Can Internet Use Increase Rural Women’s Income? Evidence from Underdeveloped Areas of China

Abstract

Although the widespread use of the Internet in rural areas provides new opportunities for economic growth, whether and how rural women benefit from it remains a question. Guided by Sustainable Development Goal (SDG) 5 and its targets, this study examines the impact of Internet use on rural women’s income by using survey data from 1384 rural households in underdeveloped areas of China. The results indicate that Internet use can significantly increase rural women’s income in underdeveloped areas. Additionally, the income effects of Internet use are heterogeneous across the different purposes of Internet use, income levels, individual characteristics, and family characteristics. Further analysis reveals that increasing labor input and enhancing capital endowment are two important channels through which Internet use increases rural women’s income. The results offer further empirical support for policymakers to utilize the Internet to increase rural women’s income and contribute to poverty alleviation in underdeveloped areas.

1. Introduction

Gender equality is a critical issue of global concern, which is related to economic growth, social fairness, and even the prosperity or decline of nations [1,2,3]. Especially in underdeveloped countries and regions, achieving gender equality and empowering women not only enhances their well-being but also holds significant value in attaining a range of economic and social development goals, such as eliminating poverty, protecting women’s rights and interests, improving children’s nutrition and education, and ensuring food security [4,5,6]. The United Nations has designated gender equality as the fifth Sustainable Development Goal (SDG) and incorporated it into the 2030 Agenda for Sustainable Development, emphasizing its significant role in achieving sustainable development [7].

Compared with men, women’s lower income is widely recognized as a vital cause of gender inequality [8,9]. Owing to traditional gender norms and physiological characteristics, women tend to be the primary caregivers in the family [10] and they also experience unfair resource allocation within the household [11]. These reasons lead to women’s lack of sufficient opportunities and capabilities to participate in the labor market [12], making it more difficult for them to obtain ideal labor income. There is even a phenomenon called the “feminization of poverty” in some low-income and middle-income countries [13]. Therefore, providing women with employment opportunities and increasing their income, thereby empowering them economically, is a crucial strategy for addressing gender inequality [9].

Enhancing the use of enabling technology, particularly information and communications technology, is a crucial target for advancing action toward SDG 5 [7]. The Internet holds great promise for increasing the income of poor women. For one thing, the development of the Internet has changed traditional working arrangements and created jobs suitable for women [14,15]. The Internet has lessened restrictions on the time and location of work, increasing employment flexibility [16], and online work arrangements provide women with the possibility of balancing work and family [17]. Additionally, Internet use affects women’s resource endowments. A few studies have demonstrated that Internet use can significantly improve women’s cognitive and non-cognitive abilities, social capital endowment, labor productivity, and information acquisition, thereby increasing their income [18,19,20]; however, these studies rarely focus on rural women in poor areas. Most of the existing research tends to focus on urban or general rural areas [15,17,21], overlooking the unique challenges faced by rural women in underdeveloped areas. While the Internet has become more accessible in these areas, there is a lack of comprehensive studies examining whether Internet use increases rural women’s income.

The reality of rural China provides an ideal research context for studying how Internet use affects the income of poor rural women. China’s traditional gender norm of “the man goes out to work while the woman looks after the family” has resulted in rural women being the main caregivers in the family and lacking sufficient opportunities to participate in economic activities outside the family. The average time that rural women in China spend on housework every day is three times that of rural men (data source: National Time Utilization Survey Bulletin (2018)); only 28.9% of rural women take part in off-farm work (data source: the fourth issue of Chinese women’s social status survey (2020)); and in the past decade, the number of rural women working outside their hometowns was only 53.8% of that of rural men (data source: Chinese migrant worker monitoring survey report (2014–2023)). Meanwhile, China is one of the world’s fastest-growing countries in terms of Internet development. In recent years, the country has invested heavily in expanding Internet access to rural areas, seeing it as a crucial tool for poverty alleviation. By the end of 2020, optical fibers were available in 98% of China’s poor villages. The e-commerce demonstration policy for rural areas has fully covered 832 nationally defined poverty counties, and the total online retail sales in these counties reached CNY 301.45 billion [22]. By the end of 2023, the Internet penetration rate in rural areas of China had reached 66.5%, with 326 million rural Internet users.

Numerous studies focusing on China have shown that the expansion of the Internet has significantly promoted economic growth in rural areas. These studies have found that Internet use can increase agricultural productivity [23,24], expand sales channels for agricultural products [25], improve financial market accessibility [26], and promote off-farm employment among rural residents [27,28]. Although many studies have emphasized the positive impact of Internet use on rural economic development [29,30,31], few have incorporated gender as a core variable for investigation. The lack of a gender perspective may lead to biases in the current understanding of the economic impact of the Internet. The gender digital divide may result in women facing more obstacles when engaging in economic activities on the Internet [32,33]. Although the Internet can significantly increase the income of rural households [30,34,35], existing studies have not investigated the benefits of this income-enhancing effect on rural women in detail. Furthermore, most previous research on Internet use has used the binary variables of “yes/no” [23,29,31], overlooking the complexity and diversity of the extent and purpose of Internet use. With the diversification and popularization of Internet applications, discussing Internet use with binary variables makes it difficult to fully explore the impact of Internet use on users’ economic activity.

This study explores the relationship between Internet use and rural women’s income in underdeveloped areas. We conducted a household survey in the remote areas of Western China in 2021 to obtain the research data. On this basis, we empirically investigate the connection between Internet use and rural women’s income. We also explore the mechanism from the perspectives of adjusting labor inputs and enhancing capital endowment.

Compared to the existing literature, our contribution to the effect of Internet use and rural women’s income is threefold. First, this study focuses on rural women in underdeveloped areas, providing valuable insights into women’s empowerment in low-income and middle-income countries, thereby expanding the research subjects in the existing literature. Second, previous studies on Internet use have not specifically examined the purposes and degrees of Internet use in detail. This study comprehensively investigates the impact of Internet use behaviors on the income of rural women, thereby enriching the limited literature. Third, this study differentiates between income sources and reveals the pathways through which the Internet influences income growth by exploring the differences in various income sources.

The rest of this paper is structured as follows. The theoretical analysis and research hypotheses are presented in Section 2. Section 3 explains the data and model specifications. The estimation results are presented and discussed in Section 4. Conclusions and policy implications are presented in Section 5.

2. Theoretical Analysis and Research Hypotheses

2.1. The Impact of Internet Use on Rural Women’s Income

Digital technologies hold significant potential in advancing sustainable development [36]. The impact of the Internet on rural women’s income has become a rising topic. First, as an effective production tool, the Internet directly contributes to increasing rural women’s income. It not only effectively weakens the rigid constraints of traditional working arrangements on time and geographical space [17] but also promotes a shift in the nature of work from relying on physical labor to focusing more on intellectual labor. These changes in employment are more consistent with female characteristics. The Internet has unleashed rural women’s work potential. Specifically, Internet use can greatly enhance rural women’s participation in off-farm employment. Applying the Internet to agricultural production can improve rural women’s production efficiency [37]. Second, the Internet helps rural women overcome information barriers and indirectly increases their income. On the one hand, abundant learning resources on the Internet create favorable conditions for the re-education of rural women, and they can reduce the cost of acquiring knowledge and improve their human capital through online learning activities [38]; on the other hand, social media can broaden their social networks. Rural women’s social interactions are primarily centered around families or their villages but the Internet can extend their scope of interactions [21]. Rural women can use online social platforms to maintain interpersonal relationships with more people. High-quality human and social capital can help rural women gain a competitive edge, indirectly increasing their income. The following research hypothesis is proposed:

Hypothesis 1.

