Introduction

Great effort has been put into explaining why some children achieve success in young adulthood while others do not. In the related literature, success is normally defined by academic achievement, occupation and income levels, among others.

The literature has assigned a key role to premarket factors in explaining success in the market. Premarket factors are broadly interpreted to represent (i) endowed ability, which is usually presented as the argument to explain the positive association between one’s educational attainment and that of one’s parents; (ii) parents’ decisions on the amount, nature and timing of resource distribution and its influence on children’s achievements;Footnote 1 (iii) additional choices made by parents that also affect children, namely neighborhood, type of school or number of siblings; and (iv) the characteristics of schools, which are not directly controlled by parents. Haveman and Wolfe (1995) and Monna and Gauthier (2008) present a review of methods and findings. Nonetheless, there has been considerable debate over exactly which inputs increase children’s achievement and to what extent. The related literature has mainly focused on two approaches: (i) gaining greater insight about determinants of parental and school investment, its various forms, and its effectiveness; and (ii) analyzing different interventions to correct disparities in children’s success.

In this paper, our aim is to contribute to the first of these approaches. Specifically, the first goal of this paper is to determine how time allocation to children has changed over time. The main contribution here is to incorporate two different sources of heterogeneity. On the one hand, we consider not only total time spent with children, but also the type of time devoted to them. We distinguish between active time, which includes activities which children benefit from directly, and passive time, which includes activities where children are present. On the other hand, we make a distinction made between children under 6 years of age, who are assumed to be time-intensive in their production and those older than six, who are assumed to be goods-intensive. The reason that children under the age of six are considered time-intensive is that school attendance becomes compulsory at about that age. The second goal of this paper is to gain a better understanding of the economic determinants of time-allocation decisions and whether such determinants have changed over time.

The rest of this paper is organized as follows. The related literature is reviewed in the following section. A theoretical model and its main implications are presented in section “Model”. In section “Data”, the time-diary data and the measurement of quality of time are described. In section “Empirical Analysis”, a descriptive analysis of the main changes in allocation of time, if any, is presented. The empirical model to identify the determinants of time allocation is then described and the results of the estimation presented. Finally, this article ended with section “Conclusions”.

Related Literature

Economic studies examine how shifts in family structure and maternal employment status contribute to the time mothers spend with children and fertility rate (e.g. Caucutt et al. 2002; Erosa et al. 2002) or to the marital dissolution rate (Weagley et al. 2007) or to children’s obesity (Miller 2011; You and Davis 2011). Most of these studies have found a significant negative effect on time with children in the case of working mothers and single-parent households. Using the 1981 National Longitudinal Survey of Youth, Baydar et al. (1999) found that working mothers spent 1 hour less in physical care, half an hour less in interactive care and 2 h less in passive supervision of children.

However, different results can be found in more recent literature. According to Hofferth and Sandberg (2001), the structural effects reflect the decrease in time with children due to demographic changes (mainly mother’s employment and marital status), but there exist also behavioral effects, that work in the opposite direction. They find that those behavioral effects play as important a role as structural effects regarding shifts in family structure and maternal employment. They report that American children spent no less time with parents in 1997 than in 1981, and that in two-parent families, time with parents actually increased over that period. Bianchi (2000) has reported that the time mothers devote to care-giving activities has actually increased slightly. Sayer et al. (2004) also found that although there were intermittent declines, both mothers and fathers from the US spent more time with children in the 1990s compared to the 1960s. Craig (2007) found that although working mothers decrease time with children, they manage to avoid a one-to-one trade-off between work and child care. Similar evidence is found in Bianchi and Robinson (1997) and Stafford and Yeung (2005), among others.

