Partnership preferences of the Belgian second generation: Who lives with whom?

Partnership

preferences

of

the

Belgian

second

generation:

Who

lives

with

whom?

Anne

Hartung

*,

Ve´ronique

Vandezande

,

Karen

Phalet

,

Marc

Swyngedouw

a_{UniversityofLeuven&CEPS/INSTEAD,Belgium}

b_{DepartmentofWell-Being,PublicHealthandFamily,FlemishGovernment,Belgium} c_{UniversityofLeuven,Belgium}

1. Introduction

Divergent bodies of theory have devoted ample attention to ethnic intermarriage, exogamy and ethnic assortative mating. Marriage is often regarded as an indicatorofthetransmissionofethnicallyspecificcultural values and practices. Song (2009: 332f) explains why intermarriageisconsideredsoimportantfortheorists:for themintermarriageisapracticewhichmayfundamentally affect the boundaries between ethnic minority groups (Barth, 1969). Specifically, increasing rates of intermar-riage might signal fading or shifting boundaries and decrease ethnic prejudices (Kalmijn, 1991). This should besimilarforinter-ethnicunionsingeneral,onwhichwe focushere. In anutshell,thisarticlelooksat patternsof

inter- and intra-ethnic cohabiting partnerships1 of the descendantsofTurkishandMoroccanimmigrants(i.e.the secondgeneration)inBelgiumbasedontheBelgiandata from the TIES project (The Integration of the European SecondGeneration).DespiteBelgium’slargeethnic popula-tionandtheextensiveliteratureintraditionalimmigration countries,thereareonly a few studiesonintermarriage andinterethnicunionsinBelgiumtodate.

From the ‘golden sixties’ onward, Belgian migration statisticsshowalargeandsteadyintakeofforeignlabour intheheavymetalandminingindustriesfromruralareas of Southern countries, such as Turkey and Morocco. Increasingly, foreign workers were also contracted by employersin other industries, construction, and menial

ARTICLE INFO

Articlehistory:

Received17November2010

Receivedinrevisedform7September2011

Accepted7September2011 Keywords: Secondgeneration Belgium Intermarriage Inter-ethnicpartnerships ABSTRAC T

Divergentbodiesoftheoryhavedevotedampleattentiontoethnicintermarriage.Using thedatafromtheBelgianTIESproject(TheIntegrationoftheEuropeanSecondGeneration), thispaperfocusesontheyoungTurkishandMoroccansecondgenerationinBelgiumand, incontrasttootherstudiesinthefield,includescohabitationinadditiontomarriages. Furthermore,itdistinguishesnotonlypartnershipstonativesversuspartnershipsto non-natives but three types of partnerships: those to first generation partners, second generationpartnersand‘native’Belgianpartners.

Our results show, first, that a large part of the second generation lives with first generationcoethnicpartners.WefindsecondlythatmostoftherelationstoBelgian-born persons are in fact relations to partners of second generation from the sameethnic background.Weconcludethatestimationsofintermarriage/cohabitingunionsbasedon relationstofirstgenerationimmigrantsseriouslyunderestimatetheextentofintra-ethnic partnerships.Thirdly,wefindthatnotonlyindividualcharacteristicsbutalsothesocial environmentimpactsonthepartnerchoice.

CrownCopyrightß2011PublishedbyElsevierLtd.Allrightsreserved.

* Correspondingauthor.Tel.:+352585855452.

E-mailaddress:anne.hartung@ceps.lu(A.Hartung).

1_{Welookatinter-ethnicpartnershipscomprisingmarriedand}

non-marriedcohabitingunions.Werefer,however,tothestrictersenseof ‘intermarriage’toquotestudies.

ContentslistsavailableatSciVerseScienceDirect

Advances

in

Life

Course

Research

j o urn a l hom e pa ge :ww w. e l s e v i e r. c om/ l o ca t e / a l cr

jobs. From the middle of the 1970s, Belgiumadopted a restrictivemigrationpolicyandthesincethenmaininflux resulted from family reunification and later family formation (Reniers, 1998). This has profoundly changed the natureofforeign populations:fromtemporary guest workers toresidinghouseholds and minority communi-ties.Duetothetimingofthismigrationandtherelatively highfertilityofthefirstgenerationimmigrantpopulation, thecurrentTurkishandMoroccancommunitiesinBelgium haveanatypicallyyoungagestructure(Lesthaeghe,2000). It is difficult to find numbers on foreign origin populations in Belgium, since most statistics are based on nationality rather than ethnic background. When considering more inclusive categorisation criteria the Turkish origin population is estimated around 2.4% of thetotalpopulationinAntwerpand3.7%inBrussels.For theMoroccanoriginpopulationthisis7.5%inand12.9%in Brussels(Vandezande,Phalet,&Swyngedouw,2011).

FindingsbasedontheBelgianCensusof1991suggest that there is ahigh shareof marriages with aco-ethnic fromthecountryoforiginamongthosewithaTurkishand Moroccan nationality who migrated toBelgium prior to marriageintheperiodbetween1960and1990.Theshare ofsuchcross-bordermarriagetoaco-ethnicishigherfor Turkishnationals(75%formenand69%forwomen)than forMoroccannationalsliving in Belgium(57%)(Lievens, 2000).Also,thesecondgenerationofTurksandMoroccans showrelativelyhighandstableratesofmarriagewitha co-ethnic partner from the country of origin (Corijn &

Lodewijckx, 2009; Lievens, 1997; Lodewijckx, 2010;

Reniers & Lievens, 1997; Reniers, 1998). Similar results were foundfortheNetherlands: Moroccansand Turkish migrantswerefoundtobe(among)thegroupsleastopen tointerethnic marriageand cohabitation(Kalmijn& van Tubergen,2007).Therefore,thisstudyexploresfurtherthe partnershippatternsofthetwolargestsecondgeneration groups in Belgium, the Moroccan and Turkish second generation. As the second generation in Belgium have come ofagein the lastyears, it isa goodtimetostudy family formation in these populations. In addition, the process of union and family formation among ethnic minorities undergoes a rapid change in Belgium: they assimilateintermsofmarriageageandnumberofchildren (Schoenmaeckers,Lodewijckx,&Gadeyne,1999).

Manyempiricalstudieshavetheshortcomingofmerely focusingon‘‘whomarrieswhom?’’,i.e.arelimitedtothe concept of intermarriage and exclude partnerships or cohabitations(Song,2009).Inotherwords,thisapproach ignores important recentsocietal changes.Over the last decades,cohabitationhasbecomearelevantformofliving together (Corijn & Klijzing, 2001; Corijn, 2010). Today many Western countries provide alternative legal alter-nativestomarriagesuchasofficiallyrecognised partner-ships (e.g. Pacte civil de solidarite´ (PACS) in France, samenlevingscontract/contratdeviecommuneinBelgium). Literatureonassortativematingoneducationalmatching betweenthepartnerstakesincreasinglyintoaccountthat young cohorts tend to live together first rather than directlygettingmarried.Corijn(2010)showsthatpeople living together before a possible marriage aregenerally highereducatedandmoreoftennon-religious.Yet,dueto

data limitations, to date there are only few studies on inter-ethnic partnering who include also cohabiting unions (Blackwell & Lichter, 2000; Corijn & Lodewijckx, 2009;Kalmijn&vanTubergen,2007).

In addition,the operational definition of inter-ethnic union often collapses different ‘‘types’’ of partnerships. Whendifferentiating by nationality, partnerships to the secondgenerationare–dependingonthecountry’sdegree ofopennesstowardsnaturalisations–oftencollapsedwith either first generation partnerships (when restricted access to nationality) or are assigned to the ‘native’ category(when open access tonationality). This second drawbackrelated tothequestion ‘‘whomarrieswhom?’’ Throwinglightontheseadditionaldimensions,ourpaper remedies this situation and thus contributes to the empirical literature in the field in two ways. First, we look at the more generally defined concept of union formation or partnership including marriages and also cohabitation.Secondly,wedistinguishpartnershipstofirst generation migrants, second generation and natives. In otherwords, weapply a more precise concept of inter-ethnicunions.

