How do the gender gaps in Japan, Korea, and China rank globally? Figure 1 shows overall ranking using the Global Gender Gap Index (GGGI). It shows that, in 2017, among the 144 countries (regions) covered, Japan ranked 114^th, Korea 118^th and China 100^th. This means that for all three countries the gender gap remains significant [1] .

Looking at trends in the overall GGGI ranking, the number of countries (regions) covered is increasing, but the order in which the three countries rank is unchanged. That said, in the four years since 2013, Japan has dropped nine positions in ranking, Korea seven and China thirty-one. While it is true that employment of women is embedded within Chinese business culture, there does seem to have been an impact due to progress made in the marketization of labor accompanying the creation of a market-orientated economy [11] . In addition, Ishizuka [12] noted, particularly in cities in central China, a slight increase in unemployed women who fulfill the role of “housewives” typical of developed countries.

Figure 2 shows the GGGI in 2017 in Japan, Korea, and China in more detail. The GGGI is divided into four fields of 1) economy, 2) education, 3) health, and 4) politics. As a reference, the scores of second-ranked Norway, 49^th-ranked US, and tenth-ranked Philippines in Asia are also shown on the line graph. As a result of gender diversity management policies, the proportion of women on corporate boards in Norway has risen to about 40%², and it is exemplary in terms of actively adopting strategies that promote a work-life balance, whereby both work and the home/children are supported. Also, while it would be incorrect to say that either factor results from any active strategy on the part of its government,

the US has maintained a fertility rate of around 2.0, the level at which the population is sustained, and women themselves are choosing to work; in addition, the US ranks at the global top in terms of GDP.

The scores in Japan, Korea, and China for 2) education and 3) health are not problematic when compared to the scores of European countries and the US, with one exception³. The fields in which Japan’s GGGI ranking has fallen significantly are 1) economics and 4) politics. These are both fields that impact the actual workings of society, and there is indeed substantial gender disparity in all three countries⁴.

As it is the theme of this study, we will look in some detail at the economic elements of the index. Japan, Korea and China all fall behind Norway in the field overall. Looking at the components of the field, the largest differentials are, as already mentioned, in the female-to-male ratio of labor force participation (Labor Quantity), the female-to-male ratio of legislators, senior officials, and managers and the female-to-male ratio of professional and technical workers (Labor Quality), and the gender wage gap. Note that, although the differentials in China are higher than in Japan and Korea, care should be exercised in reading the data. This is because, of the five countries in Figure 2, only China is not an OECD member, and as many as a half the employees within the data for China overall are employed in primary industries, most notably agriculture. This means that, for China overall, the labor market is still developing. From an analytical viewpoint, this issue is resolved in principle in the next and subsequent sections by making comparisons and drawing conclusions based only on data for China’s urban areas where employed households are concentrated, to reflect the relevant situation in China.

Firstly, Figure 3 shows labor force participation for men and for women in each age group, as well as characteristics of the labor market in all three countries. The differences between the countries were smaller for men than for women, with the male labor force participation rate highest in Japan, followed by China overall, then Korea, then China’s urban areas. For women, the highest labor force participation rate was in China overall, followed by Japan, then China’s urban areas, then Korea. However, in China overall, around half of the population is involved in agriculture. This makes it more appropriate to use China’s urban areas in any comparison with Japan and Korea, both of which are developed nations.

According to Ishizuka [2] , for regular employees in Japan and Korea, the Japanese employment practices of age-based remuneration and lifetime employment are entrenched. Retirement age in Japan has in many cases moved from 60 to 65, and enterprises are obliged by legislation that deals with the employment of older people to keep on any employee who wants to keep working until the age of 65⁵. Following two years of military service while young, Korean men tend to take up employment as regular employees and work in the same job until they retire⁶. The labor market is solid, and both countries share the problem of regular

employees being in “work-type” employment for a long time. The curve of women’s participation in the labor force, however, is M-shaped, because many women stop work to care for children, and in many cases the female employees on the right-hand side of this “M” are not regular employees.

