Praying on Friday, voting on Sunday? Mosque attendance and voter turnout in three West European democracies

ABSTRACT Several studies have demonstrated a positive association between regular church attendance and turning out to vote in established democracies. This paper examines whether the relationship holds for Muslims who regularly attend religious services. Using an original dataset of Muslim-origin citizens in Germany, the Netherlands and the United Kingdom, I find that regular mosque attendance is indeed associated with higher likelihood to vote in national elections in Germany and the United Kingdom, while among Dutch Muslims turnout is positively associated with individual religiosity. I find evidence that the proposed association between regular mosque attendance and voting is mediated through the acquisition of relevant political information and stronger associational involvement. The paper provides an individual-level analysis complementing studies of country-level institutional particularities and group-level characteristics that are conducive to higher levels of turnout among Muslims. The findings dispel the myth that mosques are sites of civic alienation and self-segregation, but can, in fact, play the role of ‘schools of democracy’.


Introduction
The decision to show up at the polling station and vote on election day is regarded as a sign of successful civic integration and engagement of ethnic minorities. Electoral participation defies the trend of increasing political apathy among the general population in liberal democracies (Blais and Rubenson 2013). This is all the more important when citizens of minority background, often suspected of alienation in their beliefs from the liberaldemocratic mainstream and practices of self-segregation from the political process, are the ones doing the participating. This paper explores the factors contributing to electoral turnout among members of an important ethnoreligious minority, Muslims with a right to vote in three West European democracies, and zeroes, in particular, into the association between attendance of religious services and electoral participation. Is attending the mosque regularly associated with indifference to one of the most important aspects of the democratic political process? Or does it confer the resources, cognitive skills and psychological dispositions that render electoral participation more likelyand if so, why?
The choice of Muslim faith as the basis of sample selection is justified, because Muslims have increasingly come under scrutiny by politicians and a number of public intellectuals for supposedly forming, as a group, an 'enemy within' liberal democracies (Cesari 2013, 6-10). More broadly, Muslim-majority democracies have been found to have lower turnout rates, when controlling for other relevant institutional and socio-economic variables (Stockemer and Khazaeli 2014). At the same time, sympathetic observers have noted that a thaw in relations between Muslim minorities and non-Muslim majorities is predicated upon a better understanding of how Muslim communities are organised and how legitimacy is generated among their members (see, for instance, March 2011).
Relatedly, the idea that Muslim faith has become, in West European countries, the basis of a 'thick' ethnic identity, in other words, that it is currently a salient source of grievances and basis of potential mobilisation, is gaining traction among sociologists and political scientists (Alba and Foner 2015). However, the hypothesis that being Muslim is associated with specific political opinions or distinct political behaviour (turnout, left-right placement, party choice) has been tested only sporadically across countries (see for instance , Dancygier 2017;Kranendonk, Vermeulen, and van Heelsum 2018). This paper contributes to this literature by looking at the most common form of political participation, electoral turnout.
I am interested here in individual-level characteristics, and in particular the influence of regular mosque attendance on individual voter turnout, among Muslim citizens with the right to vote in West European democracies. I aim to test the hypothesis that the relationship is positive, as has been argued for Muslim voters in non-European contexts (for the US, see Jamal 2005, for Australia, see Peucker and Ceylan 2017) and explore possible mechanisms behind the hypothesised relationship. Other recent individual-country studies in Europe have also advocated a positive association between religious attendance and various forms of political participation (including voting) and proposed several possible working mechanisms (Giugni, Michel, and Gianni 2014;Sobolewska et al. 2015;Oskooii and Dana 2018). These studies are heavily influenced by a more established literature on the effects of church attendance on electoral participation among Christians, particularly in the United States. Regular churchgoers have consistently been found to report higher individual turnout rates (see, for instance, Rosenstone and Hansen 1993;Gerber, Gruber, and Hungerman 2016 for a quasi-experimental design leading to similar conclusions). This empirical regularity rests on the side-benefits of religious attendance for political mobilisation. Those attending religious services are more likely to acquire, on average, resources and skills that can inter alia be applied to voting. For instance, they may be exposed to political information of interest to their religious communities, develop strong associational networks that facilitate mobilisation on election day, and develop a heightened sense of political efficacy through regular participation in congregational activities. Regular attendants of religious services are also likely to develop group consciousness, that is, they may increasingly self-identify as members of a distinct minority and vote to defend and promote the rights of their group.