Thisrelatestothefirstaim,whichisofmethodological nature, namely to compare definitions of inter-ethnic partnerships and its empirical implications: first versus secondgenerationversusBelgianpartners.Indoingso,we define partnerships – or unions – as couples living together,married orunmarried.Couples that arelegally bound but not living together are thus not considered. Althoughthe numberofnon-marriedcohabitingcouples among the Turkish second generation is still relatively small,theshareisincreasingnotonlyfornativesbutalso forMoroccans inBelgium,the Moroccansecond genera-tion and for those with ‘‘native’’ partners (Corijn & Lodewijckx,2009:22ff;Schoenmaeckersetal.,1999).

This articlefirst reviews the literature in the fieldof inter-ethnicpartnershipsand thenintroducesthe meth-odologyofthestudy.Wethenpresentdescriptiveresults on inter- and intra-ethnic partnerships in Belgium and estimate a multinomial logisticregression to modelthe factorsfacilitatinginterethnic partnershipsandconclude inthefinalsection.In otherwords,thesecondaimisto predict partner choices of the second generation and scrutinisetheassimilationargument.

2. Literaturereview

As mentioned, many studies do not distinguish between first generation partners, second generation partnersandnativepartners.Collapsingsecondgeneration partnersasfirstgenerationornativesmaycauseproblems for cross-country comparisons and for substantive rea-sons.2 _{Conceptually, such approaches must over- or}

underestimate the true rate of inter-ethnic unions. In

2_{Apartfrommethodologicalreasons,thedifferentiationbetweenthese}

differentkindsofpartnershipsalsoseemstobeimportantforsubstantive

reasons. It poses the theoretical question if partnerships between

membersofthe firstandsecond generationareto beconsideredas

countrieswhoseaccesstocitizenshipcanbecharacterised asopen–likeBelgium–thesecondgenerationwilltendto benaturalised.IfthepartnerisanativeBelgian,thiswould be categorised as intra-ethnic marriage if based on nationality, but as inter-ethnic marriage if based on country of origin. In countries with a more restrictive citizenship law, this case will turn out completely different. Itis thereforeimportantin oureyestolookat both definitions separately. We expect that the concep-tualisationmattersandthatlookingatpartnershipswith persons born in Belgium (including thus the second generation) overestimate the probability of interethnic partnerships severely (H1). As rates of cohabitation are higher thanmarriage rates (Corijn & Lodewijckx, 2009), inter-ethnic marriage is an unsatisfactory indicator of inter-ethnicpartnership.

A similarpoint is also raised by Reniers and Lievens (1997)whoclaimthat,atleastintheBelgiancase,itisnot very useful or sufficient to talk about ‘interethnic’ marriages. More specifically, they found that the vast majorityofthesecondgenerationengagedinamarriage withapartnerfromtheirparents’homecountry,Turkeyor Morocco, who comes to Belgium to marry them. The researchers found that people married with a co-ethnic fromthecountryoforiginandpeoplewhomarriedtoa co-ethnicraisedinBelgiumdiffersignificantly.Thesestudies argue that it is important to distinguish between the migrationgenerationsofthepartnersandtheyshowthat the social and symbolical functions of both types of marriages can be quite different (Lodewijckx, 2010; Reniers,1998).Thisrevealstheproblemofdefiningwhat exactly an interethnic partnership is. Song (2009: 338)

rightlypointsoutthat‘‘iftheboundariesbetweengroups areinflux andarefundamentallymessy,how areweto know which marriages count as incidents of intermar-riage?’’. She argues that a differentiation by nativity is questionable,sincedifferencesbetweennewimmigrants, secondandthirdgenerationsareperceivedassubstantial. Taking these arguments into account, a generational approachseemsmostappropriate.

After discussingdifferent conceptualisations of inter-ethnic partnerships, a next question to ask is what the underlying dynamicsofa differentialpartnerchoiceare. Three different strands of explanations can be distin-guished in the literature on inter-ethnic partnerships: preferences, opportunities and third parties (Kalmijn, 1998;Kalmijn&vanTubergen,2007).Thesestrandslook on marriage behaviour either from a micro or a macro perspective. On the micro level, it is claimed that individuals preferto marry someone with similar back-groundcharacteristicssuchasethnicity(ethnicmatching hypothesis).3Gordon(1964)arguesthatthemost assimi-latedmigrantsaremorepronetohavepartnersfromthe majority.

Blau,Blum,and Schwartz(1982)viewsintermarriage

fromamacroperspectiveandexplainstheprobabilityof intergroup contacts – and thus also intermarriage – by structural or contextual determinants, which create or limit (contact) opportunity and choices. Such structural factors are for instance the size of the ethnic group, heterogeneity(ofone’ssurrounding)(Blauetal.,1982)but alsodegreeofurbanisationoftheplaceoforigin(Huschek, Liefbroer,&deValk,2010;Lievens,1998).

StudiesoninterethnicpartnershipsinBelgiumarerare (Lievens,1998). Lievens (1998)for instance investigates marriage behaviour of the two biggest migrant groups livinginBelgium,TurksandMoroccansandfindsthatboth individualandhigher-levelfactorsdeterminethepartner choice. However, based on census data, Lievens’ study

(1998) is limited to married couples and takes merely

nationality and countryof birth as the indicator forthe ethnicgroups.Thisexcludesbydefinitionthenaturalised second generation, which is born in Belgium. This is a particularproblemwhenusingrecentdataastheaccessto Belgiannationality wasliberalised in2000.Since thena large group of the second generation in Belgium is naturalisedand onlypossible toidentifyviainformation ontheparents.

Yet, the results are of utter importance. Most of the studies include age at marriage, which is seen as an indicatorofmaturityandindependence,asapredictorand findthat thelikelihoodtointermarryincreaseswithage (Kalmijn&vanTubergen,2007;Lievens,1998).Thisisalso hypothesisedhere(H2).Huscheketal.(2010)revealthat the timingof the first union among the Turkish second generation in Europe isnot only influenced by personal characteristicsbutalsothoseofpeersandparents. Inter-ethnic contacts as well as low human capital and rural originoftheparentshaveapostponingeffectonentryto the first union.4 _{Unfortunately, their empirical analysis}

onlyincludestheTurkishsecondgenerationanddoesnot investigatetheeffectofthoseonthepropensitytohavea partner with a different ethnic background. Yet other studiesfindthat endogamous,patrilinealfamilysystems hinderintermarriage(Lucassen&Laarman,2009). Anthro-pological research on the Netherlands suggests that traditional patterns are imported from the cultures of originsandmaintainedinthe receivingsociety(Van der Hoek&Kret,1992).Thesestudiesfindthatregardingthe partnerchoice, to whicha large importanceis attached amongunmarriedyoungpersonsandtheirfamilies,strong socialcontrol, especiallyon girls, is exerted.Lodewijcks etal.(1997citedinLievens,1998)findatrendtowardsan increasing degree of liberty and participation in this decision,whichismorepronouncedintheMoroccangroup andyet that theparents largelypreserve aconsiderable degreeofinfluence.Weexpectthusthattheopinionofthe parentsimpactsonthepartnerchoice:theprobabilityof beingin aninterethnic partnershipdiminishes/increases withtheparents’rejection/approval(H3).

3_{Yet,the competing economic competitionhypothesis claims that}

individualsprefertomarrysomeoneofhighereconomicstatus(Kalmijn, 1994,assortativematingbyculturalandeconomicoccupationalstatus). Kalmijn(1994)forinstancefoundthatculturalassortativematingismore importantthanassortativematingbyoccupationalstatus.