Due to the limited availability of time-diary data, economic studies dealing with determinants of parent’s allocation of time are scarce. Using US time-diary data from 1975 to 1981, Kooreman and Kapteyn (1987) found that mothers’ allocation of time to children is positively affected by fathers’ wages, but does not respond to changes in the mother’s own wages. Using the same survey, Nock and Kingston (1988) reported that mother’s employment decreased the time spent with children during after-school hours, but this reduction was concentrated in secondary activities, while little evidence of such behavior was found among fathers. Bryant and Zick (1996) also found that the hours that mothers spent in the labor market reduced the time they devoted to child care, particularly for older children. For an Australian sample, Craig (2007) investigates how working mothers manage to avoid the one-to-one trade-off between work and child care. Using Swedish data, Hallberg and Klevmarken (2003) found that the time a spouse spends on child care has a positive impact on partner’s child care time, that neither personal nor partner’s wages affect childcare time, that own spouse’s hours worked have a negative effect on time spent on child care, while partner’s hours worked have a positive effect. Bianchi et al. (2006) reported that increased parental time on child care is the result of incorporating children into adult leisure time, and the gradually increasing effort that men are putting into housework.

Concerning the effects of family structure, Hofferth and Sandberg (2001) showed that single-parent households spend substantially less time with children. Based on British data, Kalenkoski et al. (2005) found that single parents spend more time on child care and less time working. Using data from the US, Kalenkoski et al. (2007) found no evidence that single parenthood affects total time with children. For a summary on social and economic determinants of parental time see Monna and Gauthier (2008).

Therefore, in related literature there is little evidence on how mother allocate different types of time with children over time and how the type of children affects such an allocation. Moreover, related literature does not provide how those types of time with children react to different determinants over time. In this paper, we cover both sources of heterogeneity and we also provide an analysis of how economic factors influence the mother’s allocation of time over time. Therefore we cover the gap in the literature since we try to disentangle changes in parent’s allocation of time, while controlling for children’s age (time-intensive vs. goods-intensive children), type of time (active versus passive time) and certain parental characteristics.

Model

We present a well known model that considers a household whose members are, at least, the mother and her child (Hallberg and Klevmarken 2003; Rosen 1997). To avoid bargaining issues within the household, the mother is assumed to take the decisions even when considering a two-parent household. The mother obtains utility from consumption (C), from leisure (t l ), and from the production of child services (Q). The production of child services takes into account not only the number of children (n), but also their quality (q). Child quality (q) is treated as a function of parental child time (t c ), purchased child care (t cc ), and goods (s). The total amount of goods for children is not modelled, but calculated following the OECD household equivalence scale. We assume that day care is not necessary if the mother does not work in the market at all, and then add the condition that purchased child care is equal to market work time (t cc  = t h ). However, since some additional time devoted to children is required, the mother cannot enjoy all the time after work as leisure. Given the mother’s wage per hour (w), the non-labor female income in the household (R), and the charge for one hour of out-of-home child care (p cc ), the mother maximizes utility subject to a budget and time constraint:

$$ \begin{aligned} C = & \left( {w - p_{cc} } \right)t_{h} + R - ns \\ T =\, & t_{h} + t_{c} + t_{l} \\ \end{aligned} $$

The model yields demand functions for any use of time as functions of the wage rate (w), price of purchased child care (p cc ), and other income (R). An increase in the wage rate increases the opportunity cost of market work time, and therefore increases the quantity demanded. Parental allocation of time is influenced via three effects. The first effect is the production substitution effect, which leads parents to substitute more purchased child care (i.e., market work time) for own child care. The elasticity of substitution between market work time and time with children in the production function of child quality will determine whether there is an increase or decrease in leisure time.

The consumption substitution effect functions differently. The higher price of market work time or purchased child care increases the implicit price of child quality. Hence, parents are willing to reduce demand for child quality and increase consumption and leisure time. The size of this substitution depends on the magnitude of the elasticity of substitution concerning the utility function. The decreased demand for child quality implies lower parental child care time and purchased child care time. Finally, there is an income effect. The higher wage rate means higher real purchasing power, higher demand for child quality, consumption and leisure. Theoretically, then, any effect on allocation of time could be possible. A higher price for out-of-home child care time displays similar effects to an increase in wages. An increase in non-labor income generates only an income effect, which implies an increase in consumption goods and time. Moreover, since education is positively related to the wage rate, one should expect the same effect, but we can consider that parents with different educational levels value the quality of children differently.

