The Age of Independence, Revisited: Parents and Interracial Union Formation Across the Life Course

Abstract

Romantic relationships that cross racial lines have grown since anti-miscegenation laws were deemed unconstitutional. In The Age of Independence, Rosenfeld argued that parental influence over children’s mate selection processes had waned. Rosenfeld, however, was not able to test this supposition directly because of his reliance on cross-sectional census data. Using Waves I and III of Add Health for a cohort of individuals from 1994 to 2002, we examine whether parents matter in shaping their offspring’s romantic attachments, by exploring whether adolescent reports of maternal closeness and parental control are associated with youth’s likelihood of being in an interracial relationship in emerging adulthood. We find that parental factors do influence emerging adults’ romantic relationships; these associations vary by race, ethnicity, and gender. Among white men, maternal closeness in adolescence reduces the likelihood of being in an interracial relationship in emerging adulthood. Parental control elevates the odds of being in an interracial relationship among black and Hispanic women. We also find that parental decisions on where families live shape offspring’s choices, as relative exogamous group size in adolescence is associated with interracial union formation in later life. Our findings suggest that parental influence remains salient in the partner choices made by emerging adults.

INTRODUCTION

Since 1967, when the U.S. Supreme Court ruled that anti-miscegenation laws were unconstitutional (Loving v. the State of Virginia), the number of mixed-race marriages, cohabiting unions, and dating relationships has increased rapidly (Black-well and Lichter 2000; Qian and Lichter 2007). Many view the increasing prevalence of interracial relationships as an indicator of blurring racial boundaries (King and Bratter 2007; Qian and Lichter 2007). Others (Rosenfeld 2007) attribute these dramatic shifts to the decline in parental authority and control over children’s mate selection. In his book, The Age of Independence, Rosenfeld (2007) argued that the delayed transition to adulthood experienced by contemporary young adults leads to greater autonomous decision making regarding partner choice. This, he suggested, in conjunction with youth’s residential mobility, has diminished the role parents play in their children’s choice of romantic partners, with a resulting increase in interracial and same-sex unions.

The argument set out in The Age of Independence (Rosenfeld 2007) was that young people of the early 1990s and 2000s were more independent and more willing to select romantic partners without consideration of parental approval than previous generations. Yet a sizable body of literature argues that young adults were still dependent on their parents in the turn of the twenty-first century (Hardie and Seltzer 2016; Newman 2012). In 2000, approximately 23.2% of young adults aged 18–34 were living with a parent (U.S. Census Bureau 2018). Coresidence with parents also varies by race and ethnicity, with minority youth more likely to reside with parents than white young adults (Furstenberg 2010; Hardie and Seltzer 2016). Young adults remain in school for longer, and often the more advantaged receive parental assistance in paying for school (Furstenberg 2010). Parental socialization and monitoring also vary widely by the gender of the child (Madsen 2008), with parents more lenient toward sons than daughters (Sassler, Ciambrone, and Benway 2008). Furthermore, young adults of the 2000s often report high levels of parent-child support (Hardie and Seltzer 2016).

It is therefore important to revisit Rosenfeld’s claims and broaden their focus. Rosenfeld (2007) could not directly explore the association between parental control and children’s partner choice, as his analysis relied on census data, which did not assess measures of parent-child involvement beyond coresidence. His analysis on parental involvement was also limited to whites, despite abundant evidence that parent-child relationships vary by race, ethnicity, and gender (Aquilino 1997; Foner and Dreby 2011; Hardie and Seltzer 2016; Rossi and Rossi 1990). Finally, Rosenfeld did not address other secular factors that may facilitate interracial relationship formation—such as whether one grew up in a neighborhood with a greater representation of other groups, or the rise in less formal unions, such as dating and cohabiting relationships, where partner choice is more expansive than in marital unions.

In this article, we directly test the association of adolescent reports of parentchild relationships and their likelihood of being in an interracial relationship in emerging adulthood, with a cohort of individuals from 1994 to 2002. Our article uses two measures of parental influence: maternal closeness and parental control.

Due to choice or circumstance, parents also determine the neighborhoods where they raise their children (Goyette, Iceland, and Weininger 2014); we explore whether adolescent tract-level population composition of race and ethnicity, or relative group size, is associated with interracial union formation in emerging adulthood. To further assess Rosenfeld’s argument that residential mobility facilitates independent partner choice, we also include a measure of postmobility relationship formation and examine whether it is associated with an increased likelihood of being in an interracial relationship. Finally, we expand Rosenfeld’s earlier focus on white youth by including racial and ethnic minorities and broaden the scope to explore gender variation in the association between parental factors and partner choice.

Transitions to Adulthood: Meaning, Change, and Variation

What it means to be an independent young adult has changed dramatically in the past century. For young adults of the early 2000s, many markers of adulthood, including living away from parents, completing higher education, being financially independent, getting married, and having children, occurred later and in different sequences than they did for previous generations (Furstenberg 2010; Rosenfeld 2007). Furthermore, the proportion of young adults that lived with their parents has grown over the past few decades; approximately 23% of young adults lived with their parents in the 2000s (Furstenberg 2010; U.S. Census Bureau 2018). One key aspect of the transition to adulthood—selecting a mate—has changed in many ways. In the 1980s, approximately 42% of young adults aged 18–34 were never married, but this increased to 53% by the early 2000s (U.S. Census Bureau 2018). But as the pursuit of higher education has become more protracted and stable jobs that pay enough to support a family become more difficult to find, marriage is increasingly delayed (Furstenberg 2010). The increased duration between completing high school and “settling down” provides new opportunities for relationship exploration (Sassler, Michelmore, and Holland 2016), which has expanded young adults’ opportunities to meet romantic partners from varied backgrounds (Rosenfeld 2007). In addition, cohabitation has increased, though the proportions cohabiting have grown more among whites and Hispanics than among blacks (Manning 2012). This ever-lengthening period of “emerging adulthood” (Arnett 2004) may weaken the impact of parental influence over young adults’ choices over the life course.

Parental Influence

Parents have historically steered their offspring’s mate selection, approving some partners and discouraging others (Kalmijn 1998; Miyawaki 2015). Regarding parental influence over interracial relationship and union formation, Kalmijn (1998:400) argued that parents may prevent mixed marriages because they “threaten the internal cohesion and homogeneity of the group.” Yahirun and Kroeger (2017) find that aspects of parent-child relationships in adolescence, such as perceived closeness, warmth, and care, are associated with a decreased likelihood of being in an interracial relationship in young adulthood.

Group positioning can also factor in; parents of the dominant group may view intermarriage as threatening their status position (Blumer 1958). Specifically, white parents may disapprove of interracial relationships more than racial and ethnic minority parents. White family members were found to be the least supportive of their kin involved in interracial marriages, while blacks were the most supportive of those in interracial relationships (Lewis and Yancey 1995). Compared to those in monoracial relationships, those in interracial relationships are less likely to receive and perceive kin support. Whites in interracial relationships were less likely to receive residential and financial support, as well as perceived residential, financial, and childcare support, compared to blacks and Hispanics in interracial relationships (Bratter and Whitehead 2018).

