Cross-Country Performance in Social Integration of Older Migrants. A European Perspective

Transcription

1 Document de travai Working paper Cross-Country Performance in Socia Integration of Oder Migrants. A European Perspective Caroine Berchet (LEDa-LEGOS-Université Paris-Dauphine, Irdes) Nicoas Sirven (Irdes) DT n 46 Mars 2012 Reproduction of the text on other web sites is prohibited but inks to access the document are permitted: Institut de recherche et documentation en économie de a santé IRDES - Association Loi de rue Vauvenargues Paris - Té. : Fax :

2 The IRDES Working Papers coection is estabished as a means to disseminate prepubication versions of scientific artices. This coection aims at stimuating refection and discussion with regard to economic anaysis and method appied to socia heath protection, as we as pubic poicy assessment. The opinions expressed are the responsibiity of the authors and do not necessariy refect those of IRDES. Readers are encouraged to emai authors with comments, critics and suggestions. institut de recherche et documentation en économie de a santé 10, rue Vauvenargues Paris Té: Fax: E-mai: pubications@irdes.fr Director of pubication: Yann Bourguei Scientific adviser: Mohamed Ai Ben Haima, Juien Mousquès, Nicoas Sirven Copy editing: Franck-Séverin Cérembaut - Anne Evans - Anna Marek Layout compositor: Khadidja Ben Larbi Distribution: Suzanne Chriqui, Sandrine Bequignon

3 Cross-Country Performance in Socia Integration Caroine Berchet,ab*, Nicoas Sirvenb Abstract This paper provides new empirica evidence on the reationship between migration and socia integration. It expores the hypothesis that migrants essentiay differ from non-migrants with regard to the ength of residence in the country which is a proxy of migrants socia distance to natives. The determinants of socia participation and interpersona trust are examined at both the individua and institutiona eve. Using SHARE data and macroeconomic series, we first anayse the infuence of immigrant ength of stay in the host country on socia integration indicators. We then examine the roe institutiona characteristics pay on cross-country differences in speed of socia integration (i.e. immigrants propensity to socia participation according to their ength of stay in the host country). As expected, the immigrant popuation presents a ower ikeihood than the native popuation to get invoved in socia activities and to trust other peope. Nevertheess, the more immigrants have spent time in the host country, the more they take part in socia activities. The anaysis aso reveas significant cross-country differences in immigrants speed of socia integration. Macroeconomic series ike the GINI coefficient of income inequaity and the Corruption perceived index coud expain these differences. From a pubic poicy perspective, our resuts suggest that immigrants socia integration is more rapidy achieved in fair countries i.e. those with a more favourabe socia environment where the eves of income inequaity and perceived corruption are ower. Keywords: Socia capita, Ageing, Income inequaity, Mutieve modes. JEL Codes: F220, O520, C310. This work is part of the HEAPS research program (Heath Economics of Ageing and Participation in Society, www. irdes.fr/heaps) supported by grant number ANR-09-JCJC awarded by the Nationa Research Agency, France. This paper uses data from SHARELIFE reease 1, as of November 24th 2010 or SHARE reease 2.4.0, as of March 17th The SHARE data coection has been primariy funded by the European Commission through the 5th framework programme (project QLK6-CT in the thematic programme Quaity of Life), through the 6th framework programme (projects SHARE-I3, RII-CT , COMPARE, CIT5-CT , and SHARELIFE, CIT4- CT ) and through the 7th framework programme (SHARE-PREP, and SHARE-LEAP, ). Additiona funding from the U.S. Nationa Institute on Aging (U01 AG S2, P01 AG005842, P01 AG08291, P30 AG12815, Y1-AG and OGHA , IAG BSR06-11, R21 AG025169) as we as from various nationa sources is gratefuy acknowedged (see for a fu ist of funding institutions). a LEDa-LEGOS-Université Paris-Dauphine. Pace du Marécha de Lattre de Tassigny, Paris Cedex 16. b IRDES (Institute for Research and Information in Heath Economics), 10, rue Vauvenargues, Paris. * Corresponding Author: caroine.berchet@dauphine.fr; berchet@irdes.fr Document de travai n 46 - IRDES - Mars

4 Résumé Une perspective européenne des performances d intégration sociae des migrants âgés L objet de cette étude est d anayser es reations entre a migration et intégration sociae. Notre anayse se propose d étudier hypothèse seon aquee intégration sociae des immigrés diffère essentieement de cee des natifs en raison de a durée de résidence dans e pays d accuei, facteur qui constitue dans notre anayse un indicateur de a distance sociae des migrants aux natifs. Les déterminants de a participation sociae et de a confiance interpersonnee sont anaysés au niveau individue et institutionne. À partir des données de enquête SHARE, compétées par des séries macroéconomiques, nous anaysons dans un premier temps infuence de a durée de résidence dans e pays d accuei des immigrés sur es deux indicateurs d intégration sociae. Nous étudions ensuite e rôe joué par es caractéristiques institutionnees sur es différences de vitesse d intégration entre es pays européens (i.e. a probabiité d un immigré d être intégré seon sa durée de résidence dans e pays d accuei). Les résutats indiquent que a popuation immigrée présente une pus faibe probabiité que a popuation native de participer à des activités coectives et d avoir confiance en autrui. Cependant, intégration sociae des immigrées s accroit avec a durée de résidence dans e pays d accuei mais anayse révèe, par aieurs, des différences de vitesse d intégration entre es pays européens. Ces différences sont expiquées par es séries macroéconomiques tees que e coefficient de GINI et indice de corruption. D un point de vue des poitiques pubiques, nos résutats suggèrent que intégration sociae des immigrés est pus rapide dans es sociétés caractérisées par un environnement socia favorabe, où es niveaux d inégaité de revenu et de corruption sont faibes. Mots-cefs : capita socia, vieiissement, inégaités de revenu, indice de corruption, modèe mutiniveaux JEL Cassification : F220, O520, C310. Document de travai n 46 - IRDES - Mars

