Early Family Formation among White, Black and Mexican American Women

Abstract

Using data from Waves I and III of Add Health, we examine early family formation among 6,144 White, Black, and Mexican American women. Drawing on cultural and structural perspectives, we estimate models of the first and second family transitions (cohabitation, marriage, or childbearing) using discrete time multinomial logistic regression. Complex differences by race/ethnicity and generation are partially explained by differences in attitudes and values in adolescence and family SES; marriage values are especially important in first-generation Mexican women's early entry into marriage. Examination of sequential family transitions sheds light on race/ethnic differences in the meaning and consequences of early cohabitation and pre-union births.

The changing family formation patterns of American men and women have been a major focus of demographic research in recent decades. The rising ages of first marriage and first birth, the growth of cohabitation, and the de-coupling of marriage and childbearing have been identified as significant changes in the family system (Casper & Bianchi, 2002). Perhaps the most fundamental shift is the decline of marriage as the social institution guiding family formation. Marriage has not disappeared, but its timing has shifted upward, it is regarded as more optional, and its centrality to other aspects of family formation has waned (Bramlett & Mosher, 2002).

Within the United States, these family changes are widespread, but there is also diversity in family patterns. For example, despite the trend toward later ages of marriage and childbearing, a significant share of U.S. women continue to form families in their teens and early 20s (Martin et al., 2007). Further, although cohabitation has generally supplanted marriage as the dominant type of first co-residential union, cohabitation is linked to marriage and childbearing in ways that vary across social groups (Smock, 2000). One major source of diversity in the timing and nature of family formation is race and ethnicity. Early marriage continues to be more common among Hispanics, especially Mexican Americans, than it is among non-Hispanic Whites and Blacks (Landale & Oropesa, 2007). In addition, disadvantaged racial/ethnic groups, such as Mexican Americans and Blacks, are less likely than more advantaged groups to delay childbearing (Martin et al., 2007). Births to women in their teens and early 20s are often nonmarital, but there is also variation by race/ethnicity in the share of births outside of marriage, in rates of nonmarital childbearing, and in the role of cohabitation in nonmarital births (Bumpass & Lu, 2000; Ventura & Bachrach, 2000).

One major thrust of recent research is identifying the social forces underlying the rising ages of marriage and entry into motherhood; however, equally important is research on why some women continue to form families early in the life course. The present study addresses that question with an emphasis on differences among non-Hispanic White, non-Hispanic Black, and Mexican American women (for ease of presentation, we refer to the former two groups as Whites and Blacks in the remainder of the article). Using data from Waves I and III of the National Survey of Adolescent Health (Add Health), we examine the routes to family formation and the predictors of early family transitions with discrete-time event history models of competing risks. We analyze both the first family transition in young women's lives (marriage, cohabitation, parenthood) and the second transition, conditional on the first transition made.

Our study makes several contributions to the literature. First, in contrast to most recently published work, our analysis is based on data that extend beyond 1995. Second, we provide new information on Mexican Americans, a group that is increasingly important because of its rapid growth, its relatively early marriage and childbearing, and evidence of sharp declines in the “traditional” family patterns of Mexicans as they spend more time residing in the United States. Here, we are able to examine the recent behavior of three generations of Mexican American women (foreign born; native born of foreign parentage; native born of native parentage) in their initial years of family building. Third, using a life course perspective, we explicitly model the sequential family transitions in young women's lives. This approach is important because it allows us to assess whether the family pattern that follows an initial transition varies by race and ethnicity. Indeed, we show striking race/ethnic differences in marriage and entry into motherhood from cohabitation, and in second pre-union births among women whose first transition was a birth. A fourth contribution is the inclusion of a rich set of predictors measured in adolescence and early adulthood. Within a framework that considers both cultural and structural perspectives, our models link values and beliefs, structural characteristics of the family of origin, educational investments, and histories of education, work, and sexual relations to choices about early family formation.

Race/Ethnicity and Family Formation

Recent Studies

Numerous recent studies have addressed race/ethnic differences in marriage, cohabitation, and childbearing, but firm conclusions about the nature of current patterns are limited by a lack of consistency in the time periods and age groups examined, the family transitions analyzed, and how interrelated family transitions are incorporated (or not) into models. For example, Raley, Durden and Wildsmith (2004) and Wildsmith and Raley (2006) provide analyses of nonmarital fertility and marriage, respectively, based on data from the 1995 National Survey of Family Growth (NSFG). Using retrospective histories collected from women ages 15−44 in 1995, these studies provide valuable information about family formation among Mexican American, White, and Black women. However, because the transitions in the data span the period from the late 1960s through 1995, the conclusions may not apply to the current period. Other recent studies also cover periods that do not extend beyond 1995. For example, Glick, White, and Goldscheider (2006) analyze race/ethnic differences in childbearing and marriage from 1988−1994 and Lloyd (2006) studies transitions to first marriage among Latinas between 1979 and 1994.

Comparisons across studies also are limited by differences in approaches to examining interrelated family transitions. For example, research on race/ethnic differences in marriage frequently does not consider cohabitation (Lloyd, 2006; Raley et al, 2004) or premarital childbearing (Lloyd, 2006). Studies of race/ethnic differences in nonmarital fertility increasingly incorporate cohabitation as a covariate (e.g., Wildsmith & Raley, 2006), but generally do not examine group differences in entry into cohabitation or in fertility within cohabiting unions. Even studies that incorporate multiple transitions, such as marriage and childbearing, often do not consider related family statuses or events, such as cohabitation (Glick et al., 2006). Finally, research based on cross-sectional data is often limited to examination of current family statuses (e.g., currently single, cohabiting, or married) without consideration of the family transitions that preceded individuals’ statuses at a given point in time (Brown, Van Hook & Glick, 2008).

