Digital technology access, labor market behavior, and income inequality in rural China

Abstract

This study uses China Family Panel Studies (CFPS) data from 2010 to 2018 to empirically investigate the interplay between digital technology access, labor market behavior, and income inequality in rural China. The following salient conclusions are derived. Digital technology access has a substantial negative influence on individual income inequality in rural China, with a more pronounced inhibitory effect on inequality among low-income groups, males, middle and higher professional classes, and younger cohorts. Mechanism analysis suggests that digital technology access significantly impacts a range of rural labor practices, including increasing the frequency of digital technology use among rural inhabitants, decreasing credit costs, enhancing entrepreneurial activities, and boosting rural labor mobility. Based on these findings, this study proposes accelerating digital infrastructure development in rural regions, improving digital and financial literacy among rural residents, and refining inclusive digital financial services to facilitate more stable and sustainable progress to promote common prosperity.

1. Introduction

The report from the 20th National Congress of the Communist Party of China posits that “Chinese modernization is a modernization of common prosperity for all.”1 Fostering common prosperity requires the fundamental consideration of the status quo of imbalance and inadequate development in China's new era, with a particular emphasis on significantly insufficient rural development and focusing on enhancing rural incomes and optimizing the rural income structure. To address this issue, the Outline of the 14th Five-year Plan for National Economic and Social Development and Vision 2035 of the People's Republic of China prioritized “accelerating the construction of a digital countryside,” improving rural infrastructure, invigorating the agricultural market, and enhancing rural residents' quality of life.2 According to the Statistical Report on Internet Development in China, as of June 2023, the internet penetration rate in rural areas had reached 76.4 %.3 The prevalence of digital technology in rural areas has facilitated and expedited rural residents' access to agricultural information, enabling their engagement in various online economic activities. The Ministry of Commerce of the People's Republic of China reported that by the end of September 2021, 16.4 million online rural stores had been founded nationwide, achieving an online retail sales volume of 2.13 trillion yuan. Furthermore, concerning information services integral to rural residents' lives, more than 14.89 million jobseekers have registered to use China's Employment Online platform, and the People's Bank of China indicated that 93.66 % of rural financial institutions had opened online channels, covering 519,500 administrative villages.4 Integrating digital technology in rural areas is assisting rural residents in these regions to move forward in promoting common prosperity.

In the post-poverty era, income disparities among different groups in rural areas remain pronounced. Data from the China National Bureau of Statistics show that in 2021, the average income level of rural households in the highest quintile was 8.87 times higher than that of low-income households.5 Narrowing this income gap requires using data as a crucial production factor, leveraging the endogenous dynamics of digital technology, and unleashing modern market forces for rural residents. It is essential to explore the dynamic evolution of rural income inequality and its key mechanisms in the digital era. This includes considering the various economic decision-making behaviors of farmers, who are the main income creators in rural areas, compared with the traditional labor force within the context of digital technology. Given this economic development background, this study contends that examining how digital technology access has improved the rural income structure by influencing labor market behavior is pertinent to achieving common prosperity.

Previous research has explored the long-term changes and causes of rural income inequality, focusing on trends, including Piketty et al. (2019) [1], who examined national accounts, surveys, and new tax data, finding that the income share of the top 10 % of the population within urban and rural areas in China increased significantly from 1978 to 2015. Additionally, a decreasing trend in the income share of the bottom 50 %, with rural income inequality widening substantially. Concurrently, Luo (2020) [2] discovered that the Gini coefficient of income inequality within rural areas exhibited a significant downward trend after entering the digital era. Generally speaking, in the 40 years following the implementation of reform and opening-up policies, China's rural residents' income gap has followed a pattern of rising, plateauing, or ceasing to rise and then declining with fluctuations (Chen and Zhang, 2021) [3]. Research on the factors influencing the income gap of rural residents can be divided into macro and micro perspectives.

Factors from a macro perspective include institutional elements such as differences in employment opportunities due to the household registration system, taxation, and regional variations have been found to be important causes of income disparity in China's rural areas (Li, 2003) [4]. Policy factors encompassing industrial policies (Cheng et al., 2015; Li and Ran, 2019) [5,6], investment policies (Zhang and Wan, 2016) [7], transfer payment policies such as pensions and agricultural subsidies (Yang et al., 2021) [8], and poverty alleviation policies (Zhang, 2020) [9] have significantly improved rural income distribution. In contrast, high-income groups in rural areas often engage in political elite capture during the marketization process, expanding the rural income gap (Cheng et al., 2016) [10]. From a micro perspective, human capital (Gao and Yao, 2006) [11], social capital (Song and Han, 2020) [12], labor mobility (Liu, 2018) [13], land transfers (Shi, 2020; Liu et al., 2021; Wu et al., 2022) [[14], [15], [16]], and livelihood diversity (Wang and Lv, 2021) [17] have been found to be influential factors affecting the income gap among rural residents.

In addition to the aforementioned influencing factors, some studies have begun to examine the impact of digital technology and the digital economy on rural residents' income within the context of realistic economic development. As new key elements, data and digital technology have been organically integrated with agriculture and rural areas, altering traditional production methods as well as distribution, allocation, and consumption patterns, permeating all aspects of production and life for rural residents. A significant body of literature has explored the impact of digital technology on rural residents' income and income inequality, primarily encompassing three considerations. First, digital technology improves rural residents' income and optimizes the income structure within rural areas (Cai et al., 2015; Fu et al., 2019; Liu and Han, 2018; Luo and Niu, 2019; Zhu and Song, 2020; Lin et al., 2020) [[18], [19], [20], [21], [22], [23]]. Second, unbalanced and insufficient digital technology development can exacerbate the rural income gap (Wang and Kuang, 2022; Luo and Niu, 2019) [22,24]. Third, digital technology has a complex, nonlinear relationship with income inequality (Li et al., 2019; Richmond and Triplett, 2017) [25,26].

In summary, the academic debate concerning whether digital technology can reduce rural income inequality remains unsettled. Building on contemporary research advances, this study explores the impact of changes in labor market behavior on intra-rural income inequality following digital technology access and tests the findings with micro empirical evidence, contributing theoretical significance. Our research methodology involves investigating the impact of digital technology access on income inequality among rural individuals by examining micro individuals as the target group, analyzing the heterogeneous effects, and exploring the specific ways in which digital technology access has affected income inequality among rural residents.

The marginal contributions of the current study are twofold. First, a limited number of investigations have explored the intra-rural income gap from the perspective of digital technology access (Li and Ke, 2021) [27]. These studies have generally used macro data and prioritized the effect of digital technology on rural residents' income increases. In contrast, this investigation uses micro-individual data to examine the impact of digital technology access on rural labor market behavior as a focal point, investigating its influence on rural residents' individual income inequality, which is the principal marginal contribution of the paper. Second, this study explores the theoretical mechanisms by which digital technology has affected income inequality among rural residents, analyzing how digital technology access has impacted rural laborers' behaviors concerning technology use frequency, credit availability, innovation and entrepreneurship, and mobility, subsequently affecting income inequality within rural areas.

The remainder of this paper is organized as follows. Section 2 presents the theoretical analysis and research hypotheses; Section 3 details our data sources, variable selection, and model settings; Section 4 analyzes the empirical results; Section 5 conducts mechanism testing; and Section 6 offers conclusions, policy implications and limitations.

