Across Western Europe, politicians and media express concerns about whether Muslims feel part of liberal democracies and trust national parliament. Although often assumed, research is actually inconclusive on whether Islamic religiosity hampers such political trust. The current study dives into this complex association by studying over 5,000 Muslims across 17 Western European countries. It makes two core contributions. First, instead of comparing Muslims with non-Muslims, we address Muslims specifically to gain an understanding of how various religiosity dimensions—mosque attendance, religious identification and prayer—relate to trust in national parliament. Second, we theorize and test how exclusionary conditions buffer or aggravate Islamic religiosity's impact on political trust. We theorize how formal boundaries (e.g. restrictive citizenship policies) and informal boundaries (e.g. hostile public attitudes) affect Muslims’ political trust and the role of Islamic religiosity herein. After uniquely harmonizing data from the European Social Survey, European Values Study and World Values Survey, multilevel analyses show that Islamic religiosity can relate negatively to political trust, yet its importance is gendered and contextualized. For example, Muslim men who more often attend a mosque experience less trust, and more so in exclusionary societies, whereas we find indications that mosque attendance among women stimulates such trust in the least hostile environments. Moreover, Muslims’ political trust is in general lower in societies with more hostile public attitudes towards migrants and where political participation is restricted for some. Our study has thus illustrated that Islamic religiosity matters for Muslims’ political trust in Western Europe, though it is neither clear-cut nor all-explanatory.

Despite heated Europe-wide debates on Islam and politics, little is known about if and how individual-level manifestations and reproductions of Islam feed into political belonging and behavior. Specifically, while politicians and media express concerns about whether Muslims feel part of these liberal democracies and trust parliament, the number of studies exploring the association between Islamic religiosity and political trust is limited and often solely studied by comparing Muslims to non-Muslims (Hsiung and Djupe, 2019; Superti and Gidron, 2022). In this study we therefore theorize and unpack Islamic religiosity's potential impact on trust in national parliament, also shedding light on the larger question about whether differences between Muslim and non-Muslim citizens can be ascribed to Islam, as is often done.

On this matter, we offer two main contributions. First, this manuscript stresses the need to take an intracategorical approach by studying Muslims as a heterogenous group in terms of religiosity (see Geurts, 2022), also in religious terms, and thereby going beyond the popular analytical dichotomy between the religious and the non-religious (as stressed by Patrikios and Huhe, 2022). Studying the multidimensionality of Islamic religiosity allows us to go beyond the predominant understanding in the general political engagement literature, which uses mosque attendance as proxy for religiosity and understands religiosity foremostly as an indicator of social capital (Moutselos, 2020). This approach also allows us to go beyond the literature on cultural clashes that reduce the Muslim/non-Muslim dichotomy to deeply ingrained cultural differences.

While we consider both of these theoretical mechanisms, we will theorize and empirically study mosque attendance, religious identification and individual prayer as separate dimensions of Islamic religiosity that tap into different consequences and meanings of religious practice and belonging, and that can thus relate differently to trust in national parliament. Addressing various dimensions of religiosity, both theoretically and empirically, allows us to pinpoint which mechanisms are likely at play with regards to Islamic religiosity's relationships with political trust, thus enabling further development of theory in the field. Although the concept of political trust can encompass trust in various institutions, we predominantly focus on the most widely studied and measured dimension (Newton, Stolle and Zmerli, 2018): trust in national parliament. National parliament is one of the core institutions of liberal democracy, and therefore a relevant category of political trust to study (see Van der Meer, 2010; Zmerli and Newton, 2007).

In order to assess religiosity's multidimensional impact, we combine and build on insights from large-scale socio-political and sociology of religion studies on trust, in particular among migrants and minorities (Ben-Nun Bloom and Arikan, 2012; Glas, 2020). This approach helps us to understand the potential causal qualities of Islamic religiosity and to connect this to the core mechanisms identified in the general political trust literature, contributing to both of these bodies of literature.

Our second core contribution is that we explore how cross-national differences affect Muslims’ political trust and the role of religiosity, which is currently lacking (Isani and Schlipphak, 2017). That trust reacts to stimuli provided by the environment has been illustrated in general trust studies (Hooghe and Zmerli, 2011; Jeannet, 2020; Kroknes, Jakobsen and Grønning, 2015; Marien, 2011; Noordzij, de Koster and van der Waal, 2021; Van der Meer, 2017; Voicu and Tufiş, 2017), yet the exact contextual factors are likely (partly) different for Muslims as a minority in Western Europe (Bird, Saalfeld and Wüst, 2011; Eggert and Giugni, 2011).

In that light we introduce and test contextual conditions that are likely to affect Muslims’ political trust and shape the impact of Islamic religiosity. Particularly, we build on previous work by Oskooii (2020) and Simonsen (2016) addressing exclusionary policies and practices on the national level, which dovetails with the more general institutionalist literature that partly puts forward what such boundaries could be. Concretely, we study the role of laws on citizenship and political participation, seat disproportionality and public hostility towards ethnic minorities as indicators of such exclusionary circumstances.

Our contributions are summarized in our research question: How do mosque attendance, religious identification and individual praying relate to Muslims’ trust in national parliament in Western Europe, and how do exclusionary policies and practices on the national level affect Muslims’ trust in national parliament and Islamic religiosity's impact on it? To answer this question, we make use of a unique harmonized dataset combining Muslim citizens from the European Social Survey (ESS), the European Values Study (EVS) and the World Values Survey (WVS), the only three cross-national datasets that include information on all three dimensions of religiosity and items on trust in the same political institution: parliament.

In total, we include 5,267 Muslim respondents from 17 countries, covering 2002 to 2020. These data are enriched with macro-level data from multiple sources including Gallagher's disproportionality index (Gallagher, 1991), MIPEX legal migration data (Solano and Huddleston, 2020) and the general public's attitudes derived from ESS (European Social Survey Cumulative File, 2020). Moreover, drawing from general surveys, we can control for the general level of trust in the respective countries and years. Using multilevel regression models, we show that the dimensions of Islamic religiosity do indeed have different influences on political trust. Identification plays no systematic role, whereas particularly individual praying is often linked to lower political trust. The latter also holds for attendance, although mainly among men and seemingly more so in societies where the public is more hostile towards migrants.

Established causes of political trust

Various individual drivers of political trust have been theorized and have received considerable empirical support in the existing core literature on political trust. As mentioned in the introduction, we empirically measure trust in national parliament specifically. In the theoretical section, we build on previous literature mainly addressing political trust as a broader spectrum, where possible specifying how the mentioned mechanisms relate to trust in national parliament.

To theorize the impact of religiosity, the logic of the impact of socio-economic factors and social capital/integration and civic participation are most crucial. Below we briefly outline these mechanisms, which suggest there is a relationship between religiosity and political trust (against the null hypothesis that there is none). When we then turn to specific hypotheses on Islamic religiosity, we connect these general ideas to the role of social identification and specific understanding of religious acts and formulate specific hypotheses in line with the predominant discussion of religion in the trust literature. Next, each dimension of religiosity is followed by a reflection on these mechanisms in the context of belonging to a minoritized religious group. Doing so, we build on previous research on the role of religion among migrants and their descendants and integrate this into the two core mechanisms above, leading to a discussion of alternative expectations for the religiosity dimensions.

Regarding socio-economic factors, a core mechanism is the winner-loser logic. This logic suggests that higher levels of political trust can be found among those who are rich, happy, satisfied with life, healthy, well-educated and high on the ladder of social and economic status—the privileged ones in society (Newton, Stolle and Zmerli, 2018; Röder and Mühlau, 2012). These winners have more positive first-hand experiences with the world around them, and because the government is a likely cause and protector of this privilege, it will be trusted more. One important note, however, which comes back in the migration literature (see Geurts, 2022), is that education specifically is also linked to cognitive capabilities and being trained to think more critically and independently, which can feed into less political trust (Mayne and Hakhverdian, 2017).

Social capital and civic participation have been shown to relate positively to political trust and trust in national parliament specifically, also among immigrant minorities (De Vroome, Hooghe and Marien, 2013). It is argued that more knowledge gained via social capital decreases distrust and increases the likelihood of actively engaging in politics, which feeds into more positive attitudes towards democracy going hand-in-hand with experiencing political trust (Ben-Nun Bloom and Arikan, 2012; Zmerli and Newton, 2007). In addition, being embedded in intermediary groups creates more ties with society, which has been argued to facilitate reciprocal relationships of understanding and trust in a civic setting. This trust spills over to higher trust in other contexts (Kaase, 1999; Nannestad, 2008; Putnam, 2000; Uslaner, 2002). As Voicu and Tufiş (2017, pp. 355–356) put it, “While crediting others with trust, one is more likely to trust institutions that others create.”

By connecting these ideas to those more prominent in research on the implications of religion among marginalized migrant populations in Western Europe, we provide a systematic theorizing of different elements of (Islamic) religiosity's impact, showing how various dimensions work via different processes as well as tapping multiple, sometimes potentially counteracting, mechanisms. As will become clear below, the minoritized position of Muslims in Western Europe leans towards an overall interpretation that religiosity undermines political trust, whereas the general literature on political trust would suggest a positive relationship. We will empirically test which effect is dominant for each dimension of religiosity, and in the conclusion we will reflect on the specific mechanisms, based on the overall outcomes.

Islamic religiosity: a driving force or stumbling block for political trust?

