Research on the Impact of Digital Rural Construction on Agricultural Economic Resilience

Chengkun Liu; Xiqin Pan

doi:10.20965/jaciii.2026.p1389

single-jc.php

« previous

JACIII Vol.30 No.5 pp. 1389-1403

(2026)

doi: 10.20965/jaciii.2026.p1389

Research Paper:

Views over last 60 days: 21

Research on the Impact of Digital Rural Construction on Agricultural Economic Resilience

Chengkun Liu^† and Xiqin Pan

School of Statistics and Data Science, Jiangxi University of Finance and Economics
No.168 East Shuanggang Road, Nanchang, Jiangxi 330013, China

^†Corresponding author

Received:

January 25, 2026

Accepted:

March 30, 2026

Published:

September 20, 2026

Keywords:

digital rural development, agricultural economic resilience, double machine learning model

Abstract

The resilience of the agricultural economy and the process of rural revitalization are substantially strengthened through digital rural development. Comprehensive indicator systems for digital rural development and agricultural economic resilience are constructed using inter-provincial panel data from China between 2011 and 2023. The cross-entropy unbiased weighting approach is adopted to quantify development levels, and a double machine learning framework is applied to examine the association between the two. First, agricultural economic resilience generally exhibits an upward trend across provinces, with comparatively higher levels concentrated in regions such as Northeast and North China. Second, digital rural development contributes substantially to agricultural economic resilience, primarily through improvements in agricultural labor productivity, innovation capacity, and industrial structural upgrading. Third, the effects vary across regions, with the western region experiencing a more pronounced effect than other areas; the effect is greater during 2017–2023 than during 2011–2016; and non-major agricultural provinces benefit more than major agricultural provinces. It is recommended to further promote digital rural development, expand the utilization of digital technologies across the agricultural sector, and adopt region-specific strategies to enhance agricultural economic resilience and provide solid support for rural revitalization as well as the modernization of agriculture and rural areas.

Close full text

Cite this article as:

C. Liu and X. Pan, “Research on the Impact of Digital Rural Construction on Agricultural Economic Resilience,” J. Adv. Comput. Intell. Intell. Inform., Vol.30 No.5, pp. 1389-1403, 2026.

Data files:

1. Introduction

Agriculture underpins the national economy, and its capacity to withstand risks is essential for shaping a new development paradigm and safeguarding food security. The 20th Communist Party of China (CPC) National Congress report explicitly proposed building China into an agricultural powerhouse, while the 2025 Central Document No.1 emphasized the centrality of agriculture, rural areas, and farmers. Currently, agriculture faces multiple challenges, including natural conditions and market fluctuations. Enhancing its economic resilience is key to advancing agricultural modernization and securing the supply of agricultural product. The development of digital technologies presents new opportunities for agriculture. Tools like artificial intelligence and blockchain are supporting the implementation of the Digital China strategy and driving high-quality agricultural development. As a vital component of rural revitalization, digital village construction has become a key engine for agricultural modernization. The 14th Five-Year Plan formally incorporated digital villages into the national development agenda, while the 20th CPC National Congress highlighted the deepening application of digital technologies in the real economy, thereby outlining a clear pathway for strengthening agricultural economic resilience through digital villages. The intrinsic relationship between digital villages and agricultural economic resilience is examined in the present research. It constructs a multidimensional evaluation system, applies the unbiased cross-entropy weighting method for measurement, and employs dual machine learning models to systematically investigate the effects, underlying mechanisms, and heterogeneity of digital village development on agricultural economic resilience.

2. Literature Review

The literature associated with this research is grouped into three categories. The first aspect focuses on the concept and measurement of resilience. Holling originally introduced resilience to evaluate an ecosystem’s ability to recover after a disturbance ¹. Building on this ecological foundation, resilience thinking soon extended beyond its original disciplinary boundaries. Reggiani et al. ² applied it to the economic domain, while Martin and Sunley ³ further defined economic resilience as encompassing resistance, recovery, reconstitution, and renewal capacities, laying the foundation for agricultural economic resilience research. Additionally, scholars have constructed evaluation frameworks from various dimensions, such as establishing comprehensive frameworks around economic stability, governance efficiency, market efficiency, and social development. Liu et al. assessed China’s macroeconomic resilience based on risk absorption intensity and duration ⁴. Parallel to these macroeconomic investigations, the resilience lens was increasingly turned toward sector-specific applications. As research progressed, Folke systematically reviewed resilience concepts and applied them to the agricultural sector, defining agricultural economic resilience as the capacity of farming systems for coping with external shocks, including natural disasters, policy changes, and market fluctuations ⁵. Domestic studies of agricultural economic resilience began relatively late, and evaluation systems are still under development. Jiang et al. established an evaluation framework based on the pressure-state-response (PSR) model, focusing on three dimensions: resistance-recovery, adaptation-regulation, and transformation-innovation ⁶. Zhao and Xu constructed a system centered on resistance, adaptation, and transformation capabilities ⁷.

The determinants of agricultural economic resilience are addressed in the second research category. In recent years, related studies have gradually increased, unfolding at both micro and macro levels. At the micro level, Zhang and Jiao utilized China Family Panel Studies panel data from 2012 to 2018 to examine the relationship between agricultural insurance, agricultural total factor productivity, and household economic resilience, grounded in poverty trap and nonlinear dynamics theories ⁸. Shifting from risk-coping mechanisms to growth-enabling factors, subsequent micro-level studies have explored the role of financial inclusion. Li et al. investigated how digital inclusive finance influences the economic resilience of 1,021 new agricultural business entities across nine provinces (autonomous regions) in China ⁹. Guo et al. constructed a framework to assess the economic resilience of apple growers, using 2,005 survey responses collected from the main apple-producing regions of Shaanxi, Gansu, and Shanxi provinces on the Loess Plateau ¹⁰. They systematically examined the impact of e-commerce participation on their economic resilience and its operational mechanisms. Synthesizing these threads, recent work has begun to unravel the spatiotemporal dynamics and technological underpinnings of resilience evolution. Yang et al. revealed that agricultural economic resilience exhibits fluctuating growth trends, significant regional disparities, and multi-level differentiation, with agricultural technological innovation significantly enhancing agricultural economic resilience ¹¹. Agricultural labor productivity ¹² and technological innovation with scale management ¹³ also enhance resilience.

The third category of research examines how digital village development influences agricultural and rural progress. The swift development of the digital economy, driven by next-generation information technologies, is becoming a key factor in strengthening agricultural economic resilience. Bai argues that equipping rural governance with digital technologies offers a crucial avenue for advancing the modernization of agriculture and rural areas ¹⁴. Li and Zhao investigated how digital village initiatives affect Chinese-style agricultural modernization, finding that they significantly advance this process ¹⁵. Probing deeper into the underlying mechanisms, researchers have begun to unpack how digitalization transmits its effects through various channels. Similarly, Liu conducted related research and further explored the mechanisms through which digital village initiatives shape agricultural and rural modernization ¹⁶. Cao and Wang demonstrated that digital village initiatives substantially advance high-quality agricultural development, with agricultural technological innovation serving as a positive mediating factor ¹⁷. Zhang et al. found that digital rural development strengthens the robustness of agricultural industrial chains, with technological innovation in agriculture acting as a major driver ¹⁸. Thus, digital rural development holds significant potential for strengthening agricultural economic resilience.

Literature shows limited direct research on how digital rural initiatives affect agricultural economic resilience. This paper contributes by constructing a measurement system to analyze this relationship, using dual machine learning to identify labor productivity, industrial upgrading, and innovation as key mechanisms, and exploring heterogeneous effects across region, time, and agricultural development level to inform policy.

3. Indicator System Construction and Measurement Methodology

3.1. Indicator System Construction

This study constructs an indicator system specifically for digital rural construction, capturing the application of digital technologies in rural areas for agriculture, governance, and livelihoods—distinct from general urban-focused digital economy indicators.

Two indicator systems—one representing digital rural development and the other reflecting agricultural economic resilience—are constructed in this study to systematically examine the effects and mechanisms through which digital rural development influences agricultural economic resilience.

To ensure the measurability and comparability of the results, the following principles were primarily followed in selecting indicators for each dimension. Firstly, the principle of scientific rigor: indicators are closely aligned with the core concepts of digital rural development and agricultural economic resilience, grounded in theoretical and empirical evidence to scientifically and systematically reflect actual development levels—a fundamental prerequisite for ensuring valid comparability. Secondly, the principle of systematization: the indicator system comprises multiple dimensions, each containing several interrelated indicators. Emphasis is placed on logical consistency within dimensions and synergistic complementarity between dimensions to construct a clear-layered, comprehensive integrated evaluation framework. Thirdly, the principle of operability: indicator design must serve practical evaluation and policy application. Therefore, selected indicators should possess good data availability, feasible assessment methods, and comparability across regions or time periods.

Drawing on existing research and policy definitions of digital villages, a comprehensive evaluation framework is developed in this study, covering four key aspects: digital infrastructure investment, digital information infrastructure, digital industry development, and digital service levels. Investment in digital infrastructure is quantified by fixed-asset expenditures in transportation, warehousing, and postal services, along with the information transmission, computer services, and software sectors. Digital information infrastructure is represented by measures including the number of rural broadband subscribers, mobile phone and computer ownership per 100 households, postal service outlets, and agricultural meteorological observation stations. The development of the digital industry is measured using key indicators such as e-commerce sales and procurement volumes, the digital inclusive finance index, and the number of Taobao villages. Meanwhile, digital service levels are represented by rural delivery route length, the number of agricultural technicians, and per capita transportation and communication expenditures among rural residents. Based on this, a specific indicator system for digital rural development is constructed.

