Abstract
Abstract: The literature on interracial families has examined social stigmas attached to interracial relationships but has not thoroughly documented whether crossing racial boundaries increases the risk of divorce. Using the 2002 National Survey of Family Growth (Cycle VI), we compare the likelihood of divorce for interracial couples to that of same‐race couples. Comparisons across marriage cohorts reveal that, overall, interracial couples have higher rates of divorce, particularly for those marrying during the late‐1980s. We also find race and gender variation. Compared to White/White couples, White female/Black male, and White female/Asian male marriages were more prone to divorce; meanwhile, those involving non‐White females and White males and Hispanics and non‐Hispanic persons had similar or lower risks of divorce.
Recent increases in the rate of interracial marriage point to increased social acceptance of these relationships (Joyner & Kao, 2005; Lee & Edmonston, 2005). As of 2000, nearly 6% of all married couples were interracial compared to fewer than 1% in 1970. However, a growing literature describing the challenges faced by interracial couples (e.g., Chito Childs, 2005; Dalmage, 2000; Killian, 2003; Lewis & Yancey, 1995; Root, 2001) suggests that crossing racial lines still violates enduring norms of who should and should not marry whom (Killian). Demographic evidence further supports this hypothesis. For example, Bramlett and Mosher (2002) found that 41% of interracial couples divorced by the 10th year of marriage compared to only 31% of same‐race couples. Their findings imply that, although entering an interracial marriage tends to carry less social stigma, these relationships are less likely to remain intact.
We investigated the relative marital stability of interracial and same‐race marriages. Although interracial union formation has garnered a large degree of scholarly interest (Joyner & Kao, 2005; Lee & Bean, 2004; Qian, 1997), only a few studies have explored whether these relationships are more vulnerable to divorce (Felmlee, Sprecher, & Bassin, 1990; Heaton, 2002; Kreider, 2000; Monahan, 1970; Price‐Bonham & Balswick, 1980). Prior research may have oversimplified this comparison by not attending to the specific racial‐ethnic characteristics of couples. The rise in interracial marriages by Asians and Hispanics has diversified the picture of the “typical interracial couple” that had previously been dominated by the experiences of Black/White couples (Lee & Edmonston, 2005). We address this gap by investigating the risk of marital disruption of interracial couples distinguished along the lines of race and gender (e.g., Black male/White female vs. Black female/White male) using Cycle VI of the National Survey of Family Growth (NSFG). The aim of the current investigation was to empirically test whether an increased risk of interracial divorce can be found across all types of interracial couples.
Background
Of all the ways individuals can form families across ascribed social boundaries, crossing racial lines is the most controversial within contemporary American society. Race stratifies physical, mental, and economic well‐being (Oliver & Shapiro, 1995; Williams & Collins, 1995) and is a primary component of individual and family identity (Porter & Washington, 1993; Zack, 1993). Families that cross racial lines have historically challenged the structure of racial hierarchies by demonstrating the possibilities of moving across and between racial lines in family formation and individual identity (Root, 2001). Keeping racial boundaries intact has historically been achieved through institutional means, such as laws banning cross‐racial mixing (Davis, 1991; Moran, 2001). Although such laws are now unconstitutional (see Loving v. Commonwealth of Virginia, 1967), interracial romance that crosses the Black/White divide is often still stigmatized as an inherently dysfunctional relationship, motivated by racial stereotypes of sexual virility or even psychological pathology (Foeman & Nance, 1999). Therefore, even as interracial marriage has increased in frequency (Lee & Edmonston, 2005), the notion of racial homogamy as normal and racial heterogamy as aberrant remains intact (Foeman & Nance; Killian, 2003).
Differences in the rates of divorce between interracial and same‐race couples lend insight into the ways “race” still matters in relationship stability. Rates of divorce among interracial couples compared to same‐race couples provide an indication of the consequence of violating norms of racial homogamy. The scant research on this topic finds a relatively higher propensity toward divorce but historical evidence on the stability of Black/White marriages is mixed. Monahan (1970), using data from Iowa vital statistics, showed that some Black/White unions were more stable than Black homogamous couples, although less stable than White/White marriages. Estimates that are based on the U.S. Census for a similar time frame showed less stability for Black/White marriages compared to both Black and White homogamous couples (Heer, 1974). More recently, the racial mixture of a couple has been used as a covariate in research focused on the causes of relationship and marital dissolution (e.g., Felmlee et al., 1990; Heaton, 2002). Heaton’s analysis using national survey data found that interracial marriages were 13% more likely than same‐race marriages to divorce, after controlling for social and demographic background characteristics. Using a similar list of controls, Kreider (2000) reported that interracial marriages tended to have shorter durations, but she concluded that factors such as age at marriage and educational level have more impact on whether a marriage dissolves than couple‐level racial dissimilarity.
