1 Introduction

1.1 Background

A clear division of paid and unpaid work along gender lines in households is found in every country of the world, but a continuing trend towards dual-earner families, where both husband and wife are the family breadwinners, can be detected in recent years in many advanced industrial societies. The percentage of male breadwinner families steadily decreases and it may reflect a rise in the numbers of more gender-equal couples.

However, despite the continuing rise in women’s participation in the economy over the period of industrialization and beyond in Japan as well as many Western societies, gender division of labor has been still accepted as “normal” and strong in Japanese society. While the number of households with wives entirely dependent on their spouses’ income has dramatically declined, most women in dual-earner households still earn much less than their spouses, and the number of households in which wives earn more than their husbands are very few.

As gender inequalities in the division of labor within households are closely related to gender inequalities in society at large, particularly in the labor market, understanding what determines the division of labor within couples is a key to understanding other aspects of gender stratification. Many studies have argued that women’s economic dependency on men is an important attribute of stratification systems and essential force in the maintenance of gender inequality (e.g., Sorensen and McLanahan 1987). In this paper, we examine what differentiates dual-income couples from husband sole provider couples, and how these associations have changed over the past three decades in Japan.

1.2 Hypotheses

It is widely acknowledged that the incentives and restrictions that affect women’s employment are: (a) woman’s human resources, (b) economic need for household, (c) availability of resources for balancing work and family, and (d) values (e.g., Oppenheimer 1982; Treas 1987). Do these relationships also apply to Japan?

Based on some previous studies in Table 1, hypothesis are as follows.

Table 1 Some major findings regarding the determinants of women’s participation in paid work
Full size table

1.2.1 Human Resource Hypothesis

First, we hypothesize that the women’s education may have positive effects on married women’s labor force participation. Substantial studies have shown that women with higher educational resources have a higher participation rate in the labor market in all the Western industrialized countries (e.g., Sweet 1973; Blossfeld and Drobnič 2001). Highly educated women’s risk of employment exit tends to be considerably lower and the re-entry rate is higher. In general, highly educated women appear to combine work and family by reducing their working time rather than by exiting from employment. Educational expansion and the resulting improvement in women’s educational opportunities have led to increasing female labor force participation, undermining social norms favoring the male breadwinner households in many post-industrial societies. However, the effect of a woman’s educational attainment on her employment has not been significant in Japan. It has been pointed out that Japan is an anomaly, where women are highly educated but typically barred from making full use of their education in economic and political fields (e.g., Brinton 1993; Shirahase 2003; Nakai 2009).

1.2.2 Supplement Household Income Hypothesis

Second, husband’s socio-economic status may have negative effects on married women’s labor force participation. Married women may be more likely to enter the labor market when their husbands’ income is low, so that their earnings can supplement household income. According to some empirical research, a married woman’s labor participation is negatively associated with her husband’s socio-economic status (Nakai 2011).

However, it has been found that the impact of husbands’ resources on their spouses’ employment differs according to the institutional context since around the end of the twentieth century and these differences correspond to the welfare state regimes (Esping-Andersen 1990, 1999; Sainsbury 1999; Blossfeld and Drobnič 2001; Stier and Mandel 2009; Hofacker et al. 2013). For example, in continental conservative welfare states, the effect of husband’s socio-economic status on wife’s labor force participation is negative: husbands’ high occupational resources suppress spouses’ participation in paid work, showing the traditional division of labor within couples and increasing dependency of married women on their spouses over the life course. On the other hand, in social democratic welfare states, the effect of husband’s socio-economic status is positive, which means that men’s occupational resources increase their spouses’ labor market activity. Positive effect implies that economic resource at the household level facilitates a woman’s employment also because it helps balancing work and family. More and more advanced post-industrial economies see the positive effects of husband’s occupational resources on their partner’s participation rates in recent year. Women married to well-educated husbands as well as women with high-income partners are less likely to leave the labor market than women with low-resource partners.

