Abstract

This paper examines the determinants of the demand for female workers, focusing on the role of information and communication technology (ICT) and offshoring. Estimating a system of variable factor demands for manufacturing industries between 1980 and 2011, we find that, whereas the ICT capital stock has significantly positive effects on the demand for low-, middle-high-, and high-skilled female workers, it has significantly negative effects on the demand for middle-low-skilled female workers. In contrast, offshoring has insignificant effects on the demand for female workers, which suggests that offshoring is at least neutral on the demand for female workers.

1.  Introduction

Sustaining the labor force is one of the main concerns for many developed countries. Accordingly, raising female labor participation is becoming an issue for policymakers because it can help avoid looming labor supply shortages. In Japan, the focus of this study, female labor participation is a key policy agenda for the current administration. In 2014, the female labor participation rate, at 71.8 percent for ages 25–54 years, was 20 percentage points below that of men and was ranked 24 out of 33 OECD countries (OECD 2015a, Table B).1 According to OECD (2015b) estimates, “if the female participation rate were to converge to those of men by 2030, the labor supply would decline by only 5 percent, increasing GDP by almost 20 percent compared with unchanged participation rates” (p. 15). Indeed, Japanese Prime Minister Shinzo Abe places priority on female labor participation as an important policy agenda item, which has been called Abe's Womenomics.2

Our study is closely related to a number of research strands that have focused on the issue of female labor participation in Japan. The first strand studies the supply side, asking what factors hinder female labor participation. For example, Akabayashi (2006) revealed that spousal deductions have a significant and large effect on labor supply by married women, which suggests that the tax system affects the married women's labor market decisions.

The second strand of literature asks what is responsible for the large wage gap between female and male workers in Japan. For example, Miyoshi (2008) estimates wage functions for female and male workers in Japan. His results show that the male–female gap in experience as a full-time worker is one of the most important factors that helps to explain Japan's gender wage gap.

These lines of research have made significant contributions to the literature. However, none of these studies have focused on the effects of skill-biased technological change, including the use of the information and communication technology (ICT) such as computers and other high-tech equipment, as factors that potentially influence female labor participation. Skill-biased technology may be relevant, because it often makes it possible to replace physically demanding tasks with less physically demanding tasks. This implies that the demand for female workers relative to male workers may increase through technological change. Indeed, Weinberg (2000) found that increases in computer use accounted for more than half of the growth in demand for female workers in the case of the United States.3

Note also that globalization could narrow the gender wage gap, as is noted in the third strand of research. For example, if increased product market competition should drive out costly discrimination (Becker 1957), then increased competition from foreign countries should similarly reduce firm market power in an industry, thereby creating pressures for reduced gender discrimination. If so, increased international competition will lead to increases in the relative wages and employment of female workers in the industries that face increased international competition. Indeed, several empirical studies have found that international trade has led to a narrowing the gender wage gap.4 These studies, however, did not consider the effects of offshoring nor those of ICT, even though both offshoring and ICT have expanded in the last three decades.5

The fourth strand of the literature examines the effects of ICT and offshoring on skill demand. As is well known, the effects of offshoring are qualitatively similar to the effects driven by skill-biased technological change.6 As a result, increased demand for female workers may be attributable to either skill-biased technological change or offshoring (or both). Determining which of these explanations accounts for the changes is an empirical question. Therefore, a number of studies, such as Hijzen, Görg, and Hine (2005) and Kiyota and Maruyama (2017a), have examined the effects of ICT and offshoring on the skill demand in the trade literature. Interestingly, however, none of the studies on ICT and offshoring has taken into account differential effects on female and male workers.

Building on the previous literature, this paper studies the effects of ICT and offshoring on female labor demand in Japan's manufacturing industries.7 One of the contributions of this paper is thus to shed light on the role of ICT and offshoring as determinants of female labor demand. To address this issue, we estimate a system of labor demands, controlling for the effects of skill-biased technological change and offshoring simultaneously. This framework was first proposed by Hijzen, Görg, and Hine (2005), who examined skill demand in the United Kingdom. Following their methodology, our companion paper, Kiyota and Maruyama (2017a), examined overall developments in skill demand in Japan. The current paper extends this work by focusing on the demand for female workers.

Why is it important to examine the determinants of skill demand by gender? One key reason is the difference in the distribution of skill types for male and female workers. For example, the share of female workers among part-time workers was 73 percent in 1988 at its peak, and 57 percent in 2011 (see Figure 3 later in this paper). The female share in this category is much higher than it is for any other worker type. Moreover, industries with a higher share of part-time workers tend to have a higher share of female workers. To illustrate this correlation, Figure 1 displays the relationship between the share of female workers and the share of part-time workers by industry in 2011. The correlation coefficient for these two variables was 0.687, which suggests that labor demand for part-time workers may reflect labor demand for female workers. Therefore, we need to separate effects on labor demand by gender as well as by skill.8

Figure 1.

The share of female workers and the share of part-time workers, 2011

Figure 1.

The share of female workers and the share of part-time workers, 2011

The rest of the paper is organized as follows. Section 2 describes the empirical framework. Section 3 describes the data used in this paper. Section 4 presents the estimation results, and Section 5 presents a summary and concluding remarks.

2.  Econometric methodology

2.1  Model

Let i be the index of industry (i = 1,, N); j be the index of factor (j = 1,, J); k be the index of fixed input or output (k = 1,, K); and r be the index of the proxy for technological change (r = 1,, R). For the ease of presentation, we omit time subscript t, unless otherwise noted. Assume that the industry cost function can be represented by a translog function, which is twice differentiable, linearly homogenous, and concave in factor prices. The cost function for industry i, , can be represented as follows:
formula
1
where is factor price for factor j in industry i; is fixed input or output k in industry i; and is exogenous factor r such as technological change.9
Let represent variable input j in industry i. Differentiating the translog cost function with respect to factor prices gives us the cost share of factor j in total variable costs. Adding time subscript t and error term and taking into account the industry-factor specific fixed effects provides us with our regression specification, which is written as:10
formula
2
where and .

Following Hijzen, Görg, and Hine (2005) and Kiyota and Maruyama (2017a), we treat both labor inputs and intermediate inputs as variable factors. The detailed classification of labor inputs is defined in the next section. For fixed inputs , we use the non-ICT capital stock. For , we utilize offshoring and the ICT capital stock.11 One way to control for the supply-side effects such as the declining supply of junior-high school graduates is through the use of instrumental variable methods. It is difficult, however, to use such a method in the seemingly unrelated regressions (SUR) model that we use in this analysis. We therefore include a full set of time dummies to capture the economy-wide technological change over time, and a factor-specific time trend to control for supply side effects on factor j.12

2.2  Elasticities

Without loss of generality, we remove the industry subscript i and the time subscript t for ease of exposition. The cost share for factor j is written as . The elasticity of factor demand j with respect to a change in factor prices is:
formula
3
where if j = s and if j ≠ s, , and is an estimated parameter value in equation (2). The elasticity of factor demand j with respect to a change in non-ICT capital stock or output is:
formula
4
where and is an estimated parameter value in equation (2). The elasticity of factor demand with respect to skill-biased technological change due to offshoring is:13
formula
5
where and is an estimated parameter value in equation (2).