The present article focuses on the determinants of turnout among Muslims with the right to vote in national elections in Germany, the Netherlands and the United Kingdom. All three are major countries of immigration with sizeable Muslim populations, estimated at around 6% of the general population and projected to reach higher levels over the next decades (Grim and Karim 2011). The data is drawn from a recently completed large comparative project (Finding a place for Islam in Europe -EURISLAM) which has disseminated comparable survey data on the opinions and behaviour of Muslims across West European democracies. I focus here on individual-level traits and their interaction with group-level characteristics and thus complement the work of others who consider the effect of contextual factors, such as country-level institutional and discursive opportunities facilitating Muslim participation (Cinalli and Giugni 2016).

Mosque attendance and turnout: mediating mechanisms
Voting is the most common political activity in democracies, the one that regularly grants legitimacy to candidates for government. It also leads to substantive gains for voters of minority ethnic and religious background: formal and informal political mobilisation are linked with better political and economic outcomes for minority populations (Maxwell 2012). Still, minorities are less likely to vote than majorities in most of Western Europe, including in the three countries studied here (Bird, Saalfeld, and Wüst 2011). Large inter-country and inter-group differences exist and a lot of the variation is explained by inequalities in socioeconomic status and level of education (Sandovici and Listhaug 2010). Beyond the standard individual-level, socio-economic explanations, scholars have increasingly focused on institutional characteristics within countries, strength of group organisation and relational (network) resources of minority individuals with a positive influence on the likelihood to vote on election day (Sobolewska et al. 2015;Cinalli and Giugni 2016;Pilati and Morales 2016). The present paper focuses on one such potential source of networks and resources, namely religious attendance.
Religious attendance can potentially increase the resources individuals of minority background possess. By resources I mean various types of material and immaterial capital likely to make voting less costly/more beneficial and help overcome structural disadvantages. First, mosques, like churches, provide organised opportunities and a regular meeting place for individuals to discuss and act upon communal affairs (Baggetta 2009). Exposure to political affairs of direct interest to the Muslim community can be particularly valuable for persuading mosque attendants to vote. Thus the first hypothesised link between mosque attendance and individual voter turnout is: H1: Mosque attendance increases the amount of exposure to political issues affecting the Muslim community, leading to higher voter turnout.
Second, the benefits of religious attendance for electoral participation may be indirect, through an increase in overall associational involvement. Mosques are often focal points of a constellation of civic associations, where regular mosque attendants may interact with others, make decisions about communal affairs, become exposed to political campaigns and messages, and receive social pressure to participate in upcoming elections (Djupe and Gilbert 2006). This arguments reflects the seminal work by Verba, Schlozman, and Brady (1995) who showed that religious attendance increases civic skills, such as information about communal matters, familiarity with decision-making processes inside an organisation and the likelihood to mobilise for one's group. Thus a second hypothesis regarding the link between religious attendance and turnout can be generated: H2: Mosque attendance increases overall associational involvement leading to higher voter turnout.
Third, beyond the acquisition of direct and indirect resources that encourage political participation, mosque attendance might contribute to individual turnout through fostering support for democracy. As a long tradition of research on the Black and Latino Churches in the United States has shown, places of worship can become the milieus of direct political mobilisation (Jones-Correa and Leal 2001). For example, it has been argued that religious leaders often provide scriptural interpretations that are conducive to political participation (Driskell, Embry, and Lyon 2008). Mosques might similarly provide direct support for democracy as 'the only game in town'. There is a positive empirical link between a belief in the efficacy and legitimacy of democratic institutions and political participation (Smets and van Ham 2013, 11). To that effect, West European governments have attempted to structure a dialogue with umbrella Islam Councils and the latter have generated declarations of allegiance to democratic constitutional principles (Cesari 2004;Laurence 2012). Thus: H3: Mosque attendance is conducive to positive feelings towards democracy, leading to a higher individual-voter turnout.
