Household Composition and Gender Differences in Parental Time Investments

Abstract

Recent research has documented the relatively poor performance of boys, especially those from single-mother households, on a number of outcomes. Differences in noncognitive skills are often cited as a main contributing factor. However, we still know little about the underlying mechanisms driving differences in noncognitive skills and other outcomes. This article provides empirical evidence that parental time investments, defined as the amount of time that parents spend participating in activities with their child, change differentially by child gender following a transition from a two-parent to single-mother household. Boys experience larger investment reductions following the change in household structure, which may help facilitate previously documented gender gaps in noncognitive skills for those in single-mother households. Boys lose an estimated additional 3.8 hours per week in fathers’ time investments, nearly 30% of average weekly paternal investments across the sample. The difference is increasing with age, concentrated in leisure and entertainment activities, with little to no evidence that mothers increase investments in boys relative to girls after such transitions.

Introduction

Recent research has documented the poor outcomes for boys raised in single-mother households on a number of critical dimensions, including cognitive performance (Cunha and Heckman 2008; Heckman and Mosso 2014; Kristoffersen et al. 2015), educational attainment (Autor et al. 2016; Becker et al. 2010; Fortin et al. 2015; Goldin et al. 2006; Jacob 2002; Owens 2016), labor market outcomes (Deming 2017; Figlio et al. 2019; Heckman et al. 2006, 2013), arrests (Heckman et al. 2013), and risky behaviors (Heckman et al. 2006). For example, among children born to married couples, males are more likely than females to be employed at age 30, but the opposite is true among individuals born to nonmarried couples (Figlio et al. 2019). An underlying theme of this research is that differences in noncognitive skills play an important role in generating gender gaps in other outcomes.

We still lack a clear understanding of the mechanisms that generate these gender gaps, which is important for designing policies or treatments to improve boys’ noncognitive skills and related outcomes. In particular, quantifying how much of the single-mother household gender gaps in noncognitive skills are due to differential returns to inputs versus differential levels of inputs is a first-order concern. Bertrand and Pan (2013) found little support for a differential inputs story, instead attributing the wider noncognitive skill gaps found in single-mother households to differential returns to inputs. However, because of the inability of data sets to capture all relevant inputs, the measured returns to one input can be conflated with the levels of and/or returns to omitted inputs (Figlio et al. 2019; Heckman and Mosso 2014). This article builds on prior literature by testing for differential changes in levels of inputs, measured as parental time investments, around changes in household structure from two-parent to single-mother households. Because fathers tend to spend relatively more time with boys as they age (Baker and Milligan 2013; Lundberg et al. 2007; Mammen 2011), and single-parent households are more often headed by the mother, growing up in a single-parent household could be more detrimental for boys in terms of time investments, which are important for skill formation (Coleman 1988; Cunha and Heckman 2008; Cunha et al. 2010; Heckman and Mosso 2014; McLanahan and Sandefur 1994).

Using the Panel Study of Income Dynamics (PSID) (2014) and the accompanying Child Development Supplement (CDS), I obtain direct measures of parental time investments and use within-child variation to estimate differential changes in investments by gender around changes in household composition. The emphasis is on testing whether parental time investments differ by household type for boys and girls, while abstracting from the possibility that boys and girls may have differential returns to parental time investments.¹

Although investments are lower for both boys and girls in single-mother homes relative to two-parent households, the reduction is more substantial for boys. I find large differences in paternal investments, for which boys lose about 3.8 hours per week more than girls, about 30% of the average weekly paternal investment. About 60% of the total gap comes from declines in paternal weekday investments, estimated at 27 minutes per day. The remaining 40% is due to a 42-minute decline in weekend investments per day. The investment gap is largest during adolescence, during which boys in single-mother homes lose 2.4 to 5.8 hours more per week than girls. Estimating the gaps by activity type reveals that boys experience relatively large decreases in investments through leisure activities and entertainment, which account for more of the gap than any other activity type. Furthermore, there is little to no evidence that mothers increase investments in boys relative to girls, which would offset the decline in paternal investments. Similarly, I estimate that investments from grandparents increase for boys relative to girls after the transition, but the results are statistically insignificant (Dunifon et al. 2018; Kalil et al. 2014).

