Ethnic Identity and Labor Market Integration: Evidence from Denmark (work in progress, not for quote) Cédric Gorinas 1 November 2010

Transcription

1 Ethnic Identity and Labor Market Integration: Evidence from Denmark (work in progress, not for quote) Cédric Gorinas 1 November 2010 Abstract Immigrants integration and attachment to both their country of origin and the receiving country have become central in the political debate. In this context, this paper explores ethnic identity of first and second-generation immigrants in Denmark. I use recent survey data that cover an extensive set of personal, human capital and post-migration to investigate both identity formation and the link between ethnic identity and employment. Based on Berry s (1997) acculturation framework identity is modeled through both direct and indirect measures, successively following the methodologies developed by Nekby and Rödin (2007), Battu and Zenou (2010) and Constant and Zimmermann (2008). First, I find that identification with the majority is positively associated with post-migration such as the practice of Danish at home, education from Denmark, and in some cases with years since migration. To some extent, some ethnic and religious groups tend to reject the majority culture more often than others. Second, preliminary results do not indicate large differences in employment probabilities between assimilated and integrated migrants. However, individuals with strong minority identification and no majority attachment seem to pay a penalty with regards to labor market integration. A distinctive result is that oppositional preferences for descendants are found to be positively correlated with employment. Using more objective composite indices to measure ethnic identity shows that employment is associated with local language fluency, contacts with locals, education taken in the host country and experienced discrimination, rather than with feeling of ethnic identity as such. 1 Department of Economics, Århus School of Business/Århus University and The Danish National Center for Social Research, cgorinas@asb.dk. 1

2 1. Introduction Ethnic identity has become a recurrent topic in both the public and the political arenas. This can be illustrated by the recent heated debate in France and Denmark about forbidding the wear of the Islamic burqa. Ethnic identity is even now an essential component of immigration policy in several countries such as France, Germany, the UK, Denmark, the US and Australia (Casey and Dustmann, 2010). In Denmark, this for instance illustrates by the introduction of a citizenship test in 2005, where all applicants to the Danish citizenship have to prove a minimum knowledge of Danish values, culture, history and language. In parallel with ethnic identity, the labor market integration of immigrants constitutes in Denmark an essential political topic. This gave birth to a Ministry of Refugees, Immigration and Integration Affairs in 2001 and to an increasing focus on how to accelerate the entry of immigrants into the labor market, clearly stated in the objectives of the new Integration Act of 2004 (Liebig, 2007). The country however distinguishes in the European Union as having one of the widest employment gaps between immigrants and natives (MIPEX, 2008; Bisin et al., 2010a). Several economic studies have looked at immigrant integration in Denmark (see among others Skyt-Nielsen et al., 2001; Schultz-Nielsen, 2001; Rosholm and Vejlin, 2007; Clausen et al., 2009) but to my knowledge no economic study has been conducted on the topic of ethnic identity. Looking at the relationship between immigrants identity and employment may nevertheless help to understand the poor economic integration of immigrants, as ethnic identity may induce behavioral patterns detrimental to one s economic success (Akerlof and Kranton, 2000; Battu et al., 2005; Austen-Smith and Fryer, 2005; Casey and Dustmann, 2010). A national identity is characterized by several elements such as history, customs, language and religion. Most citizens of a country can agree on common habits and cultural traits, seen as typical for this country. However, all inhabitants of a country may not share the same attachment to the same national identity. The feeling of national identity is only relevant when the individual is faced with other national cultures and identities (Gundelach, 2004). It is then easy to understand that, for ethnic minorities, i.e. individuals who left their country of origin to settle in another country and their descendants, this feeling manifests on a daily basis. Indeed, an immigrant may have mixed feelings regarding which country, national values and culture she identifies the most to. For example, an Iranian person who has permanently moved to Denmark can feel either mostly Dane or mostly Iranian. It is also possible that she simultaneously expresses a strong (weak) attachment to both the country of origin ( minority identity ) and the hosting country ( majority identity ). Identification with the country of origin and identification with the host country are independent from each other (Berry, 1980, 1997). Not feeling a strong attachment to the majority culture immediately raises possible conflicts of misunderstanding between locals and ethnic groups, due to the non-acceptance of each other s identity. One reason for this could be that, since nationals admit their nationality feeling without question, they may not understand the complexity of the identity dilemma ethnic minorities have to deal with (Gundelach, 2004). Another possible reason behind possible conflicts is given by the theories of evolutionary psychology: hostility towards outsiders, i.e. individuals from the minority, and cooperation with insiders, i.e. individuals from the majority, is a component of human nature and has been a condition for survival throughout human evolution (Akerlof and Kranton, 2000). The violent riots that occurred in 2001 in several cities in England illustrate this (Battu and Zenou, 2010). In Denmark, an example can be the debate about social cohesion and the integration of Muslims in the Danish society, a topic touched upon everyday in the media (Hellström, 2010). Hence, 2

