Background
In 2004 California introduced paid family leave (PFL), providing six weeks of paid leave for mothers with a newborn baby. Parental leave contributes to mothers’ employment attachment in cross-national contexts (Baker and Milligan, Reference Baker and Milligan2008; Charles et al., Reference Charles, Buchmann, Halebsky, Powers and Smith2001; Klerman and Leibowitz, Reference Klerman, Leibowitz, Blau and Ehrenberg1997; Kluve and Tamm, Reference Kluve and Tamm2013; Pettit and Hook, Reference Pettit and Hook2005; Ray et al., Reference Ray, Gornick and Schmitt2010); studies also find that PFL increases mothers’ labor market re-entry or attachment after childbirth in the US (Baum, Reference Baum2002; Rossin-Slater et al., Reference Rossin-Slater, Ruhm and Waldfogel2013; Waldfogel, Reference Waldfogel2001), and PFL increased mothers’ hours and weeks of work by 15-20% during their child’s second year (Baum and Ruhm, Reference Baum and Ruhm2016).
Understanding the effect of paid leave on employment among low-income mothers is of particular interest given their economic instability (Hill and Ybarra, Reference Hill and Ybarra2014) as well as the long-term consequences of economic hardships on their children’s development and health (Carneiro et al., Reference Carneiro, Loken and Salvanes2015). Paid parental leave may be more salient to vulnerable mothers’ labor force participation than advantaged mothers with bachelor’s degrees (Byker, Reference Byker2016) because the former group has limited benefits and work protections (Clemans-Cope et al., Reference Clemans-Cope, Perry, Kenney, Pelletier and Pantell2008; Hegewisch et al., Reference Hegewisch, Liepmann, Hayes and Hartmann2010) and often experiences greater financial constraints during leave-taking (Waldfogel, Reference Waldfogel2001; Klerman et al., Reference Klerman, Daley and Pozniak2012). Lower-income women are more precariously attached to the labor force at the time of pregnancy and birth than are more educated women and they have limited benefits and work protections (Hegewisch et al., Reference Hegewisch, Liepmann, Hayes and Hartmann2010) and experience financial constraints around leave-taking (Waldfogel, Reference Waldfogel2001; Klerman et al., Reference Klerman, Daley and Pozniak2012). They are likely to face difficulties in fulfilling childcare and work responsibilities (Meyers and Jordan, Reference Meyers and Jordan2016) due to expensive childcare (Giannarelli and Barsimantov, Reference Giannarelli and Barsimantov2000), relatively unstable childcare support (Gordon et al., Reference Gordon, Kaestner and Korenman2008; Hofferth and Collins, Reference Hofferth and Collins2000), and lack of work protections (Johnson and Corcoran, Reference Johnson and Corcoran2003; Waldfogel, Reference Waldfogel2001; Henly and Lambert, Reference Henly, Lambert, Bianchi, Caspar and Kind2005; Presser, Reference Presser2003; Presser and Cox, Reference Presser and Cox1997). Thus, California PFL (CA PFL) can potentially reduce labor market exits around childbirth among low-income mothers (Appelbaum and Milkman, Reference Appelbaum and Milkman2011) because it provides an affordable leave to mothers who have significantly lower access to paid leave through employer-provided benefits (Shepherd-Banigan and Bell, Reference Shepherd-Banigan and Bell2014).
However, if PFL does not narrow the stratification in access to paid maternity leave it may not have an impact on labor participation among low-income mothers. There are disparities in access to publicly funded paid leave shown in international studies (McKay et al., Reference McKay, Mathieu and Doucet2016), but empirical studies provide scant attention to the effect of CA PFL on low-income mothers’ labor force participation.
Our aim is thus to empirically examine whether CA PFL has increased low-income mothers’ labor force participation one year after childbirth. Previous studies have found that CA PFL increased the take-up rate or accessibility to the leave among low-income mothers (Milkman and Applebaum, Reference Milkman and Applebaum2004; Rossin-Slater et al., Reference Rossin-Slater, Ruhm and Waldfogel2013) but far less attention has been paid to whether PFL increases labor force participation among this group. This study contributes to the current literature by providing evidence of the effects of CA PFL on low-income mothers’ labor force participation, which can inform policy makers on ways to improve low-income families’ self-sufficiency and economic well-being around and after childbirth.