One advantage of the approach used here is that by focusing on children who underwent changes in household structure, the findings are based on comparisons of within-child changes in investments. Using child-level fixed effects controls for all time-invariant observed and unobserved child characteristics, which might otherwise confound the relationship between household structure and parental time investments in children. This builds on previous research that considered cross-sectional differences in time investments using PSID and CDS data (Lundberg et al. 2007; Yeung et al. 2001). Another advantage of this approach is the use of a direct measure of parental time investments, calculated from 24-hour time diaries collected as part of the CDS to the PSID. Although it is important to recognize that no single input measure can capture all relevant aspects of such a complex production process, the use of a direct measure of parental investments allows for transparency and clear interpretation of the results.

Literature

Although research has cataloged differences in outcomes between children from single-mother versus two-parent households (e.g., see McLanahan and Sandefur 1994), explaining the differences is more difficult. Documented advantages in behavioral outcomes for children in two-parent households support the idea that noncognitive skills facilitate differences in long-run success (Bertrand and Pan 2013; Waldfogel et al. 2010). Time investments can help explain differences in both noncognitive skills and long-run outcomes in several ways. Aside from direct transmission of human capital, parental time investments are important for generating intrafamily social capital, which describes the relationships between individuals in the family (Coleman 1988). Time investments facilitate trust between parent and child, help transmit norms across generations, and are necessary for parents to guide the child’s behavior (Coleman 1988; McLanahan and Sandefur 1994; Parcel and Menaghan 1993). Time investments are also important for developing community-level social capital: familial relationships help determine the child’s broader network. Finally, investments (or lack of) from a nonresident father could influence the resident mother’s parenting style by reinforcing or contradicting norms and/or affecting the stress, anxiety, or depression levels of the mother (McLanahan and Sandefur 1994; Osborne and McLanahan 2007).

Indeed, parental investments are empirically linked to both household structure (Hofferth 2006; Hofferth and Anderson 2003; Kalil et al. 2014) and behavioral problems, including being antisocial, anxious/depressed, headstrong, hyperactive, and having problems with peers (Cunha and Heckman 2008; Cunha et al. 2010). The resulting impacts on noncognitive skills facilitate changes in long-run outcomes, including high school graduation, educational attainment, and adult earnings. Also noteworthy is the apparent overlap between leisure and entertainment in the current study and investment measures used by Cunha and Heckman (2008) and Cunha et al. (2010), which suggests that the investment gaps found in this study are similarly important in the skill-formation process.² Taking this collection of studies as evidence that these inputs are important in the generation of noncognitive skills and long-run success, I focus on potential boy-girl differences in inputs generated by household structure.

The literature on gender-based parental biases suggests that inputs may depend on a child’s gender, which could generate the patterns found in gender gaps in single-parent versus two-parent households. For example, Dahl and Moretti (2008) provided evidence of paternal bias in favor of boys, showing that household structure is related to the sex of the firstborn child and that fathers are more likely to obtain custody of sons following a separation. Paternal bias in time investments has been documented using cross-sectional data in both the American Time Use Survey (Mammen 2011) and the CDS (Lundberg et al. 2007). Consistent with my findings, Lundberg et al. (2007) found that fathers’ preferences for spending time with sons in active and passive leisure were more pronounced and that investments from mothers to boys are not higher in single-mother households. One advantage of the current study over cross-sectional boy-girl comparisons of investments is the use of panel data to implement a fixed-effects estimation strategy. Taking advantage of the panel nature of the CDS is important because there are likely many reasons why children in single-parent households receive fewer investments, such that cross-sectional correlations may not merit a casual interpretation.

In a closely related study, Bertrand and Pan (2013) also analyzed the relationships among gender, household structure, and noncognitive skills. Using data from the Early Childhood Longitudinal Survey–Kindergarten, the authors found that gender gaps in externalizing behavior and eighth-grade school suspensions, favoring girls, are nearly twice as large among children from single-mother households.³ I build on that analysis by taking advantage of the panel nature of the CDS to test directly whether changes in household structure lead to differential changes in inputs and by using a more salient measure of investments: parental time investments.

Data

The CDS is a survey administered to a subset of PSID families with children in three waves (1997, 2002/2003, and 2007). The first wave of the CDS included children under age 13, and they were eligible for the CDS until they turned 18.⁴ For this analysis, the most critical components of the CDS are the collection of 24-hour time diaries that cataloged the activities of each child and the ability to identify parental presence in the household.