4_{Infact,theauthorsshowinasecondstepthatthisisanindirecteffect}

Due to intergenerational bonds and transmission of confessions,resultsonthefamilybackgroundmustalsobe seen in light of religious orientation. Results on the US suggest that while differences in religion diminish the propensity to intermarry, intermarriages have become morecommoninliberalreligiousgroups(Sherkat,2004). Endogamy rates were found to be particularly high in Hindu and Muslim communities (Lucassen & Laarman, 2009;compare alsovan Tubergen & Maas, 2007 on the Netherlands). However, we believe that it is not being ascribedtoareligiouscategorybuttowhatextentreligion matters in one’s personal life that affects the partner choice.Religionmaybeconsideredasanimportantpartof one’sidentityandthusculturalcontinuity.Thus, intereth-nic partnerships are less likely when partners attach differentlytoreligionandreligiousinvolvement(H4).

We also expect stark differences across the cities in Belgium.Migrationpatternsledtoadifferentprofileofthe first generation groups across these cities. Lesthaeghe (2000) noticesaso-called ‘‘persistentAntwerp-effect’’in hisstudyonmodernityinTurkishandMoroccanwomen: he finds that religious and traditional conformity are stressedsignificantlymoreinAntwerpthanelsewhere.He explains this as a reaction to a highextreme-right vote amongBelgiansin thiscity.Asserting theirreligiousand ethnic authenticity is seen as areaction against Belgian ethnocentrism. Similarly Vandezande et al. (2011) find that higher shares of the second generation Turks and Moroccansreportdiscriminationinthepublicdomainin AntwerpthaninBrussels.This perceivedhostilitymight alsodecreasethe probabilityofinterethnicpartneringin Antwerp(H5).Besides,unionformationamongtheTurkish second generation occurs earlier in Antwerp than in Brussels(Corijn&Lodewijckx,2009).Ifthisalsoaffectsthe likelihoodofhavinganinter-ethnicpartnerhasnotbeen investigatedtodate.

Inlinewithotherstudies(Corijn&Lodewijckx,2009;

Kalmijn & van Tubergen, 2007; Lievens, 1998;

Schoen-maeckersetal.,1999;Sherkat,2004),itisexpectedthatthe propensitytointermarryincreasesnotonlywithagebut also with higher educational attainments (H6). Higher educationcancreateadditionalopportunitiesforminority memberstoestablishcontactswithpeersfromadifferent ethnicbackground.

Much research investigated gender differences in (inter)marriage behaviour (Qian & Lichter, 2001). In general, men are found more likely to be in an inter-ethnicunion(Kalmijn& vanTubergen,2007).Corijnand

Lodewijckx(2009) findthat in particularMoroccanmen

have more often partners with a different ethnic back-ground.Itisreasonedthatwomentakecareofthechildren andthusareexposedtomorepressurebythirdparties.An importantrolemayalsobethefactthatformanyreligions, womenarethetransmitteroffaith.Briefly,weexpectthat men are more likely to have a partner from a different ethnicbackground(H7).

Regarding the wider definition of partnership, it is hypothesised(H8)thatbothmarriedaswellascohabiting couplesareveryhomogamousregardingethnicitybutthat mateselectionprocessesdifferforcohabitingandmarried couples:cohabitingcouplestendtobelesshomogamous

thanmarriedcouples(Blackwell&Lichter,2000).Kalmijn

andvanTubergen(2007:378)summarisethatthe‘‘norms

of endogamy will be applied less strictly to cohabiting unions than to marriages.’’ Cohabiting relationships are often seen as looser bonds, in which certain ascribed characteristicsofthe partnerareless important.From a differentperspective,cohabitingrelationshipscanbeseen astrialmarriages.Ifthemoreheterogeneouscouplestend tosplitupmoreoften,marriages as‘‘successful relation-ships’’shoulddiffersystematicallyfromcohabitingunions, soforinstanceBlackwellandLichter(2000).

3. Methodology

3.1. Data

ThepresentanalysisisbasedontheBelgianTIESdata (TheIntegrationoftheEuropeanSecondgeneration),whichis partof thecross-nationalTIESsurveyonthe trajectories and experiences of the second generation in several Europeancities(Swyngedouw,Phalet,Baysu,Vandezande, & Fleischmann, 2008). The advantage of using the TIES dataset is the large variety of variables on the ethnic background. For Belgium, the TIES data was collected interviewingtheTurkishandMoroccansecondgeneration – andaBelgiancontrolgroup–aged18–35inAntwerpand Brussels.Thesecondgenerationinthissurveyisdefinedas local-bornpersonswithatleastoneparentborninTurkey orMorocco;thenativecontrolgroupisdefinedashaving two Belgian born parents. The total sample consists of 1717persons.Asthenativepopulationwassampledinthe same urban districts, the design allows to controls for neighbourhoodandsomebackgroundcharacteristics.The sampleiscorrectedbyaweightbasedoninformationon age,genderandmunicipalityfromthepopulationregister ofthecityofAntwerp(Antwerpsubsample)andthe2001 census (Brussels subsample). Swyngedouw et al. (2008)

concludethat‘‘theTurkishsamplerepresentsthe Belgian-born population of Turkish descent in Antwerp well in termsofitsagebygenderbydistrictdistribution,whereas thenativeBelgianandMoroccansamplesdeviatefromthe populationdistributionsinsomerespects’’(Swyngedouw etal.,2008:23)and‘‘thattheunweightedTIESsamplesdo notrepresenttheagebygenderbymunicipality distribu-tions of the secondgeneration populations in Brussels.’’

(Swyngedouw et al., 2008: 24) The data used here is

weightedandadjustedforthesedeviationsandrepresents thus the population in the two cities. It is, however, importanttonotethatthesurveyisnotrepresentativefor allBelgiannatives,butfortheBelgianpopulationlivingin same neighbourhoods as the Turkish and Moroccan minorities.

3.2. Sample

livingtogetherwithapartner(N=509).Whiletheaimof theconstructionofthefirstsampleistocomparesinglesto non-singles, and Belgians to the Moroccan and Turkish second generation, the objective of the multivariate analysis is to contrast partner choices of the ethnic minorities. In other words, the sample accordingly only containscasesfromthesecondgenerationinBelgium,not from‘‘native’’Belgians.Caseswithpartnersfromanother ethnic background (neither Belgian, nor Moroccan or Turkish)werealsodisregarded.

3.3. Estimationtechnique

Multinomial logistic regressions (polytomous logistic regression) are used when the dependent variable is measuredatthenominallevelandisnotdichotomous,as for instance the present case of different forms of partnerships(seeTable1).

Multinomiallogistic regressionsestimate with which probability an expected event will occur depending on differentimpactingvariables(Backhaus,Erichson,Plinke, & Weiber, 2006:426). In thissense,it isan appropriate methodforpredictingthepartnerchoiceandthe impact-ingfactors.Putdifferently,ouraimistogiveaprofileofthe persons in inter- and intra-ethnic partnerships and investigate how different or similar these‘‘groups’’ are. In technical words, the logistic regression allows to

identifythe weightswithwhichthepredictorsinfluence theprobabilitythatanobservedindividualbelongstoone ofthedistinguishedgroups.Indoingso,oddsarereported. Ratherthandescribing the probabilityof anevent, odds indicate the relation of a probability to the counter probability:eventy=1iscomparedtoeventy=0( Back-hausetal.,2006):

Oddsðy¼1Þ¼ pðy¼1Þ 1 pðy¼1Þ

3.4. Operationalisationandvariables

Theliteraturereviewmadeclearthatitisveryusefulto lookattheindividualprofilesbutalsothecontextasitcan giveaninsightin marriage strategiesand opportunities/ constraints.Theincludedindividual-andcontextual-level variables are defined as follows (for descriptive sample statisticsseeTable1).

