Enfranchising non-citizens: What drives natives ’ willingness to share power?

is an


Introduction
International migration is on the rise, yet most democratic participation rights are still tied to traditional forms of citizenship.Hence, in many democratic countries a large and increasing share of the population has no or only limited franchise. 1An extreme case is Luxembourg, where 47 percent of residents were non-citizens in 2017 and, hence, excluded from full voting rights. 2In Germany and Switzerland, this share was 12 and 24 percent, respectively.The numbers differ across age groups.Among Swiss residents aged 30 to 40, the share of foreigners is about 40 percent.Consequently, their democratic endowment is comparable to the situation of women before their enfranchisement.
Participatory political institutions are known to generate positive effects.For instance, they foster cooperation (Acemoglu and Robinson, 2012), civic virtues (Frey, 1997), trust (Rainer and Siedler, 2009), and beneficial political outcomes (Ellis and Fender, 2010).While these findings refer to the potential of political integration through enfranchisement, the literature has mainly focused on the determinants of naturalization and successful economic integration of foreigners. 3We are only aware of three recent studies which empirically evaluated the conditions for non-citizens' enfranchisement (Earnest, 2015;Kayran and Erdilmen, 2020;Stutzer and Slotwinski, 2020) and mostly relied on cross-sectional variation.Given the limited democratic legitimacy of decisions taken by only a fraction of the taxpayers and the integrative effects of political empowerment (e.g., Koukal, 2013;Slotwinski et al., 2017;Koukal and Portmann, 2019), understanding the conditions under which political participation rights are transferred to non-citizens is important.
On the one hand, native citizens may benefit from enfranchising foreigners through integrative effects.On the other hand, they lose part of their political influence when sharing their political power with non-citizens.If the median voter positions of the new and established electorates differ, the majority of the established electorate may lose by enlarging the franchise, as political outcomes may change.With an increasing number of foreigners to be enfranchised, the potential costs and benefits of native citizens increase.A priori it is therefore unclear how ongoing immigration will affect the willingness to share political power and thereby the democratic endowment of the population.
In this paper we analyze the potential drivers of and barriers to non-citizens' enfranchisement in Switzerland during the 1992-2016 period and focus on the role of the size and composition of the foreign population.This paper is among the first to address this question and it is based on a much richer dataset than previous studies.We seek to contribute in at least two ways: First, this paper is one of the few empirical analyses of non-citizens' enfranchisement and the first to use an extensive panel dataset of natives' revealed preferences.Second, we contribute to a growing literature on the effects of immigrants' presence on natives' behavior by shedding light on how the presence of non-citizens affects the willingness of natives to share voting rights.
For two reasons, Switzerland provides an ideal setting for analyzing the enfranchisement process: First, the decision to enfranchise non-citizens is not taken by the national parliament but by the native electorate through popular referenda.This institutional setting allows an analysis of voters' revealed preferences.Second, due to Swiss federalism, non-citizens' enfranchisement is not a one-shot decision at the national level but a multilayered process at the municipal, cantonal, and federal levels.The extent of the franchise for cantonal and municipal matters is defined by the cantons, i.e., their respective voters.Therefore, the decision to hold a referendum vote is exogenous to the individual municipality.Thus, we exploit the within-municipality variations in the municipal acceptance to enfranchise non-citizens in 10 cantons over the 1992-2016 period.
Our results show that a larger share of foreigners in a muncipality decreases natives' willingness to enfranchise non-citizens.Our estimates suggest that a one percent increase in the share of foreigners reduces natives' willingness to share political power by approximately 0.20 percentage points.We argue that this effect is likely to reflect the costs of enfranchising non-citizens for two reasons: First, the negative effect of an increase in the share of foreigners on the enfranchisement of non-citizens is driven by the cultural distance between natives and foreigners.Cultural distance plausibly results in a more pronounced preference heterogeneity and, thus, a higher cost of sharing political power.Second, our results are also affected by institutional factors.The hindering effect of a larger foreigner share is more pronounced in municipalities with strong direct democratic instruments (town meetings)-where the individual political influence loss of natives is larger. 4Whether and how complementary explanations, such as discrimation against foreigners, play a role to explain these effects cannot be disentangled in our setting.Since many developed countries face an increasing share of foreign residents our results imply that natives' willingness to share political power may decline in the future.
This paper is organized as follows.The next section reviews the related literature.Section 3 introduces the Swiss institutional setting.Section 4 presents the theoretical considerations and Section 5 describes our data and variables.In Section 6 we explain the empirical strategy and the results are presented and discussed in Section 7. Section 8 provides a summary of our main results and an outlook on future research.

