Educational and Type of Day Differences in Mothers’ Time Availability for Child Care and Housework

Abstract

Objective:

The authors analyzed variation by education and type of day in the “time availability” association between U.S. mothers’ paid work hours and housework and child care, types of work that vary by their urgency, affect, and symbolic meaning.

Background:

Research shows a stronger negative association of women’s work hours with housework than child care, and interprets this as evidence to show mothers prioritize child care over housework. The authors extend this work by determining if associations of work hours with partnered mothers’ housework and child care differ by college education and type of day.

Method:

The authors used ordinary least squares regression on weekend and weekday time diaries of partnered mothers aged 18–65 (N = 22,816) from the 2003–2018 American Time Use Survey (https://timeuse.ipums.org/).

Results:

Authors found negative associations of mothers’ work hours with weekday housework and child care. They found a negative association of college degrees with weekday housework but a positive association with child care that attenuates at longer work hours. The negative work hour association, and the education gap in predicted child care time, persisted on weekends. Work hour and education associations with weekend housework were positive, and the education gap widened at longer work hours.

Conclusion:

The “time availability” constraint of employment hours applies to child care and housework, even among mothers with college degrees. Education differences in unpaid work, particularly child care, are most evident on weekends.

Introduction

All of us have 24 hours in a day. Most of us have to spend some of these hours working for wages to meet our needs and wants, an imperative reinforced by precarious jobs in an uncertain economy (Moen, 2016; Pugh, 2015). But this requirement does not eliminate the need for domestic labor and child care, which remain disproportionately unpaid “women’s work.” The quantitative research has formalized the gendered tension between paid and unpaid work as the “time availability” perspective: the more time women spend on paid work, the fewer hours are left for unpaid work in their own households (Coverman, 1983, 1985). Multiple studies across countries and time periods have documented these negative “time availability” associations between women’s employment hours and their time spent on both housework and child care (Bianchi, 2000; Connelly & Kimmel, 2009; Cooke & Baxter, 2010; Craig et al., 2014; Hook, 2017; Hsin & Felfe, 2014; Kimmel & Connelly, 2007; Sayer & Gornick, 2011).

Researchers have also studied differences between the “time availability” associations for child care and housework. While forms of labor are unpaid when performed for women’s own households, they are fundamentally different with respect to urgency, affect, social interaction, and symbolic meaning. Child care involves others by definition, and gives greater satisfaction to parents than does housework (Bianchi et al., 2006; Wang, 2013). Parents can postpone or skip some chores, but can incur criminal liability for neglecting their children. Not surprisingly, the association between women’s employment hours and their time spent on child care is thought to be weaker than the corresponding association for housework (Bianchi, 2000; Bianchi et al., 2006; England & Srivastava, 2013). Mothers appear to privilege time with children over housework, leisure activities, and sleep (Bianchi et al., 2006; Guryan et al., 2008; Sullivan, 2013). Even then, most mothers feel they spend too little time with children (see Nomaguchi & Milkie, 2000, for a review). These findings may explain why, even though higher proportions of mothers are employed now than during the 1960s, their average child care time has not decreased (Bianchi, 2000; Sayer, 2016; Sayer, Bianchi, & Robinson, 2004), even as they have reduced their time spent on housework, sleep, and leisure (Bianchi et al., 2006; Craig, 2006).

Less clear from the existing literature is whether and how the “time availability” association varies among mothers. Even if most mothers, including mothers with longer employment hours, want to protect their child care time, they may differ in their ability to do so. Using mothers’ time diaries from the 2003 to 2018 waves of the American Time Use Survey (ATUS, https://timeuse.ipums.org/), we determine how “time availability” varies by mothers’ educational level, a key measure of their socioeconomic resources. Generally, less educated people spend more time obtaining necessary goods and services (Kneebone & Holmes, 2015). Mothers with greater resources may be better able to protect child care time than less advantaged mothers. They may have more flexible employment schedules, and a greater ability to reduce their housework time with meal services or the latest household technologies. On the other hand, they may internalize expectations of “intensive mothering—placing the needs of children over all other needs and devoting all available time to children—to a greater extent than less advantaged mothers” (Dow, 2019; Milkie & Warner, 2014). Educated mothers spend more time with children across time and across Western industrialized countries (Dotti Sani & Treas, 2016). Many studies suggest a steeper education gradient for activities that enhance children’s cognitive abilities, health, and social capital (Altintas, 2016; England & Srivastava, 2013; Gracia & Esping-Andersen, 2015; Gracia & Ghysels, 2017; Hsin & Felfe, 2014; Kalil et al., 2012; Vinopal & Gershenson, 2016).

These findings suggest that the expenditure of time, like that of money, is socially stratified. Yet most of the existing research has assumed implicitly that the negative association between their paid and unpaid work time is the same for all women. To date, only Hsin and Felfe (2014) and Gracia and Ghysels (2017) have considered educational differences in the time availability association for child care, and neither has used time diary data collected from U.S. mothers. Using U.S. children’s time diary data from the Panel Study of Income Dynamics (PSID), Hsin and Felfe (2014) found that children’s time with mothers is negatively associated with noncollege educated mothers’ employment hours, but not with college educated mothers’. Gracia and Ghysels (2017) reported positive associations of education with child care time for Spanish and British mothers, but null findings in Belgium. We determine specifically for U.S. mothers whether time availability, for both housework and child care, applies differently based on their educational levels.

