Immigrants Trust in Legal Institutions: Evidence from Israel 1

Transcription

1 Immigrants Trust in Legal Institutions: Evidence from Israel 1 Noam Gidron and Chiara Superti Abstract This paper examines the determinants and malleability of immigrants trust in legal institutions using survey evidence from Israel. While existing scholarship on variations in trust focuses on static comparisons across groups, we examine the process in which immigrants update their levels of trust over time. We argue that trust in legal institutions is shaped by the quality of government rather than by deeply-held cultural legacies. We demonsrate that immigrants from more corrupt sending countries have initial lower levels of trust, but also that levels of trust are updated following exposure to the quality of government in the receiving country. Immigrants adaptation of trust is expediated by mediating factors that facilitate greater exposure to information on political institutions in the receiving country. We exploit Israel s unique naturalization policy to compare immigrants from sending countries with different levels of corruption and with wide range of time since immigration. Keywords: Immigration, political trust, legal institutions. 1 This work benefited from the advice and comments of Jim Alt, Christopher Lucas, Justin Gest, Shelby Grossman, Horacio Larreguy, Rakeen Mabud, John Marshall, Jennifer Pan, Amanda Pinkston, Margaret Roberts, Mattan Sharkansky, and Vanessa Williamson. All mistakes are our own. We also thank James Reich for the editing suggestions. 1

2 2

3 Introduction High levels of trust in democratic institutions are associated with a bundle of positive outcomes such as increased social capital, effective democratic governance, civic and political participation, and economic growth. 1 While scholars mostly agree that trust is a vital pillar of a healthy democracy, its determinants and malleability remain disputed. Recently, research on the causes and consequences of trust has gained renewed attention in the context of the surge of immigration from poorer to richer countries. 2 Do immigrants update their levels of trust following immigration, or do they hold to the levels of trust prevalent in their country of origin? The scholarship on political trust is divided between the cultural approach, according to which trust is a rather stable inherited attribute, and the experiential approach, according to which trust is shaped by accumulated experiences with governmental actors and social networks. We seek to push forward this debate, theoretically and methodologically, by shifting the focus from static comparisons across groups to the process in which immigrants trust changes over time. Building on the premise that a component of trust can be inherited but that trust can also evolve" (Algan & Cahuc, 2010) we examine the determinants of both the inherited component and the pace in which trust is updated. Theoretically, we draw on research on exposure to political institutions (Pop-Eleches & Tucker, 2011; Rothstein & Stolle, 2011) to argue that immigrants trust in the legal system immediately following immigration reflects the quality of government at the country of origin, yet also that levels of trust are gradually updated based on exposure to state institutions in the receiving country. Immigrants legacies from their country of origin therefore set the starting point for their political attitudes but do not determine the endpoint. Based on this theoretical 3

4 perspective, we derive hypotheses for variations in trust based on the quality of government and levels of exposure to state institutions in the sending and receiving countries. Methodologically, existing scholarship has struggled with establishing the causal role of institutions in shaping levels of trust and has often lacked well-reasoned identification strategies (Algan & Cahuc, 2013). Focusing on immigration can only take scholars so far in addressing these challenges, as it is reasonable that immigrants self-select into receiving countries based on pre-existing perceptions. We believe that our analysis represents a significant methodological advancement on this front. Following Fisman and Miguel, we focus on a case where self-selection based on trust in government is minimal and where individuals from diverse backgrounds interact as equals with institutions of similar quality for different numbers of years (Fisman & Miguel, 2007). Israel s unique immigration policy presents us with such a case. Israel provides access to citizenship for all Jews and their immediate relatives. As a result, the Israeli citizenry is highly diverse compared to other OECD countries. Furthermore, due to Israel s continuous openness to (Jewish) immigration since its establishment in 1948, there is a significant variation among foreign-born citizens in the time spent in Israel since naturalization and the age at which they arrived. Especially important, and as discussed in length below, there is reason to assume that Jewish immigrants do not choose Israel as a destination primarily because of its political institutions, but rather due to religious and ideological reasons or simply because they can. Israel is thus a well-suited laboratory for studying political attitudes among citizens from countries with diverse political legacies. The empirical analysis utilizes two rounds of the European Social Survey and is focused on trust in legal institutions, which should be less affected by specific political constellations compared to trust in political parties or the parliament (Rothstein & Stolle, 2011). As a proxy of 4

5 direct exposure to institutions we use the time spent in the sending country (age at immigration) and the time spent in the receiving country (years since immigration). We also examine how factors that are expected to facilitate increased exposure to state institutions in the receiving country education and knowledge of the local language shape the adaptation of trust in the legal system. Employing a variety of statistical models, we find initial lower levels of trust among immigrants from more corrupt countries. This is true only among immigrants who arrived in Israel at an older age and were therefore more exposed to low quality institutions at the country of origin. Yet immigrants from more corrupt countries also adopt higher levels of trust the longer they are exposed to higher quality institutions and eventually converge to the levels of trust found among the native population. Immigrants who speak the local language (Hebrew) and are highly-educated are the first to adopt higher levels of trust following immigration. We thus conclude that exposure to institutions of different quality is the main factor in explaining variations in immigrants of trust in the legal system. These findings carry implications for both policy-makers and future research. If interactions with high quality institutions serve as the key mechanism for the upward adaptation of trust, then fair and impartial treatment of immigrants by state representatives has the potential for facilitating high levels of trust among often-marginalized societal groups. Theoretically, our analysis challenges deterministic notions of cultural heritage and speaks more generally to the impact of historical legacies on political attitudes (Pop-Eleches & Tucker, 2011, p. 397). Theoretical Framework Scholarship in political science, sociology and economics provides mixed theoretical expectations regarding the determinants and malleability of trust in democratic institutions. In 5