In general, the current literature on race/ethnic differences in family formation can be characterized as a patchwork of studies based on different time periods, age groups, and approaches to modeling linked family transitions. Importantly, there is a lack of research on the post-1995 period, although Schoen, Landale, and Daniels (2007) provide an exception. Their study, which used multistate life tables to summarize family transitions up to age 24 through 2001−02, showed that women's behavior continues to change rapidly. Almost three fourths of early first births are nonmarital, with marked variation by race/ethnicity (66% for Whites; 96% for Blacks; 72% for Mexican Americans). Cohabitation continues to rise, with 59% of women cohabiting by age 24. White women are more likely to cohabit (62%) than Blacks (51%) and Mexican Americans (50%) as well as to experience cohabitation as their only family transition. Yet, Blacks and Mexican Americans differ greatly from each other with respect to the relationships between cohabitation, motherhood and marriage. A comparison of the Schoen et al. findings to those of other studies indicates that since 1995, cohabitation has become more widespread, transitions from cohabitation to marriage have declined, and nonmarital fertility has continued to rise. The present study extends this recent portrait by focusing in depth on the patterns and precursors of women's first two family transitions and disaggregating Mexican American women by generational status.

Early Family Transitions: Patterns and Precursors

A central issue in the study of racial/ethnic variation in family formation is the meaning of race and ethnicity. Here, we regard race and ethnicity as social categories that have meaning because they structure individuals’ identities, experiences, and opportunities. Racial/ethnic identities indicate shared histories and cultural roots that may engender group differences in values and worldviews. For example, one theme that is widespread in studies of Hispanic families is the idea that Hispanics are characterized by familism, or a strong commitment to family life that is qualitatively distinct from that of non-Hispanic whites (Vega 1995). Familism is a multi-dimensional concept, but there is general agreement that a key element is values that emphasize family roles and obligations (Landale & Oropesa, 2007). Race and ethnicity are also linked to socioeconomic position and opportunities. Thus, explanations of racial/ethnic diversity in family patterns must consider both the structural positions of groups and their cultural orientations. These include both the macro-level contexts in which people live and the micro-level characteristics of individuals. Here, we focus on the individual-level manifestations of cultural orientations and structural positions.

We also draw on the life course perspective, which places emphasis on the timing and sequencing of transitions (Elder, 1998). Transitions have different meanings, precursors, and consequences depending on when they occur in the life course and whether they occur before or after other transitions. Consistent with this perspective, we focus on the timing of family transitions, the order of family transitions, and the implications of initial transitions for subsequent transitions. Along with cultural and structural perspectives on the meaning of race/ethnicity, the life course perspective guides our discussion of potential precursors of family formation patterns.

Values and beliefs

Contemporary views of culture are complex and extend far beyond the realm of self-consciously held values and beliefs. For example, Swidler (2001: 6) suggested that “we think of cultures as ‘tool kits’ or repertoires of meanings upon which people draw in constructing lines of action.” Nonetheless, values and beliefs constitute an important part of the worldviews that help people organize their lives. Thus, we expect them to play a role in race/ethnic differences in family formation. In particular, women's ideas about the desirability of early family transitions and their beliefs about their probable futures are likely to affect their choices about early cohabitation, marriage, and childbearing. For example, East (1998) showed that favorable attitudes toward early marriage and motherhood as well as traditional gender role ideologies underlie the early family transitions of Mexican Americans. In contrast, while marriage in general is highly valued among African Americans, they often believe that marriage should be postponed until financial stability has been achieved (Tucker, 2000). This combination of beliefs contributes to a relatively high level of acceptance of early childbearing in the absence of marriage (East, 1998), which should be positively related to premarital childbearing. In contrast, young women who expect to attend college are likely to delay both marriage and motherhood, but may form less committed relationships. Religiosity also shapes values and beliefs. Adolescents who are highly religious tend to hold conservative attitudes toward family formation, favoring early family formation through marriage (Pearce & Thornton, 2007).

Structural advantages associated with the family of origin

It is well established that race/ethnic differences in family formation are due partially to differences in SES (Wu, 1996). The SES of the origin family structures the aspirations, activities, and opportunities of individuals in their teens and early 20s. Family SES is related to adolescents’ success in school and their opportunities for postsecondary education. Since children from advantaged backgrounds are likely to pursue higher education, they typically delay marriage and childbearing (Oppenheimer, 1988).

SES also influences family formation through youth's experiences in the family of origin. For example, family poverty and single motherhood are linked, and women who grow up in a single-parent family are more likely to have teenage marriages, teenage births, and premarital births than other women (McLanahan & Bumpass, 1988; Raley & Wildsmith, 2006). Efforts to understand why family structure is associated with these outcomes have emphasized socioeconomic resources and the quality of parenting, especially supervision (Thomson, Hanson, & McLanahan, 1994). It is likely that family structure also influences women's ideas about acceptable family forms and their views of their futures (Amato & Booth, 1997). Further, ideas about life strategies may be shaped by other aspects of parental experience, such as the mother's age at first marriage and the marital context of her births. In addition, a positive family climate may discourage early family transitions because it promotes adolescent adjustment and social competence (Parke & Buriel, 1998).

Blacks’ and Mexican Americans’ relatively early entry into motherhood is consistent with their socioeconomic disadvantage (Martin et al, 2007), but studies of marriage illustrate a more complex pattern. Mexican Americans are considerably more likely to marry, to marry early, and to stay married than Blacks (Landale & Oropesa, 2007). These differences are linked to multiple factors, including the employment patterns of Mexican American and Black males, but they may be due partially to differences between the groups in the educational trajectories of young women.

Educational investments

As noted, one of the mechanisms through which SES affects family formation is educational investments. Positive school experiences in adolescence enhance educational achievement and occupational goals, thereby reducing the risk of early family formation. In addition, prior research has found that a sense of attachment to the school, a high level of academic engagement, and high grades are related negatively to early/nonmarital parenthood (Furstenberg, Brooks-Gunn, & Chase-Lansdale, 1989).

