In the past few generations, cohabitation, in which an unmarried couple involved in a relationship lives together, has become increasingly common in the United States. Almost half of American young adults report ever having cohabited,¹ and according to estimates from the U.S. Bureau of the Census, in 1998, more than four million households consisted of cohabiting couples.² Furthermore, cohabitation has become the most common route into marriage, with cohabitation preceding more than half of all first marriages during the early 1990s.³

Cohabitation is more commonly the choice of first union among blacks than among whites, and almost half of Puerto Ricans cohabit as their first union.⁴ The racial gap in union formation is diminished when both cohabitation and marriage are considered rather than just marriage.⁵ Many socioeconomic characteristics that influence the behavior of cohabitors, such as income and education,⁶ differ across racial and ethnic groups. This may support the contention that cohabitation occurs more among the disadvantaged, in part, as a response to economic constraints; however, racial and ethnic differences in union formation persist even after data are controlled for economic status.⁷

What is perhaps less understood about cohabitation is that it has increasingly become a setting for family formation. Cohabiting unions with children present are arguably one of the fastest-growing family forms in the United States. Fertility during cohabitation continues to account for almost all of the recent increases in nonmarital childbearing,⁸ and since 1990 there has been a 70% increase in the number of cohabiting households with children.⁹ Twelve percent of children born in the early 1990s were born to cohabiting mothers—a 100% increase from levels reported 10 years prior.¹⁰ Moreover, these trends are likely to continue: The exact estimates vary, but somewhere between one-quarter to two-fifths of children in the United States can expect to spend some time living in a cohabiting family.¹¹

Just as there are ethnic and racial differences in patterns of cohabitation, there are also ethnic and racial differences in childbearing within cohabitation. Children are more often present in minority cohabiting families than in white families. In 1990, 70% of Hispanic and two-thirds of black cohabiting households headed by individuals in their early 30s had children present, while only one-third of white cohabiting households did so.¹² Similarly, greater proportions of minority women gave birth during cohabitation than white women. In the early 1990s, one-10th of white children were born to cohabiting women, in contrast to nearly one-fifth of black (17%) and Hispanic (18%) children.¹³

Scant direct attention has been paid to the fertility behavior of cohabitors in the United States. The bulk of prior research examining childbearing in cohabiting families has been limited to comparing the fertility of cohabiting women with that of married women or single women living alone,¹⁴ to documenting bivariate differentials¹⁵ or to examining unions in other countries.¹⁶Most of this literature on cohabitation essentially has treated cohabitation as uniform across racial or ethnic groups.¹⁷ Little is known about the characteristics of Americans who have children in cohabiting unions and about the conditions under which they conceive and decide to have children.

This article describes the first study to explicitly assess racial and ethnic differences in fertility of cohabiting couples in the United States. The study has two central goals. The first is to determine whether there are racial and ethnic differences in the timing of childbearing within cohabiting unions, and if so, to examine the extent to which socioeconomic factors accounted for these differences. The second goal is to evaluate the general hypothesis that there are racial and ethnic differences in the extent to which cohabitation is considered an "acceptable" family form for childbearing. These goals were addressed by investigating which women give birth while cohabiting, using information on the reported timing status of conceptions and births during cohabitation.

Understanding the context of childbearing during cohabitation will provide insight into the potential implications of cohabitation for children's lives. This article draws on the most recently collected nationally representative fertility data, from the National Survey of Family Growth (NSFG), and builds on previous research by explicitly examining racial and ethnic differences in childbearing among cohabitors, including controls for key mediating factors such as socioeconomic circumstances. The pattern of racial and ethnic differences in fertility during cohabitation may indicate whether and how cohabitation functions differently for each group.¹⁸

Methodology

Data

The primary data source used for this analysis is Cycle 5 of the NSFG. These data were collected in 1995 and include the birth, pregnancy, marriage, cohabitation, employment and education histories of a nationally representative sample of 10,847 women of reproductive age (15-44). In addition, the NSFG contains information about the intention and planning status of all births. The NSFG data collected in 1995 for the first time also included complete cohabitation histories. No other data source has such high-quality data on both fertility behavior and cohabitation experiences.

