As the digital economy becomes a new engine driving China’s high-quality economic development, how to bridge the digital divide between urban and rural areas through institutional innovation and unleash the digital dividends of rural areas has become a key focus of national strategy. In May 2019, the General Office of the Central Committee of the Communist Party of China and the General Office of the State Council issued the Digital Rural Development Strategy Outline, officially elevating DVC to a national strategy. The strategy established a top-level design centered on improving rural information infrastructure and promoting the digital transformation of agricultural production, and explicitly proposed stimulating endogenous rural vitality by establishing New Farmers and New Technology Entrepreneurship and Innovation Centers and cultivating distinctive e-commerce brands. Under this macro-level strategic guidance, digital rural construction quickly entered the substantial implementation phase. In July 2020, the Central Cyberspace Affairs Commission, the Ministry of Agriculture and Rural Affairs, and six other departments jointly issued the Notice on Carrying out National Digital Rural Pilot Work, and in October of the same year, they officially announced the list of 117 pilot counties covering 31 provinces. This large-scale and extensive pilot work not only constitutes institutional exploration at the national level but also brings a typical exogenous policy shock to the development of local economies, thus providing an ideal quasi-natural experimental scenario for this paper.

Specifically, each pilot region, based on the overall deployment and in accordance with its own resource endowment, has carried out construction work in two key dimensions, which also constitutes the practical path for digital technology to empower agricultural development. On the one hand, the pilot work focused on constructing an integrated rural information infrastructure combining space, air, and land, promoting the comprehensive coverage of fiber broadband, 5G networks, and the agricultural IoT, and building digital public service platforms. For example, Jilin Province and Longshi City in China created the Jindalai Silk Road public service center, which facilitated the online aggregation of agricultural product supply and sales data. The improvement of such digital infrastructure effectively broke down the information barriers between urban and rural areas, reduced the costs of information search, and provided a low-cost physical foundation for rural labor to access external employment information and engage in return-to-hometown entrepreneurship. On the other hand, the policy focused on promoting the deep integration of digital technology with agricultural production and management, accelerating the transformation of traditional agriculture into smart agriculture. Taking Xingwen County in Sichuan Province as an example, local authorities actively utilized IoT monitoring devices, breeding traceability systems, and agricultural machinery apps to automate and digitize agricultural production processes. This policy-driven technological supply and service model innovation directly promoted the iteration and upgrading of agricultural equipment. Through “machine substitution” and “precision operations,” it accelerated the profound transformation of agricultural production from labor-intensive to capital and technology-intensive methods.

DVC, as an exogenous technological shock, serves as a crucial engine for driving the growth of agricultural TFP (Lu et al., 2024; Ye et al., 2021). One of its core mechanisms lies in releasing the structural dividend of labor, breaking the low-level equilibrium of traditional agriculture (Lewis, 1954). Specifically, DVC guides rural labor toward higher productivity sectors through the dual mechanisms of information smoothing and industrial extension. Crucially, the expansion of these employment opportunities allows rural residents to diversify their income sources, leading to a significant convergence in urban–rural income levels (Adamopoulos et al., 2024; Xu et al., 2025).

First, the penetration of DVC reduces information asymmetry and search friction, enhancing the accessibility of non-farm jobs. In the traditional dual economy structure, the information barriers between urban and rural areas are the main factors hindering the effective transfer of labor (Pidduck and Kelemen, 2025). Digital rural construction significantly lowers the marginal cost for rural labor to access external market information (Yuan and Fu, 2025). This information empowerment allows surplus labor trapped in low-productivity agricultural sectors to be precisely matched with urban industrial jobs (Che et al., 2025). As rural laborers access these off-farm opportunities, their wage income increases, directly narrowing the income gap with urban residents, a key indicator of successful labor market integration (Adamopoulos et al., 2024; Li X. et al., 2024).

Second, DVC creates space for non-agricultural employment by extending industrial chains, accelerating the professional transformation of labor (Couture et al., 2021; Zhao et al., 2024). The deep integration of the digital economy with rural industries creates a large number of service and skilled non-agricultural jobs within the county (Qin et al., 2023; Zhao et al., 2024). This allows farmers to achieve professional transformation through inter-industry leapfrogging (Chen et al., 2024). The creation of these high-value opportunities facilitates the exit of surplus labor from agriculture, optimizing the land-to-labor ratio.