Postsecondary schooling and employment

Investments in postsecondary education are critical to the timing of family formation. Postsecondary education delays marriage and childbearing during the years in which a student is enrolled, but educational attainment, employment, and earnings are positively related to marriage among those who have completed their schooling (Cherlin, 2000). Mexican Americans and Blacks are less likely to complete high school and to extend their education beyond high school than are Whites. This is partially due to their inability to pay for college, but it is also due to earlier disengagement from education. Thus, Mexican Americans and Blacks may have less incentive to delay family formation than Whites. At the same time, Mexican Americans are less likely to complete high school than Blacks, and among high school graduates, they are less likely to attend and graduate from college (Sólorzano, Villalpando, & Oseguera, 2005). This may contribute to differences in family formation patterns.

Sexual activity

Early sexual involvement is associated with race, SES, and family structure (Hayes, 1987; Manning, Longmore, & Giordano, 2005). The relatively early age at first intercourse among Blacks compared to Mexicans and Whites is especially noteworthy (Abma et al., 2004). Because teens with higher levels of sexual activity are more enmeshed in romantic relationships and more likely to become pregnant, early sexual activity is a precursor to early family formation (Heaton & Miller, 1991). High levels of sexual involvement in early adulthood are also likely to lead to cohabitation, marriage, and childbearing.

Approach of the Present Study

Demographers often emphasize careful description of a phenomenon as a prerequisite to attempts to explain it (Bianchi & Casper, 2005). Similarly, others (e.g., Abbott, 1998; Goldthorpe, 2001) argue that social scientists often rush to explain phenomena that have not been carefully described. Consistent with this viewpoint, our analysis has both descriptive and analytic aims. We begin by providing up-to-date information on race/ethnic differences in women's first and second family transitions. We show striking differences among Whites, Blacks, and Mexican Americans (broken down by generation) that suggest that the meaning and long-term consequences of early transitions vary by race/ethnicity. These differences are then analyzed in multivariate models. First, we examine the roles of cultural orientations and structural positions in race/ethnic differences in early family transitions. This is a topic on which scholars continue to disagree. Edin and Kefalas (2005) conclude that among impoverished women, there are few differences by race/ethnicity in attitudes and views regarding marriage and childbearing. But others (e.g., Oropesa & Gorman, 2000) provide evidence that normative beliefs about marriage differ by race/ethnicity even after demographic, economic, and other variables are controlled. Second, we examine processes through which values and SES may influence family formation--educational investments, postsecondary schooling and employment, and early sexual activity. The predictors we include in our multivariate analyses are intertwined with race/ethnicity and may partially explain racial/ethnic variation in early family transitions. However, we do not expect our models to fully account for the observed group differences. Rather, our aim is to shed light on the processes that play a role in race/ethnic differences in early family transitions.

Method

Data

The Add Health study is based on a nationally representative sample of U.S. students in grades 7 through 12 in 1994. The data include three waves of in-home interviews, which were conducted in 1995 (Wave I), 1996 (Wave II), and 2001−02 (Wave III). The initial data collection also included in-school and parent questionnaires. The data for the present study are taken from Waves I and III. Of the 20,745 persons interviewed in 1995, 15,170 were interviewed during Wave III. At Wave III, they ranged in age from 18 to 28, with 99% in the age group 18 to 25. The present study is restricted to women because they are more likely than men to form families in this age range and because the quality of the Add Health data on fertility is better for women than for men (Schoen et al., 2007). We focus on Whites, Blacks and Mexican Americans, and use Wave I reported race and Hispanic origin to assign respondents to racial/ethnic groups.

Our restrictions on gender and race/ethnicity reduce the sample size to 6,843. An additional 401 persons were excluded because they did not have a valid sample weight, and 298 cases were lost because the first family formation event occurred prior to the Wave I interview. The present study is based on the remaining 6,144 females, who were on average age 15 at Wave I (mean = 15.30, s.d. = 1.78) and age 22 at Wave III (mean = 21.66, s.d. = 1.82). There are 584 Mexican Americans, 3,948 Whites, and 1,612 Blacks. Among the Mexican Americans, there are 96 first-generation women, 288 second-generation women, and 200 third-generation women.

Parts of our analysis are based on a person-month file that follows each woman from the Wave I interview until she has her first family formation event or is censored by the Wave III interview. The unit of analysis is the person month of observation and cohabitation, marriage, and childbearing are treated as competing routes of exit from the single, nonparent state. The analysis of the first family formation event is based on 350,128 person-month observations (225,351 for Whites, 94,113 for Blacks, and 30,664 for Mexican Americans). We also analyze the second family transition of women whose first event was cohabitation or premarital birth. There were not enough marriages to follow women who married to the next family formation event. Among Whites, we follow the experience of 1,558 cohabiters (23,443 person months) and 273 single noncohabiting mothers (5,829 person months). The figures for Blacks are 379 cohabiters (4,858 person months) and 412 single mothers (10,734 person months); among Mexican Americans there are 177 cohabiters (2,520 person months) and 86 single mothers (1,868 person months).

Measures

Family formation events

The Wave III questionnaire collected cohabitation and marriage histories, from which we determined the month and year of entry into first cohabitation and first marriage. Respondents also were asked about the dates of all births within a history of their romantic and sexual relationships. Using that history, in combination with a method that uses the household roster to correct for inadvertently omitted births, we determine the date of the first birth (see Schoen et al., 2007 for more on this procedure).

Generational status

We disaggregate Mexican Americans by generation. The first generation is the foreign born, the second generation is the native born of foreign parentage (one or both parents foreign born), and the third generation is the native born of native parentage.

Values and beliefs

Adolescents’ values and beliefs are measured with several questions asked in the Wave I interview. In an activity that involved keeping or rejecting cards, they indicated which of a series of things would happen in their ideal romantic relationship in the next year. We constructed two measures from these data, an indicator of whether they would become pregnant in their ideal relationship and an indicator of whether they would marry their partner. Women also were asked if they would consider having a child in the future as an unmarried person. These questions had only two response categories (0=no; 1=yes). In addition, women assessed the likelihood (on a 1 to 5 scale) that they would attend college and that they would marry by the age of 25. Each of these variables is included in the analysis to measure young women's views of their likely futures. Finally, we include a religiosity scale. The results of a factor analysis of four measures of religious involvement and the importance of religion indicated that three measures should comprise the scale: how often the respondent attends religious services, how important religion is to the respondent, and how often the respondent attends religious youth activities. Higher values on the additive scale indicate greater religiosity (α = .81).

