at 0.5. The 0.5 anchor defines the border between case membership and case non-membership to a condition. In this article, these con-ditions refer to the presence of a leftist or rightist government, the presence of strong right populist parties, the degree of elite division or cohesion.
The Appendix B shows the details of conditions' calibration.

Necessary Conditions
The analysis of necessity did not uncover any necessary conditions; therefore, promoters of CL policies cannot rely on possible crucial conditions in order to succeed in their purposes. In other words, a direct democratic system does not provide essential necessary conditions for CL success, a finding that is in line with our theoretical expectations. Indeed, the hurdles of the direct democratic system inevitably hinder possible necessary conditions that would allow CL to succeed in a representative democratic system. In such circumstances, promoters of CL policies must therefore not rely on direct democracy in order to pursue CL extension.

Sufficient Conditions
The analysis yielded three different solutions: Table 1  Owing to the lack of agreement among QCA scholars on which solution term is preferable and causally interpretable, I decided to briefly discuss cases covered by the intermediate solution formula before producing a final statement of causal interpretation.
The outcome that explains the CL success by referendum has the following parsimonious solution: The condition multiple-issue referendum appears in every path of the parsimonious and intermediate solution formulas. Multiple-issue referendum is also the key factor that should confirm the hidden strategy hypothesis. The political elite seems to use multiple-issue referendums in order to avoid public discussion on CL during the referendum campaign. Given the high number of cases covered by the solution paths, I decided to select a reduced number of cases. As suggested by QCA literature, I selected uniquely covered cases (Schneider & Rohlfing, 2014), because these types of cases can be explained by one unique conjunction of conditions, i.e., there cannot be other explanations for the case's outcome (Schneider & Rohlfing, 2014). In contrast, cases covered by more than one path can be explained from different angles or by a single dominant configuration. This is not an error in interpretation of the configurations but is instead linked to a specific QCA understanding of causal relationships built not on causal mechanisms, but on the counterfactual causal relationship of necessity and sufficiency. As the premise of the discussion, it is important to point out that all cases covered by the parsimonious and intermediate solution formulas refer to total cantonal constitutional revisions.
Path 1: in the case of Basel Stadt 2005 (CHBS05), a long process of elites meeting (Verfassungsrat-Constitutional Assembly) preceded the popular vote. The Verfassungsrat proposed a total constitutional revision that could meet the different political expectations of every political party involved. On March 2005, after six years of meetings, the elected Constitutional Assembly reached important inter-party agreements when the final constitutional draft was accepted by every party except the non-governmental right-wing populist parties of the Swiss People's Party (SVP) and Swiss Democrats (SD). In this instance, the political elites accepted a constitutional reform which included an article on the possibility of extension of political rights to foreigners only in the towns of Riehen and Bettingen (but not in the city of Basel). The government-elites decided to unanimously campaign for the constitutional reform without revealing the sensitive CL content to the people. During the referendum campaign, the political elite kept the existence of a bill on political rights extension to foreigners hidden from the voters.
Path 2: in CHVD02, the Constitutional Assembly tried to meet some, but not all, opposing views on extending political rights to foreigners. On March 2001, the Consti-tutional Assembly voted for the first time on the extension of political rights at the cantonal and local level. Despite the proposition receiving the majority of consent (Bolliger, 2001), the assembly members decided to revote on a nuanced version of the same issue in September 2001. This proposal gained greater support with the Liberal Democratic Party (FDP), which eventually decided to support the agreement; only the Liberal Party (LPS) and the relatively weak SVP opposed the constitutional reforms. During the political campaign, the CL object played a secondary role. The campaign actively focused not only on the extension of political rights, but also on the extension of social rights (such as the right to maternity insurance), the union of Church and State, and the reduction of the cantonal districts from 19 to 10 (Rychen, 2002). In Fribourg 2004 (CHFR04), the primary issue discussed during the referendum campaign was the article on civil unions for homosexuals and not the reform of foreigners' political rights. In both CHVD02 and CHFR04, the CL object played only a secondary role and was subtly hidden by the presence of other priority objects.
From these three cases it is possible to identify two types of hidden strategies that differ in degree: i) a hardhidden strategy, and ii) a soft-hidden strategy. The hardhidden strategy refers to CHBS05, in which the unpopular CL object was never mentioned during the propaganda campaign. The soft-hidden strategy refers to CHVD02 and CHFR04, where the unpopular CL object played only a secondary role and was subtly hidden by the presence of other priority objects.

