Abstract
Many claims are made about the significance of interethnic partnerships for individuals and for society. Such partnerships continue to be seen as a “barometer” of the openness of society and have spawned extensive analysis investigating their patterns, trends, and determinants. But we know little about the experience of children growing up in families of mixed parentage. In the United Kingdom, the increase in the self-defined “mixed” population is often celebrated. But there has been little quantitative sociological analysis that has investigated the circumstances of the children of mixed ethnicity partnerships. Using two large-scale UK datasets that cover a similar period, this article evaluates the extent to which mixed parentage families are associated with circumstances (both economic and in terms of family structure) that tend to be positive or negative for children’s future life chances and how these compare to those of children with parents from the same ethnic group. It shows that there is substantial variation according to the outcome considered but also according to ethnic group. Overall, children in mixed parentage families do not unequivocally experience the equality of outcomes with majority group children that the assimilation hypothesis implies.
Interethnic marriage has typically been considered as the end point of assimilation for immigrant groups (Wildsmith, Gutmann, and Gratton 2003, 563). Much of the early literature on the integration of immigrant groups assumed a progression from relatively segregated communities to the realization of greater social opportunities and finally to situations in which barriers to the most intimate of relationships—marriage—were no longer present. In this model, any distinction between social-cultural assimilation and achievement of economic status is often elided (Gordon 1964; see also Hwang, Saenz, and Aguirre 1997). In questioning who actually intermarries, studies tend not to question the implicit significance of interethnic relationships but, rather, have explored whether they are associated with economic advantage. For example, Meng and Gregory (2005) examine whether there are identifiable gains to be achieved for a minority group member intermarrying with the majority, though these are not evident for all groups in all countries (Muttarak 2007).
A complicating factor in ascertaining whether intermarriage does in fact bring economic benefits to the minority-group individuals concerned is the extent to which those who intermarry are typical of the minority population or are instead distinctive in that they have particular characteristics, such as an educational level not typical of their group (Duncan and Trejo 2005; Furtado 2006; Kantarevic 2004). Some studies maintain that those who intermarry are distinct from those who marry within their group, and therefore it is not possible to specify an advantage accruing to intermarriage per se (Kantarevic 2004), but other studies suggest the opposite (Meng and Gregory 2005). van Ours and Veenman (2008) account for the discrepancy by concluding that the “two hypotheses may be complementary rather than competing with each other” (p. 5). Both opportunity structures and normative criteria in partnership selection have also been extensively explored in papers on intermarriage (see, for example, Belot and Fidrmuc 2010). Studies that have plotted the incidence and trends in interethnic unions across groups and for a range of countries argue that when looking at the determinants—and consequences—of interethnic unions, both individual characteristics and structural constraints need to be taken into account (Hwang, Saenz, and Aguirre 1997; Kalmijn and van Tubergen 2006).
At the same time, there is an increasing focus on the assumed cultural barriers or obstacles to integration that minority groups confront. Though the focus of these studies has been much more on labor market participation (particularly of women) and residential segregation than on marriage, the family patterns of minorities are also closely linked to the literature on “self-segregation” (Battu and Zenou 2010; Bisin et al. 2008; see also Manning and Roy 2010; Georgiadis and Manning 2011).
Both of these perspectives imply that interethnic partnerships are positive for their participants as well as for society as a whole. In the traditional assimilation perspective, the focus was on the minority integrating, and that focus still dominates most analyses of intergroup relations. However, as well as the potential gains at the individual level, interethnic unions also are perceived as reflecting on society and societal relations more generally. Thus, they become a cause for celebration as they reveal particular societal values, such as openness and tolerance. For example, speaking of the United States, Fryer (2007) refers to “an understanding of the importance of interracial intimacy as a benchmark for race relations” (p. 73). Coleman (1994), in his analysis of intermarriage in the United Kingdom, suggested that intermarriage should be seen as an indicator of assimilation (or “amalgamation”) as well as be perceived as the route by which assimilation is achieved.
Nevertheless, Coleman (1994) goes on to question whether amalgamation also indicates economic integration, and other authors have questioned more directly the overall positive inferences drawn from the presence of interethnic relationships. For example, Berthoud (2005) has pointed out that the losses involved in assimilation through intermarriage may be substantial for minority groups. Model and Fisher (2002) make a similar point, but, focusing on the children of such unions, they argue that ethnic affiliation is diluted and that the children are more likely to perceive their identity as “socially constructed and situationally malleable” (p. 729). These analyses do not, however, link directly to the growing body of work that has directly explored the identity claims relating to children of interethnic partnerships (Edwards et al. 2012).
While “mixedness,” that is, mixed or multiple heritages, is often cited as a positive feature of societies, not all the focus on the increase in mixed populations has been positive. Edwards and Caballero (2008) discuss the tendency by commentators to regard mixed-race children as lost or in peripheral positions within a racially stratified society. These authors also draw attention to the public commentary about the appropriate identification of such children as “black,” as “black and white,” or as representing the incarnation of racial hybridity and contingent identity (Edwards et al. 2012). Mixed-ethnicity children have also been alleged to experience relatively poor social outcomes (Barn, Andrew, and Mantovani 2005), but recent analysis suggests rather that they experience more positive outcomes (Panico and Nazroo 2011).
Despite the interest in and concerns with mixedness and the issues that interethnic unions present for the self-identification of the offspring of such unions (Aspinall 2010; Duncan and Trejo 2005), little direct connection has been made between, on one hand, studies of interethnic unions and their economic implications and, on the other hand, the wide-ranging interest in the identity, experience, and development of mixed-ethnicity children. These disparities partly reflect disciplinary differences: many studies of mixedness are qualitative, while attempts to understand the determinants and consequences of interethnic unions are found mostly within the domains of quantitative sociology and economics. Nevertheless, the relative neglect of the consequences of mixedness is surprising, given the positive outcomes for intermarriage posited by the assimilation model and the relatively poor average outcomes indicated for mixed children in some cases (Platt 2012; Panico and Nazroo 2011).
