Interethnic Partnerships of Western Europeans : Between Preferences and Opportunities

Previous studies on determinants of social integration of migrants in the destination countries, and of interethnic partnerships in particular, converge in attributing importance to the same set of variables. This study aims at providing a further test of the generality of findings across different contexts using survey data of intra-European adult migrants, a group which differs in many respects from the hitherto mainly analyzed migrant groups. Highskilled labor, study, retirement, and “quality of life” migration are well represented, while low-skilled labor migration which dominates traditional research in the field is of minor importance, yet still present.


Research desiderata
There are some research gaps which this paper intends to fill: First, most previous research is based on one CoR and, thus, only enables to draw conclusions about the relative performance of different groups of migrants in one context.This does not only apply for studies which aim at explaining interethnic partnerships, but integration in the CoR in general.The study by van Tubergen (2004; see also van Tubergen and Kalmijn, 2005) is an exception in this respect in that it offers a comparison of a large array of CoO groups in several CoRs.However, it does not include interethnic partnerships as a dependent variable.Second, the studied migrant groups are selective with regard to the legal and socio-economic status of migrants.German studies concentrate on low-skilled labor migration and family reunification (Haug, 2002;Kalter and Granato, 2002).Unfortunately, high-skilled labor, study, retirement, and "quality of life" migration, which are becoming increasingly important both in a quantitative and a qualitative sense (Castles and Miller, 1998;King, 2002;King et al., 1998;Salt, 1992Salt, , 1997)), are underrepresented in existing data.The same applies when it comes to explaining interethnic partnerships in particular.The prolific literature on marriage migration focuses on women from developing or less developed countries to the highly developed world.Third, most studies, in particular those comparing different countries, are based on official data.In such data, the number of variables included is usually small, and information on language proficiency at the time of migration, previous sojourns in the CoR or third countries, and migration motives are missing altogether.Finally, previous studies concentrate on partnerships of migrants with members of the main ethnicity of the CoR.Partnerships with members from third countries living in the CoR get out of focus, though they might have a similar relevance for social integration as marriages with CoR nationals, in particular as regards integration into supra-national units, such as the European Union.
This study, thus, aims at making several contributions to the existing literature.First, it will provide a further comparative test of the generality of previous findings, using survey data of intra-European adult migrants.Due to the nature of intra-European migration, highskilled labor, study, retirement, and "quality of life" migration are well represented alongside low-skilled labor migration.In addition, partnerships between members of countries which are more or less equal in socio-economic and political terms can be analyzed.Second, it will take into account several variables which are not available in official data on which comparisons of different contexts have usually been based on.These include retrospective information on language proficiency at the time of migration, previous sojourns in the CoR and other countries, and migration motives.
For several potentially relevant variables, the direction of causation is uncertain (Stevens and Swicegood, 1987).Among them are other indicators of social integration such as having CoR or thirdcountry friends which might affect interethnic partnerships, but which are much more likely affected by interethnic partnerships.In the following, the variables used to explain interethnic partnerships will therefore be restricted to the clearly exogenous ones.