Related literature
Despite the sizeable literature on franchise extensions for men and women, we are only aware of three recent papers that empirically study the conditions of non-citizens' enfranchisement.In a cross-country study over the 1975-2010 period, Earnest (2015) analyzes the conditions for the liberalization of citizenship laws, such as non-citizens' voting rights and retraction of voting rights of non-citizen residents.The author explains the former with relation to policy constraints and the latter as an interaction between policy constraints and national characteristics.In a recent cross-country comparison of 28 countries, Kayran and Erdilmen (2020) find that governments tend to (partially) delay the enfranchisement of non-citizens when their share in the total population is high.Furthermore, Stutzer and Slotwinski (2020) use data from two Swiss cantons to focus on the power dilution hypothesis for opting-in regimes to enfranchise non-citizens. 5 Unlike the literature on non-citizens' enfranchisement, the literature on franchise extension for native men and women is well established.Acemoglu andRobinson (2000, 2001) prominently frame a theory of democratization that explains suffrage extensions as a strategic decision by the elite to prevent revolution or social unrest. 6Wars (e.g., Hicks, 2013;Polishchuk and Syunyaev, 2015) and the strategic concerns of a divided elite (e.g., Lizzeri and Persico, 2004;Llavador and Oxoby, 2005) have also been considered as potential drivers of suffrage extensions.Bertocchi (2011) finds that a smaller gender wage gap increases the likelihood of female enfranchisement across Europe.Engerman and Sokoloff (2005) empirically underline this result with data from North and South America, providing evidence that greater homogeneity (in terms of socioeconomic or ethnic attributes) drives democratization.Democratization via enfranchising more natives has also been explained by various price effects.There exists evidence that the willingness of men to enfranchise women increases in their scarcity (e.g., Kenny, 1998;Braun and Kvasnicka, 2013) and decreases with the growing political influence of the actual (male) electorate (Koukal and Eichenberger, 2017).The impact of preference heterogeneity between the old and new electorate on suffrage extensions is less well explored.The few contributions about non-citizens of which we are aware find for immigrants, compared to natives, lower status quo effects (Koukal, 2013), stronger preferences for increasing social services expenditures (Vernby, 2013), and a tendency to vote for left parties (Strijbis, 2014).Moreover, they indicate ethnicity to be a main driver of party choice (Tillie, 1998).
With growing international mobility, the interest in measuring the consequences of growing ethnic diversity is increasing (Hainmueller and Hopkins, 2014).There are two prominent opposing theories on the effect of outgroup size on ingroup attitudes toward the outgroup.On the one hand, the cultural threat hypothesis suggests that natives fear immigrants because they pose a threat to their cultural identity (Hainmueller and Hiscox, 2007;Kinder and Kam, 2009).Following this approach, the size of the foreign population has been shown to positively affect the support of far-right parties (Halla et al., 2017;Brunner and Kuhn, 2018;Edo et al., 2019) and to diminish the willingness to redistribute (Luttmer, 2001;Alesina et al., 2019;Tabellini, 2020).Furthermore, restrictive naturalization rules are more prevalent within a growing foreign population (Bertocchi et al., 2010;Mariani, 2013).On the other hand, contact theory suggests that interactions between outgroup and ingroup members reduce information asymmetries, increase trust, reduce prejudices, and may moderate perceived threat (Allport et al., 1954;Yehuda, 1998;Paluck et al., 2019).Using French data, Jolly and DiGiusto (2014) find that xenophobic attitudes decrease with a growing foreign population size, while for the Netherlands, Schlueter and Scheepers (2010) provide evidence for both threat and contact theory depending on the considered measures for the size of the foreign population.Semyonov et al. (2004) show that in Germany, perceived group size drives anti-immigrant attitudes, while actual intergroup contact reduces perceived group threat.
Besides the size of the foreign population, the cultural background and institutional context also matter in the naturalization process.Hainmueller and Hangartner (2013) provide evidence for the country of origin of the applicant being the most important determinant of naturalization decisions in Swiss municipalities.Moreover, once politicians rather than citizens decide on the naturalization applications, naturalization rates increased by about 60 percent (Hainmueller and Hangartner, 2019).Therefore, direct democracy might constitute a significant barrier to the broader integration of non-citizen residents in politics, either because it raises the price of political power-sharing or because it fosters the discrimination of minorities (Gamble, 1997;Donovan and Bowler, 1998;Haider-Markel et al., 2007;Koukal and Eichenberger, 2017;Hainmueller and Hangartner, 2019).
Beyond the impact of foreigners' population size, economic circumstances have also been shown to impact attitudes toward foreigners.In a cross-country study covering 12 countries, Quillian (1995) shows that anti-immigrant attitudes are more prevalent if economic conditions are worse.Most authors analyzing the economic threat hypothesis concentrate on the labor market situation.While some authors find evidence that anti-immigrant attitudes are positively affected by greater competition in the labor market (Scheve and Slaughter, 2001;Mayda, 2006;Hyll and Schneider, 2018), others question this result (Hainmueller and Hiscox, 2007;Sides and Citrin, 2007).

Institutional background
Across Europe, the enfranchisement of non-citizen residents has increased since the late 1970s. 7In most countries, the national parliament decides on the enfranchisement of non-citizens.However, such a setting provides only limited information about the preferences of the actual electorate.In Switzerland, non-citizens' suffrage cannot be implemented by the national or cantonal parliaments, but only by the actual electorate through referenda votes.At the federal level, non-citizens' voting rights for federal matters have never been subject to a vote.In contrast, at the cantonal level, non-citizens' voting rights for cantonal and municipal matters have been subject to approximately 40 votes in 16 of the 26 cantons.The cantonal votes on non-citizens' enfranchisement result from different procedures, which are listed in Table A1 for the votes in our dataset.As voting rights are regulated at the constitutional level, the mechanisms are the following: (i) Citizens can start a popular initiative, i.e., they formulate a constitutional amendment and have to collect a certain number of supportive signatures demanding a referendum vote on the respective proposal.The proposal is implemented if it gets a simple majority of the votes.(ii) Cantonal parliaments can design a constitutional amendment.However, in all cantons, constitutional amendments are subject to a mandatory referendum.The amendment is only implemented if it gets a simple majority of the votes.(iii) Votes on non-citizens' enfranchisement sometimes result from total or partial revisions of the cantonal constitutions, which also need the consent of the citizens.In all these cases, the decision to hold a referenda vote is taken at the cantonal level and is, thus, exogenous to individual municipalities, which are our units of observation.
Due to cantonal autonomy, various types of voting rights for non-citizens have been considered and partly installed.Table A1 in the Appendix provides an overview of the different types of enfranchisement that have been voted on in the cantons of our dataset.They range from active and passive voting rights at the cantonal level to optional voting rights at the municipal level, i.e., to allow municipalities to enfranchise non-citizens at the municipal level.Active voting rights provide them with the right to take part in the political process as voters, while passive voting rights allow them to run for office.Table 1 provides an overview of the cantons that introduced non-citizens' suffrage.Analogously to the enfranchisement of women, French-speaking cantons were the first to make the move to enfranchise non-citizens.Currently, the cantons of Neuchâtel and Jura grant non-citizens the most extensive political rights, i. e., active voting rights at the cantonal level and active and passive voting rights at the municipal level.However, cantons also differ with respect to the conditions under which non-citizens receive voting rights, most importantly their duration of stay. 8o enfranchise non-citizens in a canton, at least 50 percent of the participating voters at the cantonal level must agree.Table 1 provides an overview of accepted referenda votes and the corresponding yes shares.The cantonal decisions are imposed on municipalities where only a minority of voters agrees to enfranchise non-citizens (i.e., non-citizens are enfranchised at the municipal level against the will of the majority of municipal voters).In some cantons, an opt-in rule is used that delegates the right to enfranchise noncitizens at the municipal level to the municipalities (indicated as opt-in YES in Table 1).Currently, three cantons (Grison, Appenzell Ausserrhoden, and Basel City) have introduced opt-in rules for municipalities.Furthermore, some votes on enfranchising non-citizens are integrated into general constitutional revisions and are, therefore, part of a larger political package.9