There are a wide variety of discussions on the characteristics of Japanese welfare state as well as on East Asian welfare model (e.g., Goodman and Peng 1996; Esping-Andersen 1997). In some classification, Japan’s welfare system has often been classified as a sub-category of the conservative welfare states regime inclined towards the liberal regime or referred to as a combination of key elements of both the conservative and liberal welfare models (Esping-Andersen 1997, 1999). Japan model is regarded to fall closer to conservative welfare states, especially Southern Europe, in the sense that the welfare state is committed to traditional familialism. It implies that the family and the local community are the natural and ideal loci of welfare provision and the state’s role should be limited. If this framework can be applied, there may be a negative effect of husband’s resources on wife’s participation in paid work as some previous research showed (Blossfeld and Drobnič 2001). On the other hand, turning to the role of market-provided welfare and very limited social expenditures for social services, Esping-Andersen identifies Japan closer also to the liberal welfare regime, which is associated with the dual-earner/market career family model. Under welfare states similar to the liberal welfare regime, the effect of husband’s resources on their wives’ employment transitions might be different from that in countries belonging to the conservative and Mediterranean welfare states.

1.2.3 Modernization Hypothesis

Thirdly, we also hypothesize that values and attitudes toward the family and gender roles may affect women’s participation in labor market. Inglehart and Norris (2003) argue that the twentieth century gave rise to profound changes in traditional sex roles. But the force of this “rising tide” has varied among rich and poor societies. They demonstrate that richer, post-industrial societies support the idea of gender equality more than agrarian and industrial societies and intergenerational differences in values are largest in post-industrial societies and relatively minor in agrarian societies, suggesting that the former are undergoing intergenerational changes in values. They also argue that cohort change in gender-role attitudes in post-industrial societies is unidimensional, with newer cohorts consistently more egalitarian than older cohorts. This “increasing egalitarianism in gender-role attitudes” is attributed to modernization and generational replacement. Given that younger cohorts are more egalitarian than older cohorts, it may lead to the rise in married women’s labor force participation.

1.2.4 Cohort Hypothesis

Suppose a particular cohort experiences a change in labor market institutions that significantly affects the employment opportunities for women, such as the Equal Employment Opportunity Law (EEOL) or the Maternity Leave Act, and thus, the employment patterns of that cohort. Then, certain cohort may possess different gender norms, or more gender-egalitarian values, and tend to participate in employment more than other birth cohorts (e.g., Elder 1975, 1994; Shorrocks 2016).

The life course perspective focuses on the interplay of human lives and historical times. Especially in rapidly changing societies, differences in birth year expose individuals to different historical world, with their constraints and options. Individual life courses may well reflect these different times.

2 Data and Methods

2.1 Data

Data for this study were obtained from the past three decades of four waves of cross-sectional data: the 1985, 1995, and 2005 Social Stratification and Social Mobility (SSM) surveys of Japanese society, and the 2015 Stratification and Social Psychology (SSP) survey in Japan. All the surveys were conducted with similar approach: face-to-face interviews with a special focus on social stratification and inequality in contemporary Japan. All the surveys selected national representative respondents through multiple-stage sampling. The subjects of these surveys were men and women, aged between 20 and 69 for the surveys in 1985, 1995, and 2005, and between 20 and 64 for the 2015 SSP survey. Data were collected from 1248 men and 1405 women in 1985, 2490 men and 2867 women in 1995, 2660 men and 3082 women in 2005, and 1644 men and 1931 women in 2015. The response rates were 67.9%, 66.0%, 44.1%, and 43.0% in 1985, 1995, 2005, and 2015, respectively.

To make data comparable across the four datasets, we limit our analysis to the working-age couples: Specifically, those who are married and wives’ age is between 25 and 54, N = 9067 (994 in 1985, 3180 in 1995, 2862 in 2005, and 2031 in 2015).

2.2 Measurement of Variables

2.2.1 Dependent Variable

We focus on within-couple inequality in the family. We use a concept of wives’ contribution to household income as an aspect which reflects within-couple inequality, which is defined as: (a) income provision-role type and (b) wives’ contribution to total household income. In this study, we analyze (a) income provision-role type as a dependent variable.

Income provision-role type is measured based on whether a dominant provider exists and identifies who she/he may be. We use a five-group classification: (1) husband sole provider, (2) husband provides majority, (3) equal providers, (4) wife provides majority, (5) wife sole provider (Raley et al. 2006). Husband sole provider category consists of couples where only husband is employed. Husband provides majority category consists of couples where husbands’ earnings represent 60% or more of the combined total income of the husband and wife. Equal providers category identifies couples where wife’s earnings represent somewhere from 40% to 60%, meaning that each partner contributes between 40 and 60% of total household income. Wife provides majority category consists of couples where wives’ earnings represent 60% or more of the combined total income of the husband and wife. Wife sole provider category consists of couples where only wife is employed.Footnote 1

We analyze which factors differentiate dual provider couples (three dual-income groups) from husband sole provider couples.

