1 Introduction

Gender inequality with respect to subjective well-being (SWB) has been a focus of researchers around the world for the past several decades. The direction of the gender gap in the existing literature varies due to the various economic environments and social norms across countries (e.g., Blanchflower and Oswald 2004; Kahneman and Krueger 2006; González-Carrasco et al. 2017). In standard economic models, more time spent on labor and housework tends to lower one’s utility level given constant earnings. Conversely, leisure hours can raise utility. However, China provides an intriguing puzzle. On the one hand, in China, females, on average, are found to spend more time on housework, nurturing activities and paid work while earning less than males (Zhang et al. 2008). On the other hand, females in China exhibit higher levels of utility than males. There exists minimal literature that analyzes the gender gap in SWB from this perspective or that explains this gap based on the gender-specific differentiation in the marginal utility of leisure hours. Accordingly, this paper empirically estimates the size of the gender gap by first examining this cause in the context of China.

Enhancing residents’ well-being is one of the most important goals for policy makers. For many researchers in the social sciences, happiness (or life-satisfaction) is widely accepted as a valid measure of individuals’ SWB, and thus, its determinants have received a recent surge of attention (e.g., Easterlin 1995; Ferrer-i-Carbonell 2005; Luttmer 2005; Clark et al. 2008; Dolan et al. 2008; Chui and Wong 2016). The direction, however, of the gender gap in SWB is far from consensus in the existing literature. Some find that women report higher levels of SWB than men (e.g., Di Tella et al. 2003; Blanchflower and Oswald 2004; Graham and Chattopadhyay 2013), others find the opposite (e.g., Mroczek and Kolarz 1998; González-Carrasco et al. 2017), and some find no difference at all (e.g., Kahneman and Krueger 2006). Different economic environments and diverse cultural norms may account for this inconsistency.

Nonetheless, it is important to understand the current gender gap in SWB in China as it can provide information for improving gender equality in other Asian and developing countries with similar cultural norms. China has experienced an unprecedented economic growth since the 1980s, and this rapid economic growth has absorbed substantial numbers of females into the labor force. Intuitively, labor participation is associated with greater autonomy for women, which contributes to increases in women’s earnings and their sense of well-being. Although the female participation rate has recently experienced a slight decline in China due to discrimination and fertility policy changes (Fincher 2016; Hare 2016; Xu 2016), it is still higher than most countries around the world.Footnote 1 However, women are also still responsible for most housework and nurturing activities according to traditional Asian cultural norms, thus possibly leaving females in an even worse situation than males as they are burdened with both paid work and household chores. This situation, without a doubt, impinges on the time allocation pattern between the genders. However, few of the existing studies focus on exploring the gender gap in SWB in China and the potential causes of this gap through this perspective. Thus, the question guiding this paper is whether this allocation affects levels of SWB and induces gender-specific heterogeneity in marginal returns from leisure activities.

The empirical evidence indicates that women in China earn less in the labor market and spend much more time on housework than men (Zhang et al. 2008) and that both behaviors have a negative impact on SWB. Asadullah et al. (2015) and Liu et al. (2016), after controlling for socioeconomic covariates, conclude that there is a gender gap in SWB that favors women. If this gap is real, then why are Chinese women, on average, happier than men, given the burden they bear with respect to both paid and unpaid work? As hours spent on paid and unpaid work negatively affect happiness, one plausible mechanism is that women appreciate and enjoy their leisure hours more. Zhang et al. (2017) documented that older females are more likely to engage in social and performing-arts activities and that this leisure participation has contributed to gender differences among the elderly in Shanghai. However, different from the older females, working-age women face different trade-offs, family responsibilities and economic pressure. Accordingly, research conducted among nationwide samples is necessary before any generalized conclusions regarding the working population can be drawn.

Our study uses data from a recent nationwide household survey to estimate the gender difference with respect to SWB among the working-age population. As a novel contribution, we empirically examine one cause for the gap: leisure hours and the gender difference in marginal returns from leisure hours. First, after controlling for respondents’ demographics, personal annual income, comparison income, employment status, education attainment, health conditions, life cycle profiles and geographic fixed effects, the empirical results yield a significant conditional gender gap in SWB that favors women, a finding that is consistent with previous studies. Second, we evaluate the role of leisure activities and find a significantly positive effect of leisure hours on happiness. Third, an interaction term of leisure hours and gender is controlled to explore the gender-specific marginal returns from leisure hours. We find that this interaction terms explains between 34 and 37% of the gap. In addition, we evaluate three specific leisure activities: hours spent on television, music and broadcasts, hours spent on traditional reading, and hours spent on Internet entertainment. Consistent results are found, and they are robust to a variety of model specifications and subsamples.