3.  Trends in the labor market, ICT, and offshoring in Japan

3.1  Data

Outputs and inputs

Data on outputs, inputs, and their prices are obtained from the Japan Industrial Productivity Database 2014 (JIP Database 2014). The database runs annually from 1970 to 2011, consisting of 52 manufacturing and 56 nonmanufacturing industries.14 From the 2014 JIP database, we collect information on the use gross outputs, ICT capital stock, non-ICT capital stock, intermediate inputs, labor inputs, and labor costs.15 In the JIP database, the ICT capital stock is estimated according to OECD guidelines. The ICT capital stock consists of 39 assets such as electric computing equipment, wired and radio communication equipment, and applied electronic equipment.16 Other assets are classified as non-ICT capital stock. All of these variables are valued at constant prices (for the year 2000). We also obtain nominal intermediate inputs from the database to compute the cost shares.

The labor inputs are classified into one of six categories: (1) university graduates or higher; (2) college graduates; (3) high school graduates; (4) junior high school graduates; (5) part-time workers; and (6) self-employed workers.17 The educational level of the last two categories is not available. We exclude self-employed workers from the analysis because they are employers rather than employees. It is not clear whether the demand for employers can be estimated in the same framework as the demand for employees. In the JIP database, part-time workers are defined as workers who work fewer than 35 hours per week on average.18 Labor costs are calculated as the sum of monthly wages plus bonuses. Average wages are obtained by dividing labor costs by the product of the number of workers and hours worked.19

Figure 2 presents the shift in average wages for five worker categories for female workers and for male workers. From this figure, it is obvious that wages tend to be higher for more highly educated workers. At the same time, Figure 2 indicates that wages for part-time workers are the lowest for both genders. In addition, some differing trends are observed between male and female workers. Among female workers, there are clear wage gaps between worker categories. Further, wages of male college, high school, and junior high school graduates stay at the same level in the long term, although the wages of male junior high school graduates begin to decline in the 2000s. In light of these findings, we use separate classifications of labor by gender.

Figure 2.

Average wages by education in Japanese labor markets, 1970–2011: (a) Female workers (b) Male workers

Figure 2.

Average wages by education in Japanese labor markets, 1970–2011: (a) Female workers (b) Male workers

Both male and female workers are classified into the following four skill groups: (1) high-skilled workers; (2) middle-high-skilled workers; (3) middle-low-skilled workers; and (4) low-skilled workers. Our classification is based on wage levels, under the assumption that workers with higher skills can obtain higher wages. In forming groups (3) and (4), the same definitions are applied to both genders: middle-low-skilled workers are defined as junior high school graduates, and low skilled workers are defined as part-time workers.20 Different categories are applied to men and women in the formation of labor groups (1) and (2), however. For male workers, university graduates are classified into high-skilled workers, whereas college and high school graduates are classified into middle-high skilled workers. This classification is based on the wage differences shown in Figure 2b.

For female workers, university and college graduates are classified into high-skilled workers, while high school graduates are classified into middle-high-skilled workers. The reason for combining female university and college graduates is two-fold. First, the share of female university graduates is very small for the creation of a group in the 1980s; under this definition only 0.9 percent of female workers were included in this group in 1980, although it increased to 6.8 percent in 2011. Second, we emphasize the differential job assignments by educational background. According to the School Basic Survey conducted by the Ministry of Education, Culture, Sports, Science, and Technology in 1980 for all industries, 12.9 percent of female high school graduates were involved in the manufacturing process when they found work after finishing school, whereas 0.4 percent of college graduates and 0.6 percent of university graduates were involved in the same process.21 This structure remained unchanged in 2011; the share of female workers involved in the manufacturing process included 22.7 percent of high school graduates, 0.8 percent of college graduates, and 0.2 percent of university graduates. The fact that high school graduates tend to be more involved in the manufacturing process than university or college graduates indicates that there is a qualitative difference between high school graduates and college/university graduates. Indeed, compared with our companion study (Kiyota and Maruyama 2017a), by distinguishing activities by gender, this paper gains precision in its accounting of the relationship between job skills and education levels.

Figure 3 provides information on the share of female workers by skill group. Three aspects of the figure stand out. First, the share of female workers in total workers does not necessarily increase over the period. The female share of the workforce started at 33 percent in 1970 and increased to 36 percent in 1990. It then gradually decreased to 29 percent in 2011. Second, the share of low-skilled female workers increased until early 1990s and then declined afterwards, with some fluctuation along the way.22 The share of female workers was 65 percent in 1970, 71 percent in 1990, and 57 percent in 2011. Finally, the share of high-skilled female workers increased steadily, although it showed a slight decline in the late 2000s. The share of female high-skilled workers increased from 10 percent in 1970 to 24 percent in 2004 at its peak and then slightly declined to 21 percent in 2011.

Figure 3.

Share of female workers, by education and worker type: Manufacturing in Japan, 1970–2011

Figure 3.

Share of female workers, by education and worker type: Manufacturing in Japan, 1970–2011

Figure 3 indicates that the changes in the share of female workers were modest, although the skill structure of female employees changed significantly over the period. Figure 4a indicates the disparities in average female wages per hour across skill groups in 1980 and 2011. The share of each group in the female labor force is represented by a length of the horizontal line segment, and the average hourly wage is indicated the vertical position.

Figure 4.

Average wages in Japanese labor markets, 1980 and 2011: (a) Female workers (b) Male workers

Figure 4.

Average wages in Japanese labor markets, 1980 and 2011: (a) Female workers (b) Male workers

We highlight two results from Figure 4a. First, the shares and the increases for high- and low-skilled female workers are remarkably large. The share of high-skilled workers grew from 3.9 percent to 16.0 percent between 1980 and 2011, and that of low-skilled workers grew from 17.5 percent to 36.5 percent. Moreover, in 2011, the share of low-skilled workers was more than twice as much as that of high-skilled workers. This suggests that the size of low-skilled part-time workers is not inconsequential in the employment of female workers in Japanese manufacturing. Excluding this group makes it difficult for us to describe the overall picture of manufacturing employment. Second, the wage level for low-skilled workers is low. Although part-time workers have different educational backgrounds, the wage level is the lowest among these four categories.

Figure 4b presents the same figure as Figure 4a for the case of male workers. Similar to the profile for female workers, we note that overall shares and increases in shares over time for high- and low-skilled workers are remarkably high, although the share of low-skilled workers is much smaller for male workers than for female workers.

Offshoring

The degree of offshoring is computed using import-use matrices from manufacturing industry input–output tables between 1980 and 2011.23 The input–output table is available every five years between 1980 and 2005, and in 2011.24

There are two types of offshoring in the literature. One is the narrow offshoring and the other is the broad offshoring . Narrow offshoring is defined as the imported intermediate inputs in an industry i from the same industry (which corresponds to diagonal terms of the import-use matrix) divided by industry i’s tradable intermediate inputs :
formula
6
where stands for imported intermediate inputs in industry i in year t only; and is intermediate inputs from industry j to industry i in year t.25 Tradable intermediate inputs include both domestic and imported intermediate inputs from agricultural and manufacturing industries.26 Feenstra and Hanson (1999) referred to this measure of offshoring as the narrow measure of offshoring.27
The broad measure is defined as all the imported intermediate inputs in an industry i divided by the industry i’s total tradable intermediate inputs:
formula
7
Feenstra and Hanson (1999) prefer the narrow measure to the broad measure because the essence of fragmentation, which necessarily takes place within the industry, is closer to the narrow measure than to the broad measure. In the baseline model of our analysis, we utilize the narrow definition of offshoring. In Section 4.2, we also use the broad measure to check the robustness of our results.

