Mother’s late return home from work, family relationship, and locus of control of children: evidence from Japanese adolescents

Abstract

While previous studies have examined the link between maternal employment status and child development, the results remain inconclusive, and the underlying mechanisms are not yet well understood. A potential explanation for the mixed findings is the omission of mothers’ return home time from work, a factor that has yet to be tested in the literature. To address this gap, this study examines the relationship between mothers’ time of returning home and their children’s locus of control using a nationwide child–parent survey in Japan. The results of the entropy balancing method demonstrate that the daughters of mothers who return after 7 p.m. are more likely to believe that they lack control over their life outcomes, whereas this effect is not observed for mothers who return home by 7 p.m. This relationship is mediated by the deterioration of family relationships. Consistent with prior research, the negative association is more pronounced in households with higher socioeconomic status, while it is mitigated when fathers return home early or when children cohabit with their grandparents, highlighting the importance of caregiving by all family members. Given the increasing number of married women in full-time and managerial positions and the diffusion of teleworking, these findings are relevant for policymakers.

1. Introduction

Female labor force participation, particularly in full-time and managerial positions, has seen significant growth in OECD countries. The female full-time equivalent employment rate rose from 51 per cent in 2014 to 55 per cent in 2019 (OECD n.d.). Similarly, recent data show an increase in the participation rate of married women in these countries, climbing from 66 per cent in 2010 to 71 per cent in 2019 (ILO 2022a). These trends indicate a growing number of married women engaged in diverse work arrangements, including long working hours, flexible time, and night shifts.

Despite the positive impact of married women’s employment on household income, previous studies have demonstrated its negative impact on family relationships (e.g. divorce and poor parenting quality) and child development (e.g. cognitive achievement and behavioral problems), particularly for the children of mothers engaging in a job with nonstandard work schedules, such as night shifts (Bolino et al., 2021). Han (2005) and Han and Fox (2011) show a negative association between a mother’s nonstandard work schedule and her child’s cognitive performance in the US. Studies have also demonstrated an association with behavioral problems (Han 2008; Kaiser et al., 2019; Rönkä et al., 2017), where the effects are mediated by a deterioration of parenting quality and family relationships (Strazdins et al., 2006; Han et al., 2010).

Although insightful, existing studies leave three issues unaddressed. First, there is no consensus regarding causality, given that some studies provide counterevidence. Chase-Lansdale et al. (2003) find that maternal employment has an insignificant impact on child outcomes. If anything, it has beneficial effects on the mental health and behavioral problems of early teens. Dockery et al. (2016) also find a weak association between a nonstandard work schedule and a child’s mental/physical health. Bianchi and Milkie (2010), Bolino et al. (2021), and Lucas-Thompson et al. (2010) argue that these mixed results can be partly attributed to heterogeneity in working conditions and household composition. However, few studies have explored the conditions that trigger negative consequences.

Second, although previous studies have investigated the impact on children’s health conditions, cognitive achievement, and behavioral problems, the impact on noncognitive skills is not well understood. Analyzing noncognitive skills is relevant, given their pivotal role in achieving socioeconomic success and the importance of family relationships in childhood to develop such skills (Borghans et al., 2008). Furthermore, examining this impact may enable researchers to uncover the mechanism of the impact of maternal employment on children’s cognitive and health outcomes, given the effects on these outcomes of noncognitive skills.

Third, previous studies have mainly used samples of young children from Western countries, raising a potential issue regarding the external validity of their findings. Further studies analyzing older children and children from non-Western countries with different cultural backgrounds are required.

This study bridges these gaps in the literature by examining the association between a mother’s late return home from work and her child’s locus of control in junior and senior high school in Japan. We also examine the heterogeneity in the association between the child’s gender and household characteristics, as well as the underlying mechanisms of the association, such as the deterioration of family relationships. Locus of control is a type of noncognitive skill that refers to individuals’ beliefs about the causal relationship between their own efforts and their life outcomes (Rotter 1966, 1990). The literature shows that those who believe that they can control their life outcomes achieve higher earnings (Heineck and Anger 2010), human capital investment (Findley and Cooper 1983; Chiteji 2010), and subjective well-being (Verme 2009). The psychological literature predicts that the impact of maternal employment relies on whether the mother returns home early and has enough time to spend with her children or not (Carton and Nowicki 1994). This is crucial because previous studies on nonstandard work schedules consider mothers who work long hours under a standard work schedule as the control group, leading to an underestimation of the actual impact. Nonetheless, to the best of our knowledge, no previous study has examined the impact of time returning home.