Internet use contributes to increasing the income of rural women in underdeveloped areas.

2.2. The Mechanism of Internet Use Increasing Rural Women’s Income

The role and potential of emerging technologies in promoting gender equality are widely recognized. Di Vaio et al. (2023) conducted a bibliometric analysis of 140 articles and reports, demonstrating that emerging technologies represented by blockchain contribute to gender equality and inclusion processes [39]. Consistent with the specific target of SDG 5, Internet use can increase the income of rural women in underdeveloped areas by adjusting labor inputs and enhancing capital endowment. Labor inputs encompass employment participation and working hours, while capital endowment includes social capital and human capital. Figure 1 shows the pathways through which Internet use enhances the income of rural women in underdeveloped areas.

Internet use can affect rural women’s income by adjusting their labor inputs, with the impact primarily from two sources. First, the Internet creates a non-discriminatory workplace environment, ensuring fairness among different positions, which is a key step toward achieving gender equality [40]. The widespread use of the Internet has changed traditional employment patterns and increased rural women’s participation in the labor market [14]. The widespread use of Internet technology has created new jobs and occupations, such as short video bloggers, rural e-commerce operators, and online customer service representatives. These jobs offer flexibility with hours and location requirements, allowing rural women to balance work and family responsibilities, thus increasing opportunities for rural women to participate in off-farm employment and entrepreneurship. In addition, the Internet provides an effective platform for information dissemination, lowering search costs and enhancing the efficiency of labor market information searches [41]. The Internet breaks down information barriers for rural women in remote mountainous areas during their job search. It provides abundant employment information and significantly increases off-farm employment opportunities and incomes.

Second, the Internet offers flexible working arrangements that overcome the time and location constraints of traditional work. This flexibility allows women to balance their families and work [42]. However, when the boundary between work and non-work is broken, flexible employment relationships may extend working hours. In that case, rural women may use more fragmented time for off-farm work, thereby increasing their off-farm income. However, off-farm employment may reduce the time spent on agricultural production, potentially reducing rural women’s agricultural income. The following research hypothesis is proposed:

Hypothesis 2.

Internet use increases rural women’s income by adjusting labor inputs.

Internet use can indirectly increase rural women’s incomes in underdeveloped areas by enhancing their capital endowment. The influence is twofold: First, the Internet can enhance rural women’s social capital, thereby increasing their income. Rural women’s social interactions are primarily centered around families or their villages [21]. The Internet breaks the traditional social circles bound by geography and kinship, expanding the social interaction range of rural women [43]. Compared to traditional communication tools, online communication has the advantages of timely feedback efficiency and wide coverage, which help rural women build up more network connections with others. An individual’s social network represents social capital because members of social networks are an important avenue for acquiring resources [44]. Rural women with high-quality social capital are more likely to access various resources, which form the foundation for success. Existing research generally acknowledges that social capital holds significant value in increasing income and alleviating the economic vulnerability faced by impoverished farmers [21].

Second, the Internet is an important tool for enhancing human capital. The Internet can broaden learning channels and accelerate knowledge dissemination, thereby compensating for the shortcomings of the education system in underdeveloped areas and promoting the accumulation of human capital among rural women. Rural women can use online resources to improve their knowledge and skills, which can offset the negative impact of low education and skills on employment, thereby increasing their labor productivity [45]. Government-led online learning and e-commerce sales training have also played an essential role in reducing poverty among rural women in underdeveloped areas [46]. To some extent, the ability to use the Internet represents the level of human capital. When individuals demonstrate high Internet application and learning skills, they signal to employers that these women are highly productive and worthy of higher wages [47]. The following research hypothesis is proposed:

Hypothesis 3.

Internet use increases rural women’s income by enhancing capital endowment.

3. Research Design

3.1. Data Sources

The data used in this study were obtained from a survey conducted between July and August 2021 in rural China. The survey was designed to present situations of the population, economy, society, and women’s empowerment in the underdeveloped areas of China. The food security status reflects the living standards of residents in underdeveloped areas. Nie et al. (2010) [48] classified 592 national poverty-stricken counties in China into food-secure, relatively food-secure, and food-insecure from the perspective of food security. Considering the data availability and practical factors, we selected 7 sample counties from four provinces out of 271 food-insecure counties. These seven counties are Pan County and Zheng’an County in Guizhou Province, Zhen’an County and Luonan County in Shaanxi Province, Wuding County and Huize County in Yunnan Province, and Qingshui County in Gansu Province. These sample counties are all located in the mountainous regions of Western China, where natural disasters occur frequently, infrastructure construction is weak, and economic development lags. In these sample counties, we employed a two-stage sampling method to select the household samples. In the first stage, we utilized a probability proportional to size sampling to select the sample villages. We extracted the top 19 villages by population size from each county, with Qingshui County selecting 16 sample villages after considering factors such as willingness to cooperate and data availability, resulting in 130 sample villages. In the second stage, we employed systematic random sampling to select the household samples. We randomly selected 12 households from each sample village, resulting in a total of 1560 sample households.

The survey employed two questionnaires, one targeting farm households and the other targeting villages. During the household survey, we conducted face-to-face interviews with the household members using a structured questionnaire. Certain sections of the questionnaire were answered by female household heads. The household questionnaire focused on the demographic characteristics of household members (e.g., gender, age, and health status), employment and the off-farm income of household members, agricultural production and sales, household assets, consumption, and women’s decision-making power. We also inquired about Internet use at the household level and questioned every household member about their Internet use behavior separately. In the village survey, we used a structured questionnaire and conducted face-to-face interviews with village leaders. The village questionnaire fully accounted for the complexity of the rural context, with the main contents including the demographic and economic conditions of the village, public services, infrastructure conditions, and the basic rural governance status.

It should be noted that this study focuses on the effect of Internet use on rural women’s income, particularly the economic situation of hostesses in underdeveloped rural areas. Consequently, this study focuses on the hostesses of each household. After excluding samples with missing key variables and those without a hostess in the household, 1384 valid samples were obtained.

3.2. Variables

3.2.1. Explained Variables

The explained variable is income. We focused on labor income, which is directly related to an individual’s employment and may be influenced by Internet use. Specifically, labor income encompasses rural women’s wage income and business income. Wage income is obtained through hired labor, including a stable wage income and income from odd jobs. Business income is the net income obtained after deducting production and operating costs from sales revenue, which includes non-agricultural business income and agricultural business income.