California PFL and female employment
When California became the first state to institute PFL, so that a parent may bond with a new child or care for a sick family member, the CA PFL program extended the coverage provided by the existing Temporary Disability Insurance (TDI) program (which offered paid maternity leave only to qualifying mothers) to include almost all mothers. CA PFL benefits replace 55% of weekly earnings; in 2018, the maximum benefit was $1,216 per week. PFL claims have increased since its implementation, and a majority of claims are for bonding with a newborn or adopted baby rather than caring for a sick family member (Bedard and Rossin-Slater, Reference Bedard and Rossin-Slater2016).
International studies have documented the effect of paid parental leave on mothers’ employment retention. Maternity leave has been shown to increase women’s labor force attachment and career progress after childbirth (or caring for a loved one) by guaranteeing job security during leave time (Gornick and Meyers, Reference Gornick and Meyers2008; Kluve and Tamm, Reference Kluve and Tamm2013; Pettit and Hook, Reference Pettit and Hook2005). Providing a lengthy leave is particularly beneficial to mothers who would otherwise quit and return later to an alternative job with a lower wage if only short leave or no leave is available; mandatory leave increases job continuity for such mothers (Klerman and Leibowitz, Reference Klerman and Leibowitz1994). Women’s decisions about leave are based on the calculation of a reservation wage (Baker and Milligan, Reference Baker and Milligan2008; Klerman and Leibowitz, Reference Klerman, Leibowitz, Blau and Ehrenberg1997).
It is not easy to estimate the effect of paid leave on US employment based on these studies given that the length and wage replacement of CA PFL are based on a much less generous scale compared to those in other industrialized countries. PFL effects on employment in the short term may be ambiguous (Klerman and Leibowitz, Reference Klerman and Leibowitz1994; Baum and Ruhm, Reference Baum and Ruhm2016). In the short term, PFL decreases the numbers of people who work (Han et al., Reference Han, Ruhm and Waldfogel2009) by increasing leave-taking among mothers. After parental leave legislation was enacted, particularly FMLA (Han et al., Reference Han, Ruhm and Waldfogel2009), mothers were found less likely to be employed during the birth month and the next three months; however, maternal leave has a positive impact on employment outcomes in the long term for women overall (Baum and Ruhm, Reference Baum and Ruhm2016; Rossin-Slater et al., Reference Rossin-Slater, Ruhm and Waldfogel2013).
The availability of PFL increased job continuity, possibly improving human capital retention, and increased working hours (Baum and Ruhm, Reference Baum and Ruhm2016; Rossin-Slater et al., Reference Rossin-Slater, Ruhm and Waldfogel2013). Workers also reported that PFL positively affected their ability to care for or arrange childcare for a new baby and doubled the median duration of breastfeeding (Appelbaum and Milkman, Reference Appelbaum and Milkman2011). Byker (2016) finds a significant impact of PFL in New Jersey and California on labor force participation among mothers around childbirth, a result largely driven by less educated mothers (women with less than a bachelor’s degree). On the other hand, Das and Polachek (Reference Das and Polachek2015) find unexpected effects of CA PFL: its implementation increased the labor force participation rate as well as the unemployment rate among young mothers two years after childbirth. This may suggest that paid family leave incurs disadvantage to mothers because it does not provide job protection during the leave period.
A few studies have examined the impact of PFL on mothers’ employment in general, but less is known about how PFL affects low-income mothers’ employment consequences after a child birth. Studies on the benefits of PFL among low-income mothers are concentrated on health outcomes or take-up rates (Hamad et al., Reference Hamad, Modrek and White2019; McKay et al., Reference McKay, Mathieu and Doucet2016). This benefit of PFL on labor participation may be more important among low-income mothers with limited resources who face difficulties in fulfilling childcare and work requirements. Research is needed to examine the consequences of taking PFL on employment status among low-income mothers: that is, if PFL improve post-leave employment outcomes for mothers (Baum and Ruhm, Reference Baum and Ruhm2016; Rossin-Slater et al., Reference Rossin-Slater, Ruhm and Waldfogel2013; Klerman and Leibowitz, Reference Klerman, Leibowitz, Blau and Ehrenberg1997).
Our research focus in this study is on labor force status one year after the use of PFL. This can be an indicator of whether PFL has a beneficial effect on labor force participation among low-income mothers. Thus, we examine whether PFL produces a positive effect on labor force status among low-income mothers one year after childbirth.
