How Do Digital Skills Affect Rural Households’ Incomes in China? An Explanation Derived from Factor Allocation

Abstract

Raising rural household income is central to narrowing the rural–urban gap and advancing common prosperity. Using data from the China Family Panel Studies (CFPS), this study examines the impact of digital skills, a key for human capital, on rural Chinese households’ income and uses a fixed-effects model and the instrumental variable method to address endogeneity. The study finds that digital skills raise total household income, and each additional skill is associated with an increase of CNY 1678. By skill type, online business skills have the largest effect, followed by work–study skills, while entertainment–social skills are negatively associated with income. Heterogeneity analyses indicate larger gains for households with lower educational attainment and lower income, showing that a stronger regional digital environment amplifies these effects. Mechanism tests point to factor reallocation toward the nonfarm sector, via higher probabilities of off-farm employment and entrepreneurship and improved access to formal credit, as the primary pathway. Consistent with these channels, digital skills increase wages and operating income and reduce inequality in these components, as well as benefitting total income, but they have no detectable effect on property or transfer income or their dispersion. These findings point to key implications for boosting rural income growth and reducing inequality, namely strengthening digital skill development and optimizing the digital environment to enhance rural households’ endogenous income-generating capacity.

1. Introduction

Poverty remains a central challenge for developing countries and the global community [1,2]. In response, SDG 1 commits to “end poverty in all its forms everywhere”. Over the past three decades, China has played a pivotal role. Under the current national standard, 98.99 million rural residents have exited poverty, enabling China to achieve the SDG poverty-reduction target nearly a decade early [3]. With absolute poverty eradicated, the policy focus now shifts to sustaining income growth among rural households as a pathway to common prosperity [4,5]. Yet the traditional drivers of rural income have weakened in recent years, heightening uncertainty over how to secure durable improvements in living standards.

The advent of new-generation information and communication technologies (ICT), including big data, cloud computing, the Internet of Things, and artificial intelligence, has profoundly reshaped agricultural and rural development [6,7]. By reconfiguring factor allocation and transforming production and living conditions, these technologies have become important drivers of rural household income growth [4,5]. A growing body of literature underscores their transformative potential for rural development. Specifically, digital technologies enhance transparency in agricultural production processes and mitigate information asymmetries between farmers and markets [8,9]; rural e-commerce expands market access for agricultural products and improves the alignment of supply with consumer demand [10,11]; and digital platforms broaden opportunities for rural employment and entrepreneurship [12,13,14].

However, digital access alone is insufficient. To fully benefit from the digital economy, rural households require not only equal access to digital infrastructure but also the digital skills necessary to use these technologies effectively. Without adequate skills, individual risk exclusion from increasingly dynamic labor market [15,16]. Digital skills constitute a crucial component of human capital in the digital age, underpinning competitiveness and innovation [17]. As the digital access divide narrows, attention has shifted to the emerging “application divide”, which refers to the unequal ability to utilize digital tools and resources [18,19]. This challenge is particularly salient in China. In 2023, internet penetration reached 77.5% nationally and 66.5% in rural areas, yet the urban–rural digital skills gap remained as high as 21.2 points on a 100-point scale.

The existing literature offers valuable insights but also reveals several limitations. First, most studies have examined digital technology primarily from the accessibility rather than the application perspective. Research shows that the diffusion of information infrastructure can overcome geographical barriers, expand agricultural trade, and increase farm and nonfarm incomes [20,21,22]. Broadband development also reduces transaction costs, enhance employment opportunities, and narrow the urban–rural income gap, especially in lagging regions and among low-income groups [23,24,25,26,27]. In addition, relatively little attention has been paid to intra-rural inequality, specifically whether digital technologies mitigate or exacerbate income disparities among rural households [3]. Second, while existing studies often explore the mechanisms of digital technologies’ income impact in a fragmented manner, and focusing on information access, human capital accumulation, or social capital expansion [5,9,28], these analyses remain incomplete. For instance, though studies note that digital technologies facilitate new learning opportunities and foster social connectivity [27,29]. Yet few studies provide a systematic account of how digital skills shape rural households’ factor allocation behaviors, including participation in nonfarm employment, land transfer, and access to formal credit markets.

Against this backdrop, this study examines the income effects of digital skills among rural households using nationally representative data from the China Family Panel Studies (CFPS). The empirical results show that digital skills significantly increase total household income, including both wage and operating income. Effects are heterogeneous across skill types, online business skills contribute the most, followed by work-study skill, whereas entertainment-social skills are negatively associated with household income. Heterogeneity analysis further reveals that disadvantaged groups, such as low-income and less-educated households, benefit disproportionately, and that improvements in the regional digital environment amplify these positive effects. Mechanism analysis indicates that digital skills facilitate structural adjustments in factor allocation by increasing the likelihood of nonfarm employment and entrepreneurship, and easing access to formal credit for investment, while having no significant impact on land allocation.