3.2.2. Explanatory Variables

The main explanatory variable is Internet use. We asked the hostess of the household, “Do you use a mobile phone or computer to access the Internet?” If the answer was “yes”, the variable was assigned a value of 1; otherwise, it was 0. With the robustness test, this study examined the impact of the degree of Internet use, which was measured by the total count derived from the multiple-choice question in the questionnaire, “What are the main activities you engage in online?” The values range from 0 to 11, with larger values indicating a greater degree of Internet use. In the heterogeneity analysis, this study analyzed the impact of different purposes of Internet use, including online socialization, online entertainment, online business, and online learning. All purpose-related variables are binary variables.

3.2.3. Control Variables

The control variables were drawn from related studies on Internet use [29,30,31], household income [25,35,49], and women’s empowerment [50,51,52]. We included control variables at multiple levels: first, the individual characteristics include age, education, health, marital status, and party membership; second, family characteristics encompass household size, proportion of children, farm size, distance to the market, and poverty status; third, the village characteristics consist of the regional economy and distance from the county government.

Table 1. Variable definitions and descriptive statistics.

Variable	Definition	Mean	Std. Dev.	Min	Max
Explained Variables
Labor income	Total labor income (CNY 10,000)	1.053	1.977	−1.685	19.343
Wage income	Annual income from hired labor (CNY 10,000)	0.450	1.142	0	10
Business income	Annual net income from agricultural and non-agricultural business (CNY 10,000)	0.603	1.566	−1.685	12.458
Explanatory Variables
Internet use	Whether the hostess uses the Internet: yes = 1, no = 0	0.548	0.498	0	1
Degree of Internet use	Total number of participations in 11 major Internet activities	1.979	2.152	0	8
Online socialization	Whether the hostess uses the Internet for social activities: yes = 1, no = 0	0.516	0.500	0	1
Online entertainment	Whether the hostess uses the Internet for entertainment activities: yes = 1, no = 0	0.455	0.498	0	1
Online business	Whether the hostess uses the Internet for business activities: yes = 1, no = 0	0.393	0.489	0	1
Online learning	Whether the hostess uses the Internet for learning activities: yes = 1, no = 0	0.201	0.401	0	1
Control Variables
Age	Age of hostess (years)	53.367	11.103	23	86
Education	Schooling years of hostess (years)	4.151	3.872	0	16
Health	Health status of hostess: 0 = healthy; 1 = having a chronic disease; 2 = completely incapacitated due to illness	0.529	0.712	0	2
Marital status	Whether married: yes = 1, no = 0	0.947	0.225	0	1
Party membership	Whether the hostess is a member of CPC: yes = 1, no = 0	0.041	0.198	0	1
Household size	Number of family members	4.926	1.753	1	14
Proportion of children	Proportion of family members aged 0–16 years (%)	0.180	0.170	0	0.714
Farm size	Total cultivated area in 2020 (mu)	6.700	7.574	0	100
Distance to market	Distance from the respondent’s home to the nearest market (km)	6.250	8.155	0	200
Poor household	Whether the household has ever been registered a poor household: yes = 1, no = 0	0.348	0.476	0	1
Regional economy	Annual per capita income of the village where the respondent is located (CNY 10,000)	0.871	0.254	0.100	1.520
Distance to the county	Distance from the respondent’s village to the county government (km)	38.004	28.900	0	126
Mechanism Variables
Employment	Whether the hostess participates in employment: yes = 1, no = 0	0.823	0.382	0	1
Agricultural employment	Whether the hostess participates in agricultural employment: yes = 1, no = 0	0.595	0.491	0	1
Off-farm employment	Whether the hostess participates in off-farm employment: yes = 1, no = 0	0.346	0.476	0	1
Agricultural working hours	Average daily working hours in agriculture (hour)	3.619	4.401	0	16
Off-farm working hours	Average daily working hours in off-farm employment (h)	1.879	3.264	0	16
Social capital	Whether the hostess receives skills training: yes = 1, no = 0	0.173	0.379	0	1
Human capital	Family’s annual expenditure on social relationships (CNY 10,000)	0.414	0.594	0	5

presents the definitions of the variables.

3.2.4. Mechanism Variables

Based on the previous analysis, the mechanism variables selected in this paper include employment participation, working hours, social capital, and human capital. Specifically, “employment participation” refers to employment status, off-farm employment, or agricultural employment, while “working hours” encompasses the time spent on farming and off-farm work. In rural China, characterized by a strong sense of community and a close-knit social structure, interpersonal interactions are often guided by the social norm of reciprocity. In line with the approach taken by Ma and Yang (2011) [53], we use household expenditure to measure rural women’s social capital. According to Becker’s (1962) [54] analysis of human capital, on-the-job training represents the most critical variable associated with human capital investment theory; therefore, we employ skills training as a proxy variable for human capital.

3.3. Model Setting

We aim to estimate the potential impact of Internet use on the income of rural women in underdeveloped areas. In regression analysis, with income as the outcome variable, the Ordinary Least Squares (OLS) model is widely used [9,28,55]. The OLS model can minimize the squared errors between the observed values and the predicted values. It outputs regression coefficients that are used to predict and explain the relationships between variables. Based on the OLS model, this paper constructs the following baseline regression model:

$$IHS\left({income}_i\right)={\beta }_0+{\beta }_1{internet}_i+{\beta }_2{\mathbf{X}}_{\mathbf{i}}+{\phi }_i+{\mu }_i$$

(1)

where $i$ represents a female individual; ${income}_i$ represents the income variable of rural women, including labor income, wage income, and business income; ${internet}_i$ represents the core explanatory variable, i.e., rural women’s Internet use; ${\mathbf{X}}_{\mathbf{i}}$ is a vector of individual-level, household-level, and village-level control variables; ${\phi }_i$ denotes the province dummy variable of the region where it is located, which is used to control for the effect of regional characteristics on women’s income; ${\mu }_i$ is a random error term.

As mentioned above, we used three outcome variables (labor income, wage income, and business income) and performed separate regression statistics for each variable. We are interested in the estimated result of ${\beta }_1$. After controlling for other factors included in vector ${\mathbf{X}}_{\mathbf{i}}$, a significantly positive estimate indicates that Internet use is positively correlated with income. We performed linear estimation for all three models and calculated robust standard errors clustered at the village level.

To reduce the problem of heteroskedasticity and to bring the data closer to a normal distribution, this study applied a log transformation to the income variables. As some business income values are zero or negative, taking the natural logarithm is not feasible. To preserve the integrity of the sample, we followed the approach of Chari et al. (2021) [56] by applying the inverse hyperbolic sine (IHS) function to the income variables. Equation (2) expresses the inverse hyperbolic sine function.

$$IHS\left({income}_i\right)=\ln \left({income}_i+\sqrt{1+{{income}_i}^2}\right)$$

(2)

3.4. Descriptive Statistics

Table 1 presents the definitions and descriptive statistics of the key variables. Regarding income, the average labor income of the hostesses in the sample is CNY 10.53-thousand, with wage income accounting for CNY 4.50-thousand and business income accounting for CNY 6.03-thousand. Concerning Internet use, 54.8% of women use the Internet, with the majority using it for online socialization and entertainment. In terms of personal characteristics, the average age of the hostesses in the sample households is 53.376 years old. The average schooling years of education is 4.151 years, reflecting the generally low educational attainment among rural women in underdeveloped areas. Additionally, 94.7% of women are married. For household characteristics, the average household size is 4.93 persons, and the average farm size is 6.7 mu. The average distance to the nearest market is 6.25 km, and 34.8% of rural women’s households are officially registered as poor households. These statistics suggest that these regions have significant potential for economic growth.