The indicators in Table 1 are derived from official statistical sources: rural broadband users, mobile phone/computer ownership per rural household, and rural delivery route length are from the National Bureau of Statistics and China Rural Statistical Yearbook; Taobao villages count is from AliResearch; agricultural technicians are from China Statistical Yearbook on Science and Technology; and the digital inclusive finance index is from the Peking University Digital Financial Inclusion Index. For provincial indicators that cannot be disaggregated to rural areas, we acknowledge this limitation and control for urban-rural differences in the econometric analysis.

Table 1. Digital countryside construction indicator system.

Primary indicators	Second-level indicators	Attribute
Digital construction funding investment	Fixed asset investment in transportation, warehousing, and postal services	\(+\)
Digital construction funding investment	Fixed asset investment in information transmission, computer services, and software	\(+\)
Digital infrastructure	Number of rural broadband access users	\(+\)
	Number of mobile phones per 100 households in rural areas at year-end	\(+\)
	Number of computers per 100 rural households at year-end	\(+\)
	Number of postal service outlets	\(+\)
	Number of agricultural meteorological observation stations	\(+\)
Digital industry development	E-commerce sales	\(+\)
	E-commerce procurement volume	\(+\)
	Digital inclusive finance index	\(+\)
	Number of Taobao villages	\(+\)
Digital service level	Length of rural delivery routes	\(+\)
	Number of agricultural technicians	\(+\)
	Per capita transportation and communication expenditure of rural households	\(+\)

Development of an Agricultural Economic Resilience Indicator System. Currently, academic research on agricultural economic resilience remains relatively scarce, with no clear and unified measurement standards or evaluation systems established. Building on the studies of Zhao and Xu ⁷ and Hao and Tan ¹⁹, this study employs the PSR model to develop an agricultural economic resilience indicator system. The PSR model, grounded in dynamic systems theory, is widely applied in ecological security and disaster risk assessments. Within this framework, the study establishes an agricultural economic resilience indicator system that integrates three key components: risk resistance, adaptive adjustment, and reconstructive innovation. Risk-resistance capacity assesses the agricultural system’s capability to sustain stability and mitigate risks under external shocks, specifically comprising three aspects: economic resilience, production resilience, and ecological resilience. Economic resilience focuses on the agricultural economic system’s ability to preserve core functions and stability amid market fluctuations and price risks. It is assessed using the proportion of primary industry output; the share of township and village retail sales of consumer goods; per capita added value of agriculture, forestry, animal husbandry, and fisheries; and added value per unit area—thus reflecting the system’s buffering capacity against economic disturbances. Production resilience is indicated by total grain output, the area under effective irrigation, total mechanical power per unit area, and the ratio of disaster occurrences to their impacts. This assesses how agricultural production conditions and infrastructure support system stability and recovery under risk scenarios. Ecological resilience measures the agricultural production system’s capacity to withstand fluctuations in chemical resource usage. This study quantifies it using fertilizer, diesel, pesticide, and agricultural film consumption per unit of effectively irrigated area. Adaptive adjustment capacity reflects the agricultural system’s self-repair and dynamic regulation capabilities following risk shocks. This study measures it through economic stability indicators, focusing on the system’s rebalancing process post-shock via income distribution, consumption structure, and industrial growth. Specifically, it is evaluated using the annual increase in primary industry value-added, rural residents’ average disposable income, average household consumption, and the Engel coefficient. Innovation capacity for restructuring denotes the agricultural system’s capability to transform and innovate, enabling self-renewal and sustainable development following shocks. This study quantifies it using investments in fixed assets across agriculture, forestry, animal husbandry, and fisheries; government spending on agriculture, forestry, and water management; the number of undergraduate and graduate students registered at higher agricultural institutions; and rural electricity consumption. Thus, the constructed agricultural economic resilience indicator system is presented in Table 2.

Table 2. Agricultural economic resilience index system.

Primary indicators	Second-level indicators	Tertiary indicators	Attribute
Risk-resisting ability (\(\mathit{RD}\))	\(\mathit{ER}\)	Share of primary industry	\(+\)
		Share of retail sales of consumer goods in towns and rural areas relative to total retail sales of consumer goods	\(+\)
		Added value of agriculture, forestry, animal husbandry, and fishery / (Rural population)	\(+\)
		Added value of agriculture, forestry, animal husbandry, and fishery / (Cropland area)	\(+\)
	\(\mathit{PR}\)	Total grain output	\(+\)
		Area under effective irrigation	\(+\)
		Average total mechanical power per unit area	\(+\)
		Area affected by disasters / Affected area	\(-\)
	\(\mathit{SR}\)	Agricultural fertilizer application rate per unit of effective irrigated area (converted to pure nutrient content)	\(-\)
		Agricultural diesel consumption per unit of effective irrigated area	\(-\)
		Pesticide application rate per unit of effective irrigated area	\(-\)
		Agricultural plastic film consumption per unit of effective irrigated area	\(-\)
Adaptability (\(\mathit{AD}\))	Economic stability (\(\mathit{ES}\))	Primary industry value-added growth rate	\(+\)
		Per capita disposable income of rural residents	\(+\)
		Per capita consumption expenditure of rural residents	\(+\)
		Engel’s coefficient for rural residents	\(+\)
Redesign innovation capabilities (\(\mathit{RI}\))	Innovation enhancement (\(\mathit{IE}\))	Fixed asset investment in agriculture, forestry, animal husbandry, and fisheries	\(+\)
		Fiscal expenditures on agriculture, forestry, and water affairs	\(+\)
		Number of undergraduate and vocational students enrolled in higher agricultural institutions	\(+\)
		Rural electricity consumption	\(+\)

Note: In this study, the Engel coefficient of rural residents is treated as a positive indicator of agricultural economic resilience. The reason is that regions with stronger agricultural economic resilience can better safeguard farmers’ income and food security when facing external shocks, thereby stabilizing the proportion of food consumption expenditure and avoiding sharp fluctuations and an abnormal rise in the Engel coefficient. Therefore, in the specific context of this study, a relatively high and stable Engel coefficient reflects the resilience of the agricultural economy in resisting risks and ensuring basic living standards.

3.2. Cross-Entropy Unbiased Weighting Method

In order to assess digital rural development and agricultural economic resilience, the study employs the cross-entropy unbiased weighting method. This method combines machine learning classification algorithms with econometric multiple logit regression. Compared to traditional entropy weighting, it handles extreme indicators more reasonably, yields more robust weights, provides richer information content, and ensures unbiased weighting and practicality. Since both concepts are multidimensional and comprehensive, this method effectively corrects the biases inherent in traditional entropy weighting. The procedure is outlined in the following steps:

Firstly, to ensure data comparability despite inconsistent measurement units and the presence of both positive and negative indicators, this study employs the cross-entropy unbiased weighting method. Raw data undergoes standardization, normalizing all indicator values to the 0–1 range. Eqs. (1) and (2) illustrate the calculation formulas for positive and negative indicators, respectively:

\begin{align} &\textrm{Positive indicator:} \quad &z_{ijt} &= \dfrac{x_{ijt} - x_{\mathrm{min}}}{x_{\mathrm{max}} - x_{\mathrm{min}}}, \label{eq:eq1} \tag{1} \end{align}

\begin{align} &\textrm{Negative indicator:} \quad &z_{ijt} &= \dfrac{x_{\mathrm{max}} - x_{ijt}}{x_{\mathrm{max}} - x_{\mathrm{min}}}. \label{eq:eq2} \tag{2} \end{align}

Among these, \(x_{ijt}\) and \(z_{ijt}\) represent the pre- and post-processing values of the \(i\)-th value for the \(j\) indicator in the \(t\) year, respectively. \(x_{\mathrm{max}}\) and \(x_{\mathrm{min}}\) denote the maximum and minimum values of the \(j\) indicator across the entire sample, respectively.

Secondly, calculate the indicator weight, i.e., compute the proportion of indicator \(j\) within the entire indicator set \(z_{ijt}\): \(p_{ijt}\). The method for calculating indicator weights is consistent for both positive and negative indicators, as shown in Eq. (3):

\begin{equation} p_{ijt} = \dfrac{r_{ijt}}{\displaystyle \sum_{t=1}^{m} \sum_{i=1}^{k} z_{ijt}}. \label{eq:eq3} \tag{3} \end{equation}

Thirdly, calculate the cross-entropy value of the \(j\) indicators, as shown in Eq. (4):

\begin{equation} \mathit{ce}_{j} = -\sum_{t=1}^{m} \sum_{i=1}^{k} \sum_{j=1}^{n} \bigl(p_{ijt} \times \ln p_{ijt}\bigr). \label{eq:eq4} \tag{4} \end{equation}

Fourthly, assuming the cross-entropy values of each indicator follow a normal distribution, standardize them using Eq. (5):

\begin{equation} \mathit{std{\_}ce}_{j} = \dfrac{\mathit{ce}_{j} - \mathit{mean}(ce)}{\mathit{std}(ce)}. \label{eq:eq5} \tag{5} \end{equation}

Fifthly, map the standardized cross-entropy values to the range \([0,1]\) using the sigmoid function, as shown in Eq. (6):

\begin{equation} \mathit{w{\_}ce}_{j} = \dfrac{1}{1 + \exp\bigl(-\mathit{std{\_}ce}_{j}\bigr)}. \label{eq:eq6} \tag{6} \end{equation}

Sixthly, calculate the weights for the indicators, as presented in Eq. (7):

\begin{equation} \mathit{cw}_{j} = \dfrac{1 - \mathit{w{\_}ce}_{j}}{\displaystyle n - \sum_{i=1}^{n} \mathit{w{\_}ce}_{i}}. \label{eq:eq7} \tag{7} \end{equation}

Finally, calculate the \(j\) indicator score for each province based on the product of \(p_{ijt}\) and \(\mathit{cw}_j\). Subsequently, determine the aggregate score for each indicator, which constitutes the final Digital Rural Development and Agricultural Economic Resilience Index.