I restrict the sample to respondents who reported living with both a biological/adoptive mother and father in the first wave and who completed at least one time diary (i.e., the weekday diary, weekend diary, or both). Because I start with the sample of two-parent households, the findings are especially relevant to households that separate and do not necessarily apply to cases in which the parents were never married or cohabiting. Of the 3,563 interviews conducted in the first wave, 2,174 of the respondents reported living with both their biological/adoptive mother and father. Among those living with both parents, 1,823 (84%) completed at least one time diary. Completing the initial diary is unrelated to whether the child is a boy or girl, with response rates of 84% for both subsamples.

Of those living with both biological/adoptive parents in the first wave who completed at least one time diary, 80% (1,452) and 47% (849) completed at least one time diary in Waves 2 and 3, respectively. The Wave 3 response rate appears low, but most of the attrition is driven by respondents aging out of the CDS sample. Of the 1,823 in the initial sample, only 1,001 were eligible for the Wave 3 CDS, so 85% of the eligible participants completed a diary. The most frequently missing data are the parental time investments, which is a consequence of the way that the CDS is administered. I generally make use of two CDS modules in this study: the Primary Caregiver Survey (PCG) and the time diaries. Time diary collection was attempted for only those who completed the PCG survey in Waves 1 and 2, and it was restricted to those who completed either the PCG or the child interview in Wave 3. For this reason, respondents are more likely to be missing the parental input measures than any other piece of data.

Measures

Investments

Every child in the CDS was assigned one randomly selected weekend day and weekday to record their activities for a 24-hour period starting at midnight on the assigned day. The diary data are organized at the activity level, with information on activity duration and participants. Importantly, for every observation (i.e., unique activity), indicator variables identify whether the mother and/or father participated in the activity with the child.⁵

To construct the parental investment measures, I sum the duration of all activities in the time diary for which the mother participated in the activity with the child and likewise for the father.⁶ This is done separately for each weekend and weekday diary. In addition, I calculate total investments, Total_i, by weighting the weekday, WD_i, and weekend, WE_i, investments to construct a weekly investment summary, such that Total_i = 5 · WD_i + 2 · WE_i. In all cases, the investment measures include time with biological/adoptive parents only.

I construct similar measures for time investments received from grandparents. However, there is a single question for each activity indicating whether a grandparent was participating in the activity, so the measure is for the total time the child spent with at least one grandparent participating in activities.

Parental Presence and Household Structure

The second critical feature of the CDS is the measurement of parental presence in the household. I use the Primary Caregiver Child (PCG) file of the CDS survey to construct indicators for the household composition for each child-wave observation, including presence of the child’s biological/adoptive mother and father. The survey in the first wave of the CDS included a question about parental presence of a biological or adoptive mother (and father) in the household: Does child have a biological or adoptive mother? The three valid responses to this question are living with child, not living with child, and does not have. I create a dummy variable equal to 1 if the biological or adoptive mother is living with the child. An analogous question asks about the biological or adoptive father, and I create the analogous dummy variable indicating whether the biological or adoptive father is living with the child. The sample is restricted to respondents who indicated that both the biological/adoptive mother and father were living with the child in the first wave.

In Waves 2 and 3, two separate questions asked about the presence of biological and adoptive parents: Does child live with his/her biological mother?, to which the participant can respond yes or no; and Does child have an adoptive mother?, with response options of yes, living with child; yes, not living with child; or no, does not have. For consistency, I use the combination of these two questions to construct the dummy variables about parental presence. In Waves 2 and 3, the dummy variable indicating that the child lives with their biological/adoptive parent is equal to 1 for a response of yes to the first question or a response of yes, living with child to the second question.

To create the dummy variable for single-mother household, I use only these two variables about the presence of a biological or adoptive mother and father. Single-mother household indicates that the respondent lives with his/her biological or adoptive mother but does not live with his/her biological or adoptive father. For example, children who live in the same household as their biological/adoptive mother and a stepfather are considered to be living in a single-mother household for the purposes of this study.

I focus on the presence of biological/adoptive parents because although it is possible that stepparents and/or unmarried partners mitigate the investment gaps by investing more in boys than girls, research has suggested that children spend more time engaged in activities with biological parents than stepparents or unmarried partners (Hofferth 2006; Hofferth and Anderson 2003; Kalil et al. 2014). Similarly, behavioral problems are more likely to occur in mother/stepfather and mother/partner households. Data limitations also hinder the potential for estimating investment gaps from stepparents and unmarried partners. Because I start with the sample of children in two-parent households in the first wave, only 62 observations in my sample have a stepfather in the household. I do, however, include separate control variables for whether the child has a stepmother and stepfather in or out of the household. Finally, the CDS time diaries do not include an indicator for time spent in activity with unmarried partners of the child’s parents, so accurately measuring investments from unmarried partners is infeasible.