The samples used in this study were divided into two groups: the treatment group (Internet users) and the control group (non-users). As shown in

Table A1. Descriptive statistics of differences between treatment and control groups.

Variables	Treatment Group	Control Group	Difference 1
Explained Variables
Labor income	1.444	0.580	0.864 ***
Wage income	0.688	0.162	0.526 ***
Business income	0.756	0.418	0.338 ***
Control Variables
Age	48.730	58.981	−10.252 ***
Education	5.278	2.786	2.492 ***
Health	0.398	0.688	−0.290 ***
Marital status	0.955	0.936	0.019
Party membership	0.055	0.023	0.032 ***
Household size	4.928	4.924	0.004
Proportion of children	0.194	0.163	0.031 ***
Farm size	6.906	6.450	0.456
Distance to market	6.101	6.431	−0.331
Poor household	0.386	0.316	0.070 ***
Regional economy	0.889	0.850	0.039 ***
Distance to the county	38.611	37.272	1.339
Mechanism Variables
Employment	0.849	0.791	0.058 ***
Agricultural employment	0.542	0.660	−0.119 ***
Off-farm employment	0.452	0.217	0.235 ***
Agricultural working hours	3.396	3.888	−0.493 **
Off-farm working hours	2.512	1.113	1.399 ***
Social capital	0.234	0.099	0.135 ***
Human capital	0.475	0.339	0.136 ***

¹ *** p < 0.01, ** p < 0.05.

, the characteristics of the two groups differ significantly. First, the income of the treatment group is significantly higher than that of the control group; however, this result does not necessarily imply a causal relationship between Internet use and income as other factors may also have influenced the income of the treatment group. The next section of this paper examines the causal relationship between Internet use and income. Second, there are significant differences between women who use the Internet and those who do not, regarding age, education, health, political affiliation, number of children in the family, and village characteristics. For example, Internet users tend to be younger, healthier, and live in villages with better economic conditions.

4. Empirical Results and Analysis

4.1. Baseline Estimation

This study used a multiple linear regression analysis based on Equation (1) to assess the effect of Internet use on rural women’s income in underdeveloped areas.

Table 2. Baseline regression results.

	Labor Income	Wage Income	Business Income
	(1)	(2)	(3)
Internet use	0.263 ***	0.152 ***	0.145 ***
	(0.043)	(0.032)	(0.038)
Age	−0.008 ***	−0.007 ***	−0.001
	(0.002)	(0.002)	(0.002)
Education	0.007	0.014 ***	−0.005
	(0.007)	(0.005)	(0.006)
Health	−0.133 ***	−0.107 ***	−0.046 **
	(0.028)	(0.020)	(0.022)
Marital status	0.174 *	0.354 ***	−0.122 *
	(0.088)	(0.088)	(0.071)
Party membership	0.053	−0.046	0.059
	(0.100)	(0.066)	(0.078)
Household size	−0.036 ***	−0.013	−0.025 **
	(0.014)	(0.010)	(0.012)
Number of children	0.110	−0.108	0.234 *
	(0.160)	(0.117)	(0.135)
Farm size	0.017 ***	−0.004 *	0.022 ***
	(0.004)	(0.002)	(0.003)
Distance to market	−0.003	−0.000	−0.004 *
	(0.005)	(0.004)	(0.002)
Poor household	−0.068 *	−0.047	−0.016
	(0.040)	(0.030)	(0.038)
Regional economy	0.135	0.091	0.093
	(0.148)	(0.118)	(0.138)
Distance to the county	−0.000	−0.001 **	0.001
	(0.001)	(0.001)	(0.001)
Provincial dummies	YES	YES	YES
Constant	0.966 ***	0.802 ***	0.267 *
	(0.159)	(0.123)	(0.141)
Obs	1384	1384	1384
R2	0.160	0.146	0.139

Notes: (1) *** p < 0.01, ** p < 0.05, * p < 0.1. (2) Robust standard errors clustered at the village level are in parentheses.

presents the results. First, we assessed the overall effect of Internet use on rural women’s labor income. As shown in Column (1), Internet use leads to a 26.3 percent increase in rural women’s labor income. This study also analyzes the effect of Internet use on different sources of rural women’s incomes. According to Columns (2) and (3), Internet use significantly and positively affects both wage income and business income. Specifically, Internet use by rural women is associated with a 15.2 percent increase in wage income and a 14.5 percent increase in business income, with all coefficients being statistically significant at the 1 percent level. The results show a significant positive correlation between Internet use and rural women’s income, which is consistent with previous research. To date, several studies have emphasized the positive role of information technology adoption in enhancing agricultural income [57], increasing participation in off-farm work and non-agricultural income [26], and increasing farmers’ business income [25].

4.2. Endogeneity

The results of the baseline regression indicate that Internet use is positively correlated with rural women’s income. To examine the causal relationship between Internet use and rural women’s income more effectively, it is crucial to identify potential endogeneity. The sources of endogeneity can be broadly classified into two main categories: The first is omitted variables, which occur when relevant explanatory variables are excluded from the model, resulting in biased coefficient estimates. The second is reverse causality, where Internet use and rural women’s incomes may influence each other. In such cases, the OLS model for parameter estimation may lead to biased estimates. The instrumental variable (IV) approach was utilized in this study to solve the endogeneity issues caused by omitted variables and reverse causality.

Following Zhou and Cui (2020) [58], we employed the Internet penetration rate of other households in the villages where the respondent is located as an IV for Internet use. The rationality of the IV can be explained from two aspects. Firstly, the “peer effect” suggests that individual behavior is influenced by those in their surroundings. If the Internet penetration rate in the village where the respondent resides is higher, the respondent within such an environment is more likely to have access to and use the Internet, thereby satisfying the relevant conditions. Second, the Internet penetration rate of other households in the village does not exert a direct impact on the income of the sample, thus meeting exogenous requirements. Therefore, it is reasonable to conclude that using village Internet penetration rate as an IV for rural women’s Internet use is an appropriate method.

We performed a Two-Stage Least Squares (2SLS) estimation to quantify the impact of Internet use on the income of rural women. First of all, we analyzed the validity of the IV. The first-stage results of the 2SLS, shown in Column (1) of

Table 3. The 2SLS results.