We further determine the time availability associations separately for weekdays and weekends. Any educational differences in mothers’ ability to protect child care time should manifest more sharply on weekdays, during which paid work is concentrated. More generally, the temporal cycles of employment, schools, and care and recreational facilities produce a “rhythmic structure of the week” (Manke et al., 1994, p. 564) that reflect social conventions about appropriate times for activities like paid work, leisure, and sleep. Individuals’ time allocation, therefore, is partially determined by these institutional clocks, as well as cultural beliefs about the different nature of weekdays and weekends (e.g., weekends as “free time”). Although a few studies have examined the relationships between women’s employment and housework on weekdays versus weekends (Clarke et al., 1986; Hook, 2017; Manke et al., 1994; Robinson et al., 1972), none has done so for child care with nationally representative time diary data from parents. Trend and cross-national studies have compared time in housework and child care on weekdays and weekends but have not considered if the time availability associations vary by type of day (Altintas, 2016; Bianchi et al., 2006; Craig & Mullan, 2010; Gracia & Ghysels, 2017). The ample evidence that weekly employment hours are negatively associated with time spent on housework and child care requires re-interpretation if employed mothers shift some unpaid work time to weekends.

Such adjustments across days of the week could be especially important if they vary by mothers’ education. Mothers’ time with children has positive implications for children’s developmental and social outcomes, and for mother–child relationships (Hsin & Felfe, 2014; Kalil et al., 2012). Educational differences in mothers’ reallocation of time with children from weekdays to weekends may therefore have implications for the intergenerational transmission of socioeconomic advantage. If there are no differences by education—if college educated mothers who devote longer hours to employment are not able to protect time with children on weekdays—that suggests an unresolvable tension between the time demands of employment and child care. Mothers may address this conflict by cutting back on employment, if they can afford to do so with financial support from partners or with their own savings or wealth. But this solution may damage their economic or psychological well-being, as well as their children’s (Bianchi, 2011; Gerson, 2010; Nomaguchi & Milkie, 2020). And socioeconomically advantaged mothers may use their resources to reduce other demands on their time for the sake of their children’s education and transitions to adulthood (Milkie & Warner, 2014; Schneider et al., 2018). Differences by mothers’ education in “time availability,” therefore, could be consequential not only for children’s and mothers’ well-being, but also for inequality among families.

Background

Over three decades of research across several countries and time periods has documented the negative “time availability’ association between women’s paid and unpaid work (e.g., Bianchi et al., 2012; see Cooke & Baxter, 2010 for a review). Unlike other models of women’s unpaid labor based on earnings, gender ideology, and other factors, this association is robust across specifications, countries, and time periods. The common explanation is that paid work hours are set by employers and constrain people’s choices regarding domestic labor. Economic survival requires nonpartnered women to prioritize employment over household work (Boeckmann et al., 2014; Gornick & Meyers, 2009). Women with partners may have more flexibility than nonpartnered women in choosing to work for wages, but if employed, are subject to the time constraints of their employment. In contrast with this structural emphasis, microeconomic models suggest that those with higher wages will allocate more time to employment and less to unpaid activities like housework (e.g., Connelly & Kimmel, 2009). This may explain why employment hours have only modest or no negative associations with men’s unpaid work (Bianchi & Milkie, 2010; Hook, 2017; Sayer & Gornick, 2011).

The economic and sociological models used in these studies theorize weaker associations of employment hours with child care because of greater parental preferences for child care time. Child care time in the United States has increased since the 1960s among parents in all employment statuses, whereas housework hours have declined (Bianchi et al., 2012; Fox et al., 2013); child care time has also increased in most European countries (Dotti Sani & Treas, 2016). The composition of unpaid work has changed correspondingly, with the share of child care in all household labor increasing from 24% in 1965 to 43% in 2012 (Author calculations from ATUS and American Heritage Time Use Study [AHTUS]; Fisher et al., 2018). Dual-earner full-time, full-year employed mothers with children ages 5 and younger reported as much child care time in 2009 as nonemployed married mothers in male-breadwinner families in 1975 (Fox et al., 2013). Scholars have interpreted this trend as a reflection of educated parents’ anxiety about their children’s class standing (Milkie & Warner, 2014; Ramey & Ramey, 2009). It may also reflect the changing emphasis of gender enactment, from housework to the care and development of children, or “intensive mothering” (Bianchi et al., 2012; Craig, 2012; Hook, 2017; Sayer, 2005).

While the literature has amply documented the basic negative association between women’s paid and unpaid work time, fewer quantitative studies have examined how this relationship varies across women. We know that women with greater socioeconomic resources spend more time in employment, less time in housework, and more time in child care (Killewald, 2011; Sayer, 2016). Kimmel and Connelly (2007) found negative associations of mothers’ predicted wages on housework and leisure, and positive associations with total child care time, implying they maximize time in child care at the expense of housework and leisure time. Women may use their earnings to outsource some of their housework (Baxter et al., 2009; de Ruijter et al., 2005; Gupta, 2007). Employed mothers with more education may have greater time flexibility on the job, including the ability to work at home (Killewald & Zhuo, 2019; Landivar, 2017). Mothers’ education is positively associated with time spent on activities promoting their children’s physical and mental development (Gracia & Esping-Andersen, 2015; Hsin & Felfe, 2014; Kalil et al., 2012; Vinopal & Gershenson, 2016).

Qualitative studies have found that more educated parents practice “concerted cultivation” of their children, with activities fostering independence, creativity, and educational success (Lareau, 2003). The quantitative literature has documented positive associations of mother’s education with child-related time, in adult time-diary cross-sectional and trend studies of primary child care time (Altintas, 2016; Bianchi et al., 2006; Connelly & Kimmel, 2009; England & Srivastava, 2013; Fox et al., 2013; Hill & Stafford, 1974, 1980; Kimmel & Connelly, 2007; Leibowitz, 1974, 1977; Ramey & Ramey, 2009; Sayer, Bianchi, & Robinson, 2004), time with children (Bryant & Zick, 1996; Craig & Mullan, 2012), and using children’s time diaries, children’s time with mothers (Hofferth & Sandberg, 2001; Hsin & Felfe, 2014). Also using U.S. children’s time diaries, Bayraktar (2013) found that mothers’ education beyond high school was positively associated with children’s time spent doing homework. Positive associations of education with child-related time have been found across countries, although the magnitude of the education gradient varies (Bonke & Esping-Andersen, 2011; Gracia & Ghysels, 2017; Guryan et al., 2008; Sayer, Gauthier, & Furstenberg, 2004). Similar education gradients for routine and developmental care activities are observed in the United States (England & Srivastava, 2013; Guryan et al., 2008), for routine care in Australia (Craig et al., 2014), and for fathers in Denmark, Spain, and the United Kingdom (Gracia & Esping-Andersen, 2015). England and Srivastava (2013), and Bonke and Esping-Andersen (2011) interpreted the positive association between maternal education and child care time as the ability of higher-educated mothers to use economic resources to reduce their housework time in favor of child care.