The analytic sample was restricted in two ways. First, the analyses were restricted to women's fertility and cohabitation experiences prior to first marriage. Most women who cohabited (82%) did so before they had ever married, and the majority of children born in cohabiting unions (80%) were born to women who had never been married. The analyses in this article focus on first cohabitations because the vast majority of women who cohabited prior to marriage cohabited with only one partner. Second, the analytic sample was restricted to women cohabitating after 1980, who were less than age 30 when they started cohabiting. This restriction was necessary because of the upper age limit of the NSFG; women older than 35 in 1980 were not included in the 1995 interview because they were older than 44 years. The final sample included 2,716 women.

Data Analysis

The overriding hypothesis of this article is that cohabitation is a more acceptable context for childbearing for minority women than for white women. The acceptability of cohabitation for childbearing was measured in three manners: by the cohabiting women's childbearing behavior, by the odds of remaining in cohabiting unions once a pregnancy occurs and by views about the timing status of fertility during cohabitation.

Life tables are used to estimate the proportion of women who had had a child for each duration of cohabitation. Separate estimates are provided for the mother's race and ethnicity. Event-history models are employed to determine the timing of conception and of births within cohabitation. The event-history analyses are based on person-months; individuals either have a birth (or a conception resulting in live birth) or are censored (no longer observed) at the time of interview or at the termination of the union (marriage or separation). Finally, Cox proportional hazard techniques are used to estimate the multivariate models. An advantage of this estimation technique is that it does not require specifying a particular probability distribution.¹⁹ The proportionality assumption is less problematic for unions of short duration, such as cohabitation, than for longer unions such as marriage.

Since the central focus of this study was on decisions that lead to parenthood or family building, the analyses were restricted to live births. It should be noted that women who do not have children include women who have never become pregnant, those who have had a miscarriage and those who chose to have an abortion. Given the poor quality of data on abortion, it is not appropriate to separately analyze women who have had abortions; however, including them will most likely delay the timing of motherhood and perhaps be associated with a greater likelihood of having intended or planned births.

The timing of motherhood within cohabitation was measured at two time points: conception and birth. In the NSFG, respondents were asked directly about both the date of conception and the date of birth. It is important to present findings for the timing of both conceptions and births because decisions about marriage and about union dissolution determine whether a child is born within cohabitation. Thus, models estimated here show separately who became pregnant with a child during cohabitation and who gave birth to a child during cohabitation. In addition, logistic regression was used to estimate the odds that a woman's first child conceived during cohabitation is born into that cohabiting union. Pregnancies reported to end in abortion were not considered in these analyses; presumably, they represent women who had the lowest desires for childbearing during cohabitation.

Drawing on the above analytic sample, the analyses were further limited to 657 women who conceived a child during cohabitation and 561 women who gave birth to a child during cohabitation. Logistic regression models were used to evaluate the effects of covariates on the intention status of cohabiting births and conceptions.

Intention (or planning) status was derived from several questions from the NSFG. Women who discontinued using a contraceptive and who had not wanted to become pregnant were asked if they wanted to have a baby at some time. Those women who wanted a baby were asked the question, "Would you say you became pregnant too soon, at about the right time, or later than you wanted?" The women's pregnancy status was classified into two categories: unintended (too early or unwanted) and intended (right time, don't care or too late). Very few births were classified as being too late, and existing research typically combines them with intended births, since they are more likely to be like births that occurred "on time" than like those that occurred "too early." Several different terms are used in the literature to refer to unintended childbearing; this article uses the terms used in Best Intentions.²⁰

Two statistical models are presented in this article: a zero-order model with the primary independent variable, race and ethnicity, and a second, full model that includes all the other covariates. Log-likelihood ratio tests were used to establish whether the covariates added to the fit of the model, as well as whether the effects of race and ethnicity can be explained by the socioeconomic control variables.

