Marriage, Family Structure, and Maternal Employment Trajectories

Marriage, Family Structure, and Maternal Employment Trajectories Abstract Previous studies of maternal employment have focused on marital status differences, but the rise in nonmarital cohabiting parenthood problematizes the simple dichotomy between married and unmarried mothers. Theory and previous research yield mixed predictions as to whether cohabiting mothers’ employment will more closely resemble that of married mothers or lone unmarried mothers. Using data from the Fragile Families and Child Wellbeing Study, I examine how maternal employment varies across family structures (married parents, cohabiting unmarried parents, and lone unmarried mothers) in the five years after a birth for mothers living in urban areas in the United States. Descriptive statistics show few differences in maternal employment patterns by family structure. Controlling for human capital, however, I find that cohabiting mothers return to work earlier and work more than married mothers. Cohabiting mothers and lone mothers show very similar employment patterns. Additionally, cohabiting mothers who later marry have employment trajectories that are similar to married mothers, whereas married mothers who divorce increase their employment hours. Family characteristics, partner characteristics, and gender attitudes do not explain employment differences between married and cohabiting mothers. I speculate that cohabiting mothers work more than married mothers as a hedge against economic deprivation given high union dissolution rates for cohabiting couples. In the United States and across Europe, the normative context for sexual intimacy and childrearing has been shifting away from marriage. More than half of American women in their early forties have cohabited (Goodwin, Mosher, and Chandra 2010), and more than four in 10 births in 2013 were to unmarried mothers (Martin et al. 2015). Marriage’s monopoly on sex and reproduction has declined, and unmarried cohabitation has partially replaced it. Yet, the majority of young adults aspire to marry (Thornton and Young-DeMarco 2001), suggesting that marrriage has retained its symbolic significance. Is marriage merely symbolically important? Among families with children, are there differences by family structure in how families organize fundamental tasks, such as breadwinning and child-rearing? Mothers’ time in paid employment may be a particularly important aspect of family organization because it impacts family income, as well as potentially child development and the gendered division of housework. For decades in the United States, unmarried mothers were employed at much higher rates than married mothers, but these long-standing differences in maternal employment trends started to disappear in the early 1990s (Cohen and Bianchi 1999). In 2015, the percentage of US mothers with children under age six who were employed was very similar across marital statuses, with 59.5 percent of married mothers (with spouse present) employed compared with 60.5 percent of mothers of other marital statuses (Bureau of Labor Statistics 2017, table 6). The convergence of these trends lends itself to multiple interpretations, including that marital status is no longer a meaningful axis of comparison. If this is the case, the more important distinction may be between mothers with a resident partner and mothers without a partner. In this analysis, I examine how maternal employment differs by both marital status and coresidency (two key features of family structure) and which factors account for any observed family structure differences in maternal employment. Family structure may affect maternal employment through many possible mechanisms. These include differences in family income needs, access to instrumental and social support for the mother’s employment, the mother’s own preferences for paid work, and the mother’s perceived security to pursue her preferred level of employment. Additionally, family structure may affect employment through the influence of husband or partner preferences for the mother’s employment level. Alternatively, family structure may have no causal effect and may merely be a marker of pre-existing differences among women—such as differences in human capital characteristics and gender attitudes—that are both associated with family structure and affect maternal employment. Such characteristics may affect a woman’s opportunities and compensation for work as well as her motivation and preference for paid work. Whether family structure affects women’s employment or is merely a marker of other differences among women is an important question for scholars and policymakers. In the United States, women’s earnings now account for a sizable share of family income in married-couple families (Raley, Mattingly, and Bianchi 2006) and are often the sole source of income for single-mother families. Paid employment also potentially provides mothers with nonmonetary rewards, including social ties and a source of identity. Employment, however, can have substantial costs for mothers, including foregone time with children, high childcare costs, and stress from balancing mother and worker roles. Many mothers work full-time by necessity to support their families, even if their employment interferes with the care of their children or their own well-being. Thus, whether and to what extent family structure—here defined as married motherhood, cohabiting motherhood, or lone motherhood—affects women’s employment are compelling unanswered questions in family demography and social stratification research with implications for public policy. In this paper, I address the following research questions: How does family structure correlate with maternal employment among mothers in urban areas in the United States following a birth? Are the employment patterns of cohabiting mothers more similar to those of married mothers or lone mothers? Given the obvious impossibility of a random assignment design, researchers are left with the challenge of disentangling causal effects from selection effects and identifying mechanisms through which family structure influences maternal employment. I use a two-pronged approach that capitalizes on the rich data in the Fragile Families and Child Wellbeing Survey (FFCWS), a study of families with a birth in 1999 or 2000 in 20 US cities, to examine how maternal employment varies by family structure. First, I model the employment trajectories of mothers by family structure and try to differentiate between employment differences that might be caused by family structure versus those that can be attributed to other factors, such as differences in human capital. To do this, I use models with covariate adjustments for an unusually extensive set of characteristics of mothers, their partners, and their families. Second, I look at the employment patterns of mothers who change family structures to identify how their employment changes. Maternal Employment Patterns Family Structure Influences on Maternal Employment Unmarried cohabiting parenthood is a relatively new family form, and theory suggests both reasons to expect that cohabiting mothers’ employment would be similar to married mothers and reasons why it would differ. Below, I discuss first why married mothers are expected to have lower employment levels than lone mothers (unmarried mothers without a cohabiting partner). Second, I discuss expectations regarding cohabiting mothers’ employment (in the US context). Marriage is expected to be associated with lower levels of maternal employment than lone motherhood because having a husband provides a source of income and economic security independent of a woman’s own earnings. In contrast, in the post-1996 welfare reform period, lone mothers in the United States have few sources of income beyond their own employment1 and the generosity of family and friends. Income from government programs is a temporary source of family support, and most mothers receiving income through the Transitional Assistance for Needy Families (TANF) program are required to be employed to receive benefits. Although all fathers are legally responsible for supporting their children, receiving timely child support payments is far from a universal experience (Sorensen and Hill 2004), and the average child support payment is quite small (Nepomnyaschy and Garfinkel 2010). The legal status of marriage and the public nature of the commitment grant women both legal and socially recognized claims on their husbands’ incomes. Although high divorce rates undermine some of the guarantees of marriage, previously married women are more likely to obtain child support than never-married women (Sorensen and Hill 2004). This promise of economic support may allow married women to work part-time or take periods of absence from paid employment without incurring high risks of poverty. Thus, married mothers who desire lower levels of employment—perhaps becuase of their children’s developmental needs or their preferences for time-intensive homemaking practices—may have a better chance of realizing their preferences than unmarried mothers with similar preferences because of their stronger legal and social claims on their partners’ incomes. Alternatively, married women may have lower employment levels because their husbands prefer it. On average, men have more traditional gender role attitudes than women (see Davis and Greenstein [2009]). If marriage gives men power to gain their partners’ compliance with their preferences, we expect married women to have lower employment levels. Whether cohabiting mothers’ employment levels will more closely resemble married or lone mothers is an open question. On one hand, cohabiting mothers’ employment may be similar to that of married mothers if they have similar access to their partners’ income, share childcare and household tasks in a similar way, and have similar expected relationship durations. On the other hand, cohabiting mothers’ employment may be more similar to that of lone mothers if cohabiting couples do not pool income, share housework and childcare differently, or have more unstable relationships than married couples. Relatively few studies compare the housework, parenting time, and income pooling strategies of married and cohabiting couples. On the question of whether cohabiting men do more housework than married men, findings are mixed (see Davis, Greenstein, and Marks [2007]), as are conclusions regarding marital status differences in father involvement with children (e.g., Kalenkoski, Ribar, and Stratton 2007; Ono and Yeilding 2009). Kenney’s (2006) study of couples’ income pooling practices suggests that cohabiting women’s access to their partner’s income is more varied than married women’s access. Additionally, little is known about how cohabiting (men) partners may influence women’s employment. If cohabiting men have less traditional gender attitudes than married men, we may expect their partners to have higher employment levels than married women. Even if cohabiting partners and husbands have similar attitudes toward maternal employment, we might see differences in maternal employment if cohabiting partners and husbands differ in their ability to influence their partners. Cohabiting parenthood is a less stable family form than married parenthood in the United States (Manning, Smock, and Majumdar 2004). About half of cohabiting unions with a birth transition to marital unions within five years, and these relationships are often as stable as marriages contracted before a birth, controlling for couple characteristics (Musick and Michelmore 2015). Still, many women may be aware of the high dissolution rates of cohabiting unions, and some may maintain high levels of employment as insurance against economic deprivation if their partnership dissolves. Thus, the empirical evidence regarding cohabitating parenthood has mixed implications as to whether cohabitation may differ from marriage in its effects on maternal employment. Other Family Influences on Maternal Employment Other family characteristics—including domestic violence, children’s health, number of children, and social support—may also affect mothers’ employment. These characteristics differ in their distribution, and possibly in their effects on employment, across family structures. Previous research has found that women who experience domestic violence have more employment instability—reflecting the impacts of violence on women’s health, as well as partners’ intentional interference with women’s employment (Showalter 2016)—and domestic violence is more common among cohabiting mothers than married mothers (Kenney and McLanahan 2006). Having a child with a chronic illness or a disability is also associated with reduced maternal employment (Corman, Noonan, and Reichman 2005), and is more common for unmarried mothers (Lee, Sills, and Oh 2002). Other family characteristics that may hinder maternal employment, such as having tightly spaced births and more young children in the household, are more prevalent among married mothers (Gemmill and Lindberg 2013). Factors that facilitate maternal employment, such as social and instrumental support (Livermore and Powers 2006), also differ in their prevalence across family structures. In particular, a nearby or co-resident grandmother may provide low-cost or free childcare. Children of lone mothers are more likely to live with a grandparent than children with two co-resident parents (Fields 2003, table 3). Although these factors are unlikely to be primary mechanisms through which family structure affects maternal employment, analyses of maternal employment should account for these characteristics. In summary, family structure may affect maternal employment through mechanisms such as family income needs, family resources and constraints, partner preferences, and the mother’s perceived security to pursue her preferred level of employment. Previous research and theory do not provide a clear hypothesis regarding the employment patterns of cohabiting mothers. Cohabiting mothers may have less economic need than lone mothers, suggesting their employment would be similar to married mothers. Cohabiters, however, have shorter union durations, and cohabiting mothers have weaker legal and social claims to their partners’ incomes than married mothers, suggesting that their employment may be more similar to that of lone mothers. Family Structure “Effects”? The preceding section reviews expectations for why family structure may influence maternal employment, all other factors equal. But mothers’ characteristics are not similar across family structures, and family structures are not randomly distributed. Notably, unmarried cohabitation is the most common context for childbearing among US women without a high school degree, whereas it is still a rare context for motherhood for college-educated women (Wu 2017). Additionally, racial/ethnic differences in family structure are large. For instance, cohabiting parenthood is more common among Hispanic than non-Hispanic mothers, and lone motherhood is the modal context for births to black mothers (Monte and Ellis 2014). Accounting for educational and ethnic/racial differences in parenthood contexts is crucial for a study of maternal employment because these characteristics are highly correlated with employment. Understanding how family structure affects women’s employment requires ruling out the rival hypothesis that family structure is merely a proxy for other factors that are driving differences in maternal employment. Human capital—especially educational attainment—is arguably the most important predictor of women’s employment, so comparisons of maternal employment across family structures need to be particularly attentive to differences in human capital. An additional characteristic that might be causally related to both family structure and maternal employment (and thus could cause a spurious relationship between family structure and employment) is gender attitudes. Women with more conservative gender attitudes may be more likely than women with more progressive gender attitudes both to have their children within marriage and to aspire to full-time homemaking. Previous Research on Family Structure and Maternal Employment Few previous studies of maternal employment consider cohabiting mothers as a separate category; most often, they are combined with unmarried and unpartnered (lone) mothers. Additionally, many previous studies of maternal employment consider only labor force participation, though much of the variation in maternal employment is in how soon mothers return to work after a birth and how many hours per week they work. A study by Han et al. (2008) overcomes several of these shortcomings. The authors use data from the Early Childhood Longitudinal Study-Birth Cohort (ECLS-B), a national study of children born in 2001 in the United States, to examine how soon mothers returned to work after the birth of a child. Their analysis found only small differences by family structure in the percentage working nine months after the birth. However, once they adjusted for mother’s age, race/ethnicity, education, employment status before the birth, and birth parity, the differences across family structure were sizable: cohabiting mothers were 14 percentage points and single mothers were 11 percentage points more likely to be working nine months after the birth than married mothers. The study did not investigate whether other characteristics associated with family structure (such as exposure to domestic violence) were driving these differences, nor did it consider any of the mechanisms—such as gender attitudes or partner characteristics—that might explain family structure differences in maternal employment. Additionally, the study only examined employment immediately after the birth. My study builds on the Han et al. analysis by 1) examining several mechanisms that may account for family structure differences; 2) controlling for a much richer set of mother characteristics that might be linked to both family structure and maternal employment; 3) including measures of fathers’ characteristics; and 4) considering a longer time period (five years versus nine months) and multiple measures of employment. Methods Data and Sample Data for this analysis are from the Fragile Families and Child Wellbeing Study (FFCWS), a longitudinal study of a birth cohort of children born between 1998 and 2000 in 20 large urban areas in the United States. The study oversampled nonmarital births, yielding a sample of approximately 3,700 children born to unmarried parents (including cohabitors) and 1,200 children born to married parents. Data collection began with an in-person interview of the mother at the hospital at the time of the birth and of the father either in the hospital or in another location shortly after the birth. Other data collections included surveys at 12 months (1999–2001), 30 months (2001–2003), and 60 months (2003–2005) after the child’s birth. For a thorough description of the study, see Reichman and colleagues (2001). Data from FFCWS are uniquely suited to answer my research questions because the study includes a large sample of cohabiting mothers and very detailed information on family structure, maternal characteristics, and paternal characteristics. Supplemental table 1 shows that mothers in FFCWS are similar to mothers of the same marital status with a birth in 2000 in the total US population on age and educational characteristics. Notably, mothers in FFCWS are more racially/ethnically diverse. Of the original mothers in FFCWS (n = 4,898), approximately 75 percent (n = 3,676) were interviewed at all four waves of data collection. The baseline interviews differed slightly for two of the 20 cities, so I exclude all mothers (n = 656) in these two cities. For mothers in the remaining 18 cities who participated in all survey waves, 98 percent answered all relevant employment questions. The analytic sample is composed of these mothers who participated in all four waves and answered all relevant items about employment histories (n = 3,132). The distribution of mothers by baseline family structure barely differs between the full sample (n = 4,898) and my analytic sample (n = 3,132). Of the mothers who did not participate in all four waves, most participated in at least one follow-up wave; there were no differences in maternal work hours prior to the birth or at any of the follow-up waves between mothers included in my analysis and those excluded for nonparticipation. To address missing data on parent characteristics, I use multiple imputation procedures2 to create 20 imputed datasets. I also have rerun all of the analyses using only the subset of families with no missing data (i.e., complete cases); I note when there are substantive differences in results between analyses with imputed data and complete cases only. Subsample For all descriptive statistics and the first set of analyses, I present findings for the full sample and a theoretically interesting subsample. This subsample consists of mothers who are part of demographic groups (defined by age, race/ethnicity, and education) with the most variation in family structure and where the comparison by family structure might be the most informative. In demographic groups with both very low and very high nonmarital birth rates, women in family structures that are non-normative for their particular demographic group may be unusual in ways that are not easily captured by covariate variables. One demographic group with a very low percentage of nonmarital births is college-educated mothers; only 9 percent of births to college-educated mothers from 1997 to 2001 were nonmarital (Kennedy and Bumpass 2008). Demographic groups with high percentages of nonmarital births include women without a high school degree, women under age 20, and black women. In the late 1990s, the percentage of births that were nonmarital to mothers with these characteristics were 64 percent for mothers with less than high school (Kennedy and Bumpass 2008), 79 percent for mothers under age 20 (Ventura and Bachrach 2000), and 69 percent for black mothers (Ventura and Bachrach 2000). Excluding women in demographic groups with very high or low nonmarital birth rates yields a subsample that consists of white non-Hispanic and Hispanic mothers who are age 20 or older who have either a high school degree or some college education (n = 630). Dependent and Explanatory Variables Maternal employment Measures of maternal employment are based on the mother’s self-report of when she returned to work after the birth and her hours of work in the previous week as collected at the 12-month (1999–2001), 30-month (2001–2003), and 60-month surveys (2003–2005). I topcoded hours of work at 80 hours per week.3 Initial family structure The family structures considered in this paper are 1) married, 2) (unmarried) cohabiting, and 3) living without a cohabiting partner or husband (hereafter referred to as lone mothers). I define family structure at baseline/Time 1 in reference to the mother’s relationship to the father of the new child.4 I use the mother’s report of marital status and whether she is living with the baby’s father all or most of the time. At the time of the birth, 25.2 percent of the mothers report being married, 35.3 percent cohabiting, and 39.5 percent as not living with the baby’s father. Of the subsample, 37.1 percent are married, 39.8 percent are cohabiting, and 23.0 percent are lone mothers. Family structure trajectories Many mothers experience changes in their family structure. I classify mothers according to their family structure at the beginning of the study (married, cohabiting, lone) and whether this family structure is stable or how it changes. To do this, I use the detailed marital and