3 ethnic identity is part of individual s well-being, and it is also relevant for the society as a whole (Nekby and Rödin, 2007). Economic research on the topic of national identity is quite recent (see among others, Pendakur and Pendakur, 2005, Battu et al., 2003, Nekby and Rödin, 2007, Constant and Zimmerman, 2008). Most of this literature investigates the determinants of national identity formation and the consequences for individual labor market outcomes. As explained in Akerlof and Kranton (2000) and in Casey and Dustman (2010), identity can influence negatively or positively economic outcomes in several ways. On the one hand, an individual with a strong minority identity can be reluctant to interact with nationals due to her own choice or to peer pressure. This hinders the development of social networks with locals that can be helpful to find jobs. Minority identity may also be associated with visible behavioral patterns such as dress codes and with beliefs and preferences with regards to certain jobs. On the other hand, belonging to an ethnic group can improve one s chances to find a job within the ethnic group s own social network. All in all, the existing empirical economic studies on ethnic identity differ in terms of sample, model used and results. It is then difficult to emphasize what the implications of ethnic identity are in a clear-cut manner. Hence, in order to investigate what the implications of ethnic identity for Denmark are, an empirical investigation is necessary. Not only will an economic study on immigrants identity bring some more light on the reasons for possible conflicts arising from cultural misunderstandings in a country. It will also help to understand the employment situation of ethnic groups, and in that way it will yield useful guidance for post-immigration policies. For instance, such a study will inform whether an integration policy is more suitable than an assimilation policy in order to accelerate the labor market entry of immigrants (Battu and Zenou, 2010). Former descriptive evidence suggests that most immigrants do increasingly not feel Danish, depending on their origin (Gundelach and Nørregård-Nielsen, 2007; Necef, 1994). The purpose of this paper is to investigate what lies behind ethnic identity formation and whether identity and employment for ethnic groups are linked. The analysis will be based on direct (chosen) identity measures, and not indirect as often done (Pendakur and Pendakur, 2005). An additional contribution lays in the data itself as it covers both first and second-generation immigrants who moved to Denmark in majority for humanitarian reasons. Exploring results across ancestries is also relevant in order to build a more cohesive society and to better understand the labor market integration of descendants 2. I use a cross-section dataset collected by Gundelach and Nørregård-Nielsen in a 2006 survey. It covers an extensive set of personal, human capital and other such as opinions on diverse topics, economic and housing conditions, as well as both pre-migration and post-migration variables. Based on Berry s (1997) acculturation framework, identity is modeled through both direct and composite measures, successively following the methodologies developed by Nekby and Rödin (2007), Battu and Zenou (2010) and the Ethnosizer of Constant and Zimmermann (2008). Across the different specifications, the determinants of ethnic identity are found to be related to post-migration, such as education, fluency in the local language and time. Moreover, some ethnic and religious groups tend to identify less easily with the majority culture than others. An example is Iraqis who more often manifest oppositional preferences compared with other ethnicities such as Vietnamese. Another interesting result is the relationship between religion, particularly Islam, and acculturation. From several specifications, it comes out that in general Muslims tend to stay strongly 2 Although it is smaller than for first-generation, there is still an employment gap between second-generation migrants and Danes (Statistics, Denmark, 2010). 3

4 committed to their country of origin. They are therefore less likely to assimilate and more likely to reject the majority culture. Furthermore, results from the employment equations seem to indicate that a strong ethnic identity does not harm immigrants labor market outcomes, as long as it is coupled with a strong majority identity. More particularly, using both chosen identity and composite measures, it can be shown that what matters more than chosen identity itself is interaction with locals, language fluency, being a male, education taken in the host country as well as an index for modernization. In addition, having experienced discrimination although not in a working context is negatively correlated with employment. A final result concerns descendants, for whom oppositional preferences are found to be positively correlated with employment. All in all, these results can only in part confirm the existing evidence in Europe. They may nonetheless constitute valuable knowledge for future policy-making in the highlydebated field of labor market integration of ethnics. The rest of the paper is organized as follows. Section 2 reviews the existing literature on ethnic identity looked at from an economic perspective. Section 3 presents the data and how ethnic identity is measured. The empirical findings are then exposed in Section 4 exploring both the identity formation question and the relationship between ethnic identity and employment. Section 5 finally concludes. 2. Literature Review: Methods and Findings 2.1. Measuring Ethnic Identity In several economic studies, ethnic identity has been directly measured using survey with the following type of questions: To what extent do you consider yourself name of the nationality in the receiving country? and To what degree are you attached to your country of origin?. Answers are usually scaled not at all, a little, partially, totally. Other questions indirectly related to ethnic identity are sometimes used. For instance: Do you mind if a member of your family marries a person from the majority culture?, and Do you plan on staying permanently in this country? Depending on the available data and the method preferred by the authors, identity is modeled either linearly or on a two-dimensional scale. In a linear model, the attachment to either one of the two cultures is dependent and exclusive: the most the individual identifies to the host country, the less she identifies with her country of origin and vice-versa. In this case, measures of the personal attachment to either the host country or the home country are sufficient. This model is clearly explained in Zimmerman and Constant (2008) and is illustrated in Figure 1. 4