EthnicitydistinguishesBelgiannativesfromtheTurkish and Moroccan second generations. Belgian natives are definedaspersonswithbothparentsborninBelgium.The second generation comprises persons with at least one parentofTurkishorMoroccanorigin.

Our paper builds on and extends the insights on intermarriagein twoways.First,wedonotonlylookat

Table1

Descriptivesamplestatisticsofthedependentandindependentvariables.

Mean Standard

deviation

Description

Typeofpartner:

(ref.:‘Native’Belgianpartner) PartnerborninBelgiumandbothparentsborninBelgium

Firstgenerationpartner .692 .462 PartnerborninTurkeyorMoroccorespectively

Secondgenerationpartner .183 .387 PartnerborninBelgiumbutatleastoneparent

borninTurkeyorMorocco

City:Brussels(ref.:Antwerp) .348 .477 Dummy

Ethnicgroup:Moroccan2nd

generation(ref.:Turkish2ndgen.)

.424 .495 Ethnicgroupmembershipdefinedbytheparent’s

countryofbirth

Ageofrespondent 28.8 4.004 Inyears

Gender:Women(ref.:men) .560 .497 Dummy

Cohabitationonly(ref.:married) .076 .263 Partnersarenotmarriedbutlivetogetherinthe

samehousehold

Respondent’seducation:

(ref.:lower) Nodiploma;primaryorlowersecondaryeducation

Medium .552 .498 Highersecondary

Higher .256 .437 Tertiaryandhighereducation

Mother’seducation:

(ref.:lower) Nodiploma;primaryorlowersecondaryeducation

Medium .151 .359 Highersecondary

Higher .020 .359 Tertiaryandhighereducation

Missing .409 .492 Responsemissing

Numberofnativefriends 3.051 1.029 Categorical,rangingfrom1‘‘none’’to5‘‘mostofthem’’

Parentalpressurefordivorce(ref.:no) .120 .324 Dummy

Relativesinthecity(ref.:yes) .182 .386 Dummy

BeingBelgiannotsociallyrelevant(ref.:yes) .314 .465 Dummy

BeingTurkish/Moroccannotsocially

relevant(ref.:yes)

.100 .301 Dummy

BeingMuslimnotsociallyrelevant(ref.:yes) .096 .295 Dummy

Heritageculturemaintainedathome 1.797 1.033 5itemscalefrom‘‘totallyagree’’to‘‘totallydisagree’’

marriages but include alsocohabitation.In otherwords, partnershipismeasuredatthetimeoftheinterviewand includes cohabitating couples, either marriedor unmar-ried.Firstgenerationpartners,secondgenerationpartners, ‘native’ Belgian partners and partners with another background are distinguished via nationality and the country of origin of the parents. The ethnicity of the partneristhusdefinedinthesamewayastherespondent’s ethnicgroup.

The parents involvement in their children’s partner choicearemeasuredwithtwodichotomousquestions(no/ yes) – one on parental pressure to marry and one on parentalpressuretoseparatefromthepartner.

Twoindicators characterise thesocialsurrounding of theperson.Thefirstoneidentifiesiffamilyispresentinthe cityofresidence.Thesecondonereferstoone’snetworkin the past (in order to avoid endogeneity issues) and indicateshowmany friendsareofBelgianoriginranging fromnonevia‘oneorafew’,‘quiteafew’,‘alargenumber ofthem’to‘mostofthem’.

Thesubjectivesocialrelevanceof(1)beingBelgian,(2) beingTurkish/Moroccanand(3)beingMuslimismeasured dichotomously (yes/no). The question if the cultural heritage is maintained at home was measured on a 5-itemscalefrom‘totallyagree’to‘totallydisagree’andhad thus a neutral mid-point (‘neither agree nor disagree’). Notethat,astheseconceptsweremeasuredatthetimeof being in a specific type of partnership, causality cannot clearly be established between these independent vari-ablesandthedependentvariable.

Educationismeasuredasthehighestdiplomaattained or, for the respondents still in school, the current educational level. TheseISCED levelsareregroupedinto three categories: (1) no diploma, primary or lower secondary,(2)highersecondaryand(3)tertiaryorhigher education.

4. Inter-andintra-ethnicpartnershipsinBelgium Thissectionpresents,first,selecteddescriptiveresults and,second,amultivariateanalysispredictingthetypeof partnership for the Belgian second generation. Before describingthedifferentformsofpartnerships,thefirstpart willgiveaprofileofthepersonswhoactuallyareinunion– comparedtothosewhoarenot.Thefocusofthisfirstpart isthereforetocomparetheTurkishandMoroccansecond generation in Belgium to ‘‘native’’ Belgians – how do persons in union differ from single persons? Are there differencesacrossthe ethnicgroups?The secondpartof thedescriptivesectionwillfocusonthedifferenttypesof partnerships.Theaimisheretoputorderintheconceptual ‘messiness’ ofinterethnic partnerships by distinguishing relations to a first-generation migrant from second-generation and ‘‘native Belgian’’ partners. Then, the partnerchoiceismodelled.Inthispart,onlytheMoroccan andTurkishsecondgenerationsareincluded.

4.1. Describingthedifferenttypesofpartnerships

Lookingatthefullsample,thereisasignificantlyhigher share of persons living without partner among Belgian

natives than among the Moroccan second generation, whoseshareisagainsignificantlygreaterthanthoseofthe Turkish second generation (compare Table 2, p(Chi2_{)=.000). In this sense, the Moroccan second}

generation is more similarto Belgians than the Turkish secondgeneration.Theshareofsuch‘‘singles’’ismoreover higheramongmenthanamongwomen.Theseresultshave tobeseeninthelightofagedifferencesinformingthefirst union.Theaveragepersonis22yearswhenformingthe first union (cohabitation incl. marriage). Compared to Belgian natives, the second generations of Turks and Moroccansentersuchstablepartnershipsatasignificantly youngerage(astheyhavealowershareofpersonsliving withoutpartnerinthisagegroup).Nativesareonaverage significantly but only slightly older (24 years) when moving together for the first time than Moroccans (23 years),whoareagainsignificantlyolderthantheTurkish secondgeneration(21years).

Overall,womenenterthefirstcohabitingpartnershipat asomewhat youngeragethanmen(22 comparedto23 years,p<.05).Moroccan andBelgian maleshave avery similarbehaviourintermsofage,theyareonaverage24 when entering the first cohabitation. Moroccan and Turkish females cohabitate/marry at the youngest age (21years),whileBelgianwomenaresomewhatinbetween andcomeclosetoTurkishmales(22years).

Theempiricalmaterialconfirmsalsothattheinclusion of cohabitations with not legally married partners is indispensible, especially when distinguishing different ethnicgroups.Whileoverallafifthofallcurrentcohabiting unionsinourpopulationarenotmarriages,only6%among the Turkish second generation and 11% among the Moroccansecondgenerationarenotmarriedwhenliving together,comparedto60%amongtheBelgiannatives.The shareof unmarried cohabiting partnerships is moreover less common among women in each of these ethnic groups.WomenofTurkishdescent arein only5% ofthe cases not married (7% for the male Turkish second generation);women of the Moroccansecondgeneration in 8% of the cases compared to 16% among their male counterparts(outputomitted).

Lookingnow attheethnicbackgroundofthepartner,

Table2revealsthatBelgianslivinginthesame neighbour-hoodsasthesecondgenerationshaverarelypartnerswith a different ethnic background. Four out of five Belgian nativesintheseneighbourhoodsarelivingwithaBelgian nativepartner(ifinaunion).Thislowrateofinterethnic

Table2

Typesofpartnerships,byethnicgroup(columnpercentages).