5 1. Introduction Successfu integration strategies of migrants within the European Union rey on the one hand, on abour market poicies such as antidiscrimination and migrant entrepreneurship promotion, and on the other hand, on socia participation and socia cohesion by promoting equa access to services (European Commission, 2005). They are in ine with the broader poicy of promoting socia incusion that is a one of the five key areas of the Europe 2020 strategy (European Commission, 2009). Socia incusion is indeed the process from which individuas participate to society through professiona activity, [ ] interaction with others, participation to coective institutions (Schnapper, 2008). The iterature on socia capita provides some interesting highights on the factors that promote socia incusion. Empirica studies focusing on cutura and structura aspects of socia capita (Grootaert & Van Basteaer, 2002) respectivey indicate that migrants report ower eves of generaised trust and have a ower propensity to take part in socia activities (Kazemipur, 2004;, Breton, 2003, Aeksynska, 2011). One obvious reason is that migrants tend to face the usua socia disadvantages that hamper socia capita significant gaps persist between this group and the majority popuation in terms of poverty, income, heath, unempoyment, education and eary schoo-eaving (Berchet & Jusot, 2009). Nevertheess, recent researches (Aeksynska, 2011; Asund et a., 2009; De Pao et a., 2007) indicate that migrants economic and socia conditions (socia reationship, education, empoyment and househod status) improve with the ength of stay in the host country. Another effect of the ength of stay coud be that the onger migrants dwe in the country, the more they become famiiar with the country s forma (system, administration, aw, etc.) and informa (norms, vaues) institutions, thus faciitating participation in socia activities and enhanced generaised trust. Such a process of embededness can be seen as a reduction in the socia distance between migrants and natives that fosters socia connectedness (Akerof, 1997; Van der Vegt, 2002). In both cases, the ength of stay in the host country is often used to capture immigrant socia assimiation (Aeksynska, 2011); a process that supposes a change in migrants behaviour over time. In addition to the individua characteristics, country specific cutura and historica backgrounds, eves of economic deveopment, and expenditure on socia services are often cited to expain cross-country differences in individua decisions of invovement in vountary activities (GHK, 2010) and generaised trust. Athough institutiona features of the country of origin are found to pay a roe on migrants socia incusion, few evidence is avaiabe when it comes to the infuence of macroeconomic determinants of the host country on migrants socia assimiation process (Aeksynska, 2011). One hypothesis to be tested coud be that some countries in Europe perform better than others in terms of socia participation and generaized trust with equa eves of migrants ength of residence because of nationa systems characteristics. Put differenty, some countries with different eves of migrant s ength of residence may experience simiar eves of migrants socia capita. This being empiricay verified, the question woud be why some European countries have different speed of integration? Focus on crosscountry nationa features woud then shed ight on this issue. Data from the Survey of Heath, Ageing, and Retirement in Europe (SHARE) combined with macro-economic series meet the requirements to test this assumption. These databases aow combining individua socio-economic data with macro series of the socio-economic Document de travai n 46 - IRDES - Mars

6 context and the features of wefare state regimes in mutieve modes. SHARE data cover 14 countries in wave 2 ( ) and provide individua retrospective information on migrants year of arriva in the country of residence. SHARE focus on respondents aged 50 and over provides enough time depth to anayse the infuence of migrant s ength of residence on the usua variabes of socia capita. This popuation may be more infuenced by the socio-economic environment of the host country since they have been imbedded in for severa years. In the case of a dynamic process of socia integration, it is aso interesting to focus on ong term effects of macroeconomic conditions. The set of individua and context variabes to be incuded in the modes are discussed in the next section on the basis of the existing empirica iterature of the determinants of socia capita. Section three presents the data sources and some descriptive statistics of the main variabes of interest. Econometric modes and modes estimates are respectivey discussed in sections four and five. Summary and poicy impications are drawn in the concusion. 2. Survey of the Literature 2.1. Individuas determinants of socia capita The usua individua determinants of socia capita to be found in the iterature remain the same for each popuation of immigrants and natives. By and arge, socio-demographic characteristics such as marita status and age are found to have more ambivaent effects on socia participation and trust (see for instance Aesina & La Ferrara, 2000; Gaeser et a., 2002) than education and economic variabes. Higher education provides access to socia network, increases opportunities for socia participation, fosters communication skis and deveops vaues or mora norms which promote interpersona trust (Coeman, 1988; Brehm & Rahn, 1997; Gaeser et a., 2002; Aeksynska, 2011). Income and occupationa status shape individua resources and time (Brehm & Rahn, 1997; Gaeser et a., 2002; Aesina & La Ferrara, 2000; Rupashingha et a., 2006). There is evidence of a strong positive reationship between income eve and participation (Gaeser et a., 2002; Aeksynska, 2011) whie deprived or unempoyed peope are ikey to deveop sentiments of stigmatization or discrimination, which undermine incentives to participate in socia activity and eads to distrust others. Under this premise, immigrant popuation under-investment in socia participation with regard to the native popuation can party be expained by the fact that the former face more socia disadvantages than the atter on average (Berchet & Jusot, 2009). Another interpretation of this gap in socia participation coud be due to the socia distance between the two popuations. The assumption is that socia distance between migrants and natives tends to decrease as migrants ength of residency in the host country increases. The iterature suggests that higher ength of stay in the destination country fosters immigrant assimiation because it enabes to earn more about the forma or informa institutions which increase opportunities for socia participation (Aeksynska, 2011). Nevertheess, immigrants with onger stay of residence have a significant ower propensity of trusting than native born popuation, whie newy arriving immigrant are no different from native born regarding trust in other peope. On interpretation is that migrants positive outook dampens with time due to disiusion or barriers that immigrants may face (Aeksynska & Agan, 2010). This doube-edged effect of the ength of stay in the host country on the different aspects of socia capita requires focusing on other determinants at the institutiona eve. Document de travai n 46 - IRDES - Mars

