The Impact of Work-Life Balance Policies on the Time Allocation of Japanese Couples Miki Kohara∗ and Bipasha Maity† October 27, 2017

Abstract We analyse the impact of work-life balance policies enacted by the government of Japan on the share of time allocated by Japanese women vis-`a-vis their husbands to paid employment and home production on a typical working day. Using panel data and employing random and fixed effects estimations to control for unobserved individual heterogeneity, we find that these policies have had a somewhat limited success in altering cultural norms about the gender division of labour in Japanese households. In particular, we find that these policies increased the wife’s share in the couple’s time devoted to paid employment by only 2.4 percent relative to the mean. Further, the share of time spent by the wife in the couple’s time spent in home production has remained largely unaffected on account of these policies. Alternatively, instrumenting for the provision of leave policy at the individual’s firm using information on past leave provision at the level of the industry, yields similar results. We find that the increased share of one’s time in a typical working day devoted by married women to paid employment on account of work-life balance policies is compensated by cutting down on the share of time devoted to leisure as the fraction of time spent performing domestic chores is found to remain largely unchanged.

Keywords: Labour market policies; paid employment; home production; women; panel data; Japan JEL Codes: J08; J13; J16; J22 ∗

Osaka School of International Public Policy, Osaka University. Address: 1-31 Machikaneyama, Toyonaka, Osaka, 560-0043, Japan. Email: [email protected]. † Ashoka University. Address: Plot No. 2, Rajiv Gandhi Education City, Sonepat (National Capital Region of Delhi), Haryana 131029, India. Email: [email protected]. We are grateful to the participants of the 31st Annual Conference of the European Society for Population Economics and the 1st annual meeting of the Society of Economics of the Household for many valuable comments and suggestions. All remaining errors are our own.

1

1

Introduction

Japan is a highly industrialized country with one of the highest standards of living in the world. However, there exists significant gender gap in labour market participation in Japan. According to the Organization for Economic Co-operation and Development (OECD) statistics, the labour market participation rate for men across OECD countries is around 80 percent; while the corresponding figure for women is around 61 percent for the period 2000-2015. On the other hand, men’s labour market participation rate is around 85 percent, while that of women is around 63 percent over the same time period in Japan. This indicates a gender gap in labour market participation of 22 percent for Japan, higher than that for all OECD countries which is around 19 percent during 20002015. The gender gap in labour market participation in Japan is roughly 10 percent higher than that for Australia, Canada, France, Germany, the United Kingdom and the United States. On the other hand, the time spent by men in home production such as childcare, caring for the elderly and all other routine household tasks is rather low in Japan relative to most advanced economies. The OECD Gender Portal reports the average number of minutes spent by men and women in different activities on a typical day (weekdays and weekends) that include paid jobs as well as unpaid work like routine housework and caring for household members. We find that women’s share of the time spent in performing unpaid work in the total time spent by women and men combined in these activities was around 85 percent in Japan, compared to 68 percent for all OECD countries. Men are found to spend nearly eight times as much time in paid work than in domestic chores in Japan, while men in all OECD countries together spend around 2.3 times as much time in paid jobs as in home production. Further, the employment rate for women is found to be the lowest between the ages of 25 and 44 among all working age groups in Japan. This age range coincides with childbearing and child raising for Japanese women (Ministry of Health, Labour and Welfare, Government of Japan). Therefore, women are found to exit the labour market on getting married or giving birth and at times returning back to the labour market around the age of 45. This also implies that, unlike for men, working full-time posed to be a challenging and almost impossible task for women that could be performed alongside raising children. The working environment in Japan was characterized by lack of work-life balance such as long hours of work, large incidents of overtime work, limited availability of childcare facilities in workplaces, limited availability and uptake of family care leave. Cultural norms about gender roles in the household which emphasized women’s role in performing household tasks, childcare and providing care to the elderly; while emphasizing men to be the sole earners of income in

2

a household as well as the aforementioned characteristics of workplaces have potentially contributed to the gender gap in labour market participation and engagement in domestic chores as well as childcare in Japan. One of the earliest legislations that aimed at prohibiting gender discrimination in the labour market in Japan was the Equal Employment Opportunity Law (EEOL), which was enacted in 1986. The EEOL prohibited gender discrimination during recruitment, promotion and retirement as well as worker dismissal on account of personal life events like marriage, pregnancy, childbirth and uptake of maternity leave entitlements. The EEOL, however, had no provisions that could help make workplaces women-friendly through generous provision of family care leave system, limiting long hours of work and overtime work. Importantly, the EEOL did not have any provision that could encourage men to participate in routine household tasks and childcare. Abe (2011) finds that the EEOL has not led to an increase in regular full-time employment for Japanese women. The absence of work-life balance has been associated with Japanese women facing a trade-off between participating in paid employment and raising children, if they wished to. Cultural norms about gender roles imply that men are, often, not faced with a similar trade-off. The absence of work-life balance has also been attributed to one of the leading causes of Japan’s declining birth rate. The period from the late 1990s is marked by the enactment of a number of legislations by the government of Japan with the aim of increasing women’s participation in paid employment, men’s involvement in childcare and raising Japan’s birth rate. These policies taken together are referred to as the work-life balance policies. Unlike in other industrialized countries, work-life balance policies in Japan were not promoted to achieve balance between labour and leisure time; but especially with the purpose of balancing time allocation between husbands and wives. These policies include obligation to establish family care leave system that include childcare and care leave systems to help workers who are taking care of children as well as their elderly parents, limitations on late-night work and overtime work as well as measures to shorten working hours especially for workers who are responsible for taking care of children aged three or less, prohibition of discrimination against workers who take leave in order to promote workers to take up these paid leaves, expansion of family care leave system to part-time employees, extension of the childcare leave period as well as establishment of injured or sick child care leave system. Most of these policies were mandated to be implemented by large-sized firms, especially firms with more than 300 employees and smaller firms were required to take steps to implement these measures. Important policy measures taken by the government to increase men’s involvement in childcare include extending the childcare leave period to until the child is one year and 3

two months old if both parents take leave, men being allowed to take childcare leave again even after they have taken eight weeks leave during childbirth as well as abolishing prohibition on men from taking childcare leave if they had non-working wives. This paper focuses on two major work-life balance policies, namely the provision of childcare leave and care leave systems in firms and studies their impact on primarily time allocation between couples in households. We focus on the sample of currently working couples who are not engaged in self-employment. We use panel data from the Japanese Panel Survey of Consumers (JPSC) for our analysis. The information on work-life balance policies at the wife’s firm is available for the period 1996 until 2013; while that in the husband’s firm is available for the period 1996 until 2002. The JPSC provides information on a rich set of demographic controls as well as other employment characteristics of couples. Using random and fixed effects estimation models that plausibly account for individual unobserved heterogeneity, year fixed effects that control for overall macroeconomic trends and other demographic variables that could plausibly influence time allocation between paid employment and home production among couples, we find that work-life balance policies in the wife’s firm increase the share of wife’s time in the couple’s time spent in paid employment on a typical working day by 2.4 percent relative to the mean. However, work-life balance policies do not appear to significantly alter the share of time spent by the wife in home production in the couple’s time performing these tasks on a typical working day during the period of our analysis. Alternatively, we also instrument for the presence of work-life balance policies in the wife’s firm with the past prevalence of work-life balance policies at the industry level to which the respondent’s firm belongs. Our results from the instrumental variable estimation strategy are qualitatively similar to those from the random and fixed effects estimation models. Therefore, these policies do not appear to be successful in raising men’s share of time in domestic chores and childcare. This paper also studies the effect of work-life balance policies on the willingness to have a (or another) child for working women, controlling for the number of children the woman already has. We find that work-life balance policies at the wife’s firm reduces the unwillingness to have a child by nearly six percent. This analysis is motivated by the fact that women in various countries and particularly in Japan are often faced with a trade-off between the decisions to have a career and a family. This is in contrast to men who often do not face such a trade-off in decision making. Therefore, this paper has attempted to study whether work-life balance policies were able to reduce the necessity for a woman to trade-off participating in paid employment if she wished to have a child. The existing literature has largely studied the impact of leave policies on labour 4

