Paid Family Leave, Fathers' Leave Taking, and Leave-Sharing

Paid family leave (PFL) policies provide workers with paid time off from work to care for their newborn or newly-adopted children, as well as other sick or elderly family members. PFL policies may be especially important for new parents, who often struggle with balancing the competing needs of work and family responsibilities. PFL allows them to stay home to care for and bond with their newborn children, and then return to work with minimal career interruptions. Although historically these policies only applied to women, some modern PFL programs including the three existing state-level programs in the United States in California, New Jersey, and Rhode Island—are gender-neutral and cover both male and female workers. Yet while an extensive literature has studied the implications of policies promoting work-family balance for women, much less attention has been devoted to the corresponding effects of these policies on men, or on joint leave-taking decisions within families where both parents work. This lack of understanding about the effects of such policies for fathers is particularly salient in view of recent evidence that fathers now report equal or greater levels of work-family conflict as do mothers (Auman et al., 2011; Rehel and Baxter, 2015).

This paper begins to fill this gap by analyzing how American fathers respond to the introduction of the country’s first-in-the-nation large-scale PFL policy in California. We also break new ground by studying how fathers and mothers in dual-earner households share leavetaking responsibilities.

While advocates frequently credit work-family programs such as PFL with promoting gender equality and supporting women in their careers, increased leave-taking by fathers may have benefits that extend beyond any gains to women in the workplace. To the extent that gendered patterns of childcare provision develop early on, even relatively small changes in initial paternity leave decisions may have important long-run implications. Supporting this idea, several studies suggest that the amount of time fathers spend in childcare is correlated with the generosity of paternity leave policies (Fuwa and Cohen, 2007; Boll et al., 2014) and that fathers who take more leave around the time of birth may be more involved in childcare not only during the period of leave, but throughout the child’s life (Haas, 1990; Nepomnyaschy and Waldfogel, 2007; Tanaka and Waldfogel, 2007; Haas and Hwang, 2008). While these correlations may not be surprising given positive selection into leave-taking, recent evidence from Quebec suggests there may be a causal effect of exposure to parental leave policies on long-term paternal involvement as well (Patnaik, 2015). There could also be other benefits related to the health and well-being of family members and longer-term effects on gender norms and role-models (Ray, Gornick, and Schmitt, 2008).

The extent to which such benefits may materialize crucially depends on a first-order question of whether fathers take up leave when PFL is offered. This question remains largely unanswered, especially in the U.S. context. While evidence suggests that women increase their leave-taking in response to PFL availability (Rossin-Slater, Ruhm, and Waldfogel, 2013; Baum and Ruhm, 2015), men may be less receptive to taking time off work either because they view it as too costly or because of stigma associated with gender stereotypes and norms about childcare. Additionally, while most employers still do not offer paid parental leave, more cover paid maternity leave than paternity leave (Klerman, Daley, and Pozniak, 2012). To the extent that it is harder for them to take time off in the absence of a government-provided PFL policy, new fathers might be more responsive to widespread access to paid leave. Finally, the benefits of paternity leave may depend on whether fathers stay home with their child alone, or whether they take leave alongside mothers. We know even less about the related question of how the availability of paid leave influences the joint leave-taking behavior of fathers and mothers in households where both parents work. This paper provides some of the first quasi-experimental evidence on these questions

Our analysis exploits the introduction of California’s first-in-the-nation paid family leave (CA-PFL) program in July 2004. CA-PFL offers six weeks of paid leave to new parents, with a 55 percent wage replacement rate up to a ceiling (a maximum benefit of $1,104 per week in 2015) and with almost universal eligibility among private sector workers. We use data from the 2000 Census and the 2000-2013 waves of the American Community Survey (ACS) together with difference-in-difference (DD) and difference-in-difference-in-difference (DDD) methods to identify the causal effects of CA-PFL on paternal leave-taking. Our preferred DDD specification compares employed fathers of infants in California to employed fathers of children aged one to three, relative to corresponding fathers of the same age children in other states, before and after the introduction of California’s PFL program. We perform an analogous analysis of mothers to enable comparisons of effects across parental gender. Further, we explore within-household leave-taking behavior in households where both parents work by studying “father only” (i.e., father is on leave while mother is at work), “mother only” (i.e., mother is on leave while father is at work), “both parent” (i.e., both the father and mother are on leave at the same time), and “either parent” (i.e., either the mother or the father is on leave) leave-taking outcomes.