Fourth, sermons and regular activities within a circumscribed religious community increase a sense of ethnic community and raise group consciousness, a precondition for politicisation (Lee 2008). Politicised consciousness underpins the 'linked fate' hypothesis, according to which respondents believe that political developments that alter the conditions of their ethnic group generally are bound to have an effect on them personally (Dawson 1995). Thus mosque leaders in Western Europe might advocate strong turnout before the faithful in a context where Islam is increasingly stigmatised by political actors (Betz and Meret 2009). Controversies surrounding the building of mosques, the public wearing of the veil, the expulsion of radical imams and the legitimacy of violence in the name of Islam might increase the saliency of Muslim identity, stimulate group consciousness and induce Muslims to correct public perceptions of their community in the public domain (Phalet, Baysu, and Verkuyten 2010;Oskooii 2016). H4: Mosque attendance generates a politicized, group consciousness among Muslims, leading to higher individual voter turnout.

Data and operationalisation
The analysis rests on data from the project 'Finding a place for Islam in Europe' (EURI-SLAM), an EU Seventh-Framework project. One component of the project was a telephone survey designed to capture the characteristics and beliefs of people originating from Muslim-majority countries in the six largest West European destination countries (Belgium, Switzerland, Germany, France the Netherlands and the UK). 1 Questionnaires distributed in all countries, with the exception of France, included a question item on participation in the last national/federal election. The Swiss questionnaire did not include a screening question on having the right to vote and Switzerland was excluded, along with France, from the analysis. Belgium was also not included, because of the compulsory and strictly enforced voting participation laws in that country, which result in unusually high turnout rates.
The three remaining countries (Germany, the Netherlands and the UK), besides being major destination points for Muslim immigrants, feature different policy frameworks for citizenship acquisition, recognition of cultural rights of ethnic groups and overall approach to migrant integration. Thus, the United Kingdom and the Netherlands historically enforced less strict criteria for immigrant access to citizenship compared to Germany (Howard 2009(Howard , 2012. The Netherlands demonstrated an early willingness to institutionalise ties with ethnic and religious groups as a means of achieving integration and a more recent turn towards assimilationism (Vasta 2007). UK policy prioritised anti-discrimination and equality of opportunities over outright cultural recognition of religious groups (Modood 2006). Germany, on the other hand, focused on labour market integration, while retaining a more restrictive (ethnic) model of citizenship acquisition and cultural integration (Schönwälder and Triadafilopoulos 2016). Based on such national differences and the different institutional incentives for the self-organisation of ethnic and religious minorities, the effect of religious attendance on turnout could be expected to vary across the three countries. At the same time, the existence of a positive effect of regular mosque attendance on the likelihood to vote, irrespective of country-level policy traditions, would provide convincing evidence for an individual-level effect across national contexts.
