Migrating trust: contextual determinants of international migrants’ confidence in political institutions

This paper considers the case of the international migrants’ confidence in political institutions, from a social embeddedness perspective on political trust. We use country-level aggregates of confidence in institutions as indicators of specific cultures of trust, and by employing data from the European Values Study, we test two competing hypotheses. First, as confidence in institutions depends on the values formed during early childhood, the international migrant’s confidence in political institutions in the current country of residency will be influenced by the confidence context from the country of origin. Second, the host country may have different norms of trust in political institutions, and a process of re-socialization may occur. Therefore, the immigrants’ confidence in institutions is influenced by two confidence contexts: one from the origin country and one from the host country. The time spent in the two cultures, along with other characteristics from these contexts, shape the interaction effects we tested in multilevel cross-classified models.


Introduction
Confidence in institutions depends on three types of determinants: the socio-economic characteristics and the personality traits of the individuals, the performance of the government, and the cultural environment (Newton and Norris, 2000;Mishler and Rose, 2001). In this paper we focus on the latter, by considering culture as the general depositary of values and behavioural patterns to be learned by citizens. Our research question is whether the culture-dependent confidence in institutions travels with international migrants or not, when they move from their society of origin ('origin') to the society of their current residency ('host').
Using insights from the sociology of values (Gundelach, 1994;Inglehart, 1997;Arts, 2011), from social comparison theory (Suls and Wheeler, 2012), and from assimilation theories (Esser, 2010), one may argue that political trust, being culturally embedded, should depend on both the heritage from the country of origin and on the influence of the host culture. This paper provides empirical evidence supporting this argument. Immigrants, as we show, are exposed to institutional and cultural influences in the host society, and they start acting according to the local social norms of having more or less confidence in institutions. The cultural context of the host society, however, does not erase what the immigrants have learned during early socialization in their country of origin. The cultural context of the origin society continues to produce effects over the immigrants' entire lifespan. We test for the presence of a positive influence of the average levels of political trust in the countries of origin and in the host societies on an immigrant's level of confidence in institutions.
Our contribution adds to a recent stream of research describing immigrants' political behaviour in the country of destination and its dependency on both the host and the origin country. This was previously shown for political and civic participation (Bucker, 2005;Ester et al., 2006;White et al., 2008;Aleksynska, 2011;Voicu and Rusu, 2012;Voicu, 2014a), but not for confidence in institutions, which up to now was analysed either from a methodological perspective (Andre, 2014) or in a single-case setting (Doerschler and Jackson, 2012). We discuss political trust from a cultural perspective, following a long-standing tradition to consider culture as a key for interpreting the roles and performance of political institutions (Almond and Verba, 1965;Dogan, 2005). We make use of cross-classified multilevel models on a large-scale data set provided by the European Values Study (EVS), allowing us to address the variation in institutionalization effects across 47 European societies and in the socialization effects for 98 societies, where the immigrants in our sample were born.
The paper proceeds as follows. In the first section we present the main argument. In the data and methodology section, we show how cross-sectional data can be used for our purposes, and we then list the variables and the controls used in our models. Then, we present the findings and comment on how the hypotheses are validated. Finally, we discuss the implications for further research and the consequences for immigrant integration.

International migrants and the cultures of confidence in institutions
In order to properly work, political institutions depend on their credibility and on how much confidence they are able to elicit from citizens. Thus, in a way, institutions are the creation of the social norms that support them. The cultural perspective regarding confidence in institutions argues that confidence is created through daily interactions and cooperative relations, which lead to a specific culture of trust, reciprocity, and cooperation, in which strong and effective institutions and organizations exist and reinforce the same culture of confidence in institutions (Newton and Norris, 2000). This perspective is close to the debates surrounding civic culture (Almond and Verba, 1965) or to the internationalization of European political culture (Dogan, 2005).
Observing periods of deep societal change is useful to better understand how confidence in institutions is embedded in the cultural environment. After democratization, regardless of the economic and institutional performance of governance, a fast and stable decline of confidence in institutions occurred, both in former communist states (Sztompka, 1999;Mishler and Rose, 2001;Tufiş, 2012) and elsewhere (Seligson, 2002;Joon, 2004;Catterberg and Moreno, 2006). This decline could also be interpreted as reflecting the lack of confidence in political institutions, which characterized these societies before their democratic transitions. The local culture of mistrust, which is relatively stable, as culture is supposed to be (Dogan, 2005;Inglehart and Welzel, 2005), was temporarily replaced by the euphoria of democratic transition, by the enthusiasm and the novelty of having free elections, freedom of expression, a more diverse media, or respect for human rights. Soon after, however, the pervasive influence of the culture of mistrust became salient again, in most cases being boosted by low economic performance and the declining quality of life. Inglehart (1999) argues that cultural shifts are responsible for the decline in confidence that traditional institutions, such as the army or the church, experienced during the past decades, an argument that focusses on the cultural nature of confidence in institutions. As any cultural trait, trusting or mistrusting institutions is learned and internalized during early socialization. In fact, it acts as a social value, depending on social norms and on material security during the formative years. According to the socialization hypothesis (Inglehart, 1997;Arts, 2011), the initial formation is persistent and produces effects over the entire lifespan. At the same time, values are also adaptive to context (Inglehart and Welzel, 2005;Welzel, 2007) and are exposed to institutional influences (Gundelach, 1994;Beck and Beck-Gernsheim, 2001;Arts, 2011).
