Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Internet use and gender wage gap: evidence from China

Internet use and gender wage gap: evidence from China This study explores the influence of Internet use on the gender wage gap in China by using national longitudinal survey data. A fixed effects and instrumental variable method were employed to address individual heterogeneity and other endogeneity problems. The study contributes in the form of four key conclusions. First, considering the endogeneity problems, the return to Internet use is higher for men than for women, unlike the results derived using the ordinary least squares model, which indicates the opposite. The individual heterogeneity problem considerably affects the results, thus suggesting a bias in earlier studies. The results of robustness checks based on the Internet use frequency for different purposes confirm the conclusions. Second, the gender difference in return to Internet use is higher in the low-education group and older age cohorts. Third, both the components—the gender disparity in Inter- net access and gender difference in return to Internet use—widen the gender wage gap, with the gender difference in return to Internet use having a higher impact. Fourth, the effects of these two components on the gender wage gap vary with the educational attainment and age cohorts. Keywords: Gender wage gap, Internet use, Return to Internet use, Internet access, China JEL Classification: J16, J24, J31, O33 1 Introduction 2021; Miller and Mulvey 1997; Pabilonia and Zoghi 2005, With the progress of information and communication etc.) and employment (Alam and Mamun 2017; Atasoy technology (ICT) since the 1970s, Internet usage has 2013; Deyyling 2017; Mao and Zeng 2017, etc.), empiri- expanded worldwide (OECD 2018). The gender digital cal studies on the impact of Internet use on the gender gap in Internet access rose in developed countries in the wage gap are scarce. This study attempts to bridge this early stages of ICT development (Bimber 2000; DiM- gap by providing new evidence from China—a country aggio et  al. 2001; Fatehkia et  al. 2018) but reduced with that has seen rapid Internet diffusion and gender wage the increasing diffusion of digital technologies (Haight gap growth in the last two decades. et al. 2014; Ono and Zavodny 2007; Rice and Katz 2003). The China’s gender disparity in Internet use can be Women in developing countries have a significantly lower highlighted through the Statistical Report on the Devel- likelihood of Internet access than men, and this gender opment of the Internet in China No. 45 (CNNIC 2020), disparity in Internet use can enlarge the overall socio- which reveals that the number of Internet users in China economic gender gap (Alozie and Akpan-Obong 2017; reached 904 million in April 2020, of which 48.1% were Broadband Commission 2013; Hafkin and Huyer 2007; women (30.4% in 2000). The statistics suggests the exist - OECD 2018). Although research has established that ence of a gender disparity in Internet access in China. Internet use can affect wages (Krueger 1993; Liu et  al. Additionally, the gender wage gap in China has expanded since the 1980s (Gustafsson and Li 2000; *Correspondence: xxma@hosei.ac.jp The proportion of women in the total population in China was 48.71% in Faculty of Economics, Hosei University, 4342 Machita-shi Aiharamachi, 2020 (World Bank 2022). Tokyo 194-0298, Japan © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. 15 Page 2 of 17 X. Ma Iwasaki and Ma 2020; Ma 2018). Several empirical stud- and age cohort groups. The gender differences in Inter - ies reveal that the main determinants are gender differ - net access and return to Internet use may differ among ences in human capital and workplace discrimination different education and age cohort groups; therefore, against women (Gustafsson and Li 2000; Li and Yang the effects of Internet use on the gender wage gap may 2010; Li and Yang 2010; Ma et  al. 2013). Others indi- differ per these groups. This study compares the effects cate that occupational segregation and industry/owner- of Internet use among educational attainment and age ship sector segmentation also contribute to the creation cohort groups to provide more evidence to understand of the gender wage gap (Ge 2007; Li and Ma 2006; Liu the digital division issues in depth. Additionally, we use et  al. 2000; Ma 2018; Meng and Zhang 2001; Wang and the three latest wave national longitudinal survey data of Cai 2008; Wang 2005). However, empirical studies on the 2014–2018, which can provide new information. effects of Internet use on the gender wage gap are scarce The study resulted in four main findings. First, when (Qi and Liu 2020; Wu 2021; Zhuang et al. 2016) and lim- addressing endogeneity problems, the return to Internet ited to cross-sectional data analyses, which can result in use is higher for men than for women, unlike the results endogeneity problems. using the ordinary least squares (OLS) model, which This study contributes to the literature in four ways: indicate the opposite. In this case, the individual het- First, in contrast to earlier studies (Qi and Liu 2020; erogeneity problem considerably affects the results, thus Wu 2021; Zhuang et  al. 2016), we examine the influ - suggesting a bias in earlier studies. Second, the results ence of Internet use on the gender wage gap in China indicate that the gender disparities in return to Internet using the panel data analysis method, such as the fixed use are higher in the low-education and older age cohort effects (FE) model, random effects (RE) model, and the than in others. Third, both the components—the gender lagged variable (LV) model to address endogeneity prob- disparity in Internet access and the gender difference in lems. This study is the first to investigate the association return to Internet use—widen the gender wage gap, with between Internet use and the gender wage gap in China the gender difference in return to Internet use having a based on the panel data analysis method. Second, besides higher impact. Fourth, the effects of both components investigating the extensive margin effect of Internet use on the gender wage gap vary with the educational attain- (whether used Internet), we also ascertain the intensive ment and age cohorts. margin effects (the frequency of using the Internet) for The remainder of this paper is organized as follows. different purposes (work and study, communication, lei - Section  2 discusses the channels whereby Internet use sure) on the gender wage gap, which has been not con- influences wages and summarizes the empirical literature sidered in previous studies. Third, our study is the first on the issue. Section 3 introduces the data and the meth- to decompose the determinants of the gender wage gap odology. Section  4 presents and discusses the empirical into two components—the gender disparity in Internet results, and Sect. 5 concludes the study. access (e.g., gender disparity in the proportion of work- ers using the Internet) and gender difference in return 2 Literature review to Internet use (e.g., the gender difference in the rising 2.1 Channels of influence of Internet use on the gender range of wage level in terms of Internet use)—based on wage gap the IV method. Since policy implications differ across Several economic theories can explain the gender wage these two components, it would be meaningful to investi- gap. First, gender differences in the endowment of factors gate how these two components contribute to the effects such as human capital can contribute to a gender wage of Internet use on the gender wage gap. For example, if gap. Based on the human capital theory (Becker 1964; the gender disparity in Internet access is found to be the Mincer 1974), the individual wage level in a perfectly main contributor, a policy promoting Internet accessibil- competitive labor market is determined by the work- ity among women is expected to reduce the wage gap. In ers’ labor productivity. Labor productivity is related to a contrast, if the main contributor is the gender difference worker’s human capital (e.g., education and years of work in the return to Internet use, reducing the discrimination experience). When men have higher educational attain- against women in the workplace may reduce the wage ment than women, they may earn a higher wage. gap. Fourth, although Eastin et al. (2015), OECD (2018), Second, according to the employer discrimination Riggins and Dewan (2005), and Scheerder et  al. (2017) hypothesis (Becker 1957), whenever employers, cus- have argued the Internet use divisions in education, age, tomers, and colleagues discriminate against women, and gender groups, they have not investigated the effect it can create a gender wage gap. During the economic of Internet use on gender wage gap among education transition period (after 1978), with the enforcement Internet use and gender wage gap: evidence from China Page 3 of 17 15 of the Opening-up Policy, the private sector (i.e., the discrimination may contribute to widening the gender privately-owned enterprises, the Township and Village wage gap in China. Enterprises, etc.) has developed since the 1990s, and Fourth, the crowding hypothesis (Bergmann 1974) the proportion of workers in the private sector to total stated that there remains gender occupational segrega- workers in urban areas has increased from 27.2% in 1985 tion in the labor market, which means women concen- to 67.4% in 2020 (NBS 2021). In the planned economy trate on female-dominated occupations (e.g., staff, service period (1949–1977), women’s employment promo- job), whereas men on male-dominated (e.g., manager, tion policies, such as the equality employment policy, technician). When the wage levels of male-dominated were enforced in the public sector (e.g., the state-owned occupations are higher than those of female-dominated enterprises: SOEs; the government organizations) by occupations, the gender wage gap arises. The evidence the Chinese government, which led to less discrimina- from the empirical studies for the developed countries tion against women (Gustafsson and Li 2000; Ma 2018, has supported the crowding hypothesis (Brown et  al. 2021a, b). With the progress of SOE reforms, most SOEs 1980; Kidd and Shannon 1996; Kidd 1993; Miller 1987). have changed to privately-owned enterprises. Although Meng (1998) and Li and Ma (2006) also reported that the influences of the equality employment policies are in China, there exists gender occupational segregation, still greater in the SOEs even in the current period, when which contributes to the formation of the gender wage discrimination against women persists in the private sec- gap. tor, a gender wage gap can arise. Iwasaki and Ma (2020) Finally, the monopsony power hypothesis suggests that report that the gender wage gap is higher in privately- imperfect competition may lead to a gender wage gap. owned enterprises than in SOEs, and the gender wage When a firm has the monopsony power in the labor mar - gap expanded during the economic transition period ket and sets a lower wage level for women, the gender from the 1990s to the 2010s, suggesting that the discrimi- wage gap arises (Hirsch 2010). Vick (2017) reports that nation against women has become severe with the pro- the monopsony power hypothesis is supported in Brazil. gress of the market-oriented economic reform in China. Based on these theories or hypotheses, in this study, we Third, the statistical discrimination hypothesis (Arrow considered that Internet use could lead to the formation 1973; Phelps 1972) suggests that because of information of gender wage gap in China through two channels: (i) the asymmetry, employers must make employment and wage explained component (e.g., gender disparity in Internet decisions for employees, both men, and women, based access), and (ii) the unexplained component—workplace on the average values of some unobservable factors (i.e., discrimination against women (e.g., gender difference in work effort, probability of turnover, etc.). If the employer return to Internet use as reflected in the wages). perceives that the probability of doing housework (i.e., Regarding the first component–the effect of gender childcare, geriatric care, domestic cleaning, cooking, etc.) disparity in Internet access on the gender wage gap, (i) is higher for women than for men, they may set a lower based on human capital theory, the wage level in a per- wage level for women, thus generating a gender wage gap. fectly competitive labor market is determined by a work- In China, factors such as the deregulation of the one- er’s labor productivity; the labor productivity is related child policy, lack of formal public childcare institutions to a worker’s human capital. Internet use can be consid- (e.g., kindergarten), population aging, and insufficiency ered an element of human capital: Internet users are usu- of institutional care for the elderly have increased the ally considered to be those with higher skills (or higher responsibilities of childcare and geriatric care for women productivity) than non-users. If the percentage of Inter- (Connelly et  al. 2018; Cook and Dong 2011; Ma 2021a, net users in the women’s group is lower than that in the b). Family care has decreased the female labor force par- men’s, the gender wage gap arises. (ii) According to the ticipation and reduced women’s work efforts in China crowding hypothesis, men may occupy male-dominated (Chen and Fan 2016; Chen et al. 2016; Chen 2019; Con- occupations (e.g., manager, technician), which have a nelly et  al. 2018; Ma 2021a, b). Therefore, statistical higher likelihood of using the Internet for work, which may lead to the gender divisions in Internet access. To consider the second component–the effect of gen - In China, the Township and Village Enterprises (TVEs) were collectively der differences in return to Internet use on the gender owned enterprises and developed by the end of the 1980s. Since the 1990s, wage gap, (i) according to the employer discrimination most of them have transformed into privately owned enterprises, wherein and statistical discrimination hypotheses, when work- operations and human resource management are similar to those in privately- owned enterprises. place discrimination exists, wage levels tend to be set lower for women than men despite similar endow- In the official data (NBS 2021), the private sector includes the privately- owned enterprises, collectively owned enterprises, foreign investment enter- ments (e.g., Internet use skills). (ii) Based on the crowd- prises, and other privately-owned enterprises (e.g., the joint-stock company, ing hypothesis, if Internet use is evaluated higher for joint company, limited liability company). 15 Page 4 of 17 X. Ma 2.3 Empirical studies on Internet use and gender wage male-dominated occupations than female-dominated gap occupations by a firm, the gender wage gap arises. (iii) Empirical studies on Internet use and the gender wage According to the monopsony power hypothesis, if a firm gap are scarce in developed countries and China. has monopsony power and values Internet use more Borghans et al. (2014) and Ge and Zhou (2020) find that highly for men than women, the gender wage gap widens. Internet skills and their return affect the changes in the Although it is assumed that the two components may gender wage gap in developed countries. Beaudry and contribute to the formation of the gender wage gap, Lewis (2014) hold that changes in return to computer there is no empirical study to investigate the issue. In skills are an important factor that explains the decline in this study, we attempt to estimate the contribution rates the gender wage gap. Using industrial data, Ge and Zhou of these two components in the subsequent decomposi- (2020) report that robot use reduces the gender wage tion analysis and provide new evidence to understand the gap, while an increase in computer capital raises the gen- causes of the gender wage gap in China in depth. der wage gap in the US. However, these studies are not based on the decomposition method. Therefore, the con - 2.2 Empirical studies on the gender w age gap in China tribution rates of the gender disparity in Internet access There are numerous empirical studies on the gender and gender difference in return to Internet use in forming wage gap in developed countries and China (Iwasaki and the gender wage gap are unclear. Ma 2020). We summarize only the main ones in China Regarding China, only three studies focus on the issue. as follows. Zhuang et  al. (2016) use data from the Third Survey on Gustafsson and Li (2000), Liu et  al. (2000), Li and Chinese Women’s Social Status and apply the propensity Yang (2010), Li et al. (2014), and Ma et al. (2013) use the score matching method to estimate the wage function. Blinder–Oaxaca model (Blinder 1973; Oaxaca 1973) for They find that the Internet use wage premium for women the decomposition analysis, thereby demonstrating that is 90.6% of that for men, suggesting that the return to both the explained (e.g., gender differences in educational Internet usage is lower for women than men. Using data attainment) and unexplained (e.g., returns to education) from 2010, 2013, and 2015 Chinese General Social Sur- components affect the gender wage gap. Most stud - veys (CGSS) and the Blinder–Oaxaca decomposition ies show that the contribution of the unexplained com- method, Qi and Liu (2020) report that both Internet ponent is higher than that of the explained component, access and the return to Internet usage reduced the gen- which suggests that workplace discrimination against der wage gap in 2013 and 2015. Wu (2021) uses data from women is the primary cause of the gender wage gap in the 2017 CGSS, the OLS model, and the Blinder–Oaxaca China. The contribution rates of the unexplained compo - decomposition method to analyze the effect of Internet nent in the gender wage gap for the local urban resident use on the gender wage gap and reveals that the contribu- group were reported to be 52.49% in 1988 and 63.20% in tion rate of gender disparity in Internet access is -4.34%, 1995 (Gustafsson and Li 2000); and 52.0% in 1995, 69.0% which reduces the gender wage gap. in 2002, and 77.7% in 2007 (Li et  al. 2014). The rate for Although these studies provide some evidence on the the migrant group was 74.32%–84.38% in 2008 (Li and effects of Internet use on the gender wage gap in China, Yang 2010), and that for all residents was 49.18% in 1996 they have not addressed the individual heterogene- (Meng and Zhang 2001). In addition, the values by wage ity problem owing to the use of cross-sectional analy- percentiles were 86.08–101.80% in 2006 and 45.31– sis methods. Additionally, the intensive margin effects 91.73% in 2009 (Ma et al. 2013). A few empirical studies of Internet use (e.g., the frequency of Internet use) for also focus on the effect of segmentation by sector on the different purposes (e.g., work and study, leisure) were gender wage gap in China. Li and Ma (2006) and Wang not considered. Moreover, the differences in the effects (2005) analyze the influence of occupational segregation of Internet use by the educational attainment and age on the gender wage gap. Ge (2007), Ma (2018), and Wang cohorts were not considered. This study can address and Cai (2008) explore the impact of segmentation by these neglects. industry or enterprise ownership sector type on the gen- der wage gap in China and report that the unexplained 3 Methodology component in intra-sector differentials drives the gender 3.1 Model wage gap in China. To estimate the gender disparity in return to Internet use, we estimate the wage function. The OLS model is Iwasaki and Ma (2020) conducted a meta-analysis based on the results of expressed by Eq. (1). the inclusion of a gender dummy variable in the wage functions of 199 studies for China. For the current empirical studies on the gender wage gap in other countries, please refer to Ge and Zhou (2020) for the U.S., Biewen et al. (2020) for German, Masso et al. (2022) for Estonia. Internet use and gender wage gap: evidence from China Page 5 of 17 15 the influence of the regional telecommunication capabil - LnW = a + β Int + β Female i 0 1 i 2 i ity in the past on an individual’s recent income levels is (1) + β Female ∗ Int +γ X + v 3 i i i small, which may fit the conditions of the IV from the econometric perspective. We performed several tests, In Eq. (1), i represents the individual. Int is the Internet including the Durbin-Wu-Hausman test, the exclusion use variable, Female is the female dummy; Female ∗ Int test (Hansen’s J statistic), and the weak identification test is the interaction term of Internet use and female (Cragg–Donald Wald F statistic). The results indicated dummy variable;LnW denotes the logarithmic value of that the two IVs are appropriate (see Tables 2, 3, 4 and 5 wages; X represents a set of the factors (i.e., educa- and discussions in Sect. 4). tion, years of work experience, occupation, etc.) that Second, as v in Eq.  (1) includes the errors related to affect the wages; β and γ are the estimated coefficients; individual-specific and time-invariant factors ( ρ ) and and v denotes the error term. The total of β and β are 1 3 the idiosyncratic error ( ε ), an individual heterogeneity the return to Internet use, β is the gender differences in problem may arise in Eq. (1). We use the FE or RE model return to Internet use. to address this heterogeneity problem. We perform the However, the endogeneity problem may exist in Eq. (1) Hausman specification test to judge the validity of the FE for three reasons: first, the omitted variable may influ - and RE models. In Eq. (5), t denotes the longitudinal sur- ence the likelihood of using the Internet and the wage vey year. level. We constructed the variables to control factors that may affect wage level as much as possible. However, LnW = α + β Int + β Female ∗ Int +γ X + ρ + ε it 1 it 2 it it it some unobservable variables may also affect the results. (5) We used the IV method to address the issue expressed by Third, the endogeneity issue may arise due to reverse Eqs. (2–4). causality. It is assumed that Internet use may affect wages Pr(Int = 1) = a + β Z +β Int + β Female (explored in this study). However, wages may also affect i 1 Z i i 2 i the likelihood of Internet use. For instance, the likelihood + β Female ∗ Int + γ X + u 3 i i i of using the Internet for work and study may be greater (2) for high-wage workers than low-wage workers. Since there is a two-way relationship between Internet use and LnW = a + β Int + β Female i 0 1 i 2 i (3) wage, we use a one-period lagged variable (LV) model to + β Female ∗ Int +γ X + δ 3 i i address the potential reverse causality. We assume that Internet use in period t − 1 may affect the wage level in period t . However, wages in period t cannot influence corr(Z, δ) = 0 and corr(Z, u) �= 0 (4) the likelihood of using the Internet in period t − 1 . In In Eqs. (2–4), u , and δ denote the error terms, respec- Eq.  (6) below, t represents the recent period (e.g., 2018), tively. Z represents the IV. The internet penetration rate t − 1 represents the prior period (e.g., 2016), and Int t−1 at the regional (provincial or local) level and the impor- expresses the Internet use in the prior period ( t − 1). tance of using the Internet attitude were generally used LnW =α + β Int + β Female ∗ Int it 1 it−1 2 it−1 in previous studies (e.g., Cao and Jiang 2020; Zhao and (6) Li 2020). We performed several tests for these IVs, but +γ X + ρ + ε i it the results rejected the hypothesis that these IVs are exogenous, suggesting that they were not valid for this Then, to investigate the effect of Internet use on the study. We used two variables—(i) the provincial optical gender wage gap, we use the Blinder–Oaxaca decomposi- cable circuit in 1999 and (ii) the provincial long-distance tion method (Blinder 1973; Oaxaca 1973) to decompose cable line length in 1999 as IVs in this study. Both are the the determinants into two components: (i) the explained oldest data that we could obtain from the government’s component, comprising the gender endowment differ - official dataset. It can be assumed that Internet instal- ences (i.e., the gender disparity in Internet access, etc.), lations in recent survey years (2014, 2016, and 2018) are and (ii) the unexplained component, composed of the closely related to the regional telecommunication capa- gender differences in the evaluated price of each factor bility in the past (such as the optical cable circuits or the (i.e., the gender difference in the return to Internet use, long-distance cable line length 15 to 19 years back), while etc.). Equations (7, 8) describe the model : The constant is omitted for descriptive convenience. We conduct two The information on these two IVs were obtained from the data published in decomposition analyses based on Eqs. (7) and (8); only one has been listed China Statistical Yearbook 1999 (NBS 1999). because the results are very similar. 15 Page 6 of 17 X. Ma 2010 CFPS baseline survey data was obtained through ¯ ¯ ¯ LnW − LnW = β H − H + (β − β ) H m f m m f m f f multi-stage probability sampling with implicit stratifica - (7) tion. Multi-stage sampling reduces the operational cost ¯ ¯ ¯ LnW − LnW = β H − H + (β − β ) H of the survey and permits the analysis of the social con- m f f f m f m m text. In the 2010 baseline survey, the CFPS successfully (8) interviewed nearly 15,000 families and 30,000 individuals In Eqs. (7) and (8), LnW − LnW is the gender wage within these families, with an approximate response rate gap; β , β is coefficient of each factor calculated based of 79%. The respondents were tracked through annual on the male and female wage functions, respectively; H follow-up surveys. The CFPS 2010 covers 25 provinces denotes the mean value of each factor, including Inter- and municipalities. Only the latest three waves (2014, net use. β ( H − H ) or β ( H − H ) is the m m m f f f 2016, and 2018) of the CFPS, which include the survey explained component, and ( β − β ) H or ( β − β m m f f f item on Internet use, have been used in this study. The ) H is the unexplained component which includes the CFPS sample sizes are 37,147 (2014), 36,892 (2016), and discrimination against women in the workplace. 37,354 (2018). To compare the effects of Internet use by groups, we The logarithmic value of the hourly wage is used as the also calculate the estimations per educational attainment dependent variable. The wages for 2014, 2016, and 2018 and age cohort group. have been adjusted using the regional Consumption Price Index (CPI) in the rural and urban areas published 3.2 Data by China’s National Bureau of Statistics (NBS 1999) to Three waves of data (2014, 2016, and 2018) from the account for inflation, using the 2014 CPI as the baseline. China Family Panel Studies (CFPS 2020) dataset are used We calculate the hourly wage as a dependent variable in this study. The reasons for using the CFPS are con - based on the annual earned income and corresponding sidered as follows: first, the CFPS is a nationally repre - working hours. sentative longitudinal survey of Chinese communities, The key independent variable is an Internet use dummy families, and individuals launched in 2010 by the Insti- variable based on the questionnaire item: “Did you use tute of Social Science Survey, Peking University, China. the Internet in the past year?” (1 = has used the Internet Although Chinese General Social Survey (CGSS) data in the past year, 0 = otherwise); we primarily use the vari- was used in the previous studies (e.g., Qi and Liu 2020), able to estimate the extensive margin effect of Internet since the CGSS is the cross-sectional survey data, the use, which was estimated in the literature. Based on the individual heterogeneity problem could not be addressed. questionnaire items on the frequency of Internet use by On the contrary, this study performs the analysis using purpose (work, study, communication, shopping, enter- the panel data analysis method (e.g., the FE and RE mod- tainment), we originally constructed three indicators to els, the LV model) to deal with the endogeneity problems, investigate the intensive margin effect of Internet use in this study, based on the CFPS, can provide the robustness this study: (a) frequency of using the Internet for work, results on the issue. Second, although the China Health including work and study, (b) frequency of using the and Retirement Longitudinal Survey (CHARLS) is a Internet for communication, and (c) frequency of using longitudinal survey having information on Internet use, the Internet for leisure, including shopping and enter- the survey targets of CHARLS are individuals aged 45 tainment. Based on the five questions items in the CFPS and older, whereas the CFPS covers all age generations. as “please answer the frequency of using the Internet for Therefore, we can use the data from the CFPS to com - study, work, communication, entertainment, commer- pare the differences in the effect of Internet use among cial activity (e.g., Internet payment, shopping): (i)almost the younger, middle-aged, and older generations in this every day; (ii)3–4 times a week; (iii)1–2 times a week; study. Third, we can obtain the rich information on Inter - (iv) 2–3 times a month; (v) once a month; (vi) once few net use, such as the used Internet and the frequency of months; (vii) never use)”, we re-coded each frequency using the Internet for different purposes (e.g., working as “7 = almost every day; 6 = 3–4 times a week; 5 = 3–4 and study, communication, and leisure); the latter is the times a week; 4 = 1–2 times a week; 3 = once a month; unique question item in the CFPS which is firstly utilized 2 = once in few months: 1 = never use”. We calculated the in this study on the issue. total values for (a) and (c) and used their arithmetic mean The CFPS is designed for individual-, family-, and in the analysis. community-level longitudinal data collection in con- We constructed an interaction term of Internet use and temporary China and provides information on Internet a female dummy variable to investigate the gender differ - use, wages, and other factors (education, years of work ence in return to Internet use in wage functions. experience, sex, occupation, industry sector, etc.). The Internet use and gender wage gap: evidence from China Page 7 of 17 15 Mean: Mean: M: 2.573; F:2.262 M: 2.214;F:1.765 SD: SD: M: 1.097; F:0.997 M: 0.920;F:1.003 Fig. 1 Kernel density of the logarithm of wages for Internet using and not-using groups. Source: Authors’ calculations based on the data from CFPS of 2014, 2016, and 2018. M: male workers; F: female workers Based on the economic theories and existing studies, that there exists a gender wage gap, and the gender wage we identified a set of variables that may affect wages as gap differs among Internet using and non-using groups. specified by the wage functions, such as years of school - The proportions of individuals using the Internet by ing, years of work experience and its squared term, gender are summarized in Table 1. In general, an increase ethnicity (1 = han, 0 = minority), party membership is observed from 2014 to 2018 for both men and women. (1 = member of Communist Party of China, 0 = non- However, the percentage is approximately 5% higher for member), urban residents(1 = urban residents, 0 = rural men than women in each year of the three years, suggest- residents), occupation (1 = manager and technician, ing the existence of gender disparity in Internet access. 0 = otherwise), industry sector (1 = manufacturing In terms of the disparity by educational attainment industry sector, 0 = otherwise), workplace ownership group, the percentage of individuals using the Internet (1 = state-owned sector, 0 = otherwise), region (west, is higher for men than women in the low- and high-edu- central, east), and survey year. cation groups, whereas the opposite holds in the middle- As described in Sect.  3.1, we used two variables: the education group. Regarding the disparity by age cohort, provincial optical cable circuit and the provincial long- the percentage of individuals using the Internet is higher distance cable line length in 1999, as IVs in this study. for men than women in each age cohort. The gender dis - This analysis is limited to respondents aged 16–60 years parity in Internet access is the largest in the middle-aged and excludes missing values; the longitudinal sample size generation born during 1970–1989 and the least in the 7 9 is 18,381. younger generation born after 1990. The results indicate that the gender disparities in Internet access differ by edu - 4 Empirical results and discussion cational background and age cohort. Therefore, the het - 4.1 Descriptive statistics erogeneous group should be considered in the analysis. Figure 1 shows the kernel density of the logarithmic value of wages by Internet-using and non-using groups. The 4.2 Basic results wage level is higher for men than women in both groups, Table  2 presents the results of the wage function analy- and the gender wage gap is higher in the Internet-using sis using five models—the OLS (Model 1), IV (Model 2), group than that in the not using one. The results indicate The questionnaire item on education attainment has eight categories: illit - erate/semiliterate, elementary school, junior high school, senior high school, According to the Interim Measures of the State Council on the Placement of college, university, master, and doctor. We divided the educational groups into Elderly, Weak, Sick and Disabled Cadres (Article 4) and Interim Measures of three categories: the low- (elementary school and lower), middle- (junior and the State Council on Retirement and Retirement of Workers (Article 1) pub- senior high school), and high- (college and higher) educational groups. lished in 1978, the minimum employment age in China is 16 for both men and women. However, the mandatory retirement age differs by gender: it is We divide the age cohorts considering the Internet diffusion situation. We 60 years for both blue- and white-collar male workers, while it is 50 years for define the younger generation as the group born in the year when the Inter - blue-collar female workers and 55 years for white-collar female workers (e.g., net was used in society, the middle-aged generation who used the Internet women working as a civil servant/office executive, or working as a manager, in their adolescence, and the older generation who used it in their middle- technician, or a professor in the public sector). or older age. 15 Page 8 of 17 X. Ma Table 1 The proportion of using the Internet by sex be considered that the IVs are valid. The results of the F test and the Breusch–Pagan Lagrange multiplier test Unit: % indicate that both the FE and RE models are appropriate Total Males Females Gap (M-F) compared to the OLS model. The Hausman test results (2308.67, p = 0.000) suggest that the FE model is more Total appropriate than the RE model. The main findings are 2014 34.84 37.82 32.08 5.74 summarized as follows. 2016 49.88 52.93 46.95 5.98 First, the coefficients of Internet use are significantly 2018 62.41 65.19 59.74 5.46 positive in Models 1–5, suggesting that Internet use may Low education increase the wage levels when addressing individual het- 2014 6.61 8.37 5.61 2.77 erogeneity and other endogeneity problems. The results 2016 15.72 17.46 14.72 2.75 can be explained based on human capital theory consid- 2018 34.04 38.21 31.18 7.03 ering Internet use skill as a kind of human capital, which Middle education can potentially increase labor productivity. Addition- 2014 41.21 40.42 42.14 − 1.71 ally, Internet use may have a signaling effect whereby an 2016 56.15 55.10 57.39 − 2.29 employer may evaluate that an Internet-using worker 2018 69.52 68.60 70.60 − 2.00 possesses higher skills than a non-Internet-using worker. High education Second, in terms of the gender differences in return 2014 65.98 68.89 63.20 5.68 to Internet use, the results of interaction item of Inter- 2016 74.28 75.63 72.81 2.82 net use and female dummy variable in Model 1 (OLS), 2018 94.73 94.64 94.82 − 0.19 Model 3 (LV), and Model 4 (RE) indicate that the return Born before 1979 to Internet use is significantly greater for women than 2014 6.03 8.14 3.80 4.34 men. However, considering the endogeneity problem, 2016 11.32 12.75 9.73 3.02 the wage premium is significantly greater for men than 2018 19.45 21.65 17.01 4.64 women in Model 2 (IV). Additionally, although the result Born in 1979–1989 is non-significant in Model 5 (FE), it is a negative value. 2014 32.44 36.46 28.81 7.65 The results suggest a bias in the results derived from the 2016 47.36 51.41 43.56 7.85 OLS model. When using appropriate models (e.g., the IV 2018 58.55 61.95 55.33 6.62 or FE model) to address the endogeneity problem, the Born after 1990 return to Internet use is higher for men than women. 