partnership history questions, which pertain to the mother’s relationship with the baby’s father and any new partners. I define 10 family structure trajectories: 1) Stably cohabiting (with the baby’s father); 2) Cohabiting to married (initially cohabiting with and subsequently marrying the baby’s father); 3) Cohabiting to lone (initially cohabiting with the baby’s father, followed by union dissolution); 4) Cohabiting, multiple changes (initially cohabiting with the baby’s father, followed by multiple changes, which may include re-partnering); 5) Stably married (to the baby’s father); 6) Married to divorced (initially married to the baby’s father and later divorced, with some mothers also re-partnering5); 7) Stably lone (not living with a husband or cohabiting partner at any survey wave); 8) Lone to cohabiting (initially unmarried and not cohabiting, followed by a later period of cohabitation with either the baby’s father or a new partner; 9) Lone to married (initially unmarried and not cohabiting, followed by a later marriage to either the baby’s father or a new partner); and 10) Lone with multiple changes (initially unmarried and not cohabiting, followed by multiple changes, which could include multiple partnership or marital status changes). Other Variables Independent variables include the mother’s demographic and human capital characteristics, her family resources and constraints, her family income and assets, the baby’s father’s characteristics, and the gender attitudes of the mother and father. Variables for the mother’s characteristics, family resources, household income and financial resources, and relationship status are based on the mother’s reports at the baseline interview. The father’s characteristics are based on the father’s reports (if he participated in the study) or the mother’s proxy reports if he did not participate. I do not include characteristics, such as subsequent fertility, that might be influenced by post-birth maternal employment or family structure changes. Mother and father characteristics Maternal demographic characteristics include age (dummy variable for teen, defined as under age 20; linear variable for age in years over age 20), race or ethnicity (non-Hispanic white or other, black, non-black Hispanic), and immigrant status (1 = immigrant). Paternal demographic characteristics include his age relative to the mother’s age (difference in years), and whether he is the same race/ethnicity as the mother. Human capital characteristics included for both parents include education (less than high school, GED, high school only, some college, college or more) and whether the parent has self-reported poor health. I also include the parents’ hourly pay rate (logged) from her/his last job and, for mothers, an indicator of whether she has no work experience. If the mother had no work experience or the parent had a very low (below $3.00/hour) or high (over $100/hour) reported wage rate, I treat the wage rate as missing and impute the wage. I also include an indicator variable to denote the missing wage. Approximately 13 percent of mothers and 27 percent of fathers have an imputed wage rate. For fathers, I also include whether he had a criminal record as of the baby’s birth, whether he reported using illegal substances, and how many weeks he was unemployed in the year before the birth. For mothers, I also include variables relating to mental health—a measure of the mother’s impulsive tendencies and a variable for whether there are severe mental health problems in her natal family—and her cognitive abilities (as measured by a subset of the Revised Weschler Adult Intelligence Scale), as these may affect a mother’s ability to find and maintain steady employment. Additional variables relating to mothers’ human capital include characteristics of her natal family (as these may predict unmeasured aspects of her cognitive and noncognitive abilities), including her mother’s education (less than high school, high school, more than high school, missing) and whether she grew up with both biological parents. Gender attitudes Mothers and fathers were each asked to respond to two items about their gender attitudes: “The important decisions in the family should be made by the man of the house” and “It is much better for everyone if the man earns the main living and the woman takes care of the home and family” (4 = strongly agree; 1 = strongly disagree). An individual’s responses to these two questions were combined to form a scale where higher scores indicate more traditional attitudes. The gender attitudes scales—as well as mother’s cognitive scores—are standardized within gender (mean = 0, standard deviation = 1). Higher values indicate that the mother is more conservative on gender attitudes than other mothers in the sample. FFCWS does not include any items related to a woman’s preferences for her own employment, and most other datasets used to examine women’s employment also lack items on preferences. Although preferences from early adulthood are associated with employment patterns, preferences are unstable over the lifecourse and often change with family circumstances (Kan 2007). Family constraints and resources Variables include whether the mother experienced domestic violence in the year before the birth, whether the child had a low birth weight, whether either of the baby’s grandmothers live in the household, and the number and age of other children in the household (measured as the number of children other than the new baby who are under age 3, ages 3–5, age 6 and older). Measures of economic resources include whether the mother lives in a home that she or a family member owns, and her report of her household’s status at the end of the month: just enough money to pay bills (reference category), not enough money to pay bills, or some money left after paying bills. Analyses I focus on two aspects of maternal employment: whether a mother worked within a year of the birth and how many hours she works per week in the five years after the birth. To examine how family structure associates with maternal employment, I use a two-pronged approach. First, I model maternal employment levels controlling for characteristics associated with employment and with family structure. In this part of the analysis, I use static family structure variables to predict which mothers work within a year of the birth (using logistic regression models) and how many hours mothers work per week (using growth curve models). Second, I model employment hours per week by family structure trajectories over a five-year period (using growth curve models). Comparing employment trajectories across mothers with different family structure trajectories allows me to identify whether mothers with stable family structures have different employment histories than mothers with family structure changes. To predict whether the mother returned to work within a year of the birth, I use logistic regression models6 run in STATA, and I use the MI suite of commands to combine the estimates from the 20 imputed datasets and estimate appropriate standard errors. I cluster the standard errors by city to account for the sampling design. To examine hours worked per week over the five-year period, I use growth curve models, an extension of structural equation modeling (Bollen and Curran 2006). Growth curve models are “statistical methods that allow for the estimation of inter-individual variability in intra-individual patterns of changes over time” (Curran, Obeidat, and Losardo 2010, p. 122). This is an appropriate modeling strategy given that I expect that maternal employment will change (increasing, on average) over time as children age, and that this pattern of change will vary across mothers. Growth curve models allow estimation of differences in initial levels of an outcome and in changes over time in that outcome. In my analyses, the intercept estimates hours of employment (per week) one year after the birth and the slope describes the rate of growth in employment hours (per week) over the next four years, estimated from measurements at 30 and 60 months. Growth curve models are incredibly flexible, and researchers have many options for how to specify these models, including how to model time and group differences, and whether to allow variables to predict the intercept, the slope, or both.7 Measures of model fit8 indicated that a linear specification did not fit the trend in maternal employment hours. Maternal employment hours increase slowly from one year to 30 months and then increase more rapidly between 30 and 60 months. After testing several specifications, I chose a model with the one-year timepoint set to 0, the third timepoint (at 60 months) set to 2, and the second timepoint (at 30 months) freely estimated. For estimating the growth curve models, I divide employment hours by 10 so that the scale of the variances for the outcome variables are similar to those for the predictor variables. The baseline model (without covariates) fits the data well, as indicated by measures of model fit and by a comparison of observed and fitted values (see supplemental figures 1a and 1b). I investigated whether to use a multiple-group model with groups defined by family structure, or whether to estimate a single-group model with covariates for family structure. I find that a single group model with covariates for family structure fits the data well (see supplemental figures 1a and 1b). I also tested whether the parameterization of time should be the same across family structures. Including covariates for family structure resulted in a similar model fit as that for a model in which time parameters were freely estimated separately for each family structure. Each mother’s employment trajectory is characterized by a unique intercept (α), slope (β), and error term (ε). I follow the same style of notation as Meadows, McLanahan, and Brooks-Gunn (2008). The level-one equation is as follows:   yit=αi+βit+εit (1) In this equation, y is mother i’ s hours of work at time t, αi is mother i’s intercept, βit is mother i’s slope at time t, and εit is the mother- and time-specific error term. I allow the covariates, which are time invariant, to influence both the intercept and the slope because some factors may influence maternal employment differentially by children’s age. The level-two equations are as follows:   αi=α0+α1xi1+α2xi2+α3xi3…+αkxik+ui (2)  βi=β0+β1xi1+β2xi2+β3xi3…+βkxik+vi (3) In these equations, the x’s are the time-invariant predictor variables. I present results from the model specifications, which includes all variables shown in table 1 plus city covariates. Table 1. Mother, Father, and Family Characteristics by Family Structure (at Birth of the Child), n = 3,132   Full sample (n = 3,132)  Subsample (n = 630)  Cohab  Married  Lone  Cohab  Married  Lone  n = 1105  n = 790  n = 1237  n = 251  n = 234  n = 145  Mother’s characteristics               Age (mean, excluding teens)  24.4  29.5  23.9  25.5  29.0  25.8   Teen (%)  19.0  3.4  26.9  0  0  0   Race/ethnicity:                Non-Hispanic white/Other (%)  22.3  53.6  12.6  49.0  69.7  49.7    Black (%)  48.0  27.5  70.6  0  0  0    Hispanic (%)  29.7  18.9  16.8  51.0  30.3  50.3   Immigrant (%)  13.2  21.6  4.9  22.3  20.9  8.3   In poor health (%)  7.2  3.4  8.3  4.8  2.6  6.2   Standardized cognitive score  −0.06  0.40  −0.08  0.10  0.41  0.23   Impulsive (%)  10.7  5.0  13.0  8.4  4.3  13.8   Own education:                Less than high school (%)  35.6  12.2  38.0  0  0  0    GED (%)  6.4  2.3  5.0  0.1  0  0    High school only (%)  29.6  18.1  29.9  47.6  34.2  42.1    Some college (%)  25.9  30.0  24.1  52.3  65.8  57.9    College or more (%)  2.6  37.5  2.9  0.1  0  0   Logged hourly wage rate  2.1  3.1  2.0  2.6  2.9  2.2   Missing hourly wage rate (%)  12.3  12.7  14.7  7.2  11.5  11.7   No work experience (%)  2.3  2.4  3.5  0.4  3.0  2.8   Maternal education:                Less than high school (%)  21.5  10.4  20.9  20.7  15.0  19.3    High school only (%)  55.9  60.2  51.3  61.8  61.5  59.3    More than high school (%)  16.6  25.9  19.8  14.3  20.9  15.9    Unknown or missing (%)  6.0  3.5  8.0  3.2  2.6  5.5   Poor parental mental health (%)  14.4  12.5  13.9  10.8  13.7  13.8   Lived with both parents at 14 (%)  38.6  64.7  30.7  52.5  62.9  50.3   Mother’s gender attitudes                Standardized gender attitudes (mean)  −0.08  0.03  −0.12  −0.22  0.10  −0.23  Family constraints and resources               No. of other children under age 3  0.24  0.27  0.24  0.17  0.26  0.16   No. of children ages 3, 4, or 5  0.24  0.25  0.22  0.22  0.24  0.18   No. of children ages 6 and over  0.42  0.41  0.42  0.36  0.42  0.39   Low birth weight baby (%)  10.2  5.5  12.8  6.5  5.3  7.2   Domestic violence (%)  5.1  2.0  6.9  6.0  2.1  8.9   Grandmother in household (%)  16.0  6.8  46.1  12.4  6.5  47.6   Home owned by family (%)  25.5  55.1  36.3  28.8  56.9  49.0   Not enough money for bills (%)  11.7  4.5  21.2  10.8  3.6  22.1   Money left after paying bills (%)  43.6  63.0  34.9  47.3  60.7  34.5  Father’s characteristics               Father’s age relative to mother’s  2.7  2.4  2.7  2.4  2.1  2.3   Same race/ethnicity as mother (%)  86.9  87.5  87.2  76.5  79.1  64.0   In poor health (%)  7.9  5.4  8.9  6.4  5.1  11.3   Reports drug use (%)  9.0  4.1  11.9  8.0  6.0  20.7   Education:                Less than high school (%)  37.3  12.2  36.5  25.5  9.0  25.8    GED (%)  8.0  3.3  8.9  5.6  4.7  9.2    High school only (%)  28.3  21.1  33.5  27.4  25.2  30.5    Some college (%)  23.1  29.7  18.0  35.5  39.7  28.0    College or more (%)  3.3  33.6  3.0  6.0  20.9  6.5   Criminal record (%)  30.3  13.0  21.7  28.7  15.4  21.4   Weeks not employed (mean)  10.1  4.1  15.0  6.0  4.2  10.0   Logged hourly wage rate  2.3  2.8  2.2  2.5  2.8  2.4   Missing hourly wage rate (%)  18.5  18.1  43.7  13.1  18.8  40.7   Father’s gender attitudes               Standardized (mean)  −0.05  −0.04  −0.03  −0.20  −0.08  −0.20    Full sample (n = 3,132)  Subsample (n = 630)  Cohab  Married  Lone  Cohab  Married  Lone  n = 1105  n = 790  n = 1237  n = 251  n = 234  n = 145  Mother’s characteristics               Age (mean, excluding teens)  24.4  29.5  23.9  25.5  29.0  25.8   Teen (%)  19.0  3.4  26.9  0  0  0   Race/ethnicity:                Non-Hispanic white/Other (%)  22.3  53.6  12.6  49.0  69.7  49.7    Black (%)  48.0  27.5  70.6  0  0  0    Hispanic (%)  29.7  18.9  16.8  51.0  30.3  50.3   Immigrant (%)  13.2  21.6  4.9  22.3  20.9  8.3   In poor health (%)  7.2  3.4  8.3  4.8  2.6  6.2   Standardized cognitive score  −0.06  0.40  −0.08  0.10  0.41  0.23   Impulsive (%)  10.7  5.0  13.0  8.4  4.3  13.8   Own education:                Less than high school (%)  35.6  12.2  38.0  0  0  0    GED (%)  6.4  2.3  5.0  0.1  0  0    High school only (%)  29.6  18.1  29.9  47.6  34.2  42.1    Some college (%)  25.9  30.0  24.1  52.3  65.8  57.9    College or more (%)  2.6  37.5  2.9  0.1  0  0   Logged hourly wage rate  2.1  3.1  2.0  2.6  2.9  2.2   Missing hourly wage rate (%)  12.3  12.7  14.7  7.2  11.5  11.7   No work experience (%)  2.3  2.4  3.5  0.4  3.0  2.8   Maternal education:                Less than high school (%)  21.5  10.4  20.9  20.7  15.0  19.3    High school only (%)  55.9  60.2  51.3  61.8  61.5  59.3    More than high school (%)  16.6  25.9  19.8  14.3  20.9  15.9    Unknown or missing (%)  6.0  3.5  8.0  3.2  2.6  5.5   Poor parental mental health (%)  14.4  12.5  13.9  10.8  13.7  13.8   Lived with both parents at 14 (%)  38.6  64.7  30.7  52.5  62.9  50.3   Mother’s gender attitudes                Standardized gender attitudes (mean)  −0.08  0.03  −0.12  −0.22  0.10  −0.23  Family constraints and resources               No. of other children under age 3  0.24  0.27  0.24  0.17  0.26  0.16   No. of children ages 3, 4, or 5  0.24  0.25  0.22  0.22  0.24  0.18   No. of children ages 6 and over  0.42  0.41  0.42  0.36  0.42  0.39   Low birth weight baby (%)  10.2  5.5  12.8  6.5  5.3  7.2   Domestic violence (%)  5.1  2.0  6.9  6.0  2.1  8.9   Grandmother in household (%)  16.0  6.8  46.1  12.4  6.5  47.6   Home owned by family (%)  25.5  55.1  36.3  28.8  56.9  49.0   Not enough money for bills (%)  11.7  4.5  21.2  10.8  3.6  22.1   Money left after paying bills (%)  43.6  63.0  34.9  47.3  60.7  34.5  Father’s characteristics               Father’s age relative to mother’s  2.7  2.4  2.7  2.4  2.1  2.3   Same race/ethnicity as mother (%)  86.9  87.5  87.2  76.5  79.1  64.0   In poor health (%)  7.9  5.4  8.9  6.4  5.1  11.3   Reports drug use (%)  9.0  4.1  11.9  8.0  6.0  20.7   Education:                Less than high school (%)  37.3  12.2  36.5  25.5  9.0  25.8    GED (%)  8.0  3.3  8.9  5.6  4.7  9.2    High school only (%)  28.3  21.1  33.5  27.4  25.2  30.5    Some college (%)  23.1  29.7  18.0  35.5  39.7  28.0    College or more (%)  3.3  33.6  3.0  6.0  20.9  6.5   Criminal record (%)  30.3  13.0  21.7  28.7  15.4  21.4   Weeks not employed (mean)  10.1  4.1  15.0  6.0  4.2  10.0   Logged hourly wage rate  2.3  2.8  2.2  2.5  2.8  2.4   Missing hourly wage rate (%)  18.5  18.1  43.7  13.1  18.8  40.7   Father’s gender attitudes               Standardized (mean)  −0.05  −0.04  −0.03  −0.20  −0.08  −0.20  Findings Descriptive Statistics Table 1 shows the distribution of parent and family characteristics by family structure for the full sample and the subsample. (The descriptive statistics and analyses are unweighted.) On average, cohabiting mothers are less advantaged in human capital characteristics (education, wage rate, and work experience) than married mothers. Cohabiting mothers are also more likely than married mothers to be in poor health, have a low birth weight baby, or report experiencing domestic violence in the year before the birth, all of which are characteristics that depress employment. These differences hold for the full sample and the subsample. The differences between cohabiting and lone mothers, in both the full sample and the subsample, are less stark. Cohabiting and lone mothers have similar levels of education, cognitive scores, and hourly wage rates. A smaller share of cohabiting mothers than lone mothers live with the baby’s grandmother. Additionally, cohabiting mothers are more likely than married mothers—but less likely than lone mothers—to report not having enough money to pay bills. Table 1 also shows the distribution of father characteristics by family structure. In both the full sample and subsample, husbands are more educated, less likely to use drugs or have a criminal record, and had fewer weeks of unemployment than cohabiting fathers. On some dimensions, such as unemployment, cohabiting fathers are more advantaged than lone fathers, whereas on dimensions such as educational attainment and criminal justice involvement, cohabiting fathers appear equally or more disadvantaged. Given the differences in mother, father, and couple characteristics by family structure, we might expect to find considerable differences in maternal employment levels by family structure. But, as table 2 shows, observed differences in maternal employment by family structure among the full sample of FFCWS mothers are small. The mean number of hours worked per week in the first year is similar across family structures (approximately 21 hours). The percentage of mothers employed at the one-year interview is lower for cohabiting mothers than married mothers, though the percentage of cohabiting mothers who have worked at any point during the first year after the birth (74.2 percent) is higher than that of married mothers (65.2 percent). Among employed mothers, cohabiting mothers work approximately three hours more per week than married mothers and work the same number of hours as lone mothers. These statistics suggest relatively small differences in maternal employment in the first year after a birth. Table 2. Maternal Employment by Family Structure (at Birth of the Child)   Full sample (n = 3,132)  Subsample (n = 630)  Cohabiting  Married  Lone  Cohabiting  Married  Lone  (n = 1,105)  (n = 790)  (n = 1237)  (n = 251)  (n = 234)  (n = 145)  % employed at any point during the first year  74.2  65.2*  76.5  77.7  60.7*  78.6  Percentage employed at each survey wave               At one year  54.5  58.1  53.8  63.3  55.6  64.1   At 30 months  57.1  59.1  56.6  64.0  59.0  63.4   At 60 months  62.7  62.8  57.8*  68.9  61.8  63.4  Hours worked per week: mean and standard deviation               All mothers                At one year  21.1  20.6  20.5  23.6  19.1*  23.5    (21.4)  (20.1)  (21.1)  (20.2)  (19.6)  (20.8)    At 30 months  23.3  21.4  22.6  24.5  20.3*  25.5    (22.4)  (20.6)  (22.1)  (20.5)  (19.8)  (21.6)    At 60 months  35.8  29.8*  35.9  35.6  27.5*  37.0    (16.8)  (18.9)  (16.9)  (15.4)  (18.1)  (15.9)   All working mothers (excludes mothers with zero hours)                At one year  38.8  35.4*  38.0  37.3  34.4*  36.6    (12.7)  (13.1)  (12.9)  (11.4)  (12.9)  (13.8)    At 30 months  40.8  36.3*  40.0  38.5  34.4*  40.2    (12.9)  (13.3)  (13.1)  (11.0)  (13.3)  (11.9)    At 60 months  40.3  36.9*  40.8  38.9  34.7*  39.8    (13.0)  (14.1)  (12.9)  (11.8)  (13.0)  (13.6)    Full sample (n = 3,132)  Subsample (n = 630)  Cohabiting  Married  Lone  Cohabiting  Married  Lone  (n = 1,105)  (n = 790)  (n = 1237)  (n = 251)  (n = 234)  (n = 145)  % employed at any point during the first year  74.2  65.2*  76.5  77.7  60.7*  78.6  Percentage employed at each survey wave               At one year  54.5  58.1  53.8  63.3  55.6  64.1   At 30 months  57.1  59.1  56.6  64.0  59.0  63.4   At 60 months  62.7  62.8  57.8*  68.9  61.8  63.4  Hours worked per week: mean and standard deviation               All mothers                At one year  21.1  20.6  20.5  23.6  19.1*  23.5    (21.4)  (20.1)  (21.1)  (20.2)  (19.6)  (20.8)    At 30 months  23.3  21.4  22.6  24.5  20.3*  25.5    (22.4)  (20.6)  (22.1)  (20.5)  (19.8)  (21.6)    At 60 months  35.8  29.8*  35.9  35.6  27.5*  37.0    (16.8)  (18.9)  (16.9)  (15.4)  (18.1)  (15.9)   All working mothers (excludes mothers with zero hours)                At one year  38.8  35.4*  38.0  37.3  34.4*  36.6    (12.7)  (13.1)  (12.9)  (11.4)  (12.9)  (13.8)    At 30 months  40.8  36.3*  40.0  38.5  34.4*  40.2    (12.9)  (13.3)  (13.1)  (11.0)  (13.3)  (11.9)    At 60 months  40.3  36.9*  40.8  38.9  34.7*  39.8    (13.0)  (14.1)  (12.9)  (11.8)  (13.0)  (13.6)  Note: Data are from the Fragile Families and Child Wellbeing Study. Subsample includes non-Hispanic white and Hispanic mothers over age 20 with a high school degree or some college. Differences with cohabiting mothers that are statistically significant at p < 0.05 are indicated by *. At 30 months after the birth, the average hours of work for cohabiting mothers is two hours per week more than married mothers and one hour per week more than lone mothers. Among employed mothers, the difference between cohabiting and married mothers is larger, with cohabiting mothers working 4.5 hours more per week than married mothers. By the child’s fifth birthday, differences by family structure are quite large, with a difference of six hours per week between married mothers and unmarried mothers, with nearly identical work hours for cohabiting and lone mothers. The difference between cohabiting and married mothers in hours of work at 60 months is not driven primarily by differences in labor force participation rates; 62.8 percent of married mothers and 62.7 percent of cohabiting mothers are employed. The right side of table 2 shows employment statistics for the subsample. Mothers in this group are more similar to each other on human capital and demographic characteristics than are mothers in the full sample. Among this more homogeneous group, the employment differences between cohabiting and married mothers are considerably greater than for the full sample. Approximately 78 percent of cohabiting and lone mothers return to work in the first year after the birth, compared with 60.7 percent of married mothers. In the first year, cohabiting mothers work 4.5 hours more per week than married mothers and the same number of hours as lone mothers. By the child’s fifth birthday, the difference between cohabiting and married mothers increases to 8.1 hours more per week for cohabiting mothers. These sizable differences in maternal employment by family structure in the subsample suggest that the small differences in the full sample may reflect the offsetting influences of human capital differences and family structure on employment. Multivariate Results: Employment in the First Year after a Birth I use logistic regression models to predict which mothers were employed at some point in the first year following a birth. In models shown in table 3, I find that all else equal, cohabiting mothers are more likely to be employed in the first year than married mothers. Accounting for differences among women via covariate adjustment does not attenuate this relationship much. For the full sample, the odds of working in the first year after the birth are considerably lower for married mothers compared to the reference group of unmarried cohabiting mothers (B = −0.47, s.e. = 0.16, odds ratio = 0.63). Employment for lone mothers is not significantly different than that for cohabiting mothers (B = 0.14, s.e. = 0.13, odds ratio = 1.15). Table 3. Results from Logistic Regression Models Predicting Employment in the First Year after the Birth   Full sample (n = 3,132)  Subsample (n = 630)  Odds ratio  B  s.e.  