2. Theoretical analysis and research hypotheses

As broadband and mobile internet coverage has expanded in rural areas, rural residents' cost of using internet technology has gradually decreased.6 Online access through broadband and mobile internet allows rural households to acquire advanced agricultural technology knowledge, access credit funding information, and learn about relevant employment opportunities, driving labor mobility and affecting rural residents' incomes. Specific groups, including low-income individuals, women, the elderly, and disabled individuals, can enjoy digital dividends by using digital technology, overcoming personal or external constraints, broadening social networks, obtaining accurate information, and achieving value pursuits. Digital technology access can dissolve traditional financial barriers, allowing disadvantaged groups to quickly secure loans, improve the probability of self-employment, and increase income sources. Furthermore, the digital economy's entrepreneurial employment incubator effect encourages disadvantaged groups to use digital technology to access job information promptly, engage in flexible nonagrarian employment opportunities, increase household incomes, and reduce income disparities with high-income groups.

Moreover, leveraging their advantages, high-income rural groups can use internet-enabled transaction and employment information to secure appropriate migrant work opportunities. This labor mobility significantly increases the likelihood that laborers will permanently settle in the cities in which they relocate. In addition, following the principle of diminishing marginal effects, as digital technology becomes more widespread, the growth of digital dividends for high-income groups becomes lower than that for low-income groups. Consequently, the process of accelerated income improvement with digital technology's help for low-income groups also signifies progressive income structure optimization within rural areas. In summary, the widespread application of digital technology in rural areas can effectively increase rural residents' income and reduce the income gap between low- and high-income groups. Based on these considerations, this study proposes the following hypotheses.

Hypothesis 1

Digital technology access significantly reduces intra-rural individual income inequality.

In summary, we presented the structural changes in income distribution among different groups within rural areas following digital technology access. Next, we further examine the impact of digital technology access on rural residents' income and income inequality, contending that the effects of digital technology on rural residents' income distribution can be briefly summarized as follows. First, low-income farmers can increase their earnings and narrow income gaps with nonfarm workers by increasing digital technology use frequency and obtaining accurate and effective information. Second, low-income groups can overcome the limitations of existing social circles and broaden social networks with digital technology's network effect, increasing credit availability, reducing unnecessary intermediate links, indirectly increasing revenue, and narrowing income gaps within rural areas. Third, farmers can access entrepreneurial employment opportunities through the digital technology job creation effect, increasing income channels and raising business and wage profits, while narrowing the intra-rural income gap. Fourth, with digital technology's aid, high-income rural groups can broaden their horizons, relocate from rural areas, integrate into cities, achieve value pursuits, and indirectly narrow the intra-rural income gap.

2.1. Frequency of digital technology use

As digital technology is adopted in rural areas more widely, the digital economy will deeply penetrate all aspects of rural residents' production and lives, leading to online activities such as webcasting, online shopping, and e-government participation. These activities will compel rural workers to improve internet navigation skills and increase their use frequency. This process is expressed in several ways. First, digital technology enables farmers to promptly obtain advanced agricultural technology knowledge to continuously improve agricultural production efficiency and revenue. Second, digital technology expands the market radius in which farmers can sell agricultural products, reinforcing communication with consumers and increasing income through promotions. Third, digital technology provides rural residents with the possibility of increasing employment opportunities, compelling farmers to use digital technology to regularly explore job information, search for higher incomes and obtain suitable jobs. Fourth, digital technology is widely used in government administration services, requiring rural residents to frequent government digital platforms to stay informed about government service information to obtain benefits such as government subsidies and transfer payments, ultimately improving rural residents' welfare. Based on this, we propose the following hypothesis.

Hypothesis 2

Digital technology access improves rural residents' income and optimizes income structure by increasing the frequency of digital technology use.

2.2. Credit availability

Additionally, digital technology access significantly increases loan opportunities for rural residents by providing online investment and financing platforms, effectively overcoming the temporal and spatial limitations of traditional financing processes. Therefore, financial institutions can meet remote rural areas' credit needs more efficiently and accurately, promoting the optimal allocation of credit funds and improved rural income structure. This impact manifests in several ways. First, using digital technology, farmers can access real-time information concerning financing supply and demand, improving borrowing efficiency and optimizing household capital structure (Han et al., 2019) [28]. Second, digital technology access reduces the cost of searching for information on fund supply and demand, indirectly increasing income and improving rural income distribution. Third, the application of digital technology has given rise to models such as microcredit, lowering the loan threshold for farmers and subsequently improving incomes. Fourth, digital technology access allows financial providers to gain insights into the economic circumstances of potential loan recipients, enhancing lending efficiency. The application of digital technology enables more convenient borrowing for rural residents, overcoming time and space restrictions, and reducing information dissemination lag, ultimately contributing to improved income. Based on these points, this study proposes the following hypothesis.

Hypothesis 3

Digital technology access can increase the availability of credit for farmers, enhance their overall income, and reduce intra-rural income inequality.

2.3. Innovation and entrepreneurship

Digital technology access can foster entrepreneurial behavior among farmers from product demand and job supply perspectives. On the demand side, digital technology access facilitates the exchange of information and ideas between farmers, enriching entrepreneurial resources and promoting entrepreneurial activities. The development of digital technology encourages diversified product demand, optimizes product matching and trading, and transforms the unidirectional output flow of product supply into a two-way exchange flow between supply and demand, as demonstrated by Luo and Li (2015) [29] and Guo and Luo (2016) [30]. This shift contributes to increased product market output and variety, promoting entrepreneurial activities. Furthermore, digital technology access offers valuable insights and decision-making information for farmers' entrepreneurial activities, with a crucial role in encouraging residents to seize business opportunities and facilitating information communication during the entrepreneurial process (Zhou and Fan, 2018) [31]. From the supply side, deepening digital technology access and overall improvement of residents' digital literacy can produce more entrepreneurial opportunities by influencing knowledge spillover and factors in the enterprise mix (Wang et al., 2022) [32]. Digital technology also fosters stronger social interactions, accumulates social capital, and reinforces the demonstration effect of entrepreneurial success, influencing neighboring entrepreneurship (Zhou and Fan, 2018) [31]. Based on this, we propose the following hypothesis.

Hypothesis 4

Digital technology access enhances entrepreneurial behavior among rural residents, raises overall income, and reduces intra-rural income inequality.

2.4. Labor mobility

The rapid advancement of the digital economy mitigates spatial and temporal constraints on labor mobility, optimizing the allocation efficiency of labor production factors (Huang and Wei, 2022; Acemoglu and Restrepo, 2018) [33,34]. The digital economy heightens rural laborers' subjective willingness to relocate, increasing cross-sectoral and cross-regional labor mobility. Digital technology access encourages cross-sectoral labor mobility by catalyzing the swift growth of the consumer internet, which is exemplified by digital finance, attracting numerous low-skilled rural laborers to transition into low-skilled, nonagricultural digital industries and expand income channels. Enterprises are more likely to select low-skilled laborers with qualities such as high stress resistance than middle-skilled laborers. Digital technology development also spurs inter-regional labor mobility, significantly increasing urban migrants' willingness to remain in cities (Zhang and Liu, 2022) [35]. Metropolises, which are characterized by high inclusiveness, abundant employment opportunities, and balanced welfare levels, draw more high-skilled rural residents to relocate to urban areas, enhancing employment prospects for low-skilled rural residents and ameliorating the intra-rural income gap. Based on this, this study proposes the following hypothesis.

Hypothesis 5

Digital technology access promotes rural labor mobility, augments overall income levels among farmers, and diminishes intra-rural income inequality.