Socioeconomic background (SES)

Maternal education and family income are included as measures of SES. Maternal education is coded into four categories: less than a high school degree, high school degree, some college, and a four-year college degree or higher. Our measure of income reflects the total income of the household at Wave I, as reported by the resident parent.

Family characteristics

Two attributes of the family histories of women's mothers are considered: age at first marriage and whether the mother gave birth to the woman in our study before her first marriage.¹ These variables were derived from information in the Wave I parent questionnaire. Family structure and family protective factors, both based on Wave I in-home interview data, are additional predictors. Our measure of family structure distinguishes four family types: families with two resident biological parents; other families with two resident parents; mother-only families; and all remaining families. We measure family protective factors with a scale based on three questions: “How much do you feel that: (1) people in your family understand you? (2) you and your family have fun together? (3) your family pays attention to you?” Responses ranged from 1 to 4 on each item and the items were summed to create a scale (α = .79).

Educational investments

Three measures of educational investments at Wave I are employed: grades, school engagement, and school adjustment. The measure of grades ranges from 0 to 4 and is an average of students’ reports of their most recent grades in math, English, science, and history. School engagement is an additive scale based on responses to two items (scored 0 to 4) that measured how often the adolescent had trouble concentrating in school or completing homework (α = .78). Our school adjustment scale is based on levels of agreement with five statements: “You feel close to people at your school,” “You feel like you are a part of your school,” “You are happy to be at your school,” “The teachers at your school treat you fairly,” and “You feel safe in your school.” Responses to these items, which ranged from 1 to 5, were summed (α = .76).

Postsecondary schooling and employment

Two time-varying variables reflect the respondent's education. The first measures whether the respondent was in school in each month in her history from Wave I forward. Because the Wave III questionnaire did not collect a complete education history, we relied on several Wave III questions and some assumptions to construct this measure. One set of questions asked about the highest degree and the date the degree was obtained. We assumed that women who earned postsecondary degrees were enrolled in school continuously during the period immediately prior to degree completion. We also used responses to questions about whether the respondent was in school at each wave. Although our time-varying indicator of school enrollment undoubtedly contains measurement error, the use of plausible assumptions allows us to make maximum use of the education data.

Women also reported all degrees received by Wave III. We used the dates of the degrees earned to determine, for each month in a woman's history, her highest degree. The categories we recognize are: none; high school or g.e.d.; associate; and bachelor or higher.

Our time-varying measure of employment status is based on questions asked in Wave III about work experience from 1995 through 2001. Respondents were asked whether they had worked for pay in each year and whether the job was full-time or part-time. Our measure distinguishes between no employment, part-time employment, and full-time employment. Because of the question format, employment status changes yearly in the person-month file.

Sexual activity

In Waves I and II, adolescents were asked to report about their romantic and sexual relationships in the 18 months prior to the survey. In Wave III, they were asked about all sexual relationships from the summer of 1995 until the Wave III interview. Using these data, we constructed a complete history of sexual relationships from Wave I forward. At each age, we measure the number of sexual relationships a woman had in the prior year of age; we also measure the duration in months of her most recent sexual relationship. Women who report that they had never had sex are given a value of zero on both of these time-varying variables.

Missing data and weights

We employ Bayesian procedures for the multiple imputation of missing data to avoid erroneous inferences that might result from the rejection of cases that are not missing completely-at-random (Schafer, 1997). Five imputations were made to generate values for missing data. Each of the five datasets was then analyzed with SUDAAN to generate the correct parameter estimates and standard errors, given the complex sampling design. The results were then combined to yield estimates, standard errors, and p-values that reflect uncertainty about missing data (Schafer, 1997). In all analyses, the data are weighted by the longitudinal weight appropriate for analyses based on Waves I and III. The Add Health team incorporated information about unit nonresponse at Wave III into the longitudinal weight (Chantala et al., 2004).

Results

Descriptive statistics

Table 1 provides descriptive statistics for White, Black, and Mexican American women (by generation). In each group, the majority of women made a family transition by Wave III; nonetheless, the modal category for all groups except third-generation Mexicans was “no family formation events.” For Whites, cohabitation was the most common first event: 39% of White women (70% of those with a first event) entered cohabitation before marrying or becoming a parent. Only 10% of White women (18% of those with a first event) married before cohabiting or becoming a parent and 7% (12% of those with a first event) started family formation with a birth.

Table 1.