For these analyses, race and ethnicity were separated into four groups: Hispanics, blacks, non-Hispanic whites and others. Unfortunately, Hispanic ethnic groups had to be combined for analysis, as small sample sizes did not allow them to be separated into subgroups such as Puerto Rican, Mexican American and Cuban. (Because only 54 Puerto Rican cohabiting women are included in the sample, the resulting Hispanic sample consists primarily of Mexican Americans.) Multivariate results do not differ when Puerto Ricans are excluded from the Hispanic category. The results that refer to the "other" race and ethnic category are not discussed in this article, because the sample size is quite small and includes many different racial and ethnic groups.

Educational level and employment status at the start of cohabitation were the variables used to determine socioeconomic disadvantage. These measures seemed appropriate, given the short duration of cohabiting unions. Data on the educational level and employment status of male cohabiting partners were not available. Education was coded into a four-category variable: less than 12 years, 12 years, 13-15 years and 16 or more years of education. Employment status was categorized into full-time, part-time and not employed. Income is typically an important component of socioeconomic status, but this measure could not be included due to the lack of retrospective earnings reports.

The remaining independent variables included in analyses served largely as control variables. Measures of the respondent's background include family structure at age 14 and religiosity while growing up. The variable measuring religiosity while growing up was based on five categories that measured frequency of attending religious services. Fertility prior to the start of cohabitation was included because unmarried women who had children prior to cohabitation were assumed be more likely to have children while cohabiting. This variable is tapping parity at the time of union formation. (In previous research, though, prior fertility had no effect on the timing of fertility within Canadian cohabiting unions.²¹) The effects of age and period were measured using two variables: age at the start of cohabitation and year of cohabitation.

Results

Cohabitation and Childbearing

Figure 1 presents life-table estimates of the cumulative percentage of women entering motherhood during cohabitation at each duration of cohabitation. These estimates are more informative than observed prevalence rates because they account for the variation in the duration of exposure. White cohabiting women are slowest to make the transition to motherhood in cohabitation. Hispanic women make the transition to motherhood in cohabitation faster and at greater levels than other women, and black women enter motherhood more slowly than Hispanics but faster than whites. These differences are somewhat greater when conceptions and not births are considered.

The first column of Table 1 presents the percentage distribution of the total sample of women and the sample of women who conceived a child during cohabitation, according to the independent variables. Most of the sample (73%) was non-Hispanic white, 10% was Hispanic, 14% was black and 4% belonged to other racial or ethnic groups. At age 14, the majority of the sample (61%) lived in families with two biological parents, 18% lived in a single-parent family and 15% lived in stepfamilies. The median level of education at the start of cohabiting was 12 years, one-quarter of the sample had less than a high school degree and 15% had a college degree. More than two-thirds of the sample was employed at the start of cohabitation, and a majority of women were employed full-time.

The majority of the sample entered cohabitation childless; one-fifth of the sample had conceived a live birth before starting to cohabit and 14% had given birth to a child before cohabitation. The average age at the start of a first cohabiting union was 21. The sample was roughly equally divided among the three time intervals measuring the year that the union began.

The second column of Table 1 shows the distribution of the independent variables for women having conceived a child while cohabiting. There were larger proportions of Hispanic and black women (19% and 22%, respectively) among the women who conceived a child than among the total sample (10% and 14%, respectively). A smaller proportion of women who conceived a child during cohabitation were from families with two biological parents in comparison with the total sample (54% vs. 61%). Whether or not a woman conceived a child did not affect the mean religiosity of the sample

Women who had a child during cohabitation had lower education levels than the entire sample—almost half (46%) had less than 12 years of education, while a quarter of the total sample had not completed 12 years. Women conceiving a child during cohabitation were also less likely than the total sample to be employed full-time (44% vs. 60%).