Within an era of freer choice of partners, what explains why marriages that cross the color line appear more vulnerable to marital disruption? One explanation is that interracial marriage “selects on” persons most likely to divorce. Put another way, elevated divorce rates among interracial couples may occur because members of these couples are more likely to have characteristics other than race associated with a greater likelihood of marital dissolution. In the next section, we describe the sociodemographic factors associated with divorce and consider whether interracial couples are more likely to occupy statuses that are vulnerable to divorce.
“Nonracial” factors associated with divorce
Generally, attributes that point to other forms of instability have positive influences on the risk of divorce (Heaton, 2002). Over the past few decades, the rate of divorce (per 1,000 married women aged 15 years and older) rose from 14.9 in 1970—the year before the introduction of no‐fault divorce laws—to a peak of 22.6 in 1980 and then declined and stabilized to around 20.0 during the 1990s (U.S. Bureau of the Census, 2002). Despite this variation over time, certain attributes have consistently increased the risk of marital disruption. These can be grouped broadly into the following categories: (a) age/cohort‐specific influences, (b) premarital experiences, (c) socioeconomic resources, and (d) couple‐level characteristics.
Age/cohort‐specific influences Marriages occurring early in the life course are more likely to be of short duration than those contracted later in adulthood. For example, Bramlett and Mosher (2002) estimated that 48% of all first marriages initiated before the wife reached age 18 end in divorce by 10 years compared to 25% of first marriages begun when female spouses were at least 25 years old. (The 10‐year mark is a commonly used duration indicator in demographic research; see Bramlett & Mosher; Kreider, 2000, and other publications from the U.S. Census Bureau and the National Center for Health Statistics.) Of all predictors of divorce, age at marriage is the most well established and most consistently predictive, regardless of time period when the marriage began (Larson & Holman, 1994; Martin & Bumpass, 1989; Teachman, 2002; White, 1990).
Historically, intermarrying African Americans have tended to contract their marriages at younger ages than those marrying within race (Tucker & Mitchell‐Kernan, 1990). Joyner and Kao (2005) showed that the decline in the age at which individuals enter interracial romantic relationships reflects, to a large degree, the increased involvement of recent cohorts who matured in an era of greater tolerance of interracial relationships. We expected that part of the elevation of divorce rates for interracial marriage would be linked to lower age of entry compared to same‐race marriages. However, we further expected a countervailing influence of increased tolerance across cohorts that would narrow the gap between interracial and same‐race marriage dissolution rates over time.
Premarital experiences A second set of factors concerns experiences that predate the marriage by shaping spouses’ outlooks on marriage as a permanent bond. Experiencing a parental divorce as a young child or adolescent has a well‐established positive association with marital instability (Larson & Holman, 1994; McLanahan & Bumpass, 1988), reflecting differences in attitudes toward marriage between adult children of divorce and adults who grew up in intact families (Glenn & Kramer, 1987). Similarly, premarital family formation behaviors such as cohabiting prior to marriage (Bumpass & Sweet, 1989; Thomson & Colella, 1992) or having a child (Larson & Holman) also increase the risk of a marriage ending. According to U.S. Census estimates, cohabitations are more likely to be interracial than are marriages (Blackwell & Lichter, 2000; Harris & Ono, 2000). Kreider (2000) found a smaller gap in divorce rates between interracial and same‐race couples that did not cohabit than between those couples that did cohabit.
Socioeconomic resources Individual‐level socioeconomic status in the form of education or income at the time of marriage is consistently and negatively associated with the likelihood of marital disruption (Bumpass, Martin, & Sweet, 1991; Larson & Holman, 1994; White, 1990). It is unclear whether accounting for socioeconomic status will reduce the differences in divorce rates between interracially married and same‐race married couples. Interracial marriage tends to select on the most socioeconomically advantaged African Americans (Crowder & Tolnay, 2000) and Latinos (Lee & Edmonston, 2005); however, very few differences in education are apparent between intermarried and in‐married Asians (Lee & Edmonston). Whites, specifically White females, who intermarry tend to be less educated than those who marry other Whites (Qian, 1997).
Couple‐level characteristics Finally, couple‐level differences in age and ethnicity also are associated with increases in divorce (Felmlee et al., 1990; Heaton, 2002). These marriages potentially unite persons from disparate cultural backgrounds, differing interpersonal styles, and varying values attached to marriage and family. Therefore, maintaining that relationship may be more difficult; thus, these relationships may in turn be more vulnerable to divorce. Tucker and Mitchell‐Kernan (1990) also found that Blacks in interracial marriages were more likely to differ significantly in age from their spouses than Blacks in same‐race marriages.