One exception to this lack of attention in the literature is a study by van Ours and Veenman (2008) that compared the educational outcomes for children in the Netherlands who have one Moluccan and one Dutch parent with those having two Moluccan parents. The study found benefits for those with a Dutch mother and a Moluccan father but not for those with a Dutch father and Moluccan mother. The authors argued that the salient factor was the extent to which mothers were able to negotiate social systems to benefit their children, but they also acknowledge that their findings may not apply to parents of different ethnicities (or in other countries).
My analysis considers the structural position and well-being of children of interethnic unions compared with those of ethnically homogamous unions in the United Kingdom. It reflects on the implications of their outcomes in the context of an increasingly mixed-ethnicity population (Rees, Norman, and Boden 2011). There is clearly much to be understood about how the children of interethnic partnerships fare in a socially stratified society. The article uses two nationally representative datasets to take three complementary perspectives on this broader question in an attempt to disentangle different ways in which children’s well-being might be associated with varying outcomes when their parents are of different ethnic groups, compared with growing up in an ethnically homogamous family. The focus on children is relevant because it is in families with children where any positive or negative consequences will affect future patterns of social stratification.
First, I consider the issue of structural or economic assimilation while trying to take into account the endogeneity of interethnic partnerships and economic outcomes. Analysis of household data explores the risks of living in a workless family (in which no adults are employed) for otherwise similar families according to whether they are ethnically homogamous or heterogamous. If interethnic partnerships are markers of assimilative processes that are associated with more positive child outcomes, we would expect this to be reflected in the economic circumstances of these couples and their offspring. However, if there are gains to be had from communal resources that are lost to multiethnic families (Furtado 2006)—rather than being doubled through potential access to two networks of communities—such positive outcomes may not materialize.
Second, the analysis explores whether there are any apparent strains associated with the formation or maintenance of interethnic or homoethnic partnerships that affect children growing up in these families. It does this by looking at two indicators of potential strain that are associated with poorer outcomes for children: lone parenthood and maternal depression. It should be noted, however, that it is the poorer economic situation that frequently accompanies lone parenthood that is the major determinant of poor child outcomes on this indicator, rather than the family form itself (Schoon et al. 2011). This analysis also considers a missing part of the story for interethnic relationships where the partners bore a child together but are living apart after the birth. The selection of only those who remain as couples in analyses of interethnic (or coethnic) partnerships may lead to bias in our interpretation of the positive experiences for children associated with such partnerships. This analysis therefore helps to provide a more complete picture of not only those who maintain families with children across ethnic boundaries but also of those whose relationships are not normalized.
The final section of the analysis explores whether having parents from different ethnic groups has a direct effect on children’s adjustment, either positively or negatively, after accounting for factors associated with differences in early social-behavioral adjustment. This gives us some indication of the direct effect of the different parental environments on children, and it also allows us to examine whether differences in child outcomes begin to emerge early within the family context. Both datasets and all analyses allow specific ethnic groups to be distinguished—those that compose the main ethnic categories in the United Kingdom and that represent different histories and group experiences and characteristics. These are listed, along with their population distribution, in Table 1. The results highlight the different answers that come from considering the experience of children with a mother or a father from a particular ethnic group and the diversity between groups in their family circumstances.
The Distribution of Ethnic Groups in Great Britain, and Proportion UK-Born
SOURCE: Quarterly Labour Force Survey, pooled quarters Q4 2003–Q3 2010, weighted.
By looking at the experiences of children with parents of different ethnic groups, this article complements direct analyses of those of mixed ethnicity (see, for example, Platt 2012). By identifying whether and how early differentials in experience emerge, it also act as a barometer of the extent to which ethnicity still matters over and above the social location and interaction of individuals. These findings also provide some empirical foundation for claims that mixed-ethnicity families mark the end of an ethnically stratified society. From a policy perspective, then, the analysis can inform debates on cultural distance and whether it makes sense to so closely link the economic and the social in discourse and policy.
Overview and Data Sources
After outlining the broader patterns of interethnic partnership in the United Kingdom, this section briefly describes the two datasets that were used to explore the three issues raised above. The first dataset is derived from pooled waves of the Household Labour Force Survey (HLFS) from 2002 to 2005. This is used to estimate differences in workless family status, using propensity score matching to compare families that are similar in key respects, apart from the ethnic composition of the coresident parents. The second dataset derives from the Millennium Cohort Study (MCS), a longitudinal survey of children born around 2000 to 2001. The MCS is used to investigate their family context at nine months of age and how that varies with parental ethnicity. These MCS data draw on the information collected about nonresident as well as resident partners (fathers) from the mothers at the time of the interview. The last part of the analysis exploits the MCS’s longitudinal design to investigate whether differences in parental ethnicity at nine months affect children’s emotional-behavioral outcomes at age three, before the children experience the major change in social context and influences that comes with starting school. By bringing these different data sources and strategies together, the analysis engages with assumptions about the interconnectedness of different types of “assimilation” in an intergenerational context and their wider implications for equality and child welfare policies. The final section of the article draws some initial conclusions and indicates future lines of inquiry.
In the absence of a receptive social context, interethnic partnerships may bring additional pressures for families and family members, including choices about location in predominantly majority or more “mixed” areas, or the partnerships may bring advantages for children, as the van Ours and Veenman (2008) study showed. Within the constraints of data resources, for what is a rare event among small populations (see Table 1), this study cannot evaluate all the complexities of behavioral responses and settlement decisions. It aims instead to provide insight into the extent to which the United Kingdom is open to different ethnicities and highlight the potential costs of partnering outside one’s own group in an ethnically stratified society. Such an evaluation is not intended to suggest that interethnic partnerships are somehow good or bad but, rather, to provide an additional means to reflect on the consequences of current social organization.