Parents can play a role in their children’s romantic partner choices by offering their opinions of potential mates and withdrawing support for relationships they do not sanction (Lee and Bean 2010). Yahirun (in press) finds that blacks in interracial marriages visit their mothers less often compared to those in racially homogamous unions. Children, however, may also act against the wishes of their parents (Kalmijn 1998), highlighting the difficulty of determining the causal direction of the association—which Rosenfeld (2007) could not assess with the cross-sectional data he used.

On the other hand, parents may also be open to their children being in interracial relationships. A majority of Americans in the early 2000s were not opposed to a close relative marrying someone of a different race or ethnicity, but a hierarchy of acceptability emerged; they were the most opposed to a relative marrying a black partner, followed by a Hispanic, Asian, and white partner (Livingston and Brown 2017). Our expansion of Rosenfeld’s examination to a larger group of racial and ethnic groups allows us to assess whether group positioning and parental influence on partner choice are linked. In this article, we use measures of maternal closeness and parental control to proxy for parental influence. Maternal closeness is the degree of attachment of the child to the mother (Kapinus and Gorman 1994), while parental control is defined as “the degree and manner in which parents attempt to place constraints on their child’s behavior” (Longmore, Manning, and Giordano 2001:324). In essence, parental control and maternal closeness are two faces of a coin; if closeness is the carrot, control is the stick that parents can use to shape their children’s behaviors in ways they deem appropriate.

MATERNAL CLOSENESS

Across the life course, maternal closeness tends to be lowest during adolescence but increases in young adulthood. Using retrospective reports of closeness, Rossi and Rossi (1990) found that maternal closeness was lowest at age 16, followed by age 10, and then ages 19–29. They argued that closeness was lowest during adolescence as children “tried out their wings” in preparation for independence, but rebounded in the twenties due to maturation and a better understanding of the nature of parenting. Nonetheless, early experiences of parent-child relationships are important, even if closeness changes with age, as they set the stage for later parentchild interactions. Adolescents who experience cold and distant parenting may be less able to develop warm and close relationships with parents later in life (Aquilino 1997; Rossi and Rossi 1990).

Research examining the role of maternal closeness has generally focused on the role mothers play in deterring the sexual activity of children rather than on interracial union formation. These studies find that maternal closeness is associated with delayed sexual onset (Longmore et al. 2001; Manlove et al. 2012). Among two-parent families, high-quality relationships with mothers were associated with delayed sexual debut among boys (Manlove et al. 2012). Maternal closeness may also vary by race, ethnicity, and gender, with greater perceived parental supportiveness among whites relative to minority groups (Hardie and Seltzer 2016) and closer mother-daughter relationships relative to mother-son relationships (Suitor and Pillemer 2006). Immigration status also factors in; first-generation immigrants reported greater parental supportiveness relative to second-generation Americans (Hardie and Seltzer 2016).

The role of parent-child relationships, including parental closeness, on the formation of interracial unions has been associated with a decreased likelihood of being in an interracial relationship (Yahirun and Kroeger 2017). Therefore, we anticipate that maternal closeness reduces the likelihood of being in an interracial relationship in emerging adulthood (Hypothesis 1a). Whether the association of maternal closeness and the likelihood of being in an interracial relationship varies by gender, race, and ethnicity is an open question, though the literature suggests that maternal closeness may matter more for girls relative to boys (Hypothesis 1b) and for whites compared to racial and ethnic minorities (Hypothesis 1c).

Parental Control

A broader definition of parental control is “parental behaviors towards the child that are intended to direct the child’s behavior in a manner acceptable to the parent” (Barnes et al. 2006:1084). Among the facets of positive control are behaviors associated with the related concepts of discipline, supervision, and monitoring of adolescent behavior. Parental control and monitoring are associated with later sexual debut, and parental control declines from childhood to adolescence (Long-more et al. 2001).

Parental opportunities to exert control over their children’s choice of romantic partners may vary by race and ethnicity. There is a large body of evidence documenting differences in parenting styles by race and ethnicity. In general, white parents are more likely than black, Hispanic, or Asian parents to encourage independence among their adolescent children (Chao and Aque 2009). Parents will ostensibly have more control over children’s partner selection when they marry young than if they delay marriage into their late twenties. Furthermore, parents may have greater control over their children’s selection of marital partners than over partners in less formal unions, such as dating or cohabiting unions (Blackwell and Lichter 2000). Research on dating relationships suggests that some groups— Asian American men and black women—are less likely to be involved in any form of romantic or sexual relationship in young adulthood, though this appears to be less a function of parental control and more attributable to these groups’ placement in the racial and ethnic hierarchy of dating desirability (Balistreri, Joyner, and Kao 2015).

Parental behavior also differs by the gender of the child. Daughters report experiencing greater levels of parental monitoring and involvement in their romantic relationships compared to sons (Madsen 2008; Sassler, Ciambrone, and Benway 2008). In general, over many age groups and family arrangements, parents (often mothers) were more controlling over their daughters’ curfew, behaviors with their romantic partners, and clothing choices than they were over sons’ (Madsen 2008). We therefore anticipate a negative relationship between parental control and the likelihood of being in an interracial relationship (Hypothesis 2a), though we expect to observe a stronger association between parental control, maternal closeness, and interracial relationships among women relative to men (daughters relative to sons) (Hypothesis 2b). We also expand on Rosenfeld’s analysis by hypothesizing expected associations for racial and ethnic minorities. Prior research has found that Latinas who reported more controlling parents, in particular fathers, were more likely to be in an interracial relationship as a way to escape patriarchy (Vasquez-Tokos 2017). Given the body of research on racial and ethnic differences of parental control, we anticipate a stronger association between parental control and interracial unions among minorities than for whites (Hypothesis 2c).

Other Factors Shaping Maternal Closeness and Parental Control

Of course, other factors play important roles in the mate selection process of contemporary young adults. Social class, in particular, shapes parenting styles, as well as the structural context shaping the people with whom children come in contact. Parental social class also shapes the pursuits young adults engage in (Lareau 2003). More educated parents report exerting less control over young children but also noted less closeness and more issues with control and conflict than did less educated parents (Aquilino 1997). That may be because respondents with more educated parents also had a greater tendency to be in romantic relationships (King and Harris 2007). The family structure adolescents experience while growing up exerts considerable influence on when they begin engaging in romantic relationships and how such relationships progress (King and Harris 2007; Longmore et al. 2001), as well as their likelihood of entering into interracial relationships. Relative to those who grew up in single-parent families, those who grew up with two biological or adoptive parents were less likely to have a first sexual partner of a different race and ethnicity (King and Bratter 2007) and progressed more slowly into cohabiting relationships overall, and more specifically sexual and cohabiting relationships (Sassler, Michelmore, and Holland 2016).

