Generalised Trust among Second-Generation Muslim and Non-Muslim Minority Groups in Europe

This study compares generalised trust between second-generation Muslim and non-Muslim migrant groups in Europe, and examines the effect on trust of discrimination and cultural transmission. Analysis of data from the European Social Survey of 4,687 respondents in 32 European countries shows that second-generation Muslim groups have lower levels of trust than second-generation non-Muslim minority groups. The findings provide no evidence that Muslims’ lower levels of trust are due to discrimination and exclusion. Rather, results indicate that the differences are due to cultural transmission: Muslim groups originate more often from low-trust societies, and gener-alised trust is transmitted from one generation to the next


Introduction
In recent decades, the integration of immigrants in European countries has been a hot topic of debate in both academia and society (Drouhot and Nee, 2019).Because many immigrants originate from Muslim countries (van Tubergen and Sindradottir, 2011), researchers have devoted specific attention to Muslim migrant groups in Europe.Studies have been carried out on, among other subjects, religious change among Muslims (Voas and Fleischmann, 2012;Drouhot, 2021;Friberg and Sterri, 2021), cross-group friendships, and intermarriage (Kalmijn and van Tubergen, 2006;Carol, 2016;Leszczensky and Pink, 2017), cultural values (Bisin et al., 2008) and inequality in the labour market (Koopmans, 2016;Witteveen and Alba, 2018).This study contributes to this growing literature on Muslim migrant groups in Europe by examining their levels of 'generalised trust' .This concept has received considerable attention in the sociological literature (e.g., Putnam, 1993;Uslaner, 2002;Bjørnskov, 2007), and is typically captured with the wellknown question: 'Generally speaking, would you say that most people can be trusted, or that you can't be too careful in dealing with people?' (Nannestad, 2008;Dinesen and Sønderskov, 2018).A key element in this concept is that it relates to a general sense of trust in people, i.e., trust that goes beyond the own group and relates to members of other groups, strangers and institutions more generally.
Various scholars have argued that generalised trust is a core indicator of social cohesion in society, which has positive outcomes.For example, individuals with more generalised trust tend to be more tolerant of others, donate more to charity and volunteer more (Uslaner, 2002).Also, high-trust societies experience more economic growth (Zak and Knack, 2001), less institutional corruption (Nannestad, 2008), and less violent crime (Lederman, Loayza, and Menendez, 2002).Some scholars, therefore, argue that generalised trust is one of the core foundations of a well-functioning society (e.g., Putnam, 1993;idem, 2000).
In this study, we contribute to previous literature by studying in more depth the generalised trust found among 'second-generation' Muslim minority groups, i.e., those who are children of immigrants.This group is particularly interesting to study because they have been born and raised in Europe but have also inherited cultural values and practices from their parents.We examine two lines of argumentation.First, it can be argued that these second-generation Muslims have experienced more discrimination than second-generation members of other minority migrant groups (Savelkoul et al., 2011), which may undermine their generalised trust.If this is the case, we expect to see a clear divide between Muslim and non-Muslim minority groups in their levels of generalised trust.Second, it could also be the case that generalised trust is driven by cultural transmission, i.e., cultural traits transmitted from one generation to the next.If this is the case, differences between Muslim and non-Muslim groups in levels of generalised trust disappear once the level of trust in their country of origin is considered.
This study elaborates on and empirically tests these two explanations ('experiences of discrimination' and 'cultural transmission').We use data from the European Social Survey, a cross-national survey of 4,687 respondents from second-generation Muslim and non-Muslim minority groups.