6 the context of immigration, some scholars point to long-term continuities in political attitudes; we refer to this as the cultural approach, as it is deeply-held cultural biases that are argued to perpetuate levels of trust. In contrast, others point to patterns of integration and acculturation, in which immigrants adopt attitudes that are closer to those of the native population in the receiving country; we refer to this as the experiential approach, since it is experience with local citizens and institutions that drives the adaptation of trust. The distinction between these two perspectives has become something of a consensual starting point for analyzing the determinants of trust in general and trust among immigrants in particular. 3 The cultural approach stresses the long-term stickiness of cultural norms and their transmission across generations (Putnam, 1993). Recent work in this tradition has turned to demonstrate the causal effects of historical legacies on the persistence of corruption and trust in government through a host of geographical variables (Nunn & Wantchekon, 2009; Becker, Boeckh, Hainz & Woessmann, 2011). Another strand of research focuses on variations in individual-level behaviors and attitudes among individuals from different backgrounds. Fisman and Miguel show that even when stationed in the same city, diplomats behave in a manner highly reminiscent of government officials in the home country" (Fisman & Miguel, 2007, p. 1045). Other scholars have shown strong correlations between immigrants support for redistribution and the average support for redistribution in their country of origin, arguing that such political attitudes are transmitted across generations (Luttmer & Singhal, 2011). The intergenerational transmission of trust has received much attention, especially in the context of immigration to the United States. 4 From this perspective, where your grandparents came from shapes your values. Who you are seems to matter more than who your neighbors are" (Uslaner, 2008, p.739). 5 6

7 In contrast, the experiential perspective emphasizes how relatively short-term interactions with civic and governmental institutions or social networks may leave a deep imprint on attitudes in general and trust in particular. Soss, Rothstein and Uslaner argue that interactions with welfare programs shape levels of trust in government (Soss, 1999; Rothstein & Uslaner, 2005). With regard to immigration to Western Europe, several findings also suggest that trust may be more malleable than previously acknowledged. In their agenda-setting work, Dinesen and Hooghe show that immigrants adopt levels of trust that resemble those of the native population in the receiving country (Dinesen & Hooghe, 2012). Along similar lines, Maxwell demonstrates that immigrants levels of trust are rather similar to those of the native-born population within the same subnational region (Maxwell, 2013). Dinesen also makes a strong argument in support of the experiential approach, with evidence that immigrants adopt levels of trust that are closer to those of the receiving country than those of the sending country (Dinesen, 2012). Notwithstanding the great value of these recent advancements in the literature, some questions remain unanswered. First, many of these studies focus on static comparisons across groups (immigrants, citizens of the country of origin and citizens of the receiving country or subnational region). Yet it may very well be that individuals immigrate since their levels of trust is significantly different from those found in the country of origin or significantly closer to those of the receiving country (Maxwell, 2013, pp ). Second, these aggregate analyses say less about the process in which individuals update their attitudes over time and the mediating factors that may facilitate or hinder this process. We thus know less about temporal within-individual variations in trust and the determinants of the pace in which trust is updated. Our research design, as we explain below, seeks to address these two concerns. 7

8 Exposure to Democratic Institutions and the Malleability of Trust In examining variations in trust, we focus on trust in the legal system. While trust in parliament or in parties may be colored by ideological inclinations, trust in the legal system should be more directly associated with the quality of democratic institutions. Citizens may distrust specific parties or their governments, yet still trust the legal system as long as they believe that the judiciary and the police are able to assist those who are being mistreated (Rothstein & Stolle, 2008). Building on the life-learning model (Mishler & Rose, 2001), we conceptualize trust in legal institutions as a running tally in which past legacies matter yet not in a deterministic manner. This perspective provides a useful starting point for theorizing how immigrants update their levels of trust. What matters most for this running tally, we suggest, is exposure to governmental institutions of different quality. Research on post-communist legacies shows that exposure to, or continued interactions with, communist parties is associated with lower trust in parties after transition to democracy (Pop-Eleches & Tucker, 2011, p. 380). The longer the exposure to communist parties and the more corrupt these parties are, the deeper the distrust toward political parties following the fall of communism. However, as time passes, these legacies should be gradually superseded by the more recent performance of postcommunist political institutions" (p. 383) While scholars of post-communist legacies focus on regime transitions, immigration can be understood as the transition of individuals between different political environments, with similar mechanisms driving the adaptation of political attitudes. Based on this notion of exposure we can generate hypotheses regarding variations in trust within-individuals over-time. While we cannot directly measure specific interactions with 8