These findings suggest that the “time availability” association for mothers’ child care time varies by their education. To date, however, only Hsin and Felfe (2014) have analyzed this possibility with U.S. data, but from the vantage point of children rather than mothers. Using U.S. children’s time diaries from the PSID, they found a negative association of mothers’ weekly employment hours with time with children only among mothers with high school diplomas or some college, not among college-educated mothers. We use data from mothers’ time diaries to perform the first quantitative test of educational differences in the “time availability” association for child care.

We proceed in three stages. First, we use the same dataset to confirm the negative “time availability associations” for both child care and housework. Prior studies like Hook (2017) estimate the relationship between employment hours and housework but not child care; other work has analyzed both housework and child care but focused on earnings rather than employment hours (Connelly & Kimmel, 2009; Guryan et al., 2008; Kimmel & Connelly, 2007). We start with the understanding that both types of labor are unpaid when women do it for their own households, and therefore share the time constraint of paid work. But child care and housework are fundamentally different with respect to urgency, affect, social interaction, and symbolic meaning. Housework ranks low in enjoyment among activities (Kahneman et al., 2004) and is often a solitary activity (Flood & Genadek, 2016). Child care involves others by definition and gives greater satisfaction to parents than does housework (Bianchi et al., 2006; Wang, 2013). Time spent doing housework may be easier to reduce with technology (Gershuny & Harms, 2016). People can relax their standards for cleanliness and can postpone or skip some chores. Child care is subject to far greater social surveillance, including criminal liability for parental neglect. Given these differences between housework and child care, our first step therefore is to use the same data to test the “time availability” hypothesis for both kinds of unpaid work.

• Hypotheses 1: Time availability for housework and childcare.

• (1a) Child care: We expect a negative “time availability” association between mothers’ weekly employment hours and their child care time on the diary day.

• (1b) Housework: We expect a negative “time availability” association between mothers’ weekly employment hours and housework time on the diary day.

Next, we determine whether these associations differ by educational level, specifically by college degree attainment.

• Hypotheses 2: Interaction between time availability and college degree.

• (2a) Child care: We expect the association between weekly employment hours and child care time on the diary day to be weaker, that is, less negative, among mothers with college degrees compared to mothers without college degrees.

• (2b) Housework: We expect the association between weekly employment hours and housework time on the diary day to be stronger, that is, more negative, among mothers with college degrees compared to mothers without college degrees.

For further insight into educational differences in “time availability,” we distinguish between the associations on weekdays and weekends. Most people concentrate their paid work on weekdays, despite rises in nonstandard work, while children spend much of those days in school. People spend more time on weekends in household, social, religious, and sports activities (Bianchi, 2000; Yeung et al., 2001). Weekends may afford more time to “catch up” on housework and child care. Most studies of housework and child care time using time diaries control for weekend diary day (e.g., Chesley & Flood, 2017). Some recent studies show results descriptively, or estimate separate models, by weekday and weekend (Craig & Mullan, 2010; Gracia & Ghysels, 2017; Kolpashnikova & Kan, 2020), or examine only weekend diaries (Kalil et al., 2012) or weekday diaries (Gracia & Esping-Andersen, 2015). None of these studies interacts diary day with other covariates, including paid work hours, to determine if associations vary on weekdays and weekends. Employed women spend more time doing housework on weekends compared to weekdays (Clarke et al., 1986; Kolpashnikova & Kan, 2020). By contrast, mothers spend fewer hours on child care on weekends compared to weekdays (Altintas, 2016; Raley et al., 2012; Sayer et al., 2004) but studies of fathers’ child care time indicate higher levels on weekends compared with weekdays (Hook & Wolfe, 2012; Yeung et al., 2001).

In contrast to these previous studies, we focus on the weekday-weekend difference not in mothers’ time spent on unpaid labor on weekdays and weekends, but rather in the “time availability” association. Two studies to date have reported this association on weekdays only (Gupta & Sayer, 2015; Hook, 2017). We perform the first quantitative analyses of the weekday-weekend difference in the time availability association for mothers’ time spent on child care. The time availability perspective suggests that the association between mothers’ paid work and child care hours is negative on weekdays, just as it is for housework. Also, school-age children need less maternal care on weekdays. Given our earlier discussion of education, we expect that the time availability association on weekdays is weaker for mothers with college degrees. On weekends, given the earlier findings regarding the absence of a time availability association for housework, we do not expect one for child care.

• Hypotheses 3: Time availability for housework and child care, for mothers with and without college degrees, by type of diary day (weekday vs. weekend).

• (3a.1) Child care by college, weekdays: On weekdays, we expect the association between weekly employment hours and child care time to be weaker, that is, less negative, among mothers with college degrees compared to mothers without college degrees.

• (3a.2) Child care by college, weekends: On weekends, we do not expect a statistically significant association between mothers’ weekly employment hours and child care time, whether they have college degrees.

• (3b.1) Housework by college, weekdays: On weekdays, we expect the association between weekly employment hours and housework time to be stronger, that is, more negative, among mothers with college degrees compared to mothers without college degrees.

• (3b.2) Housework by college, weekends: On weekends, we do not expect a statistically significant association between mothers’ weekly employment hours and housework time, whether or not they have college degrees.

