The effect of citizenship on intermarriages. Evidence from a natural experiment

Transcription

1 The effect of citizenship on intermarriages. Evidence from a natural experiment Davide Azzolini FBK-IRVAPP (azzolini@fbk.eu) Raffaele Guetto University of Trento (raffaele.guetto@unitn.it) Extended abstract submitted to the 29 th Annual Conference of the European Society for Population Economics Motivation and research questions The growth of intermarriages (here defined as marriages celebrated between natives and immigrants) is often seen as an indicator of increased integration and societal openness (Adserà and Ferrer, 214) as intermarriages are hindered by the persistence of social boundaries between the native and the immigrant population. In the first place, intermarriages depend on marriage markets constraints such as the size and the sex ratio within the immigrant groups (Kalmijn, 1998; Chiswick and Houseworth, 211). But also other factors contribute to reduce the likelihood of intermarriage. Among these, an important role is played by socioeconomic differences (e.g., labor market or residential segregation) and cultural factors (e.g., individual preferences for the joint consumption of ethnic related goods in the household and the persistence of prejudices across groups; see Becker, 1981; Lam, 1988; Furtado, 212 and Kalmijn, 1998; Potarca and Mills, 212). A large number of studies have argued and documented that the more immigrants are culturally and socioeconomically assimilated, the more likely they are to entering marital unions with natives (Adserà and Ferrer, 214). Parallel to this 'assimilation hypothesis', the larger the 'social distance' between natives and immigrants, the higher the possibility that mixed unions occur as a result of an exchange between the native and the immigrant. The status exchange hypothesis was first developed by Davis (1941) and Merton (1941) and has found some empirical support in recent studies on immigrant-native marriages in Australia and the U.S. (Choi et al., 212), as well as in new immigration countries like Spain (Cortina Trilla et al., 28) and Italy (Guetto and Azzolini, 214). This hypothesis predicts that in a setting of weak integration and weak chances of socioeconomic integration for immigrants, the latter might be prone to 'trade' some of their valuable traits (e.g., young age and high education) by marrying lower rank members of the native population (e.g., older and less educated individuals). Citizenship may be one of the possible "rewards" that immigrants obtain in marrying a member of the native population, as citizenship increases individuals legal stability and access to civic and social rights in the hosting society (Bloemraad, 28). Building upon the rich literature on intermarriage, in this paper we ask whether the immigrants possession of the citizenship of the destination country works as an incentive or a disincentive of migrant-native intermarriages. Two contrasting hypotheses can be formulated: 1

2 immigrants' possession of citizenship leads to more intermarriages (H1), because it increases immigrants integration in the host society and diminishes many of the boundaries existing between them and natives e.g., citizenship might give immigrants a longer-term prospect of stability in the country and improve their access to the labor market, etc. Hence, the possession of citizenship is expected to increase immigrants likelihood to enter intimate contact with natives. An opposite prediction could instead point to the importance of citizenship as an exchange good in the mating. In this perspective, the prospect of obtaining citizenship through marriage is an incentive for migrant partners to enter unions with natives. Hence, we could expect that immigrants' possession of citizenship has a negative effect of intermarriages (H2). Evidence in favor of H1 would lend support to an assimilation explanation of intermarriages, according to which the more immigrants are assimilated in the host country, the more they are likely to cross national boundaries in marriage. In this sense, intermarriages could be considered a consequence of assimilation. In turn, evidence in favor of H2 would be supportive of a status exchange explanation, so that intermarriage would act more as a gateway to assimilation. Empirical literature on the causal relationship between citizenship and intermarriage is scarce and non conclusive (Haandrikman 214). The assessment of a causal relationship is plagued by severe endogeneity issues, as immigrants possession of the citizenship of the country of destination is correlated with a potentially long list of observable and unobservable factors that also affect intermarriages (e.g., an immigrant with citizenship is likely to have spent more years in the host country, to have acquired a higher language mastery as well as to have developed a wider network of acquaintances, relative to an immigrant without citizenship). In our contribution, we overcome this problem exploiting two European Union Enlargements (hereafter, EUEs) as two natural experiments that provide exogenous variation in the legal status of immigrants originating from the new member countries of Eastern Europe. Particularly, we focus on the effects of the 24 EUE and the 27 EUE on intermarriages celebrated between native men and foreign women in Italy Italian Husband - Foreign Bride Foreign Husband - Italian Bride Both Foreigners Figure 1 Share of mixed- and foreign-marriages out of total marriages (Italy, ). Source: Own elaboration based on data from the Italian Office of Statistics (ISTAT, 1 Intermarriages made of a native-italian woman and a foreign man are much less relevant in quantitative terms, hence we do not consider them in our work. 2