Interethnic marriages or partnerships are, as noted, still relatively uncommon despite the extensive attention paid to them. They do, however, show substantial variation by specific ethnicity. For example, in the United States, 1 percent of marriages involving a white person, 5 percent of marriages involving a black person, and 14 percent of marriages involving an Asian person were “interracial,” where the relative population proportions were 87.3, 11.3, and 1.4 percent, respectively (Fryer 2007). For the United Kingdom, rates vary according to which source is used and when they are measured, but interethnic partnerships are also uncommon.
Table 2 provides recent estimates of interethnic partnerships alongside the rates of partnership overall for each of the main UK ethnic groups. Analysis of interethnic partnerships tends to focus on marriages, but some analysis has also considered cohabiting partnerships and the differences between these two types of unions (Muttarak and Heath 2010). Muttarak and Heath (2010) found that interethnic partnerships were more common among cohabiting than married couples. This was true across groups, though to differing degrees. They suggested that this could be interpreted in two ways. First, cohabitation could avoid the additional complications in embarking on a formal marriage across ethnic group lines. Second, those who cohabit could be less conventional in both their attitudes to marriage and to ethnic homogamy. In this analysis, partnerships and marriages are considered together for two main reasons. First, in the Labour Force Survey analysis, the interest is in families’ work status, specifically the work status of the two parents who are currently living with the child. From this perspective, it is their coresidence that is of primary interest rather than the legal status of their relationship. Second, in the MCS analysis, the fragility of relationships, represented by parents being nonresident when the child is young, is part of the investigation. To differentiate between possibly more stable marriages and less stable cohabitations would clearly confound this analysis. A third, more pragmatic, reason is that combining cohabitations and marriages gives greater purchase to the research questions through slightly larger sample sizes.
Partnership Patterns by Selected Ethnic Group and Gender: Row Percentages
SOURCE: HLFS datasets October–December 2004 to April–June 2008, weighted proportions.
NOTE: These are population-based estimates for all adults age 16 and older.
A further issue when considering interethnic partnership is what constitutes a “match,” in light of variable precision in measurement of ethnicity. For some groups, notably the groups designated “other” and “mixed,” that are frequently aggregated due to their small numbers, evaluating what does and does not constitute a coethnic union is not straightforward. An “other-other” union might represent a partnership between those of similar ethnicity in some cases but not in others. For those in the mixed groups, exact matches are likely to be rare, while matches within an aggregated mixed group (as used in some census analysis) are unlikely to represent ethnically homogamous unions. One assumption is that for specific mixed groups either part of the heritage can be used to represent a homogamous union—for example, that a person of white and black Caribbean heritage can be considered “matched” with either a white person or a black Caribbean person (Belot and Fidrmuc 2010). However, such an approach tends to deny the racialization of ethnicity within the United Kingdom. For this analysis, the focus is on the larger nonmixed, non-“other” minority ethnic groups and the white British majority. These represent relatively stable categories (Simpson and Akinwale 2007). Any partnership between a person from one of these groups and a person from any other group is deemed an interethnic partnership. In most cases, for minority groups this will constitute a match with a majority group partner. However, partnerships involving a different minority group are included in the current analysis because many of the same arguments relating to social openness or cultural loss still apply.
The variation by ethnic group in the chances of being not only in an interethnic partnership but also of being single is evident in Table 2. In addition, there are clear gender asymmetries for some groups. These patterns are not dramatically altered if only those who are born in the United Kingdom are included, or if only those who are in families with children are included. Rates of interethnic partnerships are somewhat lower among minorities in couples with children than among those in couples without children, though there has been less systematic analysis of the interethnic partnerships among those with children (Platt 2009).
The Household Labour Force Survey
The HLFS is a household-level dataset constructed from the quarterly Labour Force Survey. The Labour Force Survey is a short panel study. Respondents are interviewed for five successive quarters. About sixty thousand respondents are interviewed each quarter, composing a mixture of those at their first to fifth interview. It has an unclustered sample design and is representative of the UK population. The HLFS is a derived dataset that is designed for analysis at the household (couple) rather than the individual level. Two datasets of the HLFS are constructed by the Office for National Statistics and are released each year.
To create a sufficiently large sample for analysis, eight data files covering fall 2001 to spring 2005 were pooled. Thus, the period covered in these data is roughly the same time span as the MCS data used in subsequent sections. The rates of interethnic partnerships in these data were checked against those found in the 2001 census (Office for National Statistics 2005), and there was a high degree of congruence between the two. 1 The eight files from this period have been merged and are weighted by the appropriate weights for each period. Multiple observations on the same respondent were dropped from the sample to ensure unique respondent families in the pooled data. The small number of cases missing ethnic group were dropped, as were those few cases where the “head” or “spouse” was missing from couples. In addition, same-sex couples were excluded. When restricted to families with children under age 16, the final sample comprised thirty-eight thousand cases.
The HLFS collects a range of variables related to labor-market experience and status, education and training, health status, receipt of benefits, earnings, and some demographic variables including ethnic group. A selection of these have been used to match individuals in the following analysis, namely, education (highest qualification) of both parents, region, age of the mother, age of the child, whether the mother was born in the United Kingdom, the health status of the mother, and ethnic group. The outcome of interest across the matched groups is family level worklessness (i.e., neither parent is in paid work). In all, analysis weights are applied to adjust for nonresponse.
The Millennium Cohort Study
The MCS is a birth-cohort study of 19,517 children, who are followed over time. The children were sampled from all those born in the United Kingdom between September 2000 and January 2002 (Dex and Joshi 2005). The sample was selected from a random sample of electoral wards, disproportionately stratified to ensure adequate representation of all four UK countries, deprived areas, and areas with high concentrations of families from black and South Asian minority ethnic groups. The main respondents are predominantly mothers, but resident partners are also interviewed at each wave. Main respondents are asked a few questions about nonresident partners.