5 (0,1) Ethnic loss and commitment to the host country ( ( (0,0) (1,0) Ethnic retention and commitment to the country of origin Figure 1: Ethnic identity illustrated linearly. Source: Constant and Zimmermann (2008), Figure 1. The vertical axis represents attachment and commitment to the receiving country, the horizontal axis to the country of origin. An individual can only move along the diagonal line from (0,1) to (1,0), the two extreme cases. (0,1) corresponds to the assimilation state where the person feels a very strong attachment to her new country of residence and no commitment at all to her home country. At the opposite hand, (1,0) represents the situation where individuals are purely ethnic as they only identify to their country of origin. This is called a separation acculturation state. This representation is somehow restrictive; cases where the individual feels a strong, a weak or no attachment to both the majority and the minority identities at the same time cannot be accounted for. This becomes possible in a two-dimensional model; ethnic identifications to the minority and to the majority are no longer dependent and exclusive. The two-dimensional model is based on the theoretical framework developed by Berry (1997), called acculturation framework, illustrated in the figure below. Degree of attachment the receiving country Very strong and strong Little or none Degree of attachment to the country of origin Very strong and strong Integration Separation Little or none Assimilation Marginalization Figure 2: Measuring National Identity in a two-dimensional model: the Acculturation Framework (Berry, 1997) In this framework, four acculturation states are possible. If the ethnic individual feels strongly attached to both countries, she falls into the integration category. Should the strong attachment only be felt towards the host country, the person will be referred as assimilated. When the individual reject majority identity but is deeply commitment to the country of origin, she will be seen as separated. Finally, marginalization characterizes the case where migrants feel attached neither to their home country nor to the hosting country. It has often been found that acculturation states evolve with the time spent in the hosting country (see for instance Nekby and Rödin, 2007; Gundelach and Nørregård-Nielsen, 2008). Authors of economic studies dealing with ethnic identity base their analysis either on the one-dimensional or the two-dimensional model. Most recent studies argue that the latter is more accurate, as evidence suggests that minority and majority identities are not mutually excludable and dependent (Nekby and Rödin, 2007), and as using a linear model may overestimate immigrants attachment to the host country (Constant et al., 2006a). However, in many cases a linear model can be thought satisfactory as long as the 5

6 variables from a survey are in accordance with the identity definition used in sociology the extent to which individuals define themselves as a member of a group, the definition used in Akerlof and Kranton (2000) self-perception of group belonging, as well as the definition commonly found in the public debate. Moreover, other survey questions covering related topics such as the participation in social clubs and social interaction with nationals can be used as complementary measures for identity attachment. For instance, Battu et al. (2003, 2005) construct an aggregate measure of ethnic identity of immigrants. For this they use several variables in top of self-identification with the majority culture, such as attitude towards interethnic marriage and preferences in terms of ethnic grouping at school for their own child. Battu et al. (2003) use religion and language to proxy minorities attachment to their culture of origin. In this way, they can construct a measure close to a two-dimensional model. Having access to direct identity measures for both the minority and majority identity, Constant and Zimmerman (2008) construct an aggregate twodimensional model to measure ethnic identity using language, media, self-identification, ethnic networks and future citizenship plans. The purpose is to balance the subjectivity of the direct measure of self-identity in the survey by other more indirect measures. More on these aggregate measures will be presented in the coming sections Theoretical Background Most recent economic studies use the framework developed by Akerlof and Kranton (2000) as a starting point. Akerlof and Kranton s intention was to show how identity a person s sense of self can influence economic outcomes. They develop a framework based on utility maximization and relate it to three domains where identity plays a determinant role: gender inequality at the workplace, exclusion and poverty, and the economics of the household. The issue of oppositional identities among ethnic groups comes into play directly under exclusion and poverty. In their model, identity influences individual behavior guided by possible payoffs (i.e. increase in utility level) even though it may incur self-destruction (i.e. social exclusion and unemployment). Payoffs are associated either with the individual s own actions or with others actions. Working within the dominant culture may induce a feeling of betraying oneself, and it may incur externalities for the rest of the minority group who also feels betrayed. For example, Akerlof and Kranton s identity model has been used by Fryer and Levitt (2004) in order to account for the possible utility associated with the choice of distinctive names among black communities. Moreover, Akerlof and Kranton (2000) explain that individuals chose to some extent their own identity and that it may be the most important economic decision they make. Adopting a self-destructive behavior is generally the product of low economic endowments and a high degree of social exclusion. Akerlof and Kranton (2000) use their findings to make policy recommendations: replacing minority individuals into less ethnic-concentrated areas (through schooling and special labor programs such as Job Corps) reduces possible loss in identity associated with work or decreases the level of social exclusion. Another theoretical framework has been developed by Battu et al. (2005) in order to understand the impacts of identity in term of individual labor market outcomes. They start up with Akerlof and Kranton s (2000) assumption that immigrants (or Non-whites 3 ) chose whether they want to adopt an oppositional behavior (identity), although they know it may harm their labor market outcomes ( self-destructive behavior ), i.e. they choose their location in the social space. There are several reasons for that, 3 Non-whites are defined according to their social environment (family, friends and neighbors) and their culture of origin (religion and language). 6