4. Theoretical Analysis, Model Construction, and Variable Explanation

4.1. Theoretical Analysis

4.1.1. Direct Impact of Digital Rural Development on Agricultural Economic Resilience

Firstly, the digital economy demonstrates a certain degree of autonomy. Its “moat effect” helps buffer against economic fluctuations, thus strengthening the resilience of the agricultural sector ²⁰. Simultaneously, the application of digital technologies in rural areas effectively integrates resources. Through agglomeration effects, it optimizes agricultural industry layout and fosters the convergence of primary, secondary, and tertiary industries, thereby strengthening supply chain resilience and mitigating disruption risks. Furthermore, digital technologies drive the optimization of production-supply-sales structures and extend industrial chains, consolidating the stability of the agricultural industrial system. Additionally, channels such as e-commerce platforms and social media significantly enhance the circulation efficiency of agricultural products, enabling precise matching from production to sales. The application of big data technology assists farmers and agricultural enterprises in analyzing market demand and consumer behavior, breaking down information asymmetries between rural areas and external markets. This empowers farmers to accurately grasp market dynamics and rationally adjust production decisions. By reducing market risks caused by information lags, big data technology stabilizes agricultural income and enhances the risk-resistance capacity of the agricultural economic system. Accordingly, the research hypothesis put forward is as follows:

H1:

Promoting digital rural development enhances agricultural economic resilience.

4.1.2. Indirect Impact of Digital Village Development on Agricultural Economic Resilience

Firstly, digital rural development optimizes the allocation of rural production factors, effectively guiding the flow of labor and capital toward rural areas to support agricultural modernization ²¹. This facilitates the transformation of traditional farmers into new professional farmers while enhancing the efficiency of land resource allocation and production vitality. Consequently, it significantly boosts agricultural labor productivity. Furthermore, the application of digital technologies enables precise monitoring and management of agricultural production, driving mechanization and automation. This not only improves operational efficiency and reduces costs but also enhances agriculture’s flexibility in responding to market changes, thereby enabling agriculture to better absorb external shocks.

Secondly, digital rural development can elevate innovation levels, thereby significantly impacting the resilience of the agricultural economy ²². By establishing digital agricultural technology innovation platforms and integrating research resources, it accelerates agricultural R&D and the commercialization of research outcomes, driving the industry toward intelligent and high-end upgrades. When facing external shocks, strong innovation capabilities help regions leverage their research advantages to discover new growth points, facilitating rapid recovery and adaptation of the agricultural economy post-crisis, thereby enhancing its resilience.

Finally, digital rural development has promoted the optimization and upgrading of industrial structures. As proposed by Zeng et al. ²³ argue that digital rural development has profoundly optimized the environment for rural industrial growth, creating favorable conditions for agricultural technology R&D and experimentation. Robust digital infrastructure not only facilitates the broad adoption of agricultural technologies but also promotes the modernization of traditional agricultural sectors, thus improving production efficiency and overall competitiveness. At the same time, emerging agricultural industries, including agritourism, have expanded, bringing fresh vitality to the rural economy. By optimizing industrial structures, the agricultural economy has achieved diversified and balanced development, reducing reliance on single industries while markedly improving the resilience and stability of the agricultural economy. Furthermore, digital villages drive the thorough integration of new business models with agricultural industries, providing digital momentum for agricultural supply-side structural reforms. This strengthens the internal structural resilience and self-repair capabilities of the agricultural economy. Hence, we propose the following hypothesis:

H2:

Digital rural development may enhance agricultural labor productivity, thereby promoting the resilience of the agricultural economy.
H3:

Digital rural development can enhance innovation levels, thereby boosting agricultural economic resilience.
H4:

Digital rural development can drive industrial structure upgrading, thus strengthening the resilience of the agricultural economy.

4.2. Construction of Dual Machine Learning Models

4.2.1. Benchmark Regression Model

Dual machine learning, proposed in 2018, is a flexible and efficient causal effect estimation method particularly suited for this study on agricultural economic resilience. Agricultural economic resilience is influenced by complex social factors requiring control of high-dimensional variables, while traditional methods are susceptible to multicollinearity and the curse of dimensionality. Dual machine learning integrates machine learning with regularization techniques to automatically screen variables and mitigate estimation biases. Furthermore, agricultural systems often exhibit nonlinear relationships; traditional methods relying on predefined functional forms are prone to specification errors, whereas dual machine learning effectively captures nonlinear characteristics. Based on the assumptions of conditional independence and sample partitioning, dual machine learning also alleviates endogeneity issues to yield consistent estimates.

Accordingly, a dual machine learning model is utilized to examine how digital rural development affects agricultural economic resilience. To ensure the validity of the conditional independence and sample partitioning independence assumptions, the following measures were adopted in the empirical design: Firstly, multidimensional control variables pertinent to agricultural economic resilience—such as water conservancy infrastructure and urbanization rates—were introduced to maximize fulfillment of the “no confounding” condition. Secondly, K-fold cross-validation (\(K=5\)) is employed to randomly partition the sample into training and testing sets, effectively guaranteeing data independence and preventing information leakage.

To begin with, a partially linear dual machine learning model was constructed as follows:

\begin{align} &\mathit{AE}_{it} = \alpha \mathit{DV}_{it} + g\bigl(X_{it}\bigr) + \mu_{it}, \label{eq:eq8} \tag{8} \end{align}

\begin{align} &E \bigl(\mu_{it} \mid \mathit{DV}_{it}, X_{it}\bigr)=0, \label{eq:eq9} \tag{9} \end{align}

where \(i\) denotes province; \(t\) denotes year; \(\mathit{AE}\) represents the dependent variable, the agricultural economic resilience index; \(\mathit{DV}\) represents the core independent variable, the level of digital rural development; \(\alpha\) denotes the treatment coefficient, the primary focus of this study; \(X\) is the set of high-dimensional control variables, whose specific form \(\hat{g}(X_{it})\) requires estimation via machine learning algorithms; and \(\mu\) is the error term with conditional mean of zero.

Secondly, the following auxiliary regression model is developed to optimize convergence efficiency, aiming to ensure the treatment coefficient estimator satisfies unbiasedness under finite sample conditions:

\begin{align} &\mathit{DV}_{it} = m\bigl(X_{it}\bigr) + \nu_{it}, \label{eq:eq10} \tag{10} \end{align}

\begin{align} &E\bigl(\nu_{it} \mid X_{it}\bigr) = 0, \label{eq:eq11} \tag{11} \end{align}

where \(m(X_{it})\) represents the regression function of the treatment variable on the high-dimensional control variables, and \(\nu_{it}\) denotes the error term with conditional mean of zero.

The specific estimation steps are as follows: Firstly, use the machine learning model to obtain the estimated value \(m(X_{it})\) in Eq. (10), yielding the estimated residual term \(\hat{\nu}_{it}\). Secondly, similarly use the machine learning model to obtain the estimated value \(\hat{g}(X_{it})\) in Eq. (8). Finally, obtain the unbiased estimator \(\alpha\) as shown in Eq. (12):

\begin{equation} \hat{\alpha} = \left(\dfrac{1}{n} \sum \hat{\nu}_{it} \mathit{DV}_{it}\right)^{-1} \dfrac{1}{n} \sum \hat{\nu}_{it} \left(\mathit{AE}_{it} - \hat{g}\bigl(X_{it}\bigr)\right). \label{eq:eq12} \tag{12} \end{equation}

4.2.2. Mechanism Testing Model

To examine the mechanism through which digital rural development enhances agricultural economic resilience, a dual machine learning model is utilized ²⁴. Drawing on Jiang ²⁴, the following mechanism testing model is constructed:

\begin{align} &M_{it} = \beta \mathit{DV}_{it} + g\bigl(X_{it}\bigr) + \varepsilon_{it}, \label{eq:eq13} \tag{13} \end{align}

\begin{align} &E\bigl(\varepsilon_{it} \mid \mathit{DV}_{it}, X_{it}\bigr) = 0, \label{eq:eq14} \tag{14} \end{align}

\begin{align} &\mathit{DV}_{it} = m\bigl(X_{it}\bigr) + \nu_{it}, \label{eq:eq15} \tag{15} \end{align}

\begin{align} &E\bigl(\nu_{it} \mid X_{it}\bigr) = 0, \label{eq:eq16} \tag{16} \end{align}

\begin{align} &\hat{\beta} = \left(\dfrac{1}{n} \sum \hat{\nu}_{it} \mathit{DV}_{it}\right)^{-1} \dfrac{1}{n} \sum \hat{\nu}_{it} \left(M_{it} - \hat{g}\bigl(X_{it}\bigr)\right), \label{eq:eq17} \tag{17} \end{align}

where \(M\) denotes the mechanism variable in the present study.

Equation (13) estimates the effect of \(\mathit{DV}\) on \(M\), with \(g(X)\) capturing control variable effects via machine learning. Eq. (14) states the conditional independence assumption. Eq. (15) is an auxiliary regression modeling \(\mathit{DV}\) as a function of \(X\), producing residual \(\hat{\nu}_{it}\) that isolates exogenous variation in \(\mathit{DV}\). Eq. (16) provides the corresponding assumption. Eq. (17) gives the estimator for \(\beta\), obtained by regressing residualized \(M\) on residualized \(\mathit{DV}\), which removes confounding and ensures unbiased estimation.

Equations (15) and (16) serve to construct \(\hat{\nu}_{it}\), used in Eq. (17) to obtain an unbiased estimate of \(\beta\). Although Eqs. (14) and (16) both state conditional mean assumptions, they apply to different error terms (\(\varepsilon_{it}\) vs. \(v_{it}\)) and serve distinct purposes in the estimation procedure. This separation is essential in the double machine learning framework to avoid regularization bias and ensure valid causal inference.