To summarize, the analysis shows that different uses of time are interdependent and that the sign of the cumulative effects of changes in wages, income and education on the allocation of time is unclear, making the question an empirical one.

Data

Time-Diary Data Set

Substantial methodological work has demonstrated the validity and reliability of data collected in time-diary form (Juster and Stafford 1985). The data used in this paper have been drawn from 24-h time diaries collected in two surveys of the American population: “The 1998–99 Family Interaction, Social Capital, and Trends in Time Use Study” and “The Time Use Longitudinal Panel Study, 1975–1981”.Footnote 2 The latter consists of data from 620 respondents, their spouses if they were married at the time of first contact, and up to three children between the ages of three and seventeen living in the household. The key features which characterized the 1975 time use study were repeated in 1981. The former survey contains a representative sample of 1.151 respondents aged 18 and older. Data is also available for American Use of Time for more recent years. However, the reason we have chosen the period 1981-1997 is that we wanted to add extra evidence from existing studies (Hofferth and Sandberg 2001, among others) on how different types of time with children evolve over time and if they differ in their reaction to some economic determinants.

As pointed out in Monna and Gauthier (2008), researchers often prefer time-diary surveys since respondents are not asked to focus on any specific activity. In fact, since activities are reported on a real-time basis, the risk of social desirability is reduced, providing a more accurate measure of child care time than questionnaires (Robinson and Bostrom 1994, and Sayer et al. 2004).

In both of the data collection years, time-diary data were collected for the head of the household, the spouse (if present) and children aged 3–17 and adult individuals. The respondents are asked to complete “yesterday” 24-h time diaries detailing their primary activities, their secondary activities, and with whom and where they engaged in the activities. Respondents were able to report one hundred activities, which are classified into two main groups. The non-free time activities group includes activities related to paid work, household work, child care, obtaining goods and personal needs. The free time activities group includes activities related to education and training, organizational issues, entertainment or social activities, recreation services and communication and care.

Measurement of Quality Time

The related literature classifies or defines high-quality and low-quality child care time in a variety of ways. One existing classification considers primary or high-quality child care as the time reported in care tasks as the primary activity; and the secondary or low-quality child care measure considers the time engaged in any child care task as a secondary activity (Bianchi 2000; Hofferth and Sandberg 2001; Kalenkoski et al. 2005, 2007, 2009). Alternative, we can find a classification that takes into account who was present with the respondent while engaging in a primary activity; then the active or high-quality care measure includes the time during which a person was with a child while engaging in any child-related activity while the passive or low-quality care measure refers to the remaining time engaged in other activities while children are present (Kalenkoski et al. 2007, 2009; Stafford and Yeung 2005, 2006). Some related works (Kalenkoski and Foster 2008; Stafford and Yeung 2005, 2006) distinguish between development-oriented or high-quality care versus non-development-oriented care or low-quality care. These studies consider time devoted to a development-oriented child care activity such as teaching, playing with child and physical or emotional care of child, but do not take into account if the activity was a primary or secondary activity. The remaining time up to total time with children is defined as non-development oriented. Alternatively, some other authors consider sole-tasked care as high-quality and multi-tasked care as low-quality time, where the sole-tasked care measure counts time devoted to a primary activity for any childcare task that is not simultaneously performed with any non-childcare activity (Kalenkoski and Foster 2008). Finally, additional classifications are proposed including a four-category classification that incorporates intensity and development-oriented issues, specifically developmental child care, high contact child care, low intensity child care and supervisory/passive child care (Bittman et al. 2004).

There is no absolute ranking for these quality classifications. Kalenkoski and Foster (2008) compare the four alternative classifications across various characteristics. The first characteristic concerns the “adding-up” constraint, that is, total time in all activities would add up to a 24-h-per-day constraint. The property is fulfilled when child care is a primary activity. A second characteristic of potential interest is the intensity of the time parents spend on child care, since more intense time is likely to reflect greater parental investment. The active–passive classification captures intensity: active time is time spent directly with a child, whereas passive time is simply time spent during which a parent is accessible to a child.