Descriptive statistics by race/ethnicity^*

	Whites	Blacks	Mexicans- 1st gen	Mexicans- 2nd gen	Mexicans- 3rd gen
First family formation event (%)
No events	43.8	44.1	33.1	44.4	34.5
Marriage	9.9b	3.8a	23.8ab	14.0b	10.4b
Cohabitation	39.4b	25.3a	27.9a	33.4	37.4b
Birth	6.9b	26.8a	15.2a	8.2b	17.7ab
Values and beliefs, Wave 1
Pregnancy part of ideal relationship now (%)	8.4b	15.6a	22.9a	22.5a	11.9
Nonmarital birth acceptable in future (%)	22.1b	32.8a	29.5a	31.4a	28.4
Marriage part of ideal relationship now (%)	16.7b	28.6a	48.3ab	32.2a	24.0
Chances of marriage by age 25	3.4b	2.8a	3.6b	3.2ab	3.2b
Likelihood of college	4.3	4.2	3.6ab	3.8ab	4.0a
Religiosity	7.8b	8.9a	8.4	8.2	8.0b
SES, Wave 1
Mother's education %
< high school (reference)	11.1b	20.5a	81.5ab	68.6ab	28.1a
high school	44.9	46.1	11ab	22.1ab	41.9
some college	18.5	17.3	3.8ab	5ab	19.6
college graduate	25.5b	16.1a	3.7a	4.3ab	10.4a
Family income	55,070b	36,540a	35,340a	38,750a	44,080a
Family characteristics, Wave 1
Mother's age at first marriage	20.5	21.3	18.7ab	20.7	19.8b
Respondent born premaritally (%)	6.95b	36.6a	10.9b	11.6b	16.4ab
Family structure (%)
2 bio parents (reference)	61.6b	32.7a	62.7b	68.2b	53.4b
2 parents other	18.0b	12.4a	14.2	10.1	18.3
mother only	15.1b	43.2a	10.9b	14.3b	21.4b
other	5.3b	11.7a	12.2a	7.5	7.0
Family protective factors	11.2	11.3	11.7a	10.9	11.2
Educational investments, Wave 1
Self-reported grades	3.0b	2.68a	2.8a	2.7a	2.8a
School engagement	7.7b	8a	8.2a	7.6	7.2ab
School adjustment	18.6b	17.88a	19.4ab	17.5a	18.4
Time varying characteristics (at age 22)
In school (%)	54.1b	41.1a	33.1	36.9	34.1a
Highest degree (%)
none	4.9	8.1	10.2	11.9	29.4ab
high school	74.4	79.3	89.0	80.4	64.2
associate	8.7	5.9	0.8ab	3.9	6.5
bachelor or higher	12.0	6.7	0.0ab	3.8a	0.0ab
Employed (%)	91.9b	75.3a	78.4a	75.7a	86.4
Duration recent sexual relationship (months)	14.1	16.3	12.7	15.3	13.4
Number of sexual partners last year	1.1	1.0	.8	.9a	.9
N of individuals (6,144)	3,948	1,612	96	288	200

^*Unless otherwise indicated, means are presented.

^aSignificantly different from Whites

^bSignificantly different from Blacks

Among Black women, entry into parenthood was the most common route to family formation, but cohabitation was a close second. About 27% of Black women (48% of those with a first event) had a birth before cohabiting or marrying. Although marriage was extremely rare (4%), about 25% of Black women entered cohabitation prior to marrying or becoming a parent.

First- and third-generation Mexican women were more likely to have a first family transition than Whites, Blacks, or second-generation Mexicans. Although half of first- and third-generation Mexicans formed a cohabiting or marital union, the first generation was much more likely to enter marriage as a first event (24%) than the third generation (10%). Moreover, first- and third-generation Mexicans were similar with respect to the likelihood of forming families through a birth (15% and 18%, respectively). Interestingly, second-generation Mexican women are the most similar to White women in the distribution of the type of event entered.

There are striking group differences in values and beliefs in adolescence. At Wave I, all groups except third-generation Mexicans were significantly more likely than Whites to say that marriage was part of their ideal relationship in the next year. Only 17% of Whites included marriage in their ideal relationship, compared to 48% of first-generation Mexicans, 32% of second-generation Mexicans, and 24% of third-generation Mexicans. Blacks (29%) were also more likely than Whites to include marriage in their ideal relationship even though they were significantly less likely to expect to marry by age 25. The pattern for pregnancy is similar to that for marriage: Only 8% of White adolescents considered pregnancy to be part of their ideal relationship in the next year, compared to about 23% of first- and second-generation Mexicans, 16% of Blacks, and 12% of third-generation Mexicans. All groups were more likely than Whites to be accepting of having a nonmarital birth in the future, but the difference between third-generation Mexicans and Whites was not statistically significant.

There are also large differences across the groups in maternal education. Although only 11% and 20%, respectively, of White and Black women had mothers with less than a high school education, fully 82% of first-generation Mexicans, 69% of second-generation Mexicans, and 28% of third-generation Mexicans had mothers who did not complete a high school degree. Still, family income was roughly comparable for Mexican Americans and Blacks, although somewhat higher for third-generation Mexicans. As expected, White women's families had the highest income.

Several aspects of experience in the origin family are distinct for young Black women. Fully 37% of Black respondents were born before their mother's first marriage, a figure 2−4 times higher than that for each other group. Further, only 33% of Black women lived with both biological parents during adolescence, compared with more than 60% of all other groups except third-generation Mexicans (53%). Also important is the striking increase across generations in mother-only living arrangements for Mexican Americans.

At age 22 (selected for illustrative purposes), Mexican Americans were the least likely to be in school (33−37%). They were also the most likely to have no degrees, especially third-generation women. About 29% of third-generation Mexican women had not completed any degree at age 22, compared to 5% (Whites) to 12% (second-generation Mexicans) for the other groups. Whites were the most likely to be in school and to be working, and had the highest educational attainment.

Models of the first family formation event

Table 2 presents models of the first family transition using pooled data for the racial/ethnic groups. Multinomial logistic regression is used because three competing routes of exit from the single, childless state are recognized: marriage, cohabitation, and a birth. Women contribute person months to the analysis until they experience a first family transition or are censored by the Wave III interview. Model 1 presents odds ratios that summarize the relative risks of each event among Blacks and Mexican Americans (by generation) compared to Whites. Model 2 adds measures of values and beliefs, whereas Model 3 instead adds measures of structural advantages associated with the family of origin. Inclusion of values/beliefs and structural advantages in separate steps allows us to gauge their relative importance in explaining race/ethnic differences. Model 4 includes all predictors. Although not shown, all models include time-varying dummy variables for single years of age to parameterize the baseline hazard.

Table 2.

