Abstract
Using data from the 2003–2007 American Time Use Surveys (ATUS), we compare mothers’ (N = 6,640) time spent in four parenting activities across maternal education and child age subgroups. We test the hypothesis that highly educated mothers not only spend more time in active child care than less-educated mothers but also alter the composition of that time to suit children’s developmental needs more than less-educated mothers. Results support this hypothesis: not only do highly educated mothers invest more time in basic care and play when youngest children are infants or toddlers than when children are older, but differences across education groups in basic care and play time are largest among mothers with infants or toddlers; by contrast, highly educated mothers invest more time in management activities when children are 6 to 13 years old than when children are younger, and differences across education groups in management are largest among mothers with school-aged children. These patterns indicate that the education gradient in mothers’ time with children is characterized by a “developmental gradient.”
Introduction
The destinies of children born to the most- and least-educated women have increasingly diverged over the past several decades. Historical trends in maternal employment, family structure, and education have increased inequality among children such that those born to the most-educated women have gained resources in terms of parents’ money and time, whereas those born to the least-educated women have lost such resources (McLanahan 2004). The present study investigates maternal time with children, which is a key parental “resource” for child development, especially during children’s early years (Fuligni and Brooks-Gunn 2004; Sayer et al. 2004a). To the extent that the destinies of children born to the most- and least-educated women are increasingly divergent, this particular inequality may have important implications for long-term patterns of attainment and achievement.
Whereas previous research on maternal education and time use suggests that highly educated mothers spend more time overall in active care of children than less-educated mothers, our study is the first to examine whether highly educated mothers also alter the composition of their child-rearing time for children of different ages in ways that may optimize children’s development. Although this study does not examine child outcomes, we invoke theories from developmental psychology to interpret the likely impact of both the level and composition of time on child development.
To investigate these patterns, we use data from multiple waves of the American Time Use Survey (ATUS). The rich time-use data in the ATUS, which are obtained from daily diaries, allow us to examine the time mothers spend engaged in specific activities—basic care, teaching, play, and management—that facilitate children’s development in different ways at different ages. The large ATUS sample size allows us to look separately at households with children in different developmental periods to determine whether mothers tailor the time that they spend engaged in these activities to their children’s ages and, more specifically, whether the most-educated mothers tailor their time more optimally than the least-educated mothers.
Maternal Education and Child Time: Background Literature
Both sociological and psychological research has found that parents with different education levels raise their children differently (Hill and Stafford 1974; Lareau 1989). Parental education, as an indicator of social class, has been linked to more optimal parenting behaviors across a range of domains, including more authoritative (vs. authoritarian) parenting styles (Pinderhughes et al. 2000), more sensitive and responsive mother-child interactions (NICHD ECCRN 2004), greater language stimulation (Hart and Risley 1995; Hoff 2003), and greater levels of parental management and advocacy (Lareau 1989). These patterns extend to parents’ time investments in children: numerous studies have documented that highly educated parents, and mothers in particular, spend more time in active care of children than their less-educated counterparts (Guryan, et al. 2008; Hill and Stafford 1974; Sayer et al. 2004b). Echoing McLanahan’s warning about diverging destinies, Ramey and Ramey (2010) reported that the increase in parents’ active care time with children over the past 40 years has emerged more strongly among highly educated parents than less-educated parents, suggesting that the education gradient in time with children may be widening.
Most recently, Guryan et al. (2008) reported that maternal time with children increases with education, using data from the American Time Use Survey (ATUS; the same data set we employ here). Because the “education gradient,” as they call it, for child care time differs from the education gradients for leisure and home production (which decrease with education), they argued that parental time investments in children reflect a fundamentally distinct phenomenon. Guryan et al. posited that highly educated parents (more so than less-educated parents) view time with children as an “investment behavior” with which to increase children’s human capital (for either altruistic or selfish reasons) and do not view market alternatives as highly effective substitutes for their own time investments. Yet, it remains unclear from this and previous research on maternal education and time with children whether highly educated parents’ “investment behavior” also means tailoring time with children to children’s developmental needs more so than less-educated parents.
Maternal Education and Child Time: Theoretical Frameworks
Economic models of the family view child development as an output of household production (Becker 1965). In this framework, the household derives utility from child development, or “child quality,” which parents produce by investing time and money in children’s well-being. The productivity of the household is determined by parents’ ability to make these and other investments efficiently, given time and budget constraints. All households are not equally efficient, however. We posit that efficiency in parenting can be seen, from a developmental psychological perspective, in parents’ tendency to tailor time with children in ways that optimize child development. This kind of efficiency means minimizing time on activities that are less important to children’s development and maximizing time on activities that are more important to children’s development. A key tenet in developmental theory is that an activity’s importance will vary by child age, in the ways that we describe herein.
