Gender Pay Gap in the Gig Economy

3.1 Data introduction

The dataset used in this paper is based on a survey of platform workers jointly conducted by South China University of Technology and Xinbao Technology Enterprises. The survey was conducted in April-July 2021, covering 31 provinces (municipalities and autonomous regions) in the mainland, as well as Hong Kong, Macao and Taiwan, and the survey subjects covered 21 occupations such as food delivery riders, ride-hailing drivers, freelance photographers, online anchors, freelance lawyers, self-media experts, online education and training. According to the subject of the study, the raw data are processed as follows: Only practitioners who are not less than 16 years old and whose place of work is in the mainland are retained, and 10110 samples are finally obtained. Merge similar occupations (e.g., freelance copywriters, web novel writers, and freelance translators into an online content service) to increase the sample size of specific categories of occupations. In the end, 21 occupations are merged into 14 types, namely ride-hailing services, home delivery services, delivery services, community group buying, online education, live broadcasting, program development, self-media authors, online content services, art creation, online consulting, micro-business, independent entrepreneurship, and other occupations. Of these, the first four are “geographically tethered work” because they require a specific geographic location to do it. The last three are “platform management work”, that is, using the platform as the carrier for entrepreneurship. The rest are “cloud work” which can be done online and not constrained by geospatial.

3.2 Main Variables

The variable we focus on the most is the monthly income of platform workers, identified by the question “What is your average gross income per month?”. There are fifteen income ranges for the answer options to this question, and the upper and lower bounds of each range are detailed in note (a) of Table 1. Gender is the key explanatory variable as well as a dummy variable, with respondents being female=1 and male=0. We are also concerned about other factors that may affect income, including individual characteristics, individual economic status, individual social behavior, and individual work-related situations. See Table 1 for details.

3.3 Statistical Facts

Table 1 provides descriptive statistical results for the variables covered in this paper. First of all, the average value of the monthly income variable is 3.565, that is, the average monthly income of workers is in the middle of 5000~7500 yuan (corresponding to about 6412 yuan). Further statistics show that more than two-thirds (68.12%) of workers have an income of more than 5000 yuan. In 2021, the average annual wage of urban non-private sector employees was 106837 yuan (i.e., 8903 yuan/month) and the average annual wage of private sector employees was 62884 yuan (i.e., 5240 yuan/ month) released by the National Bureau of Statistics., and the monthly income of platform workers is lower than the former and higher than that of the latter.

Figure 1 depicts the distribution of gender income in different intervals, showing a visual comparison of gender pay gaps. It can be seen that more than 95% of both men and women earn less than 20000 yuan (20K). Among them, women’s income is more inclined to the left, with the highest frequency in the [2500,5000] yuan range, while men’s income is relatively far to the right, with the highest frequency in the [5000,7500] yuan range. The KS test (P<0.01) showed that there are significant differences in income distribution between the gender. From the mean value of the interval scale, the male value is 3.735 (about 6837 yuan), which is significantly higher than the female 3.323 (about 5807 yuan), and the income gap is 1030 yuan (P<0.01).

Figure 1

Distribution of Monthly Income of Workers in the Gig Economy by Gender

According to the 2021 China Workplace Gender Pay Gap Report by the BOSS Zhipin Research Institute, the gender wage ratio of urban workers in China in 2021 was 0.77, that is, the monthly salary of women was 77% of that of men, which is comparable in time. In addition, the period comparable is the 2021 CGSS data mentioned earlier, in which the annual income of men is 76796 yuan (equivalent to 6399 yuan per month), the annual income of women is 50844 yuan (equivalent to 4237 yuan per month), and the ratio of female income to male income is 0.66. According to the survey data of this paper, the proportion of women’s monthly income to men’s monthly income in the gig economy is 5807 yuan/6837 yuan = 0.85, which is higher than 0.8, 0.77 or 0.66. This means that the gender pay gap in the new employment form of the gig economy has narrowed compared with the traditional employment form, especially when compared with the data of the same period.

4.1 Regression Model Setting

Does gender still have a significant impact on income after controlling for observables? For this reason, the regression model is set as follows:

Among them, the explained variable ln(Income_i) is the logarithm of the monthly income of individual i, and Female is the core explanatory variable, that is, the gender dummy variable (female=1). X_i is the individual characteristics and the work characteristics of i (e.g., working hours). Some characteristics (e.g., education, health, work) may be inherently gender-influenced, so they are undesirable control variables, and their inclusion in the controls may underestimate or overestimate the gender pay gap. However, we are concerned with not only the gender pay gap, but also whether the impact of gender on income (direct or unobservable) persists after individual income is explained by observable characteristics. We still try to control for more observable features, even though they may be affected by gender. In addition, IND is an occupation-fixed effect, and PRV is a region-based fixed effect, which excludes the income gap caused by individuals working in different occupations and regions. ϵ_i is a random perturbation term.