9 institutions, we follow previous research and consider the time period spent in a specific country as a proxy for levels of exposure (Pop-Eleches & Tucker, 2011). First, we expect that the lower the quality of legal and political institutions in the country of origin, the lower the trust immediately following immigration. Furthermore, a longer exposure to low-quality institutions meaning that individuals spent a longer time in the country of origin should instill greater distrust. However, we also expect that longer exposure to high-quality institutions following immigration to countries with better quality of government would gradually lead to higher levels of trust. We can summarize these expectations as follows: H1: Individuals from countries with more corrupt institutions are expected to express lower trust immediately following immigration. H2: The effect of corrupt institutions in the country of origin would be stronger for those who immigrated at an older age. H3: The longer the time spent in a receiving country with higher-quality institutions, the more immigrants update upward their levels of trust. According to our theoretical perspective, the effect of the quality of government in the country of origin on immigrants trust runs through direct exposure to state institutions. We assume that this exposure is mediated by age: children are less exposed to the quality of government compared to older individuals. We therefore expect a negligible effect of the quality of government in the country of origin on levels of trust among those who immigrated at a very young age: H4: The effect of corrupt institutions in the country of origin on trust among those who immigrated at a young age would be small. 9

10 An additional individual-level factor that should mediate the degree of exposure to legal institutions is education. In the receiving country, higher levels of education are expected to be associated with more accurate information regarding the quality of institutions only when coupled with knowledge of the local language. Well-educated individuals who mastered the local language in the receiving country are likely to be more informed about political affairs through social and professional networks, access to media, and more frequent interactions with local government officials. This generates the following prediction: H5: The adaptation of trust following immigration is most pronounced among highly educated individuals who speak the local language of the receiving country at arrival (or soon thereafter) Contextual Background: The Israeli Case An ideal experiment for studying the effects of legacies of corruption on political attitudes would involve selecting a random group of individuals from countries with different quality of government and moving them into a country with high quality of government where they immediately become full citizens. In this case, there would be no self-selection into the receiving country based on priors beliefs regarding its democratic institutions and no variations in the process of naturalization. Israel s unique immigration policy brings us closer to such an ideal experiment compared to previous research designs. Israel has a highly diverse citizenry. According to OECD data, the foreign born population accounted for around 25% in 2010 (International Migration Outlook). As mentioned above, all Jewish individuals and their immediate relatives are eligible for citizenship in Israel under the Law of Return. Opening the country s gates for Jews from all over the world has been part of 10

11 Israel s founding ethos. Israel has actively sought to encourage immigration among Jews across the world, among other things through providing immigrants with a package of welfare support upon their arrival (Hacohen, 2001). Immigration to Israel has been traditionally explained in religious-ideological terms: Jews immigrating to Israel are perceived as returning to their historical homeland, regardless of their country of origin. This was especially the case among highly-educated immigrants from rich countries, who moved to Israel in higher numbers following the 1967 war (Cohen, 2009). At the same time, since its early days, Israel has also served as a destination for Jewish refugees, as in the case of dislocated Jews following World War II and those who escaped from hostile Muslim environments (Shuval, 1998). In some of these cases, immigrants were likely to choose Israel as the destination country mostly based on their Jewishness and their ability to naturalize immediately: the resulting package in Israel of fundamentally unrestricted and unselective Jewish immigration being awarded civil rights and material help stands out when compared to the restrictive immigration policies that persist in most other countries" (Della Pergola, 2000, pp ). Especially relevant for our sample is the immigration wave from the former USSR, which was unprecedented in scope. The decision of Russian authorities to allow Jewish immigration to Israel in 1989, after years in which it was banned, opened a path for many to escape the bleak reality of the collapsing communist regime. Until then, the few Jewish immigrants escaping from the USSR had two main destinations: Israel, where they could naturalize through the Law of Return, and the United States, where they were eligible for refugee status. Yet since 1989 the United States no longer grants refugee status to this groups of immigrants, making Israel their main destination (Cohen & Haberfeld, 2007). Seen from this perspective, we can think of these immigrants as quasi-refugees with an open ticket to Israel. 11

12 The unique circumstances of Israel s immigration policy thus significantly reduce the concerns regarding self-selection among the majority of immigrants in our sample. We further examine this point in Appendix 1, where we demonstrate that individuals who immigrated to Israel are distinguished by their religious identity (Jewish). Among Jewish immigrants in the ESS data, the choice of Israel as the destination country is partially influenced by the father s education: Jews who immigrate to Israel come from a slightly less educated background. However, it is hard to draw strong conclusions from this point, considering that the number of Jewish immigrants who immigrated to countries that are not Israel in our sample is very small. Data and Measurement Our analysis builds on the European Social Survey, waves 5 and 6 (Nowegian Social Science Data Services, 2010 and 2012). 6 These are the only waves in which respondents who immigrated were asked about both country of origin and year of immigration. In wave 5, there are 761 foreign-born Israelis (33% of all Israeli respondents). In wave 6, there are 783 foreignborn Israelis (31% of all Israeli respondents). Immigrants to Israel in our sample arrived from a large variety of countries (78 countries) and their years of immigration are spread over a large period of time (82 different years of arrival, from 1928 to 2011). Appendix 2 provides a summary of the country of origins and year of immigration variables. Our main dependent variable is trust in political institutions, captured by the following question: please tell me on a score of 0-10 how much you personally trust each of the institutions I read out. 0 means you do not trust an institution at all, and 10 means you have complete trust". As explained above, we focus on trust in the legal system since it is expected to be the least biased by the specific partisan constellation at the time in which the survey was 12

