Paternity Leave-Taking and Father Involvement among Socioeconomically Disadvantaged U.S. Fathers

Abstract

In the present study, we examine the associations between the amount of time that U.S. employed fathers took off from work after the birth of a child (i.e., paternity leave-taking) and trajectories of how frequently fathers engage with their children and take responsibility for them. To do so, we analyze longitudinal data on 2,109 fathers from the Fragile Families and Child Wellbeing Study, a data set that contains information from disproportionately socioeconomically disadvantaged families from large urban areas. The results indicate that, one year after birth, paternity leave-taking and lengths of leave are positively associated with fathers’ engagement and responsibility. In addition, paternity leave-taking is positively associated with trajectories of fathers’ responsibility over the first 5 years after birth. Lengths of paternity leave are positively associated with trajectories of fathers’ engagement. Finally, there is evidence that paternity leave-taking and lengths of leave-taking are especially likely to boost fathers’ engagement and responsibility among nonresident fathers. Overall, the findings from the present study suggest that an expansion of paternity leave-taking may encourage higher subsequent levels of father involvement—especially among nonresident fathers.

Although there has been a great deal of research on work-family balance, one prominent policy that promotes work-family balance has been understudied: paternity leave. Paternity leave encourages fathers to practice parenting skills and engage with their children while also fulfilling their expectations as breadwinners. In the process, paternity leave offers support to fathers who seek to fulfill both new and traditional fathering ideals (Galinsky, Aumann, & Bond, 2011; Marsiglio & Roy, 2012; Petts, Knoester, & Li, 2018).

Paternity leave-taking may also lead to higher levels of fathers’ involvement with children. Because father involvement is associated with a number of positive outcomes for children, it is important to analyze the implications of leave-taking for father involvement in greater detail (Pragg & Knoester, 2017; Sarkadi, Kristiansson, Oberklaid, & Bremberg, 2008). Studies from Europe have indicated that fathers who took longer periods of paternity leave when their children were born engaged in childcare tasks more frequently and reported closer relationships with their young children when compared to fathers who took shorter periods of leave (Haas & Hwang, 2008; Huerta et al., 2014; Tanaka & Waldfogel, 2007). There is also evidence from the United States showing that longer periods of paternity leave were associated with higher levels of father involvement (Neponmyaschy & Waldfogel, 2007; Petts & Knoester, 2018; Pragg & Knoester, 2017; Seward, Yeats, Iftekhar, & DeWitt, 2006).

Previous work has helped to shape an initial understanding of paternity leave-taking and its implications for father involvement in the United States, but the conclusions that may be drawn from this research are limited. Overwhelmingly, these studies have focused on father involvement at one point in time, usually when children are infants, and utilized limited measures of involvement. In addition, paternity leave research has almost exclusively studied fathers who live with their children. It is not clear how paternity leave-taking patterns shape subsequent levels of father involvement among nonresident fathers.

We attempt to fill these gaps in the literature by using longitudinal data from a U.S. national sample of urban births from disproportionately disadvantaged families (i.e., high percentages of low socioeconomic status and racial-ethnic minority families) to examine whether taking paternity leave and lengths of leave-taking are associated with fathers’ involvement among both resident and nonresident fathers. It is especially important to study disadvantaged fathers because compared to more advantaged fathers, they experience heightened barriers for leave-taking (e.g., fewer job-related benefits, lower incomes, less education, racial-ethnic discrimination), they are less likely to take leave, and they tend to have more varied levels of father involvement (Milkman & Applebaum, 2013; Pragg & Knoester, 2017). In addition, disadvantaged parents are less likely to be married and live together (Edin & Nelson, 2013), and it is especially unclear how nonresident fathers experience, and may be influenced by, paternity leave (Pragg & Knoester, 2017). Nonetheless, we anticipate that the motivations and potential positive effects of leave-taking will work in similar ways for all fathers.

We also advance research through a consideration of multiple dimensions of father involvement, over time, with close attention to potential issues of selection. We focus on two key dimensions of father involvement over the first 5 years of a child’s life: engagement (measured as engagement in developmentally appropriate activities) and responsibility (measured by how often fathers took responsibility for looking after their child). Father accessibility is also tapped through an investigation of fathers’ resident and relationship statuses. We utilize propensity-score matching techniques to allay concerns about selection effects.

Paternity Leave in the United States

Public policies, gendered practices, and economic positions may constrain or enable fathers to take paternity leave when their children are born. Worldwide, 96% of countries provide paid parental leave, making the United States the only high-income country lacking a statutory paid parental leave policy (Heymann & McNeill, 2013). In the United States, the Family and Medical Leave Act (FMLA) does allow employees who meet eligibility requirements (i.e., worked 1,250 hours in the previous year for an employer with 50+ employees) to take up to 12 weeks of unpaid leave after their children’s births. Also, some workers have access to paid parental leave through their employers and a handful of states now offer paid leave (Heymann & McNeill, 2013; Petts et al., 2018). Yet, this piecemeal system prevents many workers from having access to any parental leave, either paid or unpaid. It is estimated that 40% of employees do not meet the requirements for (unpaid) leave under FMLA, only about 17% of companies offer paid paternity leave to their employees, and around 30% of all U.S. workers have some access to paid leave, through various means (Melamed, 2014; Petts et al., 2018; Society for Human Resource Management [SHRM], 2015).

Many fathers state that their most important role as fathers is to be engaged in their children’s lives (Brandth & Kvande, 1998; Edin & Nelson, 2013), and most men believe that employers should offer paid paternity leave so that fathers are better able to care for and bond with new children (Harrington, Van Deusen, Fraone, Eddy, & Hasas, 2014; Pew Research Center, 2017). Yet, despite fathers’ desires to find ways to better balance their work and family lives, paternity leave-taking by men in the United States is largely discouraged by gendered and economic practices (Acker, 1990; Aumann, Galinsky, & Matos, 2011; Coltrane, Miller, DeHaan, & Stewart, 2013; Heymann &McNeill, 2013).

However, both fathers and children benefit when fathers are able to spend time with their children from birth, making access to paternity leave especially important. Young children benefit from fathers’ care and by engaging in developmentally-appropriate activities (e.g., reading, singing, playing) with them (Edwards, Sheridan, & Knoche, 2010; Lamb & Lewis, 2010; Waldfogel, 2006). In addition, fathering can provide men with joy, a sense of purpose, feelings of generativity, and the nurturance of father identities that emphasize parent-child interactions (Lamb & Lewis, 2010; Marsiglio & Roy, 2012; McKeering & Pakenham, 2000).