3.2  Descriptive statistics

Table 1 reports summary statistics for the labor market and production data for 1980–2011. Rows of “Average cost shares” illustrate the difference in cost shares between female and male workers by skill-level, and intermediate inputs at the industry level. The information from 52 manufacturing industries from 1980 to 2011 indicates that, on average, intermediate inputs contribute the largest cost shares (), accounting for 77.3 percent of total variable costs. Second, the cost share for labor inputs vary across groups, ranging from 0.3 percent for low-skilled male workers (Male ) to 9.1 percent for middle-high-skilled male workers (Male ). Third, the cost share for female workers is smaller than that of male workers in each skill type, except for the case of low-skilled workers.

Table 1.
Average cost shares and annual percentage change, 1980–2011
Female workersMale workers
MeanStd. dev.MinMaxMeanStd. dev.MinMax
Average cost shares         
sH 0.007 0.006 0.000 0.047 0.047 0.027 0.003 0.150 
sMH 0.021 0.015 0.000 0.085 0.099 0.046 0.006 0.253 
sML 0.009 0.011 0.000 0.070 0.037 0.030 0.001 0.133 
sL 0.005 0.005 0.000 0.025 0.003 0.002 0.000 0.011 
Annual change of cost shares   
sH 0.000 0.000 0.000 0.001 0.001 0.001 0.000 0.003 
sMH 0.000 0.000 0.000 0.002 0.001 0.001 −0.001 0.004 
sML −0.001 0.000 −0.002 0.000 −0.002 0.001 −0.004 0.000 
sL 0.000 0.000 0.000 0.001 0.000 0.000 0.000 0.000 
Annual change of input quantities   
LH 0.032 0.023 −0.052 0.090 0.010 0.021 −0.044 0.075 
LMH −0.008 0.021 −0.089 0.041 0.001 0.021 −0.052 0.062 
LML −0.085 0.022 −0.177 −0.044 −0.073 0.022 −0.127 −0.011 
LL 0.008 0.020 −0.042 0.052 0.025 0.026 −0.046 0.112 
Annual change of flexible factor prices   
wH 0.024 0.008 −0.012 0.046 0.018 0.007 −0.013 0.038 
wMH 0.022 0.008 −0.011 0.049 0.017 0.007 −0.018 0.035 
wML 0.020 0.008 −0.014 0.054 0.013 0.009 −0.026 0.032 
wL 0.018 0.008 −0.009 0.042 0.019 0.008 −0.010 0.046 
  Intermediate inputs         
  Mean Std. dev. Min Max     
Average cost shares of sM 0.773 0.098 0.509 0.988     
Annual change of cost shares sM −0.001 0.001 −0.004 0.003     
Annual change of input quantities of M 0.004 0.025 −0.040 0.077     
Annual change of flexible factor prices of pM −0.001 0.012 −0.041 0.028     
 Annual change of fixed input and   
 output quantities   
  Mean Std. dev. Min Max     
ICT capital stock 0.088 0.025 0.011 0.143     
Non-ICT capital stock 0.022 0.019 −0.022 0.094     
Output 0.021 0.065 −0.093 0.282     
Female workersMale workers
MeanStd. dev.MinMaxMeanStd. dev.MinMax
Average cost shares         
sH 0.007 0.006 0.000 0.047 0.047 0.027 0.003 0.150 
sMH 0.021 0.015 0.000 0.085 0.099 0.046 0.006 0.253 
sML 0.009 0.011 0.000 0.070 0.037 0.030 0.001 0.133 
sL 0.005 0.005 0.000 0.025 0.003 0.002 0.000 0.011 
Annual change of cost shares   
sH 0.000 0.000 0.000 0.001 0.001 0.001 0.000 0.003 
sMH 0.000 0.000 0.000 0.002 0.001 0.001 −0.001 0.004 
sML −0.001 0.000 −0.002 0.000 −0.002 0.001 −0.004 0.000 
sL 0.000 0.000 0.000 0.001 0.000 0.000 0.000 0.000 
Annual change of input quantities   
LH 0.032 0.023 −0.052 0.090 0.010 0.021 −0.044 0.075 
LMH −0.008 0.021 −0.089 0.041 0.001 0.021 −0.052 0.062 
LML −0.085 0.022 −0.177 −0.044 −0.073 0.022 −0.127 −0.011 
LL 0.008 0.020 −0.042 0.052 0.025 0.026 −0.046 0.112 
Annual change of flexible factor prices   
wH 0.024 0.008 −0.012 0.046 0.018 0.007 −0.013 0.038 
wMH 0.022 0.008 −0.011 0.049 0.017 0.007 −0.018 0.035 
wML 0.020 0.008 −0.014 0.054 0.013 0.009 −0.026 0.032 
wL 0.018 0.008 −0.009 0.042 0.019 0.008 −0.010 0.046 
  Intermediate inputs         
  Mean Std. dev. Min Max     
Average cost shares of sM 0.773 0.098 0.509 0.988     
Annual change of cost shares sM −0.001 0.001 −0.004 0.003     
Annual change of input quantities of M 0.004 0.025 −0.040 0.077     
Annual change of flexible factor prices of pM −0.001 0.012 −0.041 0.028     
 Annual change of fixed input and   
 output quantities   
  Mean Std. dev. Min Max     
ICT capital stock 0.088 0.025 0.011 0.143     
Non-ICT capital stock 0.022 0.019 −0.022 0.094     
Output 0.021 0.065 −0.093 0.282     

Source:JIP database 2014.

Note:N = 364 for average cost shares. N = 54 for annual changes.

Table 1 also presents information on the average annual changes in input and output quantities and prices between 1980 and 2011. Two messages stand out from the rows labeled “Annual changes.” First, the cost shares are fairly stable over the sample period. The annual average change was less than 0.5 percent point for all the cost shares. This result is similar to that in the United Kingdom reported in Hijzen, Görg, and Hine (2005). Second, however, some of the input quantities and flexible factor prices indicate large change. For example, the demand for high-skilled female workers grew at 3.2 percent per year whereas the demand for middle-low-skilled female workers declined at 8.5 percent per year. The average wage grew at around 2.4 percent for high-skilled female workers whereas it grew at 1.7 percent for high-skilled male workers. As a result, as we confirmed in Figure 2, the gender wage gap of high-skilled workers decreased from 1980 to 2011.

Table 2 presents descriptive statistics for narrow and broad offshoring, and the share of the ICT capital stock.28 ICT capital share is the share of ICT capital stock to total capital stock. We highlight three findings. First, narrow offshoring increased steadily from 2.1 percent in 1980 to 5.7 percent in 2005, although it declined slightly to 5.3 percent in 2011. This result implies the increasing importance of offshoring from the mid 1980s. Second, broad offshoring shows a slightly different trend from narrow offshoring. It increased steadily throughout the period. This suggests that the measurement of offshoring may affect the estimation results. In Section 4.2, we examine how the results are sensitive to the measurement of the offshoring.