Overall, are selection differences in interracial marriage of “divorce‐prone” individuals enough to account for their higher rates of divorce? We have noted that previous evidence for this explanation is mixed. However, prior work has not adequately accounted for the influence of specific racial and ethnic identifications either as a key characteristic of the members of the interracial couple or as a factor that stratifies the likelihood of divorce.
Race‐ethnicity, intermarriage, and divorce
One alternate explanation is that contrasts between all interracial couples and all same‐race couples obscure important differences in interracial partnering. The negative reactions to interracial couples from strangers and the diminished social support from family and friends generally characterize the experience of Black/White couples (Chito Childs, 2005; Dalmage, 2000). However, very little comparative work has been done investigating Hispanic/White or Asian/White couples. One recent study on interracial coupling and psychological distress showed that rates of psychological distress varied considerably across different race‐ethnic couple combinations. When compared to their same‐race counterparts, intermarried White females and Hispanics individuals reported higher rates of distress, but the same did not emerge for intermarried African Americans or White males (Bratter & Eschbach, 2006).
Additionally, comparing all interracial couples and all same‐race couples overlooks very important racial variation in divorce. Studies of racial differences in risk of divorce have largely focused on the comparison between Whites and Blacks, and to a lesser extent Latinos, although these focus mostly on Mexicans (e.g., Phillips & Sweeney, 2005). These studies reported that Black marriages were generally more vulnerable than White marriages (e.g., White, 1990), although Teachman (2002) found the gap in divorce rates has closed considerably over time, mostly because of an increase in divorce among Whites. First marriages involving Asian women demonstrate a slower baseline rate of disruption compared to Whites, Blacks, or Hispanics (Bramlett & Mosher, 2002). Although some analyses do address this by adjusting for the race of the respondent (e.g., Heaton, 2002), differences between interracial couples and same‐race couples essentially reflect comparisons to one racial group, usually Whites. The question remains if the same differential would emerge if another racial group were applied as a standard of comparison.
Methods
Sample
We used data from the 2002 NSFG, Cycle VI, a nationally representative sample of 7,643 women and 4,928 men ages 15 – 44 (National Center for Health Statistics, 2004). We restricted our analyses to respondents who were ever married and who had valid information on the race of their first spouse, as well as on other predictor and dependent variables. Our final analytical sample was 1,606 males and 4,070 females.
Measures
Racial and ethnic identity of respondents and spouses As is standard for many large‐scale social surveys, including the U.S. Census, the NSFG gathered information on the race of respondents and their spouses (coded as American Indian or Alaskan Native, Asian, Native Hawaiian, White, or Black)1 separately from Hispanic origin.2 Although viewing “Hispanic” identity as an ethnic as opposed to a racial category is currently debated among social scientists and policymakers (e.g., Rodriguez, 2000), our race‐ethnicity variables followed the essential logic of the U.S. Census (Office of Management and Budget, 1997). We first coded four major non‐Hispanic (NH) racial subpopulations (NH Whites, NH Blacks, NH Asians [and Pacific Islanders and Native Hawaiians], and NH American Indians [and Native Americans]) and then, instead of grouping all Hispanics into one “pan‐racial” category, we separated them as well by their racial background (Hispanic Whites, Hispanic Blacks, Hispanic Asians [and Pacific Islanders and Native Hawaiians], and Hispanic American Indians [and Native Americans]).
Although prior studies on interracial marriage (Qian, 1997) have grouped all respondents into one pan‐ethnic racial “Hispanic” category, other works have noted that crossing the Hispanic ethnic barrier is generally less contested than crossing an explicitly racial one because many Hispanics who engage in what may be considered interracial marriage by marrying non‐Hispanics are actually marrying within race (King & Bratter, 2007; Qian & Cobas, 2004). To exploit this level of complexity in our data, as well as to elaborate on how certain types of heterogamy may have graver consequences for relationship stability than others, we modeled where differences between spouses were specifically between different racial groups in addition to between persons of different ethnic (Hispanic/NH) backgrounds.
The NSFG also allowed for the reporting of multiple races for individual spouses and respondents. For respondents who indicated multiple races for their spouse, the interviewer then asked respondents to report the race that “best describes” that spouse (wording taken from the NSFG questionnaire). We coded multiracial respondents who refused to provide a “best” race according to the first race listed, as that response reflects their initial reaction to the question. We assessed whether the couple was interracial on the basis of the spouse’s “best” reported race, as this reflected the degree to which the respondent viewed racial difference existing within the couple. We dropped cases (141 males and 40 females) where the race of spouse was not provided. Our models included dichotomous indicators for multiple‐race husband and multiple‐race wife (1 =mixed race, 0 =otherwise).