Data and Measures

We use nationally representative data from the 2003–2018 American Time Use Survey, extracted the ATUS data from the IPUMS Time Use archive (https://www.atusdata.org/atus/). The ATUS draws respondents ages 15 and over from the outgoing rotation of the Current Population Survey. Response rates range from 43% in 2018 to 57.8% in 2003 (table 3.3 of Bureau of Labor Statistics, 2019, https://www.bls.gov/tus/atususersguide.pdf). The ATUS oversamples weekend days to facilitate analysis of time use patterns by day, with 25% of the sample completing Saturday diaries, 25% Sunday diaries, and the remaining spread equally over the five weekdays. We use replicate survey weights in all analyses to account for the complex design of the ATUS, the over-sample of weekends, and nonresponse (https://www.bls.gov/tus/atususersguide.pdf, pp. 34–41).

The ATUS time diaries span 4 A.M. to 4 A.M. on the day prior to the ATUS interview. Time diaries are more accurate and reliable than compared with retrospective surveys of average weekly or monthly time spent on activities (Juster et al., 2003). They minimize social desirability bias and provide consistent coding of activities across individuals (Robinson & Godbey, 1999). However, the ATUS does not provide data on simultaneous activities, for example talking to children while doing housework, although it does collect a measure referred to as secondary child care that indicates time when parents are able to provide in-person care to children who are injured or in distress (Stewart & Allard, 2016). Additionally, the ATUS data are cross-sectional, preventing causal analyses. They may thus understate the variation in housework and child care by day, employment hours and educational level. The ATUS time diaries are collected only for a single day per respondent. Hence, we cannot analyze differences between weekdays and weekends for the same mothers, though we can estimate group differences in housework and child care time on weekends compared to weekdays.

Data

Our analysis sample consists of 11,248 weekday and 11,568 weekend diaries of married and cohabiting mothers aged 18–65 (N = 22,816). From the 112,854 time diaries corresponding to female respondents (out of a total of 201,151 available diaries), we exclude 60,375 diaries of women without children under the age of 18 in their households. To remove differences in mothers’ time associated with the availability or absence of partners, we exclude 20,138 diaries of mothers without spouses or cohabiting partners present, and control for sample mothers’ partners’ education and employment. This also limits the bias in our time availability coefficients due to associations between mothers’ education and marital or cohabiting status. We exclude 1,435 diaries of mothers reporting disabilities and those who are self-employed, retired or full-time students, or with male partners satisfying the same conditions, because their time use constraints and patterns differ substantially from other adults. These individuals have different competing time demands but also greater discretion over how they resolve time conflicts. For example, self-employed individuals have more control than other employees over their work hours and conditions of work. Individuals who have disabilities or are retired may have competing time demands from health limitations, but fewer employment obligations We cannot for all these sources of heterogeneity adequately, and therefore follow the lead of earlier research in excluding self-employed and retired individuals (England & Srivastava, 2013; Hook, 2017; Raley et al., 2012; Sayer, 2016). We exclude full-time students because they are in the process of changing their educational level, one of our primary independent variables. Excluding another 570 diaries of mothers in same-sex couple households and ages outside the range of 18–65 leaves us with 30,693 observations eligible for our analytical sample. These sample restrictions are similar to those of earlier studies of housework and child care (England & Srivastava, 2013; Hook, 2017; Raley et al., 2012; Sayer, 2016). The remaining loss of observations comes from missing values on model variables, primarily employment hours and earnings, as detailed below.

Dependent Variables

Our dependent variables are minutes spent on the diary day on child care and housework. Consistent with most quantitative research on domestic labor, our housework measure includes the four routine household tasks categorized as core or “female-typed” housework: meal preparation, meal clean up, indoor cleaning, and laundry (Bianchi et al., 2012; Chesley & Flood, 2017; Kroska, 2003). Also consistent with prior research, our child care measure consists of summed daily minutes in total primary child care, consisting of physical care of infants and toddlers, general supervision, health-related care, reading, playing, teaching, helping, and travel related to caring for or helping children (Bianchi et al., 2006; Craig & Mullan, 2010; Dotti Sani & Treas, 2016). We do not analyze the ATUS measure of secondary child care because it includes time when parents are responsible for children and may not be directly engaged with children (Schwartz, 2001; Stewart & Allard, 2016). Influences of paid work hours on secondary child care are likely to be different than those with primary child care because parents can report another activity as their main activity and indicate children were in their care, suggesting predictions based in time availability that paid work hours compete with secondary child time within the 24-hour constraints of the day may not be valid. About 30% of mothers report “all-day” secondary child care and an additional 30% report at least one 8-hour block of secondary child care, again suggesting this type of child care functions differently than primary child care. Hence, we believe it is not a useful indicator of ways that parents may reduce time in some activities, or shift them to other days, to give priority to more valued or more urgent primary activities.

Independent Variables

Our focal independent variables are usual weekly employment hours, education, and type of diary day. The first is a continuous measure based on responses to a survey question about usual work hours in the week prior to the ATUS interview. We attributed zero employment hours to diaries of respondents (and partners) reporting their employment status as “not employed,” nearly one-third of those eligible for our analytical sample. We coded “hours vary” as missing, leading to a loss of about 3% of eligible observations. Education is a binary variable for post-secondary or graduate degree attainment versus less education (1 = college degree or higher). (Models with finer divisions yielded substantively comparable results.) Obtaining college degrees provides substantially higher economic returns compared with some college and signals class-differentiated status and knowledge distinctions that affect time use preferences more strongly than some college (Dotti Sani & Treas, 2016; England & Srivastava, 2013; Gracia & Ghysels, 2017). We include comparably coded control variables for male partner’s weekly employment hours and education. Type of diary day is a binary variable that equals 1 for weekends and 0 for weekdays.