We now turn to the role of religiosity, which is no stranger in the plethora of studies on political trust, in which it is conventionally treated as a pro-democratic value. However, its causal qualities have not been systematically theorized and it is often treated as a monolithic force equated with social capital alone. Moreover, many Western European studies seem to assume that religiosity works similarly across religions, whereas our intracategorical approach does not make this assumption. Below, we acknowledge its multidimensionality, which allows us to disentangle the diverse pathways linking religiosity to trust.

Mosque attendance

From a standard political trust framework, mosque attendance taps one's involvement in religious networks as a civic organization, bringing about civic skills. During as well as before and after services, mosque attendance facilitates discussing public affairs, political information and viewpoints (Moutselos, 2020; Oskooii and Dana, 2018). Similarly, attendance is known to result in embeddedness in a social network and to enhance a greater societal connectedness, which feeds into the social capital mechanism discussed above (Putnam, 2000). Indeed, research shows a positive association between church attendance and political trust (Sobolewska et al., 2015). Given that this logic is dominant in the literature, we formulate the following hypothesis for mosque attendance:

H1 Mosque attendance has a positive effect on political trust among Muslims in Western Europe.

However, considering the Western European context of marginalization and migration, mosque attendance is more than being embedded in a civic organization, which the above hypothesis ignores. It is also about the messages given by this organization and being integrated into a group with specific norms in such areas as how one as a Muslim should engage with the political system in Western Europe (De Koning, 2013; Glas, 2020; Jamal, 2005). In this light mosque attendance may bring about more distrust, as it is likely that during services it is stressed that Allah is the only and just authority, which might undermine the idea of the national parliament standing as authority (see Ben-Nun Bloom and Arikan, 2012). Moreover, and more tangibly, in addition to driving the internalization of a certain social identity, it has been argued that religious organizations may mobilize its members through raising awareness of experienced inequalities (see Oskooii and Dana, 2018) or challenge certain policies. Given the anti-Islam discourse across Western Europe, it is likely that more-frequent mosque attendees are more aware of these messages criticizing anti-Muslim politicians and how parliaments are unwilling to act against this discourse, or at least are unable to improve Muslims’ position in society (see Jamal, 2005). Consequently, it is not unimaginable that this contrary effect might nullify or overshadow the hypothesized relationship.

Studies have stressed that the frequency and meaning of mosque attendance differs for men and women (Glas, 2020; Prickett, 2015). While we acknowledge this gender difference, the expectations formulated above apply to men and women alike. While we will empirically assess whether results vary across genders, forming a theory on these differences falls beyond the scope of our study.

Religious identification

The social integration perspective also provides a starting point to reflect on the impact of religious identification. The argument is that identifying more strongly with a societal (including religious) group leads to greater levels of both social and political trust. In general, empirical results support this expectation (Bègue, 2002; Wisneski, Lytle and Skitka, 2009). We hereby should not ignore that many claim that love for and acceptance of others are pro-social principles at the core of Islam (Glas, 2020). Following this logic, a strong identification as Muslim thus links to a trusting disposition. Based on this core logic, we formulate the following hypothesis on identification:

H2 Religious identification has a positive effect on political trust among Muslims in Western Europe.

However, considering the context and applying the winner-loser logic interpreted broadly, it should be acknowledged that Western European Muslims are a marginalized and discriminated-against minority with a high level of group identification. Altogether this might actually result in more political consciousness of grievances. Such experiences can consequently contribute unfavorable attitudes towards the country of residence and its institutions, including the national parliament (Isani and Schlipphak, 2017; Liebkind and Jasinskaja-Lahti, 2000). Again, it might be that the overall effect is actually negative.

Individual prayer

Although less often studied in relation to political attitudes and engagement, individual prayer is important to many religions, including Islam in which ritual prayer (salat) is one of the five pillars. Moreover, people who pray more are often expected to be more devout, and given that Islam as a theology might not describe definitive political institutions but does offer a vision of a just society created by God (Collins and Owen, 2012; Eickelman and Piscatori, 1996), it might be that participation in activities such as prayer has been shown to increase psychological resources and civic engagement (Calhoun-Brown, 1996). However, to hypothesize that—based on this rather thin theoretical basis—prayer relates positively to Muslims’ political trust is tricky, whereas the reflection based on being a minoritized religious group leads to a hypothesis that is more theoretically supported.

First, theological drivers of trust mentioned above might not work similarly for a minoritized group. The bond between more frequent prayer and Allah could easily take precedence: praying indicates a priority to religious life over worldly affairs, creating fewer motivations to trust and engage in politics, whereby ultimate authority is given to the deity, as discussed above (Just, Sandovici and Listhaug, 2014).

Moreover, next to these general religious principles, focusing on specific guidelines that devout Muslims are more likely to follow (Doerschler and Jackson, 2012) also suggests a negative link between prayer and political trust. For instance, Wisneski, Lytle and Skitka (2009) concluded that those who show strong moral convictions experience lower levels of political trust, as there is a greater chance that parliaments will make decisions that are not in line with one's beliefs (see also Isani and Schlipphak, 2017).

Finally, though individual prayer might be an individual and partly private act, it can still take place in public spaces. Those who pray more may also more often experience hostility, which, as argued above, can hamper trust in institutions that do not prevent these hostilities. Given these arguments, we formulate the following hypothesis for prayer:

H3 Individual prayer has a negative effect on political trust among Muslims in Western Europe.

One more reflection is in place here. The importance of the act of praying might hold particularly for Muslim women, who are sometimes (socially) barred from praying in the mosque despite the required salat. We can thus expect a gender difference in the act of prayer: although in the general sense the hypothesis applies to both men and women, the gendered impact of prayer is assessed empirically in this study.

The importance of inclusionary and exclusionary environments

Above we drew attention to the hostility in the Western European context to understand why and how Islamic religiosity might matter for political trust. Below we theorize the direct impact of exclusionary environments as well as their (reinforcing) interplay with Islamic religiosity. In doing so, we explore inclusive and exclusive environments as part of the same continuum, and we apply and build on Oskooii's (2020) theoretical perspective regarding the relationship between discrimination and political behavior. We extend this perspective to political trust, showing how it connects the political science literature on formal institutions’ impact on trust (e.g. Van der Meer, 2010; Wängnerud, 2009) with the literature on boundaries, suggesting that policies and practices at the national level can indicate who does and does not belong via societal (interpersonal) boundaries, so-called informal boundaries, and political (systematic) boundaries, so-called formal boundaries (Simonsen, 2016). The more bright such boundaries are towards Muslims, the more exclusionary we consider the national context to be.

In the case of political engagement, Oskooii (2020) proposes that facing “laws, policies, practices, symbols or political campaigns and discourse that aim to deprive some citizens of resources or rights based on group membership” (p. 868) can motivate individuals to act collectively against institutions. This engagement implies a negative assessment of the institutions based on the experience of discrimination, thus implying low political trust. And indeed, a core understanding in the trust literature, which partly underlies the winner-loser logic, is that trust is partly experience based (Uslaner, 2002). More specifically, and considering a minority context, there are two mechanisms via which such political and societal boundaries will bring about lower levels of political trust.

First, following the rejection disidentification model (Bobowik et al., 2017; Liebkind and Jasinskaja-Lahti,2000), we expect that exposure to and experience with more exclusionary circumstances results in a lower sense of belonging and a more unfavorable attitude towards Western European countries and their domestic institutions, bringing about a lower level of political trust (De Vroome, Hooghe and Marien, 2013). Second, referring back to the winner-loser logic, a low(er) standing in society links to lower trust in institutions as this low(er) standing is attributed to the political system (Newton, Stolle and Zmerli, 2018; Röder and Mühlau, 2012). Minoritized citizens are likely aware of specific exclusionary boundaries, signaling their groups’ low(er) standing in society, which negatively affects the assessed performance of political institutions and actors, including parliament, which is seemingly unable to prevent or protect those exposed to these exclusionary circumstances (De Vroome, Hooghe and Marien, 2013; Kumlin and Rothstein, 2010; Maxwell, 2010). On the other hand, more inclusive circumstances can bring about a more positive evaluation of and thus trust in national parliament.

H4 The more exclusion is present in society, the lower the levels of political trust among Muslims in Western Europe.

Importantly, we also expect that these national circumstances affect the association between Islamic religiosity and political trust, as can be deduced from the reasoning on how the different manifestations of religiosity might lead to negative effects on trust, which already refer to these exclusionary contexts. If in a context this hostility is more pronounced, if people run into these boundaries more, it is likely that the reactionary elements of religiosity are activated more. In other words, in these contexts negative messages about the political elite might be stronger in and around mosques, the feeling of a lack of privilege for one's group might be stronger among those who identify strongly as Muslim, and the hostility towards praying in public might be stronger. Particularly in these contexts, those who are more religious can be expected to have lower trust, as indicated for the Canadian case in which Muslims who faced more discrimination experienced less political trust (Dilmaghani, 2019). This implies that more exclusive circumstances bring about a stronger negative association and/or limit possible positive associations between various dimensions of Islamic religiosity and political trust.

H5 The more exclusion is present in society, the more likely is a (stronger) negative association between (any dimension of) Islamic religiosity and political trust among Muslims in Western Europe.

Synchronizing survey sources

To test our hypotheses, we need a dataset covering three religiosity dimensions, a sufficient number of macro-level contexts, and a considerable sample of Muslim respondents. We therefore created a pooled data source that combines all Muslim respondents from multiple cross-national surveys of populations in Western Europe (for full code, see Spierings, Geurts and Kollar, 2024).