3. Material and methods

3.1. Data sources and variable selection

3.1.1. Data sources

The dataset employed in this study is obtained from the China Family Panel Studies (CFPS) that is collected by the China Social Science Survey Center of Peking University, encompassing the tracking period from 2010 to 2018.7 As this research focuses on income and employment information for rural individuals, only data from rural individuals aged 16–65 years old were retained in the sample. Subsequently, the selected rural samples were cleaned by eliminating missing indicators and responses containing “don't know,” “refused to answer,” “not applicable,” and similar values. The final sample size comprises 53,385 individuals. To ensure the comparability of individual incomes across different periods, we adjust the income data using consumer price index (CPI) data for each province in the corresponding year, obtaining relevant CPI data from the China Statistical Yearbook.

3.1.2. Variable selection

Dependent Variable: The primary dependent variable in this study is the rural individual income inequality index. Referencing Zhang (2022) [36], we calculate the measure of individuals' relative deprivation based on total annual income. The income considered in this study refers to the sum of earnings, including wages, subsidies, allowances, grants, gratuities from various sources, rent, compensation, interest on deposits, dividends from stocks/funds/bonds, gifts received in RMB, and borrowed income. Specifically, we adopt Kakwani's (1984) [37] definition of relative deprivation. Suppose vector $\mathit{X}$ represents a group with a sample size of $n$, and the income of individuals in this sample ($x_i$) is arranged in ascending order to obtain the income vector $\mathit{X}=(x_1,x_2,\dots ,x_n)$. The relative deprivation experienced by individuals is denoted as $RD(x_i,x_j)$ and calculated as equation (1):

$$RD\left(x_i,x_j\right)=\frac{1}{n{\mu }_x}\sum _{j=i+1}^n(x_j-x_i)={\lambda }_{x_i}^{+}[({\mu }_{x_i}^{+}-x_i)/{\mu }_x]$$ (1)

where ${\lambda }_{x_i}^{+}$ represents the share of the total sample ($\mathit{X}$) for those in the sample with incomes above $x_i$, ${\mu }_{x_i}^{+}$ is the mean income of those in the sample with incomes above $x_i$, and ${\mu }_x$ is the mean income of the total sample $\mathit{X}$. $RD\left(x_i,x_j\right)$ falls within the range [0, 1], where a higher indicates a greater level of individual inequality.

The core explanatory variable is digital technology access, which is measured by cell phone (or computer) use, corresponding to the questionnaire questions “Have you ever used or are you currently using a cell phone?” and “Have you ever accessed the internet?” These questions assess participants' engagement with digital technology.

Mechanism Variables: The study's mechanism variables encompass the frequency of digital technology use, credit availability, entrepreneurship, and labor mobility. First, digital technology use frequency is assessed using the frequency of cell phone or internet use. The 2010 questionnaire asked, “How often do you use your cell phone when not on vacation?” with answers ranging from 1. occasionally to 4. almost every day. The 2014, 2016, and 2018 questionnaires asked, “In general, how often do you use the internet (for study/work/social/entertainment)?” with possible answers from 1. almost every day to 7. never. Due to inconsistent answer codes, we adjust the 2010 response options to align with subsequent years. Second, credit availability is measured by the presence of outstanding loans or borrowing practices, corresponding to questions regarding loans and borrowing channels in the questionnaires from 2010 to 2018. Third, entrepreneurship is determined by whether individuals started a business, with questions in the 2010 questionnaire referencing electricity usage for production, while the 2014–2018 questionnaires inquired about self-employment or private business involvement in the past 12 months.8 Fourth, labor mobility is evaluated based on household members working outside of homes, as indicated in the questionnaires from 2010 to 2018.

Notably, the frequency of digital technology use is an individual decision variable, while credit availability, labor mobility, and entrepreneurship are household decision variables. This distinction is made based on questionnaire content, variable settings, and individuals' autonomy in making decisions related to digital technology use, loans, employment, and entrepreneurship.

Control variables include individual characteristics, family environment features, and village features. Individual characteristics encompass gender, age, marital status, China Communist Party membership, occupation, education level, health status, and ethnicity. Marital status is simplified to married or not married. Occupations are categorized as higher class, middle class, and lower class.9 Village environment variables are assessed using villages' accessibility and development level, which is measured by the distance from the village office to the county town and villages' per capita income, referencing the CFPS2010 questionnaire. It is essential to note that the CFPS questionnaire only contained village residence questions in 2010 and 2014; thus, we employ interpolation methods to augment village residence data for 2016 and 2018.10

3.2. Descriptive statistics and variable characteristics analysis

3.2.1. Descriptive statistics of variables

Table 1 presents the descriptive statistical analysis of the primary variables, revealing that the average individual annual income during the studied period was 12,484 yuan, with a substantial increase from 8067 yuan in 2010 to 34,074 yuan in 2018. This growth rate surpasses the per capita GDP growth during the same timeframe. A significant improvement in digital technology access levels is also evident. Specifically, the percentage of the population using cell phones rose from 68.7 % in 2010 to 98.9 % in 2018, and the proportion of computer users increased from 11.4 % in 2010 to 25.4 % in 2018.11 Interestingly, the frequency of internet usage raised, shifting from 5.167 in 2010 to 3.491 in 2018.12

Table 1.

Descriptive statistics of selected variables.

Year			2010–2018 (N = 53,385)		2010 (N = 19,505)		2014 (N = 12,432)		2016 (N = 15,496)		2018 (N = 5952)
Variable types	Variable	Definition	Mean	Std.	Mean	Std.	Mean	Std.	Mean	Std.	Mean	Std.
Dependent variables	Income	Total annual personal income	12484	50093	8067	17604	9044	17309	12509	86211	34074	31004
	Log income	Log annual personal income	5.565	4.534	6.687	3.688	3.968	4.541	3.827	4.790	9.748	1.481
	RD K	Kakwani Index	0.698	0.325	0.661	0.295	0.776	0.332	0.788	0.309	0.420	0.262
Core explanatory variables	Mobile	Cell phone use (1 = yes)	0.828	0.377	0.687	0.464	0.858	0.349	0.920	0.272	0.989	0.104
	PC	Computer use (1 = yes)	0.174	0.380	0.114	0.318	0.223	0.417	0.181	0.385	0.254	0.435
Mechanism Variables	Loan	Loan recipient (1 = yes)	0.243	0.429	0.374	0.484	0.175	0.380	0.170	0.376	0.144	0.351
	Startup	Started a business (1 = yes)	0.101	0.301	0.0904	0.287	0.101	0.301	0.124	0.330	0.0736	0.261
	Mobility	Work outside the home (1 = yes)	0.490	0.500	0.343	0.475	0.569	0.495	0.560	0.496	0.623	0.485
	Frequency	Frequency of internet use	4.557	1.464	5.167	1.198	4.256	1.385	4.064	1.440	3.491	1.397
Individual Variables	Sex	Gender (1 = male)	0.509	0.500	0.472	0.499	0.527	0.499	0.505	0.500	0.608	0.488
	Age	Age	42.75	12.79	42.82	12.64	43.53	12.88	43.38	12.89	39.25	12.20
	Group	Age group under 40 years old (1 = yes)	0.393	0.488	0.386	0.487	0.361	0.480	0.379	0.485	0.524	0.499
	Occupation	Occupation (1–3)	1.830	0.493	1.564	0.524	1.955	0.264	1.952	0.381	2.120	0.612
	Education	Years of education	6.775	4.328	5.967	4.196	6.528	4.246	7.123	4.343	9.033	3.986
	Party	Party member (1 = yes)	0.0479	0.214	0.0382	0.192	0.0441	0.205	0.0548	0.228	0.0696	0.254
	Ethic	Ethnicity (1 = Han)	0.959	0.199	0.898	0.303	0.989	0.104	0.997	0.0584	0.996	0.0634
	Marry	Marital status (1 = married)	0.870	0.336	0.867	0.339	0.837	0.369	0.861	0.346	0.969	0.172
	Health	Health status (1–5)	2.500	1.272	1.809	1.031	2.884	1.261	2.962	1.233	2.762	1.122
Family Variables	Fasize	Family size	4.534	1.932	4.522	1.789	4.656	1.954	4.539	2.039	4.304	2.027
Village Variables	Distance	Distance from the village office to the county town	36.77	36.75	24.33	24.03	46.53	41.48	45.26	41.18	40.58	40.03
	Incper	Village development level	2044	12109	3118	3094	1508	17346	1316	15643	764.2	4540

Note: Total annual personal income is adjusted by CPI for each province using 2010 as the base period.