A greater percentage of the women who already had children went on to conceive a child while cohabiting than the total sample (21% vs. 14%). On average, the women who had a child while cohabiting were slightly younger than the total sample. Finally, women who had a child during cohabitation more often than the total sample started cohabiting prior to 1990.

Most women had only one child during cohabitation, but approximately one-fifth of women who became mothers while cohabiting gave birth two or more times (not shown). Among women who gave birth during cohabitation, one-third of Hispanic women gave birth to two or more children, while only one-quarter of black women and one-fifth of white women did so.

In this article, we can distinguish between women who had children that were born during cohabitation and those who conceived children during cohabitation. This distinction is important because nearly three-fifths (59%) of women who conceived a child during cohabitation went on to give birth to that child during cohabitation (Table 2). Thirty-four percent of the women married before the birth of their child, while only 7% of women separated from their cohabiting partner before their child was born. It is important to distinguish between births and conceptions because potentially different factors may predict whether a woman conceives her child during cohabitation or whether she also gives birth to her child while cohabiting. Furthermore, the family type at birth sharply differs among racial and ethnic groups. The overwhelming majority of Hispanic (69%) and black (77%) women who conceived a child during cohabitation also gave birth during that cohabiting union, while only half of white women did so.

Multivariate Models

The multivariate results estimating time to first conception are presented in Table 3. The hazard ratios are the exponential values of the coefficients and represent the percentage increase or decrease in the hazard of conception in contrast to the omitted group. The standard errors are presented in parentheses. Two models are shown: a zero-order model consisting of only race and ethnicity and a full model that includes all of the covariates. Identical analyses were conducted estimating the hazard of first birth in cohabitation, and any differences in results between the two outcomes are noted below.

The zero-order effects of race and ethnicity on the hazard of conception during cohabitation indicate that Hispanics have a 134% greater hazard of conceiving during cohabitation than do non-Hispanic white women, while black women have a 79% higher hazard of conceiving a child during cohabitation. The difference between black and Hispanic women is statistically significant; blacks have a 25% lower hazard of conceiving while cohabiting than do Hispanics. An examination of cohabiting births, and not conceptions, led to similar conclusions about racial and ethnic differences (not shown).

The full model shows that the racial and ethnic differences remain significant, but are somewhat reduced in magnitude when other independent variables are included. An exception is that the difference between black and Hispanic women does not remain statistically significant when the other covariates are added to the model.

The log-likelihood ratio test indicates that the covariates significantly contribute to the fit of the model (p=.000). There are no significant differences in the effects of family structure and religiosity on the hazard of a cohabiting conception or birth. Both education and employment status have substantial influences on the hazard of having a cohabiting conception or birth. The hazard of childbearing during cohabitation decreases significantly with education; women with some college education at the start of cohabitation have a lower hazard of having a conception during their cohabiting union than women with a high school education (hazard ratio, 0.7). Unemployed women and those working part-time have greater hazards of experiencing a conception than do women working full-time (1.3 and 1.5, respectively).

Women's fertility behavior prior to cohabitation significantly influences the hazard of a conception during cohabitation. Women who have conceived children prior to cohabitation make a significantly slower transition into parenthood than do women who have not had children prior to cohabitation (0.8). Also, women who cohabit at older ages have lower hazards of a conception or birth than do younger women. The period effects (the year at the start of cohabitation) do not significantly differ for the hazard of a conception, but the results suggest that women who started cohabiting in the 1990s had a significantly greater hazard of giving birth to a child during cohabitation than women who cohabited in the early 1980s (not shown). There has been no change in the proportion of cohabiting women who conceived a child during cohabitation, but the percentage of cohabiting women who give birth to a child during cohabitation has risen, largely because of declines in marriage among pregnant cohabiting women.²²

Cohabitation Status at Child's Birth

The odds that a woman who conceived during cohabitation gave birth during that cohabiting union are presented in Table 4, with the first column presenting the effects of only race and ethnicity (the zero-order model) and the second column incorporating the other covariates into the model (the full model). The multivariate results mirror those reported in Table 3: Pregnant Hispanic women have 139% greater odds of giving birth to their child during cohabitation than white pregnant women, and pregnant black women have 247% greater odds of giving birth to their child during cohabitation than white women. No statistically significant difference exists between black and Hispanic women.