The EURISLAM survey ran in 2011 and the first quarter of 2012 and questionnaire modules included, among other inquiries into patterns of 'interaction and cultural distance', items on political opinions and behaviour. The sample frame was based on name-recognition (onomastic sampling; see Schnell et al. 2017 for a description of the method) to identify Muslims of Turkish, Moroccan, Pakistani and ex-Yugoslav origin, these groups representing large enough populations from which to draw a representative sample in all six countries (Tillie et al. 2013). The purpose of the multi-group sample was to allow for testing the moderating effects of group-level characteristics. The sample frame was then applied to a digital phone directory to be subsequently used for random selection of potential respondents. Group-based and country-based sampling were combined, in order to provide a framework of testing hypotheses of minority attitudes and behaviour at the appropriate levels of analysis. Table 1 presents the numbers of respondents (natives and from each Muslim ethnic group) screened for eligibility and successfully interviewed, as well as the response rates for each group. To account for the heterogeneity of Muslim populations, seven groups based on Muslim-majority countries of origin were initially identified by the EURISLAM team as historically, linguistically and culturally distinct sample targets. However, only Turks, Pakistanis, Moroccans and ex-Yugoslavs (Muslims from Bosnia, Kosovo and FYR of Macedonia) featured high enough numbers in all targeted countries to produce a large enough sample. Bangladeshis, Algerians and Tunisians are thus excluded from the following group-level analyses. The dependent variable of interest measures electoral participation by asking respondents whether they voted in the last national/federal elections. In order to avoid conflating those ineligible to vote and those who did not vote for other reasons, the questionnaire included a filter question on eligibility to participate in national elections. The main independent variable of interest, frequency of mosque attendance, is derived from the question item: 'How often do you go to the mosque or other place of worship?'. This item is likely to evoke attendance to congregational Friday and daily prayers at the mosque or other places designed for communal worship, but should not be taken to exclude other social and religious activities taking place in these spaces (Westfall 2018). For clarity's sake, I have constructed a binary variable with value '1' representing those who answered they attended services daily or weekly and '0' standing for those who answered 'rarely' or 'never'. Mosque attendance is, of course, one of several components of individual religiosity, for instance frequency of prayer or conspicuous display of religious symbols (see Cinalli and Giugni 2016). I focus here on mosque attendance, because, much like church attendance, and unlike individual ritual prayers (salah) that are considered a duty for every Muslim, it accrues relational/networking and organisational benefits. Still, the analysis of individual countries, in particular the Netherlands, calls for examining the confounding effect of individual religiosity, broadly defined, among respondents; in this case, I include daily frequency of prayer (see Table 4 and adjacent analysis).
In exploring the association between mosque attendance and turnout, the paper proposes testing different causal mechanisms identified in the literature on political behaviour of minorities and controlling for potentially important confounding variables. I recognise that a causal link cannot be definitively established with cross-sectional, non-experimental data. Yet in providing empirical evidence for positive and statistically significant associations across countries and exploring potential mechanisms is an important first step towards studying the link between Muslim religious attendance and electoral participation in European democracies. To test the mediating mechanisms of the effect of mosque attendance, I include models with variables standing for knowledge of the national Islam Council (H1exposure to political issues), non-political associational membership (H2overall associational involvement), opinion on democracy (H3attitudes towards democracy) and self-identification as Muslim (H4group consciousness). 2 I have also included several control variables in the models that are deemed relevant in the well-established literature on electoral participation: age (in years), unemployment (whether the respondent has paid work for more than 12 hours a week), educational status (International Standard Classification of Education 5-point scale), whether the respondent was born in the host country or not (Coffé and Bolzendahl 2010;Sondheimer and Green 2010;Smets and van Ham 2013). I also include self-reported proficiency in the host-country language. Last but not least, mosque attendance in the three countries is very much a gendered experience, because congregational prayers at the mosque or other communal places of worship (the Friday or Jummah prayers) are considered a duty only for Muslim men. This makes Muslim women considerably less likely to report regular mosque attendance (only about 13% of female respondents in the three countries compared to 37% among men). For this reason gender was included as a control variable in the following analyses.
Operationalising associational involvement as a mediating mechanism for the 'civic' effect of mosque attendance on voting raises objections. Some scholars argue that those attending religious services are also likely to join other associations and vote, because they are 'joiners' more generally and thus likely to participate in various forms of public and civic life (van Ingen and van der Meer 2016). To account for that possibility I run a separate analysis with non-political associational involvement as the dependent variable. I also include the knowledge of the national Islam Council as an alternative indicator of political resources acquired through mosque attendance. Islam Councils are officially recognised interlocutors between Muslim communities and national governments. They aim to centralise and standardise the administration of issues that arise between European states and practitioners of the Islamic faith (for instance, religious education in school curricula, burial and praying practices) and to represent Muslim interests in a way that respects constitutional principles and dispels fears of radicalisation (Laurence 2012). These bodies are typically umbrella organisations for several hundreds of mosques (or federations of mosques) and feature elections or other selection procedures that familiarise mosque attendants with standard democratic processes. As they mediate on statemosque relations, they are likely to increase political awareness of and interest for national politics among Muslims who attend mosque services.