Surveys took place when children were about nine months old and at around three, five, and seven years old. The longitudinal design facilitates analysis of different outcomes and at different stages of the child’s life, and how these relate to earlier experiences and family background.
The survey is characterized by rich content from multiple participants: information on child health and development, parental prenatal and postnatal behavior, relationships within the family, cognitive outcomes, schooling, and so on. Its particular sample design allows for extensive analysis of child outcomes in relation to family, demographic and economic context, parenting practice, and parental health status.
At the nine-month sweep of the survey, information on family structure and on maternal depression is used as outcome variables. Parents’ ethnicity (whether or not a coresident with the child) that was collected at nine months was used to construct a measure of whether the couples were mixed or from the same ethnic group. This measure of interethnic partnership of parents was interacted with ethnic group in the analysis. It was included in the model alongside controls for maternal education, siblings, family income, and mother’s age. The outcome measures in both the HLFS and the nine-month sweep of the MCS are intended to capture structural experiences that have been shown to put children more at risk for poorer outcomes later in life.
The dependent variable used in this article for children’s outcomes at age three is emotional-behavioral difficulties, measured by the Strengths and Difficulties Questionnaire (SDQ). Behavioral outcomes have been selected as a specific well-being outcome, because the logic of the preceding analysis suggests that interethnic partnerships may be subject to greater social constraints and pressures than single-ethnicity partnerships. If this is the case, such pressures might be expected to reveal themselves in children’s behavioral adjustment. The age three sweep is selected because this represents a time when children are subject to family influences but are still in their preschool years. In addition to interethnic partnership interacted with ethnic group, additional controls were included from the age-nine-month sweep to establish prior conditions for the outcome of interest and thus reduce the chances of reverse causality. These controls comprised maternal depression, family structure, maternal education, sibling status, and child’s gender. All analyses using the MCS are adjusted for the complex sample design and for nonresponse.
Risks of Living in a Workless Family
Overall, in the HLFS, 7 percent of couples with children consisted of partners of different ethnic groups. Similarly, 7 percent of children living with two parents were living with parents from different ethnic groups. However, as the subsequent analysis shows, those who do not live with two parents may be somewhat more likely to have parents of different ethnicities. The overall rates of interethnic partnerships with children are quite low, though they vary considerably by ethnic group, as Table 2 shows.
The issue that I explore in this analysis is the extent to which children “benefit” from interethnic partnerships, after netting out some of the factors that traditionally account for such partnerships, such as education. That is, the approach estimates the extent to which children growing up in interethnic—as opposed to coethnic—partnerships are more or less likely to be living in a workless family at some point in time. Propensity score matching (PSM) was used to compare the outcomes of children living with parents of different ethnicities (treated) compared with those living with parents of the same ethnicity (untreated). There has been a noted increase in polarization of work at the household level, with an increase in both work-rich and work-poor households over recent decades (Gregg and Wadsworth 2001). Household or family worklessness has been identified as being detrimental for child outcomes over and above the low income it entails (Ermisch, Francesconi, and Pevalin 2004), and it varies by ethnic group in the long and short term (Platt 2010). Thus, worklessness could be seen as a significant risk factor that might be affected by compositional household factors, such as discrimination risks, faced by both partners or by dislocation from traditional opportunities for work, resulting from industrial restructuring or firm relocation. Exploring family worklessness as an outcome offers an advantage over individual measures of employment status because being in a working household does not require both parents to have jobs and, thus, in principle is not subject to debates about cultural preferences for particular gendered working-caring patterns within households (Dale, Lindley, and Dex 2006). 2
Overall, there was a higher probability of being in a workless family for children with parents of different ethnic groups. However, given the numerical dominance of the white population, these rates were driven by that group. When broken down by ethnic group and distinguishing the ethnicity of mother and father, Table 3 reveals that, with certain exceptions, interethnic partnerships appear to lower the chances that children will live in a workless family. Given what we know about selection in relation to interethnic partnerships (Kalmijn and van Tubergen 2006)—that is, that those who marry across ethnic boundaries differ in several respects from those who marry within, with minorities who marry out typically more highly educated—this may well simply represent the more highly qualified status of those parents in interethnic partnerships.
Rates of Worklessness in Families with Children by Ethnic Group and Partnership Status, Raw Rates, and Estimates from Propensity Score Matching
SOURCE: HLFS pooled datasets 2002–2005, weighted.
NOTE: Differences that do not approach statistical significance at the 10 percent level are in italics and statistically significant differences at the 5 percent level or above are in bold.
A priori, it was not clear that, once matched, the circumstances of children of interethnic parents would be more salutary than those of youths with parents from the same ethnic group. The matching was intended to reveal the extent to which those who partnered across group boundaries and had families together had access to economic advantages associated with assimilation over and above these selection effects. This could help to shed light on the nature of ethnic stratification in the United Kingdom by shifting the focus to the next generation rather than focusing primarily on labor market outcomes of adult minorities and migrants.
Drawing on Leuven and Sianesi (2003), the approach used nearest-neighbor matching, matching on age of child, mother’s age, region, mother’s education (highest qualification), father’s education, mswother’s health status, and whether born in the United Kingdom. A larger range of potential variables for matching were explored initially, but those chosen for the analysis were limited to those that had comparable distributions across matched samples. 3 Because the number of individual models estimated was substantial, the following discussion summarizes the key points of the analysis. The estimates are also illustrated in the second panel of Table 3.
Without accounting separately for ethnic group, the results showed a difference of 2.4 percentage points in worklessness experienced by children in interethnic compared with monoethnic families. That is, 8.3 percent of those in mixed-ethnicity families were living in workless families, compared with 5.9 percent of those in matched nonmixed households.