Native Turkish Moroccan N

Livingwithoutpartner 67.3 49.8 59.7 1008

Livingwithpartner,ofwhich:

Belgianpartnera _{81.4 11.1 12.6 211}

Secondgenerationpartnera _{10.8 16.3 19.9 113}

Firstgenerationpartner 7.8 70.7 61.1 364

Partnerotherorigin 0.0 2.0 6.4 20

Total(livingwithpartner) 100.0 100.0 100.0 1717

Source:TIES(Belgium),weighteddata.

partnerships among the native population is consistent withotherstudies.Theshareofinter-ethnicpartnerships to Belgian natives is approximately equally high in this age-group among the Turkish and Moroccans second generation(11and12%).Moreimportantly,71and61%of thepartnershipsofTurkishandMoroccanrespectivelyare partnerships with someone from the first generation (Table2,p(Chi2_{)=.000).Approximatelytwothirdsofthe}

partnersborninBelgiumareinfactpartnersofthesecond generationfromthesameethnicbackground.Only1outof 8partnershipsofthesecondgenerationinBelgiumistoa ‘‘native’’Belgian.Itcanthusbeconcludedthatamajorityof therelationshiptoBelgiannationalsareinfactrelationsto second generation Turkish or Moroccans. These results confirmtheexpectationsstatedinH1thatestimationsof intermarriage/partnerships based on relations to first generationimmigrantsthereforeseriouslyunderestimate theextentofintra-ethnicpartnerships.Thisshowsthatitis important tomakeaconceptualdistinctionbetweenthe differenttypesofpartnerships.

The gendered distributions (Table 3, p(Chi2_)=.000)

revealthatthemaledescendentsofTurkishandMoroccan immigants is about twice as likely as their female counterparts to be in a union with a Belgian native. Interestingly, the share of interethnic partnerships is equallyhighwhenthepartner’soriginiselsewherethan Belgiumor thecountryof origin.Womenofthe Turkish andMoroccansecondgenerationaresomewhatlesslikely to have a partner from the second generation but have more frequently a migrant partner from the country of origin(firstgeneration)thanmen.

Lievens (1997) explains that assimilation theory

assumes that marrying a first generation partner is an elementoftraditionalism.Inhisresearch,thisassumption isrefuted.HisresultsshowthatitistheTurkishwomen with the most ‘modern’ characteristics (in terms of educationandageatmarriage)thatpreferacross-border marriage with a co-ethnic. Lievens (1997: 5) cites the qualitativeresearchofEsveldtetal.(1995),whofoundthat women often prefer a cross-border marriage with co-ethnics because local-born Turks and Moroccans are considered as ‘‘too traditional, low educated and unem-ployed’’.Callaerts(1997)alsonotesinherqualitativestudy with Turkish Belgiansthat ifa Belgian bornTurkish girl

askshermalepartnertoBelgiumtocomeandlivewithher, thisactuallymeansthatthevirilocaltraditions(inwhich thebridemovesinwithherfamilyinlaw)arebrokenand evenreversed. Lievens’ evidence suggests that for Mor-occans this choice is ratherinspired by socio-economic thanby traditionalist motives. Therefore, weshould not automatically consider the choice for a partner coming fromthecountryoforiginasatraditionalpattern,thatwill diminish as the integration and modernisation proceed (Callaerts,1997;Lievens,1997).

Also,theargumentthatpartnershipstoanimmigrant takeplaceataveryyoungagecannotbeconfirmedassuch. Union formation to the local secondgeneration startat aroundthesameage(22years).Theageatenteringaunion withaBelgiannativeiswith23yearssignificantlyhigher. Theagedifferenceswithregardtothetypeofpartnership are,however,relativelysmall.Agepatternsdonotseemto correspond to what is generally seen as ‘‘modern’’ or ‘‘traditionalist’’marriagebehaviour.

Theeffectsofeducationarealsoinlinewithresultsof previousstudies.Thehighereducated,themoreprobable tolivewithoutpartner.Thetimingofthefirstunionoccurs laterwithincreasingeducation,alsowhenlookingatthe descendantsof immigrants only. Similarly, partners are less often married among the highest educated in this population(68%comparedto81%/86%amongthelowest/ medium educated). The question is then if the ethnic differencesinpartnershippatternsareduetoeducational differencesacrossthegroups.Aseducationislikelytohave agenderedeffect,Table4issplitintotwoparts–theleft onedisplayingtheresultsforthemalesecondgeneration (p(Chi2_{)=.014)andtherightonefortheirfemale}

counter-parts(p(Chi2_{)=.002).Noticeableisthatamongthehigher}

educatedmenandwomen,theshareofsinglesishigher. Moreover, the share of the second generation with a Belgian native partner is higher among the higher educated. In addition, among them the chance to live with someone from a different origin is elevated. This confirmsH6predictingthatthehighereducatedaremore likelytobeinaninterethnicunion.Thechancestolivewith apartnercoming from thecountryof originare highest amongthelowesteducated.

H3 states that theparents’ opinion hasa significant impact onthe partnerchoice. Looking nowonly at the

Table3

Typesofpartnerships,byethnicgroup(columnpercentages).

Men Women Belgian native Turkish 2ndgen. Morocc. 2ndgen. Belgian native Turkish 2ndgen. Morocc. 2ndgen.

Livingwithoutpartner 72.0 51.3 65.9 62.3 48.3 53.2

Livingwithpartner,ofwhich:

Belgianpartnera _{89.1 14.8 18.3 75.3 7.6 8.3}

Secondgen.partnera _{8.5 19.2 22.0 12.6 13.5 18.3}

Firstgenerationpartner 2.4 64.0 53.2 12.1 77.0 67.1

Partnerotherorigin 0.0 2.0 6.6 0.0 2.0 6.3

Total(livingwithpartner) 100.0 100.0 100.0 100.0 100.0 100.0

Source:TIES(Belgium),weighteddata.

sample of persons in cohabiting unions, the results suggest (Table 5) that a big majority of parents in all groupsdonotexertpressure;differencesbetweengroups are present, but they are relatively small. Marrying (much more than objecting their children’s partner choice) in general seems to be a majorconcern of the parents of allethnic groups in that age group, butnot particularly for the second generation. In general, parental pressuretoseparate is lowerformarriedthan for cohabiting couples (14% compared to 10%; output omitted).Interestingly,inparticulartheMoroccanmale secondgeneration,whoismorefrequentlylivingwithout partner, is more often exposed to parental pressure to marry and divorce than their female counterparts. Regarding theTurkish secondgeneration, parentsseem toobjectmoreoftentothepartnerchoiceofwomenthan to those of men.Putting thisinto perspective,also the shareofnative’sparentsthatpresstheirsonstomarry/ leavetheirpartnerisveryhigh.WhileMoroccansmales donotdiffermuchinthisrespectfromthenativegroup, Turkishmen andwomendo.

Withregardtothesecondgenerationparents’opinion regardingthe originofthe partner(Table6),thereis no evidencethatmarriagestofirst-generationimmigrantsare moreoftenenforcedbytheparents.Infact,thepressurefor marrying a first generation immigrant from the same ethnic background is even lower than for marrying a secondgenerationpartner.

Astrikingresultismoreoverthesubjectivelyreported, relativelydepreciativemindsetofthesecondgeneration’s parents towards inter-ethnic partnerships to Belgians. Parental pressure to discourage marriages to Belgian natives is about three times higher towards Belgian partners than towards partners with the same ethnic background.Ontheotherhand,pressuretoencourageto marriagethenativespartnerisalsohighestcomparedto othertypesofpartnerships. MarryingaBelgianseemsto invokea strong reactionby the parents– thisseems to pointintothedirectionthat theparentswouldliketheir childrentofindastablepartnership.Nativepartnersmight in this regard be scrutinised carefully by the second generation’sparents.

Table6

Pressureexertedbyparents,bytypeofpartnership(columnpercentages).