Causal Interpretation on Political Rights Extension Success
The hidden strategy has been determined to be the causally relevant strategy for CL promoters to reach success in CL policies; configurational Hypothesis 2 (CH2) is therefore confirmed even if some unexpected elements exist. In contrast to CH2, the solution formula does not display left-wing government as a causally relevant condition in the referendum's success, as constitutional revisions allow for great support on the issue from different political forces (including right-wing parties). The specific circumstance of a total constitutional revision is already sufficient to reduce the CL-related divergences of the actors participating in the direct democratic arena.
As previously observed, the hidden strategy hypothesis has a pessimistic understanding of the electorate's Table 3. Causal configurations for referendums success.

Solution formula Cases non-uniquely covered Cases uniquely covered
(1) ∼INI*MULT*ELITE CHAR95, CHNE00, CHGR03 CHBS05 (2) ∼LEFT*∼INI*MULT CHAR95, CHNE00, CHGR03 CHVD02; CHFR04 Cases not covered by the solution formula CHGE05.1; CHNE07.2; CHJU14 behavior as a veto player. Indeed, this strategy suggests that the sole explicit reference to a CL issue in the referendum question is enough for the population to act as a veto player toward the CL reform. As suggested by the solution formula, in the hidden strategy voters' preferences would not be determined by the CL object. Essentially, in a multi-issue referendum, voters would not exert their veto power over the CL object because the salience of the unpopular object would have been reduced. In this strategy, voters would not be aware of the existence of the CL reform, or when they were aware, such as in CHVD02 and CHFR04, they would be required to prioritize their preferences across a range of different topics. The causal core of the hidden strategy is condition MULT. The fact that the CL object was hidden during the referendum campaign supports the causal inference of the MULT condition on the outcome. In the specific context of a total constitutional revision, the elite cohesion (ELITE) and the presence of a mandatory referendum (∼INI) are definitional elements of the multiple-issue referendum (MULT) and have no causal inference on the outcome of referendum success. The cases' observations corroborate Thiem and Baumgartner's assumptions of the parsimonious solution: conditions ELITE and ∼INI are redundant. Essentially, multiple-issue referendum is the Boolean difference maker, and can be considered the only non-redundant and causal factor that explains the outcome of CL success in the context of extending political right to foreigners.

Necessary Conditions for Political Rights Extension Failure
Single-issue referendum (∼MULT) is the necessary condition for citizenship liberalization failure in a direct democracy; the failure of the extension of foreigners' voting rights would not occur without this crucial condition. This result corroborates the initial hypothesis: CL resistance finds fewer hurdles in the context of direct democracy than CL extension. Indeed, a sole explicit reference to a CL issue in the referendum question is necessary for the voters to activate their negative attitudes toward CL and therefore act as a veto player toward the referendum. Moreover, the necessary condition single-issue referendum also provides the ideal context for increasing conflict between the actors involved in the direct democratic debate. Whenever the explicit question on CL reform is put up for direct democratic debate, an ideological conflict over the issue at stake is difficult to avoid. Essentially, the context of the single-issue referendum question enables the actors involved to express their political views about CL during the referendum debate.

Sufficient Conditions for Political Rights Extension Failure
As displayed in Table 4, the parsimonious sufficient solution for the outcome "failure of political rights extension to foreigners" encompasses two different causal combinations of conditions; meanwhile, as displayed in Table 5 the intermediate sufficient solution encompasses three different causal combinations of conditions. Ragin (2006) and Veri (2019) consistency is high, which means that nearly all the cases fit into the patterns identified by the analysis; Ragin's (2006) and Veri's (2018) coverages are also high, which makes the solution empirically relevant and not trivial.
The parsimonious solution suggests that a popular vote will fail whenever the following criteria exist:   The first path is identical in both the parsimonious and intermediate solutions, meaning that each insufficient but non-redundant part of a condition which is itself an unnecessary but sufficient component of the causal configuration is a Boolean difference maker.