In China, the labor market is fluid as it is in Europe and the US, and “job-type” employment is common. Retirement ages do exist, however. These are usually different for men and for women, being generally set at around 50 for women and around 60 for men⁷. Thanks to equal employment for men and women that accompanied the introduction of a planned economy in 1949, it is currently normal for women in China’s urban areas to be employed⁸.

Next, to highlight trends in the male and female labor force participation rates in Japan, Korea, and China overall, Figure 4 shows trends in the labor force participation rate for only the productive population (aged 15 - 64). Limiting the data to the productive population reduces the impact of the aging population in the three countries.

The two charts show that the gender gap in the labor force participation rate in Japan and Korea has shrunk from about 25 points in 1990 to around 20 points now, as participation by women has gradually increased and participation by men has remained flat. Meanwhile, while the gender gap in labor force participation has stayed close to the 10 points seen in 1990, participation rates in China for both men and women fell after the reform of state-owned enterprises in the middle of the 1990s. Having been impacted by the economic downturn of the summer of 2008, they have increased slightly from a bottom in 2010. The current gender gap in the roughly calculated labor force participation rate has expanded to 14 points (women 70% and men 84%).

We will now outline labor market-related policies in each of the three countries.

In Japan, the Gender Equality in Employment Act was implemented in 1986. In 2003, the Act on Advancement of Measures to Support Raising the Next Generation of Children was implemented, obliging business owners and municipalities that have 301 or more regular employees to give notification of a relevant action plan. In 2011, the obligation expanded to include employers and corporations with more than 100 employees⁹. This legislation expired in March 2015, but has been extended for 10 years. In 2015, the Act of Promotion of Women’s Participation and Advancement in the Workplace was implemented, requiring enterprises with 301 or more employees to formulate a relevant action plan by April 1st, 2016. Specifically, each enterprise must: Step 1. Identify issues by analyzing 1) the proportion of women among its employees, 2) the gender gap in number of years of continuous employment, 3) working hours, 4) the proportion of women in management positions; Step 2.Formulate an action plan, file a report on it, and disseminate it within the enterprise; Step 3. Make this information widely available to the public.

In Korea, the Gender Equality in Employment Act was implemented in December 1987. Active measures to improve women’s employment (hereafter “affirmative action programs”) were added to this legislation in March 2006. When first introduced, these requirements applied to businesses and government institutions with 1000 or more regular employees, but in March 2008 they expanded to include those with 500 or more employees¹⁰. Analysis by Ishizuka [5] confirmed the impact of this initiative, with a rise in both the number of women employed and women in management positions. There was also an increase in the number of women on company boards, a factor not covered by the legislation. With anticipated further expansion of coverage to include businesses and government institutions with 100 or more employees, these positive impacts are expected to continue. The main objective of this affirmative action program is to mobilize women who are not sufficiently economic active to prevent a decline in Korea’s international competitiveness as result of the ongoing decline in the birthrate and the aging of society, similar to what is being experienced in Japan. Also, a low proportion of female workers is seen as a sign of “indirect discrimination” by the business in question, and such businesses are required to make improvements. The specific achievement target is 60% of the average ratio of regular female workers and women in management in other enterprises of the same size in the same industry. All businesses and government institutions subject to the requirements are obliged to report to the government regarding their male and female workers by type of job, position, and gender by the end of March (Step 1). If they have not achieved the criteria for female employment, they are to present a female employment improvement plan by the end of March of the following year (Step 2). The government evaluates the quality and the transitional results of the implementation plan, offers administrative and financial support to “successful” enterprises, and issues transitional guidance to unsuccessful enterprises (Step 3).