Interracial marriage therefore included any marriage in which a difference exists in the racial or ethnic background of the respondent and the spouse. Same‐race marriages in our analyses occurred between two individuals of the same race who were either both NH or both Hispanic.
Marital outcomes For descriptive statistics on first marriage, we considered three possible outcomes at the end of the first decade after entry into the marital state. At the end of this period, marriages may be (a) divorced, (b) intact, or (c) censored, meaning that they are intact at the interview date, but the interview occurred before the 10‐year duration mark was reached. For the multivariate analyses, we modeled divorce versus remaining married over time, with the estimation procedure accounting for censoring among the observations.
We used 10 years as our primary exposure period because the average duration of American marriages that end in divorce is 8 years (Kreider, 2005). A 10‐year exposure period thus covers that exposure and goes slightly farther. We did not model a longer exposure period because of concerns about censoring. Although our modeling technique handles the statistical aspects of censoring, it does not address substantive issues. For the earliest cohorts, censoring was generated only by widowhood, an extremely rare occurrence for young married couples. But beginning with the 1992 marriage cohort, lack of exposure time becomes an increasing cause of censoring. We chose the 10‐year period to balance allowing enough time for divorce to occur against the limitations of our data.
Control variables
Other background variables representing predictors of marital disruption included: respondent’s age at marriage (coded in years), marriage cohort, whether the respondent or spouse had a child prior to the marriage, whether the respondent cohabitated with the spouse or another partner prior to marriage, and whether or not the respondent’s parents’ marriage was intact at [respondent’s] age 14. We also included information on respondent’s education, coded as a series of dichotomous variables indicating degree earned (less than high school, high school, some college, and college degree or higher) to account for socioeconomic differences in propensity toward divorce. We used nominal categories rather than a continuous measure because high school and college completion have distinct socioeconomic implications regardless of years taken to attain these credentials. Finally, we compared the ages at marriage of the respondent and spouse to create a measure of age homogamy (with levels of response indicating whether the couple violated the social norm of a 2‐year age difference, and in what direction), coded to reflect whether (a) the husband and wife were within 3 years of each other, (b) the husband was 3 or more years older, or (c) the wife was 3 or more years older.
Analytical strategy
We first generated descriptive statistics for Tables 1, 2, and 3 using the survey commands in SAS to adjust for the complex sampling design of the NSFG. We report unweighted N values so that the reader can see on how many observations a statistic is based; we report weighted percents that have been adjusted for the weighted and stratified nature of the data.
| Independent Variables | Males (n = 1,606) | Females (n = 4,070) | ||
|---|---|---|---|---|
| % | N | % | N | |
| Race/ethnic background | ||||
| Non‐Hispanic (NH) White | 69.3 | 848 | 71.1 | 2,429 |
| NH African American | 10.2 | 274 | 10.1 | 600 |
| NH Asian/Pacific Islander | 3.0 | 51 | 3.6 | 144 |
| Hispanic White | 13.5 | 340 | 12.6 | 735 |
| Other non‐White | 4.0 | 93 | 2.6 | 162 |
| Racial mixture of respondent/spouse | ||||
| Respondent mixed‐race | 3.0 | 48 | 2.7 | 136 |
| Husband/Wife mixed race | 0.9 | 15 | 1.8 | 67 |
| Age at first marriagea | 24.6 | 1,606 | 23.1 | 4,085 |
| Age differences in couples | ||||
| Respondent 3+ years older | 20.8 | 316 | 3.8 | 146 |
| Respondent 3+ years younger | 13.2 | 261 | 32.3 | 1,475 |
| Age difference within 3 years | 66.0 | 1,029 | 63.9 | 2,449 |
| Education of respondent | ||||
| Less than high school | 13.3 | 263 | 11.3 | 544 |
| High school | 35.3 | 564 | 30.7 | 1,218 |
| Some college | 24.9 | 404 | 30.8 | 1,232 |
| College degree or beyond | 26.5 | 375 | 27.3 | 1,076 |
| Couple’s premarital cohabitation | ||||
| Couple cohabited prior to marriage | 41.3 | 638 | 37.2 | 1,607 |
| Respondent cohabited not with/first spouse | 14.7 | 267 | 10.7 | 465 |
| Respondent did not cohabit before marriage | 44.0 | 701 | 52.1 | 1,998 |
| Couple’s premarital birth | ||||
| One/both spouses had premarital birth | 45.3 | 805 | 27.2 | 1,260 |
| Neither spouse had premarital birth | 54.7 | 801 | 72.8 | 2,810 |
| Family background | ||||
| Parents’ marriage intact until age 14 | 71.0 | 1,116 | 6.9 | 2,768 |
- Note. Percents (not n values) are weighted using “survey” commands in SAS to adjust for complex survey design effects.
- a Indicates mean age at first marriage.