Method
Data
We use multiple years (2001–2014) of the cross-sectional dataset from the Current Population Survey (CPS) March Annual Social and Economic Supplement. CPS data is drawn from a nationally representative sample and provides a broad range of data, including income, employment status, and demographic information for each household member.
Women between the ages of 20 and 40 with income levels below 150% of the federal poverty level were selected for the study. We restricted the sample to mothers likely to be eligible for CA PFL, i.e. those who had worked during the previous year. We included only non-military civilian women and excluded those who reported having a disability because the effects of PFL could be confounded by social policies such as disability benefits; we also excluded residents of New Jersey because the state introduced PFL in 2009, which might have affected accurate estimations.
Difference-in-Difference
To identify the impact of CA PFL on low-income mothers’ employment outcomes a year after childbirth, we begin with a Difference-in-Differences (DD) methodology, which compares similar individuals who experience different policy conditions in a logistic regression, in order to estimate the effects of those policies – in this case, CA PFL (Berger and Waldfogel, Reference Berger and Waldfogel2004; Hill, Reference Hill2012; Rossin-Slater et al., Reference Rossin-Slater, Ruhm and Waldfogel2013). DD is often used to examine effects of policy change as a quasi-experimental design (Angrist and Pischke, Reference Angrist, Pischke, Angrist and Pischke2014; Berger and Waldfogel, Reference Berger and Waldfogel2004; Rossin-Slater et al., Reference Rossin-Slater, Ruhm and Waldfogel2013).
The treatment group is low-income mothers in California who worked the previous year and have a one-year-old child. The comparison group is those who are likely to have similar characteristics as those in the treatment group, but are less likely to be affected by the policy: low-income mothers with a one-year-old child who lived in other states and worked during the year they gave birth. The DD approach compares employment outcomes among mothers with children one year of age in California before and after the implementation of CA PFL to those of a comparison group of mothers with one-year-old children who did not live in California.
We estimate:
$$Yist\; = \;{\beta _0} + \;{\beta _1} \times \;{\rm{CA\;Mothe}}{{\rm{r}}_{ist}}\; \times \;{\rm{Pos}}{{\rm{t}}_{it}} + {\beta _2}{\rm{Pos}}{{\rm{t}}_{it}} + {\beta _3}{\rm{CA\;Mothe}}{{\rm{r}}_{ist}} + \;\;{\beta _x}{X_{ist}} + \;{\beta _p}{P_{st}} + \;{\beta _s}{S_s} + {\beta _t}{T_t} + \;{\varepsilon _{ist}},$$
where i indexes individuals,
$$s$$
states, and
$$t\;$$
years. Y captures whether an individual i reports that she is working in the current year. The dichotomous nature of employment status is associated with a logistic model. The coefficient of
$${\beta _1},$$
the interaction effect between mothers who have a one-year-old child in California (eligible for PFL a year ago; treatment group), and Post quantifies the effect of PFL for the treatment group. Xist is a vector of maternal and family characteristics such as age, education (less than high school, high school, some college, four-year degree or more), race (white, black, Hispanic, others), and number of family members. Pst is a vector of state policy and economic contexts such as unemployment rate, log transformed state population, minimum wage, and state Temporary Assistance for Needy Families (TANF) exemption policy. Ss represents state fixed-effects; it is included to control for any unobserved time-invariant differences between states. To control for differential time trends in public assistance program use, year fixed effects,
$${T_t}$$
are included in the model. In order to account for the nested structure of the data, robust standard errors were also calculated at the state level using a previous research method (Baum and Ruhm, Reference Baum and Ruhm2016; Hill, Reference Hill2012).
The underlying assumption in estimating causality with DD is that the employment trend for mothers with a one-year-old child in California is not differentially changing from that of mothers in others states with no PFL. DD estimates may be biased if trends in employment between the treatment and comparison group would have been different in the absence of PFL (Bartel et al., Reference Bartel, Rossin-Slater, Ruhm, Stearns and Waldfogel2018). This assumption concerning employment trends is required to interpret the treatment effects as causal. Because this assumption is untestable, we further employed two different approaches.