This study makes three principal contributions to the literature. First, it shifts the focus from digital access to digital application by providing micro-level evidence on how digital skills shape both the level and structure of rural household income. In doing so, it enriches research on the digital economy and income growth in China and offers insights for other developing economies undergoing similar digital transformations. Second, it adopts an intra-rural perspective by examining whether digital skills mitigate or exacerbate income disparities within rural areas, with particular attention to heterogeneity across education, income levels, and the regional digital environment. This perspective contributes to the broader global debate on whether digital technologies act as an equalizer or a divider in rural development. Third, it advances an integrated understanding of the underlying mechanisms by analyzing how digital skills influence factor allocation behaviors, including nonfarm employment, land transfer, and access to credit. Collectively, the findings generate policy-relevant insights of international significance for leveraging digital skills to foster inclusive and sustainable rural development.

The remainder of the paper is organized as follows: Section 2 outlines the theoretical framework, while Section 3 describes the empirical strategy and data sources. Section 4 presents the baseline regressions, robustness checks, heterogeneity analysis, and mechanism analysis, followed by Section 5, which further examines income structure and inequality. Section 6 concludes with key findings and policy recommendations.

2. Theoretical Framework and Hypothesis

2.1. Digital Skills and Rural Households’ Income

As the digital access divide narrows with continued expansion of diverse infrastructure across China, a central policy priority has shifted to cultivating 21st century competencies. Among these, digital skills—conceived as a distinct dimension of human capital—are increasingly critical for enabling rural households to capture digital dividends. By definition, digital skills entail the capacity to leverage technologies whose core function is the large-scale, intelligent collection, production, and use of information [16,30]. For rural households, this capacity strengthens the acquisition, evaluation, and application of information. In particular, digital skills lower the costs of producing, searching, and processing information, thereby mitigating information asymmetries and easing access constraints [18,31]. The overall theoretical framework is shown in Figure 1.

In the short term, rural households with greater digital skills can more efficiently acquire, filter, and process information, converting fragmented signals into actionable knowledge [4]. For instance, market signals regarding land prices, agricultural product prices, and labor wages. By correcting distortions in returns arising from information asymmetries, digital skills promote income growth [27]. Beyond short-term benefits, datafication of production factors renders knowledge and information highly replicable at near-zero marginal cost [32]. Sustained engagement with digital platforms thus facilitates cumulative learning and internalization of information, reinforcing other dimensions of human capital [13,14]. For example, through instructional videos and peer-to-peer knowledge sharing. This compounding effect strengthens the rural labor force, supports occupational mobility, and generates durable gains in household income.

Based on the above theoretical analysis, this study proposes the following hypothesis.

H1. 

Digital skills promote rural households’ income growth.

2.2. Digital Skills, Factor Allocation and Rural Households’ Income

Digital skills affect rural households’ income primarily through information effects, but the mechanism extends beyond simple information acquisition. In the classical household production framework, households allocate land, labor, and credit across markets to maximize income subject to resource and technological constraints [27]. Within this framework, information is not an end in itself; it is an input that reshapes factor allocation [33]. In other words, the information-enhancing function of digital skills enables more efficient allocation of production factors, which becomes the proximate pathway to higher income.

Complementing this view, new institutional economics emphasizes that allocation is constrained by both initial endowments and transaction costs, such as search, negotiation, and monitoring costs, and that information asymmetry inflates these costs and impairs market functioning [34]. By reducing the costs of generating, searching, and verifying information, digital skills relax these frictions [35]. Households are thereby better able to align resources with opportunities in land, credit, and labor markets, reallocating factors toward higher-return uses and raising income.

First, digital skills can facilitate land transfer and the scaling of agricultural operations. Land is both the core production factor and a multifunctional asset that provides livelihood security, employment, and old-age support. Empirical work links land transfer to income growth [36,37]. By enabling participation in online land-transaction platforms, digital skills reduce pre- and post-contract information asymmetries, strengthen contract enforceability, and lower monitoring costs [38]. Lower transaction costs allow households either to lease out land for stable rental income or to consolidate plots and realize economies of scale, thereby improving land use and earnings [39].

Second, digital skills can relax credit constraints by lowering barriers to digital financial services. Credit availability supports consumption smoothing, productive investment, and the accumulation of human and social capital, all of which raise income [40,41]. Yet rural credit markets are typically characterized by severe information asymmetry and high transaction costs. China’s digital-inclusive finance initiatives aim to bridge this gap [5,42], but effective participation requires a threshold level of digital competence. Households with adequate digital skills can reduce the search, negotiation, and enforcement costs of borrowing and access formal credit [27,43], thereby financing production, business ventures, and longer-term development.

Finally, digital skills can improve labor allocation by reducing job search costs and mitigating information asymmetry in labor markets [12]. Digital recruitment platforms widen access to off-farm opportunities and improve employer–worker matching, raising wages [26]. As large cities’ absorptive capacity for surplus labor plateaus, rural entrepreneurship becomes an increasingly salient pathway. Digital platforms help identify opportunities, coordinate resources, and overcome local market constraints [44,45], lowering the transaction costs of opportunity discovery and resource orchestration and enabling households to capture greater value.