Our measure also combines these characteristics. We only consider time spent on activities deemed to be primary activities, distinguishing whether the respondents were with children or not. We also distinguish between activities that children benefit from directly and activities they do not benefit from directly. We then define active or high-quality time as the sum of time reported in the ten child-focused activities in Child Care Activities. This group includes time devoted to the following activities: infant care, child care, helping/teaching, talking/reading, indoor playing, outdoor playing, medical care-child, other child care and travel. We calculate passive or low-quality time by adding time devoted to other activities in which children are present during the activity. Our choice is also compatible with the classification proposed in Bittman et al. (2004) since our active time is the sum of the first two categories, while our passive time is the result of the time spent in the two last categories. We consider total time with children as the sum of both active and passive time.

We measure time as a percentage of weekly time subtracting sleeping time. We also calculate market work time, leisure time and household production time as defined in the surveys with the particular feature that the respondents are not with their children. Moreover, subtracted sleeping time does not include joint sleeping time with children, not to neglect positive interactions between mother and children. In addition to the information on time use obtained from the time diaries, the project elicited information on other socioeconomic characteristics such as marital status, education, employment status, age, and others. These characteristics will be described below in the empirical model.

Empirical Analysis

Descriptive Analysis

As indicated in the introduction, our first goal is to examine how time allocated to children may have changed over the period under study. The idea is to disentangle changes in parent’s allocation of time as reported by Hofferth and Sandberg (2001), while controlling for children’s age (time-intensive versus goods-intensive children), type of time (active versus passive time) and certain parental characteristics. As pointed out above, the distinction between both sources of heterogeneity has not been considered jointly before. Table 1 reports the allocation of time per week, while Table 2 shows the allocation of time by characteristics. We have tested the null hypothesis that the differences in mean are zero. For all tests we consider up to a 10% significance level, although almost all the main results will still hold at a 5% significance level.

Table 1 Mothers’ time allocation (% of weekly time)
Full size table
Table 2 Mothers’ time allocation (% of weekly time)
Full size table

We first found that mothers devoted more time to children than to market work activities in 1997 (about 5%, p value = 0.01) and that in 1981 they devoted the same amount of time. Results showed that the distribution of total time with children is skewed to passive time; particularly in goods-intensive children (all p values lower than 0.05). Secondly, there was an increase of about 6% (p value = 0.00) in total time with children from 1981 to 1997. These results are in line with Hofferth and Sandberg (2001), although, unlike this paper, they reported hours and not percentages. A new finding is that behind this increase, we observed an increase of about 4% (p value 0.00) in active time and an increase of about 2% (p value = 0.08) in passive time. Moreover, we found that the increase had mostly occurred when considering time-intensive children (about 5%, p value = 0.06).

As in Hofferth and Sanberg (2001), we showed that working mothers devoted less time to children than non-working mothers (about 14% in 1981 and about 21% in 1997, p values = 0.00 for both). A new finding is that behind that increase in the negative effect of maternal labor force participation, we observed that non-working mothers had increased children’s time by about 7% (p value = 0.02) chiefly through passive time, while working mothers had, at least, maintained time allocation to children. Behind this negative effect of mother’s employment status, we would like to point out that working mothers increased active time with time-intensive children by about 5% (p value = 0.01), while non-working mothers increased passive time with goods-intensive children by about 7% (p value = 0.05).

In contrast to the notion that single-parent households have a negative effect on time allocated to children, we found that single mothers devoted around 6% (p value = 0.010) more time than married mothers in 1981, but found no differences among mothers in 1997. Behind these changes, we observed that both married and singled mothers devoted more active time with time-intensive children in 1981 than in 1997 (about 5%, p value = 0.01 for married mothers and 6%, p value = 0.06 for single mothers). Again, as in the related literature, we showed that mothers with a university degree increased time with children over the 1981–1997 period (about 7% with a p value of 0.01). This increase was due to active time spent with time-intensive children, which has increased by about 9% (p value = 0.00). A more detailed description of all uses of time by mother’s characteristics can be found in Moro-Egido (2005).

