Id

Jena Neb Vri

devLifeconperAlsconsetpreTimdetencfamof

immwocouLie

Advances in Life Course Research 18 (2013) 257–269

A R

Artic

Rece

Rece

Acce

Keyw

Fam

Idea

Imm

Seco

Fam

Euro

*

(Hel

104

http

eal ages for family formation among immigrants in Europe

nifer A. Holland a,*, Helga A.G. de Valk a,b

therlands Interdisciplinary Demographic Institute, Lange Houtstraat 19, 2511 CV The Hague, The Netherlands

je Universiteit Brussel, Interface Demography, Department of Sociology, Pleinlaan 5, 1050 Brussels, Belgium

The timing and sequencing of events is central to theelopment and trajectory of life courses (Billari, 2005). course researchers have stressed the importance oftext and time in shaping both the experience andception of events in the life course (Elder Jr., 1985).o in studies of immigrants, emphasis is placed on thetextual (socialization) effects of country of origin and

tlement when it comes to life course transitions andferences (De Valk, Wingens, Windzio, & Aybek, 2011).ing preferences for family life transitions are likely

ermined by individual characteristics and experi-es, but also influenced by dominant, macro-levelily formation systems. However, comparative studiesthe timing of family life events among those of

igrant origin are still largely lacking and existingrk on timing preferences mainly focuses on onentry or one migrant origin group (e.g. De Valk &

fbroer, 2007; Giuliano, 2007). With information on

diverse migrant populations across a range of countriesof settlement, we can shed light on the influences offamily formation systems in both origin and destinationthat may shape timing preferences for family lifetransitions. In this paper we fill this gap in the literatureand explore family formation ideals regarding the timingof marriage and childbearing among immigrants ofdiverse origins across Europe.

Forming a partnership, getting married, and bearingand raising children are significant transitions for individ-uals. Ideals regarding the timing of these events may beindicative of the meaning attached to these transitions andhow they should best fit into the life course. Moreover, inthe case of migrants, ideals may be a particularly usefulmeasure of attitudes toward the family life course since thetiming of actual family behaviors is often distorted by theact of migration (e.g. Andersson, 2004; Milewski, 2007;Toulemon, 2004). The perceived ideal timing of familyevents is governed by attitudes and values, and istransmitted at multiple levels, through family andcommunity socialization, and institutions. As such, idealsmay give us insight into identity, individuals’ understand-ing of their position in the social world, and processes of

T I C L E I N F O

le history:

ived 14 December 2012

ived in revised form 9 August 2013

pted 9 August 2013

ords:

ily formation

ls

igrants

nd generation

ily change

pe

A B S T R A C T

This paper investigates ideal ages for marriage and parenthood among immigrants from

over 160 countries origins living in 25 European countries. Ideals regarding the timing of

family formation are indicative of how individuals perceive the family life course and

provide insight into family-life aspirations and the meaning attached to these transitions.

Using data from the European Social Survey (Round 3, 2006; N = 6330) and a cross-

classified multilevel modeling approach, we investigate associations between the

influences of the dominant family formation timing patterns in countries of origin and

settlement, individual-level characteristics, and ideal ages. We make innovative use of a

standard demographic measure, the singulate mean age of marriage, to measure family

formation patterns. Results suggest that residential context influences are associated with

the timing ideals of all migrants, but origin influences seem to be associated with the ideals

of only the most recent migrants.

� 2013 Elsevier Ltd. All rights reserved.

Corresponding author. Tel.: +31 703 565 214; fax: +31 703 647 187.

E-mail addresses: holland@nidi.nl (J.A. Holland), valk@nidi.nl

ga A.G. de Valk).

Contents lists available at ScienceDirect

Advances in Life Course Research

jo u rn al h om epag e: ww w.els evier .c o m/lo cat e/a lcr

0-2608/$ – see front matter � 2013 Elsevier Ltd. All rights reserved.

://dx.doi.org/10.1016/j.alcr.2013.08.002

J.A. Holland, H.A.G. de Valk / Advances in Life Course Research 18 (2013) 257–269258

family change after migration among immigrant-back-ground (first- and second-generation) populations.

Using data from the third round of the European SocialSurvey, covering 25 European countries, we explore factorsshaping ideal ages for two key family life transitions,marriage and parenthood, among immigrants and theirdescendants. We analyze variation in ideal ages byindividual attributes and develop a new proxy measureto capture the dominant family formation timing patternsin countries of origin and settlement. With this measure,we are able to assess the extent to which macro-levelfamily formation regimes in countries of origin andsettlement influence individual timing preferences, andthe relative importance of these influences. Finally, weconsider how the influences of macro-level family forma-tion patterns are mediated by immigrant’s duration ofresidence in their country of settlement and migrantgeneration. Results provide new insights into the influenceof contexts on perceptions of family life, social distance,and family change among immigrant populations inEuropean societies.

1. Theory and hypotheses

First- and second-generation immigrants constitute alarge and growing share of European populations (Castles &Miller, 2003; Eurostat, 2011). Questions of social distanceand change between ethnic groups are at the core ofacademic and public discourses. It has been argued thatwhile immigrants may integrate in the public domain (e.g.labor market and education), changes in the private domain,in particular regarding the family life course, are slower tooccur. Thus, it may be particularly important to disentanglepersistent cultural influences on social distance in theprivate domain (Gordon, 1964; Lesthaeghe, 2002a). Re-search into family change among migrants has largelyfocused on actual behavior and, in particular, on intermar-riage and fertility (some exceptions are: De Valk & Liefbroer,2007; Huschek, de Valk, & Liefbroer, 2011; Sassler & Qian,2003). Although intermarriage may be the strongestindicator of the social distance between groups and thedegree of incorporation in a new country of residence (see,for instance: Bean & Stevens, 2003; Kalmijn, 1998; Pagnini &Morgan, 1990), there is a well-documented tendencytoward homogamy across a variety of characteristics (race,ethnicity, education, as well as nativity). If intermarriage israre, it may not be an ideal measure of social distance(Sassler & Qian, 2003). Studies of fertility behavior havedemonstrated that the act of migration is often closelybound up with family formation processes (Milewski, 2007;Stephen & Bean, 1992), and so it too may be a flawedmeasure of family change among first generation immi-grants (Andersson, 2004; Sobotka, 2008; Toulemon, 2004).

Where private domain behaviors may be more resistantto influence and slower to occur (as with intermarriage) orbe sensitive to the act of migration (as with fertility),family formation ideals may be an alternative indicator ofsocial distance between groups in society. Ideals are‘‘representation[s] of the attributes that someone (yourselfor another) would like you, ideally, to possess (i.e., a

for you)’’ (Higgins, 1987, pp. 320–321). Ideals contribute toan individual’s sense of self and identity, but also to one’sunderstanding of the position of the self in the social world(Higgins, 1987). Along with individual behaviors, values,and norms, ideals regarding the timing of family-formationevents may be indicative of the meaning attached to thesetransitions and how they should best fit into the life course.In addition to reflecting individual identity, ideals can belinked to behavior; for instance, according to the Theory ofPlanned Behavior, individual behaviors are the outcome ofideals and attitudes, subjective norms and perceivedbehavioral control (Ajzen, 1991; Lesthaeghe, 2002b). Assuch, the ideal timing of family life events may give a moreaccurate understanding of immigrants’ aspirations orwishes for their own family lives or for the family livesof other significant persons in their families, socialnetworks, or communities.

Ideals may be shaped by community- or country-levelinfluences, through family and community socialization,institutions, and social norms. Immigrants are unique frommajority (non-immigrant-background) populations in thatthey occupy a sociocultural middle ground between theircountries of origin and residence, with family-life idealspotentially shaped by influences on both sides (De Valk &Liefbroer, 2007; De Valk & Milewski, 2011; Foner, 1997;Glick, 2010; Nauck, 2001). The distinction between theinfluences of countries of origin and residence is often madewhen theorizing about the position of immigrants in theirnew home country and has been previously used to studystructural integration (Van Tubergen, 2005, 2010). Culturalnorms, practices, and behaviors associated with the domi-nant family life patterns in countries of origin may betransmitted and maintained by family and friends from thesame origin, in both origin and settlement countries. Familyformation patterns in the country of origin can serve as amacro-level determinant, capturing the cultural meaning offamily formation. We hypothesize the following:

Hypothesis 1. Family formation patterns in countries oforigin of immigrants will be associated with ideal ages forfamily formation among immigrants from that origin.