Odds ratio  B  s.e.  Family structure               Married  0.63  −0.47  0.16**  0.48  −0.72  0.23**   Lone  1.15  0.14  0.13  1.33  0.29  0.39  Mother’s characteristics               Age (years over 20)  0.97  −0.03  0.01**  0.96  −0.04  0.02~   Teen mother  1.30  0.26  0.13*  n.a.       Black  2.42  0.88  0.13***  n.a.       Hispanic  1.26  0.23  0.15  1.39  0.33  0.30   Immigrant  0.74  −0.30  0.11**  1.06  0.06  0.21   In poor health  0.59  −0.53  0.23*  0.46  −0.78  0.59   Standardized cognitive score  0.97  −0.03  0.04  1.01  0.01  0.08   Impulsive  0.85  −0.16  0.12  0.68  −0.38  0.34   Less than high school  0.60  −0.51  0.09***  n.a.       GED  1.14  0.13  0.19  n.a.       Some college  1.38  0.32  0.18~  1.87  0.63  0.22**   College  2.06  0.72  0.26**  n.a.       Logged hourly wage rate  1.24  0.22  0.03***  1.26  0.23  0.12*   Missing hourly wage rate  0.47  −0.76  0.11***  0.79  −0.24  0.33   No work experience  0.15  −1.91  0.44***  0.19  −1.68  0.88~   Maternal ed.—less than high school  0.93  −0.08  0.14  1.32  0.28  0.42   Maternal ed.—more than high school  1.02  0.02  0.17  1.04  0.03  0.38   Maternal ed.—missing  1.04  0.04  0.21  1.05  0.05  0.72   Poor parental mental health  1.01  0.01  0.17  0.77  −0.26  0.36   Lived with both parents at age 14  0.96  −0.04  0.10  0.77  −0.26  0.28   Gender attitudes  0.83  −0.19  0.05***  0.79  −0.24  0.10*  Family constraints and resources             # of children under age 3  0.76  −0.27  0.11*  0.94  −0.07  0.26   # of children ages 3–5  0.87  −0.14  0.13  1.16  0.15  0.23   # of children ages 6 and over  0.94  −0.06  0.07  0.96  −0.05  0.09   Low birth weight baby  0.74  −0.30  0.19  0.90  −0.11  0.33   Domestic violence  1.17  0.15  0.21  1.00  0.00  0.60   Grandmother in the household  0.97  −0.03  0.12  1.02  0.02  0.27   Home owned by mother/family  0.95  −0.05  0.09  0.97  −0.03  0.15   Reports not enough money for bills  0.70  −0.36  0.12**  0.41  −0.90  0.24***   Reports money left after paying bills  1.07  0.07  0.10  1.72  0.54  0.20**  Father characteristics               Father’s age relative to mother’s  0.98  −0.02  0.01**  0.94  −0.06  0.02***   Same race as mother  0.67  −0.40  0.13**  0.75  −0.28  0.23   In poor health  1.32  0.28  0.17  0.91  −0.09  0.43   Reports drug use  1.52  0.42  0.16**  3.07  1.12  0.42**   Less than high school  1.21  0.19  0.14  1.63  0.49  0.33   GED  1.15  0.14  0.20  0.94  −0.06  0.48   Some college  0.83  −0.19  0.13  1.03  0.03  0.19   College  0.63  −0.46  0.24~  1.11  0.10  0.38   Criminal record  1.04  0.03  0.13  1.18  0.17  0.30   Weeks not employed  0.99  −0.01  0.00*  1.00  0.00  0.01   Logged hourly wage rate  0.93  −0.08  0.11  0.67  −0.40  0.21~   Missing hourly wage rate  0.80  −0.23  0.14  1.00  0.00  0.24   Gender attitudes  0.86  −0.15  0.05***  0.91  −0.10  0.12  Intercept    1.84  0.40    2.21  0.61  Log likelihooda  −1540.31  −313.25    Full sample (n = 3,132)  Subsample (n = 630)  Odds ratio  B  s.e.  Odds ratio  B  s.e.  Family structure               Married  0.63  −0.47  0.16**  0.48  −0.72  0.23**   Lone  1.15  0.14  0.13  1.33  0.29  0.39  Mother’s characteristics               Age (years over 20)  0.97  −0.03  0.01**  0.96  −0.04  0.02~   Teen mother  1.30  0.26  0.13*  n.a.       Black  2.42  0.88  0.13***  n.a.       Hispanic  1.26  0.23  0.15  1.39  0.33  0.30   Immigrant  0.74  −0.30  0.11**  1.06  0.06  0.21   In poor health  0.59  −0.53  0.23*  0.46  −0.78  0.59   Standardized cognitive score  0.97  −0.03  0.04  1.01  0.01  0.08   Impulsive  0.85  −0.16  0.12  0.68  −0.38  0.34   Less than high school  0.60  −0.51  0.09***  n.a.       GED  1.14  0.13  0.19  n.a.       Some college  1.38  0.32  0.18~  1.87  0.63  0.22**   College  2.06  0.72  0.26**  n.a.       Logged hourly wage rate  1.24  0.22  0.03***  1.26  0.23  0.12*   Missing hourly wage rate  0.47  −0.76  0.11***  0.79  −0.24  0.33   No work experience  0.15  −1.91  0.44***  0.19  −1.68  0.88~   Maternal ed.—less than high school  0.93  −0.08  0.14  1.32  0.28  0.42   Maternal ed.—more than high school  1.02  0.02  0.17  1.04  0.03  0.38   Maternal ed.—missing  1.04  0.04  0.21  1.05  0.05  0.72   Poor parental mental health  1.01  0.01  0.17  0.77  −0.26  0.36   Lived with both parents at age 14  0.96  −0.04  0.10  0.77  −0.26  0.28   Gender attitudes  0.83  −0.19  0.05***  0.79  −0.24  0.10*  Family constraints and resources             # of children under age 3  0.76  −0.27  0.11*  0.94  −0.07  0.26   # of children ages 3–5  0.87  −0.14  0.13  1.16  0.15  0.23   # of children ages 6 and over  0.94  −0.06  0.07  0.96  −0.05  0.09   Low birth weight baby  0.74  −0.30  0.19  0.90  −0.11  0.33   Domestic violence  1.17  0.15  0.21  1.00  0.00  0.60   Grandmother in the household  0.97  −0.03  0.12  1.02  0.02  0.27   Home owned by mother/family  0.95  −0.05  0.09  0.97  −0.03  0.15   Reports not enough money for bills  0.70  −0.36  0.12**  0.41  −0.90  0.24***   Reports money left after paying bills  1.07  0.07  0.10  1.72  0.54  0.20**  Father characteristics               Father’s age relative to mother’s  0.98  −0.02  0.01**  0.94  −0.06  0.02***   Same race as mother  0.67  −0.40  0.13**  0.75  −0.28  0.23   In poor health  1.32  0.28  0.17  0.91  −0.09  0.43   Reports drug use  1.52  0.42  0.16**  3.07  1.12  0.42**   Less than high school  1.21  0.19  0.14  1.63  0.49  0.33   GED  1.15  0.14  0.20  0.94  −0.06  0.48   Some college  0.83  −0.19  0.13  1.03  0.03  0.19   College  0.63  −0.46  0.24~  1.11  0.10  0.38   Criminal record  1.04  0.03  0.13  1.18  0.17  0.30   Weeks not employed  0.99  −0.01  0.00*  1.00  0.00  0.01   Logged hourly wage rate  0.93  −0.08  0.11  0.67  −0.40  0.21~   Missing hourly wage rate  0.80  −0.23  0.14  1.00  0.00  0.24   Gender attitudes  0.86  −0.15  0.05***  0.91  −0.10  0.12  Intercept    1.84  0.40    2.21  0.61  Log likelihooda  −1540.31  −313.25  Note: Reference group is unmarried cohabiting mothers. n.a. indicates not applicable. aThe reported loglikelihood is the average log likelihood of 20 imputed datasets. Standard errors are for the coefficients. Models adjust standard errors for the clustering by city. Data are from the Fragile Families and Child Wellbeing Study. Statistical significance levels are denoted as follows: ~ p < 0.10 * p < 0.05 ** p < 0.01 *** p < 0.001 (two-tailed) In addition to differences by marital status, I find that mothers with more human capital and black mothers have a higher likelihood of returning to work in the first year. The gender attitudes variable is predictive (B = −0.19, s.e. = 0.05, odds ratio = 0.83) of maternal employment; mothers with more gender-conservative attitudes have lower odds of working in the first year. Because the distribution of mothers’ gender attitudes does not differ much across family structure, gender attitudes cannot explain family structure differences in employment. Father’s gender attitudes are also a significant predictor of her employment; mothers whose partners have more traditional gender attitudes are less likely to work in the first year. Additional tests reveal that there is an interaction between father’s gender attitudes and family structure, such that fathers’ attitudes are only predictive of married mothers’ employment. Other analyses reveal no interactions between mother’s race and family structure, mother’s education and family structure, or mother’s gender attitudes and family structure. Results from the analysis of the subsample of mothers are similar to those for the full sample. The model shows no statistically significant differences in the odds of working with an infant between cohabiting and lone mothers (B = 0.29, s.e. = 0.39) but substantial differences between cohabiting and married mothers (B = −0.72, s.e. = 0.23, odds ratio = 0.48). My results from the full sample and the subsample9 are consistent with Han and colleagues’ (2008) findings from their analysis of mothers’ employment in the nine months following a birth; they found that unmarried mothers had a higher rate of labor force participation, controlling for education. Multivariate Results: Maternal Work Hours in the Five Years after a Birth In these data, cohabiting mothers are more likely to work in the first year after a birth than married mothers, but employment patterns may change as children age because infants place different demands on families than older children. Table 4 shows how family structure associates with hours of work per week. (Recall that the employment hours variables are divided by 10 to estimate the growth curve models.) Table 4. Results from Growth Curve Models Predicting Hours Worked per Week with Family Structure as Predictor Variables   Full sample  Subsample  Intercept  Slope  Intercept  Slope  Family structure           Married  −0.38***  0.07  −0.24  −0.06    (0.10)  (0.05)  (0.19)  (0.09)   Lone  −0.03  0.01  0.19  0.02    (0.08)  (0.05)  (0.20)  (0.10)  Mother’s characteristics           Age (years over 20)  −0.01~  −0.01  −0.02  0.00    (0.01)  (0.00)  (0.01)  (0.01)   Teen mother  −0.19*  0.12*  n.a.  n.a.    (0.09)  (0.06)       Race/ethnicity:            Black  0.53***  −0.02  n.a.  n.a.    (0.10)  (0.05)        Hispanic  0.29*  −0.06  0.17  −0.05    (0.11)  (0.06)  (0.19)  (0.10)   Immigrant  0.03  0.06  −0.10  0.16    (0.12)  (0.07)  (0.22)  (0.12)   In poor health  −0.19  −0.02  −0.34  −0.16    (0.13)  (0.08)  (0.35)  (0.17)   Std. cognitive score  −0.01  0.00  −0.09  0.07~    (0.04)  (0.02)  (0.08)  (0.04)   Impulsive  −0.25*  0.03  −0.13  −0.02    (0.10)  (0.06)  (0.26)  (0.13)   Own education:            Less than high school  −0.61***  0.26***  n.a.  n.a.    (0.09)  (0.05)        GED  −0.09  −0.04  n.a.  n.a.    (0.16)  (0.09)        Some college  0.33***  −0.12*  0.37*  −0.17*    (0.09)  (0.05)  (0.15)  (0.08)    College or more  0.52***  −0.15*  n.a.  n.a.    (0.14)  (0.07)       Logged hourly wage rate  0.20***  −0.05**  0.21**  −0.10*    (0.03)  (0.02)  (0.08)  (0.04)   Missing hourly wage rate  −0.57***  0.10  −0.75**  −0.03    (0.10)  (0.06)  (0.22)  (0.13)   No work experience  −0.37*  −1.26***  n.a.  n.a.    (0.15)  (0.09)       Maternal education:            Less than high school  −0.05  0.07  −0.10  0.10    (0.09)  (0.05)  (0.20)  (0.11)    More than high school  0.10  −0.07  −0.25  0.07    (0.08)  (0.05)  (0.19)  (0.10)    Unknown or missing  −0.34*  0.00  −0.62  0.12    (0.13)  (0.08)  (0.42)  (0.23)   Poor parental mental health  −0.07  0.03  −0.01  0.00    (0.09)  (0.05)  (0.20)  (0.11)   Lived with both parents at age 14  −0.02  −0.05  −0.32*  0.05    (0.07)  (0.04)  (0.14)  (0.07)   Gender attitudes  −0.11**  −0.02  −0.16*  −0.03    (0.03)  (0.02)  (0.08)  (0.04)  Family constraints and resources           No. of children under age 3  −0.31***  0.14**  −0.22  −0.03    (0.07)  (0.04)  (0.17)  (0.09)   No. of children ages 3–5  0.03  −0.01  0.18  −0.12    (0.07)  (0.04)  (0.16)  (0.08)   No. of children ages 6+  0.02  0.01  0.02  0.06    (0.04)  (0.02)  (0.09)  (0.04)   Low-birth-weight baby  −0.19~  0.06  0.06  −0.21    (0.10)  (0.06)  (0.30)  (0.17)   Domestic violence  −0.04  0.01  −0.50  0.24    (0.16)  (0.09)  (0.34)  (0.19)   Grandmother in household  −0.03  0.08~  −0.10  0.03    (0.08)  (0.05)  (0.21)  (0.11)   Home owned by mother/family  0.00  −0.06  −0.10  −0.08    (0.07)  (0.04)  (0.16)  (0.08)   Not enough money for bills  −0.31**  0.14*  −0.40  0.22    (0.10)  (0.06)  (0.27)  (0.15)   Money left after paying bills  0.03  −0.01  0.07  −0.02    (0.07)  (0.04)  (0.15)  (0.07)  Father’s characteristics           Father’s age relative to mother’s  −0.01  0.00  −0.02  0.01    (0.01)  (0.00)  (0.02)  (0.01)   Same race/ethnicity as mother  −0.31**  0.04  −0.29~  0.08    (0.10)  (0.05)  (0.17)  (0.09)   In poor health  0.12  −0.05  0.12  −0.13    (0.14)  (0.08)  (0.32)  (0.18)   Reports drug use  0.22~  −0.07  0.16  −0.15    (0.12)  (0.07)  (0.27)  (0.13)   Education:            Less than high school  0.09  −0.02  0.05  0.04    (0.09)  (0.05)  (0.20)  (0.11)    GED  0.03  0.06  −0.09  0.06    (0.13)  (0.08)  (0.31)  (0.18)    Some college  0.03  −0.07  0.09  −0.04    (0.09)  (0.05)  (0.18)  (0.09)    College or more  −0.43**  −0.03  −0.05  −0.08    (0.14)  (0.07)  (0.26)  (0.12)   Criminal record  0.03  0.04  0.24  −0.03    (0.08)  (0.05)  (0.18)  (0.09)   Weeks not employed last year  −0.01*  0.00  0.00  0.00    (0.00)  (0.00)  (0.01)  (0.00)   Logged hourly wage rate  0.04  −0.08~  −0.16  −0.05    (0.08)  (0.05)  (0.16)  (0.08)   Missing hourly wage rate  −0.01  0.00  0.12  0.01    (0.07)  (0.04)  (0.18)  (0.09)   Gender attitudes  −0.09*  −0.02  0.00  −0.15**    (0.04)  (0.02)  (0.09)  (0.05)  Intercept  2.62  0.73  2.81  0.74    (0.26)  (0.15)  (0.52)  (0.27)  Slope with intercept  −2.18  −0.16    (0.05)  (0.09)  Residual variance  1.26  0.53  1.39  0.50    (0.09)  (0.14)  (0.18)  (0.68)  X2(df)  79.98 (62)  67.61 (51)  RMSEA  0.01  0.02  BIC  38275.6  7930.85  CFI/TLI  0.993/0.977  0.967/0.899    Full sample  Subsample  Intercept  Slope  Intercept  Slope  Family structure           Married  −0.38***  0.07  −0.24  −0.06    (0.10)  (0.05)  (0.19)  (0.09)   Lone  −0.03  0.01  0.19  0.02    (0.08)  (0.05)  (0.20)  (0.10)  Mother’s characteristics           Age (years over 20)  −0.01~  −0.01  −0.02  0.00    (0.01)  (0.00)  (0.01)  (0.01)   Teen mother  −0.19*  0.12*  n.a.  n.a.    (0.09)  (0.06)       Race/ethnicity:            Black  0.53***  −0.02  n.a.  n.a.    (0.10)  (0.05)        Hispanic  0.29*  −0.06  0.17  −0.05    (0.11)  (0.06)  (0.19)  (0.10)   Immigrant  0.03  0.06  −0.10  0.16    (0.12)  (0.07)  (0.22)  (0.12)   In poor health  −0.19  −0.02  −0.34  −0.16    (0.13)  (0.08)  (0.35)  (0.17)   Std. cognitive score  −0.01  0.00  −0.09  0.07~    (0.04)  (0.02)  (0.08)  (0.04)   Impulsive  −0.25*  0.03  −0.13  −0.02    (0.10)  (0.06)  (0.26)  (0.13)   Own education:            Less than high school  −0.61***  0.26***  n.a.  n.a.    (0.09)  (0.05)        GED  −0.09  −0.04  n.a.  n.a.    (0.16)  (0.09)        Some college  0.33***  −0.12*  0.37*  −0.17*    (0.09)  (0.05)  (0.15)  (0.08)    College or more  0.52***  −0.15*  n.a.  n.a.    (0.14)  (0.07)       Logged hourly wage rate  0.20***  −0.05**  0.21**  −0.10*    (0.03)  (0.02)  (0.08)  (0.04)   Missing hourly wage rate  −0.57***  0.10  −0.75**  −0.03    (0.10)  (0.06)  (0.22)  (0.13)   No work experience  −0.37*  −1.26***  n.a.  n.a.    (0.15)  (0.09)       Maternal education:            Less than high school  −0.05  0.07  −0.10  0.10    (0.09)  (0.05)  (0.20)  (0.11)    More than high school  0.10  −0.07  −0.25  0.07    (0.08)  (0.05)  (0.19)  (0.10)    Unknown or missing  −0.34*  0.00  −0.62  0.12    (0.13)  (0.08)  (0.42)  (0.23)   Poor parental mental health  −0.07  0.03  −0.01  0.00    (0.09)  (0.05)  (0.20)  (0.11)   Lived with both parents at age 14  −0.02  −0.05  −0.32*  0.05    (0.07)  (0.04)  (0.14)  (0.07)   Gender attitudes  −0.11**  −0.02  −0.16*  −0.03    (0.03)  (0.02)  (0.08)  (0.04)  Family constraints and resources           No. of children under age 3  −0.31***  0.14**  −0.22  −0.03    (0.07)  (0.04)  (0.17)  (0.09)   No. of children ages 3–5  0.03  −0.01  0.18  −0.12    (0.07)  (0.04)  (0.16)  (0.08)   No. of children ages 6+  0.02  0.01  0.02  0.06    (0.04)  (0.02)  (0.09)  (0.04)   Low-birth-weight baby  −0.19~  0.06  0.06  −0.21    (0.10)  (0.06)  (0.30)  (0.17)   Domestic violence  −0.04  0.01  −0.50  0.24    (0.16)  (0.09)  (0.34)  (0.19)   Grandmother in household  −0.03  0.08~  −0.10  0.03    (0.08)  (0.05)  (0.21)  (0.11)   Home owned by mother/family  0.00  −0.06  −0.10  −0.08    (0.07)  (0.04)  (0.16)  (0.08)   Not enough money for bills  −0.31**  0.14*  −0.40  0.22    (0.10)  (0.06)  (0.27)  (0.15)   Money left after paying bills  0.03  −0.01  0.07  −0.02    (0.07)  (0.04)  (0.15)  (0.07)  Father’s characteristics           Father’s age relative to mother’s  −0.01  0.00  −0.02  0.01    (0.01)  (0.00)  (0.02)  (0.01)   Same race/ethnicity as mother  −0.31**  0.04  −0.29~  0.08    (0.10)  (0.05)  (0.17)  (0.09)   In poor health  0.12  −0.05  0.12  −0.13    (0.14)  (0.08)  (0.32)  (0.18)   Reports drug use  0.22~  −0.07  0.16  −0.15    (0.12)  (0.07)  (0.27)  (0.13)   Education:            Less than high school  0.09  −0.02  0.05  0.04    (0.09)  (0.05)  (0.20)  (0.11)    GED  0.03  0.06  −0.09  0.06    (0.13)  (0.08)  (0.31)  (0.18)    Some college  0.03  −0.07  0.09  −0.04    (0.09)  (0.05)  (0.18)  (0.09)    College or more  −0.43**  −0.03  −0.05  −0.08    (0.14)  (0.07)  (0.26)  (0.12)   Criminal record  0.03  0.04  0.24  −0.03    (0.08)  (0.05)  (0.18)  (0.09)   Weeks not employed last year  −0.01*  0.00  0.00  0.00    (0.00)  (0.00)  (0.01)  (0.00)   Logged hourly wage rate  0.04  −0.08~  −0.16  −0.05    (0.08)  (0.05)  (0.16)  (0.08)   Missing hourly wage rate  −0.01  0.00  0.12  0.01    (0.07)  (0.04)  (0.18)  (0.09)   Gender attitudes  −0.09*  −0.02  0.00  −0.15**    (0.04)  (0.02)  (0.09)  (0.05)  Intercept  2.62  0.73  2.81  0.74    (0.26)  (0.15)  (0.52)  (0.27)  Slope with intercept  −2.18  −0.16    (0.05)  (0.09)  Residual variance  1.26  0.53  1.39  0.50    (0.09)  (0.14)  (0.18)  (0.68)  X2(df)  79.98 (62)  67.61 (51)  RMSEA  0.01  0.02  BIC  38275.6  7930.85  CFI/TLI  0.993/0.977  0.967/0.899  Note: Standard errors are reported in parentheses. Reference group is unmarried cohabiting mothers. All models also include variables identifying city of residence. Data are from the Fragile Families and Child Wellbeing Study. Statistical significance levels are as follows: ~ p < 0.10 * p < 0.05 ** p < 0.01 *** p < 0.001 I find sizable differences in the intercept between married mothers and cohabiting mothers. The intercept coefficient on the married variable (B = −0.38, s.e. = 0.10, p < 0.001) indicates that one year after the birth, married mothers worked an average of almost four hours fewer per week than cohabiting mothers, all other factors equal. The slope coefficient on the married variable is close to zero and is not statistically significant (B = 0.07; s.e. = 0.05), indicating that differences between married and cohabiting mothers do not change over time. Additionally, the model shows no differences in work hours between lone and cohabiting mothers, other factors equal (intercept: B = −0.03, s.e. = 0.08; slope: B = 0.01, s.e. = 0.05). Consistent with previous research, I find that age, race/ethnicity, impulsivity, education, and hourly pay rate are all significant predictors (p < 0.05 level) of work hours. All else equal, black and Hispanic mothers work more hours than white, non-Hispanic, or other mothers, and mothers with more human capital work more hours than other mothers. Mothers with more conservative gender attitudes have a lower intercept (B = −0.11, s.e. = 0.03, p < 0.01), indicating that mothers who have gender attitudes that are one standard deviation more conservative than the average mother work a modest one hour less per week than other mothers. Mothers with more children under age three at home work less in the first year (intercept: B = −0.31, s.e. = 0.07) but increase their employment at a faster rate in subsequent years (slope: B = 0.14, s.e. = 0.04). Father characteristics—including having a college degree, weeks of unemployment, and gender attitudes—are also predictive of mothers’ employment hours, and this suggests that researchers should remain attentive to the influence of men on their partners’ employment. Additionally, mothers of the same race/ethnicity as their partner work an average of three hours less per week. The “Slope with Intercept” parameter shows that mothers who start with greater employment hours increase their hours more slowly than mothers who start with fewer employment hours. In additional analyses (available upon request), I tested for interactions between family structure and mother, father, and family characteristics. I found large and statistically significant interactions between the married variable and (mother’s) black racial identity as well as between the married variable and father’s college education. Married black women worked considerably more hours per week than married women of other racial groups, and married women with college-educated husbands worked considerably less than other married women. Notably, both race and education are strongly correlated with divorce rates, and the directions of these interaction effects are consistent with what one would expect if the association of marriage with employment is partially related to family stability. The subsample results for family structure are somewhat different than those from the full sample. The intercept coefficient for marriage (B = −0.24) is in the predicted direction, showing that married mothers work an average of two and a half hours less than cohabiting mothers in the first year, but this difference is not statistically significant at conventional levels. The difference between cohabiting and lone mothers is also not statisticallly significant, though the coefficient is large (B = 0.19), indicating that lone mothers work an average of two hours more than cohabiting mothers in the first year. The slope coefficients for both married and lone mothers are close to zero and not statistically significant, suggesting that family structure differences are stable over time. In summary, the findings for the full sample show that married mothers work fewer hours per week than cohabiting mothers, and that cohabiting and lone mothers have similar employment hours. For the subsample, the coefficient for married is in the expected direction but is not statistically significant and is somewhat smaller in magnitude. Additionally, for the subsample, there is suggestive evidence that lone and cohabiting mothers may differ in their employment hours. Changes in Family Structure and Maternal Employment Trajectories Among the parents in the FFCWS, there are considerable changes in family structure in the five years following the birth. Most married mothers (80 percent) are still married to the baby’s father, but most cohabiting mothers (77 percent) and lone mothers (64 percent) have experienced at least one family structure change. Table 5 shows the share of the sample following each trajectory. Cohabiting mothers’ family structure trajectories are roughly split between the four possibilities of stably