Furthermore, individual education exhibited a substantial increase, with the average years of education advancing from 5.98 in 2010 to 9.03 in 2018. Household size demonstrated a slight but stable decline. It is noteworthy that the village development level, gauged by per capita village income, decreased during the research period; however, this does not impact this study's analysis.13

3.2.2. Empirical analysis of key variables

This section describes the impact of individuals' digital technology access on income inequality. For instance, considering whether an individual uses a cell phone, access to digital technology significantly affects income inequality. On average, individuals who used cell phones exhibited an income inequality of approximately 0.67, which is substantially lower than that of individuals who do not use cell phones (0.85). When examining age groups, income inequality for individuals under 40 years old (0.61) was significantly lower than for those over 40 years old (0.76). By gender, men (0.61) experienced a considerably lower degree of income inequality compared with women (0.79), regardless of cell phone usage.

Fig. 1 illustrates individual income inequality after further stratifying the sample into age and gender samples according to digital technology access, aligning with the previously described circumstance. Fig. 2 presents the cumulative distribution function of individual inequality based on cell phone usage, revealing individuals who used cell phones had lower income inequality than those who did not.14

Fig. 1.

Income inequality (RD) for different age and sex samples under digital technology access.

Fig. 2.

Cumulative distribution of income inequality under digital technology access.

Fig. 1, Fig. 2 present the means of individual income inequality and the distribution of income inequality under different samples, respectively, highlighting the presence of between-sample heterogeneity in income inequality. To further verify this phenomenon, we conduct a formal randomized dominance test. Table 2 presents the results of the consistent nonparametric test of Lorenz stochastic dominance, indicating significant Lorenz dominance in income distribution conditioned on individuals' gender and age. Women's income inequality distribution significantly dominates men's, and the distribution of income inequality for individuals aged 40 years or older significantly dominates that of individuals younger than 40. Likewise, a noteworthy difference is evident in occupation conditions, revealing that the distribution of income inequality for individuals in low- and middle-class occupations significantly dominates that of high-class occupations across all data samples.

Table 2.

Nonparametric Lorentz random dominance test for individual income distribution under different subsamples.

	Gender subsample		Cohort subsample		Parental occupation subsample
	Female	Male	Under40	Over40	Bottom	Middle	Top
1		$\prec FSD$		$\prec FSD$		$\prec FSD$
2	$\succ FSD$		$\succ FSD$		$\succ FSD$		$\prec FSD$
3						$\succ FSD$

Note: Row labels 1–3 correspond to the classification of gender, age group, and occupation, respectively. $\prec FSD$ denotes “first-order predominant.”

3.3. Model specification and endogeneity

3.3.1. Model specification

To investigate the influence of digital technology on income inequality among rural residents, we establish the following model as equation (2):

$$RD={\alpha }_1+{\beta }_{1}digi+{\mathit{\gamma }}_{\mathbf{1}}\mathit{X}+{\epsilon }_1$$ (2)

where $RD$ denotes the individual income inequality index, $digi$ represents digital technology access, and $\mathit{X}$ includes control variables encompassing individual characteristics, household features, and village features. ${\alpha }_1,{\beta }_1,{\mathit{\gamma }}_{\mathbf{1}},{\text{and}\epsilon }_1$ symbolize the intercept term, coefficient of digital technology access, coefficient of control variables, and residuals, respectively, with the residuals assumed to follow ${\epsilon }_1\sim N(0,{\sigma }^2)$.

Since ordinary least squares (OLS) estimation results are sensitive to the influence of sample outliers, we employ quantile regression methods in addition to OLS estimation to account for the effect of outliers on regression outcomes (Koenker and Bassett, 1978) [38]. The specific model is formulated as equation (3):

$$Q_r\left({RD}_i|{digi}_i\right)=\eta +\theta {digi}_i+\mathit{\vartheta }{\mathit{X}}_{\mathit{i}}+{\mu }_i$$ (3)

To assess the mechanism through which digital technology access impacts rural residents' income inequality, we also use a mediation effect model by modifying equation (2) as equations (4), (5):

$$digi={\alpha }_2+{\beta }_{2}mec+{\mathit{\gamma }}_{\mathbf{2}}\mathit{X}+{\epsilon }_2$$ (4)

$$RD={\alpha }_3+{\beta }_{3}digi+{\beta }_{4}mec+{\mathit{\gamma }}_{\mathbf{3}}\mathit{X}+{\epsilon }_3$$ (5)

where $mec$ represents the mechanism variables, including the frequency of internet use, household credit availability, entrepreneurial behavior, and labor mobility.

3.3.2. Endogeneity discussion and Mitigation

The central focus of this study—analyzing the impact of digital technology on income inequality—may be subject to endogeneity issues. First, a bidirectional causal relationship could exist between digital technology and income disparity. The state of income distribution influences individuals' access to digital technologies and potential participation in sharing digital dividends. Typically, higher-income groups have better access to digital technologies. Second, omitted variable bias could stem from the inability to incorporate all factors affecting income disparity in the model specification. Rural residents' degree of income inequality depends on a range of factors such as social institutional environment, cultural background, individual characteristics, household characteristics, and village environment characteristics, which are challenging to capture and measure. Finally, potential measurement errors in the core variables' calculations may exist. Although our study uses a representative sample for calculating the individual income inequality index, potential biases remain.

To address the endogeneity issues, existing literature on the influence of digital technology on income and income disparity has employed the instrumental variable (IV) approach. Based on data availability and previous literature on the digital economy's impact on income, we use the number of base stations at the provincial level as an IV for regression (Yin et al., 2021) [39].

4. Analysis of empirical results

4.1. Baseline regression results

Table 3 presents the estimated impact of digital technology access on income inequality. Models 1 and 2 represent the regression results without and with village variables, respectively. IVs are employed in both models to account for potential endogeneity issues concerning the influence of digital technology on income disparity. Using Model 2 as an example, the OLS results indicate a statistically significant marginal effect of digital technology access on individual income inequality at the 1 % level, which decreases by 3.82 % when an individual obtains access to digital technology. Increased mobile internet accessibility raises the amount of rural internet users due to lower costs and improves the convenience to search for information online.15 This subsequently advances the effective enhancement of rural labor force employment skills and human capital, narrowing the income gap. Consequently, Hypothesis 1 is confirmed.

Table 3.

Basic regression results (dependent variable: RD_K).