4.2 Benchmark Model Estimation Results

Monthly income is interval rather than continuous, and the appropriate estimation method is interval regression.^[1] Table 2 shows the benchmark model estimates. Column (1) does not have any control variables. Column (2) controls for relatively independent and exogenous factors such as age, household registration, and whether or not they are local. Column (3) further controls for observable individual characteristics. Column (4) further controls for observable work characteristics (although these may not be qualified control variables). First, the regression coefficients of the gender variables in each column are significantly negative, and the qualitative results are consistent. Using column (4) as the benchmark result, after controlling for observable individual characteristics and work characteristics, the monthly income of women in the gig economy is still significantly lower than that of men by about 10.8%. In other words, more than half of the income gap between the gender can be explained by observable characteristics. These results support the idea that gender pay gap still exist in gig work, even after controlling for observables.

The coefficients of the other control variables are not the focus of our attention, and most of them represent only correlations, but they can still be meaningfully interpreted. According to the results of column (2)~(4), the age and income of the gig workers show a significant inverted U-shaped relationship, and it is estimated that the peak income appears around 35~36 years old, which coincides with the “35-year-old phenomenon” in the traditional employment market. The income of urban workers with household registration is significantly higher than that of rural households by 6.3%, which is consistent with the existing literature in the field of traditional employment (Yu and Chen, 2012; Wu et al., 2015). The estimated coefficient of years of education suggests that an individual’s monthly income will increase significantly by 4% for each additional year of education. This is in line with the classic assertion of labor economics that education contributes to income growth, and also implies that educating human capital may still be important for income in the gig economy, rather than as found by Herrmann et al. (2023) that education is no longer important for income in the gig economy. However, compared with the Asian average of 9.6% and the global average of 10.1% (Psacharopoulos, 1994) and the average rate of return to education of migrants in China (6.66% in 2017) (Yang and Wang, 2021), the role of education in the gig economy on income seems to be less than that of traditional employment. Social relationships and social capital still have an important impact on income in the gig economy, and workers’ participation in online or offline communities, as well as employment through acquaintances, have a significant positive impact on their income. In terms of job characteristics, workers are required to invest more self-owned assets, work more days per week, sign labor or work contracts/agreements, and enjoy social insurance, all of which are significantly positively related to higher monthly income. However, there is a significant negative correlation between work stress and income. There is no significant difference in income between the two types of full-time and part-time platform workers, and the regression coefficient is also small (–0.020). This is easier to understand from the perspective of labor market equilibrium, since the entry or exit platform is relatively free, then after controlling for other factors, there should not be a large income gap between full-time and part-time workers.

We also did some heterogeneity analysis, and the basic conclusions are: In the three categories of platform jobs, such as platform management jobs, geographically bound jobs, and cloud jobs, women’s earnings are significantly lower than men’s, but in more subdivided occupations, they are different. In most occupations, including platform self-employment, online ride-hailing, delivery services, community group buying, self-media, online education, and content creation, women’s income is significantly lower than that of men. In a few occupations, such as live streaming, home service, online counseling, program development, and art creation, the gap between women’s income and men’s income is not statistically significant, and even women in the first two occupations (live streaming and home service) have slightly higher income than men (although statistically not significant). This shows that although women’s income is significantly lower than that of men in the gig economy, there is also a heterogeneity effect in different occupations, and the gender pay gap is also different for different occupations and jobs.

5.1 Is There Occupational Segregation in the Gig Economy: A Counterfactual Prediction of Occupational Entry

Through counterfactual analysis, the basic idea of gender occupational segregation in the gig economy is as follows: Firstly, the multiple discrete choice mlogit model is used to estimate the probability of the two gender entering different occupations. Then, according to the coefficient of the male occupation acquisition equation, combined with female characteristics, the occupational distribution of a woman is predicted if she becomes a man. Finally, comparing the actual occupational distribution of women with the counterfactual predicted occupational distribution, if there is a gap between the two, it indicates that women have experienced preferential treatment or discrimination when entering the workplace. In the same way, the actual occupational distribution of men and the counterfactual predicted occupational distribution can be examined to determine the preferential treatment or discrimination that men experience when entering the workplace. Table 3 shows the structural gap between men and women in terms of career access. Among the four professions of ride-hailing, home delivery service, delivery service, and program development, men are more engaged (preferential) and women are less engaged (discriminated against). In the other ten occupations, including online education, live broadcasting, community group buying, self-media authors, online content creation, art creation, online consulting, micro-business, self-employment, women are more engaged (preferential) and men are less engaged (discriminated against). The above occupational distribution characteristics are similar to those found by Galperin (2021). We can conclude that the occupational segregation of gender in the gig economy is relatively prominent in the three occupations of ride-hailing, delivery service, and online education, and is not prominent in most other occupations. We have also tried to calculate the structural disparities in occupational access between the genders using the occupational distribution of non-discrimination as a reference point, and the results are generally consistent with Table 3.