7 2.2. Institutiona features matter Athough most of the empirica iterature on the determinants of socia capita stresses the important roe of the economic, poitica, and socia context, ony few institutiona measures have been used to expain migrants investment in socia capita. For instance, immigrants residing in countries with higher GDP per capita tend to have higher chances to take part in socia activities (Aeksynska, 2011). However, measures of democratic institutions do not seem to have any significant infuence on immigrant popuation (Aeksynska, 2011) athough these features of society encourage the formation of socia participation and trust for genera popuation (Muer & Seigson, 1994 ; Sides, 1999; Knack & Zack, 2001; Rosthein & Stoe, 2003; Paxton, 2002; Fidrmuc & Gerxhani, 2005). Among the institutiona determinants of socia capita unexpored for the migrant popuation, some ight shoud be shed on three main areas: - socia heterogeneity measured by ethnic fractionaisation appears to discourage individua investment in socia capita (Sides, 1999; Aenisa & Ferrara, 2000; Kervin & Kine, 2002; Rupasingha et a., 2006; D Hombres et a., 2010). Sides (1999) argues that ethnic fractionaisation may refect a conficted society which in turn decreases both socia participation and trust. This ast hypothesis suggests a preference for homogeneity; individuas prefer to interact with peope that are simiar to themseves (Aenisa & Ferrara, 2000). - non-egaitarian or corrupted societies present a significanty ower eve of individua socia participation and interpersona trust (Sides, 1999; Aenisa & Ferrara, 2000; Knack, 2002; D Hombres et a., 2010). Immigrants hopes and attempt to integrate are ikey to vanish in unfair societies with greater eve of income inequaity or greater eve of corruption. They may encounter even more difficuties than the native popuation to take part in the socia ife because the socia environment appears not to be receptive, decreasing their sense of cooperation or their wish to integrate in the society (Breton, 2003). - wefare states regimes pay a significant roe in shaping a country s socia capita (Scheepers et a., 2002; Oorschot, 2003; Saamon & Sokoowski, 2003; Rosthein & Stoe, 2003; Oorschot & Arts, 2005; Kumin & Rosthein, 2005). The we estabished debate on whether they are compementary or substitute to civi society is aso reevant for the migrant popuation. On the one hand, wefare state institutions may impy mora hazard if they ower the cost of assimiation in the host country (Nannestad, 2007). This hypothesis suggests that the generosity of wefare states is negativey correated with migrant socia participation and generaized trust (crowding out effect). On the other hand, if wefare state institutions foster migrant generaized trust and encourage them to take part in civi society through socia participation, then the cooperation mode (crowding in effect) is more reevant. Document de travai n 46 - IRDES - Mars

8 3. Data 3.1. Sources The anaysis of the individua determinants of socia engagement and interpersona trust is based on data from the second wave of the Survey of Heath, Ageing, and Retirement in Europe (SHARE, SHARE is a mutidiscipinary and cross-nationa cohort of individua data on heath, socio-economic status and socia and famiy reationships of individuas aged 50 and over (Börsch-Supan & Jürges, 2005). They are a baanced representation of the various regions in Europe, ranging from Scandinavia (Denmark and Sweden) through Centra Europe (Austria, France, Germany, Switzerand, Begium, the Netherands, and Ireand) and Eastern Europe (Poand, The Czech Repubic), to the Mediterranean (Spain, Itay and Greece). The second wave ( ) consists of individuas surveyed in 14 countries and provides information on heath, socio-economic status, socia and famiy networks. Anayses are based on a sampe of 31,852 non institutionaised individuas aged 50 and over with non-missing observations for the variabes retained in the anaysis (fu rank data matrix). Institutiona variabes describing the socio-economic context and the generosity of wefare state regimes are taken from different sources. The OECD database ( oecd.org/) provides comparative statistics and offers an overview of recent economic trends through the presentation of a wide range of short-term economic indicators not ony for the OECD member countries but aso for non-member countries. Additiona macro-series are taken from Transparency Internationa and Aesina et a. (2003) Variabes - Dependant variabes: Indexes of socia capita Two dichotomous dependant variabes are considered to measure socia integration: invovement in socia activities and interpersona trust. Invovement in socia activities is derived from the participation to any of the six socia activities (vountary/charity work; providing hep to famiy, friends or neighbours; educationa/training courses; sport/socia cub; reigious organisation; and poitica activities). The variabe takes the vaue 1 if the respondent caims to take part in at east one of these activities and 0 if he does take part in none of them mentioned. The scope of activities is rather arge in order to best capture the idea of socia integration. Different combinations of the items have been tested and the six socia activities retained here are associated with the best scae reiabiity coefficient (α = 0.469) than any other combination of a esser number of items. To measure interpersona trust, respondent are asked to state on a scae from 0 to 10 where 0 means that one can t be too carefu in deaing with peope and 10 means that peope can be trusted. The variabe is then dichotomised, taking the vaue 1 if respondents provide a score that is higher than five and 0 if the eve of trust ranges between 0 and 5. - Migration reated measures The respondent s country of birth and age at migration are used in our anaysis to construct migration reated measures. SHARE respondents are asked whether or not they were born in the country of interview which enabes the sampe to be divided into Document de travai n 46 - IRDES - Mars