In our preferred specification, CA-PFL is estimated to raise the share of fathers of infants who are on leave by about 0.9 percentage points. Relative to the pre-PFL mean leave-taking rate among California fathers of infants of 2 percent, this represents a substantial 46 percent increase. Among households with two married and employed parents, we see that about half of this increase is driven by fathers who take leave at the same time as their children’s mothers and the other half by fathers who take leave on their own while their children’s mothers are at work.

We also document significant heterogeneity in paternal leave-taking. First, the overall effect of CA-PFL on fathers’ leave-taking is about twice as large for fathers of sons than for fathers of daughters. In contrast, the effect on mothers’ leave-taking is larger for mothers of daughters than mothers of sons. Second, we show that the effects on leave-taking are driven entirely by fathers of first-born children (i.e., those with no other siblings in the household). There are no such differences for mothers. Third, fathers and mothers in occupations with a high share of female workers experience greater increases in leave-taking than their counterparts in occupations with a lower share of female employees. While there are a number of explanations for this type of treatment effect heterogeneity, one possibility is that individuals in occupations with more women experience less stigma associated with taking leave.

Our findings are robust across a variety of alternative DD and DDD specifications, and to the inclusion of numerous individual-level and state-year controls. We demonstrate that there are no statistically significant pre-trends in leave-taking behavior in the years before CA-PFL and obtain similar findings when collapsing the data to the state-year level and using synthetic control methods with a variety of control groups. This consistency of estimates increases our confidence that we are accurately measuring causal effects of CA-PFL, although we do note several limitations of the data in subsequent sections.

Currently, all industrialized countries other than the United States have some kind of national paid parental leave policy (Earle, Mokomane, and Heymann, 2011). The most generous policies are found in Europe, where the length of paid leave varies from 3.5 to 38 months with 70 to 100 percent of wages replaced, and two-thirds of developed countries provide some (albeit generally more limited) rights for fathers to take leave (Ruhm, 2011). However, the take-up rate of parental leave by fathers is generally substantially lower than by mothers. In recent years, policy makers have made a concerted effort to encourage fathers to take more time off. Studies show that several countries including Sweden (Duvander and Johansson, 2012; Ekberg et al., 2013), Norway (Dahl et al., 2014), and parts of Canada (Patnaik, 2015) have substantially increased rates of fathers’ leave-taking by incorporating dedicated paternity leave into their general parental leave policy (sometimes referred to as “daddy quotas” or “daddy months”). There is some evidence that these reforms also increase the amount of childcare provided by fathers, both in the child’s first year of life and 18-30 months later (Schober, 2014).

However, most existing research on PFL policies is exclusively focused on mothers. Studies from Canada and Europe find very high leave take-up rates among mothers (Rønsen and Sundström, 2002; Baker and Milligan, 2008; Dustmann and Schönberg, 2012; Carneiro, Løken, and Salvanes, 2015; Liu and Skans, 2010, Rasmussen, 2010; Dahl et al., 2015). Parental leave has positive effects on the medium and long-term labor market outcomes of mothers as well. Cash benefits that encourage leave-taking do not appear to negatively affect medium-term employment outcomes, provided the benefits do not last longer than about a year (Lalive et al., 2014; Schönberg and Ludsteck, 2014). Other evidence suggests that job-protected leave rights substantially increase maternal employment rates (Ruhm, 1998; Misra, Budig, and Boeckmann, 2011).

In the U.S., the literature has largely focused on the impacts of unpaid leave, which became available at the national level after the introduction of the Family and Medical Leave Act (FMLA) in 1993. FMLA has been shown to increase leave-taking among mothers, without any detectable effects on later employment (Waldfogel, 1999; Han, Ruhm, and Waldfogel, 2009). Yet because of the strict eligibility requirements, less than 60 percent of private sector workers are eligible for FMLA (Klerman et al., 2012), and the impacts of the law are concentrated among relatively advantaged women, who are more likely to be eligible and able to afford unpaid time off work.

Prior to the introduction of CA-PFL, the only legislation that offered paid leave existed through the Temporary Disability Insurance (TDI) system in five U.S. states (California, Hawaii, New Jersey, New York, and Rhode Island). These programs allow pregnant and post-partum women to take short leaves with pay for “pregnancy-related” disabilities. TDI only applies to new birth mothers and not to new fathers. In contrast, CA-PFL allows both mothers and fathers to take paid time off to care for their newborn or adopted children. Recent work shows that California’s policy approximately doubles the leave-taking rates of new mothers, with the largest effects concentrated among disadvantaged groups (Rossin-Slater, Ruhm, and Waldfogel, 2013; Baum and Ruhm, 2015).