Family life ideals may just as well be shaped by thedominant patterns and practices in the country ofsettlement. The past three decades have been a timeof rapid change in European family systems, with respectto social norms, individual preferences, and behavior.Along with delayed home leaving, the emergence of non-marital cohabitation, increases in union dissolution andfalling fertility rates, first marriage and first births occurat older ages (Aassve, Arpino, & Billari, 2013; Billari &Liefbroer, 2010; Sobotka & Toulemon, 2008). Althoughthe experience of these changes has been nearlyuniversal in Europe, the magnitude of change is variedacross countries (Reher, 1998; Sobotka & Toulemon,2008). These diverse family formation patterns withinEuropean country contexts likely shape the ideals heldby all residents, irrespective of background. Therefore,we can expect the following:‘

Hypothesis 2. Family formation patterns in countries of

residence will be associated with ideal ages for family representation of someone’s hopes, aspirations, or wishes

formcou

to bcourelasett& NdurdomsamgenresMoclas196domsettthequeto

witof

mesamof fthesec

Hypof rmig

Hypof

mig

addbactionparspehougenDuvideprehantheSetyoucohtunweinflackvaltranenrwe

J.A. Holland, H.A.G. de Valk / Advances in Life Course Research 18 (2013) 257–269 259

ation among immigrant populations residing in thatntry.

Although immigrants’ timing preferences are expectede shaped by influences of both the country of origin andntry of destination, the relative importance may beted to immigrants’ duration of residence in thelement country, and to immigrant generation (Albaee, 1997, 2003; Lieberson & Waters, 1988). A longeration of residence implies longer exposure to theinant timing patterns in the country of residence. Thee holds for the children of immigrants (the seconderation) who are socialized in their countries of

idence across their entire life courses (Kasinitz,llenkopf, Waters, & Holdaway, 2008). According tosical theories on immigrant assimilation (Gordon,4), this would imply a growing influence of theinant family formation patterns in the country of

lement with longer durations of residence and among children of immigrants. It remains, however, an openstion whether this reflects acculturation or adaptation

the institutional, economic, policy, or social situationhin a country of residence, or if it is a result of a meldingthe diverse cultural and family orientations of allmbers of a society (Alba, 2005; Rumbaut, 1999). At thee time the influence of dominant patterns of the timing

amily formation in countries of origin would decline inir influence over duration of residence and among theond generation. We hypothesize the following:

othesis 3a. Family formation patterns in the countryesidence will be more important for the ideal ages ofrants with longer residence and the second generation.

othesis 3b. Family formation patterns in the countryorigin will be less important for the ideal ages ofrants with longer residence and the second generation.

Ideal ages for family formation are likely influenced byitional individual characteristics, apart from immigrantkground. Men and women may have different percep-s of the ideal timing of family formation events,

ticularly since women tend to make larger relationship-cific investments, taking on the larger share ofsework and childrearing responsibilities, even in moreder-equal societies (Baxter, 1997; Sundstrom &ander, 2002). Age may also influence perceptions of

al ages for family formation. On the one hand, timingferences may change as people grow older. On the otherd, age may capture cohort effects, reflecting changes in

standard timing of family (Furstenberg Jr., 2010;tersten Jr. & Ray, 2010): Older cohorts may tend to viewnger family formation as ideal, as compared to youngerorts that prefer to postpone family formation. Unfor-ately, due to the limitations of cross-sectional data (as

use in this study), we cannot distinguish between theuences of age and cohort (Testa, 2006). While wenowledge that these two interpretations may both beid, because we account for several important life-course

sitions strongly associated with age (educationalolment, marital status, and having ever had a child),

will discuss age as a cohort effect.

Family life ideals are also likely influenced by socio-economic status. More economically advantaged andhighly educated individuals face higher opportunity costsassociated with family building (Becker, 1991). As suchthey may prefer to postpone marriage and childbearing tolater ages, after completing education and establishingthemselves in the labor market. Having experienced thelife course transitions under investigation (i.e. marryingand/or becoming a parent) may influence perceptions ofthe ideal timing for these events. To the extent thatindividuals may act according to the standard that theyperceive as ideal or if individuals adjust their ideal toconform to their own lived experiences, we expect thatindividuals having already experienced marital andparental transitions should also prefer younger ages, onaverage. Finally, we account for religiosity, which has beenshown to influence both family demographic behavior andthe relative timing of family life events (Surkyn &Lesthaeghe, 2004).

2. Method

2.1. Sample

We used data from the European Social Survey (Round3, 2006), a cross-sectional survey of attitudes, beliefs andbehavior patterns (European Social Survey, 2006, 2011;Jowell, 2007). Round 3 covered 25 European countries:Austria, Belgium, Bulgaria, Cyprus, Denmark, Estonia,Finland, France, Germany, Hungary, Ireland, Latvia, theNetherlands, Norway, Poland, Portugal, Romania, theRussian Federation, Slovakia, Slovenia, Spain, Sweden,Switzerland, Ukraine, and the United Kingdom. The ESS isrepresentative of populations aged 15 or older, living inprivate households, and residing in the country for at leastone year. In addition to the main survey, Round 3 includedquestions about ideal ages for a range of family life-coursebehaviors (‘‘The Timing of Life’’ module).

In total there were 47,099 respondents, with approxi-mately 1000 to 3000 respondents per country. Responserates ranged from 46.0% in France to 73.2% in Slovakia,with an average of 63.5%. In our analysis we include onlythose respondents with an immigrant background: thosewho migrated (first generation) and the children ofimmigrants (second generation) (n = 7571). Further, weexcluded respondents with missing information aboutimmigrant status (n = 847; 1.8%), information on timesince migration for first generation immigrants (n = 47;0.1%), the indicator for whether ideal age questions wereworded with reference to men or women (split ballotassignment, discussed below) (n = 2; <0.1%), or informa-tion on other covariates (n = 345; 0.4%), for a pre-analysissample of 6330.

2.2. Dependent variables

We focused our analysis on respondents who provideda numeric response to the ideal age questions (marriage:N = 5250; parenthood: N = 5507). The ESS Timing of Lifemodule had a split ballot design, whereby respondentsreceived all questions worded in reference to either

J.A. Holland, H.A.G. de Valk / Advances in Life Course Research 18 (2013) 257–269260

women or men. For instance, one half of the respondentsreceived a question pertaining to the timing of marriage forwomen: ‘‘In your opinion, what is the ideal age for a girl orwoman to get married and live with her husband?’’ Theother half was asked the question pertaining to men. Whileaverage ages of marriage and parenthood differed by splitballot assignment (ideal ages for men were about 2 yearsolder than for women, on average), the distribution of ageswere quite similar (Table 1). We conducted exploratoryanalysis to determine if patterns of association betweencovariates and outcome variables were different by splitballot assignment, but the sign and magnitude of theassociations were nearly identical. Consequently thosewho were assigned the male and female versions of thequestionnaire were analyzed together. We accounted forwhether questions referenced women or men with adummy variable. To limit the influence of extremely lowand high ages, we bottom and top coded the dependentvariables at the 1st (18 years old) and 99th (35 years old)percentiles.

Although not specifically prompted, respondentscould also reply that there was ‘‘no ideal age’’ or thatthe behavior was ‘‘never’’ acceptable (for details onproportion of non-numeric responses see Table 3). Inorder to account for potential selection bias due to theexclusion of individuals offering a non-numeric re-sponse, we followed the approach of Rijken and Billari(2012), who investigated preferences for individualautonomy and the ideal age to become a parent usingthe same data (Round 3 of the European Social Survey).In line with their method, we predicted the probabilityof offering a non-numeric response using multilevellogistic regression, including an instrumental variablecapturing the proportion of all ESS respondents offeringa non-numeric response to ideal age for marriage andparenthood questions (respectively) per interviewer(range: 0 to 1; (Rijken & Billari, 2012, pp. 7–8), andcovariates capturing demographic, socioeconomic, lifecourse and religiosity characteristics (as detailed below).We predicted non-numeric response to questions aboutideal age for marriage and parenthood separately(results not shown, but available upon request), andincluded the predicted values as coefficients in thesubsequent models of ideal age for marriage andparenthood, respectively.

