Recent debates on welfare reform have increasingly centered on whether states should actively encourage marriage among poor, welfare-dependent women with children.¹ Female-headed families are considerably more likely than families headed by married couples to be poor and to receive public assistance.² Although encouraging marriage is an explicit goal of the 1996 welfare reform law (the Personal Responsibility and Work Opportunity Reconciliation Act), most states have focused instead on promoting work among welfare mothers, most of whom are unmarried, through Temporary Assistance for Needy Families (TANF). With reauthorization of welfare legislation looming, promoting marriage is viewed as one strategy to reduce poverty and welfare caseloads.

Unfortunately, surprisingly little is known about the marital histories of unwed mothers. In this article, we use nationally representative data from the 1995 National Survey of Family Growth (NSFG) to compare the marital behavior of women who have an out-of-wedlock birth—the targets of state TANF programs—with that of women who delay childbearing until marriage. Specifically, we ask two straightforward empirical questions: Among black, Hispanic and white women, how do marriage rates and patterns among women who have a nonmarital birth differ from those of women who remain childless before marriage; and to what extent do black, Hispanic and white women who marry after a nonmarital birth stay married? These are important questions in a welfare policy environment that increasingly views marriage as a panacea, especially for historically disadvantaged racial and ethnic minorities.

BACKGROUND

Since welfare reform passed in 1996, an estimated 7.8 million children have been born to unwed mothers; this total results from average increases of slightly more than 1% per year in nonmarital births throughout the 1990s in the United States.³ Nonmarital childbearing may be one of the most important impediments to the formation of lasting healthy marriages. Indeed, women who have nonmarital births are substantially less likely than other women to marry,⁴ and if they do, they are more likely to divorce.⁵ Understanding what motivates unwed mothers to marry is essential for informing public policies that promote two-parent families and marital stability.

As early as three decades ago,⁶ research and political discourse unveiled substantial racial and ethnic differences in family formation. For example, even though the birthrate among unmarried blacks dropped steadily after peaking in the late 1980s, the rate among unmarried whites rose in the 1970s and leveled off after peaking in the mid-1990s. By 1998, rates of childbearing were 73 births per 1,000 unmarried women among blacks and 38 births per 1,000 similar women among whites.⁷ In 1999, 22% of births to white women occurred outside of marriage, compared with 69% of births to black women and 42% of births to Hispanic women. An observed rise in the nonmarital fertility ratio (i.e., the proportion of all births that occur outside of marriage) among blacks can be explained almost completely by declines in marriage. Both the rise in the rate of nonmarital fertility and changing marital patterns, however, account for an increase in the nonmarital fertility ratio among whites.⁸

Black women are less likely than white women to marry and to stay married.⁹ The dramatic "retreat from marriage" in the United States appeared earliest (in the 1940s) and most prominently among blacks.¹⁰ Their drift away from marriage has been explained by black men's poor employment prospects,¹¹ by a greater economic independence (due to a longer history of workforce participation) among black women compared with women of other races,¹² by the receipt of welfare benefits,¹³ by cultural preferences against marriage¹⁴ and as a cultural adaptation to economic and social oppression.¹⁵ Although declines in marriage among whites accelerated in the 1960s, they never were as dramatic as the declines among blacks.

Marriage trends among Hispanics have been studied less often than trends among others. High overall marriage rates among Hispanic women (especially in the context of their low socioeconomic status) set them apart from other ethnicities,¹⁶as does their greater propensity to marry even when marriage markets are tight (i.e., when the number of potential partners is low relative to the number of unmarried women). At the same time, however, trends toward greater social and economic assimilation among Hispanic immigrants have fueled increases in divorce, as many have adopted mainstream American cultural patterns.¹⁷