The focus of prior research on maternal leave-taking partially occurs because fathers tend to take very little time off work compared to mothers. For example, in the U.S. in 2013, 14 percent of employed mothers of children under age one reported being on leave versus less than two percent of employed fathers, and these rates have remained very stable over the past decade. Low rates of paternal leave-taking may reflect a lack of access to it: a 2012 report indicates that only 14 percent of U.S. employers offer paid paternity leave to most or all of their male employees (Klerman, Daley, and Pozniak, 2012). Low rates of leave-taking by fathers also make analysis of PFL more difficult, because most data sets lack sufficient numbers of fathers who are on leave to produce reliable estimates of program effects.

Our analysis focuses on California’s paid leave program, which was the first of its kind in the United States. CA-PFL is gender-neutral in principle, although some features of the program may affect men and women differently. For instance, the 55 percent wage replacement rate applies only up to a maximum level, which may be more binding on new fathers than mothers, since males have higher average earnings than females. However, the maximum benefit threshold is fairly high compared to the average weekly benefit ($1,104 versus $542 per week in 2015), so this difference may be small. Equal treatment does not imply equal use and, in practice, a large majority of parental leave claims under CA-PFL have been made by mothers. In 2005, only 19.6 percent of all CA-PFL claims were filed by men, although by 2013, fathers were responsible for about 30 percent of claims.

To our knowledge, only one other study has examined the effect of paid leave on fathers' leave-taking in the United States. 6 Baum and Ruhm (2015) study the effects of CA-PFL on maternal and paternal leave-taking duration, using data from the National Longitudinal Survey of Youth (NLSY). They find that CA-PFL increases leave-taking by around 5 weeks for mothers, compared to two or three days for fathers. Yet an important limitation is that the NLSY has very small sample sizes—there are only 158 California fathers in the post-PFL group. Our analysis uses a much larger data set, allowing us to much more precisely estimate the effects of the paid leave program on leave-taking. In addition, because the ACS is a household survey, we are able to explore leave use patterns within households with two employed parents, including effects on joint leave-taking (by both parents) versus father-only leave-taking that occurs while mothers are at work.

We use data from the 2000 Census and the 2000-2013 waves of the American Community Survey (ACS) to estimate the effects of CA-PFL on fathers’ leave-taking. The ACS is conducted throughout the year and samples one percent of the population in most years; thus, it has the major advantage of providing the large samples needed to examine leave-taking behavior among fathers. When weighted, the ACS is a nationally-representative survey that provides information about labor market experiences and demographic factors. For our purposes, what is particularly important is that individuals are questioned about their labor market status in the week prior to the survey. The ACS does not ask about parental leave specifically, but it does identify individuals who are temporarily absent from work during some portion of the week. The main dependent variable that we examine below is whether the father (or mother) is on leave from work in the reference week. This absence could be for many reasons including parental leave, vacation, or illness. Unfortunately the survey does not inquire about length of the temporary work absence, so we do not know whether it lasted the entire week or just some portion of it (or whether the leave began in an earlier week).

In addition to not observing parental leave specifically, there are two other important limitations of the ACS data. First, fathers can only be linked to children who live in the same household. The analysis therefore excludes an important group of fathers who do not live with their children. Assuming that these fathers are less involved with their children than those who live with them, our results will overstate the increase in leave-taking for the average new father (including those not residing with their infants). Second, the ACS lacks precise information on children’s birth dates, and only reports the age of the child in years. Although CA-PFL can be used at any time in the first 12 months of the child’s life, most fathers take only brief leaves that occur soon after the birth (Baum and Ruhm, 2015). Since we observe leave-taking in only a single week, we will therefore miss most of these leaves.10 However, under the assumptions that births and the average length of leave are both uniformly distributed throughout the year, the percentage change in leave-taking estimated to result from the policy will be accurately captured, although the levels will be understated.

Table 2 shows the estimated effect of CA-PFL on parents’ leave-taking behavior in California. The first six columns show results for fathers, for each of the models described above. The first three columns show the DD estimates using a control group of fathers of infants in all other states, and control groups of fathers of 1-3 and 2-4 year olds in California, respectively. The last column shows DDD results for mothers using our preferred specification.