Discussion on the Political Rights Extension Failure
In Path 1 does not present any causally redundant condition, given that it appears exactly how it is in both the intermediate and the parsimonious solution. In Path 1, voters opposed the proposition supported by the government. The case of Bern (CHBE94.2) refers to a government counterproposal to the popular initiative CHBE94.1, which was voted on the same day in Parliament and in the ballot. In the counterproposal, CL issues were actively debated during the referendum campaign. Evidence exists that the right-wing populist parties used anti-foreigner arguments, which indicates a causal role of condition POP. Indeed, during the campaign, the SVP highlighted the risk to Switzerland of losing cultural ho-mogeneity in institutions due to foreigners' inability to properly speak the Swiss German dialect. In the middle of the campaign, the FDP suddenly decided to oppose the counterproposal (Kiefer, 1994;Van Liniger, 1994). The FDP's decision came only after they had already offered support for the counterproposal both in Parliament and at the beginning of the referendum campaign. The referendum held in Schaffhausen (CHSH01) was a variant of the total constitutional revision also voted on the same day by FDP. During the parliamentary debate, the FDP favored the extension of political rights to foreigners, but subsequently followed the SVP in the political campaign against the referendum (SDA/ATS, 2000).
Path 2: in the cases of Geneva 1993 (CHGE93.1 and CHGE93.2) and Uri (CHUR95), the cantonal governments ran a campaign primarily focused on small-step policies toward CL, and therefore opposed the popular initiatives only because they were considered too radical to have a chance to pass by popular vote. The government campaign was not focused on the CL matter.
Path 3: in Bern (CHBE10), populist party SVP's opinion on the initiative was moderate and not discernible from the government opinion on the popular initiative. The popular initiative campaign against the referendum was run by a single committee composed of the SVP, the Conservative Democratic Party (BDP), the Federal Democratic Union (EDU), and the FDP, which proposed a unifying strategy of propaganda (Wyler, 2010). The SVP, together with the right-wing parties in the government, argued that extending voting rights to foreigners should not be a tool of integration but a consequence of the integration process (Guggisberg, 2010). Indeed, naturaliza- Table 6. Cases covered by the intermediate solution.

Solution formula Cases non-uniquely covered Cases uniquely covered
(1) ∼∼LEFT*POP*∼MULT*ELITE CHBE94.2; CHGE01; CHSH01 (2) ∼LEFT*∼INI*MULT CHVD92, CHZH93, CHBE94.1, CHBS94, CHGE93.1, CHGE93.2, CHUR95 CHAG96, CHFR97, CHSO97, CHGL10, CHVD11, CHLU11, CHZH13, CHSH14 (3) POP*INI*∼MULT CHVD92, CHZH93, CHBE94.1, CHBS94, CHNE07.1, CHBS10.1, CHBE10 CHAG96, CHFR97, CHSO97, CHGL10, CHVD11, CHLU11, CHZH13, CHSH14 tion was considered the only official tool to guarantee that a person will be integrated and that s/he can understand the society and the language where s/he lives. In Basel (CHBS10.1) and Neuchâtel (CHNE07.1), the populist party was an additional actor to the government action in opposing the popular initiative. For example, in the case of Basel, the SVP's arguments were based on strong xenophobic and anti-immigrant sentiments; the group used street posters with figures of a woman wearing burqa, a young black man with sunglasses, and a suspicious-looking Arabic man in front of the ballot box with the cantonal flag (Weber, 2010). The examination of these specific cases highlights the fact that the condition POP is not uniformly causally related to the outcome. The fact that in Basel and Neuchâtel the condition POP is causally relevant and in Bern it is causally redundant indicates that the solution formula does not provide a full satisfactory understanding of the phenomenon under investigation. Essentially, from a set-theoretic perspective, this ambiguity is not acceptable because the condition POP appears to be an Insufficient but Necessary part of a condition which is itself Unnecessary but Sufficient (INUS condition) over certain cases, but not in other cases. Therefore, POP should be considered a redundant condition because statements of sufficiency are not fulfilled with the intermediate solution formula.
As has been discussed in this section, the conditions that only appear in the intermediate solutions are redundant either because they are not relevant (as in the case of right-wing governments in CHGE93.1 CHGE93.2) or because they provide ambiguous evidence (such as the presence of populist actors in CHBE10, CHBS10.1, and CHNE07.1.). As a result, the parsimonious solution path INI (presence of popular initiative) has been determined to be the only non-redundant solution formula.