Relative Exogamous Group Size

Other factors, such as relative exogamous group size of minority and ethnic groups, shape contact with other racial and ethnic groups. The relative size of one’s group matters as opportunities for interracial interaction are limited to the markets that are available for partner choice (Blau 1977; Choi and Tienda 2016). Minority groups, because of their smaller size, have fewer partners from whom to choose within their own racial and ethnic group than do whites, and this is reflected in their partnering behaviors. The research shows that when it comes to the choice of marital and cohabiting partners, Hispanic, Asian, and black immigrants were more likely to be in an interracial relationship than were native-born whites (Qian and Lichter 2007). Furthermore, consistent with assimilation theory (Gordon 1964), with increasing generation in the United States, the likelihood of being in an interracial relationship increased among racial minorities (Qian and Lichter 2007). Contact with racial and ethnic groups also shapes interracial relationship formation. In the contact hypothesis, Allport, Clark, and Pettigrew (1979) argued that the way to reduce prejudice was to increase interpersonal contact between racial and ethnic groups. Empirical evidence shows that those who live in communities that are more diverse are more likely to have interracial friendships (Vanhoutte and Hooghe 2012). Early interaction with others of different racial and ethnic backgrounds may be key; those who have had interracial relationships earlier in the life course (have a first sexual partner of a different race or ethnicity) are more likely to marry interracially (King and Bratter 2007).

Parents, due to choice or circumstance, may also select the neighborhoods in which their children reside—yet another aspect of involvement. Evidence suggests that white parents often leave racially and ethnically diverse neighborhoods and choose to live in areas that are predominantly white (Goyette et al. 2014). In selecting new neighborhoods, white parents often rely on recommendations from highstatus parents in their social networks, which may result in re-creating racially homogeneous neighborhoods (Holme 2002). Relative group size, then, may result from parental decisions in choosing neighborhoods, or parental “tastes” for remaining in diversifying neighborhoods. Alternatively, residing in a more diverse neighborhood may be due to circumstances—an indicator of economic difficulty in relocating in the face of growing neighborhood diversity (Goyette et al. 2014). Therefore, we expect that a higher relative exogamous group size in adolescence will be associated with an increased likelihood of being in an interracial relationship in emerging adulthood (Hypothesis 3).

Leaving the Nest, or Moving to a New Location

Moving away from the state of one’s birth might also be formative in shaping the partnering behaviors of young adults. In fact, Rosenfeld (2007) asserted that contemporary young adults’ greater mobility, relative to young adults in the early half of the twentieth century, had increased the likelihood of finding a partner of a different race. Unfortunately, because Rosenfeld relied on census data, it was not possible to determine why young adults changed their residence. Mobility involves changing residential location and breaking off ties from home, which may result in moving to regions that are more racially and ethnically diverse than their home states (Park 1928). However, young adults may also move to areas where they find themselves with a larger choice of partners of their own race or ethnic group. Migration may therefore exert different effects by race and ethnicity as well as gender. Finding that those who moved to a new state prior to the start of their current romantic relationship are more likely to be involved in an interracial relationship than those who did not move out of their state would support Rosenfeld’s (2007) assertion that mobility increases interracial unions (Hypothesis 4).

METHODS

Data are from Add Health (http://cpc.unc.edu/projects/addhealth), a nationally representative school-based study of adolescents in the United States. The first wave of data was collected in 1994–1995, when adolescents were 12–18 years old. The sample size for this wave was 20,745 students. Wave III was collected in 2001–2002, when respondents were between the ages of 18–28 years old at the time of interview; the sample size was 15,197. We used data from the adolescent in-home interviews for Waves I and III (response rates of 79% and 77.4%). In Wave III, 7,898 respondents reported current and most recent (if they had no current sexual partner) sexual partners at the time of interview; the remaining respondents reported no current or most recent sexual partners. We removed respondents who did not report their own or their romantic partner’s race (n = 119), who did not know the time their relationship started (n = 63), and those with missing sample weights (n = 401). We include multiracial respondents in our analysis but use the single racial category they most identify with in order to avoid ambiguity in defining the dependent variable of interracial relationships (Udry, Li, and Hendrickson-Smith 2003). We also omitted respondents who were Native American and Other races, due to small sample sizes (n = 114). We omitted respondents who had missing geocodes on the contextual variables for relative exogamous group size in Wave I (n = 68). Last, we removed respondents who had missing data on school enrollment (n = 79), parental financial assistance (n = 4), and residential status (n = 43) in Wave III. Our final analytic sample was 7,007 relationships. Our sample was approximately 46% of the original Wave III sample and 89% of the reported current and most recent sexual partnerships.

Dependent Variable

The dependent variable was a dichotomous measure indicating whether the respondent was in an interracial relationship or union in Wave III. If the respondent’s single-identified race or ethnicity reported in Wave I (Udry 2003) was not the same race or ethnicity of the romantic partner in Wave III, the respondent was designated as being in an interracial union or marriage.⁴

Main Independent Variables

Our primary independent variables are measures of maternal closeness and parental control, the relative exogamous group size of the respondent at the time of their initial interview, and whether the respondent moved prior to starting their current or most recent romantic relationship. Maternal closeness and parental control were measured at Wave I, as was our indicator of the relative exogamous group size of the respondent. We created our measure of maternal closeness at the initial survey based on responses to the following questions: How close do you feel to [current residential mother]?; for those who didn’t have a current residential mother, respondents were asked: How close do you feel to [previous residential mother]?; and for those with no previous residential mother: How close do you feel to your biological mother? Answers were reverse-coded and ranged from not close at all (1) to extremely close (5). We chose to focus on maternal closeness because 45% of the sample did not report a family structure of two biological parents in Wave I.

Parental control was a scaled measure of responses to five yes-no items asked at Wave I: whether parents allowed the respondent to choose friends, clothes, TV programs, how much TV to watch, and which food to eat (Cronbach’s alpha = .69 for men and .68 for women). Responses were reverse-coded so that higher scores indicated greater levels of parental control. The final scaled measure ranged from 0 (no parental control) to 1 (more parental control). Although Wave III also contained a measure of maternal closeness, we do not include it in our analysis as it is measured after the start of the romantic relationship and is therefore endogenous. We also included a measure for relative exogamous group size in adolescence by calculating the percentage of the population in the respondent’s census tract who was not the same race or ethnicity as the respondent’s reported race and ethnicity in Wave I.⁵

To test Rosenfeld’s (2007) hypothesis that moving to a new state is associated with interracial union formation, we include a measure of postmobility relationship formation. This was measured from the following questions in Wave III: “Have you continuously lived in your current state since the last interview year?”, “In what year did you move to {STATE}?”, and “How old were you when you first became romantically or sexually involved with {initials}?” The year the respondent started the relationship with his or her current or most recent partner was calculated by adding the age of first involvement to the year of the respondent’s birth. Those who continuously lived in their state since the last interview were coded as 0. If the respondent started the relationship before moving to a new state, the respondent was coded as 0. If the respondent started the relationship after moving, then the respondent was coded as 1.