Variables affecting interethnic partnerships
The following variables are truly exogenous in the sense that there is no reciprocal effect of interethnic partnerships going back to them.The effect of age at migration is well established in the literature as a determinant of partnerships with CoR nationals.Migrating at higher ages is connected to a higher probability for having married in the CoO already, and most likely a partner from one's own ethnicity.However, the effect cannot simply be assumed to be linear.After the age at which most people have found their (marriage) partner, the negative effect of age at migration is supposed to decelerate.With regard to partnerships with third-country nationals, age at migration should not have an effect, unless the probability to meet third-country nationals is different in the CoO and the CoR.
The longer a migrant's duration of stay in the CoR, the higher the probability of getting a partner from the CoR should be.This applies even to migrants who came to the CoR with a partner from their home country or were even married already.Many of these partnerships or marriages may dissolve for reasons as different as separation, divorce, or death of the partner as the stay in the CoR continues.However, this effect cannot simply be assumed to be linear either, as marginal effect of duration of stay is also decreasing.With regard to the probability of getting a partner from a third country, there should be no effect of duration of stay, for reasons similar as in the case of age at migration.Language knowledge at the time of migration should display a strong positive relationship with partnerships with CoR nationals.Language proficiency in the CoR language might also have facilitated partnerships with migrants from third countries, as it enhances communicative capabilities with this group in the CoR.
Interethnic partnerships also increase with the educational qualification of a migrant.For highly educated migrants, costs can be assumed to be lower (they know how to deal with cultural diversity) and mating opportunities might be larger, i.e. in institutions of higher education.This should apply both to partnerships with CoR and thirdcountry nationals.
Migration motives are usually not included in studies based on official data.Instead, ancillary measures such as visa category (Chiswick, Lee and Miller, 2006) or characteristics of the CoO such as political suppression in the CoO (van Tubergen, 2004) are employed.The present study extends the analysis of migration motives to a wider spectrum.In Europe, the free-mover type (Favell, 2008) is becoming more frequent, which to some degree resembles internal migrants in other parts of the world.Free movers, as the name indicates, make highly individualized moves, independent from chain migration.They might go to a different country out of curiosity and the desire to get immerged in a different culture.This might reveal an intrinsic interest in getting in touch with foreign people in general and a foreign partner in particular.
Study migrants, i.e. those who originally moved in order to attend a university -irrespective of whether they are still students at the time of the interview -should show the highest level of interethnic partnerships.They were exposed to the CoR marriage market relatively early in their lives, before stable same-ethnic partnerhips could develop.Individuals who migrated for family or love motives should have a higher probability of interethnic partnerships, unless they just followed their migrating partner.And for retirement migrants, the likelihood to have a CoR partner should be reduced.The effect of the motives, however, should be very much dependent on third variables, i.e. the sociodemographic composition and the actual behavior of migrants.When certain structural variables are taken into account, motives should not play a role any more, i.e. the effects of migration motives should disappear when appropriate variables, e.g.age at migration and sojourn in the CoR or a third country, are controlled for.
Variables usually not included in comparative studies are whether migrants had already been to the CoR or a third country for an extended period of time before the final move to the CoR.It is obvious that a previous sojourn in the CoR might already have brought the migrant into contact with (potential) partners from the CoR so that the relationship might even have already existed at the time of final migration.A previous stay also increases the duration of exposure and should work in the direction of the effect of duration of stay.Failing to take a previous sojourn into account necessarily leads to an underestimation of the latter.In addition, a previous sojourn might have been used for getting some formal education in the CoR.Therefore, this variable might partially capture some of the education effect which would be overestimated without taking a previous sojourn into account.With regard to partnerships with third-country nationals, a previous sojourn in the CoR should have no effect, for the reasons outlined above for age at migration.
Migrants with a previous sojourn in a third country should also have a higher probability to get a partner from the CoR, though the effect should be weaker than in the case of a stay in the CoR itself.Sojourns in third countries are assumed to improve above all the migrants' rapid adaptability to new environments and new people.A previous sojourn in a third country should, however, have a stronger direct impact on the probability to have a partner from a third country.This is for the same reasons that a previous sojourn in the CoR will increase the probability to end up in a partnership with a CoR member.Note that the relationships assumed here are not symmetric, as a previous sojourn in the CoR does not demonstrate a particular flexibility on the part of the migrant (as a previous sojourn in a third country does).
To sum up our hypotheses: For interethnic partnerships with CoR nationals, we expect the following variables to have strong effects: age at migration, the duration of the sojourn, CoR language proficiency at the time of migration, education, a previous stay in the CoR, and the family/love motive.A previous sojourn in a third country should only have a weak effect.For interethnic partnerships with nationals of third countries, only education and a previous sojourn in a third country should be important.

Data and Measures
The analyses presented in the following are based on the "European Internal Movers' Social Survey" (EIMSS), conducted as part of the PIONEUR project ("Pioneers of Europe's Integration 'from Below': Mobility and the Emergence of European Identity among National and Foreign Citizens in the EU", funded by the European Commission in the 5th Framework Programme).In each of the five countries, Britain, France, Germany, Italy, and Spain, approximately 250 telephone interviews were conducted.Interviewees were nationals of each of the other four countries who migrated there from 1974 through 2003, were 18 years of age or older at the time of migration, and have lived in the CoR at least for one year.A total of 4.902 interviews were conducted.By means of linguistic screening of names in telephone directories, only migrants belonging to the CoOs' main ethnic groups were considered.Thus, members of e.g. the Germanic minorities in Italy and France who have migrated to other countries as well as former migrants and their offspring who have returned to the home countries of their parents or grand-parents were excluded.Admittedly, the sampling strategy has its problems, e.g.undercoverage of migrants without an entry in a telephone directory.As this might have affected in particular female migrants married to CoR men, a small network-sampling component was included in the design by asking respondents for telephone numbers of women married to CoR men.
A standardized multilingual questionnaire was administered by bilingual interviewers in computer-assisted telephone interviews.The average duration of the interviews was slightly less than half an hour.Aims were to collect quantitative information of migration experiences, political behavior, attitudes, and European identity.The five national surveys were completed between May and September 2004 (except in Britain, where it ended in January 2005)1 .
There are several advantages of this survey compared to the data bases used in previous quantitative studies of migrant language acquisition: First, the study was conducted in five different countries and the same countries are used both as CoO and CoR.The studies of the five countries are comparable: in each country, random samples of the migrant populations were drawn according to the same sampling schema.In addition to that, the same questionnaire and the same kind of interviewers (bilinguals) as well as the same interviewer instructions were employed.Second, the selected countries are exclusively highly developed countries without huge differentials in economic performance, which forms a contrast to previous studies in particular with regard to the CoO of the migrants.This permits to extend the analysis to groups beyond the hitherto mainly studied groups of refugees/asylum seekers, low-skilled labor migrants, those coming as part of family reunification, and marriage migrants from developing countries.Third, the data also covers variables which usually cannot be obtained from census data, such as language skills at the time of migration and previous sojourn in the CoR and a third country.

