Expectations for family transitions in young adulthood among the UK second generation

ABSTRACT This paper explores whether family transitions among descendants of post second-world-war immigrants are converging towards those of white British young adults by examining family formation expectations among 16–21 year-olds collected within the 2009–2015 waves of Understanding Society. The paper asks: Do current adolescents’ cohabitation, marriage and parenthood expectations differ by ethnic group? Are differences similar for men and women? Are ethnic differences mediated by individual or parental socio-economic characteristics? We find that expectations for marriage and parenthood are unanimously high, but that there is greater uncertainty among white British and black Caribbean adolescents as to the age at which these transitions will occur. We find large ethnic differences in expectations for cohabitation, especially for women. There is evidence for a divergence in expectations within the south-Asian community. Second-generation Indians have lower expectations for marriage and higher expectations for cohabitation than second-generation Bangladeshis or Pakistanis. Ethnic group differences remain when religiosity, parental background and individual characteristics are controlled. Further research is required regarding the mechanisms which underlie the differential transmission of family formation attitudes across ethnic minority groups.


Introduction
Over the past 40 years family formation trajectories to adulthood within Western countries have become extended and de-standardized in terms of their sequencing and reversibility (Elzinga and Liefbroer 2007). Cohabitation has overtaken marriage as the first partnership type, with parenthood increasingly occurring outside of marriage, delayed to later ages, or foregone (Balbo, Billari, and Mills 2013). These changes are part of the second demographic transition (Van de Kaa 1987) wherein new living arrangements such as cohabitation are seen as expressions of secular sentiments of younger cohorts (Lesthaeghe 2010). In individualised Western cultures, diversity, exploration and instability are key characteristics of Arnett's emerging adults, who have left the associations, not detailed mechanisms for observed differences. We establish whether there are ethnic differences, whether these differ by sex, and whether they are mediated by parental background factors, religiosity, or the young person's educational and employment situation. The following section summarises the UK context and existing evidence. Subsequently the paper reviews mechanisms for the inter-generational transmission of family attitudes and describes the analytical framework which sees adolescents' expectations for family formation influenced by ethnic group, religiosity, parental background factors and structural integration in terms of education and employment (Plotnick 2007).