market outcomes of mothers in European countries, Canada and the United States. Many of these studies have studied the impact of parental leave policies on taking up of leaves by new parents, mothers’ post-birth return to the labour market as well as the effect of the duration of the leaves on fertility and job continuity of women (Ondrich, Spiess, Yang, & Wagner, 2003; Lalive & Zweimuller, 2009; Rossin-Slater, Ruhm & Waldfogel, 2013; Sch¨onberg & Ludsteck, 2014; Rossin-Slater, 2018). On the other hand, a growing literature has also focused on the effect of family leave policies on the uptake of leave by fathers, working hours of fathers, men’s involvement in childcare and housework in the aforementioned countries. While some studies indicate limited impact of childcare leave policies for fathers on their involvement with housework and childcare (Ekberg, Eriksson & Friebel, 2013; Kluve & Tamm, 2013); other studies have indicated that paternity leave entitlements are associated with greater involvement of men in childcare, sharing of domestic tasks among couples, likelihood that only the father takes leave so that the mother does not suffer from career interruption as well as both husband and wife taking leave among dual-earner couples and the ability of very young children to relate to their fathers (Nepomnyaschy & Waldfogel, 2007; Tanaka & Waldfogel, 2007; Almqvist, 2014; Bartel, Rossin-Slater, Ruhm, Stearns & Waldfogel, 2016; B¨ unning & Pollmann-Schult, 2016; Patnaik, 2016). However, most of the studies in the context of Japan have focused instead on the impact of work-life balance policies on outcomes of firms that implement these policies. In particular, the existing literature has found that work-life balance policies increase a firm’s sales and profit (Abe & Kurosawa, 2006) and raise the total factor productivity marginally for some firms in the long run (Yamamoto & Matsuura, 2011). Other studies have analysed the initial gender composition of the firm’s employees when firms implement maternity leave policies (Morita, 2005), proportion of female employees when firms provide work-life balance policy entitlements that are higher than those mandated by the law (Takeishi, 2006) as well as the effect of these policies on raising the proportion of women managers, turnover of female employees in Japanese firms (Hui-Yu & Takeuchi, 2009; Kato & Kodama, 2015; Yamaguchi, 2016). However to the best of our knowledge, whether work-life balance policies have been able to change the behaviour of husbands and wives in Japan in terms of their time allocation as well as women’s fertility preference after controlling for observed individual, family and firm characteristics of the couples as well as unobserved heterogeneity, has remained largely unexplored in the existing literature. In this regard, this paper is closer to the analysis of Foster and Stratton (2017) who use panel data to study whether labour market events (in their instance, promotions and terminations) influence the intra-household gender division of labour among married or 5

cohabiting mixed-gender couples in Australia. The main contribution of this paper has been to investigate the impact of work-life balance policies on time allocation between various activities within families. This is particularly important in the context of Japan as work-life balance policies were mostly implemented with a view to helping couples balance time between paid labour market activity and unpaid domestic chores, unlike in other industrialized economies. To the best of our knowledge, this is the first paper that has attempted to study the impact of worklife balance policies on time allocation among couples in Japan. As the government’s main focus was to bring about a change in time allocation among couples and thereby affect cultural norms about gender division of labour to some extent through these policies instead of how these policies can likely affect firms’ profitability, our paper contributes to the existing literature by analysing the extent to which these policies have been able to meet the aforementioned goals. Another significant contribution of the paper is using panel data for the purpose of our analysis. This paper uses the Japanese Panel Survey of Consumers, a longitudinal dataset that contains rich source of information on the demographic, educational and labour market conditions of women and their husbands as well as detailed information on their children and parents between 1993 and 2013. Further, this survey records low attrition. Therefore, this dataset is unique in the context of Japan as it largely tracks the same couple over time. Further, using panel data for analysis also helps in controlling for potential unobserved heterogeneity, which could otherwise, influence our outcome variables and thereby make it difficult to infer whether changes in the outcome variables are plausibly on account of the change in policy. As far as we are aware of, this is the first paper that uses longitudinal data from Japan to study the impact of work-life balance policies on time allocation among couples as well as fertility preference of Japanese women. This paper is organized as follows: Section 2 describes the institutional background; Section 3 describes the data used; Section 4 explains the empirical strategy used in this analysis; Section 5 presents the results and Section 6 concludes.

2

Institutional Background

Work-life balance policies in Japan are aimed at providing equal employment opportunities to men and women. This is largely because of two main reasons. On one hand, Japan faces low birthrate and rapidly ageing population. This implies that the working age population in Japan is on the decline. On the other hand, social norms dictate that men would primarily be engaged in the labour market and women in domestic work. In 6

order to raise the pool of workers, it was essential to encourage women’s participation in the labour market. However given social norms about gender roles, the government of Japan enacted a number of work-life balance policies in order to make labour market conditions conducive to women’s participation and encourage men’s participation in domestic chores. The earliest legislation aimed at preventing gender discrimination in recruitment, promotion and retirement was the Equal Employment Opportunity Law (EEOL) enacted in 1986. However, the EEOL had no provisions that could help make workplaces womenfriendly through generous provision of family care leave system, limiting long hours of work and overtime work. A major change in work-life balance policies was brought about in 2003 with the enactment of the Act on Advancement of Measures to Support Raising Next-Generation Children. The Act required firms with more than 300 employees to establish policies for reducing working hours and offering more generous leaves for caring for one’s children and elderly parents. Further, it required that these firms should encourage their workers to take up these paid leaves. The Act also prohibited discrimination against female workers. While it was mandatory for firms with more than 300 employees to take the aforementioned measures, firms with less than 300 employees were required to make attempts to fulfil the requirements of the Act. Another major change in work-life balance policies in Japan occurred in 2007 on account of the establishment of the Charter for Work Life Balance Policies. The charter declared specific action policies, and set numerical targets to be reached by 2020. These targets included that the proportion of workers working for more than 60 hours per week should be reduced to 5 percent (to be reduced to 10 percent by 2010); taking up paid leaves should be raised up to 70 percent (to be raised upto 47.4 percent by 2010); the proportion of female workers continuing to work after giving birth to the first child should be increased to 55 percent (to be raised to 38 percent by 2010); and the proportion of male workers taking parental leave should be increased to 13 percent (to be raised to 1.23 percent by 2010). The charter also stated that firms should allow workers to choose flexible working hours more easily, especially to encourage male workers with children younger than six years old to increase the time devoted to domestic chores. Therefore, work-life balance policies in Japan were mostly targeted to help balance time allocation between labour market work and domestic chores between husband and wife. This is in contrast to most other OECD countries, where work-life balance policies mostly aim at helping individuals balance time spent working and in leisure. This provides motivation to study the impact of work-life balance policies on the intra-household gender division of labour in Japan. 7

3

Data

The data used for our analysis come from the Japanese Panel Survey of Consumers (JPSC) conducted by the Institute for Research on Household Economics. This is a panel survey that initially surveyed women who were aged 24-34 years in 1993 with the survey being conducted for each year after that to track these women over time. Over the years, the survey has added women aged between 24 and 29 years to maintain the representativeness of the sample. We use the Waves 1-21 (corresponding to years 1993 until 2013) of the survey for our analysis. Further, attrition is reported to be low for the JPSC data. The JPSC data provide rich information on the employment characteristics of the women as well as those of their husbands, educational attainments of the wife and the husband, socio-demographic characteristics of the respondent’s household (such as family size, the ages of the family members, their relationship with the respondent, residence in large cities), time allocation of the couple to various activities, willingness to have children and so on. This dataset is unique as it provides detailed information of the couple and tracks the woman (and, therefore, her household) over time. To the best of our knowledge, there is no analogous micro-level survey in Japan that is comparable to the JPSC in this regard. The work-life balance policies that we consider are the provision of childcare leave and parental care leave policies in the respondent’s firm. We create a dummy variable that assumes the value 1 if the respondent’s firm has either or both of these policies and is 0 otherwise in the survey year. We do not focus on the provision of limiting unpaid overtime work as anecdotal evidence suggests that limiting overtime work is largely not adhered to on account of cultural norms about the employer-employee relationship in Japan. We also analyse the impact of work-life balance policies prevalent in the respondent’s husband’s firm on the respondent’s and her husband’s time allocation between paid employment and home production. For this purpose, we include only those husbands who are currently employed, not in self-employment and have been working for the same firm throughout the survey period. For our analysis, we restrict our sample to include only married women who are currently in the labour market and have been working for the same firm throughout the survey years. This is because we are primarily interested to study the time allocation of currently married women. Also, we want to abstract away from the possibility of selective entry and exit into the labour market or selectively choosing one’s employer on the basis of work-life balance policies 1 . Further, the purpose of work-life balance 1

There is a possibility that some individuals drop out of our sample due to leaving the labour market

8

policies is largely to ensure continuity in the labour market for women. Therefore, we focus on the impact of these policies on the intensive margin (that is, time allocation of currently working women) as opposed to the extensive margin (that is, whether or not women choose to enter the labour force). We also exclude those who are self-employed. This is because the role of work-life balance policies is difficult to analyse for those who are self-employed. Further as births outside marriage are rare in Japan, the impact of work-life balance policies on the willingness to have children are largely relevant for currently married women. However, we wish to emphasize that we do not focus heavily on willingness to have children for women and instead focus on women’s time allocation. This is largely because our sample contains currently married women, many of whom are likely to have already made their fertility decisions. An ideal sample to consider for studying the impact of work-life balance policies on fertility preferences would be a sample of younger, unmarried women who are yet to take their fertility decisions. This is a topic we wish to pursue in depth in the future. The first set of outcome variables that we consider are the share of wife’s time in the couple’s time spent in paid employment and home production (that includes domestic chores as well as childcare, parental care) on a typical working day. These outcome variables are meant to explicitly capture the time allocation among couples to various activities on a typical working day. Other outcome variables that we consider are the share of the wife’s and husband’s time spent in their respective working days on paid employment, home production and other activities (such as leisure, personal care). These outcome variables can provide some indication of the time allocation of the agents in their respective working days. We consider time allocation in a working day as opposed to a day off as the time constraints imposed by the necessity to report for work do not exist on a day off, thereby enabling anyone to flexibly allocate their time to home production on a day off. In particular, a husband’s involvement in home production activities can help the wife allocate more time to paid employment only on a working day. Therefore, increased involvement of husbands in domestic chores and childcare on a day off is unlikely to have any positive influence on the time allocated to paid employment for women as time spent in paid employment is only relevant for a typical working day. In this regard, it may be important to mention that our time-use outcome variables are not diary-based but based on recall. Now, measurement error in our time-use outcome variables could be a plausible concern as they have been recorded on the basis of recall method. However, as we are and re-enter our sample on account of re-entering the labour force after a few years. This might result in underestimation of the impact of work-life balance policies. However, this is unlikely to be of much concern as it is likely to provide a lower bound on the impact of these policies on intra-household gender division of labour.