There is no reason for confidence in institutions to remain completely resilient when exposed to different contextual forces. Regardless of the perceived efficiency of governance, personal experiences and transformations within society may lead to changes in the internalized levels of confidence in institutions. Imagine if everyone in one's social network describes political institutions as being reliable, while mass media often provides examples of people trusting institutions, along with regular, daily interactions with friends and co-workers, which include examples that marginally, but constantly, create a positive image for these institutions. The whole exposure works towards building a high level of confidence in institutions. Over time, this is likely to be gradually internalized as a personal value orientation. The opposite situation, facing an environment that is hostile to institutional confidence, should also produce effects, only this time boosting mistrust. Any intermediate situation works in a similar manner. Social comparison theory (Festinger, 1954;Suls and Wheeler, 2012) can be used to understand the role of social norms in moulding values. People constantly compare their own situation to other situations, and adjust their aspirations according to these comparisons and to what is socially desirable. At the individual level, cultural norms that characterize a society tend to impose clear guidelines and to generate values that are similar to those held by the majority of the population (Newson and Richerson, 2009).
Our argument until now is that confidence in institutions is embedded in specific cultures of confidence in institutions. One may credit political institutions with more or less trust according to personal experience, to socio-psychological characteristics, or to how one perceives governance efficiency. Behind the variation given by such factors lays a latent orientation towards political institutions, which can be considered a cultural trait. This orientation is created around values internalized during one's formative years, and it is modified through processes of adaptation to a given culture.
Within this framework, we ask if confidence in institutions changes when one suddenly changes his/her cultural context such as, for instance, in the case of international migration. Recent work on immigrants and confidence in institutions argues that political trust can be similarly measured for immigrants and native-born (Andre, 2014), that there are differences between Muslim and non-Muslim living in Germany (Doerschler and Jackson, 2012), as well as in Britain (Maxwell, 2010). Röder and Mülhau (2012) explained the higher political trust of immigrants through the difference in expectations from governmental performance. Confidence increases when facing up to lower expectations based on their experience in the country of origin.
We hypothesize a cultural effect that goes beyond performance. Migration normally leads to a decrease in the exposure to the context of confidence in institutions within the origin society and to an increase in the exposure to the context within the host society. For practical reasons we will consider migration as a simple relocation involving a single origin and a single destination. This implies that the migrant has not experienced living in any other cultures, except for the host culture and the origin culture. According to the arguments presented above, an international migrant should feel the effects of two different cultural influences.
The first influence is given by early socialization and the adjustments that occurred in the society of origin. These combine together to create a latent orientation that influences a person's confidence in institutions and which travels with the migrant to the host country. The average level of confidence in institutions in the country of origin can be interpreted as a general reflection of the values that have been internalized before migrating. Upon arriving in the host society, the immigrant's confidence in institutions is determined by the context of trust in the origin society to some extent. Once in the host society, this contextual effect will start to decrease, and the rate of this decrease will be determined by the frequency and intensity of direct contacts (visits in the origin country, mass media consumption from the origin country, discussions with relatives and friends, etc.). The second influence is derived from the exposure to the culture of confidence in institutions that characterizes the host country. This should influence the immigrant, just as the culture in the origin society has done. Nevertheless, contact between the immigrant and the host society is required and this can take any form, from employment to attending school or language classes, to interacting with locals, or to high levels of mass media consumption.
The argument presented above creates the framework for our first two hypotheses. We expect that HYPOTHESIS 1: Immigrants' confidence in institutions is positively correlated with the average levels of confidence in institutions in the origin country.
HYPOTHESIS 2: Immigrants' confidence in institutions is positively correlated with the average levels of confidence in institutions in the host country.
Similar hypotheses were formulated and tested in literature with respect to political participation (Bucker, 2005;Ester et al., 2006;White et al., 2008), civic participation (Aleksynska, 2011;Voicu and Rusu, 2012), or social trust (Dinesen, 2013). To this existing literature, we add the idea that confidence in institutions is another cultural trait that travels with the immigrants. We also discuss how the effect of the two cultural contexts changes, depending on the time spent in the two societies. According to the socialization hypothesis, if migration takes place at an older age, the immigrant would be less likely to be open to new experiences and to changing values and behaviours. If migration occurs at a younger age, however, primary socialization effects may still occur in the host society, increasing the impact of the host culture. This leads to the third hypothesis we test in this paper: HYPOTHESIS 3: The impact of the host society's culture of confidence in institutions is inversely related to the immigrant's age at the moment of migration.
Conversely, one may expect that long exposure to the institutions and the culture of the host society will overwrite the initial imprint of the origin society. This is implied by the various versions of assimilation theory, which argue that immigrants gradually adopt values, behaviours, and attitudes specific to the host society, and that the differences between immigrants and native-born individuals decrease with the time spent by the immigrant in the host society (Portes and Zhou, 1993;Alba and Nee, 2003;Esser, 2010). This leads to our fourth hypothesis: HYPOTHESIS 4: The impact of the origin society's culture of confidence in institutions is inversely related to the period of time the immigrant has spent in the host society.