Control Variables

Control variables include gender, race and ethnicity, nativity of the respondent, proxies for social class (family structure and maternal education, both measured in Wave I), romantic relationship characteristics (the type of relationship, the age at which the respondent started the relationship, whether the relationship reported in Wave III was heterosexual, and if the respondent reported being in a previous interracial relationship in Wave I), and measures of independence in Wave III (being enrolled in school, being employed, receiving financial support from parents, and living outside of the parental home). The gender of the respondent was a dichotomous variable (1 = male). The race and ethnicity of the respondent and their romantic partner was categorized into four mutually exclusive groups: non-Hispanic white (reference category), non-Hispanic black, Hispanic, and non-Hispanic Asian using Add Health’s constructed race variable, which was based on the respondent’s identified single race and only available for Wave I (Udry et al. 2003). The race and ethnicity of the respondent was measured in Wave I, and the single-identified race and ethnicity of their romantic partners was measured in Wave III. The respondent’s nativity was measured in Wave I with the question, “Were you born a U.S. citizen?” Respondents not born in the United States were classified as first-generation immigrants. Respondents who were born in the United States but had parent(s) who were not born in the United States were categorized as second-generation Americans. Respondents who were born and had both parents born in the United States were classified as third-generation Americans (reference category).

To proxy for the social class of our respondents in adolescence, we included the respondent’s family structure and maternal educational attainment, both measured in adolescence. Family structure, or the family members living in the respondent’s household, was measured at the time of interview in Wave I, using Add Health’s constructed five-category family structure variable (Harris 2009). We collapsed this variable to a four-category variable. The first category included respondents with two biological parents (reference category), the second category was two parents (any combination of step, foster, or adoptive), the third category was single parent only, and the fourth category was other (grandparents, siblings, and other kin and nonkin support). The maternal education of the respondent was measured in Wave I with the question, “How far in school did she [resident mother] go?” It was coded as a categorical variable with the following groups: did not graduate from high school (reference category), high school graduate or GED, some college, completed college and more, and don’t know. We did not include a measure of household income in Wave I because it was only asked of parents who completed the Parent Questionnaire in Wave I, of which 76% of our analytic sample completed. In supplemental analyses, we did not find that household income in Wave I was significantly associated with interracial relationship and union formation in emerging adulthood.

We also explore whether the likelihood of being in an interracial relationship varies by the relationship type, the age at which the relationship started, the sexual minority status of the respondent, and if the respondent reported being in a previous interracial relationship. For the relationship type, we used Add Health’s four mutually exclusive categories of whether the respondent was married (reference category), cohabiting, in a sexual relationship, or pregnant with his or her romantic partner (Harris 2009). Because parental control may wane over the life course (Longmore et al. 2001), we created a variable based on how old the respondent was when they met their romantic or sexual partner. The age categories for Wave III were grouped into those 18 years and younger, 19–22 years old, and older than 22 years old to reflect timing in transitions in and out of high school and college. The sexual minority status variable was created by cross-tabulating the genders of the respondents and their romantic partners. Respondents and romantic partners with the same sex (male-male and female-female) were coded as being in a homosexual (sexual minority) relationship. We control for prior interracial relationship experience in order to address reverse causality in respondents’ underlying propensity to interact across racial lines and because we do not know respondents’ racial attitudes in adolescence. Our measure of whether the respondent had been in a previous interracial relationship is based on information obtained in Wave I on the reported race and ethnicity of their romantic partner in adolescence, and if it differed from the race and ethnicity of the respondent.

We include measures of independence from parents in emerging adulthood, including school enrollment, employment, parental financial support, and living outside of the parental home. School enrollment came from the question asked in Wave III, “Are you currently enrolled in school?” Responses ranged from 0 (no) to 1 (yes). The respondent’s employment status was from the question, “Are you currently working for pay at least 10 hours a week?” with responses from 0 (no) to 1 (yes). Parental financial assistance came from the yes-no question, “Has your biological mother given you any money or paid for anything significant for you during the past 12 months? Don’t include regular birthday or holiday gifts.” The same question was asked of the current residential mother, previous residential mother, biological father, current residential father, and previous residential father. To capture an overall measure of parental financial assistance, we coded the variable as 1 if the respondent received any money from a parent in the past year, and 0 if the respondent received no money from a parent in the past year. We control for the respondent’s place of residence at the time of interview in Wave III from the question, “Where do you live now? That is, where do you stay most often?” Responses included a parents’ home, another person’s home, your own place, group quarters, homeless, and other. From the responses, we collapsed the variable to two categories: 0 if the respondent lived in a parents’ home, or 1 if the respondent lived outside of the parental home.

Analysis Plan

This study examined the role of parental factors, postmobility relationship formation, and relative exogamous group size on interracial union formation in emerging adulthood. We started with descriptive analyses of our primary measures and control variables and then explored differences by gender, race, and ethnicity. Next, we explored the linkages between parental factors and postmobility relationship formation on the likelihood of being in an interracial relationship. Because maternal closeness and parental control could be correlated with postmobility relationship formation, we also explored whether these measures were highly correlated, which we did not find evidence for. We therefore utilized all three measures in our multivariate analyses.

Logistic regressions using Stata 14.1’s multiple imputation by chained equations command (MICE) were estimated to calculated coefficients and odds ratios of being in an interracial relationship based on predictors measured in Wave I. Missing data on measures and background characteristics from Wave I were imputed using Stata 14.1’s multiple imputation by chained equations (MICE) command (Royston and White 2011). We imputed missing variables that had over 5% of the analytic sample missing (Schafer 1999). The imputed cases included maternal closeness in Wave I (n = 634 missing cases), parental control (n = 780 missing cases), maternal education (n = 1,061 missing cases), and the age at which the respondent met his or her romantic partner (n = 998 missing cases). Respondents who were missing data on these variables were less likely to have two biological parents in adolescence, reported lower maternal closeness, were more likely to be first-generation immigrants or second-generation Americans, and had lower maternal education relative to the imputed sample, but differences in descriptive characteristics did not reach conventional levels of significance. When the missing cases were excluded from the analysis, the analytic results remained similar in terms of significant results for the primary predictors. The one exception was for the association of postmobility relationship formation and the likelihood of interracial union formation for Asian women, which was no longer significant when the missing data was not included. Twenty multiply imputed data sets were used (Graham, Olchowski, and Gilreath 2007).

Logistic regressions identified the strength and role of parental factors, relative exogamous group size, and postmobility relationship formation to examine patterns of homophily among different racial and ethnic groups. Model 1 (the reduced model) included only our primary independent variables (maternal closeness, parental control, postmobility relationship formation, and relative exogamous group size), while Model 2 (the full model) included all the control variables. Separate analyses were conducted estimating the likelihood of being in an interracial relationship in Wave III by gender, and by gender, race, and ethnicity. For analyses run separately by gender, race, and ethnicity, our indicators of relative exogamous group size were included. This was the percentage of respondents living in a non-white, nonblack, non-Hispanic, and non-Asian tract in Wave I for whites, blacks, Hispanics, and Asians. Survey weights were applied for all analyses to account for the complex sampling design of Add Health.