In fact, representatives of Islam Councils often make direct pleas to the faithful to vote. For instance, the speaker of the Muslim Coordination Council (Koordinierungsrats der Muslime), formed in the context of the German Islamkonferenz, explicitly invited the 1,5 million Muslims in Germany with a right to vote to do so in 2017 and included a 'Master sermon' to be read out in several mosques before the election (Neue Osnabrücker Zeitung 2017). The Muslim Council of Britain (MCB) launched a 'Muslim Vote 2010' online platform to inform British Muslims about specific issues relating to the Muslim community and encourage them to vote in the general election. 3 For this reason I take knowledge of the national Islam Council as a potential mediator between mosque attendance and participation in national elections. Finally, I run the models including countrylevel and ethnic-group-level fixed effects to account for any unobserved characteristics in countries or groups. Although not of primary focus here, the discursive and institutional opportunity structures for Muslims vary in the three countries (Cinalli and Giugni 2016). Sobolewska et al. (2015) and Jamal (2005) also argue that the political salience of religion differs across groups depending on their historical trajectories in the host country and their organisational characteristics. The following section presents the main findings of the pooled-sample analysis controling for country-and group-level effects. The section after the next presents the results of the mediation analysis for potential mechanisms. Table 2 and Graph 1 provide descriptive statistics of electoral participation among Muslim respondents in Germany, the United Kingdom and the Netherlands. Turnout rates for Muslims with the right to vote were lower than the national average in all countries, the difference ranging from about 1.5% in the United Kingdom to about 5% in Germany. As Graph 1 demonstrates, regular religious attendance was associated with higher turnout rates both among Muslims and non-Muslims of the sample. A simple t-test operated for the two Muslim subgroups (high and low mosque attendance) shows that the difference in turnout is statistically significant at the 5% level, a first indication that mosque attendance is positively associated with showing up to vote on national election day. 4 The difference in electoral participation between Muslims reporting high and low mosque attendance exists in all three countries, although it is only 2% in the Netherlands. More generally, the Netherlands features substantially higher turnout rates than the other two countries.

Findings
I now turn to the main analysis, fitting models to the data in order to account for electoral participation in the last national/federal elections among Muslim respondents. Since the dependent variable (turnout) is binary, the results of simple logistic regressions with different specifications are presented. Table 3 displays the results (odds ratios) for the pooled 3-country sample. The variable of interest, regular mosque attendance, is, as hypothesised, positively and significantly associated with turnout in national elections (Model 1). The marginal effect is moderately strong, when controlling for other variables at their meansa little more than 5% higher probability to vote (Graph 2). The predicted effect is approximately as strong as being employed for more than 12 hours a week or  Standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1.
Graph 2. Predicted probabilities of voting in national elections for varying mosque attendance (Model 3.1).
adding 10 years of age or being fluent in host-country language, the other three variables of statistical significance at the 5% level. Results presented in Model 2 reveal that the effect of the attendance of religious services is moderated when controlling for group rather than country characteristics. 5 Yugoslav Muslims are consistently found to be less likely to vote and provide the strongest evidence for the (here: negative) effect of national group origin. This may be linked to the fact that the members of this group are the most recent arrivals compared to the rest of the sample. Those Yugoslav Muslims who have the right to vote have practiced it for a short time, in addition to experiencing a long period of legal and socio-economic uncertainty, while the organisation in ethnic and religious networks is also relatively recent dating from the Yugoslav wars of the 1990s (Behloul 2011;Paul 2016).