cohabiting, marrying, transitioning to living alone, and experiencing multiple changes. Of the lone mothers, 6 percent reported a marriage (and no other change), 24 percent reported a cohabiting partner (and no other change), and 33 percent reported multiple family changes during the first five years after the birth. Table 5. Distribution of Family Structure Trajectories Family structure trajectory  %  Stably cohabiting  8.2  Cohabiting to married  7.6  Cohabiting to lone  9.8  Cohabiting, multiple changes  9.6  Stably married  20.3  Married to divorced  4.9  Stably lone  14.4  Lone to cohabiting  9.3  Lone to married  2.5  Lone, multiple changes  13.3  Family structure trajectory  %  Stably cohabiting  8.2  Cohabiting to married  7.6  Cohabiting to lone  9.8  Cohabiting, multiple changes  9.6  Stably married  20.3  Married to divorced  4.9  Stably lone  14.4  Lone to cohabiting  9.3  Lone to married  2.5  Lone, multiple changes  13.3  Note: Data are from the Fragile Families and Child Wellbeing Study. How are these family structure trajectories expected to correlate with maternal employment? If stability is the main mechanism explaining married mothers’ lower levels of employment, one would expect both stably married mothers and stably cohabiting mothers to have lower levels of employment, all other factors equal. If mothers can predict relationship stability and anticipate changes, one might expect to see divergence on both intercepts and slopes by relationship trajectories. Under this condition, married mothers who eventually divorce are expected to have employment levels similar to unmarried mothers, and cohabiting mothers who eventually marry are expected to have employment similar to stably married mothers. Alternatively, if the mechanism is not economic security or if mothers cannot predict relationship stability, one would expect all married mothers to look similar in the earliest period, with changes in the slopes appearing as mothers divorce in the later periods. Similarly, one would expect all cohabiting mothers to look similar in the intercept levels and differences to emerge in the slopes as mothers experience changes. To test these predictions, I rerun the growth curve models of employment hours using the same full set of covariates in table 4 but using family structure trajectories instead of static measures of family structure. Table 6 shows the coefficients for the family structure trajectory variables. The model shows that stably married mothers work almost seven hours less than stably cohabiting mothers in the first year (intercept: B = −0.67, s.e. = 0.14, p < 0.001), and the gap between these mothers does not close over time (slope: B = 0.10, s.e. = 0.08). Notably, married mothers who later divorce have employment levels in the first year that are similar to stably cohabiting mothers (intercept: B = 0.07, s.e. = 0.18) and dissimilar from stably married mothers. The slope coefficient shows that mothers who divorce increase their hours of work over subsequent years (slope: B = 0.24, s.e. = 0.11, p < 0.05). Cohabiting mothers who marry look more similar to stably married mothers than to stably cohabiting mothers; they report fewer hours of work in the first year (intercept: B = −0.37, s.e. = 0.15, p < 0.05) and no convergence toward stably cohabiting mothers in later years (slope: B = 0.07, s.e. = 0.09). Cohabiting mothers whose unions dissolve increase their work hours over time (slope: B = 0.16, s.e. = 0.08, p < 0.05). Table 6. Selected Results from Growth Curve Models Predicting Hours of Work per Week with Family Structure Trajectories as Predictor Variables   Intercept  Slope  Stably cohabiting  reference group  Cohabiting to married  −0.37*  0.07    (0.15)  (0.09)  Cohabiting to lone  −0.04  0.16*    (0.14)  (0.08)  Cohabiting, multiple changes  0.08  0.07    (0.15)  (0.09)  Stably married  −0.67***  0.10    (0.14)  (0.08)  Married to divorced  0.07  0.24*    (0.18)  (0.11)  Stably lone  −0.04  0.07    (0.14)  (0.08)  Lone to cohabiting  −0.15  0.06    (0.15)  (0.09)  Lone to married  0.16  −0.07    (0.23)  (0.12)  Lone, multiple changes  −0.11  0.17*    (0.14)  (0.08)  Intercept  2.61***  0.63***    (0.27)  (0.16)  Slope with intercept  −0.22    (0.05)  X2(df)  84.92 (69)  RMSEA  0.01  BIC  38303.2  CFI/TLI  0.994/0.980    Intercept  Slope  Stably cohabiting  reference group  Cohabiting to married  −0.37*  0.07    (0.15)  (0.09)  Cohabiting to lone  −0.04  0.16*    (0.14)  (0.08)  Cohabiting, multiple changes  0.08  0.07    (0.15)  (0.09)  Stably married  −0.67***  0.10    (0.14)  (0.08)  Married to divorced  0.07  0.24*    (0.18)  (0.11)  Stably lone  −0.04  0.07    (0.14)  (0.08)  Lone to cohabiting  −0.15  0.06    (0.15)  (0.09)  Lone to married  0.16  −0.07    (0.23)  (0.12)  Lone, multiple changes  −0.11  0.17*    (0.14)  (0.08)  Intercept  2.61***  0.63***    (0.27)  (0.16)  Slope with intercept  −0.22    (0.05)  X2(df)  84.92 (69)  RMSEA  0.01  BIC  38303.2  CFI/TLI  0.994/0.980  Note: Standard errors are reported in parentheses. Reference group is stably cohabiting mothers. Model includes all of the covariates listed in table 1 and includes variables identifying city of residence. Data are from Fragile Families and Child Wellbeing Study. Statistical significance levels are as follows: ~ p < 0.10 * p < 0.05 ** p < 0.01 *** p < 0.001 (two-tailed) The findings from the models of employment by relationship trajectory are consistent with the hypothesis that mothers’ employment is sensitive to perceived relationship stability, but also that marital status is relevant, even among stably partnered mothers. Stably cohabiting mothers work considerably more hours per week than stably married mothers, suggesting that marital status matters, even among couples with stable unions. These findings suggest that differences between married and cohabiting mothers’ employment are not solely about experiences of union instability. Robustness Checks My results (for both logistic regression models and growth curve models) are robust to many model specifications. I obtain the same pattern of findings regarding family structure in models that 1) omit mother’s hourly pay rate and work experience; 2) omit father’s hourly pay rate; 3) omit the variables related to perceived financial need; 4) omit the variables measuring partner homogamy (age difference, same race); and 5) include variables to indicate subsequent births. For hours worked per week, I compare the results from growth curve models to those from OLS regression models that separately predict hours of employment per week at each timepoint; the pattern of results remains similar. Discussion Among a sample of families in US urban areas with a child born between 1998 and 2000, I find that cohabiting mothers are more likely to work in the first year after a birth and work more hours per week in the first five years after a birth than married mothers with similar human capital and demographic characteristics. Contrary to what one would expect if the presence of a partner were driving the relationship between family structure and maternal employment, I find no significant differences in employment between cohabiting and lone mothers. I speculate that cohabiting and lone mothers have similar employment levels because cohabiting mothers are not confident that they can rely on their partners’ income given high union dissolution rates. Thus, cohabiting mothers have similar employment levels to what they would have if they were unpartnered lone mothers. Marriage is associated with lower employment levels, but married women have greater education and wage rates than unmarried women on average. Thus, the employment-promoting effects of human capital almost entirely offset the lower employment associated with marriage. This may explain why aggregate employment rates for married and unmarried mothers of young children in the United States are so similar. Accounting for differences in human capital reveals that among mothers in urban areas, married mothers still work considerably less than unmarried mothers, including cohabiting mothers. The magnitude of family structure differences in maternal employment is fairly large. All other factors being equal, married mothers in this sample are only about half as likely as cohabiting or lone mothers to return to work in the first year after a birth. Is this difference in employment likely to affect child or maternal well-being? Several studies suggest small or no effects of maternal employment during infancy on child well-being for the total population (e.g., Baker and Milligan 2015), but new research suggests negative effects for children in low-income families (Herbst 2014). Additionally, greater maternal employment hours when children are infants are associated with worse maternal health and greater levels of parenting stress (Chatterji, Markowitz, and Brooks-Gunn 2013). Thus, family structure differences in maternal employment during infancy may contribute to family structure inequalities for children in low-income families and to inequalities in mothers’ well-being. Maternal employment differences by family structure continue through early childhood. My estimates suggest that stably cohabiting urban mothers work an average of 6.7 hours more per week than similar stably married urban mothers for all five years after the birth. In terms of magnitude, a difference of seven hours per week is greater than the difference in employment hours between mothers with a high school degree and mothers with a college degree in this sample (5.2 hours per week). Previous research yields competing predictions as to why and how family structure may affect maternal employment. Possible mechanisms include family demographic characteristics (such as number of children), family-related constraints on employment (such as domestic violence), a woman’s own gender attitudes, and her partner’s characteristics and gender attitudes. I find that none of these mechanisms explain much of the marriage difference in maternal employment patterns among mothers in the FFCWS. Instead, I hypothesize that marriage affects maternal employment by providing mothers with more long-term economic security and general stability than cohabitation offers mothers in the contemporary United States. Indeed, I find that characteristics that are strongly associated with marital stability—childhood family structure and racial homogamy among parents—are also strongly associated with maternal employment. Scholars can imagine many other possible tests of my hypothesis that cohabiting mothers work more than married mothers as a hedge against union dissolution and economic deprivation. Lacking an exogenous shock in family structure or random assignment to family structures, all possible tests fall short of the ideal for conclusive evidence. This study examines employment for mothers in urban areas, and mothers in rural areas may have different employment patterns. The data in this analysis were collected from 1998 to 2005, and the gap in employment hours between cohabiting and married mothers may change over time as cohabiting parenthood becomes more institutionalized. Additionally, the FFCWS dataset has limited information on wealth and attitudes regarding financial security, and no items on preferences for employment levels. Better data on these factors would allow for a more thorough test of my hypotheses. Detailed information on maternal employment collected at more frequent intervals would also improve the analysis by reducing reporting errors. To my knowledge, no existing dataset overcomes all of these limitations. Future research might fruitfully investigate employment differences between cohabiting and married mothers in other industrialized countries. The share of children born in cohabiting unions varies substantially across countries, as do the legal responsibilities of cohabitors to each other upon union dissolution (Perelli-Harris and Sánchez Gassen 2012). Moreover, the poverty rates associated with single motherhood vary substantially, largely because of differences in social safety net policies (Brady and Burroway 2012). In this paper, I argue that in the US context, the relatively high employment levels of cohabiting mothers are a form of insurance against high levels of union dissolution and the economic precarity of single-mother families. In countries with stronger social safety nets for lone mothers, more formalized legal obligations between cohabiting partners, or with less union instability among cohabiting couples, I expect that there would be smaller differences in employment levels between married and cohabiting mothers. My findings suggest that, in a key aspect of contemporary family life, cohabiting parenthood differs considerably from married parenthood in the US context. My findings of differences in maternal employment by marital status (adjusted for human capital differences) shed light on important issues in stratification, including economic inequality among women and their children. To the extent that marital childbearing is becoming more selective of advantaged women (McLanahan and Percheski 2008), women’s opportunities to optimally combine paid work and caretaking for their particular career goals and family situations may be becoming more unequal. This inequality may have consequences for women’s health, happiness, and well-being as well as that of their children and partners. Notes 1 The Earned Income Tax Credit provides a considerable income boost to single-mother families but is contingent on maternal employment. 2 I used the ICE command in STATA with an estimation model that included all of the variables included in the analysis plus two additional variables related to fertility history. The imputations were performed using the full sample. 3 The percentage of mothers with top-coded values is very low: 0.7 percent at one year, 1.1 percent at 30 months, and 1.9 percent at 60 months. 4 Thirteen of the 3,132 mothers in the sample report living with a female romantic partner at one or more survey waves. 5 Because the number of married mothers who divorce and subsequently re-partner is small, I did not further disaggregate this category. 6 Survival models predicting the month when the mother returned to work show the same pattern of results regarding family structure as do the logit models. 7 I used MPLUS software and a maximum likelihood parameter estimator (MLR) with TYPE = IMPUTATION to accommodate the imputed data. I took an inductive approach to specifying the employment trajectories. 8 To determine model fit, I examined the chi-square, Bayesian Information Criteria (BIC), comparative fit index (CFI), Tucker-Lewis index (TLI), and root-mean-square error of approximation (RMSEA). 9 Using the data from complete cases only (n = 1,727), the logit models predicting work in the first year for the full sample had a smaller coefficient for the married variable (B = −0.26) and miss conventional cutoffs for statistical significance, but the coefficient was in the expected direction. For the subsample, the estimates with complete cases only (n = 394) have coefficients for the married variable that are almost identical to the coefficients from the models with complete and imputed data. In both the full sample and the subsample, the coefficients for the lone variable were close to zero and not statistically significant. About the Author Christine Percheski is an Assistant Professor in Sociology and a Faculty Fellow at the Institute for Policy Research at Northwestern University. Her research addresses the intersection of family demography and social stratification, focusing on the United States. More specifically, she examines 1) changes in fertility, cohabitation, marriage, and women’s employment; and 2) how these family characteristics are intertwined with economic and health inequalities. Supplementary Material Supplementary material is available at Social Forces online. References Baker, Michael, and Kevin Milligan. 2015. “ Maternity Leave and Children’s Cognitive and Behavioral Development.” Journal of Population Economics  28( 2): 373– 91. Google Scholar CrossRef Search ADS   Bollen, Kenneth, and Patrick Curran. 2006. Latent Curve Models: A Structural Equation Perspective . Hoboken, NJ: John Wiley and Sons. Brady, David, and Rebekah Burroway. 2012. “ Targeting, Universalism, and Single-Mother Poverty: A Multilevel Analysis across 18 Affluent Democracies.” Demography  49( 2): 719– 46. Google Scholar CrossRef Search ADS PubMed  Bureau of Labor Statistics. 2017. Women in the Labor Force: A Databook. Report No. 1065. US Department of Labor. Chatterji, Pinka, Sara Markowitz, and Jeanne Brooks-Gunn. 2013. “ Effects of Early Maternal Employment on Maternal Health and Wellbeing.” Journal of Population Economics  26( 1): 285– 301. Google Scholar CrossRef Search ADS PubMed  Cohen, Philip N., and Suzanne M. Bianchi. 1999. “ Marriage, Children, and Women’s Employment: What Do We Know?” Monthly Labor Review  122: 22– 31. Corman, Hope, Kelly Noonan, and Nancy E. Reichman. 2005. “ Mothers’ Labor Supply in Fragile Families: The Role of Child Health.” Eastern Economic Journal  31( 4): 601– 16. Curran, Patrick, Khawla Obeidat, and Diane Losardo. 2010. “ Twelve Frequently Asked Questions about Growth Curve Modeling.” Journal of Cognition and Development  11( 2): 121– 36. Google Scholar CrossRef Search ADS PubMed  Davis, Shannon N., and Theodore N. Greenstein. 2009. “ Gender Ideology: Components, Predictors, and Consequences.” Annual Review of Sociology  35: 87– 105. Google Scholar CrossRef Search ADS   Davis, Shannon N., Theodore N. Greenstein, and Jennifer P. Gerstein Marks. 2007. “ Effects of Union Type on Division of Household Labor: Do Cohabiting Men Really Perform More Housework?” Journal of Family Issues  28( 9): 1246– 72. Google Scholar CrossRef Search ADS   Fields, Jason. 2003. “Children’s Living Arrangements and Characteristics: March 2002.” Current Population Reports P20-547. Gemmill, Alison, and Laura D. Lindberg. 2013. “ Short Interpregnancy Intervals in the United States.” Obstetrics and Gynecology  122( 1): 64– 71. Google Scholar CrossRef Search ADS PubMed  Goodwin, Paula Y., William D. Mosher, and Anjani Chandra. 2010. “ Marriage and Cohabitation in the United States: A Statistical Portrait Based on Cycle 6 (2002) of the National Survey of Family Growth.” Vital and Health Statistics  23( 28): 1– 45. Han, Wen-Jui, Christopher J. Ruhm, Jane Waldfogel, and Elizabeth Washbrook. 2008. “ The Timing of Mothers’ Employment after Childbirth.” Monthly Labor Review  131( 6): 15– 27. Google Scholar PubMed  Herbst, Chris. 2014. “Are Parental Welfare Work Requirements Good for Disadvantaged Children? Evidence from Age-of-Youngest-Child Exemptions.” Discussion Paper No. 8485. Bonn, Germany: Institute for the Study of Labor. Kalenkoski, Charlene M., David C. Ribar, and Leslie S. Stratton. 2007. “ The Effect of Family Structure on Parents’ Child Care Time in the United States and the United Kingdom.” Review of Economics of the Household  5( 4): 353– 84. Google Scholar CrossRef Search ADS   Kan, Man Yee. 2007. “ Work Orientation and Wives’ Employment Careers An Evaluation of Hakim’s Preference Theory.” Work and Occupations  34( 4): 430– 62. Google Scholar CrossRef Search ADS   Kennedy, Sheela, and Larry Bumpass. 2008. “ Cohabitation and Children's Living Arrangements: New Estimates from the United States.” Demographic Research  19: 1663. Google Scholar CrossRef Search ADS PubMed  Kenney, Catherine T. 2006. “ The Power of the Purse: Allocative Systems and Inequality in Couple Households.” Gender & Society  20( 3): 354– 81. Google Scholar CrossRef Search ADS   Kenney, Catherine T., and Sara S. McLanahan. 2006. “ Why Are Cohabiting Relationships More Violent Than Marriages?” Demography  43( 1): 127– 40. Google Scholar CrossRef Search ADS PubMed  Lee, Sunhwa, Melissa Sills, and Gi-Taik Oh. 2002. “Disabilities among Children and Mothers in Low-Income Families.” Institute for Women’s Policy Research, Publication #D449. Livermore, Michelle M., and Rebecca S. Powers. 2006. “ Employment of Unwed Mothers: The Role of Government and Social Support.” Journal of Family and Economic Issues  27( 3): 479– 94. Google Scholar CrossRef Search ADS   Manning, Wendy D., Pamela J. Smock, and Debarun Majumdar. 2004. “ The Relative Stability of Cohabiting and Marital Unions for Children.” Population Research and Policy Review  23( 2): 135– 59. Google Scholar CrossRef Search ADS   Martin, Joyce A., Brady E. Hamilton, Michelle J. Osterman, et al.  . 2015. “ Births: Final Data for 2013.” National Vital Statistics Reports  64( 1): 1– 68. McLanahan, Sara, and Christine Percheski. 2008. “ Family Structure and the Reproduction of Inequalities.” Annual Review of Sociology  34: 257– 76. Google Scholar CrossRef Search ADS   Meadows, Sarah O., Sara S. McLanahan, and Jeanne Brooks-Gunn. 2008. “ Stability and Change in Family Structure and Maternal Health Trajectories.” American Sociological Review  73( 2): 314– 34. Google Scholar CrossRef Search ADS PubMed  Monte, Lindsay M., and Renee R. Ellis. 2014. “Fertility of Women in the United States: June 2012.” Current Population Reports, P20-575. US Census Bureau, Washington, DC. Musick, Kelly, and Katherine Michelmore. 2015. “ Change in the Stability of Marital and Cohabiting Unions Following the Birth of a Child.” Demography  52( 5): 1463– 85. Google Scholar CrossRef Search ADS PubMed  Nepomnyaschy, Lenna, and Irwin Garfinkel. 2010. “ Child Support Enforcement and Fathers’ Contributions to Their Nonmarital Children.” Social Services Review  84( 3): 341– 80. Google Scholar CrossRef Search ADS   Ono, Hiromi, and Rosemary Yeilding. 2009. “ Marriage, Cohabitation, and Childcare: The US and Sweden.” Social Indicators Research  93( 1): 137– 40. Google Scholar CrossRef Search ADS   Perelli-Harris, Brienna, and Nora Sánchez Gassen. 2012. “ How Similar Are Cohabitation and Marriage? Legal Approaches to Cohabitation across Western Europe.” Population and Development Review  38( 3): 435– 67. Google Scholar CrossRef Search ADS   Raley, Sara B., Marybeth J. Mattingly, and Suzanne M. Bianchi. 2006. “ How Dual Are Dual Income Couples? Documenting Change from 1970 to 2001.” Journal of Marriage and Family  68( 1): 11– 28. Google Scholar CrossRef Search ADS   Reichman, Nancy E., Julien O. Teitler, Irwin Garfinkel, and Sara S. McLanahan. 2001. “ Fragile Families: Sample and Design.” Children and Youth Services Review  23( 4– 5): 303– 26. Google Scholar CrossRef Search ADS   Showalter, Kathryn. 2016. “ Women’s Employment and Domestic Violence: A Review of the Literature.” Aggression and Violent Behavior  31: 37– 47. Google Scholar CrossRef Search ADS   Sorensen, Elaine, and Ariel Hill. 2004. “ Single Mothers and Their Child-Support Receipt: How Well Is Child-Support Enforcement Doing?” Journal of Human Resources  39( 1): 135– 54. Google Scholar CrossRef Search ADS   Thornton, Arland, and Linda Young-DeMarco. 2001. “ Four Decades of Trends in Attitudes Toward Family Issues in the United States: The 1960s through the 1990s.” Journal of Marriage and Family  63( 4): 1009– 37. Google Scholar CrossRef Search ADS   Ventura, Stephanie J., and Christine A. Bachrach. 2000. “ Nonmarital Childbearing in the United States: 1940–99.” National Vital Statistics Report  48( 16): 1– 39. Wu, Huijing. 2017. “Trends in Births to Single and Cohabiting Mothers, 1980–2014.” Family Profiles. Bowling Green, OH: National Center for Family and Marriage Research. Author notes Address for correspondence: Christine Percheski, Northwestern University, 1810 Chicago Avenue, Evanston, IL 60208; telephone: 847-491-2697 (office); e-mail: c-percheski@northwestern.edu © The Author(s) 2018. Published by Oxford University Press on behalf of the University of North Carolina at Chapel Hill. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Social Forces Oxford University Press