2.3. Independent variables: Individual level

In addition to covariates accounting for split ballotassignment and sample selection, we accounted for a hostof individual-level characteristics likely to influence idealages for marriage and parenthood. We allowed forvariation in ideal ages by second-generation-immigrantstatus (reference) and, for first generation immigrants, byduration of residence, measured in the survey withcategories corresponding to: less than 5 years, 6 to 10years, 11 to 20 years or more than 20 years.

We accounted for demographic characteristics of therespondents, controlling for the gender of the respondentand, to capture cohort changes in timing ideals, therespondent’s age at interview with a linear and a squaredterm. We accounted for respondent’s socioeconomic statusand background. Highest level of education completed bythe respondent was standardized using the InternationalStandard Classification for Education (ISCED) and specifiedcategorically: less than secondary (reference), lowersecondary, upper secondary or some post-secondary,and tertiary education. We included a dummy variableindicating whether the respondent’s mother or fathercompleted tertiary education. Finally, we accounted foreconomic activity by including two variables indicatingeducational enrolment and paid work in the seven daysprior to interview. These last two variables were neithermutually exclusive nor collinear.

Because respondent’s stage in the family life coursemay influence or reflect ideal age preferences, wedistinguished respondent’s marital status as never marriedor in a civil partnership (reference), currently married or ina civil partnership, or previously married or in a civilpartnership. This last category included both the widowedand divorced. We included an indicator for whether therespondent ever had children.

Finally, we accounted for religiosity with a compositemeasure reflecting responses to three questions pertainingto frequency of religious practice and self-assessedreligiosity. Frequency of religious service attendance andfrequency of prayer were measured with a seven pointscale, with categories corresponding to: (1) every day, (2)more than once a week, (3) once a week, (4) at least once amonth, (5) only on special holy days, (6) less often, and (7)never. The mean value for attendance was 4.5 (SD = 2.5)

Table 1

Descriptive statistics of ideal ages for family formation events, by split ballot assignment (N = 6330).

Women Men

Meana SDa Rangea % Non-numericb Meana SDa Rangea % Non-numericb

Ideal age to marry and

live with spouse

23.7 3.6 18.0–35.0 17.7 25.8 3.5 18.0–35.0 16.4

Ideal age to become

a parent

24.9 3.4 18.0–35.0 12.4 27.0 3.5 18.0–35.0 13.7

N 3224 3106

a Mean, standard deviation and range reflect only those providing numeric responses to questions on ideal ages.b

The non-numeric category includes reports of ‘‘no ideal age,’’ ‘‘never,’’ and ‘‘refusal’’. Non-numeric categories were not prompted by the interviewer, but

were accepted as valid responses.

andaskordreliasscomquefor

comanareliact

2.4.

of-r

patres(SMspe195ava

Tab

Des

Im

De

So

Lif

Ad

N

a

b

J.A. Holland, H.A.G. de Valk / Advances in Life Course Research 18 (2013) 257–269 261

for prayer was 4.9 (SD = 3.0). Respondents were alsoed to assess how religious they are, with 11-point,ered response scale, ranging from 0 (‘‘not at allgious’’) to 10 (‘‘very religious’’). The average self-essed religiosity score was 5.5 (SD = 1.5). Principle

ponent factor analysis indicated that the threestions loaded onto a single factor, which accounted73.6% of the variation across the three variables. Thebined religiosity factor score was included in the

lyses. Higher scores corresponded to higher levels ofgiosity. Descriptive statistics for all individual char-eristics can be found in Table 2.

Independent variables: Country-of-origin- and country-

esidence -level

To measure the influences of diverse family life courseterns across countries of origin and countries ofidence, we used the singulate mean age of marriageAM). The SMAM captures the average number of years

nt single among those who marry before age 50 (Hajnal,3; Preston, Heuveline, & Guillot, 2001). The measure isilable for women and men in 214 countries across a

range of years in the United Nations World Marriage Data2008 database (United Nations Department of Economicand Social Affairs Population Division, 2009). Because ofthe temporal link between marriage and first birth(Mensch, Singh, & Casterline, 2005), the SMAM is a usefulmacro-level proxy for the country-specific ‘‘standard’’timing of family formation. Global trends in marriage aremirrored by a rising age of first birth (Mensch et al., 2005;Singh, 1998; Sobotka & Toulemon, 2008) and factorsdriving trends in the age of first marriage and first birth arelargely the same (Lesthaeghe, 2010; Mensch et al., 2005).

Capturing the dominant family formation patterns inthe 25 countries of residence was straightforward: Weused the SMAM measure from the year closest or prior tothe ESS survey year (2006). To account for the dominantfamily formation patterns in countries of origin, we usedinformation about respondent’s country of birth (firstgeneration). In the case of the second generation, followingthe convention of statistical bureaus across Europe, weused the country of birth of the mother and, if the motherwas born in the country of residence, the country of birth ofthe father. Immigrants and their descendants originated ina diverse range of countries: Approximately 70% fromcountries in continental Europe, 7% from Anglophonecountries (including the United Kingdom and Ireland, theUnited States, Canada, Australia and New Zealand), 8%from African countries, 12% from Asia, and 3.5% from LatinAmerica and the Caribbean. Unfortunately, we could notidentify the year a first-generation immigrant or theparent(s) of a second generation individual left theircountry of origin. Therefore, we used information onduration of residence, for the first generation, and year ofbirth for the second generation to choose the closestsingle-year value of the SMAM for the country of origin. For12 origin countries (404 individuals) no values of theSMAM were available in the World Marriage database. Forthese cases, we supplemented the database with averageages of marriage from other sources, assigned the SMAMthe nearest geographical and sociocultural neighbor or, inthe case of former countries, we assigned historical valuesof SMAM of the present-day territories.

Both the SMAM of country of residence and country oforigin were centered at their means, which allowed us tointerpret regression effects relative to persons residing ororiginating in countries with average levels of SMAM,respectively. We tested categorical, linear, and quadraticspecifications of the SMAM for both countries of origin andresidence variables (not shown) and linear specificationsbest fit the data. There were notably different distributionsof the SMAM in countries of residence and the morediverse countries of origin (Table 3).

2.5. Procedure

Respondents were nested in 25 countries of residence:36% in Western and Eastern Europe respectively, 12% inNorthern Europe, 9% in Anglophone countries, and 6% inSouthern Europe). At the same time, they were nested inover 160 countries of (parent’s) origin. Consequently, weconducted cross-classified multilevel regression analysis,using the maximum likelihood method to estimate the

le 2

criptive statistics, individual characteristics (N = 6330).

N %

migrant generation and time

since immigration

Second generation 2980 47.1

First generation 3350 52.9

Less than 5 years since migration 441 7.0

6–10 years since migration 357 5.6

11–20 years since migration 647 10.2

More than 20 years since migration 1905 30.1

mographic characteristics

Age Mean SD

44.3 17.6

Respondent female N %

3508 55.4

cioeconomic status

Respondent’s highest education completed

Less than lower secondary 568 9.0

Lower secondary 1150 18.2

Upper secondary 2531 40.0

Tertiary 2081 32.9

Respondent’s mother or father

completed tertiary education

1561 24.7

Enrolled in education 687 10.9

In paid work 3541 55.9

e course characteristics

Marital and civil partnership status

Never 1699 26.8

Currently 3285 51.9

Previously 1346 21.3

Children 4358 68.8

Mean SD

ditional characteristics

Religiosity scale 0.0 1.0

Scale components

Self-assessed religiositya 5.5 1.5

Frequency of religious service attendanceb 4.5 2.5

Frequency of prayerb 4.9 3.0

6330 100.0

Range: 0 (‘‘not at all religious’’)–10 (‘‘very religious’’).