Theoretical considerations
In the absence of non-citizens' suffrage, the electorate consists of Swiss voters only.They decide on politics, either via direct democratic institutions or by delegating their decision-making power to politicians.Extending the group of voters by enfranchising non-citizens potentially generates both costs and benefits for the actual electorate.In the following section, we discuss how these benefits and costs might evolve with a growing share of foreigners.
Enlarging the electorate has the potential of several benefits.Involving more and different people in the decision-making process increases the amount and quality of information on political issues, the legitimacy of political decisions, and the media's incentive to cover political topics. 10According to the Condorcet jury theorem, the quality of democratic decisions under uncertainty improves with an increase in the number of voters if their individual errors are independently distributed (i.e., if the heterogeneity of voters increases).11In addition, the political integration of non-citizens has been shown to have broader integrative effects as well (e.g., Koukal, 2013;Slotwinski et al., 2017;Koukal and Portmann, 2019).This is in line with the literature analyzing the role of participatory political institutions in fostering cooperation (Acemoglu and Robinson, 2012), civic virtue (Frey, 1997), or trust (Rainer and Siedler, 2009).A growing share of foreign residents might affect the extent to which democratic institutions unfold their positive effects.Therefore, the need for political inclusion and resulting benefits are likely to increase if the share of foreign residents grows.These considerations suggest that natives' willingness to enfranchise non-citizens increases with the share of foreigners.
Conversely, enlarging the electorate may also impose costs on the actual electorate.First, natives' individual influence on political outcomes decreases with a larger electorate, as the probability of a vote affecting the outcomes decreases.12These costs increase in preference heterogeneity between the native and foreign populations and depend on the institutional setting.Previous literature Note: In the Canton of Appenzell Ausserrhoden, the cantonal assembly has voted on non-citizens' suffrage, and thus no data on the municipal level is available.
Sources: Adler et al. ( 2016), cantonal chancelleries, cantonal constitutions.(Koukal, 2013;Vernby, 2013;Strijbis, 2014) found evidence that the preferences of non-citizens and natives differ, and, hence, non-citizens' suffrage is likely to move the median voter and change political outcomes.Regarding the enfranchisement of women, Kenny (1998) and Braun and Kvasnicka (2013) find that the scarcity of women, and, thus, their relative small weight in democratic decisions, has a positive effect on female enfranchisement.In other words, the larger the increase in the size of the electorate, the larger the power loss of the current electorate.This effect is likely to be more pronounced if the institutional context grants the actual electorate more political influence, for instance, with more effective direct democratic institutions (Koukal and Eichenberger, 2017).In addition, the literature also suggests that hostility against the outgroup increases when the size of the foreign population grows, as this threatens the native population in various dimensions, such as their cultural identity or their social and economic privileges (Halla et al., 2017;Brunner and Kuhn, 2018;Edo et al., 2019). 13Therefore, individual power loss and perceived threat suggest that natives' willingness to enfranchise non-citizens decreases with the share of foreigners.These costs are likely to be more pronounced if the preferences of the native and the foreign populations differ substantially from each other.Ex-ante, the overall effect of extending the electorate to non-citizens may be positive or negative for the actual electorate, as this depends on the relative sizes of costs and benefits.In the following sections, we analyze how the presence and composition of noncitizens impact the willingness of natives to enfranchise non-citizens in different institutional settings.

Data and variables
Our empirical analysis relies on three data sources.(1) We collected and digitized data from 33 cantonal referenda on suffrage extension between 1992 and 2016.An overview of the referenda in our dataset is provided in Table A1 in the Appendix.(2) We combined this information with a variety of sociodemographic municipal characteristics acquired from the Swiss Statistical Office and the Federal Tax Administration.(3) For information on institutional municipal characteristics, we make use of the municipal survey data provided by Andreas Ladner.This results in a novel dataset with approximately 3200 observations.We can exploit the withinmunicipality variation of 24 referenda stemming from 10 cantons that voted at least twice on non-citizens enfranchisement. 14able 2 provides the descriptive statistics of the outcome and the explanatory and control variables, including all vote types.
An empirical panel analysis at the municipality-level exhibits several advantages.It allows the examination of the effect of the foreigners' share, characteristics of the foreigners, municipal institutional features, and economic conditions on the approval of noncitizens' voting rights.Compared to cross-country data, municipal panel data enables the analysis of a richer variation in a more homogenous context.Furthermore, the decision to conduct a referendum vote is exogenous to the individual municipality, as the requirement to launch such a vote is decided at the cantonal level.Like in many other OECD countries, in Switzerland, some municipalities have merged in our period of observation (Ladner, 2011;Steiner and Kaiser, 2017).As statistical offices provide most statistical information only for the newly merged unit but not for each former subunit, the construction of a decent dataset covering a sufficient number of variables for the municipalities that merged later was not possible.Therefore, we excluded all municipalities that experienced an amalgamation in our period of observation, allowing us to include a larger variety of control variables and to work with Note: These statistics include observations from our main sample.Votes embedded in partial or total revisions and singleton municipality observations are excluded.The number of observations is smaller for the variable parliament since this variable is not observed for all municipalities.
a balanced panel.

Dependent variable
Our outcome variable yes share captures the share of votes in favor of enfranchising non-citizens in a municipality.The outcome at the municipal level is observed in the cantonal referenda votes on suffrage extensions in the 1992-2016 period.Fig. A1 in the Appendix illustrates the within-municipal variation between the first and last votes in our panel and suggests that the willingness to enfranchise non-citizens is not characterized by a positive time-trend.

Explanatory variables
Our main variable of interest is foreigner (measured as a share); it approximates the size of the affected group that is to be enfranchised in a municipality.Fig. A2 in the Appendix provides an overview of the variation of the share of foreigners within a municipality between the first and last votes in our sample.Switzerland has been an immigration destination for decades.The vast majority (more than 80 percent) of immigrants originate from European countries.While the period between the 1960s and 1970s was characterized by a wave of immigrants from Southern Europe (Italy, Portugal, and Spain), people from (former) Yugoslavia 15 constitute the largest new immigration group in our period of observation, accounting for around 24 percent of the foreign population in Switzerland in 2000 (Federal Statistical Office, 2020).