Variables	Model 1		Model 2
	OLS	IV	OLS	IV
Mobile	−0.0400***	−0.1750	−0.0382***	−0.6060**
	(0.0084)	(0.2726)	(0.0085)	(0.2497)
Individual Variables	Yes	Yes	Yes	Yes
Family Variables	Yes	Yes	Yes	Yes
Village Variables	No	No	Yes	Yes
Observations	53,385	34,584	39,584	23,502
R-squared	0.2021	0.2023	0.2045	−0.0568
F	373.8	1123	470.0	351.7
idstat		18.11		22.44
widstat		17.92		22.12

Note: *p < 0.1, **p < 0.05, ***p < 0.01. Robust standard errors are in parentheses.

Considering individual characteristics that can influence household income and access to digital technology services such as education and ability, direct OLS regression would underestimate the coefficient of digital technology access in the model; therefore, IVs are required for estimation. Using the IV approach, the Kleibergen–Paap LM statistics in regression tests uniformly reject the null hypothesis that the model is not identifiable at a 1 % significance level. Similarly, the Kleibergen–Paap F statistics in weak IVs tests consistently reject the null hypothesis of weak IVs.

After implementing the IV test and excluding endogeneity issues, digital technology access continues to exhibit a significant negative impact on individual income inequality. Upon obtaining digital technology access, individuals' income inequality is reduced by 60.6 %, confirming the robustness of Hypothesis 1.

4.2. Quantile regression results

Fig. 3, Fig. 4 present the regression results for individual income inequality at the 10th, 25th, 50th, 75th, and 90th percentiles for the entire sample. Digital technology access exhibits a negative impact on individual income inequality at the 10th, 25th, and 50th percentiles, whereas this effect turns positive and relatively minimal at the 75th and 90th percentile levels, passing the significance test at the 1 % level. Using Fig. 3 as an example, the coefficients of the influence of digital technology access on individual income at each percentile are $-$ 0.1395, $-$ 0.0880, and $-$ 0.0134.16 The effect of digital technology access on individual income inequality demonstrates an increasing trend as income inequality progressively rises. This suggests that the impact of digital technology access on income inequality declines as income inequality escalates. Considering subsamples, digital technology access exerts a significant inhibitory effect on income inequality in all cases, showing a notable upward trend concurrent with increasing income inequality (Appendix Figs. A.8–A.21).

Fig. 3.

Quantile regression results (without the village variable).

Fig. 4.

Quantile regression results (with the village variable).

4.3. Heterogeneity analysis

Examining the fundamental characteristics of the core variables reveals that the effect of digital technology access on individual income inequality varies significantly across gender, occupation, and age groups. This section explores the heterogeneous impact of digital technology access on income inequality based on these three dimensions.

First, concerning gender heterogeneity, the results in Table 4 Panel B show that the impact of digital technology access on income inequality is more substantial for men than for women, which is attributable to several factors. Rural women tend to have more leisure time at home compared with their male counterparts, and the widespread adoption of internet technology has spurred tfhe growth of online business models such as microenterprises, self-made media, and live streaming. These models align with women's needs for leisure and work flexibility, enabling them to balance family and professional roles. Additionally, these emerging occupations have inherent advantages for women, increasing their likelihood to choose online self-employment and entrepreneurship, which contributes to improved salaries and benefits and a reduction in the gender wage gap (Qi and Liu, 2020) [40]. Conversely, traditional perceptions that undermine women's access to work, coupled with physiological and psychological characteristics and lower educational levels in rural areas, limit their entry into high-income industries. Consequently, digital technology access has a smaller impact on income inequality for women in contrast to men.

Table 4.

Regression results under subsamples (dependent variable: RD_K).

Panel A Regression results under subsamples (without village variable)
	Gender subsample		Parental occupation subsample			Cohort subsample
Variables	Female	Male	Top	Middle	Bottom	Under40	Over40
Mobile	−0.0198**	−0.0668***	−0.1124***	−0.0356**	−0.0653***	−0.0840***	−0.0517***
	(0.0084)	(0.0126)	(0.0184)	(0.0130)	(0.0085)	(0.0133)	(0.0071)
Individual Variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Family Variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Village Variables	No	No	No	No	No	No	No
Observations	19,514	20,070	2728	38,843	11,814	14,599	24,984
R-squared	0.1285	0.1326	0.1615	0.2112	0.1945	0.1588	0.1612
F	268.2	331.9	127.8	274.7	305.8	92.94	150.5
Panel B Regression results under subsamples (with village variable)
Mobile	−0.0218**	−0.0676***	−0.1284***	−0.0307**	−0.0700***	−0.0816***	−0.0557***
	(0.0079)	(0.0115)	(0.0228)	(0.0129)	(0.0086)	(0.0118)	(0.0073)
Individual Variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Family Variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Village Variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	19,514	20,070	1739	28,907	8938	14,599	24,984
R-squared	0.1448	0.1475	0.1674	0.2176	0.1911	0.1664	0.1796
F	196.4	302.2	90.18	226.2	198.3	110.2	221.7

Note: *p < 0.1, **p < 0.05, ***p < 0.01. Robust standard errors are in parentheses.

Second, regarding occupational heterogeneity, digital technology access has a more significant effect on higher occupational classes' incomes, while the impact on the income gap is smaller among lower occupational classes. This is because individuals employed in the digital economy have generally attained higher education than those in other industries, and digital economy development increases the demand for highly skilled personnel and raises the skill premium (Gong, 2021) [41].

Finally, in terms of cohort heterogeneity, while the use of internet technology is increasingly permeating middle-aged and elderly populations, individuals aged 60 and above remain the primary group of non-internet users. As of December 2021, this age group constituted 39.4 % of China's noninternet user population. Middle-aged and elderly noninternet users experience considerable inconveniences related to travel, consumption, and medical treatment, and cannot fully benefit from the advantages offered by intelligent services.17 The inability or unwillingness to sell agricultural products through online channels significantly affects the incomes of “empty nest” elderly individuals in rural areas. In contrast, younger generations demonstrate greater receptiveness and adaptability to new technologies, effectively using digital technology to increase incomes, which reduces the income gap.

4.4. Robustness tests

We test the robustness of the above estimation results by employing alternative core explanatory variables, dependent variables, and regression methods. First, the results remain consistent with the benchmark empirical findings after substituting computer usage as the core explanatory variable to measure digital technology access. Second, replacing the dependent variable of individual income inequality with individual absolute income and the Yitzhaki index demonstrates that digital technology access effectively raises rural residents' income levels, further supporting the impact of digital technology access on reducing the rural residents' income gap. Finally, employing a two-way fixed-effects model confirms the previous findings, bolstering the reliability of the estimation results in our study.18

5. Mechanism analysis

The preceding discussion and analyses essentially confirmed the suppressive effect of digital technology access on income inequality in rural areas; however, the question remains: How does digital technology access reduce rural income disparity? This section delves further into this issue.

5.1. Mediating effect of frequency of digital technology usage

As new digital infrastructure progressively penetrates rural areas, China achieved “village broadband” coverage across all administrative villages in November 2021. This has historically addressed communication challenges in impoverished areas, continually promoting rural digital transformation, fostering numerous new business models, and enhancing modern agricultural informatization and production capacity. Information technology, epitomized by the internet, provides rural residents with diverse approaches to socialization and information access. As digital technology is increasingly integrated into everyday life, individuals' frequency of internet usage continues to rise. Using the internet for social and informational purposes enables farmers to access various types of information such as employment, education, agricultural industry prices, and sales channels. This promotes the exchange of implicit information, which increases rural residents' employment and educational opportunities, raises incomes, and reduces income inequality.