7 depending on individual s own decision and peer pressure 4. The central assumption is that non-whites are reluctant to assimilate, and their utility grows as they are more and more distant from the whites norm ( negative bias towards assimilation ). Hence, immigrants will adopt partially or totally an oppositional behavior towards the white s culture that may detriment their labor market outcomes. Another assumption is that the individual s utility is conditioned on the peer group s (or ethnic group s) level of utility. Moreover, the authors assume that jobs can only be found through social networks as vacancies are only communicated by word-of-mouth. Therefore, having an employed white in one s social network increases one s job finding probability. However, there is an externality of being close to a white person, which is conditioned on peer pressures, discrimination, and expected rewards from interacting with the whites (see p15-16). This externality will determine the individual location in the social space. All in all, this model shows that whites do have the most favored labor market outcomes (unemployment rate and job finding probability), since they are not discriminated against and have a wider network, while the contrary is true for those who adopt an oppositional identity (x=1 in the social space). This however does not imply that they are worse off so long there is a utility gain from not interacting with whites. Their work has several policy implications (see p 18). One of them is that increase in non-whites unemployment insurance (or guaranteed jobs) is positively associated with oppositional identity Review of Findings Economic research in the area of ethnic identity is relatively new. It is difficult to compare the results across the few existing studies as they rely on different assumptions, look at different populations and use different measures for ethnic identity. However, some similar patterns can be drawn Ethnic Identity Formation Both pre-migration and post-migration have been found to influence identity formation of immigrants. For instance in Battu, Mwale and Zenou (2003), the reject of majority identity, measured through an aggregate variable called oppositional identity, appears to be negatively correlated with language fluency, being UK-born, inter-ethnic marriage and, to a lesser extent, years since migration and owning education from outside the UK. Interestingly enough, the study does not show that tertiary education acquired in the UK makes non-whites become less oppositional. On the other hand, they find that co-ethnic concentration in the living area is positively associated with oppositional identity. Moreover, some ethnic groups more often show preferences for oppositional identity than others. Similar conclusions still hold in the more recent study of Battu and Zenou (2010). Constant and Zimmerman (2008) look at identity formation of immigrants in Germany relying on Berry s (1997) acculturation framework. They employ a different aggregate identity measure, namely the Ethnosizer 5. In their study, the feeling of ethnic identity is above all conditioned by pre-immigration, such as age at entry, religion, nationality and education in the home country. They do not find either any influence from qualifications gained in the receiving country. Nekby and Rödin (2007) use self-identification measures as directly given by the respondents to a Swedish survey. Their sample is mainly composed of second-generation migrants under 25. The authors 4 See the introduction in Battu et al. (2005) for more details 5 More details on these different measures will follow in the next section. 7

8 equally find a correlation between national background and ethnic identity. Moreover, they highlight the importance of gender, i.e. female are more attached to the host country. Quite interestingly in this study, secondary and tertiary education is associated with minority identity attachment, but not with majority. Nekby and Rödin (2007) furthermore emphasize on the no-correlation between labor market status and identity. Finally, local language proficiency positively affects strength of majority identity. Other non-economic studies have also looked at the determinants of identity. Piontkowski et al. (2000) explain that contacts inter-ethnic groups leads to integration and assimilation states. On the contrary, experiencing discrimination will be likely to fulfill the two more oppositional states: separation and marginalization. Gundelach and Nørregård-Nielsen (2007) use Berry s (1997) framework and take Piontkowski et al. s (2000) findings as starting hypotheses. For this, they compute correlations between variables related to both contacts and discrimination when looking at the determinants of national identity. They find that, in the case of Denmark, contacts matter, while discrimination is not influential. The time spent in Denmark has some effect in the sense that it decreases one s state of separation and increases the likelihood that one becomes more assimilated. However, there is no effect of time found on integration. The present paper distinguishes from Gundelach and Nørregård-Nielsen (2007) by applying several econometrics methods from the economics literature to the question of ethnic identity formation and it investigates in particular the link between identity and employment. In those studies, it is foremost pre-migration that seem to influence identity choice of ethnics. Indeed, the effect of education acquired in the receiving country has not been shown, while national and religious backgrounds appear to be determinant. However, the time spent since migration is also found to influence identity formation Ethnic Identity and Labor Market Outcomes Labor market implications of ethnic identity cannot be clear cut from the existing literature. Pendakur and Pendakur (2005) look at identity effect on labor market integration processes (i.e. the use of informal networks and the quality of jobs). They find that immigrants of European origin with a strong ethnic identity use informal networks to find a job more often than non-whites. Moreover, deep ethnic identity is associated with lower job quality, measured by the Blishen scores in the study, for men in visible minorities. Battu et al. (2003) look at oppositional identity impact on employment probability and find a negative effect (7 percent less likely to be employed). Other variables, mainly post-migration such as fluency in the local language, years since migration and owning some UK-education, have a positive influence on employment. The same authors (2005) develop a theoretical model - the oppositional identity model - that supports these results: oppositional non-whites are less likely to find a job. Again this finding is confirmed in a recent empirical study by Battu and Zenou (2010). Another study is the one of Constant et al. (2006) where ethnic identity of immigrants in Germany is modeled through Berry s (1997) two-dimensional scale. They find significant differences across genders in terms of factors affecting employment probability. For women, age at entry is determinant while years since migration is not. The opposite is true for males. Regarding schooling, females employment is significantly dependent on pre-migration education while men s probability to work is affected by postmigration schooling. In addition, religion is found to be correlated with employment, though only for men. Besides pre and post-migration, ethnic identity is significantly found to affect employment probability. For males, marginalization and separation acculturation states have a negative effect on 8