9 two categories: the native-born popuation and the immigrant popuation. In addition, immigrants were asked to report their years of migration into the country, which enabes to create a proxy of ength of residence using their age at migration (which is equa to years of migration minus years of birth). The acknowedged positive consequences (OECD 2011) on socia mobiity, job market participation, and access to socia right, make the citizenship criterion a potentia additiona candidate for our anaysis. Nonetheess, in most OECD countries an immigrant needs to be resident in the host country for a number of years before being eigibe to the nationa citizenship (OECD, 2010). As a consequence, immigrant naturaisation is highy correated with immigrant ength of stay in the host country or with their age at migration (OCDE, 2011). In addition, we beieve that the ength of stay (measured here through age at migration for oder peope) captures more accuratey the ong-term dynamic of the process of immigrant assimiation. The ast reason not to consider extensivey the citizenship variabe ies in the fact that interaction terms between migrant and citizenship ead to severe sampe size reduction. In spite of a that faws, the citizenship variabe has not been totay discarded of the anaysis since it wi eventuay be discussed in the descriptive statistics and in the robustness checks. - Other individua covariates Foowing previous studies we consider usua individua characteristics. We first consider demographic characteristics ike age (continuous) and gender. Education is divided into three different eves: primary eve of education, secondary eve and tertiary eve of education. Marita status is a dichotomous indicator which indicates whether (or not) respondents are married and finay, the abour market status enabes to distinguish between empoyed from unempoyed, retired or inactive. Descriptive statistics of these covariates are given in Tabe A1 in the appendix. - Context variabes In ine with the empirica iterature, we empoy institutiona variabes describing the socio-economic context and the generosity of wefare state regimes. We have been imited by the number of macroeconomics variabe because the sma number of European countries considered in the anaysis ead to a dramatic ack of variance. For instance, civi iberties and poitica rights indicators dispay the same maxima scores for most country 1. Finay, the foowing five indicators are considered in the anaysis 2 : -The annua growth of the Gross Domestic Product (GDP) refects changes in the voume of production for the 14 European countries in We retained the rea annua growth of GDP to remove the variation in GDP caused infation. -The GINI coefficient of income inequaities is based on equivaent househod disposabe income after taxes and transfers for the 2005 year. It vaue ranges between 0 which indicates a perfect equaity and 1 in the case of perfect inequaity (when a income goes to the individua with the highest income). - The Corruption Perception Index (CPI), computed by Transparency Internationa, The annua growth of GDP, the GINI coefficient of income inequaity and the Tota pubic expenditure indicators are taken from the OECD database whie the Corruption perception index and the ethnic fractionaisation index are taken from respectivey from the transparency internationa ( surveys_indices/cpi) and from Aberto Aenisa own cacuation ( romain.wacziarg/papersum.htm). Document de travai n 46 - IRDES - Mars

10 measures the perceived eve of pubic sector corruption. The CPI ranges between 0 indicating a highy cean and not corrupted country and 1 meaning that the country is highy corrupted. - The ethnic, inguistic and reigious fractionaisation Indexes, derived from Aesina et a. (2003) measure various forms of socia heterogeneity in each country. These indexes refect the probabiity that two randomy seected individuas from a popuation beonged respectivey to a different ethnic, inguistic or reigious group. - The tota pubic expenditure as a percentage of GDP measures the tota amount of pubic expenditures that is devoted to different socia programs such as od age, incapacity, heath, famiy, active abour market program, unempoyment, housing and other socia programs. It incudes cash benefits and benefits in kind for We then considered separatey severa type of socia program (unempoyment, od age, heath and incapacity 3 ) in order to expore a different effect of nationa pubic spending by type of expenditures. Descriptive statistics of these macroeconomic series are given in Tabe A3 in the appendix. Among the 14 European countries, Sweden and Denmark are the most income egaitarian countries whie Itay and Poand appear to be the ess egaitarian with GINI coefficients equa to 0.35 and 0.37 respectivey. Figures concerning the corruption perception index indicate that Sweden and Denmark are the owest perceived corrupted countries whie Greece, Czech Repubic and Poand appear to be the most perceived corrupted countries. The Czech Repubic and Ireand dispay the highest rate of GDP growth in 2005 whereas their tota pubic expenditure seems to be the owest among the 14 European countries. Finay, the most ethnicay and inguisticay diverse countries are Begium and Switzerand with an ethnic fractionaised index of and respectivey and a inguistic fractionaised index of and respectivey. Conversey, Sweden and Denmark are the ess ethnicay diverse countries whie Greece and Ireand are the ess heterogeneous in term of inguistic and reigious disparity Socia Capita of Oder Migrants at a Gance Tabe 1 dispays some raw descriptive statistics of the sampe. It is made of 6.7% of migrants, ranging from around 15% in the main recipient countries for immigration (Germany, Switzerand and France), to ess than 5% in Mediterranean countries (Spain, Greece, Itay) and post-communist countries (Poand, the Czech Repubic) that can be seen as sending countries. Due to potentia sampe seection, these descriptive statistics are not directy comparabe with nationa statistics. According to OECD (2010) statistics, the share of immigrant popuation in our sampe is underestimated in Austria, Sweden, Netherand, Spain, Itay, Denmark, Greece, Switzerand, Begium, Czech Repubic and Ireand whie it is overestimated in Germany, France, Poand. The same evidence can be found regarding age at migration. In SHARE, migrants generay arrive rather young in the host country (22 years od on average) whie the average age at migration according to Eurostat is 28 for women and 32 for men in 2008 (Eurostat, 2008). Notice that disparities in immigrant response rate across countries may expain the above differences. Nevertheess, adjusting for individua characteristics in the statistica modes usuay corrects such a bias 4. 3 Data not reported but avaiabe upon request. 4 The use of individua caibrated weights aready reduced such differences Document de travai n 46 - IRDES - Mars