Table 2.
Offshoring and ICT, 1980–2011
Offshoring
NarrowBroadICT capital share
1980 0.021 0.074 0.024 
1985 0.024 0.079 0.050 
1990 0.030 0.091 0.071 
1995 0.042 0.103 0.082 
2000 0.047 0.124 0.104 
2005 0.057 0.154 0.127 
2011 0.053 0.163 0.150 
Offshoring
NarrowBroadICT capital share
1980 0.021 0.074 0.024 
1985 0.024 0.079 0.050 
1990 0.030 0.091 0.071 
1995 0.042 0.103 0.082 
2000 0.047 0.124 0.104 
2005 0.057 0.154 0.127 
2011 0.053 0.163 0.150 

Source:JIP database 2014 and Ministry of Internal Affairs and Communications (various years).

Note:ICT capital share is the share of ICT capital stock to total capital stock.

Finally, the share of the ICT capital stock to total capital stock increased remarkably. The ICT capital share increased from 2.4 percent in 1980 to 15.0 percent in 2011. Because the ICT capital stock and offshoring as a whole increased over the period, the increase in the demand for high-skilled female workers can be explained by offshoring or skill-biased technological change (or both). At the same time, the correlation coefficient between ICT capital stock and offshoring by industry is 0.176. This suggests that the correlation between the two variables is low, and it is unlikely that there is a problem caused by multicollinearity. We now turn to the econometric analysis.

4.  Results and discussion

4.1  Estimation results

This section investigates how ICT and offshoring affected female labor demand. To begin, we estimate equation (2) as a baseline model. Equation (2) is estimated separately for each skill type. Therefore, there are eight equations to be estimated (for high-, middle-high-, middle-low-, and low-skilled workers for each gender), though the demand for each type of workers could be simultaneously determined. Indeed, the Breusch–Pagan test rejects the null hypothesis that the error terms across equations are contemporaneously uncorrelated. Because the error terms are correlated across equations, it is better to use the seemingly unrelated regressions (SUR) model to estimate the system of equations. We also test the null hypothesis that the industry-factor fixed effects equal zero. Because this hypothesis is rejected at the 1 percent level in all equations, it is important to control for the unobserved industry heterogeneity in our regression analysis.

Table 3 presents the estimation results. Note that symmetry constraints imply that wage coefficients across the diagonal are the same. Table 4 reports the own price elasticities, for ICT capital, and those for offshoring. Examining Table 4, we highlight three results. First, own price elasticities are negative or insignificant in general. This means that a necessary (but not sufficient) condition for concavity in factor prices is satisfied.

Table 3.
Baseline results: Coefficients
Female workersMale workers
Fixed effects seemingly unrelated regressionLHLMHLMLLLLHLMHLMLLL
Female         
wH 0.014*** 0.010* −0.009** 0.001 0.014*** −0.041*** 0.016*** −0.003** 
 (0.005) (0.005) (0.004) (0.002) (0.005) (0.007) (0.006) (0.001) 
wMH 0.010* −0.007 0.004 −0.008*** 0.002 0.005 −0.010 0.000 
 (0.005) (0.010) (0.005) (0.003) (0.006) (0.009) (0.006) (0.002) 
wML −0.009** 0.004 −0.001 0.002 −0.012** 0.011 0.002 −0.001 
 (0.004) (0.005) (0.005) (0.002) (0.005) (0.007) (0.006) (0.001) 
wL 0.001 −0.008*** 0.002 0.005** 0.008*** −0.002 −0.003 −0.004** 
 (0.002) (0.003) (0.002) (0.002) (0.002) (0.004) (0.003) (0.002) 
Male         
wH 0.014*** 0.002 −0.012** 0.008*** 0.059*** −0.034*** −0.014 0.005*** 
 (0.005) (0.006) (0.005) (0.002) (0.010) (0.011) (0.009) (0.002) 
wMH −0.041*** 0.005 0.011 −0.002 −0.034*** 0.077*** 0.021 −0.002 
 (0.007) (0.009) (0.007) (0.004) (0.011) (0.020) (0.014) (0.003) 
wML 0.016*** −0.010 0.002 −0.003 −0.014 0.021 −0.001 0.001 
 (0.006) (0.006) (0.006) (0.003) (0.009) (0.014) (0.016) (0.002) 
wL −0.003** 0.000 −0.001 −0.004** 0.005*** −0.002 0.001 0.005*** 
 (0.001) (0.002) (0.001) (0.002) (0.002) (0.003) (0.002) (0.002) 
ICT capital 0.001* 0.005*** −0.006*** 0.001*** −0.002 0.009*** −0.006*** 0.000** 
 (0.001) (0.001) (0.001) (0.000) (0.002) (0.003) (0.002) (0.000) 
Non-ICT capital −0.001 −0.004** 0.007*** −0.000 −0.000 −0.025*** 0.010*** −0.001* 
 (0.001) (0.002) (0.002) (0.001) (0.003) (0.004) (0.003) (0.000) 
Offshoring 0.012 0.014 −0.012 0.006* 0.034** 0.016 −0.000 0.000 
 (0.008) (0.010) (0.010) (0.003) (0.017) (0.028) (0.021) (0.002) 
Output −0.001 −0.004*** −0.002** −0.001*** 0.004** −0.004 −0.002 −0.000 
  (0.001) (0.001) (0.001) (0.000) (0.001) (0.002) (0.002) (0.000) 
N 364 364 364 364 364 364 364 364 
R2 0.915 0.975 0.908 0.959 0.981 0.985 0.956 0.939 
Year FE Yes Yes Yes Yes Yes Yes Yes Yes 
Industry FE Yes Yes Yes Yes Yes Yes Yes Yes 
Factor-specific time trend Yes Yes Yes Yes Yes Yes Yes Yes 
Female workersMale workers
Fixed effects seemingly unrelated regressionLHLMHLMLLLLHLMHLMLLL
Female         
wH 0.014*** 0.010* −0.009** 0.001 0.014*** −0.041*** 0.016*** −0.003** 
 (0.005) (0.005) (0.004) (0.002) (0.005) (0.007) (0.006) (0.001) 
wMH 0.010* −0.007 0.004 −0.008*** 0.002 0.005 −0.010 0.000 
 (0.005) (0.010) (0.005) (0.003) (0.006) (0.009) (0.006) (0.002) 
wML −0.009** 0.004 −0.001 0.002 −0.012** 0.011 0.002 −0.001 
 (0.004) (0.005) (0.005) (0.002) (0.005) (0.007) (0.006) (0.001) 
wL 0.001 −0.008*** 0.002 0.005** 0.008*** −0.002 −0.003 −0.004** 
 (0.002) (0.003) (0.002) (0.002) (0.002) (0.004) (0.003) (0.002) 
Male         
wH 0.014*** 0.002 −0.012** 0.008*** 0.059*** −0.034*** −0.014 0.005*** 
 (0.005) (0.006) (0.005) (0.002) (0.010) (0.011) (0.009) (0.002) 
wMH −0.041*** 0.005 0.011 −0.002 −0.034*** 0.077*** 0.021 −0.002 
 (0.007) (0.009) (0.007) (0.004) (0.011) (0.020) (0.014) (0.003) 
wML 0.016*** −0.010 0.002 −0.003 −0.014 0.021 −0.001 0.001 
 (0.006) (0.006) (0.006) (0.003) (0.009) (0.014) (0.016) (0.002) 
wL −0.003** 0.000 −0.001 −0.004** 0.005*** −0.002 0.001 0.005*** 
 (0.001) (0.002) (0.001) (0.002) (0.002) (0.003) (0.002) (0.002) 
ICT capital 0.001* 0.005*** −0.006*** 0.001*** −0.002 0.009*** −0.006*** 0.000** 
 (0.001) (0.001) (0.001) (0.000) (0.002) (0.003) (0.002) (0.000) 
Non-ICT capital −0.001 −0.004** 0.007*** −0.000 −0.000 −0.025*** 0.010*** −0.001* 
 (0.001) (0.002) (0.002) (0.001) (0.003) (0.004) (0.003) (0.000) 
Offshoring 0.012 0.014 −0.012 0.006* 0.034** 0.016 −0.000 0.000 
 (0.008) (0.010) (0.010) (0.003) (0.017) (0.028) (0.021) (0.002) 
Output −0.001 −0.004*** −0.002** −0.001*** 0.004** −0.004 −0.002 −0.000 
  (0.001) (0.001) (0.001) (0.000) (0.001) (0.002) (0.002) (0.000) 
N 364 364 364 364 364 364 364 364 
R2 0.915 0.975 0.908 0.959 0.981 0.985 0.956 0.939 
Year FE Yes Yes Yes Yes Yes Yes Yes Yes 
Industry FE Yes Yes Yes Yes Yes Yes Yes Yes 
Factor-specific time trend Yes Yes Yes Yes Yes Yes Yes Yes 