To explore differences that take into account specific ethnicities, matching was carried out in turn on the ethnic group of the father or mother. Focusing first on the majority population, differences in chances of family worklessness were estimated for those whose mothers and fathers were and were not in interethnic partnerships. Estimates showed that, comparing the matched cases, those families with a white British mother still experienced around a 1 percentage point higher risk of worklessness when in mixed parentage compared with ethnically homogenous families. Conversely, where white British fathers were living with a minority-group mother, the chances of family worklessness were also greater, by around 3 percentage points. This might indicate some form of ethnic hierarchy in terms of partnership selection.
The picture is more varied, and therefore the interpretation more complicated, for the minority groups. Children with an Indian mother had a 3 percentage point lower chance of living in a workless family if their father was of a different ethnic group, typically a majority-group member. This benefit was greater among those with a Pakistani mother, where living in a mixed-parentage family reduced the estimated risk of worklessness by 7 percentage points, from 14 to 7 percent. By contrast, those living with a Caribbean mother in a mixed-ethnicity couple had a 7 percent greater risk of living in a workless family than those with two Caribbean parents. However, when matching took place on fathers, the matched group showed no statistically significant variations in risks of worklessness based on whether the Caribbean father was or was not partnered with someone of his own ethnic group. This was also the case for Pakistani fathers. In both cases, this contrasts with the raw differences shown in Table 3 and indicates a high degree of selectivity for these two groups among those who partner with someone of a different ethnic group, albeit in opposite directions. Among Indian men, the chances of being in a workless family were lower for those with partners from a different ethnic group.
These results suggest that several factors influence the family work status of those families with children. First, the labor market position of men is key, such that white British women with children partnered with minority-group men have a higher risk of their children growing up in a workless family than minority-group women partnered with a majority-group man. The fact that white British men have a higher rate of worklessness and Indian men a lower rate of worklessness when living with someone from a different ethnic group may suggest that differences in women’s labor market position matter, with some minority-group women (especially Pakistani and Bangladeshi women) much less likely to be employed than majority-group women. However, if women’s rates of labor market participation were driving the results, we would expect to find a difference for Pakistani men partnered with a white woman (or other ethnic group), compared with Pakistani men partnered with a Pakistani woman, because Pakistani women have one of the lowest rates of labor market participation. There is also no difference for Caribbean men partnered with a Caribbean woman compared with Caribbean men partnered with a woman of a different ethnic group, even though Caribbean women have the highest labor market participation rates among women. Therefore, it is hard to make a case that women’s labor market participation rates are instrumental in the chances of children living in workless families. However, the results could reflect patterns of selection not captured in the matching, whereby exposure to those from different ethnic groups is implicated in more highly selected marriages for minority-group men and women. Conversely, for white men and women, partnership with a minority-group member may reflect intraregional variations in opportunities to marry out of their group that correspond to variations in workless patterns, possibly combined with a more conventional division of labor.
On balance, the data suggest that, rather than economic and social assimilation being coterminous, the economic position of different ethnic groups is significant in shaping family experiences, possibly regardless of the ethnic composition of partnerships. That is, parents can match across ethnic boundaries without implying that they have achieved equality in the economic sphere. The outcomes for children are thus particularly susceptible to the labor market position of their fathers, since fathers’ labor market position tends to determine family work status. While this might seem self-evident once demonstrated, it is important to note that it is at odds with both the conventional assimilation hypothesis that underlies much of the investigation of interethnic partnerships and cultural accounts of ethnic labor market positions (Bisin et al. 2008; Battu and Zenou 2010).
The story, though, is different for children with Caribbean mothers. The fact that those living with a Caribbean mother and majority-group father had a higher risk of worklessness than those living with two Caribbean parents, despite the well-known high unemployment rates for Caribbean men, would appear at odds with a structural account. It is important to note the very different incidence of interethnic partnerships among this group (see Table 2) alongside the high rates of lone parenthood among Caribbean mothers (see, for example, Table 6). In a context where marrying out is the norm rather than the exception and where partnerships are less likely to be cohabiting ones in the first place, the interpretation of differences between interethnic and coethnic cohabiting partnerships can be expected to change.
Much attention has been paid to the relatively high employment rates of Caribbean women with children, including lone mothers, compared with other mothers (see Reynolds 2005). If these patterns are shaped by partnership experience, it may mean that Caribbean women in interethnic couples are less likely to work and therefore less likely to help avoid worklessness within the family. That is, women in such partnerships may follow more conventional (normative) gendered patterns of division of labor. In addition, in the United States, there have been discussions of the extent to which parental cohabitation may be contingent on the work status of the father (Wilson 1978). This hypothesis has also been suggested for the United Kingdom, given the very different unemployment rates (and lone parenthood rates) that exist across minority ethnic groups. If rates of lone parenthood are influenced by work status of fathers, this might indicate that Caribbean women with children are more likely to have employed partners when in cohabiting coethnic relationships, as, according to the Wilson (1978) argument, they would be more likely to remain lone mothers if the father of their children were unemployed. And indeed, we can see in the data that single Caribbean men have much higher unemployment rates than cohabiting Caribbean men. The difference is found across all groups to a greater or lesser extent, but it stands out for Caribbean men: single Caribbean men have unemployment rates 10 percentage points higher than cohabiting Caribbean men. Thus, in this case, the economic sphere would be shaping observed patterns of family life rather than being independent of it. Conversely, in interethnic partnerships there may be processes of segmented assimilation (Portes and Zhou 1993) at play that indicate that the social integration is interlinked with less favorable outcomes for certain groups (Peach 2005). There would therefore appear to be group-specific ways in which family dynamics interact with structural constraints.