The inclusion of the other covariates slightly reduces the magnitude but not the statistical significance of the effects of the race and ethnicity variables. The control variables contribute to the overall fit of the model (p=.005). In this model, black and Hispanic cohabiting women who conceive children during cohabitation have significantly higher odds of giving birth to that child during cohabitation than do pregnant, cohabiting white women (odds ratios, 3.0 and 1.9, respectively). Pregnant women from "other" types of families have greater odds of giving birth during cohabitation than do women from families with two biological parents (2.4). Pregnant women who had less than 12 years of education have higher odds of giving birth to their child during cohabitation than do women with 12 years of education (2.1).

Intention Status

More than half (56%) of cohabiting women classified their first child conceived within cohabitation as intended (Table 5). Levels of timing status within cohabitation are better understood when contrasted with levels among married and unmarried women. Analyses of the NSFG indicate that the vast majority of married women (82%) had an intended first conception, whereas only 39% of mothers who conceived their child prior to marriage while living alone (that is women who were unmarried and not cohabiting) categorized that birth as intended (not shown).

Childbearing during cohabitation is most common for Hispanic women, and a relatively high proportion of those births are intended. Almost two-thirds of Hispanic women (65%) had an intended conception in cohabitation, while 56% of black women and 51% of non-Hispanic white women intended their first conception during cohabitation (Table 5). Results are similar if the timing status of births, not conceptions, within cohabitation is considered.

The odds of an intended (rather than unintended) conception resulting in a live birth are presented in Table 6. Identical models predicting the odds that a birth during cohabitation was intended were estimated, and a similar pattern of results exists for both outcomes. The zero-order models show that Hispanic women have significantly greater odds of having an intended conception than white women (odds ratio, 1.6). Black women have similar odds of having an intended conception as white women (1.2).

The multivariate model shows that Hispanic women continue to have higher odds of having an intended conception than white women, even with the potential intervening mechanisms included in the model (odds ratio, 1.7). The control variables significantly contribute to the fit of the model (p=.000). Cohabitors who are more religious have lower odds of an intended conception (0.9). Women with less than 12 years of education have higher odds of an intended conception during cohabitation than do women with 12 years of schooling (1.6). The effects of higher levels of education are not statistically different than the effect of simply having a high school degree. Women who conceived prior to cohabitation have lower odds of having an intended birth than do women who had no children prior to cohabitation. Women who started cohabiting at older ages have higher odds of having an intended conception.

Discussion

Levels of unmarried childbearing receive continued attention in the United States, and the importance of cohabitation in understanding recent fertility patterns is increasingly recognized. This article contributes to our understanding of cohabitation and family building in three ways. First, unlike prior studies,²³ it focuses solely on cohabiting couples, allowing for a detailed examination of the socioeconomic differentials in childbearing during cohabitation. To better understand the implications of cohabitation for children's lives, it is necessary to know more about the women who have children while cohabiting. Women who are the most socioeconomically disadvantaged, who have low educational levels and who are not working full-time have considerably higher levels of childbearing during cohabitation. Furthermore, women who already have had children are more likely to go on to have children while they are cohabiting. In the multivariate models, education and family background are the primary factors determining entry into motherhood.

Second, instead of ignoring racial and ethnic differences and similarities in childbearing behavior, this article focuses on the timing of childbearing during cohabitation for black, white and Hispanic women. An underlying premise of most prior work is that cohabitation serves the same family functions for each racial and ethnic group. The findings from this study indicate that the majority of women do not conceive or give birth to children while cohabiting. Cohabitation is generally a childless union for white women, but among Hispanics and blacks, childbearing in cohabitation is more common. After the effects of socioeconomic characteristics are controlled, Hispanic and black women have higher hazards of childbearing during cohabitation than white women.