Source:JIP database 2014 and Ministry of Internal Affairs and Communications (various years).

Note:Figures in parentheses are standard errors. ***Statistically significant at the 1 percent level; **statistically significant at the 5 percent level; *statistically significant at the 10 percent level. The null hypothesis that the fixed effect equals zero is rejected at 1 percent level in all equations. Shaded coefficients are constrained by symmetry constraints. Offshoring is measured by narrow measure. The own price effects are reported in italic. FE = fixed effects.

Table 4.
Baseline results: Elasticities
 Female workersMale workers
 LHLMHLMLLLLHLMHLMLLL
Own price elasticities 1.112 −1.312*** −1.069** 0.062 0.291 −0.122 −0.992** 1.059 
 (0.749) (0.470) (0.527) (0.423) (0.203) (0.197) (0.449) (0.679) 
ICT capital 0.192* 0.223*** −0.640*** 0.248*** −0.052 0.088*** −0.153*** 0.178** 
 (0.106) (0.043) (0.095) (0.066) (0.032) (0.026) (0.052) (0.086) 
Offshoring 1.743 0.648 −1.302 1.185* 0.723** 0.161 −0.012 0.139 
 (1.165) (0.467) (1.048) (0.714) (0.357) (0.283) (0.579) (0.931) 
 Female workersMale workers
 LHLMHLMLLLLHLMHLMLLL
Own price elasticities 1.112 −1.312*** −1.069** 0.062 0.291 −0.122 −0.992** 1.059 
 (0.749) (0.470) (0.527) (0.423) (0.203) (0.197) (0.449) (0.679) 
ICT capital 0.192* 0.223*** −0.640*** 0.248*** −0.052 0.088*** −0.153*** 0.178** 
 (0.106) (0.043) (0.095) (0.066) (0.032) (0.026) (0.052) (0.086) 
Offshoring 1.743 0.648 −1.302 1.185* 0.723** 0.161 −0.012 0.139 
 (1.165) (0.467) (1.048) (0.714) (0.357) (0.283) (0.579) (0.931) 

Note:For offshoring and ICT capital, estimated coefficients are reported. For other notes and sources, see Table 3.

Second, the effects of ICT capital vary across the skill groups of workers. On one hand, the ICT capital stock has a positive and significant effect on the demand for low-skilled female and male workers, middle-high-skilled female and male workers, and high-skilled female workers. On the other hand, it has negative and significant effects on the demand for middle-low-skilled female and male workers.29 On the whole, the elasticities for ICT capital stock reveal the same sign for both male and female workers present in the middle-high-, middle-low-, and low-skill groups. The absolute values of those elasticities are larger for female workers in comparison with male workers at the same skill level. This suggests that demand for female workers is more susceptible to the increase in ICT capital stock than the demand for male workers, although more detailed analysis is needed to clarify the factors behind this susceptibility.

Finally, offshoring has positive and significant effects on the demand for high-skilled male workers and low-skilled female workers. No significant effects are confirmed in other groups of workers. The absence of negative effects for the other groups suggest that offshoring is at least neutral for the demand for a number of worker skill groups.

In sum, the baseline results examining the demand for female workers reveal two primary findings. First, the effects of ICT capital on the demand for female workers varies across worker skill groups. While ICT capital stock has positive and significant effects on the demand for low-, middle-high-, and high-skilled female workers, it has negative and significant effects on the demand for middle-low-skilled female workers. Second, the effects of offshoring on the demand for female workers are positive and significant for low-skilled female workers, and insignificant for other female workers. The results suggest that offshoring is at least neutral for the demand for female workers.

Nevertheless, one may be concerned that our results are sensitive to the measurement of offshoring as well as the control and exogenous variables that have been selected. To address these concerns, we examine the robustness of our results in the following four ways: (1) adding research and development (R&D) intensity as an additional control variable; (2) using alternative measures of offshoring; (3) excluding part-time workers from the sample to focus on full-time workers; (4) excluding either of ICT and offshoring in light of potential endogeneity. The results of the robustness check are reported in Tables 5 and 6 for ICT capital stock and for offshoring, respectively.30 In sum, the results suggest that the effects of ICT on the demand for female workers in the robustness checks are qualitatively similar to those in the baseline model. In contrast, the effects of offshoring appear to be sensitive to the use of alternative offshoring measures. The significance level of the elasticities of low-skilled female workers is low and the results are insignificant when we add R&D intensity and when we use alternative measures of offshoring. This suggests that the positive effect of offshoring on female part-time workers is not necessarily robust. Nonetheless, our main results hold: Offshoring is at least neutral on the demand for female workers.