Maternal Depression and Lone Parenthood
Analysis of the MCS has highlighted a range of risk factors for subsequent behavioral and cognitive child outcomes, including maternal depression (Kiernan and Mensah 2009), low socioeconomic status (Schoon et al. 2011; Ermisch 2008), and various aspects of parenting (Ermisch 2008). Family structure has been found to have only limited explanatory power net of other factors with which it is strongly associated, such as poverty, maternal depression, and so on (Schoon et al. 2011), but in absolute terms it represents a major risk factor for children. Lone parenthood rates also differ substantially across ethnic groups and may indicate a range of obstacles to maintaining a conventional relationship with the father, which could conceivably be exacerbated for relationships that cross ethnic divides in an ethnically stratified society. Maternal depression is also known to show variation across ethnic groups and may also signal additional pressures that the family faces. These two risk factors (maternal depression and family structure) are therefore explored here, using the first survey of the MCS families, when the cohort children were nine months old.
The MCS, unlike many other surveys, collects some minimal information, including ethnic group, on nonresident partners. This means that it is possible to measure the ethnicity of the millennium babies’ parents regardless of whether they are both coresident. In turn, this allows us to examine whether the probability of lone parenthood, identified here as a risk for children’s subsequent outcomes, is more or less likely for parents of the same ethnicity compared with those with a partner of a different ethnicity.
These data are not susceptible to the matching approach that was used to address the endogeneity issues in the HLFS analysis, and thus the relationships shown in the analysis can be regarded as only indicative. However, the analysis provides a complementary perspective to the first set of results and enables a fuller insight into the experience of children growing up in mixed-ethnicity families. Probit regressions were estimated to determine the chances of millennium babies living in a lone-parent household or living with a mother experiencing maternal depression, controlling for a limited range of family variables that are recognized indicators for these two family outcomes. 4 An interethnic partnership dummy was included in the models and was interacted with mother’s ethnic group to ascertain if any association varied by ethnicity. Interethnic partnership was further interacted with maternal education to take some account of the ways in which educational selection plays out for different ethnic groups in interethnic partnerships.
Table 4 shows the results from these two probit regressions. The maternal-depression model indicates that there was no significant relationship between parents from different ethnic groups and maternal depression (controlling for lone-parent status). This indicates that we cannot reject the null hypothesis that interethnic partnership is unrelated to maternal stress levels across ethnic groups. There is no clear evidence that family stress is specifically impacted by mother’s cross-cultural partnering and societal reactions to it overall. However, the lower rates of maternal depression that minority-group mothers experienced are canceled out by being in an interethnic partnership for Indian and black African mothers. That is, Indian and black African mothers partnered with someone of the same ethnic group have significantly lower rates of depression than white majority women; but Indian and black African mothers partnered with someone from a different ethnic group have rates of depression that are not significantly different from white majority women.
Probit Regressions of Lone Parenthood and Maternal Depression (standard errors)
SOURCE: MCS sweep 1. Design weights applied.
NOTE: GCSE = General Certificate of Secondary Education, which is the set of exams that children take at the end of compulsory schooling at age 16.
p < .10. **p < .05. ***p < .01. ****p < .001.
When looking at the model for lone parenthood, we can see that having parents from different ethnic groups was significantly associated with the chances of lone parenthood. It is not possible to infer any causal relationship from this analysis, but it is tempting to speculate about the extent to which intimate relationships across ethnic boundaries are simply harder to sustain—or to normalize in conventional forms such as cohabitation or marriage—even when they result in children. To identify the scale of the effect and also to explore the interaction effects and examine the extent of variation across ethnic groups, predicted probabilities were calculated at the means of key variables and varying by maternal education (high or none). Table 5 shows these selected probabilities for white, Indian, and black Caribbean mothers. Not all of the differences are statistically significant, but they do show, once again, how the relationship between mixed-ethnicity parents and risk factors varies according to the mother’s ethnicity. The levels vary so substantially across ethnic groups that they tend to dwarf the differences across partnership type.
Predicted Probabilities of Lone Parenthood for Mothers by Ethnicity of Partner and Qualifications, Selected Examples (in percentages)
NOTE: Other characteristics were set to: sample means of maternal age and log income and to no siblings.
Results show that children of both Indian and white mothers are more likely to live in lone-parent households at nine months old if their fathers are from a different ethnic group than their mothers. This is consistent with the relationship-strain thesis and is a finding that cannot be picked up by analyses that explore the experience of only married or cohabiting couples. This pattern is also found for black Caribbean mothers who are highly educated, but the reverse relationship is found for black Caribbean mothers with lower educational qualifications. It is worth noting that the reverse relationship was obtained among those for whom lone parenthood is most common. The finding potentially enhances our understanding of the findings in the first part of the article, where it was suggested that the employment status or potential unemployment risk of the prospective partner or father might shape the extent to which cohabitation occurred for this group. Caribbean men who did not cohabit were a more economically precarious group. The most economically vulnerable fathers are thus likely to be those who partner with less well-qualified mothers, which in turn appears to reduce the chances of cohabitation, for a group where lone parenthood is much more common. This results in findings that go against the overall trend where having a partner from a different group is more likely to result in lone parenthood.
These findings amplify those from the first set of results. They show that the outcomes for those children growing up with parents of different ethnic groups may differ from other families partly as an artifact of which parents are observed as cohabiting couples. Because for most groups, lone parenthood is more likely among mothers partnered with someone from a different ethnic group, those couples that are together likely represent the best case situations.
Children’s Emotional-Behavioral Outcomes
Given these findings about risks associated with having parents from different ethnic groups, an obvious succeeding question is whether interethnic partnerships translate into poorer or better outcomes for children on direct measures. The evidence so far gives limited support to a hypothesis of relationship strain for families with children where the parents are from different ethnic groups, evidenced by a relative increase in maternal depression for some groups and greater rates of lone parenthood. We might therefore expect such strain to affect children’s own behavioral-emotional outcomes over time. However, if those couples who remain together provide slightly more advantageous outcomes for children, and if lone parenthood is itself a key mediating factor for children’s early life outcomes, any possible negative impact may already be factored out through relationship patterns, and the reverse pattern may then apply.