To summarize, total time with children has increased as a result of the increase in active time with time-intensive children. The distinction between types of time and children’s age provides new evidence on how demographic changes have affected allocation of time. We find evidence against the idea that employment status has a negative effect on time with children. Single mothers behave similarly to married mothers in terms of passive time and have reduced the gap in terms of active time. Mothers’ education displays a positive effect mainly in terms of active time, which increases over time.

In the light of these findings, we can affirm that active time with time-intensive children is the principal force behind aggregate changes. This suggests that mothers may have changed their behavior to emphasize primary and presumably higher-quality activities to protect investment in children, despite the fact that there are a large and increasing number of substitutes in the market for childcare activities that have allowed mothers either to work more or to enjoy greater leisure time.

As regards mother’s working status, our findings also support the explanation provided by Bianchi (2000) who claims that maternal time is overestimated when considering non-working mothers. We find that the distribution of time with children is more skewed to passive time in the case of non-working mothers than in the case of working mothers. In Table 2 we observe that the ratio of active to passive time is always greater for working mothers.

In the case of education, our results support Bianchi’s (2000) argument. It could be that highly educated parents (regardless of other variables that affect social class) spend more time with children (in particular, active time) because they are better informed about the benefits of investing in children. The existence of a skill-premium, that is, the fact that the return of education in terms of wages increases with educational level implying that the opportunity cost of time for high-educated workers is larger; could explain the fact that college-educated mothers increase active or high quality time instead of passive time. Since the opportunity cost of time is larger for college-educated mothers, they might devote more time to children through more productive time.

Multivariate Analyses

In order to explain these findings in terms of the allocation of time, we estimate an empirical model to examine the main determinants of such changes. We use the behavioral model presented above to set the empirical specification. From the theoretical model, we can consider the following set of behavioral equations based on time-demand equations:

$$ t_{j} = \beta_{0} + \beta_{1} t_{i \ne j} + \beta_{2} {\mathbf{X}}_{j} + \varepsilon_{j} \quad {\text{for }}j = c,\,h,\,f $$

where t c is time with children, t h is working time, t f is fathers’ total time with children, and X j is a vector of explanatory variables for each of the endogenous variables that include wage rate, other household income and some standard demographic characteristics. The corresponding β:s are unknown parameters. Before continuing, we would like to point out that we do not consider father’s active and passive time separately. The idea is to compare our results to those obtained in the related literature, such as Klevmarken and Stafford (1997) who consider only total time parents spend with children and find that father’s and mothers’ time are complementary. Maume (2011) finds evidence that during the period 1977–1997 there has been an increase in fathers’ time with children, which was three times larger on non-workdays than workdays.

Our estimations are obtained via a three-stage least-squares (3SLS) procedure. The time-diary data literature debates whether the use of censored regression models (Tobit) or linear models (OLS) are more appropriate in explaining the determinants of time spent on different activities. Given the large numbers of zeros typically found in time-diary data, the Tobit estimation would be appropriate. However, the number of zeros could represent a measurement error rather than non-participation in the activity, in which case OLS would be preferred (for a detailed discussion see Kalenkoski and Foster 2008). There are two problems that arise in the econometric specification: the identification problem and the endogeneity problem.

To deal with the identification problem, we need at least one unique identifying instrument for each of the endogenous variables. Thus, we should include at least one exogenous variable in each equation that is not in the other equations. Due to our data set is rich in time uses, but not so rich in certain other dimensions, there are not so many variables to consider as instruments. Taking this into account and following the related literature, we use age and age squared, father’s non-labor income, and mother’s wage squared as instruments of mother’s total time with children, father’s total children time and mother’s market work time respectively. When we split mother’s total children time into active and passive time, we use age and age squared and the number of children older than 13 respectively.