The observation that marriage and nonmarital childbearing vary substantially across race and ethnicity suggests that analyses must acknowledge culturally distinct family formation processes. These cultural differences raise several unresolved questions about the marriageability of single mothers. Marriage rates may be unaffected by nonmarital childbearing among groups for whom such behavior is common or even normative and who attach little or no stigma to it. Among blacks, for example, the stigma associated with nonmarital childbearing is not strong enough to motivate marriage in the event of an out-of-wedlock pregnancy or birth; among Hispanics, on the other hand, nonmarital cohabitation is common, but pregnancies often lead to marriage among cohabiting couples.¹⁸

In contrast, the stigma of nonmarital childbearing is comparatively strong among whites. Stigma and social disapproval increase the emotional and financial costs of raising a child alone and may further burden unwed mothers by making them less-desirable marital partners. At the same time, since white women—whether they are unwed mothers or not—have better access to economically attractive men than minority women,¹⁹ they have a greater financial incentive to marry following a nonmarital birth.

We also expect disproportionately high divorce rates when women marry after having a child out of wedlock. When the new husband is not the biological father, the presence of a child may strain economic resources and be a source of conflict (leading, for example, to arguments over visitation rights or resource allocation within the household). A woman who marries after having a nonmarital birth also might be more likely than others to be poorly matched with her husband if the marriage was motivated by economic exigencies rather than the quality of the relationship or a strong sense of commitment.

DATA AND METHODS

The Sample

The sample for our analysis comes from Cycle 5 of the NSFG. This data set provides detailed, retrospective life-history information—e.g., family background, marital and nonmarital relationship histories, and fertility experiences—for 10,847 women aged 15-44 in 1995. The data are nationally representative of U.S. women of childbearing age at that time.²⁰After excluding women of other races and ethnicities, we created event-history files, based on person-years, for 10,441 black, Hispanic and non-Hispanic white women who were never-married at age 14. We used adjusted NSFG poststratification weights in all analyses to avoid sample inflation.

Among white women, who accounted for 74% of the sample, 11% had had a nonmarital first birth (6% before age 20 and 5% later), and another 20% had been pregnant before marriage but had not given birth (i.e., they miscarried or aborted that pregnancy). Twenty-one percent of white women had divorced after their first marriage.

Black women made up 14% of the sample. Of these women, 46% had had a nonmarital first birth (28% were younger than 20 at the time, and 18% were 20 or older), and an additional 15% had had a nonmarital pregnancy that did not end in a live birth. Twenty percent of black women in the sample had divorced after their first marriage.

Among Hispanic women, who made up 12% of the sample, 23% had had a nonmarital first birth (14% as a teenager and 9% as an older woman), and another 20% had conceived outside of marriage but had not carried to term. For 19% of Hispanic women, their first marriage had ended in divorce.

Throughout the article, the term "nonmarital birth" refers to a nonmarital first birth only. We chose to focus on nonmarital first births exclusively because they are the first observable fertility event after which we expect divergent marital patterns, or trajectories, between unwed mothers and women who remain childless until marriage. Although subsequent nonmarital fertility is of empirical and policy interest, it awaits further research.

Analytic Methods

Because our sample draws from the population of women who represent the full range of childbearing years at one point in time (i.e., at the 1995 survey), individual women's risk of marriage could correspond to different consequences of nonmarital childbearing in different historical periods, beginning in the 1970s. A simple comparison of the percentage distributions of nonmarital births to women of each race by the timing of the births (i.e., before 1970, 1970-1974, 1975-1979, 1980-1984, 1985-1989 and 1990- 1995) showed that the proportions of these births that occurred in the 1970s were slightly higher among blacks than among whites or Hispanics. However, the proportions of nonmarital births among blacks that occurred in the 1980s were nearly the same as, or lower than, the proportions among whites or Hispanics.

Although these small racial differences in nonmarital childbearing in each time period are unlikely to bias our comparison of marriage by race and ethnicity, we performed multivariate logistic analyses, which controlled for the occurrence and timing of a nonmarital birth among white, Hispanic and black women, to uncover any period effects that may have resulted in racial and ethnic differences in the transition to marriage after a nonmarital birth.