3 Italy is a new destination of international migration and has received, since the early 2s, significant and highly feminized flows of immigrants, especially from East-European countries. Italy is also characterized by a particularly difficult socioeconomic integration of immigrants, most notably women, who display rather poor economic conditions and have quite precarious legal status (Sciortino, 24; Reyneri, 27; Reyneri and Fullin, 211). Nonetheless, marriages contracted between native men and immigrant women, have increased at unprecedented rates in recent years, rising from less than 3% in 1996 to 7.9% of total celebrated marriages in 212 (Figure 1). Taken together, these patterns stand in apparent contrast with a fully-fledged process of assimilation and open the possibility to test the validity of the status exchange theory. Therefore, we hypothesize that (H2) might hold in the Italian case, i.e., we predict intermarriages between Italian men and immigrant women to decrease in response to the access of Eastern European countries in the EU. Identification strategy To assess the causal effect of citizenship status on intermarriages, we exploit the two EUEs that took place in 24 and 27. These two enlargements involved ten East-European countries 2 : 8 EUE-24 countries: Poland, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Slovakia, Slovenia; 2 EUE-27 countries: Romania and Bulgaria. In our framework, these ten East-European countries represent the "treated units", as upon EU accession citizens of these countries became EU citizens and thus experienced a sudden and relevant change in their legal status. Table 1 provides an overview of the timing of the EUEs. 3 From the table, it is clear that the new members' access did not occur without preannouncement, opening up the possibility of anticipation effects. In our analyses, we take into account the likely existence of anticipation effects by setting the date of two enlargements one year before the actual access date. Table 1 Overview and timing of the last three EUEs EUE EU Treaty Council Signature Decision Accession 13/12 16/4 1/5 27 EUE EU Council Treaty Decision Signature Accession 16/12 25/4 1/1 213 EUE EU Parlament Decision and Treaty Signature Accession 1/12 1/7 2 The 24 EUE also involved Cyprus and Malta, but we do not consider these countries as they are not part of the East European area. 3 The table also includes the 213 EUE, which involved Croatia. We could not analyze this last EUE with our data because it occurred too recently. 3

4 As said above, when the EUE-24 and the EUE-27 countries became new member states of the European Union, their citizens became EU citizens. One might argue that, for some respects, having the EU citizenship may not be the same as having the Italian citizenship for immigrants residing in Italy. For example, the former does not allow to vote at the national elections. But the EU citizenship ensures nonetheless important advantages relative to having a non-eu citizenship. Among them, we could mention the possibility to stay in the country independently on immigrants permit of stay as well as an increased freedom of movement across other EU countries and the lower restrictions in the labor markets. Hence, we do not expect substantial differences. If differences exist, though, our analysis will yield lower-bound estimates of the effect of citizenship on intermarriages. Empirical set up Data We use data from the Italian Register of Marriages (IRM) relative to 18 years ( ). This dataset contains exceptionally high-quality and rich information on all marriages celebrated yearly in Italy. The dataset contains detailed information on women's citizenship status, as well as both partners' age, levels of education and labor market condition at the moment of marriage. Such information will be exploited in order to assess whether the EUE had heterogeneous effects across different types of couples. Table 2 shows the countries included in our dataset. Beyond EUE-24 and EUE-27 countries (our "treated units"), we also include as "control countries" non-treated East-European countries (i.e., countries that were not members of the EU during our time-span) and other nontreated countries outside Europe. We do not consider European countries that were already members of the EU. Table 2 Treated and control countries Treated countries (1) Control countries (38) EUE-24 Other East-European countries Poland Albania Macedonia Croatia Czech Republic Ukraine Moldova Bosnia-Herzegovina Estonia Russia Belarus Serbia-Kosovo-Montenegro Hungary Latvia Other countries Lithuania Switzerland Tunisia Ethiopia Slovakia China Canada Morocco Slovenia Philippines Cuba Mauritius Japan Dominica Nigeria EUE-27 Iran Peru Colombia Bulgaria Thailand El Salvador Ecuador Romania Turkey Mexico Dominican Republic Algeria USA Venezuela Cape Verde Argentina Australia Chile Brazil 4