First, we listed all surveys—both for the general population and for specific migrant or ethnic minorities—that include multiple Western European countries (for comparability), items on political trust and the three religiosity dimensions. Only trust in parliament turned out to be available across surveys. Given these criteria, we could include the European Social Survey (ESS), European Values Study (EVS) and World Values Survey (WVS). Other datasets we found unsuitable include CILS4EU, CSES, EURIslam, EES, FRA's EU-MIDIS and 2,000 Families data.

After selecting all adult Muslims (based on self-identified denomination) with valid scores on our core variables, we were left with 5,267 respondents from 17 countries, covering a time span of 19 years (see Online Appendix A). To the best of our knowledge, this presents the largest existing dataset on European Muslims to study the relationship between Islamic religiosity and trust in national parliament.1

Combining data to enable our novel analyses evidently comes with challenges. Most importantly, the different datasets use different specific measurements for the same concepts.2 We build on procedures introduced in previously published work on Middle Eastern data (Spierings, 2019): aligning answer categories and applying the logic of standardized effects (which corrects for different dispersions on variables). Below and in the appendices, more detailed information is provided.3 Given the procedures, these variables should not be used to provide descriptive insights, but do allow for positioning each respondent relative to the others on all our core variables, which is crucial for assessing Islamic religiosity's relation to political trust across Western Europe.

Trust in parliament

Each of the surveys includes items on trust in one or more political institutions, whereby the surveys applied here ask about trust in parliament specifically. In line with our theories and the literature we relate to, trust in parliament focusses on the functioning of the system (as opposed to specific parties, incumbent government, or the notion of democracy as such; Van der Meer, 2010; Zmerli and Newton, 2007), and scores thereon seem to represent differences in this larger underlying concept rather well, also among minority groups (Spierings and Vermeulen, 2024). Moreover, Online Appendix G, which assesses our results on the same but smaller sample on trust in politicians and in political parties as well as an index of all three, leads to the same conclusions and the three items load on one single factor (loadings > .8) in an exploratory factor analysis.

The ESS data include an 11-point variable for which the question is phrased as follows (ESS, 2020): “Please tell me on a score of 0–10 how much you personally trust each of the institutions I read out. A score of 0 means you do not trust an institution at all, and 10 means you have complete trust. Firstly, [country]’s parliament.” EVS and WVS use a slightly different question (e.g. EVS, 2020a, 2020b; Inglehart et al., 2020): “Tell me, for each item listed, how much confidence you have in them, is it a great deal, quite a lot, not very much or none at all. Parliament?” (A great deal, quite a lot, not very much, none at all).

Combining these data, two issues need addressing: the number of answering options and the question phrasing. Regarding the latter, the questions are very similar, particularly so when we realize that “confidence” and “trust” translate in many languages the same. Focusing on the four languages that cover the countries with the most respondents (Belgium, France, Germany, the Netherlands and Switzerland), ESS, EVS and WVS questionnaires use the same words: confiance (French), Vertrauen (German), fiducia (Italian) or vertrouwen (Dutch). Regarding the answer categories, we find two very similar dispersions when both are linearly rescaled from 0 to 10, with means of 5.5 and 5.2 and standard deviation of 2.5 and 2.8, respectively. Still, the 11-category item has a somewhat lower spread, so we use the standardized z-scores based on our total sample of Muslim citizens in Europe. Alternatively, we could rescale linearly (see above) or non-linearly to two 4-category variables.4 The correlation between these options is .970 and .998. Using these alternatives does not lead to substantially different results. Table 1 provides the descriptive statistics on our variables.

Table 1.

Descriptive statistics.

NMinimumMaximumMeanStd. dev.
Core micro-level variables      
Trust towards parliament (z-values) 5,267 -2.18 1.78 -.00 1.00 
Strong distrust1  0 1 .12  
Distrust1  0 1 .20  
Neither distrust, neither trust1  0 1 .18  
Trust1  0 1 .29  
Strong trust1  0 1 .21  
Mosque attendance 5,267 1.20 .95 
Never  0 1 .24  
Irregularly  0 1 .45  
Weekly  0 1 .18  
Multiple times a week  0 1 .13  
Religious identification (z-values) 5,267 -4.06 1.18 .01 .97 
How religious are you?2 4,359 0 10 7.19 2.37 
Are you a religious person?2 894 0 2 1.84 .40 
How important is God in your life?2 899 0 9 7.68 2.16 
Praying outside religious services 5,267 2.02 1.11 
Core macro-level variables      
Hostile public attitudes 207 4.13 7.55 6.08 .67 
Disproportionality of parliament 207 .63 21.95 5.17 4.86 
Restrictions on political participation of migrants 207 80 35.87 20.00 
Excluding citizenship and equality laws 207 9.00 65.33 34.19 16.76 
Control variables      
General public's trust in parliament 207 -2.68 1.99 .04 .98 
Female (ref. = no) 5,267 .45 .50 
Age (in years) 5,267 15 80 35.62 13.50 
Migration generation      
Second generation 5,267 .30  
First generation 5,267 .65  
Unknown 5,267 .05  
Education level      
Lower than secondary 5,267 .20  
Secondary education 5,267 .57  
Tertiary education 5,267 .23  
Main daily activity      
Employed 5,267 .48  
In education 5,267 .16  
Other 5,267 .36  
Relationship status      
Never married 5,267 .27  
In durable relationship 5,267 .60  
Other (incl. divorced/widowed) 5,267 .12  
NMinimumMaximumMeanStd. dev.
Core micro-level variables      
Trust towards parliament (z-values) 5,267 -2.18 1.78 -.00 1.00 
Strong distrust1  0 1 .12  
Distrust1  0 1 .20  
Neither distrust, neither trust1  0 1 .18  
Trust1  0 1 .29  
Strong trust1  0 1 .21  
Mosque attendance 5,267 1.20 .95 
Never  0 1 .24  
Irregularly  0 1 .45  
Weekly  0 1 .18  
Multiple times a week  0 1 .13  
Religious identification (z-values) 5,267 -4.06 1.18 .01 .97 
How religious are you?2 4,359 0 10 7.19 2.37 
Are you a religious person?2 894 0 2 1.84 .40 
How important is God in your life?2 899 0 9 7.68 2.16 
Praying outside religious services 5,267 2.02 1.11 
Core macro-level variables      
Hostile public attitudes 207 4.13 7.55 6.08 .67 
Disproportionality of parliament 207 .63 21.95 5.17 4.86 
Restrictions on political participation of migrants 207 80 35.87 20.00 
Excluding citizenship and equality laws 207 9.00 65.33 34.19 16.76 
Control variables      
General public's trust in parliament 207 -2.68 1.99 .04 .98 
Female (ref. = no) 5,267 .45 .50 
Age (in years) 5,267 15 80 35.62 13.50 
Migration generation      
Second generation 5,267 .30  
First generation 5,267 .65  
Unknown 5,267 .05  
Education level      
Lower than secondary 5,267 .20  
Secondary education 5,267 .57  
Tertiary education 5,267 .23  
Main daily activity      
Employed 5,267 .48  
In education 5,267 .16  
Other 5,267 .36  
Relationship status      
Never married 5,267 .27  
In durable relationship 5,267 .60  
Other (incl. divorced/widowed) 5,267 .12  

Notes: 1As z-values are standardized, they give little background information on the spread and scale of trust in parliament. Therefore, this descriptives table includes descriptives of a categorical variant of the trust variable. While it provides a general grasp of the dispersion on the variable, it should be interpreted carefully as the categorization is partly artificial. The survey items with five categories are copied to these categories as is; those with four answering categories do not have a middle answer and were recategorized in either the distrust or trust categories here; finally, those with 11 categories were recoded as follows: 0–2; 3–4; 5; 6–7; 8–10. 2Of these three items, the z-values are used to create the final index; the scores here allow for a more substantive descriptive assessment of the level of religious identification, while the descriptives in the index should be used for model interpretation.

Three dimensions of Islamic religiosity

Regarding Islamic religiosity, we focus on mosque attendance, affective religiosity (particularly religious identification) and individual prayer. Each of three included data sources includes items that tap these concepts. Below we describe how we created the three indices used in the regression models. Not only are they conceptually different, they are so empirically too. The correlations between the three indices is between .34 and .50—well below the standard correlation cut-off in terms of multicollinearity (i.e. >.7/>.8). To illustrate, of those who pray daily or more frequently, over half do not go to mosque weekly or more. And among those who hardly ever pray, about half visit the mosque weekly or more often.

Mosque attendance is asked with three different but highly similar items across surveys (see Online Appendix B). All questions are variations on “How often do you attend religious services?” and have seven or eight answering options ranging from never to daily. Across surveys we could regroup the answers to capture 0 (never to less than yearly), 1 (yearly to monthly), 2 (weekly) and 3 (more than weekly). Each of the categories represents at least 13% of our sample and close to 700 cases or more. Comparing results using an interval variant of this variable with those using a set of four dummy variables shows that the interval variable captures the relationship well. These results are presented below.

For respondents’ degree of identification with their faith, all surveys include a standard item measuring the self-identified religiousness of a respondent (see Online Appendix B). In EVS and WVS, the number of answering categories (three) is very limited, but both surveys also include another item that conceptually fits affective religiosity (see Glas, 2020) and has a more fine-grained answering scale (10-point): the importance they attach to their God. Empirically, the discussed items also fit together, as other studies have illustrated that they tap one underlying concept (Spierings, 2019). In our data we find that from the lowest to the highest of three categories of religiousness, the average on the 0-to-9 variable on god's importance is respectively 1.1, 5.8 and 8.1. This lends further support for our decision to take them together into a single indicator for the more abstract concept of degree of identification.