9

largely interested in examining how the “burden” of different activities fall on the wife vis-`a-vis the husband as measured by the share of time spent by the wife in performing the activity in the couple’s time instead of absolute magnitude of the time spent by each of the individuals, the issue of measurement error in time use variables is unlikely to be a cause of serious concern. Further, fixed effects, random effects and instrumental variable estimation specifications are used, which can plausibly account for measurement error in our time use outcome variables (we provide details of our estimation strategy in Section 4). We also consider the willingness to have a (or another) child for women in the sample as an indicator of fertility preference among women. Table 1 provides descriptive statistics of our key outcome variables. Panel A of Table 1 here presents the summary statistics for the estimation sample; whereas Panel B presents the summary statistics for the entire sample for the purpose of comparison. We find that the average share of time devoted by the wife in paid employment in the couple’s time spent in paid employment on a typical working day is around 41 percent; whereas the analogous value for the entire sample is 22 percent. Further, on an average, 30 percent of a typical working day is spent by the wife in paid employment in the estimation sample (which is around 7.2 hours); while the corresponding figure for the entire sample is 16 percent (which is around 3.8 hours). This is not surprising as the entire sample contains women who are not currently employed as well as women who are engaged in self-employment and women who have changed their employers. Interestingly, there does not appear to be large differences in the share of wife’s time in the couple’s time spent in home production on a typical working day between the estimation sample and the entire sample. Specifically, wife’s share in the couple’s time spent in home production is 90 percent for the estimation sample; whereas it is 92 percent on an average for the sample as a whole. Also, the average share of a wife’s typical working day spent in home production is 18 percent for the estimation sample (which is around 4.3 hours) while the corresponding figure for the entire sample is 29 percent (which is around 6.9 hours). The share of wife’s time on a typical day spent doing other activities for the estimation sample and entire sample are 50 percent and 54 percent respectively. We find that, on an average, 44 percent of the husband’s working day is spent in paid employment and two percent in home production in the estimation sample. The corresponding values for the entire sample do not appear to be significantly different. This is not surprising as the labour market participation of men is nearly universal in contrast to that for women. The share of a typical working day devoted to other activities is 50 percent for men in the estimation sample and 51 percent for the entire sample. 51 percent of wives in the estimation sample are unwilling to have a (or another child). In 10

Table 1: Descriptive Statistics of Outcome Variables Panel A: Sample for Analysis Variable

Mean

SD

Obs

0.41 0.90

0.14 0.16

8,192 8,026

0.30 0.18 0.50

0.10 0.12 0.10

8,209 8,209 8,209

0.44 0.02 0.50

0.09 0.04 0.09

18,043 18,043 18,043

0.51

0.50

8,398

Mean

SD

Obs

0.22 0.92

0.22 0.15

24,836 24,654

0.16 0.29 0.54

0.16 0.18 0.14

25,059 25,059 25,059

0.43 0.02 0.51

0.09 0.04 0.14

24,947 24,947 24,792

0.42

0.49

25,756

Share of Wife’s Time in Couple’s Time on a Working Day in: Paid Employment Home Production Share of Wife’s Time on a Working Day in: Paid Employment Home Production Other Activities Share of Husband’s Time on a Working Day in: Paid Employment Home Production Other Activities For Women: Unwillingness to Have a Child Panel B: Entire Sample Variable Share of Wife’s Time in Couple’s Time on a Working Day in: Paid Employment Home Production Share of Wife’s Time on a Working Day in: Paid Employment Home Production Other Activities Share of Husband’s Time on a Working Day in: Paid Employment Home Production Other Activities For Women: Unwillingness to Have a Child

Note: Data source is JPSC (1996-2013). Panel A of Table 1 is restricted to include currently married women who are employed (not self-employed) and working for the same firm throughout the sample period (1993-2013) for the outcome variables corresponding to wife’s share in the couple’s time, proportion of wife’s time in her typical working day and unwillingness to have children. The sample is restricted to husbands who are employed (not self-employed) and working for the same firm throughout the sample period (1993-2013) for the outcome variables corresponding to share of husband’s time in a typical working day. The variable “unwillingness to have a child” is a dummy variable that assumes the value 1 if the variable description is true and is 0 otherwise.

11

the entire sample, we find that 42 percent wives are unwilling to have a (or another) child. Therefore, the unwillingness to have children does not appear to be remarkably different between the estimation and entire samples. Table 2: Descriptive Statistics of Explanatory Variables Variable

Mean

Childleave/Careleave in firm Firm Less than 100 Employees Works for Less than 43 Hours per Week Full-Time Employee Wife’s Age in Years Husband’s Age in Years Number of Children Number of Children Younger than 6 Lives in a Big City Wife is atleast college educated Husband is atleast college educated Wife’s/Husband’s Parents Live with Couple Childleave/Careleave in Husband’s Firm Husband’s Firm Less than 100 Employees Husband Works for Less than 43 Hours per Week Husband is a Full-Time Employee

0.49 0.45 0.73 0.43 38.31 40.70 1.70 0.30 0.83 0.35 0.37 0.36 0.37 0.38 0.16 0.75

Standard Deviation 0.50 0.50 0.44 0.50 6.54 7.66 1.02 0.67 0.37 0.47 0.48 0.48 0.48 0.48 0.37 0.43

Observations 8,398 8,398 8,398 8,398 8,398 8,398 8,398 8,398 8,398 8,398 8,398 8,398 7,488 18,508 18,508 18,508

Note: Data source is JPSC (1996-2013). Table 2 is restricted to include currently married women who are employed (not selfemployed) and working for the same firm throughout the sample period (1996-2013) for all variables except those pertaining to the husband’s employment characteristics. Summary statistics for husband’s employment characteristics are computed by restricting the sample to include currently married men who are employed (not self-employed) and working for the same firm throughout the sample period. The information on the availability of child/parental care leave policy at the husband’s firm is only available until 2002.

The JPSC data provide a number of socio-demographic variables that have been used as controls in our analysis. Table 2 presents the summary statistics of these variables. The information on the wife’s employment characteristics as well as child and parental care leave at the wife’s firm for the estimation sample are available from the years 1996 until 2013. For our estimation sample, we find that, on an average, 49 percent wives reported that their firm has a child or parental care leave system or both throughout the sample period. The information on the availability of child and parental care leave system at the husband’s firm is only available for the years 1996-2002. All other variables pertaining to the husband’s employment characteristics are available for the years 19962013. During the years 1996-2002, 37 percent of the husbands are found to work for firms that had some leave system. 45 percent of the currently employed wives who have not changed their employer are found to work in firms with less than 100 employees. On the other hand, 38 percent currently employed husbands who have not changed their employer are found to be working for firms with less than 100 employees. 73 percent wives in the estimation sample work for less than 43 hours per week; whereas only 16 percent husbands are found to work for less than 43 hours per week. This indicates the 12

overall higher working hours prevalent in Japan among men. On an average, 43 percent wives are full time workers. The corresponding figure for the husbands is 75 percent. All other demographic variables are available throughout the sample period (19932013). The average age of the wife and the husband in the estimation sample is found to be 38 years and 40 years respectively. The average number of children is found to be 1.7 and the proportion of households with children who are six years or younger is around 30 percent. 83 percent of the couples live in a large city. The proportion of wives who have completed at least college education is 35 percent; while 37 percent of the husbands have completed at least college education. Thus, on an average, large differences in educational attainment between husbands and wives are not found. Also, 36 percent of the women in the estimation sample reported that their/their husband’s parents lived in the same house/lot as the couple during the sample period.