13 conducted. Appendix 3 provides descriptive statistics of this and other individual-level variables included in our analysis. Among the key explanatory variables is the quality of government in the country of origin. We use the Corruption Perceptions Index [CPI] from the 2011 Transparency International dataset to code respondents into two categories: those migrating from high corruption countries and those migrating from low corruption ones (see Appendix 2 for the classification of countries). We use the median CPI score from our sample s distribution as the threshold. Ideally, we would assign a corruption score to each immigrant based on the country of origin s level of corruption in the specific year of immigration. However, the CPI scores go back only until 1995 and for a very limited number of countries. By dichotomizing this variable we address the concern of changes in countries levels of corruption over time: countries are likely to move within their respective categories, but it is very unlikely that highly corrupt countries have constructed high-quality institutions within our time span, and vice versa. 7 Put differently, we assume that the quality of government is stable enough so that highly corrupt countries in 2011 were also highly corrupt 20 or 50 years ago. As shown in Appendix 4, this assumption is supported when looking at changes over : most countries have retained very similar CPI scores over this time period. 8 Relevant to mention, Israel stands firmly within the low corruption category, ranked 36 out of 182 countries coded in the data. Israel is therefore also less corrupt than many of the countries within the category of low corruption, including countries that have been an important source of immigration to Israel. This suggests that immigration to Israel would entail an exposure to better quality institutions for many of those arriving from low corruption countries. 13

14 Empirical Strategy We argue that exposure to institutions of different quality is the key factor in shaping trust in the legal system. Unfortunately, we do not have access to direct measures of exposure to institutions. Furthermore, measures of direct interactions with institutions, available in some surveys, may be endogenous to levels of trust. To overcome this problem, we use what we believe is the next best metric of exposure to institutions: the time spent in both the sending and receiving countries. We assume that the longer time an adult spends in a country the more this adult is exposed to its institutions. Furthermore, we also identify factors, namely early knowledge of the local language and education prior to immigration, which should facilitate exposure to institutions in both the sending and receiving countries. In identifying the causal relationship between exposure to institutions considering both their quality and the length of exposure and immigrants trust in the legal system, we utilize two empirical strategies. First, we interact the length exposure to institutions in the sending country and the length of exposure to institutions in the receiving country, while holding constant the quality of government in the country of origin. 9 We claim that for each individual the combination of the length of exposure in the sending and receiving countries is exogenous. As discussed above, it is difficult to believe that migrants choose the exact date of migration and, even more difficult to think that they would select the timing of the survey relative to when they migrated to Israel. Even if a reader would worry about the influence of some unobservable factors on the choice of the timing of migration, we claim that one element of this combination the time spent in the receiving country or the time spent in the hosting country will always be exogenous. Let us assume for example that one of our respondents decided to immigrate after 30 years in the country of origins due to distrust in local institutions 14

15 (making the first component of the treatment endogenous). In this case the exact length of time spent in Israel before the surevy could be considered orthogonal to any features of the respondent (for more details about this see balance checks in Appendix 10). In this strategy, we utilize variations in immigrants different years of arrival and ages of arrival within countries of origin (within and across level of corruption). The second strategy addresses another potential concern: some readers might worry that our results are driven by differences between immigrants arriving in different years. In order to address this concern, we exploit the time difference between the two survey waves used in the empirical analysis. By comparing individuals from the same sending country and same time of arrival, we identify the average impact of extra two years of exposure to the institutions in the receiving country (Israel). This second strategy shows that our results do not simply capture a cohort effect but that the length of exposure matters. Finally, to show that there is a causal effect of exposure to institutions on trust we test the interaction of two key variables that are expected to increase exposure: early knowledge of the language in the receiving country and higher education. For this test, we focus on levels of trust at the time of immigrants arrival, since we want to use only pre-arrival (or early-afterarrival) measure of language knowledge and we do not have the required information for respondents who immigrated long time ago. Given our empirical strategy, we do not control for variables that we consider postarrival, such as income or ideology, which are all features that length of exposure to a country s institutions could have influenced. However, robustness checks for the main models, with a number of control included, can be found in Appendix 9. 15