- Source: 2002 National Survey of Family Growth Cycle VI Male and Female Respondent Files.
| (A) Respondent’s Background (Male Sample, n = 1,606) | |||||
|---|---|---|---|---|---|
| NH White | NH Black | NH Asian | H White | Other Non‐White | |
| Wife’s background | |||||
| NH White | 91.7 | 14.6 | 6.3 | 11.6 | 22.0 |
| NH Black | 0.6 | 77.3 | 0.0 | 0.4 | 3.8 |
| NH Asian/PI | 1.5 | 3.3 | 88.4 | 0.3 | 2.2 |
| H White | 4.6 | 2.8 | 0.9 | 72.5 | 10.6 |
| Other non‐White | 1.6 | 2.0 | 4.4 | 15.2 | 61.3 |
| Total N | 848 | 274 | 51 | 340 | 433 |
| (B) Respondent’s Background (Female Sample, n = 4,070) | |||||
|---|---|---|---|---|---|
| NH White | NH Black | NH Asian | H White | Other Non‐White | |
| Husband’s background | |||||
| NH White | 92.1 | 7.9 | 22.6 | 17.4 | 16.7 |
| NH Black | 1.4 | 85.1 | 3.0 | 1.4 | 11.3 |
| NH Asian/PI | 1.0 | 2.5 | 70.9 | 0.3 | 1.6 |
| H White | 4.0 | 1.1 | 1.8 | 67.7 | 12.7 |
| Other non‐White | 1.5 | 3.4 | 1.7 | 13.2 | 57.7 |
| Total N | 2,429 | 600 | 144 | 735 | 162 |
- Note. Percents (not n values) are weighted using “survey” commands in SAS to adjust for complex survey design effects. NH = non‐Hispanic; H = Hispanic; PI = Pacific Islander.
| Year of entry into marriage | Same Race | Interracial | ||||
|---|---|---|---|---|---|---|
| Divorced | Intact | Censored | Divorced | Intact | Censored | |
| Before 1980 | 51.1 | 47.6 | 1.2 | 46.3 | 53.7 | 0.0 |
| 1980 – 1984 | 39.6 | 59.9 | 0.5 | 40.2 | 59.4 | 0.3 |
| 1985 – 1989 | 35.6 | 64.0 | 0.4 | 55.0 | 42.3 | 2.6 |
| 1990 – 1994 | 25.6 | 35.9 | 38.5 | 34.0 | 27.0 | 39.0 |
| 1995 – 1999 | 13.2 | 0.0 | 86.8 | 20.7 | 0.0 | 79.3 |
| 2000+ | 4.0 | 0.0 | 96.0 | 6.0 | 0.0 | 94.0 |
- Note. Percents are weighted using “survey” commands in SAS to adjust for complex survey design effects.
The subsequent multivariate analysis employs event history modeling using a person‐year file (35,139 person‐years) to account for the changing likelihood of marital disruption over the marital life course and the censoring of recent marriages in our data. We modeled the likelihood of a divorce by the 10th year of marriage using a complementary log‐log model for continuous time processes through PROC GENMOD in SAS (Allison, 1995). We adjusted for the effects of the complex sampling design using weighted generalized estimation equations, which allowed us to produce robust variance estimates and appropriate significance tests. Cole (2001) described this method for application in SAS and specifically for PROC GENMOD. The coefficients of this model can be interpreted similar to those from a proportion hazards model, so that the exponentiated coefficient translates into the percent increase in the hazard of divorcing by 10 years given a one‐unit increase in the independent variable (Allison).
When cell sizes for combinations of race, ethnicity, and gender were too small for multivariate analysis, we created combined groups using the umbrella term “non‐White.” We include these marriages in the models for consistency, but we do not interpret the results because of the heterogeneous nature of this group.
Results
Sample characteristics
Table 1 shows the descriptive statistics (mean or the percent distribution, as appropriate) of the respondents by gender. The percentages are weighted to account for the complex sampling design of the NSFG and the N values represent the actual number of cases in the data. The distribution of race‐ethnicity is similar across both groups. As we specified earlier, we note the racial background of non‐Hispanics and Hispanics. The majority of males are NH Whites (69.3%), followed by Hispanic Whites (13.5%) and NH African Americans (10.2%). Females show the same pattern, as well as a sizable number of NH Asian or Pacific Islanders (n = 144; 3.6%) because of the overall larger sample size. A small minority of respondent reported multiple races (3.0% of men and 2.7% of women).