To investigate these competing hypotheses, the analysis exploited the SDQ for children at the second sweep of the MCS, when the children were around three years old. At this age, children are developing within the family context and able to display and be evaluated for problems in adjustment or behavior. At the same time, they are not yet in school and thus not susceptible to the large range of influences and social judgments and expectations that occur after that transition. Models were therefore estimated to investigate whether there was an effect related to interethnic partnerships (again interacted with ethnic group) over and above known risk factors relating to the differences in the SDQ score. Given the inclusion of lone parenthood and maternal depression as potential mediating factors and the strong association between lone parenthood and low income (and to a lesser extent with maternal depression and ethnic group), income was not included in these models.
The SDQ is a suite of twenty-five questions that is designed to capture the behavioral-emotional difficulties of children ages 3 to 17. It also measures prosocial behavior (Goodman 1997). It can be administered to parents or teachers or (from around the age of 11) to children for self-evaluation. In the MCS, the SDQ was administered to parents and covered in the self-completion section of the questionnaire so that responses were not influenced by the presence of an interviewer. The survey asks five questions in each of the five domains: emotional symptoms, conduct problems, hyperactivity/inattention, peer relationship problems, and prosocial behavior. Each question is scored 0 (not at all true), 1 (partly true), or 2 (certainly true), with some reverse scoring. Scores are summed across the domains, and a total-difficulties score can be created by adding together the first four domains. The SDQ has been well validated for use in community samples (Goodman and Scott 1999) and is widely applied across different countries.
There are two ways of using the total-difficulties score in the SDQ. At the top end of the distribution, a level covering around 10 percent of children has been defined at which the child can be considered “abnormal” on the particular domain or for total difficulties. Much analysis also examines differences across the whole distribution, transformed into z-scores (e.g., Del Bono and Ermisch 2009), rather than depending on a binary division of children into “normal/borderline” and “abnormal.” For robustness, models were estimated using both of these approaches. 5 Thus, a probit model was estimated that regressed the binary variable for “abnormal” difficulties against key regressors while a linear regression regressed z-scores of difficulties on the same key characteristics to explore variation across the range of difficulties.
Because there is substantial and significant variation between boys and girls in their chances of experiencing difficulties, the gender of the child was controlled in the model. Whether parents were from different ethnic groups was included in the model and interacted with mother’s ethnic group. In addition, the key risk factors for negative outcomes that were explored in the previous section (lone parenthood and maternal depression) were also modeled, along with maternal education as a summary measure of family background. For robustness, given the reduction in sample size across the two sweeps and some nonresponses in the self-completion, 6 ethnic group was reduced to a smaller set of categories, combining Pakistani and Bangladeshi mothers into one group and excluding the smaller and “other” ethnic groups.
Table 6 shows the results from the probit regression of difficulties above the threshold and of the linear regression estimates of difficulties. The results were largely consistent across the two approaches. Having parents from different ethnic groups had no overall independent effect on children’s chances of scoring high on the SDQ. Thus, it is neither a specific risk factor nor protective factor for this outcome once other characteristics are controlled, as we might expect. As anticipated, risks of difficulties are strongly associated with partnership status (being a lone mother), maternal depression, and maternal education at nine months. In line with other research, children in Indian and Pakistani-Bangladeshi families face higher risks of difficulties at this age, though this association attenuates over time (Hansen et al. 2010), and children in black African families experience lower levels of difficulties. There was some evidence from the continuous model that for Pakistani and Bangladeshi families, the negative effect of ethnicity is not found in interethnic partnerships; that is, where the mother is partnered with someone from a different ethnic group, the outcomes are equivalent to those in white (or Caribbean) families. Because the number of such partnerships is small, and because the probit estimates do not confirm these results, these estimates should be treated with caution. Nevertheless, they do provide some indication that where interethnic partnerships are rare and for those groups that face high levels of disadvantage, interethnic unions may afford a protective effect on children’s behavior. This protective effect may derive from unobserved characteristics of the partners in such unions that are not fully captured in the analysis and may pertain to different contexts of upbringing. Nevertheless, it suggests a further potential positive association with interethnic partnerships, at least for those groups that face substantial overall disadvantage, and warrants further investigation.
Probit and Linear Regressions Exploring Children’s Experience of Behavioral Difficulties at Age Three, Controlling for Family Circumstances at Age Nine Months (standard errors)
SOURCE: MCS sweeps 1 and 3. Design weights applied.
p < .10. **p < .05. ***p < .01. ****p < .001.
Discussion and Conclusion
Interethnic unions are a relatively neglected topic in the United Kingdom. Yet they are potentially of great relevance to how we think about society and the future, and exploring these unions can lead to questioning models of interaction and assumed assimilative processes and the approaches we bring to these processes.
This article has shown that, overall, children living in families with mixed parentage may be at greater risk of living in a workless family than those living with ethnically homogamous (couple) parents. However, when taking into account the ethnicity of the parents, this is shown to be the case for only certain groups. For those with white British or Caribbean mothers, having a father from a different ethnic group may increase their risks of growing up in a workless family, and the same is true for those with a white British father. For those with Indian or Pakistani mothers or Indian fathers, the presence of a parent from a different ethnic group appeared protective in relation to risks of worklessness. However, there were no differences associated with partnering out compared with coethnic partnerships for Pakistani and Caribbean fathers. These distinctions suggest that differences in outcomes are not a consequence of purely selection effects and that the disadvantage associated with minority status, particularly for earners (fathers), matters for children’s exposure to worklessness. For Indian fathers, whom we already know to have high employment rates, the differences between those in families with Indian wives and those with majority (or other ethnic group) partners may reflect unobserved selectivity or labor market context of this group.