We also include a control for reporting paid work on the diary day, to adjust for any variability in competing time use. The ATUS does not have an indicator of usual work schedules, so we are not able to determine if mothers routinely do paid work on weekends rather than weekdays. More consequential for our analyses, our argument about weekday and weekend distinctions does not rest on whether paid work is done on weekends. First, educational, religious, and recreational and cultural institutions operate with different time clocks on weekdays and weekends, and social norms about weekdays and weekends frame the latter as time for relaxation and time with family (Craig & Mullan, 2010; Sorokin & Merton, 1937). Second, many mothers who are employed in salaried jobs that ostensibly have standard weekday employment hours routinely do some paid work on weekends (Presser & Gornick, 2005). Hence, even mothers who engage in some paid work on weekends experience different individual and structural constraints on weekdays and weekends.

Control Variables

Control variables include earnings, number and age of children, race/ethnicity, and age of respondent and partner because these have associations with housework and child care (Sayer, 2016). To account for paid employment on weekends, we controlled for any paid work performed on the diary day. We also controlled for any paid work on the diary day. To reduce multicollinearity, we used a summed measure of mothers’ and partners’ weekly earnings rather than specifying them separately; this left us with just over 24,000 observations. Following Hook (2017), we do not adjust earnings for inflation and include dummy variables for each cross-sectional survey year. The indicator variables for each year also control roughly for any period effects. We also include a binary variable for home ownership (coded 1) as a rough proxy for home size and economic resources. A binary variable distinguishes children under age 6 in the household (coded 1) from those aged 6 to 17; we also include a continuous measure of the number of children under age 18 in the household. Regression models in a prior study predicted less time spent by Black mothers, compared to white ones, on both housework and child care (Pepin et al., 2018). We code race/ethnicity into five broad categories: White, non-Hispanic; Black, non-Hispanic; Latina; Asian and Pacific Islanders, and Native Americans and Multiracial individuals. The sample size of the latter group does not allow construction of separate categories for Native Americans and Multiracial individuals. Ages of mothers and male partners are continuous variables ranging from 18 to 65 years.

Analytic Strategy

We employed the “survey regression” procedures in Stata 16 (https://www.stata.com/manuals/svy.pdf) to account for the complex survey design of the ATUS (https://www.bls.gov/tus/atususersguide.pdf, pp. 34–41). To test the time availability Hypotheses 1 for childcare and housework, we used the models below:

$$K={\alpha }_K+{\beta }_E^{K}E+{\beta }_C^{K}C+{\beta }_S^{K}S+{\beta }^{K}X+{\epsilon }^K$$ (1a)

$$H={\alpha }_H+{\beta }_E^{H}E+{\beta }_C^{H}C+{\beta }_S^{H}S+{\beta }^{H}X+{\epsilon }^H$$ (1b)

where K = minutes spent on childcare during the diary day; H = minutes spent on housework during the diary day; E is total employment hours in the week; S is diary day type (1 = weekend, 0 = weekday); C = college (1 = BA and 0 = no BA); and X is a vector of the covariates.

For Hypotheses 2, we tested for differences in the time availability association by college using the models below.

$$K={\alpha }_K+{\beta }_E^{K}E+{\beta }_C^{K}C+{\beta }_{CE}^K(C\times E)+{\beta }_S^{K}S+{\beta }^{K}X+{\epsilon }^K$$ (2a)

$$H={\alpha }_H+{\beta }_E^{H}E+{\beta }_C^{H}C+{\beta }_{CE}^H(C\times E)+{\beta }_S^{H}S+{\beta }^{H}X+{\epsilon }^H$$ (2b)

These models add the interaction term (C × E) used to test the hypotheses by college.

Finally, we used the models below to test Hypotheses 3 regarding time availability by level of education and type of diary day.

$$K={\alpha }_K+{\beta }_E^{K}E+{\beta }_C^{K}C+{\beta }_{CE}^K(C\times E)+{\beta }_S^{K}S+{\beta }_{SE}^K(S\times E)+{\beta }_{SC}^K(S\times C)+{\beta }_{SCE}^K(S\times C\times E)+{\beta }^{K}X+{\epsilon }^K$$ (3a)

$$H={\alpha }_H+{\beta }_E^{H}E+{\beta }_C^{H}C+{\beta }_{CE}^H(C\times E)+{\beta }_D^{H}S+{\beta }_{SE}^H(S\times E)+{\beta }_{SC}^H(S\times C)+{\beta }_{SCE}^H(S\times C\times E)+{\beta }^{H}X+{\epsilon }^H$$ (3b)

These models add the interaction terms (C × E) and (S × E), as well as the three-way interaction term (S × C × E).

Results

Table 1 displays the weighted descriptive statistics by type of diary day. On average, mothers spent 25 fewer minutes on child care, and 12 minutes more doing housework, on weekends compared to weekdays. On the independent measures, the means and standard deviations for weekdays were very close to the values on weekends, reflecting the random assignment of respondents across days. Mothers’ weekly employment hours averaged just over 24, whereas their male partners averaged nearly 40. About 40% of mothers had college degrees, and over one third had partners with college degrees. Half of mothers had at least one child under age 6 in the household.

Table 1.

Mothers’ Time Use, Own Characteristics, and Household Characteristics: Descriptive Statistics by Diary Day Type and Overall (Average of the 160 Samples Using Replicate Weights, N = 22,816)

	Weekday		Weekend		Total
Variables	M or %	SD	M or %	SD	M or %	SD
Daily minutes spent on childcare	114.71	124.75	90.41	121.13	107.73	124.21
Daily minutes spent on housework	128.61	124.17	141.24	130.97	132.24	126.29
Weekly employment hours	24.01	19.79	24.32	19.67	24.10	19.76
Any work on diary daya	52.80%		17.97%		42.79%
Bachelor’s degreeb	39.08%		38.97%		39.05%
Mother’s and partner’s combined weekly earnings (USD)	1,486.65	1,037.04	1,513.42	1,049.46	1,494.35	1,040.71
Homeownerc	75.68%		74.73%		75.41%
Child under age 6 in householdd	50.96%		51.46%		51.10%
Number of children under age 18 in household	1.93	0.97	1.94	0.96	1.94	0.97
Blacke	6.90%		7.17%		6.98%
Latinaf	21.52%		21.57%		21.54%
Asian/Asian Americang	5.97%		6.08%		6.00%
Native American/Multiracialh	1.44%		1.57%		1.47%
Age in years	37.47	8.46	37.51	8.38	37.47	8.44
Partner has bachelor’s degreei	36.78%		36.46%		36.69%
Partner’s age in years	40.05	9.07	39.87	8.95	40.00	9.03
Partner’s weekly employment hours	39.47	17.09	39.41	16.56	39.45	16.94
Diaries	11,248		11,568		22,816

^a0 = no work on diary day, 1 = any work on diary day.