This article is organized as follows. Section 2 reviews the related literature on subjective well-being and the related gender gap. The theoretical framework of this paper is presented in Sect. 3, and the data, a statistical summary and empirical models are presented in Sect. 4. Section 5 then presents the main results and robustness checks and discusses heterogeneity across subsamples. Section 6 concludes the paper.

2 Literature Review

Since Easterlin (1974) proposed the concept of the Easterlin paradox, literature discussing SWB (happiness or life-satisfaction) and its determinants has substantially increased.Footnote 2 The general determinants of SWB include but are not limited to income, comparison income, age, gender, health status, employment status, and marital status.

Similar to the gender wage gap and other gender inequalities, the gender gap in SWB has attracted many researchers, but the results are far from consensus. For example, some surveys find that women report higher levels of SWB than men (e.g., Di Tella et al. 2003; Blanchflower and Oswald 2004; Graham and Chattopadhyay 2013), some find the opposite (e.g., Mroczek and Kolarz 1998; González-Carrasco et al. 2017), and some conclude there is no difference (e.g., Kahneman and Krueger 2006). By analyzing the trends of SWB with respect to both genders, Easterlin (2003) states that in America, women at a young age are happier than their male counterparts. However, this relationship will reverse as the individuals become older. Stevenson and Wolfers (2009) conclude that the level of happiness among women in the U.S. has declined for the past 35 years. Herbst (2011) further finds that between 1985 and 2005, the levels of well-being of both men and women exhibit a similar decrease in the U.S. Some recent studies have gone beyond the one-country paradigm and conducted global analyses. For example, Graham and Chattopadhyay (2013) conclude that in most developing countries (including China), women exhibit higher levels of SWB. Lima (2011), although reporting similar results, also finds a reverse result in some European countries and other developed countries.Footnote 3 Despite this inconsistency, studies unilaterally confirm that women are happier than men in China, a developing country (e.g., Appleton and Song 2008; Jiang et al. 2012).

Tesch-Römer et al. (2008) categorize the origins of the gap into two parts. The first, which is based on studies by biologists and psychologists, argues the effects of universal sex differences (e.g., Lippa 2005). The other, conducted by sociologists and economists, investigates the different living environments of men and women. Equal rights are one of the most interesting factors. For instance, Pezzini (2005) estimates the effects of implementing birth control rights on the welfare of women in 12 European countries. In those countries, abortion rights and the birth control pill are both substantially correlated to an increase in women’s life satisfaction, whereas neither has an impact on the level of satisfaction of men. In addition, family characteristics and cultural diversity are explored to explain this inequality of happiness between men and women. Margolis and Myrskylä (2011) study the correlation of happiness and fertility and note that happiness declines with the number of children. When women are young, having children has a negative effect on their level of happiness. However, this negative effect gradually disappears and finally becomes positive as women grow older. Tesch-Römer et al. (2008) investigate the effects of a country’s cultural and societal conditions and conclude that “the larger the gender societal inequality, the larger the gender differences in SWB.” Zuckerman et al. (2017) discuss the role of self-esteem in determining the gender gap, and other studies research the impacts of religions and ethnicity on SWB (Meisenberg and Woodley 2014).

Some newly emerged studies explain the gender gap in SWB through daily activities. Della Giusta et al. (2011) find that hours of paid work have a positive effect on well-being, whereas housework hours only affect retired men and women in Britain. The current study attempts to connect leisure activities to the gender gap in SWB. Many scholars find that leisure activities have non-negligible effects on SWB and that these effects can either be positive (e.g., Van Praag et al. 2003; Pagán 2015; Brajša-Žganec et al. 2011; Becchetti et al. 2012) or negative (e.g., Schmiedeberg and Schröder 2016; Wang and Wong 2014). The closest studies to ours are Boye (2009) and Zhang et al. (2017). The former attributes the lower levels of happiness in females in European countries to labor hours. The latter focuses on the elderly in Shanghai and finds that males and females have different preferences with respect to leisure activities. For example, males favor detachment-recovery and aesthetic activities while females favor performing-arts activities. Consistent with this branch, we focus on leisure and gender differences in SWB.