Conceptual Framework

The conceptual framework for our study draws upon Lamb and colleagues’ (1985) dimensions of father involvement, father identity theory (Pragg & Knoester, 2017; Stryker, 1968), and the implications of socioeconomic status and other factors for leave-taking and father involvement (Milkman & Applebuam, 2013; Petts et al., 2018; Pragg & Knoester, 2017). According to Lamb and colleagues, there are three dimensions of involvement: engagement (direct interaction with children), responsibility (making decisions about and/or arrangements for children), and accessibility (being available to children). In a majority of studies on father involvement, engagement is the only dimension of involvement measured (Pleck, 2010; Sarkadi et al., 2008). To present a more holistic view of father involvement, we highlight two of the dimensions of father involvement in the present study: engagement and responsibility. We also implicitly consider father accessibility by comprehensively studying the implications of fathers’ resident and relationship statuses.

Based within symbolic interactionism, father identity theory posits that father identities are developed through men’s lived experiences and social interactions, which contribute to the meanings, significance, and perceptions that men attribute to fatherhood and its accompanying roles. As such, when men perceive fatherhood to be a particularly important status in their life, their father identities are likely to be salient. Men’s identities as fathers are especially influential when they have high levels of perceived salience and they are bolstered by their relationship commitments (Pasley, Petren, & Fish, 2014; Pragg & Knoester, 2017; Stryker, 1968).

By providing devoted time to care for children, leave-taking may enhance the experience, bonding, and confidence that men derive from father involvement. Thus, we anticipate that leave-taking will encourage the development, strengthening, and salience of father identities that prioritize father involvement. Relationship commitments to significant others, such as one’s child and co-parent, may further solidify the nature and salience of father identities.

In the absence of paternity leave, the competing demands of traditional and new fatherhood ideals may limit men’s abilities to be as actively involved in their children’s lives as they want (Aumann et al., 2011; Petts & Knoester, 2018). Leave-taking allows fathers to provide care and develop bonds with their children, without much of the burden of balancing work and family responsibilities simultaneously (Rehel, 2014; Tanaka & Waldfogel, 2007). More specifically, leave-taking may give fathers opportunities to learn how to engage in parenting tasks, especially those traditionally done by mothers (Petts & Knoester, 2018; Rehel, 2014). As a result, fathers may become comfortable doing routine and decision-making tasks and view themselves (and be viewed by others) as competent caregivers, which may lead them to be involved fathers. Developing attachments to infants and gaining confidence in parenting does have a lasting influence on fathering behaviors (Cabrera, Fagan, & Farrie 2008). Yet, not all fathers may use their leaves to substantially increase involvement with their new child. They may help their families by working on household chores or taking care of other children that need tending, for example (Brady, Stevens, Coles, Zadoroznyj, & Martin, 2017; Patnaik, 2018).

A number of studies have found that paternity leave-taking is positively associated with father involvement. European and Australian studies have found that fathers who took paternity leave were more likely than were fathers who did not take paternity leave to be involved with playing, feeding, bathing, changing diapers, getting up at night, and taking primary responsibility for children (Hosking, Whitehouse, & Baxter, 2010; Huerta at al., 2014; Tanaka & Waldfogel, 2007). Leave-taking also seemed to increase the likelihood that fathers engaged in developmental tasks and caretaking when children were 2–3 years-old (Huerta et al., 2014). Similarly, studies of American fathers have suggested that fathers who took paternity leave were more likely to change diapers, prepare food for, and help dress their infant children. They were also more likely to take young children to the doctor, read to 2–3 year-old children, and engage with 5 year-old children, as compared to fathers who did not take paternity leave (Huerta et al., 2014; Pragg & Knoester, 2017; Seward et al., 2006).

A father identity perspective also anticipates that longer lengths of paternity leave will lead to higher levels of father involvement. The amount of exposure to opportunities to experience, practice, and become confident in parenting tasks is likely to be associated with higher degrees of nurturance, salience, and commitments to father identities. There is evidence that lengths of paternity leave are positively associated with father involvement. Haas and Hwang (2008) found that the number of days of paternity leave taken by Swedish fathers was positively associated with closer relationships with children; the frequency of taking responsibility for childcare, childcare tasks, and physical care tasks; as well as the number of hours spent with children. Pragg and Knoester (2017) found that the number of weeks of paternity leave taken was positively associated with levels of fathers’ engagement when children were 1 and 5 years-old. Neponmyaschy and Waldfogel (2007) found that taking 2 or more weeks of leave was associated with more frequent father engagement in caretaking tasks. Finally, Petts and Knoester (2018) reported positive associations between length of leave-taking and father engagement in developmental activities and caretaking.

Relationship commitments to significant others may also play a role in shaping father identities and the implications of paternity leave-taking (Petts & Knoester, 2018; Pragg & Knoester, 2017). In particular, the extent to which paternity leave is associated with subsequent father involvement may depend on family structures (i.e., relationship and resident statuses) that are in place. On the one hand, married residential fatherhood is often regarded as part of a package deal of fathering commitments (Townsend, 2002). As such, paternity leave may help to further solidify father identities for married fathers. In contrast, nonresident fathers may become markedly more likely to disengage from involved fathering practices, even if they take paternity leave. This distancing may be especially likely if fathers are no longer romantically involved with the mother. Furthermore, new romantic partners may introduce competing loyalties and encourage even more declines in father involvement (Tach, Mincy, & Edin, 2010).

On the other hand, paternity leave-taking may be especially likely to encourage the salience of father identities among nonresident fathers and boost father involvement compared to married fathers. This may be due to a relative lack of other opportunities to solidify father identities because the family structure and relationship commitments that are embedded in residential (married) fatherhood, which typically encourage the salience of father identities, are absent. Also, paternity leaves may serve as unique opportunities for nonresident fathers to receive structural supports that enable them to experience conveniences in fathering that otherwise may be absent. Further, paternity leave-taking may signal fathers’ willingness to be an engaged parent, reducing the likelihood that mothers act as gatekeepers to limit nonresident father involvement (Fagan & Cherson, 2017). These processes may prove to be especially influential for nonresident fathers who are no longer romantically involved with their child’s mother because father involvement often occurs as part of the dynamics of a romantic relationship with the mother (Edin & Nelson, 2013; Goldberg, 2015).

Nonetheless, father identities that were developed prior to the birth of a child may also shape paternity leave-taking practices and father involvement. Fathers may have preconceived notions of how they expect to be involved in the care and development of their child. Then, they may especially look to participate in their child’s life in ways that help them to fulfill their own expectations and preferred identities as fathers (Pasley et al., 2014; Stryker, 1968). Fathers who have more salient father identities that promote father involvement, as well as have greater commitments to these identities through relationships with significant others, may be more likely to take paternity leave and take longer lengths of leave than will men who view engaged fathering as less salient in their lives. For example, fathers who live with and are married to a co-parent are especially likely to take paternity leave and to take longer lengths of leave (Petts et al., 2018; Pragg & Knoester, 2017). Similarly, paternity establishment is likely to function as a relationship commitment. In turn, having more salient father identities that prioritize father involvement, supported by relationship commitments, is associated with higher levels of engagement throughout children’s lives (Goldberg, 2015; Pragg & Knoester, 2017). Thus, the relationships between paternity leave-taking and father involvement trajectories may be a function of pre-birth characteristics, including indicators of father identities. We consider these associations in our analytic models.