16 Finally, some readers may point to other possible mechanisms that are consistent with our findings. Perhaps the most important one is the fact that length of time spent in a country also might increase the integration into the local society. For this reason we include in the appendix models in which we control for generalized trust ( trust in people ) and for subjective sense of discrimination (see Appendix 9). We show that our key findings regarding exposure to institutions remain intact. Empirical Analysis We expect that immigrants from more corrupt countries will express lower levels of trust immediately following immigration compared to immigrants from countries with high-quality institutions (H1), and that levels of trust would be updated upward the longer these immigrants are exposed to high quality institutions (H3). The effect of corrupt institutions in the country of origin on initial levels of trust was predicted to be especially strong for those who immigrated at an older age (H2) and negligible for those who immigrated at a young age (H4). We use the age of 7, the first quartile of the age at immigration distribution, as our cut-off point for immigrating at a young age. We use the age of 16, the median age at immigration, as our cutoff point for immigration as an adult. Figure 1.a plots the relationship between time spent in Israel and trust in the legal system for four groups of immigrants: 1. those who immigrated at a young age from high corruption countries; 2. those who immigrated at a young age from low corruption countries; 3. those who immigrated as adults from high corruption countries; 4. those who immigrated as adults from low corruption countries. The lines represent the slopes and intercepts from simple bi-variate regressions (see Appendix 5 for the full regression results). In line with our expectations, adult immigrants from high-corruption countries have lower initial levels of trust immediately 16

17 following immigration compared to adult immigrants from low corruption countries. This difference, which in the figure is represented by the distance between the intercepts of groups 3 and 4, is also confirmed as statistically significant by a series of regressions reported in Appendix 6. Despite their differences at arrival, adults from both low corruption and high corruption countries update their levels of trust upward as time passes. However, adults from high corruption countries show a faster (steeper slope) recovery from their low initial score. The fact that also individuals from low corruption countries update upward their level of trust is expected, considering our choice of the median of the corruption scores as the threshold between low- and high-corruption countries. As noted above, Israel s institutions are scored higher than most countries within the low corruption category. If we take a less conservative threshold, the 3rd quartile of the CPI distribution (3.6, instead of the median 2.9 used before), we find that the slope of group 3 gets even steeper while the slope of group 4 gets flatter and becomes statistically insignificant (see Appendix 5, Table A11). 17

18 Figure 1: Immigrant s age, trust and corruption level (predicted values from models in Appendix 5). In contrast, regardless of their country of origin, those respondents who immigrated before the age of 7 express higher initial levels of trust compared to older immigrants and no clear trend of change over time. This not only supports the claim that the main determinant of trust is direct exposure to state institutions, but also excludes the existence of significant intergenerational transmission of political trust. We find no evidence for the claim that parents and grandparents cultural biases regarding political trust are transmitted to younger generations of immigrating families. In order to further examine how exposure to state institutions shapes levels of trust, we add a series of control variables for these regressions and interact age at immigration with time spent in Israel, the former being a proxy of the exposure to sending countries institutions and the latter being the proxy for exposure to receiving countries institutions. Using country 18

19 dummies, we are measuring the variation between immigrants from the same country. We also control for individual level features that may influence the level of trust. These variables could be considered pre-treatment, if the treatment is the time spent in Israel: gender, age at immigration, and father s education. We use the father s education as a proxy of the socioeconomic status of the family prior to immigration (for the coding of father s education, see Appendix 7). In addition, we include a measure of self-reported religiosity (on a 10-point scale). We control for religiosity, since some readers may be worried that more religious immigrants would be more suspicious of the legal system, given the periodical tensions in Israel between religious and legal authorities (Ben-Porat, 2013, p. 55). We assume that religiosity is a rather constant feature and remains stable after immigration. 10 These two variables, religiosity and father s education, are also the covariates that may play a role in immigrants selection into Israel (Appendix 1). As in all other regressions, we also include survey weights. The results are presented in Table Before discussing the variables of theoretical interest, we note that male immigrants express lower levels of trust. The effect of father s education is non-significant across all the regressions. More religious immigrants express lower levels of trust in the Israeli legal system, a result that is consistent across the different statistical models. Most importantly for us, these analyses confirm that those who immigrated from more corrupt countries at an older age also express lower levels of trust in the legal system at arrival. The relationship with age is non-linear: after a certain amount of exposure to corrupt institutions, additional years in countries of origin with low quality of government make only a negligible difference on levels of trust. Age at immigration is therefore better captured in a logarithmic form. We present also the linear operationalization of age, and the results are consistent across all the regression specifications. 19

20 We further examine our argument by testing the convergence of immigrants trust toward the levels of trust among native Israelis. The dependent variable in columns 5 and 6 is the absolute difference in trust between each respondent and the average trust among native Israelis of similar age (+/- 2 years) and years of education (+/- 1 years). Note that in this analysis we do not distinguish between immigrants from low- and high-corruption countries, which allows us to make sure that our previous results were not driven by the coding of quality of government in the sending country. We find that indeed a longer time period spent in Israel corresponds to a smaller gap of trust with native Israelis. Table 1: Age at arrival, years in Israel, and trust Trust in the Legal system Abs Diff from Natives High corr Low corr High corr Low corr All All Years in Israel ** *** * (0.022) (0.022) (0.015) (0.014) (0.009) (0.006) Log (age migr + 1) *** *** (0.284) (0.349) (0.136) Age at migr ** ** (0.017) (0.020) (0.008) Male *** *** *** *** (0.248) (0.253) (0.248) (0.254) (0.112) (0.112) Edu father (0.061) (0.061) (0.061) (0.061) (0.027) (0.027) Religiosity * ** (0.044) (0.042) (0.044) (0.042) (0.019) (0.019) Years in Israel * Log (age ** at migr + 1) * Years in Israel * age at migr (0.007) (0.007) (0.003) * (0.001) (0.001) (0.0002) Constant *** *** *** *** *** (1.162) (3.079) (0.803) (2.919) (0.549) (0.390) Observations ,019 1,019 R Country of origin yes yes yes yes yes yes dummies Survey Wave dummies yes yes yes yes yes yes Note: * p<0.1; ** p<0.05; *** p<