As regards the endogeneity problem, we are aware that many variables are endogenously determined, such as leisure, marital status, fathers’ income, and others. For example, in Kalenkoski et al. (2007), marital status is endogenized to consider the possibility that some unobserved characteristics may affect not only time spent on childcare but also marital status.

Given the nature of our data set, and following Hallberg and Klevmarken (2003), we choose to consider as endogenous only those variables presented previously, that is t c , t h and t f . In the empirical analysis we name these variables as Total Children Time to measure the percentage of time that a mother devotes to children per week; Market Work Time, to measure the percentage of time that a mother devotes to market work per week and Fathers’ Children Time, that is the percentage of time that a father devotes to children per week respectively. When we consider different types of time with children, we define the variable Active Time as the percentage of active time with children per week; and Passive Time, as the percentage of passive time with children per week.

Some discussion on explanatory variables included in vector X j is now in order. We include variable Leisure, that measures the percentage of time that a mother devotes to leisure per week, and variable Fathers’ Household Time, that is the percentage of time that a father devotes to household production per week. This variable was denoted by (t f ) in the model. We have done this firstly because our data are percentages and once work time and children’s time are determined, the rest is considered here as leisure time; and secondly, fathers’ allocation of time is fairly constant across time and does not appear to react with mothers’ characteristics (see Kooreman and Kapteyn 1987).

We include variable Wage Rate, that measures mothers’ net wage per hour; and the variable Other Income, that considers any income in the household different from mothers’ labor income, to ascertain the net empirical effect on allocation of time to children since the theoretical sign of the effects is unclear. By including variables reflecting mother’s level of education (Mother’s Education), we are able to clarify whether or not this variable produces its own effect, and not only the indirect effect by means of wages. Equivalently, we include father’s level of education (Fathers’ Education). Finally, we include a group of variables to cover other demographic characteristics. Young Children is a dummy that takes the value of 1 if there is at least one child in the household younger than 6 years old to reflect the different effect of having children that are time-intensive or goods-intensive. Single is a dummy that takes the value of 1 if the mother is unmarried to reflect that it is likely that single mothers differ from married mothers in their behavior. Finally, for the case of pooled data we include a dummy variable (Dumyear) to control for the effect of the year under consideration.

When dealing with female labor participation, it is important to note that a change in an exogenous variable could imply that the female decides not to work (a selection effect) and a change in the amount of hours devoted to work. The selection effect would not arise if the female’s labor participation in the labor market were completely random and independent from time with children. The theory states that females compare an offered wage (expected marginal product) and their own reservation wage; and that difference is reflected by an unobserved latent variable L*. However, what we observe is whether or not the female participates in the labor market, which is modeled by a dummy variable L that takes the value of 1 if the female is working. Consequently, we use the following reduced-form model:

$$ L = \left\{ {\begin{array}{ll} 0 &\,\,\, {{\text{if }}L^{ * } \le 0} \\ 1 &\,\,\, {\text{otherwise}} \\ \end{array} } \right. $$

\( {\text{where}}\;L^{ * } = \gamma^{\prime } \bf{Z} + \eta\, \) ,and Z is a vector of exogenous variables explaining the female’s decision on labor market participation. All random errors are assumed to be multivariate normal. There is no a priori assumption of zero correlation. Therefore, we will estimate this model by means of a two-step procedure: we first estimate the participation equation and then the equations explaining allocation of time. We define the variable Labout Market Participation that is a dummy that is activated if the mother is working. To deal with the identification problem, we include a variable reflecting the number of children younger than 6 years of age (Under 6) in Z. In an alternative specification for the second step, we will separately consider an equation for active time and another one for passive time. In order to ascertain whether these determinants could possibly have changed their influence over time, we have estimated the specifications by year and also with the pooled data. The main statistics and definitions for the demographic variables we will consider are reported in Table 3.