In China, in the household registration system that divides the population into two groups, employee households are classified as “non-agricultural” (urban). Reform of state-owned enterprises started in earnest in the mid-1990s but, until a “labor market” was formed, only a limited number of employees were able to move out of the region in which their household was registered and change classifications from agricultural to non-agricultural [13] . However, as the population is set to peak in around 2030, the household registration system is currently being partially relaxed. In 2015, the one-child policy was completely abolished, but the impact of this action seems limited. In practice, with the transition from a situation marked by the lingering impact of the planned economy to a market economy, the gender gap has tended to increase in terms of work-life balance, promotion at work, wages, types of jobs, and the application of new technology [11] . That is, female employment, including the employment of working mothers, is found on China’s eastern seaboard, where many major cities are located and there has been significant economic development. However, in the center of the country, where economic development is only middling, there has been a slight increase, as in Japan and Korea, in the number of housewives staying at home to look after children [12] . Western China has failed to benefit sufficiently from economic development. Therefore, in this study, the focus is on China’s urban areas and, in the empirical analysis in particular, comparisons with Japan and Korea are made using only enterprises located in the three large cities of the eastern seaboard.

Finally, falling birthrates and an aging society are having a significant impact on the labor markets in the three countries. In all three countries, the birthrate is between 1.0 and 1.5. It is highest in Japan, followed by Korea, and then China’s urban areas. There has recently been a slight increase in the birthrate in Japan, but this is mainly a blip that has appeared as second-generation baby boomers pass the age of 40 and take their last opportunity to become parents. Looking only at the urban areas of China, the reduction in the birth rate is even more significant than in Korea [14] . Economic policies must secure the income required in the face of an expansion in expenditure. Specifically, pressing problems include improving welfare benefits for the elderly and securing the labor supply (in other words, the working population) needed in the near future. Looking at the public pension system, there is universal coverage in Japan, but in Korea and China, system reform is still ongoing.

There are various debates regarding substitution and complementarity elasticity, the first stage in our analysis. Stern [15] describes five types of elasticity of substitution and five types of elasticity of complementarity and notes differences in estimation results. Yamaguchi [16] also establishes four types of substitution and complementarity elasticity and conducts relevant estimations. There are various ongoing debates but, by recommending the use of multiple models, Hicks and Allen [17] , as well as others, have further broadened the potential of elasticity analysis. Fundamental elements are Hicks’ [17] partial elasticity of substitution and Allen’s [15] partial elasticity of complementarity.

There is hardly any prior research estimating male and female labor demand using substitution and complementarity elasticity. Although Chinhui and Kim [18] did not use the elasticity analysis mentioned above, they estimated substitution effect model for gender labor supply based on wages using time series data in US. It was shown that female labor supply was substituted for less-educated male labor supply, and reducing the gender wage gap in the 1980s. However, it noted that labor demand decisions are made from the perspective of labor demand. Also, there is no evidence that an increase of educated women reduced men’s wages.

Using 2005 and 2006 data for Japanese enterprises, Yamaguchi [16] estimated Hicks’ partial elasticity of substitution and Allen’s partial elasticity of complementarity and two other elasticities for regular and non-regular employees, and found a relationship of substitution. The relationship of substitution weakened in 2006. However, interpretation of the results was based exclusively on the historical background.

In summary, it is necessary to estimate multiple elasticities including “Hicks’ partial elasticity of complementarity” and “Allen’s partial elasticity of substitution”. The analysis of substitution and complementarity reveals the relationship between capital and two types of labor, such as male and female labor demand, or regular and non-regular employees individually. To conduct a more detailed investigation, a second stage is an analysis of the wage gap.

In the first stage of analysis (substitution and complementarity), a substitution relationship is that a factor decreases (increases), when the other factor increases (decreases).And a complementarity relationship is that both factors rise or fall together. However, push factors and pull factors are assumed to make up a complementarity relationship.

A push factor is an increase in female labor demand based on an “economic rational evaluation of male and female labor as equal” that causes an increase in male labor demand based on an “economic rational evaluation of male and female labor as equal”. A pull factors is an increase in female labor demand based on an “economic irrational and discriminate evaluation of female labor than male labor”.