21 In order to further examine change over-time in levels of trust across age groups, in Figure 1.b we zoom in on immigrants from high-corruption countries. The figure plots the predicted values from Table 1 column 1 for four different age groups. 12 Again we see that the line for those who were children at immigration has a less steep slope and a higher intercept. In contrast, those who arrived at an older age have a significantly lower initial level of trust. Around 20 years after immigration to Israel, trust in legal institutions almost converges across immigrants from different age groups. Change over time: Beyond cohort effects So far we have shown that time spent in Israel, depending on age at arrival and country of origin, is associated with an increase in trust. We interpreted this as an effect of longer exposure to democratic institutions of higher quality. However, these results may be driven by a cohort-of-arrival effect. An alternative argument may suggest that those who immigrated to Israel few decades ago are different in some parameters (besides those we control for) compared to more recent immigrants and that these differences drive variations in trust. In order to reject this alternative explanation, we need to demonstrate that each year in Israel is associated with an increase in trust regardless of the year of arrival. We test this by including year of arrival dummies and exploiting the two-year difference between the two survey waves we analyze (Borjas & Friedberg, 2009). With this model we examine the impact of additional two years ( ) for immigrants who arrived at the same time. The first two columns of Table 2 use only the sample of individuals coming from high corruption countries. We find that additional two years in Israel are correlated with an increase in trust. These results hold with country dummies (column 1) and without country 21

22 dummies (column 2). In columns 3-6 we again take as the dependent variable the absolute difference between immigrants trust and the group of reference among native Israelis (similar age and education, as in the regressions above). We subset the sample by time of immigration to Israel. We expect that additional two years in Israel would matter the most for immigrants who arrived in Israel recently because, as can be observed in Figure 1.b, the adaptation of trust takes place mostly within the first 20 years after immigration. We therefore examine the convergence of trust for immigrants that arrived in Israel after 1985, 1989, 1993, and 1997, after which the sample gets incrementally too small. The largest wave of immigration in Israel s recent history arrived from the former-ussr during the early 1990s; we therefore use the years right before (1989) and after (1993) as our initial cut-off points, and then add two additional time points (1985, 1997) to make sure that our results are not driven by the years we chose. As expected, additional two years in Israel are associated with a convergence in levels of trust, especially among more recent immigrants. This allows us to reject the alternative explanations of a cohort effect. It also shows that a relatively short exposure to high quality institutions can make a significant impact on trust, in contrast to cultural arguments that emphasize long-term continuities of political attitudes among immigrants. Relevant to mention, some may suggest that specific developments during these two years ( ) could explain variations in trust among immigrants. We note that large scale military operations in Gaza between the Israeli Defense Forces and Hamas indeed took place during time period. Yet it is not clear whether and how these events should affect trust in the legal system and why they would do it differentially based on period of arrival as shown in the last four columns of Table 2. 22

23 Table 2: The effect of additional two-years in Israel on trust Extra two years Age at migr Trust in the Legal System Abs Diff from Natives High Corr High Corr After 85 After 89 After 93 After ** ** ** ** (0.289) (0.258) (0.189) (0.194) (0.273) (0.426) * * * (0.009) (0.008) (0.005) (0.006) (0.008) (0.013) Edu father (0.066) (0.054) (0.047) (0.049) (0.071) (0.119) Male ** *** ** ** ** (0.265) (0.256) (0.174) (0.179) (0.277) (0.447) Religiosity ** * (0.047) (0.042) (0.030) (0.031) (0.047) (0.074) Constant * *** *** *** (4.917) (4.846) (1.388) (0.489) (0.800) (1.452) Observatio ns R Country orig dummies Year of arrival dummies Note: Yes No Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes * p<0.1; ** p<0.05; *** p<0.01 Facilitating Factors and Exposure to Institutions: Education and Language So far, we provided evidence that greater exposure to corrupt institutions in the country of origin is associated with lower levels of trust immediately following immigration. We also showed that every additional year spent in Israel is associated with an increase in trust among those who immigrated as adults. Lastly, we provided evidence that these results do not reflect a 23