First, we ran a DD model with a selected group of states similar to California before the reform as a comparison group that satisfied the parallel trend assumption, following previous research (Jardim et al., Reference Jardim, Long, Plotnick, van Inwegen, Vigdor and Wething2017; Stanczyk, Reference Stanczyk2019). Also, we employed a synthetic control group method (Abadie et al., Reference Abadie, Diamond and Hainmueller2011; Pac et al., Reference Pac, Bartel, Ruhm and Waldfogel2019) that allows each state to be weighted based on their statistical similarity to California in terms of pre-treatment trends. This method can be useful if the parallel trends assumption between the treatment and comparison group is violated when employing a standard difference-in-difference methodology. In order to employ the synthetic control method in this analysis, the unit of analysis must be state level. A set of variables analysis are used (e.g. race, education, age, state-level unemployment rate, state minimum wage, state level income per capita and state log-transformed population). Individual level variables are aggregated into state-year cells except state-level variables.
Second, we used two sets of control groups combined into a difference-in-difference-in-differences (DDD) model (Bartel et al., Reference Bartel, Rossin-Slater, Ruhm, Stearns and Waldfogel2018) that compares mothers of a one-year-old child to mothers of a three-year-old (with no one-year-old child) in California versus the same categories in other states, before and after the policy implementation. This DDD specification allows for differential trends in employment across states and by age of youngest child as long as the difference in the rate of change between mothers with a one-year-old and a three-year-old in California would have been the same as that in other states in the absence of CA PFL. Under this assumption, we estimate the DDD equation:
$$Yist = {\beta _0} + {\beta _1} \times mothers\,with\,1 - year - ol{d_{ist}} + {\beta _2} \times mothers\,with\,1 - year - ol{d_{ist}} \times C{A_s} + {\beta _a} \times C{A_s} \times Pos{t_{it}} + {\beta _4} \times mothers\,with\,1 - year - ol{d_{ist}} \times C{A_s} \times Pos{t_t} + {\beta _5} \times mothers\,with\,1 - year - ol{d_{ist}} \times Pos{t_t} + {\beta _x}{X_{ist}} + {\beta _p}{P_{st}} + {\beta _s}{S_s} + {\beta _t}{T_t} + {\varepsilon _{ist}},$$
where
$$i$$
indexes individuals,
$$s$$
states, and
$$t\;$$
years.
$${Y_{ist}}$$
captures whether an individual i reports that she is working, and
$$C{A_s}$$
indicates whether an individual lives in California or not. Postst indicates the availability of PFL in state s and year t. The coefficient of
$${\beta _4},\;$$
or the interaction effect between mothers who were eligible for PFL (treatment group), Postt, and residence in California quantifies the effects of PFL for the treatment group. Mothers with a one-year-old as their youngest child equal 1; if the youngest child is three years old, the respondent is coded 0.
Variables
Employment
We analyzed employment consequences using the current working status. Respondents are coded as “employed” when they report they were employed and working the week prior to the survey or employed even if not working at the time of survey. Employment variables are based on the reference year: that is, one year after giving birth. It should be noted that a specific date of birth is not provided in the data set, leading to potential measurement errors that will be discussed in the limitation.
Individual variables
A set of demographic characteristics is controlled for: mother’s age, marital status (married, not married), race (white, black, Hispanic, other), educational level (less than high school, high school completion, some college but no degree, bachelor’s degree or higher), and number of family members.
State-level variables
We also controlled for state characteristics such as unemployment rate, population (log transformed), income per capita (log transformed), and minimum wage (Bartel et al., Reference Bartel, Rossin-Slater, Ruhm, Stearns and Waldfogel2018). State unemployment rates, income per capita, and minimum wage reflect state economic conditions and were retrieved for each year from the Bureau of Labor Statistics, Bureau of Economic Analysis, and Department of Labor Historical Wage and Hour Division.
We included state TANF exemption policy for work-related activities when participants are caring for a child. Having no work exemption can possibly be associated with a full-time work rate among mothers with lower education levels (Hill, Reference Hill2012). TANF work exemption policies (the length of exemptions) were adapted from the Welfare Rules Databook for the years 2002–2014 (Huber et al., Reference Huber, Kassabian and Cohen2014, Reference Huber, Cohn, Briggs and Kassabian2015; Kassabian et al., Reference Kassabian, Vericker, Searle and Murphy2011, Reference Kassabian, Whitesell and Huber2012, Reference Kassabian, Huber, Cohen and Giannarelli2013; Rowe and Murphy, Reference Rowe and Murphy2007, Reference Rowe and Murphy2009; Rowe et al., Reference Rowe, Murphy and Williamson2006a, 2006b, 2008, 2010; Rowe and Russell, Reference Rowe and Russell2004; Rowe and Versteeg, Reference Rowe and Versteeg2005).