Table 3 Descriptive statistics
Full size table

Estimation Results

For all cases in the selection equation representing female labor participation, we obtain the expected results, namely that the more educated, the older and the fewer number of younger children the mother has, the more likely she is to work. In Table 4 we report the estimated coefficients for the system of equations concerning allocation of time.

Table 4 3SLS estimates of total time with children
Full size table

We first find that decisions on market work time (Market Work Time) and children’s total time (Total Children Time) are interdependent. The estimated parameter is negative, thus reflecting the production substitution effect. Moreover, time with children is less prone to change than market work time for any change in the exogenous variables. For example, in the pooled data, 1% more of market work time or leisure time (Leisure) decreases by around 0.058 or 0.059 (respectively) times the percentage of total children’s time. In contrast, an increase of 1% of total children time or leisure time will decrease by around 17.25 and 1 (respectively) times the percentage of market work time. The lower sensitivity displayed by decisions on children’s time could be interpreted as mothers protecting time with children and being able to enjoy a smaller amount of leisure in exchange for market work time, but not lower time with children. Moreover, we observe that total children time sensitiveness decreased over time. We have tested whether the effects of some exogenous variables have changed over time. In this section we include only discussions on changes that are statistically significant.

The second finding concerns the interdependence between parents’ time allocation. Variations in mothers’ market work time or time with children have a positive effect on fathers’ time with children (Fathers’ Children Time) and vice versa. The sensitiveness of mothers’ market work time is larger than that of total children’s time. The positive effect of one parent’s time with children on the other parent’s time with children; this is a common result in the literature (Hallberg and Klevmarken 2003; Klevmarken and Stafford 1997). This positive effect does not necessarily imply that they spend all these hours together, although such an interpretation is likely. These findings also provide evidence in favor of the hypothesis outlined by Bianchi (2000) in that females’ status has increased fathers’ involvement and has also resulted in free time for mothers to work more. Finally, fathers’ household production time (Fathers’ household time) has no effect on any other allocation of time.

The distinction between active and passive time provides some new effects that are important to highlight. The results are reported in Table 5. First, although active time and passive time interact negatively with mothers’ market work time in 1981, passive time was more sensitive. In 1997, however, active time was not found to interact with market work time. This suggests that mothers start considering both times differently: mothers now protect active time more than passive time than previously. Secondly, there was no relationship between active and passive time in 1981. In 1997, however, the relationship is positive; an interaction that works in both directions and is of the same magnitude. Third, fathers’ time with children interacts positively with passive time both years, while the negative interaction with active time appears only in 1997.

Table 5 3SLS estimates of types of time with children
Full size table

Mothers’ wage rate (Wage rate) displays no effect on mothers’ allocation of time in 1981. The influence of wage rate is different in 1997. First, an increase in wage rate implies a decrease in market work time. Second, wage rate has a positive effect on active time, while a negative effect on passive time. This can be interpreted as a production substitution effect, where mothers, perceiving active time to be more productive, substitute more active time for passive time. Fathers increase time with children when mothers’ wage rate increases since the relative opportunity cost of fathers’ time decreases. Non-labor female income (Other Income) displays no effect on mothers’ market work time, and if any, it is negative as expected.

Parents’ education displays a specific effect aside from the effect of wage increases. In 1981, mothers’ education (Mothers’ Education) negatively affects active time and market work time (this last effect being of larger magnitude). However, in 1997, mothers’ education only affects mothers’ decisions on how much to work. This finding does not support the hypothesis found in the literature that parents with a higher level of education could have preferences in favor of higher quality children. Fathers’ education (Fathers’ Education) only has a negative significant effect in 1997 on fathers’ time with children. The number of children under 6 years of age (Young Children) displays a positive effect through both types of time in 1981, but only through active time in 1997. Being single (Single) implies more market work time and total children time. As shown by the negative estimated effect, the selection process is relevant only through active time in 1981.