RESULTS

Table I shows descriptive statistics for our measures of maternal closeness and parental control, relative exogamous group size, postmobility relationship formation, and control variables for the analytic sample. Most respondents reported being very close to their mothers. Reported parental control during adolescence was low, meaning that on average, adolescents could make their own decisions regarding their friendships and aspects of their daily lives, including TV programs to watch and clothes to wear. Adolescents lived in census tracts that were predominantly white. Only 15% of respondents had started their romantic relationships after moving to a new state. At the time of their initial interview, approximately 10% of adolescents reported being in an interracial relationship, but by emerging adulthood, this share had nearly doubled, to 18% by the time of their Wave III interview.

Table I.

Summary Statistics for the Analytic Sample

Variable	Mean	SD	Range	α
Maternal closeness (W1)	4.50	0.02	1 to 5
Parental control scale (W1)	0.14	0.01	0 to 1	0.68
Relative Exogamous Group Size Measures
Avg. % living in a nonwhite census tract (W1)	0.19	0.02	0 to 1
Avg. % living in a nonblack tract (W1)	0.87	0.02	0 to 1
Avg. % living in a non-Hispanic tract (W1)	0.93	0.01	0 to 1
Avg. % living in a non-Asian tract (W1)	0.98	0.01	0 to 1
Postmobility relationship formation	0.15	0.01	0 to 1
Female	0.56	0.01	1 to 2
Race/ethnicity			1 to 4
NH white	0.71	0.03
NH black	0.14	0.02
Hispanic	0.12	0.02
NH Asian	0.03	0.01
Immigrant generation status			1 to 3
1st generation	0.05	0.01
2nd generation	0.26	0.01
3rd generation	0.69	0.01
Age of respondent at interview (W3)	22.32	0.12	18 to 28
Family structure (W1)			1 to 4
Two biological parents	0.55	0.01
Two parents (step or bio)	0.18	0.01
Single parent	0.22	0.01
Other family arrangement	0.05	0.00
Mother’s education (W1)			1 to 5
Less than HS	0.16	0.01
HS grad or GED	0.36	0.01
Some college	0.19	0.01
Completed college+	0.24	0.02
Don’t know	0.04	0.00
Relationship type			1 to 4
Marriage	0.26	0.01
Cohabitation	0.32	0.01
Sexual relationship	0.39	0.02
Pregnancy	0.04	0.00
Age when respondent met partner			1 to 3
<19 years old	0.50	0.01
19–22 years old	0.42	0.01
22+ years old	0.08	0.01
Homosexual relationship	0.02	0.00	0 to 1
Enrolled in school (W3)	0.34	0.01	0 to 1
Employed (W3)	0.72	0.01	0 to 1
Parental financial assistance (W3)	0.52	0.01	0 to 1
Lived out of the parental home (W3)	0.67	0.01	0 to 1
% in an interracial relationship (W1)	0.10	0.01	0 to 1
% in an interracial relationship (W3)	0.18	0.01	0 to 1
n	7,007

Interracial Union Formation by Race, Ethnicity, and Gender

Table II highlights how interracial union formation patterns vary by race, ethnicity, and gender in Wave III. Although the prevalence of interracial relationships increased from adolescence to emerging adulthood, racial and ethnic homogamy was the dominant pattern. White and black young adults were most likely to be in racially homogamous unions, as noted by the shaded boxes. However, gender disparities were evident. Black women had the lowest proportion in racially heterogamous unions (11%), followed by white men (12%), white women (16%), and black men (24%). Hispanic men and Asian women were more likely to be in interracial relationships (39% and 44%, respectively, relative to Hispanic women and Asian men (37% and 36%), but these differences were not significant.

Table II.

Distribution of Race and Ethnicity of Respondent and Partner, by Gender in Wave III

	Race/Ethnicity of Respondent
	NH White		NH Black		Hispanic		NH Asian
Wave III Race/ethnicity of partner	Males	Females	Males	Females	Males	Females	Males	Females
NH white	0.88	0.84	0.11	0.03	0.27	0.21	0.18	0.21
NH black	0.02	0.04	0.76	0.89	0.04	0.11	0.02	0.09
Hispanic	0.05	0.05	0.06	0.04	0.61	0.03	0.08	0.09
NH Asian	0.02	0.01	0.01	0.01	0.02	0.02	0.64	0.56
NH Native American	0.02	0.03	0.02	0.02	0.02	0.02	0.03	0.01
NH Other	0.01	0.02	0.04	0.02	0.04	0.02	0.05	0.04
In an interracial relationship	0.12	0.16	0.24a	0.11a	0.39b	0.37b	0.36c	0.44c
n	1,664	2,421	513	791	541	625	192	260

Gray boxes indicate racially homogamous unions.

Underlined coefficient represents significant differences by gender respective to racial and ethnic groups, at p < .05.

^aBlack-white differences

^bHispanic-white differences

^cAsian-white differences, within gender, significant at p < .05.

Main Predictors by Race, Ethnicity, and Gender

Table III shows descriptive statistics for our independent variables of interest by race, ethnicity, and gender. Gender differences emerged in men’s and women’s reports of maternal closeness. In Wave I, boys reported being closer to their mothers than girls, a finding that is apparent across all four racial and ethnic groups. Contrary to expectations, adolescent girls’ reports of parental control in adolescence did not differ significantly from the reports of adolescent boys. Relative exogamous group size in adolescence differed across race and ethnic groups. White respondents reported the lowest relative exogamous group size (9%), while Asians had the highest relative exogamous group size (78%). Postmobility relationship formation did not significantly vary by gender but was significantly lower among Hispanic respondents relative to white respondents.

Table III.

Descriptives of Means for Main Predictors by Race, Ethnicity, and Gender

	Wave III
	Men				Women
	NH White	NH Black	Hispanic	NH Asian	NH White	NH Black	Hispanic	NH Asian
Maternal closeness (W1)	4.56	4.70a	4.66	4.60	4.43	4.50	4.37	4.26
Parental control scale (W1)	0.13	0.17	0.20b	0.19	0.13	0.15	0.20b	0.18
Relative exogamous group size^ (W1)	0.09	0.48a	0.67b	0.78c	0.09	0.43a	0.67b	0.78c
Postmobility relationship formation (W3)	0.17	0.15	0.10b	0.14	0.17	0.10a	0.08b	0.20
n	1,664	513	541	192	2,421	791	625	260

Underlined value represents significant differences by gender respective to racial and ethnic groups, at p <.05.

^aBlack-white differences

^bHispanic-white differences

^cAsian-white differences, within gender, significant at p < .05.