Causal Interpretation of Political Rights Extension Failure
The sufficient and necessary conditions corroborate the principle behind our hypothesis on CL failure, which considers the non-acceptance of a CL bill to be relatively easy due to the presence of the hurdles of the direct democratic arena. Indeed, the presence of the necessary condition single-issue referendum and the sufficient conditions activates the negative attitude of the population toward the reform and increases conflict among political actors.
The solution formulas also suggest the same pessimistic view of voters on CL matters, given that the mere presence of a question on a CL issue plays a necessary role in the activation of the voters' veto power. In Path 1, the presence of strong populist actor and a divided government deprives the government of the necessary resources to influence voters' preferences on the issue through a strong one-sided influx of information. In Path 2, the popular initiative object itself, which is always a single issue, and as pointed out above is more radical and transversal to the various socio-cultural cleavages, is sufficient to trigger the veto power of the electorate toward the popular vote.
Parsimonious solution paths corroborate that CL reforms are opposed by the voters because of the high level of conflict inherent in the direct democratic arena. Indeed, in the first path, the conflict is inflated by the presence of populist parties and by a divided government. Populist parties raise the level of conflict inside the government during the referendum campaign, which is an interactive condition that triggers the elite split. Meanwhile, in the second Path, the conflict is high due to the absence of a deliberative arena (such as the parliament), which would allow compromises to be found between the popular vote promoters and other political actors; such deliberation could also increase support for the popular vote. The popular initiative condition INI is by itself sufficient for popular vote failure because popular initiatives are a more divisive vector of policy promotion than mandatory or counterproposal referendums. This divisiveness occurs because the promoters of popular initiatives decide to put the object as it is to a popular vote without engaging in consensus negotiations with the political elite. In a popular initiative, support for the object is usually limited to the popular initiative's promoters, which implies substantial opposition from the government and other political actors. Basically, popular initiatives provide little to no room for deliberation among the political actors involved. In contrast, counterproposal and mandatory referendums are usually subject to a legislative deliberation process in which the political actors' conflicts are reduced; in such circumstances, the majority of the parliament usually must agree on the object to be put into referendum.

Conclusion
This article demonstrates that a system with direct democratic instruments is open to introducing CL reform proposals; however, such a system also makes it almost impossible to gain the majority of the established electorate's support for such reforms. If the lightening of citizenship is to be approved by those who possess the status of a citizen (i.e., nationals), the CL object must be hidden from public scrutiny and embedded in a general constitutional reform. For supporters of CL, this study contains two further messages: introducing CL either through a popular initiative or in a situation in which a strong far-right populist party will stimulate a right-wing government party to oppose the initiative is a recipe for failure in the direct democratic context. This research also suggests a pessimistic view of voters' behavior toward CL reforms. The visibility of the object plays a central role, while at the same time the level of conflict between political actors involved in the direct democratic arena contributes to the referendum outcome. When the object is not visible to the voters and the level of conflict is reduced, the voters will accept the object; meanwhile, when the CL object is visible to the voters and the level of conflict is high, the voters will oppose the object and exert their veto power.
This conceptualization also indicates that conflicts and the nature of the object are mutually dependent in a direct democracy. Indeed, only the conjunction of these two axes produce the outcome of interest; condition(s) located along these two axes are not enough to produce the outcome of interest.
This research also contains certain limitations due to the fsQCA methodology and the research design used that have affected the interpretation of findings. Specifically, the fsQCA analysis usually produces solution formulas that rarely cover the whole population of cases analyzed; as a result, there are cases that are not covered by the solution formulas. The cases that remain uncovered can probably be explained by other conditions that have not been considered by our explanatory model. In order to identify such conditions, further specific case analysis should be conducted. Moreover, the fsQCA analysis only produces causal static statements on sufficiency and necessity; it does not add inferential value by tracing how the combination of necessary and sufficient conditions affects the outcomes. In other words, the fsQCA analysis identifies sufficient and necessary conditions but not how these conditions interact to produce the outcome of interest. The temporary succession of conditions and the causal mechanisms that link the conditions together and lead to the outcome remain unexplained. Finally, the fsQCA analysis reveals statements of sufficiency and necessity with only the conditions available in the explanatory model, implying that it might be possible to miss conditions that might not appear in the solution formula.
Despite the limitations represented by the employed method, this research provides a comprehensive picture of the fate of CL proposals in the context of direct democracy by bringing to the forefront the question of CL in the context of direct democracy and identifying non-trivial sufficient and necessary conditions for the success or failure of CL policy in the direct democratic context.