Results
Descriptive analysis
Table 1 shows the weighted population characteristics between California low-income mothers of a one-year-old child and comparable mothers in other states before and after CA PFL implementation. Estimates for overall work rates are statistically significant by treatment and comparison group status prior to CA PFL implementation, but not post-CA PFL implementation. Before the PFL introduction, the proportion of those working in California (58.18%) differed substantively from those in other states (70.59%). After PFL, the proportion of those at work was greater in California (75.24%) than in other states (70.39%), but this difference was not statistically significant.
Descriptive characteristics of treatment and comparison group pre and post CA- PFL
Source: Current Population Study, March Economic Supplement, 2000–2014.
Note: 1YO=1-year-old. Values are percentages (%) except mean values. + p < .1; * p < .05; ** p < .01; *** p < .001.
Chi-square for categorical variables, t-test for continuous variable. + p < .1; * p < .05; ** p < .01; *** p < .001.
As Figure 1 shows, starting in 2001 the overall trend of employment among low-income mothers with a one-year-old child in California parallels that of other states until 2004, when PFL was implemented in California. An initial rise after PFL implementation occurred for employment among low-income mothers in California, whereas employment slightly decreased among low-income mothers in all other states.
Employment Trends among low-income mothers with a 1-year-old child.
Source: Current Population Study, March Economic Supplement, 2000–2014.
Employment Trends among low-income mothers with a 1-year-old child (CA and Synthetic CA).
Table 2 shows the weighted population characteristics for DDD analysis of four low-income groups (California mothers of a one-year-old, California mothers of a three-year-old, non-California mothers of a one-year-old, and non-California mothers of a three-year-old) during the pre-CA PFL period (2000-2004) and the post-CA PFL period (2005-2014). There are some noteworthy differences in work rates between groups before CA PFL implementation. California mothers were less likely to be at work (55.74%, for those with a one-year-old and 69.28% for those with a three-year-old) than were non-California mothers (71.08% for those with a one-year-old and 71.73% for those with a three-year-old). However, after CA PFL implementation these differences were no longer statistically significant. After CA PFL, those who were working increased from 55.74% (pre-PFL) to 74.47% (post-PFL), which was greater than the work rate of California mothers with a three-year-old (71.69%) and that of other states’ mothers of a one-year-old (70.93%) or a three-year-old (72.90%).
Descriptive characteristics of treatment and comparison group pre and post CA- PFL for DDD analysis
Source: Current Population Study, March Economic Supplement, 2000–2014.
Note: Values are percentages (%) except mean values. + p < .1; * p < .05; ** p < .01; *** p < .001.
Chi-square for categorical variables, t-test for continuous variable. 1YO = 1-year-old, 3YO = 3-year-old.
+ p < .1; * p < .05; ** p < .01; *** p < .001.
Both mothers of a one-year-old and a three-year-old in California were more likely to be less educated than were mothers outside of California. The proportion of those with less than a high school degree was greater in California (32.14% for mothers of a one-year-old, 34.96% for mothers of a three-year-old) than for those in other states (19.43% for mothers of a one-year-old, 23.05% for mothers of a three-year-old). The proportion of Hispanics was significantly larger in California. After CA PFL, in California 61.8% of mothers of a one-year-old and 63.61% of mothers of a three-year-old were Hispanic. Overall, after PFL began in 2004 the proportion of Hispanics increased across all states.
Regression estimates
Table 3 shows the estimated differences in labor force status as a function of California residence and PFL implementation. The DD coefficients estimate the effects of CA PFL on low-income mothers of a one-year-old child in California compared to those of a mother of a one-year-old in other states. These models estimate differences in labor force status outcome between the treatment group and the comparison group post-CA PFL, after accounting for differences in employment between the treatment group and the comparison group before CA PFL. The DD coefficients (Post
$$ \times $$
Mom in Table 3) suggest that CA PFL led to a significant increase in the likelihood of working (b = .58, p<0.05). The DDD specification (CA
$$ \times $$
Post
$$ \times $$
Mom in Table 3) also suggests that CA PFL improves the working status of low-income mothers a year after childbirth. DDD coefficients indicate that the introduction of CA PFL increased the likelihood of working (b = .78, p<0.05).