11 Tabe 1 Description of the Sampe Country N. Obs. Share of migrants Age at migration Raw Weighetd OECD data Average Median Austria Germany Sweden Netherands Spain Itay France Denmark Greece Switzerand Begium Czechia Poand Ireand Tota Source: SHARE wave2 (reease 2.3.0) and OECD (2010). Note: Figures presenting nationa statistics are taken from OECD (2010); data cover the year 2007 for most countries, whie data for Poand are ony avaiabe for 2001, and 2002 for Itay and Greece. Descriptive statistics (see Tabe A2 in the appendix) provide a cross-country overview of the composition of oder Europeans socia capita in its cutura and structura dimensions. On average, 45.8% have high eves of generaised trust (weighted), and 40.1% of the popuation aged 50 and over are invoved in socia activities (weighted). This ast figure is higher than the ones recorded by the Eurobarometer because our measure incudes more diverse items. By and arge, it seems that the two dimensions of socia capita are compementary since countries with high proportion of their oder popuation invoved in socia activities, aso have the highest rate of peope trusting each other. Not surprisingy, there is a north-south gradient in socia capita where northern countries ike Sweden, Netherand or Denmark have the highest eve of socia capita. Document de travai n 46 - IRDES - Mars

12 Figure 1.1. Profie of Socia Capita Among Natives Ony Note : Weighted Statistics (average weights for Ireand) Figure 1.2. Profie of Socia Capita Among Migrants Ony Share of the 50 + popuation with eves of trust > 5 Share of the 50 + popuation with eves of trust > 5 Note : Weighted Statistics (average weights for Ireand) Document de travai n 46 - IRDES - Mars

13 Figures 1.1 and 1.2 suggest that this main scheme remains true beween the migrant popuation and the natives. However, consistenty with previous research on immigrant socia capita 5 (Breton, 2003; De Pao et a., 2007; Aeksynska, 2011), the migrant popuation seem to systematicay under-invest in socia capita. From this perspective, internationa migration is considered as disturbing for popuation because it impies an adaption into a new environment. Immigrant popuation shoud, for instance, adjust their cutura habits, their socia bonds and their institutiona knowedge and skis (Breton, 2003). As the descriptive statistics indicate, such transformations may reduce opportunities to participate in society and may be obstace to socia integration. 4. Method 4.1. Overview The anaysis of the determinants of socia capita foows a two-step mutieve strategy. In the first step, individua determinants of socia engagement and of interpersona trust are anaysed with specia attention to the infuence of the ength of residence in the host country. In the second step, we intend to expain cross-country differences in the speed of integration, i.e. which institutiona variabes describe best the previous reationship between migration statuses and socia capita. Since this ast concept is approached here by two variabes, we computed two-stage equations of socia participation and trust. Considering that invovement in socia activities and interpersona trust are correated which each other (Brehm & Rahn, 1997), the two stage equation strategy enabes to estimate simutaneousy both equations which gives not ony more efficient measure of the coefficients but aso gives more accurate standard-errors. More specificay, the first step of the anaysis aims at studying the infuence of migrants ength of residence on the probabiity to get invoved in socia activity and to trust in other peope Individua eve modes Mode 1 thereafter heps investigate the first assumption that, ceteris paribus, migrants differ from the native popuation by their ength of residence in the host country (H1). Formay: with k=1, 2. [1] where y* 1i and y* 2i are atent variabes representing respectivey individua i s socia participation and interpersona trust, and A i stands for age at migration. Respondents age, gender, marita status, education, abour market status and sef-assessed heath status and the constant are inserted in X i. Finay, d ij represents the J-1=13 country dummies ( j = 1,, 14) and ε 1i and ε 2i are the error terms that are assumed to be normay distributed. Mode 1 thus consists of a system of two Probit equations (aso Bivariate Probit) with the same independent variabes, so that coefficients α 1, β 1, γ 1 and α 2, β 2, γ 2 can be simutaneousy estimated with Maximum Likeihood. Notice that the correation coefficient of the error terms ε 1 and ε 2 wi be denoted ρ (rho) thereafter 6. 5 Apart from Poand and Itay, but the ow rates of migrants in these countries suggest that there may be some statistica imprecision about the previousy mentioned rates. 6 Notice that various specifications for Mode 1 have been tested. For instance, we expored the possibiity that the ength of residence coud actuay concea some generationa effects. Distinction of the ength of residence for peope who migrated before and after 1970 was considered and no generationa effects was reveaed. A modes specification and statistica programs (.do) avaiabe upon request. Document de travai n 46 - IRDES - Mars