We also describe characteristics that have established links to divorce. The average age at first marriage was 25 for males and 23 for females. More than 60% of males and females had spouses that are within 3 years of their age. Both male and female respondents are equally well educated, with nearly identical percentages earning a college education or more (26.5% of men and 27.3% of women). We also found a similar percentage of male and female respondents reporting premarital cohabitation with their first spouse (41.3% of men and 37.2% of women). A smaller segment of males (14.7%) and females (10.7%) premaritally cohabited with someone other than their first spouse. Less than half of male respondents (45.3%) and only 27% of female respondents who reported that a birth (for either partner) preceded their first marriage; the majority of these births were conceived within the (eventual) marital relationship.
Patterns of intermarriage
To identify the intermarried couple combinations most well suited for analysis, Table 2 shows all combinations of race‐ethnic background of couples for male respondents (shown in Panel A) and female respondents (shown in Panel B). As documented extensively elsewhere, the majority of first interracial marriages involve one NH White husband or wife because of this racial‐ethnic group’s dominant numbers in the population (Blau & Schwartz, 1984; Lee & Edmonston, 2005). However, the vast majority of NH Whites marry within their group. Among NH White respondents, a small minority had spouses who were racially similar (i.e., identify as “White”), although ethnically Hispanic (4.6% of men and 4.0% of women). The remaining marriages for this group most prevalently involved NH Asian (1.5% of men and 1% of women) first spouses; only 1% of NH White wives and even fewer NH White husbands reported first spouses who were NH Black.
Among non‐White respondents, racial‐ethnic homogamy was also common: 77% of NH Black men, 88% of NH Asian men, and 73% of Hispanic Whites reported homogamous first marriages. Not surprisingly, the most common identities of interracial spouses were also members of the largest demographic group, NH Whites. We observe the same patterns in reports from female respondents, although gender differences in intermarriage are also apparent. Racial homogamy was relatively greater among Black female respondents than males (85.1% vs. 77.3%) and less common among NH Asian or Pacific Islander female respondents (71% vs. 88%) and Hispanic White females (67.7% vs. 72.5%) than among their male counterparts.
These data provide confirmation that considerable variability exists in patterns of crossing the racial divide for marriage by race‐ethnicity and gender. Once these relationships form, does violating norms of homogamy affect their overall stability and do similar patterns of racial‐ethnic and gender variation appear?
Patterns of divorce by marriage cohort
Table 3 shows the percent of marriages that had disrupted, remained intact, or became censored (by the death of a spouse or by remaining intact at the most recent observation, which was before 10 years had passed) by the end of their first decade. This information provides data across several marriage cohorts (listed in rows). Within each cohort, the number of years available to be observed in the data is roughly similar. To explore the impact of crossing racial lines on stability of a marriage, we stratified by whether the couple was interracially or married within race, labeled “same‐race married,” as we defined earlier.
Consistent with results from other national data (Kreider, 2000), the percentage of couples divorcing by Year 10 was elevated among interracial marriages compared to marriages in which couples are of the same race, particularly among later marriage cohorts. Among those marrying before 1980, the likelihood of divorce for same‐race marriages was actually slightly higher than for interracial marriages (51.1% vs. 46.3%). However, in all succeeding cohorts, the percentage of interracial couples divorcing by Year 10 exceeded that of same‐race couples. The biggest disparity occurred among marriages initiated between 1985 and 1989, where 55% of interracial marriages divorced by their 10th year compared to 35.6% of same‐race marriages. This increased risk relative to previous marriage cohorts is interesting in light of the substantial increase in the number of intermarriages during the 1980s. Although this trend in incidence is often cited as evidence of improving racial relationships, these data suggest that many of these marriages did not endure. Subsequently, although the gap narrows somewhat among the most recent cohorts, it remains substantial.
Multivariate analyses of divorce
Table 4 shows the results of event history models predicting the hazard of divorce by Year 10. We present exponentiated coefficients whose interpretation is similar to odds ratios. We first assessed baseline differences in the hazard of divorcing for different race‐ethnic combinations of husbands and wives (see Model I) and then adjusted for the influence of background characteristics (see Model II). We examined the hazard of divorce by Year 10 for the race‐ethnic couple combinations of marriages between NH Whites (from this point on, referred to simply as “Whites”) and other groups relative to the hazard of divorcing for White/White couples (the reference category). We also include covariates for homogamous marriages within the three largest minority racial/ethnic groups—NH Blacks, NH Asians, and Hispanic Whites—to compare the relative stability of interracial marriages with the stability of other non‐White homogamous marriages. We include indicators for the remaining non‐White/White intermarriages and non‐White/non‐White marriages, but we were unable to break these categories down further because of small sample sizes. Finally, we include controls for whether the respondent characterized himself, herself, or the spouse as racially mixed by reporting multiple races. The effects of all control variables in Model II are consistent with the findings in the previous literature, so we do not discuss them here.