The dependent variables
As we are interested in two dependent variables, interethnic partnerships with CoR nationals and with nationals from third countries, we estimate two logistic regressions.In the first, the baseline is constituted by those who have a partner from their CoO or a third country, in the second by those who have a partner from their CoO or the CoR.Migrants having no partner are excluded from the regressions.

Independent variables
Age at migration and duration of sojourn are included together with their squared terms.Unfortunately, the duration of stay in the CoR is confounded with the period of migration, e.g.those who have stayed already for a long time are those who came in an earlier period.In the EIMSS data set, we have no additional variables which we could use to control for this.There is no age variable included in the models, as this variable exhibits perfect linear dependence with age at migration and duration of sojourn.Language knowledge at the time of migration is measured as a self-assessment on a 5-point scale: "And how well do you speak [language of CoR] now?" Response categories offered were "almost as well as native language", "quite well", "just so-so", "poorly" and "no knowledge".This variable is assumed to be linearly quantitative.Education is entered as two dummy variables for upper secondary and university education (with those having a lower secondary education or less constituting the baseline).Previous sojourns in the CoR and a third country are included as two dummy variables.Migration motives were measured by an open question.Four pure types were extracted from the responses: work, family/love, quality of life and study reasons, and a mixed type which mostly represents combinations of quality of life with one of the other reasons.Thus, four dummy variables were entered in the regressions for the pure types, with the mixed type serving as the baseline.
Gender and marital status and their interaction are used mainly as controls, as our study might exhibit certain distortions with regard to the former and clearly has insufficient information of the timing of marriages affecting the latter.As we commented earlier, the sampling design of the PIONEUR study, selecting migrants from telephone directories, might have contributed to an under-representation of women married to CoR men.The main effects and the interaction are included to take this potential bias into account.Gender is a dummy variable with men as the baseline category, i.e. the effects presented pertain to women.Marital status is included by four dummy variables: separated, divorced, widowed, and never married.Married migrants constitute the baseline.
The results section will start with an analysis of the proportions of migrants with a partner from the CoO, the CoR, a third country, and those without any partner in the different CoO/CoR combinations.Then, logistic multilevel models will be estimated.Multilevel or hierarchical linear models (Raudenbush and Bryk, 2002;Snijders and Bosker, 1999) are appropriate when variables pertain to two levels, an individual level (the individual migrants) and a group level (the CoO/CoR combinations) and when the higher-level units can be assumed to constitute a sample from a population (all migrants from any of the EU countries to one of the other countries).Stata Version 10 (Rabe- Hesketh and Skrondal, 2008;StataCorp, 2007) is used for all analyses.

Distribution of partner statuses in the migrant groups
Table 1 presents the proportions of the different partner statuses for the 20 CoO/CoR combinations.The proportion of migrants having a partner from their own CoO is particularly high for Britons in France, Italians in Germany, and Britons and Germans in Spain.It is particularly low for French, Germans, and Spaniards in Britain and for Britons and Spaniards in Italy.On the opposite, the proportion of migrants having a partner from their CoR is particularly high for Italians in France, French and Britons in Germany, French and Germans in Britain, and all migrant groups in Italy.It is particularly low for Britons in France, Italians in Germany, and French, Germans, and Britains in Spain.Partners from a third country are more frequently found among Spaniards in Britain and French and Italians in Spain, and particularly rare among Britons in France, Germany, Italy, and Spain, and Italians in Germany.Finally, the proportion of migrants having no partner is highest for Britons in Italy and lowest for Italians in Germany.
The two Southern European countries are positioned at opposite ends with regard to their foreign migrants having a partner from the country of origin: In Italy, relatively few migrants have conationals as partners, whereas in Spain the vast majority does.For migrants to Britain, having a partner from the own community is not very common.
Mate selection can be analyzed by reference to rational-choice type calculations (Becker, 1991) and opportunity structures shaped by geographical and demographic constraints (Blau, 1994;Blau, Blum, and Schwartz, 1982).The very low proportion of Italians in Germany having a German partner might be due to the latter: the large number of Italians in this country from which these people can choose and the physical proximity to their home country.The result is in sharp contrast to results found on binational marriages in Germany conducted for first-generation migrants (Schroedter, 2004), which show that Italians and Spaniards are the best integrated of the traditional guest-worker populations in this respect.However, these analyses do not include migrants from France and Britain and put a special emphasis on the low integration of Turks in Germany.In addition, they are based on a different definition of first-generation migrants, which includes all those who migrated to Germany after the age of 6.Given the fact that the Italian migrants in our sample came to Germany at an average age of 26, it is obvious that even they were exposed to a high "risk" to marry or at least getting a partner from their home country before they arrived.The logistic multilevel models for partner status Table 2 shows two logistic random-intercept models for partner status.We present odds ratios and z-statistics as indicators of effect size.The first model contrasts having a partner from the CoR with having a partner from either the CoO or a third country.The second model contrasts having a partner from a third country with having a partner from either the CoO or the CoR.