14 Mode 2 woud determine whether the effect of the ength of residence on socia capita varies by country (H2). To test this assumption, we have created interaction terms between coefficients associated to the 14 country dummies and the one associated with age at migration. The third mode is simpy estimated from the foowing expression: Notice that the standard specification of equation [2] woud suppement the interaction term A ij xd j with both of its components A ij and d j in order to isoate the pecuiar effect of each term. Notice that N i pays the same roe A ij, woud pay in mode [2] since both variabes have by construction the same variance. However, the sma sampe size prevent us from such an approach since the high number of country dummies woud capture most of the inter-country variance. As a stopgap soution, we propose to specify a country-custered type of variance-covariance estimator that woud produce robust standard-errors in the case of such a misspecification. This robust estimator aows for intragroup correation, reaxing the usua requirement that the observations be independent. In other words, the observations are independent across countries but not necessariy within countries. Such a specification is systematicay appied to modes 1 and Country eve regressions with k=1, 2. [2] Estimations from Mode 2 wi be used to expain differences across countries in immigrant speed of socia integration. In order to expain these differences, we wi use the coefficients associated to the interaction terms as a new dependant variabe that wi be regressed on a set of country institutiona variabes. In this ast step, we test for the impact of the economic and socia context (measured by the growth of GDP, the GINI coefficient of income inequaity, the Corruption perceived index and the Fractionaised indexes) and the impact of the resources devoted to wefare programs (measured by socia expenditure as a percentage of GDP) on countries performance in speed of integration (H3). Formay: with k=1, 2. [3] where μ 1j and μ 2j are the vectors of the coefficients previousy estimated in Mode 2. They represent cross-country disparities in speed of socia integration (i.e. the correation between socia capita variabes and immigrants ength of residence). Our concept of speed of socia integration attempts to highight differences between countries in the necessary ength of stay to achieve socia integration of immigrant through socia participation and interpersona trust. Mode 3 anayses the respective infuence of each institutiona variabes (Z j ) at a time to expain these differences. For each country, five Z j specifications are retained: the 2005 growth in GDP, the 2005 vaue of the GINI coefficient, the 2005 corruption perception index, the fractionaisation index (for anguage, ethnicity and reigion respectivey) and finay the 2005 socia expenditure as a percentage of GDP. Notice that equations for μ 1j and μ 2j in Mode 3 are not estimated simutaneousy. Since μ 1j and μ 2j are inear, coefficients a 1, b 1 and a 2, b 2 in Mode 3 are simpy estimated using Ordinary Least Squares. Document de travai n 46 - IRDES - Mars

15 5. Resuts 5.1. Individua eve mode estimates concur with findings in previous studies Tabe 2 dispays the resuts for Modes 1 and 2. Notice that the correation coefficient (ρ) is significanty different from zero at 1% eve, which confirms the need to estimate simutaneousy both equations. Estimations of baseine Mode 1 concur with previous empirica studies on the individua determinants of socia participation and interpersona trust. The coefficients are significant apart from gender and they dispay the expected coefficients. Notice that resuts from Mode 2 are simiar to Mode 1 with regards to the effect of age, gender, educationa eve, marita status, market abour status and sef assessed heath status. Tabe 2 Individua Determinants of Socia Capita MODEL 1 Dependant var. Socia Participation Generaised Trust Indep. var. Coef. Robust S.E. Coef. Robust S.E. Socio-Demo. Migrant Non-migrant (Native) ** Age at migration *** Age (years) *** Gender (1=man) * Married or coupe (1=yes) *** ** Education Primary Ref. Ref. Ref. Ref. Secundary 0.199*** *** Tertiary 0.464*** *** 0.02 Empoyement status Occupied *** *** 0.02 Other Ref. Ref. Ref. Ref. Sef-assessed Heath Exceent or very good 0.202*** *** Other Ref. Ref. Ref. Ref. Country dummies FR-France Ref. Ref. Ref. Ref. AT-Austria *** *** DE-Germany *** *** SE-Sweden 0.546*** *** NL-Netherands 0.353*** *** ES-Spain *** *** IT-Itay *** *** DK-Denmrk 0.344*** *** GR-Greece *** SW-Switzerand 0.261*** *** BE-Begium 0.152*** *** CZ-Czech Rep *** *** PL-Poand *** *** IE-Ireand 0.206*** *** Constant 0.877*** *** Rho 0.107*** 0.01 Obs Source: SHARE wave2 (reease 2.3.0). Legend: * p<0.10, ** p<0.05, *** p<0.010 Document de travai n 46 - IRDES - Mars