It is particularly relevant to analyze the case of Japan for three reasons. First, existing studies rely on evidence from the USA and Europe, while evidence from a society with strong traditional gender roles, such as East Asian countries, is still scarce.¹ In Japan, mothers are perceived as the primary caregivers of their children, and mealtime is seen as a time for communication. Second, the maternal labor force participation in this country has increased by 8.9 percentage points from 61.7 per cent in 2010 to 70.6 per cent in 2019 (ILO 2022b). Third, while the Japanese government has encouraged female labor force participation, government support for accelerating maternal employment, such as childcare programs, remains insufficient. This is exacerbated by the fact that men in Japan work longer hours than men in other OECD countries (OECD 2021a), and it has been less common for married couples to reside with their parents. Furthermore, the use of nannies is not a common means of parenting support in Japan from an early age, with only 1.8 per cent of households using a babysitter for their first child by age three as of 2015–8 (IPSS 2021). Although after-school programs such as cram schools are widely available for Japanese junior and senior high school students, these programs do not provide after-school care after 7 p.m. for more than half of them (MHLW 2014). Therefore, the impact of mothers’ late return home from work on children may be particularly severe in Japan and empirical evidence is required.

2. Background

2.1. Definition and determinants of child locus of control

According to Rotter (1966), locus of control refers to a generalized attitude, belief, or expectancy regarding the nature of the causal relationship between an individual’s own behavior and its consequences. This belief is termed “internal control” if individuals perceive that an outcome of their behavior is contingent on their own behavior or personal characteristics. Conversely, it is labeled “external control” if they interpret the outcome as not entirely contingent on their actions but rather determined by chance, luck, fate, or powerful others.

Internal control beliefs are critical to achieving socioeconomic success in life. Those with internal control beliefs may exert more effort than those with external control beliefs. Earlier studies provide evidence that internal control belief is significantly associated with individuals’ earnings (Heineck and Anger 2010), human capital investment (Findley and Cooper 1983; Chiteji 2010), and subjective well-being (Verme 2009).

The significant impact of locus of control raises an important question regarding its formation process. Social learning theory, as proposed by Rotter (1966), suggests that belief in internal control develops through the experience of perceiving that an outcome is contingent on the individual’s behavior. Once such a belief is established, the outcome reinforces the belief that the particular behavior or event will be followed by the outcome in the future. By contrast, the failure of the outcome to occur weakens the belief. In particular, experiences involving powerful external forces can influence individuals’ perceptions of external control. Locus of control changes over time as these life experiences accumulate.

In line with this theory, studies have empirically demonstrated three major determinants of child locus of control. First, socioeconomic adversity, such as poverty and disaster experiences, is associated with external control belief (Stephens and Delys 1973; Culpin et al., 2015; Shoji 2023). Shoji (2023) finds that experiencing exogenous income shocks in childhood has negative impacts on adult locus of control, which persists for decades. In addition, living in a wealthy neighborhood could positively affect the child’s locus of control (Catterson and Hunter 2010; Ahlin 2014; Ahlin and Lobo Antunes 2015).

Second, parenting styles and family relationships are important determinants of a child’s locus of control. Studies have found that children are more likely to perceive internal control if their parents are warm, emotionally supportive, accepting, and nurturant. Furthermore, consistent parental discipline and reward are also associated with children’s internal control (Ahlin and Lobo Antunes 2015). By contrast, they grow up to believe in external control if their parents are overprotective, and controlling, and frequently use physical punishment (Spokas and Heimberg 2009; Ahlin 2014).

Third, parents’ locus of control may be intergenerationally transmitted to children, although, to date, there is no consensus in the literature regarding this possibility. On the one hand, some studies find positive spillover effects, while the magnitude of the effect varies with the gender of parents and children (Tully et al., 2016). On the other, many studies demonstrate weak or no association (Hoffman and Levy-Shiff 1994; Morton 1997).

2.2. Potential influence of a mother’s late return home from work on child locus of control

Given the arguments in the previous subsection, we consider four channels through which a mother’s late return home from work affects the child’s locus of control in both positive and negative ways: increase in household income, deterioration of the family relationship, change in mother’s locus of control, and decrease in children’s free time.