2.2.2 Independent Variables

To capture the effects of human resources of women, we include wife’s education. Wife’s education is collapsed into four categories: (1) less than high school, (2) high school graduate, (3) two-year college, and (4) four-year tertiary education or more. Wife’s age is coded into six categories: (1) 25–29, (2) 30–34, (3) 35–39, (4) 40–44, (5) 45–49, and (6) 50–54, where 30–34-year-old group is the reference category. Wife’s birth cohort is coded into five categories: (1) 1931–1943, (2) 1944–1953, (3) 1954–1963, (4) 1964–1973, and (5) 1974–1995. The birth cohort group of 1931–1943 were born before WWII. The birth cohort group of 1944–53 were born in early postwar period and this group includes the first baby boomer generation. The group who were born in 1954–1963 is a cohort who entered the labor market prior to the Equal Employment Opportunity Law (EEOL) enforcement, whereas the group who were born in 1964–1973 is a post-EEOL cohorts, those who were 22 or younger in 1986, when EEOL went into effect in Japan. The group who were born in the period 1974–1995 are the most recent cohort and often referred to as the second baby boom generation and post-bubble who finished schooling in the recession period, when the labor demand was weak.

Married couples division of labor within household may vary systematically also with regards to household level characteristics. The household level explanatory variables include age and the number of children within a household, husband’s income, and the couples’ relative education. We create variables to indicate the number of children and the presence of a preschooler. The four categories measuring number of children are: (1) no, (2) one, (3) two, and (4) three or more children, where no children is the reference category. Husband income level is measured by income decile (ten groups) in each survey year. The couples’ relative education-level variable measures whether wife has higher or lower education than her spouse and has three categories: (1) husband and wife have equal educational attainment, (2) wife has higher education than her spouse (hypogamy), and (3) husband has higher education than his spouse (hypergamy), where equal educational attainment is the reference category.

2.3 Method

Although the level of the original dependent variable is four ordinal categories except wife sole provider category, we treat the response variable as dichotomous, whether the married couple is dual provider or husband sole provider couple. The focus here is on as to which factors are determinants for wives in playing an important role by contributing to household income because not all research has consistent findings with regard to the impact of various factors on women’s participation until recently. We examine what differentiates dual-income couples, which consist of three dual-income groups (where the husband provides the majority of income, equal providers, and the wife provides the majority), from husband sole provider couples. This is very important since the traditional division of roles within households where a man is the sole provider and a woman is the main care provider for the family remains relatively unchanged despite the increasing rate of female labor force participation in Japan. The effects of individual-level and household level characteristics on the pattern of couple’s income provision role, or breadwinner type, are analyzed using logistic regression analysis. Since we want to take account of differences across birth cohorts with respect to the effect of wife’s education on couples income provision type, assuming that associations of women’s human capital with work increase over time, we add interaction terms to our models.

3 Results

3.1 Descriptive Statistics

We first examined how families are trying to allocate time to market work and household production within couples. Table 2 shows how bread-winning patterns among married couples have changed over the past three decades. Until 2015, an overwhelming majority of couples were dual providers. The proportion of the households with husbands as sole provider have declined from 42.8% in 1985 to 29% of the observations in 2015. However, compared to Western post-industrial societies, excluding a couple of continental European countries such as Spain and Italy, male breadwinner is still much higher (Harkness 2010). Moreover, equally shared income provisioning portrayed only 14% in Japan even in 2015.

Table 2 Trends in percent distribution of household types of couples and wife’s economic contribution to household income: 1985–2015
Full size table

The table also shows how married women’s earnings contribution of household income changed between 1985 and 2015. As expected, the percentage of earnings that comes from wives has increased quite dramatically recently, but still roughly only a quarter (25.6% and 23.1%) of family income in 2015.

3.2 Determinants of Wife’s Participation in Paid Work

Table 3 displays the results of a logistic regression estimating the likelihood of being in paid employment for wives of married couples. It investigates whether employment of married women is influenced by their own human capital, whether household level factors have a significant influence, and whether the effects have changed over time. Women’s human capital measured by their own education is not a determinant of labor participation for married women in Japan. Even having a college degree does not lead to women’s higher participation in the labor force.