Marriage, Family Structure, and Maternal Employment Trajectories

Loading next page...
 
/lp/ou_press/marriage-family-structure-and-maternal-employment-trajectories-xcgd20vWli
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the University of North Carolina at Chapel Hill. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
ISSN
0037-7732
eISSN
1534-7605
D.O.I.
10.1093/sf/sox094
Publisher site
See Article on Publisher Site

Abstract

Abstract Previous studies of maternal employment have focused on marital status differences, but the rise in nonmarital cohabiting parenthood problematizes the simple dichotomy between married and unmarried mothers. Theory and previous research yield mixed predictions as to whether cohabiting mothers’ employment will more closely resemble that of married mothers or lone unmarried mothers. Using data from the Fragile Families and Child Wellbeing Study, I examine how maternal employment varies across family structures (married parents, cohabiting unmarried parents, and lone unmarried mothers) in the five years after a birth for mothers living in urban areas in the United States. Descriptive statistics show few differences in maternal employment patterns by family structure. Controlling for human capital, however, I find that cohabiting mothers return to work earlier and work more than married mothers. Cohabiting mothers and lone mothers show very similar employment patterns. Additionally, cohabiting mothers who later marry have employment trajectories that are similar to married mothers, whereas married mothers who divorce increase their employment hours. Family characteristics, partner characteristics, and gender attitudes do not explain employment differences between married and cohabiting mothers. I speculate that cohabiting mothers work more than married mothers as a hedge against economic deprivation given high union dissolution rates for cohabiting couples. In the United States and across Europe, the normative context for sexual intimacy and childrearing has been shifting away from marriage. More than half of American women in their early forties have cohabited (Goodwin, Mosher, and Chandra 2010), and more than four in 10 births in 2013 were to unmarried mothers (Martin et al. 2015). Marriage’s monopoly on sex and reproduction has declined, and unmarried cohabitation has partially replaced it. Yet, the majority of young adults aspire to marry (Thornton and Young-DeMarco 2001), suggesting that marrriage has retained its symbolic significance. Is marriage merely symbolically important? Among families with children, are there differences by family structure in how families organize fundamental tasks, such as breadwinning and child-rearing? Mothers’ time in paid employment may be a particularly important aspect of family organization because it impacts family income, as well as potentially child development and the gendered division of housework. For decades in the United States, unmarried mothers were employed at much higher rates than married mothers, but these long-standing differences in maternal employment trends started to disappear in the early 1990s (Cohen and Bianchi 1999). In 2015, the percentage of US mothers with children under age six who were employed was very similar across marital statuses, with 59.5 percent of married mothers (with spouse present) employed compared with 60.5 percent of mothers of other marital statuses (Bureau of Labor Statistics 2017, table 6). The convergence of these trends lends itself to multiple interpretations, including that marital status is no longer a meaningful axis of comparison. If this is the case, the more important distinction may be between mothers with a resident partner and mothers without a partner. In this analysis, I examine how maternal employment differs by both marital status and coresidency (two key features of family structure) and which factors account for any observed family structure differences in maternal employment. Family structure may affect maternal employment through many possible mechanisms. These include differences in family income needs, access to instrumental and social support for the mother’s employment, the mother’s own preferences for paid work, and the mother’s perceived security to pursue her preferred level of employment. Additionally, family structure may affect employment through the influence of husband or partner preferences for the mother’s employment level. Alternatively, family structure may have no causal effect and may merely be a marker of pre-existing differences among women—such as differences in human capital characteristics and gender attitudes—that are both associated with family structure and affect maternal employment. Such characteristics may affect a woman’s opportunities and compensation for work as well as her motivation and preference for paid work. Whether family structure affects women’s employment or is merely a marker of other differences among women is an important question for scholars and policymakers. In the United States, women’s earnings now account for a sizable share of family income in married-couple families (Raley, Mattingly, and Bianchi 2006) and are often the sole source of income for single-mother families. Paid employment also potentially provides mothers with nonmonetary rewards, including social ties and a source of identity. Employment, however, can have substantial costs for mothers, including foregone time with children, high childcare costs, and stress from balancing mother and worker roles. Many mothers work full-time by necessity to support their families, even if their employment interferes with the care of their children or their own well-being. Thus, whether and to what extent family structure—here defined as married motherhood, cohabiting motherhood, or lone motherhood—affects women’s employment are compelling unanswered questions in family demography and social stratification research with implications for public policy. In this paper, I address the following research questions: How does family structure correlate with maternal employment among mothers in urban areas in the United States following a birth? Are the employment patterns of cohabiting mothers more similar to those of married mothers or lone mothers? Given the obvious impossibility of a random assignment design, researchers are left with the challenge of disentangling causal effects from selection effects and identifying mechanisms through which family structure influences maternal employment. I use a two-pronged approach that capitalizes on the rich data in the Fragile Families and Child Wellbeing Survey (FFCWS), a study of families with a birth in 1999 or 2000 in 20 US cities, to examine how maternal employment varies by family structure. First, I model the employment trajectories of mothers by family structure and try to differentiate between employment differences that might be caused by family structure versus those that can be attributed to other factors, such as differences in human capital. To do this, I use models with covariate adjustments for an unusually extensive set of characteristics of mothers, their partners, and their families. Second, I look at the employment patterns of mothers who change family structures to identify how their employment changes. Maternal Employment Patterns Family Structure Influences on Maternal Employment Unmarried cohabiting parenthood is a relatively new family form, and theory suggests both reasons to expect that cohabiting mothers’ employment would be similar to married mothers and reasons why it would differ. Below, I discuss first why married mothers are expected to have lower employment levels than lone mothers (unmarried mothers without a cohabiting partner). Second, I discuss expectations regarding cohabiting mothers’ employment (in the US context). Marriage is expected to be associated with lower levels of maternal employment than lone motherhood because having a husband provides a source of income and economic security independent of a woman’s own earnings. In contrast, in the post-1996 welfare reform period, lone mothers in the United States have few sources of income beyond their own employment1 and the generosity of family and friends. Income from government programs is a temporary source of family support, and most mothers receiving income through the Transitional Assistance for Needy Families (TANF) program are required to be employed to receive benefits. Although all fathers are legally responsible for supporting their children, receiving timely child support payments is far from a universal experience (Sorensen and Hill 2004), and the average child support payment is quite small (Nepomnyaschy and Garfinkel 2010). The legal status of marriage and the public nature of the commitment grant women both legal and socially recognized claims on their husbands’ incomes. Although high divorce rates undermine some of the guarantees of marriage, previously married women are more likely to obtain child support than never-married women (Sorensen and Hill 2004). This promise of economic support may allow married women to work part-time or take periods of absence from paid employment without incurring high risks of poverty. Thus, married mothers who desire lower levels of employment—perhaps becuase of their children’s developmental needs or their preferences for time-intensive homemaking practices—may have a better chance of realizing their preferences than unmarried mothers with similar preferences because of their stronger legal and social claims on their partners’ incomes. Alternatively, married women may have lower employment levels because their husbands prefer it. On average, men have more traditional gender role attitudes than women (see Davis and Greenstein [2009]). If marriage gives men power to gain their partners’ compliance with their preferences, we expect married women to have lower employment levels. Whether cohabiting mothers’ employment levels will more closely resemble married or lone mothers is an open question. On one hand, cohabiting mothers’ employment may be similar to that of married mothers if they have similar access to their partners’ income, share childcare and household tasks in a similar way, and have similar expected relationship durations. On the other hand, cohabiting mothers’ employment may be more similar to that of lone mothers if cohabiting couples do not pool income, share housework and childcare differently, or have more unstable relationships than married couples. Relatively few studies compare the housework, parenting time, and income pooling strategies of married and cohabiting couples. On the question of whether cohabiting men do more housework than married men, findings are mixed (see Davis, Greenstein, and Marks [2007]), as are conclusions regarding marital status differences in father involvement with children (e.g., Kalenkoski, Ribar, and Stratton 2007; Ono and Yeilding 2009). Kenney’s (2006) study of couples’ income pooling practices suggests that cohabiting women’s access to their partner’s income is more varied than married women’s access. Additionally, little is known about how cohabiting (men) partners may influence women’s employment. If cohabiting men have less traditional gender attitudes than married men, we may expect their partners to have higher employment levels than married women. Even if cohabiting partners and husbands have similar attitudes toward maternal employment, we might see differences in maternal employment if cohabiting partners and husbands differ in their ability to influence their partners. Cohabiting parenthood is a less stable family form than married parenthood in the United States (Manning, Smock, and Majumdar 2004). About half of cohabiting unions with a birth transition to marital unions within five years, and these relationships are often as stable as marriages contracted before a birth, controlling for couple characteristics (Musick and Michelmore 2015). Still, many women may be aware of the high dissolution rates of cohabiting unions, and some may maintain high levels of employment as insurance against economic deprivation if their partnership dissolves. Thus, the empirical evidence regarding cohabitating parenthood has mixed implications as to whether cohabitation may differ from marriage in its effects on maternal employment. Other Family Influences on Maternal Employment Other family characteristics—including domestic violence, children’s health, number of children, and social support—may also affect mothers’ employment. These characteristics differ in their distribution, and possibly in their effects on employment, across family structures. Previous research has found that women who experience domestic violence have more employment instability—reflecting the impacts of violence on women’s health, as well as partners’ intentional interference with women’s employment (Showalter 2016)—and domestic violence is more common among cohabiting mothers than married mothers (Kenney and McLanahan 2006). Having a child with a chronic illness or a disability is also associated with reduced maternal employment (Corman, Noonan, and Reichman 2005), and is more common for unmarried mothers (Lee, Sills, and Oh 2002). Other family characteristics that may hinder maternal employment, such as having tightly spaced births and more young children in the household, are more prevalent among married mothers (Gemmill and Lindberg 2013). Factors that facilitate maternal employment, such as social and instrumental support (Livermore and Powers 2006), also differ in their prevalence across family structures. In particular, a nearby or co-resident grandmother may provide low-cost or free childcare. Children of lone mothers are more likely to live with a grandparent than children with two co-resident parents (Fields 2003, table 3). Although these factors are unlikely to be primary mechanisms through which family structure affects maternal employment, analyses of maternal employment should account for these characteristics. In summary, family structure may affect maternal employment through mechanisms such as family income needs, family resources and constraints, partner preferences, and the mother’s perceived security to pursue her preferred level of employment. Previous research and theory do not provide a clear hypothesis regarding the employment patterns of cohabiting mothers. Cohabiting mothers may have less economic need than lone mothers, suggesting their employment would be similar to married mothers. Cohabiters, however, have shorter union durations, and cohabiting mothers have weaker legal and social claims to their partners’ incomes than married mothers, suggesting that their employment may be more similar to that of lone mothers. Family Structure “Effects”? The preceding section reviews expectations for why family structure may influence maternal employment, all other factors equal. But mothers’ characteristics are not similar across family structures, and family structures are not randomly distributed. Notably, unmarried cohabitation is the most common context for childbearing among US women without a high school degree, whereas it is still a rare context for motherhood for college-educated women (Wu 2017). Additionally, racial/ethnic differences in family structure are large. For instance, cohabiting parenthood is more common among Hispanic than non-Hispanic mothers, and lone motherhood is the modal context for births to black mothers (Monte and Ellis 2014). Accounting for educational and ethnic/racial differences in parenthood contexts is crucial for a study of maternal employment because these characteristics are highly correlated with employment. Understanding how family structure affects women’s employment requires ruling out the rival hypothesis that family structure is merely a proxy for other factors that are driving differences in maternal employment. Human capital—especially educational attainment—is arguably the most important predictor of women’s employment, so comparisons of maternal employment across family structures need to be particularly attentive to differences in human capital. An additional characteristic that might be causally related to both family structure and maternal employment (and thus could cause a spurious relationship between family structure and employment) is gender attitudes. Women with more conservative gender attitudes may be more likely than women with more progressive gender attitudes both to have their children within marriage and to aspire to full-time homemaking. Previous Research on Family Structure and Maternal Employment Few previous studies of maternal employment consider cohabiting mothers as a separate category; most often, they are combined with unmarried and unpartnered (lone) mothers. Additionally, many previous studies of maternal employment consider only labor force participation, though much of the variation in maternal employment is in how soon mothers return to work after a birth and how many hours per week they work. A study by Han et al. (2008) overcomes several of these shortcomings. The authors use data from the Early Childhood Longitudinal Study-Birth Cohort (ECLS-B), a national study of children born in 2001 in the United States, to examine how soon mothers returned to work after the birth of a child. Their analysis found only small differences by family structure in the percentage working nine months after the birth. However, once they adjusted for mother’s age, race/ethnicity, education, employment status before the birth, and birth parity, the differences across family structure were sizable: cohabiting mothers were 14 percentage points and single mothers were 11 percentage points more likely to be working nine months after the birth than married mothers. The study did not investigate whether other characteristics associated with family structure (such as exposure to domestic violence) were driving these differences, nor did it consider any of the mechanisms—such as gender attitudes or partner characteristics—that might explain family structure differences in maternal employment. Additionally, the study only examined employment immediately after the birth. My study builds on the Han et al. analysis by 1) examining several mechanisms that may account for family structure differences; 2) controlling for a much richer set of mother characteristics that might be linked to both family structure and maternal employment; 3) including measures of fathers’ characteristics; and 4) considering a longer time period (five years versus nine months) and multiple measures of employment. Methods Data and Sample Data for this analysis are from the Fragile Families and Child Wellbeing Study (FFCWS), a longitudinal study of a birth cohort of children born between 1998 and 2000 in 20 large urban areas in the United States. The study oversampled nonmarital births, yielding a sample of approximately 3,700 children born to unmarried parents (including cohabitors) and 1,200 children born to married parents. Data collection began with an in-person interview of the mother at the hospital at the time of the birth and of the father either in the hospital or in another location shortly after the birth. Other data collections included surveys at 12 months (1999–2001), 30 months (2001–2003), and 60 months (2003–2005) after the child’s birth. For a thorough description of the study, see Reichman and colleagues (2001). Data from FFCWS are uniquely suited to answer my research questions because the study includes a large sample of cohabiting mothers and very detailed information on family structure, maternal characteristics, and paternal characteristics. Supplemental table 1 shows that mothers in FFCWS are similar to mothers of the same marital status with a birth in 2000 in the total US population on age and educational characteristics. Notably, mothers in FFCWS are more racially/ethnically diverse. Of the original mothers in FFCWS (n = 4,898), approximately 75 percent (n = 3,676) were interviewed at all four waves of data collection. The baseline interviews differed slightly for two of the 20 cities, so I exclude all mothers (n = 656) in these two cities. For mothers in the remaining 18 cities who participated in all survey waves, 98 percent answered all relevant employment questions. The analytic sample is composed of these mothers who participated in all four waves and answered all relevant items about employment histories (n = 3,132). The distribution of mothers by baseline family structure barely differs between the full sample (n = 4,898) and my analytic sample (n = 3,132). Of the mothers who did not participate in all four waves, most participated in at least one follow-up wave; there were no differences in maternal work hours prior to the birth or at any of the follow-up waves between mothers included in my analysis and those excluded for nonparticipation. To address missing data on parent characteristics, I use multiple imputation procedures2 to create 20 imputed datasets. I also have rerun all of the analyses using only the subset of families with no missing data (i.e., complete cases); I note when there are substantive differences in results between analyses with imputed data and complete cases only. Subsample For all descriptive statistics and the first set of analyses, I present findings for the full sample and a theoretically interesting subsample. This subsample consists of mothers who are part of demographic groups (defined by age, race/ethnicity, and education) with the most variation in family structure and where the comparison by family structure might be the most informative. In demographic groups with both very low and very high nonmarital birth rates, women in family structures that are non-normative for their particular demographic group may be unusual in ways that are not easily captured by covariate variables. One demographic group with a very low percentage of nonmarital births is college-educated mothers; only 9 percent of births to college-educated mothers from 1997 to 2001 were nonmarital (Kennedy and Bumpass 2008). Demographic groups with high percentages of nonmarital births include women without a high school degree, women under age 20, and black women. In the late 1990s, the percentage of births that were nonmarital to mothers with these characteristics were 64 percent for mothers with less than high school (Kennedy and Bumpass 2008), 79 percent for mothers under age 20 (Ventura and Bachrach 2000), and 69 percent for black mothers (Ventura and Bachrach 2000). Excluding women in demographic groups with very high or low nonmarital birth rates yields a subsample that consists of white non-Hispanic and Hispanic mothers who are age 20 or older who have either a high school degree or some college education (n = 630). Dependent and Explanatory Variables Maternal employment Measures of maternal employment are based on the mother’s self-report of when she returned to work after the birth and her hours of work in the previous week as collected at the 12-month (1999–2001), 30-month (2001–2003), and 60-month surveys (2003–2005). I topcoded hours of work at 80 hours per week.3 Initial family structure The family structures considered in this paper are 1) married, 2) (unmarried) cohabiting, and 3) living without a cohabiting partner or husband (hereafter referred to as lone mothers). I define family structure at baseline/Time 1 in reference to the mother’s relationship to the father of the new child.4 I use the mother’s report of marital status and whether she is living with the baby’s father all or most of the time. At the time of the birth, 25.2 percent of the mothers report being married, 35.3 percent cohabiting, and 39.5 percent as not living with the baby’s father. Of the subsample, 37.1 percent are married, 39.8 percent are cohabiting, and 23.0 percent are lone mothers. Family structure trajectories Many mothers experience changes in their family structure. I classify mothers according to their family structure at the beginning of the study (married, cohabiting, lone) and whether this family structure is stable or how it changes. To do this, I use the detailed