Measures of cultural distance
Cultural distance-which we understand as a proxy of preference heterogeneity between the native and the foreign population-can be measured by various approaches that vary widely between disciplines (Beugelsdijk et al., 2019).To meet this challenge, we follow two approaches that are well-established in the economic literature and compare the respective results.First, we apply the approach by Inglehart and Baker (2000) which is based on data from the Word Values Survey (WVS).Second, we follow authors who emphasize the role of common language and geographical proximity as an important indicator of common cultural traits (Guiso et al., 2009;Jasso, 2009;Bisin and Verdier, 2011).Inglehart and Baker (2000) locate 65 countries on two central dimensions of cross-cultural variation, which explain around 70 percent of the variation in the applied WVS items. 16The first dimension "survival vs. self-expression" captures the importance of economic and physical integrity relative to personal self-fulfillment and self-expression.The second dimension "traditional vs. secular-rational" considers the relative importance of traditional values that concentrate on the importance of family, religion, or respect for authorities in comparison with secular values emphasizing beliefs such as tolerance of human diversity or gender equality.We use the results of Inglehart and Baker (2000) to classify the countries of origin in our sample as culturally similar or distant to Switzerland.In a next step we use this information to create an indicator of cultural distance between natives and non-citizen residents on the Swiss municipal level.The variable culturally distant captures the share of foreigners which are classified as cultural distant to Switzerland among the total foreign population.The group of culturally similar countries contains countries with a value system similar to the Swiss one with respect to the two dimensions explained above (survival vs. self-expression and traditional vs. secular-rational), such as historically Catholic and Protestant countries, or English-speaking OECD states.A list of countries classified as culturally similar, or culturally distant respectively, is provided in Table A2 in the Appendix.To further ensure the reliability of our classification we compared it with the composite bilateral measure of cultural distance by Kaasa et al. (2016).It strongly supports our categorization of European countries. 17In a second step, we keep following Inglehart and Baker's (2000) classification but focus on Switzerland's wave of immigrants from former Yugoslavia, which forms the largest immigration group in our period of observation.Moreover, immigrants from former Yugoslavia got major attention in the Swiss immigration debate (Hainmueller and Hangartner, 2013).The share of non-citizens from former Yugoslavia among the total foreign population is captured in the variable former Yugoslavia.
Our second approach, which emphasizes the importance of geographical and linguistic proximity, is implemented by categorizing immigrants from neighboring countries that share a common language with the Swiss population (Austria, France, Germany, Italy, and Liechtenstein) as culturally similar and foreigners from all other countries as culturally (more) distant.The variable not neighbor indicates the share of non-citizens among the total foreign population in a municipality who are not originally from a neighboring state of Switzerland.Table A2 in the Appendix lists the states classified as culturally distant and culturally similar in the three measures.

Measure of institutional price differences
Following Koukal and Eichenberger (2017), we hypothesize that more representative (instead of direct democratic) instruments at the municipal level can-under specific conditions-foster the enfranchisement of non-citizens.The intuition behind this approach is that the existing electorate is more willing to share their political power with a new group of voters if their individual influence, and, thus, their influence loss, is smaller.One way of measuring the strength of representative democracy on the local level is by using the municipal legislative institutions, which either consist of a town meeting (direct democracy) or a parliament (representative democracy). 18In direct democratic municipalities, voters meet directly (one to four times a year) and discuss and decide on municipality topics themselves.In contrast, in municipalities with a parliament, voters delegate a part of their political power to politicians.We construct the dummy variable parliament, which is equal to zero when a municipality has a town meeting (direct democracy) and equal to one if the municipality has a parliament (representative democracy).We rely on survey data by Ladner to classify the municipal legislative institution.Note that changes in the municipal legislative institutions are possible over time but are rather uncommon.We identify institutional switchers in our sample based on information provided by Funk and Litschig (2020) and the survey data by Ladner. 19By excluding the switchers from our analysis, we treat the variable parliament as being fully time-invariant.

Control variables
We introduce a set of control variables that cover a broad range of economic and non-economic factors that might shape native's willingness to enfranchise non-citizens.To account for the different sizes of municipalities, we introduce the variable population that constitutes the natural logarithm of the population in a municipality.A channel that has been mentioned in the literature on how natives perceive immigrants is political ideology (Mayda, 2006;Knoll et al., 2011).The variable Social Democratic Party reflects the vote share of the largest left party in Switzerland in the most recent national elections and aims to proxy for prevalence of left leaning voters in a municipality.To further control for the prevalence of (conservative) attitudes in the population, we introduce the variable pensioner that measures the share of the municipal population above 64 years and captures age-related preferences.Furthermore, we generate the variable agriculture as the number of farms relative to the population in a municipality.This variable captures the prevalence of traditional and conservative norms, the economic structure of a municipality and the location on the rural-urban continuum, which has shown to be an important determinant of voting behavior (Garcia and Davidson, 2013;Scala and Johnson, 2017).To control for the general openness toward foreigners in a municipality we include naturalization, which is constructed as the number of naturalizations relative to the resident population in the year of observation.The information for the former four variables is provided by the Swiss Federal Statistical Office.We would have liked to further analyze how the stock of naturalized individuals in a municipality affects the acceptance to enfranchise non-citizens.However, this information is not available on the municipality level.An important economic factor that has shown to influence native attitudes towards foreigners are labor market concerns (Scheve and Slaughter, 2001;Mayda, 2006;Hainmueller and Hiscox, 2007;Hainmueller and Hopkins, 2014).To consider this prominently discussed channel, we introduce the variables unemployment foreign and unemployment Swiss, which reflect the number of unemployed individuals relative to the foreign or domestic population in our set of control variables. 20Besides economic competition on the labor market, also the fiscal burden has shown to impact native attitudes towards foreigners (Hanson et al., 2007).To account for this factor, we proxy the financial situation within the municipality with data from the Federal Tax Administration and include the variable mean income (measured in logs), which reflects the mean income of natural persons in a municipality.Finally, we also introduce the variable Gini coefficient to account for the prevalence of inequality in a municipality.There are further variables we would have liked to include in our analysis, such as information about the educational attainment, income, crime rate or religious affiliation, specifically for the native and the foreign population.Unfortunately, information on these variables is not available at the municipality level, which limits our analysis.