5 The Synthetic Control Method To estimate the causal effects of the EUEs we apply the "synthetic control" (SC) method (Abadie and Gardeazabal 23; Abadie, Diamond, and Hainmueller 21; Abadie, Diamond, and Hainmueller 214, Dube and Zipperer, 214). The SC method allows to construct a 'counterfactual' by calculating a weighted average of the control units. More precisely, the method assigns different weights to our 38 control countries (or, the countries included in the so called "donor pool") in order to approximate the outcome predictors. Hence, the weights distribution among the control units creates the so called "synthetic" unit, which is an estimate of what would have been observed for the treated unit if it had not received the treatment. The technique has at least two advantages over a "Difference-in-Differences" (DiD) estimator. First, the SC performs well also when the treated units are few or even only one. This is true especially when considering that the estimated standard errors obtained from DiD regressions with very few treated units are not correct as they rely on asymptotic assumptions which obviously do not hold with small number of units (Conley and Taber 21). Second, the SC does not require the "parallel trend" condition. In turn, as we are going to see, the SC is demanding as it requires a wider pre-treatment time window to reduce the bias due to unobservable factors affecting the outcome. Outcome variable Our outcome variable is obtained dividing (a) the yearly number of marriages celebrated between women of each of the above listed countries with Italian men by the product of two components: (b) the number of immigrants residing in Italy and originating again form each of the countries considered; (c) the female ratio in each specific national group. Hence, this measure incorporates the main marriage market structural constraints identified in the theory (Kalmijn, 1998; Chiswick and Houseworth, 211) and allows to compare intermarriages across women's countries of origin net of the size and the female balance in immigrants flows. Predictors The SC method assigns weights to the different control units on the basis of the set of 'predictors' (or matching variables) that are included in the analysis. In the preliminary results that follow, we considered lagged measures (i.e., measured in the pretreatment period) of our outcome variable (which already incorporates two important predictors of intermarriages, namely size of the immigrant group and female ratio in the group), per-capita GDP of the countries, kilometric distance from Italy. We are currently collecting additional indicators relative to other socioeconomically relevant attributes of the countries and the immigrant groups, like the levels of education, literacy rates, labor market indicators. Indexes of linguistic or cultural distance between Italy and the sending countries considered will also be considered. Preliminary results Figure 2 shows the trends of intermarriages by the four groups of countries identified according to their 'treatment' status. The red line identifies the 24 EUE while the blue line represents the 27 EUE. Both As mentioned above, we allow for anticipation effects: i.e. we set the 24 EUE on instead on and the 27 EUE on instead on , because the enlargement decisions were taken before the actual access of the new member states. Although the graph pictures a merely descriptive and aggregate analysis of the number of intermarriages, some patterns are worth being underlined. All groups, although with different intensity, show a growth in intermarriages in the period Between 22 and 25, marriages between 5

6 native-italian men and women from EUE-24 countries (the treated countries, in this period) experience a decrease in the number of intermarriages and the number remains stable throughout the rest of the time window considered (hence till 212). Marriages with women of non-european and non-treated countries (bottom-right panel) also decrease after 22 but then tend to increase again. Intermarriages with women from EUE-27 countries, instead, increase sharply in the time window and then suddenly and dramatically decrease after the 27 EUE. 6 Non-treated East-European Countries EUE-24 Countries EUE-27 Countries Other Non-treated Countries Figure 2 Number of intermarriages by group of countries of origin of the bride Note: Own elaboration based on IRM data. The red line identifies the 24 EUE while the blue line represents the 27 EUE. EUEs are anticipated relative to the actual access of the new member states as the enlargement decisions were taken before (as shown above in Table 1). 2 Non-treated East European Non-treated Non-European EUE Figure 3 Trends in the weighted number of intermarriages: EUE-27 vs. Non-treated East European and Non-Treated non-european countries. Note: Own elaboration on IRM data. The blue line represents the 27 EUE. 6