First, if a mother’s late return is attributed to longer working hours, it may lead to an increase in household income. This may cause children to develop an internal control belief if the increased income is used for the child’s allowance or education expenditure. Higher allowances improve children’s sense of free choice over their daily spending and activities, while children with access to better educational resources have more opportunities to perceive that they can change their future through their efforts. However, the impact of income gain may be small for wealthy households or when parents spend their income on other items, such as parents’ private consumption.

Second, mothers who return home late might not be able to spend enough time or use good parenting methods for their children (Johnson et al., 2013; Kaiser et al., 2019). For example, they may have fewer opportunities to talk to their children about problems and concerns at school; some mothers may become more anxious and micromanage them, while others may become more neglectful. Subsequently, the deterioration in parenting quality and family relationships may negatively affect the child’s locus of control.

Third, mothers’ late return from work may have positive or negative effects on their own locus of control, which may spill over to their children’s locus of control. On the one hand, there is the negative effect of having little free time available for themselves, making them feel they have little control over their own lives (Castillo et al., 2020). Even if they do manage their free time, they may be accused by family members, such as husbands or parents (in-laws), of not spending time on household chores or child-rearing. Such third-party pressure or deviation from a social norm could result in a lower locus of control of mothers, particularly in Japan, where the norm of gender roles is strong. Subsequently, the deterioration in the mother’s locus of control could lead to low locus of control in children. On the other hand, working late may also positively affect mothers’ locus of control if it leads to greater responsibility at work and better career prospects, thus empowering such mothers. Therefore, their children could exhibit a higher locus of control.

Fourth, children of mothers who return home late may have to perform household chores. Subsequently, the reduction of children’s free time may lead them to feel that they have little control over their lives.

These four channels are not mutually exclusive, but there is a trade-off among them. Specifically, the advantages of working late, such as income gains and positive effects on a mother’s locus of control, depend on the cost to the family relationship. In addition, children, especially daughters, may have their free time reduced when they do housework on behalf of mothers. Mitigating these opportunity costs requires that mothers spend most of their time between income-earning activities, household chores, and parenting, which in turn has negative effects on mothers’ locus of control.

Intriguingly, these arguments suggest heterogeneous impacts on child locus of control by household socioeconomic status and composition. On the one hand, in households with higher socioeconomic status, the positive impact of the mother’s late return through income gain may be marginal compared to the negative impacts, such as poor parenting quality and more amounts of children’s time spent on household chores. On the other, the effects of income gain may be critical and exceed the negative effects among lower socioeconomic status households. Evidence from previous studies supports this conjecture (Bianchi and Milkie 2010; Lucas-Thompson et al., 2010). Similarly, the negative impacts may be mitigated in households in which other adult members, such as the father and grandparents, are available to spend time with children.

However, it is theoretically unclear whether the households with higher socioeconomic status suffer from severe negative impacts on children’s locus of control. On the one hand, wealthy parents may allocate resources to compensate for the decline in parenting quality (e.g. housekeeping service). On the other hand, they may not take such a measure if they are unaware of the negative effects of returning home late on their children. Furthermore, such measures may not mitigate the negative effects if parents and housekeeping services are not substitutable for children.

3. Method

3.1. Survey design

This study uses a nationwide parent–child survey conducted by The Institute of Social Science of the University of Tokyo and Benesse Educational Research and Development Institute of Education. This survey was designed to collect information on elementary, junior high, and senior high school students (between 6 and 17 years) and their parents from all 47 prefectures in Japan. This feature is crucial because it enables us to collect accurate data, such as the mothers’ and children’s locus of control and the socioeconomic conditions of households, including education expenditure. Although prior studies have analyzed children’s outcomes evaluated by their parents (Nes et al., 2014), parents who work late may be less able to accurately assess their children, causing the estimation results to be biased.

The survey was conducted every year between 2015 and 2018, and the sample size varies across years between approximately 14,000 and 17,000 child–parent pairs. These respondents were selected from a large database that covers more than one-half of the child population in Japan.² The representativeness of the sample is carefully tested and discussed by Kimura (2020) and Okabe (2020). They conclude that the sample selection problem is unlikely to be serious in this dataset.

The questionnaire for child survey includes questions on children’s academic outcomes, motivation for learning, noncognitive skills, relationship with family members and other individuals, attitude to career and study, and daily habits.³ The parent survey covers the questions for socioeconomic conditions, parenting attitude, attitude to the surveyed child’s education and career, and relationship with the surveyed child. The child and a parent separately complete the survey, except for children in or under grade three, whose parents provide the answers on their behalf.