Other determinants of confidence in institutions
Four other explanations of how political trust is generated are salient in the existing literature. We control their presence in order to depict the contextual effects that represent our focus. The first mechanism emphasizes the role of social trust (or interpersonal trust) in other people as a source of confidence in political institutions (Gabriel, 1995;Mishler and Rose, 2001;Zmerli and Newton, 2008). While crediting others with trust, one is more likely to trust the institutions that others create. Nevertheless, the relation may imply reversed causation as well (Brehm and Rahn, 1997); therefore, in the analysis we will treat social trust only as a covariate. A second mechanism represents an extension of the argument on the effect of cultures of confidence in institutions. Such cultures are transmitted to the population through social institutions, such as school, mass media, and voluntary organizations. For instance, the more time one has spent in the educational system, the higher the chances are of adopting the dominant culture of confidence in institutions to which one has been exposed (Kaase and Newton, 1995;Newton, 1999). Interest in politics and media consumption related to political topics might also have a positive effect on confidence in political institutions, as they increase the knowledge and understanding of how institutions work (Norris, 1996). Membership in voluntary organizations is also associated with higher levels of confidence in institutions (Keele, 2007;Maloney et al., 2008).
The third mechanism describes confidence in institutions as primarily depending on the evaluations of institutional performance (Miller and Listhaug 1999;Newton, 2006;Kelleher and Wolak, 2007) and on the performance as such (Norris, 1999). EVS does not include any items for measuring evaluations of institutional performance and, as a result, we can control for this mechanism only at the second level. For the individual level, we use a life satisfaction measure, said to represent a proxy for evaluations of institutional performance (Comşa and Tufiş, 2016) and shown to be influenced by perceptions of government performance (Whiteley et al., 2010).
Fourth, one may consider the opportunities that immigrants have for successful integration in the host societies. These depend on the electoral and political system, including the effective number of parties, being a federal of a unitary state, and the majoritarian or proportional vote (Norris, 1999;Anderson et al., 2005). They are also shaped by how institutions and individuals foster immigrant incorporation (Bloemraad et al., 2008).

Data and method
The ideal design for testing our hypotheses would be to follow individuals over a long period of time, to periodically assess their confidence in institutions, to record if they move from one society to another, and to measure the norms of political trust in the society of residence. With such a design, immigrants and non-migrants could be compared and the impact of living in a specific culture could be easily isolated. To our knowledge, such a large-scale panel data set is not available yet. Our best available solution is to use comparative large-scale surveys that include information on both political trust and the country of origin. From all the comparative surveys conducted in the past decade, only the EVS and the European Social Survey include all the information we need. We have selected the 2008 wave of EVS, because it is more comprehensive in terms of coverage: there are 47 societies, including all the European states, except for Andorra, Lichtenstein, Monaco, San Marino, and the Vatican, plus distinct samples for Northern Ireland and Northern Cyprus. The 67,786 respondents came from 154 origin countries, with 9.3% of them being foreign-born. This leads to the typical cross-classified multilevel design: cases (individuals) are nested in both origin countries and host countries, but the two second-level contexts (origin and host countries) are not hierarchically related.
Our dependent variable is built starting from the four-point scales of confidence in the three main political institutions (parliament, cabinet, and political parties) that are available in the 2008 wave of EVS. Using Mplus7, we have tested and shown that the three variables are explained by a factor of political trust that fulfils the requirements of partially scalar invariance across the 47 host societies (the results, not presented here, can be obtained from the authors). For convenience reasons, we computed the dependent variable used in our models as the average of the respondents' scores on the three questions.
The key independent variable in our study is a measure of political trust, both in the origin and in the host countries. Finding such a measure for the host countries was easy, as EVS already provides such information for the three institutions that we have considered. Finding a similar measure for all 154 origin countries was more difficult, and we had to use multiple sources of information to cover all of these countries. The 2005-06 wave of World Values Survey (WVS) provides data in this respect, but not for all origin countries. For the missing countries, we used information from the Latinobarometer, the Asian-Barometer, the Afrobarometer, and the Arab-Barometer, using data collected between 2004 and 2010. All these large-scale surveys are based on representative national samples, are well documented on their webpages, and use four-point scales to measure confidence in institutions. Unfortunately, not all of them include the three items that we use. As the three measures are highly correlated, we have assumed that using only one item is reliable enough and we have selected confidence in government as the information for our estimate for the culture of political trust in the origin countries, and we were able to do so for the host societies as well. Even after this decision, we still had to deal with the fact that the Arab-Barometers do not include measures for the trust in government, but in the prime minister, while the Afrobarometers measure confidence in the president. As all three institutions belong to the executive power, we have assumed them to be interchangeable. In the countries where WVS provided information about confidence in government, the president, and the prime minister, confidence in government was close to the estimates being obtained from the Arab-Barometer and the Afrobarometer. We concluded that these different measures are likely to indicate similar latent traits, and we can use them as more or less equivalent indicators. Thus, for origin countries we computed the percentage of respondents who declared that they have confidence in the government (or president/prime minister if the government indicator was not available) and used this as a measure of the culture of political trust in origin countries.
The effort to include immigrants in the analysis from as many sources as possible is related mainly to testing the second hypothesis. We were interested to show the impact of the host culture of political trust on confidence in institutions, irrespective of the country of origin.