The DD coefficients in columns (1)-(3) suggest that CA-PFL leads to a 0.88-1.25 percentage point increase in fathers’ leave-taking, representing a 44 to 63 percent increase from the pretreatment mean of 1.99 percent. Figures 2 and 3 show the corresponding event-study plots for the models using fathers of infants in other states and fathers of 1-3 year olds in California as control groups, with the coefficients normalized to zero in 2004. Although the estimates are somewhat noisy, there is no evidence of an upward trend in leave-taking among fathers of infants in California, relative to the control group, prior to the introduction of CA-PFL. However, there is an indication of an upward trend after the policy takes effect. Such an increase over time (rather than an immediate jump) might occur if fathers are learning about the availability of CA-PFL, and this is consistent with the 146 percent increase in male “bonding” claims (from 24,021 to 59,256) filed in California between 2005 and the 2012-2013 fiscal year.

Our preferred DDD specification, shown in column (4) of Table 2, suggests that the policy causes a 0.9 percentage point, or 46 percent, increase in leave-taking from the pre-treatment baseline of 1.99 percent. This model compares fathers of infants to fathers of youngest children aged 1-3, in California versus other states, before and after the introduction of the policy. Results are very similar if fathers of youngest children aged 2-4 are used instead, and therefore are not shown. Columns (5) and (6) of Table 2 show that the DDD results are robust to the inclusion of state-specific linear time trends and the exclusion of state-year fixed effects. Since the DDD model allows for both national trends in leave-taking among fathers with infants, as well as statespecific trends in leave-taking overall, subsequent results are only presented for the DDD model. However, results from the DD models are similar.

We cannot directly translate the estimated PFL program effects into the number of additional days of leave taken, because the ACS contains only binary information about temporary absences from work during the prior week, not information about the total amount of time off during that week. However, if we assume that these men were off the job for the full week and births of infants were uniformly distributed throughout the year, our preferred estimates suggest that the program added approximately 2.4 days of leave (0.00915 x 52 weeks x 5 days/week) from a pre-treatment baseline of around 5.2 days

The last column of Table 2 shows the effect of CA-PFL on maternal leave-taking behavior using our preferred DDD specification. 24 Similar to Rossin-Slater, Ruhm, and Waldfogel (2013) and Baum and Ruhm (2015), we find that mothers are also more likely to take leave after the introduction of CA-PFL. CA-PFL leads to a 2.3 percentage point increase in leave-taking among mothers with infants, which represents a 13 percent increase from the pre-treatment baseline. In percentage terms, this is much smaller than the effect estimated for men (13 vs. 46 percent) but it is much larger in absolute terms, since new mothers are more likely to be on leave. Assuming that temporary work absences last for the full reference week, CA-PFL is estimated to increase the leave-taking of new mothers by 6 days from a base level of around 46 days.

Because CA-PFL provides paid parental leave to any eligible parent and not just the primary caregiver, Table 3 explores effects on joint leave-taking by employed parents in the household. The sample in this analysis is limited to fathers with employed spouses, so that both parents are working and potentially eligible to take paid leave, and demographic controls for both spouses are included in the regressions. As expected, the program has large effects on both mothers and fathers. Column (1) shows that the policy increases leave-taking by either parent by four percentage points, or 22 percent. The increase in fathers’ leave-taking is driven both by a 0.41 percentage point rise in the probability that both parents are on leave at the same time (a 28 percent increase) and a 0.53 percentage point increase in father-only leave, while the mother is at work (a 50 percent increase). The increase in father-only leave-taking means that providing CAPFL to fathers—in addition to mothers— increases the total number of days that at least one parent stays home with the infant. Married employed mothers are substantially more likely to take leave after the CA-PFL program comes into effect as well, and they are almost always on leave while the father is at work.

An important limitation of a difference-in-difference analysis is that one must rely on an assumption that the outcomes in treatment and control groups would have followed parallel trends in the absence of the policy reform. While this assumption is inherently untestable, the fact that our results are very consistent across the DD and DDD specifications and robust to different sets of controls is reassuring. We also perform a variety of other robustness tests that lend credibility to the identifying assumption.