Building on this theoretical reasoning, this study proposes the following hypothesis.

H2. 

Digital skills increase rural households’ income by optimizing the allocation of land, capital, and labor factors.

3. Empirical Strategy and Data

3.1. Data

The data used in this paper come from the China Family Panel Studies (CFPS), a nationally representative household survey initiated by the Social Science Research Center of Peking University and fielded biennially since 2010. The CFPS covers 162 districts or counties across 25 provinces, municipalities, and autonomous regions in China and provides rich information on household living conditions, assets, income and expenditures, and individual-level characteristics. For this study, we use the 2014 and 2018 waves, the most recent public releases that contain the variables required to measure digital skills and household income.

We restrict the sample to households residing in rural areas, as identified by the CFPS urban–rural classification of current residence. Within each wave, we cross-sectionally merge the household, adult, and child files to construct the necessary variables. We exclude observations with missing or implausible values for income or digital-skill measures and retain the resulting household-level samples for each wave. Finally, we link households across the 2014 and 2018 waves to form a two-period panel dataset for analysis.

3.2. Variable Measurement

3.2.1. Rural Household’ Income

The household income variable in the CFPS data is measured at the household level, with the dependent variable being the total income of rural household. In the robustness test part, this study also uses per capita household income as the dependent variable to verify the consistency of results.

3.2.2. Digital Skills

Skills are commonly defined as proficiencies developed through practice and imitation. Accordingly, we conceptualize digital skills as the extent to which rural households can apply digital technologies in daily life, production, employment, and education [16,18]. Following prior work [18,46], we construct a composite index that captures the “breadth of digital exposure” rather than “depth of digital ability”. Specifically, we code five binary indicators for whether the household exhibits: (1) e-learning skills (use of digital tools for online learning/educational activities), (2) e-work skills (use of digital tools for work-related tasks), (3) e-socialization skills (use of digital platforms for social interaction, e.g., WeChat), (4) e-entertainment skills (use of online entertainment such as video/music streaming), and (5) online business skills (use of digital technologies for commercial transactions, e.g., online purchasing/sales). The composite digital skills index equals the sum of these five indicators (range: 0–5). For ancillary analyses, we group the five items into three domains: work–learning skills (1–2), entertainment–social skills (3–4), and online business skills (5).

3.2.3. Control Variables

We control for both individual and household level characteristics. Individual controls include age, age squared, gender, marital status, self-reported health, political identity, and years of schooling. Because our outcome is household income, we further adjust for household characteristics: share of older members, household size, social capital, receipt of government subsidies, receipt of social-insurance contributions, and total household assets.

Table 1. Descriptive statistics.

Variables	Definition	Obs.	Mean	S.D.
Rural households’ income	Total household income, unit: thousand yuan	9257	40.844	38.260
Digital skills	Sum of digital skills: 0–5	9257	0.652	1.359
Entertainment-social skill	Use of digital technology for entertainment and social activities: 1 = yes; 0 = otherwise	9257	0.218	0.413
Work-learning skill	Use of digital technologies for work and learning activities: 1 = yes; 0 = otherwise	9257	0.109	0.312
Online business skill	Use of digital technology for online commerce activities: 1 = yes; 0 = otherwise	9257	0.102	0.303
Age	Year	9257	51.222	13.270
Age square	Year square/100	9257	27.997	13.713
Gender	1 = male; 0 = female	9257	0.548	0.498
Marriage	1 = married; 0 = otherwise	9257	0.874	0.332
Health	1 = very unhealthy; 5 = very healthy	9257	2.969	1.291
Political identity	1 = communists; 0 = otherwise	9257	0.041	0.199
Education in years	Years of education, units: year	9257	6.167	4.137
Percentage of elderly	Elderly population as a proportion of total household size	9257	0.202	0.307
Household size	Total number of family members	9257	4.061	1.908
Social capital	Expenditures on family favors, unit: thousand yuan	9257	35.708	49.780
Government subsidies	Receipt of government subsidies:1 = yes; 0 = otherwise	9257	0.667	0.471
Social contributions	Receipt of social contributions: 1 = yes; 0 = otherwise	9257	0.016	0.127
Total assets	Total household assets, units: thousand yuan	9257	296.748	1010.900

Notes: Considering the problem of income and social capital outliers, a 1% reduction in the upper and lower tails is used for such variables.

reports summary statistics for all variables. Mean total household income is 40.844 thousand yuan. The average value of the digital skills index is 0.652, consistent with relatively limited digital proficiency in rural China. The mean respondent age is 51 years; 55% of respondents are male and 87% are married. Approximately 5% report a political identity. On average, self-rated health is relatively good, and mean schooling is 6 years.