Sample

I start with the subsample of CDS participants who were in two-parent households in the first wave and focus on comparing changes in time investments for children who do and do not transition to single-mother households in subsequent waves. Restricting the sample to observations with no missing data yields about 1,800 time diaries from children in two-parent households in the first wave and almost 4,000 total observations in the final sample.

Data Summary

Figures 1 and 2 display weekly investments by gender and household type from the different waves of the CDS. The left panel in Fig. 1 displays locally smoothed means of total investments by age, for boys and girls who were in two-parent households during the first wave.⁷ A different locally smoothed mean is estimated for those who eventually transition to a single-mother household as well as for those who were in a two-parent household in each survey that they completed. Weekly time spent with mothers decreases dramatically with age for all gender and household type combinations. The average investments across these groups are within about five hours of each other at every age in Fig. 1. However, at ages where investments differ, it is generally true that mothers invest more time in daughters than sons and that mothers in two-parent households spend more time participating in activities with their children. The age trends in weekly maternal investments are similar between girls who remain in a two-parent household and those who transition to a single-mother household, but a two- to three-hour gap emerges at around age 7. Although this gap varies in size, it persists throughout Waves 2 and 3 of the CDS as shown in the right panel of Fig. 1. Note that all sample individuals were in a two-parent household in Wave 1, but those who had a change in household structure were in a single-mother household in a least one (or both) of Waves 2 and 3. Surprisingly, the gap in maternal weekly investments that emerges in the first wave between those who do and those who do not undergo a change in household structure does not appear to widen in Waves 2 and 3 after the changes have taken place.

Interestingly, the pattern in maternal investments in boys differs from the pattern in maternal investments in girls. For example, there is a two- to three-hour investment gap between boys who remain in two-parent households and those who eventually transition to a single-mother household from birth to about age 7, at which point the gap disappears. Carrying over to Waves 2 and 3, the maternal investments in boys in both household types are strikingly similar. Figure 1 demonstrates the importance of the child’s age when considering time investments: investments decrease sharply with age. For example, mothers invest roughly 35 hours to their infant and toddler daughters, but that number is only 15 to 20 hours per week for 12- to 13-year-olds.

Figure 2 graphs weekly paternal investments. Fathers generally invest less time than mothers do across all household types and genders, but the decrease in investments with age is less drastic, especially for boys. In fact, Fig. 2 shows that paternal investments are similarly low for both boys and girls in single-mother households during Waves 2 and 3. On the other hand, a gender gap in paternal investments for those in two-parent households emerges around age 4 and increases to about four hours per week by age 12. That gap represents a large fraction of total time, considering that the average weekly investment is less than 20 hours for every group in Fig. 2. This gap continues throughout adolescence to varying degrees.

One implication of the investment patterns presented in Figs. 1 and 2 is that for boys in two-parent households, the proportion of total investments that come from fathers increases with age. This is shown more directly in Fig. 3, which graphs the proportion of total parental investments that come from the father for four groups: (1) boys in two-parent households in Wave 2 or 3, (2) boys who are in a single-mother household in Wave 2 or 3, (3) girls in two-parent households in Wave 2 or 3, and (4) girls who are in a single-mother household in Wave 2 or 3. I measure proportion of total parental investments for each parent as the weekly measure of time spent with each parent divided by the total weekly measure. The proportion from mothers and fathers add to 1 by construction, so Fig. 3 displays paternal investment proportions only.

The proportions are roughly the same across all four groups for infant and toddlers, with each receiving roughly 30% of total parental investments from their father. However, the relationship between proportion of investments from their father and age clearly depends on the gender of the child. For girls in two-parent households, paternal investments are essentially constant with age. However, paternal investments increase to more than 45% of total parental investments by about age 13 for boys in two-parent households. On the other hand, paternal investments for both boys and girls who transition to a single-mother household decline drastically to 15% to 20% by age 15. Another implication of the gender differences in the investments-age relationship is that one might expect the differential effect of household composition on parental time investments to differ by age. The increasing relative importance of paternal investments for boys, apparent in Fig. 3, suggests that the potential for investment losses relative to girls increases with age.