DD and DDD estimates for the effects of CA PFL on a-year-after employment consequences among low-income mothers
This analysis includes state level characteristics and state fixed effects as controls. State-level characteristics = log state population, minimum wage, TANF exemption.
+ p < .1; * p < .05; ** p < .01; *** p < .001
Because DD and DDD estimates in logistic regression are not easy to interpret, Table 4 presents the marginal effect of PFL on low-income mothers’ employment one year after childbirth. Marginal effects are calculated as changes in the outcome variable as the categorical variable changes from 0 to 1, holding all other variables at their means. Marginal effects in Table 4 indicate that PFL has increased the probability of employment by 11% ([0.004, .221] in DD estimate), 17%([.002, .348] in DD with New York, Alabama, Louisiana, Texas, and Maine or 15% ([.011, .287] in DDD estimate). Using a group of states showing a similar trend in low-income mothers’ employment as a comparison group, the magnitude of CA PFL seems to be larger than others.
Estimated marginal effects of PFL effects
+ p < .1; * p < .05; ** p < .01; *** p < .001
We also found an increase in employment among low-income mothers with 1-year-old children using the synthetic control group method (Figure 2). Based on the previous trend before the CA-PFL, the five states of Texas, Arizona, Nevada, Maryland, and Connecticut were selected as constituents of the synthetic control group (Appendix 1). The figure shows an increase in employment among low-income mothers with 1-year-old children in CA compared to synthetic CA. Despite some fluctuations, the gaps in employment between synthetic CA and CA after the CA PFL tend to be narrower than the gaps before the implementation.
To check the robustness of our findings, we ran several sensitivity tests (Table 5). First, we further restricted the sample to those who worked and earned more than $300 in the previous year (which is the CA PFL eligibility criteria). The CA PFL outcome shows a significant increase in employment in both DD (b = .12, p<0.05) and DDD estimates (b = .16, p<0.05).
Sensitivity test (marginal effects)
+ p < .1; * p < .05; ** p < .01; *** p < .001;
+300: the sample to those who worked and earned more than 300 dollars in the previous year; Checking placebo with an alternative comparison – DD: Mothers with 3-year-old child in California; DDD mothers with 5-year-old child
Second, the definition used to identify one or more comparison groups is a key factor in the internal validity of DD estimates. We chose another alternative comparison using mothers of a three-year-old in California who would be less likely to use PFL for a newborn, but are likely to have similar demographic characteristics as mothers with a one-year-old in California. This can be understood a similar approach to check Placebo effect as mothers of a three-year-old are not likely to be affected. We also examined if there is any CA PFL effect on mothers of a three-year-old in CA compared to mothers in other states. We also used alternative comparison groups of mothers with a five-year-old in California for DDD estimates. Overall, the direction of coefficients is similar to that in the main analysis, but the degree of statistical significance is attenuated. CA PFL increased employment in both the DD (b = .15 p<.05) and DDD models (b = .11, p<.10). This can be understood as a similar approach to checking for the placebo effect, as mothers of three-year-olds are not likely to be affected. We also examined whether there is any CA PFL effect on mothers of three-year-olds in CA compared to mothers in other states. We also used alternative comparison groups of mothers with five-year-olds in California for DDD estimates. Overall, the direction of coefficients is similar to that in the main analysis, but the degree of statistical significance is attenuated. CA PFL increased employment in both the DD (b = .15 p < .05) and DDD models (b = .11, p < .10).
When the state-specific year linear trend is added, the treatment effect of CA PFL is still significant in the DDD estimation (b = .78, p < .05). This can be translated into a 14.6 percentage point increase in employment probability (not shown). However, although we did not find a significant CA PFL treatment effect in the DD estimation, the association was still positive.
Finally, our study period covers the Great Recession, the largest economic downturn since the Great Depression, with an increase in unemployment rates, a decline in median household and real per capita incomes, and slowed growth in the GDP (Blumberg et al., Reference Blumberg, Garrett and Holahan2016; Holahan, Reference Holahan2011). Despite findings that the Great Recession had little impact on PFL program participation (Bedard and Rossin-Slater, Reference Bedard and Rossin-Slater2016), it may have had a confounding effect on employment consequences. We examined whether or how the recession in 2008 influenced the effects of PFL on caregivers’ employment during various time periods: during and post-Great Recession (2007-2014), during the Great Recession (2007-2010), and post-Great Recession (2011-2014). We found a significant PFL effect on at-work status during the Great Recession (b = .20, p<0.01 for DD, b = .25, p<0.01 for DDD), before the Great Recession (b = .20, p<0.05 for DD), and after the Great Recession (b = .14, p<0.05 for DD, b = .17, p<.05 for DDD). Both during and after the Great Recession the effects of PFL were significant, but during the Great Recession, PFL might have played a more significant role in increasing employment for low-income mothers than after this period. (For the period before the Great Recession, the effect of PFL using DDD was not statistically significant.)