	Internet Use	Labor Income	Wage Income	Business Income
	(1)	(2)	(3)	(4)
Internet use		0.360 **	0.188 **	0.262
		(0.176)	(0.094)	(0.200)
Internet penetration rate	0.748 ***
	(0.045)
Control variables	YES	YES	YES	YES
Constant	0.997 ***	0.804 ***	0.748 ***	0.092
	(0.104)	(0.276)	(0.193)	(0.279)
Obs	1384	1384	1384	1384
R2	0.333	0.157	0.145	0.134

Notes: (1) *** p < 0.01, ** p < 0.05. (2) Robust standard errors clustered at village level are in parentheses. (3) The control variables are the same as in Table 2.

, indicate a significant correlation between the IV, “Internet penetration rate”, and the core explanatory variable, “Internet use”. Furthermore, the Cragg-Donald Wald F-value is 34.016, and the Kleibergen-Paap rk Wald F-value is 67.051, both of which exceed the threshold of 10, thereby rejecting the null hypothesis of a weak IV. In addition, the p-value of the Sargan test is higher than 0.1, indicating that the IV does not have a non-exogenous problem. Therefore, selecting the Internet penetration rate as an IV for Internet use is valid.

After addressing the problem of endogeneity, as presented in Table 3, rural women’s labor income and wage income are positively impacted by Internet use. Furthermore, the regression coefficients are larger than those in the baseline regression, indicating a stronger relationship between Internet use and rural women’s income. This suggests that the impact of Internet use on rural women’s income may have been underestimated because of endogeneity. Moreover, the impact of Internet use on business income is no longer statistically significant, which is consistent with Fang and Zhu’s (2024) [59] findings. A possible explanation is that, on the one hand, the Internet lowers the barriers to entrepreneurship, helping rural women start businesses and increasing their non-agricultural business income [25,60]; on the other hand, the Internet has facilitated the transformation of the employment structure from agricultural to non-agricultural [28]. Rural women who were originally engaged in agricultural production shifted to off-farm sectors, resulting in a decrease in agricultural business income. These two effects may offset one another, leading to an insignificant effect of Internet use on rural women’s business income.

4.3. Robustness Tests

4.3.1. Change in Sample Range: Focusing on the Labor Force Samples

The above examination does not impose age restrictions on the samples because farmers do not have a strict retirement age and tend to continue working as long as their health permits. However, Huang et al. (2023) [61] argued that while the Internet creates many informal job opportunities and removes barriers related to age and geography, the income-enhancing effects are more likely to be concentrated within the active working population. Therefore, we reset the sample range and performed regressions on women aged 16–55 and 16–65 years.

As demonstrated in

Table 4. Change in sample range: focusing on the labor force samples.

	Labor Income		Wage Income		Business Income
	Aged 16–55	Aged 16–65	Aged 16–55	Aged 16–65	Aged 16–55	Aged 16–65
	(1)	(2)	(3)	(4)	(5)	(6)
Internet use	0.265 ***	0.268 ***	0.171 ***	0.154 ***	0.131 ***	0.147 ***
	(0.059)	(0.045)	(0.046)	(0.034)	(0.048)	(0.041)
Control variables	YES	YES	YES	YES	YES	YES
Constant	0.816 **	1.031 ***	0.764 ***	0.900 ***	0.138	0.253
	(0.345)	(0.208)	(0.251)	(0.152)	(0.267)	(0.172)
Obs	808	1169	808	1169	808	1169
R2	0.120	0.136	0.146	0.143	0.125	0.128

Notes: (1) *** p < 0.01, ** p < 0.05. (2) Robust standard errors clustered at village level are in parentheses. (3) The control variables are the same as in Table 2.

, the coefficients of Internet use remain significantly positive across all the models. These coefficients are consistent with the direction and significance of the baseline regression results. In other words, the income-enhancing effects of Internet use on rural women in underdeveloped areas are consistently significant, and the previous conclusion is reliable.

4.3.2. Substitution of the Explanatory Variable: Degree of Internet Use

Internet use is the key explanatory variable used in this study. In the baseline regression, Internet use is a binary variable measured by asking respondents whether the hostess uses the Internet. Although most previous studies have adopted this measurement method, given the widespread adoption of the Internet and the expansion of usage scenarios, a simple binary measure of “whether or not” is no longer adequate to capture the influence of varying degrees of Internet use on users’ income.

In this study, Internet use by rural women was surveyed across 11 scenarios. We construct a new variable representing the degree of rural women’s Internet use by counting the scenarios in which they were engaged. Therefore, we can use “degree of Internet use” to replace “Internet use” for the robustness test. As shown in

Table 5. Substitution of the explanatory variable: degree of Internet use.

	Labor Income	Wage Income	Business Income
	(1)	(2)	(3)
Degree of Internet use	0.074 ***	0.043 ***	0.044 ***
	(0.012)	(0.009)	(0.010)
Control variables	YES	YES	YES
Constant	0.924 ***	0.773 ***	0.223
	(0.168)	(0.127)	(0.152)
Obs	1384	1384	1384
R2	0.169	0.152	0.145

Notes: (1) *** p < 0.01. (2) Robust standard errors clustered at village level are in parentheses. (3) The control variables are the same as in Table 2.

, the coefficients of the degree of Internet use are all significantly positive at the 1 percent level. This indicates that an increase in the degree of Internet use significantly boosts rural women’s income in underdeveloped areas. In other words, a higher degree of Internet use corresponds to a greater income-enhancing effect. Overall, the above findings are robust after changing the explanatory variables.

4.3.3. Replacing the Test Model: Propensity Score Matching (PSM) Method

Theoretically, the decision to use the Internet is not random. For instance, younger and better-educated groups are more inclined to choose to use the Internet. In this context, rural women who use the Internet are highly likely to earn higher incomes. If the aforementioned self-selection process is overlooked, directly using the OLS model for parameter estimation may lead to biased estimates (self-selection bias). Thus, the impact of Internet use on rural women’s income may be inaccurately estimated. The instrumental variable (IV) method and propensity score matching (PSM) method are two common approaches for addressing self-selection bias in econometric analysis models [23,62]. In the endogeneity section, we address the endogeneity issue by using the IV method. Here, we employ the PSM method as a robustness test to address the estimation bias arising from sample self-selection problems.

Firstly, a counterfactual framework is constructed, and the samples are divided into a treatment group (Internet users) and a control group (non-users) to obtain the average treatment effect for the treatment group (ATT):

$$ATT=E\left\{E\left[Y_{1i}{D}_i=1,p\left(X_i\right)\right]-E\left[Y_{0i}{D}_i=0,p\left(X_i\right)\right]\right\}$$

(3)

where $i$ represents a female individual; $Y_{1i}$ denotes the income of rural women who use the Internet; $Y_{0i}$ denotes the income of rural women who do not use the Internet; $D_i=1$ represents rural women who use the Internet; $D_i=0$ represents rural women who do not use the Internet; $X_i$ represents other control variables; and $p\left(X_i\right)$ is the conditional probability (propensity score) of whether a given sample uses the Internet, which can be expressed as follows:

$$p\left(X_i\right)=\mathrm{P}\mathrm{r}\left[D_i=1|X_i\right]$$

(4)

Subsequently, the treatment group and control group are matched. After matching, if there are no significant differences in individual characteristics, the income difference between the two groups represents the net effect of Internet use. When conducting propensity score matching, the focus is on the ATT values.