4

Estimation Strategy

We estimate the following equation: yit = µi + θt + βW LBit + γXit + εit

(1)

Here, yit is the outcome variable for individual i in year t. Our sample is restricted to include only those individuals who were currently working in the survey year (but not in self-employment) and in the same firm throughout the survey period. The outcome variables that we consider are the share of wife’s time devoted to paid employment and home production in the couple’s time devoted to these activities on a typical working day. Other outcome variables we consider are the share of one’s typical working day devoted to paid employment, home production and other activities (including personal care, leisure) for both wife and husband separately. We also consider the unwillingness of having a (or another) child for women in the sample as an outcome variable as a proxy for fertility preference. W LBit is a dummy variable that assumes the value 1 if the individual i ’s firm has either childcare leave policy or parental care leave policy or both in year t ; and is 0 otherwise. This is our explanatory variable of interest and thus β is our coefficient of interest. Xit is the vector of controls used in our analysis. In particular, we control for demographic characteristics such as ages of the wife and the husband, the number of children in the household, whether the wife has at least college education, the husband has at least college education, if the couple lives in a large city, the wife’s/husband’s parents live in the same house/lot as the couple. Other control

13

variables include the employment characteristics of the wife. Specifically, we control for whether the wife is a full-time worker, works in a firm with less than 100 employees and works for less than 43 hours per week. We also include analogous controls for the husband’s employment characteristics along with the presence of leave (childcare and parental care) in his firm in some specifications. When our outcome variables are the shares of wife’s time in the couple’s time in various activities, the share of the wife’s time in her typical working day, the unwillingness to have children; our primary set of controls corresponding to labour market characteristics are that of the wife. We also control for her husband’s employment characteristics in other specifications. Here we further restrict our sample to include those wives whose husbands are currently employed (but not in self-employment). When our outcome variables are the share of the husband’s time in a typical working day performing various activities, we primarily control for his employment characteristics. In addition, we control for his wife’s employment characteristics in other specifications by restricting the sample to include only those husbands whose wives are currently employed (but not in self-employment). We control for the husband’s employment characteristics when our outcome variables correspond to those of the wife because we wish to understand whether the husband’s employment characteristics affect the outcome variables corresponding to the wife. θt are the year fixed effects that are included to control for overall macroeconomic trends that could likely influence time allocation to various activities as well as fertility preferences. µi are time-invariant individual/family fixed effects 2 . These are included to control for time-invariant factors at the level of the individual/families such as cultural norms about the household division of labour, fertility preferences. The panel nature of our dataset enables us to control for these fixed effects. εit is the regression disturbance term that is clustered at the level of the individual/family to account for serial correlation in the errors within families over time. Our primary estimation strategy is the fixed effects model as outlined in equation (1). However, we also estimate (1) without family fixed effects as well as by using the random effects model. We present the results from these models for the purpose of comparison of the estimates from these models to those obtained from the fixed effects model. In this context, we note that a limitation of our data is that the women respondents are answering about their husband’s time allocation. Therefore, it is possible that this gives rise to plausibly biased responses where women may be more likely to report that their 2 The unit of observation in the JPSC data is the woman. The data collect information on the respondent’s household (socio-demographic characteristics as well as employment characteristics of her husband) from the respondent. Therefore, individual fixed effects are analogous to family/household fixed effects in the JPSC data.

14

husbands spend more time working and less time in home production than they actually do. However, as time allocation variables are our outcome variables, measurement error in husband’s time allocation is unlikely to bias the estimates of β because there is not enough reason to believe that biased responses about one’s husband’s time allocation is systematically correlated with work-life balance policies in the respondent’s firm. Further, controlling for individual random and fixed effects can potentially control for unobserved individual characteristics that could be correlated with biased responses about one’s husband’s time allocation and the likelihood of giving biased responses are unlikely to vary over time. Alternatively, we also present results from instrumental variable (IV) estimation. Our estimation equations are as follows: W LBif ht = ηt + αZih + δXif ht + υif ht

(2)

yif ht = θt + β W\ LBif ht + γXif ht + εif ht

(3)

Here we instrument for the availability of work-life balance policies at the respondent i’s firm f in industry h in year t (denoted by W LBif ht ) by Zih , the fraction of firms in the respondent’s industry that had work-life balance policies before the start of the JPSC survey (that is, prior to 1993). Therefore, we use past prevalence of work-life balance policies at the industry level as an instrument for current work-life balance policies at the firm level of the respondent. Equation (2) is the first stage equation where we regress W LBif ht on the proposed instrument Zih . Equation (3) is the second-stage equation of the IV estimation. The IV estimation results on the impact of work-life balance policies on time allocation of couples as well as fertility preferences of women are used to investigate whether findings on the impact of work-life balance policies differ across alternative estimation strategies. A pertinent question that might arise is why we are not using a difference-in-difference estimation strategy to estimate the impact of work-life balance policies on time allocation of working women. We do not use the difference-in-difference estimation strategy in our analysis as the period of our analysis was marked by gradual implementation of worklife balance policies across firms over the years. Specifically, the period of 1996-2013 was marked by implementation of a number of policies by the government of Japan. Therefore, there is no clear cut-off year that we can take as the year of starting of the work-life balance policies. Also, we do not have information on leave policies in the husband’s firm for periods after 2002. Further, the implementation of work-life 15

Share of Wife's Firm with Some Leave Policies

.2

.3

Share of Firms .4 .5 .6

.7

By Firm Size

1997

1999

2001

2003

2005 Year

Less than 100 Employees

2007

2009

2011

2013

More than 100 Employees

Source: JPSC Data Note: Sample Consists of Currently Employed (Non-Self-Employed) Married Women

Figure 1: Availability of Work-Life Balance Policies by Firm Size balance policies was made mandatory for large firms (especially those with more than 300 employees). However, firms with less than 100 employees were required to attempt to implement these policies. Therefore, using firms with less than 100 employees or 300 employees as a control group would not be appropriate for the analysis. In fact, as Figure 1 here shows the period was marked by increase in the proportion of both large and small firms in the availability of work-life balance policies. However, our data provide us with a rich set of controls that enable us to control for observable characteristics that could likely influence our outcome variables of interest. Further, the panel structure of our data enables us to control for likely unobservables that could also affect the outcomes we study through the random and fixed effects estimation strategies. In addition to random and fixed effects models, we also conduct instrumental variable estimation. Therefore, we are able to account for several confounding factors that influence our outcomes as much as possible even though we do not pursue a difference-in-difference estimation strategy. It is because of these reasons we focus on the impact of work-life balance policies on time allocation and fertility preferences using fixed effects and random effects estimation models as well as instrumental variable estimation instead of a difference-in-difference estimation strategy.

5 5.1

Results Baseline Results

We present the baseline estimation results in Tables 3, 4, 5 and 6. Table 3 reports the results on the impact of the work-life balance policies on the share of wife’s time in the couple’s time in paid employment (Panel A) and home production (Panel B) 16

Table 3: Share of Wife’ Time in Couple’s Time on a Typical Working Day in: Panel A: Paid Employment Childleave/Careleave in Wife’s Firm

(1)

(2) 0.01** (0.004)

(3) RE 0.004 (0.004)

(4) RE 0.01*** (0.004)

(5) FE 0.004 (0.004)

(6) FE 0.01** (0.004)

0.004 (0.005)

Wife’s Firm Has Less Than 100 Employees

0.006 (0.005)

0.0005 (0.004)

0.005 (0.004)

-0.00002 (0.003)

0.005 (0.005)

-0.0004 (0.004)

Wife Works for Less Than 43 Hours Per Week

-0.04*** (0.005)

-0.04*** (0.005)

-0.03*** (0.005)

-0.02*** (0.004)

-0.02*** (0.006)

-0.02*** (0.005)

Wife is a Full-Time Worker

0.05*** (0.005)

0.05*** (0.005)

0.05*** (0.005)

0.04*** (0.005)

0.04*** (0.01)

0.03*** (0.01)

Husband is a Full-Time Worker

-0.02 (0.01)

-0.002 (0.01)

-0.02 (0.01)

Husband Works for Less Than 43 Hours Per Week

0.05*** (0.004)

0.03*** (0.004)

0.02*** (0.004)

Husband’s Firm Has Less Than 100 Employees

0.01*** (0.004)

0.01*** (0.005)

0.01 (0.005)

Observations Panel B: Home Production Childleave/Careleave in Wife’s Firm

8,192 (1)

7,331 (2) -0.02*** (0.01)

8,192 (3) RE -0.01** (0.005)

7,331 (4) RE -0.01** (0.005)

8,192 (5) FE -0.001 (0.01)

7,331 (6) FE 0.002 (0.01)

-0.02*** (0.01)

Wife’s Firm Has Less Than 100 Employees

0.002 (0.01)

0.01 (0.01)

0.01 (0.005)

0.02* (0.005)

0.004 (0.01)

0.005 (0.01)

Wife Works for Less Than 43 Hours Per Week

0.02** (0.01)

0.02** (0.01)

0.02** (0.01)

0.01** (0.01)

0.01 (0.01)

0.01 (0.01)

Wife is a Full-Time Worker

-0.04*** (0.01)

-0.04*** (0.01)

-0.03*** (0.01)

-0.03*** (0.01)

-0.01 (0.01)

-0.005 (0.001)

Husband is a Full-Time Worker

-0.03* (0.01)

-0.01 (0.01)

-0.01 (0.02)

Husband Works for Less Than 43 Hours Per Week

-0.05*** (0.01)

-0.03*** (0.005)

-0.02*** (0.01)

Husband’s Firm Has Less Than 100 Employees

-0.01 (0.01)

-0.005 (0.01)

0.0003 (0.01)