Table 5.
Robustness check: ICT elasticities
Female workersMale workers
LHLMHLMLLLLHLMHLMLLL
Baseline 0.192* 0.223*** −0.640*** 0.248*** −0.052 0.088*** −0.153*** 0.178** 
 (0.106) (0.043) (0.095) (0.066) (0.032) (0.026) (0.052) (0.086) 
Adding R&D as a control variable 0.148 0.239*** −0.628*** 0.261*** −0.052* 0.083*** −0.220*** 0.121 
 (0.103) (0.044) (0.098) (0.072) (0.029) (0.026) (0.050) (0.082) 
Broad measure 0.204* 0.230*** −0.655*** 0.260*** −0.050 0.091*** −0.158*** 0.184** 
 (0.105) (0.043) (0.095) (0.065) (0.032) (0.026) (0.052) (0.085) 
Relative to all intermediate inputs 0.194* 0.224*** −0.641*** 0.249*** −0.051 0.088*** −0.154*** 0.178** 
 (0.106) (0.043) (0.095) (0.066) (0.032) (0.026) (0.052) (0.086) 
Relative to gross output 0.199* 0.226*** −0.641*** 0.253*** −0.049 0.089*** −0.152*** 0.182** 
 (0.106) (0.043) (0.095) (0.065) (0.032) (0.026) (0.052) (0.085) 
Excluding low skilled workers 0.188* 0.219*** −0.634***  −0.045 0.090*** −0.154***  
 (0.106) (0.044) (0.095)  (0.032) (0.026) (0.052)  
Excluding offshoring 0.200* 0.227*** −0.647*** 0.256*** −0.047 0.088*** −0.153*** 0.176** 
 (0.106) (0.043) (0.095) (0.066) (0.032) (0.026) (0.052) (0.085) 
Female workersMale workers
LHLMHLMLLLLHLMHLMLLL
Baseline 0.192* 0.223*** −0.640*** 0.248*** −0.052 0.088*** −0.153*** 0.178** 
 (0.106) (0.043) (0.095) (0.066) (0.032) (0.026) (0.052) (0.086) 
Adding R&D as a control variable 0.148 0.239*** −0.628*** 0.261*** −0.052* 0.083*** −0.220*** 0.121 
 (0.103) (0.044) (0.098) (0.072) (0.029) (0.026) (0.050) (0.082) 
Broad measure 0.204* 0.230*** −0.655*** 0.260*** −0.050 0.091*** −0.158*** 0.184** 
 (0.105) (0.043) (0.095) (0.065) (0.032) (0.026) (0.052) (0.085) 
Relative to all intermediate inputs 0.194* 0.224*** −0.641*** 0.249*** −0.051 0.088*** −0.154*** 0.178** 
 (0.106) (0.043) (0.095) (0.066) (0.032) (0.026) (0.052) (0.086) 
Relative to gross output 0.199* 0.226*** −0.641*** 0.253*** −0.049 0.089*** −0.152*** 0.182** 
 (0.106) (0.043) (0.095) (0.065) (0.032) (0.026) (0.052) (0.085) 
Excluding low skilled workers 0.188* 0.219*** −0.634***  −0.045 0.090*** −0.154***  
 (0.106) (0.044) (0.095)  (0.032) (0.026) (0.052)  
Excluding offshoring 0.200* 0.227*** −0.647*** 0.256*** −0.047 0.088*** −0.153*** 0.176** 
 (0.106) (0.043) (0.095) (0.066) (0.032) (0.026) (0.052) (0.085) 

Note:For notes and sources, see Table 3.

Table 6.
Robustness check: Offshoring elasticities
 Female workersMale workers
LHLMHLMLLLLHLMHLMLLL
Baseline 1.743 0.648 −1.302 1.185* 0.723** 0.161 −0.012 0.139 
 (1.165) (0.467) (1.048) (0.714) (0.357) (0.283) (0.579) (0.931) 
Adding R&D as a control variable 1.108 0.121 −1.299 0.782 0.588* −0.071 0.265 −0.084 
 (1.114) (0.484) (1.075) (0.780) (0.315) (0.288) (0.552) (0.879) 
Broad measure 1.143 −0.452 −0.509 0.072 0.362 −0.531*** 0.576 −0.408 
 (0.796) (0.321) (0.725) (0.493) (0.245) (0.193) (0.398) (0.636) 
Relative to all intermediate inputs 2.509 0.908 −2.086 1.954* 1.006* 0.171 0.063 0.308 
 (1.856) (0.746) (1.674) (1.140) (0.570) (0.451) (0.925) (1.488) 
Relative to gross output 1.526 0.416 −3.447 1.245 0.396 −0.674 −0.628 −2.132 
 (2.823) (1.133) (2.534) (1.729) (0.868) (0.686) (1.400) (2.237) 
Excluding low skilled workers 1.817 0.666 −1.289  0.733** 0.161 −0.016  
 (1.164) (0.477) (1.048)  (0.359) (0.289) (0.579)  
Excluding ICT 1.890 0.846* −1.830* 1.370* 0.666* 0.248 −0.167 0.244 
 (1.165) (0.485) (1.106) (0.717) (0.359) (0.286) (0.581) (0.944) 
 Female workersMale workers
LHLMHLMLLLLHLMHLMLLL
Baseline 1.743 0.648 −1.302 1.185* 0.723** 0.161 −0.012 0.139 
 (1.165) (0.467) (1.048) (0.714) (0.357) (0.283) (0.579) (0.931) 
Adding R&D as a control variable 1.108 0.121 −1.299 0.782 0.588* −0.071 0.265 −0.084 
 (1.114) (0.484) (1.075) (0.780) (0.315) (0.288) (0.552) (0.879) 
Broad measure 1.143 −0.452 −0.509 0.072 0.362 −0.531*** 0.576 −0.408 
 (0.796) (0.321) (0.725) (0.493) (0.245) (0.193) (0.398) (0.636) 
Relative to all intermediate inputs 2.509 0.908 −2.086 1.954* 1.006* 0.171 0.063 0.308 
 (1.856) (0.746) (1.674) (1.140) (0.570) (0.451) (0.925) (1.488) 
Relative to gross output 1.526 0.416 −3.447 1.245 0.396 −0.674 −0.628 −2.132 
 (2.823) (1.133) (2.534) (1.729) (0.868) (0.686) (1.400) (2.237) 
Excluding low skilled workers 1.817 0.666 −1.289  0.733** 0.161 −0.016  
 (1.164) (0.477) (1.048)  (0.359) (0.289) (0.579)  
Excluding ICT 1.890 0.846* −1.830* 1.370* 0.666* 0.248 −0.167 0.244 
 (1.165) (0.485) (1.106) (0.717) (0.359) (0.286) (0.581) (0.944) 

Note:For notes and sources, see Table 3.

4.2  Discussion

As we have seen, the effects of offshoring on labor demand is neutral in general, and only two skill groups obtain positive results. The question that follows is, why does offshoring affect only the demand for two groups positively? In the case of high-skilled male workers, assignments to managerial, professional, and technical occupations—which are classified into abstract, non-routine tasks—may explain the outcome. For example, the increase of offshoring raises relative demand for workers who coordinate transactions for global value chains or workers who can manage foreign operations. In addition, the increases in R&D or technical centers may accompany professional and technical occupations. In Japan, these tasks are concentrated among highly educated male workers. For low-skilled female workers, the reason behind this trend is not clear, because the educational background of part-time workers is diverse and tasks assigned to them do not fall into obvious patterns.

The difference between the effects of ICT and offshoring may be explained by the demand for tasks. Acemoglu and Autor (2011) explain that the distribution of tasks may differ by gender and by skill level. For instance, the share of clerical/sales occupations is high among female workers, and the share of production/operative occupations is high among male workers. These middle-skill routine tasks tend to be substituted by machines. Besides, tasks are different in their offshorability. Acemoglu and Autor (2011) find that clerical/sales occupations and professional, managerial, and technical occupations are much more offshorable than production/operative or services occupations. Differences in tasks between female and male workers may help explain the differences of the effects between ICT and offshoring.