The UK context: ethnic differences in family formation
The UK experienced considerable post-war immigration: black Caribbeans primarily from Jamaica in the 1950s and 1960s, Indians from India and Africa in the 1960s and 1970s; Pakistanis and Bangladeshis in the 1970s and 1980s (Peach 2006). First generation migrants from India, Pakistan and Bangladesh tended to replicate traditional patterns of early childbearing within marriage (Ballard 1990;Jejeebhoy 1998), but with significant differences in the overall level of fertility. Family sizes of first generation Indian migrants, including highly educated Indians who migrated from East Africa, were significantly lower (around 2.7) at the outset, than those of women arriving from (predominantly rural areas of) Pakistan or Bangladesh at around 5 and 7 births per woman respectively (Dubuc 2012).
Traditionally, the Caribbean family system was highly individualised with a family centred on two or three generations of women and their children (Shaw 2017). Foner (1977) describes how marriage in rural Jamaica took place at a relatively late age. Premarital relationships were common, and often did not involve common residence, men remaining in their parents' household. First generation Caribbean migrants to Britain are thought by Foner (1977) to have been rather more traditional in their family formation patterns, with shorter durations of non-marital cohabitation and a younger age at marriage, so as to comply with expectations for early marriage dominant in the UK in the 1950s and 1960s.
Research on the experience of the second-generation born in the UK in the 1950-1970s suggests a rapid convergence across ethnic minority groups in overall family size towards the UK average. Fertility levels during the 1990s and early 2000s among second-generation Indian women were similar, or even a little lower than the UK average. Second-generation Pakistani and Bangladeshi women had fewer children at younger ages as compared to their mothers' generation, but continued to have slightly larger completed family sizes than the UK average due to more third and fourth births (Kulu et al. 2017). Existing findings suggest there has been rather less convergence across ethnic groups in the types of partnership (Berrington 1994;Hannemann and Kulu 2015). Among the second generation, rates of cohabitation were significantly higher among black Caribbeans, and much lower for south Asians, amongst whom marriage continued to be the norm (Berrington 1994). New evidence from the UK Household Longitudinal Study (UKHLS) describing the average timing and sequencing of family formation for cohorts born in the UK in [1960][1961][1962][1963][1964][1965][1966][1967][1968][1969][1970][1971][1972][1973][1974][1975][1976][1977][1978][1979] (Table 1) is consistent with this literature. Only one in five British-born black Caribbean men and women had married by age 30. However, almost two thirds had lived with a partner, and 80% of first births were prior to marriage. In contrast, British-born south Asians were far more likely to have marriedonly a minority living with their partner beforehand, and the vast majority made the transition to parenthood within marriage. Previous research highlighted how second-generation Indian women were postponing marriage and childbearing to later ages (Berrington 1994), associated with their older age at leaving full time education and a greater likelihood of paid employment (Dale et al. 2002). This postponement can also be seen in Table 1 where 62% of Indian men and 74% of women had married by age 30. Nevertheless the likelihood of marriage among British Indians was significantly higher than for the white majority.
The literature emphasises differences in marriage practices within south Asians. Ballard (1990) highlighted differing marriage rules whereby Sikhs and Hindus were barred from marrying their close kin, whereas Muslims were encouraged to do so. These rules were associated with very different levels of transnational marriage. Although early cohorts of Sikhs born in the UK sought spouses from the Punjab, locally arranged marriages (or matches with the overseas Sikh Diaspora) were the norm by the 1980s (Ballard 1990). In contrast, marriage with Pakistan-based partners remained much more frequent amongst UK-based Mirpuris, not least due to parents' obligations to their Pakistani-based biraderi (Shaw 2017). Whilst parents traditionally play an important role in the selection of a spouse among south Asian minorities, there has been a shift in the perceptions of arranged marriages (Qureshi, Charsley, and Shaw 2014) with second-generation Asians wanting to have a say in whom they marry (Ahmad 2012;Din 2006). Young women especially are using their greater educational achievements and financial autonomy as leverage in marriage decisions (Ahmad 2012;Shaw 2017). Additionally, researchers have discussed how Islam has increasingly been used as a resource for young women to Note: NA refers to situation where the denominator is less than 50 respondents. Weighted percentages and unweighted sample size. Source: UKHLS 2009/10. a Includes those who arrived in the UK before age 6. b Percentages are based on the denominator who had entered a first co-residential partnership by age 30. c Percentages are based on the denominator who had become a parent by age 30. d Sample refers to the overall number in the sample, irrespective of whether they had already married or become a parent. assert their rights in terms of marriage decisions (Qureshi, Charsley, and Shaw 2014;Din 2006). Thus the role of parents is increasingly described in terms of 'assisted' (Ahmad, 2012) or 'introduced' marriage (Twamley 2014), rather than 'arranged'.
It is important to distinguish between the behaviour of second-generation young people born in the 1950-1970s (the focus of past research), and the expectations of more recent birth cohorts. Transitions in young adulthood have generally become more uncertain and risky (Furlong and Cartmel 2007). Current adolescents face greater difficulty in securing employment with sufficient stability and level of income to allow residential independence and family formation. Ethnic capital (Shah, Dwyer, and Modood 2010) may act as protection for some ethnic groups by promoting educational achievement. The increase in the age at leaving full time education has been more rapid among ethnic minorities (Crawford and Greaves 2015). Thus the timing of transitions to the labour market and partnership formation among recent cohorts of ethnic minorities are likely to be delayed as compared with earlier cohorts (Ní Bhrolcháin and Beaujouan 2012). Finally, in comparison to earlier cohorts of second generation, attachment to the labour force among south Asian women is now stronger (Arcarons 2020), due to increased human capital and language proficiency (Khattab, Johnston, and Manley 2018). Therefore, in this paper we examine ethnic group differences in expectations for family formation among cohorts born in 1989-1999, now aged 16-21.