Observations Number of Children Parents Live with Couple Live in a Big City Ages of Wife & Husband Wife is Atleast College Educated Husband is Atleast College Educated Year Fixed Effects

8,026 Yes Yes Yes Yes Yes Yes Yes

7,190 Yes Yes Yes Yes Yes Yes Yes

8,026 Yes Yes Yes Yes Yes Yes Yes

7,190 Yes Yes Yes Yes Yes Yes Yes

8,026 Yes Yes Yes Yes Yes Yes Yes

7,190 Yes Yes Yes Yes Yes Yes Yes

Note: Data source is JPSC (1996-2013). The sample is restricted to married, currently employed (not self- employed) women working in the same firm throughout the period. Robust standard errors clustered at the individual/household level are reported in parentheses. ***, **,* indicate statistical significance at the 1%, 5% and 10% level of significance respectively. In columns (2),(4) and (6), the sample is further restricted to include only married, currently employed (not self- employed) women whose husbands are also currently employed (not self-employed). Columns (1) and (2) do not include household/individual fixed or random effects. Columns (3) and (4) present results from the random effects estimation model; Columns (5) and (6) present results from the fixed effects estimation model.

on a typical working day. Columns (1) and (2) report the results from the pooled OLS

17

model; whereas Columns (3) and (4) report the results from the random effects estimation model and Columns (5) and (6) from the fixed effects estimation model. All columns include demographic and wife’s other employment characteristics. In addition to these controls, husband’s employment characteristics are included in Columns (2),(4) and (6). All columns also include year fixed effects. As such Columns (4) and (6) are our preferred regression specifications as they include all possible controls including individual random and fixed effects respectively. We will focus on Columns (4) and (6) for the purpose of interpretation of our results. Column (2) in Panel A of Table 3 shows us that presence of work-life balance policies in the wife’s firm is associated with an increase in the share of time spent by the wife in the couple’s time in paid employment on a typical working day. However, as Column (2) does not account for possible unobserved individual heterogeneity, we focus on Columns (4) and (6) instead for the purpose of interpretation of results. From Column (4) (random effects model) we find that presence of work-life balance policies in the wife’s firm results in an increase in the share of the wife in the couple’s time spent in paid employment by 1 percentage point. Relative to the mean, this implies an increase in the share of time spent by the woman in the couple’s time in paid employment by 2.4 percent. We find analogous results (in terms of direction and magnitude) from the fixed effects model in Column (6). Column (2) in Panel B of Table 3 finds that the presence of work-life balance policies in the wife’s firm is associated with a decrease in the share of time spent by the wife in home production in the couple’s time in home production on a typical working day. As before, we focus on Columns (4) and (6) for the purpose of interpretation of our results as they account for possible unobserved individual heterogeneity unlike Column (2). Column (4) indicates that presence of work-life balance policies in the wife’s firm reduces the wife’s share in the couple’s time spent in home production by 1 percentage point from the random effects estimation model. Relative to the mean, this implies a decline in the share of the wife’s time in the couple’s time in home production by 1.1 percent. However, Column (6) which presents the results from the fixed effects estimation model shows us that work-life balance policies in the wife’s firm has no significant impact on the share of wife’s time in the couple’s time in home production. Therefore, it appears that although the share of wife’s time in paid employment has increased; the share of wife’s time in the couple’s time spent in home production on a typical working day has remained largely unaffected. We intend to explore how married couples allocate their own time on a typical working day between paid employment, home production and other activities (notably, leisure 18

and personal care). Table 4 here estimates the impact of work-life balance policies in the wife’s firm on the share of her typical working day spent in paid employment, home production and other activities. We report the coefficient on the presence of work-life balance policies in the wife’s firm for brevity. We report the estimations from the random and fixed effects models that include the full set of controls as in Columns (4) and (6) of Table 3 here. Table 4: Fraction of a Typical Working Day Spent by Wife in: Panel A: Paid Employment Childleave/Careleave in Wife’s Firm

(1) RE 0.01*** (0.003)

(2) FE 0.01*** (0.003)

Panel B: Home Production Childleave/Careleave in Wife’s Firm

(1) RE -0.001 (0.004)

(2) FE -0.005 (0.004)

Panel C: Other Activities (Leisure, Personal Care) Childleave/Careleave in Wife’s Firm

(1) RE -0.01*** (0.004)

(2) FE -0.003 (0.004)

Demographic Controls Wife’s Other Employment Characteristics Husband’s Employment Characteristics Year Fixed Effects Observations

Yes Yes Yes Yes 7,329

Yes Yes Yes Yes 7,329

Note: Data source is JPSC (1996-2013). Regression specification is analogous to Columns (4) and (6) of Table 3. For details on all other controls, regression specifications, sample restrictions and clustering of standard errors, please see table notes of Table 3.

Column (1) in Panel A of Table 4 shows that work-life balance policies in the wife’s firm is found to increase the share of time on a typical working day devoted by the wife to paid employment. The random effects model also shows that although work-life balance policies at the wife’s firm do not significantly influence the share of time on a typical working day the wife spends in home production, it is found to reduce the share she devotes to other activities such as leisure and personal care (Column (1) in Panels B and C). The random effects model shows us that the increase in the share of time spent in paid employment is exactly compensated by the decline in time spent in other activities on a typical working day. In particular, the share of time in one’s working day spent in paid employment is found to go up by 1 percentage point; whereas the share of time spent in pother activities is found to decline by the same magnitude. Relative to the mean, share of time spent in paid employment is found to increase by 3 percent, while 19

that spent in performing other activities is found to decline by 2 percent. Importantly, the share of the wife’s working day spent in home production is not found to be affected due to work-life balance policies in her firm. Column (2) of Table 4 estimates the impact of work-life balance policies in the wife’s firm on the share of time she spends in paid employment, home production and in other activities on a typical working day using the fixed effects estimation model. Panel A shows that work-life balance policies in the wife’s firm increase the share of a typical working day spent by the wife in paid employment; but it does not appear to significantly reduce the share devoted to home production or other activities (Column (2) in Panels B and C)3 . Specifically, the magnitude of the impact of work-life balance policies in the wife’s firm on the share of time spent on a typical working day spent by the wife in paid employment is analogous to that obtained from the random effects model in Column (1) of Panel A. We next turn to the husband’s time allocation on a typical working day and examine whether leave policies prevalent in his firm have any influence on the share of time on a typical working day he spends in paid employment, home production and other activities; controlling for potential factors that could influence this time allocation as well as unobserved individual heterogeneity in Table 5. Column (1) estimates the random effects model and Column (2) estimates the fixed effects model. The information on leave policies in the husband’s firm is only available for years between 1996 and 2002. The sample is restricted to include husbands who are currently employed (not self-employed), working in the same firm during the sample period and whose wives are also currently employed (not self-employed). As before, we only present the results from the random and fixed effects estimation models here. After controlling for potential factors, including individual heterogeneity, we find that leave policies in the husband’s firm have no significant impact on the share of time spent by the husband on a typical working day in paid employment, home production and other activities in both the random and fixed effects models (Columns (1) and (2) in Panels A, B and C of Table 5). Therefore, work-life balance policies do not appear to have any significant impact on men’s time allocation on a typical working day. These findings are interesting as they indicate that work-life balance policies have not been particularly effective in Japan during the period of our analysis in changing cultural norms about gender roles, specifically in the sphere of home production, which is largely believed to 3 However, if we add the coefficients in Column (2) across Panels B and C, the sum of the coefficients is approximately around -0.01. It is possible that inclusion of a large of individual fixed effects renders individual coefficients on the share of time spent in home production and other activities statistically insignificant.

20

Table 5: Fraction of a Typical Working Day Spent by Husband in: Panel A: Paid Employment Childleave/Careleave in Husband’s Firm

(1) RE 0.0004 (0.004)

(2) FE 0.004 (0.005)

Panel B: Home Production Childleave/Careleave in Husband’s Firm

(1) RE 0.002 (0.002)

(2) FE -0.001 (0.002)

Panel C: Other Activities (Leisure, Personal Care) Childleave/Careleave in Husband’s Firm

(1) RE -0.002 (0.004)

(2) FE -0.002 (0.005)

Demographic Controls Husband’s Other Employment Characteristics Wife’s Employment Characteristics (including leave policies at her firm) Year Fixed Effects Observations

Yes Yes Yes

Yes Yes Yes

Yes 2,626

Yes 2,626

Note: Data source is JPSC (1996-2002). The sample is restricted to married, currently employed (not self- employed) men working in the same firm throughout the period and whose wives are also currently employed (not self-employed). Robust standard errors clustered at the individual/household level are reported in parentheses. ***, **,* indicate statistical significance at the 1%, 5% and 10% level of significance respectively. Other controls included, but not shown are ages of wife and husband in years, dummies for whether the wife and husband have attained at least college education, whether the household lives in a large city, the number of children in the household and whether the couple’s parents live with them. Husband’s other employment characteristics include whether he works for less than 43 hours per week, if he is a full-time worker and if the firm he works for has less than 100 employees. Analogous variables are included as controls for wife’s employment characteristics.