The questions for locus of control were only asked of junior and senior high school students who participated in the 2017 survey (6,497 observations). Excluding the single-parent households from the sample to control for the difference in demographic and socioeconomic conditions, 5,994 observations remain. We also restrict the sample to the cases in which the respondent of the parent survey is the mother (5,409 observations). Finally, after excluding the samples with missing values, the final sample size is 4,757 children and their mothers. The missing values are found mainly in the variable of the father’s time of returning home.

3.2. Measures

3.2.1. Child locus of control

A challenge in measuring child locus of control is the trade-off between the (internal) reliability of the scale and the representativeness of the sample. To obtain a scale with high reliability, children are requested to answer many similar questions, as undertaken by Connell (1985) and Nowicki and Strickland (1973). However, the respondents may refuse to answer all the questions or may not answer them seriously, if they are asked too many and overly complicated questions. This issue is particularly crucial for multipurpose surveys—which consist of various modules and contain many questions—such as this survey. To address this issue, some studies measure locus of control using fewer questions. For example, Cobb-Clark and Schurer (2013) employ the Household, Income, and Labor Dynamics in Australia, which includes seven questions. Cebi (2007) and Verme (2009) use scales derived from four and one questions, respectively.

Given these arguments, our measure of child locus of control is generated using the following three items: 1. “What happens in my life is my responsibility,” 2. “I can do most things if I try,” and 3. “I think I can handle most things.” Similar types of questions are used frequently in the literature (Coleman and DeLeire 2003; Cobb-Clark et al., 2014; Caliendo et al., 2015; Abay et al., 2017; Heywood et al., 2017). The answer options for these items employ a Likert scale from 1 (strongly disagree) to 4 (strongly agree).

To generate a composite index of locus of control from these items, we perform principal component analysis. Supplementary Table A1 in Supplementary Appendix A presents the results. Considering the eigenvalues, their confidence intervals, and the scree plot, we keep the first component (eigenvalue = 1.41), which explains 47 per cent of the variation in the original three items. The second item demonstrates the highest factor loading (0.756), followed by the third (0.714) and first (0.573) items. The obtained composite index is standardized to have a mean zero and a standard deviation one. We test the robustness to the usage of alternative measures in Section 5 and Supplementary Appendix B.

3.2.2. Mother’s time of returning home from work

In the parent survey, respondents are asked about both parents’ time of returning home from work on a typical day. The answer options are in a categorical format, including “before 3 p.m.,” “between 3 p.m. and 5 p.m.,” “between 5 p.m. and 7 p.m.,” “between 7 p.m. and 9 p.m.,” “between 9 p.m. and 11 p.m.,” and “after 11 p.m.” We construct a binary variable that takes unity if the mother returns home from work after 7 p.m. (a proxy for a typical dinner time), and zero otherwise. Nonworking mothers are included in the control group, as they are expected to be at home by 7 p.m.. In Supplementary Appendix B, we drop the sample of nonworking mothers and test the other thresholds to compare the effect size.

3.2.3. Family relationship

This study uses two types of measures for the family relationship. The first measure directly elicits the extent to which children are satisfied with the relationship with their family. The answer options are based on a Likert scale ranging from (1) not satisfied at all to (4) very satisfied. The second measure is drawn from the child survey, which contains fourteen questions on children’s perceptions of their relationship with their parents, as listed in Supplementary Table A2. The answer options use a Likert scale ranging from (1) strongly disagree to (4) strongly agree. We use these items to conduct exploratory factor analysis and retain three factors: emotionally supportive (eigenvalue = 4.72), providing guidance for study (1.85), and equal relationship (1.24). These factors explain 48 per cent, 19 per cent, and 13 per cent of the variance in the data. Supplementary Appendix A2 discusses the methodology and results of the factor analysis in detail.

3.3. Empirical strategy: entropy balancing method

A potential issue in estimating the impact of having a mother returning home late from work on the child locus of control is sample selection. Confounders, such as parents’ educational background and family composition, could determine both the mother’s working conditions and the child’s locus of control.⁴ To address this issue, we use the entropy balancing method proposed by Hainmueller (2012), given the unavailability of a suitable instrumental variable.⁵ This method estimates a weighted least squares model in which the weight is computed to balance the observed characteristics between mothers who return home late and those who do not.