The differences in measurements impose the need for various robustness checks. In order to avoid possible biases caused by the difference in measuring political trust in origin societies, we have also estimated all models on a subsample of European-born respondents. For these respondents we can use the same measure of confidence in government as indicator for the culture of trust.
Another important decision for combining the macro-level information with the micro-level data from EVS was related to the year we should consider for the macro-level data. Controlling for economic and political performance, the cultures of political trust should be relatively stable due to the early socialization hypothesis. Therefore, when controlling for age, the current average levels of confidence in institutions of the 'stayers'the natives in the emigration countries, who still live there, should reflect the culture of confidence in institutions in which their fellow immigrants grew up. The same should be valid for hosts as well.
The EVS data was collected for 41 host countries in 2008, and for the remaining six countries in 2009. We attempted to obtain data for the origin countries as close as possible to these years. For 87 origin countries, we managed to find data from the same years or from the previous ones. For 12 more origin countries, the reference year is 2006, for another two we have the data from 2005, and in one case we use data collected in 2004. For 52 origin countries we lack information about the culture of political trust. These 52 countries account for 5.6% of the 6264 immigrants in the data set, including the immigrants for which we lack information.
As indicators for economic performance, we used GDP/capita and GDP growth rate, measured during the year the EVS data was collected, either from the World Bank's World Development Indicators or from the UN statistical databases. We also added the Freedom in the World Index from Freedom House as indicator of political performance. Alternative models were run with the Polity2 indicator (Marshall et al., 2014). However, the Polity indicator, when it comes to European countries has the disadvantage of lacking variation, most countries being placed in a very narrow range, and only a few authoritarian countries standing as outliers. Therefore, we ran the main analyses using the Freedom House indicator and we checked for robustness using Polity2 as predictor as well. When controlling for quality of government both at origin and host, one also implicitly controls for the difference between the two. Just for illustrative purposes we also estimated a set of models controlling for the difference in quality of government instead of Polity2 in the country of origin (as Röder and Mühlau, 2012 did). All the results are enclosed in the supplementary file that accompany this paper and show no substantial difference with respect to our hypotheses. The findings also remain unchanged in the models that control if both societies (host and origin) are an 'old democracy' or not.
As perceptions of performance can be more important than actual levels, we have also employed an average value of such subjective representations. It has been computed for each country in the EVS data file from the item 'People have different views about the system for governing this country. Here is a scale for rating how well things are going: 1 means very bad; 10 means very good'.
The indicators for the type of political regime are dummy variables: having a majoritarian or proportional electoral system, being a federal state. When using them, the sample of host-origin countries splits in two, and the number of cases in each subsample becomes low. We preferred to control for these predictors (retrieved from the Quality of Government database), as well as for the effective number of parties (Gallagher, 2015) in separate models, assessing the changes to the results that are relevant for our hypotheses. We included these models in the supplementary file. They reveal very little, if any, difference by comparison with the main models that we discuss in the next section. The same happens when controlling for another element of the structure of opportunities given by the legislative framework of immigrant political participationmeasured through the MIPEX political participation measure (Niessen et al., 2007). The indicator includes information about electoral rights, political liberties, consultative policy elaboration, and immigrants' involvement in policy implementation in 27 European countries, for the year 2007 (Table 1).
At the individual level, we included the usual controls for socio-economic status in the models: gender, age (measured in years), and education (measured using seven ISCED categories). We also used a series of variables that have been linked in the literature to political trust: religiosity (measured via frequency of attending religious services and the importance of God), interpersonal trust (measured on a 10-point scale), life satisfaction (measured on a 10-point scale), membership in voluntary associations (number), and interest in politics (measured as interest in politics and as frequency of following politics in the media). The models we estimated also include variables relating to immigration: age at arrival in the host country, time spent in the host country, and a variable indicating whether the respondent was born in a foreign country or in the country of residence from two native-born parents, two foreign-born parents, or one native-born parent and one foreign-born parent.
Considering the cases with missing information on one or more variables, listwise deletion leads to a reduction in the total number of cases down to 56,559, out of which 4796 were foreign-born. This represents 84% of the total cases in the sample, 78% of the subsample of immigrants with known origin, and 80% of the subsample of European-born immigrants. Taking into account only cases with full information, the respondents come from 98 societies of origin.
Cross-classified models were estimated using the lmer package in R. First, we estimated the null models to assess how much of the variance is located at the second levels. In the second set of models, we included individual-level socio-economic and evaluation predictors. In the third group of models, we added the host and origin determinants in order to test the first two hypotheses. Finally, we also added in the model controls for the immigration status: being foreign-born, having one parent born abroad, and having both parents born abroad (having both parents native-born is the reference category). To test the third and the fourth hypotheses, we re-estimated the models for subsamples of immigrants, where it makes sense to control for the time spent in the host society and the age at migration. The third hypothesis requires an interaction effect between the confidence in government in the host country and the respondent's age when migrating. For testing our fourth hypothesis, we have included a cross-level interaction between the culture of confidence in institutions in the origin society and the length of time spent in the host society (also measured in years).
There are two measurement challenges that require special attention. The first refers to how we have computed the indicator for the culture of political trust in the host society, using the same items that were used to construct the predicted variables. We guarantee reliability of the findings in two ways. We separately predicted the components of the political trust indicator (confidence in government, parliament, and, correspondingly, in political parties). As the main independent variable is based solely on confidence in government, the problem is solved at least when predicting confidence in the two other components of the political area. Running the models on subsamples of foreign-born respondents also removes the problems, because we computed the independent variable considering the whole population.