As the three items used were of different types, not each available across all surveys, and had different ranges, we could not recode the variables as was done for attendance. Building on the same logic as calculating the relative impact in a regression model via standardized Beta-coefficients, we calculated the z-values of respondents on each of the items (all of them having at least 1,200 respondents in our master Muslim-only database) and took the average of the available standardized scores. Doing so, each and every respondent is ascribed a score on identification relative to the other respondents, taking into account that some survey items might return higher scores because of the formulation or options offered. The final variable does not allow precise descriptive analyses, but it does enable us to assess the relationship between affective religiosity and political trust.

For individual praying, we used the recategorizing approach as discussed for attendance. In all three surveys the question stem reads, “How often do you pray?” EVS and ESS explicitly added “apart from or outside of religious services” to the question, while WVS made this distinction clear in the answering option (see Online Appendix B). Across surveys, all with seven or eight answering options, we have been able to distinguish the following: 0 ([practically] never), 1 (infrequent), 2 (at least weekly) and 3 (daily or more often).

Macro level: exclusionary environments

Our focus on exclusionary contexts encompasses formal political and informal societal exclusion, whereby, given existing macro-level data, we try to capture different manifestations of hostile public attitudes, disproportionality of parliament, restrictions to migrants’ political participation, and exclusionary citizenship and equality laws. A factor analysis on the 10 final variables used (see Online Appendix C) indeed shows this clustering, supporting that we combine our five public attitudes items and our three citizenship and equality laws in two respective indices. The items related to disproportionality and political participation did not clearly load on those factors and are included singularly.

First, hostile public attitudes were based on the European Social Survey data (2002–18; ESS). Per country-year we calculated the average score on the full general nationally representative dataset for five items: allow many/few immigrants of (i) different race/ethnic group from majority/(ii) poorer countries outside Europe; (iii) immigration bad or good for country's economy; (iv) country's cultural life undermined or enriched by immigrants; and (v) immigrants make country worse or better place to live. To calculate the overall index, the 4- and 11-category means were linearly rescaled to run from inclusive (0) to exclusionary (10) and averaged. Missing country-year scores were interpolated linearly if data were available for the same country prior to and after the missing data point.

Second, disproportionality of the parliament (i.e. a stronger discrepancy between the percentage of votes parties receive and their percentage of seats in parliament) has been shown to lead to more skewed (descriptive) representation (Wängnerud, 2009), signifying political exclusion in terms of underrepresentation of Muslim minority citizens (on which no direct data across years and countries are available). We used the least square index as provided by Gallagher (1991). For election years we took the score after the election and used that for all years up until a new election. A higher score indicates a higher degree of disproportionality. Of the macro-level variables, this is the only one that has a distribution that is neither normal nor evenly spread; however, using a rescaled variable based on the natural logarithm (i.e. leading to a better distribution) leads to the same conclusions (see Online Appendix I for the details).

Third, restrictions on the political participation of migrants are captured in the widely used Migrant Integration Policy Index (MIPEX) data (Solano and Huddleston, 2020). The index makes use of expert surveys and covers migrant voting rights, the strength and presence of migrant consultative bodies, and funding of immigration organizations—in other words, enfranchising, informing, consulting and involving people with a migration background. While, evidently, not all Muslims are migrants, almost all Muslims in the countries studied here will have at least one migrant parent. We reversed the 0–100 index so that a higher score indicates more exclusion. MIPEX presents data as of 2009, but it turned out to be rather stable within countries over the years; therefore, we imputed the missing years by the country average for a country.

Last, we combined three MIPEX 0–100 indices on citizenship and equality laws, with a higher score indicating more exclusion, imputed the same way as discussed above. The three items focus on access to permanent residency, access to nationality and having effective legal protection against anti-discrimination.

None of the correlations between the macro-level variables is higher than .8 (or .6 for that matter), making multicollinearity issues highly unlikely. For more information, see Online Appendix I, which reports and discusses the correlations as well as additional robustness checks.

Control variables

The ESS, EVS and WVS all provide data on core socio-economic, demographic and migration-related control variables for the study at hand; these variables are often linked to trust (Uslaner, 2002; Newton, Stolle and Zmerli, 2018). By recategorizing these as, discussed above, we include the following variables.

Age is measured in years. Gender was included by a dummy indicating identification as female. We included migration status by considering whether the respondent was born in the country of residence (so-called first generation), in another country (second generation) or whether this is unknown (which we include in order to not lose respondents). Regarding education, we distinguish between no education, primary education, secondary education and tertiary education. For main daily activity, we make a distinction between being employed (for a considerable number of hours), being in education, or being non- or unemployed. Lastly, relationship status distinguishes married/partnered, never married and other (including divorcees and widows).

At the macro level, we included the level of parliamentary trust among the general public for each country-year as a control. This measure is calculated by taking the weighted average of the original full surveys used here. As discussed in detail in the results section and Online Appendix I, we also ran the core models without this variable. Overall this approach leads to highly similar conclusions. Of the country-year outcomes, we took the standardized values (see above on the dependent variable).

Modelling strategy

In line with the data structure, and given our goal of assessing the context-dependency of micro-level relationships, we estimate three-level multilevel models: individuals nested in country-years nested in countries, with random intercepts for both country-year and country. To account for survey differences and temporal trends, we include dummies for the different source surveys and a linear time variable. The main models have been rerun for women and men (Table 2) and for first- and second-generation respondents (Online Appendix D) separately. This analysis assesses whether results hold across groups or whether they apply to a specific gender for instance (see theory section). To assess Hypothesis 5, we include cross-level interaction terms in the later models. As recommended (Giesselmann and Schmidt-Catran, 2019; Heisig and Schaeffer, 2019), the slope of the individual-level religiosity variables is part of the random part of the model at the country-year level for the interaction models, which prevents false positives on the interaction terms.5 Where these models suggest meaningful interaction effects, we have graphically depicted them based on the models presented in Online Appendix E, by plotting the relationship between the religiosity dimension and trust for different levels of the macro-level variable with confidence intervals for the estimated level of trust (panel a in Figures 1 and 2) as well as the effect of the religiosity dimension for the different levels of the macro-level variable, with the confidence interval of the estimated effect (panel b).
Table 2.

Multilevel regression models estimating the impact of Islamic religiosity on trust in national parliament among self-identified Muslim citizens in Western Europe (2002–20).

Model 3a,Model 3b,
Model 1Model 2women onlymen only
Fixed effects     
MICRO-LEVEL VARIABLES OF INTEREST     
Mosque attendance (0-3) -.02 -.02 .02 -.05* 
Religious identification .02 .02 .01 .03 
Individual praying -.03* -.03* -.04# -.02 
MICRO-LEVEL CONTROL VARIABLES     
Female (ref. = no) -.03 -.02 – – 
Age (in years) .00 .00 -.003# .004* 
Migration generation (ref. = 2nd)     
First generation .27*** .27*** .28*** .27*** 
Unknown .29** .29** .24* .35** 
Education level (ref. = low)     
Secondary education -.06 -.06 -.17** .02 
Tertiary education .03 .03 -.06 .09 
Main activity (ref. = employed)     
In education .22*** .23*** .11# .34*** 
Other -.01 -.01 .05 -.08# 
Relationship status (ref. = never married)     
In durable relationship .09* .09* .03 .11# 
Other (incl. divorced/widowed) -.00 .00 -.01 -.00 
MACRO-LEVEL VARIABLES     
Hostile public attitudes .04 -.10# .04 .02 
Disproportionality parliament .00 -.01 -.01 .01 
Restrictions political participation migrants -.003# -.005* -.00 -.005* 
Excluding citizenship and equality laws -.00 -.00 -.00 -.00 
Trust in parliament (general population) .25***  .22*** .26*** 
SYSTEM VARIABLES     
Time (in years; 2002 = 0) .01** .01** .01# .02** 
Source survey (ref = ESS)     
WVS -.14 -.14 -.29* .05 
EVS -.01 -.02 -.08 .03 
Intercept -.42 .51 -.22 -.48 
Random effects     
Country-level intercept .02# .02# .02# .01 
Country-year-level intercept .01* .02*** .01 .01 
Model statistics     
BIC 14,571.406 14,613.053 6,484.792 8,153.148 
Nindividual 5,267 5,267 2,876 2,391 
Model 3a,Model 3b,
Model 1Model 2women onlymen only
Fixed effects     
MICRO-LEVEL VARIABLES OF INTEREST     
Mosque attendance (0-3) -.02 -.02 .02 -.05* 
Religious identification .02 .02 .01 .03 
Individual praying -.03* -.03* -.04# -.02 
MICRO-LEVEL CONTROL VARIABLES     
Female (ref. = no) -.03 -.02 – – 
Age (in years) .00 .00 -.003# .004* 
Migration generation (ref. = 2nd)     
First generation .27*** .27*** .28*** .27*** 
Unknown .29** .29** .24* .35** 
Education level (ref. = low)     
Secondary education -.06 -.06 -.17** .02 
Tertiary education .03 .03 -.06 .09 
Main activity (ref. = employed)     
In education .22*** .23*** .11# .34*** 
Other -.01 -.01 .05 -.08# 
Relationship status (ref. = never married)     
In durable relationship .09* .09* .03 .11# 
Other (incl. divorced/widowed) -.00 .00 -.01 -.00 
MACRO-LEVEL VARIABLES     
Hostile public attitudes .04 -.10# .04 .02 
Disproportionality parliament .00 -.01 -.01 .01 
Restrictions political participation migrants -.003# -.005* -.00 -.005* 
Excluding citizenship and equality laws -.00 -.00 -.00 -.00 
Trust in parliament (general population) .25***  .22*** .26*** 
SYSTEM VARIABLES     
Time (in years; 2002 = 0) .01** .01** .01# .02** 
Source survey (ref = ESS)     
WVS -.14 -.14 -.29* .05 
EVS -.01 -.02 -.08 .03 
Intercept -.42 .51 -.22 -.48 
Random effects     
Country-level intercept .02# .02# .02# .01 
Country-year-level intercept .01* .02*** .01 .01 
Model statistics     
BIC 14,571.406 14,613.053 6,484.792 8,153.148 
Nindividual 5,267 5,267 2,876 2,391 

Notes: Nyear = 17; Ncountry-year = 207. ***p < .001, **p < .01, *p < .05, #p < .1 (macro-level effects and split model only).