Two important points of Figs. 1, 2, and 3 are that investments generally decline with age and that the relationship between investments and age differs by gender. In most cases, across all age groups and household structures, mothers spend a little more total time with girls, and fathers spend a little more with boys, on average. The figures demonstrate how important age is when evaluating time investments and suggests that using flexible controls for age is necessary in the following analysis.

Table 1 summarizes the time investment variables and covariates by gender and household type. In particular, I separate the individuals who underwent a change in household structure because estimation relies in part on the comparison of these two groups. Columns 3 and 4 display average characteristics for boys who underwent a change at some point. Column 3 displays the Wave 1 summary statistics for boys who lived with both parents in the first period but eventually transitioned to a single-mother household. Column 4 displays the summary statistics from Waves 2 and 3 of the CDS for the same group. Columns 7 and 8 display the analogous summary statistics for girls who underwent a change in household structure.

The first row of Table 1 summarizes total weekly maternal investments in hours. Girls receive larger maternal investments than boys across all household types. Girls who were always in two-parent households received about 26.6 hours in maternal investments per week relative to 24.8 hours per week for boys. The gap in maternal investments for children in two-parent households that eventually split, comparing column 7 with column 3, is about 2 hours per week, with girls receiving more investments. Both boys and girls in households that eventually split received larger maternal investments than those who were always in a two-parent household. As shown in column 3, on average, boys in families that eventually split received 25.5 hours per week, and those in households that never split received 24.8 hours per week. Similarly, girls in two-parent households that experienced a change in composition received 27.4 hours per week in maternal investments, but those in two-parent households that never split received 26.6 hours per week.

It is important to consider age along with investments because the household structure categories are also correlated with age. For example, girls in two-parent households that never experience a change are just under 7 years old, on average, but the average age of girls in two-parent households that eventually split is under 5 years. The difference is similar for boys. This, along with Figs. 1, 2, and 3, demonstrates why it is important to control flexibly for age when estimating gender gaps in the relationship between investments and household structure. Age is correlated not only with investments differentially by gender but also with household structure. With that in mind, it is similarly true that boys in two-parent households that eventually split received more paternal investments than those who were always in two-parent households: 16.7 and 16.3 hours per week, respectively. However, the opposite is true for girls, with girls who were always in a two-parent household receiving nearly two hours more per week in paternal investments, despite being roughly two years older on average.

Figures A3 and A4 in the online appendix display the main result unconditionally by comparing investments for boys and girls who remain in two-parent households with those who transition to single-mother households in the second or third wave of the CDS.⁸ Fig. A3 displays average paternal investments across CDS waves. All individuals represented in the figure were in a two-parent household in Wave 1, but the averages are separated by whether the individual transitioned to a single-mother household in Wave 2 or 3.

Both the boy-girl differences for those who stay in two-parent households, and the boy-girl differences for those who transition to single-mother households contribute to the relatively large decrease in paternal investments in boys after transitioning to a single-mother household. First, the decrease in investments for boys who transition to a single-mother household is large relative to girls who move to a single-mother household. In Wave 1, among boys and girls who eventually transition to a single-mother household, boys received more than 16 hours per week in paternal investments relative to only 12.5 hours per week for girls. In Waves 2 and 3, paternal investments for both boys and girls who transition from a two-parent to single-mother household decrease dramatically to about 3.6 hours per week. This comparison suggests that boys experience a more substantial decrease in paternal investments after transitioning from a two-parent to single-mother household. Furthermore, among those who remain in a two-parent household, boys receive about an hour more in investments in the first wave. In subsequent waves, that gap increases to almost two hours per week. In other words, the drop from Wave 1 to Waves 2 and 3 in paternal investments for boys who remain in a two-parent household is smaller than that drop for girls who stay in a two-parent household. Again, this comparison suggests a more substantial decrease in paternal investments for boys from transitioning to a single-mother household than for girls.

In contrast, Fig. A4 (online appendix) displays the analogous graph for maternal investments. Although maternal investments are generally higher for girls than boys, the unconditional differences suggest that girls may experience more substantial decreases in maternal investments than boys from the transition to a single-mother household. However, the between-group differences are stable relative to the paternal investment patterns, and conditioning on characteristics of the individuals—and age in particular—may be important for further exploring this possibility.