Empirical strategy
To gain a precise understanding on how the share of foreigners and other factors impact the willingness to enfranchise non-citizens, we need to account for the complex and multifactorial setting.Therefore, a cross-sectional analysis is not suitable to answer our research question.Due to our rich panel dataset, we choose a model with municipality fixed effects as our preferred option.Hence, we estimate the following model: where X mt denotes a vector of controls and δ, γ, and ϕ are fixed effects.By applying municipality fixed effects δ in our base model, we take into account the omitted variable biases stemming from the municipality level and control for time-invariant municipality 18 For simplification, we denote municipalities with a local parliament as representative democratic.Strictly speaking, also municipalities with a local parliament have access to direct democratic means on the cantonal and national levels. 19The municipal surveys of Ladner are available at http://www.andreasladner.ch/uebersicht.htm.In most cantons, municipalities can choose their legislative form (parliament or town meeting).Population thresholds exist for municipalities in the cantons of Vaud, Fribourg, Valais, and Zurich (Funk and Litschig, 2020).We include these cantons in our fixed-effect analysis, as the number of municipalities around these cut-offs is low, and the assignment rule is not stringent (i.e., mandating a town meeting below the cut-off as well as allowing choice above).Only the canton of Vaud uses a sharp assignment rule and had a regulation change during our period of observation.Thus, in Table A7 we perform a robustness check excluding the canton of Vaud from the analysis. 20It would be optimal to calculate the unemployment rate with the number of unemployed individuals relative to the working population.Unfortunately, the number of the working population at the municipality level is in many cases associated with high uncertainty (data from SECO).Hence, we decided to divide the number of unemployed by the total number of foreign and native residents in a municipality based on information from the Federal Statistical Office.
A.M. Koukal et al. characteristics, such as municipal institutions, general openness toward foreigners, culture, or geographical location.By including referenda fixed effects γ , we account for the different types of voting rights that have been debated and absorb cantonal specific time effects.Additionally, by adding time fixed effects ϕ , we control for general time effects across cantons.
Even after including municipality, time, and referenda fixed effects, a consistent estimation of β 1 is only possible under the assumption that the yes share is uncorrelated with the error term.As discussed in Section 4, it is not clear ex ante if we expect β 1 < 0 (costs exceeding benefits) or β 1 > 0 (benefits exceeding costs).As is always true, our estimate of β 1 can be biased due to causality and simultaneity issues.The most plausible argument is that not only natives' preferences react to high foreigner shares (our thesis), but also foreigners react to the natives' preferences, that is, they migrate to municipalities where they are welcome, which is then also mirrored in natives' support for enfranchising foreigners.However, if β 1 < 0, this chain of reactions would not inflate but deflate our estimates, which would make them rather conservative and immune to being falsely interpreted as being statistically significant.
Nevertheless, we try to control for potential endogeneity by an instrumental variable approach.One possible instrument is the 'shift-share' methodology which has been extensively used in the field of labor economics (Card, 2001;Peri, 2012).However, the instrument has also been criticized (Jaeger et al., 2018;Goldsmith-Pinkham et al., 2020) and in the Swiss setting, shift-share instruments have, to the best of our knowledge, only been used to instrument the share of foreigners at the regional levels (local labor markets or cantons). 21Therefore, we follow Brunner and Kuhn (2018) and instrument the share of foreigners at the municipal level with the foreigner share at the MS-regional level (MS is the abbreviation of "mobilité-spatiale" or spatial mobility). 22The boundaries of the 106 MS regions are determined by the Federal Statistical Office and follow the idea of small labor market areas, as the regions are classified by the level of local economic activities.Fig. A3 provides an overview of the distribution of MS regions. 23In order for the MS foreigner share to be a valid instrument, two identifying assumptions must be fulfilled.First, the foreigner share in an MS region must be determined by factors other than natives' attitudes toward foreigners.A possible argumentation in favor of this assumption is that immigrants settle in a specific region due to factors such as closeness to their family or employment possibilities.Second, the exclusion restriction requires that the share of foreigners in the broader region have no direct effect on the yes share to enfranchise non-citizens in a municipality.The second assumption may not be fulfilled if, for instance, an individual is working in another municipality in the same MS region.In this case, the attitudes toward foreigners may not merely be driven by the share of foreigners in their living municipality but also by the share of foreigners in the workplace municipality.
Besides our base model (1), we estimate models ( 2) and ( 3) to analyze potential heterogenous effects of foreigner on natives' willingness to enfranchise non-citizens.We investigate the interaction effect between the share of foreigners and the degree of cultural distance between natives and foreigners (2), as well as the interaction effect of the share of foreigners and the institutional structure, measured by the variable parliament (3).We hypothesize that (among others) the effect of foreigner on natives' willingness to enfranchise non-citizens depends on the extent of preference heterogeneity between the two groups.For proxying the extent of preference heterogeneity between the foreign and the native population, we rely on three different measures: culturally distant, former Yugoslavia, and not neighbor.We estimate the following model: where X mt denotes again a vector of controls and δ, γ, and ϕ are fixed effects.If preference heterogeneity, proxied by the share of culturally distant non-citizens, is a moderating factor, we expect β 3 to be negative.As mentioned in Section 5, information on other relevant characteristics of the foreign population-such as education, income, crime rates, or language proficiency-is not available on the municipality level.Therefore, we focus on the measures of cultural distance on the municipality level.Following Koukal and Eichenberger (2017), we further hypothesize that representative democracy, when compared to direct democracy can-under specific conditions-lower the price to enfranchise non-citizens, as the individual power loss is less pronounced.We thus estimate the following model: where X mt denotes analogous to Model (1) and (2) a vector of controls and δ, γ, and ϕ are fixed effects.We expect the boosting effect of representative democracy (compared to direct democracy) to grow in the share of foreigners and therefore expect β 3 to be positive.

Results and discussion
This section presents and discusses the results of our main estimations and robustness checks.