marital and partnership history questions, which pertain to the mother’s relationship with the baby’s father and any new partners. I define 10 family structure trajectories: 1) Stably cohabiting (with the baby’s father); 2) Cohabiting to married (initially cohabiting with and subsequently marrying the baby’s father); 3) Cohabiting to lone (initially cohabiting with the baby’s father, followed by union dissolution); 4) Cohabiting, multiple changes (initially cohabiting with the baby’s father, followed by multiple changes, which may include re-partnering); 5) Stably married (to the baby’s father); 6) Married to divorced (initially married to the baby’s father and later divorced, with some mothers also re-partnering5); 7) Stably lone (not living with a husband or cohabiting partner at any survey wave); 8) Lone to cohabiting (initially unmarried and not cohabiting, followed by a later period of cohabitation with either the baby’s father or a new partner; 9) Lone to married (initially unmarried and not cohabiting, followed by a later marriage to either the baby’s father or a new partner); and 10) Lone with multiple changes (initially unmarried and not cohabiting, followed by multiple changes, which could include multiple partnership or marital status changes). Other Variables Independent variables include the mother’s demographic and human capital characteristics, her family resources and constraints, her family income and assets, the baby’s father’s characteristics, and the gender attitudes of the mother and father. Variables for the mother’s characteristics, family resources, household income and financial resources, and relationship status are based on the mother’s reports at the baseline interview. The father’s characteristics are based on the father’s reports (if he participated in the study) or the mother’s proxy reports if he did not participate. I do not include characteristics, such as subsequent fertility, that might be influenced by post-birth maternal employment or family structure changes. Mother and father characteristics Maternal demographic characteristics include age (dummy variable for teen, defined as under age 20; linear variable for age in years over age 20), race or ethnicity (non-Hispanic white or other, black, non-black Hispanic), and immigrant status (1 = immigrant). Paternal demographic characteristics include his age relative to the mother’s age (difference in years), and whether he is the same race/ethnicity as the mother. Human capital characteristics included for both parents include education (less than high school, GED, high school only, some college, college or more) and whether the parent has self-reported poor health. I also include the parents’ hourly pay rate (logged) from her/his last job and, for mothers, an indicator of whether she has no work experience. If the mother had no work experience or the parent had a very low (below $3.00/hour) or high (over $100/hour) reported wage rate, I treat the wage rate as missing and impute the wage. I also include an indicator variable to denote the missing wage. Approximately 13 percent of mothers and 27 percent of fathers have an imputed wage rate. For fathers, I also include whether he had a criminal record as of the baby’s birth, whether he reported using illegal substances, and how many weeks he was unemployed in the year before the birth. For mothers, I also include variables relating to mental health—a measure of the mother’s impulsive tendencies and a variable for whether there are severe mental health problems in her natal family—and her cognitive abilities (as measured by a subset of the Revised Weschler Adult Intelligence Scale), as these may affect a mother’s ability to find and maintain steady employment. Additional variables relating to mothers’ human capital include characteristics of her natal family (as these may predict unmeasured aspects of her cognitive and noncognitive abilities), including her mother’s education (less than high school, high school, more than high school, missing) and whether she grew up with both biological parents. Gender attitudes Mothers and fathers were each asked to respond to two items about their gender attitudes: “The important decisions in the family should be made by the man of the house” and “It is much better for everyone if the man earns the main living and the woman takes care of the home and family” (4 = strongly agree; 1 = strongly disagree). An individual’s responses to these two questions were combined to form a scale where higher scores indicate more traditional attitudes. The gender attitudes scales—as well as mother’s cognitive scores—are standardized within gender (mean = 0, standard deviation = 1). Higher values indicate that the mother is more conservative on gender attitudes than other mothers in the sample. FFCWS does not include any items related to a woman’s preferences for her own employment, and most other datasets used to examine women’s employment also lack items on preferences. Although preferences from early adulthood are associated with employment patterns, preferences are unstable over the lifecourse and often change with family circumstances (Kan 2007). Family constraints and resources Variables include whether the mother experienced domestic violence in the year before the birth, whether the child had a low birth weight, whether either of the baby’s grandmothers live in the household, and the number and age of other children in the household (measured as the number of children other than the new baby who are under age 3, ages 3–5, age 6 and older). Measures of economic resources include whether the mother lives in a home that she or a family member owns, and her report of her household’s status at the end of the month: just enough money to pay bills (reference category), not enough money to pay bills, or some money left after paying bills. Analyses I focus on two aspects of maternal employment: whether a mother worked within a year of the birth and how many hours she works per week in the five years after the birth. To examine how family structure associates with maternal employment, I use a two-pronged approach. First, I model maternal employment levels controlling for characteristics associated with employment and with family structure. In this part of the analysis, I use static family structure variables to predict which mothers work within a year of the birth (using logistic regression models) and how many hours mothers work per week (using growth curve models). Second, I model employment hours per week by family structure trajectories over a five-year period (using growth curve models). Comparing employment trajectories across mothers with different family structure trajectories allows me to identify whether mothers with stable family structures have different employment histories than mothers with family structure changes. To predict whether the mother returned to work within a year of the birth, I use logistic regression models6 run in STATA, and I use the MI suite of commands to combine the estimates from the 20 imputed datasets and estimate appropriate standard errors. I cluster the standard errors by city to account for the sampling design. To examine hours worked per week over the five-year period, I use growth curve models, an extension of structural equation modeling (Bollen and Curran 2006). Growth curve models are “statistical methods that allow for the estimation of inter-individual variability in intra-individual patterns of changes over time” (Curran, Obeidat, and Losardo 2010, p. 122). This is an appropriate modeling strategy given that I expect that maternal employment will change (increasing, on average) over time as children age, and that this pattern of change will vary across mothers. Growth curve models allow estimation of differences in initial levels of an outcome and in changes over time in that outcome. In my analyses, the intercept estimates hours of employment (per week) one year after the birth and the slope describes the rate of growth in employment hours (per week) over the next four years, estimated from measurements at 30 and 60 months. Growth curve models are incredibly flexible, and researchers have many options for how to specify these models, including how to model time and group differences, and whether to allow variables to predict the intercept, the slope, or both.7 Measures of model fit8 indicated that a linear specification did not fit the trend in maternal employment hours. Maternal employment hours increase slowly from one year to 30 months and then increase more rapidly between 30 and 60 months. After testing several specifications, I chose a model with the one-year timepoint set to 0, the third timepoint (at 60 months) set to 2, and the second timepoint (at 30 months) freely estimated. For estimating the growth curve models, I divide employment hours by 10 so that the scale of the variances for the outcome variables are similar to those for the predictor variables. The baseline model (without covariates) fits the data well, as indicated by measures of model fit and by a comparison of observed and fitted values (see supplemental figures 1a and 1b). I investigated whether to use a multiple-group model with groups defined by family structure, or whether to estimate a single-group model with covariates for family structure. I find that a single group model with covariates for family structure fits the data well (see supplemental figures 1a and 1b). I also tested whether the parameterization of time should be the same across family structures. Including covariates for family structure resulted in a similar model fit as that for a model in which time parameters were freely estimated separately for each family structure. Each mother’s employment trajectory is characterized by a unique intercept (α), slope (β), and error term (ε). I follow the same style of notation as Meadows, McLanahan, and Brooks-Gunn (2008). The level-one equation is as follows:   yit=αi+βit+εit (1) In this equation, y is mother i’ s hours of work at time t, αi is mother i’s intercept, βit is mother i’s slope at time t, and εit is the mother- and time-specific error term. I allow the covariates, which are time invariant, to influence both the intercept and the slope because some factors may influence maternal employment differentially by children’s age. The level-two equations are as follows:   αi=α0+α1xi1+α2xi2+α3xi3…+αkxik+ui (2)  βi=β0+β1xi1+β2xi2+β3xi3…+βkxik+vi (3) In these equations, the x’s are the time-invariant predictor variables. I present results from the model specifications, which includes all variables shown in table 1 plus city covariates. Table 1. Mother, Father, and Family Characteristics by Family Structure (at Birth of the Child), n = 3,132   Full sample (n = 3,132)  Subsample (n = 630)  Cohab  Married  Lone  Cohab  Married  Lone  n = 1105  n = 790  n = 1237  n = 251  n = 234  n = 145  Mother’s characteristics               Age (mean, excluding teens)  24.4  29.5  23.9  25.5  29.0  25.8   Teen (%)  19.0  3.4  26.9  0  0  0   Race/ethnicity:                Non-Hispanic white/Other (%)  22.3  53.6  12.6  49.0  69.7  49.7    Black (%)  48.0  27.5  70.6  0  0  0    Hispanic (%)  29.7  18.9  16.8  51.0  30.3  50.3   Immigrant (%)  13.2  21.6  4.9  22.3  20.9  8.3   In poor health (%)  7.2  3.4  8.3  4.8  2.6  6.2   Standardized cognitive score  −0.06  0.40  −0.08  0.10  0.41  0.23   Impulsive (%)  10.7  5.0  13.0  8.4  4.3  13.8   Own education:                Less than high school (%)  35.6  12.2  38.0  0  0  0    GED (%)  6.4  2.3  5.0  0.1  0  0    High school only (%)  29.6  18.1  29.9  47.6  34.2  42.1    Some college (%)  25.9  30.0  24.1  52.3  65.8  57.9    College or more (%)  2.6  37.5  2.9  0.1  0  0   Logged hourly wage rate  2.1  3.1  2.0  2.6  2.9  2.2   Missing hourly wage rate (%)  12.3  12.7  14.7  7.2  11.5  11.7   No work experience (%)  2.3  2.4  3.5  0.4  3.0  2.8   Maternal education:                Less than high school (%)  21.5  10.4  20.9  20.7  15.0  19.3    High school only (%)  55.9  60.2  51.3  61.8  61.5  59.3    More than high school (%)  16.6  25.9  19.8  14.3  20.9  15.9    Unknown or missing (%)  6.0  3.5  8.0  3.2  2.6  5.5   Poor parental mental health (%)  14.4  12.5  13.9  10.8  13.7  13.8   Lived with both parents at 14 (%)  38.6  64.7  30.7  52.5  62.9  50.3   Mother’s gender attitudes                Standardized gender attitudes (mean)  −0.08  0.03  −0.12  −0.22  0.10  −0.23  Family constraints and resources               No. of other children under age 3  0.24  0.27  0.24  0.17  0.26  0.16   No. of children ages 3, 4, or 5  0.24  0.25  0.22  0.22  0.24  0.18   No. of children ages 6 and over  0.42  0.41  0.42  0.36  0.42  0.39   Low birth weight baby (%)  10.2  5.5  12.8  6.5  5.3  7.2   Domestic violence (%)  5.1  2.0  6.9  6.0  2.1  8.9   Grandmother in household (%)  16.0  6.8  46.1  12.4  6.5  47.6   Home owned by family (%)  25.5  55.1  36.3  28.8  56.9  49.0   Not enough money for bills (%)  11.7  4.5  21.2  10.8  3.6  22.1   Money left after paying bills (%)  43.6  63.0  34.9  47.3  60.7  34.5  Father’s characteristics               Father’s age relative to mother’s  2.7  2.4  2.7  2.4  2.1  2.3   Same race/ethnicity as mother (%)  86.9  87.5  87.2  76.5  79.1  64.0   In poor health (%)  7.9  5.4  8.9  6.4  5.1  11.3   Reports drug use (%)  9.0  4.1  11.9  8.0  6.0  20.7   Education:                Less than high school (%)  37.3  12.2  36.5  25.5  9.0  25.8    GED (%)  8.0  3.3  8.9  5.6  4.7  9.2    High school only (%)  28.3  21.1  33.5  27.4  25.2  30.5    Some college (%)  23.1  29.7  18.0  35.5  39.7  28.0    College or more (%)  3.3  33.6  3.0  6.0  20.9  6.5   Criminal record (%)  30.3  13.0  21.7  28.7  15.4  21.4   Weeks not employed (mean)  10.1  4.1  15.0  6.0  4.2  10.0   Logged hourly wage rate  2.3  2.8  2.2  2.5  2.8  2.4   Missing hourly wage rate (%)  18.5  18.1  43.7  13.1  18.8  40.7   Father’s gender attitudes               Standardized (mean)  −0.05  −0.04  −0.03  −0.20  −0.08  −0.20    Full sample (n = 3,132)  Subsample (n = 630)  Cohab  Married  Lone  Cohab  Married  Lone  n = 1105  n = 790  n = 1237  n = 251  n = 234  n = 145  Mother’s characteristics               Age (mean, excluding teens)  24.4  29.5  23.9  25.5  29.0  25.8   Teen (%)  19.0  3.4  26.9  0  0  0   Race/ethnicity:                Non-Hispanic white/Other (%)  22.3  53.6  12.6  49.0  69.7  49.7    Black (%)  48.0  27.5  70.6  0  0  0    Hispanic (%)  29.7  18.9  16.8  51.0  30.3  50.3   Immigrant (%)  13.2  21.6  4.9  22.3  20.9  8.3   In poor health (%)  7.2  3.4  8.3  4.8  2.6  6.2   Standardized cognitive score  −0.06  0.40  −0.08  0.10  0.41  0.23   Impulsive (%)  10.7  5.0  13.0  8.4  4.3  13.8   Own education:                Less than high school (%)  35.6  12.2  38.0  0  0  0    GED (%)  6.4  2.3  5.0  0.1  0  0    High school only (%)  29.6  18.1  29.9  47.6  34.2  42.1    Some college (%)  25.9  30.0  24.1  52.3  65.8  57.9    College or more (%)  2.6  37.5  2.9  0.1  0  0   Logged hourly wage rate  2.1  3.1  2.0  2.6  2.9  2.2   Missing hourly wage rate (%)  12.3  12.7  14.7  7.2  11.5  11.7   No work experience (%)  2.3  2.4  3.5  0.4  3.0  2.8   Maternal education:                Less than high school (%)  21.5  10.4  20.9  20.7  15.0  19.3    High school only (%)  55.9  60.2  51.3  61.8  61.5  59.3    More than high school (%)  16.6  25.9  19.8  14.3  20.9  15.9    Unknown or missing (%)  6.0  3.5  8.0  3.2  2.6  5.5   Poor parental mental health (%)  14.4  12.5  13.9  10.8  13.7  13.8   Lived with both parents at 14 (%)  38.6  64.7  30.7  52.5  62.9  50.3   Mother’s gender attitudes                Standardized gender attitudes (mean)  −0.08  0.03  −0.12  −0.22  0.10  −0.23  Family constraints and resources               No. of other children under age 3  0.24  0.27  0.24  0.17  0.26  0.16   No. of children ages 3, 4, or 5  0.24  0.25  0.22  0.22  0.24  0.18   No. of children ages 6 and over  0.42  0.41  0.42  0.36  0.42  0.39   Low birth weight baby (%)  10.2  5.5  12.8  6.5  5.3  7.2   Domestic violence (%)  5.1  2.0  6.9  6.0  2.1  8.9   Grandmother in household (%)  16.0  6.8  46.1  12.4  6.5  47.6   Home owned by family (%)  25.5  55.1  36.3  28.8  56.9  49.0   Not enough money for bills (%)  11.7  4.5  21.2  10.8  3.6  22.1   Money left after paying bills (%)  43.6  63.0  34.9  47.3  60.7  34.5  Father’s characteristics               Father’s age relative to mother’s  2.7  2.4  2.7  2.4  2.1  2.3   Same race/ethnicity as mother (%)  86.9  87.5  87.2  76.5  79.1  64.0   In poor health (%)  7.9  5.4  8.9  6.4  5.1  11.3   Reports drug use (%)  9.0  4.1  11.9  8.0  6.0  20.7   Education:                Less than high school (%)  37.3  12.2  36.5  25.5  9.0  25.8    GED (%)  8.0  3.3  8.9  5.6  4.7  9.2    High school only (%)  28.3  21.1  33.5  27.4  25.2  30.5    Some college (%)  23.1  29.7  18.0  35.5  39.7  28.0    College or more (%)  3.3  33.6  3.0  6.0  20.9  6.5   Criminal record (%)  30.3  13.0  21.7  28.7  15.4  21.4   Weeks not employed (mean)  10.1  4.1  15.0  6.0  4.2  10.0   Logged hourly wage rate  2.3  2.8  2.2  2.5  2.8  2.4   Missing hourly wage rate (%)  18.5  18.1  43.7  13.1  18.8  40.7   Father’s gender attitudes               Standardized (mean)  −0.05  −0.04  −0.03  −0.20  −0.08  −0.20  Findings Descriptive Statistics Table 1 shows the distribution of parent and family characteristics by family structure for the full sample and the subsample. (The descriptive statistics and analyses are unweighted.) On average, cohabiting mothers are less advantaged in human capital characteristics (education, wage rate, and work experience) than married mothers. Cohabiting mothers are also more likely than married mothers to be in poor health, have a low birth weight baby, or report experiencing domestic violence in the year before the birth, all of which are characteristics that depress employment. These differences hold for the full sample and the subsample. The differences between cohabiting and lone mothers, in both the full sample and the subsample, are less stark. Cohabiting and lone mothers have similar levels of education, cognitive scores, and hourly wage rates. A smaller share of cohabiting mothers than lone mothers live with the baby’s grandmother. Additionally, cohabiting mothers are more likely than married mothers—but less likely than lone mothers—to report not having enough money to pay bills. Table 1 also shows the distribution of father characteristics by family structure. In both the full sample and subsample, husbands are more educated, less likely to use drugs or have a criminal record, and had fewer weeks of unemployment than cohabiting fathers. On some dimensions, such as unemployment, cohabiting fathers are more advantaged than lone fathers, whereas on dimensions such as educational attainment and criminal justice involvement, cohabiting fathers appear equally or more disadvantaged. Given the differences in mother, father, and couple characteristics by family structure, we might expect to find considerable differences in maternal employment levels by family structure. But, as table 2 shows, observed differences in maternal employment by family structure among the full sample of FFCWS mothers are small. The mean number of hours worked per week in the first year is similar across family structures (approximately 21 hours). The percentage of mothers employed at the one-year interview is lower for cohabiting mothers than married mothers, though the percentage of cohabiting mothers who have worked at any point during the first year after the birth (74.2 percent) is higher than that of married mothers (65.2 percent). Among employed mothers, cohabiting mothers work approximately three hours more per week than married mothers and work the same number of hours as lone mothers. These statistics suggest relatively small differences in maternal employment in the first year after a birth. Table 2. Maternal Employment by Family Structure (at Birth of the Child)   Full sample (n = 3,132)  Subsample (n = 630)  Cohabiting  Married  Lone  Cohabiting  Married  Lone  (n = 1,105)  (n = 790)  (n = 1237)  (n = 251)  (n = 234)  (n = 145)  % employed at any point during the first year  74.2  65.2*  76.5  77.7  60.7*  78.6  Percentage employed at each survey wave               At one year  54.5  58.1  53.8  63.3  55.6  64.1   At 30 months  57.1  59.1  56.6  64.0  59.0  63.4   At 60 months  62.7  62.8  57.8*  68.9  61.8  63.4  Hours worked per week: mean and standard deviation               All mothers                At one year  21.1  20.6  20.5  23.6  19.1*  23.5    (21.4)  (20.1)  (21.1)  (20.2)  (19.6)  (20.8)    At 30 months  23.3  21.4  22.6  24.5  20.3*  25.5    (22.4)  (20.6)  (22.1)  (20.5)  (19.8)  (21.6)    At 60 months  35.8  29.8*  35.9  35.6  27.5*  37.0    (16.8)  (18.9)  (16.9)  (15.4)  (18.1)  (15.9)   All working mothers (excludes mothers with zero hours)                At one year  38.8  35.4*  38.0  37.3  34.4*  36.6    (12.7)  (13.1)  (12.9)  (11.4)  (12.9)  (13.8)    At 30 months  40.8  36.3*  40.0  38.5  34.4*  40.2    (12.9)  (13.3)  (13.1)  (11.0)  (13.3)  (11.9)    At 60 months  40.3  36.9*  40.8  38.9  34.7*  39.8    (13.0)  (14.1)  (12.9)  (11.8)  (13.0)  (13.6)    Full sample (n = 3,132)  Subsample (n = 630)  Cohabiting  Married  Lone  Cohabiting  Married  Lone  (n = 1,105)  (n = 790)  (n = 1237)  (n = 251)  (n = 234)  (n = 145)  % employed at any point during the first year  74.2  65.2*  76.5  77.7  60.7*  78.6  Percentage employed at each survey wave               At one year  54.5  58.1  53.8  63.3  55.6  64.1   At 30 months  57.1  59.1  56.6  64.0  59.0  63.4   At 60 months  62.7  62.8  57.8*  68.9  61.8  63.4  Hours worked per week: mean and standard deviation               All mothers                At one year  21.1  20.6  20.5  23.6  19.1*  23.5    (21.4)  (20.1)  (21.1)  (20.2)  (19.6)  (20.8)    At 30 months  23.3  21.4  22.6  24.5  20.3*  25.5    (22.4)  (20.6)  (22.1)  (20.5)  (19.8)  (21.6)    At 60 months  35.8  29.8*  35.9  35.6  27.5*  37.0    (16.8)  (18.9)  (16.9)  (15.4)  (18.1)  (15.9)   All working mothers (excludes mothers with zero hours)                At one year  38.8  35.4*  38.0  37.3  34.4*  36.6    (12.7)  (13.1)  (12.9)  (11.4)  (12.9)  (13.8)    At 30 months  40.8  36.3*  40.0  38.5  34.4*  40.2    (12.9)  (13.3)  (13.1)  (11.0)  (13.3)  (11.9)    At 60 months  40.3  36.9*  40.8  38.9  34.7*  39.8    (13.0)  (14.1)  (12.9)  (11.8)  (13.0)  (13.6)  Note: Data are from the Fragile Families and Child Wellbeing Study. Subsample includes non-Hispanic white and Hispanic mothers over age 20 with a high school degree or some college. Differences with cohabiting mothers that are statistically significant at p < 0.05 are indicated by *. At 30 months after the birth, the average hours of work for cohabiting mothers is two hours per week more than married mothers and one hour per week more than lone mothers. Among employed mothers, the difference between cohabiting and married mothers is larger, with cohabiting mothers working 4.5 hours more per week than married mothers. By the child’s fifth birthday, differences by family structure are quite large, with a difference of six hours per week between married mothers and unmarried mothers, with nearly identical work hours for cohabiting and lone mothers. The difference between cohabiting and married mothers in hours of work at 60 months is not driven primarily by differences in labor force participation rates; 62.8 percent of married mothers and 62.7 percent of cohabiting mothers are employed. The right side of table 2 shows employment statistics for the subsample. Mothers in this group are more similar to each other on human capital and demographic characteristics than are mothers in the full sample. Among this more homogeneous group, the employment differences between cohabiting and married mothers are considerably greater than for the full sample. Approximately 78 percent of cohabiting and lone mothers return to work in the first year after the birth, compared with 60.7 percent of married mothers. In the first year, cohabiting mothers work 4.5 hours more per week than married mothers and the same number of hours as lone mothers. By the child’s fifth birthday, the difference between cohabiting and married mothers increases to 8.1 hours more per week for cohabiting mothers. These sizable differences in maternal employment by family structure in the subsample suggest that the small differences in the full sample may reflect the offsetting influences of human capital differences and family structure on employment. Multivariate Results: Employment in the First Year after a Birth I use logistic regression models to predict which mothers were employed at some point in the first year following a birth. In models shown in table 3, I find that all else equal, cohabiting mothers are more likely to be employed in the first year than married mothers. Accounting for differences among women via covariate adjustment does not attenuate this relationship much. For the full sample, the odds of working in the first year after the birth are considerably lower for married mothers compared to the reference group of unmarried cohabiting mothers (B = −0.47, s.e. = 0.16, odds ratio = 0.63). Employment for lone mothers is not significantly different than that for cohabiting mothers (B = 0.14, s.e. = 0.13, odds ratio = 1.15). Table 3. Results from Logistic Regression Models Predicting Employment in the First Year after the Birth   Full sample (n = 3,132)  Subsample (n = 630)  Odds ratio  B  s.e.  