Bianchi and Robinson (1997) argued that the education gradient in time spent with children is evidence that parental education is the predominant predictor of parents’ human and social capital investments in children. Sociological theory posits that education operates this way because through educational attainment, parents acquire values about social mobility and standards of success that in turn motivate certain parental behaviors (Sewell et al. 1970). However, a simple linear relationship between time investments in children and parental education does not necessarily reflect the concept of “developmentally effective” parenting. Presumably, for example, there is a point of diminishing returns to parental time for promoting child development, particularly for specific activities at specific developmental stages, examples of which we list later. If education enhances a person’s efficiency or effectiveness in other realms of nonmarket human behavior (Michael 1972), and if these realms include parenting, we would expect highly educated parents not only to spend more time with children overall but also to tailor their child care time more effectively than less-educated parents. Specifically, we would expect to see not only an education gradient in time spent with children but also a “developmental gradient,” such that highly educated parents vary the composition of their time according to their children’s developmental needs.
Although this proposition has not been explored in quantitative studies of parental time use, Lareau’s (2003) qualitative study of family life reported that middle-class parents target their time with children toward developmentally enhancing activities. In her study, middle-class families (whose jobs, by her definition, require college-level skills) engage in a pattern of “concerted cultivation” to actively develop children’s talents and skills. By contrast, in lower-class families, Lareau identified a pattern that she calls “the accomplishment of natural growth,” wherein parents attend to children’s material and emotional needs but presume that their talents and skills will develop without concerted parental intervention. Although Lareau (2002:72) argued that middle-class parents tend to engage their children in “numerous age-specific activities,” her thesis was based on case studies of 12 families with school-aged children. It is not clear whether her results generalize to a larger population or whether a “developmental gradient” exists in how middle-class parents orchestrate their children’s time.
Developmental Framework for Parental Time Investments
Developmental psychology offers a framework for understanding parenting behavior in terms of both effectiveness and Lareau’s concept of “concerted cultivation.” Specifically, developmental theory proposes that children at different developmental stages require different types of parental investments (although, of course, the importance of certain investments, such as warmth, adequate monitoring, and nourishment, remain constant across childhood). In the parenting literature, “sensitivity” is the defining construct of effective parenting; in its broadest conceptualization, the hallmark of sensitive parenting is responding contingently to children’s needs (Adamson and Bakeman 1984; Bornstein 2002; Carew 1980). Sensitivity in parents’ time investments thus means tailoring child-rearing time to the specific challenges, tasks, and capacities that dominate a child’s developmental period. In the case of a 5-year-old child, for example, one hour of time spent exclusively on “basic care” may confer fewer benefits to child development than an hour of time spent “teaching,” whereas an hour of “basic care” may confer greater benefits to an infant than time spent in formal teaching activities.
We identify four categories of active parenting that reflect types of parental investments best suited to a particular developmental period. These are (1) basic care, which refers to routine tasks, such as feeding, bathing, and physically caring for the child (including attention to health needs); (2) play, which includes playing games, doing art projects, and pretending; (3) teaching, which includes reading to the child or helping with homework; and (4) management, which refers to planning, organizing, and monitoring the child’s life outside the home. The developmental stages for which each activity is best suited are infancy, toddlerhood, the preschool period, and middle childhood, respectively. Although parents engage in many other activities that surely influence children’s development, we examine mothers’ time in these activities because their relative importance to children’s well-being varies by developmental period; thus, examining them allows us to test our hypothesis that the education gradient in time with children is characterized by a developmental gradient.
During infancy (0 to 12 months), parents spend most of their child care time performing basic caregiving tasks, such as feeding, bathing, and comforting (Bornstein 2002). Through the provision of basic care, parents help infants establish regular sleeping and eating routines, the two central challenges of infancy. According to attachment theory, it is also through warm, consistent, and sensitive responses to infants’ basic physical and emotional needs that parents establish secure emotional bonds or attachments with their infants. These attachments serve as children’s mental model of future relationships and form the foundation of their socioemotional development (Ainsworth et al. 1978; Bowlby 1969). Although attachment theory specifies that the quality of mothers’ responses during basic care shapes mother-child attachments, a greater quantity of time in basic care increases mothers’ opportunities to demonstrate and practice their responsiveness. During infancy, then, basic care is the paramount parenting activity.
During toddlerhood (12 to 35 months), as babies become capable of initiating and maintaining social interactions and acquire the capacity for representational thought, they begin to engage in “symbolic” or pretend play (Piaget 1952). Engaging in pretend play not only reflects children’s increased capacities but also promotes their cognitive and social skills, including attention, memory, logical reasoning, vocabulary, creativity, and behavioral and emotional regulation (Bergen and Mauer 2000; Berk 2001; Elias and Berk 2002; Lindsey and Mize 2000; Ruff and Capozzoli 2003). According to sociocultural theory, toddlers learn most during play when a parent (or other adult) facilitates and structures their activities (Keren et al. 2005; Rogoff 2003) in ways that help children explore their environment, grasp concepts, express curiosity about novel objects, and gain competence motivation (Hubley and Trevarthen 1979; Sigel 1986). By actively guiding children’s play, parents also positively foster compliance (Parpal and Maccoby 1985) and stimulate language development (Duckworth 1972). In toddlerhood, then, developmental theory suggests parents respond most contingently to children’s developmental needs through child-directed play.