Socioeconomic status (SES) and other factors may also enable or constrain men’s decisions and abilities to take paternity leave and the amount of time that men devote to their children. Because there is no U.S. federal paid parental leave mandate and most workplaces do not offer paid paternity leave to their employees, it is clear that low-income workers, less-educated workers, and those in low-prestige occupations have the least access to paid paternity leave (Melamed, 2014; SHRM, 2015). Not only are these relatively disadvantaged workers the least likely to have access to paid paternity leave, but they are also the least likely to be able to afford to take unpaid leave (Milkman & Appelbaum, 2013).

Furthermore, paternity leave-taking is often stigmatized, especially among low-wage workers and men in low-prestige occupations (Williams, Blair-Loy, & Berdahl, 2013). Indeed, men who request paternity leave, especially men with relatively low levels of SES, receive lower performance ratings, are viewed as inferior workers, and may earn less income in the future (Coltrane et al., 2013; Rege & Solli, 2013). Relatedly, fathers with low levels of education may feel particularly vulnerable to a potential job loss if they take paternity leave (Williams et al., 2013). The SES of partners may also be influential; prior research has found some evidence that paternity leave-taking may be more common when mothers are employed and when mothers earn more income, but these findings are equivocal (Petts & Knoester, 2018; Pragg & Knoester, 2017; Raley, Bianchi, & Wang, 2012).

SES may also be associated with father involvement. Men with higher incomes and levels of education may work in jobs that are highly demanding of their time and reduce the time they are able to spend with their children, and men who work in low-wage sectors may be subjected to unpredictable schedules and inflexible hours (Aumann et al., 2011; Petts et al., 2018). There is also evidence that mothers’ employment activities may be positively associated with father involvement (Raley et al., 2012).

Race/ethnicity and age may also shape leave-taking practices and may have implications for father involvement (Edin & Nelson, 2013; Petts et al., 2018; Pragg & Knoester, 2017). White fathers are more likely to have access to paternity leave and are less likely to be penalized for taking leave than are Black or Hispanic fathers (Nepomnyaschy & Waldfogel, 2007; Pragg & Knoester, 2017; Rudman & Mescher, 2013). Yet, racial/ethnic minority fathers are more likely to be involved in taking responsibility for their children than White fathers and are no different in their levels of engagement and accessibility on weekdays (Hofferth, 2003; Yeung et al., 2001). Also, older fathers may be more able to take paternity leave than younger fathers because they are more established in their careers (Huerta et al., 2014; Nepomnyaschy & Waldfogel, 2007), but having a more established career may limit the amount of time that older fathers spend with their children (Edin & Nelson, 2013; Pragg & Knoester, 2017; Yeung et al., 2001).

Hypotheses

Our conceptual framework and extant research lead to three hypotheses. First, fathers who take paternity leave will have higher levels of subsequent father involvement than will fathers who do not take paternity leave (Hypothesis 1). Second, length of paternity leave will be positively associated with levels of subsequent father involvement (Hypothesis 2). Third, we expect that paternity leave-taking and lengths of paternity leave length will lead to higher levels of father involvement among nonresident fathers compared to resident fathers (Hypothesis 3). We anticipate that paternity leaves may be especially likely to bolster father involvement among nonresident fathers who are no longer romantically involved with their child’s mother, but we will explore the salience of other demographic and relationship variables as well.

Method

Data for the present study come from the Fragile Families and Child Wellbeing Study (FFCW) conducted in the United States. The FFCW is a longitudinal birth cohort study that follows 4,898 children born between 1998 and 2000 and their parents. Fragile families are defined as unmarried parents and their children, and these data consist of an urban sample of births, with high percentages of low-income, racial/ethnic minority, and unmarried parents. The FFCW is meant to be representative of all nonmarital births in cities with populations over 200,000 (Reichman et al., 2001). Parents were interviewed at the hospital shortly after birth (Wave 1) and for follow-up interviews when children were approximately one (Wave 2), three (Wave 3), five (Wave 4), nine (Wave 5), and 15 years-old (Wave 6). Data from Waves 1–4 are used for the present study.

Participants

The sample is restricted to families in which mothers and fathers were interviewed at both Wave 1 and Wave 2 in order to have information on key variables (n = 3,056). The sample is further restricted to families in which fathers were employed at Wave 1 in order to be eligible for paternity leave (n = 2,586). Finally, our sample is limited to fathers who were included in the 18-city subsample because questions on paternity leave were not asked in the first two cities. These conditions result in a sample size of 2,109 families for analyses that predict father involvement when children are infants (Wave 2); for trajectory analyses, the sample is further restricted to families in which fathers and mothers completed the questionnaires in Waves 1–4 (n = 1,494).

Paternity Leave-Taking

Fathers reported on whether they took any time off from work after the birth of the focal child, and how many weeks of leave they took, in the Wave 2 survey. Paternity leave indicates whether fathers took leave (1 = yes). Length of paternity leave is an ordinal-level variable indicating whether fathers took no leave, one week or less of leave, 1–2 weeks of leave, or more than 2 weeks of leave (range: 0–3).

Father Involvement

Fathers were asked about their involvement with the focal child in Waves 2–4. Engagement indicates how many days per week fathers reported playing games with their children, reading to their children, telling stories to their children, and singing songs or nursery rhymes to their children. The mean is used as the scale score (range = 0––7 days per week). Responsibility is taken from mother’s reports on how often the father looked after the focal child when she needed to do things (1 = never to 4 = often).

Father Identity Indicators

We include a number of variables that reflect the nature and salience of father identities. Positive father attitudes is drawn from fathers’ levels of agreement at Wave 1, rated from 1 (strongly disagree) to 4 (strongly agree), on whether (a) “Being a father and raising children is one of the most fulfilling experiences for a man,” (b) “I want people to know that I have a new child,” and (c) “Not being a part of my child’s life would be one of the worst things that could happen to me.” The mean is used as the scale score (α = .73) wherein higher scores indicate more strongly positive attitudes.

Fathers were also asked to identify which aspects of fatherhood were most important to them in the Wave 1 survey. The options were: to provide regular financial support, to teach child about life, to provide direct care (such as feeding, dressing, and childcare), to show love and affection to the child, to provide protection for the child, or to serve as an authority figure and discipline the child. New fathering attitudes denotes fathers who identified either providing direct care or showing love and affection to children as most important (1 = yes). Traditional fathering attitudes identifies (1 = yes) whether fathers agreed at Wave 1 that it is much better for everyone if the man earns the main living and the woman takes care of the home and family. Prenatal involvement indicates (1 = yes) fathers who reported engaging in both of these tasks prior to the birth of the child: (a) gave the mother money to buy things for the baby during the pregnancy and (b) helped in other ways such as providing transportation to the prenatal clinic or helping with chores.