The result for black Caribbean mothers needs to be understood in the context that, for this group, interethnic partnerships are not as uncommon as they are for other minority groups. Moreover, in the context of high rates of lone parenthood, the observed coethnic cohabiting partnerships may represent particular sets of circumstances that are more conducive to positive outcomes. Since, only those family circumstances that are relatively stable were observed, there are clearly questions about the positive selectivity of cohabitating coethnic partnerships if more vulnerable situations (lone parenthood) that were not observed are also associated with greater risks for children growing up amid worklessness.
The analysis, therefore, next considered the risks of living in a lone-parent household or of maternal depression in explaining whether interethnic partnerships may be associated with greater stress, given that social or cultural proscriptions are potentially associated with such partnerships. Because both of these factors affect child outcomes, it was thus pertinent to establish whether they were affected by partnership status. The analysis found no evidence for increased maternal depression overall where parents were of different ethnic groups. However, the rates of maternal depression among Indian and black African mothers, which were typically lower than those of white majority mothers, became comparable to majority levels when Indian and black African mothers were in interethnic partnerships. They thus lost their “advantage” when in such partnerships and became similar to white majority mothers in their risks, other things being equal. The analysis also showed that for most groups, the chances of lone parenthood were greater for mothers of young children, when the parents were of different ethnicities. However, the story for black Caribbean mothers followed a more complex pattern, varying by level of education and the differential access to resources (and opportunities for types of partnerships) that goes with education level. While better-educated black Caribbean mothers had higher rates of lone parenthood when the father was of a different ethnic group, lower-educated black Caribbean mothers had higher chances of lone parenthood when the father was of the same ethnic group. This is consistent with the speculative interpretation of the first part of the analysis: that the circumstances of the potential partners may be highly related to who is observed in couples, especially where neither interethnic partnership nor lone parenthood is uncommon. This is consistent with arguments relating to forms of segmented assimilation in the UK context (Peach 2005). Overall, it is important to note that the chances of lone parenthood vary dramatically across groups and dwarf the differences associated with mixed-ethnicity parentage.
The third section of the analysis focused directly on children’s outcomes at age three: their rates of difficulties revealed by the SDQ. Results indicate that children’s outcomes are not systematically sensitive to whether parents are from the same or different ethnic groups. However, there are indications that for children growing up in what are typically the most disadvantaged families the increased risk of experiencing behavioral-emotional difficulties is moderated when a majority-group parent is present.
There is clearly not a single process by which interethnic unions lead smoothly to assimilative outcomes for minorities. It is unlikely that either the matching or the regression approaches explored here take account of the full range of relevant factors that might help us to understand differences in the contexts in which children grow up and how they differ according to patterns of parental partnership. The local context of families is particularly hard to capture, given the relatively small sizes of specific minority ethnic groups and the relative rarity of mixed-ethnicity partnerships for most groups; and it is not within the scope of the analysis to unpick the mechanisms further. Nevertheless, for some groups it could be argued that economic and social “integration” do not go together and that economic context and ethnic labor market stratification trump a context in which social barriers are becoming increasingly porous.
Overall, the evidence that interethnic partnerships bring additional stresses and difficulties is not substantial; in most cases, however, rates of lone parenthood are greater for mothers with a partner of a different ethnic group. This suggests that it may be easier for families to remain together if the parents share ethnicity or where there are more proscriptions against alternative family forms. This may help to explain an additional dimension of selectivity in interethnic partnerships: it is the most successful that are typically observed, since analysis can rarely account for noncohabiting relationships or nonresident parents.
Lone parenthood and maternal depression are important factors in children’s subsequent outcomes and life chances. Once those mediating factors are taken into account, there is little evidence that interethnic partnerships cause further stress that manifests itself in children’s social-behavioral outcomes. Conversely, there is a slight suggestion that for groups that are most disadvantaged and most susceptible to poor social-behavioral outcomes when children are three years old, and where interethnic parenting is rare, such partnerships may be protective in terms of child outcomes.
Minority-group mothers who both make and maintain cohabiting interethnic partnerships tend to be positive for children, but this is likely to be in part a consequence of those specific factors that characterize such parents, since being able to sustain these relationships tends to require more positive environments. Moreover, where interethnic partnerships are more common, there is less that is special about interethnic partnerships, and thus their positive aspects are not as evident. Finally, the findings for majority-group mothers parenting with minority-group partners indicate that this is not an equal process. Commentators may be right to celebrate mixed relationships as an indication of social openness and a changing social world, but at the same time, the existence of such relationships does not imply that children growing up in such families will necessarily have the same life chances as do those from the majority. Social openness on its own will not bring or reflect economic equality. A continued focus on redressing the structural inequalities that face minority-group members is necessary to bring dividends in terms of better future outcomes for children regardless of an ethnic group’s patterns of partnership and rates of intermarriage.
Footnotes
Note:
I am grateful for comments and contributions on earlier versions of this article from participants in seminars at Stirling University, the Policy Studies Institute, the University of Kent, and the APPAM and INSIDE 2011 conferences. I would particularly like to thank Kate Choi and Randy Capps for their insightful discussions of the article and the editors of this issue for their careful reading and constructive comments. I started this article at the University of Essex and completed it at the Institute of Education, but I carried out much of the work on it during an EC-funded Fellowship at the Institute for Economic Analysis (IAE) in Barcelona. I am grateful to all institutions for their support.
Notes
Lucinda Platt is a professor of sociology at the Institute of Education, University of London. She is director of the Millennium Cohort Study, a survey of more than nineteen thousand children born between 2000 and 2001, who are followed over time. She has published widely in the areas of ethnicity, poverty, and inequality.