Odds ratio  B  s.e.  Family structure               Married  0.63  −0.47  0.16**  0.48  −0.72  0.23**   Lone  1.15  0.14  0.13  1.33  0.29  0.39  Mother’s characteristics               Age (years over 20)  0.97  −0.03  0.01**  0.96  −0.04  0.02~   Teen mother  1.30  0.26  0.13*  n.a.       Black  2.42  0.88  0.13***  n.a.       Hispanic  1.26  0.23  0.15  1.39  0.33  0.30   Immigrant  0.74  −0.30  0.11**  1.06  0.06  0.21   In poor health  0.59  −0.53  0.23*  0.46  −0.78  0.59   Standardized cognitive score  0.97  −0.03  0.04  1.01  0.01  0.08   Impulsive  0.85  −0.16  0.12  0.68  −0.38  0.34   Less than high school  0.60  −0.51  0.09***  n.a.       GED  1.14  0.13  0.19  n.a.       Some college  1.38  0.32  0.18~  1.87  0.63  0.22**   College  2.06  0.72  0.26**  n.a.       Logged hourly wage rate  1.24  0.22  0.03***  1.26  0.23  0.12*   Missing hourly wage rate  0.47  −0.76  0.11***  0.79  −0.24  0.33   No work experience  0.15  −1.91  0.44***  0.19  −1.68  0.88~   Maternal ed.—less than high school  0.93  −0.08  0.14  1.32  0.28  0.42   Maternal ed.—more than high school  1.02  0.02  0.17  1.04  0.03  0.38   Maternal ed.—missing  1.04  0.04  0.21  1.05  0.05  0.72   Poor parental mental health  1.01  0.01  0.17  0.77  −0.26  0.36   Lived with both parents at age 14  0.96  −0.04  0.10  0.77  −0.26  0.28   Gender attitudes  0.83  −0.19  0.05***  0.79  −0.24  0.10*  Family constraints and resources             # of children under age 3  0.76  −0.27  0.11*  0.94  −0.07  0.26   # of children ages 3–5  0.87  −0.14  0.13  1.16  0.15  0.23   # of children ages 6 and over  0.94  −0.06  0.07  0.96  −0.05  0.09   Low birth weight baby  0.74  −0.30  0.19  0.90  −0.11  0.33   Domestic violence  1.17  0.15  0.21  1.00  0.00  0.60   Grandmother in the household  0.97  −0.03  0.12  1.02  0.02  0.27   Home owned by mother/family  0.95  −0.05  0.09  0.97  −0.03  0.15   Reports not enough money for bills  0.70  −0.36  0.12**  0.41  −0.90  0.24***   Reports money left after paying bills  1.07  0.07  0.10  1.72  0.54  0.20**  Father characteristics               Father’s age relative to mother’s  0.98  −0.02  0.01**  0.94  −0.06  0.02***   Same race as mother  0.67  −0.40  0.13**  0.75  −0.28  0.23   In poor health  1.32  0.28  0.17  0.91  −0.09  0.43   Reports drug use  1.52  0.42  0.16**  3.07  1.12  0.42**   Less than high school  1.21  0.19  0.14  1.63  0.49  0.33   GED  1.15  0.14  0.20  0.94  −0.06  0.48   Some college  0.83  −0.19  0.13  1.03  0.03  0.19   College  0.63  −0.46  0.24~  1.11  0.10  0.38   Criminal record  1.04  0.03  0.13  1.18  0.17  0.30   Weeks not employed  0.99  −0.01  0.00*  1.00  0.00  0.01   Logged hourly wage rate  0.93  −0.08  0.11  0.67  −0.40  0.21~   Missing hourly wage rate  0.80  −0.23  0.14  1.00  0.00  0.24   Gender attitudes  0.86  −0.15  0.05***  0.91  −0.10  0.12  Intercept    1.84  0.40    2.21  0.61  Log likelihooda  −1540.31  −313.25    Full sample (n = 3,132)  Subsample (n = 630)  Odds ratio  B  s.e.  Odds ratio  B  s.e.  Family structure               Married  0.63  −0.47  0.16**  0.48  −0.72  0.23**   Lone  1.15  0.14  0.13  1.33  0.29  0.39  Mother’s characteristics               Age (years over 20)  0.97  −0.03  0.01**  0.96  −0.04  0.02~   Teen mother  1.30  0.26  0.13*  n.a.       Black  2.42  0.88  0.13***  n.a.       Hispanic  1.26  0.23  0.15  1.39  0.33  0.30   Immigrant  0.74  −0.30  0.11**  1.06  0.06  0.21   In poor health  0.59  −0.53  0.23*  0.46  −0.78  0.59   Standardized cognitive score  0.97  −0.03  0.04  1.01  0.01  0.08   Impulsive  0.85  −0.16  0.12  0.68  −0.38  0.34   Less than high school  0.60  −0.51  0.09***  n.a.       GED  1.14  0.13  0.19  n.a.       Some college  1.38  0.32  0.18~  1.87  0.63  0.22**   College  2.06  0.72  0.26**  n.a.       Logged hourly wage rate  1.24  0.22  0.03***  1.26  0.23  0.12*   Missing hourly wage rate  0.47  −0.76  0.11***  0.79  −0.24  0.33   No work experience  0.15  −1.91  0.44***  0.19  −1.68  0.88~   Maternal ed.—less than high school  0.93  −0.08  0.14  1.32  0.28  0.42   Maternal ed.—more than high school  1.02  0.02  0.17  1.04  0.03  0.38   Maternal ed.—missing  1.04  0.04  0.21  1.05  0.05  0.72   Poor parental mental health  1.01  0.01  0.17  0.77  −0.26  0.36   Lived with both parents at age 14  0.96  −0.04  0.10  0.77  −0.26  0.28   Gender attitudes  0.83  −0.19  0.05***  0.79  −0.24  0.10*  Family constraints and resources             # of children under age 3  0.76  −0.27  0.11*  0.94  −0.07  0.26   # of children ages 3–5  0.87  −0.14  0.13  1.16  0.15  0.23   # of children ages 6 and over  0.94  −0.06  0.07  0.96  −0.05  0.09   Low birth weight baby  0.74  −0.30  0.19  0.90  −0.11  0.33   Domestic violence  1.17  0.15  0.21  1.00  0.00  0.60   Grandmother in the household  0.97  −0.03  0.12  1.02  0.02  0.27   Home owned by mother/family  0.95  −0.05  0.09  0.97  −0.03  0.15   Reports not enough money for bills  0.70  −0.36  0.12**  0.41  −0.90  0.24***   Reports money left after paying bills  1.07  0.07  0.10  1.72  0.54  0.20**  Father characteristics               Father’s age relative to mother’s  0.98  −0.02  0.01**  0.94  −0.06  0.02***   Same race as mother  0.67  −0.40  0.13**  0.75  −0.28  0.23   In poor health  1.32  0.28  0.17  0.91  −0.09  0.43   Reports drug use  1.52  0.42  0.16**  3.07  1.12  0.42**   Less than high school  1.21  0.19  0.14  1.63  0.49  0.33   GED  1.15  0.14  0.20  0.94  −0.06  0.48   Some college  0.83  −0.19  0.13  1.03  0.03  0.19   College  0.63  −0.46  0.24~  1.11  0.10  0.38   Criminal record  1.04  0.03  0.13  1.18  0.17  0.30   Weeks not employed  0.99  −0.01  0.00*  1.00  0.00  0.01   Logged hourly wage rate  0.93  −0.08  0.11  0.67  −0.40  0.21~   Missing hourly wage rate  0.80  −0.23  0.14  1.00  0.00  0.24   Gender attitudes  0.86  −0.15  0.05***  0.91  −0.10  0.12  Intercept    1.84  0.40    2.21  0.61  Log likelihooda  −1540.31  −313.25  Note: Reference group is unmarried cohabiting mothers. n.a. indicates not applicable. aThe reported loglikelihood is the average log likelihood of 20 imputed datasets. Standard errors are for the coefficients. Models adjust standard errors for the clustering by city. Data are from the Fragile Families and Child Wellbeing Study. Statistical significance levels are denoted as follows: ~ p < 0.10 * p < 0.05 ** p < 0.01 *** p < 0.001 (two-tailed) In addition to differences by marital status, I find that mothers with more human capital and black mothers have a higher likelihood of returning to work in the first year. The gender attitudes variable is predictive (B = −0.19, s.e. = 0.05, odds ratio = 0.83) of maternal employment; mothers with more gender-conservative attitudes have lower odds of working in the first year. Because the distribution of mothers’ gender attitudes does not differ much across family structure, gender attitudes cannot explain family structure differences in employment. Father’s gender attitudes are also a significant predictor of her employment; mothers whose partners have more traditional gender attitudes are less likely to work in the first year. Additional tests reveal that there is an interaction between father’s gender attitudes and family structure, such that fathers’ attitudes are only predictive of married mothers’ employment. Other analyses reveal no interactions between mother’s race and family structure, mother’s education and family structure, or mother’s gender attitudes and family structure. Results from the analysis of the subsample of mothers are similar to those for the full sample. The model shows no statistically significant differences in the odds of working with an infant between cohabiting and lone mothers (B = 0.29, s.e. = 0.39) but substantial differences between cohabiting and married mothers (B = −0.72, s.e. = 0.23, odds ratio = 0.48). My results from the full sample and the subsample9 are consistent with Han and colleagues’ (2008) findings from their analysis of mothers’ employment in the nine months following a birth; they found that unmarried mothers had a higher rate of labor force participation, controlling for education. Multivariate Results: Maternal Work Hours in the Five Years after a Birth In these data, cohabiting mothers are more likely to work in the first year after a birth than married mothers, but employment patterns may change as children age because infants place different demands on families than older children. Table 4 shows how family structure associates with hours of work per week. (Recall that the employment hours variables are divided by 10 to estimate the growth curve models.) Table 4. Results from Growth Curve Models Predicting Hours Worked per Week with Family Structure as Predictor Variables   Full sample  Subsample  Intercept  Slope  Intercept  Slope  Family structure           Married  −0.38***  0.07  −0.24  −0.06    (0.10)  (0.05)  (0.19)  (0.09)   Lone  −0.03  0.01  0.19  0.02    (0.08)  (0.05)  (0.20)  (0.10)  Mother’s characteristics           Age (years over 20)  −0.01~  −0.01  −0.02  0.00    (0.01)  (0.00)  (0.01)  (0.01)   Teen mother  −0.19*  0.12*  n.a.  n.a.    (0.09)  (0.06)       Race/ethnicity:            Black  0.53***  −0.02  n.a.  n.a.    (0.10)  (0.05)        Hispanic  0.29*  −0.06  0.17  −0.05    (0.11)  (0.06)  (0.19)  (0.10)   Immigrant  0.03  0.06  −0.10  0.16    (0.12)  (0.07)  (0.22)  (0.12)   In poor health  −0.19  −0.02  −0.34  −0.16    (0.13)  (0.08)  (0.35)  (0.17)   Std. cognitive score  −0.01  0.00  −0.09  0.07~    (0.04)  (0.02)  (0.08)  (0.04)   Impulsive  −0.25*  0.03  −0.13  −0.02    (0.10)  (0.06)  (0.26)  (0.13)   Own education:            Less than high school  −0.61***  0.26***  n.a.  n.a.    (0.09)  (0.05)        GED  −0.09  −0.04  n.a.  n.a.    (0.16)  (0.09)        Some college  0.33***  −0.12*  0.37*  −0.17*    (0.09)  (0.05)  (0.15)  (0.08)    College or more  0.52***  −0.15*  n.a.  n.a.    (0.14)  (0.07)       Logged hourly wage rate  0.20***  −0.05**  0.21**  −0.10*    (0.03)  (0.02)  (0.08)  (0.04)   Missing hourly wage rate  −0.57***  0.10  −0.75**  −0.03    (0.10)  (0.06)  (0.22)  (0.13)   No work experience  −0.37*  −1.26***  n.a.  n.a.    (0.15)  (0.09)       Maternal education:            Less than high school  −0.05  0.07  −0.10  0.10    (0.09)  (0.05)  (0.20)  (0.11)    More than high school  0.10  −0.07  −0.25  0.07    (0.08)  (0.05)  (0.19)  (0.10)    Unknown or missing  −0.34*  0.00  −0.62  0.12    (0.13)  (0.08)  (0.42)  (0.23)   Poor parental mental health  −0.07  0.03  −0.01  0.00    (0.09)  (0.05)  (0.20)  (0.11)   Lived with both parents at age 14  −0.02  −0.05  −0.32*  0.05    (0.07)  (0.04)  (0.14)  (0.07)   Gender attitudes  −0.11**  −0.02  −0.16*  −0.03    (0.03)  (0.02)  (0.08)  (0.04)  Family constraints and resources           No. of children under age 3  −0.31***  0.14**  −0.22  −0.03    (0.07)  (0.04)  (0.17)  (0.09)   No. of children ages 3–5  0.03  −0.01  0.18  −0.12    (0.07)  (0.04)  (0.16)  (0.08)   No. of children ages 6+  0.02  0.01  0.02  0.06    (0.04)  (0.02)  (0.09)  (0.04)   Low-birth-weight baby  −0.19~  0.06  0.06  −0.21    (0.10)  (0.06)  (0.30)  (0.17)   Domestic violence  −0.04  0.01  −0.50  0.24    (0.16)  (0.09)  (0.34)  (0.19)   Grandmother in household  −0.03  0.08~  −0.10  0.03    (0.08)  (0.05)  (0.21)  (0.11)   Home owned by mother/family  0.00  −0.06  −0.10  −0.08    (0.07)  (0.04)  (0.16)  (0.08)   Not enough money for bills  −0.31**  0.14*  −0.40  0.22    (0.10)  (0.06)  (0.27)  (0.15)   Money left after paying bills  0.03  −0.01  0.07  −0.02    (0.07)  (0.04)  (0.15)  (0.07)  Father’s characteristics           Father’s age relative to mother’s  −0.01  0.00  −0.02  0.01    (0.01)  (0.00)  (0.02)  (0.01)   Same race/ethnicity as mother  −0.31**  0.04  −0.29~  0.08    (0.10)  (0.05)  (0.17)  (0.09)   In poor health  0.12  −0.05  0.12  −0.13    (0.14)  (0.08)  (0.32)  (0.18)   Reports drug use  0.22~  −0.07  0.16  −0.15    (0.12)  (0.07)  (0.27)  (0.13)   Education:            Less than high school  0.09  −0.02  0.05  0.04    (0.09)  (0.05)  (0.20)  (0.11)    GED  0.03  0.06  −0.09  0.06    (0.13)  (0.08)  (0.31)  (0.18)    Some college  0.03  −0.07  0.09  −0.04    (0.09)  (0.05)  (0.18)  (0.09)    College or more  −0.43**  −0.03  −0.05  −0.08    (0.14)  (0.07)  (0.26)  (0.12)   Criminal record  0.03  0.04  0.24  −0.03    (0.08)  (0.05)  (0.18)  (0.09)   Weeks not employed last year  −0.01*  0.00  0.00  0.00    (0.00)  (0.00)  (0.01)  (0.00)   Logged hourly wage rate  0.04  −0.08~  −0.16  −0.05    (0.08)  (0.05)  (0.16)  (0.08)   Missing hourly wage rate  −0.01  0.00  0.12  0.01    (0.07)  (0.04)  (0.18)  (0.09)   Gender attitudes  −0.09*  −0.02  0.00  −0.15**    (0.04)  (0.02)  (0.09)  (0.05)  Intercept  2.62  0.73  2.81  0.74    (0.26)  (0.15)  (0.52)  (0.27)  Slope with intercept  −2.18  −0.16    (0.05)  (0.09)  Residual variance  1.26  0.53  1.39  0.50    (0.09)  (0.14)  (0.18)  (0.68)  X2(df)  79.98 (62)  67.61 (51)  RMSEA  0.01  0.02  BIC  38275.6  7930.85  CFI/TLI  0.993/0.977  0.967/0.899    Full sample  Subsample  Intercept  Slope  Intercept  Slope  Family structure           Married  −0.38***  0.07  −0.24  −0.06    (0.10)  (0.05)  (0.19)  (0.09)   Lone  −0.03  0.01  0.19  0.02    (0.08)  (0.05)  (0.20)  (0.10)  Mother’s characteristics           Age (years over 20)  −0.01~  −0.01  −0.02  0.00    (0.01)  (0.00)  (0.01)  (0.01)   Teen mother  −0.19*  0.12*  n.a.  n.a.    (0.09)  (0.06)       Race/ethnicity:            Black  0.53***  −0.02  n.a.  n.a.    (0.10)  (0.05)        Hispanic  0.29*  −0.06  0.17  −0.05    (0.11)  (0.06)  (0.19)  (0.10)   Immigrant  0.03  0.06  −0.10  0.16    (0.12)  (0.07)  (0.22)  (0.12)   In poor health  −0.19  −0.02  −0.34  −0.16    (0.13)  (0.08)  (0.35)  (0.17)   Std. cognitive score  −0.01  0.00  −0.09  0.07~    (0.04)  (0.02)  (0.08)  (0.04)   Impulsive  −0.25*  0.03  −0.13  −0.02    (0.10)  (0.06)  (0.26)  (0.13)   Own education:            Less than high school  −0.61***  0.26***  n.a.  n.a.    (0.09)  (0.05)        GED  −0.09  −0.04  n.a.  n.a.    (0.16)  (0.09)        Some college  0.33***  −0.12*  0.37*  −0.17*    (0.09)  (0.05)  (0.15)  (0.08)    College or more  0.52***  −0.15*  n.a.  n.a.    (0.14)  (0.07)       Logged hourly wage rate  0.20***  −0.05**  0.21**  −0.10*    (0.03)  (0.02)  (0.08)  (0.04)   Missing hourly wage rate  −0.57***  0.10  −0.75**  −0.03    (0.10)  (0.06)  (0.22)  (0.13)   No work experience  −0.37*  −1.26***  n.a.  n.a.    (0.15)  (0.09)       Maternal education:            Less than high school  −0.05  0.07  −0.10  0.10    (0.09)  (0.05)  (0.20)  (0.11)    More than high school  0.10  −0.07  −0.25  0.07    (0.08)  (0.05)  (0.19)  (0.10)    Unknown or missing  −0.34*  0.00  −0.62  0.12    (0.13)  (0.08)  (0.42)  (0.23)   Poor parental mental health  −0.07  0.03  −0.01  0.00    (0.09)  (0.05)  (0.20)  (0.11)   Lived with both parents at age 14  −0.02  −0.05  −0.32*  0.05    (0.07)  (0.04)  (0.14)  (0.07)   Gender attitudes  −0.11**  −0.02  −0.16*  −0.03    (0.03)  (0.02)  (0.08)  (0.04)  Family constraints and resources           No. of children under age 3  −0.31***  0.14**  −0.22  −0.03    (0.07)  (0.04)  (0.17)  (0.09)   No. of children ages 3–5  0.03  −0.01  0.18  −0.12    (0.07)  (0.04)  (0.16)  (0.08)   No. of children ages 6+  0.02  0.01  0.02  0.06    (0.04)  (0.02)  (0.09)  (0.04)   Low-birth-weight baby  −0.19~  0.06  0.06  −0.21    (0.10)  (0.06)  (0.30)  (0.17)   Domestic violence  −0.04  0.01  −0.50  0.24    (0.16)  (0.09)  (0.34)  (0.19)   Grandmother in household  −0.03  0.08~  −0.10  0.03    (0.08)  (0.05)  (0.21)  (0.11)   Home owned by mother/family  0.00  −0.06  −0.10  −0.08    (0.07)  (0.04)  (0.16)  (0.08)   Not enough money for bills  −0.31**  0.14*  −0.40  0.22    (0.10)  (0.06)  (0.27)  (0.15)   Money left after paying bills  0.03  −0.01  0.07  −0.02    (0.07)  (0.04)  (0.15)  (0.07)  Father’s characteristics           Father’s age relative to mother’s  −0.01  0.00  −0.02  0.01    (0.01)  (0.00)  (0.02)  (0.01)   Same race/ethnicity as mother  −0.31**  0.04  −0.29~  0.08    (0.10)  (0.05)  (0.17)  (0.09)   In poor health  0.12  −0.05  0.12  −0.13    (0.14)  (0.08)  (0.32)  (0.18)   Reports drug use  0.22~  −0.07  0.16  −0.15    (0.12)  (0.07)  (0.27)  (0.13)   Education:            Less than high school  0.09  −0.02  0.05  0.04    (0.09)  (0.05)  (0.20)  (0.11)    GED  0.03  0.06  −0.09  0.06    (0.13)  (0.08)  (0.31)  (0.18)    Some college  0.03  −0.07  0.09  −0.04    (0.09)  (0.05)  (0.18)  (0.09)    College or more  −0.43**  −0.03  −0.05  −0.08    (0.14)  (0.07)  (0.26)  (0.12)   Criminal record  0.03  0.04  0.24  −0.03    (0.08)  (0.05)  (0.18)  (0.09)   Weeks not employed last year  −0.01*  0.00  0.00  0.00    (0.00)  (0.00)  (0.01)  (0.00)   Logged hourly wage rate  0.04  −0.08~  −0.16  −0.05    (0.08)  (0.05)  (0.16)  (0.08)   Missing hourly wage rate  −0.01  0.00  0.12  0.01    (0.07)  (0.04)  (0.18)  (0.09)   Gender attitudes  −0.09*  −0.02  0.00  −0.15**    (0.04)  (0.02)  (0.09)  (0.05)  Intercept  2.62  0.73  2.81  0.74    (0.26)  (0.15)  (0.52)  (0.27)  Slope with intercept  −2.18  −0.16    (0.05)  (0.09)  Residual variance  1.26  0.53  1.39  0.50    (0.09)  (0.14)  (0.18)  (0.68)  X2(df)  79.98 (62)  67.61 (51)  RMSEA  0.01  0.02  BIC  38275.6  7930.85  CFI/TLI  0.993/0.977  0.967/0.899  Note: Standard errors are reported in parentheses. Reference group is unmarried cohabiting mothers. All models also include variables identifying city of residence. Data are from the Fragile Families and Child Wellbeing Study. Statistical significance levels are as follows: ~ p < 0.10 * p < 0.05 ** p < 0.01 *** p < 0.001 I find sizable differences in the intercept between married mothers and cohabiting mothers. The intercept coefficient on the married variable (B = −0.38, s.e. = 0.10, p < 0.001) indicates that one year after the birth, married mothers worked an average of almost four hours fewer per week than cohabiting mothers, all other factors equal. The slope coefficient on the married variable is close to zero and is not statistically significant (B = 0.07; s.e. = 0.05), indicating that differences between married and cohabiting mothers do not change over time. Additionally, the model shows no differences in work hours between lone and cohabiting mothers, other factors equal (intercept: B = −0.03, s.e. = 0.08; slope: B = 0.01, s.e. = 0.05). Consistent with previous research, I find that age, race/ethnicity, impulsivity, education, and hourly pay rate are all significant predictors (p < 0.05 level) of work hours. All else equal, black and Hispanic mothers work more hours than white, non-Hispanic, or other mothers, and mothers with more human capital work more hours than other mothers. Mothers with more conservative gender attitudes have a lower intercept (B = −0.11, s.e. = 0.03, p < 0.01), indicating that mothers who have gender attitudes that are one standard deviation more conservative than the average mother work a modest one hour less per week than other mothers. Mothers with more children under age three at home work less in the first year (intercept: B = −0.31, s.e. = 0.07) but increase their employment at a faster rate in subsequent years (slope: B = 0.14, s.e. = 0.04). Father characteristics—including having a college degree, weeks of unemployment, and gender attitudes—are also predictive of mothers’ employment hours, and this suggests that researchers should remain attentive to the influence of men on their partners’ employment. Additionally, mothers of the same race/ethnicity as their partner work an average of three hours less per week. The “Slope with Intercept” parameter shows that mothers who start with greater employment hours increase their hours more slowly than mothers who start with fewer employment hours. In additional analyses (available upon request), I tested for interactions between family structure and mother, father, and family characteristics. I found large and statistically significant interactions between the married variable and (mother’s) black racial identity as well as between the married variable and father’s college education. Married black women worked considerably more hours per week than married women of other racial groups, and married women with college-educated husbands worked considerably less than other married women. Notably, both race and education are strongly correlated with divorce rates, and the directions of these interaction effects are consistent with what one would expect if the association of marriage with employment is partially related to family stability. The subsample results for family structure are somewhat different than those from the full sample. The intercept coefficient for marriage (B = −0.24) is in the predicted direction, showing that married mothers work an average of two and a half hours less than cohabiting mothers in the first year, but this difference is not statistically significant at conventional levels. The difference between cohabiting and lone mothers is also not statisticallly significant, though the coefficient is large (B = 0.19), indicating that lone mothers work an average of two hours more than cohabiting mothers in the first year. The slope coefficients for both married and lone mothers are close to zero and not statistically significant, suggesting that family structure differences are stable over time. In summary, the findings for the full sample show that married mothers work fewer hours per week than cohabiting mothers, and that cohabiting and lone mothers have similar employment hours. For the subsample, the coefficient for married is in the expected direction but is not statistically significant and is somewhat smaller in magnitude. Additionally, for the subsample, there is suggestive evidence that lone and cohabiting mothers may differ in their employment hours. Changes in Family Structure and Maternal Employment Trajectories Among the parents in the FFCWS, there are considerable changes in family structure in the five years following the birth. Most married mothers (80 percent) are still married to the baby’s father, but most cohabiting mothers (77 percent) and lone mothers (64 percent) have experienced at least one family structure change. Table 5 shows the share of the sample following each trajectory. Cohabiting mothers’ family structure trajectories are roughly split between the four possibilities of stably cohabiting, marrying, transitioning to living alone, and experiencing multiple changes. Of the lone mothers, 6 percent reported a marriage (and no other change), 24 percent reported a cohabiting partner (and no other change), and 33 percent reported multiple family changes during the first five years after the birth. Table 5. Distribution of Family Structure Trajectories Family structure trajectory  %  Stably cohabiting  8.2  Cohabiting to married  7.6  Cohabiting to lone  9.8  Cohabiting, multiple changes  9.6  Stably married  20.3  Married to divorced  4.9  Stably lone  14.4  