We designated two cohorts—women whose nonmarital birth occurred before 1985 and those whose nonmarital birth occurred in 1985 or later. Women in both of these cohorts were compared with all women who had not had a nonmarital birth; any attenuation of racial or ethnic effects created by adding the birth-year cohort terms to the models would indicate that an association between the birth's timing and the mother's race or ethnicity would explain racial or ethnic differences in marriage. Also, interaction terms tested for any differential effects of the year of the nonmarital birth by race or ethnicity. The results from these models are reported as odds ratios.

We then performed discrete-time event history analyses, which are not shown, using logistic regression techniques to obtain transition hazard rates. In turn, we used these hazard rates to generate life tables of transitions to marriage and, for the ever-married, of transitions to divorce. Our strategy approximates a Cox regression in a continuous-time framework,²¹ where time-dependent covariates are treated as step functions and where bias in the estimated effects of covariates is minimized by specifying a semiparametric form of the baseline hazard function. The data are right-censored at marriage, divorce or the survey date.

Following the method described by Guilkey and Rindfuss,²² we used these life-table estimates to derive the cumulative proportions of women who would marry by specific ages, for each race and ethnicity. These proportions were calculated for two main groups of women—those who had a nonmarital birth, whom we divided into women who did so as a teenager and those who did so when they were older; and those who avoided a nonmarital birth, including those who became pregnant nonmaritally but had a miscarriage or abortion, whom we analyzed separately. For all women who gave birth out of wedlock, we used the same methodology to generate the cumulative proportions marrying by the number of years since the birth.

Clearly, unwed mothers differ from other women on many characteristics, observed and unobserved, that could affect their likelihood of marrying. Thus, in our decision to compare the cumulative proportions marrying among women who avoided nonmarital childbearing and the subset who conceived nonmaritally but did not give birth, we assumed that these groups would share characteristics, since they are drawn from the same population (i.e., women who became pregnant before marriage). Differences in marriage rates between them would suggest that nonmarital childbearing has "causal" effects on the likelihood of marriage; Hotz and colleagues conducted a similar natural experiment to study the consequences of teenage childbearing.²³ Alternatively, if unwed mothers and women who conceived nonmaritally but did not give birth follow similar marriage trajectories, nonmarital childbearing might not have a causal effect, and unwed mothers' relatively low marriage rates might stem from other factors (i.e., factors that would bias their "selection" into groups more likely to bear children outside of marriage or to have poor marital prospects).

Finally, we used the same methodology to derive the cumulative proportions of ever-married women who made the transition to divorce. (Although racial and ethnic differences in the probability of divorce following separation may increase the importance of considering separations as well as divorces, we focus on formal divorce alone because it has legal ramifications for family interactions and financial obligations.) We calculated the proportions divorcing by specific ages for women who avoided nonmarital childbearing (and for the subset of these women who were pregnant nonmaritally), and estimated the proportions divorcing among women who gave birth out of wedlock as teenagers and among those who gave birth when they were older.

In addition, we estimated the same cumulative proportions for women who married within six months of having given birth out of wedlock, no matter how old they were at the time of the birth. Women who wed within six months of a nonmarital birth are the most likely to marry the child's biological father; these women, therefore, have a lower risk of divorce than other unwed mothers. Such couples may best represent what might be expected when so-called fragile families—the targets of current welfare reform efforts to promote marriage—are formed. Further, if the divorce patterns of women who conceived premaritally but did not give birth are similar to those of unmarried women who bore a child, then the high divorce rates associated with nonmarital childbearing can be attributed to other, unmeasured factors, and causal interpretations would be unwarranted.