4.3. Variable Descriptions

4.3.1. Dependent Variables and Explanatory Variables

The dependent variable and independent variables in this paper are the agricultural economic resilience index (\(\mathit{AE}\)) and the digital rural development index (\(\mathit{DV}\)), respectively, both calculated earlier using unbiased cross-entropy weighting.

4.3.2. Mechanism Variables

Drawing on the theoretical analysis above, three instrumental variables are selected:

(1)

Agricultural labor productivity (\(\mathit{ALP}\)), defined as the ratio of production output created by workers within a specific timeframe to the corresponding labor input. It is quantified using the ratio of primary industry value-added to the primary industry labor force across regions.
(2)

Innovation capacity (\(\mathit{IA}\)) is quantified based on the number of patent applications granted in each province, following the methodology of Li et al. ²⁵.
(3)

Industrial structure upgrading (\(\mathit{IND}\)), defined as the process of transitioning from a low-level to a high-level industrial structure, is quantified using the ratio of tertiary industry value-added to secondary industry value-added.

4.3.3. Control Variables

Beyond digital rural development, numerous other factors influence agricultural economic resilience. Drawing on Zhao and Xu ⁷ and Hao and Tan ¹⁹, the control variables selected for this study are as follows:

(1)

Water infrastructure (\(\mathit{WI}\)), measured by agricultural water consumption.
(2)

Power infrastructure (\(\mathit{EI}\)), measured by agricultural electricity consumption.
(3)

Transportation infrastructure (\(\mathit{TI}\)), measured by rural road density, defined as the ratio of road mileage to total population.
(4)

Ecological environment (\(\mathit{ECO}\)), measured by wetland area.
(5)

Urbanization rate (\(\mathit{UR}\)) is quantified using the ratio of urban population to total population. Urbanization rates may interact with digital rural development: the former is primarily driven by regional economic development and urbanization processes, while the latter’s advancement may be influenced by urbanization levels. Simultaneously, urbanization itself impacts agricultural economic resilience. Therefore, to mitigate potential reverse causality effects, the urbanization rate is treated as a control variable in the model.
(6)

Urban-rural income gap (\(\mathit{IG}\)) is quantified using the ratio of per capita disposable income between urban and rural residents. The urban-rural income gap may interact with digital rural development: the latter helps narrow the gap, while the former influences rural areas’ capacity to adopt and apply digital technologies. Given the strong endogeneity of the urban-rural income gap—which may be affected by digital rural development while itself impacting agricultural economic resilience—this gap is treated as a control variable in the model, with potential endogeneity issues addressed in the analysis.
(7)

Fiscal expenditure (\(\mathit{FE}\)), measured by general fiscal budget expenditure.

Based on data availability and validity, panel data from 31 provincial-level administrative regions in mainland China spanning 2011 to 2023 are employed to construct a comprehensive dataset covering economic, social, and technological dimensions. The required data primarily originates from the China Statistical Yearbook, China Rural Statistical Yearbook, Peking University Digital Inclusive Finance Index, and the EPS database. To ensure data quality, linear interpolation was applied for addressing isolated missing values within the dataset. For variables with extensive missing data, imputation was performed by combining data from other years and trend characteristics to maintain continuity and plausibility. Following interpolation, the processed data was cross-validated against the original data to ensure high consistency in trends and distributions. After preliminary processing of the original data, Table 3 presents the descriptive statistics of the variables.

Table 3. Descriptive statistics of variables.

Variable	Variable description	Sample size	Mean	Standard deviation	Minimum	Maximum
\(\mathit{AE}\)	Agricultural economic resilience	403	3.616	2.227	-1.727	16.252
\(\mathit{DV}\)	Digital rural development	403	2.894	1.604	0.335	10.187
\(\mathit{ALP}\)	Agricultural labor productivity [10,000 yuan/person]	403	3.555	2.035	0.632	15.509
\(\mathit{IA}\)	Innovation capacity [10,000 units]	403	7.381	11.854	0.012	87.221
\(\mathit{IND}\)	Industrial structure upgrade	403	1.295	0.732	0.518	5.690
\(\mathit{WI}\)	Water conservancy infrastructure [billion cubic meters]	403	121.169	105.830	2.500	563.600
\(\mathit{EI}\)	Electric power infrastructure [100 million kWh]	403	272.400	381.125	0.870	2011.000
\(\mathit{TI}\)	Transportation infrastructure [km/10,000 people]	403	47.317	48.071	5.129	336.197
\(\mathit{ECO}\)	Ecological environment [100 million mu]	403	164.913	195.110	3.436	814.360
\(\mathit{UR}\)	Urbanization rate	403	0.592	0.130	0.227	0.896
\(\mathit{IG}\)	Urban-rural income gap	403	2.538	0.384	1.794	4.857
\(\mathit{FE}\)	Fiscal expenditures [100 million yuan]	403	5477.833	3233.342	705.910	18533.080

5. Empirical Findings and Analysis

5.1. Measurement Results and Analysis

Based on the research methodology introduced earlier, we measured digital rural development and agricultural economic resilience. Taking the agricultural economic resilience measurement as an example, China’s 31 provinces experienced a steady increase in their agricultural economic resilience index from 2011 to 2023. Provinces with higher agricultural economic resilience indices are primarily concentrated in Northeast China and North China, while those with lower indices are mainly located in Western China. Furthermore, a significant gap exists between provinces with high and low agricultural economic resilience. Among them, Guizhou Province exhibits the fastest growth rate in its agricultural economic resilience index, while some provinces show relatively slow growth, indicating substantial room for improvement.

5.2. Benchmark Regression Results and Analysis

A dual machine learning model, based on the random forest algorithm with a \({1:4}\) sample split ratio, is utilized to assess how digital village construction affects agricultural economic resilience. To enhance result robustness, robust standard errors clustered at the provincial level were used to account for heteroskedasticity and individual fixed effects. Specific regression outcomes are reported in Table 4. Column (1) shows the benchmark regression model controlling for fixed effects of year and province, while including the first-order effect of the control variables. The findings show that digital rural development has a strong positive effect on agricultural economic resilience, with a regression coefficient of 0.735, significant at the 1% level. These findings suggest that digital rural development enhances agricultural economic resilience through improved resource allocation efficiency, deeper integration between agriculture and related industries, and greater market information symmetry. This evidence underscores the importance of digital rural development for strengthening agricultural economic resilience, thereby providing empirical support for research hypothesis H1. Considering the potential nonlinear relationships among variables in agricultural economic development, column (2) incorporates quadratic terms of control variables with column (1) serving as the baseline. Column (2) indicates that, after incorporating the quadratic terms, digital rural development continues to have a statistically positive association with agricultural economic resilience, with only negligible changes in coefficient magnitude. This indicates strong robustness of column (1)’s conclusions, further validating hypothesis H1. Evidence from the analysis shows that digital rural development affects agricultural economic resilience both through direct channels and via more complex nonlinear mechanisms.

Table 4. Baseline regression results.

Variable	(1) AE	(2) AE	(3) ER	(4) PR	(5) SR	(6) ES	(7) IP
\(\mathit{DV}\)	0.735\(^{***}\) (0.211)	0.736\(^{***}\) (0.222)	0.023\(^{*}\) (0.014)	0.023\(^{*}\) (0.014)	0.026 (0.027)	0.054\(^{*}\) (0.031)	0.098\(^{***}\) (0.031)
Control variable terms	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Control variable quadratic term	No	Yes	Yes	Yes	Yes	Yes	Yes
Provincial fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year fixed effect	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Constant term	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Sample size	403	403	403	403	403	403	403

Note: ***, **, and * indicate significance levels at the 1%, 5%, and 10% levels, respectively. The values in parentheses are robust standard errors.

Furthermore, to explore in greater depth how digital rural development relates to specific dimensions of agricultural economic resilience, columns (3)–(7) of Table 4 report the estimated results for economic resilience, production resilience, ecological resilience, economic stability, and innovation enhancement. Digital rural development shows a favorable association with both economic resilience and production resilience, with the effects being statistically significant at the 10% level. Moreover, its role in enhancing regional innovation capacity is even more pronounced, achieving statistical significance at the 1% level. This indicates that digital rural development not only effectively strengthens the risk resistance and production stability of agricultural economic systems but also significantly drives the improvement of regional innovation capabilities. Specifically, the effect of digital rural development on economic resilience is notably greater than its influence on production resilience or innovation enhancement. This may stem from digital rural development serving as a crucial breakthrough in traditional rural development. The comprehensive application of digital technologies in rural development substantially enhances farmers’ labor productivity and the overall efficiency of rural industries, thereby strongly promoting economic resilience. Regarding production resilience, while digital technologies can indirectly boost agricultural productivity, their role as specialized production factors makes it challenging to directly enhance risk-resistance capabilities within agricultural production processes, resulting in relatively weaker impacts. In terms of innovation enhancement, digital rural development significantly strengthens agricultural innovation capacity through establishing digital agricultural technology innovation platforms and promoting agricultural mechanization, thereby positively influencing agricultural economic resilience. However, innovation enhancement typically requires prolonged accumulation and sustained investment, limiting its short-term effectiveness. Furthermore, while digital rural development yields positive estimates for both ecological resilience and economic stability, these fail to pass significance tests. Regarding ecological resilience, despite promoting green and sustainable agricultural models, improvements remain limited due to insufficient technology adoption and the need to enhance farmers’ environmental awareness. Regarding economic stability, although digital rural development has improved the overall competitiveness of the agricultural economy, the intrinsic vulnerability of the sector and the complexity of the external environment indicate that enhancing economic stability likely requires sustained efforts and comprehensive policy support. Consequently, its impact has not yet reached a significant level in the short term.