The main results can be seen unconditionally in Figs. A3 and A4 (online appendix), but as previously noted, with age, it is important to consider other differences displayed in Table 1. As shown in column 3 of Table 1, boys in two-parent households who eventually see a change in household composition have an average of a little over one sibling in the household; those in two-parent households who experience no change have about 1.3 siblings in the household. The difference is similar for girls. Girls who experience a change in household structure but who lived with both parents in the first period had 1.1 siblings in the household at Wave 1, and those in a two-parent household and do not experience a change had 1.3 siblings in the household. The racial compositions of boys and girls in two-parent households that eventually split are also similar. In both cases, there are roughly equal percentages of Black and White individuals in the subsamples, and the percentage of Hispanic individuals is relatively small. Last, among those in households that eventually split, the percentage of children in households in which their parents are married is about 86% for girls and nearly 90% for boys.

Estimation

The left-side variable in Eq. (1), T_it, represents some measure of parental time investments that child i received in wave t. The investment measures are the amount of time that child i spent participating in activities with his or her mother/father from the weekday/weekend 24-hour time diary, measured in hours. In the main specification, I report estimates for the weekday and weekend investments as well as a total weekly investment constructed as a weighted sum of the weekday and weekend investments. I construct the total investment by summing the weekday investment multiplied by five with the weekend investment multiplied by two. For ease of reporting and because using the weekly measure better reflects the effects on total investments, I focus on reporting estimates for total investments in alternate specifications.

The independent variable of interest is the interaction term, M_i· MO_it, where M_i represents a dummy variable for being male, and MO_it represents a dummy variable indicating that child i was in a single-mother household at wave t. I focus on estimating gender gaps in investments for those living in single-mother households because children are more likely to live with their mother if the family is separated.⁹ Furthermore, if fathers invest relatively more in boys as they get older, not having their father in the household could hinder development for boys, even if not for girls.

The coefficient on the interaction term, β_M, is a measurement of the change in the gender difference in investments for those who transitioned to single-mother households relative to those in two-parent households. I estimate β_M using individual-level fixed effects. One advantage of using fixed effects is to control for the time-constant, child-level characteristics indicated by c_i. Unobservable characteristics will bias the estimator to the extent that they are time-variant and correlated with household structure differentially by gender.

This estimator also leads to a convenient interpretation: it can be thought of as a triple-difference estimator. The first difference is from individual changes in investments over time for boys who transition from a two-parent to a single-mother household. The second difference is between boys who transitioned to a single-mother household and boys who remained in a two-parent household in all periods. If I were to stop there and not include female observations in the sample, this would be a difference-in-differences estimator of the changes in individual investments over time for boys who transitioned to single-mother households relative to boys who remained in a two-parent household. Including girls in the sample and using an interaction term for being male and in a single-mother household, M_i· MO_it, adds a third difference between males and females. The end result, $β̂M$⁠, is then based on differential changes in investments that boys receive in single-mother households relative to those remaining in two-parent households, after removing the analogous difference in investments for girls. An estimated β_M< 0 suggests that boys receive relatively low levels of investments in single-mother households, and β_M> 0 suggests that boys in single-mother households are relatively well off in terms of time investments.

One disadvantage of the estimation strategy is that I do not estimate coefficients on time-constant characteristics. However, X_it represents a vector of time-varying observable characteristics, including child’s age and age squared interacted with gender; the number of biological siblings in the household; and indicators for CDS wave, presence of stepparents in or out of the household, and marital status of parents in the household. The gender-specific age terms are important because investments are differentially related to age by gender.¹⁰ The inclusion of gender-specific age terms effectively restricts the counterfactual to similarly aged respondents of the same gender, but it does not dramatically change the interpretation of the estimates.

After presenting the main results, I estimate the gender difference for changes in investments from grandparents. To do this, I estimate Eq. (1) using total, weekday, and weekend investments from grandparents as the outcome variable. This is relevant for assessing the possible role that other, nonparental investments could play. Next, I examine parental investment gap heterogeneity by age, and decompose the gaps into specific activities to determine which are the main contributors to the differential investment losses found in the main results.