In addition, the accelerated pace and depth of information dissemination effectively address information asymmetry in rural areas, providing additional employment and educational opportunities and business prospects for agricultural product production and marketing, ultimately reducing income inequality. Columns (4)–(6) in Table 5 present the specific empirical regression results, demonstrating that digital technology access significantly increases residents' frequency of digital use, which subsequently diminishes income inequality,19 confirming Hypothesis 2 of this study.

Table 5.

Mechanism analysis: Mediating effect of frequency of digital technology usage.

Variables	Model 1			Model 2
	(1)	(2)	(3)	(4)	(5)	(6)
	RD_K	Frequency	RD_K	RD_K	Frequency	RD_K
Mobile	−0.0400***	−0.5672***	−0.0590	−0.0382***	−0.5810***	−0.0282
	(0.0084)	(0.1882)	(0.0611)	(0.0085)	(0.2046)	(0.0762)
Frequency			0.0199***			0.0134***
			(0.0022)			(0.0025)
Individual Variables	Yes	Yes	Yes	Yes	Yes	Yes
Family Variables	Yes	Yes	Yes	Yes	Yes	Yes
Village Variables	No	No	No	Yes	Yes	Yes
Observations	53,385	26,319	26,319	39,584	18,588	18,588
R-squared	0.2021	0.1468	0.1463	0.2045	0.0868	0.1502
F	373.8	122.5	390.2	470.0	56.27	148.4

Note: *p < 0.1, **p < 0.05,***p < 0.01. Robust standard errors are in parentheses.

5.2. Mediating effect of credit availability

The demand for credit in China's rural areas faces substantial constraints, leading to insufficient demand from farmers for formal financial sector funds. The inadequate natural demand results from a low degree of rural commercialization and farmers' high subsistence consumption, reducing the transactional demand for funds. The demand deficiency also arises from institutional supply shortages that are primarily characterized by lagging consumer credit services, weakened farmers' demand for consumption funds, reduced farmers' investment demand due to agricultural and rural marketization challenges, difficulty in obtaining loans from formal financial institutions, and crowding-out effects of informal financial organizations (Ma and Lan, 2003) [42]. In addition to rural economic characteristics and institutional disincentives, Han et al. (2007) [43] contend that household incomes and wealth underpin farmers' borrowing repayment capabilities as fundamental determinants of borrowing demand. Moreover, Ye et al. (2004) [44] determined that social capital positively affects loan availability and farmers with higher social capital have better access to loans from formal financial institutions.

The integration of digital technologies into rural infrastructure appears to have mitigated credit constraints for farm households at both information and technology levels. By accessing the funds required for production activities and daily life, farm households' income-generating capacity has been substantially enhanced, reducing income inequality. Table 6 presents the specific empirical regression results. For instance, Model 2 in column (4) demonstrates that digital technology access significantly reduces income inequality for farm households. Column (5) reveals that digital technology access effectively improves credit availability for farm households, and column (6) confirms the mediating role of digital technology access in decreasing the income gap by promoting farm households' credit availability.20 Therefore, Hypothesis 3 is confirmed.

Table 6.

Analysis of mechanisms: Mediating effect of credit availability.

Variables	Model 1			Model 2
	(1)	(2)	(3)	(4)	(5)	(6)
	RD_K	Loan	RD_K	RD_K	Loan	RD_K
Mobile	−0.0400***	−0.0262***	−0.0398***	−0.0382***	−0.0241**	−0.0381***
	(0.0084)	(0.0085)	(0.0084)	(0.0085)	(0.0115)	(0.0084)
Loan			0.0072			0.0019
			(0.0080)			(0.0074)
Individual Variables	Yes	Yes	Yes	Yes	Yes	Yes
Family Variables	Yes	Yes	Yes	Yes	Yes	Yes
Village Variables	No	No	No	Yes	Yes	Yes
Observations	53,385	53,385	53,385	39,584	39,584	39,584
R-squared	0.2021	0.0290	0.2022	0.2045	0.0299	0.2045
F	373.8	84.74	366.2	470.0	111.0	442.6

Note: *p < 0.1, **p < 0.05, ***p < 0.01. Robust standard errors are in parentheses.

5.3. Mediating effect of innovation and entrepreneurship

With the proliferation of digital information technology and the advanced information infrastructure in rural areas, rural households are increasingly using digital technology for entertainment, socializing, acquiring information, and engaging in market transactions, which has consequently fostered entrepreneurship (Mao et al., 2019; Song and He, 2021; Yin et al., 2021) [[45], [46], [47]]. Zhou and Fan (2018) [31] determined that the internet, as an effective communication medium, facilitates information dissemination, enabling access to and exploration of additional business opportunities and reinforcing the demonstration effect of entrepreneurial success. For farmers, entrepreneurship can facilitate the transition from traditional agriculture, promote the growth of nonagricultural industries, accelerate surplus rural labor transfers, and contribute to agricultural modernization, rural industrialization, and urbanization. With farmers as the primary entrepreneurial agents, rural entrepreneurship is emerging as an irreplaceable driving force in rural economic development (Zhang et al., 2015) [48]. Moreover, rural entrepreneurship allows farmers to alleviate poverty and unemployment by providing job opportunities and increasing household income (Bruton et al., 2013) [49]. Accordingly, the government has implemented numerous policy measures to optimize the entrepreneurial environment and encourage farmers' entrepreneurship. Table 7 presents the overall impact of digital technology access on rural residents' entrepreneurial behavior, revealing that digital technology access significantly enhances entrepreneurial behavior and reduces income inequality among rural residents, confirming Hypothesis 4.21

Table 7.

Mechanism analysis: Mediating effect of innovation and entrepreneurship.

Variables	Model 1			Model 2
	(1)	(2)	(3)	(1)	(2)	(3)
	RD_K	Startup	RD_K	RD_K	Startup	RD_K
Mobile	−0.0400***	0.0192***	−0.0414***	−0.0382***	0.0110**	−0.0387***
	(0.0084)	(0.0048)	(0.0086)	(0.0085)	(0.0052)	(0.0086)
Startup			0.0745***			0.0487***
			(0.0123)			(0.0101)
Individual Variables	Yes	Yes	Yes	Yes	Yes	Yes
Family Variables	Yes	Yes	Yes	Yes	Yes	Yes
Village Variables	No	No	No	Yes	Yes	Yes
Observations	53,385	53,385	53,385	39,584	39,584	39,584
R-squared	0.2021	0.0141	0.2068	0.2045	0.0121	0.2065
F	373.8	72.07	350.6	470.0	69.64	444.0

Note: *p < 0.1, **p < 0.05, ***p < 0.01. Robust standard errors are in parentheses.

5.4. Mediating effect of labor mobility

As the digital economy advances, new models and business approaches arising from this progress provide ample opportunities for novel employment alternatives. Emerging employment patterns such as the gig economy, nonstandard employment, and freelance work have emerged, offering a vast array of job prospects for the migrating population, including errand services, delivery workers, and self-made media operations. Digital economy development generates various new employment patterns for the migrating population such as platform employment and self-employment, accelerating urban–rural population mobility (Qi and Liu, 2021) [50].

The rapid growth of the digital economy and the emergence of innovative models and industries generate additional opportunities for new jobs, effectively driving employment and promoting cross-sector and cross-regional labor mobility. According to the 2019 White Paper on Digital Economy Development and Employment in China, 191 million new jobs were generated within China's digital economy in 2018, comprising 24.6 % of the total employment for that year. Despite a 0.07 % annual decline in total national employment, digital economy jobs experienced a robust 11.5 % annual growth.22 The digital transformation of traditional industries has absorbed more jobs, rapidly expanding and becoming an essential channel for stable employment in China.