Therefore, in the second stage of analysis, the factor decomposition of wage gap is used to investigate which factors are applicable. In neoclassical labor economics, “discrimination” is the economic distance that exists between economic groups that possess the same productivity. In this study, the economic groups are men and women, and the economic distance is the wage gap¹². When there is a wage gap, it is possible to verify the appropriateness of the factors that cause the wage gap by clarifying whether the gender gap economic rational based on Becker’s “human capital theory” or if it is economic irrational based on Becker’s [19] “discriminatory preference theory.”¹³ Yamaguchi [20] analyzed the productivity of men and women based on “discriminatory preference theory.”

As an estimation method, Ishizuka [11] and others utilized Blinder-Oaxaca factor decomposition to clarify the impact of China’s marketization on the gender wage gap, separating the state-owned sector (representative of planned economy tendencies) from the non-state owned sector (representative of market economy tendencies), using 2004 Chinese survey data. For state-owned enterprises, this analysis revealed differences between men and women consistent with economic rationality as a factor, but not with a factor for sexual discrimination. At private businesses, which have increased in number as a result of marketization, factors were roughly evenly split between those that were discriminatory and those that were consistent with economic rationality. In this study, which is based on Neumark [19] and Oaxaca and Ransom [20] , gender wage gap factors are divided into those consistent with economic rationality and those that are discriminatory. Discriminatory factors are then further divided in two, allowing estimates regarding both underpayment of women and overpayment of men.

The model for calculating the relationship of substitution and of complementarity of the three factors of production for each country and each enterprise is outlined below. Where male regular employees = X₁, female regular employees = X₂, and capital = X₃, we have the linear homogeneous production function:

Q = f ( X 1 , X 2 , X 3 ) (1)

Further, specification with a translog production function gives:

ln Q = ln α 0 + ∑ i = 1 3 α i ln X i + 1 2 ∑ i = 1 3 ∑ j = 1 3 γ i j ln X i X j , ( i = 1 , 2 , 3 ) ( j = 1 , 2 , 3 ) ( i ≠ j ) (2)

Here, α₀, α_i, γ_ij are coefficients showing production techniques.

Assuming an enterprise determines the scale of the input of each factor of production in a way that maximizes profit Q − ∑ i = 1 3 W i W i , this gives rise to:

W i = ∂ Q ∂ X i ( i = 1 , 2 , 3 ) (3)

Here, Ｗ_i is the real cost of the factor of production. Also, taking the wages that are the cost of Ｘ₂ as the real cost of the factor of production, and assuming a perfectly competitive market, the cost share versus production volume of the factor of production i becomes:

S i = ∂ ln Q ∂ ln X i = α i + ∑ j = 1 3 γ i j ln X j , ( i = 1 , 2 , 3 ) ( j = 1 , 2 , 3 ) ( i ≠ j ) (4)

Assuming the production function has linear homogeneity:

∑ i = 1 3 S i = 1 , ∑ i = 1 3 α i = 1 , ( i = 1 , 2 , 3 ) (5)

∑ j = 1 3 γ i j = 0 　 , ( i = 1 , 2 , 3 ) ( j = 1 , 2 , 3 ) ( i ≠ j ) (6)

Here,

γ i j = γ j i (7)

is produced. In this study, based on Equation (8) and Equation (9) below, α_i, γ_i1, γ_i2, (i = 1, 2) are estimated and, based on Equation (6), γ₃₁, γ₃₂ are estimated_.

Ｃ i i = 1 S i ( S i − 1 + γ i i S i ) , ( i = 1 , 2 , 3 ) (8)

C i j = 1 S i ( S j + γ i j S i ) , ( i = 1 , 2 , 3 ) ( j = 1 , 2 , 3 ) ( i ≠ j ) (9)

It is possible to derive the value of Hicks’ [17] partial elasticity of complementarity by substituting within Equation (8) and Equation (9) the estimated value of the cost share and coefficient obtained by applying an SUR model¹⁴ to Equation (4), with Equation (7) as a constraint.

Further, using

E i j = 1 S i S j H i j | H | , H = [ 0 1 1 1 1 c 11 c 12 c 13 1 c 21 c 22 c 23 1 c 31 c 32 c 33 ] , ( i = 1 , 2 , 3 ) ( j = 1 , 2 , 3 ) ( i ≠ j ) (10)

it is possible to estimate the value of Allen’s [15] partial elasticity of substitution.