During the preschool period (ages 3 to 5), children’s advancing conversation and attention skills typically increase their appetite for didactic activities, such as book reading, puzzles, and problem solving (Hoff 2006). These kinds of teaching activities help children master many cognitive skills that influence early academic outcomes, such as distinguishing print in books, recognizing letters and numbers, and identifying words (Snow 2006). A large body of research has established that the frequency with which parents engage in these kinds of teaching activities is associated with language and literacy development (e.g., Bus et al. 1995; Roberts et al. 2005) as well as early math and reading scores (e.g., Bradley et al. 1988). Thus, while child-directed play still provides important opportunities for enrichment—and although some types of teaching activities, such as reading, are important to initiate well before the preschool period—parents’ time in teaching activities may be particularly important during the years just prior to school entry.
During middle childhood (ages 6 to 13), children’s lives extend beyond the family to include schools, peers, and extracurricular activities. As children’s worlds expand, parents spend less time interacting directly with them and more time planning and monitoring their academic and social networks (Collins et al. 2002). Through this management, parents ensure that children form positive relationships, learn self-management, and adopt a sense of personal responsibility in their extrafamilial lives (Collins et al. 2002). Early on, management entails learning about and arranging for enriching academic, recreational, and social opportunities (Dryfoos 1999; Vuchinich et al. 1992). As children near adolescence, management also involves monitoring social networks and activities to minimize children’s exposure to violence, substance use, and delinquent peers (Dishion et al. 1999; Dubow et al. 1997). During middle childhood, then, parents respond contingently to children’s developmental needs by actively managing their increasingly complex lives.
Hypotheses
The hypothesis that highly educated mothers tailor their child-rearing time to children’s developmental needs more so than less-educated mothers could manifest in two ways. First, highly educated parents could invest more time in the activity most appropriate for their child’s developmental stage relative tothe timethat theyinvest inthese activitiesduring otherstages ofdevelopment. For example, when their child is an infant or a toddler, highly educated parents should invest more time in basic care and play than they do later in a child’s development. Second, highly educated parents could tailor their child-rearing time to their children’s developmental needs to agreater degreethan less-educated parents. Importantly, demonstrating that highly educated parents concertedly cultivate across time periods does not mean that less-educated parents do not also tailor their time accordingly. Thus, we hypothesize that highly educated parents will not only invest more time in basic care and play during infancy and toddlerhood relative to other ages, but that the difference in the amount of mothers’ time in basic care and play across education groups (the education gradient) should be largest during infancy and toddlerhood, respectively. Likewise, the education gradient in teaching activities should be largest during the preschool years, and it should be largest for management during middle childhood. These patterns would indicate that the education gradient is characterized by a developmental gradient.
Method
Data
Data are drawn from the 2003–2007 American Time Use Surveys (ATUS). The ATUS is an ongoing, monthly, national survey conducted by the U.S. Census Bureau for the Bureau of Labor Statistics, fielded in conjunction with the Current Population Survey (CPS). Households are selected so that comparisons can be made across demographic groups defined by marital status, children’s age, or other criteria. Telephone interviews (aided by CATI technology) ask one randomly selected person aged 15 or older within each household to recount how that person’s time was spent during the preceding day. The sample covers all days of the week and all months of the year. We select as respondents women with at least one “own child” (biological or adopted) aged 13 or younger in her home. The ATUS collects information on the respondent’s stated primary activity, which could be a child care activity, and whether child care was performed as a secondary activity. We examine only primary child care activities and not secondary child care time because only primary activity codes identify the specific developmentally enriching activity in which the parent is engaging.
These data offer several advantages for the proposed work. First, the survey’s diary approach is a relatively flexible and accurate method for measuring time use, especially time devoted to caregiving. Few other U.S. nationally representative surveys have collected any information on caregiving, and not in as great detail as the ATUS. Second, the survey’s large size and sampling methodology are designed to facilitate the types of comparisons between mothers with different levels of education as are proposed here. Third, the ATUS collects data on parents’ demographic characteristics, such as maternal age and number of children in the household. These measures are likely associated with education and time in child care, thus their inclusion allows us to control for the influence of characteristics that distinguish parents of different education levels.
We define child age groups by the age of the youngest child in the household to ensure that at least one child of the target developmental period lives in the household (an approach also used in Aguiar and Hurst (2009)). We then compare all time that mothers spend in each child care activity across the resulting age groups. We hypothesize that having an infant in the household will increase mothers’ time spent in basic care relative to having only preschool or school-aged children, that having a preschool-aged child in the household will increase mothers’ time spent in teaching relative to having only school-aged children, and so on. Our classification means that if the youngest child in the household is in middle childhood, the mother cannot report time with infants, toddlers, or preschoolers. However, if the youngest child is an infant or a toddler, the mother may also report time with preschool- or school-aged children. We address the implications of this ambiguity in the discussion. Another limitation is that the data are pooled from cross-sectional waves, thus obviating longitudinal comparisons.