3 Theoretical Analysis

SWB is currently accepted as a valid measure of a person’s state of well-being (Clark et al. 2008). Thus, SWB is widely used as a proxy for utility (e.g., Ferrer-i-Carbonell 2005; Böckerman et al. 2016; Peng 2016). Formally,

$$SWB = h\left( U \right) + \varepsilon$$
(1)

where \(\upvarepsilon\) represents the homoscedastic transformation error when the respondent declares his happiness level based on his internal utility. A sufficiently large sample can remove the error when evaluated as a whole. Furthermore, \(h\left( \cdot \right)\) is a positive monotonic transformation such that for any U 1 > U 2, SWB 1 > SWB 2.

Our interest is to evaluate the effects of leisure activities on SWB and on other daily activities. Consequently, SWB is stated as follows:

$$SWB = h\left( {U\left( {H_{labor} , H_{housework} ,H_{leisure} ,X} \right)} \right)$$
(2)

where \(H_{labor} , H_{housework} ,H_{leisure}\) represents hours spent on labor, housework, and leisure activities, respectively. X is a set of socioeconomic and demographic variables. Females, on average, possess less leisure time and engage in more housework than males in China. Nonetheless, they exhibit higher levels of SWB:

$$\left. {SWB} \right|_{female} > \left. {SWB} \right|_{male}$$
(3)

Similar to many other countries in the world, the traditional culture in China states that men are responsible for supporting the family, which results in pressure from the competitive labor market. The pressure from the labor market is generally expected to be greater than the pressure associated with family chores, which traditionally is borne primarily by females (Cao and Chai 2008). Therefore, it is plausible that females are, on average, happier than males. During the past several decades, however, females in many regions have largely assumed the roles of males in society, and thus, they bear the pressure from the labor market as well. The double burden of labor and housework simultaneously reduces the well-being of women in China and throughout the world.Footnote 4 On the contrary, most leisure activities directly promote physical and mental health, reduce stress related to life and work, and improve interpersonal relationships. More leisure hours can generate positive feelings, alleviate negative emotions, and enhance quality of life (Brajša-Žganec et al. 2011).

Consider two respondents, one female and one male, who share the same socio-economic and demographic characteristics (same age, residential area, income, human capital, physical health, etc.). The one with fewer leisure hours will exhibit a lower level of life satisfaction than the one with more leisure hours. In the scenario of China, women in the labor force must engage in both paid work and unpaid housework, and as a consequence, they have fewer leisure hours. Interestingly, however, Chinese women are still happier than men. One plausible explanation is that females derive higher marginal utility from leisure hours than do men, and females are more perceptual and appreciate leisure hours more than males. In other words, when leisure time is at a minimum, it is valued and appreciated in a more intense way.

Moreover, men and women in China as well as in many other countries throughout the world are traditionally required to play different roles and behave differently in society. Thus, it is reasonable to assume that this marginal difference between men and women may vary across different types of leisure activities. In addition, Zhang et al. (2017) find heterogeneity in the tastes for leisure activities between men and women in Shanghai. The dataset used in this study includes hours spent on three specific types of leisure activities: television, music and broadcast; traditional reading; and Internet entertainment. Accordingly, it is valuable to test the effects of these specific activities on men and women. In addition, the marginal difference should be more significant in urban areas because in comparison to rural women, the labor force participation rates of urban women are much higher, and urban women are also responsible for most of the housework. Single respondents have much less housework than do couples, and therefore, leisure time for married women is significantly less than it is for single women. Accordingly, marginal differentiations should be more pronounced for the married subsample.

Based on the above analysis, our hypotheses are summarized as follows:

Hypothesis 1:

On average, while females can derive higher marginal utility from leisure hours than males, this heterogeneity varies depending on the specific type of leisure activity.

Hypothesis 2:

The gender-specific differentiated marginal utility of leisure hours is more significant for urban and married subsamples.

4 Empirical Model and Data

4.1 Empirical Model

Based on the literature and Eq. (1), the estimation function is formulated as follows:

$$SWB_{i} = \beta_{0} + \beta_{1} Female_{i} + \beta_{2} LeisureHours_{i} + \beta_{3} LaborHours_{i} + \beta_{4} HouseworkHours_{i} + X_{i} + \varepsilon_{i}$$
(4)

in which subscript i represents individual i; X i is the vector of control variables; Female i is the gender indicator and equals 1 if individual i is female (0 otherwise). We apply both Ordered probit and OLS estimations. Ferrer-i-Carbonell and Frijters (2004) find that these two approaches should generate similar results. As leisure hours, labor hours and housework hours are the key explanatory variables, their coefficients can reveal marginal effects of time allocation decisions. β 1 reveals the conditional gender gap in SWB controlling for a series of covariates.