5.3. Robustness Testing

To test robustness, this study conducted multiple analyses. Firstly, the sample scope was adjusted: based on digital rural development metrics from 2011 to 2023, four provinces—Jiangsu, Zhejiang, Shandong, and Guangdong (high values)—and four provinces—Hainan, Tibet, Qinghai, and Ningxia (low values)—were excluded. The remaining provincial samples were re-estimated, and the corresponding results are reported in column (1) of Table 5. As the second step, to mitigate the impact of outliers on parameter estimates, the original data underwent preprocessing using two-tailed trimming at the 1% and 5% percentiles. The adjusted regression outcomes are presented in column (2) of Table 5. Additionally, given the homogeneity of provinces within China’s governance system and the potential influence of internal policy environments and regional characteristics, a province-by-year interaction fixed effect was introduced into the baseline regression model. The adjusted regression outcomes are reported in column (3) of Table 5.

Table 5. Robustness test results.

Variable	(1)	(2)		(3)
Variable	Adjusted research sample	1% truncation	5% truncation	Include province-year interaction fixed effects
\(\mathit{DV}\)	0.724\(^{***}\) (0.246)	0.339\(^{*}\) (0.203)	0.254\(^{*}\) (0.155)	0.770\(^{***}\) (0.267)
Control variable term	Yes	Yes	Yes	Yes
Control variable quadratic term	Yes	Yes	Yes	Yes
Provincial fixed effects	Yes	Yes	Yes	Yes
Year fixed effects	Yes	Yes	Yes	Yes
Province \(\times\) year interaction fixed effect	No	No	No	Yes
Constant term	Yes	Yes	Yes	Yes
Sample size	299	403	403	403

Note: ***, **, and * indicate significance levels at the 1%, 5%, and 10% levels, respectively. The values in parentheses are robust standard errors.

Secondly, to validate the robustness of the dual machine learning model design, tests were conducted from three perspectives: adjusting the sample split ratio, replacing the machine learning algorithm, and constructing an interactive model. Firstly, the baseline sample split ratio of \({1:4}\) was modified to \({1:3}\) and \({1:8}\). Secondly, LASSO regression, gradient boosting (GBM), and neural networks were used to replace the random forest algorithm in the baseline model. Finally, an interactive dual machine learning model was constructed for re-estimation. Under all scenarios, digital rural development had a statistically significant favorable effect on agricultural economic resilience. Although the estimated coefficients fluctuated slightly, the core conclusions remained robust, validating the model’s reliability, with the test results reported in Table 6.

Table 6. Robustness test results of the double machine learning model.

Variable	(1)		(2)			(3)
	Changing the sample splitting ratio		Replace machine learning model			Interactive model
	\(\boldsymbol{1:3}\)	\(\boldsymbol{1:8}\)	LASSO regression	Gradient boosting algorithm	Neural network model	Interactive model
\(\mathit{DV}\)	0.563\(^{**}\) (0.271)	0.544\(^{***}\) (0.212)	0.733\(^{***}\) (0.254)	0.497\(^{**}\) (0.235)	0.139 (0.137)	1.684\(^{***}\) (0.167)
Control variable linear terms	Yes	Yes	Yes	Yes	Yes	Yes
Control variable quadratic term	Yes	Yes	Yes	Yes	Yes	Yes
Provincial fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Year fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Constant term	Yes	Yes	Yes	Yes	Yes	Yes
Sample size	403	403	403	403	403	403

Note: ***, **, and * indicate significance levels at the 1%, 5%, and 10% levels, respectively. The values in parentheses are robust standard errors.

5.4. Endogeneity Test

Given the multifaceted nature of factors influencing agricultural economic resilience, endogeneity issues are difficult to fully avoid in empirical research. To overcome endogeneity bias and enhance the credibility of research conclusions, this paper adopts the theoretical framework of Liu et al. ²⁶. Considering the characteristics of the panel data, the present research also references the methodology proposed by Nunn and Qian to construct a robust instrumental variable system ²⁷.

Specifically, the instrumental variable in this study is constructed as the interaction between the number of broadband access ports (in ten thousand) and the lagged values of digital rural development levels. This selection is grounded in the following theoretical rationale: broadband access ports, as a key indicator of digital infrastructure, are theoretically highly correlated with digital rural development. Simultaneously, the number of broadband access ports exhibits a significant positive association with digital rural development, satisfying the endogeneity requirement for instrumental variables. Moreover, the 2006 data on broadband access ports, being historical, lacks direct causality with the agricultural economic resilience development process from 2011 to 2023, effectively ensuring its exogeneity. To further mitigate estimation biases from reverse causality and omitted variables, this study incorporates the lagged digital rural development level into the instrumental variable system, enhancing the robustness and explanatory power of the instruments.

As shown in Table 7, after accounting for endogeneity, digital rural development retains a statistically meaningful effect on agricultural economic resilience. Firstly, the underidentification test results show the KP LM for columns (1) and (2) to be 83.401 and 86.742, rejecting the null hypothesis of instrumental variable underidentification at the 1% significance level. This confirms a significant correlation between the constructed instrumental variables and the core explanatory variable (digital rural development level), satisfying the relevance condition. Secondly, the weak instrumental variable test yields KP F of 71.025 (column (1)) and 98.250 (column (2)), both far exceeding the Stock–Yogo critical values at the 10% significance level (15.166 and 15.814). The null hypothesis of weak instrumental variables is thus rejected, fully verifying the validity of the selected instrumental variables with no weak identification issues. Since the model is exactly identified (one instrumental variable for one endogenous regressor), an overidentification test is not applicable. This conclusion aligns with theoretical expectations and robustly supports the benchmark regression results. Therefore, by constructing a reasonable and effective instrumental variable system and leveraging panel data characteristics for estimation, this study effectively mitigates the potential interference of endogeneity on research conclusions. It confirms the significant positive effect of digital rural development on agricultural economic resilience after effectively addressing the endogeneity problem of the model through a valid instrumental variable system, and the research findings have both theoretical value and policy implications.

Table 7. Endogeneity test results.

Variable	(1) IV: Constructing an interaction term between early 2006 broadband internet access ports and the digital rural development level one period prior	(2) IV: Digital rural development level one period lagged
\(\mathit{DV}\)	0.157\(^{**}\) (0.837)	0.134\(^{**}\) (0.065)
Control variable first-order term	Yes	Yes
Control variable quadratic term	Yes	Yes
Provincial fixed effects	Yes	Yes
Kleibergen-Paap rk LM statistic	83.401 [0.000]	86.742 [0.000]
Kleibergen-Paap rk Wald F statistic	71.025 {15.166}	98.250 {15.814}
Year fixed effects	Yes	Yes
Constant term	Yes	Yes
Sample size	372	372

Note: ***, **, and * indicate significance levels at the 1%, 5%, and 10% levels, respectively. The values in parentheses are robust standard errors.

5.5. Testing the Mechanism of Action

As demonstrated by the preceding theoretical analysis, digital rural development not only exerts direct impacts on agricultural economic resilience but may also influence it indirectly through pathways such as enhancing agricultural labor productivity, strengthening innovation capabilities, and optimizing industrial structures. Therefore, building upon the previously established mechanism testing model, a dual machine learning framework is utilized to examine the mechanisms through which digital rural development influences agricultural economic resilience. Specifically, to examine the multiple transmission pathways through which digital rural development affects agricultural economic resilience, this study adopts the LASSO regression model, referencing the research by Farbmacher et al. ²⁸. The findings are presented in Table 8.

Table 8. Mechanism test results.

Variable	(1) ALP	(2) IA	(3) IND
\(\mathit{DV}\)	1.326\(^{***}\) (0.528)	1.415\(^{***}\) (0.512)	1.415\(^{***}\) (0.509)
Control variable linear terms	Yes	Yes	Yes
Control variable quadratic term	Yes	Yes	Yes
Provincial fixed effects	Yes	Yes	Yes
Year fixed effects	Yes	Yes	Yes
Constant term	Yes	Yes	Yes
Sample size	403	403	403

Note: ***, **, and * indicate significance levels at the 1%, 5%, and 10% levels, respectively. The values in parentheses are robust standard errors.

The significant positive effects of DV on ALP, IA, and IND are confirmed. Drawing on established findings that these factors enhance agricultural economic resilience ¹²,¹³, these results support the mediating pathways proposed in H2–H4.

5.5.1. Agricultural Labor Productivity Enhancement Effects

Column (1) of Table 8 shows that digital rural development significantly enhances agricultural economic resilience through its indirect effect via increased agricultural labor productivity. The specific impact mechanism is as follows: the introduction of digital technologies (including mobile payments, e-commerce, and smart devices) optimizes the rural employment and investment environment, driving the transformation of traditional agriculture toward modernization and scale expansion, thereby directly improving production efficiency. For instance, deploying drones and the Internet of Things optimizes resource utilization. Concurrently, digital technologies enhance agriculture’s risk resilience by leveraging data analytics, predictive models (e.g., smart irrigation systems), and supply chain management technologies (e.g., blockchain) to mitigate natural disasters and market volatility, thereby reducing production costs. Moreover, increased agricultural labor productivity creates a “moat effect” in driving agricultural economic growth. This refers to the agricultural sector’s enhanced buffering capacity and resilience when facing external shocks. Therefore, digital rural development, by boosting agricultural labor productivity, not only improves output efficiency but also strengthens the resilience of the agricultural economy, validating research hypothesis H2.