9 employment in comparison to assimilation, while integration has no significant effect. The story is quite different for females; only integration appears significant and positively influences employment probability. This supports the authors hypothesis, which states that integrated individuals have more chances to find a job than assimilated, as they can exploit a bigger network. Casey and Dustmann (2010) also look at German panel data through a two-dimensional scale. Quite distinctively from other studies, they find little evidence of a relationship between the attachment of firstgeneration immigrants with either the receiving or the home country and labor market outcomes. This is mainly the case for females. However, as far as the second generation is concerned, a positive relationship between ethnic identity and employment and a negative relationship with unemployment are found. Also based on the acculturation framework of Berry (1997), Nekby and Rödin (2007) study ethnic identity and employment relationships for second and middle generation (born of foreign parents in Sweden or who moved at a very early age) immigrants in Sweden. The main common feature between their and the present study is the comparison of results based succesively on self-reported identity measures and on a composite index. They find that what counts for one s employment probability is the strength of identification with the majority. In this case, the extent to which an immigrant identifies with her ethnic background does not matter. It should be kept in mind that their sample is mainly composed of descendants. In a more recent study in 2010, using the same sample, the same authors show only small differences in terms of employment between assimilated and integrated migrants regardless of gender. Moreover, they are able to confirm their 2007 results as separated individuals are significantly less likely to have a job than assimilated. This finding holds however only for men after controlling for human capital. Finally, Bisin, Pattachini, Verdier and Zenou (2010a) study the relationship between ethnic identity and employment for 20 EU countries using data from the European social survey. They are not able to use direct self-identity measures but instead resort to a composite index combining the language spoken at home, the importance of religion and of cultural norms and traditions. They are able to show that having a strong ethnic identity induces a penalty in the labor market. This is also the case for descendants, although they integrate better than their parents when not controlling for a strong ethnic identity. The authors are able to estimate a causal effect using an instrumental variable strategy for part of their study, and this allows them to conclude that identity influences employment. Moreover, they look at the impact of labor market and integration policies on the relationship between identity and employment. Those policies may improve immigrants integration depending on the strength of minority identification. All in all, the existing literature seems to indicate that identification with the majority may matter for an immigrant s labor market situation contrarily to the attachment to the minority identity. It is though difficult to conclude on a clear-cut relationship between ethnic identity and labor market outcomes as the studies adopt various model assumptions and identity measures. Moreover, national labor market and integration policies may influence this relationship (Bisin et al., 2010). In this paper, both the factors behind ethnic identity formation as well as the relationship between ethnic identity and employment are explored for Denmark. The main difference with the existing studies is the measurement of ethnic identity using several measurement methods for ethnic identity. Another reason why this study is relevant lays in the composition of the sample used. This one mainly consists of non-western immigrants who moved for humanitarian reasons, i.e. the group with the greatest difficulties to enter the Danish labor market, and their descendants. Finally, other variables related to experienced discrimination, religion and other ethnic values are also available and constitute a novelty in comparison to 9