Sample
Our analytic sample is limited to our target population: that is, to women with at least one “own child” (biological or adopted) aged 13 or younger in her home. We further eliminate any women who have more household children (children living in the household who are not necessarily the mother’s own) than own children (n = 1,756) to limit the number of blended families included because maternal time allocation might differ in unknown ways in these families, particularly if the youngest child is not the mother’s own. Because mothers may have greater flexibility in their weekend schedules, the amount and distribution of time that they choose to devote to children on weekend days is of greater interest than time on weekdays. Therefore, we further restrict the sample to mothers whose time diary captured a weekend day. These restrictions leave us with 6,640 cases for analysis. As per ATUS instructions, weights adjusted to the 2006 population are used for all years in all analyses.
Table 1 displays descriptive statistics for our analytic sample. Roughly 12 % of mothers fitting our criteria have less than a high school education (including those with a GED), 28 % have a high school diploma, 29 % have some college or an associate’s degree, and 31 % have a college degree. The mothers’ average age is just under 36 years. Our sample is predominately white/Asian/other (69 %), with 12 % African American and 18 % Hispanic. Most families have two children.
Almost one-third of the sample (32 %) have a child younger than 3 (0 to 2 years), and 23 % have a youngest child older than 2 but younger than 6 (3 to 5), and 45 % have a youngest child 6 or older (6 to 13). Although we distinguish theoretically between infancy and toddlerhood (i.e., the first year of life vs. the second and third years), we combined these groups in analyses because distinguishing between households with youngest children in the first and second-to-third years of life yielded prohibitively small age group by education cell sizes.
Measures
Child Care Time
We examine the amount of time mothers spend in active child care per weekend day. By active care, we mean time spent directly engaging with children. This time is broken into four domains—basic care, play, teaching, and management—corresponding to our developmental hypothesis. As described earlier, basic care includes direct physical care of household children. Thus, if a mother spent time feeding (including nursing), bathing, physically comforting, physically attending to health needs, or putting a child to sleep, she was coded as engaging in basic care. Play includes sports, arts and crafts, and general play with household children. If a mother was playing a game, “pretend playing,” or using clay with a child, as examples, she was coded as playing. Teaching children includes reading, talking, and helping with homework or any other educational activity directly related to children. Thus, for example, if a mother was reading a book, having a conversation, or solving a puzzle with a child, she was coded as teaching. Management of children includes attending events, traveling, and planning activities on children’s behalf. Thus, for example, if a mother was attending a child’s performance, driving a child to an activity, or arranging for a child to participate in an activity, she was coded as managing. We based these categories on those of Guryan et al. (2008), but altered them somewhat to fit our hypotheses and limited them to time spent with household children. These activity codes are listed in the Appendix. We also use a more global measure of primary child care time that records the amount of time spent in any of these activities.
Independent Variables
Our main independent variable is maternal education. We calculated education level based on mothers’ highest educational degree attained. Four mutually exclusive levels of education are used: less than a high school diploma (including GED recipients), high school diploma, some college (including an associate’s degree), and college degree or higher. In models, three dummy variables are entered with high school education as the reference category. The other main independent variable is age of youngest child. Models are run on the full sample with two dummy variables for child age group and 0 to 2, 3 to 5, or 6 to 13 as the reference category, depending on which activity is under examination.
Covariates
We control for a set of characteristics that may predict maternal education level and may covary with maternal time with children. Specifically, mothers’ race/ethnicity is entered to control for any possible cultural differences in parenting styles (two dummy variables are used for black and Hispanic, with white as the reference category). We also control for mothers’ age at the time of the interview (mean-centered to ease interpretation of the intercept). We also enter the number of children in the household in these models, even though maternal education level arguably affects family size. Because we cannot identify which child the mother is interacting with, mothers with more children in the household should report more time with children, all else being equal. If family size varies by age of youngest child, then family size will confound the basic patterns we aim to investigate. Having multiple children could also change the nature of time spent with the youngest children because the oldest could do some child care, especially play. The number of children in the household is centered at 1.
Analysis Plan
We begin by examining unadjusted differences in mothers’ time in child care by education level, age of youngest child, and activity type. The mean minutes for each subgroup will indicate whether, unadjusted for any exogenous characteristics, child time patterns are characterized by both education and developmental gradients.