The culture of political trust in the society of origin raises the second measurement issue. Our measure relies on contemporary data, while immigrants may have left the country long ago. Comparing with stayers provides a guarantee that the indicator for the culture of confidence in institutions in origin countries stands for our intended purposes, and includes a whole range of contextual experiences, even if they are collected at the time the survey was conducted, and not at the moment the migration episode took place. Still, this might be a weak assumption. Therefore, we have repeated all the models for those who migrated in the past 10 years, as a way to ensure that the contemporary culture of political trust in the country of origin is related at least to the one at the time of the migration. When controlling for the interaction effects with time spent in the society of origin and the age at time of migration, we also partly check the possibility that the impact of the early socialization vanishes over time.
Our measure of being an immigrant may also be subject to inquiry. Some respondents may have been born abroad but from parents which are native in the current society of residence. One may therefore wrongly label them as immigrants. To avoid such suspicion, we have repeated the models removing these respondents from the sample. None of the effects that we have noticed with respect to our hypotheses significantly differs in this scenario as compared with the other models.
The literature argues that confidence in political institutions is linked to the electoral cycle (Sandu, 1999;Banducci and Karp, 2003). Recently elected governments have been shown to enjoy higher levels of trust for a short period of time. Moreover, confidence in political institutions also seems to increase as new elections draw closer. In order to control for the possible cyclicity, we estimated a set of models that included predictors, both for origin countries and for host countries, measuring for the time passed from the previous elections and the time remaining until the next elections. These predictors have been constructed from Psephos -Adam Carr's Election Archivethe PARLINE database from the Inter-Parliamentary Union, and Wikipedia. Due to space constraints, we present these models only in the supplementary file, but we notice that the findings remained unchanged.

Findings
We have first considered the country-average levels of our indicator of confidence in institutions for foreign-born and native-born respondents. Across Europe, immigrants tend to adopt the culture of confidence in institutions that characterizes the host society. In most countries, there is no significant difference between immigrants and non-immigrants (at P < 0.05), the exception being three Eastern countries (Estonia, Kosovo, and Slovakia), where natives have more political trust than foreign-born respondents, and six societies (Belgium, Great Britain, Iceland, Luxembourg, Spain, and Switzerland) where the opposite holds true. The average Pearson correlation between the confidence in government in host societies for natives and for immigrants is 0.77. Comparing native-born respondents with emigrants of the same origin, the correlation drops to 0.26. In most of the countries, the results show no significant difference between the stayers and emigrants. In several Eastern societies (Azerbaijan, Belarus, Northern Cyprus, Slovakia, Slovenia, and Russia) emigrants display lower levels of political trust than those who did not migrate, while emigrants with higher levels of confidence in institutions can be found in Armenia, Belgium, Germany, Italy, Portugal, and Romania.
The empty cross-classified models we estimated on the whole population show that most of the variance resides at the individual level (0.445 in absolute terms), while the between-hosts variance is 0.068, and the between-origins variance is 0.014. This partitioning of the total variance indicates that the host and origin characteristics account for about 15% of the total variance, and there is almost five times more variance between-hosts than between-origins. This suggests that one should expect the characteristics of the host country to be better predictors of individual political trust than the characteristics of the country of origin. The other null models, estimated only on the immigrant subsamples, show similar variances.
The control variables have the expected effects. Women tend to have higher levels of confidence in political institutions, but the gender difference is rather small in size. The relationship between age and political trust is curvilinear, suggesting that political trust decreases with age in the case of young respondents, but it starts to increase with age for respondents over 40 years old. Education has a very weak significant effect in the model estimated on the whole population, but it has no significant effects in the models estimated on the immigrant subsamples. This difference may be determined by the fact that immigrants and non-immigrants have different distributions on the education variable. Life satisfaction has a significant positive effect on confidence in political institutions. One indicator of religiosity, the frequency of attending religious services, has a significant positive effect in all models. Membership in voluntary organizations has a significant and positive effect in the model estimated on the whole population, but loses its significance in the models estimated on immigrant subsamples. The two indicators of political involvement (interest in politics and mass media consumption of political information) have significant positive effects in all models, suggesting that a politically active population is beneficial for confidence in institutions. Lastly, social trust also has a significant positive effect on political trust. Table 2 displays some of the results from our analyses. Due to space constraints, we did not include the effects of the control variables, which we briefly introduced All models include controls for gender, age, age-squared, education, life satisfaction, importance of God, church attendance, association membership, mass media consumption, interested in politics, social trust. The Satterthwaite approximation was used. 'Immigrants' indicates foreign-born respondents. ***P ⩽ 0.001; **P ⩽ 0.01; *P ⩽ 0.05; #P ⩽ 0.10.
above, choosing to present here and to discuss only the models we used to directly test the four hypotheses (the full results are available from the authors upon request). The first two models presented in Table 2 are estimated on the entire sample, including immigrant and native-born respondents. The next group of models (M5 and M7) use only the subsample of immigrants. They add controls for time-related measures of immigration. Then, they are repeated for the subsample of European-born immigrants (M6 and M8). Table 2 does not show the models that include only individual-level predictors. When controlling only for the socio-economic status, the residual variance is reduced by a very modest 2%. Another 2% decrease comes from adding evaluation variables to the model, and a supplementary 1% is observed when political activism variables are included in the model. Including the characteristics of the host and origin countries in the models does not increase the variability explained at the individual level, but it helps to explain about half of the variances recorded in the null model between-origins and between-hosts.