To enhance the robustness of the results, we selected three matching methods, including radius matching, kernel matching, and nearest neighbor matching. As shown in

Table 6. ATT of Internet use on rural women’s labor income.

Matching Methods	Differences	Standard Error	T-Value
Radius matching (caliper = 0.01)	0.297 ***	0.051	5.69
Kernel matching (bandwidth = 0.05)	0.294 ***	0.051	5.75
Nearest neighbor matching (1:5)	0.309 ***	0.055	5.25

Notes: (1) *** p < 0.01. (2) Robust standard errors clustered at village level are in parentheses. (3) The control variables are the same as in Table 2.

, the ATT values derived from all matching methods are close to the baseline regression model’s results. In addition, all ATT values are statistically significant at the 1% level. Therefore, the estimation results of the PSM model are consistent with those of the baseline regression model. These findings are supported by the fact that Internet use can greatly boost the labor income of rural women.

To assess the quality of the matching, a balance test was required to confirm whether the distribution of the explanatory variables was even between the treatment group and the control group before and after matching. The explanatory variables are selected based on the control variables of the baseline regression model.

Table A2. Balance test results before and after matching.

Variables	Before Matching (U)	Mean		Standard Deviation	Error Deviation	T-Test
	After Matching (M)	Treatment Group	Control Group	(%)	(%)	t	p > |t|
Age	U	48.702	58.962	−103.2		−19.29	0.000
	M	48.968	48.353	1.8	98.2	0.39	0.694
Education	U	5.366	2.817	70.2		12.94	0.000
	M	5.329	5.281	1.3	98.2	0.24	0.809
Health	U	0.393	0.688	−42.0		−7.83	0.000
	M	0.394	0.400	−0.9	97.8	−0.19	0.849
Marital status	U	0.057	0.024	16.9		3.07	0.002
	M	0.053	0.044	4.6	72.7	0.81	0.415
Party membership	U	0.954	0.935	8.2		1.53	0.126
	M	0.956	0.944	5.1	38.3	1.03	0.305
Household size	U	4.928	4.916	0.7		0.13	0.893
	M	4.935	4.971	−2.0	−181.9	−0.44	0.663
Proportion of children	U	0.192	0.162	18.1		3.35	0.001
	M	0.192	0.191	0.8	95.3	0.16	0.871
Farm size	U	6.983	6.476	6.7		1.23	0.219
	M	6.997	6.866	1.7	74.1	0.33	0.740
Distance to market	U	6.047	6.439	−4.6		−0.88	0.379
	M	6.055	6.497	−5.3	−13.1	−1.19	0.233
Poor household	U	0.313	0.386	−15.5		−2.88	0.004
	M	0.314	0.323	−1.8	88.2	−0.36	0.717
Regional economy	U	0.625	0.604	14.6		2.70	0.007
	M	0.625	0.607	12.3	15.7	2.30	0.022
Distance to the county	U	38.384	37.128	4.4		0.80	0.422
	M	38.335	37.402	3.2	25.8	0.63	0.530

presents the results of the balance test conducted before and after matching when utilizing the radius-matching method. Before matching, the treatment and control groups differed significantly in terms of their individual and family characteristics; however, following the matching process, these differences are markedly diminished and the proportion of deviation in the control variables is reduced to below 10%. The t-test results further indicate that the absolute t-values of the samples decrease substantially after matching, with the significance tests for the variables no longer showing significant differences between the treatment and control groups. This indicates that the samples have passed the balance test after matching, thereby confirming the effectiveness of the PSM method in addressing self-selection bias.

4.4. Heterogeneity Analysis

4.4.1. Heterogeneity of Internet Use Purposes

Rural women’s adoption and application of the Internet (e.g., online socialization, entertainment, business, and learning) is influenced by many factors, including educational attainment, health status, family status, and accessibility to Internet services [63]. The baseline regression mainly explores the influence of Internet use on rural women’s income. This section comprehensively analyzes the impact of different Internet use purposes on rural women’s income based on Internet application scenarios.

Table 7. The impact of different purposes of Internet use on rural women’s labor income.

	(1)	(2)	(3)	(4)
Online socialization	0.223 ***
	(0.046)
Online entertainment		0.209 ***
		(0.043)
Online business			0.253 ***
			(0.048)
Online learning				0.296 ***
				(0.061)
Control variables	YES	YES	YES	YES
Constant	0.973 ***	0.962 ***	1.024 ***	1.253 ***
	(0.163)	(0.160)	(0.171)	(0.173)
Obs	1384	1384	1384	1384
R2	0.160	0.166	0.159	0.151

Notes: (1) *** p < 0.01. (2) Robust standard errors clustered at village level are in parentheses. (3) The control variables are the same as in Table 2.

demonstrates that all purposes of Internet use significantly increase rural women’s income. Notably, the coefficients for online business and online learning are higher than those for online socialization and online entertainment, indicating that the income-enhancing effects of online business and learning are more pronounced. In recent years, rural areas have seen a dramatic transformation in lifestyle and industry due to the rapid expansion of the Internet. The advent of online sales, purchases, and payments has had a profound impact on rural livelihoods. Some new business models such as rural e-commerce and live streaming have emerged, which have significantly increased farmers’ income [50,64]. Moreover, Internet platforms are becoming increasingly crucial channels for farmers to access information and improve their production and business skills. Overall, the Internet has become a key tool in guiding the development of rural women and increasing their income.

4.4.2. Heterogeneity of Income Level

This study employs the quantile regression (QR) model to reveal the heterogeneous impact of Internet use on the income growth of rural women with different income levels. Furthermore, it assesses the potential effect of Internet use on income distribution fairness in underdeveloped areas.

Table 8. Quartile regression results.

	0.1 Quartile	0.25 Quartile	0.5 Quartile	0.75 Quartile	0.9 Quartile
Internet use	0.025 *	0.029 **	0.132 **	0.379 ***	0.553 ***
	(0.014)	(0.014)	(0.052)	(0.088)	(0.121)
Control variables	YES	YES	YES	YES	YES
Constant	−0.048	0.008	0.568 ***	1.808 ***	2.504 ***
	(0.059)	(0.040)	(0.195)	(0.326)	(0.511)
Obs	1384	1384	1384	1384	1384

Notes: (1) *** p < 0.01, ** p < 0.05, * p < 0.1. (2) Robust standard errors clustered at the village level are in parentheses. (3) The control variables are the same as in Table 2.

presents the results. The effect of Internet use on rural women’s income increases with higher quantiles, suggesting the existence of a “Matthew Effect” in the income-enhancing impact of the Internet. This suggests that those in higher-income brackets tend to gain greater advantages from Internet use. While the marginal contribution of Internet use is significantly positive across all income levels, it also appears to widen the income gap. One potential explanation for this phenomenon is that low-income women often belong to the “digitally disadvantaged” group [63]. On the one hand, they may lack the necessary skills to effectively utilize the Internet; on the other hand, their lower resource endowment may restrict their capacity to fully leverage the resources and opportunities afforded by the Internet. The digital divide within rural areas must be bridged urgently, and Internet accessibility for disadvantaged groups must be enhanced to increase the income of low-income rural women and narrow the income gap.