Next, in Tobit models, we regress time in each activity type, as well as our global measure of all types of child care time, on dummy variables for maternal education level and child age group, controlling for maternal race, age, and family size. Six variables interacting maternal education and child age are entered to determine whether differences by education level vary significantly across age groups. The reference categories in these models are high school education and the childage groupfor whichthe activityis mostimportant developmentally. Thus, in models predicting time in basic care and play, 0 to 2 is the reference category because basic care and play are the primary parenting activities when children are infants and toddlers. In the model predicting time in teaching, the reference category is ages 3 to 5 because teaching becomes paramount during the preschool years; in the model predicting management, the reference category is ages 6 to 13 because management becomes paramount during middle childhood. Because we hypothesize that educational differences will be largest in the age group for which each activity is most important, and that group is always the reference, significant negative regression coefficients for the interactions between college education (and some college) and both non-omitted age groups would indicate that educational differences are indeed largest for the omitted age group.
We use Tobit models because they are designed to estimate linear relationships between variables when there is extreme censoring in the dependent variable (Breen 1996; Greene 2003). Tobit models are often applied to time-use data because time is necessarily left-censored at 0, and a substantial proportion of cases often report 0 time spent in any given activity (Kalenkoski et al. 2008; Kimmel and Connelly 2007). Indeed, a large number of cases are censored at 0 minutes in each activity in our data. For example, nearly 80 % of mothers do no teaching, and nearly 80 % do no playing on the diary day (see Table 1). The Tobit model allows one estimation method to predict the probability of being censored and the expected value of the uncensored portion, conditional on observation (Breen 1996; Greene 2003).
We present marginal effects on the unconditional expected value for each type of care generated from the Tobit models (Greene 2003). We then generate predicted means from the marginal effects for the number of minutes mothers spend in each of these types of care at specific levels of maternal education and child age.
Results
Bivariate Results
Table 2 displays the average time spent in each of the four activities, and in total care, as minutes per weekend day. Consistent with Guryan et al.’s findings, a clear education gradient emerges at all ages in terms of total care time as well as time spent in most activities. However, the magnitude of the gradient varies by child age and activity type. In terms of total care time, the largest differences by mothers’ education emerge in the youngest and the oldest age groups. College-educated mothers in households with youngest children aged 0 to 2 spend 67 more minutes in total care than those with only a high school diploma. By contrast, college-educated mothers in households with youngest children aged 3 to 5 spend 21 more minutes in total care than high school–educated mothers. Among mothers in households with youngest children aged 6 to 13, those with a college education spend 22 more minutes in total care than their counterparts with only a high school education.
More interestingly, this overall education gradient masks important variations in the pattern by activity type. For basic care and play, the largest education gaps emerge when youngest children are aged 0 to 2, with college-educated mothers spending about 42 % more time in basic care (30 more minutes) and approximately 94 % more time in play (32 more minutes) than those with only a high school diploma. In contrast, when youngest children are aged 3 to 5, college-educated mothers spend only 5 minutes more in basic care (11 % more time) and play (19 % more time) than those with only a high school diploma. Finally, when youngest children are aged 6 to 13, college-educated mothers spend about 4 more minutes per weekend day in basic care than mothers with only a high school education and about the same amount of time in play.
For time spent teaching, the education gradient is more consistent across age groups, although a developmental gradient does emerge. When youngest children are aged 0 to 2, college-educated mothers spend 5 extra minutes (95 % more time) teaching children than mothers with only a high school diploma. In contrast, when youngest children are aged 6 to 13, college-educated mothers spend 3 extra minutes (59 % more time) teaching than mothers with only a high school diploma. The largest gap emerges when youngest children are aged 3 to 5: in this group, college-educated mothers spend about 8 more minutes per weekend day teaching (155 % more time) than high school–educated mothers. Nevertheless, it should be noted that all mothers spend the least amount of time in teaching at every developmental stage, so even large percentage differences equal small differences in actual minutes per weekend day.
Finally, the education gradient for management widens as youngest children age. When youngest children are 0 to 2 years old, college-educated mothers spend roughly the same amount of time in management as those with a high school diploma. In contrast, when youngest children are 6 to 13 years old, college-educated mothers spend 14 more minutes per weekend day in management (130 % more time) than mothers with a high school diploma.
Multivariate Results
To determine whether the bivariate patterns obtain when demographic differences are held constant, and to test whether education differences are statistically significant and different across age groups, we run Tobit regression models for time in all care and in each four activity types with maternal education levels and child age groups interacted, controlling for mothers’ race/ethnicity, age, and family size. In each model, significant negative effects for the interactions between college (and some college) and child age groups would support the hypothesis that college-educated mothers (and those with some college) tailor their child-rearing time to children’s developmental needs more so than high school–educated mothers. For example, because youngest child aged 0 to 2 is the reference category in models predicting time in basic care and play, a significant negative interaction between college educated and youngest child aged 6 to 13 would indicate that the gaps in basic care and play between college- and high school–educated mothers are smaller when children are in middle childhood than when they are infants and toddlers. Likewise, because youngest child aged 6 to 13 is the reference category in models predicting time in management, a significant negative interaction between college-educated and youngest child aged 0 to 2 would indicate that the gap in management time between college-educated and high school–educated mothers is smaller when children are infants and toddlers than when they are in middle childhood.