10 some earlier studies. 3. Ethnic Identity and Data 3.1. Available Sources As discussed above, the investigation of ethnic identity requires survey data. In Denmark, several are accessible. The two main sources are the Ethnic Groups Values (thereafter EGV) survey conducted by Gundelach and Nørregård-Nielsen in 2006, and the Institute for Social Research (SFI) survey collected in two waves: 1988 and The EGV survey is preferred as it contains information on both immigrants and descendants, and as it covers direct measures of ethnic identity regarding both the country of origin and the receiving country, just as in other studies (Nekby and Rödin, 2007; Constant and Zimmermann, 2008; Battu and Zenou, 2010). Hence, the two-dimensional acculturation framework can directly be applied to the EGV dataset 6. Information relative to ethnic identity from these two datasets is summarized in the two following tables in order to provide a broad picture of the topic. The tables present the share of the respondents according to their answers to the direct self-identity questions in each survey. The EGV survey shows that most immigrants and descendants express a strong attachment to both their country of origin and to Denmark, and this holds across gender (Table 1). Moreover, it can be seen that a larger share reports a little minority attachment than a majority attachment. Table 1: Repartition of the answers to the ethnic self-identity questions in the EGV survey, percentage of total respondents. Own calculations. 7 Attachment to Denmark Attachment to the country of origin Overall M F Overall M F Very strong Quite strong Minor None Don t know / no response As a comparison, the answers collected in the SFI surveys are also presented (Table 2). It should be kept in mind that almost all respondents are second-generation immigrants who have never lived in their country of origin. Table 2 shows that ethnic identity changes with time towards a stronger majority attachment, a weaker minority attachment or a mix of the two. 6 More on the dataset follows in Section 4.The use of the SFI-surveys is planned for future research focusing on the economic achievement of second-generation foreigners only. 7 Ethnic Danes, included in the sample for control purposes, are not included in these calculations. 10

11 Table 2: Answers to the question: Do you see yourself mostly as a Dane or as a (ethnic origin)? in percentage of the total sample Mostly Dane Mostly ethnic Equally Imprecise mix Neither 3 2 Don t know 2 2 In addition, results from another survey conducted in 1994 conducted among young immigrants of Turkish origin, follow. Here, ethnic identity is modeled linearly. Quite interestingly, this survey shows that this ethnic group tends to identify mostly with their origin and not with the majority, or express an equal commitment to both cultures. Table 3: Answers to the question: Do you see yourself mostly as a Dane or as a Turk/Kurd? in percentage of the total sample? 9 Only as a Dane 2 Mostly Dane 12 Equally 42 Only as a Turk/Kurd 20 Mostly Turk/Kurd 19 Don t know /no response 5 It is interesting to notice that a stronger feeling of identification with the minority is depicted in the SFI surveys at least in the first wave (Table 2) than in the EGV survey (Table 1). This is certainly due to the different sampled population across the two databases. SFI looked at descendants from guest-workers immigrants, while the EGV dataset mostly counts family-reunification and asylum-seeking immigrants as respondents. This is confirmed in Table 3 with young Turkish descendants Data Used The data used in the present study originates from the EGV survey counting 129 variables and covering ethnic minorities values. The survey was conducted in 2006 through phone and internet interviews and counts 4478 respondents. Only particular ethnic groups are represented in the survey: immigrants and descendants from Turkey and Pakistan, immigrants from Vietnam, West Balkan countries, Iraq and Iran. Moreover, 520 ethnic Danes were interviewed. As indicated in the summary statistics below most of the sample encompasses refugees or family reunion migrants. The selection of the sample was random and delivered upon the Danish Ministry of Refugees, Immigration and Integration Affairs request by the central registration office. The dataset was furthermore weighted to correct for the bias with regards to gender 8 The 1988 and 1999 surveys cover the same population, observed in two waves, augmented of around 300 new respondents in 1999 due to attrition. 9 This survey was conducted 594 among young people of Turkish origin in Denmark in Results are freely accessible on Dansk Data Arkiv (first investigator: Mehmet Ümit Necef DDA-5272, 1. edition (with Birgitte Grønlund Jensen and Heidi Wittendorff Sørensen) 11