4.4.3. Heterogeneity of Individual Characteristics

This study delineates three distinct groups: a young group (aged 16–44), a middle-aged group (aged 45–59), and an elderly group (aged 60+). Subsequently, subgroup regression analyses are conducted for each group. As demonstrated in

Table 9. Heterogeneity analysis: grouped by age.

	Labor Income			Wage Income			Business Income
	Young Group	Middle-Aged Group	Elderly Group	Young Group	Middle-Aged Group	Elderly Group	Young Group	Middle-Aged Group	Elderly Group
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Internet use	0.406 ***	0.171 ***	0.196 **	0.258 ***	0.075 *	0.166 **	0.189 *	0.120 **	0.054
	(0.127)	(0.059)	(0.088)	(0.086)	(0.041)	(0.065)	(0.097)	(0.051)	(0.065)
Control variables	YES	YES	YES	YES	YES	YES	YES	YES	YES
Constant	0.007	1.862 ***	−0.243	0.288	1.505 ***	−0.291	−0.360	0.651 *	0.094
	(0.647)	(0.394)	(0.409)	(0.490)	(0.355)	(0.271)	(0.537)	(0.359)	(0.344)
Obs	286	708	390	286	708	390	286	708	390
R2	0.134	0.148	0.143	0.153	0.144	0.094	0.119	0.157	0.154
Difference in coefficients between groups	(1)–(2)	(2)–(3)	(1)–(3)	(4)–(5)	(5)–(6)	(4)–(6)	(7)–(8)	(8)–(9)	(7)–(9)
	0.235	−0.024	0.210	0.183	−0.091	0.092	0.069	0.066	0.135
p values	0.123	0.822	0.194	0.160	0.237	0.408	0.587	0.433	0.298

Notes: (1) *** p < 0.01, ** p < 0.05, * p < 0.1. (2) Robust standard errors clustered at the village level are in parentheses. (3) The control variables are the same as in Table 2.

, the findings indicate that Internet use has a markedly positive influence on rural women’s labor income, wage income, and business income across all groups. This encouraging result highlights the potential of digital technology in rural development, particularly for women. The tests comparing the coefficients between age groups reveal no statistically significant differences in the regression coefficients. This suggests that the income-enhancing effect of Internet use is universal, benefiting rural women across all age groups.

These findings are corroborated by the field research conducted in specialized e-commerce villages in Qingshui County, Gansu Province. In these villages, young women are typically engaged in online sales, middle-aged groups focus on agricultural production, and elderly women grade and label agricultural products for online stores. This age-appropriate division of labor within rural e-commerce activities demonstrates how the Internet provides income-enhancing opportunities tailored to different age groups.

The descriptive statistics in Table 1 indicate that the average years of education for the samples is only 4.151, reflecting a generally low educational attainment. Therefore, this study divides rural women into two groups according to whether they have finished the complete nine-year compulsory education: “low-educated” (less than 9 years of education) and “highly educated” (nine years of education or more). As presented in

Table 10. Heterogeneity analysis: grouped by education level.

	Labor Income		Wage Income		Business Income
	Low-Educated Group	Highly Educated Group	Low-Educated Group	Highly Educated Group	Low-Educated Group	Highly Educated Group
	(1)	(2)	(3)	(4)	(5)	(6)
Internet use	0.239 ***	0.483 ***	0.138 ***	0.319 ***	0.122 ***	0.253 **
	(0.049)	(0.133)	(0.034)	(0.106)	(0.041)	(0.098)
Control variables	YES	YES	YES	YES	YES	YES
Constant	0.992 ***	0.368	0.785 ***	0.255	0.332 *	0.073
	(0.189)	(0.549)	(0.136)	(0.446)	(0.170)	(0.471)
Obs	1097	287	1097	287	1097	287
R2	0.169	0.188	0.104	0.169	0.177	0.078
Difference in coefficients between groups	−0.244 *		−0.181 *		−0.131 *
p values	0.065		0.087		0.075

Notes: (1) *** p < 0.01, ** p < 0.05, * p < 0.1. (2) Robust standard errors clustered at the village level are in parentheses. (3) The control variables are the same as in Table 2.

, compared with the low-educated group, the income-enhancing effect of Internet use is more pronounced in the highly educated group.

The potential reasons for this difference include the fact that the highly educated group are generally more adept at accepting new ideas and utilizing new technologies. They typically possess stronger Internet use skills, higher proficiency in Internet applications, and a greater likelihood of accessing information or conducting work online. As a result, Internet use has a more significant marginal impact on the income of highly educated women.

4.4.4. Heterogeneity of Family Characteristics

China’s childcare service system remains underdeveloped, which makes it much harder for women to enter the workforce [65]. The samples in this study are split into two groups according to whether or not preschoolers are included in the family members. The regression results in

Table 11. Heterogeneity analysis: grouped by whether or not they are raising preschoolers.

	Labor Income		Wage Income		Business Income
	No Preschoolers	With Preschoolers	No Preschoolers	With Preschoolers	No Preschoolers	With Preschoolers
	(1)	(2)	(3)	(4)	(5)	(6)
Internet use	0.329 ***	0.133	0.173 ***	0.113 **	0.191 ***	0.042
	(0.057)	(0.082)	(0.043)	(0.048)	(0.046)	(0.077)
Control variables	YES	YES	YES	YES	YES	YES
Constant	0.996 ***	0.776 **	0.899 ***	0.495 **	0.206	0.385
	(0.194)	(0.353)	(0.164)	(0.221)	(0.153)	(0.291)
Obs	971	413	971	413	971	413
R2	0.197	0.146	0.172	0.136	0.182	0.103
Difference in coefficients between groups	0.196 *		0.060		0.149 *
p values	0.054		0.367		0.088

Notes: (1) *** p < 0.01, ** p < 0.05, * p < 0.1. (2) Robust standard errors clustered at the village level are in parentheses. (3) The control variables are the same as in Table 2.

indicate that rural women without childcare responsibilities are more likely to benefit from Internet use than those with preschoolers in their families. Rural women without preschoolers experience more pronounced labor and business income growth.

This phenomenon may be attributed to the fact that rural women without childcare responsibilities have more discretionary time to engage in activities such as learning, working, or starting a business through the Internet. Conversely, rural women with childcare responsibilities, even when presented with the opportunity to utilize the Internet, frequently lack the requisite energy to engage in Internet-related activities or participate in the labor market.

4.5. Mechanism Analysis

4.5.1. Adjustment of Labor Inputs

The empirical analysis in this section examines the causal relationship between Internet use and labor input. If rural women participate in employment and extend their working hours, their incomes will naturally increase significantly. Therefore, it is not necessary to test the causal relationship between labor input and income [61]. To verify the role of Internet use in increasing rural women’s income by adjusting labor inputs, we examined two aspects: promoting rural women’s employment participation and increasing their working hours.