Next, we introduce the interaction term of gender and leisure hours into Eq. (4). The estimation function is as follows:

$${\text{SWB}}_{i} = \beta_{0} + \beta_{1} LeisureHours_{i} + \beta_{2} Female_{i} + \beta_{3} LeisureHours_{i} *Female_{i} + X_{i} + \varepsilon_{i}$$
(5)

In Eqs. (4) and (5), X i includes labor hours, housework hours, and other control variables. Hypothesis 1 is tested by evaluating the following inequality:

$$\left. {\frac{{\partial {\text{SWB}}}}{{\partial {\text{leisure}}}}} \right|_{female = 1} > \,\left. {\frac{{\partial {\text{SWB}}}}{{\partial {\text{leisure}}}}} \right|_{female = 0}$$
(6)

After taking simple derivatives of (5), it is noted that \(\left. {\frac{{\partial {\text{SWB}}}}{{\partial {\text{leisure}}}}} \right|_{female = 1} = \beta_{1} + \beta_{3}\) and \(\left. {\frac{{\partial {\text{SWB}}}}{{\partial {\text{leisure}}}}} \right|_{female = 0} = \beta_{1}\). If β 3, the coefficient of the interaction term, is statistically significant, we can deduce that women derive different marginal utility from leisure hours. In addition, we replace the total leisure hours with the hours of the three different leisure activities mentioned above. We include the interaction terms of female and the hours of each activity into the regression. In this way, we can further analyze the effects of different leisure activities on SWB and determine how the marginal utility from each activity differs between males and females.

4.2 Data and Measurement

The main empirical sample is drawn from the Chinese Family Panel Study (CFPS). We use the 2010 wave since it is the only wave that contains information on respondents’ leisure activities. The CFPS is a large-scale, nationally representative survey covering 162 counties from 25 provinces in China. To estimate how labor hours, housework hours and different leisure activities influence SWB, our sample is narrowed to individuals between the ages of 16 and 65 years and excludes full-time students and retired and disabled persons. The survey provides information on individuals’ annual income, self-reported health condition, education attainment, ethnicity, gender, and residential information. Details regarding the individual’s daily time allocations are also gathered.

4.2.1 Dependent Variables

There are two types of SWB questions on this CFPS questionnaire.Footnote 5 The first is “How happy are you overall?” and the second is “On the whole, are you satisfied with the life you lead?” The answers are arranged in cardinal order by asking the respondents to choose from 1 to 5, where 5 equates to very happy or very satisfied, 3 means ok or not happy but not unhappy and 1 denotes not at all happy or not at all satisfied. These two measures, i.e., happiness and life satisfaction are both widely used and acknowledged in the existing literature. In our work, while we adopt the happiness variable as the benchmark (Budria 2013; Liu and Shang 2012; McBride 2001), the other variable is also estimated and presented.

4.2.2 Interested Explanatory Variables

Our interested explanatory variables include female, leisure hours, labor hours and housework hours. The survey asks questions such as “How many hours on average per weekday do you spend on work”; “How many total hours per weekday do you spend on leisure activities”; “Of the total leisure hours, how much time do you spend on television, music and broadcasts”; “Of the total leisure hours, how much time do you spend on traditional reading”; “Of the total leisure hours, how much time do you spend on Internet entertainment”; and “How many hours per weekday do you spend on housework”. Accordingly, we can examine the effects of the total leisure hours as well as the hours spent on each leisure activity.

4.2.3 Other Control Variables

Individual demographic and economic characteristics are controlled in our estimations. First, age and squared age are included to control for the U-shaped link between reported well-being and age.Footnote 6 Second, we partial out other confounding factors, including income, comparison income, employment status, marital status, self-reported health condition, hukou,Footnote 7 ethnicity, education attainment and regional heterogeneity. Regional heterogeneity is controlled by controlling an urban dummy and 162 county indicators. Income is taken as a natural logarithm for the normal distribution transformation. Comparison income is measured as the corresponding county’s average annual income. Ethnicity is a dummy that indicates whether the respondent belongs to the Han, the most common ethnic group in China (approximately 92% of the total population). Employment status is measured by a dummy indicating whether the individual is currently employed.