^b0 = no college degree, 1 = college degree.

^c0 = not homeowner, 1 = homeowner.

^d0 = children in household are over six, 1 = presence of at least one child under six.

^e0 = not Black, 1 = Black.

^f0 = not Latina, 1 = Latina.

^g0 = not Asian, 1 = Asian.

^h0 = not Native American/Multiracial, 1 = Native American/Multiracial.

ⁱ0 = partner does not hold college degree, 1 = partner holds college degree.

Tables 2 and 3 present our multivariate findings, which we discuss in detail below with respect to our hypotheses. Our use of the ATUS’s replicate weights means standard regression diagnostics are not available in common statistical packages, such as Stata, version 16 which we used for analyses, and diagnosis using complex survey data remains under-theorized (West et al., 2018). We did use ATUS probability weights to test for multicollinearity, and found high variance inflation factors only for age and age-squared, as anticipated. Heteroscedasticity tests do not apply straightforwardly to weighted data, but in any case they yielded null results for our models. Given the large sample size, bias from outliers is unlikely.

Table 2.

Summary of Simple Regression Analyses for Variables Predicting Mothers’ Time Spent on Childcare (N = 22,816)

	Model 1		Model 2		Model 3
Variable	β	SE β	β	SE β	β	SE β
Employment hours	−0.804***	(0.059)	−0.671***	(0.070)	−0.872***	(0.092)
Bachelor’s degree	21.642***	(2.105)	30.763***	(3.745)	32.796***	(4.629)
Weekend	−40.882***	(1.986)	−40.993***	(1.984)	−54.459***	(3.537)
Employment hours × BA			−0.348***	(0.101)	−0.568***	(0.126)
Employment hours × Weekend					0.406***	(0.104)
Weekend × BA					−8.017	(5.616)
Employment hours × Weekend × BA					0.798***	(0.159)
Any work on diary day	−46.715***	(2.389)	−46.959***	(2.391)	−41.309***	(2.502)
Mother’s + Partner’s earnings	0.003**	(0.001)	0.004**	(0.001)	0.004**	(0.001)
Homeowner	0.690	(2.363)	0.482	(2.365)	0.407	(2.365)
Child under 6	68.932***	(1.893)	68.942***	(1.890)	69.023***	(1.897)
Number of children under 18	10.099***	(0.982)	10.156***	(0.982)	10.141***	(0.983)
Black	−25.212***	(3.064)	−25.346***	(3.060)	−25.126***	(3.045)
Latina	−26.564***	(2.661)	−25.829***	(2.660)	−25.767***	(2.672)
Asian/Asian American	−1.753	(4.084)	−2.282	(4.081)	−2.511	(4.066)
Native American/Multiracial	−22.023**	(7.067)	−21.811**	(7.057)	−21.654**	(7.110)
Age	3.677**	(1.227)	3.515**	(1.239)	3.456**	(1.242)
Age-squared	−0.062***	(0.015)	−0.060***	(0.015)	−0.059***	(0.015)
Partner bachelor’s degree	10.118***	(2.046)	9.434***	(2.058)	9.454***	(2.081)
Partner age	−1.718	(1.201)	−1.733	(1.202)	−1.655	(1.195)
Partner age-squared	0.012	(0.013)	0.012	(0.013)	0.011	(0.013)
Partner employment hours	0.200***	(0.056)	0.186***	(0.056)	0.193***	(0.056)
Constant	84.328***	(22.399)	84.960***	(22.374)	88.068***	(22.375)
R2	0.279		0.280		0.284

Notes. All models include binary variables for each year of data, 2003 to 2018; these are not jointly significant in any of the models.

^*p ≤ .05.

^**p ≤ .01.

^***p ≤ .001.

Table 3.

Summary of Simple Regression Analyses for Variables Predicting Mothers’ Time Spent on Housework (N = 22,816)

	Model 1		Model 2		Model 3
Variables	β	SE β	β	SE β	β	SE β
Employment hours	−0.608***	(0.071)	−0.823***	(0.085)	−1.490***	(0.111)
Bachelor’s degree	−10.787***	(1.956)	−25.591***	(3.842)	−31.259***	(4.810)
Weekend	−15.271***	(2.211)	−15.091***	(2.203)	−55.917***	(4.334)
Employment hours × BA			0.565***	(0.105)	0.627***	(0.131)
Employment hours × Weekend					1.650***	(0.138)
Weekend × BA					18.098**	(6.229)
Employment hours × Weekend × BA					−0.156	(0.185)
Any work on diary day	−80.725***	(2.678)	−80.330***	(2.679)	−67.604***	(2.939)
Mother’s + partner’s earnings	−0.005***	(0.001)	−0.006***	(0.001)	−0.006***	(0.001)
Homeowner	−7.681**	(2.699)	−7.345**	(2.728)	−7.520**	(2.678)
Child under 6	0.694	(2.188)	0.679	(2.173)	1.109	(2.160)
Number of children under 18	10.812***	(1.166)	10.721***	(1.165)	10.655***	(1.155)
Black	−8.789*	(3.671)	−8.571*	(3.684)	−8.065*	(3.744)
Latina	39.462***	(2.729)	38.269***	(2.699)	38.396***	(2.680)
Asian/Asian American	31.056***	(3.526)	31.914***	(3.523)	32.048***	(3.540)
Native American/Multiracial	−4.777	(7.091)	−5.121	(7.068)	−5.417	(6.950)
Age	3.355**	(1.203)	3.618**	(1.205)	3.573**	(1.213)
Age-squared	−0.029*	(0.015)	−0.032*	(0.015)	−0.031*	(0.015)
Partner bachelor’s degree	−5.807**	(2.218)	−4.698*	(2.225)	−4.847*	(2.187)
Partner age	1.703	(1.104)	1.728	(1.106)	1.917	(1.117)
Partner age-squared	−0.008	(0.013)	−0.008	(0.013)	−0.010	(0.013)
Partner employment hours	0.298***	(0.068)	0.320***	(0.068)	0.329***	(0.068)
Constant	26.058	(20.260)	25.032	(20.220)	32.884	(20.430)
R2	0.217		0.219		0.231