Results from the model predicting total child care reflect patterns that emerge in bivariate analyses (see Table 3). An education gradient exists such that college-educated mothers spend more time in all care than high school–educated mothers. Moreover, the gradient differs by age of youngest child. The significant negative effects of college interacted with youngest child aged 3 to 5 and college interacted with youngest child aged 6 to 13 (and some college interacted with youngest child aged 3 to 5) indicate that differences by maternal education in total child care time are smaller when youngest children are preschool and school aged compared with when children are very young. Results from this model are also displayed in Fig. 1. It shows that the gap (between college-educated mothers and high school–educated mothers) in total care time is over 60 minutes when youngest children are aged 0 to 2 but roughly 20 minutes when youngest children are aged 3 to 5, and 6 to 13.
For mothers’ time in basic care, significant and negative interactions emerge between higher maternal education and youngest child age group that suggest a developmental gradient in mothers’ time in basic care. College-educated mothers spend more time in basic care than high school–educated mothers during all developmental periods; however, this difference is significantly larger when youngest children are infants and toddlers than when youngest children are preschool or school aged, as the significant negative interactions between college and youngest child aged 3 to 5, and college and youngest child aged 6 to 13 indicate. Figure 2 displays predicted means for this model. It shows that the gap (between college- and high school–educated mothers) in basic care time is lower by 22 minutes when youngest children are 3 to 5 versus 0 to 2, and by 26 minutes when youngest children are 6 to 13 versus 0 to 2 (see Fig. 2).
A similar pattern emerges for play. Here again, the reference group is youngest children aged 0 to 2 years. The significant, negative interactions between college-educated and youngest child aged 3 to 5 (and youngest child aged 6 to 13) indicate that the differences between college-educated mothers and high school–educated mothers in play are significantly smaller when youngest children are preschool or school aged than when they are very young (see Table 3). Specifically, the difference between college- and high school–educated mothers’ time in play is lower by 25 minutes when youngest children are preschool aged and by 31 minutes when youngest children are school aged compared with when youngest children are infants and toddlers (see Fig. 3). Thus, college-educated mothers, to a greater extent than less-educated mothers, emphasize play when youngest children are infants and toddlers but de-emphasize play as children grow older, suggesting that the education gradient for time in play, as in basic care, is characterized by a developmental gradient.
Table 3 also shows results for models predicting mothers’ time spent teaching. In these models, the reference category is youngest child aged 3 to 5. Thus, significant negative interactions between college- (and some college-) educated and youngest child age groups 0 to 2 and 6 to 13 would suggest the education gradient is largest when youngest children are 3 to 5. Indeed, the interaction between college-educated and youngest child ages 6 to 13 is negative and significant, indicating that the difference between college- and high school–educated mothers’ time in teaching is largest when youngest children are aged 3 to 5 (the difference in predicted means in the two groups is 8 minutes) and smallest when youngest children are aged 6 to 13 (the difference in predicted means in the two groups is 2 minutes) (see Fig. 4). Although no other significant interactions emerge, the interaction between college and youngest child aged 0 to 2 is negative, as are the interactions between some college and both non-omitted child age groups. Taken together, these results suggest a pattern consistent with the hypothesis that mothers with college and some college spend more time in teaching when youngest children are preschoolers than when they are infants, toddlers, or school age, again supporting our notion of a developmental gradient.
Finally, patterns by age diverge sharply for mothers’ time spent in management when compared with time spent in all other activities. In these models, the reference category is youngest child aged 6 to 13 because the education gradient is hypothesized to be largest when youngest children are that age. Indeed, significant negative interactions between college-educated and youngest child aged 0 to 2 and 3 to 5 indicate that although mothers with a college degree spend more time in management than high school–educated mothers overall, those differences are far larger when youngest children are 6 to 13 years than when children are infants and toddlers or preschool age. Figure 5 displays predicted mean minutes in management for mothers by education level and youngest child age. When youngest children are infants and toddlers, mothers with college, some college, and high school education spend almost equal time in management (mean minutes are 11, 13, and 14, respectively). However, when youngest children are 6 to 13 years, mothers with a college education spend 25 minutes in management versus 13 minutes for high school–educated mothers. Similar differences emerge when we compare mothers with some college and those with only a high school education, although the interaction terms are smaller and only approach statistical significance. These results indicate that for time spent in management, the education gradient is largest when youngest children are 6 to 13, a result that supports the hypothesis that college-educated mothers attend to a developmental gradient in management more so than less-educated mothers.