Odds ratios from discrete time event history models of first family formation event

	Model 1			Model 2			Model 3			Model 4
	marriage	cohab	birth	marriage	cohab	birth	marriage	cohab	birth	marriage	cohab	birth
Demographic characteristics1
Race/ethnicity
White (reference)	.---	.---	.---	.---	.---	.---	.---	.---	.---	.---	.---	.---
Black	.35***	.62***	3.91***	.33***	.63***	3.49***	.35***	.50***	2.89***	.29***	.50***	2.21***
Mexican 1st generation	2.05**	.66	2.26	1.16	.58	1.52	1.22	.54	1.36	.94	.66	1.39
Mexican 2nd generation	1.32	.82	1.17	1.06	.76	.87	.93	.68	.74	.85	.77	.74
Mexican 3rd generation	1.30	1.16	3.02***	1.31	1.08	2.59***	.92	1.00	2.27***	1.18	.99	2.17**
Attitudes and values, Wave 1
Pregnancy part of ideal relationship now				1.34	1.10	1.75***				1.39	1.03	1.51**
Nonmarital birth acceptable				.83	1.41***	1.55***				.77	1.22***	1.39**
Marriage part of ideal relationship now				1.59**	1.01	1.12				1.39*	.97	1.02
Chances of marriage by age 25				1.43***	1.05	1.07				1.41***	1.07*	1.07*
Likelihood of college				.81***	.90***	.77***				.89*	1.06	.93
Religiosity				1.11***	.95***	.99				1.14***	.99	1.03
SES, Wave 1
Mother's education
< high school (reference)							.---	.---	.---	.---	.---	.---
high school							.80	.96	.77	.75	1.02	.94
some college							.86	.91	.49**	.90	1.03	.63*
college graduate							.65	.70**	.34***	.70	.89	.53**
Family income							.94*	.98	.97	.96*	.98	.97
Family characteristics, Wave 1
Mother's age at first marriage							.89***	.96**	.95**	.91***	.97*	.97
Respondent born premaritally							1.61	1.29*	1.11	1.47	1.24	1.05
Family structure
2 bio parents (reference)							.---	.---	.---	.---	.---	.---
2 parents other							1.25	1.70***	1.56***	1.39*	1.52***	1.35*
mother only							.78	1.57***	1.83***	1.02	1.47***	1.69***
other							1.11	1.70***	1.89***	1.30	1.43**	1.50*
Family Protective factors							1.02	.91***	.95**	.99	.95***	.99
Educational investments, Wave 1
Self-reported grades										1.05	.80***	.73***
School engagement										1.06	1.02	1.07*
School adjustment										.98	.97***	.97*
Time varying characteristics
In school										.56***	.56***	.41***
Highest degree
none (reference)										.---	.---	.---
high school or g.e.d.										1.46*	1.02	.99
associate										3.34***	.99	.62
bachelor or higher										2.44*	2.02***	.23*
Employment status
not employed (reference)										.---	.---	.---
part time										.63*	1.10	.69**
full time										.87	1.48***	.53***
Duration recent sexual relationship										1.02***	1.01***	1.02***
Number of sexual partners last year										1.07	1.20***	1.40***
N of person months	350,128			350,128			350,128			350,128

^*p<.05

^**p<.01

^***p<.001 two tailed tests

¹All models include a set of time varying dummy variables for single years of age

The findings in Model 1 show that the odds of marriage as a first event vary greatly across groups. The odds ratios range from .35 for Black women to 2.05 for first-generation Mexican women. Comparing first-generation Mexicans and Blacks, the odds of marriage for the former group are almost 6 times higher than for the latter (2.05/.35=5.86). Black women's low propensity to form co-residential unions is also evident in the pattern for cohabitation, where the odds ratio for Blacks relative to Whites is .62. Although Blacks and Mexican Americans differ greatly with respect to marriage, their behavior is more similar with regard to forming families through births. The odds ratio for first-generation Mexicans is 2.26 (p=.051), and those for third-generation Mexicans and Blacks are 3.02 and 3.91, respectively.²

In Model 2, which adds values and beliefs at Wave I, first-generation Mexican women no longer differ from Whites with respect to marriage. Clearly, their propensity to marry early (and to move directly into marriage) is related to their favorable attitudes toward early marriage. Other differences by race/ethnicity remain significant in Model 2, despite some attenuation of the odds ratios for births for Blacks and third-generation Mexicans. Importantly, the attitudinal variables exhibit strong relationships with family formation. Adolescents who view marriage as part of their ideal relationship at Wave I and those who expect to be married by age 25 are more likely than others to marry by Wave III. Religiosity is also positively related to marriage, but negatively associated with cohabitation. Further, young women who regard pregnancy as part of their ideal relationship are more likely than others to form a family through a birth, as are women who are accepting of nonmarital childbearing. Adolescents who believe it is highly likely that they will attend college are less likely than others to marry, to cohabit, or to have a child by Wave III.

The findings in Model 3 suggest that, in addition to values and beliefs, structural conditions in the family of origin underlie the pattern of early marriage among first-generation Mexicans. Controlling for SES and family characteristics (but not values and beliefs), first-generation Mexicans are no different from Whites with respect to early marriage. Taken together, the findings in Models 2 and 3 suggest that both cultural and structural differences play a role in the early marriage of foreign-born Mexican women. In Model 3, the odds ratios for forming a family through a birth are also smaller for Blacks and third-generation Mexicans than in Model 1.

Model 3 shows that maternal education beyond high school reduces the odds of cohabitation and births, and family income is negatively related to marriage. There is also strong evidence of the inter-generational transmission of family patterns. Women whose mothers married relatively early are more likely to form families through each route than other women. And all family structures other than the two-parent biological family propel women into early cohabitation and births.

In the full model (Model 4), the odds ratios for the race/ethnic groups remain similar to those in Models 2 and 3 except that the odds ratio for family formation through a birth is reduced substantially for Blacks. Net of all predictors, Black women are considerably less likely to form a family through marriage (odds ratio=.29) or cohabitation (odds ratio=.50) than White women, and considerably more likely to form a family through having a birth (odds ratio=2.21). Among Mexican Americans, only third-generation women differ from White women, and the difference lies in their much greater likelihood (odds ratio=2.17) of forming a family through a birth.

In the full model, the relationships between values/beliefs and family formation remain largely the same as in Model 2, as do those involving family characteristics. However, SES is somewhat more weakly related to family transitions. High grades and school adjustment reduce the odds of early cohabitation and births, and remaining in school reduces the likelihood of all early family transitions. At the same time, degree completion encourages marriage and cohabitation (for the latter, only college degrees). And as one would expect, longer sexual relationships and greater numbers of recent sexual partners increase early family formation.