Belgianpartner Secondgenerationpartner Firstgenerationpartner Otherpartner

Pressuretoseparate No 68.5 89.9 90.4 87.8

Yes 31.5 10.1 9.6 12.2

Pressuretomarry No 78.3 80.6 84.2 69.5

Yes 21.7 19.4 15.8 30.5

Source:TIES(Belgium),weighteddata,TurkishandMoroccansecondgenerationonly.

Table5

Pressureexertedbyparents,byethnicgroupandgender(columnpercentages).

Men Women

Belgiannative Turkish2ndgen. Morocc.2ndgen. Belgiannative Turkish2ndgen. Morocc.2ndgen.

Pressuretoseparate No 92.5 92.3 85.8 94.6 84.1 88.6

Yes 7.5 7.8 14.3 5.4 15.9 11.4

Pressuretomarry No 86.9 85.9 73.3 90.5 85.1 81.9

Yes 13.1 14.2 26.7 9.5 15.0 18.1

Source:TIES(Belgium),weighteddata.

Table4

Typesofpartnershipofthedescendantsofimmigrants,byeducationandgender(columnpercentages).

Men Women Nodipl./prim./low secondary Higher secondary Tertiary/ higher Nodipl./prim./low secondary Higher secondary Tertiary/ higher

Livingwithoutpartner 45.6 60.7 61.5 38.8 45.1 62.0

Livingwithpartnerofwhich:

PartnerBelgiannative 11.0 13.8 25.5 5.8 6.2 13.5

Partner2.gen. 14.5 22.8 19.4 6.9 16.5 18.1

Partner1.gen. 72.8 59.2 49.6 87.3 72.2 64.9

Partnerbornelsewhere 1.7 4.2 5.5 0.0 5.1 3.6

Total(livingwithpartner) 100.0 100.0 100.0 100.0 100.0 100.0

Table7

Multinomiallogisticregressionpredictingpartnershipoutcomes(oddsratios).

Model1 Model2 Model3 Model4 Model5

Outcome:Belgiannativepartner

Cityofresidence:Brussels(ref:Antwerp) 2.541** _2.147* _2.236* _2.162* _1.958

(.857) (.751) (.829) (.838) (.806)

Ethnicgroup:Moroccan(ref:Turkish) .925 .792 .709 .863 .858

(.322) (.288) (.274) (.336) (.353)

Ageinyears 1.018 1.021 1.014 .974 .942

(.044) (.045) (.048) (.048) (.049)

Female(ref:male) .448* _.412* _.351** _.316** _.280**

(.158) (.150) (.136) (.129) (.122)

Cohabitationonly(ref.:married) 37.809*** _36.286*** _29.992*** _26.002*** _29.179***

(18.758) (18.337) (15.880) (13.682) (16.202)

Education(ref:low)

Mediumeduc. 1.611 1.519 1.674 2.197 2.238

(.801) (.756) (.877) (1.227) (1.313)

Highereduc. 4.742** _4.588** _5.019** _5.518** _4.448*

(2.488) (2.404) (2.766) (3.315) (2.825)

Educationofmother(ref.:primary)

Mothersecondaryeducation 2.446 2.470

(1.233) (1.261)

Mothertertiaryeducation 9.611** _3.550

(8.391) (3.527)

Mother’seducationmissing 1.709 1.583

(.676) (.653)

Numberofnativefriends 1.275 1.231 1.117

(.219) (.224) (.218)

Parents:pressuretoseparate(ref:no) 3.520** _4.121** _5.014***

(1.588) (1.888) (2.428)

Relativesinthecity(ref.:yes) 1.126

(.546)

BeingBelgiansociallynotrelevant(ref.:yes) 1.000 1.131

(.440) (.528)

BeingTurkish/Moroccansociallynotrelevant(ref.:yes) 2.580 1.907

(1.353) (1.087)

BeingMuslimsociallynotrelevant(ref.:yes) 4.336** _4.329**

(2.173) (2.347)

Heritageculturemaintainedathome

(fromtotallyagreetototallydisagree)

1.462*

(.257)

Outcome:Secondgenerationpartner

Cityofresidence:Brussels(ref:Antwerp) 2.264** _2.078** _2.252** _2.267** _2.355**

(.577) (.549) (.616) (.602) (.638)

Ethnicgroup:Moroccan(ref:Turkish) 1.552 1.615 1.648 1.524 1.592

(.402) (.430) (.445) (.400) (.424)

Ageinyears .894*** _.904** _.903** _.889*** _.877***

(.029) (.030) (.031) (.030) (.030)

Female(ref:male) .440** _.447** _.425** _.450** _.465**

(.117) (.122) (.119) (.124) (.130)

Cohabitationonly(ref.:married) 1.331 1.179 1.092 .940 .887

(.957) (.862) (.801) (.676) (.641)

Education(ref:low)

Mediumeducation 2.415* _2.689* _2.738* _2.412* _2.399*

(.924) (1.055) (1.084) (.932) (.935)

Highereducation 2.445* _2.542* _2.670* _{2.336 2.190}

(1.053) (1.117) (1.182) (1.020) (.960)

Educationofmother(ref.:primary)

Mothersecondaryeducation 2.630** _2.486**

(.904) (.872)

Mothertertiaryeducation .000 .000

(.002) (.001)

Mother’seducationmissing .872 .853

(.256) (.252)

Numberofnativefriends 1.106 1.115 1.048

(.140) (.140) (.136)

Parents:pressuretoseparate(ref:no) 1.300 1.144 1.189

4.2. Explanatoryanalysis

Table 7 shows the final model of the multinomial

logisticregressionsestimatedbasedonthesampleofthe descendentsof migrantswho dolivewiththeirpartner. Comparedtopartnershipswithafirstgenerationpartner, the determinants differwhen lookingat partnerships to nativeBelgiansandtoasecondgenerationmember.

One of the most important results in Table 7 is that Moroccans do not significantly differ from Turkish with regard to their marriage behaviour (p>.05). Also, the interactionbetween sexand ethnicgroupwasnotfound significant. Regarding geographic disparities, it is most likelyinAntwerptohaveafirstgenerationimmigrantas partner(oddsratio>1).Itismorethantwiceaslikelyto chooseforasecondgenerationornativeBelgianpartnerin Brussels. This confirms H5 that interethnic partnering is moreprobableinAntwerp.Oneexplanationmightindeed be the ‘‘ethnicconflict’’ argumentraised earlierand that Brussels’ population is by and large francophone, which could be a factor facilitating inter-ethnic partnerships (Lievens,2000).Itcouldalsobethatthecontacthypothesis ofBlauetal.(1982)andotherauthorsholdstrue:theethnic compositioninBrusselsismuchmoreheterogeneousthan inAntwerp,whichcouldcreatemoreopportunitiestomeet inter-ethnicpartnersandlimitthechancestomeetsomeone fromthesameethnicbackground.

The hypothesis that with increasing age it becomes morelikelytolivewithaBelgianpartner(H2)cannotbe confirmed. The reason for this discrepancy with the pertinentliteraturemightbethattheagegroupisalready limited but also that the data includes the second generation only.Theassimilationeffectthat comeswith ageaffectingintermarriagebehaviourmightholdtruefor firstgenerationmigrantsbutnotforthesecondgeneration, who is born in Belgium.On the otherhand,it becomes morelikelywithagetolivewithafirstgenerationpartner overasecondgenerationpartner.

ThedataseemstoconfirmthepredictionthatBelgian partnersaremorecommonincohabitationsratherthanin

marriages.Inotherwords,non-marriedcouplesseemtobe less homogamous (H8). (Note, however, that this vast effectisduetothesmallnumberofcohabitingcouplesin thereferenceoutcome–duetounreliabilityinscarcecells, theseoddsmustbeinterpretedwithcaution.)