Conclusions

The analysis of the data for the US in 1981 and 1997 reveals a main finding, which mothers have increased the time they devote to their children, mainly through active time with time-intensive children. This finding provides evidence regarding a change in mothers’ behavior to avoid variations in children’s time and/or that there has been a change in what a child needs. Besides, both parents encourage each other to devote time to children; and the time fathers spend with children frees time for mothers to work. These behavioral changes might have been influenced either by public policies, or by changes in social attitudes, or by both phenomena.

There has been a change in social norms and expectations concerning not only what parenting involves, but also what children need. A culture of intensive mothering seems to have emerged. Mothers feel that they need to be experts on child development and spend ever-increasing time interacting with their children. Due to a lack of time, however, they must ensure that the time spent with children is quality time and juggle family duties by including children in their own leisure and free-time activities. Moreover, free time has been trimmed increasingly as multi-task; which is the idea behind passive time. These changes have had an effect on families’ consumption patterns insofar as they demand different leisure and home-production goods (e.g. going to the cinema with children instead of going to the opera).

Additionally, greater concern is now shown for child development, together with an unfixed concept of childhood. Due to an explosion of activities that a child is perceived to need, there has been a rise in the demand for goods that were not previously needed. We observe that more preschool-age children spend time outside the home in school-like settings regardless of their mothers’ employment status. At the same time, older children need time and monetary investment for an extended number of years as more attend college. There may simply be a greater “preference” for spending time on child care.

Investing in institutional and parental child care has usually been proposed as a policy to promote female labor participation. Since external child care is often focused on supervisory and low-contact care, parents can thus devote their scarce time to developmental activities and high-contact activities. Furthermore, as shown in Bernal and Keane (2006), informal child care (child care provided by relatives, friends, etc.…) considerably reduces children’s scores; while formal child care programs (public or private day care facilities or pre-school centers) increase these scores. Thus, policies that encourage female labor participation should also promote formal child care institutions for all working mothers, particularly single mothers who are the most likely users of informal care. Child care subsidy policies should also be directed at providing access to high-quality child care. In related literature, we find evidence that there is an inverse relationship between the costs or availability of care services and female labor market decisions (Berry et al. 2008; Borra and Palma 2009; Forry 2009). We also find that among low-income parents (particularly single parents) child care plays a critical role in employment, although being unable to afford or access high-quality care can serve as a barrier to employment (Forry 2009).

More recently, policies have focused on enhancing children’s development and reducing work-family stress among parents. Policies that help families to reconcile paid work and family responsibilities might, in principle, allow mothers and fathers to share household tasks and time with children more equitably, while at the same time facilitating participation in the work force. From our findings, we know that these effects do partially occur. We find that whenever one of the parents devotes more time to children, the other also does so, although this does not necessarily imply that they spend all these hours together. We also know that fathers’ involvement creates free time for mothers to work rather than to devote to children.

A closely related policy measure is to increase the flexibility of working hours. There is some evidence that many organizations having “family-friendly” policies actually remain fairly gendered. Such programs are mostly for mothers, and fathers often feel less entitled to use parental leave and other family-friendly options, even when these are provided as a matter of public policy. Public policy related to reconciling family and work duties and flexibility of working hours should therefore continue to insist on greater involvement by fathers.

Policy areas that might benefit from our findings, and from future analyses of determinants of time with children, include day care subsidies, curriculum development or teaching practices targeted at children from different backgrounds, subsidies for in-kind support for working parents, encouraging fathers’ involvement, measures for reconciling paid work, or making working hours more flexible, among others.

As with all research, there are caveats. One of them is data limitations, in the sense, that time use surveys is really rich in time use information, but not so rich in certain other dimensions. Therefore the analysis of how active and passive times evolve with some economic determinants could be improved in several directions. Another natural extension to this research is to consider a theoretical model with more complex individual decisions, which could improve the empirical strategy as well. However, we face again the limitation of the data availability, that is, there are few number of covariates apart from time uses. Another possible extension to this research would be to consider a larger period of analysis, since there are Time Use Surveys for the period 2003–2007. One difficulty could be to assure the comparability of all these surveys.