The Wakasugi, Ito, and Tomiura (2008) investigation of Japanese offshoring based on a survey of Japanese manufacturing firms showed that the most frequently offshored tasks were directly related to manufacturing activities such as the production of intermediate and final assembly. Meanwhile, Ikenaga (2009) has looked for evidence of job polarization in the Japanese labor market. She finds that ICT complemented workers who perform non-routine analytic tasks while substituting for workers who perform routine tasks. That study, however, includes the service sector and it did not account for the effects of offshoring. To address the issue of tasks further, more detailed data and analysis are needed, which goes beyond the scope of this paper.31

Moreover, our estimation reveals the only negative effects of ICT on labor demands are concentrated among middle-low-skilled male and female workers are negative. One may worry that ICT harms some types of worker though the effects on middle-low-skilled workers should not be overestimated. The share of this skill category, namely, junior high school graduates, is shrinking. By 2011, it accounted for 5.1 percent of female workers and 5.7 percent of male workers. On the grounds that ICT and offshoring affect labor demands positively or neutrally for other groups, governments should not hesitate to implement policies to encourage ICT and/or offshoring.

5.  Summary and concluding remarks

In light of the increasing importance of female labor participation in Japan, this paper examines the determinants of the demand for female workers. One novel contribution of this paper is that we shed light on the role of information and communication technology and offshoring as determinants of female labor demand based on the estimation of a system of variable factor demands for manufacturing industries between 1980 and 2011.

Our major findings are two-fold. First, the effects of the ICT capital on the demand for female workers vary across worker skill groups. Whereas the ICT capital stock has positive effects on the demand for low-, middle-high-, and high-skilled female workers, it has negative effects on the demand for middle-low-skilled female workers. Second, the effects of offshoring on the demand for female workers are generally insignificant. Thus, the results suggest that offshoring is at least neutral on the demand for female workers. A part of the increasing demand for female workers can be attributable to ICT, which contributes to narrowing of the gender wage gap in Japan.

In conclusion, there are several research issues for the future that are worth mentioning. First, it is important to examine the relationship between tasks and gender. As mentioned earlier, tasks may be different between female and male workers, which may result in the difference of the effects between ICT and offshoring. The use of individual-level micro data may allow us to address such an issue.

Second, a more detailed analysis of the cost function within each industry presents another important issue. Our analysis is based on industry-level data and thus estimates the system of cost functions for manufacturing as a whole. The industry differences are captured by the industry-fixed effects.

Finally, services offshoring is of interest, although we could not include it in this analysis because of data limitations. The inclusion of service offshoring is important because the share of female workers is generally higher in the services than in the manufacturing sector. Further, the offshoring of service tasks in both manufacturing and services is increasing. Thus, future inclusion of service offshoring may provide an expanded view offshoring in global value chains.

References

Acemoglu
,
Daron
, and
David
Autor
.
2011
.
Skills, Tasks and Technologies: Implications for Employment and Earnings
. In:
Handbook of Labor Economics
,
edited by
D.
Card
and
O.
Ashenfelter
, pp. 4:
1043
1171
.
Amsterdam
:
North Holland
.
Akabayashi
,
Hideo
.
2006
.
The Labor Supply of Married Women and Spousal Tax Deductions in Japan—A Structural Estimation
.
Review of Economics of the Household
4
(
4
):
349
378
.
Autor
,
David H.
2015
.
Why Are There Still So Many Jobs? The History and Future of Workplace Automation
.
Journal of Economic Perspectives
29
(
3
):
3
30
.
Becker
,
Gary S.
1957
.
The Economics of Discrimination
.
Chicago
:
University of Chicago Press
.
Black
,
Sandra E.
, and
Elizabeth
Brainerd
.
2004
.
Importing Equality? The Impact of Globalization on Gender Discrimination
.
Industrial and Labor Relations Review
57
(
4
):
540
559
.
Black
,
Sandra E.
, and
Alexandra
Spitz-Oener
.
2010
.
Explaining Women's Success: Technological Change and the Skill Content of Women's Work
.
Review of Economics and Statistics
92
(
1
):
187
194
.
Feenstra
,
Robert C.
2010
.
Offshoring in the Global Economy: Microeconomic Structure and Macroeconomic Implications
Cambridge, MA
:
MIT Press
.
Feenstra
,
Robert C.
, and
Gordon H.
Hanson
.
1999
.
The Impact of Outsourcing and High-Technology Capital on Wages: Estimates for the United States, 1979–90
.
Quarterly Journal of Economics
114
(
3
):
907
940
.
Hijzen
,
Alexander
,
Holger
Görg
, and
Robert C.
Hine
.
2005
.
International Outsourcing and the Skill Structure of Labor Demand in the United Kingdom
.
Economic Journal
115
(
506
):
860
878
.
Ikenaga
,
Toshie
.
2009
.
Polarization of the Japanese Market: Adoption of ICT and Changes in Task Required
.
The Japanese Journal of Labor Studies
51
(
2–3
):
73
90
[In Japanese]
.
Juhn
,
Chinhui
,
Gergely
Ujhelyi
, and
Carolina
Villegas-Sanchez
.
2014
.
Men, Women, and Machines: How Trade Impacts Gender Inequality
.
Journal of Development Economics
106
(
C
):
179
193
.
Kiyota
,
Kozo
, and
Sawako
Maruyama
.
2017a
.
ICT, Offshoring, and the Demand for Part-time Workers: The Case of Japanese Manufacturing
.
Journal of Asian Economics
48
(
C
):
75
86
.
Kiyota
,
Kozo
, and
Sawako
Maruyama
.
2017b
.
On the Demand for Female Workers in Japan: The Role of ICT and Offshoring
. Keio-IES Discussion Paper Series No. 2017-027.
Tokyo
:
Keio University
.
Kiyota
,
Kozo
,
Takanobu
Nakajima
, and
Kiyohiko G.
Nishimura
.
2009
.
Measurement of the Market Power of Firms: The Japanese Case in the 1990s
.
Industrial and Corporate Change
18
(
3
):
381
414
.
Miyoshi
,
Koyo
.
2008
.
Male–Female Wage Differentials in Japan
.
Japan and the World Economy
20
(
4
):
479
496
.
Nishimura
,
Kiyohiko G.
, and
Masato
Shirai
.
2003
.
Can Information and Communication Technology Solve Japan's Productivity-Slowdown Problem
?
Asian Economic Papers
2
(
1
):
85
136
.
OECD
.
2015a
.
OECD Economic Outlook
2015
.
Paris
:
OECD Publishing
.
OECD
.
2015b
.
OECD Economic Surveys Japan: April 2015 Overview
.
Paris
:
OECD Publishing
.
UNCTAD
.
2011
.
World Investment Report
2011
.
New York and Geneva
:
United Nations
.
Wakasugi
,
Ryuhei
,
Banri
Ito
, and
Eiichi
Tomiura
.
2008
.
Offshoring and Trade in East Asia: A Statistical Analysis
.
Asian Economic Papers
7
(
3
):
101
124
.
Weinberg
,
Bruce A.
2000
.
Computer Use and the Demand for Female Workers
.
Industrial and Labor Relations Review
53
(
2
):
298
308
.