5.5.2. Innovation Enhancement Effect

As shown in Table 8, column (2), digital rural development exerts a robust positive effect on innovation capacity, suggesting that it plays an important role in strengthening innovation. Moreover, the magnitude of the estimated coefficient indicates that digital rural development exerts the strongest promotional effect on innovation capacity. According to Cao and Wang ¹⁷, digital rural development primarily drives agricultural technological innovation by improving regional innovation ecosystems and optimizing innovation financing environments. On one hand, digital infrastructure development, including 5G networks and data centers, provides a more favorable environment for agricultural innovation entities. On the other hand, the proliferation of digital finance alleviates financing challenges for agricultural technological innovation, offering capital support for innovative activities. By encouraging the uptake and accessibility of digital finance in rural areas, digital village development provides more convenient financial support and services to agricultural innovation entities, further facilitating the smooth progress of agricultural technological innovation activities. Regarding the effect of technological innovation on economic resilience, industries engaged in relatively cutting-edge R&D demonstrate greater resistance to external shocks. Agricultural technological innovation not only enhances product quality and stabilizes economic growth but also strengthens the resilience of the agricultural economy. Therefore, digital village development significantly enhances agricultural economic resilience by boosting innovation capacity, thereby validating research hypothesis H3.

5.5.3. Industrial Structure Upgrading Effects

Column (3) of Table 8 shows that, similar to agricultural labor productivity and innovation capacity, digital rural development significantly promotes industrial structure upgrading. This finding demonstrates the strategic importance of digital rural development in advancing industrial optimization and upgrading. Firstly, grounded in the application of modern information technologies, digital rural initiatives substantially drive the growth of high-end manufacturing and service industries. The extensive adoption of digital technologies has led to a gradual reallocation of land and labor from the primary sector to the secondary and tertiary sectors, thus improving the industrial structure. Secondly, by enhancing farmers’ digital literacy, digital rural development further stimulates innovation and entrepreneurship in rural areas, injecting new momentum into rural economic growth. This catalytic effect not only drives the transformation of agricultural production toward scale, intensification, and digitization but also fuels the diversification of rural economies. Boschma indicates that the restructuring and advancement of industrial structure significantly enhances the economic system’s resilience by promoting diversified industrial development ²⁹. During economic recovery phases, industrial upgrading offers more diversified pathways for economic development. Consequently, it significantly enhances economic resilience. Drawing on the above analysis, digital rural development strengthens agricultural economic resilience by driving industrial upgrading, thereby validating research hypothesis H4.

5.6. Heterogeneity Analysis

After analyzing the impact of digital rural development on agricultural economic resilience and its underlying mechanisms, this section further investigates the heterogeneity of these effects across the following dimensions.

5.6.1. Regional Heterogeneity

Firstly, following the regional classification standards issued by the National Bureau of Statistics, China’s 31 provinces are categorized into four major regions: northeast, east, central, and west. heterogeneity analysis of the sample data is conducted accordingly, with results shown in column (1) of Table 9. Estimation results indicate that the influence of digital rural development on agricultural economic resilience differs markedly across regions: it is strongly positive in the western, northeast, and central regions, with a sequentially decreasing magnitude, whereas the effect in the eastern region is statistically insignificant. This phenomenon can be explained economically through the law of diminishing marginal returns. As China’s most economically dynamic region with the most complete industrial chains and strongest industrial foundations, the eastern region already possesses a high level of informatization and relatively advanced agricultural production methods. Consequently, as digital rural development progresses, its marginal contribution to enhancing agricultural economic resilience gradually diminishes, making it difficult to generate substantial new effects. Conversely, the Western region trails the other three regions in development, characterized not only by weaker infrastructure but also by predominantly traditional agricultural production methods. Consequently, the advancement of digital village construction has a stronger impact on boosting agricultural economic resilience in these regions. This regional disparity suggests that the marginal effects of digital village development vary considerably among different regions, mainly as a result of differences in development stages and initial conditions. This finding provides crucial reference for policymakers in implementing differentiated digital village development strategies across distinct regions.

Table 9. Heterogeneity test results.

Variable	(1)				(2)		(3)
Variable	Northeast region	Eastern region	Central region	Western region	2011–2016	2017–2023	Non-agricultural provinces	Agricultural provinces
\(\mathit{DV}\)	0.225\(^{*}\) (0.102)	0.095 (0.107)	0.186\(^{*}\) (0.078)	1.379\(^{**}\) (0.539)	0.601\(^{***}\) (0.231)	1.210\(^{***}\) (0.410)	0.889\(^{***}\) (0.326)	0.470\(^{***}\) (0.135)
Control variable linear terms	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Control variable quadratic term	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Provincial fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year fixed effect	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Constant term	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Sample size	39	130	78	156	186	217	273	130

Note: ***, **, and * indicate significance levels at the 1%, 5%, and 10% levels, respectively. The values in parentheses are robust standard errors.

5.6.2. Period Heterogeneity

Secondly, considering that General Secretary Xi Jinping formally proposed the initiative to advance the “Digital China” initiative in December 2015, the development process of digital villages before and after this proposal shows distinct differences. This suggests that digital village development may influence agricultural economic resilience in a phased manner. Therefore, the analysis reclassifies all samples into two periods—2011–2016 and 2017–2023—for heterogeneity testing, with outcomes displayed in column (2) of Table 9. The findings suggest that from 2017 to 2023, the enhancement of agricultural economic resilience through digital village development was markedly stronger than during 2011–2016. This divergence primarily stems from differences in the depth and breadth of policy implementation across the two phases. During 2011–2016, although China made certain progress in agricultural modernization and informatization, its contribution to enhancing agricultural economic resilience was relatively limited compared to subsequent digital rural development efforts. After 2016, however, the launch of the “Digital China” initiative propelled digital rural development into a phase of rapid advancement. Digital technologies became more deeply integrated into agricultural production, management, and distribution processes, thereby enhancing digital rural development’s contribution to agricultural economic resilience. This finding indicates that as policy intensity increased and digital technologies were more widely adopted, the marginal effects of digital rural development exhibited an upward trend, with policy outcomes becoming more fully realized in the later stages.

5.6.3. Heterogeneity in Agricultural Development Levels

Finally, the promotional effect of digital rural development on agricultural economic resilience may additionally depend on the level of agricultural development itself. Therefore, this study classifies 31 provinces based on differences in agricultural resource endowments, agricultural output scale, agricultural product quality and safety levels, and farmers’ income status, referencing Gao et al. ³⁰, which classifies China’s 31 provinces. Specifically, Heilongjiang, Shandong, Henan, Jiangsu, Hunan, Sichuan, Anhui, Jilin, Hebei, and Inner Mongolia are designated as major agricultural provinces, while the remaining 21 provinces are categorized as non-major agricultural provinces ²⁹,³⁰. This classification enables an analysis of the heterogeneity in agricultural development levels across these regions with respect to how digital rural initiatives affect agricultural economic resilience. Based on the estimation results in column (3) of Table 9, the effect of digital rural development on agricultural economic resilience is significantly positive for both agricultural and non-agricultural provinces, although the regression coefficients vary. Notably, the coefficient representing the effect on agricultural economic resilience is higher in non-agricultural provinces. This indicates that these provinces have comparatively lower agricultural economic resilience, and digital rural development has a stronger marginal effect in enhancing it. This disparity primarily stems from the relatively weaker agricultural infrastructure and more traditional production methods in non-agricultural major provinces. The application of digital technologies can thus more significantly improve the stability and risk-resistance capabilities of their agricultural economies. Consequently, the impact of digital rural development varies across regions with different agricultural development levels, offering important evidence for designing differentiated digital rural development policies suited to these diverse contexts.

6. Research Findings and Policy Recommendations

Drawing on panel data from China’s 31 provinces spanning 2011–2023, this study constructs indicator systems for digital rural development and agricultural economic resilience. It innovatively employs a dual machine learning model to examine how digital rural initiatives influence agricultural economic resilience. Firstly, it conducts theoretical analysis from both direct and indirect perspectives of digital rural development’s impact on agricultural economic resilience, innovatively introducing a dual machine learning model to investigate this relationship. Secondly, the dual machine learning model validates the transmission mechanism by which digital rural development enhances agricultural economic resilience, namely through improvements in agricultural labor productivity, reinforcement of innovation capacity, and optimization of industrial structure. Finally, heterogeneity analysis is conducted across three dimensions: region, time, and whether the province is an agricultural powerhouse. The study concludes as follows. Firstly, digital rural development has a significantly positive effect on agricultural economic resilience, with similarly strong positive influences on its sub-dimensions—economic resilience, production resilience, and innovation enhancement—especially in fostering innovation advancement. Secondly, findings from the mechanism test suggest that enhancing agricultural labor productivity, strengthening innovation capacity, and upgrading industrial structure are the three main pathways through which digital rural development fosters the growth of agricultural economic resilience. Thirdly, the heterogeneity analysis demonstrates that digital rural development significantly affects agricultural economic resilience, with notable spatiotemporal variability and regional disparities. Specifically, it generates a stronger beneficial effect on agricultural economic resilience in western regions during the 2017–2023 period and in provinces not dominated by agriculture. Therefore, to effectively enhance agricultural economic resilience and advance rural revitalization and agricultural modernization, the study puts forward the following policy recommendations.

Firstly, addressing digital infrastructure gaps to bolster its support for agricultural economic resilience. Priority should be given to increasing rural internet penetration and postal network coverage to lay the groundwork for digital technology adoption. Concurrently, targeted support for underdeveloped regions is essential to channel technology, talent, and capital toward areas with lower digitalization levels. Agricultural technology training and crop variety improvement should be implemented to tangibly enhance agricultural productivity and output quality. Furthermore, strengthen the coupling of digital technologies and agriculture by enhancing policy backing for “Internet Plus Agriculture.” Building digital distribution platforms and leveraging big data for market monitoring and forecasting will enhance agriculture’s risk-coping capacity, thereby systematically boosting the resilience of the agricultural economy.