4.2.4 Descriptive Statistics

Descriptive statistics are reported in Table 6 of the “Appendix”. Our raw data indicate that females are, on the whole, significantly happier than males. On average, the mean happiness (life satisfaction) score for females is 0.07 (0.095) point higher than that for males on the 5-point scale. Figure 1 illustrates this gap. Further, we investigate and graph the average happiness levels over the life-cycle by gender in Fig. 2. U-shapes for life-cycle SWB are found for both genders, with females exhibiting greater happiness than males over almost all age cohorts. This is consistent with studies of developing countries (e.g., Meisenberg and Woodley 2014). In Fig. 3, we document an inverted u-shaped age-labor profile and a u-shaped age-leisure profile for both genders. On average, males spent more time on both paid work and leisure activities than females.

Fig. 1
figure 1

Distribution of subjective well-being by gender

Full size image
Fig. 2
figure 2

Subjective well-being over life-cycle by gender

Full size image
Fig. 3
figure 3

Time allocation over life-cycle by gender

Full size image

Table 6 indicates that approximately 51% of the observations are of females and approximately 43% of the whole are urban residents. Females, on average, are 0.8 year younger than males. On average, they earn less, have less education, and self-report worse health conditions. Whereas there are more single men than women, there are more widowed women than men. The average working hours per weekday are five for females and seven for males. With respect to leisure time, males report an average 3.44 h per weekday compared to 3.14 h per weekday for females. Reported daily hour(s) spent on housework are 2.3 for females, but only one for males. While females report lower participation rates in the labor market than males, the total number of hours spent on paid and unpaid work (housework and nurturing activities) are almost equal between the genders.

There is no significant difference between the genders in hours spent on television, music and broadcasts. However, females spend fewer hours on traditional reading and Internet entertainment. Figure 4 presents the details of these leisure activities over age cohorts. It is evidenced that females spend more time on traditional television, music and broadcasts than males when they are young, but the opposite holds when the two genders get older. Interestingly, the gender gap with respect to traditional reading increases with age, while it decreases with respect to Internet entertainment.

Fig. 4
figure 4

Time allocations on leisure activities and housework by gender

Full size image

Descriptive statistics clearly indicate that women in China receive less income, have less education, engage in less leisure time and bear a greater burden regarding housework. Furthermore, all of these factors have negative effects on one’s level of well-being. Thus, it is natural to consider the heterogeneity in valuing leisure hours between the genders as one potential explanation for the current gender gap in SWB in favor of women.

5 Empirical Results and Heterogeneities

5.1 Main Results

Table 1 presents the estimation results of Eq. (4). We sequentially control for leisure hours, labor hours and housework hours to reveal their roles in influencing individual happiness and life satisfaction. The OLS and Ordered probit models are estimated separately. Consistent with previous studies, there is a U-shaped age-happiness profile, and education attainment and health conditions are found to be positively correlated with happiness. Marital instability is associated with lower levels of well-being. Moreover, we consistently find positive marginal effects of leisure hours and negative marginal effects of housework and labor hours.

Table 1 The effects of leisure and labor hours on subjective well-being
Full size table

In regression (1) of the OLS estimation (the results of the Ordered probit estimation are similar, as illustrated by Ferrer-i-Carbonell and Frijters 2004), the positive coefficient for female indicates that females are, on average, happier than males by 0.125 point on the 5-point scale. Furthermore, the coefficient of the natural log of income is found to be insignificant, indicating that income plays a negligible role in affecting SWB. Labor hours, leisure hours and housework hours are controlled in regression (2). As expected, one more leisure hour on weekday can significantly increase SWB by 0.028 point on the 5-point scale, while one extra labor hour decreases SWB by 0.005 point, and one extra hour of housework decreases SWB by 0.020 point. In regression (2), the effect of one’s own income on happiness is small but significant. Meanwhile, comparison income generates a large and significant negative effect on happiness in both regression (1) and (2). This is consistent with the findings of Asadullah et al. (2015), who use the 2010 wave of the Chinese General Social Survey, and Knight and Gunatilaka (2010, 2011), who use the 2002 Chinese Household Income Project Survey.Footnote 8 The negative role of comparison income is, overall, consistent with many studies in developing countries (Clark et al. 2008; Asadullah and Chaudhury 2012; Carlsson et al. 2008; Guillen-Royo 2011). Because a decrease in comparison income indicates an improvement in social status, it can result in a large utility gain. It was first proposed by Duesenberry (1949) and observed by Easterlin (1974) that individuals care about their income relative to others’ income as well as the income itself. There are, however, some inconsistences that suggest that a decrease in comparison income could lead to a lower level of satisfaction (Kingdon and Knight 2007; Linssen et al. 2011; Jiang et al. 2012). In addition, in the case of life satisfaction, we find insignificant estimates of comparison income for the OLS estimation but significantly negative effects are observed in the Ordered probit models. This may be due to discrepancies between the two measures. We suggest that future studies explore these findings.