Although these analyses include mothers interviewed on weekend days only, we replicate all analyses with mothers interviewed on weekdays to assess the sensitivity of our results to diary day. In the weekday sample, the signs of all significant maternal education by child-age interaction terms in the models predicting all child care time, basic care, play, and management are the same as in the weekend sample. In models predicting all child care and basic care, the interactions that are significant in the weekend sample are also significant in the weekday sample at the .05 level. Only in models predicting time in teaching do the patterns in the weekday sample not resemble the developmental gradient that emerged in the weekend sample. This similarity suggests that the educational and developmental gradient patterns hold throughout the week, although they emerge more strongly on weekend days (these results are available upon request).
Discussion
In his 1972 book, Robert Michael wrote that “a distinguishing characteristic of human capital is that it is embedded in an individual and therefore goes with him wherever he goes—into the labor market, into the theater, into the voting booth, and into the kitchen” (Michael 1972:3). Michael argued that human capital enhanced one’s capabilities as an efficient consumer (of purchased goods and services, in that research). The present study illustrates this idea by showing that a mother’s human capital (i.e., her education) influences not only the amount of time she spends with her children (which may not be related to efficiency in a linear fashion) but also the composition of that time with her different-aged children.
Specifically, we found (as have others before us) an “education gradient” in mothers’ use of time: in almost all cases, highly educated mothers spend more time than less-educated mothers in the broad categories of child time investments that promote development. However, we also identified a “developmental gradient,” such that highly educated mothers shift the composition of their time in ways that specifically promote children’s development at different developmental stages. Specifically, the education gradient in basic care and play is greatest when youngest children are infants and toddlers (0 to 2), which is precisely when children most require parents’ time on such basic activities as bathing and feeding and also precisely the age when parent–child play is at its most developmentally appropriate. The education gradient for teaching is greatest when youngest children are preschool aged (3 to 5), which is precisely when time spent in learning activities (such as reading and problem solving) best prepare children for school entry. Conversely, the education gradient in management is greatest when youngest children are between the ages of 6 and 13—precisely the ages when parental management is a key, developmentally appropriate input.
Most of our significant results emerged in the comparison between mothers with a college education and those with only a high school education; the differences between mothers with some college and those with only a high school education reflected the hypothesized direction, but were seldom statistically significant. This pattern suggests the developmental gradient may be particularly pronounced among college-educated mothers. This phenomenon echoes the findings reported in Ramey and Ramey (2010), who described the “rugrat race” among highly educated parents. Such parents spend an ever-increasing amount of time in child care to (in the Rameys’ view) increase the chances their children will gain entry into a good college. At the same time, whereas the Rameys emphasize increasing divergence by parental education in time in “management-type” activities with school-aged children, our findings show that with respect to total child care time, the educational gradient is most apparent in households with the youngest children, a point also made by Hurst (2010), who argued that nearly all the increase over time in educational differences in time use is driven by households with young children (see also Sacks and Stevenson 2010). Thus, although college-educated mothers in our sample may be attempting to build their children’s resumes by enrolling them in a plethora of extracurricular activities, as described by Ramey and Ramey, they appear to have also adopted the mantra, championed in academic research (e.g., Heckman 2011) as well as in the popular press, that parental investments in the earliest years are key ingredients in children’s long-run success.
Before considering the implications of our results, it is important to note their limitations. Most notably, we do not identify which household child the mother is interacting with during a particular activity. Thus, we cannot argue that mothers are spending more time in basic care with infants and toddlers than they are with 6- to 13-year-olds, only that when an infant or toddler lives in the house, mothers are spending more time in basic care and play than they do when only 6- to 13-year-olds live in the house. However, if this ambiguity biased our estimates, we believe that it could only have minimized the developmental gradient in mothers’ time use that we identified. The inclusion of older children in the infant and toddler subgroup families theoretically dilutes age group variation, making differences in mothers’ time use across age groups, as well as differences in the education gradient across age groups, harder to detect.
In sensitivity analyses, we assessed the impact of mothers’ interacting with children of different ages in the same household by restricting our sample to families with only one child. In these families, we can safely assume that mothers report all time with the youngest child. For basic care, teaching, and management, the estimated interactions between maternal education and child age groups are comparable in sign and size. The developmental gradient in play does not emerge, however, among one-child families. It is difficult to interpret this difference from the full sample results because the one-child sample, which is much smaller (N = 2,184), may differ from multiple-child families on a number of unobservable characteristics that could influence time spent in play. One possibility, however, is that college-educated mothers in one-child families maintain higher levels of play throughout children’s school years than their counterparts with multiple children because the basic care of younger children does not monopolize their time for play with older children. Because the possible uniqueness of one-child families would limit the generalizability of our findings, these sensitivity analyses are not our preferred results even though they clarify the interpretation of mothers’ time.