Figure 1a.

Estimated relation between attendance and trust towards parliament for different levels of “hostile public attitudes,” with 95% confidence intervals.

Note: Only countries with at least 100 cases are included here.

Figure 1a.

Estimated relation between attendance and trust towards parliament for different levels of “hostile public attitudes,” with 95% confidence intervals.

Note: Only countries with at least 100 cases are included here.

Close modal

As there are restrictions in terms of the effective number of variables that can be added to our models at the contextual level and that the results at this level might be at risk of being influenced by outlier cases, we (i) run the macro-level models adding one macro-level variable at a time and (ii) in the case of finding potential interaction effects, rerun the direct effect models per country for the countries with sufficient respondents and plot the B-coefficients for the respective religiosity dimension against the macro-level variables (panel c in Figures 1 and 2). If results are similar across these approaches, we can draw rather robust conclusions on the existence of context-dependent effects. Details are provided in the results section where relevant.

Control variables and overall model

Before turning to testing the hypotheses, let us briefly assess the results for the control variables and model as a whole. The level of trust in parliament among the general population is a strong explanatory factor of the trust among Muslim citizens (e.g. Table 2, Model 1). Importantly, political trust and for instance anti-migrant attitudes are clearly related among general populations (Kokkonen and Linde, 2023). Consequently, the strong effect of macro-level trust might indicate that to some extent we now overcontrol for the macro-level exclusionary context variables. Therefore, we also ran Model 1 as well and the interaction models without this variable (see Model 2 and Online Appendix I), and we will return to these results where relevant below.

Our time variable suggests a rather stable increase in trust among Muslim citizens in Western Europe, and this result remains if only time is included as an explanatory variable or when it is estimated curvilinearly.6 Substantively, this result contrasts an often mentioned decline in political trust. We find a positive trend among one of the least welcomed communities across Western Europe, a result that deserves further analysis.

At the micro level, most notably, we do not find a clear positive impact of education, which is found in a majority of studies on high-income consolidated democracies (Mayne and Hakhverdian, 2017). However, also in Mayne and Hakhverdian's overview there are a few studies (on the general population) that find no significant results, and among Muslim populations in the Middle East a negative relationship with (generalized) trust has been found as well (Spierings, 2019). This result indicates the relevance of analyzing minoritized ethnic groups in Western countries to shed more light on why and how antecedents exert influence.7 Also, we find that so-called second-generation citizens are less trusting than first-generation citizens; this result is in line with previous research studying the evaluation of public institutions (Röder and Mühlau, 2012).

While these results are not at the core of our study, we highlighted them as they help to assess our data and model as well as support the larger claims underlying our study: bringing together the political trust and literature on minority inclusion processes provides fruitful ground for better understanding the causes and mechanisms, also stressing that we should be careful in extending conclusions of “general population” studies in Western Europe to specific minority groups (who are part of those general populations), with education offering a prime example.

Average effects for the religiosity dimensions

In Models 1 and 2, we do not find a statistically significant relationship between mosque attendance and political trust. Hypothesis 1 can be considered falsified, the more so because zooming in shows a nuanced picture, with negative associations between Islamic religiosity being present under specific circumstances.

Particularly, Model 3b shows a statistically significant, albeit substantially limited, negative relationship between mosque attendance and trust in parliament among men. Additional models, including attendance as a set of dummies, show that citizens who attend mosque weekly are less trusting than those who go less often and that particularly men who go multiweekly have lower trust. Among women this result is not found; it deserves notice that fewer than one in five women in our sample visit the mosque weekly or more often, whereas this number is over two in five among men. The particularities of women's mosque attendance, including not having access to mosques in their neighborhood (Nyhagen, 2019), might interfere with the mechanisms at work.

Also, as will be discussed in more detail below, we find that the attendance-trust link is conditioned by context. In contexts where the public is more hostile we find negative effects, while in the least hostile contexts we even find positive effects.

The results for the degree of religious identification are rather clear: we find no systematic relationship, not in the overall model (Table 2), not when included separately, non-linearly or only using ESS with the 11-point identification variable (in additional analyses),8 and not in only the more inclusive or more exclusive societies (see Online Appendix E). Thus, overall Hypothesis 2 needs to be rejected. Still, it might be possible that the discussed counteracting mechanisms might be at play and balance each other out.

Regarding the link between prayer and political trust, our results are conditionally supporting Hypothesis 3, predicting a negative relationship (Table 2). The negative effect is reproduced across models with similar B-coefficients (among men and women, see Table 2, and migrant generation, see Online Appendix D). The clearest negative and significant effect is found among women. In combination with the gender-split results for attendance, this might suggest that praying taps into similar mechanisms as attendance (e.g. indicating a stronger group consciousness and discrimination as well as commitment to more conservative guidelines that are not enacted by parliament). Women who do not attend mosque frequently might revert to praying, and indeed on daily praying women score 8 percentage points higher than men in our data.

Context-level exclusion effects

Acknowledging that exclusion can take different shapes, we studied four different manifestations added together in the model (Table 2) as well as one by one. However, before adding those, the macro-variance is estimated with only time and the general publics’ trust being included. The country-year-level variance was significant (p = .025), and the variance at the country level had a p-value of .051 (see Table 2, Model 1), which we consider meaningful given that 17 countries are included.

Part of the contextual variance in trust in parliament among Muslims can be ascribed to our indicators of hostility. Clearly not all tested relationships are statistically significant, but those that are all show negative effects. Also, comparing Model 1 with Model 2 shows that it matters whether the trust levels of the general public are included. With trust levels included, the existence of more exclusionary laws and policies towards the migrant participation—arguably the factor most closely linked to the political domain—shows a negative relationship to Muslims’ trust in parliament. The coefficient seems small, but the variable's scores run from 0 to 80. Without trust among the general population included (Model 2), the effects of the macro-level variables turn more strongly negative, and we also find that the public’s level of anti-migrant attitudes relate negatively to trust. Moreover, only including this substantive macro-level variable shows a negative relationship (b = .16) that is significant (p < .01). We should be careful in drawing strong conclusions on this result, but given the inter-relatedness of political trust and anti-migrant attitudes among the larger population, it deserves consideration that macro-level trust partly overcontrols for the impact of anti-migrant populations. Put differently, there is some support for Hypothesis 4, whereby we find indications of both formal and informal boundaries playing a role in Muslims’ political trust, on top of the overall political trust in a certain context.

Do exclusionary contexts aggravate religiosity's impact?

Lastly, we theorized that the contextual exclusion might also shape the way in which individual-level religiosity relates to Muslims’ political trust (Hypothesis 5). As shown in Online Appendix E, only 1 out of 12 tested interaction effects is statistically significant, regardless of testing them separately or together. In other words, the impact of religiosity dimensions on political trust is rather stable across contexts studied here, largely falsifying our hypotheses on cross-level moderations.

Which effect did show significant variation by level of exclusion? The results for attendance indicate a particularly negative impact of attendance in contexts with more hostile native populations (interaction effect is negative with p < .05;9 see Online Appendix E), as is depicted in Figure 1a, showing the downward slope for a context 1 standard deviation more hostile than average. Figure 1b additionally shows the effect across the range of contexts in terms of public hostility. As shown, the effect of attendance is negative and statistically significant (the confidence interval, also presented in the figures, does not include 0) from around a hostility level of 6.5, which is around half a standard deviation above the mean context on public hostility (see Table 1). The positive coefficient for lower hostility contexts is not significant (and less hostile contexts do not exist in our sample; see Table 1). Lastly, Figure 1c shows the B-coefficient of attendance (vertical axis) against the level of public hostility (horizontal axis) per country (each marker is a country), based on re-estimating the model for attendance per country separately.10 This also shows that in a more hostile environment, a stronger negative impact of frequent attendance exists. The only three cases with a positive B-coefficient (above 0 vertically) are among the most inclusive countries (on the left of the panel). This includes Denmark, which based on the figure could be considered one of two potentially influential or outlier cases; the other being Sweden in the upper left-hand corner. Without these two cases, the results remain to show a negative significant interaction effect (p = .018).
Figure 1b.

Estimated effect of attendance on trust per level of hostility, with 95% confidence intervals.

Figure 1b.

Estimated effect of attendance on trust per level of hostility, with 95% confidence intervals.

Close modal
Figure 1c.

B-coefficient of attendance on trust towards parliament per country, plotted against the score on “hostile public attitudes.”

Figure 1c.

B-coefficient of attendance on trust towards parliament per country, plotted against the score on “hostile public attitudes.”