The use of computer-assisted self-interviews in Cycle 5 of the NSFG reduced, but did not eliminate, the underreporting of abortion,²⁴ and we assumed that the underreporting of a nonmarital pregnancy that ended in miscarriage would be no greater than the underreporting of one that ended in abortion. If a married woman in our sample failed to report a premarital pregnancy that ended in an abortion or miscarriage, she would be included among those with no nonmarital birth, but not in the group of women who miscarried or aborted a nonmarital pregnancy. If women who did not report these events share characteristics with those who accurately reported them, their exclusion from the nonmaritally pregnant group will not bias our comparison between women who gave birth out of wedlock and those who conceived but did not give birth nonmaritally. On the other hand, their inclusion in the group with no nonmarital birth will bias our overall results in a conservative direction, because the marital transitions of that group of women will be downwardly biased.

In the unexpected case that women who did not report a nonmarital pregnancy are similar to unwed mothers, the exclusion of those who did not report the loss of a nonmarital pregnancy from the nonmaritally pregnant group could underestimate the selection effects. More important, however, any observed differences in marriage rates between unwed mothers and women who delay motherhood until marriage are likely to be conservative estimates.

RESULTS

Multivariate Analyses

We present the results of our regressions of the transition to marriage as odds ratios (i.e., the odds of being married compared with the odds of being unmarried). Estimates from the first model, which assessed the effects of the woman's age (i.e., the number of years since she was 14) and race and ethnicity only, indicate, not surprisingly, that Hispanic and black women have significantly lower odds of marrying than white women (odds ratios, 0.9 and 0.4, respectively—Table 1). Results from a model assessing the effects of the woman's age and the period of her nonmarital birth (i.e., birth-year cohort) suggest that nonmarital childbearing significantly lowers the odds of marriage, and that the effect has increased over time (0.5 for the earlier cohort and 0.4 for the later one).

Our third model shows, however, that nonmarital childbearing cannot fully explain marriage differences between blacks and whites. Black women still have significantly lower odds of marrying than white women (0.5) even when we control for nonmarital childbearing. However, nonmarital childbearing appears to explain completely the difference in the odds of marriage between Hispanics and whites (i.e., the odds ratio for Hispanics loses significance in this model).

The final model includes the interaction between race and nonmarital childbearing, to estimate whether the effects of nonmarital childbearing differ significantly by race and ethnicity. The results show that the effects of nonmarital childbearing are stronger for blacks and Hispanics than for whites (i.e., the odds ratios reflecting the interaction effects are below 1.0). The negative effects of a nonmarital birth on the likelihood of marriage have increased over time for whites; they have decreased for Hispanics; and they have not changed significantly for blacks. Thus, the racial difference in the effect of a nonmarital birth remains, but has generally become weaker over time. The implication is that nonmarital childbearing has had a growing negative effect on marriage among white women, and this change has been relatively greater than that for black women.

Life-Table Estimates of Marriage

The probabilities of marriage, or cumulative proportions of women ever married by specific ages (Table 2, page 289), for each racial and ethnic group suggest that nonmarital childbearing is associated with reduced rates of marriage, a finding that is consistent with the data in Table 1. Among whites, for example, women who had a nonmarital birth are significantly less likely to be married by age 40 (82%) than are women who conceived out of wedlock but did not give birth (85%) or who had no nonmarital birth (89%). However, selection may partly account for the lower marriage rates among women who gave birth out of wedlock, since the difference between those with a nonmarital birth and those with a nonmarital conception is relatively small.

Whereas the cumulative proportion ever married by age 40 among whites who had a nonmarital birth as a teenager was the same as that among women who avoided a nonmarital birth (89%), only 70% of women who had a nonmarital birth when they were 20 or older married by age 40. Teenage unwed mothers may simply have more time than older ones to find a spouse.

Among Hispanic women, only 62% of those who gave birth out of wedlock had ever married by age 40, compared with 93% of women who avoided nonmarital childbearing. The cumulative proportion ever marrying by age 40 among Hispanic women who miscarried or aborted a nonmarital pregnancy was 82%. Clearly, unmeasured factors cannot explain the large difference between the marriage trajectories of unwed mothers and women who had no nonmarital birth; the presence of the child or stigma surrounding a nonmarital birth appears to be a major impediment to marriage among Hispanics. Those who avoid nonmarital childbearing have a significantly higher risk of marriage than other women at almost all ages. However, Hispanic women who give birth outside of marriage are significantly less likely to marry by age 40 than are similar white women (62% vs. 82%; p<.01—not shown).