Results

Parental Investment Gaps

Table 2 reports fixed-effects estimates of the gender gap in time investments based on Eq. (1) for maternal/paternal total, weekday, and weekend investments. The full set of controls are included.¹¹ The estimates suggest that paternal investments drop for boys and girls after going to single-mother households, but the decrease is relatively large for boys. As shown in column 1, the estimated gender difference in total weekly paternal investments is –3.8, which is statistically significant at the 5% level. This suggests that paternal investments drop by 3.8 more hours per week for boys in single-mother homes than they do for girls in single-mother homes, which is nearly 30% of average paternal investments across gender and household types over the entire sample. About 60% of the gap is driven by weekday investments, for which the estimated gap is –0.45 with a standard error of 0.25. That equates to roughly 27 minutes per weekday, which is about 29% of average paternal weekday investments during Wave 1. The gap in paternal weekend investments is also negative, –0.70 hours and statistically significant at the 10% level. Despite being larger in magnitude, the weekend gap drives about 40% of the overall gap because it has a lower weight in the makeup of the total weekly investment measure.

Although the estimates suggest large boy-girl differences in paternal investments, it is possible that single mothers increase investments in boys relative to girls, meaning that aggregated change in investments is not necessarily negative. Columns 4–6 of Table 2 show the estimated differences in total, weekday, and weekend maternal investments. Interestingly, the estimated gap in total weekly investments is positive, suggesting that mothers do increase investments in boys relative to girls. However, the magnitude of the estimated difference is much smaller than the losses in paternal investments, and none of them are near statistical significance. For example, the estimated gap in total weekly maternal investments suggest that maternal investments increase by about 0.38 hours per week for boys relative to girls. However, the 95% confidence interval spans from –3.7 to 4.5 hours per week, so we should not draw any conclusions about changes in mothers’ investments based on that difference.

Investments From Grandparents

To examine the potential role of grandparents, I estimate the boy-girl difference in investments from grandparents around changes in household structure. The estimates in Table 3 suggest that relative to girls, boys experience a 1.8 hours per week increase in investments from grandparents after transitioning to a single-mother household. The magnitude is about 50% of the main estimate for differences in paternal time, which suggests that total investments for boys may decline by a smaller amount than the paternal investment gap would suggest. However, all the estimated changes in investments from grandparents are statistically insignificant. For the remaining analyses, I focus on parental investments only.

Investment Gaps by Age

Next, I consider heterogeneity in the gender gaps in parental investments. Because of the strong correlation between investments and age, and differential age trends by gender and household structure, one might expect that the boy-girl investment gap in single-mother households depends on the age of the child. For example, Fig. 3 shows that the proportion of total parental investments that come from the father increases with age for boys in two-parents households, suggesting that paternal investments become increasingly important for boys as they get older. Figure 2 provides insight to that trend by showing that gaps arise in boy and girl investments with age, with fathers spending relatively more time with boys. Furthermore, the steep decline in maternal investments from Fig. 1 means that paternal investments become increasingly important in the makeup of total investments for boys but less so for girls. To compare the gender gaps across ages, I group observations into three age bins (6–10, 11–15, and 16 and older) and estimate the gap for each.¹²

Panel A of Table 4 shows the estimates for total paternal investments by age bin. Interestingly, the estimated gap for 6- to 10-year-olds is positive, although paternal investments in both boys and girls decrease drastically for this age group. The estimate suggests that paternal investments in girls decline by 1.8 hours more than boys, but the difference is not statistically significant, with a standard error of 2.5. From column 2, the estimated gap for 11- to 15-year-olds is negative, suggesting that paternal investments in boys decreases by 2.4 hours per week more than for girls. However, this estimate is not statistically significant, given the decline in observations directly contributing to this estimate. The decline for boys relative to girls is even larger for the age 16 and older group, who receive about 5.8 hours per week less in investments than girls. This estimate is statistically significant at the 5% level.

In considering the estimated gaps by age bins, the decline in investments in boys relative to girls is strongly correlated with age, with the gap widening with age. The pattern in these estimated differences supports the idea that paternal investments become increasingly important for boys as they get older, leading to relatively large investment losses during adolescence. The second row of Table 4 shows that the main effect (i.e., the decline for girls) shrinks in magnitude with age. The estimated change in paternal investments for those aged 6–10 is –10.2 hours per week and significant at the 1% level, but the estimated main effect is 0.13 for those age 16 and older, with a standard error of 1.5. This suggests that paternal investments shrink similarly for girls in both two-parent and single-mother households as they age.