Secondly, focus on three key pathways to inject sustained momentum into agricultural economic resilience. Firstly, enhance agricultural labor productivity through digital technologies by promoting smart agricultural machinery, digital breeding, and big data analysis to drive the transformation of agriculture toward digitalization and scale. Secondly, strengthen agricultural technological innovation capabilities by supporting R&D and adoption through diverse investment mechanisms, including attractive platforms for returning entrepreneurs, dedicated development funds, and tax incentives. Thirdly, advance industrial restructuring and upgrading by adjusting crop patterns, extending industrial chains, and promoting integration across primary, secondary, and tertiary sectors to elevate the overall agricultural value chain.

Thirdly, advance coordinated regional agricultural resilience tailored to local conditions. The influence of digital rural development differs markedly by region, time period, and development level, highlighting the need for differentiated approaches. To address regional disparities, Eastern regions should focus on technologies like the industrial internet and big data to incubate high-quality agricultural projects and leverage their radiating influence; northeastern and central regions can draw on eastern experiences to attract scientific talent and develop distinctive agricultural models; western regions should prioritize improving information infrastructure, using industrialization and urbanization to drive agricultural modernization. Addressing temporal differences requires seizing digital economic opportunities and policy dividends while sustaining investment. Regarding agricultural development disparities, major agricultural provinces can explore new production models empowered by digital technologies; non-agricultural provinces should deepen digital integration with secondary and tertiary industries to capture digital dividends.

References

[1] C. S. Holling, “Resilience and stability of ecological systems,” Annu. Rev. Ecol. Syst., Vol.4, pp. 1-23, 1973. https://doi.org/10.1146/annurev.es.04.110173.000245
[2] A. Reggiani, T. De Graaff, and P. Nijkamp, “Resilience: An evolutionary approach to spatial economic systems,” Netw. Spat. Econ., Vol.2, No.2, pp. 211-229, 2002. https://doi.org/10.1023/A:1015377515690
[3] R. Martin and P. Sunley, “On the notion of regional economic resilience: Conceptualization and explanation,” J. Econ. Geogr., Vol.15, No.1, pp. 1-42, 2015. https://doi.org/10.1093/jeg/lbu015
[4] X. Liu, X. Zhang, and S. Li, “Measurement of China’s macroeconomic resilience: A systemic risk perspective,” Soc. Sci. China, Vol.42, No.1, pp. 12-32+204, 2021 (in Chinese).
[5] C. Folke, “Resilience (republished),” Ecol. Soc., Vol.21, No.4, Article No.44, 2016. https://doi.org/10.5751/ES-09088-210444
[6] H. Jiang, C. Zhang, and H. Jiang, “Study on effect and mechanism of China’s agricultural economic resilience on agricultural high-quality development,” Agric. Econ. Manag., No.1, pp. 20-32, 2022 (in Chinese).
[7] W. Zhao and X. Xu, “Effect and mechanism of digital economy on the resilience of agricultural economy,” J. South China Agric. Univ. (Soc. Sci. Ed.), Vol.22, No.2, pp. 87-96, 2023 (in Chinese).
[8] D. Zhang and Y. Jiao, “Agricultural insurance, total factor productivity in agriculture and the economic resilience of farm households,” J. South China Agric. Univ. (Soc. Sci. Ed.), Vol.21, No.2, pp. 82-97, 2022.
[9] M. Li, X. Ke, K. Kang, and P. Guo, “Digital inclusive finance and economic resilience of new agricultural operators: Empowerment or negative energy?,” J. China Agric. Univ., Vol.29, No.4, pp. 40-53, 2024 (in Chinese).
[10] Y. Guo, X. Yan, and X. Huo, “Can e-commerce improve the economic resilience of apple farmers?—Evidence from the micro-survey in the Loess Plateau region,” J. Arid Land Resour. Environ., Vol.38, No.8, pp. 167-178, 2024 (in Chinese). https://doi.org/10.13448/j.cnki.jalre.2024.174
[11] H. Yang, Y. Zou, and J. Yuan, “A study of the impact of technological innovation on the resilience of the agricultural economy,” Collect. Essays Finance Econ. (J. Zhejiang Univ. Finance Econ.), No.4, pp. 29-40, 2025 (in Chinese). https://doi.org/10.13762/j.cnki.cjlc.2025.04.001
[12] A. A. Vărzaru, “The relationship between economic performance, sustainability, and agricultural productivity: Empirical evidence from the European Union,” Agriculture, Vol.16, No.2, Article No.217, 2026. https://doi.org/10.3390/agriculture16020217
[13] H. Zhang, J. Li, and T. Quan, “Strengthening or weakening: The impact of an aging rural workforce on agricultural economic resilience in China,” Agriculture, Vol.13, No.7, Article No.1436, 2023. https://doi.org/10.3390/agriculture13071436
[14] Q. Bai, “Digital technology empowering resilient rural governance and its action framework,” J. Fujian Norm. Univ. (Philos. Soc. Sci. Ed.), No.3, pp. 35-43+169, 2024 (in Chinese).
[15] L. Li and W. Zhao, “Mechanism and path of digital rural construction boosting the modernization of Chinese-style agriculture,” J. South China Agric. Univ. (Soc. Sci. Ed.), Vol.23, No.3, pp. 12-22, 2024 (in Chinese).
[16] Y. Liu, “The impact and mechanism of digital countryside construction on the modernization of agriculture and rural areas,” J. Shanxi Univ. (Philos. Soc. Sci. Ed.), Vol.47, No.2, pp. 152-160, 2024 (in Chinese). https://doi.org/10.13451/j.cnki.shanxi.univ(phil.soc.).2024.02.018
[17] J. Cao and Y. Wang, “Digital village, technology innovation, and high-quality agricultural development,” J. Manag., Vol.37, No.3, pp. 128-142, 2024 (in Chinese). https://doi.org/10.19808/j.cnki.41-1408/F.2024.0028
[18] Q. Zhang, Y. Li, M. Li, and N. Shan, “Research on the impact of digital rural construction on the resilience of agricultural industry chain,” J. Southwest Univ. (Nat. Sci. Ed.), Vol.47, No.3, pp. 129-140, 2025 (in Chinese). https://doi.org/10.13718/j.cnki.xdzk.2025.03.011
[19] A. Hao and J. Tan, “Empowering agricultural resilience by rural industrial integration: Influence mechanism and effect analysis,” J. Agrotech. Econ., Vol.2023, No.7, pp. 88-107, 2023 (in Chinese). https://doi.org/10.13246/j.cnki.jae.20220418.002
[20] J. Han and Y. Gao, “Measurement of industrial correlation effects of Chinese provincial digital economy development,” J. Quant. Tech. Econ., Vol.39, No.4, pp. 45-66, 2022 (in Chinese). https://doi.org/10.13653/j.cnki.jqte.2022.04.009
[21] Z. Li and J. Zhuang, “Internet usage, non-agriculture and agricultural machinery socialized services: Empirical analysis based on data from CLDS,” J. Agro-For. Econ. Manag., Vol.20, No.2, pp. 166-175, 2021 (in Chinese). https://doi.org/10.16195/j.cnki.cn36-1328/f.2021.02.18
[22] J. Li, L. Teng, H. Ma, and Y. Wang, “Spatial heterogeneity and influencing factors of agricultural economic resilience in Anhui Province,” East China Econ. Manag., Vol.36, No.11, pp. 75-84, 2022 (in Chinese). https://doi.org/10.19629/j.cnki.34-1014/f.220409001
[23] Y. Zeng, Y. Song, X. Lin, and C. Fu, “Some humble opinions on China’s digital village construction,” Chin. Rural Econ., No.4, pp. 21-35, 2021 (in Chinese).
[24] T. Jiang, “Mediating effects and moderating effects in causal inference,” China Ind. Econ., Vol.22, No.5, pp. 100-120, 2022 (in Chinese). https://doi.org/10.19581/j.cnki.ciejournal.2022.05.005
[25] X. Li, W. Wu, and G. Gu, “Does the development of digital economy improve regional economic resilience? A research based on intermediary effect, threshold effect and spatial spillover effect,” J. Zhejiang Univ. (Humanit. Soc. Sci.), Vol.52, No.12, pp. 21-39, 2022 (in Chinese).
[26] C. Liu, Y. Chen, and M. Li, “Digital rural construction empowering rural high-quality development: Theory and empirical evidence,” Contemp. Econ. Res., No.3, pp. 80-93, 2024 (in Chinese).
[27] N. Nunn and N. Qian, “US food aid and civil conflict,” Am. Econ. Rev., Vol.104, No.6, pp. 1630-1666, 2014. https://doi.org/10.1257/aer.104.6.1630
[28] H. Farbmacher, M. Huber, L. Lafférs, H. Langen, and M. Spindler, “Causal mediation analysis with double machine learning,” Econom. J., Vol.25, No.2, pp. 277-300, 2022. https://doi.org/10.1093/ectj/utac003
[29] R. Boschma, “Towards an evolutionary perspective on regional resilience,” Reg. Stud., Vol.49, No.5, pp. 733-751, 2015. https://doi.org/10.1080/00343404.2014.959481
[30] E. Gao, J. Zhu, and J. Zheng, “The impact of agricultural socialized services on the reduction of fertilizer: Double inspection based on panel data of 31 provinces in China,” Chin. J. Eco-Agric., Vol.31, No.4, pp. 632-642, 2023 (in Chinese). https://doi.org/10.12357/cjea.20220637

This article is published under a Creative Commons Attribution-NoDerivatives 4.0 Internationa License.