Specifically, we estimate the following weighted least square model:

where, $Yip$ denotes the locus of control of child i in prefecture p. $Dip$ is a binary indicator that takes unity if the mother returns home from work after 7 p.m., and zero otherwise. $Xip$ includes the covariates listed in controls in Table 1. Finally, $δp$ denotes the prefecture dummies and $Xip$ is an error term. A parameter of interest is $β1$⁠.

In this model, the observations are weighted by 1 for children whose mothers return home from work after 7 p.m., that is, $Dip=1$⁠. The weight for those whose mothers are at home by the time (⁠$Dip=0$⁠) is computed by the entropy balancing method, so that the first and second moments of observed characteristics, $Xip$ and $δp$⁠, are balanced between them. This estimator provides a consistent estimate of $β1$⁠, given the assumption of selection-on-observables, that is, no omitted variable bias conditional on $Xip$ and $δp$⁠. We employ this method because it is preferred over the other methods frequently used for impact evaluation, such as the propensity score method.

We also extend the model by adding the interaction term between the child’s gender and $Dip$ (Section 4.2). Furthermore, to uncover the underlying mechanisms of association, we reestimate this model using different dependent variables, such as the mother’s locus of control and a composite index of family relationships (Section 4.3). Finally, we split the sample based on household characteristics to examine the heterogeneity of association (Section 4.4).

A remaining issue in this method is the violation of the selection-on-observables assumption, that is, unobserved heterogeneity. We test the robustness to this issue in the next section and Supplementary Appendix B.

4. Results

4.1. Summary statistics

Columns (1) to (4) in Table 1 show the summary statistics of used variables by mothers’ time of returning home from work. The first three rows present the three items that construct the child locus of control. The fourth row is the standardized composite index of the child’s locus of control. Children whose mothers return home after 7 p.m. exhibit 0.07 standard deviation lower locus of control than the other children. The rest of the variables are parents and child characteristics. Columns (5) to (8) present the summary statistics of the weighted sample. The table shows that the parents’ characteristics significantly differ from the mother’s time of returning home, such as the birth year, permanent job, and educational attainment. However, after adjusting the sampling weight, they are balanced except for the birth years of the parents.

4.2. Association between the mother’s time of returning home and the child’s locus of control

4.2.1. Main results

Column (1) of Table 2 presents the results of the entropy balancing method. It shows that a mother’s returning home from work after 7 p.m. is associated with a 0.115 standard deviation decrease in child locus of control compared to that of children whose mothers are at home by 7 p.m. Column (2), which reports the OLS results for comparison, shows qualitatively the same results, but the point estimate is larger in Column (1).

Columns (3) and (4) of Table 2 examine the heterogeneity across child gender by incorporating the interaction terms with the indicator of “boy” and “girl.” Both columns show significant differences between sons and daughters. In Column (3), mothers returning home from work after 7 p.m. is significantly associated with a decrease in their daughters’ locus of control by 0.149 standard deviations, while the association for sons is smaller in magnitude and statistically insignificant. Again, using OLS leads to an underestimation of the coefficient.

4.2.2. Robustness

Supplementary Appendix B shows five tests that we conduct for the validity of our empirical strategies, namely, (B1) exclusion of nonworking mothers from the sample; (B2) alternative measures of child locus of control; (B3) additionally controlling for twenty variables on parents’ subjective attitude to child education; (B4) falsification test using alternative measures of the mother’s time of returning home; and (B5) the coefficient stability test of Cinelli and Hazlett (2020) to test the robustness to the violation of the selection-on-observables assumption. These tests support the validity of our strategy.

4.3. Tests for underlying mechanisms

Given the arguments in Section 2.2, three out of four channels are in line with the negative association between mothers’ late returning home time and daughters’ locus of control: deterioration of mothers’ locus of control, decrease in children’s free time, and deterioration of the family relationship. This subsection tests the relevance of these channels.

4.3.1. Deterioration of the mother’s locus of control

Our parent survey contains two out of three questions that are used to elicit child locus of control: “I can do most things if I try” and “What happens in life is the responsibility of the person.” Using these measures as proxies for the mother’s locus of control, we examine the association between the mother’s time of returning home and her locus of control based on the specification in Equation (1). As the mother’s locus of control is not child-specific but a household-level outcome, the estimation model does not include the interaction terms with the child’s gender. Supplementary Table C1 of Supplementary Appendix C shows that the coefficients are small in magnitude and statistically insignificant, ruling out this channel.