24 cohort effect. We now turn to examine whether and how mediating factors that are expected to condition exposure to political institutions are also associated with variations in trust. As discussed in the theoretical section, we assume that immigrants who speak the local language and are more highly educated are also more exposed to the quality of local institutions in the receiving country (H5). 13 These immigrants are more likely to have access to information about local institutions through the consumption of local media, embeddedness in social and professional networks and direct interactions with state institutions. These two factors are expected to work in tandem: higher education would facilitate exposure to local state institutions mostly among those who speak the local language. Since education and speaking Hebrew cannot be considered pre-treatment variables, there might be possible concerns of endogeneity. For instance, it might be that the Israeli education system instills trust or that those who trust choose to learn Hebrew. In the case of education, we solve this by only using a the sub-sample of respondents who arrived in Israel at the age of 26 (third quartile of the distribution for age at arrival) and older and are likely to have completed their studies outside Israel. While we could not find a similar solution for Hebrew, we are able to exclude some of the most obvious paths of reverse causality. The first one is a causal path that would go from trust in institutions to speaking Hebrew, which does not seem to be possible since the level of trust does not predict speaking Hebrew. The second path could go from education to Hebrew and, at the same time, from education to trust in institutions. This also seems highly unlikely since by itself education predicts neither trust nor speaking Hebrew. We further discuss these issues in Appendix 8. Table 3 presents the results of several tests for the relationship between speaking Hebrew, education and trust. In some of the models we include only those who immigrated at the age of 26 or older, who have most likely received their education outside of Israel. In some of the 24

25 models we subset to respondents who immigrated more recently (after 1993), a group for which the slope for changing levels of trust has still not plateaued (see Table 2). Columns 1-4 confirm that among immigrants from corrupt countries who speak Hebrew, higher education is indeed associated with higher trust in legal institutions. The interaction of higher education and Hebrew for this group is also associated with a smaller gap in trust between immigrants and the relevant reference group of native Israelis. These results are robust to subsetting the data by age at immigration and when including all or only recent immigrants in the analysis. In congruence with previous results, the effect is especially pronounced among more recent immigrants. We illustrate the relations between education, language and trust in Figure 2, which presents the predicted values from column 4 in Table 3. For those who speak Hebrew, higher education is associated with a smaller difference in trust compared to the native group of reference. For Hebrew speaking immigrants with more than 12 years of education, the gap in trust compared to the local group of reference gets smaller. In contrast, such convergence in trust does not take place among highly educated immigrants who do not speak Hebrew. As a robustness check, Appendix 8 (Table A14) offers different combinations of country and year of arrival dummies. The results for both within and across countries analyses are similar in substantive terms. They all point in the same direction: all else equal, speaking the local language of the receiving country and being more educated are associated with higher trust, in line with the expectation that these factors facilitate increased exposure to state institutions. 25

26 Table 3: Local language, education, and trust Trust in the Legal System HC HC, 26+ HC, 26+, after 93 Abs Diff from Natives all all, 26+ all, 26+, after 93 Years of Edu *** *** * (0.048) (0.072) (0.105) (0.028) (0.041) (0.087) Hebrew *** * *** *** ** (0.846) (2.043) (5.131) (0.475) (0.988) (3.468) Male ** *** *** (0.226) (0.400) (0.697) (0.101) (0.206) (0.463) Religiosity (0.040) (0.072) (0.120) (0.017) (0.038) (0.079) Years in Israel (0.011) (0.028) (0.103) (0.004) (0.011) (0.060) *** ** *** *** * ** Years of Edu * Hebrew (0.060) (0.136) (0.355) (0.033) (0.066) (0.235) Constant *** *** ** * (0.912) (1.483) (2.731) (0.509) (0.811) (1.739) Observations , R Country Orig dummies Survey dummies Note: Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes * p<0.1; ** p<0.05; *** p<

27 Figure 2: Language, Education, and Trust Conclusions The empirical analysis of survey data from Israel provides support for our argument that immigrants trust in the legal system is shaped by exposure to political institutions of different quality. Through disentangling the inherited and malleable components of trust, we have shown that immigrants from more corrupt countries have initial lower levels of trust immediately following immigration but also that this distrust can be superseded by exposure to high quality institutions in the receiving country. We demonstrated variations in trust among 27

28 immigrants to Israel based on country of origin, age at immigration, time since immigration, and facilitators of exposure to local institutions in the form of education and knowledge of the local language. Within the scholarly debate between the cultural approach and the experiential perspective, our results provide strong support for the emphasis on exposure to political institutions in studying political attitudes. Historical legacies in our case, the quality of government in the country of origin set the starting point for political attitudes and trust in government, but do not determine their endpoint. In addition, our results suggest that governments can facilitate the integration of newly-naturalized citizens from countries with low quality of government by making sure that they are treated fairly by governmental institutions, promoting access to education and encouraging the learning of the local language. Our results support the claim that public policy and institutional design can play a role in the integration of minorities, including often-marginalized groups such as immigrants from countries with low quality of government. For future research, there is a place to consider whether and how the eligibility of immigrants for immediate naturalization a unique feature of Jewish immigration to Israel affects levels of trust. Do similar variations in trust by age at immigration and country of origin appear also when examining trust among non-citizen immigrants and asylum seekers? There is also room for future research that would more accurately measure exposure to state institutions. In this analysis, we assume that a longer time spent in a country is associated with greater exposure to its political institutions, notwithstanding the role of mediating factors. However, experimental research may be able to manipulate exposure to state institutions of different quality and measure its effect on trust. Examining these questions should be high on the agenda of scholars who are interested in the diversifying nature of Western societies. 28