Relationship commitments that fathers have to significant others frequently shape father identities. Relationship status at Wave 1 is categorized as (a) married to the mother (used as reference category), (b) cohabiting with the mother, (c) not residing with the mother, but romantically involved, or (d) not residing with the mother and not involved with her at the time of the birth. Indicators from Wave 1 are also included to assess paternity not being established, number of other children, number of other adults in [child’s] household, and father’s religious participation (i.e., attendance at religious services, ranging from 0 = never to 4 = at least once a week).

Socioeconomic Status and Other Demographic Factors

Measures of SES at Wave 1 include indicators of father’s educational attainment, ranging from 1 (did not complete high school) to 4 (college degree); father’s income, ranging from 0 (less than $10,000) to 8 ($75,000 or more); and mother’s income, ranging from 0 (less than $5,000) to 4 ($20,000 or more). Father’s hours worked is categorized as (a) part-time (less than 35 hours a week), (b) full-time (35–54 hours a week, used as reference category), and (c) more than full time (55 hours a week or more). Mother’s hours worked is coded in a similar way, but an additional category is included to indicate mothers who do not work in paid employment. Father’s occupation type is categorized as (a) professional, (b) labor (used as reference category), (c) service, (d) sales, or (e) other occupational type.

Father’s race/ethnicity is coded as White, Black, Hispanic, or other racial/ethnic group, with White used as the reference category. Other control variables taken from Wave 1 are father’s age (in years), whether the father was born in the U.S. (1 = yes), child gender (1 = male) and whether the child had low birth weight (1 = yes). Three additional control variables are first measured at Wave 2. Child age and an indicator of the length of mother’s time off of work are measured in months. Mother’s involvement is indicated by the mean frequency from 0 (never) to 3 (daily) that mothers spend playing with, singing to, reading to, and telling stories to their child per week.

Time-Varying Characteristics

We also incorporated a number of time-varying variables in the trajectory analyses. These variables were coded in the same way as previously described, but were taken from Waves 2–4 and allowed to vary over these waves. Time-varying variables include: relationship status with the mother (an additional category, single father, was included for resident fathers who were not living with the birth mother in Waves’ 2–4 interviews), new partner for the father, new partner for the mother, number of children, number of other adults in child’s household, religious participation, father’s and mother’s income, father’s and mother’s hours worked (an additional category, does not work, was created for fathers who are not working during Waves’ 2–4 interviews), father’s occupation type, father’s age, child age, and mother’s involvement.

Analytic Strategy

We used ordinary least squares (OLS) and multilevel growth curve models for our analyses. We also run supplementary models designed to address selection concerns. First, OLS regression models are used to analyze whether paternity leave-taking and lengths of paternity leave are associated with each indicator of father involvement at Wave 2. Then, we incorporate interaction terms to assess whether the relationship between paternity leave and father involvement varies by family structure and relationship status.

Second, we used multilevel growth curve models to assess whether paternity leave-taking and lengths of paternity leave are associated with longitudinal patterns of father involvement from Wave 2 to Wave 4. Unconditional growth models, with years after birth used as the indicator of time, were first examined to determine the shape of the growth curve trajectories by testing model fit with linear, quadratic, and cubic specifications of the time indicator. A quadratic model was the best fitting model to predict trajectories of engagement, and a linear model was the best fitting model to predict trajectories of responsibility. A random effect term for years after birth was also included in all models (tests for additional random effects and slopes did not improve the fit of the models). To estimate the intercept, years after birth was rescaled so that 0 is used to estimate the intercept (0 for Wave 2, 2 for Wave 3, and 4 for Wave 4). All continuous variables (with the exception of length of paternity leave) were mean-centered to allow for easier interpretation.

Substantively, we first examine multilevel models that include only indicators of paternity leave-taking as predictors. We then analyze full multilevel models that include all control variables. In these models, time-invariant variables are taken from Wave 1 (i.e., race/ethnicity, nativity, education, child gender, low birth weight, father identity variables, and length of maternity leave). Time-varying variables are taken from each wave (Waves 2-–4) to allow these to vary over time. Similar to the Wave 2 analyses, we incorporate interaction terms to assess whether the relationship between paternity leave and father involvement varies by family structure and relationship status. In all models, multiple imputation is used to account for missing data (combined results from five imputed models are used).

Finally, there may be factors that select fathers into both taking paternity leave and being highly involved in their children’s lives. Unfortunately, eliminating potential selection effects due to unobserved factors (i.e., variables not measured in the FFCW) is challenging. To attempt to minimize the selection problem, we took two approaches to address selection based on observed factors in supplementary models, in addition to including control variables in all models. First, we utilize propensity score matching (PSM) to examine whether relationships between paternity leave-taking and father involvement may be due to selection. Specifically, we first ran a logistic regression model using the Wave 1 control variables to predict paternity leave-taking and then matched fathers in the treatment (took leave) and control (did not take leave) groups with the closest propensity scores. We then used the propensity scores in regression models predicting each indicator of father involvement to estimate average treatment effects. These tests are reported in the text.

Second, we utilize augmented inverse propensity weighted (AIPW) estimators to examine whether associations between lengths of paternity leave-taking and father involvement are due to selection. Similar to PSM, this approach estimates average treatment effects accounting for factors that may select people into certain treatments (e.g., lengths of paternity leave), yet differs from PSM in that AIPW estimators can be used when there are multiple treatments (Cattaneo, 2010). AIPW estimators combine aspects of inverse probability of treatment and regression adjustment weighting and are more efficient than other estimators when models are estimated correctly (Cattaneo, 2010). All Wave 1 controls are included in models that simultaneously predict length of paternity leave and each indicator of father involvement, and results from AIPW models that estimate the average treatment effects of length of paternity leave are reported in the text. In the AIPW models, it is necessary to reclassify the variable for lengths of paternity leave-taking as a series of dummy variables, with no leave as the reference category, in order to focus the analysis on specific “treatment effects” of different lengths of leave.

Results

To further describe our sample and measures, summary statistics for all Wave 1 variables are presented in Table 1. Results show that fathers in this sample actively engage with their children approximately 4 days per week, on average. In addition, the majority of the fathers “often” look after their child. Paternity leave-taking is also common, with 79% (n = 1,670) of fathers taking paternity leave. Yet, just over 70% (n = 1,235) of the leave-takers take only one week of leave or less (results not shown). Thus, nearly 80% (n = 1,674) of fathers can be classified as taking one week of paternity leave or less—if one includes the 21% of fathers who took no leave (n = 439). As anticipated from the sampling strategy of the FFCW (see Reichman et al., 2001), the families are also disproportionately low SES, comprised of racial/ethnic minority fathers (e.g., 73% non-White), and embedded in unmarried family structures (e.g., 66% of fathers not married to the mothers).