12 and age within each ethnic group. The original purpose of the database was to provide material for a sociological study ordered by the Danish Ministry of Refugees, Immigration and Integration Affairs 10. The 129 questions cover individual opinions and values across diverse domains such as democracy, children education, work and ethnic identity. First, the survey is perfectly suited for the two-dimensional framework presented above in Figure 2. Indeed, the question of self-identity is approached from both a minority and a majority perspective, i.e. individual identification to both the country of origin and the receiving country. Second, the survey contains a large number of other variables indirectly related to ethnic identity and similar to the ones used in other economic studies. Examples of these variables are ethnic preferences in terms of media, partner, children s acquaintances as well as language. As a result, this survey enables measuring ethnic identity through the acculturation framework using direct self-identification reports as in Nekby and Rödin (2007). In addition, it covers the necessary information to construct aggregate indices such as the Ethnosizer of Constant and Zimmermann (2008) and the Oppositional Identity variable of Battu and Zenou (2010). Hence, the models developed in these three economic studies will be applied to the Danish data. The variables used in the present study are presented in the table below. These summary statistics are restricted to the sample including ethnic minorities only (ethnic Danes are excluded). The mean age was 33 overall, 36 for immigrants and 23 for descendants. The respondents had in average spent around 15 years in Denmark at the time of the interview. Two third are married and around the same proportion has at least one child. Descendants represent one third of the sample. They are either of Turkish or Pakistani origin, these two countries being hence the two most represented in the sample (all in all representing around 50 percent). Iran, Iraq, Vietnam and West Balkan countries each represent about 12 percent of the sample. These latter proportions reflect the shares in the whole Danish population since 2006 (Statistics Denmark, 2010). Relative to the other religions, Islam is overrepresented; almost 75 percent of the sample. This has to be kept in mind when analyzing the findings. As already exposed almost all respondents migrated to Denmark for asylum seeking or family reunification motives. These have for years been major reasons of migration to Denmark for numbers of non-western newcomers in comparison to other EU countries (Liebig, 2007). Work and study migrants only represent six percent of the respondents. Let us now turn to post-migration. About 60 percent of the whole sample state that they rarely or never experience problem with the Danish language (Fluent in Danish). However, only 53 percent speak at Danish at home. Around 10 percent have a Danish partner. About one third lives in an area where more than 40 percent of the population is from an ethnic minority. Although this information is self-reported, it is reasonable to believe that this is representative of the non-western immigrant population in Denmark, as 25 percent of foreign-ethnics live in socially-excluded areas (Gundelach and Nørregård Nielsen, 2007). With regards to education, 40 percent went to a Danish high school institution, 46 percent has undergone some education in Denmark, whereas only 23 percent attended school in their home country. As far as economic are concerned 60 percent of the foreign respondents are either employed or self-employed and about 30 percent own their accommodation. Finally, it can be seen that most of the interviewees live in or around Copenhagen (54 percent) or in mid- and northern Jutland (18 percent), where the second and fourth biggest cities Århus and Ålborg are located. 10 See Gundelach and Nørregård-Nielsen (2007) 12

13 Table 4: Means by gender and ancestry, and definition of the variables. Source: EGV survey and own calculations All M F Imm Desc Definition: Equal to 1 if the respondent Number of observations Individual and pre-migration Male is male Age Age of in 2006 Descendant is descendant Children has children Married is married Schooling home went to school in the country of origin Origin and religion Turkish has a Turkish background Pakistan has a Pakistani background West Balkan countries is from a West Balkan country Iraq is from Iraq Iran is from Iran Vietnam is from Vietnam Christian is Christian Muslim is Muslim Atheist is Atheist Other religion believes in another religion or doesn t know Reasons for migration Refugee migrated as a refugee Family reunification migrated for family reunification motives Study or work migrated in order to work or to study Post-migration YSM Years since migration Employment is in employment Fluent in Danish is fluent in the Danish language Danish spoken at home speaks Danish at home Ethnic minorities in the living area lives in an area where more than 40 percent of the population is from an ethnic minority 11 Having a Danish partner has a Danish partner High School in Denmark has gone to high school or equivalent in Denmark Higher education in Denmark has been enrolled in a tertiary education in Denmark Large Copenhagen lives in Copenhagen and around Rest of Zealand lives in the rest of Zealand Fyn lives on Fyn South Jutland lives in the Southern part of Jutland Mid and North Jutland lives in the Northern or mid part of Jutland When looking at summary statistics across genders, a few numbers stand out. The most striking difference is, as one could expect, that the main motive for migration of females is family reunification. Refugees are much less represented among females. Ethnic females are also remarkably less in employment than their male counterpart: 48 against 71 percent respectively. Moreover, females are less often fluent in Danish, and they have more rarely a Danish partner (7 percent against 12 percent of foreign males). Otherwise, the 11 This variable directly relies on the interviewees answers but reflects register data. 13