3.3. Estimation Model

We estimate the income effects of digital skills using a household fixed-effects (FE) model that absorbs time-invariant unobserved heterogeneity:

$${Income}_{it}={\beta }_0+{\beta }_1{DS}_{it}+\sum \gamma X_{it}+{\lambda }_t+{\mu }_i+{\epsilon }_{it}$$

(1)

where ${Income}_{it}$ denotes income in households $i$ and year $t$, income includes total, property, wage, operating, and transfer income. ${DS}_{it}$ is a ranking variable indicating the digital skills score of households $i$ in year t. $X_{it}$ includes individual and household covariates; ${\lambda }_t$ are time-fixed effects, ${\mu }_i$ denotes individual fixed effects of rural households, controlling for the fixed traits of rural households that do not change over time. ${\epsilon }_{it}$ is a random interference term. The parameter of interest, ${\beta }_1$ measures the within-household association between digital skills and income.

The baseline regression employs a fixed-effects model, which addresses individual heterogeneity but fail to resolve potential endogeneity issues, such as reverse causality and omitted variables. Digital skills reflect an individual’s ability to use digital technology. However, rural households with higher incomes have a “head start” advantage, they are more likely to purchase digital devices and access digital technology in their productive activities. Additionally, unobservable factors that are not controlled for may still introduce estimation bias, such as competence. To address this, we implement two-stage least squares (2SLS) with instrumental variables (IVs) [4].

In economic research, geographic distance is widely used as an IV because it is highly correlated with the variable of interest while exerting no interference on individuals’ specific social characteristics, thus satisfying relevance and exogeneity [47]. Drawing on existing literature [35,38], this study selects the spherical distance between the rural household’s residential area and Hangzhou as the first IV. This IV correlates with digital skills. Hangzhou is home to internet companies led by Alibaba Group, so rural households’ digital skills are negatively correlated with their distance to Hangzhou. Moreover, this geographic distance exerts no direct impact on rural households’ income, satisfying the exogeneity requirement. Notably, as individual behavior is shaped by cognition, this study also selects rural households’ evaluation of the Internet’s importance as an information access channel as the second IV [46]. This evaluation is highly correlated with rural households’ digital skills; however, as a form of subjective cognition, it bears no direct correlation with rural households’ income, further satisfying the exogeneity requirement for IV.

4. Results

4.1. Effects of Digital Skills on Farmers’ Income

4.1.1. Baseline Results

Table 2. Effects of digital skills and sub-dimensions on rural households’ income.

Variables	Rural Households’ Income
	(1)	(2)	(3)	(4)	(5)
Digital skills	1.6782 *** (0.5160)
Entertainment-social skill		0.8001 (1.5453)			−3.6052 ** (1.7552)
Work-learning skill			5.7408 *** (2.0200)		3.8478 * (2.2788)
Online business skill				10.8773 *** (1.9807)	11.2640 *** (2.1769)
Control variables	YES	YES	YES	YES	YES
Regional FE	YES	YES	YES	YES	YES
Time FE	YES	YES	YES	YES	YES
Observations	9257	9257	9257	9257	9257

Notes: Table 2 shows the baseline regression. In Columns (1)–(5), the dependent variables are total income, property income, wage income, operating income, and transfer income. Levels of significance: *** 1%, ** 5%, * 10%. Robust standard errors are reported in parentheses.

reports the estimations of Equation (1) on the impact of digital skills on rural household income. Column (1) presents the baseline fixed-effects result, each additional digital skill is associated with an increase of CNY 1678 in total household income, a magnitude that is both economically and statistically significant, providing initial support for H1. Relative to the sample mean income (CNY 40,844), this implies an effect of roughly 4.1% per additional skill. To probe heterogeneity, we decompose digital skills. Columns (2)–(4) report results for the three domains, and column (5) further examines all skill items. Columns (5) shows that online business skills yield the largest gains, households possessing these skills earn, on average, CNY 11,264 more, and approximately 27.6% of mean income. By contrast, entertainment-social skills are negatively associated with income when other skill domains are held constant. This pattern suggests that time and budget allocated to entertainment or purely social online activities may crowd out productive uses of digital technology, yielding limited productivity benefits once “productive” skill sets are accounted for.

These findings are consistent with a broader literature documenting the transformative role of digitalization in restructuring economic activity and improving household welfare. Prior studies emphasize macro-level drivers as contributors to rural income growth, such as digital village programs, inclusive digital finance, and information-infrastructure roll-outs [48,49,50]. Our results complement this perspective with micro-level evidence, households’ own digital capabilities, particularly online business skills, are critical for converting the digital ecosystem into income gains. As digital diffusion lowers traditional frictions in moving goods and information between rural and urban markets [11,19], the returns to online business competencies rise through broader market reach, stronger distribution channels, and greater scope for product and activity diversification.

4.1.2. Endogenous Analysis

Table 3. Effects of digital skills on rural households’ income (IV method).