Implications of the first event for the family life course

In some cases, the first family formation event is the only event or the defining event of the early family life course. In others, it is followed quickly by another transition that tempers or amplifies the consequences of the first. To examine race/ethnic differences in the consequences of the first event for the unfolding family life course, we examine the second family formation event among women whose first event was cohabitation or a birth (parallel analyses for women whose first event was a marriage were not possible because there were too few marriages).

Transitions out of cohabitation

The most common first family transition across all groups combined is cohabitation. Of women who had a first transition, 58% entered a cohabiting union. The top panel of Table 3 shows that most of those women moved out of the cohabiting nonparent state by Wave III. Among Whites (41%), Blacks (40%), and third-generation Mexicans (45%), the most common outcome was dissolution of the union. Having a birth was the most common second transition for first- and second-generation Mexican cohabiters (38% and 29%), but among first- generation Mexican women the percent marrying (36%) was almost as high as that for a birth.

Table 3.

Second family formation event by race/ethnicity

	Whites	Blacks	Mexicans- 1st gen	Mexicans- 2nd gen	Mexicans- 3rd gen
Second family formation event- women whose first event was cohabitation (%)
No second event- remained cohabiting nonparent	24.99	19.54	7.87	16.63	13.78a
Marriage	17.62b	8.67a	36.20b	22.32b	4.53a
Birth	16.14b	31.81a	38.02a	28.85a	36.41a
Dissolution	41.24	39.98	17.91	32.20	45.28
Second family formation event- women whose first event was birth (%)
No second event- remained single parent	24.31b	40.20a	33.67	29.45	46.07
Marriage	18.32b	6.74a	18.14	23.09b	5.96
Cohabitation	50.00b	30.72a	27.72a	46.62	45.13
Second Birth	7.36b	22.34a	20.46	.84ab	2.85b

^aSignificantly different from Whites

^bSignificantly different from Blacks

Table 4 (columns to the left) shows results from multivariate models of the second family transition for women who first cohabited. The monthly event history file follows cohabiters from the time they begin cohabiting until they marry, have a birth, dissolve the union, or are censored by the interview (whichever comes first). Because the patterns for race/ethnicity and other variables changed little across model specifications, only the full model is shown.

Table 4.

Odds ratios from discrete time event history models of second family formation event

	Second from cohabitation			Second from birth
	marriage	birth	dissolve	marriage	cohab	birth
Demographic characteristics1
Race/ethnicity
White (reference)	.---	.---	.---	.---	.---	.---
Black	.47*	1.97***	1.06	.32**	.52***	2.41**
Mexican 1st generation	3.81**	2.48***	.81	1.33	.60	2.36
Mexican 2nd generation	1.68	1.43	.96	3.20	1.73	.22
Mexican 3rd generation	.30	1.99**	1.37	.38	.99	.58
Attitudes and values, Wave 1
Pregnancy part of ideal relationship now	.67	1.07	1.20	.57	1.43	.86
Nonmarital birth acceptable	.86	1.16	1.11	1.26	1.06	1.19
Marriage part of ideal relationship now	1.60*	1.01	.74*	1.03	.77	.96
Chances of marriage by age 25	.99	1.03	1.02	1.40**	.97	1.13
Likelihood of college	.93	.96	1.08	1.03	1.11	.98
Religiosity	1.04	1.01	1.01	1.05	1.00	.98
SES, Wave 1
Mother's education
< high school (reference)	.---	.---	.---	.---	.---	.---
high school	1.41	1.24	1.11	2.08	1.00	.89
some college	1.92*	.98	1.48*	2.28	.96	1.28
college graduate	1.75	.88	1.53*	.89	.76	1.05
Family income	.97	.99	1.02	.97	1.00	.92
Family characteristics, Wave 1
Mother's age at first marriage	.98	.96	1.01	.98	.94	1.00
Respondent born premaritally	1.02	1.32	1.08	1.61	.94	.90
Family structure
2 bio parents (reference)	.---	.---	.---	.---	.---	.---
2 parents other	.72	1.12	1.14	.61	1.28	.98
mother only	1.01	1.53*	1.25	.64	1.05	.77
other	.91	1.40	1.08	.77	1.26	1.14
Family protective factors	1.02	.99	.98	1.14*	1.00	.99
Educational investments, Wave 1
Self-reported grades	1.17	.82	1.06	1.13	.74*	.58*
School engagement	.99	.94	.94*	1.00	.94	1.02
School adjustment	1.00	1.03	1.02	1.02	1.02	.99
Time varying characteristics
In school	.63**	.56***	.99	.56*	.61*	.40***
Highest degree
none (reference)	.---	.---	.---	.---	.---	.---
high school and above	1.55	1.05	.85	1.35	1.11	.57
Employment status
not employed (reference)	.---	.---	.---	.---	.---	.---
part time	.75	.51***	1.3	.49	1.03	.90
full time	.56*	.24***	1.21	.50	.80	.77
Duration recent sexual relationship	1.01	1.00	.99	1.00	1.00	1.01
Number of sexual partners last year	.99	1.18*	.97	1.05	.94	.98
N of person months	34,042			18,431

^*p<.05

^**p<.01

^***p<.001 two tailed tests

¹All models include a set of time varying dummy variables for single years of age

Among women whose first event was entry into cohabitation, first-generation Mexican women's strong proclivity toward marriage is again apparent. The odds ratio for marriage is almost 4 times higher for first-generation Mexicans than for Whites and more than 8 times higher than for Blacks, net of the full set of predictors. Another striking finding is that the risk of a birth as a second event is much lower for White cohabiters than it is for cohabiters in all other groups except second-generation Mexicans. The odds ratios range from 1.97 (for Blacks) to 2.48 (for first-generation Mexicans). This indicates clearly that cohabitation is more of a family-building state for Blacks and Mexican Americans than it is for Whites.