Theory
A key idea in the literature on trust is 'experiential theory' (Dinesen and Hooghe, 2010;Dinesen, 2012;Nannestad et al., 2014), which argues that generalised trust is at least partly experience-based.This means that trust is an evaluation of experiences of others in a specific setting.When people repeatedly feel they are treated fairly, either personally or as a group, their generalised trust will improve.Conversely, when people perceive themselves to be treated unfairly, feeling, for example, that they are discriminated against, or that the group to which they belong is somehow excluded, they develop lower trust in others (Brehm and Rahn, 1997;Schildkraut, 2005;Uslaner 2008;Rothstein and Stolle, 2008;Dinesen and Hooghe, 2010).
Elaborating on this idea, one could argue that, in Europe, second-generation Muslim minority groups differ from non-Muslim minority groups in their generalised trust.First, studies show that Muslim minority groups are strongly discriminated against in Western labour markets (Bartkoski et al., 2018) and face significant barriers in finding employment and earnings (Adida, Laitin, and Valfort, 2016).Many Muslims report experiencing discrimination (Kislev, 2018), and anti-Muslim sentiments are widespread in the populations of Western countries (Strabac and Listhaug, 2008;Strabac, Aalberg, and Valenta, 2014;Savelkoul et al., 2011).Scholars have therefore argued that, in the European context, Muslim boundaries are particularly salient (Alba, 2005;Alba, and Foner, 2015).We argue that these discriminatory sentiments and actions impact the development of trust among the children of immigrants who belong to such Muslim groups.We, therefore, hypothesise that: In Europe, second-generation Muslim groups have lower levels of generalised trust than second-generation non-Muslim groups (H1).
Scholars have argued that some Muslim groups may experience more discrimination than other Muslim groups.This is because some Muslim groups are larger in size, which makes them more visible in the host country.According to group threat theory (Blumer, 1958;Blalock, 1967;Stephan and Stephan, 2000), ethnic majority members perceive larger groups as more threatening (Coenders, Gijsberts, and Scheepers, 2004a;eisdem, 2004b;Freeman, 2009), particularly if the cultural values and practices of these bigger groups are different from mainstream culture (Coenders, 2001;Scheepers, Gijsberts, and Coenders, 2002).The large group of Germans living in Austria, for example, are hardly ever noticed by majority members, and neither are the sizeable migrant group of Germans in the Netherlands.Similarly, smaller Muslim groups may not often encounter discrimination, but the larger Muslim groups in society (e.g., Turks in Germany), are often exposed to discrimination.We, therefore, hypothesise, that: In Europe, the size of the migrant group has a stronger negative effect on generalised trust among second-generation Muslim groups than among non-Muslim groups (H2).
We test these hypotheses -derived from experiential theory -against a competing idea.According to this line of thought, what matters is the generalised trust inherited from the parents and the wider community.Muslim minority groups may have lower levels of trust than non-Muslim minority groups because there is a lower level of generalised trust in their parents' country of origin.Previous studies have shown that immigrants born and raised in lowtrust societies tend to have lower levels of generalised trust.Such 'origin effects' are even present among the second-generation, i.e., when immigrant parents originate from a low-trust society, their children, on average, have lower levels of trust (Uslaner, 2008;Dinesen, 2013;Ljunge, 2014) This suggests that generalised trust is a cultural trait that is transmitted from one generation to the next within the migrant group.We test the idea that Muslim minority groups may have lower levels of trust because there are lower levels of trust in the countries from which they come.We, therefore, hypothesise the following: In Europe, lower levels of generalised trust among second-generation Muslim minority groups disappear once the level of trust in the (parents') country of origin is considered (H3).

Data
For the analyses of this study, we use wave 2-91 of the European Social Survey (ESS), which cover the period between 2004 and 2018.The ESS is a nationally representative cross-sectional survey collected in 32 different countries in Europe.Our sample consists of the second generation, i.e., we excluded those born abroad and excluded those whose parents were not born abroad.To make sure we could clearly define the country of origin in our sample, we only included respondents whose parents were both born in the same foreign country.We also excluded Israel from the ESS sample (as our focus is on Europe), and Turkey (because the context for Muslim migrants is entirely different).

Operationalization 3.2.1
Dependent Variable The dependent variable in this study, generalised trust, was measured with the question: 'Generally speaking, would you say that most people can be trusted, or that you can't be too careful in dealing with people?'This question was measured on an eleven-point scale ranging from 0 'You can't be too careful' to 10 'Most people can be trusted' .Previous research has also used this specific question (Nannestad, 2008;Dinesen and Sønderskov, 2018), and it has been shown to be invariant across countries and cultures (Reeskens and Hooghe, 2008;Freitag and Bauer, 2013).

3.2.2
Muslims in Migrant Community (%) In this study, we compare Muslim with non-Muslim minority groups.This means that we are not focusing on people's subjective identity and their individual religious affiliation, but rather on their membership of a (non)Muslim minority group, regardless of their individual religious practice or belief.Dividing migrant groups into a simple dichotomy of Muslim versus non-Muslim groups would be great a simplification.In some groups, for example, over 90% identify as Muslim, whereas in other groups, this is 70% or 30%.Therefore, rather than using a dichotomy, we measure the percentage of Muslims in each migrant group.A 'migrant group' is defined as the specific combination between the country of origin of an immigrant (or that of his/ her parents), and his/her country of residence (van Tubergen, Maas, and Flap, 2004).Those of Turkish origin who moved to Germany are a migrant group, for example, and so are Pakistanis living in the UK.
To estimate the proportion of Muslims in each migrant group, we aggregated individual-level data from the ESS.In the ESS, respondents were asked to report their self-identified religion.Using this variable, we calculated the proportion of first-and second-generation members in each migrant community that considered themselves Muslim.Because some migrant groups are very small, with only a few respondents in the ESS, we implemented two measures to reduce measurement error.First, we pooled waves 2-9 to increase the number of respondents per migrant group.Second, we excluded groups with less than 20 first-and second-generation migrants present in our dataset.This resulted in 220 migrant groups, of which 102 groups had no Muslims, and 24 groups had a very high percentage of Muslims in their community (80% or more), such as the Turkish and Moroccan groups in the Netherlands and Germany, the Pakistani group in the UK and the Kosovar group in Slovenia.