Inter-generational transmission of expectations for family formation
Assimilation theories (Alba and Nee 1997;Portes and Zhou 1993) explain how family formation behaviours of ethnic minorities change according to the cultural distance between origin and destination country, time since arrival, and generation (Kulu et al. 2017). Where the norms and family formation preferences dominant in the country of origin contrast with those of the country of destination, second generations are exposed to both parents' preferences and behaviour, but also ideas and norms gained as a result of being socialised in the destination country (De Valk and Liefbroer 2007). The behaviour of second-generation ethnic minorities will thus be more like that prevailing in the destination country than was the case for their parents. Such 'linear assimilation' theories have been critiqued for ignoring the continuing importance of transnational relationships among ethnic minorities (Qureshi, Charsley, and Shaw 2014;Reynolds 2006;Shaw 2017) and for assuming convergence towards a single (Western) model of individualised family system (Ahmad 2012;Shaw 2017). Nevertheless, most researchers agree that 'normative conflict' between second-generation youth and their parents is especially likely in the area of sexual activity and partnership formation (Giguère, Lalonde, and Lou 2010;LaLonde et al. 2004).
The extent to which young adults from more collectivistic cultures are likely to adopt more individualistic attitudes towards family formation depends upon their socialisation, religion, and structural integration in to the host society (Gravel et al. 2016). Parents play the most important role in socialisation, influencing young people via their behaviour, parenting styles and attitudes (Peterson and Bush 2013). Parents act as role models (Arcarons 2020) and their own experiences of marriage, separation and childbearing are associated with the family formation trajectories of their young adult children (Axinn and Thornton 1996). Young people who have been brought up in a lone parent family tend to view marriage with more uncertainty and have more positive attitudes towards cohabitation (Crissey 2005), choosing cohabitation as a way to test their own relationship before committing themselves (Berrington, Perelli-Harris, and Trevena 2015b).
In south Asian families where traditional cultural values emphasise the authority of elders and the well-being of the wider kin group, parenting styles tend to be more authoritarian with greater surveillance and control over young adults (Peterson and Bush 2013), particularly girls (Din 2006). Religiosity is generally associated with more traditional patterns of family formation (Berrington, Perelli-Harris, and Trevena 2015b), with cohabitation and non-marital childbearing associated with increased secularisation (Lesthaeghe 2010;Van de Kaa 1987). Part of the explanation for continued early marriage in south Asia is the desire to protect daughters from unsanctioned sexual activity prior to marriage (Kamal et al. 2015). Religious identity continues to shape the lives of Muslim women in particular because traditional gender roles and chastity prior to marriage are linked to prestige and family honour/izzat (Shaw 2017). In the UK, south Asian parents' expectations regarding chastity prior to marriage are often a source of tension and conflict as second generations are influenced by the more liberal values of the majority white British culture. The intergenerational transmission of cultural norms about sexual behaviour and partnership formation thus involves 'actively managing conflicting messages about acceptable behaviour' (Sen Das 2018, 2), and there are variations e.g. according to ethnicity, religion and education in the degree to which more traditional rules regarding chastity are enforced. Ahmad (2012, 205) suggests that 'getting to know someone for the purposes of marriage' (but not dating in the western sense of 'boyfriend' or 'girlfriend') is increasingly acceptable among university-educated British Muslims. Dating and premarital sexual relationships appear to be increasing among second-generation British Indians (French et al. 2005) among whom 'love marriages' are often preferred (Twamley 2014).
Social interactions also occur outside of the parental homewith other relatives, peers and social institutions. Kulu et al. (2017) suggest that high levels of residential and school segregation, particularly among the UK Pakistani and Bangladeshi communities, help maintain traditional patterns of family formation. Residential independence from the parental home in young adulthood is also likely to be important. Twamley (2014) argues that the anonymity and financial independence of young Gujaratis living in London allows young couples to have sexual relationships without being watched over by their elders.
The intergenerational transmission of family attitudes among second generations will be influenced by the level of structural integration of young adults into British society, particularly their educational and employment experiences (Li and Heath 2020). For both the white majority and ethnic minorities prolonged educational enrolment delays family formation due to role incompatibility (Ní Bhrolcháin and Beaujouan 2012). Rising levels of attainment additionally influences family formation largely through increased economic independence for women and rising opportunity costs of reducing hours spent in the labour market for childrearing. For second generations socialised within more traditional, collectivistic cultures, participation in higher education and in the labour force provides 'greater cultural capital and access to material resources, making it easier for them to make their own choices, sometimes against their parent's wishes' (Ferrari and Pailhé 2017, 36).