16 Tabe 2 (cont d): Individua Determinants of Socia Capita MODEL 2 Dependant var. Socia Participation Generaised Trust Indep. var. Coef. Robust S.E. Coef. Robust S.E. Socio-Demo. Age (years) ** ** Gender (1=man) Education Primary Ref. Ref. Ref. Ref. Secundary 0.237*** ** Tertiary 0.592*** *** Empoyement status Occupied *** Other Ref. Ref. Ref. Ref. Sef-assessed Heath Exceent or very good 0.351*** *** Other Ref. Ref. Ref. Ref. Migration Non-migrant (Native) *** Interaction terms Age at migration X FR-France Ref. Ref. Ref. Ref. Age at migration X AT-Austria *** Age at migration X DE-Germany *** Age at migration X SE-Sweden 0.014*** *** Age at migration X NL- Netherands * *** Age at migration X ES-Spain *** Age at migration X IT-Itay *** *** Age at migration X DK-Denmark *** Age at migration X GR-Greece *** Age at migration X SW-Switzerand ** Age at migration X BE-Begium Age at migration X CZ-Czech Rep *** *** Age at migration X PL-Poand *** *** Age at migration X IE-Ireand *** Constant *** 0.17 Rho 0.160*** 0.03 Obs Source: SHARE wave2 (reease 2.3.0). Legend: * p<0.10, ** p<0.05, *** p< As expected in the case of the socia participation equations, age is a decreasing factor whie higher educationa eve is one of the most important drivers of socia participation. Exceent/very good sef-assessed heath status aso improves the ikeihood to get invoved in socia activities. However, being married/iving as a coupe is a decreasing factor of socia participation, just ike being in empoyment. One reason coud be the individua time constraints reduce the eisure time that socia participation requires. Document de travai n 46 - IRDES - Mars

17 Country dummies corroborate the existence of a north-south gradient in participation in socia activities that aready appeared in descriptive statistic. In the case of the equations for generaised trust, age does not appear to have any significant infuence on the probabiity to report high eve of trust. Notice that there is a significant gender difference in favour of women with regards to generaised trust. A high eve of education, being married, and being in empoyment, aso increases the probabiity to report a high eve of trust. Reporting exceent/very good sef assessed heath status is associated with a higher ikeihood to trust other peope. Finay, coefficients associated with country dummies indicate that iving in France greaty decreases the probabiity to trust other peope. Previous researches having estabished a simiar resut have argued that the shortage of trust in France may resut from extensive corporatism and State contro (Agan & Cahuc, 2007) Differences in migrants socia capita are associated with the ength of residence In ine with previous studies (Aeksynska, 2011), we suspect a different infuence of migration on socia integration according to the ength of residence in the host country. The equation for socia participation in Mode 1 confirms our intuition: peope having migrated at oder ages present a ower ikeihood to get invoved in socia activities. As predicted by the assimiation process, the onger is the ength of residence in the host country, the higher is the probabiity for immigrant to get invove in socia activity. Looking at Mode 1 equation for generaised trust, age at migration does not seem to be significanty associated with the probabiity to report higher eves of trust. Therefore, we do not confirm a different infuence of migration on trust according to the immigrant ength of residence in the host country. Nevertheess, being a migrant is significanty associated with interpersona trust so that the native-born popuation presents a higher probabiity to trust other peope, but this difference is not due to the ength of residence and shoud therefore be investigated esewhere. In this ater perspective, it may be that some country specific effects are competing so that the overa coefficient for the poo of European countries may be ni. It is the purpose of Modes 2 and 3 to expore this issue. Document de travai n 46 - IRDES - Mars

18 Figure 2.1. Cross-Country Differences in Speed of Socia Integration (Socia Participation Equation) Note: Fixed effects controed for a set of individua variabes. Reference is FR (France) Figure 2.2 Cross-Country Differences in Speed of Socia Integration (Generaised Trust Equation) Coeffs. of Crossed-terms from Bi-Probit Coeffs. of Crossed-terms from Bi-Probit Note: Fixed effects controed for a set of individua variabes. Reference is FR (France) With the aim to anayse more specificay the effect of ength of residence in each country, we inserted interaction terms between age at migration and country dummies in Mode 3 to expore differences in immigrant socia integration through European countries. Figure 2.1 and 2.2 dispay the associated coefficients of Tabe 2 (mode 3) with the 95% confidence intervas. Estimates from the socia participation and the Document de travai n 46 - IRDES - Mars

19 generaised trust equations confirm a other things being equa that there are crosscountries differences in the speed of socia integration: Scandinavian countries (Sweden and Denmark) have high rates of speed of socia integration, whie most Mediterranean countries (Spain and Greece) have the owest rates Speed of integration is higher in fair societies Resuts dispayed in Tabe 3, show that neither GDP per capita, nor the eve of socia expenditure (as % of GDP), nor any of the fractionaisation indexes seem to expain differences in speed of socia integration across countries whether through socia participation or through generaised trust. With regard to the atter dimension, one may notice that the concept of interpersona trust is argey disputed. Athough no difference in the interpretation of the generaized trust construct is to be found between natives and immigrants (Dinensen, 2011), Eoi (2009) warns that it suffers from a ack of internationa reiabiity. Some researchers (Aesina & La Ferrara, 2000), have even guarded about interpretations that resut from the anaysis of interpersona trust in an internationa comparison. Tabe 3 Cross-country Determinants of «Speed of Integration» MODEL 3 Dependant var. Socia Participation Generaised Trust Indep. var. Coef. S.E. R² Coef. S.E. R² GINI Index *** GDP per capita Socia Expenditures Corruption Index *** Ethnic Fractionaisation Language Fractionaisation Reigious Fractionaisation Obs Source: SHARE wave2 (reease 2.3.0) and Macroeconomic series. Legend: * p<0.10, ** p<0.05, *** p<0.010 Document de travai n 46 - IRDES - Mars

20 Figures 3.1 Speed of Socia Integration and Income Inequaity (Socia Participation Equation) Figure 3.2 Speed of Socia Integration and Perceived Corruption (Socia Participation Equation) Coeffs. of Crossed-terms from Bi-Probit Coeffs. of Crossed-terms from Bi-Probit Document de travai n 46 - IRDES - Mars