Specifically, digital technology development and the consequent emergence of digital finance can reduce borrowing costs and increase information transparency, alleviating financing constraints and lowering enterprises' operating and intermediation costs. This fosters the growth of small, medium, and micro enterprises (Shahrokhi, 2008) [51], consequently leading to increased employment opportunities and attracting an influx of rural labor (Ma and Hu, 2022) [52]. Meanwhile, the diffusion and use of digital technologies in rural areas can release surplus labor, diversify opportunities, and increase flexibility in employment options, fostering greater mobility between regions, sectors, and industries. Table 7 reveals that digital technology access can significantly promote rural labor outflow and reduce income inequality among rural residents, confirming Hypothesis 5.23

6. Conclusions, policy implications, and limitations

6.1. Conclusion and policy implications

The transformation of production methods in rural areas from digital technology access and digital economy development has inevitably caused corresponding changes in income distribution among rural residents, impacting the income gap.

This study examines the effects of digital technology access, labor market behavior, and income inequality in rural China. Our theoretical analysis reveals that digital technology access influences rural labor force practices and decisions, subsequently affecting the rural income distribution structure. The study determines that access to digital technology effectively increases farmers' frequency of digital technology use, enhances credit availability, fosters innovation and entrepreneurship, and promotes sectoral mobility, which raises income levels and narrows income disparities in rural areas. Employing CFPS microdata from 2010 to 2018 yields the following relevant conclusions.

First, digital technology access significantly reduces individual income inequality by 60.6 % after accounting for endogeneity and employing alternative variables in the regression. Second, the impact of digital technology usage on income inequality exhibits heterogeneous effects across different income inequality levels, gender, occupational classes, and age cohorts. Third, our mechanism analyses demonstrate that digital technology access effectively improves various rural labor force practices, increasing digital technology use frequency, reducing credit costs, enhancing rural residents' entrepreneurial behavior, and promoting rural labor mobility, ultimately increasing rural residents' incomes and reducing income inequality.

Based on these findings, we propose three recommendations to reduce the income gap among rural residents.

• (1)Accelerate rural information infrastructure optimization and upgrading by accelerating rural broadband construction plans and ensuring the completion of the last mile of rural broadband connectivity. Implement the construction of 4G base stations in rural areas to cover blind spots, and promote the expansion of 5G and gigabit fiber networks in suitable rural areas. Continuously advance the same network and speed policy for urban and rural regions, optimizing and improving the quality of rural broadband networks as outlined in the Digital Rural Development Action Plan (2022–2025). Ensuring that rural residents can access digital technology and participate in the digital economy will raise overall income levels and narrow income disparities.
• (2)Improve rural residents' digital information application capabilities by providing comprehensive training on the essential aspects of the digital economy and enhancing digital and financial literacy. Organize remote online public service tutorials to cultivate farmers' understanding of the internet, improve their ability to access and identify accurate information, and enhance their capability to employ the internet for learning, starting businesses, and engaging in e-commerce activities. As a result, rural residents' information acquisition costs are reduced, and farmers' income-generating channels and capabilities will be enhanced, narrowing the rural income gap.
• (3)Accelerate the construction of an inclusive digital financial service system for rural areas and enhance the availability of credit for rural residents. The integration of digital technology into rural infrastructure can effectively alleviate farmers' credit constraints from informational and technological perspectives. By obtaining the necessary funds for production activities and daily living, rural residents' income-generating capabilities will significantly improve, reducing income inequality.

6.2. Limitations and future recommendations

Despite the insights provided by this study, some limitations should be acknowledged. First, while the Kakwani index (1984) is a relative income ranking index that measures individual income inequality, as the dependent variable, it does not adequately account for intragroup income inequality. Second, the core independent variable, digital technology access, is a broad concept; however, due to the constraints of the CFPS questionnaire structure, we only considered internet and cell phone access. Third, while four different mechanisms for the impact of digital technology access on income inequality are examined, additional channels may warrant exploration. Future research on the impact of digital technology access on income inequality could benefit from using city- and county-level data, which would allow for the inclusion of more relevant variables related to digital technology access and income inequality and also enable comparative analyses at the macroaggregate level.

Funding information

This work was supported by the Chongqing Technology and Business University Foundation (950321064).

Data availability statement

Has data associated with your study been deposited into a publicly available repository? No.

Additional information

No additional information is available for this paper.

CRediT authorship contribution statement

Jie Zhang: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Resources, Methodology, Funding acquisition, Formal analysis, Data curation, Conceptualization. Mengna Li: Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix.

A1. Key Variable Characteristics

Fig. A.1.

Cumulative distribution of inequality under digital technology access (Male)

Fig. A.2.

Cumulative distribution of inequality under digital technology access (Female)

Fig. A.3.

Cumulative distribution of inequality under digital technology access (over 40 years old)

Fig. A.4.

Cumulative distribution of inequality under digital technology access (under 40 years old)

Fig. A.5.

Cumulative distribution of inequality under digital technology access (top occupations)

Fig. A.6.

Cumulative distribution of inequality under digital technology access (middle occupations)

Fig. A.7.

Cumulative distribution of inequality under digital technology access (bottom occupations)

A2. Quantile Regression Results Under Each Subsample

Fig. A.8.

Regression results for female subsample quartiles (without village variable)

Fig. A.9.

Regression results for female subsample quartiles (with village variable)

Fig. A.10.

Regression results for male subsample quartiles (without village variable)

Fig. A.11.

Regression results for male subsample quartiles (with village variable)

Fig. A.12.

Regression results for quartiles of the subsample under 40 years old (without village variable)

Fig. A.13.

Regression results for quartiles of the subsample under 40 years old (with village variable)

Fig. A.14.

Regression results for subsample quartiles over 40 years old (without village variable)

Fig. A.15.

Regression results for subsample quartiles over 40 years old (with village variable)

Fig. A.16.

Regression results for higher occupational subsample quartiles (without village variables)

Fig. A.17.

Regression results for higher occupational subsample quartiles (with village variable)

Fig. A.18.

Regression results for the medium occupation subsample quartiles (without village variable)

Fig. A.19.

Regression results for the medium occupation subsample quartiles (with village variable)

Fig. A.20.

Regression results for the lower occupation subsample quartile (without village variables)

Fig. A.21.

Regression results for lower occupation subsample quartiles (with village variables)

A3. Robustness Test

A3.1 Independent Variables: PC

Table A1.

Basic regression results (dependent variable: RD)

	Model 1		Model 2
	OLS	FE	OLS	FE
PC	−0.0906∗∗∗	−0.0126∗∗	−0.0815∗∗∗	−0.0116
	(0.0117)	(0.0064)	(0.0116)	(0.0081)
Individual variables	Yes	Yes	Yes	Yes
Family Variables	Yes	Yes	Yes	Yes
Village Variables	No	No	Yes	Yes
Observations	53,384	53,384	39,583	39,583
R-squared	0.2084	0.1916	0.2088	0.1993
F	362.3	124.8	366.4	.

Note: *p < 0.1, ∗∗p < 0.05, ∗∗∗p < 0.01. Robust standard errors are in parentheses.

A3.2 Dependent Variable: Log of income and RD_Y

Table A2.