Family expectations in young adulthood
Although plans regarding marriage and family formation shift during adolescence (Willoughby 2010), expectations reported in young adulthood have been shown to be useful predictors of behaviour (Clarkberg, Stolzenberg, and Waite 1995;Willoughby 2014). In the US expectations to marry have remained very high (Manning, Longmore, and Giordano 2007), although Black young people perceive a lower likelihood of marriage compared to Whites (Crissey 2005). However, we might not anticipate as high expectations for marriage in the UK since cohabitation has replaced marriage to a greater extent (Berrington, Perelli-Harris, and Trevena 2015b). Cohabitation is not institutionalised and lacks a symbolic event (a wedding), meaning that expectations for cohabitation may be more tentative and more uncertain than for marriage (Manning, Longmore, and Giordano 2007). Previous UK research suggests that fertility preferences are formed early on in the life course, and that intended childlessness among British men and women is relatively rare (Berrington and Pattaro 2014). Nevertheless, just under a fifth of the female population remains childless at the end of their reproductive years (Berrington, Stone, and Beaujouan 2015a), and thus some uncertainty among adolescents as to the likelihood of becoming a parent would be expected.

Data and methods
The UKHLS is a longitudinal survey of the members of approximately 30,000 (private) UK households (Knies 2016;University of Essex et al. 2018). Adults in participating households are interviewed annually. Individuals joining original households, together with children of original households who reach age 16 ('rising 16s'), become part of the sample and complete an adult interview. Data collection for each wave is scheduled across 24 months, wave 1 taking place in 2009 and 2010. An Ethnic Minority Boost (EMB) sample was designed to provide at least 1000 adults from each of five groups: Indian, Pakistan, Bangladesh, Caribbean, and African (Knies 2016). The EMB was achieved by oversampling areas with a higher density of ethnic minority participants (Lynn 2009). From wave 2, members of the British Household Panel Survey were also included. For further details about the survey see Platt and Nandi (2020). The main analysis sample consists of young adults born between 1989 and 1999 either in the UK or born abroad but who arrived in the UK prior to age 6. These 16-21 yearolds were questioned within a special 'young adult module' which asked about expectations for cohabitation, marriage and parenthood. Young adults who gave a full interview in wave 2, 3 or 5, when the relevant questions were asked (n = 7366) are included. Only the first wave in which they were a respondent is included in the analysis (even if the response was 'don't know'). In this way respondents appear only once in the analysis, and their responses are not affected by panel conditioning. The implication of including the first response is that the age distribution is biased towards the younger end of the age range. At wave 6 two new questions were included in the 'young adult module' asking respondents their expected age at marriage and parenthood. Thus for a sample of 3240 men and women aged 16-21 in 2014/15 we highlight ethnic group differences in the expected timing of family formation.
Dependent variables: expectations for cohabitation, marriage and parenthood Young adults are asked a series of probabilistic expectation questions (Manski 2004). Past research based on such quantitative expectations suggests that young adults are able to form reasonable beliefs about significant life events , and that adolescents' expectations are positively related to their actual experiences (de Bruin, Parker, and Fischhoff 2007). The UKHLS survey asks: 'Please tell me how likely it is that the following events will happen in your life in the future. If any of the following events have already happened, just let me know. On a scale from 0% to 100%, where 0% means "No chance of happening" and 100% means "Totally likely to happen". The respondents are asked 'How likely is it that you will … … Marry at some time / Live together unmarried with a partner / Have a child?' The respondent is presented with a card showing a horizontal line from 0% to 100% with end points labelled: 0% labelled "No chance will happen" and 100% labelled "Totally likely to happen". The very small number (maximum 3% found for the white British group) who have already experienced the event, are given a value of 100. A potential challenge is the heaping of responses on 0, 50, or 100. However, Manski (2004) among others has found that respondents do use the full expanse, often rounding to the nearest five percent in the middle of the range, and reportomg one percent intervals at the extremes. There is a debate as to whether the value of 50 represents a real probability, or epistemic uncertainty (Fischhoff and De Bruin 1999;Bruine de Bruin et al. 2000).