The optimization of labor allocation ultimately drives the growth of agricultural TFP through the de-involution effect (Lewis, 1954). According to the Lewis dual economy theory, traditional agriculture has long been trapped in the overcrowding dilemma, where excessive labor input leads to diminishing marginal returns (Knight et al., 2011). The labor transfer induced by DVC is essentially a de-redundancy process. As surplus labor exits, the remaining labor force experiences a significant improvement in per capita resource endowment (Shen et al., 2025). This optimization implies that the narrowing urban–rural income gap is a robust signal of the effective utilization of non-farm employment opportunities (Yang et al., 2016).

Based on the above theoretical analysis, this paper proposes the following research hypothesis:

Another core mechanism through which DVC enhances agricultural TFP lies in stimulating the production technical intensification effect (Hayami and Ruttan, 1971). According to the theory of induced technological change, in response to changes in relative factor prices caused by labor loss, agricultural production methods inevitably seek capital to replace labor (Yangchen et al., 2025). However, due to the indivisibility of traditional agricultural capital and the fragmentation of land, this substitution process often lags behind (Wang et al., 2020). Digital rural construction, through the dual paths of smart equipment supply and socialized service configuration, breaks the barriers to factor substitution and drives the transformation of agricultural production towards a high-tech, intensive model (Qiu et al., 2021).

On the supply side, DVC promotes the embodied upgrade of technological factors, improving the marginal quality of capital input (Yangchen et al., 2025). The deep integration of digital technology with agricultural machinery has led agricultural capital goods through an intergenerational leap from mechanization to intelligentization (Barnes et al., 2019). The incremental equipment used in agricultural production now integrates high-precision navigation and sensing systems (Kolady et al., 2021). This embodied technological progress means that the unit mechanical power no longer merely represents the increase in physical work but also represents the enhancement of data processing and decision-making capabilities. This embodied technological progress implies that modern capital deepening is not merely an accumulation of iron and steel, but a qualitative leap in energy-driven technical density (Sheng et al., 2020).

On the usage side, DVC reshapes the allocation pattern of technological factors, reducing the transaction costs of intensive farming (Qing et al., 2019). The agricultural socialized service system built on digital platforms effectively addresses the scale mismatch between small farmers and advanced technological equipment (Shen et al., 2025). Through algorithmic matching and sharing models, digital technology enables the servitization and separation of ownership of large agricultural machinery, significantly reducing the search friction and institutional barriers for farmers to access machinery services (Liu et al., 2023). This allows small farmers, who traditionally engage in dispersed farming, to become part of the socialized division of labor network (Qiu et al., 2021). At the micro level, this is reflected in a significant increase in agricultural mechanization intensity, i.e., the improvement in technical intensiveness.

This technology intensification, led by digitalization, structurally drives the growth of agricultural TFP through the precise operation path. Unlike traditional extensive input, the technology intensification enabled by digital technology features significant precision. The application of smart equipment allows agricultural production to break free from experience-based constraints, enabling real-time monitoring and precise control of inputs. This precision agriculture model not only reduces redundant inputs to achieve cost savings, but also improves output stability through standardized operations, thereby directly transforming the increase in mechanization intensity into the growth of agricultural TFP. Based on the above theoretical analysis, this paper proposes the following research hypothesis:

This paper investigates the impact of the implementation of the DVC on agricultural TFP using a Difference-in-Differences (DID) model. The model Equation (1) is as follows:

Where i represents the county, and t represents the year. The dependent variable TFP_it represents the agricultural TFP in county i during year t; the core explanatory variable DVC is a dummy variable representing the implementation of the Digital Village Policy at the county level; controls_it represents the set of control variables; η_i and μ_t represent county and year fixed effects, respectively; ϵ_it is the random error term. Given the potential correlation of residuals and heteroscedasticity that may cause estimation bias, all standard errors are clustered at the county level.

The data used in this study are derived from three main sources. First, county-level socioeconomic panel data spanning from 2009 to 2023 are obtained from the China County Statistical Yearbook. During the data cleaning process, we excluded samples with severe missing values or unidentifiable administrative codes, and imputed minor missing data using linear interpolation. Second, the list of the National Digital Village Pilot Areas is compiled from the official notification published by the relevant government authorities. Third, the terrain ruggedness data used for the endogeneity test were derived from the 1:1,000,000 public version of the basic geographic information dataset provided by China’ s National Catalogue Service for Geographic Information. Finally, Digital Financial Inclusion Index of China for districts and counties comes from the Digital Finance Research Center at Peking University (with some district and county data missing). Furthermore, to mitigate the potential influence of outliers on the regression results, all continuous variables are winsorized at the 1st and 99th percentiles.