Composition of Investment Gaps

I also estimate gender gaps by activity type to determine which activities are the most important contributors to the overall differences. Table 5 displays fixed-effects estimates of Eq. (1) by activity category.¹³ Column 1 provides estimates on total paternal investments. The largest gaps in paternal investments are in passive leisure (e.g., watching television), active leisure (e.g., active sports, walking), and entertainment (e.g., attending events/movies), which are estimated at –1, –0.95, and –0.84 hours per week, respectively. Together, these three activities account for about 70% of the total estimated gap in paternal investments from Table 2. However, the estimated gaps in paternal investments are negative for 9 of the 10 investment categories listed.

Focusing on maternal investment gaps in column 2 of Table 5, active leisure decreases by 0.65 hours per week more for boys in single-mother households. This is the largest negative gap (suggesting a relative decline in investments for boys) of all maternal investments. When considering investment gaps from both parents together, boys experience the largest decline in active leisure. On the other hand, there is an increase in maternal investments in boys relative to girls in tending to needs (e.g., providing care to the child his or herself),¹⁴ meaning that mothers increase time with boys by about 1 hour per week relative to girls in this activity after transitioning to a single-mother household. When considering this in conjunction with the negative gap in tending to needs in paternal investments, –0.49 hours per week, boys experience an overall increase in this particular activity type.¹⁵ However, both estimates are noisy and not statistically different from 0.

Conclusion

Determining the mechanisms that lead to gender gaps in noncognitive skills is largely an open question (Autor et al. 2016; Bertrand and Pan 2013; Jacob 2002). Gender gaps in noncognitive skills among adolescents could arise for several reasons, including gender differences in returns to and levels of inputs in single-mother and two-parent households. Disentangling the contributions of mechanisms leading to noncognitive differences is complicated by the presence of many inputs, most of which are correlated with household structure and are difficult to measure and interpret.

Using time diary data from the Child Development Supplement to the PSID, I obtain direct measures of parental time investments and estimate gender differences in time investments around changes in household composition. Although both boys and girls see significant reductions in parental investments when living in single-mother homes, boys experience larger decreases in paternal investments of 3.8 hours per week, about 30% of the average paternal investment over the sample. The additional loss in paternal investments operates through both weekday and weekend investments. A difference of 27 minutes per day in weekday investments explains about 60% of the total loss, and the rest comes from a difference of 42 minutes per day in weekend investments. Additional losses for boys are generally increasing with age, with the estimated gap in total paternal investments of 2.4–5.8 hours per week for boys during adolescence. This decrease is especially important given that noncognitive skills continue to develop through adolescence (Heckman and Mosso 2014). Differences in leisure activities and entertainment account for about 70% of the gender gap in paternal investment losses.

Considering changes in investments from other sources is also important because the differences in paternal time could over- or underestimate the gap in total investments. I find little to no evidence that mothers increase investments in boys relative to girls following a change in household structure. Grandparents are another potentially important sources of investments, given the substantial portion of children and adolescents that receive grandparental investments (Dunifon et al. 2018; Kalil et al. 2014). I find that investments from grandparents increase for boys relative to girls after the transition to a single-mother household. The magnitude is one-half the size of the estimated gap in paternal investments, suggesting that grandparents could play a significant role, but the estimated differences are statistically insignificant.

Finding differential investment losses by gender suggests that living in a single-mother household could have a large negative impact on the quantity of investments that boys receive relative to girls. All this evidence considered together with existing research suggests that time investments are another potential mechanism that may help explain the noncognitive skill gender gap, specifically for those in single-mother households. Future research should consider a more direct link between time investments and outcomes, focusing on measures of noncognitive skills.

Acknowledgments

I would like to thank Todd Elder, Stacy Dickert-Conlin, Scott Imberman, Michelle Maxfield, Kelly Vosters, Keith Teltser, Alex James, and seminar participants at Michigan State University for helpful comments. Any opinions expressed here are those of the author and do not necessarily represent the views of the U.S. Department of Education.

Funding Information

This research was supported by a Pre-Doctoral Training Grant from the IES, U.S. Department of Education (Award #R305B090011) to Michigan State University.

Data Availability

The data used in this study are from the Panel Study of Income Dynamics (https://simba.isr.umich.edu/data/data.aspx#gsc.tab=0).

Compliance With Ethical Standards

Conflict of Interest

The author reports no conflict of interest.

Notes

References

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.