Dependent variables: expectations for age at marriage and parenthood
In wave 6, respondents are asked 'At what age do you want to get married/ would you like to start a family?' If the respondent does not want to get married/become a parent the interviewer is instructed to enter a zero. A 'don't know' response is permitted, and is given as a response by a significant minority of young men and women (See Supplementary materials, Tables S4 and S5 for details). Mean expected age at marriage and parenthood are calculated based on the sample who say they intend to get married/become a parent and give a numerical response. Table 2 shows the distribution of independent variables for the main sample.

Key independent variable: ethnic group
Ethnic group was collected in the main adult interview through self-identification from 18 ethnic categories on a showcard. These categories have been collapsed into: white British, Indian, Pakistani, Bangladeshi, black African, black Caribbean & black Caribbean/white British mixed, and a final group including other and mixed. Information on ethnic group for 'rising 16s' (i.e. those without a prior adult interview) was gathered from their response to a youth questionnaire completed when they were aged between 10 and 15. Information on ethnic group was not available for 210 young people, predominantly 'rising 16s' from Northern Ireland who are part of the BHPS. They are included within a 'Not known' group. Mixed black Caribbean/white British respondents are added to those of black Caribbean heritage since high rates of inter-ethnic marriage between the white and black Caribbean population in the UK mean that there is a growing mixed population of young adults, and a relatively small group of young people who identify as black Caribbean (Voas 2009). Initial analyses of expectations among these two groups demonstrated similarity.

Other independent variables
Sex is entered as a dummy variable. Age in completed years is a continuous variable. Generation differentiates those who have at least one parent who was born in the UK (2.5 and 3rd generation) from those whose parents were both born overseas (second generation). Religiosity is based on the respondents' response to the question (included in wave 1 and wave 4) 'How much difference would you say religious beliefs make to your life?' The variable contrasts those who say that religion makes a great or some difference with those reporting that it makes little or no difference. Young people not present at either wave 1 or wave 4 are coded as 'not known'. Parental separation is a dummy variable identifying those whose parents had either never lived together or who had separated. Since a significant minority of adolescents was not living with their father at the time of interview, we focus on maternal characteristics. For those no longer resident with their mother, we use responses to questions about their mother's level of education and whether she was working when the respondent was 14. Maternal education is coded as no qualifications, less than degree, degree. Mother in paid work is coded as 0 = no, 1 = yes. Mother's age at first birth is grouped as: under 20, 20-24, 25-29 and 30+. Item non-response is dealt with by the inclusion of not known categories. Educational aspirations, coded as either 'low' or 'high', are used instead of attainment in order to include young people who are still studying. Among those who have left education, achieving A levels and above qualifications puts the young person in the 'high' aspiration group, whilst those whose highest qualification is either AS levels, GCSEs or below are coded as having 'low' aspirations. Those currently in further or higher education are coded in the 'high' group. For those still enrolled in school or college we code those who aspire to achieve A levels and equivalent as 'high', with those who only aspire to achieve GCSE/ AS /AVCE level qualifications as 'low'. Economic activity identifies whether the respondent is currently employed, a full time student, unemployed or economically inactive. Region of residence distinguishes those living in London, from those living in the North, the Midlands, the rest of the south and East, and those living in Scotland, Wales and Northern Ireland.