2014 81.28 81.94 80.69 1.25 2016 89.31 90.24 88.41 1.83 4.3 Estimations on the intensive margin effect of Internet 2018 93.14 93.45 92.83 0.62 use Source: Authors’ calculations based on the data from CFPS of 2014, 2016, and As shown in Table  2, although we analyzed the exten- sive margin effects of Internet use on the gender wage Gap = percentage of male Internet user-percentage of female Internet user gap using a binary variable (whether used Internet), Age range: 16–60 years the intensive margin effect of Internet use (e.g., the fre - Low education: elementary school and lower; Middle education: junior and senior high school; High education: college and higher quency of using the Internet) on the wage gap should also be considered. Additionally, the effects of Internet use on the wage gap may differ for different purposes of Inter - LV (Model 3), RE (Model 4), and FE (Model 5). Regard- net use. For example, when comparing the group using ing the appropriateness of the IVs, the endogeneity test the Internet frequently for leisure (e.g., entertainment, result (Durbin–Wu–Hausman test) is statistically signifi - shopping) to the group using it frequently for work or cant at the 1% level; therefore, the null hypothesis con- study, it is assumed that the group using the Internet fre- sidering that all the variables are exogenous is rejected. quently for work or study is likely to obtain higher earned The Hansen J statistic is not significant, thus revealing income. To investigate the intensive margin effects of that the IV is exogenous in the second stage estima- Internet use, we used the three indices: (i) the frequency tion. Furthermore, the Cragg-Donald Wald F statistic is of using the Internet for work and study, (ii) the fre- 26.393, which is larger than 10, suggesting that the weak quency of using the Internet for communication, and (iii) identification problem can be neglected. Therefore, it can the frequency of using the Internet for leisure including shopping and entertainment, to replace the binary varia- ble of Internet use in Table 2 and re-estimate the models. Due to space limit constraints, results, including all control variables, were The results are presented in Table 3. reported in Additional file 1: Table S1. Internet use and gender wage gap: evidence from China Page 9 of 17 15 Table 2 The gender differences in return to Internet use (1) OLS (2) IV (3) LV (4) RE (5) FE Coef t Coef t Coef t Coef t Coef t Internet use 0.603*** 6.68 14.900*** 5.59 0.107 0.98 0.606*** 6.65 0.405** 2.26 Female − 0.849*** − 11.29 4.054*** 4.40 − 0.697*** − 5.76 − 0.844*** − 10.90 – Internet use × Female 0.540*** 5.19 − 8.881*** − 5.04 0.360** 2.51 0.531*** 5.02 − 0.106 − 0.42 Control variables Yes Yes Yes Yes Yes First stage estimation IV1 0.083*** 3.47 IV2 − 0.243*** − 11.58 Observations 18,381 18,381 7777 18,381 18,381 R-squared 0.476 0.367 12,876 12,876 R-sq. Between 0.190 0.197 Within 0.574 0.017 Overall 0.476 0.008 F-test (Prob > F) 0.000 BP test (Prob > chibar2) 66.51 (p = 0.000) Hausman test (Prob > chi2) 2308.67 (p = 0.000) Endogeneity test (DWH) 69.216 (p = 0.000) Hansen J statistic p = 0.224 Cragg-Donald Wald F statistic 26.393 Source: Authors’ Calculations Based on the data from CFPS of 2014, 2016 and 2018 Control variables, including years of schooling, years of work experience and its squared term, party membership, urban, occupation, industry sector, workplace ownership, region, and survey year variables, have been calculated, but the results are not listed in the table owing to space limit constraints OLS: ordinary least squares; IV: instrumental variable method; LV: lagged variable model; RE: random effects model; FE: fixed effects model; BP test: Breusch–Pagan Lagrange multiplier test; DWH: Durbin–Wu–Hausman test; IV1: the provincial optical cable circuit in 1999; IV2: the provincial long-distance cable line length in 1999 ***p < 0.01; **p < 0.05; *p < 0.1 15 Page 10 of 17 X. Ma Table 3 Frequency of Internet use for different purposes (1) Work and study (2) Communication (3) Living IV FE IV FE IV FE Internet use 2.020*** − 0.003 2.540*** 0.056* 1.385*** 0.010 (4.96) (−0.14) (5.59) (1.94) (6.58) (0.45) Female 2.709*** 2.987*** 1.922*** (3.77) (4.21) (4.44) Internet use × Female − 1.317*** − 0.008 − 1.540*** − 0.016 − 0.797*** − 0.015 (− 4.66) (− 0.29) (− 5.11) (− 0.39) (− 5.93) (− 0.54) First stage estimation IV1 0.204*** 0.151*** 0.411*** (4.66) (4.91) (9.78) IV2 − 0.351*** − 0.249*** − 0.612*** (− 9.55) (− 9.63) (− 17.32) Control variables Yes Yes Yes Yes Yes Yes Observations 18,605 18,605 18,605 18,605 18,605 18,605 Groups 12,920 12,920 12,920 R-sq. Between 0.192 0.198 0.192 Within 0.015 0.030 0.015 Overall 0.007 0.033 0.006 F test (Prob > F) 0.000 0.000 0.000 BP test (Prob > chibar2) 73.70 (p = 0.000) 77.77 (p = 0.000) 77.22 (p = 0.100) Hausman test 2406.81 (p = 0.000) 2355.36 (p = 0.000) 2380.03 (p = 0.000) Endogeneity test (DWH) p = 0.000 p = 0.000 p = 0.000 Hansen J statistic p = 0.917 p = 0.082 p = 0.482 Cragg-Donald Wald F statistic 18.682 26.134 47.343 Source: Authors’ Calculations Based on the data from CFPS of 2014, 2016 and 2018 Control variables, including years of schooling, years of work experience and its squared term, party membership, urban, occupation, industry sector, workplace ownership, region, and year variables have been calculated but are not listed in the table owing to space limit constraints IV: Instrumental variable method; FE: fixed effects model; BP test: Breusch–Pagan Lagrange multiplier test; DWH: Durbin–Wu–Hausman test ***p < 0.01; **p < 0.05; *p < 0.1. t-values are shown in parentheses The results from the Hausman specification test, F-test, female dummy variables are negative values and signifi - and the Breusch–Pagan Lagrange multiplier test indicate cant at the 1% level in Models 1–3, but they are not sig- that the FE model is more appropriate than the OLS and nificant in the results from the FE model. These results RE models. Regarding the validity of the IV method, the are consistent with those in Table 2. results in the first stage estimations indicate that both IVs significantly affect the likelihood of using the Internet at 4.4 Estimations considering heterogeneous group the 1% level; the results from the Durbin–Wu–Hausman The results for the heterogeneous group based on edu - test, Hansen J statistic, and Cragg-Donald Wald F statistic cational attainment and age cohort are summarized in values indicate that the IV method is valid. Therefore, we Tables 4, 5. The results from the Hausman specification report the results using the FE and IV models in Table 3. test, F-test, and the Breusch–Pagan Lagrange multiplier First, the results from the IV method indicate that the test indicates that the FE model is more appropriate than coefficients of Internet use are positive and significant at the OLS and RE models. Regarding the validity of the IV the 1% level in Models 1–3. It is also positive and signifi - method, the results in the first stage estimations indicate cant at the 10% level in Model 2 from the FE model. The that both IVs significantly affect the likelihood of using conclusions are consistent with the results in Table 2. the Internet at the 1% level; the results from the Durbin– Second, the results from the IV methods show that the Wu–Hausman test, Hansen J statistic, and Cragg-Donald coefficients of the interaction term of Internet use and Wald F statistic values indicate that the IV method is 11 12 Due to space limit constraints, results, including all control variables, were Due to space limit constraints, results, including all control variables, were reported in Additional file 1: Table S2. reported in Additional file 1: Tables S3 and S4. Internet use and gender wage gap: evidence from China Page 11 of 17 15 Table 4 The gender differences in return to Internet use by educational group (1) Low (2) Middle (3) High IV FE IV FE IV FE Internet use − 11.427 0.496 10.277*** 0.475** 19.852** − 0.083 (−0.94) (0.93) (4.60) (2.22) (2.37) (− 0.14) Female − 2.598 3.403*** 16.978** (− 1.41) (3.50) (2.25) Internet use × Female 9.648 0.139 − 6.856*** − 0.370 − 18.215 − 0.033 (− 0.97) (0.21) (− 4.40) (− 1.21) (− 2.28) (− 0.04) First stage estimation IV1 − 0.167*** 0.165*** 0.466*** (− 3.71) (5.40) (4.58) IV2 − 0.060 − 0.277*** − 0.472*** (− 1.23) (− 10.83) (− 5.70) Control variables Yes Yes Yes Yes Yes Yes Observations 4652 4652 10,284 10,284 3445 3445 Groups 3289 7070 1905 R-sq. Between 0.017 0.200 0.200 Within 0.008 0.004 0.004 Overall 0.000 0.001 0.001 F test (Prob > F) 0.000 0.000 0.000 BP test (Prob > chibar2) 0.000 (p = 0.987) 17.67 (p = 0.000) 19.58 (p = 0.000) Hausman test 248.75 (p = 0.000) 874.72 (p = 0.000) 874.72 (p = 0.000) Endogeneity test (DWH) p = 0.131 p = 0.000 p = 0.000 Hansen J statistic p = 0.001 p = 0.652 p = 0.127 Cragg-Donald Wald F 1.199 21.914 19.93 statistic Source: Authors’ Calculations Based on the data from CFPS of 2014, 2016 and 2018 Low: elementary school and lower; middle: junior and senior high school; high: college and higher Control variables, including years of schooling, years of work experience and its squared term, party membership, urban, occupation, industry sector, workplace ownership, region, and survey year variables, have been calculated, but the results are not listed in the table owing to space limit constraints IV: Instrumental variable method; FE: Fixed effects model; BP: test Breusch–Pagan Lagrange multiplier test; DWH: Durbin–Wu–Hausman test ***: p < 0.01; **: p < 0.05; *: p < 0.1. t-values are shown in parentheses valid in most cases. Therefore, we report the results using First, in terms of the return to Internet use by the educa- the FE and IV models in Tables 4, 5. tional group, the results from the IV method show that the Table  4 presents the low-, middle- and high-educa- coefficients of Internet use are positive and significant at tional group results. To secure enough samples for the the 1% level in the middle- and high-educational groups, analysis and consider the distribution of workers by edu- whereas it is not significant for the low-education group. The cation attainment levels, we distinguished the samples results from the FE model reveal that the Internet use coef- into three groups (i) the low-education group (elemen- ficient is significant only for the middle education group at tary school and lower); (ii) the middle-education group the 5% level. They suggest the effects of the return to Internet (junior and senior high school); and (iii) the high-educa- use on wages are much more significant for the middle- and tion group (college and higher). The main findings are as high-educational groups. Additionally, the individual hetero- follows. geneity problem considerably affects the results. Second, in terms of the gender difference in return to Internet use by the educational group, the results from the IV method show that the coefficient of the interaction of Internet use and female dummy is negative and signifi - The distribution proportion of workers by the educational group in the samples is 25.3% for the elementary school and lower group, 55.9% for the cant at the 5% level for the middle-and high-educational junior and senior high school group, and 18.8% for the college and above group. 15 Page 12 of 17 X. Ma Table 5 The gender disparity in return to Internet use by age cohort (1)Born before 1969 (2) Born from 1970–1989 (3) Born after1990 IV FE IV FE IV FE Internet use 28.375*** − 0.822 12.618*** 0.341* − 24.696 1.322 (3.41) (− 1.32) (4.88) (1.75) (− 0.90) (2.62) Female 0.077 3.510*** − 22.009 (0.28) (3.78) (− 0.97) Internet use × Female − 23.690*** 1.404 − 8.217*** 0.011 24.273 − 1.207 (− 3.25) (1.14) (− 4.45) (0.04) (0.96) (− 1.25) First stage estimation IV1 0.290*** 0.088*** − 0.032 (5.09) (3.32) (− 0.44) IV2 − 0.383*** − 0.236*** − 0.155* (− 11.13) (− 9.99) (− 1.91) Control variables Yes Yes Yes Yes Yes Yes Observations 3.512 3.512 12,701 12,701 2.168 2.168 Groups 2256 9229 1.676 R-sq. Between 0.230 0.198 0.200 Within 0.002 0.030 0.004 Overall 0.000 0.033 0.001 F test (Prob > F) 0.999 0.000 0.000 BP test (Prob > chibar2) 0.000 (p = 1.000) 49.00 (p = 0.000) 1.63 (p = 0.100) Hausman test 67.74 (p = 0.000) 2736.48 (p = 0.000) 57.86 (p = 0.000) Endogeneity test (DWH) p = 0.000 p = 0.000 p = 0.083 Hansen J statistic p = 0.554 p = 0.604 p = 0.625 Cragg-Donald Wald F statistic 11.590 20.405 11.201 Source: Author’s calculations based on the data from CFPS of 2014, 2016 and 2018 Control variables, including years of schooling, years of work experience and its squared term, party membership, urban, occupation, industry sector, workplace ownership, region, and survey year variables, have been calculated, but are not listed in the table owing to space limit constraints IV: Instrumental variable method; FE: Fixed effects model; BP test: Breusch–Pagan Lagrange multiplier test; DWH: Durbin–Wu–Hausman test ***p < 0.01; **p < 0.05; *p < 0.1. t-values are shown in parentheses groups, while it is insignificant for the low-education among the high-education group is serious much more group, thus indicating when the other factors are held than that in the low- and middle-educational groups. consistent, the return to Internet use is lower for female Table 5 summarizes the results by three age cohorts: the workers than male workers, and the gender difference in younger (born after 1990), middle-aged (born in 1970— return to Internet use is higher in the middle-and high- 1989), and older (born before 1969) generations. First, educational groups than the low-education group. How- the results from the IV method show that the coefficients ever, the results from the FE model are not significant in of Internet use are positive and significant at 1% levels in three educational groups. both the middle-aged and older cohort groups, while they Third, comparing the magnitude of the coefficients are not significant for the younger cohort. The FE model from the IV method show that the gender difference in results reveal that the coefficients of Internet use are sig - return to Internet use is larger in the high-education nificant for both the younger and middle-aged cohort group (−  18.215) than in the middle education group groups at 5 and 10% levels, while it is not significant for the (− 6.856). The possible reasons can be considered as fol - older cohort. The results suggest that, in general, there is a lows: the discrimination against high-educated women in positive effect of Internet use on wages in each age cohort the workplace, which may be caused by the glass-ceiling group, and the influence of individual heterogeneity on the problem, or due to the gender occupational segregation return to Internet use is greater for the older cohort. Internet use and gender wage gap: evidence from China Page 13 of 17 15 Table 6 Decomposition results of Internet use and the gender Third, the gender difference in return to Internet use wage gap was found higher in the older cohort than the other age cohorts when compared using the magnitude of the coef- Value Percentage ficients based on the IV method. It may be caused by Explained Unexplained Explained Unexplained that the discrimination against female workers is serious (%) (%) much more among older generations than those among Total 0.471 0.449 51.2 48.8 the younger and middle-aged generations. Internet use 0.461 0.665 50.1 72.4 Education − 0.052 0.005 − 5.7 0.5 4.5 Decomposition results Experience − 0.044 1.087 − 4.8 118.2 The results in Table  1 indicate that Internet access dif- Ethnicity − 0.002 − 0.107 − 0.2 − 11.6 fers by gender, and those in Tables 2, 3 suggest that the Party − 0.027 − 0.029 − 3.0 − 3.2 return to Internet use is different for men and women. Occupation − 0.018 − 0.005 − 2.0 − 0.5 However, how the two components affect the forma - Industry 0.021 − 0.160 2.3 − 17.4 tion of the gender wage gap is unclear. Therefore, we State- 0.000 − 0.188 0.0 − 20.4 conduct a decomposition analysis to calculate the con- owned tribution rates of these two components (Table  6). We Urban 0.067 0.037 7.3 4.0 also perform the decomposition analyses based on the Region 0.001 0.219 0.1 23.8 educational attainment and age cohort (Tables 7, 8). Year 0.064 0.002 7.0 0.2 Table  6 reports the decomposition results for the Constant 0.000 − 1.077 0.0 − 117.1 total sample. First, the explained and unexplained com- Source: Authors’ Calculations Based on the data from CFPS of 2014, 2016 and ponents contribute to the formation of the gender wage The decomposition is based on the results from the IV method gap. The influence is slightly less for the former (46.3%) than the latter (53.7%). Second, in terms of the effects of Internet use, it is Second, in terms of the gender difference in return to shown that both the gender disparity in Internet access Internet use by age cohorts, the results from the IV method and gender difference in the return to Internet use show that the interaction coefficients of Internet use and contribute to widening the gender wage gap, and the female dummy are negative values and significant at 1% or contribution rates of both are higher than the other 5% level for both middle-aged and older cohorts, and not factors (e.g., education, occupation). Additionally, the significant for the younger cohort. The results from the FE contribution rate is larger for the return to Internet use model are not significant in the three age cohorts. Table 7 Decomposition results of Internet use and the gender wage gap by educational group Value Percentage Explained Unexplained Explained (%) Unexplained (%) (a) Low (N = 4652) Total 0.579 1.019 36.2 63.8 Internet use − 0.842 − 2.097 − 52.7 − 131.3 Other variables 1.421 3.116 88.9 195.1 (b) Middle (N = 10,284) Total 0.051 0.357 12.4 87.6 Internet use − 0.224 − 0.484 − 54.9 − 118.7 Other variables 0.275 0.841 67.3 206.3 (c) High (N = 3445) Total − 0.113 0.229 − 97.3 197.3 Internet use − 0.255 3.480 − 219.6 2996.8 Other variables 0.142 − 3.251 122.3 − 2799.5 Source: Authors’ Calculations Based on the data from CFPS of 2014, 2016 and 2018 The decomposition is based on the results from the IV method Low: elementary school and lower; Middle: junior and senior high school; High: college and higher “Other variables” include years of schooling, years of work experience, party membership, urban, occupation, industry sector, workplace ownership, region, and years 15 Page 14 of 17 X. Ma Table 8 Decomposition results of Internet use and the gender wage gap by age cohort Value Percentage Explained Unexplained Explained (%) Unexplained (%) (a) Born before 1969 (N = 3512) Total 1.000 1.162 46.3 53.7 Internet use 1.023 − 2.401 47.3 − 111.0 Other variables − 0.023 3.563 − 1.0 168.3 (b) Born between 1970–1989 (N = 12,701) Total 0.538 0.445 54.7 45.3 Internet use 0.536 − 0.004 54.6 − 0.4 Other variables 0.002 0.449 0.1 45.7 (c) Born after 1990 (N = 2168) Total 0.132 0.136 49.1 50.9 Internet use 0.044 − 3.022 16.3 − 1129.4 Other variables 0.088 3.158 32.8 1180.3 Source: Author’s Calculations Based on the data from CFPS of 2014, 2016 and 2018 The decomposition is based on the results from the IV method “Other variables” include years of schooling, years of work experience, party membership, urban, occupation, industry sector, workplace ownership, region, and years (72.4%) than the Internet access (50.1%). The results Second, the gender difference in return to Internet use indicate that although both components drive the gen- reduces the gender wage gap in three cohorts; the effects der wage gap in China, the effect of the gender differ - are greater for the younger age cohorts. ence in return to Internet use is greater, suggesting that The results of Table  8 suggest that the gender disparity workplace discrimination against women in terms of in Internet access in the middle/older age cohorts and the Internet use is a serious issue in China. workplace discrimination against the younger women in Table 7 shows the decomposition results based on the terms of Internet use widens the gender wage gap in China. These results contradict those of Qi and Liu (2020), educational attainment group. The results indicate that who report that both components reduce the gender the effects of gender differences in both the Internet wage gap in the younger, middle-aged, and older groups. access and return to Internet use differ across different There are two reasons: first, the method of analysis dif educational groups. - First, the gender disparity in Internet access reduces fers. This study conducts the decomposition analysis the wage gap in three educational groups; its effect is based on the IV method, whereas Qi and Liu (2020) use greater in the high-education group than those in the the OLS model; it can be concluded that an endogeneity low- and middle-education groups. problem such as the unobservable omitted variable issue Second, the gender difference in return to Internet use might exist in the earlier studies. Second, the period of widens the gender wage gap in the high-education group, analysis in this study is 2014–2018, whereas it is 2010– whereas they reduce the wage gap in the low- and mid- 2015 in Qi and Liu (2020). With Internet diffusion and dle-education groups. Internet technology progressing, the effects of both com - Hence, these results suggest that the workplace dis- ponents may change over time. The results in this study crimination against female workers is greater for the high suggest that the gender divisions in Internet accessibility education group in terms of Internet use than the other might become severe in the middle-aged and older age group, which widens the gender wage gap in China. cohorts in the recent period (2014–2018). Table  8 presents the results of the decomposition analysis for three age cohorts: (i) the older cohort (born 4.6 Further discussions on the limitations of this study before 1969), (ii) the middle-aged cohort (born between It should be