Variables	First-Stage	Second-Stage
	(1)	(2)
Digital skills		5.8667 *** (0.7010)
IV 2	−0.0003 *** (0.0001)
IV 1	0.4043 *** (0.0085)
Control variables	YES	YES
Regional FE	YES	YES
Time FE	YES	YES
Underidentification test		1789.1090
Weak identification test		1121.9490
Overidentification test		0.4730
Observations	8670	8670

Notes: Table 3 shows the estimation results of the instrumental variable method. In this case, the dependent variable in the Columns (1) is numerical skills, and the dependent variable in the Columns (2) is total farmer income. Levels of significance: *** 1%.

reports two-stage least squares estimate of the effect of digital skills on rural households’ income. In the first stage, the dependent variable is the digital-skills index; in the second stage, it is household income. Before interpreting coefficients, we assess instrument validity and strength. The overidentification test yields a Hansen–J statistic of 0.473 with a p-value > 0.10, so we fail to reject the null that the instruments are jointly exogenous. The Cragg–Donald Wald F-statistic equals 1121.949, which far exceeds conventional thresholds (e.g., 10), ruling out weak-instrument concerns. After addressing endogeneity, digital skills have a positive and statistically significant effect on household income at the 1% level, reaffirming H1. The 2SLS coefficient is larger than the baseline FE estimate, a pattern consistent with downward bias in the FE specification, thereby strengthening the case for a positive causal effect under the IV assumptions.

Because the acquisition of digital skills is conditional on digital access and reflects household choice, estimates may suffer from self-selection on observables. We therefore implement propensity score matching. Households that have mastered at least one digital skill constitute the treatment group ($D_i=1$); those with no digital skills form the control group ($D_i=0$). Propensities are estimated using observed covariates described in Section 3.2.3, and matching is performed within the region of common support. We report the average treatment effect on the treated (ATT), using four algorithms, including nearest-neighbor matching (k = 4), caliper matching, radius matching, and kernel matching. As shown in

Table 4. Effects of digital skills on rural households’ income (PSM method).

Matching Methods	ATT	Bootstrap S.E.	Z-Stat
Near neighbor matching (k = 4)	8.2228 ***	2.2638	3.63
Caliper Matching	8.2909 ***	2.2676	3.66
Radius Matching	8.5881 ***	2.0808	4.13
Kernel Matching	8.3837 ***	1.9760	4.24

Notes: Levels of significance: *** 1%.

, ATT estimates are positive and statistically significant at the 1% level across all four methods, reinforcing the conclusion that digital skills raise rural household income.

4.1.3. Robustness Tests

To strengthen the scientific validity of the baseline results, this study conducts a series of robustness checks, including alternative measures of digital skills and income, sample restrictions, and additional data validation.

First, alternative measure of the independent variable. We proxy digital skills with digital attitude, the perceived importance of digital technology in learning, work, socializing, entertainment, and online business, aggregated analogously to the skills index [51]. Column (1) of

Table 5. Results of robustness tests.

Variables	Robustness 1	Robustness 2	Robustness 3	Robustness 4	Robustness 5
	(1)	(2)	(3)	(4)	(5)
Digital skills		0.3571 ** (0.1566)	1.2036 ** (0.5269)	1.6008 *** (0.5824)	4.0203 *** (0.8002)
Digital attitude	2.2240 *** (0.6606)
Control variables	YES	YES	YES	YES	YES
Regional FE	YES	YES	YES	YES	YES
Time FE	YES	YES	YES	YES	NO
Observations	9257	9257	8212	8129	3302

Notes: Table 4 shows the results of the three robustness tests. Columns (1) and (2) are the two-stage least squares estimation results. Column (1) replaces the independent variable with numerical attitudes. Column (2) replaces the dependent variable as household income per capita. Column (3) excludes sample farmers in Internet-developed regions. Robust standard errors are reported in parentheses. Levels of significance: *** 1%, ** 5%.

shows a positive, 1%-significant effect on household income, indicating that our findings are not sensitive to how digital capability is measured and supporting H1.

Second, alternative measure of the dependent variable. Because total income scales with household size, we replace the dependent variable with per-capita household income. Column (2) indicates that digital skills remain positively associated with income at the 1% level, corroborating the baseline.

Third, exclusion of households from highly digitalized regions. To reduce concerns that external digital supply conditions drive results, we drop households in the five most digitalized provinces (Guangdong, Beijing, Shanghai, Zhejiang, Fujian). Column (3) shows a positive, 1%-significant effect, suggesting that regional digital intensity is not the sole driver of our results.

Fourth, retention to high-quality samples. To mitigate measurement error from rushed interviews, we retain observations with CFPS “urgency to finish” scores < 4 (on a 7-point scale). Column (4) again shows a positive, 1%-significant effect, reinforcing the accuracy of the baseline estimates.

Finally, validation using CFPS2020 data. Given rapid technological change, we replicate the analysis using CFPS2020. Although the questionnaire items differ, we construct a four-dimension skills index (entertainment, learning, social, shopping) and keep other variables consistent with prior specifications. Column (5) reports a positive, 1%-significant effect, reaffirming the baseline results and further supporting H1.

Across all checks, the estimated effects remain positive and statistically significant at the 1% level, strengthening the conclusion that digital skills raise rural household income.