Analytical strategy
First we present descriptive analyses from the probabilistic expectations for cohabitation, marriage and parenthood, and for expected age at marriage and parenthood. Since a significant proportion of the answers to the expected age at marriage and parenthood were uncertain, and given the relatively small sample sizes within particular ethnic groups (see Tables S4 and S5) we do not attempt multiple regression analyses of this outcome. However, f we undertake proportional odds ordinal regression of grouped expectation (0-24%, 25-49%, 50%, 51-75% and 76-100%) of cohabiting, marrying and becoming a parent. Two models are fitted for each outcome, the first contains just ethnic group, and sex, age and immigrant generation. We test for the significance of a two-way interaction between ethnic group and sex, but only find it to be significant in the regressions of cohabitation expectations. The second model includes religiosity, parental background factors, and the respondent's educational and employment characteristics. Comparison of the odds ratios for ethnic group in model 1 and in model 2 provides some indication as to whether ethnic group differences are mediated by these additional factors. All analyses are weighted to be nationally representative using the cross-sectional weight corresponding to the wave in which the data were collected and all analyses take into account the clustering in the survey design (Knies 2016;Lynn 2009  Note: 1 Includes those who arrived in the UK before age 6.

Results
Expectations for cohabitation, marriage and parenthood Figure 1(a,b) shows the average expectations for cohabitation, marriage and parenthood by sex and ethnic group. (See Supplementary materials tables S1-S3 for detailed distributions). Few respondents (less than 5%) gave themselves a zero probability of marrying or becoming a parent, though a significant minority of non-white British adolescents do not expect to cohabit. Just over half of Pakistani and Bangladeshi men, and almost three quarters of Pakistani and Bangladeshi young women gave an expectation of cohabitation of zero. By contrast only around one third of Indian men and one half of Indian women give report a probability of zero. All these groups differ from white British youth: only 5% and 4% of men and women give a zero expectation (Table S1). White British youth -17% of men, 25% of womenwere significantly more likely to give a response of 100% that they would cohabit. Black Caribbeans expect to cohabit more than south Asian youth. However, 8% of Indian men had a 100% expectation of cohabiting (Figure 1(a,b)). Mean expectations for marriage and parenthood are generally high (Table S2), but there are differences by ethnic group: the mean expectation of marriage for south Asian men was 80-90% compared with an average of 69% for white British and black Caribbean men. The lowest average expectation for marriage is found for black Caribbean women (63%). It is only among black Caribbean youth that we see a greater expectation for childbearing, than for either cohabitation or marriage. In sum, ethnic differences are greatest for cohabitation and smallest for parenthood. We find some evidence of diversity within the south Asian groups: Expectations to cohabit are higher for young Indians, than Pakistanis or Bangladeshis. Ethnic differences in expectations to marry or become a parent are similar by sex, but differences in expectations to cohabit according to ethnicity are larger for women than for men. Figure 2(a,b) shows the mean expected age at marriage and parenthood by sex and ethnic group. Given the small sample sizes caution is required in interpretation, especially because a high percentage of young people (around one third of men, and one quarter of women) report that they 'don't know' when they might get married (Table S4). Similarly, almost a third of young men and one fifth of young women are uncertain as to when they will start a family (Table S5). Whilst many are uncertain, relatively few (less than four percent) said they did not want to marry or have a childconsistent with the results from the probabilistic questions.