be a woman’s task. Table 6 estimates the impact of work-life balance policies on the unwillingness to have a (or another child). We include the full set of controls (including existing number of children and the ages of the respondent and her husband). Columns (1) and (2) of Table 6 provides the estimates from the random and fixed effects models respectively. Panel A includes currently working women (that is, our sample for analysis). In addition, we wanted to investigate whether work-life balance policies in the husband’s firm has any effect on fertility preferences of non-working women. We present those results in Panel B of Table 6. In Panel A of Table 6 we find that work-life balance policies appear to decrease the unwillingness to have a (or another child) in both the random and fixed effects models (Columns (1) and (2)); however, the coefficient on the work-life balance policies is only statistically significant for the random effects model. In particular, work-life balance policies in the wife’s firm is found to reduce the unwillingness to have a (or another child) by 3 percentage points from the random effects model (Column (1) of Panel A). Relative to the sample mean, this signifies a 5.9 percent decline in the unwillingness to have a (or another child). 21

Table 6: Unwillingness to Have a (or Another) Child among Women Panel A: Currently Employed Women Childleave/Careleave in Wife’s Firm

(1) RE -0.03*** (0.01)

(2) FE -0.01 (0.01)

Demographic Controls Wife’s Other Employment Characteristics Husband’s Employment Characteristics Observations Panel B: Women who are Not Working Childleave/Careleave in Husband’s Firm

Yes Yes Yes 7,484 (1) RE 0.005 (0.01)

Yes Yes Yes 7,484 (2) FE 0.01 (0.02)

Demographic Controls Husband’s Other Employment Characteristics Year Fixed Effects Observations

Yes Yes Yes 4,794

Yes Yes Yes 4,794

Note: Data source is JPSC (1996-2013) for Panel A and JPSC (1996-2002) for Panel B. All controls, sample restrictions, regression specifications, clustering of standard errors in Panel A are analogous to Table 4. All controls, sample restrictions, regression specifications, clustering of standard errors in Panel B are analogous to Table 5.

We also compare whether work-life balance policies in husband’s firm has any effect on fertility preferences of currently married women who are not working in Panel B of Table 6. We find that work-life balance policies in the husband’s firm do not have any significant effect on the unwillingness to have a (or another child) for non-working women in both the random and fixed effects models (Columns (1) and (2) in Panel B of Table 6). Therefore, work-life balance policies in one’s own firm appear to be more important in influencing fertility preferences of women. Thus, these policies appear to have some potential in checking Japan’s declining birth rate.

5.2

Results by Different Subgroups

The previous analysis pertains to the entire sample of currently working women. However, there might be heterogeneity in the impact of work-life balance policies. The analysis conducted on the entire sample may mask potentially important differences in the impact of work-life balance policies on time allocation of currently working women. We consider the differential impact of these policies for full-time and part time working women and women for whom their/their husband’s parents live with the couple and for those whose parents/parents-in-law do not live them. We consider these different subgroups because of two main reasons. Firstly, the government mandated the expansion of work-life balance policies to cover part time employees. Secondly, the presence of one’s parents is

22

fairly common in Japan relative to other OECD countries and this can have important implications for the availability of childcare from one’s parents as well as the need to take care of one’s parents. We report the results from the specification where we include all potential controls (such as the specification in Columns (4) and (6) of Table 3) and use random and fixed effects models for our estimation. 5.2.1

Full-time vs Part time Working Women

In this section we study the implications of work-life balance policies in the wife’s firm on her time allocation depending on whether she is a full-time or part time worker. The odd numbered columns in Table 7 comprise the sample of currently working women who are full-time workers; whereas the even numbered columns pertain to the sample of part time working women. Further, the first two columns report the coefficients on worklife balance policies in the wife’s firm using the random effects model and the last two columns report the results from the fixed effects model. Table 2 shows that around 43 percent of the sample used for analysis comprises of full-time working women. Therefore, more than half of the sample used for analysis comprises of women who are part time workers. This is not surprising as the share of part time working women in the sample of women who are currently working is high in Japan. Around 70 percent of the pool of part time workers in Japan was found to be women in 2014 according to the International Labour Organization. Panel A of Table 7 estimates the impact of work-life balance policies in the wife’s firm on the share of time spent by the wife in the couple’s time spent in paid employment for full-time and part-time working women. Columns (1) and (2) in Panel A of Table 7 include the results from the random effects estimation model. We find that work-life balance policies in the wife’s firm raise the share of time spent by the wife in the couple’s time in paid employment for part-time workers only from the random effects estimation model. This is likely because part-time workers are likely to have greater flexibility in raising their working hours. However, we do not find statistically significant impact of work-life balance policies on the share of wife’s time in the couple’s time spent in paid employment irrespective of full-time working status from the fixed effects estimation model in Columns (3) and (4). Panel B of Table 7 estimates the impact of work-life balance policies in the wife’s firm on the share of time spent by the wife in the couple’s time spent in home production for full-time and part-time working women. Columns (1) and (2) show that there appears to be no significant effect of work-life balance policies on the share of time spent by the wife in the couple’s time spent in home production irrespective of full-time working status 23

Table 7: Share of Wife’s Time in Couple’s Time in Different Activities and Unwillingness to Have a Child: Full-time vs Part time Working Women Panel A:

Full-Time Worker RE -0.002 (0.01)

Part time Worker RE 0.01** (0.004)

Full Time Worker FE 0.01 (0.01)

Part time Worker FE 0.005 (0.005)

3,121 Full-Time Worker RE -0.02 (0.01)

4,210 Part time Worker RE -0.005 (0.01)

3,121 Full Time Worker FE 0.01 (0.02)

4,210 Part time Worker FE 0.002 (0.01)

Unwillingness to Have a Child Childleave/Careleave in Wife’s Firm

3,045 Full-Time Worker RE -0.06*** (0.02)

4,145 Part time Worker RE -0.01 (0.01)

3,045 Full Time Worker FE -0.05** (0.02)

4,145 Part time Worker FE 0.005 (0.02)

Observations Demographic Controls Wife’s Other Employment Characteristics Husband’s Employment Characteristics Year Fixed Effects

3,183 Yes Yes Yes Yes

4,301 Yes Yes Yes Yes

3,183 Yes Yes Yes Yes

4,301 Yes Yes Yes Yes

Paid Employment Childleave/Careleave in Wife’s Firm Observations Panel B: Home Production Childleave/Careleave in Wife’s Firm Observations Panel C:

Note: Data source is JPSC (1996-2013). Regression specification is analogous to Columns (4) and (6) of Table 3. For details on all other controls, regression specifications, sample restrictions and clustering of standard errors, please see table notes of Table 3.

from the random effects estimation model. We obtain analogous fining from the fixed effects estimation model in Columns (3) and (4). Lastly, Panel C of Table 7 estimates the impact of work-life balance policies in the wife’s firm on the unwillingness to have a (or another) child for full-time and part-time working women. From both the random and fixed effects estimation models, we find that the presence of work-life balance policies reduces the unwillingness to have a child for full-time working women only. This has important implications for labour market participation of women as well as birth rate in Japan. In particular, working full-time is associated with higher wages and better job security and benefits. However, the burden of domestic duties that are exclusively borne by women in Japan can be a serious challenge to participating full-time in the labour market for women. Therefore, working full-time and having children were often deemed as roles that were incompatible with each other for women. However, the presence of work-life balance policies appears to help women work full-time as well as reduce their undesirability to have children. 5.2.2

Parents live with the Couple vs Parents do not Live with the Couple

In this section, we estimate the impact of work-life balance policies on women’s time allocation on a typical working day depending on whether or not her/her husband’s 24

parents live with the couple in the same house/lot. Japan is a country with a large population of individuals who are aged 65 years and above. Table 2 shows that around 36 percent of working women have their/husband’s parents living in the same house/lot as the couple. This is higher than other non-Asian OECD countries. The presence of one’s parents or parents-in-law can have important implications for labour market participation and fertility. Parents/parents-in-law can provide help in performing domestic duties and can also serve as informal sources of childcare for couples. However, co-residing parents could also require care from their adult children. In Japan, social norms often dictate that women would be the caregivers for their parents/parents-in-law. Thus, the presence of parents/parents-in-law could have opposite effects on women’s participation in paid employment and performance of domestic duties. Now, work-life balance policies include parental care leave. Clearly, absence of work-life balance policies could adversely affect women’s time spent in paid employment. Therefore, it might be interesting to understand whether work-life balance policies can have any impact on a working woman’s time allocation as well as her fertility preferences when her/her husband’s parents co-reside with the couple. Table 8 reports the results of the impact of work-life balance policies on time allocation and fertility preferences of currently working women using the random and fixed effects models. The coefficients from the random effects model are presented in the first two columns and those from the fixed effects model are presented in the last two columns of Table 8. The odd numbered columns include working women whose parents/husband’s parents live in the same house/lot as them; whereas the even numbered columns include those whose parents/parents-in-law do not live with them. Panel A of Table 8 shows that work-life balance policies in the wife’s firm increase the share of wife’s time in the couple’s time spent in paid employment on a typical working day for women whose parents/husband’s parents live with them in both the random and fixed effects models. Although it is possible that one’s parents/parents-in-law are likely providing informal care to one’s children; but as co-residence often implies that women need to provide care to their parents/parents-in-law, work-life balance policies which also include parental care leave appear to be helping women with co-residing seniors spend more time in paid employment on a typical working day. Panel B of Table 8 shows that work-life balance policies in the wife’s firm decreases the share of wife’s time in the couple’s time spent in home production on a typical working day; but the coefficient is only significant in the random effects model. Further, work-life balance policies are found to reduce the wife’s time in the couple’s time spent in home production for all women; though marginally higher for working women whose 25