From Table 1, we find that leisure hours increase levels of well-being, whereas housework and labor hours negatively affect emotions. Because females enjoy fewer leisure hours and have more household responsibilities than males, the conditional gender gap increases when we consider these decisions in regression (2), 0.152. Consistent results are observed in both the OLS and the Ordered probit estimation when using either the happiness or the life satisfaction measure.

Next, to explore whether the conditional gender gap in SWB can be explained by leisure hours at margins, we include the interaction term of female and leisure hours (Eq. 5). The results are reported in regression (2) in Table 2. The significantly positive estimates, 0.019 (happiness as the dependent variable) and 0.015 (life satisfaction as the dependent variable), indicate that one leisure hour marginally results in increased well-being for females than for males. The unexplained gender gap shrinks from regression (1) to (2) in Table 2, as the coefficient for female decreases from 0.169 (0.154) to 0.107 (0.103) with happiness (life satisfaction) as the dependent variable.

Table 2 The effects of leisure activities on the gender SWB gap
Full size table

Furthermore, we provide the results of the predictions and the average marginal effects in Table 3. In panel A, we predict the probabilities of being happy (happiness equals 4 or 5) at mean values of all other controlled variables for females and males with models without and with the interaction term, respectively. Before controlling the interaction term, the conditional gender gap is approximately 0.069, but after, it decreases to 0.038 with the interaction term. Similar conclusions are found when investigating the alternative dependent variable, life satisfaction. In general, the results imply that females better appreciate leisure hours, and moreover, this heterogeneity accounts for 30–45% of the unexplained conditional gender gap in SWB.

Table 3 Average marginal effects and the conditional gender gap
Full size table

In panel B, we compute the average marginal effects of the variable of interest (female) on the probability of feeling happy (happiness equal to 4 or 5) or satisfied (life satisfaction equal to 4 or 5). Regarding the marginal average effect of being very happy (happiness equals 5), the coefficient for female declines from 0.54 to 0.34 after controlling the interaction term. Following previous covariate analyses (Jennison and Turnbull 1997; Nunn 2008), this implies that the marginal gender differentiation accounts for 37% of the unexplained variation of exhibiting a happiness level of 5. Similarly, it explains 34% of the unexplained variation of having a life satisfaction level of 5. Regarding the marginal average effect of reporting a happiness level of 4, the coefficient of female declines by 39% after controlling the interaction term. The results of other variables, i.e., labor hours, leisure hours, the interaction term and housework hours, are presented in Table 9 in the “Appendix”.

Furthermore, we consider three specific leisure activities that are available in the survey: hours spent on television, music and broadcasts; hours spent on traditional reading; and hours spent on Internet entertainment (per weekday). Similarly, we explore how these leisure activities contribute to the conditional gender gap in SWB by adding the interaction terms of each activity to the female indicator. In general, all of these leisure activities contribute to higher levels of well-being (Table 2). With respect to the happiness measure, hours spent on television, music, and broadcasts contribute to more happiness for both genders, but the effect is significantly greater for females. Meanwhile, hours spent on traditional reading have an equal effect on happiness for both males and females. Finally, hours spent on Internet entertainment only affect the level of happiness for females. The decreasing coefficient for female after controlling the interaction terms (from 0.160 to 0.106) implies that marginal differences from these specific leisure activities contribute to the gender gap in SWB. Regarding the life satisfaction measure, the findings are generally similar, with only slightly different results. We also explore the average marginal effects and the conditional gender gap, as in Table 3, using hours of different types of leisure activities. As the results are consistent, they are omitted.