On the other hand, Muslim respondents of Moroccan origin report the highest turnout rates, a finding that is surprising given the scholarly consensus on lower levels of Morroccan organisation and mobilisation , and might reflect a longer-term integration process of these populations. Less surprisingly, Turkish Muslims also report higher than average turnout rates. Turkish mosques abroad enjoy organisational cohesion and resources through the Interior Ministry's Directorate of Religious Affairs (Diyanet). For example in 2014, the German (Turkish Islamic Union for Religious Affairs -DITIB) and Dutch (Turkish Islamic Cultural Federation in the Netherlands -TICF) branches of Diyanet organised about 900 and 140 mosques respectively, more than half of the overall number of Turkish mosques and a much larger number than any other Muslim religious organisation in these two countries (Nielsen and Otterbeck 2015, 34, 55). Tight organisation of mosques by the Turkish government often involves appointing imams directly for a few years and using them as a means of indirect foreign policy. It also means that exhortations to vote in host-country elections are likely to reach large audiences and in a consistent manner. Diyanet imams have consistently advocated deeper political integration of Turkish immigrants and their descendants (including voting), because it increases the visibility and provides a springboard for representing Turkish interests in the host country (Warner and Wenner 2006, 466). For instance, the head of the German DITIB Izzet Er issued a press release before the 2013 Bundestag elections, in which he stressed that 'Voting is a citizen's duty. One should absolutely make use of this democratic right' (cited in Gorzewski 2015, 170).
Pakistanis are also considered a well-organised religious community, simultaneously combining high levels of identification with Islam and with their host countries, especially in the UK (Jacobson 2006, 32, 34). They were the first South Asian communities to become politically mobilised during the Salman Rushdie Affair, but also in response to the Kashmir conflict and other cases where Muslims around the world were perceived to be targeted (Gulf War, Israel-Palestine conflict). In fact, Pakistani diasporic communities have been at the forefront of the politicisation of Islam (Werbner 2005, 480-483) and would thus be expected to exert pressure on their members to become politically active. In Britain, the country where the Pakistani community is the most numerous, they have provided leadership for the national Muslim Council. At the same time, Pakistani diaspora mosques have also suffered from sectarian fragmentation (Werbner 2005, 481-482) and do not enjoy the organisational grip of a foreign government that characterises Turkish mosques.
The pooled-sample analysis yields a strong and significant country-level effect on turnout for the Netherlands compared to the UK and Germany. In order to account for this empirical outlier, I ran separate country-level analyses for all countries (Table 4). The effect of mosque attendance on turnout is found to be stronger for the country-level UK and Germany analyses, albeit weakly significant as the number of observations decreases substantially. The effect disappears in the analysis specific to the Netherlands. What accounts for the Dutch exceptionalism? Interestingly, among Dutch-Muslim respondents there is a strong and statistically significant positive association between regular prayer and voting in the national elections (Model 4). Alternative specifications, including running models with group-level effects and other mediating mechanisms do not alter the strong effect of regular prayer on the likelihood to vote in the Netherlands. I do not find a similar positive link between regular prayer and voting in Germany and the UKin both countries the positive association between mosque attendance and voting is positive, strong and robust in alternative specifications, including regular prayer. 6 It seems then that religiosity is a better predictor of turnout on election day than mosque attendance among Dutch Muslims. Given this result, more research is needed to gauge the effect of the Dutch national integration model on the accommodation of Islam, in particular with regard to how institutional and discursive opportunities lead to increased political participation among devout Muslims. The very public debates in the Netherlands about the role of Islam in public life since the assassinations of controversial Islam critics Theo Van Gogh and Pim Fortuyn are likely to have increased the group consciousness of religious Muslims and induced them to participate in national politics. At the same time, the earlier recognition and inclusion of ethnic and religious associations in the Netherlands (Vasta 2007) means that devout Dutch Muslims can rely on a wealth of resources for mobilisation (minority schools, associations, ethnic networks), other than the mosques.