^{^}Relative exogamous group size is defined as the percentage of respondents who are not the race or ethnicity of the respondent in the respondent’s census tract in Wave I.

Factors Shaping Interracial Union Formation by Gender

Results of our multivariate analysis exploring the associations between parental factors, relative exogamous group size, and the formation of relationships following migration on being in an interracial relationship or union in Wave III are presented in Table IV. In both sets of analyses, two models were estimated: the reduced model included only the main predictors, while the full model included all control variables. We subsequently include interaction terms for variables of interest (gender by maternal closeness and parental control, and race and ethnicity by maternal closeness and parental control), not shown in Table IV, before exploring gender-differentiated models to determine which measures differentially predict the likelihood of forming interracial relationships by gender.

Table IV.

Logistic Regressions of the Likelihood of Being in an Interracial Relationship in Wave III

	Reduced Model		Full Model
Variables	B	OR	B	OR
Maternal closeness (W1)	−0.13 (0.05)	0.88**	−0.09 (0.06)	0.91
Parental control scale (W1)	0.04 (0.17)	1.04	0.03 (0.18)	1.03
Postmobility relationship formation (W3)	0.11 (0.12)	1.12	0.16 (0.13)	1.17
Relative exogamous group sizea (W1)	0.33 (0.22)	1.39	−0.51 (0.30)	0.60
Gender			0.08 (0.10)	1.08
Race and ethnicity - NH White; NH Black			0.31 (0.22)	1.36
Hispanic			1.36 (0.17)	3.90***
NH Asian			1.48  (0.26)	4.39***
Immigrant generation status - 3rd gen; 1st gen			−0.36 (0.19)	0.70
2nd generation			0.11 (0.12)	1.12
Relationship type - Marriage; Cohabitation			0.70 (0.13)	2.01***
Currently dating			0.46 (0.15)	1.58**
Pregnancy			0.71 (0.22)	2.03**
Age met partner; <19 years old; 19–22 years old			0.06 (0.11)	1.06
22+ years old			−0.05 (0.16)	0.95
Homosexual relationship			0.25 (0.38)	1.28
In an interracial rel. (W1)			1.68 (0.10)	5.37***
Intercept	−1.01 (0.22)	0.36***	−2.24 (0.34)	0.11***
n	7,007		7,007

^*p < .05

^**p < .01

^***p < .001.

Standard errors in parentheses.

^aRefers to the percentage of nonwhites in the respondent’s census tract during adolescence.

Note: Family structure, maternal education, school enrollment, employment status, residing outside of the parental home, and parental financial support were included as controls for the full model but not significant.

Rosenfeld posited that parental influence no longer mattered in an era of increased independence, while postmobility relationship formation would elevate the likelihood of being in an interracial relationship. However, results from Table IV contradict Rosenfeld’s argument. Maternal closeness in adolescence in adolescence is negatively associated with forming an interracial union in emerging adulthood. Although the coefficients are in the expected direction for our other indicators, postmobility relationship formation by emerging adulthood does not elevate the risks of young adults being in an interracial relationship to conventional levels of significance. Including our other controls in the full model reduces the association between maternal closeness in Wave I and being in an interracial relationship to nonsignificance, though the coefficient is still negative. For our total sample, then, we do not find support for either Hypothesis 1a or 2a; neither maternal closeness nor parental control is associated with the likelihood of being in an interracial relationship. We also did not find evidence that relative exogamous group size (the proportion of nonwhite census tracts in adolescence) was associated with interracial union formation for the overall sample (Hypothesis 3). Finally, we did not find that starting a romantic relationship after moving to a new state was significantly associated with being in an interracial relationship (Hypothesis 4) for the overall sample. The results from our overall sample, with the exception of postmobility relationship formation, support Rosenfeld’s argument.

Do the associations of maternal closeness and parental control matter more for women than for men, as much of the existing literature would suggest (Hypotheses 1b and 2b)? Models that separately interact gender by maternal closeness and parental control yield no evidence that either measure exerts any greater influence on daughters’ likelihood of partnering across racial lines than they do for sons’ chances of forming an interracial relationship (results available upon request). In other words, daughters who reported high maternal closeness or very controlling mothers were no less (or more) likely to have entered interracial relationships than sons who report high levels of maternal closeness or parental control.

We also examined whether maternal closeness and parental control mattered more for racial and ethnic minority groups compared to whites (Hypotheses 1c and 2c). We interacted race and ethnicity by maternal closeness and parental control in the full model. Consistent with our expectations (Hypothesis 1c), we found that the interaction of being white and maternal closeness in adolescence was significantly associated with a decreased likelihood of being in an interracial relationship. Contrary to our expectations (Hypothesis 1a), we found that the interaction of being Hispanic and maternal closeness was associated with an increased likelihood of being in an interracial relationship (results available upon request). In other words, relative to white respondents, Hispanic respondents who were closer to their mothers in adolescence were more likely to be in an interracial relationship in emerging adulthood. We did not find that the interaction of race and ethnicity and parental control on interracial union formation differed significantly for whites compared to other racial and ethnic minority groups, so we could not confirm Hypothesis 2c for the full sample.

Other measures operate largely as expected. Hispanics and Asians are significantly more likely to be in interracial relationships than their non-Hispanic white counterparts, and those in less formal (cohabiting, dating, or pregnant) relationships are more likely to be in interracial unions than respondents who were in more formal (marital) relationships. We do not find that relationships formed at later ages are any more likely to be interracial.⁶ We also do not find that measures of independence, such as being enrolled in school, being employed, receiving financial assistance from parents, or living outside of the parental home, are associated with being in an interracial relationship. Perhaps the largest predictor of being in an interracial union in Wave III was having been in one in Wave I. In fact, those with previous experience having a partner of another racial or ethnic background are over five times more likely to be in an interracial relationship at Wave III than their counterparts who were not in interracial relationships as adolescents.

Racial and Ethnic Variations in Interracial Union Formation in Emerging Adulthood

Of course, Rosenfeld’s examination of the role of parental factors focused only on whites, and while we find support for his argument upon looking at our total sample, associations may vary by race, ethnicity, and gender. Table V presents regression results, run separately by race, ethnicity, and gender, for our four groups in order to assess the variation in the magnitude of parental factors, postmobility relationship formation, and relative exogamous group size. The results from our regression analysis limited to non-Hispanic Whites do not support Rosenfeld’s assertions. We found that among our cohort, maternal closeness was a strong predictor of being in an interracial relationship in emerging adulthood among white men. Maternal closeness in adolescence was associated with a 29% decreased likelihood of being in an interracial relationship (OR = 0.71) among white men, though the association did not reach conventional levels of significance among other groups. At the same time, we did not find evidence that parental control was significantly associated with interracial relationship formation for white men. Among white women, neither maternal closeness nor parental control was significantly associated with interracial union formation.

Table V.