Expected age at marriage and parenthood
Among those who gave a numerical answer, women reported expected ages at marriage that were on average one year lower than those reported by men. Among both men and women in this sample, ethnic group variations in expected age at marriage are greater than ethnic variations in expected age at parenthood. Pakistani and Bangladeshi youth reported younger expected ages at marriage (around age 25 and 24 among men and women) as compared white British youth (28 and 27, respectively). Expected age at marriage among the small group of black Caribbean men is highalmost 30 years.
The mean reported expected age for entry into parenthood is 30 years for men and 28 years for women. Expected age at entry into motherhood is similarly high at around 27 Note: 1 Includes those who arrived in the UK before age 6. years for young women from all of the south Asian ethnic groups, suggesting a future postponement fertility from the earlier profile of entry into motherhood observed for the 1960-1979 cohorts (Table 1), particularly for women of Pakistani and Bangladeshi heritage. Whilst among south Asian ethnic groups the expected mean age at parenthood is somewhat higher than that of marriage, among men and women of Black Caribbean heritage the expected mean age at marriage is actually higher than for parenthood. Given the small size of the sample and high levels of uncertainty, we do not go further in examining how these differences are mediated by individual and family background characteristics.
Are ethnic differences in expectations for family formation mediated by individual and parental background factors? Table 3 shows the proportional odds ratios from ordered logistic regression models of individual expectations for cohabitation, marriage and parenthood. The odds ratios tell us how a particular category of a covariate, relative to the reference category, is associated with the likelihood of being in a higher category of expectation, ceteris paribus. In the first model for marriage expectations, the odds of being in the highest category for expectation (76% and higher), versus being in any of the lower categories (0-24. 25-49, 50, and 51-75 combined) are 2.26 times higher for Indian than for white British young adults (the reference group). The corresponding odds ratios for Pakistanis (2.61) are little higher, whilst the odds of being in the highest expectation category are four times higher for Bangladeshi youth than for white British youth. Expectations for parenthood are also higher among young south Asians; the odds of being in the highest expectation group (76-100% likelihood of parenthood), as compared to one of the combined lower groups are 79% and 74% higher for Indian and Pakistani youth, as compared with white British young people, and nearer three times higher for Bangladeshi youth. For the model of cohabitation expectations, we find a significant interaction between sex and ethnic group. Among white British youth, women are more likely than men to expect to cohabit, but that this is not the case for non-white ethnic groups among whom men are often more likely to expect to cohabit. Expectations for cohabitation are particularly low among south Asian women, but are higher among those who have at least one parent born in the UK, than among those whose parents were both born overseas.
Comparison of the first and second models in Table 2 shows that ethnic differences in expectations for family formation are not accounted for by religiosity, parental background or the respondent's educational aspirations and employment experience. The odds ratios for ethnic group remain largely unchanged in the model of cohabitation expectation. For the models of expectation for marriage and parenthood, the odds ratios associated with ethnic group either stay the same or increase slightly when other factors are controlled.
As anticipated, religiosity is associated with family expectationsyoung adults who reported that religion makes little or no difference to their lives are one third more likely to be in the highest expectations group for cohabitation, and are significantly less likely to be in the highest expectations group for either marriage (odds ratio 0.59) or parenthood (odds ratio 0.79), as compared those for whom religion makes some or a great difference.  Parental demographic behaviour is significantly associated with young adults' expectations. Young adults whose parents had stopped living together are less likely to have the highest expectations for marriage or parenthood. Those whose mothers had started childbearing in their teens are less likely to have the highest expectations for marriage, but also less likely to have the highest expectations for parenthood. Maternal education is positively associated with marriage and parenthood, but not significantly associated with expectations for cohabitation.
Young adults' with low educational aspirations were less likely to expect to form a coresidential partnership. This is not the case for parenthood however, where educational aspirations did not seem to be relevant. Unemployed young adults had lower expectations for all threecohabitation, marriage and parenthood. Current enrolment in education is associated with lower expectations for cohabitation and marriage but not associated with expectations for parenthood.