Basic regression results

Dependent variable	Model 1				Model 2
	Log income		RD_Y		Log income		RD_Y
	OLS	FE	OLS	FE	OLS	FE	OLS	FE
PC	0.8242∗∗∗	0.2296∗∗∗	−0.2741∗∗∗	−0.0283∗	0.7977∗∗∗	0.2783∗∗	−0.2396∗∗∗	−0.0278
	(0.1195)	(0.0855)	(0.0385)	(0.0162)	(0.1335)	(0.1106)	(0.0339)	(0.0207)
Individual variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Family Variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Village Variables	No	No	No	No	Yes	Yes	Yes	Yes
Observations	53,384	53,384	53,384	53,384	39,583	39,583	39,583	39,583
R-squared	0.1474	0.2902	0.1470	0.3697	0.1564	0.3149	0.1595	0.3610
F	697.2	212.4	405.4	369.0	1407	.	318.1	.

Note: ∗p < 0.1., ∗∗p < 0.05, ∗∗∗p < 0.01. Robust standard errors are in parentheses.

Table A3.

Basic regression results

Dependent variable	Model 1				Model 2
	Log income		RD_Y		Log income		RD_Y
	OLS	FE	OLS	FE	OLS	FE	OLS	FE
Mobile	0.1119	0.0972	0.0070	−0.0661∗∗∗	0.0515	0.1607∗	−0.0123	−0.0684∗∗∗
	(0.1570)	(0.0786)	(0.0151)	(0.0095)	(0.1596)	(0.0897)	(0.0163)	(0.0109)
Individual variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Family Variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Village Variables	No	No	No	No	Yes	Yes	Yes	Yes
Observations	53,385	53,385	53,385	53,385	39,584	39,584	39,584	39,584
R-squared	0.1441	0.2900	0.1338	0.3703	0.1535	0.3148	0.1497	0.3617
F	759.0	212.2	314.9	373.8	1095	.	289.1	.

Note: ∗p < 0.1, **p < 0.05, ∗∗∗p < 0.01. Robust standard errors are in parentheses.

A4. Results of Mediating Effects Test

Table A4.

Results of mediating effects test based on Table 5

	Model 1				Model 2
	Coef	Std Err	Z	P>|Z|	Coef	Std Err	Z	P>|Z|
Sobel	−0.0113	0.00469	−2.408	0.0160	−0.00781	0.00370	−2.112	0.0347
Goodman-1 (Aroian)	−0.0113	0.00470	−2.403	0.0163	−0.00781	0.00372	−2.098	0.0359
Goodman-2	−0.0113	0.00468	−2.414	0.0158	−0.00781	0.00367	−2.126	0.0335
a coefficient	−0.567	0.232	−2.442	0.0146	−0.581	0.266	−2.183	0.0290
b coefficient	0.0199	0.00137	14.55	0	0.0134	0.00161	8.357	0
Indirect effect	−0.0113	0.00469	−2.408	0.0160	−0.00781	0.00370	−2.112	0.0347
Direct effect	−0.0590	0.0516	−1.143	0.253	−0.0282	0.0583	−0.483	0.629
Total effect	−0.0703	0.0518	−1.357	0.175	−0.0360	0.0584	−0.616	0.538

Table A5.

Results of mediating effects test based on Table 6

	Model 1				Model 2
	Coef	Std Err	Z	P>|Z|	Coef	Std Err	Z	P>|Z|
Sobel	−0.000189	8.74e-05	−2.166	0.0303	−4.52e-05	7.87e-05	−0.575	0.565
Goodman-1 (Aroian)	−0.000189	8.89e-05	−2.129	0.0332	−4.52e-05	8.11e-05	−0.557	0.577
Goodman-2	−0.000189	8.59e-05	−2.205	0.0275	−4.52e-05	7.61e-05	−0.594	0.552
a coefficient	−0.0262	0.00549	−4.777	1.80e-06	−0.0241	0.00615	−3.910	9.20e-05
b coefficient	0.00723	0.00297	2.430	0.0151	0.00188	0.00324	0.581	0.561
Indirect effect	−0.000189	8.70e-05	−2.166	0.0303	−4.50e-05	7.90e-05	−0.575	0.565
Direct effect	−0.0398	0.00377	−10.55	0	−0.0381	0.00396	−9.623	0
Total effect	−0.0400	0.00377	−10.60	0	−0.0382	0.00396	−9.637	0

Table A6.

Results of mediating effects test based on Table 7

	Model 1				Model 2
	Coef	Std Err	Z	P>|Z|	Coef	Std Err	Z	P>|Z|
Sobel	0.00143	0.000300	4.759	1.95e-06	0.000536	0.000207	2.585	0.00973
Goodman-1(Aroian)	0.00143	0.000301	4.752	2.02e-06	0.000536	0.000208	2.574	0.0101
Goodman-2	0.00143	0.000300	4.766	1.88e-06	0.000536	0.000206	2.597	0.00939
a coefficient	0.0192	0.00388	4.939	7.90e-07	0.0110	0.00412	2.675	0.00748
b coefficient	0.0745	0.00419	17.79	0	0.0487	0.00483	10.09	0
Indirect effect	0.00143	0.000300	4.759	1.90e-06	0.000536	0.000207	2.585	0.00973
Direct effect	−0.0414	0.00376	−11.01	0	−0.0387	0.00396	−9.784	0
Total effect	−0.0400	0.00377	−10.60	0	−0.0382	0.00396	−9.637	0

Table A7.

Results of mediating effects test based on Table 8

	Model 1				Model 2
	Coef	Std Err	Z	P>|Z|	Coef	Std Err	Z	P>|Z|
Sobel	−0.00191	0.000279	−6.840	0	−0.00392	0.000395	−9.926	0
Goodman-1 (Aroian)	−0.00191	0.000280	−6.829	0	−0.00392	0.000396	−9.914	0
Goodman-2	−0.00191	0.000279	−6.852	0	−0.00392	0.000395	−9.939	0
a coefficient	0.0961	0.00627	15.33	0	0.0996	0.00684	14.56	0
b coefficient	−0.0199	0.00260	−7.644	0	−0.0394	0.00290	−13.56	0
Indirect effect	−0.00191	0.000279	−6.840	0	−0.00392	0.000395	−9.926	0
Direct effect	−0.0380	0.00378	−10.08	0	−0.0342	0.00396	−8.644	0
Total effect	−0.0400	0.00377	−10.60	0	−0.0382	0.00396	−9.638	0

Table 8.

Mechanism analysis: Mediating effect of labor mobility.

Variables	Model 1			Model 2
	(1)	(2)	(3)	(1)	(2)	(3)
	RD_K	Mobility	RD_K	RD_K	Mobility	RD_K
Mobile	−0.0400***	0.0961***	−0.0380***	−0.0382***	0.0996***	−0.0342***
	(0.0084)	(0.0167)	(0.0088)	(0.0085)	(0.0178)	(0.0083)
Mobility			−0.0199			−0.0394***
			(0.0135)			(0.0110)
Individual Variables	Yes	Yes	Yes	Yes	Yes	Yes
Family Variables	Yes	Yes	Yes	Yes	Yes	Yes
Village Variables	No	No	No	Yes	Yes	Yes
Observations	53,384	53,384	53,384	39,583	39,583	39,583
R-squared	0.2021	0.0661	0.2030	0.2045	0.0773	0.2082
F	374.3	91.47	396.6	474.5	90.10	701.2

Note: *p < 0.1, **p < 0.05, ***p < 0.01. Robust standard errors are in parentheses.