| Predictor | Baseline Model | Full Model | ||||
|---|---|---|---|---|---|---|
| B | SE B | e B | B | SE B | e B | |
| Husband race/wife race | ||||||
| White/H White | .07 | .11 | 1.07 | −.05 | .16 | 0.95 |
| H White/White | .31 | .29 | 1.37 | .13 | .31 | 1.14 |
| Black/White | .81*** | .08 | 2.26 | .73*** | .05 | 2.08 |
| White/Black | −.46† | .27 | 0.67 | −.58** | .21 | 0.56 |
| Asian/White | .35 | .25 | 1.42 | .47*** | .02 | 1.60 |
| White/Asian | .06 | .04 | 1.06 | .04† | .02 | 1.04 |
| All other/White | .97*** | .02 | 2.63 | .51*** | .08 | 1.67 |
| White/All other | .99*** | .15 | 2.70 | .76*** | .07 | 2.14 |
| H White/H White | −.36*** | .10 | 0.70 | −.43*** | .09 | 0.65 |
| Black/Black | .35*** | .07 | 1.42 | .08 | .17 | 1.08 |
| Asian/Asian | −1.21* | .53 | 0.30 | −.80† | .42 | 0.45 |
| All other combinations | −.07 | .15 | 0.93 | −.34 | .21 | 0.71 |
| Racially mixed husband | −.66* | .26 | 0.52 | −.91*** | .15 | 0.40 |
| Racially mixed wife | .45*** | .05 | 1.57 | .41*** | .07 | 1.51 |
| Constant | −1.02 | .66 | ||||
| Log likelihood | −76670029.08 | −72884700.26 | ||||
| df | 35,110 | 34,611 | ||||
- Note. Controls adjusted in full model are marriage cohort, age at marriage, education of respondent, status of parents’ marriage when respondent was age 14, age differences between spouses, premarital cohabitation, premarital birth. H = Hispanic.
- † p < .10. *p < .05. **p < .001. ***p < .0001 (according to a two‐tailed test).
Although results from Table 3 showed an elevated likelihood of divorce across marriage cohorts for intermarried couples, the results from multivariate analyses showed that this experience was not shared across all combinations of intermarriage. The types of differences that create the greatest risk of divorce were race‐, ethnicity‐, and gender specific. Intermarriages that did not cross a racial barrier, which was the case for White/Hispanic White couples, had statistically similar likelihoods of divorcing as White/White marriages. Racial differences in marriage, on the other hand, correspond to higher divorce rates but mostly in marriages where the White spouse is female. NH Black husband/White wife marriages were twice as likely to divorce as White/White couples, and NH Asian husband/White wife couples were 59% more likely, according to Model II. Highlighting the role of gender in interracial dynamics, the reverse combinations actually showed a lower or similar risk of divorce. White husband/NH Black wife couples were 44% less likely to divorce than White/White couples, and White husband/NH Asian wife couples were only 4% more likely to divorce by Year 10. Notably, we also found strong gender differentials were the association of being racially mixed and marital disruption, as marriages involving racially mixed women were 51% more likely to disrupt than couples in which both spouses are monoracial, but couples involving racially mixed men were 60% less likely. The meanings of these findings are unclear, but it must be borne in mind that racially mixed persons are both racially and socially heterogeneous groups (Campbell & Eggerling‐Boeck, 2006). Therefore, any generalization of these patterns to the experience of the multiracial population is premature.
How do interracial marriages compare to homogamous marriages between non‐Whites? The effects for in‐married Hispanic Whites and NH Asians revealed, as prior research has documented (see Bramlett & Mosher, 2002) that these groups were less likely than Whites to divorce by the 10th year of marriage. In analyses not shown, we found, not surprisingly, that Hispanic White/White marriages were substantially more likely to disrupt than Hispanic White homogamous marriages. Similar analyses of NH Asian/White couples and NH Asian/NH Asian couples demonstrated that the former were more likely to divorce by their 10th year, even White male/NH Asian female marriages, which were only slightly more likely than White/White marriages to disrupt. Finally, we compared NH Black/White unions to NH Black/NH Black marriages, which showed a persistently higher risk of divorce than White/White unions. We found again that crossing the racial divide for marriages between NH Black men and White women coincided with even greater marital instability than NH Black homogamous marriages experience. Generally, non‐Whites who intermarry experienced less marital stability than their same‐race married coethnics.
Summary and discussion
The current study examined the likelihood of divorce among interracial couples compared to same‐race couples. We find that although interracial marriages overall are more vulnerable to divorce, this reflects the experience of some but not all couples. According to the adjusted models predicting divorce as of their 10th year of marriage, interracial marriages that are most vulnerable involve White females and non‐White males (with the exception of White females/Hispanic White males) relative to White/White couples. Conversely, White men/non‐White women couples show either very little or no differences in divorce rates; or, as in the case of White men and Black women, are substantially less likely than White/White couples to divorce by their 10th year.