Our estimates for black women are consistent with findings from previous research, which has shown that overall, lower proportions of black women than of other racial groups ever marry.²⁵ Only 59% of black women who had a nonmarital birth married by age 40, compared with 76% of black women who avoided a nonmarital birth. Among those who lost or terminated a nonmarital pregnancy, 66% married by age 40. Thus, although nonmarital childbearing reduces the likelihood of marriage among black women, their overall low rates of marriage—even among childless women—suggest that other factors, such as cultural attitudes and values or the shortage of economically attractive men, also play a large role.

Teenage unwed mothers are more likely than older unwed mothers to wed by midlife. Because the teenage mothers in our sample have had more time, on average, to marry than the older mothers, it is important to compare these two groups by their time to marriage following the birth event, rather than by the time since age 14. Figure 1 shows the trajectory of first marriages among unwed mothers, beginning with the year following the birth, for whites, Hispanics and blacks. Except for whites, the marital trajectories do not differ appreciably between teenage unwed mothers and those who had a nonmarital birth when they were older. Among whites, however, unwed teenage mothers marry more quickly after an out-of-wedlock birth than unwed older mothers; older unwed mothers may find fewer potential mates than younger mothers. Still, roughly one-half of unwed white mothers have married by the time their child is five years old, compared with one-third or fewer of minority unwed mothers.

Interestingly, whites and Hispanics are more likely than blacks to marry in the first few years following a nonmarital birth. For example, by the third year afterward, approximately one-fifth of Hispanic women and one-third of white women have married, compared with about one in seven black women. Previous studies have found that a birth often precipitates marriage among white and Hispanic couples who cohabit.²⁶ Indeed, many of these white and Hispanic women and their children might have been living in families headed by unmarried couples, although we did not assess the extent of cohabitation in our analysis.

Life-Table Estimates of Divorce

While our findings call for greater understanding of the various incentives and barriers to marriage among unwed mothers—especially if marriage is to be viewed as a route out of poverty for single mothers—getting married is only part of the story. If the benefits of marriage promotion initiatives are to have lasting effects on women and children, staying married also is important. In Table 3, for women of each race, we compare the cumulative probabilities of divorce at three, five and 10 years following a first marriage among five groups of ever-married women according to their nonmarital fertility status. Having a child out of wedlock is associated with greater marital instability, at least among whites and Hispanics, even when the spouse is most likely to be biologically related to the child (i.e., among those who marry within six months of a nonmarital birth). Blacks who marry within their child's first six months of life have lower rates of divorce than others. (Fewer than 5% of black women, however, marry within a year of a nonmarital birth—see Figure 1.)

Only among Hispanics is the likelihood of divorce after three years the same for teenage and older unwed mothers who marry. Very early motherhood and marriage may be culturally more normative among Hispanics, which may reduce their risk of divorce. Moreover, although a younger age at marriage raises couples' risk of divorce, not all teenage unwed mothers marry at very young ages (see Figure 1).

Still, regardless of the timing of a first birth in the unwed mother's life course and her race or ethnicity, at least one-quarter of marriages among unwed mothers overall end within five years (not shown). Further, more than one-half of unwed black mothers who marry, and two-fifths of their white counterparts who do so, will divorce within 10 years.