5.2 Robustness Checks

In this section, we rescale the measures of both happiness and life satisfaction and transform them into dummies. If one feels happy or very happy, i.e., happiness equals 4 or 5, respectively, then the dummy is equal to 1, and it equals 0 otherwise. We perform the same transformation on the life satisfaction measure. In Table 4, we consistently find positive effects of leisure hours and negative effects of housework hours on happiness and life satisfaction. Moreover, given one additional leisure hour, females can derive more utility at margins than males. As a determinant of the conditional gender gap in SWB, this marginal differentiation accounts for 40% (22.4%) of the unexplained variation when evaluating the happiness (life satisfaction) dummy. We also estimate the effects of specific leisure activities and find consistent outcomes (see Table 7 in “Appendix”).

Table 4 Robustness with alternative measurement
Full size table

5.3 Heterogeneity and Discussion

We separate the sample according to the respondents’ current marital and residential statuses. The results, based on the measure of happiness, are presented in Table 5 (results using the measure of life satisfaction are provided in Table 8). We consider these subsamples because both marital and residential statuses may substantially affect the social expectations of females. For example, traditional social norms expect females to be housewives. In general, married and urban females are expected to assume more responsibilities in either paid or unpaid work.

Table 5 Heterogeneous effects across subsamples
Full size table

First, for single respondents, i.e., never married, divorced or widowed, the coefficient of the interaction term is insignificant. For married respondents, females can derive a higher marginal return than males. As illustrated herein, married females in China not only imitate the social role of males in the labor market, but they are also responsible for the housework. As a response, they derive utility from leisure hours more efficiently than males. This finding is not reflected in the single population because both single males and single females must take care of themselves and do their own housework, whereas the housework burden of married females is disproportionately large, and their leisure hours are relatively rare.

Second, with respect to the urban subsample, the interaction term is statistically significant and positive. However, for the rural subsample, it is not significant. This suggests that while urban females’ marginal return from leisure hours is higher than that of urban males, this is not the case in rural areas. The reason is similar to that cited previously. That is, most rural families are still influenced by the traditional norm that females are not expected to have their own careers, and as a result, rural females do not value leisure hours more than males. The analysis of the married and urban subsamples reinforces our proposition that the double burden is the primary cause for this heterogeneity on marginal utility from leisure hours.

6 Concluding Remarks

This current study contributes to the literature that aims to explain the gender gap in SWB (Pezzini 2005; Margolis and Myrskylä 2011; Tesch-Römer et al. 2008; Meisenberg and Woodley 2014). We provide evidence of gender-specific differentiated marginal utility of leisure hours to explain this gap in China. We estimate not only the effect of leisure hours on SWB but also how this effect contributes to the gender gap in SWB.

Consistent with many other studies (e.g., Appleton and Song 2008; Jiang et al. 2012), a gender gap in SWB favoring women is observed in our Chinese dataset. After controlling for regular socio-economic characteristics, we find that leisure hours contribute to better well-being and account for approximately 0.15–0.17 of the conditional gender gap on the 5-point scale. Moreover, women have higher marginal returns from leisure hours than men, which helps to explain approximately 30–40% of the current estimated conditional gender gap in SWB. We also consider details of three specific leisure activities identified in the survey, namely, television, music and broadcasts; traditional reading; and Internet entertainment, and investigate their contributions to the gap. The results suggest that for any of these three leisure activities, females’ marginal utility is no less than that of males. Finally, we explore the effects across different sets of subsamples. The estimates indicate that single women and rural women are relatively happier than single men and rural men, compared with their counterparts (married women and urban women). The gender-specific marginal utility from leisure becomes insignificant for single and rural residents because that women among those subsamples do not bear the double burden of paid and unpaid work as single women do not need to do housework for their spouses and rural women are not expected to have a career. Accordingly, these concepts reinforce our proposition that the double burden is the primary cause for the heterogeneity on marginal utility from leisure hours.

Our results can serve as a reference for the government when considering policies to support leisure activities and increase expenditures on leisure-related infrastructures, such as public libraries, parks, Internet cafes, etc. Such policies not only help to accelerate human capital accumulation, but they also improve the SWB of the population, especially that of females. Furthermore, our findings shed light on the heterogeneous intrinsic nature of feelings between females and males. According to Guven et al. (2012), marital instability correlates to a family with a relatively unhappy wife. Thus, the happiness or life satisfaction level of the wife is a critical factor in maintaining a stable marriage. Finally, the ongoing migration in China may have a non-negligible effect on the gender gap in SWB.Footnote 9 In future studies, the impact of migration and the social status of the female should be considered to facilitate the understanding of the gender gap in SWB.