These patterns raise several interesting questions about what drives higher divorce rates for certain interracial marriages. As posited earlier, some of the elevation in divorce rates is because of couples being selected into divorce‐prone states. Most notably, a substantial portion of the differential risk of divorce for Black male/White female couples is explained by adjusting for background factors. However, in some cases, we find that these background factors suppressed differences between the interracial couples and the reference group. These results support the position that interracial divorce is not driven by spouses for whom marriage is “more risky” in other aspects.
First, the degree of social distance between groups likely sets an important context for durability of these unions. Prohibitive norms against crossing the Black/White divide have a long history whose impact on social interaction and marital selection remain today (Moran, 2001); as such, Whites remain least likely to marry Blacks, whereas Blacks have the lowest rates of interracial marriage of any non‐White racial group (Lee & Edmonston, 2005). Our data show that these marriages, specifically those involving Black men and White women, have the highest likelihood of disruption of any White/non‐White marriages. By comparison, Asian and Hispanic groups, whose rates of intermarriage are higher and experience less social distance from Whites in other arenas (see Yancey, 2003), have interracial marriages whose risk of disruption is either not as high or comparable to White/White marriages. Marriages between Whites and Hispanic Whites are likely less often perceived as “interracial” and may not experience the accompanying social stigmas.
The influence of gender, specifically the elevated divorce rates for White female interracial marriages, highlights important aspects of what makes interracial marriage socially distinctive. We consider two explanations. First, differences in divorce probabilities may be linked to the differences in the incidence of intermarriage. Since the 1960s, NH Black men have married White women more often than NH Black women have married White men, which suggests that the intermarriage barriers for NH Black women are greater than those facing NH Black men. Therefore, NH Black women and White men who choose to intermarry may be selective of an especially high degree of commitment to their relationship that reduces the potential for divorce. This is beyond the scope of the data at hand but should be investigated in future research. However, this calculus does not explain the propensities toward divorcing for other types of intermarriages involving White females, which are more prevalent than those involving White males and the equivalent female coethnics.
A second explanation posits that the “problems” interracial couples face vary in severity for different race or gender groups. A substantial literature, mostly focusing on Black/White unions, has shown that these couples experience negative reactions from strangers and diminished support from family and friends (Chito Childs, 2005; Dalmage, 2000; Hill & Thomas, 2000; Killian, 2003). Our findings suggest that these issues may be more common for White female/non‐White male couples. Stigmas attached to interracial interaction are strongly gendered as well as racially specific. In particular, White females have been perceived as a threat to Black women’s marriage opportunities (Chito Childs; Dalmage). Furthermore, White mothers experience the stigma of being perceived as unqualified to raise and nurture their putatively non‐White offspring (Twine, 1999) because of their lack of experience in navigating American culture as a minority. We suggest that this provides a particularly “unwelcoming context” for interracial marriages involving White females, which may increase the likelihood of their disruption.
We also note a few limitations in our analysis. Unfortunately, these data do not provide information on relationships with anyone beyond the couple (e.g., extended family members, friends, and neighbors). Other studies have shown that NH Black/White couples tend to isolate themselves, engage in fewer leisure activities to buffer against negative experiences (Hibbler & Shinew, 2002), or move away from their home communities (Tucker & Mitchell‐Kernan, 1990). Future research on this topic needs to incorporate broader measures of social support to address the argument that higher rates of interracial divorce are in some ways linked to greater social isolation or diminished social support. An additional limitation is that our data did not allow us to capture the relative influence of other types of assortative mating, such as class or religion. Interracial couples tend to be homogamous on education (Qian, 1997), but this does not speak to differences in other forms of social affiliation. Within‐couple religious or social class differences may speak to different interpersonal styles that enhance the level of conflict and strain, beyond what can be directly linked to identity as an interracial couple.
Implications for practice
The current study provides a necessary empirical basis for future explorations into the lives of interracial couples. In response to the challenges faced by interracial couples, a small but growing literature has outlined useful interventions (Hill & Thomas, 2000; Killian, 2003; Rockquemore & Laszloffy, 2003). Recently, maintaining healthy marriages has become a public policy priority. The Healthy Marriage Initiative was crafted under the current presidential administration to “help couples develop the skills and knowledge to form and sustain healthy marriages” (Administration for Children and Families, 2007). Several policies have been designed to promote healthy communication among at‐risk families. Interestingly, many of these policies target the development of healthy families within racial communities and craft curricula that are relevant to these specific groups. However, none of these policies addresses “healthy marriage” for families that span racial communities. We suggest that this may represent a gap that, as our findings suggest, requires more attention as interracial marriage becomes more common.