Likelihood of Being in an Interracial Relationship for White and Black Respondents

	NH Whites				NH Blacks
	Men		Women		Men		Women
Wave III Variables	B	OR	B	OR	B	OR	B	OR
Maternal closeness (W1)	−0.34 (0.11)	0.71**	−0.14(0.08)	0.87	−0.13 (0.08)	0.88	−0.05(0.23)	0.95
Parental control scale (W1)	0.03 (0.49)	1.03	−0.05 (0.31)	0.95	−0.74 (0.60)	0.48	1.23 (0.56)	3.42*
Postmobility relationship formation	0.14 (0.28)	1.15	0.05 (0.19)	1.05	0.44 (0.50)	1.55	−0.11 (0.54)	0.90
Relative exogamous group size (W1)	2.12 (0.64)	8.33**	1.59 (0.59)	4.90**	2.23 (0.50)	9.30***	0.27 (0.62)	1.31
Immigrant generation status - 3rd gen; 1st gen	0.54 (1.05)	1.72	0.38 (0.67)	1.46	−0.07 (1.19)	0.93	1.22 (0.47)	3.39
2nd generation	0.88 (0.24)	2.41**	−0.15 (0.22)	0.86	0.14 (0.52)	1.15	0.26 (0.38)	1.30
Marriage; Cohab	0.35 (0.28)	1.42	0.99 (0.19)	2.69***	0.28 (0.63)	1.32	−0.55 (0.43)	0.58
Sexual relationship	−0.30 (0.32)	0.74	0.82 (0.25)	2.27**	0.25 (0.73)	1.28	–0.44 (0.48)	0.64
Pregnancy	−0.62 (0.73)	0.54	1.84 (0.37)	6.30***	−0.14 (0.99)	0.87	−0.90 (0.64)	0.41
Age met partner; <19 years old; 19–22 years old	0.23 (0.22)	1.26*	−0.07 (0.18)	0.93	0.18 (0.39)	1.20	0.40 (0.35)	1.49
22+ years old	−0.16 (0.37)	0.85	0.04 (0.27)	1.04	0.08 (0.58)	1.08	0.58 (0.39)	1.79
Homosexual relationship	0.75 (0.70)	2.12	−1.76 (0.93)	0.17	3.48 (1.13)	32.50**	2.56 (0.89)	12.90**
In an interracial rel. (W1)	1.49 (0.28)	4.44***	1.24 (0.18)	3.46***	1.67 (0.47)	5.31**	2.06 (0.44)	7.85***
Intercept	−0.62 (0.66)	0.54	−2.20 (0.40)	0.11	−1.86 (1.32)	0.16	−3.01 (1.36)	0.05*
n	1,664		2,421		513		791
	Hispanics				NH Asians
	Men		Women		Men		Women
Variables	B	OR	B	OR	B	OR	B	OR
Maternal closeness (W1)	0.12 (0.31)	1.13	0.12 (0.16)	1.13	0.45 (0.49)	1.57	0.20 (0.21)	0.82
Parental control scale (W1)	0.60 (0.63)	0.55	1.11 (0.55)	3.03*	0.12 (1.11)	1.13	0.14 (0.86)	1.15
Postmobility relationship formation	−0.44 (0.55)	0.64	−0.45 (0.43)	0.64	0.08 (0.94)	1.08	1.34 (0.58)	3.82*
Relative exogamous group size (W1)	2.30 (0.67)	9.97**	3.10 (0.86)	22.20***	2.81 (1.68)	16.60	1.40 (0.80)	4.01
Immigrant generation status - 3rd gen; 1st gen	−0.42 (0.44)	0.66	−0.57 (0.45)	0.57	−0.69 (0.80)	0.50	−0.14 (0.42)	0.87
2nd generation	−0.48 (0.46)	0.62	−0.29 (0.43)	0.75	−1.67 (0.71)	0.19*	0.31 (0.58)	1.36
Marriage; Cohab	1.83 (0.54)	6.23**	1.30 (0.43)	3.67**	1.64 (1.05)	5.16	−0.28 (0.59)	0.76
Sexual relationship	1.20 (0.51)	3.32*	1.32 (0.48)	3.74**	0.73 (1.15)	2.08	0.79 (0.55)	2.20
Pregnancy	1.34 (0.91)	3.82	−0.86 (1.12)	0.42	5.01 (1.55)	150**	−3.25 (1.20)	0.04*
Age met partner; <19 years old; 19–22 years old	0.12 (0.40)	1.13	0.29 (0.30)	1.34	−1.10 (0.75)	0.33	0.13 (0.49)	1.14
22+ years old	0.40 (0.49)	1.49	−0.44 (0.72)	0.64	−0.10 (1.08)	0.90	−1.59 (1.24)	0.20
Homosexual relationship	−0.64 (1.09)	0.53	2.10 (0.90)	8.17*	1.28 (2.02)	3.60	—	—
In an interracial rel. (W1)	2.12 (0.38)	8.33***	1.38 (0.31)	3.97***	0.81 (0.86)	2.25	1.14 (0.43)	3.13**
Intercept	−3.67 (1.58)	0.03*	−5.25 (1.37)	0.01***	−5.77 (4.00)	0.00	−1.33 (1.88)	0.26
n	541		625		192		259

^*p < .05

^**p < .01

^***p < .001.

Standard errors in parentheses. Family structure, maternal education, school enrollment, employment status, residing outside of the parental home, and parental financial support were included as controls but not significant for all groups except for white men, who were less likely to be in interracial relationships if they grew up with a single parent, and black women, who were more likely to be in interracial relationships if their mothers had completed <HS degree.

Standard errors in parentheses. Sample size varies for NH Asian women because none were in homosexual relationships. Family structure, maternal education, school enrollment, employment status, residing outside of the parental home, and parental financial support were included as controls but not significant except for Asian women. Asian women who grew up with two bio parents and had a mother who attended some college in adolescence were more likely to be in an interracial relationship, while those who received any parental financial support in emerging adulthood were less likely to be in an interracial relationship.

We do find that parental control plays a role in interracial union formation but only among black women and Latinas (Hypothesis 2c). Black women and young Latinas who reported experiencing the greatest levels of parental control in adolescence were over three times more likely to be in an interracial relationship at Wave III than black women and Latinas with parents who exerted less control, providing evidence for Vasquez-Tokos’s (2017) finding that Latinas who grew up with more controlling fathers were more likely to be in interracial relationships. We ran these models separately by immigrant generation status to see if parental control differed by generation status, due to literature on parent-child relations that documents how tensions may arise between first-generation immigrant parents and their second-generation American children (Foner and Dreby 2011). Relative to first-and third-generation Latinas, second-generation Latinas who had controlling parents were more likely to be in interracial relationships, but this was only significant at the p < .10 level. Among black women, parental control was significantly associated with interracial union formation but only among third-generation Americans.