Table 8: Share of Wife’s Time in Couple’s Time in Different Activities and Unwillingness to Have a Child: Wife’s/Husband’s Parents live with the Couple or not Panel A:

Parents Live With Couple RE 0.02*** (0.004)

Parents Do Not Live With Couple RE 0.01 (0.005)

Parents Live With Couple FE 0.01*** (0.005)

Parents Do Not Live With Couple FE 0.01 (0.01)

2,671 Parents Live With Couple RE -0.02** (0.01)

4,660 Parents Do Not Live With Couple RE -0.01* (0.01)

2,671 Parents Live With Couple FE -0.01 (0.01)

4,660 Parents Do Not Live With Couple FE 0.003 (0.01)

Unwillingness to Have a Child Childleave/Careleave in Wife’s Firm

2,611 Parents Live With Couple RE -0.02 (0.02)

4,579 Parents Do Not Live With Couple RE -0.04*** (0.01)

2,611 Parents Live With Couple FE 0.01 (0.02)

4,579 Parents Do Not Live With Couple FE -0.02 (0.02)

Observations Demographic Controls Wife’s Other Employment Characteristics Husband’s Employment Characteristics Year Fixed Effects

2,744 Yes Yes Yes Yes

4,740 Yes Yes Yes Yes

2,744 Yes Yes Yes Yes

4,740 Yes Yes Yes Yes

Paid Employment Childleave/Careleave in Wife’s Firm Observations Panel B: Home Production Childleave/Careleave in Wife’s Firm Observations Panel C:

Note: Data source is JPSC (1996-2013). Regression specification is analogous to Columns (4) and (6) of Table 3. For details on all other controls, regression specifications, sample restrictions and clustering of standard errors, please see table notes of Table 3.

parents/husband’s parents live with them in the random effects model. Panel C of Table 8 shows that work-life balance policies have no significant impact on the unwillingness of having a (or another) child, irrespective of whether the parents or parents-in-law live with her in the fixed effects model; although it is found to reduce the unwillingness to have a child for working women whose parents or parents-in-law do not live with her only in the random effect model. We find that although the presence of one’s parents/parents-in-law increases the share of the woman’s time in the couple’s time in paid employment, it does not reduce time spent in home production (such as childcare and other domestic chores) and fertility preferences. This indicates that co-residing parents are likely in need of care and are less likely to be able to provide informal childcare or aid in performing domestic chores for the household. Therefore, these results likely indicate that the provision of parental care leave system is likely helping women spend more time in paid employment vis-`a-vis their husbands. Therefore, despite social norms that dictate that women are the primary caregivers of their parents/parents-in-law which could adversely affect their engagement in paid employment, the provision of parental care leave system as part of work-life balance policies appears to aid women spend more time in the labour market.

26

5.3

Instrumental Variable Estimation Results

We present the results from the instrumental variable (IV) estimation here as an alternative to the random and fixed effects estimation strategies. We instrument for the presence of child leave or care leave policies at the respondent’s firm by the proportion of firms in the industry that had implemented work-life balance policies prior to 1993 (that is, the starting year of the JPSC). The conjecture is that the higher is the proportion of firms in an industry that had a history of work-life balance policies, the higher is the probability that the respondent’s firm (that belongs to that industry) has child leave or care leave policies. Further, it is unlikely that individuals choose which industry (as opposed to the firm) to work in on the basis of work-life balance policies prevalent in that industry. This is because the choice of industry to work in largely depends on one’s education and skills and is less likely to depend on the work-life balance policies that the industry attempts to provide. Further, we use information on the prevalence of work-life balance policies at the industry level before the start of the JPSC. It is, therefore, unlikely that unobserved individual characteristics is likely to be correlated with the proportion of firms in each industry that had work-life policies prior to the JPSC. We use firm-level data from the Social Science Japan Data Archive at the University of Tokyo to construct the proportion of firms in each industry that had childcare or parental care leave policies before 1993, which was the starting year of the JPSC. We merged this information with the JPSC data at the industry level. We present the first stage results of the IV estimation where we regress the presence of work-life balance policies at the respondent’s firm on the instrument, the fraction of firms in the respondent’s industry that had work-life balance policies prior to the start of the JPSC, and other controls and report this finding in the lower panel of Table 9. We find that higher fraction of firms in the respondent’s industry that had work-life balance policies prior to the start of the JPSC is associated with a higher likelihood that the respondent’s firm has work-life balance policies. Further, the F-stat on the excluded instrument is 18.50. We report our findings from the instrumental variable estimation in Table 9 here. Panel A reports the IV results of the impact of leave policies at the respondent’s firm on the share of time spent by her in the couple’s time on a typical working day on paid employment and home production. We find that, while leave policies at the respondent’s firm have resulted in an increase in the share of time spent by the wife in the couple’s time in paid employment; these policies appear to have no impact on the share of time spent by the wife in the couple’s time spent in home production on a typical working day. In particular, the IV results indicate that the presence of leave policies at the respondent’s firm raises the share of the wife’s time spent in paid employment in the couple’s time on 27

Table 9: Instrumental Variable Results Panel A: Share of Wife’s Time in Couple’s Time in: Childleave/Careleave in Wife’s Firm

Paid Employment 0.09* (0.05)

Home Production -0.07 (0.08)

Observations Panel B: Fraction of Wife’s Time in Working Day in: Childleave/Careleave in Wife’s Firm

7,302

7,164

Paid Employment 0.09** (0.04)

Home Production -0.03 (0.05)

Other Activities (leisure etc.) -0.09* (0.05)

Observations Panel C:

7,301 Unwillingness to have a Child -0.37** (0.18)

7,301

7,301

Yes Yes Yes Yes

Yes Yes Yes Yes

Childleave/Careleave in Wife’s Firm Observations First Stage:

7,452 Childleave/Careleave in Wife’s Firm

Fraction of firms in Wife’s Industry with WLB Policies before 1993

0.11*** (0.01)

F-Stat on Excluded Instrument Demographic Controls Wife’s Other Employment Characteristics Husband’s Employment Characteristics Year Fixed Effects

18.50 Yes Yes Yes Yes

Note: Data source is JPSC (1996-2013). Regression specification is analogous to Columns (4) and (6) of Table 3. However, individual fixed or random effects are not included in estimation. Here, the presence of child leave or care leave policies at the wife’s firm is instrumented by the proportion of firm’s in the respondent’s industry that had work-life balance policies prior to 1993 (that is, start of JPSC). For details on all other controls, regression specifications, sample restrictions and clustering of standard errors, please see table notes of Table 3.

a typical working day by nearly 9 percentage points (which corresponds to an increase by 22 percent relative to the mean). This estimate is significantly higher than the estimates obtained from random and fixed effects estimation models reported in Columns (4) and (6) respectively of Panel A of Table 3. However, as in Panel B of Table 3 (in particular, the fixed effects estimate reported in Column (6)), we find that leave policies at the respondent’s firm have no significant effect on the share of her time in the couple’s time spent in home production on a typical working day using the IV estimation strategy. Panel B of Table 9 reports the IV estimation results of the impact of work-life balance policies in the respondent’s firm on the share of time she spends on a typical working day in paid employment, home production and other activities. We find that leave policies at the wife’s firm result in an increase in the share of her time spent in paid employment; however, this appears to be exactly compensated by a decline in the share of time devoted by her to other activities (such as leisure) on a typical working day from the point estimates. Specifically, the IV estimation shows that presence of work-life balance policies in the wife’s firm raises the share of time spent in paid employment by the 28

wife on a typical working day by 9 percentage points; while it results in a decline in the share of time spent in other activities by the same magnitude from the point estimates. Further, the share of time spent in home production on a typical working day appear to be unaffected on account of work-life balance policies. Although the point estimates from the IV estimation are larger in magnitude than those obtained from the fixed and random effects estimations for these outcome variables, the findings are qualitatively similar to the fixed and random effects estimations reported in Table 4. Panel C of Table 9 presents the IV estimation results of the impact of leave policies at the wife’s firm on the unwillingness to have a (or another) child, controlling for the number of children the respondent already has. We find that presence of work-life balance policies in the respondent’s firm reduces the unwillingness to have a (or another) child. This result is analogous to the finding in Panel A of Table 6 where the impact of work-life balance policies on fertility preferences of working women were reported using random and fixed effects estimation models.