4.3.2. Decrease in children’s free time

Mothers’ late returning home time may increase their children’s time spent on household chores and decrease their free time. These may drive a negative association with the locus of control. To test this conjecture, we examine children’s time spent on nine activities, including TV/DVDs, games, and music using Equation (1). Supplementary Table C2 of Supplementary Appendix C shows no significant correlations for daughters in most columns. These results are contrary to the conjecture.

4.3.3. Deterioration of the family relationship

Mothers returning home late from work may negatively impact the family relationship and parenting quality (Kaiser et al., 2019), which may lead to their daughters’ low locus of control. To test this possibility, this subsection examines whether children, particularly daughters, of mothers who return home after 7 p.m. have poorer perceptions of the family relationship. Table 3 presents the estimation results. Column (1) shows that there is a negative and significant association between mothers’ time of returning home from work and their daughters’ satisfaction with their family relationships. Regarding the three dependent variables derived from the exploratory factor analysis, mothers’ time of returning home is negatively associated with daughters’ perception of emotional support from parents, the factor that explains approximately half of the variance (Column (2)). Table 3 also reports the false discovery rate q-values (Anderson 2008) to adjust the P-values of the four coefficients of each outcome, confirming a robust association. By contrast, the coefficients for the interaction terms with sons are smaller in magnitude and statistically insignificant.

To explore the plausibility of this channel more carefully, we additionally control for the indicators of perceptions about family relationships and emotional support in the model of child locus of control in Equation (1). If these perceptions are a major channel, we should observe significantly positive coefficients of perception variables and a smaller coefficient of mothers’ late returning home time than that for the results in Column (3) of Table 2.⁶Table 4 provides supporting evidence. Both perception variables are positively and significantly associated with the child’s locus of control. Furthermore, the coefficient of mothers’ late returning home time for daughters is 10 per cent smaller than the results of Table 2. These findings suggest that deterioration in family relationships, especially lack of emotional support from parents, maybe one of the channels for daughters’ poor locus of control.

4.3.4. Increase in household income

Although the baseline results in Table 2 show a negative association between the mother’s time of returning home and the child’s locus of control, this does not necessarily rule out any positive impacts through income gain. Mothers’ late return home time may lead to an increase in resources spent on children, improving their locus of control. To investigate this effect, we estimate the association between a mother’s time of returning home and the outcomes—such as annual household income, education expenditure on the child, an indicator for private tutoring, and the monthly allowance amount for the child—using Equation (1).

Supplementary Table C3 of Supplementary Appendix C provides counterevidence. While we find a positive association between mothers’ time of returning home and household income (Column (1)), it is uncorrelated with expenditure on their daughters (Columns (2) to (4)). These results confirm that the effects through income gain are unlikely to be critical in our sample.

4.4. Roles of family support and socioeconomic status

The argument in Section 2.2 suggests there is a particularly more severe negative impact on children from households with higher socioeconomic status and no support from other adult members. To test this possibility, this section conducts subsample analyses for children who cohabit with grandparents, fathers’ time of returning home from work, and the educational background of parents.

Table 5 presents the results, which are in line with our conjectures. Columns (1) to (4) demonstrate that the coefficients of the interaction terms for daughters are significantly negative only among households with no adult support available. The negative association of a mother’s late returning home time disappears if the father returns home before 7 p.m. or if the child cohabits with grandparents. Furthermore, Column (2) shows that not only daughters but also sons suffer from lower locus of control if both parents return home after 7 p.m. These findings suggest that the existence of a primary caretaker, regardless of who in the family it is, is important for an adolescent’s locus of control.

Similarly, Columns (5) and (6) of Table 5 show that mothers’ late return home time is negatively associated with daughters’ locus of control only among households where parents have a high educational background. This is in line with the arguments of previous studies (Bianchi and Milkie 2010; Lucas-Thompson et al., 2010).

5. Discussion

Using a nationwide parent–child survey in Japan, this study investigates the association between mothers’ late return home from work and the locus of control of their adolescent children. The results show that children, particularly daughters, whose mothers return home from work after 7 p.m. are more likely to believe that they cannot control their life outcomes than are children whose mothers are at home by 7 p.m. This difference is explained by the time of returning home and not by the difference in the mothers’ employment status. We also provide suggestive evidence that this association is driven by the deterioration of family relationships. Finally, as the theory predicts, the negative association is larger among households with higher socioeconomic status and without other adult members actively engaging with children.