Size of the foreign population
Table 3 reports the OLS estimates of the base model (1) with a listwise introduction of control variables.Referenda votes that were part of constitutional revisions and, thus, embedded in a broader political package are excluded in Table 3.All specifications of Table 3 display a negative coefficient of foreigner, which remains statistically significant at the 1 percent level and robust in terms of size.As we estimate a model with municipality-fixed effects, β 1 captures the within-municipality variation of the share of foreigners.Our estimates suggest that a one percent increase of foreigners in one's municipality is associated with a decreased willingness to share political power by approximately 0.20 percentage points. 24he results presented in Table 3 suggest that, overall, the benefits for citizens to enfranchise foreigners seem to be overcompensated by the respective costs, for instance, the loss of political power (Braun and Kvasnicka, 2013).Another possible interpretation for β 1 < 0 is that increasing foreigner shares induce the feeling of threat among natives, which would be consistent with a recent literature on the role of migration for the support for the far right (Halla et al., 2017;Brunner and Kuhn, 2018;Edo et al., 2019).Unfortunately, we cannot disentangle the threat hypothesis from other potential explanations.However, in the further analysis we try to gain a better understanding of the conditions under which the negative relationship of granting non-citizens the right to vote and the share of foreigners is more pronounced.
In a variety of robustness exercises the negative coefficient of foreigner stays robust in terms of size and significance.To address concerns of endogeneity, we present IV estimates as discussed in Section 6.Table A3 presents an overview of the first and second stage results of our IV estimations, which confirm the negative association of foreigner share and natives' willingness to enfranchise noncitizens, however the coefficient is larger.As a next step we estimate our model using the full sample including all referenda, also those related to partial or total constitutional revisions (Table A4, Spec.1).
As many Swiss municipalities are small, we control for the role of municipality size in the following ways: First, we estimate our model with a non-linear effect of the absolute size of population (A4, Spec.2).Second, we exclude the five percent smallest municipalities in our sample (A4, Spec.3), and third, we additionally exclude the five percent largest municipalities from our sample (A4, Spec.4).To consider the different scopes of referenda with respect to the demanded franchise, we perform an analysis with a reduced sample of only the quasi-identical referenda from the canton of Zurich (A4, Spec.5).As we observe different cantons and regions in our sample, we want to rule out that our coefficient of interest is driven by specific regional time trends.Therefore, in Table A5, we provide robustness checks where we account for differential linear time trends among language regions (A5, Spec.1), MS-regions (A5, Spec.2), second observation of a municipality may vary.
A.M. Koukal et al. and cantons (A5, Spec.3).Again, the coefficient of foreigner stays robust to this exercise.Throughout all robustness checks, the effect of the foreign population on the yes share remains similar both in terms of size and statistical significance-only in the IV estimates the coefficient is somewhat larger when compared to the OLS estimates.Although these estimations provide stable results, we cannot fully rule out that unobserved time variant factors are affecting our results.Similar to the recent literature that analyzed the enfranchisement of non-citizens across countries (Kayran and Erdilmen, 2020) or by opt-in rules (Stutzer and Slotwinski, 2020), our estimations provide robust and convincing evidence from a large municipality-level panel that an increasing share of foreigners in a municipality decreases the willingness of natives to share political power with them.Fig. A4 in the Appendix depicts the development of the share of foreigners in the cantons of our sample and for the whole of Switzerland over the 1980-2018 period.During our observed period, the share of foreign residents in Switzerland increased by seven percentage points, from 17.43 to 24.60 percent.Thus, for a municipality that faced an average increase of foreign residents, the estimates of our base model point to a decrease in the willingness to share power by 1.40 percentage points over the 1992-2016 period. 25his effect is considerable, as the mean yes share is 29.33 percent and some of the referenda ended in close decisions (see Table A1).

Composition of the foreign population
So far, we have only considered the average effect of the foreigners' share on natives' willingness to enfranchise non-citizens.In the following section, we analyze potential heterogenous effects to gain a better understanding of the results provided in Table 3.Following the preference and the cultural-threat hypotheses, larger cultural distance (and, hence, larger preference heterogeneity) between the foreign and the native population could lead to a decreasing willingness to enfranchise non-citizens.In Section 5, we introduced three measures of cultural distance between the native and the foreign population.The measures account for the share of non-citizens among the total foreign population in a municipality that is (i) culturally distant based on Inglehart and Baker (2000), (ii) originally from former Yugoslavia, and (iii) not migrating from a neighboring country of Switzerland.We hypothesize that the negative effect of foreigner on natives' willingness to enfranchise foreigners is more pronounced if cultural distance is larger and interact the share of culturally distant foreigners with the total share of foreigners.In Table 4, the moderator variables are the shares of culturally distant foreigners (three measures) among the total foreign population in a municipality, where the share of foreigners is our main predictor.
Note that the base effect of foreigner is not negative and statistically significant in all specifications in Table 4, since it depicts the effect of foreigner in a municipality without culturally distant foreigners.The same accounts for the base effects of the three measures of cultural distance, which yield positive estimates but relate to municipalities without foreigners (which only applies for a neglectable number of municipalities).Our coefficient of interest is β 3 , which reflects the interaction of the share of residents from culturally more distant countries with the total share of foreigners.β 3 is negative and significant throughout all specifications and all measures in Table 4.In terms of size, β 3 in Table 4 varies between − 0.003 and − 0.007, which seems small.However, considering the average increase of non-citizen share in Switzerland of seven percent in our period of observation (see Fig. A4), a one percentage point increase of culturally distant would account for − 0.028 (7 * − 0.004) percentage points of the yes share.The sample mean of culturally distant is 28.91 percent, hence cultural distance to the native population seems to be an important channel to explain the conditions under which the presences of immigrants negatively shapes the natives' willingness to enfranchise non-citizens.
Fig. 1 provides a graphical illustration of this interaction term.The slopes of the yes share are computed as a reaction to the varying foreigner share while holding the value of the moderator variable (culturally distant) constant at values running from 0 to 100 percent.The interaction of culturally distant * foreigners is represented by the slopes, which are consistently steeper for municipalities with a larger share of culturally distant foreigners (Fig. 1).
As Fig. 1 illustrates, the cultural composition of the immigrant population seems to be an important moderator on how the share of foreigners affects natives' willingness to enfranchise non-citizens.To further test the reliability of this finding, Table A6 in the Appendix yields the results of various robustness checks.We perform the estimations with the full sample including all referenda, also those related to partial or total constitutional revisions (Table A6, Spec. 1 to 3), with a non-linear effect of the absolute size of population (A6, Spec. 4 to 6), and exclude population outliers (A6, Spec.7 to 9).The results remain robust in most specifications and provide convincing evidence that with an increasing share of culturally distant foreigners, the negative effect of the share of foreigners is more pronounced.There are at least three explanations for this result, which cannot be disentangled in this paper: First, the hindering effect of cultural distance may be a consequence of larger preference heterogeneity between the native and the (potential) foreign electorate.Moreover, the negative coefficient of the not neighbor * foreigner interaction may also point to the importance of common languages and geographical proximities in fostering trust, thereby accelerating the speed of integration (Guiso et al., 2009;Bisin and Verdier, 2011).Second, β 3 < 0 may also be explained by the role of cultural threat (Hainmueller and Hiscox, 2007;Kinder and Kam, 2009).Third, we cannot rule out that β 3 < 0 also captures discrimination against culturally distant foreigners (Hainmueller and Hangartner, 2013).