Range: 1 (‘‘every day’’)–7 (‘‘never’’).

J.A. Holland, H.A.G. de Valk / Advances in Life Course Research 18 (2013) 257–269262

variance components (Hox, 2010; Rabe-Hesketh & Skrondal,2012). We assessed how ideals are shaped by influences atboth the individual-, country-of-residence-, and country-of-origin-levels. Our baseline model (Model 1) accounted onlyfor structural characteristics of the data: The variableaccounting for sample selection (predicted probability ofoffering a non-numeric response), an indicator for beingassigned questions pertaining to women (vs. men, i.e. splitballot assignment), as well as country of origin, country ofresidence and residual variance components. In Model 2, weaccounted for immigrant generation and duration ofresidence, individual demographic, socioeconomic, and lifecourse stage characteristics, and religiosity. Model 3 furtherincorporated the influences of family formation regimes incountries of origin and residence (SMAM). Finally, in Model4, we allowed for the influences of the family formationregimes to vary by immigrant generation and duration ofresidence by incorporating interaction terms.

3. Results

3.1. Ideal ages for marriage

Tables 4a and 4b present estimates from cross-classified multilevel regression models of ideal age formarriage. Model 1 was a baseline model, accounting onlyfor structural characteristics of the data: An indicator forbeing assigned questions pertaining to women (vs. men,i.e. split ballot assignment), our variable accounting forsample selection (predicted probability of offering a non-numeric response), as well as country of origin, country ofresidence and residual variance components. The idealage for marriage for women was about two years youngerthan men. In this model without covariates, there wasevidence of selection bias—those with a higher predictlikelihood of offering a non-numeric response to ques-tions about ideal ages of marriage tend to prefer 10 monthyounger ideal age for marriage. Turning to the randomeffects, the estimated residual standard deviation of theintercept (i.e. average ideal age for marriage) acrosscountries of origin was a little over 6 months(ffiffiffiffiffiffiffiffiffiffiffiffiffi

0:306p

¼ 0:55 years), and this variation accounted forabout 2.5% (0:306=ð0:306 þ 0:920 þ 10:852Þ) of the totalvariation in ideal ages for marriage. The estimatedresidual standard deviation of the intercept acrosscountries of residence was 11.5 months(ffiffiffiffiffiffiffiffiffiffiffiffiffi

0:920p

¼ 0:959 years), and it accounted for 7.6% of thetotal variation in ideal ages for marriage. The remaining

of countries of origin and residence, was estimated as 3.3years (

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

10:852p

) (90.0% of the variation in the ideal age formarriage).

In Model 2, individual characteristics were introducedinto the model. The two-year difference in ideal ages formen and women remained, but selection bias due to theexclusion of respondents who offer non-numericresponses was fully accounted for by individual character-istics. There is a positive gradient in marriage ideals andimmigrant generation and time since arrival. Secondgeneration immigrants preferred the oldest ages formarriage but were statistically indistinguishable fromtheir long duration of residence (>20 year), first generationcounterparts. The most recent immigrants (those arrivingin the 5 years previous to the survey) preferred theyoungest ages for marriage, 11 months younger than theirsecond generation counterparts.

Female respondents tended to prefer older ages ofmarriage, net of whether they receive questions wordedabout men or women (split ballot assignment). Age waspositively associated with ideal age preferences and thispositive association increased at older ages, as indicated bythe positive and statistically significant coefficient on thepolynomial term. Respondent’s education, parent’s tertiaryeducation, and employment were also positively associat-ed with ideal ages for marriage. While the coefficient waspositive, being enrolled in education in the 7 days prior tothe survey was not statistically significantly associatedwith marriage timing preferences. Having ever marriedwas negatively associated with ideal ages for marriage:The currently married reported 10 month younger idealages as compared to the never married, while thepreviously married (either divorced or widowed) reported5.5 months younger ideal ages. Ever having children wasnot statistically significantly associated with ideal age formarriage, although the coefficient was negative, asexpected. Finally, religiosity was negatively associatedwith ideal age preferences, although the magnitude of thecoefficient was not large: A one-standard-deviationincrease in a respondent’s religiosity was associated withan approximately 2 month decrease in the ideal age formarriage. Including these individual covariates reducedthe estimate of the residual standard deviation of theintercept across countries of origin, from 6.6 months toabout 5 months (1.7% of the total variation in the ideal agefor marriage), however variation across countries ofresidence actually increased from 11.5 to 13 months(10.2% of the total variation). The remaining residual

Table 3

Descriptive statistics: family formation regimes, countries of origin and residence, by gender (N = 6330).

Women Men

Mean SD Range Mean SD Range

SMAM, countries of origin 23.1 3.0 15.6–33.3 26.6 3.1 21.1–37.5

N, countries of origin 142 146

SMAM, countries of residence 28.3 3.0 23.1–32.2 30.7 2.8 25.9–34.3

N, countries of residence 25 25

N, by split ballot assignment 3224 3106

standard deviation of the average ideal age for marriage,

residual standard deviation, not due to the additive effects

notresthe

corpatestioriginsHowbetandimm2.5witformnot

Tab

Cros

(N =

Fix

In

Ra

Lo

Es

Note

unst

*

**

**

J.A. Holland, H.A.G. de Valk / Advances in Life Course Research 18 (2013) 257–269 263

due to the additive effects of countries of origin andidence, was reduced slightly to about 3.1 years (88.0% of

total variation).In Model 3 we introduced the SMAM measures,responding to the dominant family formation timingterns in countries of origin and of residence. Themate of the association between SMAM in country ofin and ideal age for marriage was small and statistically

ignificant; thus, we fail to confirm our Hypothesis 1.ever, we found evidence of a positive association

ween family formation regime in countries of residence ideal ages: A 1-year increase in the SMAM of anigrant’s country of settlement was associated with a

-month increase in ideal ages for marriage, consistenth Hypothesis 2. By and large, accounting for family

ation regimes in countries of origin and residence did significantly influence the associations between

individual characteristics and ideal ages for marriage, asestimated in Model 2. The residual standard deviationacross countries of residence was reduced from 13 monthsto 11 months and accounted for 7.9% of the total variationin the ideal age for marriage. Given the small andstatistically insignificant estimate of the associationbetween the SMAM in countries of origin and ideal ages,it is unsurprising that the residual standard deviation ofthe average ideal age for marriage attributable to countriesof origin was not greatly reduced when comparing Models2 and 3.

In Model 3, the association between family formationregimes in countries of origin and residence and ideal agesfor marriage was an average effect, across all respondentsregardless of generation and time since immigration. InModel 4, we relaxed that assumption, including interac-tions between immigrant background characteristics and

le 4a

s-classified multilevel models predicting ideal ages for marriage from individual-, country-of-origin-, and country-of-residence-level factors

5250).

Model 1 Model 2 Model 3

b SE b SE b SE

ed effects

Constant 26.58 0.360*** 24.25 0.540*** 24.27 0.528***

Split ballot assignment: Female-worded question �2.06 0.092*** �2.05 0.088*** �2.07 0.088***

Predicted probability of non-response �0.83 0.309** �0.35 0.300 �0.36 0.299

Country-of-origin-level

SMAM, country of origin 0.04 0.024

Country-of-residence-level

SMAM, country of residence 0.21 0.067***

dividual-level

Migrant background

2nd generation 0.00 0.00

1st gen, �5 years residence �0.93 0.198*** �1.04 0.206***

1st gen, 6–10 years residence �0.41 0.207* �0.50 0.212*

1st gen, 11–20 years residence �0.38 0.157* �0.43 0.159**

1st gen, >20 years residence �0.16 0.113 �0.18 0.114

Female 0.75 0.092*** 0.75 0.092***

Age 0.06 0.017*** 0.06 0.017***

Age2 0.00 0.000*** 0.00 0.000***

Respondent’s highest education completed

Less than lower secondary 0.00 0.00

Lower secondary 0.68 0.185*** 0.68 0.185***

Upper secondary 1.04 0.174*** 1.04 0.174***

Tertiary 1.53 0.182*** 1.54 0.182***

Mother or father completed tertiary education 0.46 0.114*** 0.47 0.114***

Enrolled in education 0.32 0.182 0.31 0.182

In paid work 0.42 0.110*** 0.41 0.110***

Marital status

Never married 0.00 0.00

Married �0.83 0.156*** �0.82 0.156***

Previously married �0.46 0.181* �0.46 0.181*

Any children �0.20 0.140 �0.19 0.140

Religiosity �0.14 0.047** �0.14 0.047**

ndom effects

Variance (country of origin) 0.306 0.116 0.198 0.094 0.200 0.095

Variance (country of residence) 0.920 0.305 1.158 0.367 0.848 0.281

Variance (residual) 10.852 0.215 10.004 0.198 9.993 0.198

g likelihood �13773.97 �13556.93 �13550.72

timated parameters 6 23 25

: Analysis samples include only those respondents offering numeric responses to questions about timing preferences for marriage. b-coefficients are

andardized.

p < 0.05.

p < 0.01.