Table 1.

Sample Characteristics

	M or %	(SD)	Min	Max
Father Involvement (Wave 2)
Engagement	3.93	(1.96)	0	7
Responsibility	3.47	(0.92)	1	4
Paternity Leave
Paternity Leave-taking	79%	-	0	1
Length of Paternity Leave	1.07	(0.78)	0	3
Father Identity Indicators
Positive Father Attitudes	3.76	(0.40)	1	4
New Fathering Attitudes	66%	-	0	1
Traditional Fathering Attitudes	39%	-	0	1
Prenatal Involvement	93%	-	0	1
Relationship Status
Marrieda	34%	-	0	1
Cohabiting	41%	-	0	1
Nonresident, Relationship with Mother	16%	-	0	1
Nonresident, No Relationship with Mother	9%	-	0	1
Paternity Not Established	15%	-	0	1
Number of Other Children	1.02	(1.18)	0	5
Other Adults in Household	0.59	(0.93)	0	3
Religious Participation	1.91	(1.35)	0	4
SES and Other Demographic Factors
Father’s Characteristics
Education	2.31	(1.00)	1	4
Income	3.36	(2.18)	0	8
Employment
Works Part-Time	11%	-	0	1
Works Full-Timea	71%	-	0	1
Works more than Full-Time	18%	-	0	1
Professional Occupation	17%	-	0	1
Labor Occupationa	49%	-	0	1
Sales Occupation	8%	-	0	1
Service Occupation	24%	-	0	1
Other Occupation	2%	-	0	1
Does Not Work	5%	-	0	1
Works Part-Time	30%	-	0	1
Works Full-Timea	62%	-	0	1
Works more than Full-Time	3%	-	0	1
Age	28.18	(6.93)	16	57
Race/Ethnicity
Whitea	27%	-	0	1
Black	42%	-	0	1
Latino	26%	-	0	1
Other Race	5%	-	0	1
U.S. Native	84%	-	0	1
Child Characteristics
Child Age	15.60	(3.90)	5	30
Child is Male	52%	-	0	1
Low Birth Weight	9%	-	0	1
Mother Characteristics
Mother’s Income	1.43	(1.66)	0	4
Length of Mother’s Time Off	2.46	(3.14)	0	12
Mother Involvement	2.27	(0.57)	0	3

Note. n = 2109. Min and Max refer to minimum and maximum actual values.

^aUsed as reference category.

We now turn to our OLS and growth curve analyses. First, to examine whether paternity leave-taking is associated with father involvement one year after birth, as hypothesized, OLS regression models are used to predict father involvement at Wave 2 (more basic versions of our models are available in the online supplement Table 1s – Table 3s). Results are presented in Table 2. As shown in Model 1, fathers who took paternity leave were more engaged with their children than were fathers who did not take paternity leave (b = 0.28, p = .008). This result is consistent when using PSM (b = 0.31, p = .031), suggesting that the relationship between paternity leave-taking and engagement was not due to selection on observed factors (PSM results are available in the online supplement Table 4s). Similarly, in Model 3, fathers who took paternity leave were more likely to look after their children (b = 0.13, p = .007) than were fathers who did not take paternity leave. Yet, this result becomes nonsignificant with a standard two-tailed test in PSM models (b = 0.19, p =.054). Overall, the results from predicting father involvement one year after birth provide some support for our first hypothesis that paternity leave-taking is positively associated with subsequent father involvement.

Table 2.

Results from OLS Regression Models Predicting Father Involvement One Year after Birth (Wave 2)

	Engagement				Responsibility
	Model 1		Model 2		Model 3		Model 4
Variables	B	p	B	p	B	p	B	p
Paternity leave	0.28	.008			0.13	.007
Length of Paternity Leave			0.24	<.001			0.04	.089
Father Identity Indicators
Positive Father Attitudes	0.22	.029	0.21	.042	0.00	.881	0.00	.901
New Fathering Attitudes	0.02	.780	0.02	.771	0.02	.533	0.03	.525
Traditional Fathering Attitudes	0.01	.902	0.03	.764	0.02	.724	0.02	.691
Prenatal Involvement	0.41	.024	0.41	.023	0.02	<.001	0.33	<.001
Cohabiting	0.01	.939	0.02	.900	0.31	.001	−0.18	.001
Nonresident, in relationship	−0.28	.062	−0.27	.070	−0.18	<.001	−0.36	<.001
Nonresident, no relationship	−0.62	.001	−0.60	.001	−0.35	<.001	−0.90	<.001
No Paternity Established	−0.59	<.001	−0.58	<.001	−0.89	.135	−0.10	.101
Number of Other Children	−0.14	.001	−0.13	.001	−0.09	.011	−0.05	.010
Other Adults in Household	−0.10	.070	−0.09	.088	−0.05	.498	−0.01	.534
Religious Participation	0.10	.003	0.09	.003	−0.01	.905	−0.00	.896
SES and Other Demographic Factors
Education	0.00	.900	−0.01	.942	−0.02	.486	−0.02	.465
Works Part-Time	−0.05	.674	−0.06	.599	0.00	.942	−0.00	.864
Works more than Full Time	0.01	.752	0.01	.760	−0.07	.150	−0.07	.140
Professional Occupation	0.15	.284	0.13	.357	0.00	.943	−0.00	.870
Sales Occupation	0.27	.075	0.28	.067	0.02	.814	0.02	.799
Service Occupation	0.04	.764	0.02	.900	−0.09	.046	−0.09	.041
Other Occupation	−0.24	.386	−0.26	.355	−0.16	.152	−0.16	.258
Income	0.02	.756	0.01	.786	0.03	.012	0.03	.022
Does Not Work	−0.59	.005	−0.62	.003	−0.05	.557	−0.06	.529
Works Part-Time	0.02	.855	0.02	.817	−0.01	.727	−0.01	.729
Works more than Full-Time	0.23	.441	0.23	.431	0.07	.568	0.07	.572
Mother’s Income	0.01	.714	0.01	.780	0.01	.560	0.01	.599
Age	0.00	.638	0.00	.600	0.00	.359	0.00	.350
Black	−0.12	.255	−0.12	.256	0.16	.004	0.15	.006
Latino	−0.08	.487	−0.08	.462	0.17	.006	0.16	.007
Other Race	−0.04	.873	−0.07	.799	−0.15	.256	−0.16	.216
U.S. Native	0.29	.022	0.30	.019	−0.04	.424	−0.05	.407
Child Age	0.01	.444	0.01	.436	−0.02	.001	−0.02	.001
Child is Male	0.03	.660	0.03	.676	−0.02	.742	−0.02	.737
Low Birth Weight	−0.01	.899	−0.02	.866	−0.09	.191	−0.09	.187
Length of Mother’s Time Off	−0.01	.681	−0.01	.680	−0.00	.564	−0.00	.583
Mother Involvement	0.35	<.001	0.35	<.001	0.05	.136	0.05	.132
R2	0.11		0.12		0.15		0.15