Note. All models include binary variables for each year of data, 2003 to 2018; these are not jointly significant in any of the models.

^*p ≤ .05.

^**p ≤ .01.

^***p ≤ .001.

Overall Time Availability Effect

The results of Models 1 in Tables 2 (childcare) and 3 (housework) confirm the basic time availability hypotheses for both childcare (H1a) and housework (H1b). Each additional weekly employment hour was associated with a statistically significant decrease of 0.8 minutes spent on childcare, and of 0.6 minutes spent on housework.

Time Availability by College Education

Model 2 adds education. For women with and without college degrees, the time availability associations for both childcare and housework were statistically significant and negative. Each additional weekly employment hour was associated with 0.7 fewer daily minutes spent on childcare for mothers without bachelor’s degrees, and 1.0 fewer daily minutes spent on childcare for mothers with bachelor’s degrees (obtained by adding the main effect of employment hours and the interaction effect of employment hours with college degree). Wald chi-squared tests showed that the difference in the education coefficient is statistically significant. From Table 3, each additional weekly employment hour was associated with 0.8 fewer daily minutes spent on housework for mothers without college degrees, and 0.2 fewer daily minutes spent on housework for mothers with college degrees.

To summarize: the time availability association for childcare was statistically significant and amplified (more negative) for college educated mothers. However, the one for housework was reduced (less negative) for college educated mothers. These findings were the opposite of what we predicted under H2a and H2b. However, Model 2 also showed that among women with zero employment hours, mothers with college degrees were predicted to spend 31 minutes more each day on childcare and 26 fewer minutes on housework than their less educated counterparts. The steeper slope for college educated women reduced this difference considerably at high employment hours. The time availability associations for childcare and housework must be interpreted in light of these differences. We discuss this further with regard to the results of Model 3, which accounts for type of diary day.

Time Availability by College Education, on Weekdays and Weekends

Model 3 adds an indicator for type of diary day, weekday or weekend. On weekdays, for both childcare and housework, Table 2 shows statistically significant and negative time availability associations, for women with and without college degrees. For college educated mothers, an additional employment hour was associated with 1.5 fewer daily minutes spent on childcare; for mothers without college degrees it was associated with 0.9 fewer daily minutes. That is, the weekday time availability association for childcare was stronger, or more negative, for college educated mothers. As discussed above, this result contradicts our Hypothesis 3a.1; however, it does accord with Model 2, which does not specify type of day. These results are illustrated in the top left panel of Figure 1. On weekdays, mothers with college degrees had a steeper time availability effect than mothers without college degrees. However, mothers with college degrees spent more time on child care on weekdays than mothers without college degrees throughout the range of employment hours, though that difference narrowed at higher employment hours.

Figure 1.

Predicted Daily Minutes Spent on Childcare and Housework, Weekdays versus Weekends, Mothers with and without Bachelor’s Degrees.

Note. With 99% confidence intervals. All variables other than weekly employment hours are at their means.

On weekends, the time availability association for childcare was negative and statistically significant for mothers with and without college degrees. Each additional employment hour during the week was associated with 0.3 fewer daily minutes spent on childcare on weekends by college educated mothers, and 0.5 fewer daily minutes for mothers with less education. The time availability effect on weekends was stronger (i.e. more negative) for women without college degrees. (Wald chi-squared tests showed that the difference between the predictions for mothers with and without college degrees was statistically significant.) These results contradicted our expectation for weekends of no association between mothers’ employment hours and child care time; they are illustrated in the top right panel of Figure 1.

Model 3 in Table 3 shows the results for housework. The results for weekdays mirror the results from Model 2 which does not incorporate type of day. On weekdays, each additional weekly employment hour was associated with 0.9 fewer daily minutes spent on housework by mothers with college degrees, and 1.5 fewer daily minutes by women who without college degrees. Therefore, hypothesis 3b.1 was not supported, as the time availability effect for housework during the week was statistically significantly stronger for women who did not have bachelor’s degrees. These results are illustrated in the bottom left panel of Figure 1, which helps us contextualize these results. On weekdays, mothers without college degrees had a steeper time availability effect during the week than mothers with college degrees. However, for most of the range of weekly employment hours, mothers without college degrees were predicted to spend more time on housework than mothers with college degrees. Thus, although the time availability effect was steeper for mothers without college degrees, it is only at very high employment hours (over 50) that their predicted daily weekday minutes spent on housework fell below that of mothers with college degrees.

On weekends, the time availability association for housework was positive for both women with and without college degrees. It was small (0.2) and not statistically significant for mothers without college degrees. However, it was statistically significant for mothers with college degrees, contradicting our expectation for them. Each additional weekly employment hour was associated with 0.6 additional daily minutes spent by college educated mothers, meaning those employed 35 hours during the week were predicted to spend 21 additional minutes on housework during weekends. (Wald chi-squared tests showed that the difference in the associations for women with and without college degrees is statistically significant.) This suggests that mothers with college degrees “catch up” on housework during the weekend. Again Figure 1 helps interpret these results. Among mothers with fewer than 30 weekly employment hours, those without college degrees were predicted to spend more time on housework on weekends than those with college degrees. This is because, on weekends, as during the week, the number of minutes that nonemployed mothers without college degrees spent doing housework exceeded that of nonemployed mothers with college degrees.