7 We now turn to the analysis of the impact of 27 EUE on the weighted number of intermarriages. We aggregated the two countries involved in the EUE, Romania and Bulgaria, summing the number of marriages of the two countries. Separate analyses by country led to qualitatively similar results and are available upon request. Due to unavailability of data on the number of immigrant residents before 22 for some nationalities included in the donor pool, we had to restrict our pretreatment period to the years Figures 3 compares the trends in the number of weighted marriages of the EUE-27 countries and the control countries belonging both to the non-treated East- European group and the other non-treated countries. It is evident that, while the levels differ quite a lot across the groups--with EUE-27 countries showing a much higher incidence of marriages with Italians, the trends in the pre-treatment period are similar across groups. Figure 4 shows the results obtained with the SC method. More precisely, Figure 4 displays the trends of the weighted number of intermarriages for the 'real' EUE-27 countries and for the 'synthetic' EU-27 countries. Weighted Marriages treated unit synthetic control unit Figure 4 Trends in the weighted number of intermarriages: EUE-27 vs. synthetic EUE-27 Note: Own elaboration on IRM data. Figure 4 makes clearly evident that EU access led to a marked drop in the number of intermarriages between native-italian men and women from EUE-27 countries (namely, Romania and Bulgaria). The SC performed well as the real pre-treatment trend is fairly well approximated by the synthetic unit one. The Root Mean Squared Prediction Error in the pretreatment period is minimized by the choice of weights and is Out of the 38 countries included in the donor pool, the SC assigns non-zero weights to three countries, namely Brazil (.323), Japan (.425) and Moldova (.226). Hence, the synthetic unit is given by the weighted average of these three countries. To assess the significance of these estimates, we conduct a series of "placebo tests" by applying the SC method used to estimate the EUE effects on EUE-27 countries to each of the other 38 countries included in the donor pool. This study is accomplished by assigning the "EU enlargement intervention" to each control countries as if they were treated as EUE-27 countries. The results of these tests are presented in Figure 5 and confirm that the EUE in 27 had strongly significant effects on intermarriages celebrated between Italian men and Romanian and Bulgarian women. 7

8 1 EU 27 Enlargement - Weighted N of marriages 5 gap anno Figure 5 Estimated gap between real EUE-27 and synthetic EUE-27 (bold black line) and placebo gaps in the remaining 38 countries (grey lines) Note: Own elaboration on IRM data. The significance of the effect is visibly high as the placebo gaps (grey lines) almost never overlap with the EUE-27 gap line in the post-treatment period. The only exception is observable in 26 and it is given by only one placebo line. The probability of estimating a gap larger than the one estimated for the EUE-27 countries is very low (1/38=.263), hence a lower level than the level typically used in conventional tests of statistical significance (i.e.,.5). Concluding remarks and further steps The preliminary results provide clear support for H2. An exogenous change in the legal status of immigrant women from Eastern Europe (who became EU citizens with 27 EUE) led to a very sharp drop in intermarriages in Italy. This result points to the relevance of 'status exchange' in intermarriages formation in a context like Italy, where immigrant integration is poor and naturalizations are quite difficult and mostly accessible via marriages with a native citizen. Further analysis will be performed to provide robustness to the results: a longer pretreatment period will be analyzed, other predictors will be included in the analysis, and different donor pools will also be considered. Most importantly, to leave out the possibility that the impact of EUE is only specific to the two countries which accessed the EU in 27, we will assess the impact of the 24 EUE. Finally, we will perform analyses aimed at assessing the heterogeneity of EUE effects on different types of mixed couples indentified by the combination of three theoretically relevant partners' characteristics--i.e., the level of education, the occupational status and age. 8