6

Conclusion

This paper has studied the impact of work-life balance policies on the time allocation and fertility preferences of currently working, married Japanese women. The motivation for this study comes from the situation that participation of women in the labour market and men in home production is particularly low in Japan relative to most OECD countries. There are several possible factors that make women’s labour market participation particularly difficult in Japan. These include long working hours, preponderance of overtime work, socio-cultural norms that dictate women to be the primary caregivers of their children and elderly parents as well as women to be primarily responsible for performing domestic chores. Also, women were found to exit the labour market on account of marriage and childbirth. This would not only lower their earnings, but the experience lost on account of these personal events would also make re-entry into the labour market difficult. Further, the aforementioned labour market practices also imply that women can largely work part time as it offers them greater flexibility in terms of working hours and help them balance work and domestic duties. However, part time employment is less remunerative and has fewer social security benefits unlike full-time employment. One of the major labour market policies that aimed at increasing women’s labour market participation was the EEOL. The EEOL made it illegal for firms to practice discrimination against women in the labour market. This was because firms would often be inclined to not employ women as women would exit the labour market on account of 29

childcare responsibilities. However, a number of studies have indicated that the EEOL has not been successful in raising women’s labour market participation. Thereafter, the government of Japan introduced a number of policies that aimed at helping workers balance their domestic and work-related responsibilities. These work-life balance policies aimed at not only raising women’s labour market participation; but also attempted at raising men’s involvement in domestic duties and check Japan’s falling birth rate through the provision of childcare/parental care leave policies. Unlike the EEOL, these work-life balance policies are not merely aimed at making gender discriminatory practices by firms illegal. But these policies aimed at providing an environment for workers and firms in which performing one’s work-related responsibilities do not come at a cost of foregoing one’s domestic duties and vice-versa. In particular, childcare and parental care leave provisions are also available to men to help them participate in domestic chores. In general, work-life balance policies have the potential of changing cultural norms about gender roles. This motivated us to study whether these policies have indeed been successful in changing the time allocation among couples and fertility preferences of women. We use panel data for the period 1996-2013 and employ random and fixed effects as well as instrumental variable estimation methods to estimate the impact of work-life balance policies on the aforementioned outcomes. We find that work-life balance policies in the wife’s firm raise the share of time devoted by the wife in the couple’s time spent in paid employment on a typical working day. However, we do not find any significant impact of these policies on the share of the wife’s time in the couple’s time spent in home production on a typical working day as well as fertility preferences of women, especially in the fixed effects estimation model. Further, we find that the rise in the share of time spent by married women on a typical working day in paid employment is largely compensated by a decline in the share of time spent in other activities (such as leisure, personal care); while the share of time spent in domestic chores is found to remain unaffected on account of work-life balance policies in the respondent’s firm. Instrumental variable estimation method also provide findings largely analogous to the findings obtained from the random and fixed effects estimation models. We also study whether these policies are likely to have differential impacts on different subgroups of women. We find that these policies are likely to lower the unwillingness to have children among full-time working women. Further, the provision of parental care leave policies are likely to raise the time allocated by the wife in the couple’s time spent in paid employment for couples whose parents co-reside with them. On the other hand, work-life balance policies in the husband’s firm have no effect on the share of time on a typical working day a husband spends in home production. 30

Our results have important implications for labour market policies in Japan. On one hand, our analysis points towards the importance of work-life balance policies and the limited success these policies are found to have achieved regarding women’s time spent in paid employment. So far, however, these policies have not been successful in raising men’s involvement in domestic chores. This indicates that work-life balance policies may have a limited impact on cultural norms that dictate gender division of labour within the household, at least during the period of our analysis. Now, it will be interesting to analyse the impact of work-life balance policies on women’s subjective well-being as well as their physical and mental health. This is because, although work-life balance policies can raise women’s labour market participation, they do not appear to reduce the burden of domestic duties on them. Women are found to reduce their leisure time to accommodate for the increased time spent working. Therefore, it might be important to analyse how the welfare gains on account of greater involvement in paid employment compares with the loss in welfare on account of reduction in time for leisure and personal care. This analysis can indicate that work-life balance policies would need to be complemented with gender sensitization campaigns that can help alter cultural norms about intra-household gender division of labour and increase men’s involvement in domestic work. It is likely through the impact of these policies together that women would not have to spend more time working by cutting down their time devoted to leisure and personal care. These are important issues for future research.

References [1] Abe, Y. (2011). The Equal Employment Opportunity Law and Labor Force Behavior of Women in Japan. Journal of the Japanese and International Economies, 25(1), 39-55. [2] Abe, M., & Kurosawa, M. (2006). Work-Life Balance Support and Company Performance. In Report of the Study Group on Work-Life Balance Support and Company Performance, Issued by NLI Research Institute, Tokyo. [3] Almqvist, A.L. (2014). Changes in Gender Equality? Swedish Fathers’ Parental Leave, Division of Childcare and Housework. Journal of Family Studies, 20(1), 19-27. [4] Bartel, A.P., Rossin-Slater, M., Ruhm, C.J., Stearns, J., & Waldfogel (2016). Paid Family Leave, Fathers’ Leave-Taking, and Leave-Sharing in Dual-Earner Households. Unpublished Manuscript. 31

[5] Bunning, M., & Pollmann-Schult, M. (2016). Family Policies and Fathers’ Working Hours: Cross-National Differences in the Paternal Labour Supply. Work, employment and society, 30(2), 256âĂŞ274. [6] Ekberg, J., Eriksson, R., & Friebel, G. (2013). Parental leave- A Policy Evaluation of the Swedish “Daddy-Month” Reform. Journal of Public Economics, 97, 131-143. [7] Foster, G., & Stratton, L.S. (2017). Do significant labor market events change who does the chores? Paid work, housework, and power in mixed-gender Australian households. Journal of Population Economics, forthcoming. [8] Hui-Yu, C., & Takeuchi, M. (2009). The Effect of Work-Life Balance Policies on Female Employee Turnover. Unpublished Manuscript. [9] Kato, T., & Kodama, N. (2015). Work-Life Balance Practices, Performance-Related Pay, and Gender Equality in the Workplace: Evidence from Japan. IZA Discussion Paper No. 9379. [10] Kluve, J., & Tamm, M. (2013). Parental Leave Regulations, Mothers’ Labor Force Attachment and Fathers’ Childcare Involvement: Evidence from a Natural Experiment. Journal of Population Economics, 26, 983âĂŞ1005. [11] Lalive, R., & Zweimuller, J. (2009). How Does Parental Leave Affect Fertility and Return to Work? Evidence from Two Natural Experiments. The Quarterly Journal of Economics, 124(3), 1363-1402. [12] Morita, Y. (2005). The Child-Care Leave Law and the Demand for Female Labour. The Japanese Journal of Labour Studies, 536, 123-136. [13] Nepomnyaschy, L., & Waldfogel, J. (2007). Paternity Leave and Fathers’ Involvement with their Young Children: Evidence from the American ECLSâĂŞB. Community, Work & Family, 10(4), 427-453. [14] Organization for Economic Co-operation and Development. (2017). LFS - by age and sex: Labour Force Participation Rate. Retrieved from https://stats.oecd.org/Index.aspx?DataSetCode=LFS_SEXAGE_I_R [15] OECD Gender Data Portal. (2016). Time Use Across the World. Retrieved from http://www.oecd.org/gender/data/

32

[16] Ondrich, J., Spiess, C.K., Yang, Q., & Wagner, G.G. (2003). The Liberalization of Maternity Leave Policy and the Return to Work after Childbirth in Germany. Review of Economics of the Household, 1(1), 77-110. [17] Patnaik, A. (2016). Reserving Time for Daddy: The Consequences of Fathers’ Quotas. Unpublished Manuscript. [18] Rossin-Slater, M., Ruhm, C., & Waldfogel, J. (2013). The Effects of California’s Paid Family Leave Program on Mothers’ Leave-Taking and Subsequent Labor Market Outcomes. Journal of Policy Analysis and Management, 32(2), 224-245. [19] Rossin-Slater, M. (2018). Maternity and Family Leave Policy. In S.L. Averett, L.M. Argys & S.D. Hoffman (Eds.), Oxford Handbook on the Economics of Women (forthcoming). New York: Oxford University Press. [20] Schonberg, U., & Ludsteck, J. (2014). Expansions in Maternity Leave Coverage and Mothers’ Labor Market Outcomes after Childbirth. Journal of Labor Economics, 32(3), 469-505. [21] Takeishi, E. (2006). The Significance of Work-Family Policies for Companies. Japanese Journal of Labour Studies, 553, 19-33. [22] Tanaka, S., & Waldfogel, J. (2007). Effects of Parental Leave and Work Hours on Fathers’ Involvement with their Babies: Evidence from the Millennium Cohort Study. Community, Work & Family, 10(4), 409-426. [23] Yamaguchi, K. (2016). Determinants of the Gender Gap in the Proportion of Managers among White-Collar Regular Workers in Japan. Japan Labor Review, 13(3), 7-31. [24] Yamamoto, I., & Matsuura, T. (2011). Do Work-Life Balance Policies Increase a Firm’s Total Factor Productivity?: Evidence from panel data of Japanese firms. RIETI Discussion Paper Series 11-J-032.

33