Furthermore,theresultssuggestthatwomenaretwice aslikelyasmentochoosefirst-generationpartnersover native Belgians and second generation partners. Other empiricalstudies onBelgiumconclude that women and men have different motivations for marrying a first generation partner. For the Turkish minority, ‘‘marriage of a man to an import partner is motivated out of traditionalism,whereaswomendosotoaccomplishtheir ownagenda.[...]Forwomen,thehighestchanceofbeing marriedtoanimportpartnerisfoundamongthosewith the most modern characteristics, for men among those with the least modern characteristics.’’ (Lievens, 1997: 18)5 Yet, in our study the interactions of gender with education and ethnicity were not found significant.6 As women aremore educated thismight offset the gender effect.Thefindingthatwomenarelesslikelythanmento haveaBelgiannativepartnerconfirmsH7butraisesthe questionif,asCorijnandLodewijckx(2009)argue,thisis indeedduetoincreasedparentalinterference.

Many studies suggest that the context matters for partnerchoices. The social network (also referred to as third parties, see e.g. Kalmijn & van Tubergen, 2007) impactsonthematchingprocessoftwopersonsthrough creatingopportunitiesormediationoffamilyandfriends. Yet, the models including variables on the parental positiontowardstheunion showthat pressuretomarry didnotinfluencepartnerchoice(outputomitted).Onthe otherhand,rejectionofthepartnerbytheparentshasan

Table7(Continued)

Model1 Model2 Model3 Model4 Model5

Relativesinthecity(ref.:yes) 2.147*

(.684)

BeingBelgiansociallynotrelevant(ref.:yes) .669 .715

(.200) (.219)

BeingTurkish/Moroccansociallynotrelevant(ref.:yes) 1.054 .836

(.517) (.424)

BeingMuslimsociallynotrelevant(ref.:yes) 1.584 1.484

(.791) (.760)

Heritageculturemaintainedathome(fromtotallyagreetototallydisagree) 1.376*

(.178) N 500 500 494 494 487 Loglikelihood 342.51 331.60 319.52 318.63 304.21 Chi2 _{138.16 159.97 162.35 164.13 180.80} Degreesoffreedom 14 20 26 24 27 PseudoR2 _{.17 .19 .20 .20 .23}

Source:TIES(Belgium),TurkishandMoroccansecondgenerationwithapartneronly.

Notes:Referencecategory:firstgenerationpartner.

* _p<0.05. ** _p<0.01. *** _p<0.001.

5_{Theindicatorofattitudestowardsgenderequalitydidnotturnoutas}

arelevantfactordeterminingthepartnerchoice.

6_{Sinceseparatemodelsformaleandfemalesubsampleswerevery}

effectontheoddsoflivingwithaBelgianratherthanafirst generationpartner(Models3–5).ConfirmingH3, partner-shipswithBelgiannativesgohandinhandwithpressure toseparatebytheparents.

The odds of living with a second generation partner comparedtolivingwithafirstgenerationpartneraremore thantwo timeshigherformediumand highereducated than for lower educated (Models 1–3). This multiplier effect is only significant for tertiary education when comparing first generation versus Belgian partners. The effect of tertiary education is on the other hand more pronounced (odds ratio>4). Briefly, H6 that education increases the odds of marrying/cohabitating a native Belgianholdstrue.

Humancapitaloftheparentsseemstopartlyimpacton the partner choice. Although the father’s educational backgrounddoesnotmakeadifference,tertiaryeducation of the mother increases as expected the probability of having a native partner (Model 2). Yet, this effect disappearswhencontrollingforparents’attitudetowards the partner (Model 3, parental pressure to separate). Secondary education compared to lower educational attainments of the mother promotes having a second generationratherthanafirstgenerationpartner(Models2 and3).

Surprisingly, also having native friends does not increasetheoddsofmarryingaBelgiannativeratherthan afirstgenerationpartner(Models3–5).Thecompositionof thesocialnetwork,however,mattersinthesensethatthe presenceofrelatives,whichcouldhaveasimilarinfluence onmatingofthesecondgenerationastheparents,hasa positiveeffectonfindingasecondgenerationpartner.Yet, controllingfortheeffectofthesocialsurrounding(friends, familyand relatives)in Model3,women donot become morelikelytolivewithaBelgiannative.Parentalinference andothersocialcapitalcharacteristicsdoapparentlynot explainthegendergapregardingpartnerchoices.

While this analysis gives some evidence on the importance of the parents in the mating process, the religiousinvolvementoftheparentsisnotrelevant(output omitted).Yet,inlinewithH4,the importanceofbeinga Muslimforthesecondgenerationrespondenthim/herself has a strong negative effect on having a Belgian native partnervis-a-vishavingafirstgenerationpartner(Models 4and5).Ontheotherhand,otherethnicboundaries(i.e. moreprecisely, beingBelgian orTurkish/Moroccanis an importantsocialcategory)donothaveanyimpactonthe partner choice. Similar findings were reported for the Netherlands by van Tubergen and Maas (2007). Briefly, religion is a relevant predictor but not if the person is raisedreligiousbutifs/hethinksitisanimportantsocial category.Ontheotherhand,whentheculturalheritageis maintained at home, the probability to have a Belgian native–butalsoasecondgeneration–partnerdecreases.

5. Discussion

The analyses presented here have thrown additional lightonthepartnershippatternsofthesecondgeneration in Belgiumby including cohabitationand distinguishing three different types of partnerships. A first conclusion

fromthe empiricalanalysis is thatthe inclusion of non-marriedbutcohabitingunionsisindispensiblefor obtain-ing an accurate and complete picture of partnership patterns. Family studies should thus broaden the scope beyond marriages, especially when looking at ethnic differences. The descents of Turkish and Moroccan immigrantsin Belgium, the second generation,differ in their marital and cohabiting behaviour crucially from thoseofBelgiannatives.

In linewith assimilationist ideas,it isoftenassumed that successive immigrant generations assimilate and show higher rates of ethnic intermarriage than their parentsandgrandparents(e.g.Alba&Nee,2003;Qian& Lichter, 2001). Language, culture and opportunity struc-turesareheldresponsible.Previousresearchon intermar-riage in Belgium suggests that the second generation women and men have lower chances to have a first generation partner than the first and 1.5 generation (Lievens, 1997). Ourstudy contributesto this literature byraisingthequestionhowtoconceptualiseinter-ethnic unions. We have showed that most of the relations to Belgians are in fact unions with partners of the second generationfromthesameethnicbackgroundandconclude that estimations of inter-ethnic partnerships based on relationstofirstgeneration immigrantsseriously under-estimatetheextentofintra-ethnicpartnerships.Although it was not possible to perform an inter-generational comparisonhere,thehighshareofintra-ethniccohabiting unions points towardsthe conclusions ofother authors, namelythattheassimilationthesis(understanding assim-ilation roughly as decreasing disparities or increasing similaritiesbetweenlifechancesandlifestyles between ‘immigrantorigin’groupsandarelevant‘non-immigrant origin’referencegroup)doesnot necessarilyhold inthe Belgian context (Callaerts, 1997; Lievens, 1996, 1997; Reniers&Lievens,1997;Reniers,1998;seealsoMilewski& Hamel,2010).

Regardingthedeterminantsofthepartnerchoice,our findings,whicharebasedondata from theBelgian TIES project,aretoalargeextentconsistentwithotherstudies on inter-ethnic partnerships. This study confirmed the positive effect of higher education and the mother’s education as well as regional and gendered differences ofearlierstudiesonintermarriage.Resultsthatcouldnot beconfirmedrelatetoageandgenderedeffectsofethnicity and education. Also language skills were not found relevant. Reasons for these discrepancies could be the homogeneity of the sample, which comprises only the secondgenerationborninBelgium.

ratheranindicatorofthequality(conflictlevel/instability) oftherelationshipthantheparents’preferencestowards theirchildren’spartner.