Partnerships with CoR nationals
Age at migration has a curvilinear relationship with partnerships with CoR nationals.The probability to have a partner from the CoR sharply decreases with age at migration but this trend flattens down at higher ages.The opposite applies to the duration of the stay, which increases the probability first and then flattens down.Language proficiency at the time of migration also increases considerably the probability to have a partner from the CoR.Those who have an upper secondary education are not distinct from those with less education, but having a university education heightens the chance to have a CoR partner.However, more important than having a university education is a previous sojourn in the CoR which might have been used to attend a university there.A previous sojourn in a third country has no effect, contrary to our expectation.The only migration motive which makes a difference is having migrated for family/love reasons.This indicates that this motive is predominantly related to coming to or following a CoR spouse, and not to follow a CoO national.Among intra-EU movers, traditional family reunification with already migrated partners is obviously a less common pattern of migration than movements to join a foreign partner in his or her home country.
The remaining variables were mainly included to control for the potential underrepresentation of women married to a CoR (or thirdcountry) national.The effect for women pertains to married women, and shows that these have in fact a lower likelihood to be in a partnership with a CoR national.The main effect of marital status pertains to men and shows that divorced men have a higher probability to have a CoR partner than married men.Finally, the interaction effects between gender and marital status show that divorced, widowed, and never married women have a much higher probability to have a CoR partner.The marital-status effects and their interaction with gender can partly be interpreted in terms of a different (or changed) opportunity structure compared to married people and partly reflect the limitations of our sampling procedure.Divorced men and widowed women might have recently lost the CoO partner with whom they might have originally migrated to the CoR, making them available for a partnership with a CoR national.The (changed) opportunity structure is also relevant for divorced and never married women, but at the same time they might have been easier to contact, given our sampling procedure.

Partnerships with third-country nationals
Neither age at migration, the duration of the stay, language proficiency at the time of migration nor a previous sojourn in the CoR have an impact on having partners from third countries.These variables are largely CoR specific and, thus, it does not come as a surprise that they have an impact on CoR partners, only.Upper secondary and university education and a previous sojourn in a third country, however, do both have considerable and independent effects, i.e. both a general attitude of open-mindedness and opportunities to meet third country nationals (during the previous stay in a third country) seem to matter.The only migration motive which leads to a particularly high incidence of partnerships with third-country nationals is, again, family and love.The effect for women is again negative, but less pronounced than for CoR partnerships.Finally, the separated are more likely to have third-country partners.The interaction between gender and marital status is not significant (due to the lower number of cases in the positive category of the dependent variable), nevertheless divorced and never married women tend to have a higher probability of having a third-country partner.

Conclusions
The migrant groups analyzed here are different from those dealt with by the bulk of studies on interethnic marriages or partnerships.They are more homogenous in so far as they all stem from a western EU member state and cultural and linguistic distance to their CoR can be safely assumed to be relatively modest.On the contrary, the literature has concentrated on marriages either between members of a "guestworker" type of immigrant population or typical marriage migrants from developing (or at least less developed) countries with the main ethnicity.The only migrant group considered in this study which contains remnants of the typical "guestworker" migration are Italians in Germany.
With regard to partnerships with CoR nationals, the study could show how similar processes are operating with quite different migrant groups.This is not only true in terms of structural and demographic factors but especially in terms of migration motives.Personal relationships with individuals of different nationality are a cause and not only an effect of migration in Western Europe.This finding is not circumscribed to some specific combinations of nationalities in selected countries, pointing at the generality of underlying processes.

Table 1 :
Proportions of migrants with a partner from the CoO, the CoR, a third country, and of those without any partner for the different CoO/CoR combinations

Table 2 :
Random-intercept model for partner status (reference category: all other partner statuses except no partner)