Specifically, S: the cost share numerical value, which is equal to the average wage × number of employees ÷ value added (sales basis), Q: the sales-based value added (sales ? intermediate input costs; in millions of yen in Japan, tens of thousands of won in Korea, and thousands of yuan in China), Ｗ_i: annual wages per employee(including bonuses, tax, and public pension premiums), Ｘ₁ : the number of male regular employees, Ｘ₂: the number of female regular employees, and Ｘ₃: the value of fixed assets (in millions of yen in Japan, tens of thousands of won in Korea, and thousands of yuan in China). In only model (4), Ｘ₁: the number of regular male and female employees, and Ｘ₂: the number of non-regular male and female employees.

We will start with an explanation regarding how to read Tables 2-4. Table 2 shows Hicks’ partial elasticity of complementarity C_ij calculated by substituting the values in Table 1 in Equation (8) or Equation (9). The values in Table 1 are the results of estimating a simultaneous equation of S₁ (male regular employees) and S₂ (female regular employees) with an SUR model, based on Equation (4) in this study¹⁵. A positive value signifies that the factors of production have a relationship of complementarity, while a negative value signifies a relationship of substitution. In addition, the higher the numerical value, the stronger the relationship between the two factors, and the lower the numerical value, the weaker the relationship. Table 3 shows Allen’s partial elasticity of substitution E_ij, calculated using numerical values including those in Table 2. A positive value signifies that the factors of production have a relationship of substitution, while a negative value signifies a relationship of complementarity. Thus, the same result is obtained if the signs of C and E are reversed, which implies that the estimations have been produced in a consistent manner. The interpretations in this study are, in principle, based on C. Table 4 is limited to the estimation results of the substitution/complementarity relationship of male and female labor demand, which have been extracted from Table 2 and Table 3.

In Table 4, all values of C are positive and all values of E are negative, confirming a relationship of complementarity in all three countries between male and female labor demand. The two models give consistent results. In other words, if the labor demand of female regular employees increases, so does the labor demand of male regular employees. Seen overall, C is highest in China and

Data Source: “Corporate Survey on the Personnel Utilization of Men and Women (Japan 2015)”; “Corporate Survey on the Personnel Utilization of Men and Women (China, Rep. of Korea) 2013” by RIETI. Note 1. Based on Equation (4) in this paper, the values in Appendix Table 1, which are the results of estimates from the simultaneous equations for male regular employees S₁ and female regular employees S₂ using the SUR model, have been calculated by substituting into Equation (8) or (9).

Data Source: Same as Table 2. Note 1. Using the values in Table 2, results have been calculated by substituting into Equation (10). As in Table 2, results are estimated by industry and enterprise scale. However, only the overall estimation results are displayed in this paper. Other estimations have generally been in line with the results from Table 2.

E is highest in Korea. However, the largest and second largest values occur in China and Korea in both C and E.

Looking at the numerical value of C, the strongest relationship of complementarity is in China, with an overall value of 0.983, with the relationship particularly strong for large enterprises (1.075) and tertiary sectors (0.948). The next strongest relationship is in Korea, with an overall value of 0.941, with the relationship particularly strong in medium enterprises (0.952) and secondary sectors

Data Source: Same as Table 2. Note 1. Excerpted from Table 2 and Table 3. Note 2. Significance is same as significance using the SUR model in Appendix Table 1. †: 0.1%; ***: 1%; **: 5%; *: 10%.