Note. n = 2109

Next, we consider whether there is initial support for our second hypothesis that predicted positive associations between lengths of leave-taking and father involvement. Models 2 and 4 of Table 2 present the associations between lengths of paternity leave-taking and subsequent father involvement at Wave 2. The results in Model 2 suggest that longer periods of leave were associated with higher levels of engagement (b = 0.24, p < .001). AIPW models (available in the online supplement Table 4s) provide additional support for these findings. In these models, fathers who took 1–2 weeks of leave engaged in activities 13% more frequently (b = 0.50, p = .004), and fathers who took more than 2 weeks of leave engaged in activities 20% more frequently (b = 0.74, p = .001), than did fathers who did not take leave. In Model 4, there seems to be a lack of evidence with a standard two-tailed test that longer periods of leave were positively associated with taking responsibility for children (b = 0.04, p =.089), but there is some support for a significant association in the AIPW models; fathers who took one week of leave or less were 5% more likely to look after their children than were fathers who did not take leave (b = 0.16, p = .007). Overall, the results from predicting father involvement one year after birth generally provide support for our second hypothesis that longer periods of paternity leave are associated with higher levels of subsequent father involvement.

In our final OLS analysis, we consider whether there is initial support for our third hypothesis that anticipated that nonresident fathers would be more likely to have their father involvement patterns altered by paternity leave-taking experiences. In Table 3, interaction terms are added to the full models in order to assess whether the relationships between paternity leave-taking and father involvement at Wave 2 varied by family structure and relationship status. As shown in Model 3, compared to married fathers, paternity leave is especially likely to lead to increases in responsibility at Wave 2 among fathers who were nonresident at birth, but in a romantic relationship with the mother (b = 0.47, p < .001). Thus, there is some evidence in support of our third hypothesis that anticipated that paternity leave may be especially beneficial in encouraging nonresident fathers to increase their involvement with their children.

Table 3.

Family Structure Moderating Influences in Predicting Father Involvement One Year after Child’s Birth (Wave 2)

	Engagement				Responsibility
	Model 1		Model 2		Model 3		Model 4
Variables	B	p	B	p	B	p	B	p
Paternity leave	0.07	.748			0.02	.837
Length of Paternity Leave			0.16	.089			0.02	.554
Cohabiting	−0.15	.563	−0.06	.755	−0.20	.100	−0.15	.072
Nonresident, with Mother	−0.64	.031	−0.52	.020	−0.70	<.001	−0.49	<.001
Nonresident, not with Mother	−0.83	.007	−0.76	.003	−0.96	<.001	−0.95	<.001
Paternity Leave x Cohabiting	0.19	.499			0.01	.937
Paternity Leave x Nonresident,with Mother	0.46	.150			0.47	<.001
Paternity Leave x Nonresident, Not with Mother	0.26	.468			0.05	.735
Length of Leave x Cohabiting			0.06	.637			−0.03	.609
Length of Leave x Nonresident, with Mother			0.25	.123			0.14	.057
Length of Leave x Nonresident, Not with Mother			0.17	.428			0.06	.564
R2	0.11		0.12		0.16		0.15

Note. n = 2109. Each model also includes all of the other predictor variables (i.e., father identity indicators, SES, and other demographic factors) presented in Table 2.

We now turn to our growth curve analyses of whether paternity leave-taking experiences are associated with trajectories of father involvement over the first 5 years of children’s lives. To test our first hypothesis that predicted positive associations between paternity leave-taking and father involvement, we first estimated multilevel models that include only indicators of paternity leave-taking as substantive predictors of trajectories of father engagement. Then, we add control variables to the models and display them in Table 4. At the bivariate level (results available in the online supplement Table 2s), the results indicate that fathers who took paternity leave had a higher initial level of engagement at Wave 2 than did fathers who did not take leave (b = 0.55, p < .001), that this advantage increased from Wave 2 to Wave 3 (b = 0.22, p = .038), and that it then narrowed between Wave 3 and Wave 4 (b = −0.05, p = .042). Once control variables were entered into the model (as shown in Model 1 of Table 4), paternity leave-taking was no longer significantly related to initial status (i.e., Wave 2) differences in engagement. Supplementary analyses (results available in the online supplement Table 2s) suggest that this relationship is explained away by father identities. That is, fathers with more salient father identities were more likely to take paternity leave and controlling for this explains away the relationship between leave-taking and trajectories of engagement.

Table 4.

Results from Growth Curve Models Predicting Trajectories of Father Involvement with Children Ages 1–5 (Waves 2–4)

	Engagement				Responsibility
	Model 1		Model 2		Model 3		Model 4
Variables	B	p	B	p	B	p	B	p
Initial Status	3.60	<.001	3.52	<.001	3.36	<.001	3.41	<.001
Paternity leave	0.10	.398			0.10	.043
Length of Paternity Leave			0.20	.001			0.02	.390
Linear Rate of Change	−0.53	.014	−0.38	.064	−0.02	.706	−0.01	.814
Paternity leave	0.21	.053			0.01	.462
Length of Paternity Leave			−0.01	.681			0.00	.796
Quadratic Rate of Change	0.15	.006	0.12	.020
Paternity Leave	−0.04	.095
Log-likelihood	−6899		−6896		−3843		−3846

Note. n = 1494 (4482 person-years). Each model also includes all of the other predictor variables (i.e., father identity indicators, SES, and other demographic factors) presented in Table 2, as well as the time-varying characteristics that are listed in the Method section.

In relation to responsibility, as shown in Model 3 of Table 4, fathers who took paternity leave were more likely to take responsibility for their children at Wave 2 than did fathers who did not take leave; this difference persists over time, even after all control variables are included (b = 0.10, p = 043). This difference also persists in PSM models (results available in the online supplement Table 4s), suggesting that the relationship between leave-taking and trajectories of responsibility were not due to selection on observed characteristics (b = 0.13, p = .001). Thus, we find partial support in the growth curve analysis for our first hypothesis that paternity leave-taking would be associated with relatively higher levels of father involvement.