Table 7 shows that the findings are similar if the data are collapsed down to the state-year level or if control groups are chosen to best match pre-policy trends in fathers’ leave-taking using synthetic control methods (Abadie, Diamond, and Hainmueller, 2010). The first column shows the DD estimate of the effect of the policy on the share of fathers of infants on leave in California compared to all other states. The Donald and Lang (2007) two-step approach is used to obtain the estimates and standard errors, so inference is conducted using a t-distribution with 12 degrees of freedom. The effect is very similar to the estimate obtained using individual-level data, with CA-PFL predicted to raise paternal leave taking by 0.8 percentage points in both cases. The Donald and Lang approach for calculating standard errors is often used to conduct inference when there are a small number of clusters. While 50 clusters is normally assumed to be large enough for the asymptotic results of cluster-robust standard errors to apply, recent work shows that the “effective” number of clusters is smaller when the number of observations per cluster varies across groups, as is true in the uncollapsed data (Carter, Schnepel, and Steigerwald, 2015), so that the cluster-robust standard errors could be under-estimated. However, the fact that inference is the same using the Donald and Lang two-step approach and when using clusterrobust standard errors on the uncollapsed data suggests this is not a major issue.

While paid parental leave programs have traditionally focused on mothers, there are potential benefits to having fathers, as well as mothers, take time off work shortly after the birth of a new infant (Haas, 1990; Nepomnyaschy and Waldfogel, 2007; Tanaka and Waldfogel, 2007; Haas and Hwang, 2008). California’s paid family leave program is the first source of governmentprovided paid parental leave available to fathers in the United States, and our results show that the policy raised fathers’ estimated leave-taking during the first year of a child’s life by a substantial and statistically significant 46 percent. In relative terms, this increase is much larger than the 13 percent increase estimated for mothers, although since mothers take so much more leave, the absolute increase is only about 40 percent as large (at around two days). Interestingly, the predicted increase in male leave-taking is similar to the estimate found in Baum and Ruhm’s (2015) analysis of CA-PFL, using a much smaller sample, while the rise in mother’s leave-taking is considerably smaller than that obtained by either Baum and Ruhm (2015) or Rossin-Slater, Ruhm and Waldfogel (2013).

These results, when combined with the relative lack of employer-provided paternity leave in the United States, suggest that many new fathers respond to expanded opportunities to take paid family leave. Furthermore, among married households where both parents work, father-only leave-taking increases by a substantial 46 percent, implying that these fathers may become more actively involved in childcare, spending more time alone with their infants than they would have in the absence of the policy.

The increase in paternal leave-taking may also have important implications for addressing the gender wage gap. Although women currently make up nearly half of the United States labor force, the wage gap still persists, with full-time female workers earning 77 percent of what their male counterparts earn.32 Further, mothers have traditionally performed a disproportionate share of childcare and housework, and this disparity also persists today (Hochschild and Machung, 1989; Blair and Lichter, 1991; Bianchi, 2011; Bianchi et al., 2012). The unequal burden faced by women in the home, combined with a lack of flexibility in work schedules at most jobs, may be an important explanation for why the gender wage gap still exists despite tremendous progress in women’s educational and labor market performance over the last half century (Goldin, 2014). Our results suggest that a gender-neutral PFL policy can increase the amount of time fathers of newborns spend at home—including time they spend at home while the mothers work—and may therefore be seen as one way to promote gender equality. Future research may explore the impacts of CA-PFL on the gender wage gap as well as time use patterns in the home.

Our results also call attention to important limitations about the extent to which policymakers can use paid parental leave programs similar to the one implemented in California to increase paternal involvement in childcare. CA-PFL almost doubles the rate of leave-taking among first parity fathers, but has almost no effect on leave-taking after higher-order births. While fathers take leave from work after the birth of their first child, they do not take advantage of the CA-PFL benefit when subsequent children are born. Moreover, fathers of boys are much more likely to take leave as a result of CA-PFL than fathers of girls. Although the reasons for these results are not yet understood, they suggest that making family-friendly policies genderneutral might not be enough to change men’s long-run attitudes about gender equality, or their level of participation in household duties. In addition, these patterns may reflect joint decisions made by mothers and fathers together, if for example they feel there is less need for exclusive parental care for a second or later child, or for a girl. On the other hand, our finding that men who work in female-dominated jobs are more likely to take leave raises the possibility that attitudes existing in the workplace may play an important role and that changes in societal norms about the importance of fathers, as well as mothers, spending time at home with newborns may change behaviors.

Regardless of the longer-term implications, California’s PFL policy has led to a large relative increase in leave-taking among new fathers when compared to the low pre-PFL mean. Although the average number of days spent on leave is small compared to mothers, there has been a substantial increase in the share of fathers who take at least some time off work. In fact, out of all CA-PFL “newborn-bonding” claims filed in the 2013-2014 fiscal year, one-third were filed by fathers. As more states start to adopt gender-neutral parental leave policies similar to CA-PFL, we might expect to see fathers’ leave-taking become even more common.