Lone to cohabiting  9.3  Lone to married  2.5  Lone, multiple changes  13.3  Family structure trajectory  %  Stably cohabiting  8.2  Cohabiting to married  7.6  Cohabiting to lone  9.8  Cohabiting, multiple changes  9.6  Stably married  20.3  Married to divorced  4.9  Stably lone  14.4  Lone to cohabiting  9.3  Lone to married  2.5  Lone, multiple changes  13.3  Note: Data are from the Fragile Families and Child Wellbeing Study. How are these family structure trajectories expected to correlate with maternal employment? If stability is the main mechanism explaining married mothers’ lower levels of employment, one would expect both stably married mothers and stably cohabiting mothers to have lower levels of employment, all other factors equal. If mothers can predict relationship stability and anticipate changes, one might expect to see divergence on both intercepts and slopes by relationship trajectories. Under this condition, married mothers who eventually divorce are expected to have employment levels similar to unmarried mothers, and cohabiting mothers who eventually marry are expected to have employment similar to stably married mothers. Alternatively, if the mechanism is not economic security or if mothers cannot predict relationship stability, one would expect all married mothers to look similar in the earliest period, with changes in the slopes appearing as mothers divorce in the later periods. Similarly, one would expect all cohabiting mothers to look similar in the intercept levels and differences to emerge in the slopes as mothers experience changes. To test these predictions, I rerun the growth curve models of employment hours using the same full set of covariates in table 4 but using family structure trajectories instead of static measures of family structure. Table 6 shows the coefficients for the family structure trajectory variables. The model shows that stably married mothers work almost seven hours less than stably cohabiting mothers in the first year (intercept: B = −0.67, s.e. = 0.14, p < 0.001), and the gap between these mothers does not close over time (slope: B = 0.10, s.e. = 0.08). Notably, married mothers who later divorce have employment levels in the first year that are similar to stably cohabiting mothers (intercept: B = 0.07, s.e. = 0.18) and dissimilar from stably married mothers. The slope coefficient shows that mothers who divorce increase their hours of work over subsequent years (slope: B = 0.24, s.e. = 0.11, p < 0.05). Cohabiting mothers who marry look more similar to stably married mothers than to stably cohabiting mothers; they report fewer hours of work in the first year (intercept: B = −0.37, s.e. = 0.15, p < 0.05) and no convergence toward stably cohabiting mothers in later years (slope: B = 0.07, s.e. = 0.09). Cohabiting mothers whose unions dissolve increase their work hours over time (slope: B = 0.16, s.e. = 0.08, p < 0.05). Table 6. Selected Results from Growth Curve Models Predicting Hours of Work per Week with Family Structure Trajectories as Predictor Variables   Intercept  Slope  Stably cohabiting  reference group  Cohabiting to married  −0.37*  0.07    (0.15)  (0.09)  Cohabiting to lone  −0.04  0.16*    (0.14)  (0.08)  Cohabiting, multiple changes  0.08  0.07    (0.15)  (0.09)  Stably married  −0.67***  0.10    (0.14)  (0.08)  Married to divorced  0.07  0.24*    (0.18)  (0.11)  Stably lone  −0.04  0.07    (0.14)  (0.08)  Lone to cohabiting  −0.15  0.06    (0.15)  (0.09)  Lone to married  0.16  −0.07    (0.23)  (0.12)  Lone, multiple changes  −0.11  0.17*    (0.14)  (0.08)  Intercept  2.61***  0.63***    (0.27)  (0.16)  Slope with intercept  −0.22    (0.05)  X2(df)  84.92 (69)  RMSEA  0.01  BIC  38303.2  CFI/TLI  0.994/0.980    Intercept  Slope  Stably cohabiting  reference group  Cohabiting to married  −0.37*  0.07    (0.15)  (0.09)  Cohabiting to lone  −0.04  0.16*    (0.14)  (0.08)  Cohabiting, multiple changes  0.08  0.07    (0.15)  (0.09)  Stably married  −0.67***  0.10    (0.14)  (0.08)  Married to divorced  0.07  0.24*    (0.18)  (0.11)  Stably lone  −0.04  0.07    (0.14)  (0.08)  Lone to cohabiting  −0.15  0.06    (0.15)  (0.09)  Lone to married  0.16  −0.07    (0.23)  (0.12)  Lone, multiple changes  −0.11  0.17*    (0.14)  (0.08)  Intercept  2.61***  0.63***    (0.27)  (0.16)  Slope with intercept  −0.22    (0.05)  X2(df)  84.92 (69)  RMSEA  0.01  BIC  38303.2  CFI/TLI  0.994/0.980  Note: Standard errors are reported in parentheses. Reference group is stably cohabiting mothers. Model includes all of the covariates listed in table 1 and includes variables identifying city of residence. Data are from Fragile Families and Child Wellbeing Study. Statistical significance levels are as follows: ~ p < 0.10 * p < 0.05 ** p < 0.01 *** p < 0.001 (two-tailed) The findings from the models of employment by relationship trajectory are consistent with the hypothesis that mothers’ employment is sensitive to perceived relationship stability, but also that marital status is relevant, even among stably partnered mothers. Stably cohabiting mothers work considerably more hours per week than stably married mothers, suggesting that marital status matters, even among couples with stable unions. These findings suggest that differences between married and cohabiting mothers’ employment are not solely about experiences of union instability. Robustness Checks My results (for both logistic regression models and growth curve models) are robust to many model specifications. I obtain the same pattern of findings regarding family structure in models that 1) omit mother’s hourly pay rate and work experience; 2) omit father’s hourly pay rate; 3) omit the variables related to perceived financial need; 4) omit the variables measuring partner homogamy (age difference, same race); and 5) include variables to indicate subsequent births. For hours worked per week, I compare the results from growth curve models to those from OLS regression models that separately predict hours of employment per week at each timepoint; the pattern of results remains similar. Discussion Among a sample of families in US urban areas with a child born between 1998 and 2000, I find that cohabiting mothers are more likely to work in the first year after a birth and work more hours per week in the first five years after a birth than married mothers with similar human capital and demographic characteristics. Contrary to what one would expect if the presence of a partner were driving the relationship between family structure and maternal employment, I find no significant differences in employment between cohabiting and lone mothers. I speculate that cohabiting and lone mothers have similar employment levels because cohabiting mothers are not confident that they can rely on their partners’ income given high union dissolution rates. Thus, cohabiting mothers have similar employment levels to what they would have if they were unpartnered lone mothers. Marriage is associated with lower employment levels, but married women have greater education and wage rates than unmarried women on average. Thus, the employment-promoting effects of human capital almost entirely offset the lower employment associated with marriage. This may explain why aggregate employment rates for married and unmarried mothers of young children in the United States are so similar. Accounting for differences in human capital reveals that among mothers in urban areas, married mothers still work considerably less than unmarried mothers, including cohabiting mothers. The magnitude of family structure differences in maternal employment is fairly large. All other factors being equal, married mothers in this sample are only about half as likely as cohabiting or lone mothers to return to work in the first year after a birth. Is this difference in employment likely to affect child or maternal well-being? Several studies suggest small or no effects of maternal employment during infancy on child well-being for the total population (e.g., Baker and Milligan 2015), but new research suggests negative effects for children in low-income families (Herbst 2014). Additionally, greater maternal employment hours when children are infants are associated with worse maternal health and greater levels of parenting stress (Chatterji, Markowitz, and Brooks-Gunn 2013). Thus, family structure differences in maternal employment during infancy may contribute to family structure inequalities for children in low-income families and to inequalities in mothers’ well-being. Maternal employment differences by family structure continue through early childhood. My estimates suggest that stably cohabiting urban mothers work an average of 6.7 hours more per week than similar stably married urban mothers for all five years after the birth. In terms of magnitude, a difference of seven hours per week is greater than the difference in employment hours between mothers with a high school degree and mothers with a college degree in this sample (5.2 hours per week). Previous research yields competing predictions as to why and how family structure may affect maternal employment. Possible mechanisms include family demographic characteristics (such as number of children), family-related constraints on employment (such as domestic violence), a woman’s own gender attitudes, and her partner’s characteristics and gender attitudes. I find that none of these mechanisms explain much of the marriage difference in maternal employment patterns among mothers in the FFCWS. Instead, I hypothesize that marriage affects maternal employment by providing mothers with more long-term economic security and general stability than cohabitation offers mothers in the contemporary United States. Indeed, I find that characteristics that are strongly associated with marital stability—childhood family structure and racial homogamy among parents—are also strongly associated with maternal employment. Scholars can imagine many other possible tests of my hypothesis that cohabiting mothers work more than married mothers as a hedge against union dissolution and economic deprivation. Lacking an exogenous shock in family structure or random assignment to family structures, all possible tests fall short of the ideal for conclusive evidence. This study examines employment for mothers in urban areas, and mothers in rural areas may have different employment patterns. The data in this analysis were collected from 1998 to 2005, and the gap in employment hours between cohabiting and married mothers may change over time as cohabiting parenthood becomes more institutionalized. Additionally, the FFCWS dataset has limited information on wealth and attitudes regarding financial security, and no items on preferences for employment levels. Better data on these factors would allow for a more thorough test of my hypotheses. Detailed information on maternal employment collected at more frequent intervals would also improve the analysis by reducing reporting errors. To my knowledge, no existing dataset overcomes all of these limitations. Future research might fruitfully investigate employment differences between cohabiting and married mothers in other industrialized countries. The share of children born in cohabiting unions varies substantially across countries, as do the legal responsibilities of cohabitors to each other upon union dissolution (Perelli-Harris and Sánchez Gassen 2012). Moreover, the poverty rates associated with single motherhood vary substantially, largely because of differences in social safety net policies (Brady and Burroway 2012). In this paper, I argue that in the US context, the relatively high employment levels of cohabiting mothers are a form of insurance against high levels of union dissolution and the economic precarity of single-mother families. In countries with stronger social safety nets for lone mothers, more formalized legal obligations between cohabiting partners, or with less union instability among cohabiting couples, I expect that there would be smaller differences in employment levels between married and cohabiting mothers. My findings suggest that, in a key aspect of contemporary family life, cohabiting parenthood differs considerably from married parenthood in the US context. My findings of differences in maternal employment by marital status (adjusted for human capital differences) shed light on important issues in stratification, including economic inequality among women and their children. To the extent that marital childbearing is becoming more selective of advantaged women (McLanahan and Percheski 2008), women’s opportunities to optimally combine paid work and caretaking for their particular career goals and family situations may be becoming more unequal. This inequality may have consequences for women’s health, happiness, and well-being as well as that of their children and partners. Notes 1 The Earned Income Tax Credit provides a considerable income boost to single-mother families but is contingent on maternal employment. 2 I used the ICE command in STATA with an estimation model that included all of the variables included in the analysis plus two additional variables related to fertility history. The imputations were performed using the full sample. 3 The percentage of mothers with top-coded values is very low: 0.7 percent at one year, 1.1 percent at 30 months, and 1.9 percent at 60 months. 4 Thirteen of the 3,132 mothers in the sample report living with a female romantic partner at one or more survey waves. 5 Because the number of married mothers who divorce and subsequently re-partner is small, I did not further disaggregate this category. 6 Survival models predicting the month when the mother returned to work show the same pattern of results regarding family structure as do the logit models. 7 I used MPLUS software and a maximum likelihood parameter estimator (MLR) with TYPE = IMPUTATION to accommodate the imputed data. I took an inductive approach to specifying the employment trajectories. 8 To determine model fit, I examined the chi-square, Bayesian Information Criteria (BIC), comparative fit index (CFI), Tucker-Lewis index (TLI), and root-mean-square error of approximation (RMSEA). 9 Using the data from complete cases only (n = 1,727), the logit models predicting work in the first year for the full sample had a smaller coefficient for the married variable (B = −0.26) and miss conventional cutoffs for statistical significance, but the coefficient was in the expected direction. For the subsample, the estimates with complete cases only (n = 394) have coefficients for the married variable that are almost identical to the coefficients from the models with complete and imputed data. In both the full sample and the subsample, the coefficients for the lone variable were close to zero and not statistically significant. About the Author Christine Percheski is an Assistant Professor in Sociology and a Faculty Fellow at the Institute for Policy Research at Northwestern University. Her research addresses the intersection of family demography and social stratification, focusing on the United States. More specifically, she examines 1) changes in fertility, cohabitation, marriage, and women’s employment; and 2) how these family characteristics are intertwined with economic and health inequalities. Supplementary Material Supplementary material is available at Social Forces online. References Baker, Michael, and Kevin Milligan. 2015. “ Maternity Leave and Children’s Cognitive and Behavioral Development.” Journal of Population Economics  28( 2): 373– 91. Google Scholar CrossRef Search ADS   Bollen, Kenneth, and Patrick Curran. 2006. Latent Curve Models: A Structural Equation Perspective . Hoboken, NJ: John Wiley and Sons. Brady, David, and Rebekah Burroway. 2012. “ Targeting, Universalism, and Single-Mother Poverty: A Multilevel Analysis across 18 Affluent Democracies.” Demography  49( 2): 719– 46. Google Scholar CrossRef Search ADS PubMed  Bureau of Labor Statistics. 2017. Women in the Labor Force: A Databook. Report No. 1065. US Department of Labor. Chatterji, Pinka, Sara Markowitz, and Jeanne Brooks-Gunn. 2013. “ Effects of Early Maternal Employment on Maternal Health and Wellbeing.” Journal of Population Economics  26( 1): 285– 301. Google Scholar CrossRef Search ADS PubMed  Cohen, Philip N., and Suzanne M. Bianchi. 1999. “ Marriage, Children, and Women’s Employment: What Do We Know?” Monthly Labor Review  122: 22– 31. Corman, Hope, Kelly Noonan, and Nancy E. Reichman. 2005. “ Mothers’ Labor Supply in Fragile Families: The Role of Child Health.” Eastern Economic Journal  31( 4): 601– 16. Curran, Patrick, Khawla Obeidat, and Diane Losardo. 2010. “ Twelve Frequently Asked Questions about Growth Curve Modeling.” Journal of Cognition and Development  11( 2): 121– 36. Google Scholar CrossRef Search ADS PubMed  Davis, Shannon N., and Theodore N. Greenstein. 2009. “ Gender Ideology: Components, Predictors, and Consequences.” Annual Review of Sociology  35: 87– 105. Google Scholar CrossRef Search ADS   Davis, Shannon N., Theodore N. Greenstein, and Jennifer P. Gerstein Marks. 2007. “ Effects of Union Type on Division of Household Labor: Do Cohabiting Men Really Perform More Housework?” Journal of Family Issues  28( 9): 1246– 72. Google Scholar CrossRef Search ADS   Fields, Jason. 2003. “Children’s Living Arrangements and Characteristics: March 2002.” Current Population Reports P20-547. Gemmill, Alison, and Laura D. Lindberg. 2013. “ Short Interpregnancy Intervals in the United States.” Obstetrics and Gynecology  122( 1): 64– 71. Google Scholar CrossRef Search ADS PubMed  Goodwin, Paula Y., William D. Mosher, and Anjani Chandra. 2010. “ Marriage and Cohabitation in the United States: A Statistical Portrait Based on Cycle 6 (2002) of the National Survey of Family Growth.” Vital and Health Statistics  23( 28): 1– 45. Han, Wen-Jui, Christopher J. Ruhm, Jane Waldfogel, and Elizabeth Washbrook. 2008. “ The Timing of Mothers’ Employment after Childbirth.” Monthly Labor Review  131( 6): 15– 27. Google Scholar PubMed  Herbst, Chris. 2014. “Are Parental Welfare Work Requirements Good for Disadvantaged Children? Evidence from Age-of-Youngest-Child Exemptions.” Discussion Paper No. 8485. Bonn, Germany: Institute for the Study of Labor. Kalenkoski, Charlene M., David C. Ribar, and Leslie S. Stratton. 2007. “ The Effect of Family Structure on Parents’ Child Care Time in the United States and the United Kingdom.” Review of Economics of the Household  5( 4): 353– 84. Google Scholar CrossRef Search ADS   Kan, Man Yee. 2007. “ Work Orientation and Wives’ Employment Careers An Evaluation of Hakim’s Preference Theory.” Work and Occupations  34( 4): 430– 62. Google Scholar CrossRef Search ADS   Kennedy, Sheela, and Larry Bumpass. 2008. “ Cohabitation and Children's Living Arrangements: New Estimates from the United States.” Demographic Research  19: 1663. Google Scholar CrossRef Search ADS PubMed  Kenney, Catherine T. 2006. “ The Power of the Purse: Allocative Systems and Inequality in Couple Households.” Gender & Society  20( 3): 354– 81. Google Scholar CrossRef Search ADS   Kenney, Catherine T., and Sara S. McLanahan. 2006. “ Why Are Cohabiting Relationships More Violent Than Marriages?” Demography  43( 1): 127– 40. Google Scholar CrossRef Search ADS PubMed  Lee, Sunhwa, Melissa Sills, and Gi-Taik Oh. 2002. “Disabilities among Children and Mothers in Low-Income Families.” Institute for Women’s Policy Research, Publication #D449. Livermore, Michelle M., and Rebecca S. Powers. 2006. “ Employment of Unwed Mothers: The Role of Government and Social Support.” Journal of Family and Economic Issues  27( 3): 479– 94. Google Scholar CrossRef Search ADS   Manning, Wendy D., Pamela J. Smock, and Debarun Majumdar. 2004. “ The Relative Stability of Cohabiting and Marital Unions for Children.” Population Research and Policy Review  23( 2): 135– 59. Google Scholar CrossRef Search ADS   Martin, Joyce A., Brady E. Hamilton, Michelle J. Osterman, et al.  . 2015. “ Births: Final Data for 2013.” National Vital Statistics Reports  64( 1): 1– 68. McLanahan, Sara, and Christine Percheski. 2008. “ Family Structure and the Reproduction of Inequalities.” Annual Review of Sociology  34: 257– 76. Google Scholar CrossRef Search ADS   Meadows, Sarah O., Sara S. McLanahan, and Jeanne Brooks-Gunn. 2008. “ Stability and Change in Family Structure and Maternal Health Trajectories.” American Sociological Review  73( 2): 314– 34. Google Scholar CrossRef Search ADS PubMed  Monte, Lindsay M., and Renee R. Ellis. 2014. “Fertility of Women in the United States: June 2012.” Current Population Reports, P20-575. US Census Bureau, Washington, DC. Musick, Kelly, and Katherine Michelmore. 2015. “ Change in the Stability of Marital and Cohabiting Unions Following the Birth of a Child.” Demography  52( 5): 1463– 85. Google Scholar CrossRef Search ADS PubMed  Nepomnyaschy, Lenna, and Irwin Garfinkel. 2010. “ Child Support Enforcement and Fathers’ Contributions to Their Nonmarital Children.” Social Services Review  84( 3): 341– 80. Google Scholar CrossRef Search ADS   Ono, Hiromi, and Rosemary Yeilding. 2009. “ Marriage, Cohabitation, and Childcare: The US and Sweden.” Social Indicators Research  93( 1): 137– 40. Google Scholar CrossRef Search ADS   Perelli-Harris, Brienna, and Nora Sánchez Gassen. 2012. “ How Similar Are Cohabitation and Marriage? Legal Approaches to Cohabitation across Western Europe.” Population and Development Review  38( 3): 435– 67. Google Scholar CrossRef Search ADS   Raley, Sara B., Marybeth J. Mattingly, and Suzanne M. Bianchi. 2006. “ How Dual Are Dual Income Couples? Documenting Change from 1970 to 2001.” Journal of Marriage and Family  68( 1): 11– 28. Google Scholar CrossRef Search ADS   Reichman, Nancy E., Julien O. Teitler, Irwin Garfinkel, and Sara S. McLanahan. 2001. “ Fragile Families: Sample and Design.” Children and Youth Services Review  23( 4– 5): 303– 26. Google Scholar CrossRef Search ADS   Showalter, Kathryn. 2016. “ Women’s Employment and Domestic Violence: A Review of the Literature.” Aggression and Violent Behavior  31: 37– 47. Google Scholar CrossRef Search ADS   Sorensen, Elaine, and Ariel Hill. 2004. “ Single Mothers and Their Child-Support Receipt: How Well Is Child-Support Enforcement Doing?” Journal of Human Resources  39( 1): 135– 54. Google Scholar CrossRef Search ADS   Thornton, Arland, and Linda Young-DeMarco. 2001. “ Four Decades of Trends in Attitudes Toward Family Issues in the United States: The 1960s through the 1990s.” Journal of Marriage and Family  63( 4): 1009– 37. Google Scholar CrossRef Search ADS   Ventura, Stephanie J., and Christine A. Bachrach. 2000. “ Nonmarital Childbearing in the United States: 1940–99.” National Vital Statistics Report  48( 16): 1– 39. Wu, Huijing. 2017. “Trends in Births to Single and Cohabiting Mothers, 1980–2014.” Family Profiles. Bowling Green, OH: National Center for Family and Marriage Research. Author notes Address for correspondence: Christine Percheski, Northwestern University, 1810 Chicago Avenue, Evanston, IL 60208; telephone: 847-491-2697 (office); e-mail: c-percheski@northwestern.edu © The Author(s) 2018. Published by Oxford University Press on behalf of the University of North Carolina at Chapel Hill. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.

Journal

Social ForcesOxford University Press

Published: Mar 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 12 million articles from more than
10,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Unlimited reading

Read as many articles as you need. Full articles with original layout, charts and figures. Read online, from anywhere.

Stay up to date

Keep up with your field with Personalized Recommendations and Follow Journals to get automatic updates.

Organize your research

It’s easy to organize your research with our built-in tools.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve Freelancer

DeepDyve Pro

Price
FREE
$49/month

$360/year
Save searches from Google Scholar, PubMed
Create lists to organize your research
Export lists, citations
Read DeepDyve articles
Abstract access only
Unlimited access to over
18 million full-text articles
Print
20 pages/month
PDF Discount
20% off