Close modal
Given the gendered effect of attendance on trust, we also estimated the interaction model for men and women separately. The interaction term is negative for both (B = -.07 and -.05 for men and women, respectively), but only statistically significant for men (p < .1). Figure 2 shows plots similar to those discussed above. For men, the negative effect is a bit more pronounced than in general (as shown by comparing the slopes of Figure 1a and Figure 2a), and there is a statistically significant negative effect around being reached in contexts that score about 6 on public hostility (i.e. the mean; Figure 2b). Also, the country-based plot (Figure 2c) shows that the Danish case is no longer an outlier. In brief, the negative attendance effect in hostile contexts seems more certain and stronger among men. Noteworthy here is also that the only two per-country (marginally) significant effects that were found among women were both positive relationships between attendance and trust, and both in relatively less exclusive countries (Denmark and Sweden).
Figure 2a.

Estimated relation between attendance and trust towards parliament for different levels of “hostile public attitudes,” among men only, with 95% confidence intervals.

Figure 2a.

Estimated relation between attendance and trust towards parliament for different levels of “hostile public attitudes,” among men only, with 95% confidence intervals.

Close modal
Figure 2b.

Estimated effect of attendance on trust per level of hostility, among men only, with 95% confidence intervals.

Figure 2b.

Estimated effect of attendance on trust per level of hostility, among men only, with 95% confidence intervals.

Close modal
Figure 2c.

B-coefficient of attendance on trust towards parliament per country, plotted against the score on “hostile public attitudes,” among men only.

Figure 2c.

B-coefficient of attendance on trust towards parliament per country, plotted against the score on “hostile public attitudes,” among men only.

Close modal

All in all, there is some indication that attendance's negative influence on trust is particularly present among men and in contexts in which the public is more hostile towards migrants. At the same time, among women we find some indications that attendance might actually have a positive effect in the least hostile environments. Given the sparsity of clear interaction results, we do not corroborate Hypothesis 5, but fully refuting it also seems too harsh, given the results discussed above. In the conclusion, we come back to the interpretation and implication of this set of results.

This study has contributed to the scarce amount of research studying the association between Islamic religiosity and political trust (Hsiung and Djupe, 2019). We do so by studying Muslims in Western Europe and offer two key contributions. First, instead of comparing Muslims with another group and assuming differences can be ascribed to Islamic religiosity or studying the impact of religiosity as similar across religions,11 we study differences among Muslims in Western Europe and distinguish several dimensions of Islamic religiosity to study its importance for trust in national parliaments. Second, we take a cross-national approach to find overall patterns while realizing that political trust is subject to its surroundings, and in doing so illustrate how certain exclusionary policies and practices on the national level affect Muslims’ political trust.

In general, we find that Muslims’ political trust is lower in countries where public attitudes towards migrants are more hostile and where political participation is restricted for some. Muslim men who more often attend a mosque experience on average lower trust in national parliament, and even more so when the public is more hostile towards migrants. Among women, a negative association between praying and trust is found. Surprisingly, we find no association between religious identification and trust in parliament. Below, we reflect on what this combination of results implies for the theories and literature brought together in this study.

With respect to mosque attendance, we could initially conclude that the negative relationship we found among men does not support the social capital literature, at least in terms of mosque attendance being considered a form of civic activity or integration, although bonding capital could of course have been strengthened by mosque attendance. However, considering the indication we found of context-dependency, considering the more general observation that the West European context has been rather hostile towards Islam at least since 2001 (Vermeulen, 2018), and considering that we even found some positive relationships in the least hostile contexts all draw out a more nuanced story that might bear larger significance in the political trust literature.

Among men, we find support for the idea that integration in the religious community mobilizes grievances towards parliament, which is mostly likely to work via consumption of what is preached at the mosque or exchanged in communication with others in the context of mosque attendance. Additionally, and also in line with the negative effect of praying for women, the discrepancy between the moral convictions in the country of residence and those dominant in the “average” mosque or among devout prayers seems to feed into some distrust, too. Part of this story might also be that the often praying women are actually more likely to be partnered with men who attend mosque often. Considering the strong interhousehold political socialization (Iyengar, Konitzer and Tedin, 2018; Stoker and Jennings, 1995), this theory would further support our interpretation. However, it also flags the importance of more research taking gendered effects, including partner effects, into account and trying to disentangle the group consciousness about marginalization from the impact of dissatisfaction with the moral direction a country takes.

Moreover, the gendered effects we found, including a positive impact of attendance in the least hostile environments for women, suggest that attending mosque also leads to the transfer of civic skills and political knowledge, particularly among women who attend regularly. These women are likely to be more conservative and potentially less involved in networks that bridge minority groups and general society, including politics. Then, in a relatively low-hostility setting, mosques might provide them with political information, which outweighs the negative effect between devout Muslims and parliament's policies due to moral discrepancies.

Phrased more generally, and referring back to the larger literature on political trust, we find some support for the positive effects mentioned in the social capital literature (De Vroome, Hooghe and Marien, 2013; Putnam, 2000). Yet for a problematized and minoritized religious group, these effects seem to be largely overruled by a stronger impact of mosque attendance via mechanisms of group consciousness and experiences of marginalization that are assumed to be communicated via mosque attendance. The latter realization actually can be considered a manifestation of the privilege or winner-loser mechanism (Newton, Stolle and Zmerli, 2018; Röder and Mühlau, 2012). The negative formulation here seems to be the dominant process in Western Europe today when it comes to linking Islamic religiosity, and mosque attendance in particular, to political trust. This logic also points to new questions for other minorities and how they relate to the political system: the winner-loser logic often focused more on the winners and less on the marginalized groups. Our results suggest that for other minorities that are mobilized in organizations, similar processes are likely. Moreover, people belong to or identify with different groups and marginalization is an intersectional process, raising questions like whether this process works differently for different ethnic groups, such as Muslim people with a Middle Eastern or a Bosnian background. Such ethnic differences, and complexities more generally, were impossible to study here but do deserve more attention in future work.

Part of our starting point and interpretation above stresses the importance of the context. This context is partly comparable across Western Europe, yet it partly varies between the countries and years studied. Besides indications of an interplay between public attitudes and mosque attendance, we also show that restriction in the political participation of ethnic minorities and the general publics’ trust level translate to the political trust of Muslims in a country. At the same time, our results suggest that not all forms of such exclusionary policies and practices have the same power and that they relate to general trust in society, which deserves further study among Muslim citizens specifically. Again, this might apply to other groups as well. For instance, the social boundary of public opinion on LGBTIQ+ people might also inform LGBTIQ+ citizens’ political trust. We hope to have provided directions for studying the interactions between national contexts and minority members in shaping their political attitudes.

Still, we should acknowledge that we found only a little interplay between Islamic religiosity on the individual level and the differences in exclusionary boundaries in European societies, which is in line with a recent study conducted on how such contextual factors shape the link between Islamic religiosity and voting (intentions; Kollar, Geurts and Spierings, 2024). The limited power on the macro level may be part of why there are hardly any significant interactions between these two, although previous studies with a similar number of cases have illustrated the context-dependency of religiosity on trust in other regions and on other outcomes in Europe (Glas and Alexander, 2020). A missing link here might be that of subjective experiences. Indeed, previous research has illustrated that it is not so much the objective level of exclusion that intersects with Islamic religiosity, but that individual perceptions of exclusion are key (Röder and Spierings, 2022). We recommend future research to dive into this issue, as well as into other underlying mechanisms that may explain findings like why prayer relates negatively to political trust, for which qualitative research is key. Our study has illustrated that Islamic religiosity matters for political trust, but the results are neither clear-cut nor all-explanatory. Its importance is gendered and contextualized: we found a negative impact of attendance among men and of praying among women. These findings are both explainable by a focus on group consciousness of exclusion indicating Muslims’ non-belonging in society, which may strengthen moral convictions that are ill-fitting in Western European societies.

No AI-assisted technologies were used to produce this manuscript.

We hereby confirm that data used for this manuscript are publicly available. A replication package including the syntax used for this study is available online on OSF: https://doi.org/10.17605/OSF.IO/PNWYX

We have no conflicts of interest to disclose.

Using publicly available data, no ethics review was required for this study.

This paper was written in the context of the NWO-funded research project SPIRiT: the socio-political impact of Islamic religiosity among Muslim citizens in Europe today (project identification: VI.Vidi.191.023).

Supplemental data for this article can be accessed online at https://doi.org/10.1162/euso_a_00002.

1

Additional analyses, as reported and discussed in Online Appendix F, using the same data sources and modelling strategy show that the effects of the different religiosity variables are substantively different for Muslims and Christians in Western Europe, which above all warrants an intracategorical approach, studying Muslims as a heterogenous group.

2

The analyses have been rerun per country to assess whether differences exist and whether these can be linked to methodological differences between the surveys (e.g. different items used to measure a concept). These analyses are discussed in more detail in Online Appendix H. They show that only one of nine potential differences between the survey sources regarding religiosity was statistically significant. More importantly, this difference could not be linked to methodological issues, and the variation found did not undermine the overall patterns found and conclusions drawn.

3

Exact codes are available online on OSF: https://doi.org/10.17605/OSF.IO/PNWYX.

4

Rescaling: 0–2 = 0; 3–4 = 1; 5 = 1.5; 6–7 = 2; 8–10 = 3.

5

Based on Giesselmann and Schmidt-Catran (2019), one might argue that when testing cross-level interactions in models nesting individuals in repeated cross-sectional multicountry surveys, it is also wise to check whether adding the slope of the country-year variable in the random part of the country level of the model matters to avoid false positives (i.e. Type II errors). As we find a potentiation cross-level interaction only between public hostility (varies at the country-year level) and attendance, we also re-estimated that model including the slope of public hostility in the random part at the country level. The cross-level interaction remains statistically significant (p = .018 instead of p = .015).