4.2. Heterogeneity Analysis

This paper assesses group differences in the impact of digital skills on rural households’ incomes by analyzing heterogeneity across three dimensions. (1) Education. Educational attainment, an important dimension of human capital, is used to split the sample at nine years of schooling (≤9 vs. >9). This allows us to test whether returns to digital skills depend on households’ educational foundation. (2) Income. While digitalization has boosted rural economies, it may also generate a new rural dualism. To examine distributional implications, we estimate subgroup regressions for low- and high-income households using the sample mean of household income as the cutoff. (3) Digital environment. Because external conditions shape the returns to capability, we classify regions into high- and low- digitization areas based on the mean value of the regional internet development level and compare effects across the two groups.

4.2.1. Heterogeneity of Education

If human capital exhibits diminishing marginal returns, schooling may attenuate the incremental payoff to basic digital skills. Consistent with this mechanism,

Table 6. Digital skills and rural households’ income: the heterogeneity of education.

Variables	Low-Education	High-Education
	(1)	(2)
Digital skills	2.0155 *** (0.5616)	−1.7387 (2.1041)
Control variables	YES	YES
Regional FE	YES	YES
Time FE	YES	YES
Observations	8141	1116

Notes: Table 5 shows the heterogeneity of education. Column (1) is a subsample with a low education level (education years less than or equal to 6), and Column (2) is a subsample with a high education level (education years more than 6). Robust standard errors are reported in parentheses. Levels of significance: *** 1%.

shows that the effect of digital skills is positive and statistically significant for the low-education group (≤9 years) but statistically insignificant for the high-education group (>9 years). This pattern suggests that digital-skills formation complements, rather than substitutes for, formal education, helping less-educated households capture digital dividends and raise income.

The channels are intuitive. Individuals with lower schooling are typically disadvantaged in the labor market and face fewer employment and entrepreneurial opportunities. By acquiring stronger digital skills, they can match to better jobs via digital platforms and improve career progression [12], and initiate entrepreneurial activities online [26], thereby increasing both wage and operating income. By contrast, higher-educated households already possess substantial human capital from formal schooling; for them, basic digital skills function mainly as auxiliary tools, so the marginal contribution to income is smaller. For this group, income growth likely hinges more on the deepening of professional or occupation-specific competencies than on the breadth-type digital skills analyzed here.

4.2.2. Heterogeneity of Income

Table 7. Digital skills and rural households’ income: the heterogeneity of income.

Variables	Low-Income	High-Income
	(1)	(2)
Digital skills	0.6233 ** (0.2905)	0.9976 (1.2092)
Control variables	YES	YES
Regional FE	YES	YES
Time FE	YES	YES
Observations	6042	3215

Notes: Table 6 analyses the effect of digital skills on households at different income levels using panel quantile regressions. Column (1) shows the analysis results on the 10 quantiles. Column (2) is 50 quantiles, and Column (3) is 90. Robust standard errors are reported in parentheses. Levels of significance: ** 5%.

reports subgroup regressions by income. Digital skills have a positive and statistically significant effect on the low-income group, whereas the effect for the high-income group is comparatively weak. This pattern is consistent with an inclusive development interpretation: strengthening digital capabilities appears especially consequential for poverty reduction in rural areas.

The mechanisms mirror those in the education analysis. Low-income households are more exposed to “information poverty” arising from geographic isolation and thin social networks, which pushes them toward traditional farming and fragmented off-farm work [27]. By lowering the costs of information acquisition and resource coordination, digital skills help these households identify better jobs and integrate into broader markets, thereby raising both operating and wage income. By contrast, higher-income households rely more on non-digital assets, which can substitute for basic digital competencies, such as resource endowments, social ties, and established business networks. As a result, the marginal payoff to the breadth-type skills examined here is limited for the high-income group; effects may be larger for more advanced, task-specific digital capabilities not captured by our index.

4.2.3. Heterogeneity of the Digital Environment

Table 8. Digital skills and rural households’ income: the heterogeneity of digital environment.

Variables	Low-Digitization	High-Digitization
	(1)	(2)
Digital skills	1.8522 ** (0.6092)	2.7543 ** (1.0965)
Control variables	YES	YES
Regional FE	YES	YES
Time FE	YES	YES
Observations	6375	2882

Notes: Table 7 shows the heterogeneity of the digital environment. Column (1) is a subsample with a low digitization level (digital inclusion index is less than or equal to mean), and Column (2) is a subsample with a high digitization level (digital inclusion index is above mean). Robust standard errors are reported in parentheses. Levels of significance: ** 5%.

examines how the regional digital environment moderates the income returns to digital skills. Digital skills are positively associated with household income in both settings, but the effect is substantially larger in high-digitization regions than in low-digitization regions. This pattern indicates a complementarity between individual capability and external supply conditions: a stronger digital environment amplifies the payoff to digital skills.