noted that this study has several limita- 1979 and 1989), and (iii) the younger cohort (born after tions. First, although we used the IV, LV, RE, and FE 1990). models to attempt to address the endogeneity problem, First, the gender disparity in Internet access widens future research should also explore the causal association the gender wage gap in three age cohorts; the effects are between Internet use and the gender wage gap. greater for the middle-aged (54.6%) and older age cohorts Second, the gender gap in the return to Internet (47.3%) than for the younger age cohorts (16.3%). usage may also result from the gender disparity in Internet use and gender wage gap: evidence from China Page 15 of 17 15 Internet-using abilities (or skills). Ge and Zhou (2020) Fourth, the influence of Internet usage on the gender report that the firm attributes (e.g., capital, trade expo - wage gap varies with the educational attainment and age sure) and firm technology level (e.g., robot or computer cohort groups. For example, the gender difference in the use situations) affect the gender wage gap. Future studies return to Internet use widens the gender wage gap in the can conduct a detailed survey on individuals’ Internet use high-education group while they reduce the wage gap in skills and workplace technology levels. the low- and middle-education groups; the influences Third, although the occupation, industry sector, and of gender disparity in Internet access on the formation enterprise ownership type were used to control the influ - of gender wage gap are greater for the middle-aged and ence of the workplace on the gender wage gap, other factors older generations than the younger generation. (e.g., wage and employment systems, enterprise attrib- The study highlights two policy implications. First, utes) may also affect the wage gap. We should conduct an there exists a large gender difference in return to Internet employer-employee survey on the issue in the future. use even when the other factors (e.g., education, occupa- Fourth, some studies have found that the work prefer- tion) are held constant, which highlights the prevalence ences differ by gender (e.g., Beblo and Görges 2018). This of workplace discrimination against women in terms of gender preference disparity in using new technology can Internet use, especially for the higher-educated women. also affect Internet access and return to Internet use. Fur - The enforcement of the implementation of equality poli - ther research should consider the influence of personality cies, such as the equality employment policy and “equal factors and self-selection on the issue. pay for equal work” policy, can be expected to reduce the Finally, due to the limitation of the survey period, we gender wage gap. Moreover, the discrimination against only investigated the issue in the current period (2014, women may be caused by the more family responsi- 2016 and 2018), and the longer-term analysis has become bilities for women than men (Connelly et  al. 2018; Ma a new challenge in the future. 2021a, b). Thus, policies that reduce the responsibili - ties of childcare and geriatric care, such as the one to promote the establishment of public kindergartens and 5 Conclusions long-term care insurance, are also expected to close the Using national longitudinal data from CFPS of 2014, gender wage gap in the long run. Second, the results 2016, and 2018, this study empirically analyzed the influ - suggest that the policies aimed at the reduction of the ence of Internet use on the gender wage gap in China, Internet access disparities among various groups, such considering the endogeneity problems. It yields the fol- as women in the middle-aged and older generations, may lowing four main conclusions. contribute to reducing the gender wage gap. First, according to the results derived from the OLS Despite these limitations mentioned above, this study model, the return to Internet usage is higher for women investigated the influence of Internet use on the gen - than men, meaning the results are similar to those of earlier der wage gap and provided new evidence on the deter- studies. However, when longitudinal survey data is used to minants of the gender wage gap in the era of the digital address heterogeneity and other endogeneity issues based economy from China. We anticipate that the insights on the IV and FE models, the results show that the return about the gender disparities in Internet access and gen- to Internet usage is higher for men than women. The indi - der difference in return to Internet use (including the vidual heterogeneity problem considerably affects the esti - discrimination against women in terms of Internet use mations, thus suggesting an estimation bias in the existing in the workplace), that contribute to the formation of the literature. The results based on the frequency of Internet gender wage gap in China, can provide valuable lessons use for different purposes confirm the conclusions. for other countries as well. Second, the gender difference in return to Internet use differs by heterogeneous groups: it is higher in the mid - Supplementary Information dle-/high- education groups and middle-aged/older age The online version contains supplementary material available at https:// doi. cohorts than those in low education and younger age org/ 10. 1186/ s12651- 022- 00320-9. cohorts. Third, the decomposition results indicate that, in gen - Additional file 1: Table S1. The gender differences in return to Internet use. Table S2. Results using frequency of Internet use for different eral, the two components (the gender disparity in Inter- purposes. Table S3. The gender differences in return to Internet use by net access and the gender difference in return to Internet educational group. Table S4. The gender differences in return to Internet use) drive the gender wage gap in China; the effect of use by age cohort. gender difference in the return to Internet use is greater. 15 Page 16 of 17 X. Ma Acknowledgements Beaudry, P., Lewis, E.: Do male–female wage differentials reflect differences in The authors are grateful to professor Boris Hirsch, and anonymous reviewers the return to skill? Cross-city evidence from 1980–2000. Am. Econ. J. Appl. for their helpful comments and suggestions. Econ. 6, 178–194 (2014) Beblo, M., Görges, L.: On the nature of nurture: The malleability of gender dif- Author contributions ferences in work preferences. J. Econ. Behav. Organ. 151, 19–41 (2018) X. MA designed the study, collected the data, performed the formal analyses, Becker, G.S.: The economics of discrimination. University of Chicago Press, and wrote the original manuscript; the author read and approved the final Chicago (1957) manuscript. Becker, G.S.: Human capital: a theoretical and empirical analysis, with special reference to education. Columbia University Press, New York (1964) Author information Bergmann, B.R.: Occupational segregation, wages and profits when employers Dr. X. MA is a professor at the Faculty of Economics, Hosei University, Japan. discriminate by race and sex. East. Econ. J. 1, 103–110 (1974) She was the editor of the Japanese Journal of Comparative Economics, Asian Biewen, M., Fitzenberger, B., Seckler, M.: Counterfactual quantile decomposi- Studies, and the Journal of Chinese Economics. Her articles have been published tions with selection correction taking into account Huber/Melly (2015): in peer-reviewed journals such as China Economic Review, Journal of Asian Eco- an application to the German gender wage gap. Labour Econ. 67, 101927 nomics, Economic Systems, Education Economics and Journal for Labor Market (2020) Research. She is the author of Economic Transition and Labor Market Reform in Bimber, B.: Measuring the gender gap on the Internet. Soc. Sci. Q. 81, 868–876 China (Palgrave Macmillan 2018), Female Employment and Gender Gap in China (2000) (Springer 2021), and the single editor of Employment, Retirement and Lifestyle in Blinder, A.: Wage discrimination: reduced form and structural estimation. Hum. Aging East Asia (Palgrave Macmillan 2021). Resour. Manag. Res. 8, 436–455 (1973) Borghans, L., Ter Weel, B.T., Weinberg, B.A.: People skills and the labor-market Funding outcomes of underrepresented groups. ILR Rev. 67, 287–334 (2014) This research was supported by JSPS (Japan Society for the Promotion Sci- Broadband commission: Doubling digital opportunities: enhancing the inclu- ence) Grant-in-Aid for Scientific Research (Grant Numbers: 20H01512 and sion of women and girls in the information society. UNESCO ITU (2013). 20H01489). Brown, R.S., Moon, M., Zoloth, B.S.: Incorporating occupational attainment in studies of male/female earnings differentials. J. Hum. Resour. 15, 3–28 Data availability (1980) The data can be obtained through the website https:// opend ata. pku. edu. cn/ Cao, J., Jiang, T.: The impact of Internet usage on women’s income: An analysis datav erse/ CFPS? langu age= en. The datasets constructed by the author are based on CFPS data. J. Tianjin Univ. Fin. Econ. 371, 79–95 (2020). (曹景林, available upon request. 姜甜:互联网使用对女性收入的影响-基于CFPS数据的经验证据.《天 津财经大学学报》, 371(12), 79–95 (2020)) Chen, X.: The evolution of female labour force participation in urban China. Declarations Econ. Transit. 27, 267–299 (2019) Chen, L., Fan, H.: Does family elderly care reduce female employment? Two- Ethics approval and consent to participate Stage Residual Inclusion Approach. Popul. Res. 40, 71–81 (2016). (陈璐,范 The dataset used in this study, the China Family Panel Studies (CFPS), is pub- 红丽:家庭老年照料会降低女性劳动参与率吗?—基于两阶段残差介 licly available (http:// opend ata. pku. edu. cn/ en). The medical ethics committee 入法的实证分析.《人口研究》,40(3), 71–81(2016)) approved the CFPS study, and all interviewees were required to sign informed Chen, L., Fan, H., Zhao, N., Chu, L.: The impact of informal care on employ- consent. Ethics approval for the data collection in CFPS was obtained from ment for women in China. Econ. Res. J. 51, 176–189 (2016). (陈璐,范红 the Biomedical Ethics Review Committee of Peking University (IRB00001052- 丽,赵娜,褚兰兰:家庭照料对女性就业的影响.《经济研究》,51(3), 14010). Ethics approval for the use of CFPS data was obtained from the Univer- 176–189(2016)) sity of Newcastle Human Research Ethics Committee. And all the participants China Family Panel Studies (CFPS): https:// opend ata. pku. edu. cn/ datav erse/ gave their written informed consent before any study procedures began. CFPS? langu age= en. Accessed 20 May 2020. Connelly, R., Dong, X., Jacobsen, J., Zhao, Y.: The care economy in post-reform Consent for publication China: feminist research on unpaid and paid work and well-being. Fem. The author of this study declares that this manuscript does not infringe Econ. 24, 1–30 (2018) copyright, moral rights, or other intellectual property rights of any other Cook, S., Dong, X.Y.: Harsh choices: Chinese women’s paid work and unpaid person. The author of this study testifies that this paper is original and has not care responsibilities under economic reform. Dev. Change 42, 947–965 incurred any type of plagiarism. The author of this study declares that the arti- (2011) cle is original, has not already been published in a journal, and is not currently Deyyling, L.J.: Broadband in the labor market: the impact of residential high- under consideration by another journal. The author of this study agrees to the speed internet on married women’s labor force participation. Ind. Lab. terms of the Springer Open Copyright and License Agreement. Relat. Rev. 70, 451–482 (2017) DiMaggio, P., Hargittai, E., Neuman, W.R., Robinson, J.P.: Social implications of Competing interests the internet. Annu. Rev. Sociol. 27, 307–336 (2001) The author declares no conflicts of interest associated with this study. Eastin, M.S., Cicchirillo, V., Mabry, A.: Extending the digital divide conversation: examining the knowledge gap through media expectancies. J. Broadcast. Received: 19 March 2022 Accepted: 10 August 2022 Electron. Media 59, 416–437 (2015) Fatehkia, M., Kashyap, R., Weber, I.: Using Facebook ad data to track the global digital gender gap. World Dev. 107, 189–209 (2018) Ge, Y.: Sectoral choice and the gender wage gap: 1988–2001. China Econ. Q. 2, 607–628 (2007). (葛玉好: 部门选择对工资性别差距的影响:1988– References 2001年. 《经济学季刊》,2,607–628 (2007)) Alam, K., Mamun, S.A.K.: Access to broadband Internet and labour force out- Ge, S., Zhou, Y.: Robots, computers, and the gender wage gap. J. Econ. Behav. comes: a case study of the Western Downs Region. Queensland. Telemat. Organ. 178, 194–222 (2020) Inform. 34, 73–84 (2017) Gustafsson, B., Li, S.: Economic transformation and the gender earnings gap in Alozie, N.O., Akpan-Obong, P.: The digital gender divide: confronting obstacles urban China. J. Popul. Econ. 13, 305–329 (2000) to women’s development in Africa. Dev. Policy Rev. 35, 137–160 (2017) Hafkin, N.J., Huyer, S.: Women and gender in ICT statistics and indicators for Arrow, K.J.: The theory of discrimination. In: Asehenfelter, O., Rees, A. (eds.) Dis- development. Inf. Technol. Dev. 4, 25–41 (2007) crimination in Labor Market. Princeton University Press, Princeton (1973) Haight, M., Quan-Haase, A., Corbett, B.A.: Revisiting the digital divide in Atasoy, H.: The effects of broadband Internet expansion on labor market Canada: the impact of demographic factors on access to the internet, outcomes. ILR Rev. 66, 315–345 (2013) Internet use and gender wage gap: evidence from China Page 17 of 17 15 level of online activity, and social networking site usage. Inf. Commun. Organization for Economic Co-operation and Development (OECD): Bridging Soc. 17, 503–519 (2014) the digital gender divide: include, upskill, innovate. OECD, Paris (2018) Hirsch, B.: Spatial Monopsony and the Gender Pay Gap. Springer, New York Pabilonia, S.W., Zoghi, C.: Returning to the returns to computer use. Am. Econ. (2010) Rev. 95, 314–317 (2005) Iwasaki, I., Ma, X.: Gender wage gap in China: a large meta-analysis. J. Lab. Phelps, E.S.: The statistical theory of racism and sexism. Am. Econ. Rev. 62, Mark. Res. 54, 17 (2020) 659–661 (1972) Kidd, M.P.: Sex discrimination and occupational segregation in the Australian Qi, Y., Liu, C.: Does the use of internet reduce the difference in gender pay in labour market. Econ. Rec. 69, 44–55 (1993) the context of digital economy? Econ. Theory and Econ. Admi. 9, 70–87 Kidd, M.P., Shannon, M.: Does the level of occupational aggregation affect esti- (2020). (戚聿东,刘翠花:数字经济背景下互联网使用是否缩小了性别 mates of the gender wage gap. Ind. Lab. Relat. Rev. 49, 317–329 (1996) 工资差异.《经济理论与经济管理》, 9, 70–87 (2020)) Krueger, A.B.: How computers have changed the wage structure: evidence Rice, R.E., Katz, J.E.: Comparing internet and mobile phone usage: Digital from microdata, 1984–1989. Q. J. Econ. 108, 33–60 (1993) divides of usage, adoption, and dropouts. Telecommun. Policy 27, Li, S., Ma, X.: Occupational segregation and gender wage differentials in urban 597–623 (2003) China. Chin. J. Popul. Sci. 5, 2–13 (2006). (李实,马欣欣:中国城镇职工 Riggins, F.J., Dewan, S.: The digital divide: Current and future research direc- 的性别工资差异与职业分割的经验分析.《中国人口科学》5, 2–13 tions. J. Assoc. Inf. Syst. 6, 298–337 (2005) (2006)) Scheerder, A., van Deursen, A., van Dijk, J.: Determinants of internet skills, uses Li, S., Yang, X.: The determinants of gender wage gaps in migrants Comp. Econ. and outcomes. A systematic review of the second- and third-level digital Soc. Syst. 5, 82–89 (2010). (李实,杨修娜:农民工工资的性别差异及其影 divide. Telemat. Inform. 34, 1607–1624 (2017) 响因素.《经济社会体制比较》,5, 82–89 (2010)) Vick, B.: Measuring links between labor monopsony and the gender pay gap in Li, S., Song, J., Liu, X.: Evolution of the change of the gender wage gap among Brazil. IZA J. Dev. Migr. 7, 10 (2017) China’s urban employees. Manag. World 3, 53–65 (2014). (李实, 宋锦, 刘 Wang, M.: Work chance and wage differentials in the urban labor market: Work 小川:中国城镇职工性别工资差距的演变.《管理世界》, 3, 53–65 and wage of migrants. Chin. Soc. Sci. 5, 36–46 (2005). (王美艳: 城市劳动 (2014)) 力市场上的就业机会与工资差异-外来劳动力就业与报酬研究. 《中 Liu, P., Meng, X., Zhang, J.: Sectoral gender wage differentials and discrimina- 国社会科学》,5, 36–46 (2005)) tion in the transitional Chinese economy. J. Popul. Econ. 13, 331–352 Wang, M., Cai, F.: Gender earnings differential in urban China. Rev. Dev. Econ. (2000) 12, 442–454 (2008) Liu, S., Zhang, X., Yang, Z.: Impact of Internet use on rural residents’ income. World Bank, http:// data. workd bank. org/ indic ator. Accessed 30 Jun 2022. Quant. Econ. Tech. Econ. Res. 38, 103–119 (2021). (刘生龙,张晓明,杨 Wu, E.: Analysis of Internet usage and gender wage difference. Coop. Econ. 竺松:互联网使用对农村居民收入的影响.《数量经济技术经济研 Technol. 16, 138–140 (2021). (吴恩喆:互联网使用与性别工资差异分 究》,38(4), 103–119 (2021)) 析.《合作经济与科技》, 6, 138–140 (2021)) Ma, X.: Ownership sector segmentation and the gender wage gap in urban Zhao, Y., Li, X.: The impact of internet use on individual health: Based on the China during the 2000s. Post-Communist Econ. 30, 775–804 (2018) evidence analysis of the Chinese Family Panel. Stud. Sociol. 5, 139–144 Ma, X.: Female employment and gender gap in China. Springer, Singapore (2020). (赵颖智,李星颖:互联网使用对个体健康的影响-基于中国家庭 (2021a) 追踪调查数据的实证分析.《江汉论坛》, 5, 139–144 (2020)) Ma, X.: Grandchildren care and labor supply of middle-aged grandmothers: Zhuang, J., Liu, A., Sun, C.: Does Internet reproduce gender inequality? A study evidence from China. J. Fam. Econ. Issues (2021b). https:// doi. org/ 10. on Internet use wage premium. Social 36, 88–106 (2016). (庄家炽,刘爱, 1007/ s10834- 021- 09794-0. (Early published online) 玉孙超:网络空间性别不平等的再生产:互联网工资溢价效应的性别 Ma, C., Gu, H., Li, J.: Change of gender wage gaps in Chinese labor market: evi- 差异 以第三期妇女地位调查为例.《社会》, 36 (5), 88–106 (2016)) dence from quantile decomposition analysis of panel data. World Econ. Pap. 2, 96–108 (2013). (马超, 顾海, 李佳佳: 中国劳动力市场上的性别 Publisher’s Note 工资差异变化研究:来自面板分位数回归分解方法的证据.《世界经 Springer Nature remains neutral with regard to jurisdictional claims in pub- 济文汇》, 2, 96–108(2013)) lished maps and institutional affiliations. Mao, Y., Zeng, X.: Does Internet use promote female employment? An empiri- cal study based on the data from the CGSS. Econ. Dyn. 6, 21–131 (2017). (毛宇飞,曾湘泉:互联网使用是否促进了女性就业-基于CGSS数据的 经验分析.《经济学动态》, 6, 21–131 (2017)) Masso, J., Meriküll, J., Vahter, P.: The role of firms in the gender wage gap. J. Comp. Econ. 50, 454–473 (2022) Meng, X.: Gender occupational segregation and its impact on the gender wage differential among rural-urban migrants: a Chinese case study. Appl. Econ. 30, 741–752 (1998) Meng, X., Zhang, J.: Two-tier labor market in urban China: occupational segregation and wage differentials between urban residents and rural migrants in Shanghai. J. Comp. Econ. 29, 485–504 (2001) Miller, P.W.: The wage effect of the occupational segregation of women in Britain. Econ. J. 97, 885–896 (1987) Miller, P., Mulvey, C.: Computer skills and wages. Aust. Econ. Pap. 36, 106–113 (1997) Mincer, J.: Schooling, experience and earning. Columbia University Press, New York (1974) National Bureau of Statistics NBS: China statistical yearbook 1999. China Statistics Press, Beijing (1999).. (国家统计局: 中国统计年鉴1999.中国统 计出版社,北京(1999)) National Bureau of Statistics NBS: China statistical yearbook 2021. China Statis- tics Press, Beijing (2021).. (国家统计局: 中国统计年鉴2021.北京:中国 统计出版社,北京(2021)) Oaxaca, R.: Male–female wage differentials in urban labor markets. Int. Econ. Rev. 14, 693–709 (1973) Ono, H., Zavodny, M.: Digital inequality: a five country comparison using microdata. Soc. Sci. Res. 36, 1135–1155 (2007) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal for Labour Market Research Springer Journals

Internet use and gender wage gap: evidence from China

Journal for Labour Market Research , Volume 56 (1) – Dec 1, 2022

Loading next page...