Next, we test our hypothesis that anticipated positive associations between lengths of leave and trajectories of father involvement. In the bivariate model (results available in the online supplement Table 2s), longer leaves were associated with initially higher levels of father engagement (b = 0.41, p < .001), and this advantage persisted over time. As shown in Model 2 of Table 4, after control variables are included in the models, lengths of paternity leave continued to be significantly associated with trajectories of more frequent engagement (b = 0.20, p < .001). These results persist in AIPW models (results available in the online supplement Table 4s) and indicate that fathers who took one week of paternity leave or less engaged in activities 6% more frequently (b = 0.19, p = .027), fathers who took 1–2 weeks of leave engaged in activities 12% more frequently (b = 0.40, p = .001), and fathers who took more than 2 weeks of leave engaged in activities 13% more frequently (b = 0.46, p = .007) than did fathers who did not take leave.

At the bivariate level (results available in the online supplement Table 3s), longer leaves were also associated with initially higher levels of responsibility (b = 0.15 p < .001), and this advantage persisted over time. Yet, after control variables are added, as shown in Model 4 of Table 4, this association is not statistically significant. Supplementary analyses suggest that the significant bivariate relationship is partially explained away by father identities. That is, fathers with more salient father identities are more likely to take longer periods of paternity leave, and controlling for this variable explains away some of the association between lengths of leave and trajectories of responsibility. Nonetheless, the AIPW findings (results available in the online supplement Table 4s) contribute some support for a significant relationship between lengths of paternity leave and father responsibility. Compared to taking no leave, taking a week or less of paternity leave is associated with a 4% increase in taking responsibility for children (b = .13, p = .001). Overall, we find general support in the growth curve models for our second hypothesis about greater lengths of paternity leave predicting more elevated trajectories of father involvement.

Finally, we consider whether the relationships between leave-taking experiences and trajectories of father involvement depend on family structure, as hypothesized. The growth curve interaction models are presented in Table 5. As shown in Model 1, among nonresident fathers who were not involved with the mother, paternity leave was more positively associated with levels of engagement with children (b = 0.91, p = .005), but this advantage declined over time (b = −0.23, p = .015). In further support of our hypothesized interaction effects, as shown in Model 2 of Table 5, initial levels of engagement are higher for nonresident fathers who were not involved with the mother and who took longer periods of leave (b = 0.40, p = .012) compared to married fathers, and these advantages persisted over time. However, the relationships between paternity leave experiences and trajectories of responsibility did not seem to vary by family structure, as shown in Models 3 and 4. Overall, there is some evidence in support of our third hypothesis: Although resident fathers had higher levels of involvement than nonresident fathers did, regardless of leave-taking practices, paternity leave-taking was seemingly more effective at encouraging father involvement among nonresident fathers relative to resident fathers. At least in terms of trajectories of father engagement, nonresident fathers who were not in a relationship with the mother appeared to be most likely to be affected by paternity leave-taking experiences.

Table 5.

Family Structure Moderating Influences in Predicting Father Involvement with Children Ages 1–5 (Waves 2–4)

	Engagement				Responsibility
	Model 1		Model 2		Model 3		Model 4
Variables	B	p	B	p	B	p	B	p
Initial Status	3.93	<.001	3.54	<.001	3.40	<.001	3.42	<.001
Paternity leave	−0.08	.703			0.05	.546
Length of Paternity Leave			0.16	.046			0.01	.786
Cohabiting	0.10	.675	0.17	.339	0.02	.850	0.01	.993
Single Father	−0.37	.521	−0.47	.240	0.07	.748	−0.08	.647
Nonresident, with Mother	−0.93	.003	−0.34	.145	−0.74	<.001	−0.51	<.001
Nonresident, not with Mother	−1.95	<.001	−1.68	<.001	−1.24	<.001	−1.23	<.001
Father has Other Partner	−0.15	.530	−0.16	.465	−0.43	<.001	−0.44	<.001
Mother has Other Partner	−0.45	.091	−0.36	.136	−0.53	<.001	−0.51	<.001
Paternity Leave x Cohabiting	0.05	.859			−0.01	.897
Paternity Leave x Single Father	−0.32	.619			−0.42	.094
Paternity Leave x NR, with mom	0.61	.080			0.27	.067
Paternity Leave x NR, not w/ mom	0.91	.005			0.15	.226
Length of Leave x Cohabiting			−0.04	.744			0.01	.840
Length of Leave x Single Father			−0.03	.917			−0.16	.257
Length of Leave x NR, with mom			0.02	.908			−0.03	.702
Length of Leave x NR, not w/mom			0.40	.012			0.11	.109
Linear Rate of Change	−0.65	.015	−0.39	.062	−0.02	.794	−0.01	.877
Paternity leave	0.15	.272			0.01	.757
Length of Paternity Leave			0.01	.815			−0.00	.875
Cohabiting	0.01	.947	0.00	.959	−0.04	.288	−0.04	.201
Single Father	0.07	.779	0.13	.407	−0.05	.596	0.01	.893
Nonresident, with Mother	0.20	.116	0.10	.272	0.02	.677	−0.05	.267
Nonresident, not with Mother	0.15	.105	0.10	.156	0.01	.851	0.01	.693
Father has Other Partner	−0.01	.850	−0.02	.737	0.03	.412	0.03	.356
Mother has Other Partner	−0.02	.838	−0.01	.858	0.10	.015	0.10	.020
Paternity Leave x Cohabiting	−0.04	.693			0.04	.401
Paternity Leave x Single Father	0.21	.416			0.16	.161
Paternity Leave x NR, w/mom	−0.21	.023			−0.03	.614
Paternity Leave x NR, not w/ mom	−0.23	.015			−0.03	.564
Length of Leave x Cohabiting			−0.02	.643			0.02	.297
Length of Leave x Single Father			0.09	.457			0.05	.437
Length of Leave x NR, w/mom			−0.06	.378			0.05	.166
Length of Leave x NR, not w/mom			−0.09	.053			−0.03	.213
Quadratic Rate of Change	0.15	.015	0.12	.023
Paternity leave	−0.01	.621

Note. n = 1494 (4482 person years). Each model also includes all of the other predictor variables (i.e., father identity indicators, SES, and other demographic factors) presented in Table 2, as well as the time-varying characteristics that are listed in the Method section.

Discussion

The current study adds to our understanding of paternity leave practices in the United States and the implications that they have for trajectories of father involvement. Our findings provide support for our hypotheses that paternity leave-taking and longer lengths of leave are likely to lead to higher levels of subsequent father involvement. In addition, nonresident fathers seem to be especially likely to become more involved with their children as a result of taking paternity leave and taking longer leaves. Overall, the findings of our study are supportive of the sparse previous research on patterns of paternity leave-taking and their implications for subsequent levels of father involvement.

First, consistent with previous research (Nepomnyaschy & Waldfogel, 2007; Pragg & Knoester, 2017), we find evidence that paternity leave-taking is very common. Even among a sample of disadvantaged urban fathers, 79% (n = 1,670) of employed fathers took time off of work when new children were born, yet paternity leave durations were quite short. Specifically, from among the 79% (n = 1,670) of the fathers who took leave, over 70% (n = 1,235) took only one week or less of paternity leave.