Built upon the institutionalization argument, our second hypothesis is supported by all the models. People living in a culture of high political trust have a higher propensity to display confidence in institutions. One point of increase on the scale of political trust in the host society leads to an increase of about 0.6 points in political trust at the individual level. This is considerable, given the fact that the scale of political trust that we predict ranges from 1 to 4. The effect is to be found irrespectively if we run the models on immigrants only, if we include the native-born as well, or if we refer to the subsample of intra-European migrants. It is more stable than the effect of any contextual effect at the host level. If one considers the models run on immigrant subsamples, the host's culture of political trust competes only with the weaker impact of the economic output. All other contextual variables are insignificant. When contrasting against natives and stayers (M3 and M4), the widespread representation upon government performance adds as significant impact, but its effect is not as strong as the one of the culture of political trust. It might be the case that the effect of other predictors is mediated by the cultural factor, but our argument on the importance of cultures of trust remains unchanged.
Our first hypothesis is built upon the socialization argument, and refers to the second group of predictors in Table 2. According to the results, the culture of political trust that characterizes the origin society travels with the migrants and has a positive effect upon individual political trust in the host society. This is obvious when comparing with natives (M3 and M4), or when considering the smaller subsample of cross-European migrants (models 6 and 8; N = 3435). However, the relation is not significant when adding the non-European-born respondents (M5, M7age). In these models it remains positive, but becomes significant at P < 0.10 only when controlling for the time spent in the host society.
The non-significant result obtained for the subsample of all immigrants may be due to measurement issues. As we have discussed in the previous section, due to data availability issues for non-European origin countries, we have used slightly different indicators for the cultures of confidence in institutions: they come from surveys other than the EVS, and might be affected by the different positioning in the questionnaire or by the content of the questionnaire as a whole. In several cases, the indicator we used is not based on confidence in the government, but on confidence in the prime minister or in the president. A second explanation might reside in the methodology used by the EVS, which requires that questionnaires be translated into the respondent's native language, only when this language is spoken by at least 10% of the residents in the country. Immigrants that are not very fluent in the languages used in the host country would therefore have a lower probability of answering the survey. At the same time, these immigrants are precisely those who are more likely to have stronger attachments to their origin societies, and their absence from the sample would actually decrease the size of the effect. Considering all these factors, we suggest that the Hypothesis 1 would have been confirmed in these models too.
The effect of the context of trust in the origin society is weaker than the one in the host society: a variation of 1 unit on the scale of average political trust in the origin society is converted to an average expected increase of 0.13 points on the scale of the dependent variable in the models estimated on all respondents. If the model is only restricted to immigrants coming from European countries, the effect is stronger than the one based on the model estimated on the whole population, but remains half of the one given by the host's culture.
Even if it remains weaker than the culture of political trust in the host society, the culture at origin has stronger impact than any other contextual predictor in our models. At origin level, only GDP/capita also matters. Its impact is negative: those coming from more performing countries are likely to be more critical towards the political system in the host society. The impact of the performance in the society of origin is stronger than the one in the host country, but both are less important than the cultures of political trust. The latter ones also outstrip the contextual effect of the representations about government's efficiency. This stresses once more the importance of the culture of political trust.
A supplementary set of models considered only respondents that came to the host country at a maximum 10 years before the survey. The samples become smaller (734 immigrants, of which 547 are European-born), therefore some of the effects tend to become less significant. However, the impact of the cultures of political trust remains important, and all the relations described by the models in Table 2 keep their meaning. As already argued, this is of particular importance for the effect of the political trust in the country of origin, stated by the first hypothesis.
The robustness of the impact of the host culture is remarkable. It maintains significance regardless of other contextual effects that we include in the equation, and of the subsample we use to test the corresponding model. The full results, included in the supplementary file, also confirm the effect of the culture of origin, which remains the same as reported in Table 2. M4 shows no difference between foreign-born and native-born, when controlling for other effects. However, the second generation of migrants tend to have lower political trust as compared with natives. This may be due to the socialization into the culture of political trust of their parents. It may also result from unpleasant personal interactions with host institutions, or from different expectations. Testing such hypotheses would be a natural development of our argument. However, it would require setting up and explaining a set of multiple memberships combined with cross-classified multilevel models (each respondent will be nested into three non-hierarchical classifications: host, origin, multiple origins of parents). This is beyond the limited space of this paper.
For the moment, we focussed on the impact of the length of exposure and when this exposure occurs. Models 7 and 8 test for our third and fourth hypotheses. First, we have estimated the cross-level interaction between age at migration and political trust in the host society. It turns out that migrating at an older age is not associated with people being less likely to follow the norms of trust in the host society. The effect is not significant (the estimate of the P value is 0.10004). Even if it would have been significant, it would have been very weak: the marginal effect of the host's culture of confidence in institutions is deterred by 0.04 points for every 10 years of age at migration, on a dependent variable that ranges from 1 to 4. Even if the migration took place at a very advanced age, the effect would still be very small by comparison with the direct effect of the political trust in the host society. If we decompose the dependent variable in its components and predict them individually, the effect becomes significant only when it comes to confidence in parliament (P < 0.05), but it remains much weaker (b = −0.006) than the direct effect of the host's culture of political trust. Consequently, older migrants are not resilient to change, as our third hypothesis stated.