29 Appendix 1: Self-Selection When investigating self-selection of immigrants into Israel, two questions call for special attention: 1. Who, among all immigrants, choose Israel? 2. Who, among all Jewish immigrants, choose Israel? We answer these questions through extracting information on immigrants from all survey waves of the ESS. The first question is addressed by Table A1: being a Jew and being religious appear to be the main predictors of immigration to Israel. This is only to be expected, in light of Israel s immigration policy. Table A1: Selection into Israel, logit regression Into Israel" Dummy Religiosity ** *** (0.105) (0.045) Jewish *** *** (1.179) (0.301) Year of Birth *** (0.015) (0.005) Education father ** ** (0.135) (0.043) Male ** (0.502) (0.174) Religiosity * Jewish *** *** (0.173) (0.060) Constant *** (29.309) (10.077) Corr level dummies Yes No Survey wave dummies Yes No Observations 4,720 14,733 Log Likelihood Akaike Inf. Crit , Note: * p<0.1; ** p<0.05; *** p<

30 The second question, in contrast, is almost impossible to answer given that within all the ESS waves, only 71 Jewish respondents who identified as immigrants did not choose Israel as a destination, as shown in Table A2. Father s education is a small but predictive variable for immigration destination, as shown in table A3: Jews that immigrate to Israel come from a slightly less educated background. Figure A3 reports the distributions of the father s education among Jewish immigrants, depending on the receiving country. This figure, together with an analysis that considers the father s education variable as a factor, suggests that differences by father s education in selection into the receiving country are driven by the lack of category four in the sample of immigrants to Israel and less by the difference in highly educated respondents (see appendix 7 for father s education coding). Table A2: Jewish and non-jewish immigrants in our sample. Non-jewish Jewish Non-Israel Israel

31 Table A3: Father s levels of education, logit regression Into Israel" dummy (among Jews) Religiosity (0.040) (0.040) Year of birth (0.007) (0.008) Education Father * (0.063) Male (0.263) (0.263) Education father (0.757) Education father (0.772) Education father * (0.771) Education father ** (0.823) Education father ( ) Education father * (0.742) Education father (1.082) Constant (14.429) (14.684) Observations 2,251 2,251 Log Likelihood Akaike Inf. Crit Note: * p<0.1; ** p<0.05; *** p<

32 Figure A1: Father s levels of education 32

33 Appendix 2: Country of origin and years of immigration This appendix reports the number of respondents per country of origin in Table A4 and per year of arrival in Israel in Table A5. In bold are the countries classified as highly corrupt, meaning their corruption score is below the median of the CPI (2011) corruption score distribution. Table A4: Number of immigrants per country of origin Country nu m Country nu m USSR 81 Jamaica 1 Czechoslovakia 1 Jordan 4 Yugoslavia 2 Kenya 1 Afghanistan 2 Kyrgyzstan 1 Armenia 2 Korea, Republic of 1 Argentina 21 Kazakhstan 11 Austria 3 Lebanon 7 Australia 1 Lithuania 13 Azerbaijan 11 Latvia 4 Belgium 5 Libyan Arab Jamahiriya 16 Bulgaria 10 Morocco 154 Brazil 5 Moldova 16 Belarus 28 Mexico 3 Canada 5 Netherlands 1 Congo 1 Panama 1 Switzerland 2 Peru 1 Chile 3 Poland 59 Colombia 2 Paraguay 1 Cyprus 1 Romania 95 Czech Republic 6 Russian Federation 205 Germany 18 Sweden 2 Denmark 1 Slovakia 1 Algeria 15 El Salvador 1 Ecuador 1 Syrian Arab Republic 24 Egypt 27 Tajikistan 4 Eritrea 1 Turkmenistan 2 Spain 2 Tunisia 38 Ethiopia 53 Turkey 38 France 30 Ukraine 148 United Kingdom 15 United States 40 Georgia 32 Uruguay 6 Greece 1 Uzbekistan 26 Guatemala 1 Yemen 48 Hungary 17 South Africa 6 Ireland 1 Italy 2 India 17 Libya 10 Iraq 74 Moldova, Republic of 6 Iran, Islamic Republic of 42 Venezuela, Bolivarian Republic of 1 33

34 Table A5: Number of immigrants per year of arrival Year Immigrants Year Immigrants Year Immigrants

35 Appendix 3: Summary statistics In this appendix we report the summary statistics of the variables used in the analysis, from wave 5 and 6 of the ESS. Table A6: Foreign-born and Jewish, ESS waves 5 and 6 variable mean min max # of NAs Trust legal system Religiosity Year of arrival Years of education Male Jewish Year of birth Hebrew spoken in the house Age at migr High corruption country orig Years in Israel Level education father Table A7: All Israeli respondents, ESS waves 5 and 6 variable mean min max # of NAs Trust legal system Religiosity Years of education Male Jewish Year of birth Hebrew spoken in the house

36 Appendix 4: Corruption scores over time The two figures below demonstrate the strong correlations between CPI scores of Transparency International in 1995, 2001, and CPI CPI 1995 Figure A2: Corruptions scores from different years: 1995, 2001,