In addition to previous research, we were able to incorporate attitudinal determinants of partner choice, whichallowsustochallengeandrefinepreviousfindings. Inlinewithotherstudies(vanTubergen&Maas,2007),not religionbutitssubjectiveimportanceasasocialcategory decreasedtheoddsofaninterethnic-partnership.Yet,this did not hold true for other ethnic boundaries (Belgium versus country of origin). Finally, if the maintenance of culturalheritageathomewasattachedmuchimportance, itbecamelesslikelytochooseaBelgiannativeovera first-generationpartner.This,togetherwiththefindingofthe importanceofreligion,posesquestionsaboutthe interde-pendence and reciprocity of ethnic identity and inter-ethnicpartnerships,whichshouldbeaddressedbyfuture research.However,limitedbydatarestrictions,therather crude operationalisation of the theoretical concepts ‘‘parentalinvolvement’’,‘‘religiosity’’and ‘‘cultural main-tenance at home’’ only represents a first step towards answeringsuchquestionsbutneedstoberefinedinfuture studies.

References

Alba,R.,&Nee,V.(2003).RemakingtheAmericanMainstream.Cambridge: CambridgeUniversityPress.

Backhaus,K., Erichson,B., Plinke,W., & Weiber,R.(2006). Multivariate Analysemethoden.Berlin:Springer.

Barth,F.(1969).Ethnicgroupsand boundaries:Thesocialorganization of culturaldifference.Boston:Little,Brown.

Blackwell,D.L.,&Lichter,D.T.(2000).Mateselectionamongmarriedand cohabitingcouples.JournalofFamilyIssues,21,275–302.

Blau,P.M.,Blum,T.C.,&Schwartz,J.E.(1982).Heterogeneityand intermar-riage.AmericanSociologicalReview,47(1),45–62.

Callaerts, T. (1997). ‘Import’ huwelijken bij JongeTurkse Migranten in VlaanderenenBrussel:eenKwalitatieveBenadering.VrijeUniversiteit Brussel,WorkingPapersEtnischeMinderhedeninBelgie(1997-2).

Corijn,M.(2010).Hetprofielvandeniet-gehuwdsamenwonendeninhet

VlaamsGewest.SVR-Webartikel(2010/18).

Corijn,M.,&Klijzing,E.(Eds.).(2001).TransitionstoadulthoodinEurope.

London:KluwerAcademicPublishers.

Corijn,M.,&Lodewijckx,E.(2009).DestartvandegezinsvormingbijdeTurkse enMarokkaansetweedegeneratieinhetVlaamseGewest.SVRRapport.S. v.d.V.Regering.Brussel:VlaamseOverheid.

Esveldt,I.,Kulu-Glasgow,I.,Schoorl,J.J.,&vanSolinge,H.(1995). Migra-tiemotieven,MigratienetwerkenenPartnerkeuzevanTurkenenMarkkanen inNederland.DenHaag:NIDI.

Gordon,M.M.(1964).AssimilationinAmericanlife:Theroleofrace,religion, andnationalorigins.NewYork:OxfordUniversityPress.

Huschek,D.,Liefbroer,A.C.,&deValk,H.A.G.(2010).Timingoffirstunion amongsecond-generationTurksinEurope:Theroleofparents,peers andinstitutionalcontext.DemographicResearch,22(16),473–504.

Kalmijn,M.(1991).StatushomogamyintheUnitedStates.TheAmerican JournalofSociology,97(2),496–523.

Kalmijn,M.(1994).Assortativematingbyculturalandeconomic occupa-tionalstatus.TheAmericanJournalofSociology,100(2),422–452. Kalmijn,M.(1998).Intermarriageandhomogamy:Causes,patterns,trends.

AnnualReviewofSociology,24,395–421.

Kalmijn, M., & van Tubergen, F. (2007). Ethnic intermarriage in the Netherlands:Confirmationsandrefutationsofacceptedinsights. Euro-peanJournalofPopulation,22,371–397.

Lesthaeghe,R.(2000).TransnationalIslamiccommunitiesinamultilingual secular society.InR.Lesthaeghe(Ed.),Communitiesand generations. TurkishandMoroccanpopulationsinBelgium.Brussels:VUBPress.

Lievens, J. (1996). Krachtlijnen van gezinsvormende migratie. Working

Papers‘‘EtnischeMinderhedeninBelgie’’(1996-3).

Lievens,J.(1997).TheThirdWaveofImmigrationfromTurkeyandMorocco: Determinants and Characteristics. Universiteit Gent, Interuniversity papersindemography(IPD)(1997-2).

Lievens,J.(1998).Interethnicmarriage:Bringinginthecontextthrough multilevelmodelling.EuropeanJournalofPopulation,14,117–155. Lievens,J.(2000).ThethirdwaveofimmigrationfromTurkeyandMorocco.

InR.Lesthaeghe(Ed.),Communitiesandgenerations.Turkishand Mor-occanpopulationsinBelgium.Brussels:VUBPress.

Lodewijckx,E. (2010).Gezinsvormingbij detweedegeneratieTurken en Marokkanen.Eenverschillendestartalnaargelangzehuwenmeteen huwelijksmigrantofmetiemandvandetweedegeneratie?SVR-Webartikel (2010/22).

Lucassen,L.,& Laarman, C.(2009). Immigration,intermarriageand the changingfaceofEuropeinthepostwarperiod.Historyofthefamily, 14,52–68.

Milewski,N.,&Hamel,C.(2010).Unionformationandpartnerchoiceina transnationalcontext:ThecaseofdescendantsofTurkishimmigrantsin France.InternationalMigrationReview,44(3),615–658.

Qian,Z.,&Lichter,D.T.(2001).Measuringmaritalassimilation:Intermarriage amongnativesandimmigrants.SocialScienceResearch,30,289–312. Reniers,G.(1998).Post-migrationsurvivaloftraditionalmarriagepatterns:

ConsanguineousmarriageamongTurkishandMoroccanimmigrantsin

Belgium.Universiteit Gent,Interuniversitypapersindemography(IPD) (1998-1).

Reniers,G.,&Lievens,J.(1997).Stereotypeninperspectief.Deevolutievan enkeleaspectenvanhethuwelijkbijdeTurkseenMarokkaanse mind-erheden in Belgie. Working Papers ‘‘Etnische Minderheden in Belgie’’ (UniversiteitGent,VakgroepBevolkingswetenschappen)(1996-3). Schoenmaeckers,R.C.,Lodewijckx,E.,&Gadeyne,S.(1999).Marriagesand

fertilityamongTurkishandMoroccanWomeninBelgium:Resultsfrom censusdata.InternationalMigrationReview,33(4),901–928.

Sherkat,D.E.(2004).ReligiousintermarriageintheUnitedStates:Trends, patterns,andpredictors.SocialScienceResearch,33,606–625. Song,M.(2009).Isintermarriageagoodindicatorofintegration?Journalof

EthnicandMigrationStudies,35(2),331–348.

Swyngedouw,M.,Phalet,K.,Baysu,G.,Vandezande,V.,&Fleischmann,F. (2008).TechnicalReport.TIES2007–2008Belgium—ExtendedTrajectories andExperiencesofTurkish,MoroccanandNativeBelgiansinAntwerpand Brussels.Leuven:CenterforSociological(CeSo),UniversityofLeuvenand CenterforSocialandCulturalPsychology(CSCP).

vanTubergen,F.,&Maas,I.(2007).Ethnicintermarriageamongimmigrants intheNetherlands:Ananalysisofpopulationdata.SocialScience Re-search,36,1065–1086.

Vandezande,V.,Phalet,K.,&Swyngedouw,M.(2011).Dofeelingsof dis-criminationexplaintheriotsinBrussels?AcomparisonofMoroccanand TurkishgroupsinBrusselsandAntwerp.BrusselsStudies47.