Discussion
We examined whether CA PFL has had an effect on low-income mothers’ labor force participation a year after childbirth, providing empirical evidence of the program’s effects on improving low-income mothers’ employment participation.
First, we found that PFL increases employment among low-income mothers. Our results are consistent across the DD, DD with similar states, synthetic control group and DDD analyses and robust after conducting several different sensitivity tests. This finding is aligned with previous studies on the positive effects of CA PFL (and a similar paid family leave program in Rhode Island) for employment of mothers overall (Baum and Ruhm, Reference Baum and Ruhm2016; Silver et al., Reference Silver, Mederer and Djurdjevic2016) and low-educated mothers (Byker, 2016). Our study suggests that PFL can be a good policy option to maintain low-income mothers’ labor force participation and prevent them from exiting the labor market after childbirth. It appears that PFL makes it more likely that low-income mothers who might have decided to exit the labor market will remain employed or return to the labor market. The implication is that CA PFL improves labor market outcomes of women by increasing employment attainment and that is expected to preserve job-specific human capital.
Only a few studies have examined CA PFL’s effect on vulnerable mothers, and those studies have not focused on low-income mothers’ employment. Our results contribute to the current literature by suggesting that even though CA PFL has been of short duration compared to international standards, the provision has had a considerable effect on labor participation among low-income mothers. Of course, the increase of labor force participation may not be directly beneficial to low-income mothers given recent trends in labor market polarization and growing precariousness in employment (Kalleberg et al., Reference Kalleberg, Reskin and Hudson2000). However, the implications of the findings can be understood in the broader context of American welfare states. Work and family balance are a challenge for every worker (Barrow, Reference Barrow1999), but achieving such a balance can be more difficult for those who are economically disadvantaged, particularly less educated and low-income workers and immigrants (Baum, Reference Baum2002) in the US – a liberal welfare regime where limited social social support is provided. The limited public family support that is available has differing consequences for female workers in different classes (Mandel, Reference Mandel2009; Mandel and Shalev, Reference Mandel and Shalev2009; Korpi et al., Reference Korpi, Ferrarini and Englund2013), creating a high level of stratification (Esping-Andersen, Reference Esping-Andersen1990; Katz, Reference Katz2001). In combination with limited public social benefits and privatization, increased work-family conflicts and work absences are likely to have negative consequences for mothers, including wage reduction or job loss (Callan and Dolan, Reference Callan and Dolan2013; Lahaie et al., Reference Lahaie, Earle and Heymann2012) and the necessity of forgoing employment, especially when wages are low (Berry et al., Reference Berry, Katras, Sano, Bauer and Lee2008). Increased labor force participation among low-income mothers is likely to be associated with greater economic security based human capital retention and job continuity (Appelbaum and Milkman, Reference Appelbaum and Milkman2011). In this sense, paid family leave can be a potential policy recommendation for promoting gender equality and reversing extreme poverty and inequality in the US (Lein et al., Reference Lein, Romich and Sherraden2016). Nonetheless, the US is the only developed country that does not offer paid maternity leave.