 
/lp/springer-journals/internet-use-and-gender-wage-gap-evidence-from-china-h4XwXIbMuw
Publisher
Springer Journals
Copyright
Copyright © The Author(s) 2022
ISSN
1614-3485
eISSN
2510-5027
DOI
10.1186/s12651-022-00320-9
Publisher site
See Article on Publisher Site

Abstract

This study explores the influence of Internet use on the gender wage gap in China by using national longitudinal survey data. A fixed effects and instrumental variable method were employed to address individual heterogeneity and other endogeneity problems. The study contributes in the form of four key conclusions. First, considering the endogeneity problems, the return to Internet use is higher for men than for women, unlike the results derived using the ordinary least squares model, which indicates the opposite. The individual heterogeneity problem considerably affects the results, thus suggesting a bias in earlier studies. The results of robustness checks based on the Internet use frequency for different purposes confirm the conclusions. Second, the gender difference in return to Internet use is higher in the low-education group and older age cohorts. Third, both the components—the gender disparity in Inter- net access and gender difference in return to Internet use—widen the gender wage gap, with the gender difference in return to Internet use having a higher impact. Fourth, the effects of these two components on the gender wage gap vary with the educational attainment and age cohorts. Keywords: Gender wage gap, Internet use, Return to Internet use, Internet access, China JEL Classification: J16, J24, J31, O33 1 Introduction 2021; Miller and Mulvey 1997; Pabilonia and Zoghi 2005, With the progress of information and communication etc.) and employment (Alam and Mamun 2017; Atasoy technology (ICT) since the 1970s, Internet usage has 2013; Deyyling 2017; Mao and Zeng 2017, etc.), empiri- expanded worldwide (OECD 2018). The gender digital cal studies on the impact of Internet use on the gender gap in Internet access rose in developed countries in the wage gap are scarce. This study attempts to bridge this early stages of ICT development (Bimber 2000; DiM- gap by providing new evidence from China—a country aggio et  al. 2001; Fatehkia et  al. 2018) but reduced with that has seen rapid Internet diffusion and gender wage the increasing diffusion of digital technologies (Haight gap growth in the last two decades. et al. 2014; Ono and Zavodny 2007; Rice and Katz 2003). The China’s gender disparity in Internet use can be Women in developing countries have a significantly lower highlighted through the Statistical Report on the Devel- likelihood of Internet access than men, and this gender opment of the Internet in China No. 45 (CNNIC 2020), disparity in Internet use can enlarge the overall socio- which reveals that the number of Internet users in China economic gender gap (Alozie and Akpan-Obong 2017; reached 904 million in April 2020, of which 48.1% were Broadband Commission 2013; Hafkin and Huyer 2007; women (30.4% in 2000). The statistics suggests the exist - OECD 2018). Although research has established that ence of a gender disparity in Internet access in China. Internet use can affect wages (Krueger 1993; Liu et  al. Additionally, the gender wage gap in China has expanded since the 1980s (Gustafsson and Li 2000; *Correspondence: xxma@hosei.ac.jp The proportion of women in the total population in China was 48.71% in Faculty of Economics, Hosei University, 4342 Machita-shi Aiharamachi, 2020 (World Bank 2022). Tokyo 194-0298, Japan © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. 15 Page 2 of 17 X. Ma Iwasaki and Ma 2020; Ma 2018). Several empirical stud- and age cohort groups. The gender differences in Inter - ies reveal that the main determinants are gender differ - net access and return to Internet use may differ among ences in human capital and workplace discrimination different education and age cohort groups; therefore, against women (Gustafsson and Li 2000; Li and Yang the effects of Internet use on the gender wage gap may 2010; Li and Yang 2010; Ma et  al. 2013). Others indi- differ per these groups. This study compares the effects cate that occupational segregation and industry/owner- of Internet use among educational attainment and age ship sector segmentation also contribute to the creation cohort groups to provide more evidence to understand of the gender wage gap (Ge 2007; Li and Ma 2006; Liu the digital division issues in depth. Additionally, we use et  al. 2000; Ma 2018; Meng and Zhang 2001; Wang and the three latest wave national longitudinal survey data of Cai 2008; Wang 2005). However, empirical studies on the 2014–2018, which can provide new information. effects of Internet use on the gender wage gap are scarce The study resulted in four main findings. First, when (Qi and Liu 2020; Wu 2021; Zhuang et al. 2016) and lim- addressing endogeneity problems, the return to Internet ited to cross-sectional data analyses, which can result in use is higher for men than for women, unlike the results endogeneity problems. using the ordinary least squares (OLS) model, which This study contributes to the literature in four ways: indicate the opposite. In this case, the individual het- First, in contrast to earlier studies (Qi and Liu 2020; erogeneity problem considerably affects the results, thus Wu 2021; Zhuang et  al. 2016), we examine the influ - suggesting a bias in earlier studies. Second, the results ence of Internet use on the gender wage gap in China indicate that the gender disparities in return to Internet using the panel data analysis method, such as the fixed use are higher in the low-education and older age cohort effects (FE) model, random effects (RE) model, and the than in others. Third, both the components—the gender lagged variable (LV) model to address endogeneity prob- disparity in Internet access and the gender difference in lems. This study is the first to investigate the association return to Internet use—widen the gender wage gap, with between Internet use and the gender wage gap in China the gender difference in return to Internet use having a based on the panel data analysis method. Second, besides higher impact. Fourth, the effects of both components investigating the extensive margin effect of Internet use on the gender wage gap vary with the educational attain- (whether used Internet), we also ascertain the intensive ment and age cohorts. margin effects (the frequency of using the Internet) for The remainder of this paper is organized as follows. different purposes (work and study, communication, lei - Section  2 discusses the channels whereby Internet use sure) on the gender wage gap, which has been not con- influences wages and summarizes the empirical literature sidered in previous studies. Third, our study is the first on the issue. Section 3 introduces the data and the meth- to decompose the determinants of the gender wage gap odology. Section  4 presents and discusses the empirical into two components—the gender disparity in Internet results, and Sect. 5 concludes the study. access (e.g., gender disparity in the proportion of work- ers using the Internet) and gender difference in return 2 Literature review to Internet use (e.g., the gender difference in the rising 2.1 Channels of influence of Internet use on the gender range of wage level in terms of Internet use)—based on wage gap the IV method. Since policy implications differ across Several economic theories can explain the gender wage these two components, it would be meaningful to investi- gap. First, gender differences in the endowment of factors gate how these two components contribute to the effects such as human capital can contribute to a gender wage of Internet use on the gender wage gap. For example, if gap. Based on the human capital theory (Becker 1964; the gender disparity in Internet access is found to be the Mincer 1974), the individual wage level in a perfectly main contributor, a policy promoting Internet accessibil- competitive labor market is determined by the work- ity among women is expected to reduce the wage gap. In ers’ labor productivity. Labor productivity is related to a contrast, if the main contributor is the gender difference worker’s human capital (e.g., education and years of work in the return to Internet use, reducing the discrimination experience). When men have higher educational attain- against women in the workplace may reduce the wage ment than women, they may earn a higher wage. gap. Fourth, although Eastin et al. (2015), OECD (2018), Second, according to the employer discrimination Riggins and Dewan (2005), and Scheerder et  al. (2017) hypothesis (Becker 1957), whenever employers, cus- have argued the Internet use divisions in education, age, tomers, and colleagues discriminate against women, and gender groups, they have not investigated the effect it can create a gender wage gap. During the economic of Internet use on gender wage gap among education transition period (after 1978), with the enforcement Internet use and gender wage gap: evidence from China Page 3 of 17 15 of the Opening-up Policy, the private sector (i.e., the discrimination may contribute to widening the gender privately-owned enterprises, the Township and Village wage gap in China. Enterprises, etc.) has developed since the 1990s, and Fourth, the crowding hypothesis (Bergmann 1974) the proportion of workers in the private sector to total stated that there remains gender occupational segrega- workers in urban areas has increased from 27.2% in 1985 tion in the labor market, which means women concen- to 67.4% in 2020 (NBS 2021). In the planned economy trate on female-dominated occupations (e.g., staff, service period (1949–1977), women’s employment promo- job), whereas men on male-dominated (e.g., manager, tion policies, such as the equality employment policy, technician). When the wage levels of male-dominated were enforced in the public sector (e.g., the state-owned occupations are higher than those of female-dominated enterprises: SOEs; the government organizations) by occupations, the gender wage gap arises. The evidence the Chinese government, which led to less discrimina- from the empirical studies for the developed countries tion against women (Gustafsson and Li 2000; Ma 2018, has supported the crowding hypothesis (Brown et  al. 2021a, b). With the progress of SOE reforms, most SOEs 1980; Kidd and Shannon 1996; Kidd 1993; Miller 1987). have changed to privately-owned enterprises. Although Meng (1998) and Li and Ma (2006) also reported that the influences of the equality employment policies are in China, there exists gender occupational segregation, still greater in the SOEs even in the current period, when which contributes to the formation of the gender wage discrimination against women persists in the private sec- gap. tor, a gender wage gap can arise. Iwasaki and Ma (2020) Finally, the monopsony power hypothesis suggests that report that the gender wage gap is higher in privately- imperfect competition may lead to a gender wage gap. owned enterprises than in SOEs, and the gender wage When a firm has the monopsony power in the labor mar - gap expanded during the economic transition period ket and sets a lower wage level for women, the gender from the 1990s to the 2010s, suggesting that the discrimi- wage gap arises (Hirsch 2010). Vick (2017) reports that nation against women has become severe with the pro- the monopsony power hypothesis is supported in Brazil. gress of the market-oriented economic reform in China. Based on these theories or hypotheses, in this study, we Third, the statistical discrimination hypothesis (Arrow considered that Internet use could lead to the formation 1973; Phelps 1972) suggests that because of information of gender wage gap in China through two channels: (i) the asymmetry, employers must make employment and wage explained component (e.g., gender disparity in Internet decisions for employees, both men, and women, based access), and (ii) the unexplained component—workplace on the average values of some unobservable factors (i.e., discrimination against women (e.g., gender difference in work effort, probability of turnover, etc.). If the employer return to Internet use as reflected in the wages). perceives that the probability of doing housework (i.e., Regarding the first component–the effect of gender childcare, geriatric care, domestic cleaning, cooking, etc.) disparity in Internet access on the gender wage gap, (i) is higher for women than for men, they may set a lower based on human capital theory, the wage level in a per- wage level for women, thus generating a gender wage gap. fectly competitive labor market is determined by a work- In China, factors such as the deregulation of the one- er’s labor productivity; the labor productivity is related child policy, lack of formal public childcare institutions to a worker’s human capital. Internet use can be consid- (e.g., kindergarten), population aging, and insufficiency ered an element of human capital: Internet users are usu- of institutional care for the elderly have increased the ally considered to be those with higher skills (or higher responsibilities of childcare and geriatric care for women productivity) than non-users. If the percentage of Inter- (Connelly et  al. 2018; Cook and Dong 2011; Ma 2021a, net users in the women’s group is lower than that in the b). Family care has decreased the female labor force par- men’s, the gender wage gap arises. (ii) According to the ticipation and reduced women’s work efforts in China crowding hypothesis, men may occupy male-dominated (Chen and Fan 2016; Chen et al. 2016; Chen 2019; Con- occupations (e.g., manager, technician), which have a nelly et  al. 2018; Ma 2021a, b). Therefore, statistical higher likelihood of using the Internet for work, which may lead to the gender divisions in Internet access. To consider the second component–the effect of gen - In China, the Township and Village Enterprises (TVEs) were collectively der differences in return to Internet use on the gender owned enterprises and developed by the end of the 1980s. Since the 1990s, wage gap, (i) according to the employer discrimination most of them have transformed into privately owned enterprises, wherein and statistical discrimination hypotheses, when work- operations and human resource management are similar to those in privately- owned enterprises. place discrimination exists, wage levels tend to be set lower for women than men despite similar endow- In the official data (NBS 2021), the private sector includes the privately- owned enterprises, collectively owned enterprises, foreign investment enter- ments (e.g., Internet use skills). (ii) Based on the crowd- prises, and other privately-owned enterprises (e.g., the joint-stock company, ing hypothesis, if Internet use is evaluated higher for joint company, limited liability company). 15 Page 4 of 17 X. Ma 2.3 Empirical studies on Internet use and gender wage male-dominated occupations than female-dominated gap occupations by a firm, the gender wage gap arises. (iii) Empirical studies on Internet use and the gender wage According to the monopsony power hypothesis, if a firm gap are scarce in developed countries and China. has monopsony power and values Internet use more Borghans et al. (2014) and Ge and Zhou (2020) find that highly for men than women, the gender wage gap widens. Internet skills and their return affect the changes in the Although it is assumed that the two components may gender wage gap in developed countries. Beaudry and contribute to the formation of the gender wage gap, Lewis (2014) hold that changes in return to computer there is no empirical study to investigate the issue. In skills are an important factor that explains the decline in this study, we attempt to estimate the contribution rates the gender wage gap. Using industrial data, Ge and Zhou of these two components in the subsequent decomposi- (2020) report that robot use reduces the gender wage tion analysis and provide new evidence to understand the gap, while an increase in computer capital raises the gen- causes of the gender wage gap in China in depth. der wage gap in the US. However, these studies are not based on the decomposition method. Therefore, the con - 2.2 Empirical studies on the gender w age gap in China tribution rates of the gender disparity in Internet access There are numerous empirical studies on the gender and gender difference in return to Internet use in forming wage gap in developed countries and China (Iwasaki and the gender wage gap are unclear. Ma 2020). We summarize only the main ones in China Regarding China, only three studies focus on the issue. as follows. Zhuang et  al. (2016) use data from the Third Survey on Gustafsson and Li (2000), Liu et  al. (2000), Li and Chinese Women’s Social Status and apply the propensity Yang (2010), Li et al. (2014), and Ma et al. (2013) use the score matching method to estimate the wage function. Blinder–Oaxaca model (Blinder 1973; Oaxaca 1973) for They find that the Internet use wage premium for women the decomposition analysis, thereby demonstrating that is 90.6% of that for men, suggesting that the return to both the explained (e.g., gender differences in educational Internet usage is lower for women than men. Using data attainment) and unexplained (e.g., returns to education) from 2010, 2013, and 2015 Chinese General Social Sur- components affect the gender wage gap. Most stud - veys (CGSS) and the Blinder–Oaxaca decomposition ies show that the contribution of the unexplained com- method, Qi and Liu (2020) report that both Internet ponent is higher than that of the explained component, access and the return to Internet usage reduced the gen- which suggests that workplace discrimination against der wage gap in 2013 and 2015. Wu (2021) uses data from women is the primary cause of the gender wage gap in the 2017 CGSS, the OLS model, and the Blinder–Oaxaca China. The contribution rates of the unexplained compo - decomposition method to analyze the effect of Internet nent in the gender wage gap for the local urban resident use on the gender wage gap and reveals that the contribu- group were reported to be 52.49% in 1988 and 63.20% in tion rate of gender disparity in Internet access is -4.34%, 1995 (Gustafsson and Li 2000); and 52.0% in 1995, 69.0% which reduces the gender wage gap. in 2002, and 77.7% in 2007 (Li et  al. 2014). The rate for Although these studies provide some evidence on the the migrant group was 74.32%–84.38% in 2008 (Li and effects of Internet use on the gender wage gap in China, Yang 2010), and that for all residents was 49.18% in 1996 they have not addressed the individual heterogene- (Meng and Zhang 2001). In addition, the values by wage ity problem owing to the use of cross-sectional analy- percentiles were 86.08–101.80% in 2006 and 45.31– sis methods. Additionally, the intensive margin effects 91.73% in 2009 (Ma et al. 2013). A few empirical studies of Internet use (e.g., the frequency of Internet use) for also focus on the effect of segmentation by sector on the different purposes (e.g., work and study, leisure) were gender wage gap in China. Li and Ma (2006) and Wang not considered. Moreover, the differences in the effects (2005) analyze the influence of occupational segregation of Internet use by the educational attainment and age on the gender wage gap. Ge (2007), Ma (2018), and Wang cohorts were not considered. This study can address and Cai (2008) explore the impact of segmentation by these neglects. industry or enterprise ownership sector type on the gen- der wage gap in China and report that the unexplained 3 Methodology component in intra-sector differentials drives the gender 3.1 Model wage gap in China. To estimate the gender disparity in return to Internet use, we estimate the wage function. The OLS model is Iwasaki and Ma (2020) conducted a meta-analysis based on the results of expressed by Eq. (1). the inclusion of a gender dummy variable in the wage functions of 199 studies for China. For the current empirical studies on the gender wage gap in other countries, please refer to Ge and Zhou (2020) for the U.S., Biewen et al. (2020) for German, Masso et al. (2022) for Estonia. Internet use and gender wage gap: evidence from China Page 5 of 17 15 the influence of the regional telecommunication capabil - LnW = a + β Int + β Female i 0 1 i 2 i ity in the past on an individual’s recent income levels is (1) + β Female ∗ Int +γ X + v 3 i i i small, which may fit the conditions of the IV from the econometric perspective. We performed several tests, In Eq. (1), i represents the individual. Int is the Internet including the Durbin-Wu-Hausman test, the exclusion use variable, Female is the female dummy; Female ∗ Int test (Hansen’s J statistic), and the weak identification test is the interaction term of Internet use and female (Cragg–Donald Wald F statistic). The results indicated dummy variable;LnW denotes the logarithmic value of that the two IVs are appropriate (see Tables 2, 3, 4 and 5 wages; X represents a set of the factors (i.e., educa- and discussions in Sect. 4). tion, years of work experience, occupation, etc.) that Second, as v in Eq.  (1) includes the errors related to affect the wages; β and γ are the estimated coefficients; individual-specific and time-invariant factors ( ρ ) and and v denotes the error term. The total of β and β are 1 3 the idiosyncratic error ( ε ), an individual heterogeneity the return to Internet use, β is the gender differences in problem may arise in Eq. (1). We use the FE or RE model return to Internet use. to address this heterogeneity problem. We perform the However, the endogeneity problem may exist in Eq. (1) Hausman specification test to judge the validity of the FE for three reasons: first, the omitted variable may influ - and RE models. In Eq. (5), t denotes the longitudinal sur- ence the likelihood of using the Internet and the wage vey year. level. We constructed the variables to control factors that may affect wage level as much as possible. However, LnW = α + β Int + β Female ∗ Int +γ X + ρ + ε it 1 it 2 it it it some unobservable variables may also affect the results. (5) We used the IV method to address the issue expressed by Third, the endogeneity issue may arise due to reverse Eqs. (2–4). causality. It is assumed that Internet use may affect wages Pr(Int = 1) = a + β Z +β Int + β Female (explored in this study). However, wages may also affect i 1 Z i i 2 i the likelihood of Internet use. For instance, the likelihood + β Female ∗ Int + γ X + u 3 i i i of using the Internet for work and study may be greater (2) for high-wage workers than low-wage workers. Since there is a two-way relationship between Internet use and LnW = a + β Int + β Female i 0 1 i 2 i (3) wage, we use a one-period lagged variable (LV) model to + β Female ∗ Int +γ X + δ 3 i i address the potential reverse causality. We assume that Internet use in period t − 1 may affect the wage level in period t . However, wages in period t cannot influence corr(Z, δ) = 0 and corr(Z, u) �= 0 (4) the likelihood of using the Internet in period t − 1 . In In Eqs. (2–4), u , and δ denote the error terms, respec- Eq.  (6) below, t represents the recent period (e.g., 2018), tively. Z represents the IV. The internet penetration rate t − 1 represents the prior period (e.g., 2016), and Int t−1 at the regional (provincial or local) level and the impor- expresses the Internet use in the prior period ( t − 1). tance of using the Internet attitude were generally used LnW =α + β Int + β Female ∗ Int it 1 it−1 2 it−1 in previous studies (e.g., Cao and Jiang 2020; Zhao and (6) Li 2020). We performed several tests for these IVs, but +γ X + ρ + ε i it the results rejected the hypothesis that these IVs are exogenous, suggesting that they were not valid for this Then, to investigate the effect of Internet use on the study. We used two variables—(i) the provincial optical gender wage gap, we use the Blinder–Oaxaca decomposi- cable circuit in 1999 and (ii) the provincial long-distance tion method (Blinder 1973; Oaxaca 1973) to decompose cable line length in 1999 as IVs in this study. Both are the the determinants into two components: (i) the explained oldest data that we could obtain from the government’s component, comprising the gender endowment differ - official dataset. It can be assumed that Internet instal- ences (i.e., the gender disparity in Internet access, etc.), lations in recent survey years (2014, 2016, and 2018) are and (ii) the unexplained component, composed of the closely related to the regional telecommunication capa- gender differences in the evaluated price of each factor bility in the past (such as the optical cable circuits or the (i.e., the gender difference in the return to Internet use, long-distance cable line length 15 to 19 years back), while etc.). Equations (7, 8) describe the model : The constant is omitted for descriptive convenience. We conduct two The information on these two IVs were obtained from the data published in decomposition analyses based on Eqs. (7) and (8); only one has been listed China Statistical Yearbook 1999 (NBS 1999). because the results are very similar. 15 Page 6 of 17 X. Ma 2010 CFPS baseline survey data was obtained through ¯ ¯ ¯ LnW − LnW = β H − H + (β − β ) H m f m m f m f f multi-stage probability sampling with implicit stratifica - (7) tion. Multi-stage sampling reduces the operational cost ¯ ¯ ¯ LnW − LnW = β H − H + (β − β ) H of the survey and permits the analysis of the social con- m f f f m f m m text. In the 2010 baseline survey, the CFPS successfully (8) interviewed nearly 15,000 families and 30,000 individuals In Eqs. (7) and (8), LnW − LnW is the gender wage within these families, with an approximate response rate gap; β , β is coefficient of each factor calculated based of 79%. The respondents were tracked through annual on the male and female wage functions, respectively; H follow-up surveys. The CFPS 2010 covers 25 provinces denotes the mean value of each factor, including Inter- and municipalities. Only the latest three waves (2014, net use. β ( H − H ) or β ( H − H ) is the m m m f f f 2016, and 2018) of the CFPS, which include the survey explained component, and ( β − β ) H or ( β − β m m f f f item on Internet use, have been used in this study. The ) H is the unexplained component which includes the CFPS sample sizes are 37,147 (2014), 36,892 (2016), and discrimination against women in the workplace. 37,354 (2018). To compare the effects of Internet use by groups, we The logarithmic value of the hourly wage is used as the also calculate the estimations per educational attainment dependent variable. The wages for 2014, 2016, and 2018 and age cohort group. have been adjusted using the regional Consumption Price Index (CPI) in the rural and urban areas published 3.2 Data by China’s National Bureau of Statistics (NBS 1999) to Three waves of data (2014, 2016, and 2018) from the account for inflation, using the 2014 CPI as the baseline. China Family Panel Studies (CFPS 2020) dataset are used We calculate the hourly wage as a dependent variable in this study. The reasons for using the CFPS are con - based on the annual earned income and corresponding sidered as follows: first, the CFPS is a nationally repre - working hours. sentative longitudinal survey of Chinese communities, The key independent variable is an Internet use dummy families, and individuals launched in 2010 by the Insti- variable based on the questionnaire item: “Did you use tute of Social Science Survey, Peking University, China. the Internet in the past year?” (1 = has used the Internet Although Chinese General Social Survey (CGSS) data in the past year, 0 = otherwise); we primarily use the vari- was used in the previous studies (e.g., Qi and Liu 2020), able to estimate the extensive margin effect of Internet since the CGSS is the cross-sectional survey data, the use, which was estimated in the literature. Based on the individual heterogeneity problem could not be addressed. questionnaire items on the frequency of Internet use by On the contrary, this study performs the analysis using purpose (work, study, communication, shopping, enter- the panel data analysis method (e.g., the FE and RE mod- tainment), we originally constructed three indicators to els, the LV model) to deal with the endogeneity problems, investigate the intensive margin effect of Internet use in this study, based on the CFPS, can provide the robustness this study: (a) frequency of using the Internet for work, results on the issue. Second, although the China Health including work and study, (b) frequency of using the and Retirement Longitudinal Survey (CHARLS) is a Internet for communication, and (c) frequency of using longitudinal survey having information on Internet use, the Internet for leisure, including shopping and enter- the survey targets of CHARLS are individuals aged 45 tainment. Based on the five questions items in the CFPS and older, whereas the CFPS covers all age generations. as “please answer the frequency of using the Internet for Therefore, we can use the data from the CFPS to com - study, work, communication, entertainment, commer- pare the differences in the effect of Internet use among cial activity (e.g., Internet payment, shopping): (i)almost the younger, middle-aged, and older generations in this every day; (ii)3–4 times a week; (iii)1–2 times a week; study. Third, we can obtain the rich information on Inter - (iv) 2–3 times a month; (v) once a month; (vi) once few net use, such as the used Internet and the frequency of months; (vii) never use)”, we re-coded each frequency using the Internet for different purposes (e.g., working as “7 = almost every day; 6 = 3–4 times a week; 5 = 3–4 and study, communication, and leisure); the latter is the times a week; 4 = 1–2 times a week; 3 = once a month; unique question item in the CFPS which is firstly utilized 2 = once in few months: 1 = never use”. We calculated the in this study on the issue. total values for (a) and (c) and used their arithmetic mean The CFPS is designed for individual-, family-, and in the analysis. community-level longitudinal data collection in con- We constructed an interaction term of Internet use and temporary China and provides information on Internet a female dummy variable to investigate the gender differ - use, wages, and other factors (education, years of work ence in return to Internet use in wage functions. experience, sex, occupation, industry sector, etc.). The Internet use and gender wage gap: evidence from China Page 7 of 17 15 Mean: Mean: M: 2.573; F:2.262 M: 2.214;F:1.765 SD: SD: M: 1.097; F:0.997 M: 0.920;F:1.003 Fig. 1 Kernel density of the logarithm of wages for Internet using and not-using groups. Source: Authors’ calculations based on the data from CFPS of 2014, 2016, and 2018. M: male workers; F: female workers Based on the economic theories and existing studies, that there exists a gender wage gap, and the gender wage we identified a set of variables that may affect wages as gap differs among Internet using and non-using groups. specified by the wage functions, such as years of school - The proportions of individuals using the Internet by ing, years of work experience and its squared term, gender are summarized in Table 1. In general, an increase ethnicity (1 = han, 0 = minority), party membership is observed from 2014 to 2018 for both men and women. (1 = member of Communist Party of China, 0 = non- However, the percentage is approximately 5% higher for member), urban residents(1 = urban residents, 0 = rural men than women in each year of the three years, suggest- residents), occupation (1 = manager and technician, ing the existence of gender disparity in Internet access. 0 = otherwise), industry sector (1 = manufacturing In terms of the disparity by educational attainment industry sector, 0 = otherwise), workplace ownership group, the percentage of individuals using the Internet (1 = state-owned sector, 0 = otherwise), region (west, is higher for men than women in the low- and high-edu- central, east), and survey year. cation groups, whereas the opposite holds in the middle- As described in Sect.  3.1, we used two variables: the education group. Regarding the disparity by age cohort, provincial optical cable circuit and the provincial long- the percentage of individuals using the Internet is higher distance cable line length in 1999, as IVs in this study. for men than women in each age cohort. The gender dis - This analysis is limited to respondents aged 16–60 years parity in Internet access is the largest in the middle-aged and excludes missing values; the longitudinal sample size generation born during 1970–1989 and the least in the 7 9 is 18,381. younger generation born after 1990. The results indicate that the gender disparities in Internet access differ by edu - 4 Empirical results and discussion cational background and age cohort. Therefore, the het - 4.1 Descriptive statistics erogeneous group should be considered in the analysis. Figure 1 shows the kernel density of the logarithmic value of wages by Internet-using and non-using groups. The 4.2 Basic results wage level is higher for men than women in both groups, Table  2 presents the results of the wage function analy- and the gender wage gap is higher in the Internet-using sis using five models—the OLS (Model 1), IV (Model 2), group than that in the not using one. The results indicate The questionnaire item on education attainment has eight categories: illit - erate/semiliterate, elementary school, junior high school, senior high school, According to the Interim Measures of the State Council on the Placement of college, university, master, and doctor. We divided the educational groups into Elderly, Weak, Sick and Disabled Cadres (Article 4) and Interim Measures of three categories: the low- (elementary school and lower), middle- (junior and the State Council on Retirement and Retirement of Workers (Article 1) pub- senior high school), and high- (college and higher) educational groups. lished in 1978, the minimum employment age in China is 16 for both men and women. However, the mandatory retirement age differs by gender: it is We divide the age cohorts considering the Internet diffusion situation. We 60 years for both blue- and white-collar male workers, while it is 50 years for define the younger generation as the group born in the year when the Inter - blue-collar female workers and 55 years for white-collar female workers (e.g., net was used in society, the middle-aged generation who used the Internet women working as a civil servant/office executive, or working as a manager, in their adolescence, and the older generation who used it in their middle- technician, or a professor in the public sector). or older age. 15 Page 8 of 17 X. Ma Table 1 The proportion of using the Internet by sex be considered that the IVs are valid. The results of the F test and the Breusch–Pagan Lagrange multiplier test Unit: % indicate that both the FE and RE models are appropriate Total Males Females Gap (M-F) compared to the OLS model. The Hausman test results (2308.67, p = 0.000) suggest that the FE model is more Total appropriate than the RE model. The main findings are 2014 34.84 37.82 32.08 5.74 summarized as follows. 2016 49.88 52.93 46.95 5.98 First, the coefficients of Internet use are significantly 2018 62.41 65.19 59.74 5.46 positive in Models 1–5, suggesting that Internet use may Low education increase the wage levels when addressing individual het- 2014 6.61 8.37 5.61 2.77 erogeneity and other endogeneity problems. The results 2016 15.72 17.46 14.72 2.75 can be explained based on human capital theory consid- 2018 34.04 38.21 31.18 7.03 ering Internet use skill as a kind of human capital, which Middle education can potentially increase labor productivity. Addition- 2014 41.21 40.42 42.14 − 1.71 ally, Internet use may have a signaling effect whereby an 2016 56.15 55.10 57.39 − 2.29 employer may evaluate that an Internet-using worker 2018 69.52 68.60 70.60 − 2.00 possesses higher skills than a non-Internet-using worker. High education Second, in terms of the gender differences in return 2014 65.98 68.89 63.20 5.68 to Internet use, the results of interaction item of Inter- 2016 74.28 75.63 72.81 2.82 net use and female dummy variable in Model 1 (OLS), 2018 94.73 94.64 94.82 − 0.19 Model 3 (LV), and Model 4 (RE) indicate that the return Born before 1979 to Internet use is significantly greater for women than 2014 6.03 8.14 3.80 4.34 men. However, considering the endogeneity problem, 2016 11.32 12.75 9.73 3.02 the wage premium is significantly greater for men than 2018 19.45 21.65 17.01 4.64 women in Model 2 (IV). Additionally, although the result Born in 1979–1989 is non-significant in Model 5 (FE), it is a negative value. 2014 32.44 36.46 28.81 7.65 The results suggest a bias in the results derived from the 2016 47.36 51.41 43.56 7.85 OLS model. When using appropriate models (e.g., the IV 2018 58.55 61.95 55.33 6.62 or FE model) to address the endogeneity problem, the Born after 1990 return to Internet use is higher for men than women. 