The mechanism is intuitive. By providing infrastructure and ecosystem services, such as broadband coverage, digital payments, platform intermediation, the regional digital environment lowers search, matching, and enforcement costs, thereby converting digital competencies into market reach and monetizable opportunities. Conversely, underdeveloped environments constrain the value realization of digital skills [52] (Bai et al., 2024). Policy wise, investments that upgrade local digital infrastructure and platform ecosystems are likely to be skill-complementary, enhancing the income effects of household digital upskilling.

4.3. Mechanism Analysis: Allocation of Production Factors

Guided by the theoretical framework, digital skills should raise income by improving the allocation of land, capital (credit), and labor via information effects. To test these channels, we follow the mechanism-testing approach in Jiang (2022) [53], widely applied in economics [4,37,41]. Based on data availability, we construct binary indicators for: land transfer-out, land transfer-in, access to formal credit, access to informal credit, nonfarm employment, and nonfarm entrepreneurship. We estimate probit models for each mechanism variable using the same covariates as in the baseline. Results are reported in

Table 9. Mechanism analysis.

Variables	Land Transfer Out	Land Transfer In	Formal Credit	Informal Credit	Non-Farm Jobs	Entrepreneurship
	(1)	(2)	(3)	(4)	(5)	(6)
Digital skills	0.0012 (0.0054)	−0.0000 (0.0058)	0.0113 ** (0.0048)	0.0127 * (0.0067)	0.0259 *** (0.0059)	0.0118 *** (0.0039)
Control variables	YES	YES	YES	YES	YES	YES
Regional FE	YES	YES	YES	YES	YES	YES
Time FE	YES	YES	YES	YES	YES	YES
Observations	9257	9257	9257	9257	9257	9257

Notes: Table 9 shows the effect of digital skills on the allocation behavior of the three main factors of production of farmers. In Columns (1) and (2), the dependent variables are land shifting out or not and land shifting in or not (dummy variables, with shifting behavior = 1, otherwise equal = 0). In Columns (3) and (4), the dependent variable is credit status, which includes both formal and informal credit (dummy variable, with borrowing behavior = 1, otherwise = 0), with formal credit being bank loans and informal credit referring to borrowing on behalf of relatives, friends, etc. The dependent variable in Column (5) is also a dummy variable indicating whether the individual is engaged in nonfarm work (dummy variable, having nonfarm work = 1, otherwise = 0), including entrepreneurial activities, working outside the home, and other work of non-agricultural production operations character. Robust standard errors are reported in parentheses. Levels of significance: *** 1%, ** 5%, * 10%.

.

4.3.1. Land Factor Allocation

Columns (1)–(2) show that digital skills have no statistically significant effect on either land transfer-out or land transfer-in. Hence, we find no evidence that digital skills increase property income through leasing-out activity or raise operating income by facilitating scale expansion. In short, the land-allocation channel is not supported in our data.

4.3.2. Capital Factor Allocation

Columns (3)–(4) indicate that digital skills significantly raise the likelihood of obtaining both formal and informal credit. On the supply side, inclusive digital finance has expanded access to formal lending; on the demand side, digital competencies lower participation, search, and enforcement costs, improving take-up. In parallel, stronger digital engagement can broaden social networks and information flows, easing access to informal credit. These results support the credit-access channel.

4.3.3. Labor Factor Allocation

Columns (5)–(6) show that digital skills significantly increase the probability of nonfarm employment and of engaging in nonfarm entrepreneurship. This points to a reallocation of labor from traditional agriculture toward higher-return off-farm activities, with corresponding gains in wage and operating income. These findings support the labor-reallocation channel.

Taken together, the evidence suggests that digital skills raise rural household income primarily through credit and labor reallocation rather than through land adjustment.

5. Further Discussion

5.1. Digital Skills and Rural Households’ Income Sources

The composition of income is tightly linked to the urban–rural income gap, which widens when rural income sources are skewed—most notably because rural households derive far less property income than urban households. Following the National Bureau of Statistics (NBS) classification, we decompose rural income into four components: (1) property income, (2) wage income, (3) operating income, and (4) transfer income.

Table 10. Effect of digital skills on rural households’ income sources.

Variables	Total Income	Property Income	Wage Income	Operating Income	Transfer Income
	(1)	(2)	(3)	(4)	(5)
Digital skills	1.6782 *** (0.5160)	0.0052 (0.0245)	1.2873 ** (0.5095)	0.4437 * (0.2575)	0.0175 (0.1753)
Control variables	YES	YES	YES	YES	YES
Regional FE	YES	YES	YES	YES	YES
Time FE	YES	YES	YES	YES	YES
Observations	9257	9257	9257	9257	9257

Note: Table 10 shows the impact of digital skills on different sources of rural households’ income, including property income, wage income, operational income, and transfer income, Column (1) shows the total income, which serves as the baseline result. Robust standard errors are reported in parentheses. Levels of significance: *** 1%, ** 5%, * 10%.

reports fixed-effects estimates for each component.