6

Including as nine categories gives the following coefficients (ref: first two years): .06; .09; .17; .04; .14; .15; .15; .26. Quadratic term, if added is insignificant (p > .8).

7

Explanations for finding no positive effect included being socialized in authoritarian regimes (see Spierings, 2019) and being more aware of ethno-religious-based exclusion, often perceived more by the higher educated (Geurts, 2022). Mayne and Hakhverdian (2017) do indeed theorize this more.

8

Only including identification: p = .67. Adding a quadratic term (Table 2): p-values > .45. Estimation of Model 2 on ESS only: p = .249.

9

Also p < .05 when only this interaction term is included.

10

This also allows the control variables’ impact to vary between countries, creating an additional robustness test: finding the same result decreases the risk of a Type II error. Countries > 100 respondents are included.

11

Online Appendix F shows that the effects of the dimensions studied here do indeed differ by religion.

Bègue
,
L.
(
2002
) ‘
Beliefs in justice and faith in people: just world, religiosity and interpersonal trust
’,
Personality and Individual Differences
,
32
(
3
), pp.
375
382
. https://doi.org/10.1016/S0191-8869(00)00224-5
Ben-Nun Bloom
,
P.
and
Arikan
,
G
. (
2012
) ‘
A two-edged sword: the differential effect of religious belief and religious social context on attitudes towards democracy
’,
Political Behavior
,
34
(
2
), pp.
249
276
. https://doi.org/10.1007/s11109-011-9157-x
Bird
,
K.
,
Saalfeld
,
T.
and
Wüst
,
A.
(
2011
)
The political representation of Immigrants and minorities: voters, parties, and parliaments in liberal democracies
.
New York
:
Routledge
.
Bobowik
,
M.
et al. (
2017
) ‘
“Healthy” identities? Revisiting rejection-identification and rejection-disidentification models among voluntary and forced immigrants
’,
European Journal of Social Psychology
,
47
(
7
), pp.
818
831
. https://doi.org/10.1002/ejsp.2306
Calhoun-Brown
,
A.
(
1996
) ‘
African American churches and political mobilization: the psychological impact of organizational resources
’,
Journal of Politics
,
58
(
4
), pp.
935
953
. https://doi.org/10.2307/2960144
Collins
,
K.
and
Owen
,
E.
(
2012
) ‘
Islamic religiosity and regime preferences: explaining support for democracy and political Islam in Central Asia and the Caucasus
’,
Political Research Quarterly
,
65
(
3
), pp.
499
515
. https://doi.org/10.1177/1065912912450554
De Koning
,
M.
(
2013
) ‘
How should I live as a “true” Muslim? Regimes of living among Dutch Muslims in the Salafi Movement
’,
Etnofoor
,
25
(
2
), pp.
53
72
.
De Vroome
,
T.
,
Hooghe
,
M.
and
Marien
,
S.
(
2013
) ‘
The origins of generalized and political trust among immigrant minorities and the majority population in the Netherlands
’,
European Sociological Review
,
29
(
6
), pp.
1336
1350
.
Dilmaghani
,
M.
(
2019
) ‘
Perceived religious discrimination and confidence in Canadian institutions
’,
Ethnic and Racial Studies
,
42
(
15
), pp.
2603
2622
. https://doi.org/10.1080/01419870.2018.1539505
Doerschler
,
P.
and
Jackson
,
P. I.
(
2012
) ‘
Do Muslims in Germany really fail to integrate? Muslim integration and trust in public institutions
’,
Journal of International Migration and Integration
,
13
(
4
), pp.
503
523
.
Eggert
,
N.
and
Giugni
,
M.
(
2011
) ‘
The impact of religion on the political participation of migrants
’, in
L.
Morales
and
M.
Giugni
(eds.)
Social capital, political participation and migration in Europe
.
London
:
Palgrave
, pp.
219
237
.
Eickelman
,
D. F.
and
Piscatori
,
J.
(
1996
)
Muslim politics
.
Princeton
:
Princeton University Press
.
European Social Survey Cumulative File, ESS 1-9
(
2020
)
Data file edition 1.0. NSD - Norwegian Centre for Research Data, Norway - Data archive and distributor of ESS data for ESS ERIC
.
Round 7: file: ESS1-7e01 (no doi available
);
Round 8: Available at
: https://doi.org/10.21338/NSD-ESS8-2016;
Round 9: Available at
: https://doi.org/10.21338/NSD-ESS9-2018
EVS
(
2020a
) ‘
European values study longitudinal data file 1981–2008 (EVS 1981-2008)
’.
GESIS Data Archive, Cologne
.
ZA4804 Data file version 3.1.0
. Available at: https://doi.org/10.4232/1.13486
EVS
(
2020b
) ‘
European values study 2017: integrated dataset (EVS 2017)
’.
GESIS Data Archive, Cologne
.
ZA7500 Data file version 4.0.0
. Available at: https://doi.org/10.4232/1.13560
Gallagher
,
M.
(
1991
) ‘
Proportionality, disproportionality and electoral systems
’,
Electoral Studies
,
10
(
1
), pp.
33
51
. https://doi.org/10.1016/0261-3794(91)90004-C
Geurts
,
N.
(
2022
)
Puzzling pathways: the integration paradox among migrants in Western Europe
.
Nijmegen
:
Radboud Repository
.
Giesselmann
,
M.
and
Schmidt-Catran
,
A. W.
(
2019
) ‘
Getting the within estimator of cross-level interactions in multilevel models with pooled cross-sections: Why country dummies (sometimes) do not do the job
’.
Sociological Methodology
,
49
(
1
), pp.
190
219
. https://doi.org/10.1177/0081175018809150
Glas
,
S.
(
2020
).
Where are the Muslim feminists? Religiosity and support for gender equality in the Arab region
.
Nijmegen
:
Radboud Repository
.
Glas
,
S.
and
Alexander
,
A.
(
2020
) ‘
Explaining support for Muslim feminism in the Arab Middle East and North Africa
’,
Gender & Society
,
34
(
3
), pp.
437
466
.
Heisig
,
J.
and
Schaeffer
,
M.
(
2019
) ‘
Why you should always include a random slope for the lower-level variable involved in a cross-level interaction
’,
European Sociological Review
,
35
(
2
), pp.
258
279
. https://doi.org/10.1093/esr/jcy053
Hooghe
,
M.
and
Zmerli
,
S.
(
2011
) ‘
Introduction: the context of political trust
’, in
M.
Hooghe
and
S.
Zmerli
(eds.)
Political trust: why context matters
.
Colchester
:
ECPR Press
, pp.
1
12
.
Hsiung
,
B. O.
and
Djupe
,
P. A.
(
2019
) ‘
Religion and the extension of trust
’,
Political Behavior
,
41
(
3
), pp.
609
631
. https://doi.org/10.1007/s11109-018-9466-4
Inglehart
,
R.
et al. (eds.)
2020
. ‘
World values survey: all rounds – country-pooled datafile
’.
Madrid, Spain and Vienna, Austria
:
JD Systems Institute and WVSA Secretariat
.
Available at
: http://www.worldvaluessurvey.org/WVSDocumentationWVL.jsp
Isani
,
M.
and
Schlipphak
,
B.
(
2017
) ‘
In the European Union we trust: European Muslim attitudes toward the European Union
’,
European Union Politics
,
18
(
4
), pp.
658
677
. https://doi.org/10.1177/1465116517725831
Iyengar
,
S.
,
Konitzer
,
T.
and
Tedin
,
K.
(
2018
) ‘
The home as a political fortress: family agreement in an era of polarization
’,
Journal of Politics
,
80
(
4
), pp.
1326
1338
. https://doi.org/10.1086/698929
Jamal
,
A.
(
2005
) ‘
The political participation and engagement of Muslim Americans: mosque involvement and group consciousness
’,
American Politics Research
,
33
(
4
), pp.
521
544
. https://doi.org/10.1177/1532673X04271385
Jeannet
,
A. M.
(
2020
) ‘
Immigration and political distrust in Europe: a comparative longitudinal study
’,
European Societies
,
22
(
2
), pp.
211
230
. https://doi.org/10.1080/14616696.2019.1694162
Just
,
A.
,
Sandovici
,
M. E.
and
Listhaug
,
O.
(
2014
) ‘
Islam, religiosity, and immigrant political action in Western Europe
’,
Social Science Research
,
43
, pp.
127
144
. https://doi.org/10.1016/j.ssresearch.2013.09.005
Kaase
,
M.
(
1999
) ‘
Interpersonal trust, political trust and non-institutionalised political participation in Western Europe
’,
West European Politics
,
22
(
3
), pp.
1
21
. https://doi.org/10.1080/01402389908425313
Kokkonen
,
A.
and
Linde
,
J.
(
2023
) ‘
A nativist divide? Anti-immigration attitudes and diffuse support for democracy in Western Europe
’,
European Journal of Political Research
,
62
(
3
), pp.
977
988
. https://doi.org/10.1111/1475-6765.12551
Kollar
,
R.
,
Geurts
,
N.
and
Spierings
,
N.
(
2024
) ‘
Religiosity and electoral turnout among Muslims in Western Europe
’,
European Political Science Review
,
16
(
3
), pp.
393
412
. https://doi.org/10.1017/S1755773923000334
Kroknes
,
V. F.
,
Jakobsen
,
T. G.
and
Grønning
,
L. M.
(
2015
) ‘