Controls

Reporting any paid work on the diary day was associated with 41 fewer minutes spent doing child care, and 67 fewer minutes doing housework. Notably, male college educated partners added 9.5 minutes to mothers’ child care time in Model 3. Our findings for the other control variables generally concur with those of earlier research. Total earnings had a positive and statistically significant, but small, association with child care time, amounting to 6 additional minutes at the mean of total earnings. Their association with mothers’ housework time was negative and statistically significant, amounting to 9 fewer minutes at the mean of total earnings. Compared to white mothers, Black mothers were predicted to spend 25 fewer minutes on child care and 8 fewer minutes on housework. Latina mothers were predicted to spend 27 fewer minutes on child care and 38 more minutes on housework. Asian/Asian-American mothers were predicted to spend 32 more minutes on housework. Finally, there was no evidence of a period effect, with a joint test of all binary variables for each year of the ATUS failing to reject the null hypothesis at the 5% significance level.

Discussion

Mothers know that whatever else it is, child care is work that competes for their time with employment, and the existing research has confirmed a negative relationship between the time they spend on the two activities. Here we have turned our attention to differences among mothers, by educational level, in this relationship. We expected the “time availability” association to be less pronounced for mothers with bachelors’ degrees, given their greater economic and social resources, and perhaps their “intensive” approach to mothering. But we found that on weekdays, these mothers experience an even sharper tradeoff between their time spent on employment and child care than do mothers without bachelors’ degrees. College-educated mothers may have greater resources to offset the negative association between their time spent on employment and child care, but may also experience greater pressure for, or perhaps greater rewards from, longer employment hours. Also, the availability of children at the end of mothers’ workdays may be comparable for mothers across educational levels given the typical schedules of schools, care and recreational facilities.

Yet there are clear differences by education in mothers’ child care time. Figure 1 shows that on weekdays, mothers with bachelors’ degrees were predicted to spend more time on child care than are mothers without degrees, even though the former’s more negative slope shrank this gap quickly at higher employment hours. And on weekends, when mothers are typically less constrained by employment, this education gap in child care time was consistent across the range of mothers’ weekly paid hours. These “education gaps” in the predicted levels of child care time are consistent with the “intensive mothering” and parental anxiety suggested by previous scholars. On weekdays, the gap was highest among nonemployed mothers. Mothers with college degrees may choose no or little employment due to intensive mothering norms (Stone, 2007). The additional 9 minutes added to mothers’ child care time by their male partners’ college degrees may be evidence of educated parents’ heightened anxiety about their children’s futures (Lareau, 2003; Ramey & Ramey, 2009).

We were surprised by the persistent negative “time availability” association on weekends for child care time. Perhaps mothers’ employment responsibilities spill over into the weekends in the form of paperwork, preparation for the next week, and other unpaid employment-related activities. Child care behaviors during the week could persist on weekends as habit or routine. Possibly child care time does not capture other categories of time spent with children, such as recreation, entertainment, socializing. Time spent on such activities, if it replaces time on activities reported as child care, could lead to the negative association with weekly employment hours even on weekends. This possibility is consistent with earlier findings that U.S. mothers spend most of their leisure time with their children, on average about 3.5 hours a day (Craig & Mullan, 2012). It is also possible that fathers’ child care on weekends is substituting for mothers’ child care time, as found with children’s time diaries and parents’ diaries in European contexts (Craig & Mullan, 2010; Hook & Wolfe, 2012; Yeung et al., 2001) and suggested by Negraia et al. (2018) using the ATUS. Perhaps mothers facilitate fathers’ involvement with children on weekends by reducing their own time with them. Figure 1 suggests that whatever the reasons, they appear to operate comparably for mothers with and without bachelors’ degrees.

Our results for housework further illuminate the educational differences in mothers’ “time availability.” Their paid work hours were negatively associated with their time spent on both housework and child care, both of which are time-bounded, everyday aspects of mothers’ lives. But the differences by education in the weekday associations for housework were the reverse of those for child care. Employed college-educated mothers may have greater flexibility or resources to shave their weekday housework than their child care, whether by outsourcing, buying substitutes, or using household appliances (Gershuny & Harms, 2016). They may have greater flexibility compared to child care in standards for some household tasks, including skipping them altogether. As with child care, however, the size of this education gap in housework is progressively smaller for mothers at higher employment hours. And in contrast to the persistent negative association on weekends between all mothers’ child care time and weekly employment hours, the weekend association for housework was weakly positive, and statistically significant only for those with college degrees. Male partners may be less inclined on weekends to share housework compared to child care. Some of mothers’ catch-up on housework during weekends may also be a strategy to reduce time for necessary housework during the week, such as by preparing and freezing weekly dinners.

Our results are limited because they are based on cross-sectional one-day diaries for one person. Thus, we are not able to draw causal inferences or investigate dyadic or parental-child time use patterns. Additionally, the ATUS does not include data on gender ideology or preferences regarding daily activities. And time-use data do not tell us the exact content of mothers’ time spent with their children, let alone the emotional quality or longer-term impact of this time; these may differ considerably by education. With these caveats, our findings argue for an expanded conception of “time availability.” They demonstrate simultaneously the overarching constraint of employment on mothers’ unpaid work time and the marked differences by education in its operation. “Time availability” expresses differently for housework and child care even though both are types of unpaid work. Its distinct manifestations on weekdays and weekends show that it is a social rather than solely individual constraint. Our findings suggest that maternal time spent on child care remains a critical brake on the “gender revolution.” Even college-educated mothers may need to restrict their employment on weekdays to maximize their child care time. That may have positive consequences for their children’s well-being but negative ones for their own economic well-being, and may contribute to the persistence of socioeconomic inequality across families.