Second, consistent with our first hypothesis, our results suggest that fathers who took paternity leave became relatively more involved with their children compared to fathers who did not take paternity leave. These findings build on previous research from the United States and elsewhere (Huerta et al., 2014; Petts & Knoester, 2018; Pragg & Knoester, 2017). For example, we find that paternity leave-taking was positively associated with levels of fathers’ engagement one year after birth. Also, paternity leave-taking was positively associated with more elevated trajectories of responsibility over the first 5 years of a child’s life. It is notable that PSM models support these findings.

Third, in support of our second hypothesis, we found that longer lengths of paternity leave led to relatively higher levels of subsequent father involvement. For example, we found that lengths of leave-taking were positively associated with father’s engagement one year after the birth of a new child, and there was also evidence for increases in father’s responsibility in the PSM model. In addition, longer lengths of leave predicted elevated trajectories of father engagement over the first 5 years of a child’s life. There was broad support for these findings in the AIPW models, suggesting that these associations were not due to selection on observed characteristics. These findings also support and extend research in the United States and elsewhere (Haas & Hwang, 2008; Petts & Knoester, 2018; Pragg & Knoester, 2017).

Finally, as anticipated in our third hypothesis, family structure and relationship status moderated these relationships. Specifically, paternity leave-taking and lengths of leave were especially likely to encourage nonresident fathers to be more involved in their children’s lives. For example, among fathers who were not residing with the mothers, but were romantically involved with them at the time of their child’s births, taking leave led to higher levels of father responsibility one year after birth compared to the implications of taking paternity leave among married fathers. Similarly, taking paternity leave led to significantly elevated trajectories of father engagement among fathers who were not living with their children and not romantically involved with the mothers compared to fathers who were married to the mothers and residing with their children. In addition, longer lengths of paternity leave predicted more elevated trajectories of father engagement among nonresident fathers who were not involved with the mothers compared to married fathers. These results are particularly noteworthy because previous research has only just begun to consider the patterns and implications of leave-taking among nonresident fathers (Pragg & Knoester, 2017).

Limitations and Future Research Directions

There are some limitations to acknowledge in our study. First, we have little knowledge of the programs that fathers used to take leave or the options that they might have had at their disposal. Fathers may have taken leave through workplace parental leave programs (which may be paid or unpaid), FMLA, or by using other ways of taking time off (e.g., vacation or sick days). Unfortunately, fathers in the FFCW were not asked about the options that were available to them or about how they were able to take leave. More research is needed to better understand the paternity leave options that are available to fathers and the programs that they use when they do take paternity leave.

Second, we do not present extensive comparisons between the implications of paid and unpaid leave practices in the current study. These analyses are beyond the scope of the present study, yet supplementary analyses utilizing information on whether fathers took paid leave suggest that there are few differences in the relationships between paternity leave-taking (and length of leave) and father engagement after one disaggregates paid and unpaid leave. However, there was some evidence that lengths of paid leaves are more likely to be significantly associated with taking responsibility for children compared to lengths of unpaid leaves. Future research should focus on how type of leave (i.e., paid vs. unpaid) may influence family well-being.

Third, we were unable to consider the potential influences of maternity leave practices to the extent that we would have liked. The FFCW does not contain ideal maternity leave information, but we were able to incorporate indicators of how much time off from work that mothers took around the time of their children’s births. The packages of maternity and paternity leave that families use, and the implications of these combinations, are also important topics for future research.

Finally, we should note that our indicator of father responsibility is limited, that paternal bonding may be a feature of father involvement that should be more fully explored, and that nonresident father attrition patterns offer potential challenges to the validity of our findings. More comprehensive measures of the extent and types of father responsibility should be used in future research. In addition, our conceptual framework assumes processes of paternal bonding that may not automatically occur; future research should more fully explore these processes and their links to both paternity leave-taking and father involvement (Brady et al., 2017; Habib & Lancaster, 2005). In terms of nonresident father selectivity, in supplementary analyses, we found that 22% of nonresident fathers who worked at Wave 1 did not complete the survey at Wave 2 (when paternity leave information was given). However, among nonresident fathers who completed the Wave 2 survey, there was not a statistically significant difference in leave-taking between fathers who participated in later waves and those who dropped out (64% vs. 62% took leave). Moreover, there was not a statistically significant difference in involvement at Wave 2 between nonresident fathers who participated in later waves and those who dropped out based on leave-taking.

Despite these limitations, our study offers a number of important contributions to extant research. Our study offers new and more convincing evidence of the relationships between paternity leave-taking practices and subsequent levels of father involvement. Our study is unique in its focus on an U.S. urban sample of disadvantaged families who are followed over time. It is also unique in its use of more advanced statistical techniques (e.g., propensity score matching and growth curve analyses) that are used to consider the implications of potential selection effects and associations of paternity leave-taking practices with trajectories of father involvement. Finally, our study contributes the first known in-depth analysis of how family structures moderate the implications of paternity leave-taking practices for subsequent trajectories of father involvement.

Practice Implications

The present findings may be of special interest to families, practitioners, and policymakers who are trying to assuage work-family conflicts and encourage increased levels of father involvement. The results should also be meaningful for concerned citizens who are seeking to restructure U.S. practices of work and family so that families with children are more generously supported, opportunities for parental involvement are more equally distributed between mothers and fathers, and people with diverse family forms all have access to, and support for, leave-taking.

Specifically, the findings of our study suggest that paternity leave-taking opportunities, and more generous opportunities, may be especially valuable for promoting subsequent levels of father involvement. Paternity leave-taking experiences may be especially important and influential for nonresident fathers—and perhaps nonresident fathers who are no longer romantically involved with the mother of their child in particular. Paternity leaves offer unique supports for fathers who seek to be more involved with their children. Taking leave may also lead to gate-opening practices that include facilitating and encouraging father involvement from mothers (Fagan & Cherson, 2017; Pragg & Knoester, 2017). As such, results from our study suggest that expanding paternity leave opportunities and encouraging leave-taking may help to strengthen (especially disadvantaged) U.S. families by encouraging more father involvement.

Conclusion

The results of the present study offer consistent and rigorous evidence that paternity leave-taking and longer lengths of paternity leave seem to boost levels of father involvement for up to five years after a child’s birth—especially among nonresident fathers. It appears that a devoted window of time that allows employed fathers to focus on family interactions after the arrival of a new child can encourage fathers to bond with their children, engage in fathering activities, and help in developing father identities that lead to greater levels of subsequent father involvement. Such opportunities seem to be especially beneficial to nonresident fathers, who may lack the relationship commitments and structure of resident (married) fatherhood. As a result, paternity leave may especially help to facilitate the development of father identities among nonresident fathers.