Third, this project moves beyond documenting differentials in childbearing by attempting to determine whether cohabitation serves as an acceptable family form for childbearing. Some women who become pregnant during cohabitation may decide that cohabitation is not an appropriate place for raising children and either may subsequently marry their partner (so their child is born into a formal marriage) or may separate from their partner and raise the child alone. If a pregnancy is associated with continuation of a cohabiting union (not marriage or separation), then cohabitation can be viewed as operating as an acceptable arena for childrearing. The findings indicate that Hispanic and black women are more likely than white women to remain cohabiting once they become pregnant. These results suggest that cohabitation serves as a more acceptable venue for raising children for Hispanic and black women than for white women.

Another indicator of how women feel about childbearing during cohabitation is whether they report a birth as being intended. More than half of women who conceived a child during cohabitation reported that it was intended. The contention that women who are more likely to have children during cohabitation are also more likely to report those children as intended is true only for Hispanic women. The higher levels of childbearing among Hispanic women during cohabitation are not the result of unintended births. Even though the level of childbearing during cohabitation is twice as high among blacks as among whites, the percentage of women who report unintended conceptions and births is roughly the same. Taken together, these results suggest that childbearing during cohabitation is more acceptable for Hispanic women than for blacks or whites.

The study has three key shortcomings. First, the analyses focus on single-sex models of fertility and cohabiting outcomes. These decisions occur within a dyadic context, and future work should incorporate crucial information about the cohabiting partners' characteristics; however, the NSFG did not collect retrospective data about the characteristics of a cohabiting partner whom the respondent did not marry.

Another limitation is the limited sample size, which precluded distinguishing between Hispanic ethnic groups (e.g., Cubans, Puerto Ricans, and Mexican-Americans). Consequently, the Hispanic categorization includes individuals with many different ethnic roots. The acceptability of cohabitation as an arena for family building could differ among Latino subgroups. Puerto Ricans represent the most disadvantaged Hispanic group—having an economic status on par with blacks. The family experiences of Puerto Ricans appear to be quite different from those of other Hispanic subgroups; in fact, cohabitation appears to operate more akin to marriage among Puerto Ricans than among other racial and ethnic groups.²⁴ Thus, both economic and cultural factors may influence how cohabitation operates among Hispanic subgroups.

Third, this article could not establish the mechanisms through which race and ethnicity influence fertility decisions because racial and ethnic differences are not explained by mere socioeconomic variation. Therefore, other aspects of the social context must contribute to racial and ethnic patterns of childbearing in cohabitation. It is beyond the scope of this analysis and of the data used to determine the specific mechanisms. Nonetheless, understanding the mechanisms remains a critical question. Differences in intention status of childbearing and decisions to remain living with cohabiting partners at the time of the birth provide some hints to possible explanations for racial and ethnic differentials. Other research on attitudes about nonmarital childbearing and cohabitation also provide some clues about sources of differences.²⁵ An important next step will be to specifically examine the racial and ethnic cultural and social backdrop under which family building decisions are made.

Researchers often try to place cohabitation on a continuum, with singlehood at one end and marriage at the other end. The questions often posed are whether cohabitation is a substitute for marriage or whether cohabitation replaces premarital courtship. These are reasonable questions, but they are confounded by the fact that the boundaries between marriage and singlehood have become less distinct, largely due to increases in nonmarital fertility. It does not appear that we will ever be able to strongly argue that cohabitation functions in one particular manner for all participants. We need instead to recognize and understand the sources of variability and change within cohabitation.

This article has attempted to improve our knowledge about some sources of variability in cohabiting unions. Moreover, our understanding of racial and ethnic differences in family formation often ignores cohabitation. These results suggest that accurate portrayals of recent family change among racial and ethnic groups require acknowledging cohabitation.