Net of other predictors, the attitudinal variables are largely unrelated to cohabiters’ second family transitions, with one exception. Women who indicated as adolescents that marriage was part of their ideal relationship are more likely to move from cohabitation to marriage and less likely to dissolve a cohabiting union. There are irregular patterns for maternal education, but in general higher levels of maternal education are positively related to both marriage and union dissolution. Also noteworthy is that being in school and working are associated with lower odds of marriage and childbirth among cohabiters.

Transitions out of single parenthood

Returning to Table 3, the bottom panel shows the second family transitions of women whose first transition was a birth. The possible transitions are entry into cohabitation, entry into marriage, and going on to a second pre-union birth. Among women who first had a birth, the most common second family transition for all race/ethnic groups is entry into cohabitation. At the same time, relatively high percentages of Blacks (22%) and first-generation Mexicans (20%) have a second pre-union birth. This pattern is far less common among Whites (7%), second-generation Mexicans (1%) and third-generation Mexicans (3%).

Table 4 (columns to the right) provides results from the full model of the second transition for women whose first transition was a birth. Women are followed from the time of their pre-union birth until they enter a cohabiting union, marry, have a second pre-union birth, or are censored by the interview. There are two key findings in this analysis. First, Black mothers are markedly less likely to enter any kind of co-residential union than are White mothers, even when an extensive set of other factors is controlled. Second, Black single mothers are substantially more likely than White single mothers to have a second pre-union birth. The odds ratio is 2.4 for Black mothers, relative to White mothers. First-generation Mexicans exhibit a similar pattern, but the difference is not statistically significant due to small cell sizes. Apart from race/ethnicity, few predictors are significant. Single mothers’ assessments in high school of their chances of early marriage are positively related to marriage. And remaining in school is negatively related to each transition.

Discussion

Using recent longitudinal data, our analysis provides new information on racial/ethnic variation in family formation in the teens and early 20s. Up-to-date information is especially important in the current era because of the rapid changes in family formation and the shifting racial/ethnic composition of the U.S. population, especially the growing share of Mexican Americans. Using data from the Add Health Study, we charted the course of family formation among three generations of Mexican-origin women as well as White and Black women.

For White women, the period before the mid-20s typically includes no family formation events or is limited to a cohabitation that does not progress to marriage or childbearing. In contrast, young Black women are less likely to cohabit than Whites and much more likely to begin family building with a birth. Moreover, Black women who do cohabit are more likely to have a birth as their next transition than are their White counterparts, and Blacks who begin family building with a birth are much more likely than Whites who do so to progress to a second pre-union birth. At the same time, similar percentages of Black and White women (44%) in our sample had no family formation events. Because studies of Black-White differences in family formation have focused on potentially problematic behaviors, little is known about differences in the outcomes associated with postponement for Blacks and Whites.

Mexican American women's early family formation behavior is different from that of Whites and Blacks and varies greatly by generation. First-generation Mexican women are the most likely to enter marriage as a first family event, a pattern that is linked to both values and beliefs supporting early marriage and to low SES. Further, among women who cohabit before marriage and childbirth, first-generation Mexicans are much more likely to marry their partner than Whites, Blacks, or later-generation Mexicans. Although support for marriage as an institution is part of Mexican culture, young women's behavior suggests that it weakens across generations: Second- and third-generation Mexicans are roughly similar to Whites with respect to the odds of marriage as a first event. Still, regardless of generation and across all models, Mexican Americans are significantly more likely to marry than Blacks (models not shown), a fact that is particularly noteworthy given that first- and second-generation Mexican women have dramatically lower maternal education and roughly similar family income compared to Blacks.

Although third-generation Mexican women are similar to Whites with respect to cohabitation and marriage, like Black women, they are highly likely to begin their family lives with a birth. Furthermore, like their counterparts in the first and second generations, third-generation Mexican women who cohabit before marrying or giving birth are highly likely to have a child within the cohabiting union. Some scholars have shown that unmarried motherhood brings psychological benefits to impoverished women (Edin & Kefalas, 2005), but the cycle of economic disadvantage that is perpetuated by childbearing among young unmarried women may override such benefits in the long run. The generational pattern in our data suggests that movement toward single motherhood (both within and outside of cohabitation) across generations may impede socioeconomic progress among Mexican Americans.

Our findings also underscore the complexity of the interplay between race/ethnicity, values and beliefs, SES, and early family formation. Values and beliefs about family formation that are expressed in adolescence clearly vary by race/ethnicity and play a key role in the distinct marriage behavior of foreign-born Mexican women. At the same time, race/ethnicity and SES are closely related and together shape young women's access to opportunities in adolescence and early adulthood. Low SES also plays a role the distinct marriage behavior of first-generation Mexican women; however, other racial/ethnic differences in family formation are reduced only modestly after SES is controlled. In addition, as has been found in prior studies, young Black and Mexican American women (as well as Mexican American women of various generations) make different choices about their early family lives, even though both Blacks and Mexican Americans are economically disadvantaged. Thus, our study finds support for both cultural and structural perspectives on racial/ethnic differences in early family formation.

One limitation of the Add Health data, like many other quantitative data sets, is that there are few measures of attitudes about family formation, especially how young women think about family choices and their relationships with other aspects of their lives. Recent qualitative research has contributed to a deeper understanding of how disadvantaged women think about motherhood and marriage (e.g., Edin & Kefalas, 2005), but we need to learn more about the attitudes, expectations and worldviews that underlie racial/ethnic and generational differences in the pathways to early family formation. Research on generational differences among Mexican American women would be especially useful in this regard.

Our understanding of early adulthood would also benefit from further study of racial/ethnic differences in the experiences of young women who delay family formation. Because the Add Health respondents had only reached their mid-twenties by the time of the Wave III interview, it was not possible to compare women who formed families early to women who formed families in their mid-to-late twenties. The extent to which groups differ in their later family formation patterns and in their ability to translate postponement of family responsibilities into meaningful development and career preparation is an important topic for future research.