* p < 0.001.

J.A. Holland, H.A.G. de Valk / Advances in Life Course Research 18 (2013) 257–269264

the SMAM of origin and residence, respectively, in ourrandom coefficient model. Allowing for the influence offamily formation patterns of the country of origin to varyby time since immigration and generation revealed a clearnegative gradient of the influence of this macro-level factoron ideal ages for marriage. For those who had recentlyarrived in their country of settlement (in the 5 years priorto survey), a one year older country-of-origin SMAM wasassociated with a 2-month older ideal age preference for

marriage (as compared to those of the second generation).For those who arrived in the previous 6 to 10 years, theassociation between the SMAM and ideal ages was reducedto about 1.7 months. For longer residing first generationimmigrants, the coefficients were reduced to non-signifi-cance, suggesting that family formation patterns incountries of origin were not associated with the idealsof these respondents nor the second generation. This isconsistent with our Hypothesis 3a. In contrast, we did not

Table 4b

Cross-classified multilevel models predicting ideal ages for marriage from individual-, country-of-origin-, and country-of-residence-level factors

(N = 5250).

Model 4

b SE

Fixed effects

Constant 24.30 0.529***

Split ballot assignment: Female-worded questions �2.07 0.088***

Predicted probability of non-response �0.35 0.299

Country-of-origin level

SMAM, country of origin 0.00 0.033

SMAM, country of origin � migrant background

1st gen, �5 years � SMAM, origin 0.17 0.066**

1st gen, 6–10 years � SMAM, origin 0.14 0.064*

1st gen, 11–20 years � SMAM, origin 0.05 0.053

1st gen, >20 years � SMAM, origin �0.01 0.042

Country-of-residence-level

SMAM, country of residence 0.20 0.070***

SMAM, country of residence � Migrant background

1st gen, �5 years � SMAM, residence �0.01 0.083

1st gen, 6–10 years � SMAM, residence 0.08 0.081

1st gen, 11–20 years � SMAM, residence 0.02 0.055

1st gen, >20 years � SMAM, residence 0.04 0.037

Individual-level

Migrant background

2nd generation 0.00

1st gen, �5 years residence �1.23 0.239***

1st gen, 6–10 years residence �0.69 0.231**

1st gen, 11–20 years residence �0.42 0.161**

1st gen, >20 years residence �0.15 0.116

Female 0.75 0.092***

Age 0.05 0.017**

Age2 0.00 0.000***

Respondent’s highest education completed

Less than lower secondary 0.00

Lower secondary 0.67 0.185***

Upper secondary 1.03 0.174***

Tertiary 1.52 0.182***

Mother or father completed tertiary education 0.47 0.114***

Enroled in education 0.33 0.183

In paid work 0.41 0.110***

Marital status

Never married 0.00

Married �0.82 0.157***

Previously married �0.46 0.181*

Any children �0.20 0.140

Religiosity �0.13 0.047**

Random effects

Variance (country of origin) 0.208 0.096

Variance (country of residence) 0.866 0.286

Variance (residual) 9.962 0.197

Log likelihood �13543.65

Estimated parameters 33

Note: Analysis samples include only those respondents offering numeric responses to questions about timing preferences for marriage. b-coefficients are

unstandardized.

* p < 0.05.

** p < 0.01.

*** p < 0.001.

findmopatmacal

termThi

3.2.

preideresmaagewotho

Tab

Cros

(N =

Fix

In

Ra

Lo

Es

Note

unst

*

**

**

J.A. Holland, H.A.G. de Valk / Advances in Life Course Research 18 (2013) 257–269 265

evidence that duration of residence and generationderated the association between family formationterns of the country of residence and ideal ages forrriage: No individual interaction term reached statisti-significance, nor did the inclusion of the interactions improve the fit of the model (Wald test; not shown).

s finding stands in contrast to our Hypothesis 3b.

Ideal ages for parenthood

Findings regarding ideal ages to become a parent,sented in Tables 5a and 5b, are very similar to those foral ages for marriage. Consequently, we discuss only keyults and those results that differ from models ofrriage ideals. Similarly to models of marriage, the ideal

for becoming a parent was two years younger formen than for men. In the case of parenthood, excludingse respondents who offered non-numeric responses did

not seem to be associated with bias in reported ideal ages.The estimated residual standard deviation of the intercept(i.e. average ideal age for parenthood) across countries oforigin was smaller than found for marriage, just under 5months (

ffiffiffiffiffiffiffiffiffiffiffiffiffi

0:159p

¼ 0:40 years) and accounted for 1.4% ofthe total variation in the ideal age for parenthood. Theestimated residual standard deviation across countries ofresidence was about 12 months (

ffiffiffiffiffiffiffiffiffiffiffiffiffi

1:039p

¼ 1:02 years), andaccounted for about 8.9% of the variation in the totalvariation. The remaining residual standard deviation of theintercept, not due to the additive effects of countries oforigin and residence, was estimated to be 3.3 years(ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

10:459p

; 90.0% of the total variation).Model 2 incorporated the individual characteristics of

respondents. Associations between these characteristicsand ideal ages for parenthood were very similar to thosefound for ideal ages for marriage. Again, we found apositive gradient in the ideal ages for parenthood and

le 5a

s-classified multilevel models predicting ideal ages for parenthood from individual-, country-of-origin-, and country-of-residence-level factors

5507).

Model 1 Model 2 Model 3

b SE b SE b SE

ed effects

Constant 27.13 0.389*** 25.09 0.543*** 25.11 0.532***

Split ballot assignment: Female-worded questions �2.04 0.088*** �2.04 0.085*** �2.06 0.085***

Predicted probability of non-response �0.20 0.339 �0.03 0.328 �0.03 0.328

Country-of-origin level

SMAM, country of origin 0.02 0.020

Country-of-residence-level

SMAM, country of residence 0.23 0.067***

dividual-level

Migrant background

2nd generation 0.00 0.00

1st gen, �5 years residence �0.49 0.187** �0.55 0.193**

1st gen, 6–10 years residence �0.33 0.197 �0.39 0.201

1st gen, 11–20 years residence �0.13 0.151 �0.16 0.152

1st gen, >20 years residence �0.02 0.108 �0.03 0.109

Female 0.71 0.088*** 0.71 0.088***

Age 0.04 0.017** 0.04 0.017**

Age2 0.00 0.000*** 0.00 0.000***

Respondent’s highest education completed

Less than lower secondary 0.00 0.00

Lower secondary 0.55 0.178** 0.56 0.178**

Upper secondary 0.94 0.167*** 0.94 0.167***

Tertiary 1.57 0.175*** 1.57 0.174***

Mother or father completed tertiary education 0.33 0.110** 0.33 0.110**

Enroled in education 0.59 0.176*** 0.57 0.176***

In paid work 0.41 0.106*** 0.40 0.106***

Marital status

Never married 0.00 0.00

Married �0.33 0.149* �0.33 0.149*

Previously married �0.12 0.171 �0.12 0.171

Any children �0.38 0.134** �0.37 0.134**

Religiosity �0.18 0.045*** �0.17 0.045***

ndom effects

Variance (country of origin) 0.159 0.080 0.088 0.053 0.084 0.052

Variance (country of residence) 1.039 0.344 1.169 0.371 0.893 0.300

Variance (residual) 10.459 0.202 9.725 0.187 9.716 0.187

g likelihood �14335.35 �14130.27 �14123.97

timated parameters 6 23 25

: Analysis samples include only those respondents offering numeric responses to questions about timing preferences for parenthood. b-coefficients are

andardized.

p < 0.05.

p < 0.01.