Appendix B. Calibration Procedure Details
(for raw data please request information to the author)

Left Wing Government Calibration (LEFT, ∼LEFT)
The condition left-wing (LEFT) versus rightwing (∼LEFT) government has a fuzzy set membership. The full membership of 1 means coincides with the ideal typical leftwing government; meanwhile the full non-membership 0 coincides with the ideal type of rightwing government.
I decided to calibrate the ideology of the government in two steps. (i) Firstly I calibrated each government party's ideology fuzzy score (data used are from the Sotomo Institute of Zurich); (ii) then I aggregated these scores considering the party government composition by using the Arithmetic Mean Based on Compensatory Fuzzy Logic as suggested by Bouchet, Pastore, Andrade, Brun and Ballarin (2011) and Veri (2017).  This condition has a complex definition which has to be adapted to our scope. Therefore, in order to fuzzify this condition, I decided to use the Fuzzy Multiple Attribute Condition (FMAC) strategy (Veri 2017).
From the theory it is possible to extrapolate at least two attributes related to our FMAC: 1) the strength of the populist party inside the country and 2) the political agenda of the populist party, which must be far right and anti-immigration.
A third element of our FMAC is determined by the causal context itself, or the CL related referendum's context. In such context, despite that the theory only required strong far right populism I added the position of each populist party in respect of the referendums. This attribute also has the role of activating the whole condition POP in the context of direct democracy. Indeed, the outcome may change depending on whether a given populist party participates in the propaganda against CL or not.
In summary our FMAC is made by 3 distinctive attributes that are necessary and jointly sufficient in determining the final concept. These attributes are: i) the electoral strength of the populist party, The electoral strength has been measured by considering each populist party electoral performance during the election that precede the referendum. I have defined the qualitative breakpoints 0.5 at 4%. ii) far-right political agenda: The empirical information is collected from the Sotomo Institute of Zurich which refers to political parties' ideology at Swiss Cantonal level use above. iii) the populist party campaigning for or against the referendum. In relation to the populist party attitude during the referendum campaign, I decided to use the following scale. If the populist party is against the referendum it is given a value of 1, if it is in favor it is given 0. The cut-off point of maximum ambiguity occurs when there is no position taken towards the referendum or if 50% of the political party is in favor of the referendum.

Popular Initiated Referendums (INI, ∼INI)
The Popular initiated referendum has a clear binary nature and therefore it will have a crisp value of 0 or 1. Whenever the object is put into popular initiative and opposed by the government the score assigned will be 1 and whenever it is not a popular initiative, or it is supported by the government the score assigned will be 0. Popular votes with multiple or single issues have a perfectly binary nature and are based on a perfect symmetrical concept. Whenever a popular vote is expressed in terms of two or more policies it is member of the set multiple issue and will receive a score 1. In contrast whenever a popular vote only engulfs a single policy issue it is a member of the set single issue and will received a score 0. The degree of elite cohesion/division over a popular vote is a fuzzy concept. The degree of elite division refers to an ordinal scale in which the division is calculated as a percentage of the seats that are opposing or supporting a referendum. In relation to each cantonal coalition party government, the inter-party division plays an important role in the determination of elite division. Indeed, the government can have an official position, but one or more political party members of the government can dissent and campaign against the government's official position. As a consequence, the government members could campaign on different fronts and give divergent messages to the electorate. Table B7 refers to the calibration of elite division in the case the elite support the referendum. The tables also consider whether the government is formed by a coalition or a single party. Table B8 refers to the calibration of elite division in case the elite opposes the referendum.   Government unified with radical right campaign the government position AND/OR ELITE, radical right party in the government intra-divided

0.4
Government internally split with only part of the right party opposing the government ∼ELITE official position

0.2
Government internally split with only the right party opposing the government official ∼ELITE position 0 Government internally split between right left and center coalition parties ∼ELITE The two tables calibrate the condition ELITE considering the outcome of CL failure, meaning that a positive value of score > 0.5 should contribute to the failure of the outcome CL. The degree of membership will be calculated considering ideological opposition. Indeed, the opposition from the radical left is less dangerous than opposition from the radical right (Papadopoulos, 1991): although electorally weak, the radical right is capable of determining some popular votes in its favor. Moreover, the greater the lack of cohesiveness in the right, the lower the level of support and the lower the passage rate of the government's proposals (Kriesi, 2006, p. 606).