On one important dimension that Rosenfeld did not explore, relative exogamous group size in adolescence, our results provide strong support for the role of parents. Parents, whether due to choice or circumstance, can determine whether their children live in neighborhoods with higher levels of racial and ethnic diversity in adolescence, which shapes their children’s partner market in emerging adulthood. In fact, we see that for most groups (except black women and Asians) relative exogamous group size elevated the odds of interracial relationship formation, providing support for Hypothesis 3. These relative exogamous group size effects are particularly large for white and Hispanic men and women, and black men. The results suggest that rather than being unimportant, parents continue to be salient in factors shaping the partner choices of their children.

For only one group—Asian women—did postmobility relationship formation, or moving to a new state prior to the start of the current or most recent romantic relationship, elevate the odds of being in an interracial relationship in emerging adulthood. We therefore find only limited support for Hypothesis 4. Our sample size for Asians in relationships is relatively small; this may indicate the selectiveness of these women, who are both willing to move and are engaged in relationships, when a sizable share of Asians is not engaged in a romantic relationship by emerging adulthood (Balistreri et al. 2015).

Few of our other measures exert uniform associations across all groups. Prior experience with being in an interracial relationship was the most salient across the most groups; with the exception of Asian men, all others who had been in an interracial relationship at Wave I were significantly more likely to be in an interracial relationship by Wave III. We also see some evidence of the “winnowing effect” (Blackwell and Lichter 2004) among white women and Hispanics, where being in less formalized relationships (cohabiting, dating, or pregnancy relationships) were also strongly associated with having a partner of a different race or ethnic background. Finally, none of our measures of independence from parents were significantly associated with interracial relationship formation, except for Asian women, who were less likely to be in an interracial relationship if they received parental financial assistance in Wave III.

CONCLUSION

We revisited Rosenfeld’s (2007) The Age of Independence by examining whether parental factors, postmobility relationship formation, and relative exogamous group size were associated with interracial relationship and union formation during emerging adulthood. We expanded Rosenfeld’s focus to include youth who were racial and ethnic minorities and retested Rosenfeld’s claim on whether parents mattered in interracial relationship formation for a cohort of individuals from 1994 to 2002. While we do find some evidence to support Rosenfeld’s claims, we also find that parents do indeed shape their children’s partner choices. Among white men, maternal closeness was a salient factor. For racial and ethnic minorities, parental selection of the neighborhood in which children grow up (relative exogamous group size) emerged as an important factor shaping the partner choices of young adults.

In general, we argue that three important lessons can be gleaned from our study. The first is that on some dimensions, parents continue to exert an influence on their children’s choice of romantic partners via closeness and control, though this is most evident for white men, black women, and Hispanic women. Second, parents’ decisions on where their families live shapes the relationship choices of their children in adolescence and increases their likelihood of being in earlier interracial relationships. Therefore, these factors also have longer-term impacts as adolescents transition into emerging adults. The third finding is that postmobility relationship formation matters but only minimally. In other words, rather than being autonomous and completely independent of their parents, emerging adults continue to be influenced by them.

Although our descriptive results revealed higher levels of reported maternal closeness among black, Hispanic, and Asian men than for white men, the only group for whom maternal closeness reached conventional levels of significance was white men. Studies of partner choice have found that despite a growing acceptance of unions that cross racial lines, racial status hierarchies persist (Feliciano, Robnett, and Komaie 2009). Why this persists, when men from other groups also exhibit traits desirable in partners such as high levels of educational attainment or occupational achievement, suggests the persistence of group positioning (Blumer 1958). Our results suggest that a way group positioning may operate is via parent-child ties. On the other hand, we found that parental control did not exert the expected effect, increasing the likelihood of interracial relationship formation among black women and Latina women.

Even as we found evidence that parents matter in interracial relationship for-mation, what matters even more is relative exogamous group size in adolescence and prior interracial relationship experience. In the face of constraints and circumstances, parents generally choose the neighborhoods in which their children grow up (Goyette et al. 2014), suggesting that they play important roles in shaping the receptivity of their offspring to partners from different racial or ethnic backgrounds. While previous studies have examined relative exogamous group size and interracial union formation in adolescence (Strully 2014; Vanhoutte and Hooghe 2012), this study provides an extension into emerging adulthood and confirms that relative exogamous group size in adolescence indeed matters. This finding held only among white men and women, black men, and Hispanic men and women. The potential mechanisms driving this association need further exploration and include decreased prejudice through greater contact with other racial and ethnic groups, thus affecting preferences (Allport et al. 1979).

We also explored whether respondents had moved prior to the start of their most recent or current relationship, and if this increased their likelihood of partnering across racial or ethnic lines. Previous literature has suggested that those who are geographically mobile are more likely to have diverse friendship networks and be in interracial relationships (Rosenfeld 2007; Vanhoutte and Hooghe 2012). We did not find evidence for postmobility relationship formation being associated with a greater likelihood of being in an interracial relationship among all respondents, but we did find that this association only reached conventional levels of significance for Asian women when differentiating our sample by race, ethnicity, and gender.

Our study was not free from limitations. First, we examine only one cohort of individuals from 1994 to 2002, so we cannot directly assess whether parental influence has declined over time. Also, our measures of parental influence were limited. Our maternal closeness measure does not capture maternal preferences; mothers may encourage their children to date and marry across racial and ethnic lines. Parental control may differ across the life course, and our measure does not address control over relationship decisions. Although we control for prior interracial relationship experience in adolescence, we are not able to directly measure respondents’ own racial and ethnic partner preferences or attitudes, leading to potential endogeneity bias. Our measure of postmobility relationship formation did not assess whether states that respondents moved to were more or less diverse than their home states, because we did not know which states respondents moved to. Finally, our focus on respondents from Wave III may be a limiting factor—we included those in current sexual relationships. This led to an increase in the reported percentage of respondents in interracial relationships compared to the Census data, which only looks at those in married and cohabiting relationships (Qian and Lichter 2007), as well as small sample sizes for Asians, who are less likely to be in romantic relationships during these stages (Balistreri et al. 2015). Relative to the Census data of 2000, our Add Health sample was less racially and ethnically diverse for the age cohort (U.S. Census Bureau 2018).

Although the prevalence and tolerance of interracial relationships has increased in the United States since anti-miscegenation laws were declared unconstitutional (Livingston and Brown 2017), our study shows that parents remain an important third-party influence that perpetuate racial and ethnic hierarchies in the United States. Given stark racial, ethnic, and gender differences in the prevalence of interracial relationships (Livingston and Brown 2017), this study provides a muchneeded intersectional lens to examine interracial relationship formation that Rosenfeld (2007) overlooked. Our results indicate that white parents continue to have power in maintaining the white-nonwhite color line in the United States, though this operates mainly through maternal closeness with adult children and parental choice of neighborhoods in adolescence. White families tend to live in majority-white neighborhoods (Goyette et al. 2014), which serves to limit children’s interactions with other racial and ethnic groups, thus maintaining social distance from minorities and perpetuating the white-nonwhite color line. Our findings contradict Rosenfeld’s argument that parents no longer have influence over their children’s romantic relationship choices, and calls into question the extent to which children are fully “independent” from their parents in emerging adulthood.