Our results also suggest that the characteristics of unwed Hispanic and white women predispose them to higher rates of marital instability when they do marry. Indeed, women who miscarried or aborted a nonmarital pregnancy have cumulative 10-year divorce rates that are more similar to those of unwed mothers than to those of women who avoided nonmarital childbearing. One interpretation is that for Hispanics and whites, marital instability may be associated less with out-of-wedlock childbearing per se than with other, unobserved risk factors. Among black women, by contrast, those who avoided nonmarital childbearing have the same probabilities of divorce as those who lost or aborted a nonmarital pregnancy, and teenage unwed mothers who marry are more likely to divorce. Thus, among blacks, a nonmarital birth may destabilize a subsequent marriage, unless the marriage involves the child's biological father.

DISCUSSION

The debate over reauthorization of the 1996 welfare law has raised considerable interest in whether marriage can be a panacea for unwed mothers. Our study demonstrates that nonmarital childbearing shapes marital trajectories and patterns of marital dissolution. Thus, marriage promotion policies might begin by promoting lower rates of nonmarital childbearing.

Whether nonmarital childbearing per se is responsible for lower marriage rates and higher divorce rates is hard to know. On the one hand, nonmarital childbearing may restrict women's opportunities to marry, and even if unwed mothers eventually marry, an early out-of-wedlock birth may create conflict or problems (especially financial hardship) that lead to divorce. On the other hand, unwed mothers may have characteristics (observed and unobserved) that reduce their likelihood of marriage and increase their likelihood of divorce even in the absence of an out-of- wedlock birth.

Our study compared the marriage patterns of unwed mothers with those of presumably similar women who experienced a nonmarital pregnancy that ended in miscarriage or abortion. The observed lower marriage probabilities among unwed mothers are not explained by selection alone, because the marital trajectories of women who gave birth out of wedlock differed significantly from those of women who were pregnant nonmaritally but did not carry to term.

Patterns of divorce appear to be different, however, with nonmarital childbearing itself having less of an impact. The higher probabilities of divorce among unwed mothers appear to be largely due to selection—at least for whites and Hispanics, since cumulative proportions of divorce among women who miscarried or aborted a nonmarital pregnancy were similar to those among women who gave birth outside of marriage. Marriage counseling and relationship classes, which are being considered in some current proposals to promote marriage and have been incorporated into others,²⁷ may improve marriage chances for women whose personal characteristics undermine their ability to have stable intimate relationships. However, for blacks, personal characteristics appear to matter less for maintaining a marriage, so simply improving interpersonal or relationship skills may be less relevant than efforts to improve the economic foundation of marriage.

Our results indicate that nonmarital childbearing contributes to lower marriage rates, especially among black women. But out-of-wedlock childbearing is not the entire story. Black women who avoided nonmarital childbearing still had lower rates of marriage than white women who bore children out of wedlock. Further, the divorce rates of blacks who avoided a nonmarital birth and ultimately married are much higher than the rates for similar white women, while their divorce rates are the same as those of white women who gave birth out of wedlock.

Clearly, other factors help explain the racial gaps in marriage, including the tenuous economic position of many low-income minority women and their families.²⁸ For example, blacks are less likely than others to legitimate a premarital conception by marrying: In the early 1990s, only 4% of black women who conceived premaritally had married by the time they gave birth, compared with 19% and 18% of whites and Hispanics, respectively.²⁹ Thus, both the quality of male-female interrelationships and men's employment are critical factors in the marriage process;³⁰ policies to address these factors may help to close the racial gap in marriage.

We consider the current emphasis on marriage to be at once important and misplaced. It is important because marriage surely contributes to the economic and emotional well-being of many women and their children. For some women, however, it may also create new dependencies on men that are ultimately unhealthy for everyone involved—for example, in the case of abusive relationships.³¹ Yet the emphasis on marriage may be misplaced if earlier nonmarital childbearing militates against the success of initiatives that promote marriage, especially among teenagers.³² States have a limited track record with policies that promote marriage and staying married. Marriage promotion might best begin with expanded efforts to reduce nonmarital childbearing. Indeed, policies that aim to reduce nonmarital childbearing in the first place may have the indirect benefit of increasing the incidence and stability of marriage, as well as reducing the number of needy families in the United States.