3.2.3
Group Size The relative population size of each migrant group was also measured using ESS data.Specifically, we counted the number of first-and second-generation members of each migrant group as a percentage of the total number of ESS respondents in the destination country.Again, we pooled all the ESS waves used in this study.Our measurement shows a strong correlation (r=.923, p<.001) with another measure of group size, which we constructed by combining data on migrant group size from the 'International Migration Outlook 2017' (OECD, 2017) and data on the total (migrant and non-migrant) population from the 'World Population Prospects' (United Nations, 2017).Because the variable was highly left-skewed, we took the logarithm.

3.2.4
Generalised Trust in the Country of Origin To examine the role of cultural transmission and the impact of inherited trust, we developed two measures.First, we look at the average generalised trust in the country of origin of the respondents' parents.To construct this measure, we used all surveys from the World Value Survey (WVS) and European Value Survey (EVS) up to 2014.In the WVS and the EVS, respondents were asked the same question about generalised trust as in the ESS, but they had to choose between 'You can't be too careful' or 'Most people can be trusted' .Although generalised trust can vary over time, previous studies have shown that on the country-level, they remain rather stable (Uslaner, 2002).Therefore, the measurement error resulting from pooling 12 years of data is limited.We computed the proportion of the population who stated that most people can be trusted.This practice has most commonly been used in previous literature (e.g., Soroka, Helliwell, and Johnston, 2003;Dinesen, 2012).
A potential drawback of this measure is that immigrants from the same country of origin may have different levels of trust, depending on the destination country in which they settled.For example, it might be that Moroccans who migrated to France had higher levels of trust than Moroccans who settled in the Netherlands.Consequently, second-generation Moroccans born in France inherit more trust from their migrant group than Moroccans born and raised in the Netherlands.
We therefore created an even more direct measure of the generalised trust that is embedded in the migrant group in which children of immigrants are socialised.For this, we looked at the average generalised trust among the foreign-born members of the migrant group.To construct this variable, we put the generalised trust question to the foreign-born respondents in the ESS.Thus, for each migrant group, we include the average generalised trust among the first generation.

3.2.5
Control Variables We control for several variables.First, we take into account the average generalised trust in the destination country.Scholars have documented several 'destination effects' , i.e., that immigrants and their children change their trust levels depending on characteristics of their receiving society.For example, immigrants who migrate to high-trust societies, and societies with less corruption, develop more generalised trust, although this process has only been examined among first-generation migrants (Rothstein, 2000;Dinesen, 2012;Nannestad et al., 2014;Voicu, 2014;Helliwell, Wang, and Xu, 2016;Shaleva, 2016).To construct this measure, we use only those respondents who have no migration history, i.e., they and their parents were all born in the destination country.At the individual level, we take into account well-known correlates to generalised trust.We controlled for respondents' education, measured by the 'International Standard Level of Education' (ISLED) scheme (Schröder and Ganzeboom, 2014).We also take into account gender, age, whether the respondent has citizenship in the destination country, and ESS wave.
Since generalised trust can be highly volatile and unreliable in childhood (Dawson, 2019), we have excluded respondents younger than 18 from our dataset.Furthermore, we have removed respondents older than 80, to get a realistic sample range to test our hypotheses.After listwise deletion of missing data for the dependent, independent and control variables, the sample consists of 4,687 second-generation respondents in 32 countries in Europe.Their parents originate from 66 different countries, and there are a total of 220 communities2 (i.e., origin by destination combinations) present in our data.
Table 1 shows the descriptive statistics.All continuous variables are standardised before inclusion in the analyses.