(0.960). Of the three countries, the relationship was weakest in Japan, with an overall value of 0.937. The figures do not vary much, but the overall figure and the figures by sector and by size of enterprise, all show the same order: first China, followed by Korea, the Japan. This is likely because, although it has risen gradually (Figure 2), China has the smallest gender gap among the three countries, China’s GGGI ranking is the highest among the three countries (Figure 1), and the labor market structure in China has historically normalized the labor demand of women (see Subsection 2.2 of this study). Korea ranks next likely because of the positive impact of the introduction of measures to promote expansion in female labor demand (Ishizuka [5] ). These measures were introduced before similar measures were introduced in Japan. The figures for Japan have probably been enhanced by the promotion of policies for the expansion of labor demand as part of the “Promotion of Women” since 2012, and also by the underlying strength of the economy in 2015 when the survey data was collected (Ishizuka [4] , [9] ).

Estimation results of model (4) in Table 4 confirm a relationship of complementarity between regular and non-regular employees in Japan. The numerical value for the overall relationship of complementarity C is 0.866. In other words, when the labor demand of regular employees increases, so does the labor demand of non-regular employees. The context includes an expansion in demand for both categories of labor thanks to relative scarcity in the Japanese labor market in 2015 in comparison to the previous year, and the continuing increase in the labor demand of non-regular employees, more of whom are women than men. Yamaguchi [16] found a declining trend for the relationship of substitution between the two in 2006. The results in this study are from ten years later, 2015, and are likely to have been impacted by the implementation of policies to reform working practices since 2012 and by the strength of the economy.

In addition, Table 2 and Table 3 show the relationship between demand for male and female labor and capital in the three countries. In China, as seen in Table 2, a relationship of substitution between male labor and capital and between female labor and capital was confirmed overall and for all sectors and all enterprise sizes. A substitution relationship between female labor demand and capital in China can also be derived from Table 3. For Korea, the overall and medium-sized enterprise values in Table 2, as well the results in Table 3, confirmed a relationship of substitution between male labor and capital and between female labor and capital. In Japan, for all values in Table 2, and for female labor and capital in Table 3, a relationship of substitution was confirmed. The background to this is likely to be that, in China, there is still a high volume of supply of labor including farmers (migrant workers), allowing the development of labor-intensive operations. In Japan, the rapid decline in the population has encouraged business development through the investment of capital. Korean businesses can be considered to engage in a relatively diverse range of operations.

In this section, a relationship of complementarity was verified between increases in female labor demand and increases in male labor demand for all three countries. In the next section, the factors within the relationship of complementarity are verified using factor decomposition of the gender wage gap.

In the previous section, a relationship of complementarity was derived for male labor demand and for female labor in the three countries. A relationship of complementarity means that an increase in female labor demand is accompanied by an increase in male labor demand. However, within the relationship of complementarity, as shown in Subsection 4.1, both push factors and pull factors are involved. Push factors are those in which female labor demand increases due to the equal evaluation of men and women, consistent with economic rationality. Pull factors are those which increase female labor demand due to treatment of women in an economic irrational manner that is discriminatory compared to that of men.

To resolve this point, earnings functions for men and women were estimated, and factor decomposition of the gender wage gap was carried out. Estimation is based on Neumark [19] and Oaxaca and Ransom [20] , with the factors within the gender wage gap divided into elements consistent with economic rationality and discriminatory elements (underpayment of women and overpayment of men).

First, the Mincer earnings function as used in general gender data was estimated for men and for women for a situation without discrimination:

ln W i = X i B i , ( i = m , f , a l l ) (11)

Here, lnW is the natural logarithm of hourly wages. On the basis of “human capital theory,” X is the number of graduates of a four-year university course or higher, the number of years of continuous employment, and age. It also includes two dummies to represent whether an enterprise belongs to a tertiary sector and is a large enterprise, distinctions made using corporate attributes. Within i, m represents male, f represents female, and all the combined male and female figure.

In addition:

L n ( W m − W f ) = X m ( B m − B * ) ＋ X f ( B * − B f ) ＋ ( X m − X f ) B * (12)

Here, the first term on the right of the equation represents “overpayment of men,” and the second “underpayment of women,” the third represents “the gap between men and women that is consistent with economic rationality.”