Economic performance and political trust: the impact of the financial crisis on European citizens
’,
European Societies
,
17
(
5
), pp.
700
723
. https://doi.org/10.1080/14616696.2015.1124902
Kumlin
,
S.
and
Rothstein
,
B.
(
2010
) ‘
Questioning the new liberal dilemma: immigrants, social networks, and institutional fairness
’,
Comparative Politics
,
43
(
1
), pp.
63
80
. https://doi.org/10.5129/001041510X12911363510394
Liebkind
,
K.
and
Jasinskaja-Lahti
,
I.
(
2000
) ‘
The influence of experiences of discrimination on psychological stress: a comparison of seven immigrant groups
’,
Journal of Community and Applied Social Psychology
,
10
(
1
), pp.
1
16
. https://doi.org/10.1002/(SICI)1099-1298(200001/02)10:1%3c1::AID-CASP521%3e3.0.CO;2-5
Mayne
,
Q.
and
Hakhverdian
,
A.
(
2017
) ‘
Education, socialization, and political trust
’, in
S.
Zmerli
and
T.
van der Meer
(eds.)
Handbook on political trust
.
Cheltenham
:
Edward Elgar
, pp.
176
196
.
Marien
,
S.
(
2011
) ‘
The effect of electoral outcomes on political trust: a multi-level analysis of 23 countries
’,
Electoral Studies
,
30
(
4
), pp.
712
726
. https://doi.org/10.1016/j.electstud.2011.06.015
Maxwell
,
R.
(
2010
) ‘
Trust in government among British Muslims: the importance of migration status
’,
Political Behavior
,
32
(
1
), pp.
89
109
. https://doi.org/10.1007/s11109-009-9093-1
Moutselos
,
M.
(
2020
) ‘
Praying on Friday, voting on Sunday? Mosque attendance and voter turnout in three West European democracies
’,
Journal of Ethnic and Migration Studies
,
46
(
11
), pp.
2275
2292
. https://doi.org/10.1080/1369183X.2019.1662283
Nannestad
,
P.
(
2008
) ‘
What have we learned about generalized trust, if anything?
’,
Annual Review of Political Science
,
11
, pp.
413
436
. https://doi.org/10.1146/annurev.polisci.11.060606.135412
Newton
,
K.
,
Stolle
,
D.
and
Zmerli
,
S.
(
2018
) ‘
Social and political trust
’, in
E. M.
Uslaner
(ed.)
The Oxford handbook of political and social trust
.
London
:
Oxford University Press
, pp.
37
56
.
Noordzij
,
K.
,
de Koster
,
W.
and
van der Waal
,
J.
(
2021
) ‘
The micro–macro interactive approach to political trust: quality of representation and substantive representation across Europe
’,
European Journal of Political Research
,
60
(
4
), pp.
954
974
. https://doi.org/10.1111/1475-6765.12436
Nyhagen
,
L.
(
2019
) ‘
Mosques as gendered spaces: the complexity of women's compliance with, and resistance to, dominant gender norms, and the importance of male allies
’,
Religions
,
10
(
5
), pp.
321
336
. https://doi.org/10.3390/rel10050321
Oskooii
,
K. A. R.
(
2020
) ‘
Perceived discrimination and political behavior
’,
British Journal of Political Science
,
50
(
3
), pp.
867
892
. https://doi.org/10.1017/S0007123418000133
Oskooii
,
K. A. R.
and
Dana
,
K.
(
2018
) ‘
Muslims in Great Britain: the impact of mosque attendance on political behaviour and civic engagement
’,
Journal of Ethnic and Migration Studies
,
44
(
9
), pp.
1479
1505
. https://doi.org/10.1080/1369183X.2017.1330652
Patrikios
,
S.
and
Huhe
,
N.
(
2022
) ‘
Neither religious nor rational: heterodoxy and institutional trust
’,
West European Politics
,
47
(
2
), pp.
255
279
. https://doi.org/10.1080/01402382.2022.2145101
Prickett
,
P. J.
(
2015
) ‘
Negotiating gendered religious space: the particularities of patriarchy in an African American mosque
’,
Gender and Society
,
29
(
1
), pp.
51
72
. https://doi.org/10.1177/0891243214546934
Putnam
,
R.
(
2000
)
Bowling alone: the collapse and revival of American community
.
New York
:
Touchstone
.
Röder
,
A.
and
Mühlau
,
P.
(
2012
) ‘
Low expectations or different evaluations: what explains immigrants' high levels of trust in host-country institutions?
’,
Journal of Ethnic and Migration Studies
,
38
(
5
), pp.
777
792
. https://doi.org/10.1080/1369183X.2012.667986
Röder
,
A.
and
Spierings
,
N.
(
2022
) ‘
What shapes attitudes toward homosexuality among European Muslims? The role of religiosity and destination hostility
’,
International Migration Review
,
56
(
2
), pp.
533
561
. https://doi.org/10.1177/01979183211041288
Simonsen
,
K. B.
(
2016
) ‘
How the host nation's boundary drawing affects immigrants’ belonging
’,
Journal of Ethnic and Migration Studies
,
42
(
7
), pp.
1153
1176
. https://doi.org/10.1080/1369183X.2016.1138854
Sobolewska
,
M.
et al. (
2015
) ‘
Understanding the effects of religious attendance on political participation among ethnic minorities of different religions
’,
European Journal of Political Research
,
54
(
2
), pp.
271
287
. https://doi.org/10.1111/1475-6765.12081
Solano
,
G.
and
Huddleston
,
T.
(
2020
) ‘
Migrant integration policy index (MIPEX) 2020
’.
Barcelona/Brussels
:
CIDOB and MPG
.
Spierings
,
N.
(
2019
) ‘
Social trust in the Middle East and North Africa: the context-dependent impact of citizens’ socio-economic and religious characteristics
’,
European Sociological Review
,
35
(
6
), pp.
894
911
. https://doi.org/10.1093/esr/jcz038
Spierings
,
N.
,
Geurts
,
N.
and
Kollar
,
R.
(
2024
) ‘
European Societies Geurts & Spierings 2024
’.
Syntax for data harmonization and analyses of trust in parliament using ESS, EVS, WVS, parts A and B
.
Nijmegen
:
Radboud University
,
Radboud Social Cultural Research
,
SPIRiT project
. Available at: https://doi.org/10.17605/OSF.IO/PNWYX
Spierings
,
N.
and
Vermeulen
,
F.
(
2024
) ‘
Minding the political engagement gap: how discrimination and religion impact Dutch voters with an immigrant background
’,
Acta Politica
,
59
, pp.
619
642
. Available at: https://doi.org/10.1057/s41269-023-00286-7
Stoker
,
L.
and
Jennings
,
M. K.
(
1995
) ‘
Life-cycle transitions and political participation: the case of marriage
’,
American Political Science Review
,
89
(
2
), pp.
421
433
. https://doi.org/10.2307/2082435
Superti
,
C.
and
Gidron
,
N.
(
2022
) ‘
Too old to forget: the dynamics of political trust among immigrants
’,
Political Studies
,
70
(
3
), pp.
624
654
. https://doi.org/10.1177/0032321720980899
Uslaner
,
E. M.
(
2002
)
The moral foundations of trust
.
Cambridge
:
Cambridge University Press
.
Van der Meer
,
T.
(
2010
) ‘
In what we trust? A multi-level study into trust in parliament as an evaluation of state characteristics
’,
International Review of Administrative Sciences
,
76
(
3
), pp.
517
536
.
Van der Meer
,
T.
(
2017
) ‘
Democratic input, macroeconomic output and political trust
’, in
S.
Zmerli
and
T.
van der Meer
(eds.)
Handbook on political trust
.
Cheltenham
:
Edward Elgar
, pp.
270
284
.
Vermeulen
,
F.
(
2018
) ‘
The paradox of immigrant political participation in Europe amidst crises of multiculturalism
’, in
C.
Menjívar
,
M.
Ruiz
and
I.
Ness
(eds.)
The Oxford handbook of migration crises
.
Oxford
:
Oxford University Press
, pp.
801
817
.
Voicu
,
B.
and
Tufiş
,
C. D.
(
2017
) ‘
Migrating trust: contextual determinants of international migrants’ confidence in political institutions
’,
European Political Science Review
,
9
(
3
), pp.
351
373
. https://doi.org/10.1017/S1755773915000417
Wängnerud
,
L.
(
2009
). ‘
Women in parliaments: Descriptive and substantive representation
’,
Annual Review of Political Science
,
12
, pp.
51
69
. https://doi.org/10.1146/annurev.polisci.11.053106.123839
Wisneski
,
D. C.
,
Lytle
,
B.
and
Skitka
,
L. J.
(
2009
) ‘
Gut reactions: moral conviction, religiosity, and trust in authority
’,
Psychological Science
,
20
(
9
), pp.
1059
1063
. https://doi.org/10.1111/j.1467-9280.2009.02406.x
Zmerli
,
S.
and
Newton
,
K.
(
2007
) ‘
Trust in people, confidence in political institutions and satisfaction with democracy
’, in
J. W.
van Deth
,
J. Ramón
Montero
and
A.
Westholm
(eds.)
Citizenship and involvement in European democracies: a comparative analysis
.
London
:
Routledge
, pp.
34
65
.
This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) license, which permits unrestricted use, distribution, and reproduction in any medium, provided the use is non-commercial and the original work is properly cited. For a full description of the license, please visit https://creativecommons.org/licenses/by-nc/4.0/legalcode.

Supplementary data