3.3
Analyses To test our hypotheses, we use cross-classified multilevel regression models, in which respondents (level 1) are nested in communities (level 2), which are nested in both origin and destination countries (level 3, crossed factors).If the multilevel structure were ignored, the standard errors of the parameters would be underestimated, creating a possible bias in testing the hypotheses (Hox, Moerbeek, and van de Schoot, 2010).Because the country of origin and destination country are not nested within each other, cross-classified models were used (for more information on cross-classified models, see Goldstein and Leckie, 2011).3 3 We also examined the impact of influential cases at the migrant group level, because this is the level at which we formulated our hypotheses, and there are not many observations

Findings
Results of the multilevel models are presented in Table 2.We find a statistically significant association between the percentage of Muslims in the migrant group and generalised trust in models in which we control for gender, age, education, citizenship, ESS wave, migrant group size, and average trust among the non-migrant population in the receiving country (Model 0a).Thus, the larger the share of Muslims in the migrant group to which the second generation belongs, the lower the level of trust they have in people in general.This finding remains when we remove influential cases (Model 0b).We find, therefore, that on average, second-generation Muslim groups have lower levels of trust than non-Muslim groups.This may possibly be due to experiences of discrimination among Muslim groups.However, we do not find any support for this idea.First, we do not find that the interaction term between group size and percentage of Muslims is statistically significant, nor do we find a main effect of group size (Model 2a and 2b).This means that larger Muslim groups, which are most often targeted by xenophobia and discrimination, do not have lower levels of trust than Muslim groups that are smaller and less visible.Second, in additional models, we included a measure of average perceived discrimination at the group level, but it appears not to correlate with generalised trust, let alone explain away the difference between Muslim and non-Muslim groups (results available upon request).
Instead, we find more evidence for the idea that generalised trust is transmitted from generation to generation.Model 1a shows that the average trust among the foreign-born in the migrant group correlates positively (and significantly) with generalised trust.In other words, when the level of trust is higher among the first generation of the migrant group, the second generation of that migrant group also has higher trust levels.Comparing Models 0a and 0b with Models 1a and 1b reveals that the initial difference in generalised trust between Muslim and non-Muslim groups disappears once this variable is taken into account.This suggests that second-generation Muslim groups in Europe have lower levels of trust than non-Muslim groups, not because of discrimination and exclusion but rather as a result of cultural transmission of trust from one generation to the next.
(N=220).We used the method proposed by Belsley, Kuh, and Welsh (2005;also Snijders and Bosker, 1999;van der Meer, Grotenhuis, and Pelzer, 2010).We calculated the standardised change in estimates when we removed one migrant group at the time from our analyses.We detected 25 influential cases at the migrant group level and report the results when excluding these observations.2 Multilevel regression of generalised trust among the second generation in Europe (cont.)Discussion The societal significance of generalised trust has long been recognised by scholars.With growing levels of ethnic diversity in Western countries, the topic of trust among minority groups has become increasingly relevant.We have elaborated on previous research in this area by studying trust among secondgeneration Muslim groups in Europe -about which little is known.We compared generalised trust among 4,687 respondents of second-generation Muslim and non-Muslim groups across 32 countries in Europe.
A key finding of this study is that second-generation members of Muslim minority groups have lower levels of trust than those from non-Muslim groups.However, the evidence from our analyses suggests that this is not the result of discrimination and exclusion -as expected by experiential theory.Bigger Muslim groups, who experience more discrimination, do not have lower trust levels than smaller Muslim groups.Also, the average perceived discrimination in the migrant group is not related to generalised trust.
Instead, our study finds evidence for the role of cultural transmission.We find that, when the foreign-born members of the migrant group have higher levels of generalised trust, the second generation also has higher levels of trust.Once we take into account the levels of trust in parents' country of origin, and that among the first generation in the migrant group, the second-generation members of Muslim groups have the same level of trust as those in non-Muslim groups.
It seems, therefore, that cultural transmission is key to understanding why second-generation members of Muslim groups have lower levels of generalised trust.Lower levels of trust are more common in Muslim countries, such as Morocco, Turkey and Algeria, than in non-Muslim societies, and these culturally learned expectations are transmitted from parents to children.Even when these children are born and raised in a different country from their parents, they acquire the lower levels of trust from their parents and their first-generation ethnic community more broadly -i.e., their foreign-born family, neighbours and friends.
Adding further support to the idea of cultural transmission is the finding that the average level of trust among the majority population impacts the level of trust among the second-generation.We find that the latter 'adjust' to the trust levels among the non-migrant population of the receiving country.In very-high trust societies in Europe, such as Denmark and Norway, the second-generation develops more generalised trust than in European societies with lower levels of trust, such as Italy and Greece.This is cultural transmission, too, but in this case from the ethnic majority group to the second-generation.

Table 2
Multilevel regression of generalised trust among the second generation in Europe