Table 3 Logistic regression estimates for the likelihood that a couple is dual provider: 1985–2015
Full size table

Having said that, a four-year college degree may have somewhat different meaning by cohorts. Cohort itself does not make differences in the likelihood of being dual provider household, but, according to the interaction term, highly educated women born in the postwar period may have different gender-role attitudes in terms of participation in paid work and division of labor within household across birth cohorts. Women who were born between 1944 and 1953 may have relatively stronger gendered division of household labor than other birth cohorts and their economic participation tend to be inactive. Many of them got married and settled down in the 1960s and 1970s, when Japan’s economy grew. Being housewife symbolized middle-class status back then and many female baby boomers stayed home as housewives. In contrast, highly educated women who were born between 1964 and 1973 tend to commit to their career than other cohorts. They experienced a social context that was significantly different from that of the earlier period in their youth, when economic growth stagnated after the oil shock in 1973. These women came of age in the Bubble Era and entered the labor market as the Equal Employment Opportunity Law was being implemented, which might lead the college graduates to active participation in the labor force.

Age effects show a different pattern from Western countries. It is not reverse U-shaped or hump-shaped pattern, which implies the highest employment level in mid-career, or in their 30s, and lower probability for older age group. In Japan, wives of older age group, or in their 40s, have higher probability of being in paid work than those in their 30s.

Compared to the women’s own variables, household level and spouses’ variables have significant effects. We find significant negative effects of husbands’ income on wives’ labor force participation. This means that high occupational resources of husband drive wife out of the labor market, which is the relationship known as “Douglas-Arisawa effect,” which hypothesizes that the decision as to whether or not a woman becomes a paid worker is influenced by her husband’s income; the lower the husband’s income, the more the wife tends to choose paid work (Douglas 1934; Arisawa 1956). It is sometimes claimed nowadays that the argument of negative association between husband’s income and wife’s participation in paid work does not hold true anymore. However, in reality, the results of the analysis show that the relationship mentioned in the “Douglas-Arisawa effect” is still valid in Japan. This relationship has been found in conservative and Mediterranean welfare states in some previous research (Blossfeld and Drobnič 2001). Unlike in the case of Social democratic welfare states, whose tax system promotes wives’ employment as taxes are individual-based rather than household-based, the tax and pension policies in Japan are geared toward discouraging wives to work.

Quite a few studies have investigated the associations between wives’ labor force participation and income inequality of society level (e.g., Treas 1987; Breen and Salazar 2010; Breen and Andersen 2012; Shin and Kong 2015). Past studies have referred that wives’ earnings once decreased income inequality among households in a society, but the increasing positive association between spouses’ earnings due to educational assortative mating, which helps to maintain greater economic equality within marriage on the one hand, contributes to growing earnings inequality among married couples in recent years (Schwartz 2010; Shin and Kong 2015). In Japan, however, the traditional division of labor in the family has not changed a lot and therefore wives’ earnings still help to reduce income inequality across married couple families.

As for relative education level of a couple, wives with higher educational attainment than their husbands are not necessarily more likely to participate in work, but wives’ labor participation is restricted among couples where wives have lower educational level than their husbands. This suggests that values related to household context influence gendered arrangement for work and care in the household. Educational hypergamous couples, in which women marry men of higher status than themselves, may prefer more traditional marriage practices, with women being mainly responsible for caring for children and housework, with men having bread-winning role.

The presence of preschool children strongly negatively affects wife’s labor participation. However, mothers’ participation increases as the number of child increases. In Western European countries and the United States, there has been a steadily increasing trend toward paid employment by married women, especially among those with young children. On the other hand, the employment rate for women with young children in Japan is currently one of the lowest among advanced industrial countries (OECD 2016).

4 Conclusion and Discussion

To summarize, household level and partner variables have stronger influences, whereas women’s own human capital has little influence on women being in paid work, which would suggest a more traditional division of labor.

After all, unlike in the case of Western advanced societies, Japanese women seem to adjust their working style primarily in response to household needs. These findings are consistent with previous studies (Nakai 2009, 2011). In the Western advanced societies, wife’s college degree pushes a couple toward dual providing, but in Japan, women’s own education has not been a determinant of labor participation for married women until recently.

Analyzing cohort differences more thoroughly in future research could enrich theory and evidence about how introduction of policy package might affect employment of married women, especially mothers of preschool children, as well as societal level of gender equality. Further research must explore the factors that differentiate couples where the husband provides the majority of the couple’s income from equal providers.