* p < 0.001.

J.A. Holland, H.A.G. de Valk / Advances in Life Course Research 18 (2013) 257–269266

immigrant generation and time since arrival, although onlythe coefficient for the most recent arrivals (within the 5years prior to the survey) reached statistical significance(about 6 months younger than second generation respon-dents). The second generation reported the oldest idealages for parenthood, all else equal. Unlike with the idealage for marriage, we found that enrolment status wassignificantly associated with preferring older ages forparenthood. As with marriage, having ever been married

was negatively associated with ideal ages for parenthood,although only the coefficient for currently married reachedstatistical significance. Having children was negatively andstatistically significantly associated with parenthoodtiming preferences. Including individual covariates re-duced the residual standard deviation of the average idealage for parenthood across countries of origin, from 5months to about 3.5 months (0.8% of the total variation inthe ideal age for parenthood), but slightly increased the

Table 5b

Cross-classified multilevel models predicting ideal ages for parenthood from individual-, country-of-origin-, and country-of-residence-level factors

(N = 5507).

Model 4

b SE

Fixed effects

Constant 25.13 0.532***

Split ballot assignment: Female-worded questions �2.05 0.085

Predicted probability of non-response �0.03 0.328

Country-of-origin level

SMAM, country of origin 0.00 0.028

SMAM, country of origin � migrant background

1st gen, �5 years � SMAM, origin 0.14 0.062*

1st gen, 6–10 years � SMAM, origin 0.09 0.061

1st gen, 11–20 years � SMAM, origin 0.05 0.050

1st gen, >20 years � SMAM, origin �0.04 0.039

Country-of-residence-level

SMAM, country of residence 0.23 0.070***

SMAM, country of residence � migrant background

1st gen, �5 years � SMAM, residence �0.04 0.080

1st gen, 6–10 years � SMAM, residence 0.01 0.078

1st gen, 11–20 years � SMAM, residence 0.01 0.053

1st gen, >20 years � SMAM, residence 0.03 0.036

Individual-level

Migrant background

2nd generation 0.00

1st gen, �5 years residence �0.73 0.225***

1st gen, 6–10 years residence �0.50 0.221**

1st gen, 11–20 years residence �0.17 0.154

1st gen, >20 years residence �0.03 0.110

Female 0.71 0.088***

Age 0.04 0.017**

Age2 0.00 0.000***

Respondent’s highest education completed

Less than lower secondary 0.00

Lower secondary 0.54 0.178**

Upper secondary 0.92 0.167***

Tertiary 1.55 0.175***

Mother or father completed tertiary education 0.33 0.110**

Enroled in education 0.59 0.176***

In paid work 0.40 0.106***

Marital status

Never married 0.00

Married �0.32 0.149*

Previously married �0.12 0.171

Any children �0.37 0.134**

Religiosity �0.17 0.045***

Random effects

Variance (country of origin) 0.075 0.049

Variance (country of residence) 0.896 0.299

Variance (residual) 9.699 0.187

Log likelihood �14118.07

Estimated parameters 33

Note: Analysis samples include only those respondents offering numeric responses to questions about timing preferences for parenthood. b-coefficients are

unstandardized.

* p < 0.05.

** p < 0.01.

*** p < 0.001.

stanto 1resaddma(88

of

formfindstroconassof rof awitparstanacrwhresaccage

levresthemaorigin

paryeaThialthcal

ideinflformandfindcon

4. D

cougraIdeinsaboinvualforminvin

gra(SMpatfamandinfl

J.A. Holland, H.A.G. de Valk / Advances in Life Course Research 18 (2013) 257–269 267

dard deviation across countries of residence, from 123 months (10.6% of the total variation). The remaining

idual standard deviation of the intercept, not due to theitive effects of countries of origin and residence, was

rginally reduced, by about 1.4 months to 3.1 years.6% of the variation).In Model 3 we introduced variables capturing countryorigin and country of residence dominant family

ation timing patterns. As with marriage, we did not evidence that the SMAM in country of origin wasngly associated with the ideal age for a first birth, intrast to Hypothesis 1. Again, there was a positiveociation between family formation timing in countriesesidence and ideal ages: A 1-year increase in the SMAMn immigrant’s country of settlement was associated

h an almost 3 month increase in ideal ages forenthood, consistent with Hypothesis 2. The residualdard deviation of the average ideal age for parenthood

oss countries of origin was reduced only marginally,ile the residual standard error across countries ofidence was reduced from 13 months to 11.3 months,ounting for about 8.5% of the total variation in the ideal

for parenthood.In Model 4, we tested whether the influences of macro-el family formation patterns in countries of origin andidence were constant across respondents regardless ofir generation and time since immigration. As withrriage, we found a negative gradient in the influence ofin across duration of residence: For each year increase

the SMAM in the country of origin, ideal ages forenthood among recent migrants (arriving in the fivers prior to the survey) increased by about 1.7 months.s association declined with duration of residence,ough no other interaction coefficient reached statisti-significance. Also in parallel to the results for marriageals, duration of residence and generation did notuence the association between the macro-level family

ation pattern in a respondent’s country of residence individual-level ideal ages for parenthood. Theseings conform to our Hypothesis 3a but we fail tofirm Hypothesis 3b.

iscussion

In this paper we examined ideal ages for two key liferse events, marriage and parenthood, among immi-nt-background individuals in 25 European countries.als regarding the timing of family formation provideight into individual preferences and understandingut family life among migrants. In particular thisestigation highlights the importance of both individ-

characteristics as well as broader, macro-level familyation timing patterns in shaping ideals. We

estigated the influence of family formation contextorigin and destination, utilizing a standard demo-phic measure, the singulate mean age of marriageAM), as a proxy for dominant family formation

terns. By having a complimentary measure of theily formation patterns in both countries of residence

origin, we show the importance of both spheres ofuence on individual ideals for family life, improving

upon studies considering migrants within one country ofresidence or a single migrant group across differentcountries of residence.

We found that the contextual influence of the countryof residence matters for family formation ideals. There wasa positive association between dominant family formationpatterns across countries of residence and ideal ages,consistent with Hypothesis 2. Although we did not fullyaccount for cross-country-of-residence variance in idealsand there was not a one-to-one relationship betweenfamily formation patterns and ideal ages, the SMAM incountries of residence did in part explain differences acrossEuropean countries. Contrary to our Hypothesis 3b,however, we did not find that the influence of familyformation patterns in the country of residence varies bytime since arrival or immigrant generation; rather thisassociation was proportional in its influence on ideal agesfor marriage and parenthood. This stands in contrast withclassical theories of assimilation, which would predict agrowing influence of the dominant family formationpatterns in the country of residence with longer durationsof residence and among the children of immigrants. Asnoted in recent theoretical debates, straight-line theoriesof assimilation neglect the possibility that (particularlywithin diverse societies) immigrants may not adapt tobecome indistinguishable from the majority population.Rather, more complex processes of social change mayoccur, with both immigrant-background and majoritypopulations interacting to produce new, hybrid familysystems (Alba, 2005; Rumbaut, 1999). As such, the SMAMin countries of residence may reflect an averaging of amultitude of diverse family behaviors, and thus beproportionally associated with immigrants’ ideals, regard-less of duration of residence.