[B1] [1] C. S. Holling, “Resilience and stability of ecological systems,” Annu. Rev. Ecol. Syst., Vol.4, pp. 1-23, 1973. https://doi.org/10.1146/annurev.es.04.110173.000245

[B2] [2] A. Reggiani, T. De Graaff, and P. Nijkamp, “Resilience: An evolutionary approach to spatial economic systems,” Netw. Spat. Econ., Vol.2, No.2, pp. 211-229, 2002. https://doi.org/10.1023/A:1015377515690

[B3] [3] R. Martin and P. Sunley, “On the notion of regional economic resilience: Conceptualization and explanation,” J. Econ. Geogr., Vol.15, No.1, pp. 1-42, 2015. https://doi.org/10.1093/jeg/lbu015

[B4] [4] X. Liu, X. Zhang, and S. Li, “Measurement of China’s macroeconomic resilience: A systemic risk perspective,” Soc. Sci. China, Vol.42, No.1, pp. 12-32+204, 2021 (in Chinese).

[B5] [5] C. Folke, “Resilience (republished),” Ecol. Soc., Vol.21, No.4, Article No.44, 2016. https://doi.org/10.5751/ES-09088-210444

[B6] [6] H. Jiang, C. Zhang, and H. Jiang, “Study on effect and mechanism of China’s agricultural economic resilience on agricultural high-quality development,” Agric. Econ. Manag., No.1, pp. 20-32, 2022 (in Chinese).

[B7] [7] W. Zhao and X. Xu, “Effect and mechanism of digital economy on the resilience of agricultural economy,” J. South China Agric. Univ. (Soc. Sci. Ed.), Vol.22, No.2, pp. 87-96, 2023 (in Chinese).

[B8] [8] D. Zhang and Y. Jiao, “Agricultural insurance, total factor productivity in agriculture and the economic resilience of farm households,” J. South China Agric. Univ. (Soc. Sci. Ed.), Vol.21, No.2, pp. 82-97, 2022.

[B9] [9] M. Li, X. Ke, K. Kang, and P. Guo, “Digital inclusive finance and economic resilience of new agricultural operators: Empowerment or negative energy?,” J. China Agric. Univ., Vol.29, No.4, pp. 40-53, 2024 (in Chinese).

[B10] [10] Y. Guo, X. Yan, and X. Huo, “Can e-commerce improve the economic resilience of apple farmers?—Evidence from the micro-survey in the Loess Plateau region,” J. Arid Land Resour. Environ., Vol.38, No.8, pp. 167-178, 2024 (in Chinese). https://doi.org/10.13448/j.cnki.jalre.2024.174

[B11] [11] H. Yang, Y. Zou, and J. Yuan, “A study of the impact of technological innovation on the resilience of the agricultural economy,” Collect. Essays Finance Econ. (J. Zhejiang Univ. Finance Econ.), No.4, pp. 29-40, 2025 (in Chinese). https://doi.org/10.13762/j.cnki.cjlc.2025.04.001

[B12] [12] A. A. Vărzaru, “The relationship between economic performance, sustainability, and agricultural productivity: Empirical evidence from the European Union,” Agriculture, Vol.16, No.2, Article No.217, 2026. https://doi.org/10.3390/agriculture16020217

[B13] [13] H. Zhang, J. Li, and T. Quan, “Strengthening or weakening: The impact of an aging rural workforce on agricultural economic resilience in China,” Agriculture, Vol.13, No.7, Article No.1436, 2023. https://doi.org/10.3390/agriculture13071436

[B14] [14] Q. Bai, “Digital technology empowering resilient rural governance and its action framework,” J. Fujian Norm. Univ. (Philos. Soc. Sci. Ed.), No.3, pp. 35-43+169, 2024 (in Chinese).

[B15] [15] L. Li and W. Zhao, “Mechanism and path of digital rural construction boosting the modernization of Chinese-style agriculture,” J. South China Agric. Univ. (Soc. Sci. Ed.), Vol.23, No.3, pp. 12-22, 2024 (in Chinese).

[B16] [16] Y. Liu, “The impact and mechanism of digital countryside construction on the modernization of agriculture and rural areas,” J. Shanxi Univ. (Philos. Soc. Sci. Ed.), Vol.47, No.2, pp. 152-160, 2024 (in Chinese). https://doi.org/10.13451/j.cnki.shanxi.univ(phil.soc.).2024.02.018

[B17] [17] J. Cao and Y. Wang, “Digital village, technology innovation, and high-quality agricultural development,” J. Manag., Vol.37, No.3, pp. 128-142, 2024 (in Chinese). https://doi.org/10.19808/j.cnki.41-1408/F.2024.0028

[B18] [18] Q. Zhang, Y. Li, M. Li, and N. Shan, “Research on the impact of digital rural construction on the resilience of agricultural industry chain,” J. Southwest Univ. (Nat. Sci. Ed.), Vol.47, No.3, pp. 129-140, 2025 (in Chinese). https://doi.org/10.13718/j.cnki.xdzk.2025.03.011

[B19] [19] A. Hao and J. Tan, “Empowering agricultural resilience by rural industrial integration: Influence mechanism and effect analysis,” J. Agrotech. Econ., Vol.2023, No.7, pp. 88-107, 2023 (in Chinese). https://doi.org/10.13246/j.cnki.jae.20220418.002

[B20] [20] J. Han and Y. Gao, “Measurement of industrial correlation effects of Chinese provincial digital economy development,” J. Quant. Tech. Econ., Vol.39, No.4, pp. 45-66, 2022 (in Chinese). https://doi.org/10.13653/j.cnki.jqte.2022.04.009

[B21] [21] Z. Li and J. Zhuang, “Internet usage, non-agriculture and agricultural machinery socialized services: Empirical analysis based on data from CLDS,” J. Agro-For. Econ. Manag., Vol.20, No.2, pp. 166-175, 2021 (in Chinese). https://doi.org/10.16195/j.cnki.cn36-1328/f.2021.02.18

[B22] [22] J. Li, L. Teng, H. Ma, and Y. Wang, “Spatial heterogeneity and influencing factors of agricultural economic resilience in Anhui Province,” East China Econ. Manag., Vol.36, No.11, pp. 75-84, 2022 (in Chinese). https://doi.org/10.19629/j.cnki.34-1014/f.220409001

[B23] [23] Y. Zeng, Y. Song, X. Lin, and C. Fu, “Some humble opinions on China’s digital village construction,” Chin. Rural Econ., No.4, pp. 21-35, 2021 (in Chinese).

[B24] [24] T. Jiang, “Mediating effects and moderating effects in causal inference,” China Ind. Econ., Vol.22, No.5, pp. 100-120, 2022 (in Chinese). https://doi.org/10.19581/j.cnki.ciejournal.2022.05.005

[B25] [25] X. Li, W. Wu, and G. Gu, “Does the development of digital economy improve regional economic resilience? A research based on intermediary effect, threshold effect and spatial spillover effect,” J. Zhejiang Univ. (Humanit. Soc. Sci.), Vol.52, No.12, pp. 21-39, 2022 (in Chinese).

[B26] [26] C. Liu, Y. Chen, and M. Li, “Digital rural construction empowering rural high-quality development: Theory and empirical evidence,” Contemp. Econ. Res., No.3, pp. 80-93, 2024 (in Chinese).

[B27] [27] N. Nunn and N. Qian, “US food aid and civil conflict,” Am. Econ. Rev., Vol.104, No.6, pp. 1630-1666, 2014. https://doi.org/10.1257/aer.104.6.1630

[B28] [28] H. Farbmacher, M. Huber, L. Lafférs, H. Langen, and M. Spindler, “Causal mediation analysis with double machine learning,” Econom. J., Vol.25, No.2, pp. 277-300, 2022. https://doi.org/10.1093/ectj/utac003

[B29] [29] R. Boschma, “Towards an evolutionary perspective on regional resilience,” Reg. Stud., Vol.49, No.5, pp. 733-751, 2015. https://doi.org/10.1080/00343404.2014.959481

[B30] [30] E. Gao, J. Zhu, and J. Zheng, “The impact of agricultural socialized services on the reduction of fertilizer: Double inspection based on panel data of 31 provinces in China,” Chin. J. Eco-Agric., Vol.31, No.4, pp. 632-642, 2023 (in Chinese). https://doi.org/10.12357/cjea.20220637

Research on the Impact of Digital Rural Construction on Agricultural Economic Resilience

Chengkun Liu† and Xiqin Pan

1. Introduction

2. Literature Review

3. Indicator System Construction and Measurement Methodology

3.1. Indicator System Construction

3.2. Cross-Entropy Unbiased Weighting Method

4. Theoretical Analysis, Model Construction, and Variable Explanation

4.1. Theoretical Analysis

4.1.1. Direct Impact of Digital Rural Development on Agricultural Economic Resilience

4.1.2. Indirect Impact of Digital Village Development on Agricultural Economic Resilience

4.2. Construction of Dual Machine Learning Models

4.2.1. Benchmark Regression Model

4.2.2. Mechanism Testing Model

4.3. Variable Descriptions

4.3.1. Dependent Variables and Explanatory Variables

4.3.2. Mechanism Variables

4.3.3. Control Variables

5. Empirical Findings and Analysis

5.1. Measurement Results and Analysis

5.2. Benchmark Regression Results and Analysis

5.3. Robustness Testing

5.4. Endogeneity Test

5.5. Testing the Mechanism of Action

5.5.1. Agricultural Labor Productivity Enhancement Effects

5.5.2. Innovation Enhancement Effect

5.5.3. Industrial Structure Upgrading Effects

5.6. Heterogeneity Analysis

5.6.1. Regional Heterogeneity

5.6.2. Period Heterogeneity

5.6.3. Heterogeneity in Agricultural Development Levels

6. Research Findings and Policy Recommendations

Chengkun Liu^† and Xiqin Pan