As mentioned in Subsection 4.1, the third term represents the gender wage gap consistent with economic rationality, that is, the gap between the marginal productivity of men and of women due to differences in attributes such as number of years of continuous employment. This term is based on “human capital theory.” The first and second terms represent the gender gap that is not consistent with economic rationality, based on Becker’s [19] “discriminatory preference theory.” This is the gender wage gap attributable to “discrimination.”At enterprises in Japan and Korea in particular, attempts to reduce costs result in a difference between marginal productivity and marginal wage rate, and “overpayment of men” (the first term) and “underpayment of women” (the second term) are generated.

The coefficients resulting from the estimation signify that, when X increases by one unit, W wages increase [B × 100]%. For statistical significance, only results from the third decimal place are taken into account.

Table 5 shows that, among Japanese regular employees, men’s wages rise as their number of years of continuous employment increases, and women’s wages are higher for older women, women at large enterprises, and women in the service sector. Among Korean regular employees, both men and women see their wages rise as their number of years of continuous employment increases, but there is a negative correlation between age and wages. For women, wages are higher in the service sector. Among Chinese regular employees, for both men and women, wages are pushed up by length of labor demand and by working in the service sector. For women, wages are higher at medium-sized enterprises.

Table 6 shows the estimation results of factor decomposition of the gender wage gap. Simple average values for the gender wage gap are, in ascending order, 6.0% in China, 16.2% in Korea, and 17.0% in Japan. Looking in more detail at the figures reveals a large contribution to the gap from the “discriminatory element” (after deduction of the economic rational contribution from the gender

Note 1. The natural log of the hourly wage has been estimated as the explained variable using OLS. Note 2. p values for statistical significance are as follows: †: 0.1%; ***: 1%; **: 5%; *: 10%.

gap in marginal productivity) of, in ascending order, 65.1% in China, 95.1% in Korea, and 98.4% in Japan. In detail, “underpayment of women” is 61.7% and “overpayment of men” is 3.4% in China, with the equivalent figures, 68.8% and 26.3% in Korea, 77.3% and 21.2% in Japan.

In Japan, the gender gap in average wages for regular employees is relatively large at 17.0%. The most significant factor expanding the gender wage gap based one conomic irrational discrimination is the 0.40 gender wage gap attributable to number of years of continuous employment. Meanwhile, it seems that the most significant factor reducing the gender wage gap (at −1.79) is the system whereby wages increase as employees get older. Indeed, the breakdown of the age contribution shows significant negative contributions from underpayment of women (−1.47) and from overpayment of men (−0.34). In addition, large enterprises and tertiary sectors reduce the discriminatory impact of the underpayment of women.

In Korea, the biggest contributor to the gender wage gap (at 0.23) is the number of years of continuous employment, and it is clear that underpayment of women is particularly important (0.12). On the other hand, economic rational causes of disparity are similar to those in Japan.

In China, the gap in average wages themselves is very small at 6.0%, and even components such as the underpayment of women in Table 6 have low values. Also, as mentioned in the previous section, historically, the structure of the Chinese labor market has normalized the female labor demand (Subsection2.2 of this study), further confirming that the theory being discussed is not applicable to China.

In summary, in the same order as in the substitution and complementarity analysis in the previous section, the contribution from the economic irrational discriminatory component is largest in Japan, followed by Korea, and then China. Also, the contribution from the underpayment of women is significant. The proportion of the gender wage gap attributable to discrimination with no basis in economic rationality is 98.4% for regular employees at Japanese enterprises and 95.1% at Korean enterprises. This study found that this was largely because, consistent with “discriminatory preference theory,” women are employed at wages lower than implied by marginal productivity in human capital theory. This is either due to cost-cutting or, in Japan and Korea, because of female labor demand policies.

In other words, the labor demand of male and female regular employees expands in a complementary manner. However, this is largely attributable to an increase in labor demand for male regular employees causing an increase in labor demand for female regular employees under conditions in which female regular employees are evaluated in an economic irrational, discriminatory fashion in comparison to men (the pull factor in Subsection 4.1). According to Ishizuka [6] , raising the proportion of regular employee management roles held by women to above 40% would eliminate this discriminatory component from the labor market¹⁶.