At the same time, the fact that we found no differenti-ation between immigrants by durations of residence couldreflect selection processes associated with migration flowsand stages. Network theories of migration suggest thatinterpersonal relationships and shared cultural or idea-tional ties, will preserve and perpetuate migration flows(Massey et al., 1998). Individual migrants may thus selectdestination countries based on values and norms theyshare with (or that are similar to) established immigrantcommunities and/or majority populations within a coun-try of settlement (Van Tubergen, 2010). So too mightselection operate through differential survey non-contact-ability and non-response due to residential mobility, issuesof legal status, a lack of trust, language difficulties, or othersources of social exclusion (Barnes, 2008; Stoop, Billiet, &Koch, 2010). As a consequence, we may overestimate thepositive association between macro-level family formationtiming patterns in the country of residence and individualideal age preferences.

Our findings regarding the association between macro-level family formation patterns in countries of origin andideal ages for marriage and parenthood were moreconsistent with theories predicting a declining influenceof origin over time and generation. There was no evidencethat origin family formation regimes had an averageinfluence on immigrants (in contrast to Hypothesis 1); thisinfluence was only evident in relation to the ideals of the

J.A. Holland, H.A.G. de Valk / Advances in Life Course Research 18 (2013) 257–269268

most recent migrants, consistent with Hypothesis 3a.However, also here it seems that a straight-line assimila-tion view is too simple and it may be only those whorecently arrived for whom country of origin remainscrucial for their ideals. Changes in migrant flows acrosstime likely produce different country-of-origin composi-tions across categories of duration of residence in oursample. This might imply that we capture the influences ofdifferent countries of origin rather than a change acrossduration of residence and generation. This is underscoredby the fact that including the interactions betweenduration of residence and generation and the dominantfamily formation patterns of countries of origin anddestination actually increases the gradient found for themain effect of duration of residence and generation.Unfortunately, we cannot disentangle the differentialinfluences of countries of origin and duration of residenceon family formation ideals without having longitudinaldata for immigrants of diverse origin residing in Europe. Inthe case of country of origin influences, it is likely that ourestimates are conservative, since the processes of selectionin survey response discussed above may mean that thosemost closely holding to the family formation scripts oftheir countries of origin may be under-represented in thesample. However, selection operating through differentialmigration flows (i.e. as predicted by network theory)would not influence the estimates.

Issues of selection notwithstanding, these findings arenotable in that they highlight the unique position ofimmigrants, whose family life ideals are shaped by thedominant family formation patterns of both countries oforigin and residence. We are not able to disentangleprocesses underlying the associations demonstrated here.Indeed, it is likely that institutional characteristics, such asthe labor market, educational structures, housing markets,etc., which have been shown to influence the familybehaviors of non-immigrant-background populations,underlie individual ideal age preferences of immigrants.These factors also play a role in shaping the broader,macro-level patterns of family formation captured in ourSMAM measures. Future research investigating the relativeinfluences of these other society-level influences on thefamily life ideals and behaviors of immigrants, relative totheir non-immigrant-background peers, will help usunderstand family life transitions among diverse popula-tions of Europe.

Finally, we aimed to assess the relative importance ofcontextual and individual characteristics for timingideals. Respondent’s demographic characteristics, socio-economic status, stage in the family life course, andreligiosity all contribute to our understanding of familylife ideals among immigrant groups. We found that themagnitude of these influences on ideals was quite similarto that of our contextual measures of patterns of familyformation, underscoring the importance of giving atten-tion to the influences of both contextual- and individual-level influences in shaping immigrant’s views of the lifecourse.

The strength of the European Social Survey data is thatit provides a broad overview of the great diversity of

European populations. These data are unique, offering usthe possibility to explore family life course ideals acrossEurope, among immigrant subgroups with origins in aglobal range of countries. At the same time, we werelimited in our ability to explore individual country-of-origin and sub-country-of-origin patterns of ideal ages bysmall sample sizes. Moreover, our contextual measure, thesingulate mean age of marriage (SMAM), is available onlyfor countries, not for sub-regions or sub-populations,limiting the precision with which we can measure thesecontextual influences. As immigrants constitute an everlarger share of Western populations, future research intoimmigrant family life and family change necessitatesbetter attention to maximize response-rates and over-sampling immigrant sub-populations. Nevertheless, theresults presented here offer a valuable starting point forcross-national investigations of immigrant populationideals, attitudinal and value orientations toward the lifecourse, and family change.

Acknowledgments

This work is supported by the European ResearchCouncil Starting Grant project ‘‘Families of migrant origin:a life course perspective’’ (project no. 263829). The authorsgratefully acknowledge helpful suggestions from ElizabethThomson, Dana Garbarski, Kimberly Turner, and twoanonymous reviewers. A previous version of this paperwas presented at the EUROCORES (ESF) Cross-National andMulti-Level Analysis of Human Values, Institutions andBehaviour Programme (HumVIB) Final Conference (BerlinSeptember 2011), the Dutch Demography Day (UtrechtNovember 2011), and the 4th Dutch ESS Workshop (theHague September 2012).

References

Aassve, A., Arpino, B., & Billari, F. C. (2013). Age norms on leaving home:Multilevel evidence from the European Social Survey. Environment andPlanning A, 45(2), 383–401 http://dx.doi.org/10.1068/a4563

Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior andHuman Decision Processes, 50(2), 179–211 http://dx.doi.org/10.1016/0749-5978(91)90020-T

Alba, R. (2005). Bright vs. blurred boundaries: Second-generation assimila-tion and exclusion in France, Germany, and the United States. Ethnicand Racial Studies, 28(1), 20–49 http://dx.doi.org/10.1080/0141987042000280003

Alba, R., & Nee, V. (1997). Rethinking assimilation for a new era of immigra-tion. International Migration Review, 31, 826–874 http://dx.doi.org/10.2307/2547416

Alba, R., & Nee, V. (2003). Remaking the American mainstream: assimilationand contemporary immigration. Cambridge: Harvard University Press.

Andersson, G. (2004). Childbearing after migration: Fertility patterns offoreign-born women in Sweden. International Migration Review, 38(2),747–774 http://dx.doi.org/10.1111/j.1747-7379.2004.tb00216.x

Barnes, W. (2008). Improving the participation of migrants in the labourforce survey: A review of existing practices in EU member states. SurveyMethodology Bulletin63.

Baxter, J. (1997). Gender equality and participation in housework: A cross-national perspective. Journal of Comparative Family Studies, 28(3), 220.

Bean, F. D., & Stevens, G. (2003). America’s newcomers and the dynamics ofdiversity. New York: Russell Sage Foundation.

Becker, G. S. (1991). A treatise on the family (Enl. ed.). Cambridge, MA: HarvardUniversity Press.

Billari, F. C. (2005). Life course analysis: Two (complementary) cultures?Some reflections with examples from the analysis of the transition to

adulthood. Advances in Life Course Research, 10, 261–281 http://dx.doi.org/10.1016/S1040-2608(05)10010-0 backgrounds, attitudes, ideals, and opinions among

Billa

Cast

De V

De V

De V

Elde

Euro

Euro

Euro

Fone

Furs

Giul

Glic

Gord

Hajn

Higg

Hox

Hus

Jowe

Kalm

Kasi

Lest

Lest

Lest

Lieb

Mas

Men

J.A. Holland, H.A.G. de Valk / Advances in Life Course Research 18 (2013) 257–269 269

ri, F. C., & Liefbroer, A. C. (2010). Towards a new pattern of transition toadulthood? Advances in Life Course Research, 15(2–3), 59–75 http://dx.doi.org/10.1016/j.alcr.2010.10.003les, S., & Miller, M. J. (2003). The age of migration. New York: GuilfordPress.alk, H. A. G., & Liefbroer, A. C. (2007). Timing preferences for women’s

family-life transitions: Intergenerational transmission among migrantsand Dutch. Journal of Marriage and Family, 69(1), 190–206 http://dx.doi.org/10.1111/j. 1741-3737.2006.00353.x

alk, H. A. G., & Milewski, N. (2011). Family life transitions amongchildren of immigrants: An introduction. Advances in Life Course Re-search, 16(4), 145–151 http://dx.doi.org/10.1016/j.alcr.2011.10.001alk, H. A. G., Wingens, M., Windzio, M., & Aybek, C. (2011). Immigrant

settlement and the life course: an exchange of research perspectives andoutlook for the future. In M. Wingens, M. Windzio, H. De Valk, & C. Aybek(Eds.), A life-course perspective on migration and integration. Dordrecht,NL: Springer Publishers.r, G. H., Jr. (1985). Perspectives on the life course. In G. H. Elder, Jr., (Ed.).Life course dynamics (pp. 23–49). Ithaca, NY: Cornell University Press.pean Social Survey. (2006). Ess Round 3: European Social Survey Round 3

data. Data file edition 3.3. Retrieved from: http://www.europeansocial-survey.org/.pean Social Survey. (2011). Ess Round 3: European Social Survey (2011):