Table 12. Mechanism: adjustment of labor inputs.

	Employment Participation			Working Hours
	Employment	Agricultural Employment	Off-Farm Employment	Agricultural Working Hours	Off-Farm Working Hours
	(1)	(2)	(3)	(4)	(5)
Internet use	0.130	0.061	0.289 ***	−0.395 ***	0.578 ***
	(0.091)	(0.080)	(0.089)	(0.134)	(0.188)
Control variables	YES	YES	YES	YES	YES
Constant	1.271 ***	−1.444 ***	1.555 ***	3.516 ***	5.890 ***
	(0.377)	(0.349)	(0.353)	(1.058)	(0.759)
Obs	1384	1384	1384	1384	1384
Pseudo R2/R2	0.102	0.110	0.150	0.093	0.135

Notes: (1) *** p < 0.01. (2) Robust standard errors clustered at the village level are in parentheses. (3) The control variables are the same as in Table 2.

reports the regression results. According to Columns (1)–(3), Internet use promotes the off-farm employment of rural women, while showing no discernible effect on overall or agricultural employment. Columns (1) and (2) indicate that rural women actively engage in labor activities, regardless of Internet use, which aligns with real-world observations and supports the causal relationship between Internet use and off-farm employment. Thus, Internet use behaviour helps to increase the probability of rural women’s off-farm employment participation, thereby increasing their income from wages. This study also explores the effect of Internet use on rural women’s working hours in underdeveloped areas. As shown in Columns (4) and (5) of Table 12, Internet use reduces rural women’s agricultural working hours significantly while raising their off-farm working hours. The findings show that, as predicted, there has been a labor force transfer from the agricultural to the non-agricultural sectors as a consequence of Internet use. This phenomenon has two impacts. First, Internet use has extended the time spent completing off-farm work, thereby increasing wage income. Second, Internet use has reduced the time spent on agricultural work, leading to a decline in agricultural business income. These findings confirm the hypothesis that Internet use affects the income of rural women by adjusting labor inputs. Hypothesis 2 is verified.

4.5.2. Enhancement of Capital Endowment

Enhancing rural women’s capital endowment is a crucial strategy for raising their incomes [3]. This capital-accumulation effect not only facilitates closer links between rural women and the labor market but also significantly enhances their competitiveness in it [21]. Specifically, extensive social networks have provided rural women with more diversified channels of information and access to resources, helping them grasp employment opportunities and reduce information asymmetry. At the same time, the improvement in their work skills directly enhances their suitability and efficiency in the labor market, thus further consolidating their competitive advantages. All of these factors lead to an increase in rural women’s income. To verify the enhancement effect of Internet use on rural women’s capital endowment, we conducted tests on two aspects: rural women’s endowment of social capital and human capital.

Table 13. Mechanism: enhancement of capital endowment.

	Social Capital	Human Capital
	(1)	(2)
Internet use	0.055 ***	0.378 ***
	(0.018)	(0.103)
Control variables	YES	YES
Constant	0.375 ***	−1.492 ***
	(0.090)	(0.422)
Obs	1384	1384
Pseudo R2/R2	0.136	0.131

Notes: (1) *** p < 0.01. (2) Robust standard errors clustered at the village level are in parentheses. (3)The control variables are the same as in Table 2.

presents the impact of Internet use on the capital endowment of rural women. The estimated coefficient of Internet use on rural women’s social capital (Column (1)) is significantly positive. It can be proposed that the Internet provides convenient communication channels, thereby overcoming the geographical limitations of traditional communications. This fosters external communication among rural women in underdeveloped regions, thereby enhancing their social capital. Moreover, the impact of Internet use on rural women’s human capital is also markedly positive at the 1% level. The Internet platform not only provides a vast amount of learning resources but also disseminates information from free government training [46]. These opportunities can compensate for the lack of formal education available to rural women in underdeveloped areas, thereby augmenting their professional skills. Thus, Hypothesis 3 is verified.

5. Conclusions and Policy Implications

Increasing women’s income is a critical strategy for achieving SDG 5. Meanwhile, the advancement of specific targets under SDG 5 provides women with more opportunities and resources, thereby increasing their income. The Internet is considered to have significant potential for improving women’s income. This study surveyed 1384 rural households in underdeveloped areas of Western China in 2021 to analyze how Internet use affects rural women’s income. These findings suggest that Internet use has significantly increased the income of rural women in underdeveloped areas. The effect of income growth exhibits heterogeneity owing to variations in the purposes of Internet use, income levels, individual characteristics, and family characteristics. Internet use has increased the income of rural women by facilitating the transfer of agricultural labor to non-agricultural sectors. Specifically, this increases the possibility of off-farm employment for rural women, extends off-farm working hours, and reduces agricultural working hours. In addition, Internet use increases rural women’s labor income by enhancing their capital endowment. Specifically, Internet use enhances rural women’s social capital and human capital.

These findings not only confirm the positive role of Internet use in increasing rural women’s income and alleviating poverty but also provide valuable policy insights for management practices focused on achieving SDG 5.

First, the government should continue to support the coverage of Internet services in rural areas. This study found that the Internet is the most commonly used information and communication tool for rural women, playing a crucial role in their daily lives and production activities. The government should increase investment in information infrastructure in rural areas, improve the accessibility of Internet services, and reduce the cost of Internet services in rural areas to enable rural women in remote and impoverished regions to access high-speed, affordable Internet services.

Second, the government should support Internet-based training courses to increase the digital literacy of rural women. This study confirms that the extent to which rural women benefit from Internet use is significantly influenced by the degree and purpose of Internet use. Offering training programs on Internet skills for rural women in underdeveloped areas is an important approach to enhancing their abilities. During the training process, the government should combine the production and living needs of rural women, providing practical courses such as online shopping and payment as well as employment-oriented courses such as engaging in e-commerce activities and being online customer service representatives.

Third, the government should create employment opportunities for rural women in underdeveloped areas. The government can strengthen cooperation with relevant Internet enterprises (e.g., Alibaba or TikTok) to promote digital industry projects (e.g., Fighting Mulan, AI Bean Plan) to be implemented locally. These projects aim to help women find online jobs, which can offer job opportunities for rural women. In addition, the government can support the development of Internet-related industries, such as rural e-commerce. The development of rural e-commerce can not only create more jobs but also promote the sales of local products.

Fourth, rural women should actively adopt emerging digital technologies. By utilizing the Internet and other digital technologies, rural women can expand their social networks and acquire new knowledge and skills, which can enhance their social and human capital. This has significant practical implications for improving their competitiveness in the labor market and increasing their income.

This study has several limitations. First, as the empirical evidence is based on one-year cross-sectional data, the dynamic relationship between Internet use and rural women’s income cannot be captured. Second, this study focuses on wage income and business income from women’s participation in work; however, due to the lack of relevant data, we were unable to explore how Internet use affects rural women’s property income and transfer income.