Institutional context
As elaborated in Section 6, we estimate Model (3) to account for the strength of representative democracy on the municipal level.We expect a higher willingness to enfranchise non-citizens in municipalities with a local parliament (as compared to a town meeting), as we assume the costs of political power sharing to be lower.Moreover, we assume this effect to grow in the share of foreigners as the mentioned cost mechanism is more pronounced with a growing new electorate.If costs of sharing political power are lower in representative (parliament) than direct democracy (town meeting), we expect parliament * foreigner (β 3 ) to be positive.
As the municipal legislative institution is time-invariant, it is not possible to display the base effect of parliament when applying the municipality fixed effects in Table 5, Specifications (3) and (4).We start with a lean model considering cantonal fixed effects in Specifications (1) and (2) of Table 5.The introduction of a dummy for parliament in the cross-sectional analysis in Specification (1) does not display a significant effect on the willingness to enfranchise non-citizens.From Specification (2) onward, we introduce the interaction of the municipal legislative with the foreign share.When applying cantonal fixed effects in Specification (2), β 3 is positive but not statistically significant.A graphical illustration of Specification (2) can be found in Fig. A5 in the Appendix.However, when including municipality-fixed effects and further control variables in Specification (4), β 3 becomes positive and significant at the five percent level.
This result suggests that the share of foreign residents exhibits a heterogenous effect on the willingness to share power with respect to the strength of representative democracy.An increase in the foreign share by one percentage point in municipalities with a parliament is associated with a 0.23 higher yes share when compared to municipalities with a town meeting.Note that this effect stays positive and significant in most robustness exercises; these are presented in Table A7 in the Appendix.Robustness tests in Table A7 include estimates for the whole sample (including total revisions), samples excluding outliers based on population size, a sample excluding the canton of Vaud (as they apply sharp population threshold for the municipal legislative), a sample only of the canton of Zurich, and an estimation with cantonal time trends.The boosting effect of representative democracy, when compared to direct democracy, can be explained by two mechanisms: first, by the lower price of sharing political influence in a representative democratic setting as compared to a more direct democratic environment (Koukal and Eichenberger, 2017) and second, by the controversially debated tendency of direct democracy to be harmful to minorities (Gamble, 1997;Donovan and Bowler, 1998;Haider-Markel et al., 2007;Koukal and Eichenberger, 2017;Hainmueller and Hangartner, 2019).Unfortunately, the data structure does not allow to further disentangle these two potential mechanisms but underline the importance of local institutions in political power sharing.

Further discussion
This section discusses interesting results obtained for our control variables presented in Table 3.With the growth in municipality size, a citizen's vote is less likely to be decisive.Following the cost argument, population growth can translate into a higher willingness to share political rights with non-citizens in larger municipalities.However, when we look at the estimates in Table 3, we do not find evidence that population size has a significant effect on the approval of non-citizens' suffrage.Another factor which has gained a lot of attention is the role of labor market competition.From Specification (2) onward, we introduce the municipal unemployment rate of the  3) and ( 4) due to the introduction of municipality fixed effects, which drops singleton municipality observations.*** p < 0.01, ** p < 0.05, * p < 0.1.
A.M. Koukal et al. native and foreign populations as our proxy for economic competition.Throughout all specifications, the unemployment rate of natives exhibits a statistically significant negative effect on the citizens' willingness to enfranchise non-citizens.A one percent increase in the unemployment rate of natives is associated with a decrease in the willingness to enfranchise non-citizens by approximately one percentage point.This is in line with the related literature which explains anti-immigrant attitudes with increased labor market competition and economic downturn (Scheve and Slaughter, 2001;Mayda, 2006;Hainmueller and Hiscox, 2007;Hainmueller and Hopkins, 2014).The negative coefficient of unemployment Swiss hints to the explanation of economic threat, which may hinder the political power sharing with non-citizens.In contrast, the unemployment of foreigners in one's municipality does not seem to impact the willingness of natives to enfranchise non-citizens.Furthermore, in Table 3, from Specification 3 onward, we control for the vote shares of the Social Democratic Party, the most important left-wing party.As expected, the coefficient of the share of Social Democratic Party voters exhibits a positive and significant coefficient throughout all specifications in Table 3.Our estimates suggest that a one percent increase of the voters of the Social Democratic Party in a municipality increases the approval of non-citizens' suffrage by approximately 0.30 percentage points.There are multiple explanations for this relation.Given that foreigners have shown to have preferences for more leftist political positions (Vernby, 2013;Strijbis, 2014), the enfranchisement decision may be driven by a strategic calculus to shift the median voter to an income bracket that supports a political position (Meltzer and Richard, 1981).However, stronger support for the Social Democratic Party may also proxy for more leftist policy preferences within the native population, which may include the political integration of non-citizens.
In Table 3, the pensioners' share exhibits a robust negative and significant effect on the willingness to share political rights with non-citizens.Our estimates suggest that a one percent increase of pensioners in a municipality reduces the willingness to share political power with non-citizens by approximately 0.22 percentage points.A simple explanation for this observation constitutes the more conservative preferences of the elderly population.Furthermore, it can also point to the role of preference heterogeneity: If preference differences between elderly natives and foreigners (who are, on average, younger) are larger, enfranchising non-citizens will result in higher costs for elderly natives than for young natives.