Ess-3 2006 documentation report Edition 3.3.stat. (2011). Migrants in Europe—a statistical portrait of the first and secondgeneration Luxembourg. Publications Office of the European Union.r, N. (1997). The immigrant family: Cultural legacies and cultural

changes. International Migration Review, 31(4), 961–974.tenberg, F. F., Jr. (2010). On a new schedule: Transitions to adulthoodand family change. Transition to Adulthood20(1).iano, P. (2007). Living arrangements in Western Europe: Does culturalorigin matter? Journal of the European Economic Association, 5(5), 927–952 http://dx.doi.org/10.1162/JEEA.2007.5.5.927k, J. E. (2010). Connecting complex processes: A decade of research onimmigrant families. Journal of Marriage and Family, 72(3), 498–515http://dx.doi.org/10.1111/j.1741-3737.2010.00715.xon, M. M. (1964). Assimilation in American life. New York: Oxford

University Press.al, J. (1953). Age at marriage and proportions marrying. Population Studies,

7(2), 111–136 http://dx.doi.org/10.1080/00324728.1953.10415299ins, E. T. (1987). Self-discrepancy: A theory relating self and affect.

Psychological Review, 94(3), 319., J. (2010). Multilevel analysis: techniques and applications. New York:Routledge.chek, D., de Valk, H. A. G., & Liefbroer, A. C. (2011). Gender-role behaviorof second-generation turks: The role of partner choice, gender ideologyand societal context. Advances in Life Course Research, 16(4), 164–177http://dx.doi.org/10.1016/j.alcr.2011.09.005ll, R. (2007). European Social Survey 2006/2007: Technical report. London:

Centre for Comparative Social Surveys, City University.ijn, M. (1998). Intermarriage and homogamy: Causes, patterns, trends.

Annual Review of Sociology, 24, 395–421 http://dx.doi.org/10.1146/annurev.soc.24.1.395nitz, P., Mollenkopf, J., Waters, J., & Holdaway, J. (2008). Inheriting thecity: the children of immigrants come of age. Cambridge: Harvard Univer-sity Press.haeghe, R. (Ed.). (2002a). Meaning and choice: value orientations and lifecourse decisions. The Hague, The Netherlands: NIDI/CBGS.haeghe, R. J. (2010). The unfolding story of the second demographictransition. Population and Development Review, 36(2), 211–251 http://dx.doi.org/10.1111/j.1728-4457.2010.00328.xhaeghe, R. J. (Ed.). (2002b). Meaning and choice: value orientations and lifecourse decisions. The Hague, The Netherlands: NIDI.erson, S., & Waters, M. C. (1988). From many strands: ethnic and racialgroups in contemporary America. New York: Russell Sage Foundation.sey, D. S., Arango, J., Hugo, G., Kouaouchi, A., Pellegrino, A., & Taylor, J. E.(1998). Worlds in motion: understanding international migration at the endof the millennium. Oxford: Clarendon Press.sch, B., Singh, S., & Casterline, J. (2005). Trends in the timing of firstmarriage among men and women in the developing world. In C. B. Lloyd,

J. R. Behrman, N. P. Stromquist, & B. Cohen (Eds.), The changing transitionsto adulthood in developing countries: selected studies (pp. 118–171).Washington, DC: The National Academies Press.

Milewski, N. (2007). First child of immigrant workers and their descendantsin West Germany: Interrelation of events, disruption, or adaptation?Demographic Research, 17(29), 859–896 http://dx.doi.org/10.4054/DemRes.2007.17.29

Nauck, B. (2001). Intercultural contact and intergenerational transmission inimmigrant families. Journal of Cross-Cultural Psychology, 32(2), 159–173http://dx.doi.org/10.1177/0022022101032002004

Pagnini, D. L., & Morgan, S. P. (1990). Intermarriage and social distanceamong U.S. immigrants at the turn of the century. The American Journal ofSociology, 96(2), 405–432.

Preston, S. H., Heuveline, P., & Guillot, M. (2001). Demography: measuring andmodeling population processes. Malden, MA, USA: Blackwell Publishing.

Rabe-Hesketh, S., & Skrondal, A. (2012). (3rd ed.). Multilevel and longitudinalmodeling using stata (3). College Station, TX: Stata Press.

Reher, D. S. (1998). Family ties in Western Europe: Persistent contrasts.Population and Development Review, 24(2), 203–234 http://dx.doi.org/10.1007/1-4020-2837-7_3

Rijken, A., & Billari, F. (2012). Individual autonomy and timing of childbearing:A cross-European comparison. Stockholm, Sweden: Paper presented atthe European Population conference.http://epc2012.princeton.edu/abstracts/120223.

Rumbaut, R. (1999). Assimilation and its discontents: Ironies and paradoxes.The handbook of international migration (pp. 172–195).

Sassler, S., & Qian, Z. (2003). Marital timing and marital assimilation:Variation and change among European Americans between 1910 and1980. Historical Methods, 36(3), 131 http://dx.doi.org/10.1080/01615440309601606

Settersten, R. A., Jr., & Ray, B. (2010). What’s going on with young peopletoday? The long and twisting path to adulthood. Transition to Adult-hood20(1).

Singh, S. (1998). Adolescent childbearing in developing countries: A globalreview. Studies in Family Planning, 29(2), 117–136.

Sobotka, T. (2008). Overview chapter 7: The rising importance of migrantsfor childbearing in Europe. Demographic Research, 19(9), 225–248 http://dx.doi.org/10.4054/DemRes.2008.19.9

Sobotka, T., & Toulemon, L. (2008). Changing family and partnership behav-iour: Common trends and persistent diversity across Europe. Demo-graphic Research, 19(6), 85–138 http://dx.doi.org/10.4054/DemRes.2008.19.6

Stephen, E. H., & Bean, F. D. (1992). Assimilation, disruption and the fertilityof Mexican-origin women in the United States. International MigrationReview, 26(1), 67–88.

Stoop, I. A. L., Billiet, J., & Koch, A. (2010). Improving survey response: lessonslearned from the European Social Survey. Chichester, UK: John Wiley &Sons Ltd.

Sundstrom, M., & Duvander, A. Z. E. (2002). Gender division of childcare andthe sharing of parental leave among new parents in Sweden. EuropeanSociological Review, 18(4), 433–447 http://dx.doi.org/10.1093/esr/18.4.433

Surkyn, J., & Lesthaeghe, R. J. (2004). Value orientations and the seconddemographic transition (sdt) in northern, western and southern Europe:An update. Demographic Research, S3(3), 45–86 http://dx.doi.org/10.4054/DemRes.2004.S3.3

Testa, M. R. (2006). Childbearing preferences and family issues in Europe SpecialEurobarometer (Vol. 253/Wave 65.1—TNS Opinion & Social). Brussels:European Commission.

Toulemon, L. (2004). Fertility among immigrant women: New data, newapproach. Population & Societies, 400, 1–4, April 2004.

United Nations Department of Economic and Social Affairs PopulationDivision. (2009). World marriage data 2008 pop/db/marr/rev2008.

Van Tubergen, F. (2005). Self-employment of immigrants: A cross-nationalstudy of 17 western societies. Social Forces, 84(2), 709–732 http://dx.doi.org/10.1353/sof.2006.0039

Van Tubergen, F. (2010). Determinants of second language proficiencyamong refugees in The Netherlands. Social Forces, 89(2), 515–534http://dx.doi.org/10.1353/sof.2010.0092