Another limitation of the ATUS is one shared by all time diary surveys: namely, that child care time is self-reported. Thus, it is possible that college-educated mothers emphasize different activities when reporting their time rather than actually spending their time differently. For instance, perhaps college-educated mothers believe that they “should” spend time engaging in enriching activities with their children. Therefore, when talking to their children while doing other household tasks, they are likely to report their “primary” activity as “talking to children.” In the same situation, high school–educated mothers may report the household task as their “primary” activity, with talking to their children coded as “secondary” child care time. In the absence of observational data, it is difficult to determine the extent to which differences in reporting drive differences by parent education level. However, observational studies comparing the quantity and quality of parents’ language use with children by parent education (Hart and Risley 1995), as well as parents’ time spent in child-directed activities (Lareau 2003), suggest that the education gradient found in time diary data is not strongly biased by reporting differences.
Other particularities of the ATUS limited our analyses in less obvious but still important ways. For one, we could not determine whether mothers were interacting with their biological or nonbiological household children. It is plausible that mothers allocate time differently between biological and nonbiological children; thus, distinguishing between them might reveal more precise time-use patterns. However, by excluding mothers who reported more household children than own children, thereby reducing the proportion of blended or stepfamilies in the analytic sample, we reduced the possibility that mothers’ time with nonbiological children altered our findings. It also may be true that examining time with children as a primary activity ignores meaningful time that mothers spend with children in other types of activities, such as family meals, travel not directly related to children, or conversations during household tasks. However, it is impossible in the ATUS to determine whether children are the focus of attention during tasks such as meals, travel, or household chores—only that they were with the mother in the same room or location—so to include these kinds of activities would likely capture time not directly relevant to children’s development. Finally, data on fathers’ education level (as reported by mothers) is available only if mothers lived with the father or another partner; therefore, we could not control for fathers’/partners’ education in our analyses.
Despite these limitations, the rich information on mothers’ time use in the ATUS allows us to identify two important gradients in maternal time with children: one educational and one developmental. As a next step, future research should test our theoretical proposition that children at different developmental stages benefit from different types of parental investments using data that includes child-outcome measures. Specifically, these analyses could investigate whether, in fact, time invested in basic care and play are most strongly linked to child outcomes during children’s infancy and toddlerhood, whether teaching activities are most strongly linked to child outcomes during the preschool years, and whether time spent in management is most strongly linked to child outcomes during middle childhood. Moreover, these analyses could explore the idea that the “ideal” amount of time spent in different child care tasks may differ for children with different developmental characteristics or needs, by maternal education level or socioeconomic circumstances, as well as by the goals that parents in different contexts may have for their children (see, e.g., Edin and Kefalas 2007; Furstenberg et al. 1998).
Although we have identified a developmental gradient in mothers’ time with children by education level, we cannot know whether the developmental gradient arises directly or indirectly from maternal education. It is possible that mothers learn about child development during college—such that highly educated mothers are consciously acting on knowledge attained in formal schooling—or that college alters mothers’ goals for their children and thus their parenting values, as sociological theory would suggest (Sewell et al. 1970). Alternatively, mothers with parenting values that make them more responsive to children’s emerging skills and abilities simply may select to receive more education or may progress educationally because of stronger academic ability.
The patterns that we observe could also reflect the numerous correlates of maternal high education. For instance, highly educated mothers are more likely to have higher incomes, to be married, and to have spouses who are more involved in child rearing; they may also have more flexible work schedules due their different types of employment. All these demographic correlates of maternal education represent plausible pathways through which maternal education is associated with time use. Finally, highly educated mothers could differ from less-educated mothers in their consumption motives for spending time with children (Sacks and Stevenson 2010).
Highly educated mothers could also differ in their preferences for or beliefs about “investing in children” and its later-life economic returns for children (Heckman 2011). This explanation supports the view that highly educated parents “concertedly cultivate” their children’s development in ways that may optimize children’s long-run success. Lareau (2003) described highly educated parents concertedly cultivating children through a set of behaviors, such as scheduling and monitoring enriching extracurricular activities as well as advocating for children with teachers, which constitute “management” by our definition. Lareau interviewed families with school-aged children; therefore, the prominence of management in her description jibes with developmental theory and our findings. However, our results also suggest that highly educated parents concertedly cultivate children in different ways at different ages and that the parenting strategies Lareau identified may reflect broader patterns established much earlier in children’s lives—patterns that could have long-term implications for children’s achievement and attainment. To the extent that highly educated parents increasingly adopt these patterns of investing in their children, the destinies of the children of college-educated parents may diverge even farther from those of their less-advantaged peers.
Acknowledgments
The present study was supported in part by a grant to the second author from the National Institute of Child Health and Human Development (NICHD) (#1F32HD54044), and by grants to the third author from the National Institute on Aging (NIA) (#5T32AG000243) and the Agency for Healthcare Research and Quality (AHRQ) (5T32HS000084). We would like to thank Joseph Price, Sandra Hofferth, Robert Michael, and numerous attendees of the 2009 American Time Use Research Conference for helpful feedback on an earlier version of this article. We are also indebted to three anonymous reviewers for their invaluable suggestions. While we appreciate these sources of support and assistance, we remain responsible for the contents of and conclusions in this article.