Considering the model M7time, the impact of the origin society is not resistant to exposure to other cultural influences. The cross-level interaction between the time spent in the host society and the culture of confidence in the institutions in the origin society is significant at P < 0.10. The impact is as low as that of the one reported for the other cross-level interactions that we described above. The effect is not visible when considering the European-born migrants. The effect is stronger only if considering all immigrants and just the confidence in government as dependent variable. Otherwise, the impact is maintained as non-significant. Thus, our fourth hypothesis is rather rejected by the data.
The combined consequence of considering the third and fourth hypotheses is that the dual influence of the cultures of political trust, relevant for an immigrant, depends very little on the time-related variables. It is less important how much time one spent in the origin society before migrating or how much time one has lived in the host society since migrating. The former is, in fact, just a weak measure of exposure to the culture of the origin society. People often migrate with their family and grow up in a culture specific to the origin of their parents. Quite often, immigrants marry within their own ethnic group, remaining, thus, in their origin culture. As a result, at least some immigrants continue to be exposed to the values that characterize the culture of their origin society, even if they do not live in that society anymore. The weak impact of the length of time spent in the host society suggests that the origin-host dual contextuality may apply to native-born respondents as well.
The models estimated on three distinct populations (all respondents, all immigrants, and only European immigrants) do not lead to large differences in the results. On the one hand, this suggests that our findings are quite robust. On the other, the only supplementary variables we have added are the time-related measures and their interactions with the measures for cultures of political trust. They do not help in reducing unexplained variance very much. They do not substantially change the impact of the host's and origin's culture of confidence in institutions. Moreover, the host's effect remains salient in all models. This suggests that the cultural variables measured at the origin's level, and particularly at the host's level, apply both to immigrants and to native-born residents, and supports our basic assertion that political trust is fostered by cultures of trusting political parties, parliaments, and governmental cabinets.
When treating each of the three components of political trust (confidence in parliament, in government, and in political parties) as dependent variables, the culture of the host society continues to manifest a significant positive impact irrespective of the model, proving the reliability of the finding. The culture of political trust in the country of origin exerts a significant impact for each dependent variable when considering the entire sample, and maintains this influence for the sample of European migrants. The impact of origin is significant only with respect to confidence in parliament in the subsample of all immigrants and in the subsample of recent immigrants, while in the subsample of European-born recent migrants the effect of origin is significant in the case of the aggregated index. With respect to effect size, this is similar in all models, with a higher impact in the models in which we predict confidence in parliament. On one hand, one should remember that the subsamples of recent migrants are smaller: 734, out of which 547 are European-born. This may inflate the standard errors, making the effects insignificant. However, one can be confident in the transferability of political trust from origin to host, in particular with respect to confidence in parliament, but also for the aggregate index. The exposure mechanisms remain unchanged and significant, as noticed. The interactions with age when migrating and time spent in the host society prove to be insignificant in all tested scenarios.

Discussion and implications
We have shown in this paper that, independently of economic and political performance in the host society, immigrants do not build their confidence in institutions in a vacuum, but within the context defined by the culture of political trust that characterizes the country in which they live. By living in a country where everybody is confident in political institutions, one is more likely to follow the norm of trusting and to have a higher level of political trust than what one would have had if one were living in a low-confidence society. This finding is true, not only for immigrants, but also for native-born residents as well. They are exposed to the same norms of confidence, they interact with the same people who trust the government and other institutions, and they share the same cultural values that directly affect their attitudes and behaviours towards institutions.
Our second assumption was that early socialization plays an important role in driving the individuals' attitudes towards political institutions. While the support offered by our models is not as strong as in the case of the above-discussed consequences of institutionalization hypothesis, most of our models reveal that the higher the average levels of political trust in the society of origin, the more likely the immigrants will have higher levels of confidence in institutions. This remains true, even when controlling for age and the time spent in the host society, indicating that early socialization is pervasive in influencing how immigrants interact with institutions. The effect is weaker than the one of the culture of trust characterizing the host society. Continuous exposure to the host culture is likely to be the cause. However, the available data does not allow us to investigate what happens if one lives in one country, but has a strong interaction with another culture. This might be the case for immigrant communities in which people live in ethnically homogeneous neighbourhoods, who make intensive use of internet and satellite TV to stay connected to the origin society, receive frequent visits from relatives and friends sharing a similar cultural background, or often visit their home country. Would the culture of political trust from the country of origin be more influential in such cases? Answering this question requires additional information about an individual's exposure to cultural influences and should be part of the future research agenda.
On the other hand, one needs to consider with a certain level of caution our findings. We could not control for the actual culture of political trust at the time of early socialization or at least at the time of migrating. Finding historical data for so many countries so far back in time is an impossible task. To overcome this difficulty we have employed two strategies. First, we have checked the robustness of the results by restraining the sample to the recent immigrants, with the assumption that the levels of political trust displayed by those left behind reflect the type of socialization of the migrated ones as well. Nevertheless, one may argue that it is not impossible to have different effects at 10 years after migration has occurred. The cross-level interaction with time spent in the host society partly account for such an effect. To further check if the effect does not disappear in time, one may focus on the case of fewer societies of origin, and trace back in time levels of confidence in institutions, then trying to predict current political trust of immigrants of that specific origin. This is an endeavour in itself that should be considered by future research, but goes beyond the limited space of this paper.