2014 81.28 81.94 80.69 1.25 2016 89.31 90.24 88.41 1.83 4.3 Estimations on the intensive margin effect of Internet 2018 93.14 93.45 92.83 0.62 use Source: Authors’ calculations based on the data from CFPS of 2014, 2016, and As shown in Table  2, although we analyzed the exten- sive margin effects of Internet use on the gender wage Gap = percentage of male Internet user-percentage of female Internet user gap using a binary variable (whether used Internet), Age range: 16–60 years the intensive margin effect of Internet use (e.g., the fre - Low education: elementary school and lower; Middle education: junior and senior high school; High education: college and higher quency of using the Internet) on the wage gap should also be considered. Additionally, the effects of Internet use on the wage gap may differ for different purposes of Inter - LV (Model 3), RE (Model 4), and FE (Model 5). Regard- net use. For example, when comparing the group using ing the appropriateness of the IVs, the endogeneity test the Internet frequently for leisure (e.g., entertainment, result (Durbin–Wu–Hausman test) is statistically signifi - shopping) to the group using it frequently for work or cant at the 1% level; therefore, the null hypothesis con- study, it is assumed that the group using the Internet fre- sidering that all the variables are exogenous is rejected. quently for work or study is likely to obtain higher earned The Hansen J statistic is not significant, thus revealing income. To investigate the intensive margin effects of that the IV is exogenous in the second stage estima- Internet use, we used the three indices: (i) the frequency tion. Furthermore, the Cragg-Donald Wald F statistic is of using the Internet for work and study, (ii) the fre- 26.393, which is larger than 10, suggesting that the weak quency of using the Internet for communication, and (iii) identification problem can be neglected. Therefore, it can the frequency of using the Internet for leisure including shopping and entertainment, to replace the binary varia- ble of Internet use in Table 2 and re-estimate the models. Due to space limit constraints, results, including all control variables, were The results are presented in Table 3. reported in Additional file 1: Table S1. Internet use and gender wage gap: evidence from China Page 9 of 17 15 Table 2 The gender differences in return to Internet use (1) OLS (2) IV (3) LV (4) RE (5) FE Coef t Coef t Coef t Coef t Coef t Internet use 0.603*** 6.68 14.900*** 5.59 0.107 0.98 0.606*** 6.65 0.405** 2.26 Female − 0.849*** − 11.29 4.054*** 4.40 − 0.697*** − 5.76 − 0.844*** − 10.90 – Internet use × Female 0.540*** 5.19 − 8.881*** − 5.04 0.360** 2.51 0.531*** 5.02 − 0.106 − 0.42 Control variables Yes Yes Yes Yes Yes First stage estimation IV1 0.083*** 3.47 IV2 − 0.243*** − 11.58 Observations 18,381 18,381 7777 18,381 18,381 R-squared 0.476 0.367 12,876 12,876 R-sq. Between 0.190 0.197 Within 0.574 0.017 Overall 0.476 0.008 F-test (Prob > F) 0.000 BP test (Prob > chibar2) 66.51 (p = 0.000) Hausman test (Prob > chi2) 2308.67 (p = 0.000) Endogeneity test (DWH) 69.216 (p = 0.000) Hansen J statistic p = 0.224 Cragg-Donald Wald F statistic 26.393 Source: Authors’ Calculations Based on the data from CFPS of 2014, 2016 and 2018 Control variables, including years of schooling, years of work experience and its squared term, party membership, urban, occupation, industry sector, workplace ownership, region, and survey year variables, have been calculated, but the results are not listed in the table owing to space limit constraints OLS: ordinary least squares; IV: instrumental variable method; LV: lagged variable model; RE: random effects model; FE: fixed effects model; BP test: Breusch–Pagan Lagrange multiplier test; DWH: Durbin–Wu–Hausman test; IV1: the provincial optical cable circuit in 1999; IV2: the provincial long-distance cable line length in 1999 ***p < 0.01; **p < 0.05; *p < 0.1 15 Page 10 of 17 X. Ma Table 3 Frequency of Internet use for different purposes (1) Work and study (2) Communication (3) Living IV FE IV FE IV FE Internet use 2.020*** − 0.003 2.540*** 0.056* 1.385*** 0.010 (4.96) (−0.14) (5.59) (1.94) (6.58) (0.45) Female 2.709*** 2.987*** 1.922*** (3.77) (4.21) (4.44) Internet use × Female − 1.317*** − 0.008 − 1.540*** − 0.016 − 0.797*** − 0.015 (− 4.66) (− 0.29) (− 5.11) (− 0.39) (− 5.93) (− 0.54) First stage estimation IV1 0.204*** 0.151*** 0.411*** (4.66) (4.91) (9.78) IV2 − 0.351*** − 0.249*** − 0.612*** (− 9.55) (− 9.63) (− 17.32) Control variables Yes Yes Yes Yes Yes Yes Observations 18,605 18,605 18,605 18,605 18,605 18,605 Groups 12,920 12,920 12,920 R-sq. Between 0.192 0.198 0.192 Within 0.015 0.030 0.015 Overall 0.007 0.033 0.006 F test (Prob > F) 0.000 0.000 0.000 BP test (Prob > chibar2) 73.70 (p = 0.000) 77.77 (p = 0.000) 77.22 (p = 0.100) Hausman test 2406.81 (p = 0.000) 2355.36 (p = 0.000) 2380.03 (p = 0.000) Endogeneity test (DWH) p = 0.000 p = 0.000 p = 0.000 Hansen J statistic p = 0.917 p = 0.082 p = 0.482 Cragg-Donald Wald F statistic 18.682 26.134 47.343 Source: Authors’ Calculations Based on the data from CFPS of 2014, 2016 and 2018 Control variables, including years of schooling, years of work experience and its squared term, party membership, urban, occupation, industry sector, workplace ownership, region, and year variables have been calculated but are not listed in the table owing to space limit constraints IV: Instrumental variable method; FE: fixed effects model; BP test: Breusch–Pagan Lagrange multiplier test; DWH: Durbin–Wu–Hausman test ***p < 0.01; **p < 0.05; *p < 0.1. t-values are shown in parentheses The results from the Hausman specification test, F-test, female dummy variables are negative values and signifi - and the Breusch–Pagan Lagrange multiplier test indicate cant at the 1% level in Models 1–3, but they are not sig- that the FE model is more appropriate than the OLS and nificant in the results from the FE model. These results RE models. Regarding the validity of the IV method, the are consistent with those in Table 2. results in the first stage estimations indicate that both IVs significantly affect the likelihood of using the Internet at 4.4 Estimations considering heterogeneous group the 1% level; the results from the Durbin–Wu–Hausman The results for the heterogeneous group based on edu - test, Hansen J statistic, and Cragg-Donald Wald F statistic cational attainment and age cohort are summarized in values indicate that the IV method is valid. Therefore, we Tables 4, 5. The results from the Hausman specification report the results using the FE and IV models in Table 3. test, F-test, and the Breusch–Pagan Lagrange multiplier First, the results from the IV method indicate that the test indicates that the FE model is more appropriate than coefficients of Internet use are positive and significant at the OLS and RE models. Regarding the validity of the IV the 1% level in Models 1–3. It is also positive and signifi - method, the results in the first stage estimations indicate cant at the 10% level in Model 2 from the FE model. The that both IVs significantly affect the likelihood of using conclusions are consistent with the results in Table 2. the Internet at the 1% level; the results from the Durbin– Second, the results from the IV methods show that the Wu–Hausman test, Hansen J statistic, and Cragg-Donald coefficients of the interaction term of Internet use and Wald F statistic values indicate that the IV method is 11 12 Due to space limit constraints, results, including all control variables, were Due to space limit constraints, results, including all control variables, were reported in Additional file 1: Table S2. reported in Additional file 1: Tables S3 and S4. Internet use and gender wage gap: evidence from China Page 11 of 17 15 Table 4 The gender differences in return to Internet use by educational group (1) Low (2) Middle (3) High IV FE IV FE IV FE Internet use − 11.427 0.496 10.277*** 0.475** 19.852** − 0.083 (−0.94) (0.93) (4.60) (2.22) (2.37) (− 0.14) Female − 2.598 3.403*** 16.978** (− 1.41) (3.50) (2.25) Internet use × Female 9.648 0.139 − 6.856*** − 0.370 − 18.215 − 0.033 (− 0.97) (0.21) (− 4.40) (− 1.21) (− 2.28) (− 0.04) First stage estimation IV1 − 0.167*** 0.165*** 0.466*** (− 3.71) (5.40) (4.58) IV2 − 0.060 − 0.277*** − 0.472*** (− 1.23) (− 10.83) (− 5.70) Control variables Yes Yes Yes Yes Yes Yes Observations 4652 4652 10,284 10,284 3445 3445 Groups 3289 7070 1905 R-sq. Between 0.017 0.200 0.200 Within 0.008 0.004 0.004 Overall 0.000 0.001 0.001 F test (Prob > F) 0.000 0.000 0.000 BP test (Prob > chibar2) 0.000 (p = 0.987) 17.67 (p = 0.000) 19.58 (p = 0.000) Hausman test 248.75 (p = 0.000) 874.72 (p = 0.000) 874.72 (p = 0.000) Endogeneity test (DWH) p = 0.131 p = 0.000 p = 0.000 Hansen J statistic p = 0.001 p = 0.652 p = 0.127 Cragg-Donald Wald F 1.199 21.914 19.93 statistic Source: Authors’ Calculations Based on the data from CFPS of 2014, 2016 and 2018 Low: elementary school and lower; middle: junior and senior high school; high: college and higher Control variables, including years of schooling, years of work experience and its squared term, party membership, urban, occupation, industry sector, workplace ownership, region, and survey year variables, have been calculated, but the results are not listed in the table owing to space limit constraints IV: Instrumental variable method; FE: Fixed effects model; BP: test Breusch–Pagan Lagrange multiplier test; DWH: Durbin–Wu–Hausman test ***: p < 0.01; **: p < 0.05; *: p < 0.1. t-values are shown in parentheses valid in most cases. Therefore, we report the results using First, in terms of the return to Internet use by the educa- the FE and IV models in Tables 4, 5. tional group, the results from the IV method show that the Table  4 presents the low-, middle- and high-educa- coefficients of Internet use are positive and significant at tional group results. To secure enough samples for the the 1% level in the middle- and high-educational groups, analysis and consider the distribution of workers by edu- whereas it is not significant for the low-education group. The cation attainment levels, we distinguished the samples results from the FE model reveal that the Internet use coef- into three groups (i) the low-education group (elemen- ficient is significant only for the middle education group at tary school and lower); (ii) the middle-education group the 5% level. They suggest the effects of the return to Internet (junior and senior high school); and (iii) the high-educa- use on wages are much more significant for the middle- and tion group (college and higher). The main findings are as high-educational groups. Additionally, the individual hetero- follows. geneity problem considerably affects the results. Second, in terms of the gender difference in return to Internet use by the educational group, the results from the IV method show that the coefficient of the interaction of Internet use and female dummy is negative and signifi - The distribution proportion of workers by the educational group in the samples is 25.3% for the elementary school and lower group, 55.9% for the cant at the 5% level for the middle-and high-educational junior and senior high school group, and 18.8% for the college and above group. 15 Page 12 of 17 X. Ma Table 5 The gender disparity in return to Internet use by age cohort (1)Born before 1969 (2) Born from 1970–1989 (3) Born after1990 IV FE IV FE IV FE Internet use 28.375*** − 0.822 12.618*** 0.341* − 24.696 1.322 (3.41) (− 1.32) (4.88) (1.75) (− 0.90) (2.62) Female 0.077 3.510*** − 22.009 (0.28) (3.78) (− 0.97) Internet use × Female − 23.690*** 1.404 − 8.217*** 0.011 24.273 − 1.207 (− 3.25) (1.14) (− 4.45) (0.04) (0.96) (− 1.25) First stage estimation IV1 0.290*** 0.088*** − 0.032 (5.09) (3.32) (− 0.44) IV2 − 0.383*** − 0.236*** − 0.155* (− 11.13) (− 9.99) (− 1.91) Control variables Yes Yes Yes Yes Yes Yes Observations 3.512 3.512 12,701 12,701 2.168 2.168 Groups 2256 9229 1.676 R-sq. Between 0.230 0.198 0.200 Within 0.002 0.030 0.004 Overall 0.000 0.033 0.001 F test (Prob > F) 0.999 0.000 0.000 BP test (Prob > chibar2) 0.000 (p = 1.000) 49.00 (p = 0.000) 1.63 (p = 0.100) Hausman test 67.74 (p = 0.000) 2736.48 (p = 0.000) 57.86 (p = 0.000) Endogeneity test (DWH) p = 0.000 p = 0.000 p = 0.083 Hansen J statistic p = 0.554 p = 0.604 p = 0.625 Cragg-Donald Wald F statistic 11.590 20.405 11.201 Source: Author’s calculations based on the data from CFPS of 2014, 2016 and 2018 Control variables, including years of schooling, years of work experience and its squared term, party membership, urban, occupation, industry sector, workplace ownership, region, and survey year variables, have been calculated, but are not listed in the table owing to space limit constraints IV: Instrumental variable method; FE: Fixed effects model; BP test: Breusch–Pagan Lagrange multiplier test; DWH: Durbin–Wu–Hausman test ***p < 0.01; **p < 0.05; *p < 0.1. t-values are shown in parentheses groups, while it is insignificant for the low-education among the high-education group is serious much more group, thus indicating when the other factors are held than that in the low- and middle-educational groups. consistent, the return to Internet use is lower for female Table 5 summarizes the results by three age cohorts: the workers than male workers, and the gender difference in younger (born after 1990), middle-aged (born in 1970— return to Internet use is higher in the middle-and high- 1989), and older (born before 1969) generations. First, educational groups than the low-education group. How- the results from the IV method show that the coefficients ever, the results from the FE model are not significant in of Internet use are positive and significant at 1% levels in three educational groups. both the middle-aged and older cohort groups, while they Third, comparing the magnitude of the coefficients are not significant for the younger cohort. The FE model from the IV method show that the gender difference in results reveal that the coefficients of Internet use are sig - return to Internet use is larger in the high-education nificant for both the younger and middle-aged cohort group (−  18.215) than in the middle education group groups at 5 and 10% levels, while it is not significant for the (− 6.856). The possible reasons can be considered as fol - older cohort. The results suggest that, in general, there is a lows: the discrimination against high-educated women in positive effect of Internet use on wages in each age cohort the workplace, which may be caused by the glass-ceiling group, and the influence of individual heterogeneity on the problem, or due to the gender occupational segregation return to Internet use is greater for the older cohort. Internet use and gender wage gap: evidence from China Page 13 of 17 15 Table 6 Decomposition results of Internet use and the gender Third, the gender difference in return to Internet use wage gap was found higher in the older cohort than the other age cohorts when compared using the magnitude of the coef- Value Percentage ficients based on the IV method. It may be caused by Explained Unexplained Explained Unexplained that the discrimination against female workers is serious (%) (%) much more among older generations than those among Total 0.471 0.449 51.2 48.8 the younger and middle-aged generations. Internet use 0.461 0.665 50.1 72.4 Education − 0.052 0.005 − 5.7 0.5 4.5 Decomposition results Experience − 0.044 1.087 − 4.8 118.2 The results in Table  1 indicate that Internet access dif- Ethnicity − 0.002 − 0.107 − 0.2 − 11.6 fers by gender, and those in Tables 2, 3 suggest that the Party − 0.027 − 0.029 − 3.0 − 3.2 return to Internet use is different for men and women. Occupation − 0.018 − 0.005 − 2.0 − 0.5 However, how the two components affect the forma - Industry 0.021 − 0.160 2.3 − 17.4 tion of the gender wage gap is unclear. Therefore, we State- 0.000 − 0.188 0.0 − 20.4 conduct a decomposition analysis to calculate the con- owned tribution rates of these two components (Table  6). We Urban 0.067 0.037 7.3 4.0 also perform the decomposition analyses based on the Region 0.001 0.219 0.1 23.8 educational attainment and age cohort (Tables 7, 8). Year 0.064 0.002 7.0 0.2 Table  6 reports the decomposition results for the Constant 0.000 − 1.077 0.0 − 117.1 total sample. First, the explained and unexplained com- Source: Authors’ Calculations Based on the data from CFPS of 2014, 2016 and ponents contribute to the formation of the gender wage The decomposition is based on the results from the IV method gap. The influence is slightly less for the former (46.3%) than the latter (53.7%). Second, in terms of the effects of Internet use, it is Second, in terms of the gender difference in return to shown that both the gender disparity in Internet access Internet use by age cohorts, the results from the IV method and gender difference in the return to Internet use show that the interaction coefficients of Internet use and contribute to widening the gender wage gap, and the female dummy are negative values and significant at 1% or contribution rates of both are higher than the other 5% level for both middle-aged and older cohorts, and not factors (e.g., education, occupation). Additionally, the significant for the younger cohort. The results from the FE contribution rate is larger for the return to Internet use model are not significant in the three age cohorts. Table 7 Decomposition results of Internet use and the gender wage gap by educational group Value Percentage Explained Unexplained Explained (%) Unexplained (%) (a) Low (N = 4652) Total 0.579 1.019 36.2 63.8 Internet use − 0.842 − 2.097 − 52.7 − 131.3 Other variables 1.421 3.116 88.9 195.1 (b) Middle (N = 10,284) Total 0.051 0.357 12.4 87.6 Internet use − 0.224 − 0.484 − 54.9 − 118.7 Other variables 0.275 0.841 67.3 206.3 (c) High (N = 3445) Total − 0.113 0.229 − 97.3 197.3 Internet use − 0.255 3.480 − 219.6 2996.8 Other variables 0.142 − 3.251 122.3 − 2799.5 Source: Authors’ Calculations Based on the data from CFPS of 2014, 2016 and 2018 The decomposition is based on the results from the IV method Low: elementary school and lower; Middle: junior and senior high school; High: college and higher “Other variables” include years of schooling, years of work experience, party membership, urban, occupation, industry sector, workplace ownership, region, and years 15 Page 14 of 17 X. Ma Table 8 Decomposition results of Internet use and the gender wage gap by age cohort Value Percentage Explained Unexplained Explained (%) Unexplained (%) (a) Born before 1969 (N = 3512) Total 1.000 1.162 46.3 53.7 Internet use 1.023 − 2.401 47.3 − 111.0 Other variables − 0.023 3.563 − 1.0 168.3 (b) Born between 1970–1989 (N = 12,701) Total 0.538 0.445 54.7 45.3 Internet use 0.536 − 0.004 54.6 − 0.4 Other variables 0.002 0.449 0.1 45.7 (c) Born after 1990 (N = 2168) Total 0.132 0.136 49.1 50.9 Internet use 0.044 − 3.022 16.3 − 1129.4 Other variables 0.088 3.158 32.8 1180.3 Source: Author’s Calculations Based on the data from CFPS of 2014, 2016 and 2018 The decomposition is based on the results from the IV method “Other variables” include years of schooling, years of work experience, party membership, urban, occupation, industry sector, workplace ownership, region, and years (72.4%) than the Internet access (50.1%). The results Second, the gender difference in return to Internet use indicate that although both components drive the gen- reduces the gender wage gap in three cohorts; the effects der wage gap in China, the effect of the gender differ - are greater for the younger age cohorts. ence in return to Internet use is greater, suggesting that The results of Table  8 suggest that the gender disparity workplace discrimination against women in terms of in Internet access in the middle/older age cohorts and the Internet use is a serious issue in China. workplace discrimination against the younger women in Table 7 shows the decomposition results based on the terms of Internet use widens the gender wage gap in China. These results contradict those of Qi and Liu (2020), educational attainment group. The results indicate that who report that both components reduce the gender the effects of gender differences in both the Internet wage gap in the younger, middle-aged, and older groups. access and return to Internet use differ across different There are two reasons: first, the method of analysis dif educational groups. - First, the gender disparity in Internet access reduces fers. This study conducts the decomposition analysis the wage gap in three educational groups; its effect is based on the IV method, whereas Qi and Liu (2020) use greater in the high-education group than those in the the OLS model; it can be concluded that an endogeneity low- and middle-education groups. problem such as the unobservable omitted variable issue Second, the gender difference in return to Internet use might exist in the earlier studies. Second, the period of widens the gender wage gap in the high-education group, analysis in this study is 2014–2018, whereas it is 2010– whereas they reduce the wage gap in the low- and mid- 2015 in Qi and Liu (2020). With Internet diffusion and dle-education groups. Internet technology progressing, the effects of both com - Hence, these results suggest that the workplace dis- ponents may change over time. The results in this study crimination against female workers is greater for the high suggest that the gender divisions in Internet accessibility education group in terms of Internet use than the other might become severe in the middle-aged and older age group, which widens the gender wage gap in China. cohorts in the recent period (2014–2018). Table  8 presents the results of the decomposition analysis for three age cohorts: (i) the older cohort (born 4.6 Further discussions on the limitations of this study before 1969), (ii) the