14 repartition between the two genders is rather fair between the remaining variables. Differences in terms of means can also be observed across ancestries. First, one will notice that there is a higher share of Muslims among descendants than among immigrants. This is due to the fact that the descendants sample is only composed of persons of Turkish or Pakistani origin, two countries from which most respondents are Muslims. Moreover, it is interesting to note that the proportion of atheists is higher among immigrants than descendants. Finally, second-generation respondents own more advanced abilities in Danish and they use the language in their everyday life in much larger proportions (76 percent vs. 47 percent for their first-generation counterparts). As explained by Gundelach and Nørregård-Nielsen (2008), this is induced by the fact that most descendants go to school in Denmark, and they hence use the Danish language every day. Indeed, despite their young age in the sample, descendants show higher schooling rate than immigrants (64 vs. 31 percent for high school, 72 vs. 37 percent for higher education for descendants and immigrants, respectively) Construction of additional variables I define other variables that were not included in earlier economic studies on ethnic identity. Summary statistics and definitions of these four additional dummies are given in Table 5. First and similar to Bevelander and Veenman (2006), a measure of modernization is constructed. The question: To what extent do you think the following is acceptable? is asked about three elements: homosexuality, divorce and abortion. The answers are scaled from 1 to corresponding to: It is totally acceptable. When a respondent reports an answer graded 7 or above at least once, the dummy variable modernization is equal to 1. To see if this variable is relevant for ethnic people, one can compare the responses between ethnic Danes and ethnic minorities. For homosexuality, 83 percent of Danes score 7 or above while this is only the case for 29 percent of foreign ethnics. For acceptance of abortion, the respective shares for ethnic Danes and non-ethnic Danes are 75 percent vs. 32 percent, for divorce 77 percent against 42 percent. Second, a discrimination variable is defined. Across four questions interviewees were asked how often they experience racial discrimination at the police, in transports, in the street and at the bank. If they answered often or very often for at least one of the four locations, the dummy discrimination_exp takes on the entire value. 16 percent of the sample often experience discrimination, and there is no much variation across genders. However, descendants feel that they often experience discrimination in much larger proportions than immigrants (25 against 13 percent, respectively). A possible explanation is the fact that this group is much more sensible to discrimination experience, as in many aspects such as language ability and education, they look much alike with natives. The difference in terms of physical appearance is the main difference (Gundelach and Nørregård-Nielsen, 2008). Third, another variable called marriage_parentsdecision is constructed and set equal to 1 if the respondent thinks that who to marry should remain a parental decision. This is the case for one third of the sample, and interestingly enough this variable is more often coded 1 for the young second-generation respondents than firs-generation migrants. Fourth, the dummy variable badon_democracy is defined. It takes on the entire value when the respondent reports a disagreement with the fact that Denmark is governed by a democratic system. This variable s mean is very low (4 percent) and, while it is the same for men and women, it is slightly higher for descendants. Finally, not_stay_indk is a dummy taking on the unit value when the interviewee does not 14

15 manifest the wish to stay in Denmark for the rest of her life. Its overall average is 18 percent, and it is slightly higher for males than for females, and for immigrants than for descendants. Table 5: Summary statistics and definition of additional variables Variables Mean M F Imm Desc Definition Number of observations Modernization Equals 1 if the person complies the index constructed at least across one of the dimensions used (see above for more details) Marriage_parentsdecision Equals 1 if the respondent thinks that whom to marry should only be a parental decision Badon_democracy Equal to 1 if the respondent has a bad opinion about the fact that Denmark is governed by a democratic system Discrimination_exp Equal to 1 if the respondent has ever felt discriminated in public places as transports, in the street, at the police or at the bank Not_stay_inDK Takes the unit value if the respondent wants to leave Denmark at some point in the future 3.4. Ethnic Identity Measurement for Denmark Chosen identity As exposed above, the survey used for the present study covers questions relative to both majority and minority identity where the respondents are directly asked to evaluate their own feeling of national identity. More explicitly, the questions asked are: i) How strong is your attachment to Denmark? ii) How strong is your attachment to your country of origin?. Such a measure will be referred either as direct, selfreported or chosen identity 12. Respondents had the choice between five answers. The availability of such questions direclty enable the application of the model used in Nekby and Rödin (2007) to our data. The shares of each answer by ancestry and origin for each of the questions are presented in the two following tables. 12 The term chosen identity is used in Nekby and Rödin (2010) by opposition to composite measures, i.e. forced identity. 15

16 Table 6: Direct self-identity measure: How strong is your attachment to Denmark? Shares by answers, ancestry and country of origin. Source: EGV Database. Ancestry Origin Very strong Quite strong Minor No attachment Don t know Immigrants Turkish Pakistan Iraq Iran Vietnam West Balkan Descendants Turkish Pakistan Table 7: Direct self-identity measure: How strong is your attachment to your country of origin? Shares by answers, ancestry and country of origin. Source: EGV Database. Ancestry Origin Very strong Quite strong Minor No attachment Don t know Immigrants Turkish Pakistan Iraq Iran Vietnam West Balkan Descendants Turkish Pakistan The above tables depict that most immigrants report a strong attachment to both Denmark and their country of origin. Regarding commitment to Denmark, between 29 to 38 percent of the population of immigrants manifests a very strong attachment. Shares are even higher among descendants. Quite distinctively, the share of migrants who report a minor attachment only varies between 10 and 23 percent. When turning to the country of origin, the shares of very strong attachment are slightly lower, ranging from 20 to 36 percent except for the Iraqis. Indeed, half of them report a very strong degree of attachment to Iraq. This is interesting as this is the only group for which the strength of minority identity is more important than the strength of majority identity. Another interesting ethnic group is Iranians, since most of them report a minor attachment to Iran. In addition, it should be noticed that descendants remarkably more often express a minor identification with minority identity than their first-generation counterparts (14 and 22 percent). The presence of the two questions related to self-identity in the EGV survey makes it possible to categorize each respondent in one of the four acculturation state of Berry s (1997) model (see Figure 2): integration, assimilation, separation or marginalization. This is depicted in the figure below for immigrants 13. Integration is by far the acculturation state that is mostly reported, followed by assimilation, separation and marginalization. 13 Including descendants gives a very similar pattern. Using only the first-gerenation here is done for comparison purposes with the other identity measures presented below. 16