Intriguingly, the gendered impact has not been observed in previous studies on maternal employment in Western countries, although it is in line with those of previous studies on the gender difference in the locus of control (Dyal 1984; Fisman and O’Neill 2009). However, we should not interpret this result to be the consequence of strong gender norms in Japan, because further exploration shows that mothers’ late return from work does not increase daughters’ time allocation for household chores.

Our findings are most closely related to that of Johnson et al. (2013), who examine the impact of parents’ working hours in Australia. However, this study differs from theirs in several aspects. First, they investigate children’s behavioral problems reported by their parents, while we use a self-reported measure of the child’s locus of control. Second, their study finds that fathers’ long working hours lead to their sons’ (but not daughters’) behavioral problems, while that of mothers is uncorrelated with children’s behavior. This study suggests that the late return of both parents affects daughters’ locus of control. Third, they analyze children aged between 5 and 10 years in Australia, while our sample consists of those aged between 12 and 17 years in Japan. Therefore, these results and the present study should be considered as complementary.

This study also makes two contributions to the psychological literature. First, previous studies have demonstrated significant gender gaps in adult outcomes, such as labor market outcomes and academic performance (Altonji and Blank 1999). Recent studies explain these gaps by differences in noncognitive skills (Semykina and Linz 2007; Fisman and O’Neill 2009; Churchill et al., 2020). However, the sources of gender differences in noncognitive skills are not well understood (Falk and Hermle 2018). This study contributes to the relevant literature by uncovering the drivers of gender differences in locus of control. Second, to the best of our knowledge, this is the first study to examine the association between mothers’ time of returning home from work and child locus of control and its underlying mechanisms.

However, we should exercise caution in interpreting our results because they hinge on the validity of our data and identification strategy. First, the usage of observational data is common in the literature on maternal employment. Hence, we cannot fully rule out the possibility of unobserved confounders, although we conduct various robustness checks. Second, the validity of our locus of control measure has not been verified in the literature, although it is commonly employed in studies that use multipurpose surveys. Third, our dataset does not contain detailed information about the work schedule of each respondent including the total working hours. Therefore, this study cannot distinguish whether the observed association is attributed to long working hours or late return home from work. Future research is required to disentangle these mechanisms. Fourth, the analysis of single-parent households may also be important because the trade-off is potentially more severe than that of double-parent households. However, our dataset does not allow us to explore this question due to the small sample size of single-parent households. Finally, this study fails to identify the mechanism of association between a mother’s late return home from work and her daughter’s perceived family relationship, given the data limitation. It would be ideal to analyze more detailed data on the working conditions and time use of parents and children for further analysis of the mechanisms.

6. Conclusion

Our results suggest that adolescents, not only preschool children, benefit from having quality time with their families to develop a locus of control. Furthermore, fathers and grandparents can substitute for mothers to provide emotional support to children. From these findings, the following policy implications can be derived. First, parents who return home late because of long working hours may benefit from policy interventions to reduce working hours. Common policy measures in OECD countries include regulations on working hours and the introduction of remote work (OECD 2021b). Teleworking enables parents to reduce their commuting time and reallocate it to family time. Meanwhile, it should be kept in mind that such solutions may not apply equally to society as a whole, as teleworking opportunities continue to be unevenly distributed by employment status, income, and education (Araki 2023).

Second, for mothers engaging in an evening shift or night shift job, combining multiple types of work schedules (i.e. rotating shifts) may be effective. Similarly, a four-day workweek, a policy measure that has emerged recently, may also be effective (Messenger 2018; Pastore et al., 2019). However, earlier evidence suggests that flexible working conditions may increase only mothers’ time with their children, but not fathers’ (ILO 2021), highlighting a gender disparity in unequal time allocation between mothers and fathers. Therefore, it may also be important to encourage fathers’ work-life balance.

Third, the literature suggests that parents’ mental stress from work could trigger the use of negative parenting methods, such as physical punishment and inconsistent or less kindly enforced discipline (Pinderhughes et al., 2000; Oburu and Palmérus 2003). These arguments suggest that potential solutions involve mental health support for working parents.

Funding

This work was supported by the Japan Society for the Promotion of Science Grants-in-Aid for Scientific Research Grant Number [23K01411]. We thank the reviewers for their time spent on reviewing our manuscript, careful reading and insightful comments and suggestions that lead to improve the quality of this manuscript.

Conflict of interest statement. None declared.

References

Footnotes