Conclusion and outlook
In several developed countries, foreigners without political participation rights represent a large and growing percentage of the population.Therefore, the political integration of non-citizens is a major political challenge.Given the limited democratic legitimacy of decisions taken only by a fraction of the taxpayers and the potential positive economic and societal effects of political integration of non-citizens, it is important to understand the conditions under which native voters are willing to share political rights with noncitizens.While this paper is among the first to consider this question, it relies on a much richer dataset than previous studies.We explore a new municipality level dataset of Swiss cantonal referenda on the enfranchisement of non-citizens.The Swiss setting provides a unique laboratory for capturing the drivers of the enfranchisement of non-citizens, as it enables measuring the actual electorate's revealed preferences.To explain the willingness to enfranchise non-citizens, we focus on the role of the size and composition of the foreign population in different institutional settings.
Our estimates reveal that the approval of non-citizens' voting rights is negatively affected by the share of foreigners present in a municipality.Thus, when confronted with a larger foreign population, the support to extend suffrage to non-citizens decreases.Our analysis reveals that this relationship is more pronounced in municipalities with strong direct democracy and larger cultural distance between the native and foreign population.There are at least three complementary explanations for this finding: First, the costs to enfranchise non-citizens are larger if the actual electorate loses more political influence (direct democracy and preference heterogeneity).Second, our results may also indicate a tendency for direct democracy to discriminate against outsiders.Third, mechanisms based on attitudes may also explain this result.The larger and more culturally distant the outgroup is, the more threatened the natives may feel in their cultural identity and the more prevalent the anti-immigrant attitudes may become.Furthermore, our results support the hypothesis that a tense situation on the labor market has a negative effect on the enfranchisement of non-citizens.
Since many developed countries face an increasing share of foreign residents and cultural differences are likely to expand, our results imply that, in the future, citizens' willingness to enfranchise non-citizens may decline and the lack of democratic legitimacy is not likely to be washed away automatically.Moreover, political integration via suffrage extension seems more likely in times of economic prosperity and phases of low unemployment.For future research on determinants of non-citizens' enfranchisement, individual data of the actual electorate might be helpful for a clearer distinction of instrumental power loss or discrimination channels.

Appendix
Tables A1, A2, A3, A4, A5, A6 and A7 Figs.A1, A2, A3, A4, and A5 Note: Liechtenstein does not participate in the WVS and was therefore not included in Inglehart and Baker's (2000) classifications.We still added Liechtenstein to the culturally similar countries, since they are arguably culturally very close to the Swiss population (i.e., geographical proximity, language, Swiss francs, labor market, Swiss customs union).

Table A3
IV Estimations with the Foreigner Share of the MS-Regions as an Instrument.The dependent variable is the yes share mtr in votes on non-citizens' enfranchisement.Robust standard errors in parentheses are clustered at the municipal level.Votes embedded in a political package are included in Spec.(1).Spec.(2) controls for population 2 and population instead of population (log).In Spec.
(3) the smallest 5% of municipalities are excluded and in Spec.(4) also the largest 5% are excluded.In Spec.( 5

Fig. 1 .
Fig. 1.Graphical Illustration of the Interaction of Culturally Distant Foreigners with Share of Foreigners.
The dependent variable is the yes share mtr in votes on non-citizens' enfranchisement.Robust standard errors in parentheses are clustered at the municipal level.Control variables include population size (log), unemployment share of Social Democratic Party, pensioner share, agriculture share, naturalization share, Gini coefficient, and mean income (log).*** p<0.01, ** p<0.05, * p<0.1.

Fig. A1 .
Fig. A1.Intermunicipal Variation of Yes Share between First and Last Vote in the Panel.Note: Only the first and the last vote per canton is considered.Therefore, the variation is not identical with the variation used in our empirical analysis.

Fig. A2 .
Fig. A2.Intermunicipal Variation of Foreigner Share between First and Last Vote in the Panel.Note: Only the first and the last vote per canton is considered.Therefore, the variation is not identical with the variation used in our empirical analysis.

Fig. A4 .
Fig. A4.Development of Foreigner Share in Switzerland from 1980 to 2020.Source: Federal Statistical Office.

Fig. A5 .
Fig. A5.Graphical Illustration of the Interaction of Legislative Institution with Foreigner Share.Note: As base effects of time-invariant institutions cannot be displayed for our preferred specification with municipality fixed effects, this figure reflects predicted margins for a specification with cantonal fixed effects.

Table 1
Accepted Referenda on Non-citizens' Suffrage in Swiss Cantons.

Table 3
Explaining the Yes Share in Referenda on Non-Citizens' Voting Rights (OLS).
The dependent variable is the yes share mtr in votes on non-citizens' enfranchisement.Robust standard errors in parentheses are clustered at the municipal level.Votes embedded in a political package are excluded from the sample.*** p < 0.01, ** p < 0.05, * p < 0.1.

Table 4
Interaction of Culturally Distant Foreigners with Share of Foreigners (OLS).
The dependent variable is the yes share mtr in votes on non-citizens' enfranchisement.Robust standard errors in parentheses are clustered at the municipal level.Control variables include population size (log), unemployment share, share of Social Democratic Party, pensioner share, agriculture share, naturalization share, Gini coefficient, and mean income of natural persons (log).*** p < 0.01, ** p < 0.05, * p < 0.1.

Table 5
Interaction of Parliament with the Share of Foreigners (OLS).The dependent variable is the yes share mtr in votes on non-citizens' enfranchisement.Robust standard errors in parentheses are clustered at the municipal level.Control variables include population size (log), unemployment share, share of Social Democratic Party, pensioner share, naturalization share, Gini coefficient, and mean income (log).Votes embedded in a political package are excluded from the sample.The number of observations is smaller in Spec.(

Table A1
List of Referenda on Non-Citizens' Enfranchisement in our Dataset.
The dependent variable is the yes share mtr in votes on non-citizens' enfranchisement.Robust standard errors in parentheses are clustered at the municipal level.The foreigner share on the municipal level is instrumented by the foreigner share in the MS-region.Control variables include population size (log), unemployment shares, share of the Social Democratic Party, pensioner share, agriculture share, naturalization share, Gini coefficient, and mean income (log).In Spec.(5) and (6) the smallest and largest 5% of the municipalities (in terms of population size) are excluded from the sample.

Table A5
Robustness Checks with Different Linear Time Trends for Base Model.The dependent variable is the yes share mtr in votes on non-citizens' enfranchisement.Robust standard errors in parentheses are clustered at the municipal level.Votes embedded in a political package are excluded from the sample.Spec.(3) includes no vote FE as they are captured in the cantonal time trends.

Table A6 Robustness
Checks of the Interactions of Culturally Distant Foreigners with the Share of Foreigners.The dependent variable is the yes share mtr in votes on non-citizens' enfranchisement.Robust standard errors in parentheses are clustered at the municipal level.Spec.(1) to (3) include votes embedded in a political package.Spec.(4) to (6) include population (total) and population (squared) instead of population (log).In Spec.(7) to (9) the smallest and largest 5% of the municipalities (in terms population size) are excluded from the sample.Control variables include population size (log), unemployment shares, share of Social Democratic Party, pensioner share, agriculture share, naturalization share, Gini coefficient, and mean income (log).

Table A7
Robustness Checks of the Interaction of Parliament with the Share of Foreigners.