Second, the magnitude of CA PFL’s effects on employment may be differentiated by income level. Not only US but some international studies have found that the take-up rate of PFL is higher among high-income females (McKay et al., Reference McKay, Mathieu and Doucet2016) and benefits of the expansion of paid family leave have concentrated among economically advantaged women in the US (Hamad et al., Reference Hamad, Modrek and White2019) as well as in international contexts (Hanratty and Trzcinski, Reference Hanratty and Trzcinski2009). Family policies are most relevant to women without university-level education, whereas it is difficult to find substantial negative family policy effects for women with tertiary education (Korpi et al., Reference Korpi, Ferrarini and Englund2013). Even though this study did not directly compare the PFL effect between low-income mothers and mothers overall, findings from previous studies can be used to compare the varying influence of PFL by economic status. Using a linear probability model we found an 11.9 percentage point increase in employment among low-income mothers. This is smaller than CA PFL’s effect on all eligible mothers’ employment (18.3 percentage point increase; Baum and Ruhm, Reference Baum and Ruhm2016). We speculate that this may be because low-income mothers are less likely to have access to PFL. Public awareness of PFL remains limited, according to a survey by Appelbaum and Milkman (Reference Appelbaum and Milkman2011), and low-wage workers are the least likely to be aware of the program. Worry over losing their job and other potential negative employment consequences remain major barriers in using PFL, and even when low-income mothers are aware of PFL, they report fear that they might be disadvantaged by their employers for taking leave (Silver et al., Reference Silver, Mederer and Djurdjevic2016; Appelbaum and Milkman, Reference Appelbaum and Milkman2011). In this regard, future study can extend to whether the effect size of CA PFL differs between socioeconomic groups and whether it disadvantages a certain group of women – in particular, low-income mothers.
In addition, even though PFL may help alleviate economic difficulties during leave-taking, the low wage-replacement rate (55%) is cited as a reason for not applying (Silver et al., Reference Silver, Mederer and Djurdjevic2016). In fact, a considerable group of low-income mothers are better off with TANF than with PFL benefits (Ybarra, Reference Ybarra2013). Recently enacted legislation (CA A.B. 908, 2016) may improve low-income mothers’ access to the leave by raising the CA PFL wage-replacement rate for low-income workers to 70% in 2018.
Third, PFL was more important among low-income mothers during the Great Recession than it was afterward. We speculate that the role of CA PFL in protecting low-income female workers from the challenges of balancing work and family responsibilities became greater during economic hardship. This finding suggests that CA PFL may become even more important to low-income mothers in the future, given the recent trends in labor market polarization and the growing precariousness of employment (Kalleberg, Reference Kalleberg2012), as well as the reinforcement of work requirements through sanctions in TANF. These economic conditions may make the economic hardships of disadvantaged low-wage workers more critical.
The current study, however, has limitations. First, the absence of data on childbirth dates may have created potential measurement errors. Second, despite the strength of CPS as a nationally representative dataset, the small sample size by restricting only low-income single mothers with certain aged children may have skewed the results and decreased the power of the study. Over the fifteen years, across forty-nine states (except New Jersey), the total sample size amounts only to 5207 low-income single mothers who are likely to use TANF. Third, when California implemented PFL, it also provided temporary disability insurance (TDI), and thus it is not possible to examine the net of CA PFL. It operates in addition to TDI, and therefore employees who use TDI for pregnancy can extend their time off to bond with a newborn. For those who are eligible for TDI as paid maternity leave, this leave can be extended to twelve weeks through PFL. Examination of PFL’s effects in California in this study can be considered to provide estimates for the effect of PFL among other states with TDI but should be understood to be a conservative estimate of PFL effects. For those states with no TDI, the introduction of PFL may have a much more significant effect than those found in this study. Future studies may investigate to what extent PFL improves employment consequences for low-income mothers in states with no previous paid maternity leave program. Finally, this study focuses on the effects of PFL on employment for new mothers, but PFL also provides wage replacement for care of family members. This might attenuate our assumptions about a comparison group being less likely to be affected by PFL. One of our comparison groups, mothers with a three-year-old child in California, may be using PFL to care for an ill child. Such cases have not been common (Bedard and Rossin-Slater, Reference Bedard and Rossin-Slater2016), but the findings in this study present a conservative estimate of the net effects of paid maternity leave. Third, this study does not take account for the long-term effect of PFL on low-income mothers. We examined whether mothers with two-year-old children had changes in their labor force participation after the PFL implementation, but found no significant increase. This suggests that PFL might not have a lingering effect, as the duration of paid leave is relatively short. This may be because the current DD analysis is not suitable for this question. Labor force participation among mothers with two-year-old children and mothers with three-year-old children was compared with DD analysis, but mothers with two-year-old children might have already been influenced by other social changes, but necessarily by CA PFL. Future research could explore whether CA PFL has a long-term effect on low-income mothers with available datasets (panel data) and methodology.
Despite these concerns, our findings on CA PFL provide important insights for low-income mothers’ labor force participation and for those states with growing interest in instituting paid family leave programs.
Competing interests
The author(s) declare none.
Acknowledgements
We thank two anonymous reviewers for their constructive comments and Jessica Pac for her helpful advice about synthetic control method.