Notes

*

We thank Shigeyuki Abe, Sarra Ben Yahmed, Yukiko Ito, Naoto Jinji, Sung-Chun Jung, Ayako Obashi, Fumio Ohtake, Tsunehiko Otsuki, seminar participants at Kobe University and Osaka University, and participants of Asian Economic Panel Keio Meeting, EHESS—Keio conference on globalization and labor market outcomes, the JAE Meeting, and the JSIE Meeting for their helpful comments. We are grateful to Tatsuji Makino for providing us with the disaggregated JIP labor data and to Taiji Hagiwara and Kozo Miyagawa for providing us with data set of the Japanese input-output tables. Kiyota gratefully acknowledges financial support received from the MEXT-Supported Program for the Strategic Research Foundation at Private Universities. Kiyota gratefully acknowledges financial support received from the JSPS Grant-in-Aid (JP16H02018, JP26285058). The usual disclaimers apply.

1 

Data for Czech Republic in 2014 are not available in OECD (2015a).

2 

“Mr. Abe has made ‘womenomics’ a core part of his ‘Abenomics’ policies, in the hope that bringing more women into the workforce will raise Japan's growth potential.” The Financial Times, 7 December 2015.

3 

Similarly, using the data from West Germany, Black and Spitz-Oener (2010) found that technological change contributed to the women's task change, from routine to non-routine tasks. The change explains a large fraction of the closing of the gender wage gap.

4 

See, for example, Black and Brainerd (2004) for the case of the United States, and Juhn, Ujhelyi, and Villegas-Sanchez (2014) for the case of Mexico.

5 

The trends of offshoring and ICT will be discussed in Section 3.2.

6 

For the detailed explanations about the relationship between offshoring and skill-biased technological change, see Feenstra (2010).

7 

We focus on manufacturing in this paper for the following two reasons, even though we recognize the importance of services offshoring. First, offshoring in the manufacturing sector has developed widely and already been observed in the 1980s, whereas offshoring in services is rather a new phenomenon that started to develop with ICT technology in the 1990s (UNCTAD 2011, 137). Second, the JIP database is not fine enough to zoom in on those services that are heavily offshored, such as consultancy and accounting services. We also recognize the importance of analyzing services for gender issues. Analyzing services is a key question for future research.

8 

In Kiyota and Maruyama (2017a), ICT capital stock affected the demand for part-time workers positively, whereas the effect of offshoring was insignificant.

9 

For a cost function to be well behaved, assume further that the cost function is homogeneous of degree one in prices for a given output. This implies the following restrictions: and . Without loss of generality, symmetry implies .

10 

We assume that the industry-factor specific fixed effects are time-invariant. Industry characteristics can be time-variant, although we cannot introduce time-variant industry effects in our estimation because it corresponds with residuals. To apply time variance, variables that indicate industrial characteristics need to be identified. We leave this for our future study.

11 

Nishimura and Shirai (2003) estimated the translog cost function, treating ICT capital stocks as variable factors to examine the substitutability between ICT capital stocks and labor inputs. In order to estimate the cost function, they compiled industry-level capital and labor inputs data between 1980 and 1998, including the price deflator on the ICT investment. To the best of our knowledge, however, such data are not available in Japan after 2000.

12 

We also assume that these variables explain the shift of the demand curve: The shift of the intercept can be controlled for by year dummies and/or trend variables.

13 

Following Hijzen, Görg, and Hine (2005), we call and as elasticity.

14 

The database is downloadable from www.rieti.go.jp/en/database/JIP2014/index.html.

15 

ICT capital stock and non-ICT capital stock are used as fixed inputs because the user cost of capital is not available by the type of capital. Additionally, it is common to assume that capital is a (quasi-)fixed input in estimating a production function. See, for example, Kiyota, Nakajima, and Nishimura (2009) for the case of Japanese firms. We checked the correlation between ICT capital stock and non-ICT capital stock and found that it was 0.6878. The positive coefficient suggests that there is no substitutive relationship between ICT and non-ICT capital stock.

16 

Because the JIP database follows the coverage of the System of National Accounts 1993, own-account software and prepackaged software are not included in ICT capital stock.

17 

University graduates are defined as graduates of four-year universities or graduate schools. College graduates are defined as graduates of higher professional schools and junior colleges, having obtained two years of education after finishing high school.

18 

The coverage of the non-regular workers is wider than that of the part-time workers because some of the non-regular workers work more than 35 hours per week.

19 

In the JIP database, the wage of male part-time workers is computed from that of the female part-time workers. This in turn implies that there is basically no wage difference between female and male part-time workers in the JIP database.

20 

In general, firms assign jobs that require fewer skills to part-time workers rather than full-time workers. For more detail, see Kiyota and Maruyama (2017a). One may argue that we exclude part-time workers because that group consists of workers who have different educational backgrounds. As we will see subsequently, the share of part-time workers in total female workers is non-negligibly high. Therefore, we keep this category in our baseline analysis. We exclude this category as a robustness check in Section 4.2, however.

21 

Data for manufacturing is not available.

22 

Table A2 of Kiyota and Maruyama (2017b) presents the share of part-time workers between 1980 and 2011 by industry.

23 

The construction of the import-use matrices is explained in Kiyota and Maruyama (2017a). We focus on the imports of intermediate goods in order to capture trade flows in global value chains. Feenstra (2010) defines offshoring as consisting of two types of foreign production that shapes a part of global value chains: in-house foreign production by multinationals and foreign outsourcing. Imports of intermediate goods are accompanied by these foreign productions. Note that the imports of final goods are not included because it is difficult to distinguish imports due to offshoring from imports due to other reasons such as direct consumption by the household sector. Note also that exports of intermediate goods from Japan are excluded, although they constitute a part of offshoring. To capture this type of offshoring, it is necessary to develop an alternative offshoring measure, which goes beyond the scope of this paper.

24 

Note that the import-use matrix is not available every year in many countries. Some studies such as Hijzen, Görg, and Hine (2005) used linear extrapolation (or interpolation) for the missing years to fill the gaps. In this paper, however, we do not use linear extrapolation (or interpolation). The changes in imports seemed to be nonlinear because the missing years include such years as the 1997–98 Asian financial crisis and the 2008–09 global financial crisis.

25 

Note that there is a slight abuse of notation where both i and j stand for industries. To maintain the consistency, stands for the imported intermediate inputs from industry j to industry i, which is opposite from the standard notation in the input–output analysis.

26 

For some industries such as seafood products and livestock products, inputs mainly come from agricultural industries. If we focus on manufacturing intermediate inputs for , these industries tend to show a high offshoring index because their manufacturing inputs are low. In the baseline model, therefore, we take into account agricultural intermediate inputs. To check the robustness of our results, Section 4 utilizes different measures in the denominator.

27 

Strictly speaking, Feenstra and Hanson (1999) utilized non-energy intermediate inputs for the denominator.

28 

Table A3 and A4 of Kiyota and Maruyama (2017b) present narrow offshoring and the share of ICT capital stock to total capital stock between 1980 and 2011 by industry, respectively.

29 

The positive effects of ICT on middle-high- and low-skilled workers and the negative effects on middle-low-skilled workers may imply ICT-based “job polarization,” in which “automation and new technology were going to wipe out large numbers of middle class jobs” (Autor 2015, 3). Although our analysis excludes services, our result might suggest a common feature of tasks assigned to low-skilled workers in the manufacturing sector, which are “challenging to automate” and “infeasible to offshore” (Acemoglu and Autor 2011, 1077).

30 

See Kiyota and Maruyama (2017b) for details of the robustness check.

31 

We cannot obtain the wage information by occupation and by industry in JIP database, which prevents us from estimating labor demand function by occupation.