Column (1) indicates a positive but statistically insignificant association between digital skills and property income, consistent with the mechanism results showing no effect on land transfers. By contrast, Columns (2)–(3) show that digital skills significantly raise wage and operating income. These patterns align with the labor-reallocation and entrepreneurship channels: stronger digital capabilities facilitate transitions out of agriculture, improve job matching, and support business formation, thereby expanding earned income. Column (4) shows no significant effect on transfer income, which is largely policy-determined and not directly shaped by household skills [54].

Taken together, the evidence suggests that digital skills optimize income composition by shifting rural households toward endogenous, productivity-linked sources—wages and operating income—rather than exogenous transfers. Such compositional upgrading is consistent with a narrowing of the urban–rural income gap [11,23].

5.2. Digital Skills and Rural Households’ Income Inequality

Advancing common prosperity ultimately requires closing development gaps, with income inequality at the core. The preceding analyses focused on mean income effects; here we examine distributional implications by relating digital skills to within-rural income inequality. Group-level inequality is commonly summarized by the Gini coefficient, while individual inequality is captured through relative deprivation. Following Liu and Yi (2023) [55], we compute an individual-level Kakwani index that measures each household’s relative deprivation and then aggregate as needed. We construct these measures for total income and for each income component.

Table 11. Effect of digital skills on rural households’ income inequality.

Variables	Total Income Gap	Property Income Gap	Wage Income Gap	Operating Income Gap	Transfer Income Gap
	(1)	(2)	(3)	(4)	(5)
Digital skills	−0.0089 ** (0.0037)	0.0016 (0.0027)	−0.0128 *** (0.0043)	−0.0092 ** (0.0046)	0.0068 (0.0049)
Control variables	YES	YES	YES	YES	YES
Regional FE	YES	YES	YES	YES	YES
Time FE	YES	YES	YES	YES	YES
Observations	9257	9257	9257	9257	9257

Note: Table 11 presents the estimated results of the effect of digital skills on the rural households’ income inequality in terms of total income, property income, wage income, operational income, and transfer income. Rural households’ income inequality is calculated using an index based on the village level, taking values of 0–1. Robust standard errors are reported in parentheses. Levels of significance: *** 1%, ** 5%.

shows the impact of digital skills on different types of income inequality. Column (1) shows that digital skills are associated with a significant reduction in total-income inequality among rural households. Thus, beyond raising average income, digital capabilities appear to have an equalizing effect within rural areas. For income components, Columns (2)–(5) indicate that digital skills reduce inequality in operating and wage income, but have no statistically detectable effect on property or transfer income disparities. A plausible interpretation is that gains in property income depend on prior asset accumulation; consequently, households with larger asset endowments capture a disproportionate share of any technology-enabled returns [54,56]. Transfer income, being policy-determined, is likewise less sensitive to household capabilities.

In sum, digital skills are linked to narrower gaps in total, operating, and wage income, consistent with an inclusive-growth channel operating through labor-market and household-business activities, rather than through property or transfer income.

6. Conclusions and Policy Implications

Using CFPS panel data from 2014 and 2018, this study examines how digital skills affect rural households’ income. The study finds that digital skills raise total household income, and each additional skill is associated with an increase of CNY 1678. By skill type, online business skills have the largest effect, followed by work–study skills, while entertainment–social skills are negatively associated with income. Heterogeneity analyses indicate larger gains for households with lower educational attainment and lower income, also showing that a stronger regional digital environment amplifies these effects. Mechanism tests point to factor reallocation toward the nonfarm sector, via higher probabilities of off-farm employment and entrepreneurship and improved access to formal credit, as the primary pathway; we find no effect on land allocation. Consistent with these channels, digital skills increase wage and operating income and reduce inequality in these components as well as in total income but have no detectable effect on property or transfer income or their dispersion.

Our study has two important policy implications. The first is linked to investing in people via targeted digital upskilling. Because returns are largest for low-education and low-income households, policy should prioritize basic and applied digital training for disadvantaged groups, with practical modules in online commerce, job searching, and enterprise tooling. In particular, online business skills merit emphasis given their outsized income effects and scope for self-employment and market expansion. The second is linked to investing in place and skill-complementary digital ecosystems. The regional digital environment is strongly complementary to household capabilities. Continued “digital village” initiatives can lower search, matching, and enforcement costs and translate skills into monetizable opportunities, including broadband and mobile coverage, e-commerce logistics, digital payment rails, and inclusive digital finance. Expanding accessible, low-friction credit products for digitally capable households will further catalyze productive investment.

This article has some limitations. First, constrained by available items, our digital-skills measure breadth rather than depth. Future work should integrate multi-source data to construct richer skill indices and re-evaluate income effects, such as behavioral traces, task-based assessments. Second, we focused on income outcomes; subsequent research should explore broader welfare margins—quality of life and subjective well-being—especially as entertainment devices diffuse. Third, we did not deploy emerging empirical tools, for example, machine learning estimators and new heterogeneity frameworks [57,58]. Incorporating these methods could sharpen identification and uncover nuanced treatment heterogeneity.