Mediating mechanisms
Finally, Table 5 displays the results of the multivariable logistic regression models that include the variables chosen to operationalise the potential mediating mechanisms. Direct and indirect resources in the form of associational involvement and knowledge of the national Islam Council are found to be strong and statistically significant predictors of electoral participation; they also decrease the effect size of mosque attendance (Models 1-2). One can tentatively conclude that informational and network resources accumulated through mosque attendance are indeed an important contributing factor to individual electoral participation. Self-perception as a Muslim is more strongly correlated with mosque attendance and thus decreases its effect on turnout more substantially, but is itself not as significant a predictor of individual turnout as associational activity and increased exposure to political issues (Model 3). Thus I find weaker evidence for a 'group consciousness' mechanism for mosque attendance; to bolster this conclusion I have run an alternative model substituting perception of discrimination against Muslims (a commonly used operator for 'group consciousness') for self-perception as Muslim. This question was unfortunately not available in the Dutch questionnaire, but when running the models for Germany and the UK (not shown here) the effect of mosque attendance remained unchanged and statistically significant and perception of discrimination was statistically insignificant. Lastly, there is no evidence that overall attitudes towards democracy have a positive or negative effect on turnout (Model 4). 7 To be sure, associational involvement might not reflect the positive influence of religious attendance on a Muslim's participation in community life and broader networks, Standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1.
but might be the result of a confounding variable, for instance the respondent's overall, pre-existing civic-mindedness that makes them more likely to join an association and to attend congregational prayers. In order to preclude the possibility of such omitted variable bias, I have run a multivariable regression analysis on reporting being a member of an association. Table 6 presents the results; the relationship between regular mosque attendance and joining an association is positive and statistically significant even after controlling for standard socio-economic indicators (such as level of education and employment status) and, importantly, for positive attitudes towards democracy that would indicate higher levels of civic-mindedness. The results do not change when including country and group fixed-effects (not shown here). In a similar vein, a separate analysis of the factors associated with knowledge of the national Islam Council was run, in order to preclude for the possibility that reporting high levels of politically valuable resources is the effect of individual characteristics other than mosque attendance. The results are presented in Table 7. They show a very strong and statistically significant positive association between attending the mosque regularly and reporting knowledge of the national Islam Council, but no link of such knowledge with education level, associational membership or religiosity measured by frequency of prayer. The analysis here, in conjunction with the analysis presented at Table 5, points at a distinct mechanism for the effect of attending a mosque on electoral participation: regular mosque-goers are more likely to acquire political information that is pertinent to the Muslim community, making them more likely to participate in the political process. This may happen through direct canvassing in mosques or an indirect increase of perceived political efficacythe exact mechanism can unfortunately be tested using the available EURISLAM databut it seems to be independent from whether the respondent is, generally speaking, someone interested in civic life or religious matters. The analysis also demonstrates that Turkish respondents are more likely to have knowledge of the national Islam Council, a finding underlying the strong participation of the representatives of the Turkish community in these bodies. The German Islamkonferenz is also betterknown than its UK and Dutch equivalents perhaps suggesting higher levels of consolidation and representativity of this body in Germany or better media coverage.