middle-aged cohort (born between It should be noted that this study has several limita- 1979 and 1989), and (iii) the younger cohort (born after tions. First, although we used the IV, LV, RE, and FE 1990). models to attempt to address the endogeneity problem, First, the gender disparity in Internet access widens future research should also explore the causal association the gender wage gap in three age cohorts; the effects are between Internet use and the gender wage gap. greater for the middle-aged (54.6%) and older age cohorts Second, the gender gap in the return to Internet (47.3%) than for the younger age cohorts (16.3%). usage may also result from the gender disparity in Internet use and gender wage gap: evidence from China Page 15 of 17 15 Internet-using abilities (or skills). Ge and Zhou (2020) Fourth, the influence of Internet usage on the gender report that the firm attributes (e.g., capital, trade expo - wage gap varies with the educational attainment and age sure) and firm technology level (e.g., robot or computer cohort groups. For example, the gender difference in the use situations) affect the gender wage gap. Future studies return to Internet use widens the gender wage gap in the can conduct a detailed survey on individuals’ Internet use high-education group while they reduce the wage gap in skills and workplace technology levels. the low- and middle-education groups; the influences Third, although the occupation, industry sector, and of gender disparity in Internet access on the formation enterprise ownership type were used to control the influ - of gender wage gap are greater for the middle-aged and ence of the workplace on the gender wage gap, other factors older generations than the younger generation. (e.g., wage and employment systems, enterprise attrib- The study highlights two policy implications. First, utes) may also affect the wage gap. We should conduct an there exists a large gender difference in return to Internet employer-employee survey on the issue in the future. use even when the other factors (e.g., education, occupa- Fourth, some studies have found that the work prefer- tion) are held constant, which highlights the prevalence ences differ by gender (e.g., Beblo and Görges 2018). This of workplace discrimination against women in terms of gender preference disparity in using new technology can Internet use, especially for the higher-educated women. also affect Internet access and return to Internet use. Fur - The enforcement of the implementation of equality poli - ther research should consider the influence of personality cies, such as the equality employment policy and “equal factors and self-selection on the issue. pay for equal work” policy, can be expected to reduce the Finally, due to the limitation of the survey period, we gender wage gap. Moreover, the discrimination against only investigated the issue in the current period (2014, women may be caused by the more family responsi- 2016 and 2018), and the longer-term analysis has become bilities for women than men (Connelly et  al. 2018; Ma a new challenge in the future. 2021a, b). Thus, policies that reduce the responsibili - ties of childcare and geriatric care, such as the one to promote the establishment of public kindergartens and 5 Conclusions long-term care insurance, are also expected to close the Using national longitudinal data from CFPS of 2014, gender wage gap in the long run. Second, the results 2016, and 2018, this study empirically analyzed the influ - suggest that the policies aimed at the reduction of the ence of Internet use on the gender wage gap in China, Internet access disparities among various groups, such considering the endogeneity problems. It yields the fol- as women in the middle-aged and older generations, may lowing four main conclusions. contribute to reducing the gender wage gap. First, according to the results derived from the OLS Despite these limitations mentioned above, this study model, the return to Internet usage is higher for women investigated the influence of Internet use on the gen - than men, meaning the results are similar to those of earlier der wage gap and provided new evidence on the deter- studies. However, when longitudinal survey data is used to minants of the gender wage gap in the era of the digital address heterogeneity and other endogeneity issues based economy from China. We anticipate that the insights on the IV and FE models, the results show that the return about the gender disparities in Internet access and gen- to Internet usage is higher for men than women. The indi - der difference in return to Internet use (including the vidual heterogeneity problem considerably affects the esti - discrimination against women in terms of Internet use mations, thus suggesting an estimation bias in the existing in the workplace), that contribute to the formation of the literature. The results based on the frequency of Internet gender wage gap in China, can provide valuable lessons use for different purposes confirm the conclusions. for other countries as well. Second, the gender difference in return to Internet use differs by heterogeneous groups: it is higher in the mid - Supplementary Information dle-/high- education groups and middle-aged/older age The online version contains supplementary material available at https:// doi. cohorts than those in low education and younger age org/ 10. 1186/ s12651- 022- 00320-9. cohorts. Third, the decomposition results indicate that, in gen - Additional file 1: Table S1. The gender differences in return to Internet use. Table S2. Results using frequency of Internet use for different eral, the two components (the gender disparity in Inter- purposes. Table S3. The gender differences in return to Internet use by net access and the gender difference in return to Internet educational group. Table S4. The gender differences in return to Internet use) drive the gender wage gap in China; the effect of use by age cohort. gender difference in the return to Internet use is greater. 15 Page 16 of 17 X. Ma Acknowledgements Beaudry, P., Lewis, E.: Do male–female wage differentials reflect differences in The authors are grateful to professor Boris Hirsch, and anonymous reviewers the return to skill? Cross-city evidence from 1980–2000. Am. Econ. J. Appl. for their helpful comments and suggestions. Econ. 6, 178–194 (2014) Beblo, M., Görges, L.: On the nature of nurture: The malleability of gender dif- Author contributions ferences in work preferences. J. Econ. Behav. Organ. 151, 19–41 (2018) X. MA designed the study, collected the data, performed the formal analyses, Becker, G.S.: The economics of discrimination. University of Chicago Press, and wrote the original manuscript; the author read and approved the final Chicago (1957) manuscript. Becker, G.S.: Human capital: a theoretical and empirical analysis, with special reference to education. Columbia University Press, New York (1964) Author information Bergmann, B.R.: Occupational segregation, wages and profits when employers Dr. X. MA is a professor at the Faculty of Economics, Hosei University, Japan. discriminate by race and sex. East. Econ. J. 1, 103–110 (1974) She was the editor of the Japanese Journal of Comparative Economics, Asian Biewen, M., Fitzenberger, B., Seckler, M.: Counterfactual quantile decomposi- Studies, and the Journal of Chinese Economics. Her articles have been published tions with selection correction taking into account Huber/Melly (2015): in peer-reviewed journals such as China Economic Review, Journal of Asian Eco- an application to the German gender wage gap. Labour Econ. 67, 101927 nomics, Economic Systems, Education Economics and Journal for Labor Market (2020) Research. She is the author of Economic Transition and Labor Market Reform in Bimber, B.: Measuring the gender gap on the Internet. Soc. Sci. Q. 81, 868–876 China (Palgrave Macmillan 2018), Female Employment and Gender Gap in China (2000) (Springer 2021), and the single editor of Employment, Retirement and Lifestyle in Blinder, A.: Wage discrimination: reduced form and structural estimation. Hum. Aging East Asia (Palgrave Macmillan 2021). Resour. Manag. Res. 8, 436–455 (1973) Borghans, L., Ter Weel, B.T., Weinberg, B.A.: People skills and the labor-market Funding outcomes of underrepresented groups. ILR Rev. 67, 287–334 (2014) This research was supported by JSPS (Japan Society for the Promotion Sci- Broadband commission: Doubling digital opportunities: enhancing the inclu- ence) Grant-in-Aid for Scientific Research (Grant Numbers: 20H01512 and sion of women and girls in the information society. UNESCO ITU (2013). 20H01489). Brown, R.S., Moon, M., Zoloth, B.S.: Incorporating occupational attainment in studies of male/female earnings differentials. J. Hum. Resour. 15, 3–28 Data availability (1980) The data can be obtained through the website https:// opend ata. pku. edu. cn/ Cao, J., Jiang, T.: The impact of Internet usage on women’s income: An analysis datav erse/ CFPS? langu age= en. The datasets constructed by the author are based on CFPS data. J. Tianjin Univ. Fin. Econ. 371, 79–95 (2020). (曹景林, available upon request. 姜甜:互联网使用对女性收入的影响-基于CFPS数据的经验证据.《天 津财经大学学报》, 371(12), 79–95 (2020)) Chen, X.: The evolution of female labour force participation in urban China. Declarations Econ. Transit. 27, 267–299 (2019) Chen, L., Fan, H.: Does family elderly care reduce female employment? Two- Ethics approval and consent to participate Stage Residual Inclusion Approach. Popul. Res. 40, 71–81 (2016). (陈璐,范 The dataset used in this study, the China Family Panel Studies (CFPS), is pub- 红丽:家庭老年照料会降低女性劳动参与率吗?—基于两阶段残差介 licly available (http:// opend ata. pku. edu. cn/ en). The medical ethics committee 入法的实证分析.《人口研究》,40(3), 71–81(2016)) approved the CFPS study, and all interviewees were required to sign informed Chen, L., Fan, H., Zhao, N., Chu, L.: The impact of informal care on employ- consent. Ethics approval for the data collection in CFPS was obtained from ment for women in China. Econ. Res. J. 51, 176–189 (2016). (陈璐,范红 the Biomedical Ethics Review Committee of Peking University (IRB00001052- 丽,赵娜,褚兰兰:家庭照料对女性就业的影响.《经济研究》,51(3), 14010). Ethics approval for the use of CFPS data was obtained from the Univer- 176–189(2016)) sity of Newcastle Human Research Ethics Committee. And all the participants China Family Panel Studies (CFPS): https:// opend ata. pku. edu. cn/ datav erse/ gave their written informed consent before any study procedures began. CFPS? langu age= en. Accessed 20 May 2020. Connelly, R., Dong, X., Jacobsen, J., Zhao, Y.: The care economy in post-reform Consent for publication China: feminist research on unpaid and paid work and well-being. Fem. The author of this study declares that this manuscript does not infringe Econ. 24, 1–30 (2018) copyright, moral rights, or other intellectual property rights of any other Cook, S., Dong, X.Y.: Harsh choices: Chinese women’s paid work and unpaid person. The author of this study testifies that this paper is original and has not care responsibilities under economic reform. Dev. Change 42, 947–965 incurred any type of plagiarism. The author of this study declares that the arti- (2011) cle is original, has not already been published in a journal, and is not currently Deyyling, L.J.: Broadband in the labor market: the impact of residential high- under consideration by another journal. The author of this study agrees to the speed internet on married women’s labor force participation. Ind. Lab. terms of the Springer Open Copyright and License Agreement. Relat. Rev. 70, 451–482 (2017) DiMaggio, P., Hargittai, E., Neuman, W.R., Robinson, J.P.: Social implications of Competing interests the internet. Annu. Rev. Sociol. 27, 307–336 (2001) The author declares no conflicts of interest associated with this study. Eastin, M.S., Cicchirillo, V., Mabry, A.: Extending the digital divide conversation: examining the knowledge gap through media expectancies. J. Broadcast. Received: 19 March 2022 Accepted: 10 August 2022 Electron. Media 59, 416–437 (2015) Fatehkia, M., Kashyap, R., Weber, I.: Using Facebook ad data to track the global digital gender gap. World Dev. 107, 189–209 (2018) Ge, Y.: Sectoral choice and the gender wage gap: 1988–2001. China Econ. Q. 2, 607–628 (2007). (葛玉好: 部门选择对工资性别差距的影响:1988– References 2001年. 《经济学季刊》,2,607–628 (2007)) Alam, K., Mamun, S.A.K.: Access to broadband Internet and labour force out- Ge, S., Zhou, Y.: Robots, computers, and the gender wage gap. J. Econ. Behav. comes: a case study of the Western Downs Region. Queensland. Telemat. Organ. 178, 194–222 (2020) Inform. 34, 73–84 (2017) Gustafsson, B., Li, S.: Economic transformation and the gender earnings gap in Alozie, N.O., Akpan-Obong, P.: The digital gender divide: confronting obstacles urban China. J. Popul. Econ. 13, 305–329 (2000) to women’s development in Africa. Dev. Policy Rev. 35, 137–160 (2017) Hafkin, N.J., Huyer, S.: Women and gender in ICT statistics and indicators for Arrow, K.J.: The theory of discrimination. In: Asehenfelter, O., Rees, A. (eds.) Dis- development. Inf. Technol. Dev. 4, 25–41 (2007) crimination in Labor Market. Princeton University Press, Princeton (1973) Haight, M., Quan-Haase, A., Corbett, B.A.: Revisiting the digital divide in Atasoy, H.: The effects of broadband Internet expansion on labor market Canada: the impact of demographic factors on access to the internet, outcomes. ILR Rev. 66, 315–345 (2013) Internet use and gender wage gap: evidence from China Page 17 of 17 15 level of online activity, and social networking site usage. Inf. Commun. Organization for Economic Co-operation and Development (OECD): Bridging Soc. 17, 503–519 (2014) the digital gender divide: include, upskill, innovate. OECD, Paris (2018) Hirsch, B.: Spatial Monopsony and the Gender Pay Gap. Springer, New York Pabilonia, S.W., Zoghi, C.: Returning to the returns to computer use. Am. Econ. (2010) Rev. 95, 314–317 (2005) Iwasaki, I., Ma, X.: Gender wage gap in China: a large meta-analysis. J. Lab. Phelps, E.S.: The statistical theory of racism and sexism. Am. Econ. Rev. 62, Mark. Res. 54, 17 (2020) 659–661 (1972) Kidd, M.P.: Sex discrimination and occupational segregation in the Australian Qi, Y., Liu, C.: Does the use of internet reduce the difference in gender pay in labour market. Econ. Rec. 69, 44–55 (1993) the context of digital economy? Econ. Theory and Econ. Admi. 9, 70–87 Kidd, M.P., Shannon, M.: Does the level of occupational aggregation affect esti- (2020). (戚聿东,刘翠花:数字经济背景下互联网使用是否缩小了性别 mates of the gender wage gap. Ind. Lab. Relat. Rev. 49, 317–329 (1996) 工资差异.《经济理论与经济管理》, 9, 70–87 (2020)) Krueger, A.B.: How computers have changed the wage structure: evidence Rice, R.E., Katz, J.E.: Comparing internet and mobile phone usage: Digital from microdata, 1984–1989. Q. J. Econ. 108, 33–60 (1993) divides of usage, adoption, and dropouts. Telecommun. Policy 27, Li, S., Ma, X.: Occupational segregation and gender wage differentials in urban 597–623 (2003) China. Chin. J. Popul. Sci. 5, 2–13 (2006). (李实,马欣欣:中国城镇职工 Riggins, F.J., Dewan, S.: The digital divide: Current and future research direc- 的性别工资差异与职业分割的经验分析.《中国人口科学》5, 2–13 tions. J. Assoc. Inf. Syst. 6, 298–337 (2005) (2006)) Scheerder, A., van Deursen, A., van Dijk, J.: Determinants of internet skills, uses Li, S., Yang, X.: The determinants of gender wage gaps in migrants Comp. Econ. and outcomes. A systematic review of the second- and third-level digital Soc. Syst. 5, 82–89 (2010). (李实,杨修娜:农民工工资的性别差异及其影 divide. Telemat. Inform. 34, 1607–1624 (2017) 响因素.《经济社会体制比较》,5, 82–89 (2010)) Vick, B.: Measuring links between labor monopsony and the gender pay gap in Li, S., Song, J., Liu, X.: Evolution of the change of the gender wage gap among Brazil. IZA J. Dev. Migr. 7, 10 (2017) China’s urban employees. Manag. World 3, 53–65 (2014). (李实, 宋锦, 刘 Wang, M.: Work chance and wage differentials in the urban labor market: Work 小川:中国城镇职工性别工资差距的演变.《管理世界》, 3, 53–65 and wage of migrants. Chin. Soc. Sci. 5, 36–46 (2005). (王美艳: 城市劳动 (2014)) 力市场上的就业机会与工资差异-外来劳动力就业与报酬研究. 《中 Liu, P., Meng, X., Zhang, J.: Sectoral gender wage differentials and discrimina- 国社会科学》,5, 36–46 (2005)) tion in the transitional Chinese economy. J. Popul. Econ. 13, 331–352 Wang, M., Cai, F.: Gender earnings differential in urban China. Rev. Dev. Econ. (2000) 12, 442–454 (2008) Liu, S., Zhang, X., Yang, Z.: Impact of Internet use on rural residents’ income. World Bank, http:// data. workd bank. org/ indic ator. Accessed 30 Jun 2022. Quant. Econ. Tech. Econ. Res. 38, 103–119 (2021). (刘生龙,张晓明,杨 Wu, E.: Analysis of Internet usage and gender wage difference. Coop. Econ. 竺松:互联网使用对农村居民收入的影响.《数量经济技术经济研 Technol. 16, 138–140 (2021). (吴恩喆:互联网使用与性别工资差异分 究》,38(4), 103–119 (2021)) 析.《合作经济与科技》, 6, 138–140 (2021)) Ma, X.: Ownership sector segmentation and the gender wage gap in urban Zhao, Y., Li, X.: The impact of internet use on individual health: Based on the China during the 2000s. Post-Communist Econ. 30, 775–804 (2018) evidence analysis of the Chinese Family Panel. Stud. Sociol. 5, 139–144 Ma, X.: Female employment and gender gap in China. Springer, Singapore (2020). (赵颖智,李星颖:互联网使用对个体健康的影响-基于中国家庭 (2021a) 追踪调查数据的实证分析.《江汉论坛》, 5, 139–144 (2020)) Ma, X.: Grandchildren care and labor supply of middle-aged grandmothers: Zhuang, J., Liu, A., Sun, C.: Does Internet reproduce gender inequality? A study evidence from China. J. Fam. Econ. Issues (2021b). https:// doi. org/ 10. on Internet use wage premium. Social 36, 88–106 (2016). (庄家炽,刘爱, 1007/ s10834- 021- 09794-0. (Early published online) 玉孙超:网络空间性别不平等的再生产:互联网工资溢价效应的性别 Ma, C., Gu, H., Li, J.: Change of gender wage gaps in Chinese labor market: evi- 差异 以第三期妇女地位调查为例.《社会》, 36 (5), 88–106 (2016)) dence from quantile decomposition analysis of panel data. World Econ. Pap. 2, 96–108 (2013). (马超, 顾海, 李佳佳: 中国劳动力市场上的性别 Publisher’s Note 工资差异变化研究:来自面板分位数回归分解方法的证据.《世界经 Springer Nature remains neutral with regard to jurisdictional claims in pub- 济文汇》, 2, 96–108(2013)) lished maps and institutional affiliations. Mao, Y., Zeng, X.: Does Internet use promote female employment? An empiri- cal study based on the data from the CGSS. Econ. Dyn. 6, 21–131 (2017). (毛宇飞,曾湘泉:互联网使用是否促进了女性就业-基于CGSS数据的 经验分析.《经济学动态》, 6, 21–131 (2017)) Masso, J., Meriküll, J., Vahter, P.: The role of firms in the gender wage gap. J. Comp. Econ. 50, 454–473 (2022) Meng, X.: Gender occupational segregation and its impact on the gender wage differential among rural-urban migrants: a Chinese case study. Appl. Econ. 30, 741–752 (1998) Meng, X., Zhang, J.: Two-tier labor market in urban China: occupational segregation and wage differentials between urban residents and rural migrants in Shanghai. J. Comp. Econ. 29, 485–504 (2001) Miller, P.W.: The wage effect of the occupational segregation of women in Britain. Econ. J. 97, 885–896 (1987) Miller, P., Mulvey, C.: Computer skills and wages. Aust. Econ. Pap. 36, 106–113 (1997) Mincer, J.: Schooling, experience and earning. Columbia University Press, New York (1974) National Bureau of Statistics NBS: China statistical yearbook 1999. China Statistics Press, Beijing (1999).. (国家统计局: 中国统计年鉴1999.中国统 计出版社,北京(1999)) National Bureau of Statistics NBS: China statistical yearbook 2021. China Statis- tics Press, Beijing (2021).. (国家统计局: 中国统计年鉴2021.北京:中国 统计出版社,北京(2021)) Oaxaca, R.: Male–female wage differentials in urban labor markets. Int. Econ. Rev. 14, 693–709 (1973) Ono, H., Zavodny, M.: Digital inequality: a five country comparison using microdata. Soc. Sci. Res. 36, 1135–1155 (2007)

Journal

Journal for Labour Market ResearchSpringer Journals

Published: Dec 1, 2022

Keywords: Gender wage gap; Internet use; Return to Internet use; Internet access; China; J16; J24; J31; O33

References