The cultures of political trust that are likely to influence one prove to have stronger influence than any other contextual predictor. This implies a high stability of cultures of political trust. They are the first contextual element to determine how much confidence individuals put in political institutions. In turn, this reinforces the local culture of political trust, beyond economic performances and political efficiency.
Time-related variables prove to have very little impact on the effect of the cultures of political trust. It is true that when migrating at an older age, one may seem less likely to adopt the norms of trusting institutions that characterize the host society easily. However, the effect proved to be insignificant or very weak, and it does not impede older migrants from be contaminated by the culture of the host society and to let it become more important than the culture of the origin society. One should also keep in mind the measurement that we have used in the case of the origin's culture of confidence in institutions. It is compiled, not as a reflection of the culture of trust at the moment the immigrant was socialized, but as an indicator of the current level of confidence in institutions of the stayers in the country of origin. Controlling for age, and age at migration, eliminates most of the errors generated by this measurement issue. Repeating the analysis on recent immigrants helps in the same direction. Comparing only the European-born respondents also removes uncertainty about the results. The results point in the same direction. With regard to political trust, its transferability and adaptation due to exposure to new institutional and cultural environments are less dependent on the time spent in the host society or the age when migrating. It may be the case that assimilation effects occur immediately after migration, within a timespan of months, shorter than the one that we could observe using EVS data. Acculturation may even occur in a pre-migration phase, during preparations for moving to a new society. After that, according to our findings, very little if any change occurs, except for changes in the culture of political trust of the society as a whole.
Even if we cannot tell when exactly the exposure to the new culture produces its effects, the results support the idea that immigrants tend to follow the cultural norms of political trust common in their host society. This is consistent with other scholars' arguments with respect to social trust (Dinesen and Hooghe, 2010) and membership in associations (Voicu, 2014a). When moving from one context to another, people tend to fit into their new social landscape and replicate its cultural values, including those on relating to others or to institutions. On the other hand, we have shown that they also transfer their initial culture of trusting, in the same way in which it happens with respect to trusting people (Voicu, 2014b) or civic participation (Voicu and Rusu, 2012). The outcome is a mixture of cultural influences, in which the host's ones are the more powerful.
There are four points to be aware of in interpreting our results. First, the results mainly refer to those immigrants that are included in the samples. This is a wellknown limitation of most comparative surveys: they usually either fail to include or under-represent illegal immigrants, newcomers, and sometimes non-citizens. Therefore, our results mainly refer to the remaining immigrants. However, the robustness of our findings makes us believe that they would correctly describe any foreign-born resident in the studied countries.
Second, we have assumed that countries are homogeneous enough with respect to their culture of political trust. However, this might not always be entirely true. Therefore, future research might focus on the regional cultures of confidence in institutions, at least when considering the host region. Internal migration may also be considered in this case.
Third, immigrants may build their confidence in local institutions depending on how they interact with the institutional system. This may go beyond laws and regulations, and may include personal experiences with street-level bureaucracy (Bloemraad, 2002). Although native-born also interact with the public clerks, they may be treated differently, and therefore the mere comparison native-foreign born in a comparative perspective is not enough to disentangle the effect. Data limitations leave the question unanswered for the moment.
Finally, a question of endogeneity may apply to the relation between country-level cultures of political trust and the individual-level confidence in institutions. This paper built its argument upon the socialization hypothesis and the institutionalization perspective. Both root individual-level phenomena into their context and pay less attention to the social change due to individual-level transformations. However, it might be the case that migration itself triggers changes in the culture of the host society, not necessarily in the sense suggested by the hunkering down claim (Putnam, 2007), but through bringing agents of different cultural norms into the body of the local culture of political trust. It may also impact on the culture of political trust in the society of origin through contagion due to transnational relationships. Future research may consider previous measurements of political trust in countries of origin and in the host societies, as predictors for the current levels, when controlling for the average confidence in institutions of new immigrants and the size of their flows.
Beyond these limitations and directions for future research, we believe that the findings we presented contribute to the existing knowledge in at least two ways. On one hand, they provide support and complete parts of the existing literature. The impact of the host's culture shows that assimilation theorists are right to believe that immigrants become similar to native-born individuals. The results also show that assimilation is subject to effects from the origin's culture of political trust, which never completely fades. Showing that the socialization and the institutionalization explanations do not exclude each other, and that they are complementary explanations, is useful for the emerging multilevel theory of values, even if our findings do not address values, but their consequences. On the other hand, we have documented how immigrants adapt their confidence in institutions under the impact of the cultural norms in the host country. The double contextuality is therefore confirmed for political trust. The findings add to previous work on civic and political participation and bring a broader picture on what triggers political attitudes and behaviours of foreign-born residents.

Supplementary Materials
For supplementary materials referred to in this article, please visit http://dx.doi.org/ 10.1017/S1755773915000417 370 B O G D A N V O I C U A N D C L A U D I U D . T U F I Ş