Conclusion
This article set out to investigate the link between regular religious attendance among Muslims with the right to vote in three West European democracies and voting in (national) elections. The evidence drawn from an original dataset of citizens of Muslim background suggests that attending the mosque does not lead to political alienation. In no model and under no specification was there a negative association between regular mosque attendance and individual turnout. In fact, regular mosque attendance was mostly found to be positively associated with this most common form of political participation in consolidated democracies. This empirical finding, well-established for churchattending Christians and Muslims in the United States (Jamal 2005), seems to be confirmed in the case of West European Muslims as well, most significantly for Muslims in the UK and Germany. Of the possible mechanisms that mediate between regular mosque attendance and individual turnout, evidence is strongest for the exposure to political issues relevant for the Muslim community (such as knowledge of the national Islam Council) and an increase in overall associational involvement that indirectly makes Standard errors in parentheses. ***p < 0.001, **p < 0.01, *p < 0.05. a mosque attendant more likely to vote. One can thus interpret mosque attendance as a religious source of political and social capital that can be invested in electoral participation and can partly compensate for structural disadvantages of minority individuals. Unlike Sobolewska et al. (2015) I do not find evidence for mediation through increased psychological resources, such as heightened group consciousness or an individual's higher trust in democratic politics. The findings presented here do not imply that country-level or group-level dynamics are not important. To be sure, private religiosity in the Netherlands is more robustly associated with voting than attending the mosque, while some groups among Muslims (notably Yugoslavs) report lower levels of participation, while others (notably Moroccans and Pakistanis) vote almost as often as native respondents, with differences between these groups not disappearing when controlling for mosque attendance. Such findings should caution against hasty generalisations of a mosque-attendance effect on electoral participation in other countries or groups. Relatedly, one of the limitations of the present study is the inability to test for the ethnic composition of the mosque attended by the respondent as a potential mediating mechanism between mosque attendance and electoral participation. The argument, developed most prominently by Jones-Correa and Leal (2001) in the study of Latino Churches in the United States and tested for places of worship in Britain by Sobolewska et al. (2015), posits that ethnic homogeneity in the place of worship makes the politicisation of ethno-religious identities more likely, increasing electoral and non-electoral participation. Therefore, mosques act as repositories of ethnic social capital, which is itself associated with higher political participation (Jacobs and Tillie 2004). Although the brief discussion of group-level effects above suggests that most Muslims sampled are likely to attend ethnically homogeneous mosques, the EURISLAM questionnaire does not include a specific question item on the ethnic make-up of their places of worship.
In future studies, the research design can also be extended to non-electoral participation; as well as the interaction between party identity and mosque attendance. The tendency of Muslim voters to identify with and vote for left-wing parties is documented in the American literature (see Barreto and Bozonelos 2009). It is not simply the foundational democratic procedures, such as elections, that devout Muslims might be induced to endorse, but also specific parties, in particular left-wing and center-left parties (Dancygier 2017). Through support for and strong identification with such parties, minority political alienation decreases and the entire democratic process benefits. There exists little evidence for this link in European democracies as well, as whether it is solidified in religious congregations. Future international, comparative studies would also benefit from more standardised and precise survey items on the nature and sources of civic skills (political information, communication skills, political trust) acquired by Muslim-origin individuals in European democracies. knaw.nl/ui/datasets/id/easy-dataset:62447/tab/2 2. Associational membership: 'Do you join any associations such as sports clubs, religious organisations, labour unions, parents associations, etc. in, where you meet other people?'. Knowledge of the national Islam Council: 'Are you aware of Deutsche Islam Konferenz (Germany)/Mosques and Imams National Advisory Board (the UK)/ Contactorgaan Moslims en Overheid or the Contactgroep Islam? (the Netherlands)'. Opinion on Democracy: 'Could you please tell me if you agree strongly, agree, disagree or disagree strongly: Democracy may have problems, but it's better than any other form of government'. Selfdefinition as Muslim: 'To what extent do you see yourself as Muslim?' 3. 'Muslim Vote 2010' accessed 5 November 2019 at http://archive.mcb.org.uk/muslim-vote-2010/ 4. The t-test result is significant at the 5% level for the pooled sample and for any 2-country combination. For the individual-country samples, it is only significant at the 10% level for the UK. 5. Tables 3 and 4 of the Online Appendix present results including both country-fixed effects and group-fixed effects, along with the variables from the analysis for Tables 3 and 5. The size of the hypothesised effect of regular mosque attendance on voting remains substantial, but the predicted coefficient loses statistical significance because of the reduced power of the models. 6. See Online Appendix, Table 5. 7. In order to control for the possibility that the coefficient for mosque attendance decreases due to the presence of non-respondents, I have run each model in Table 5 without the mediating variables and excluding non-respondents corresponding to each mediating variable. The coefficient size and significance for mosque attendance did not change compared to the full-sample analysis.

Disclosure statement
No potential conflict of interest was reported by the author.