Discussion
Theories of family change highlight the role of generational succession as younger cohorts embracing new secular norms regarding family formation (Lesthaeghe 2010). Understanding young peoples' expectations can provide insights into the new normative order (Manning, Longmore, and Giordano 2007), and the likely ways in which white-British and second-generation ethnic minority youth in the UK will make their demographic transitions to adulthood. This paper finds significant ethnic group differences in expectations for the timing and type of family formation which remain once religiosity, parental background and individual characteristics are controlled. Whilst religiosity and experience of parental separation were associated with family formation expectations in the anticipated direction, they do not seem to mediate ethnic group differences.
Expectations among second generation born in the 1980s are to some extent consistent with ethnic group differences observed for first generation migrants, and second generations born in the 1960s and 1970s. Marriage continues to be the expected partnership form for the majority of south Asian men and women, whilst cohabitation is normative among black Caribbean and white British youth. However some generational changes are observed, not least the anticipated delay to later ages of marriage and parenthood, especially among south Asian women. Moreover, there appears to be divergence within south Asian groups as to expectations for cohabitation, which are far higher among young second-generation Indians, than Pakistani or Bangladeshi young adults. Additionally expectations for cohabitation among south Asian men are considerably higher than for south Asian women.
A striking finding from this research is the continued expectation for marriage across ethnic groups. Despite increases in divorce and childbearing outside of marriage, young people are not rejecting marriage. Even among those groups who tend to marry lessthe white British and Black Caribbeanvery few men and women did not expect to marry. Whilst it could be the case that young adults feel normative pressure to report expectations for marriage, they are consistent with previous qualitative research suggesting that marriage continues to be seen as an ideal partnership form and the ultimate form of commitment.
In this paper, we found no evidence to suggest that the idea of marriage is rejected by black Caribbean adolescents, which raises questions as how only a very few second-generation black Caribbeans born in the UK in the period 1960-1979 had their children within marriage. The continued desire for marriage co-exists, however, with considerable uncertainty as to the age at which it will actually happen, especially among white British and black Caribbean youth. The uncertainty reported by white British and second-generation black Caribbean youth as to whether they will marry contrasts with the firmer expectations found among south Asian young adults.
In the US, lower marriage rates among the Black population tend to be explained in terms of their disadvantaged economic position, and the idea that Black women place greater emphasis on economic support for marriage to be in place, as compared their white counterparts (Manning, Longmore, and Giordano 2007). Economic uncertainty may also be an important factor deterring marriage among second-generation Black Caribbeans in the UK who continue to face disadvantages in education and employment. Evidence from qualitative work suggests that for many UK adults, there is often ambivalence about marriage which can be a low priority relative to other financial and family commitments such as securing stable accommodation and financing children's upkeep (Berrington, Stone, and Beaujouan 2015b).
Much has been documented about the role of cohabitation as a way of finding out more about a partner, and of marriage as the next natural step in the evolution of a partnership (Berrington, Perelli-Harris, and Trevena 2015b). The individualised norms of the white majority emphasise 'love' in the selection of a spouse. Marriage decisions generally come after a prolonged period of courtship and period of cohabitation. This is in contrast to marriages traditionally formed within south Asian communities where parents are often involved in the search for a spouse and the period of acquaintance prior to marriage tends to be short (Ahmad 2012;Shaw 2017). However, marriage practices have been changing among Indian, Pakistani and Bangladeshi communities in contemporary Britain, with young adults increasingly being involved in the choice of their spouse (Ahmad 2012). Twamley (2014) finds that among second-generation Gujarati Indians living in London, just as for the white majority, 'love marriages' are likely to be preceded by a period of cohabitation as the relationship grows in a series of natural steps. This evidence from qualitative interviews is consistent with the findings of this paper whereby a significant minority of second-generation Indians expect to cohabit. The role of cohabitation in the family life course varies, however, across individuals and over time. Whilst cohabitation as a precursor to marriage may well increase among second-generation Indians, cohabitation as an alternative to marriage and a setting for childbearing is less common (at least among the cohorts born [1960][1961][1962][1963][1964][1965][1966][1967][1968][1969][1970][1971][1972][1973][1974][1975][1976][1977][1978][1979], in contrast to the experience of the white majority among whom childbearing prior to marriage is increasingly normative.
The findings from this paper suggest that the 'explorations in love' that Arnett (2000) describes as being a key feature of emergent adulthood, may not be equally present across ethnic groups. Of particular interest is the possible divergence in the family formation experience of second-generation young Indians, and young Pakistanis and Bangladeshis. Further research is required to elucidate reasons for this differentiation, which are likely to include religious beliefs and traditional marriage rules (Ahmad 2012). Pakistani and Bangladeshi communities in the UK have higher rates of consanguineous and transnational marriage than Indian communities in the UK (Shaw 2017;Qureshi, Charsley, and Shaw Departments, with scientific leadership by the Institute for Social and Economic Research, University of Essex, and survey delivery by NatCen Social Research and Kantar Public. The research data are distributed by the UK Data Service. Neither the data collectors nor depositors are responsible for the analyses and interpretations in this paper.

Disclosure statement
No potential conflict of interest was reported by the author.

Funding
This work was supported by Economic and Social Research Council grants ES/K003453/1, RES-625-28-0001, ES/K007394/1 and carried out in the ESRC Centre for Population Change (CPC).