According to the traditional “occupational crowding” hypothesis (Bergmann, 1971), occupations with high status, high income, and prestige are mostly occupied by men, and female workers are often crowded in occupations with low wages, low social status, and low professional prestige. This hypothesis no longer seems to hold true in the gig economy. Figure 2 plots the average income bracket for each occupation (represented by bars), as well as the average income bracket for men and women (represented by solid and dashed lines, respectively), with the horizontal axis being the average income bracket (3.565, equivalent to 6412 yuan). Taking online education as the boundary (online education income is almost equal to the average income), 7 occupations on the left have a higher than average monthly income and can be defined as high-income occupations, and 6 occupations on the right, such as community group buying, have a lower monthly income than average and can be defined as low-income occupations.

Figure 2

Comparison of Gender Concentration and Average Monthly Earnings by Occupation

Note: The bars in the figure represent the average income levels of each occupation (sorted by occupational income level from high to low they are: 1.online consultation, 2. program development, 3. live broadcast, 4.art creation, 5.self-media authors, 6. independent enterpreneurship, 7. ride-hailing, 8. online education, 9.community group buying, 10. online content services, 11. delivery services, 12. micro-business, 13. home delivery services, 14. other occupations), and the horizontal axis is the average income level of all occupations. The color of the bar corresponds to the concentration of entry between the genders, with dark colors representing occupations where men are more concentrated and light colors representing occupations where women are more concentrated. The black solid line indicates the average income bracket for men, and the grey line indicates the average income bracket for women.

From Figure 2, the following characteristic facts can be obtained: (1) Neither men nor women are more concentrated in high- or low-income occupations. Concentrated entry and insufficient entry are scattered in occupations with different incomes, so the traditional “occupational crowding hypothesis” no longer holds. (2) Men are more concentrated in a few occupations (the dark bars corresponded to the occupations in Figure 2), while women are more concentrated in most occupations (the occupations corresponding to the light bars in Figure 2). The gender pay gap in which women earn lower exists in low- and middle-income occupations, and in some occupations the gap is even larger. This shows that although the gig economy has “broken” the occupational crowding hypothesis, there is still a prominent phenomenon of “unequal pay for equal work” within the occupation, and this phenomenon mainly exists in low- and middle-income occupations. The gender pay gap in the gig economy may not primarily come from occupational segregation, but from gender pay gap within occupations.

5.2 The Relative Importance of Each Factor to Pay Gap: Decomposition of Pay Gap

In order to evaluate the importance of different factors on the gender pay gap in the gig economy, the income gap is decomposed by using Appleton decomposition method. Table 4 gives the results of the decomposition: The proportion of intra-occupational income gap in the total income gap is as high as 91.88%. The part of intra-occupation that can be explained (by age, years of education, household registration) accounted for 54.28%. The gap in the unexplainable part of intra-occupation is 37.6%. This gap stems from the unobservable preferential treatment of men (16.95%) on the one hand, and the unobservable discrimination against women (20.65%) on the other hand. The inter-occupational income gap accounted for only 8.12% of the total income gap. This is consistent with the intuitive results in Figure 2 above, in which intra-occupational factors are the main source of gender pay gaps, and the influence of inter-occupational factors is very limited.

In the last two items of Table 4, the explainable income gap accounts for 58.96% and the unexplained income gap accounts for 41.04%. Academics often classify the latter as implicit discrimination. This means that the income gap caused by implicit discrimination in the gig economy accounts for about 40%, which is much lower than the estimated results in China’s traditional employment patterns. The proportion of unexplained income gap in the existing literature is basically not lower than 60% (Luo et al., 2019), let alone less than 40%. We should also note that there is a question of comparability between different data sources, different time periods, and different disaggregation methods. Columns (3) and (4) of Table 4 provide the results of the same decomposition of CGSS2021 data that are relatively more comparable in terms of both time and control variables. It can be seen that the unexplained gender pay gap in the CGSS data accounts for 77.23%. In summary, it can be concluded that implicit gender discrimination still exists in the gig economy, but it is much better than that of traditional employment.