Before presenting the regression results, this subsection provides a brief descriptive analysis of the Digital Village Pilot and the evolution of agricultural TFP. According to the official list of the National Digital Village Pilot, the pilot counties are widely distributed across China, covering eastern, central, and western regions, and involving different administrative types such as counties, county-level cities, and urban districts. This broad spatial and institutional coverage suggests that the policy was not selectively implemented in a specific type of region, providing a suitable basis for empirical analysis. Figure 1 plots the evolution of average agricultural TFP for pilot and non-pilot counties over time. Several noteworthy patterns emerge. First, prior to the implementation of the Digital Village Pilot policy, the average TFP levels of the treatment and control groups follow broadly similar trends, with no persistent divergence between the two groups. Second, after 2020, when the Digital Village Pilot policy was formally launched, agricultural TFP in pilot counties exhibits a more pronounced upward trajectory, while the growth of TFP in non-pilot counties remains relatively moderate. As a result, a visible gap between the two groups gradually emerges in the post-policy period. It should be emphasized that this descriptive evidence does not imply a causal relationship but provides an intuitive overview of the data patterns. These preliminary observations motivate the subsequent difference-in-differences analysis, which formally examines whether the observed divergence in agricultural TFP can be causally attributed to digital village construction.

Table 1 reports the baseline regression results regarding the impact of digital village construction on agricultural TFP. To ensure the robustness of our estimates, we employ a stepwise regression strategy. Specifically, Column (1) presents the results from a Pooled OLS regression without controlling for any covariates or fixed effects. The estimated coefficient of the DVC is significantly positive at the 1% level, preliminarily suggesting a positive effect of digital village construction on agricultural productivity. Column (2) augments the model in Column (1) by incorporating Year Fixed Effects and County Fixed Effects. This specification controls for time-variant macro shocks and time-invariant regional heterogeneity. The results show that the coefficient of DVC remains significantly positive, indicating the importance of controlling for individual and time effects to isolate the net policy effect. Column (3) presents our preferred specification, which further incorporates a set of control variables into the model from Column (2) to mitigate potential omitted variable bias. The results indicate that, after controlling for factors such as natural endowment, economic development, and industrial structure, the estimated coefficient of digital village construction is statistically significant at the 1% level. In summary, across all model specifications, the coefficient of the core explanatory variable consistently remains significantly positive. This provides robust empirical evidence supporting our hypothesis that DVC acts as a significant driver for improving agricultural TFP.

The validity of the DID model relies on the Parallel Trends Assumption, which requires that, in the absence of policy intervention, the agricultural TFP of the treatment and control groups would have followed the same evolutionary trajectory. To rigorously verify this assumption and capture the dynamic effects of digital village construction, we employ the Event Study Approach by estimating the Equation (2):

Where DVCit represents the interaction term between the treatment group dummy and the year dummy for year j. The coefficient βj captures the difference in agricultural TFP between pilot and non-pilot counties in year j relative to the base period. The summation ranges from the beginning of the sample period (2009) to the end (2023). To avoid strict multicollinearity, we exclude the year immediately prior to the policy implementation (2019) as the reference base period.

Figure 2 plots the estimated coefficients βj along with their 95% confidence intervals. The results reveal two key findings: First, regarding the pre-treatment trends: For the years prior to the policy implementation (j < 2020), the estimated coefficients βj fluctuate around zero and are statistically insignificant. This indicates that there were no systematic differences in agricultural productivity trends between the pilot and non-pilot areas before the policy shock, thereby validating the parallel trends assumption. Second, regarding the dynamic policy effects: A structural break is observed after 2020. The coefficients for the post-treatment years (j > 2020) exhibit a clear upward trajectory and become statistically significant. It is worth noting that the effect in the implementation year (2020) is positive but not statistically significant, suggesting a time-lag effect. This is consistent with the reality that digital infrastructure construction requires an accumulation cycle before translating into productivity gains.

To ensure the reliability of our baseline findings, we conduct a series of robustness checks organized by the potential identification threats they address, including selection bias, omitted variable bias, measurement errors, and random chance.

The baseline results establish that digital village construction significantly enhances agricultural TFP. To uncover the underlying drivers of this causality, we explore two potential channels: structural transformation of labor and capital deepening through mechanization. We regress the outcome of these two mechanism variables on the core explanatory variable, with the results reported in Table 8.

Given the significant spatial disparities across Chinese regions in terms of geographic location, administrative hierarchy, market environment, and digital infrastructure, the empowering effect of digital village construction on agricultural TFP may exhibit asymmetric characteristics. To identify the boundary conditions for policy effectiveness and potential digital divides, we conduct sub-sample regressions based on distinct heterogeneity dimensions.