21 Nevertheess, in the case of socia participation, estimates for the GINI coefficient and the corruption index are both negative and significant. Figures 3.1 and 3.2 show that both modes nicey fit the data, despite the scarce number of observations. These resuts are necessariy expanatory since the sma number of countries in the anaysis does not provide extensive statistica power. The negative sope indicates a decreasing reationship between crossed effects and both macroeconomic indicators. Accordingy, immigrants in more egaitarian countries have a higher propensity to get invoved in socia activity when one immigrates at oder ages. Put differenty, resuts indicate that the necessary ength of stay to achieve immigrant socia integration through socia participation woud be ower in more egaitarian countries. Simiary, immigrants residing in more corrupted country have a ower propensity to get invoved in socia activity when one immigrates at oder age. This atter resut aso suggests that immigrant socia integration through socia participation woud be more ikey to be achieved in countries where perceived corruption is ower. In this respect, income inequaity and state eve perceived corruption appear to exert a deeterious effect on immigrants speed of socia integration in the host country. One interpretation coud be that unequa environments increase reative deprivation of immigrant, reduce their wish or their opportunity to take part in society. In ine with Raws (1971) theory of justice, in a fair society each person shoud benefit from equa basic rights and iberties so that each person has the same right, freedom and capacity to access services and resources (see aso Sen, 2000). Societies characterised by ow eve of income inequaities or ow eve of perceived corruption may thus be considered as fair society in providing a suitabe environment to foster immigrant socia integration Robustness checks With the aim to gain confidence from the resuts, different specifications for Mode 2 have been tested though restricted to the socia participation equation ony. We modified the popuation of interest to check whether the infuence of macroeconomic variabes remains significant in considering different sub-popuations: (i) men versus women, and (ii) citizen versus non-citizen. Tabe 4 Robustness Cheks for «Speed of Integration» with Subsampes MODEL 3 Dependant var. Socia Participation Indep. var.\ Subsampes Citizen Non-Citizen Men Women GINI Index * ** GDP per capita Socia Expenditures Corruption Index ** *** Ethnic Fractionaisation Language Fractionaisation Reigious Fractionaisation Obs Source: SHARE wave2 (reease 2.3.0) and Macroeconomic series. Legend: * p<0.10, ** p<0.05, *** p<0.010 Tabe 4 indicates that the infuence of the GINI coefficient and the perceived corruption index remain significant in the socia participation equation ony in the sub-sampes of women and Document de travai n 46 - IRDES - Mars

22 citizens. Additiona resuts for other macroeconomic variabes concur with previous resuts since no effect of GDP, socia expenditures, and any of the fractionaisation indexes are found to be significant. Notice that non-significant resuts in the case for men and non-citizens may be due to the ack of statistica power. Nevertheess, the opportunity to interpret more in the detais the above resuts shoud be taken with great care; the interpretation beow are just given as to iustrate of possibe future research paths. In the case of socia participation, (i) women s speed of integration is more infuenced by a change in income inequaity (coeff. = ) than the tota popuation of migrants (coeff. = ); and (ii) citizen s speed of integration is ess infuenced by a change in income inequaity (coeff. = ) than the tota popuation of migrants (coeff. = ). Put differenty, for a given eve of income inequaity, the speed of socia integration in a country wi therefore be higher for migrant who benefit from the citizenship of the host country, suggesting that citizenship is a substitute for reduction in income inequaity. The same reasoning appies to resuts for the perceived corruption index that remains significant with the expected sign in the case of women and citizen subsampes. A decrease in the eve of the perceived corruption index woud ead to higher speed of socia integration for women (coeff. = ) and citizen (coeff. = ) than for the tota popuation of migrants (coeff. = ), meaning that state eve of corruption exerts a more deeterious effect on women and citizen immigrant. 6. Concusion This paper provides new empirica evidence on the reationship between migration and socia integration. It expores the hypothesis that migrants essentiay differ from non-migrants with regard to the ength of residence in the country. This time difference is argued to be a proxy of migrants socia distance to natives. In order to capture the important time-depth dimension that is required here, the study makes use of data from the wave 2 of SHARE ( ) on individuas aged 50 or more in 14 European countries. Focus on oder peope heps investigate the infuence of respondents migratory status and their age at migration on two aspects of socia integration: participation in socia activities and high eves of generaised trust. Athough age at migration does not seem to have any significant infuence on generaised trust, socia participation increases with migrants ength of stay in the host country, suggesting that migrant s socia behaviour becomes over time simiar to that of natives. In other words, socia integration of immigrant in Europe generay takes time to be effective. The anaysis based on crossed effect between age at migration and country dummy reveas some differences in the speed of socia integration across European countries that foow a North-South gradient. Scandinavian countries seem more abe to incorporate rapidy migrants; whie Mediterranean countries seem to perform much worse. Institutiona determinants were further considered in the anaysis in order to investigate these cross-country differences in the speed of socia integration. It appears that the socia integration of migrants is much faster in fair countries with ower eves of income inequaity and ower eves of corruption. From a pubic poicy perspective, our resuts suggest that mainstreaming migrants socia integration coud be more difficut to achieve in countries with higher eves of income inequaity and corruption. For a given eve of migratory infux, countries abiity to socia absorption is more important as the eve of income inequaity and corruption is ow. May this resut be confirmed in further research, it woud suggest that some of the potentia negative externaities of migration due to some socia distance between migrants and natives are more ikey to be downsized (i.e. compensated or internaised) as societies grow more equa and more free. Document de travai n 46 - IRDES - Mars