How does digital economy affect food system resilience? A quasi-natural experiment on China’s smart city pilot policy using double machine learning

Abstract

Introduction:

Digitalization has facilitated the transformation and modernization of traditional agricultural inputs, leading to the fast expansion of the digital economy in recent years, which presents a crucial opportunity to bolster food system resilience (FSR). This study utilizes a double machine learning (DML) approach and exploits China’s Smart City Pilot (SC) policy as a quasi-natural experiment to explore the internal mechanisms through which the digital economy impacts FSR, thereby offering valuable implications for global food security.

Methods:

This study considers China’s SC policy as an exogenous shock to denote the advancement of the digital economy. It uses a DML model to evaluate the influence of the digital economy on FSR and its underlying mechanisms, as well as a spatial econometric model to investigate potential spillover effects.

Results and discussion:

The digital economy significantly enhances FSR. The development of digital inclusive finance, enhancement of agricultural labor productivity, and the deepening of agricultural product processing constitute critical pathways through which the digital economy contributes to FSR. The digital economy exerts a more pronounced influence on enhancing the adaptive and transformative dimensions of FSR, particularly by fostering synergies with major grain-producing regions. Moreover, the positive effects of the digital economy on FSR are more pronounced in regions characterized by higher farmer income levels, more developed economies, and greater degrees of urbanization. Further analysis indicates that the digital economy generates positive spatial spillover effects on FSR, indicating that it not only strengthens local resilience but also promotes improvements in neighboring regions.

Conclusion:

To further strengthen FSR, it is imperative to enhance digital infrastructure and develop a solid digital basis. Systematically drive comprehensive digital transformation throughout the whole food industry chain to boost food production along with quality. A coordinated digital economy development strategy should be devised.

1 Introduction

Ensuring food security is essential for fostering economic growth, preserving social order, and protecting national stability. According to the latest FAO report, despite a 4.2% increase in global cereal production between 2019 and 2024, the population facing severe food insecurity grew by 709 million during the same period. In the current global context, frequent uncertainties, including the Russia-Ukraine conflict and US-China trade tensions, have posed compound threats to food security worldwide. Meanwhile, increased agricultural output has not consistently translated into enhanced food security. Consequently, there is an urgent need to enhance food security and establish a highly resilient food system. However, dependence on conventional input-intensive agricultural approaches offers limited scope for innovation and fails to address the difficulties presented by a progressively dynamic external environment.

The digital economy has grown rapidly recently, and as a result, big data and cloud computing technologies are being progressively incorporated into agriculture, creating new prospects for revolutionary advancements in food management and production (Yang et al., 2024). Since 2015, China has actively advanced its digital economy. The 2024 Report indicates that the scale of China’s industrial digitalization attained 43.84 trillion Yuan in 2023, with a penetration rate of 10.78% in agriculture, the highest worldwide. China has proposed a variety of particular initiatives to expedite the profound integration of agriculture and the digital economy. For example, China issued official guidelines in 2012 to launch the Smart City Pilot (SC) policy, which was subsequently expanded nationwide in phases from 2012 to 2014. China’s smart city development emphasizes integrating digital technologies to promote coordinated digital transformation across urban and rural areas. This process includes advancing digital infrastructure and facilitating the movement of digital production elements, hence being intricately connected to the digital economy. Consequently, as an empirical framework of note, China’s SC policy allows for analysis of both how the digital economy affects food system resilience (FSR) and the mechanisms underpinning this dynamic. This study lays a theoretical and empirical foundation for refining relevant policies and offers insights for countries grappling with food security and sustainability challenges.

The literature pertinent to this study may be generally classified into three primary domains. The first domain pertains to FSR. The notion of FSR was first proposed by Tendall et al. (2015). It denotes the ability to absorb, recover from, learn from, and adapt to shocks, which is essential for sustainable food growth (Karan et al., 2023). Most measurement approaches rely on composite indicator frameworks capturing resistance, adaptability, and transformability (Zeng et al., 2025), while a smaller body of work uses economic indicators such as agricultural output and farm income. Additionally, certain studies decompose FSR into specific components, such as production resilience (Jensen and Orfila, 2021), supply chain resilience (Poo et al., 2024), and demand-side resilience (Han et al., 2024). The second domain of literature is concerned with the digital economy. Driven by the swift expansion, it has had a profound worldwide influence. Most scholars regard it as a positive force, although some hold divergent views, reflecting the field’s multidimensional and increasingly diversified nature. The impact of the digital economy constitutes a primary focus of the current research. On a macro scale, it is considered a revolutionary paradigm that reconfigures growth models and drivers (Sadiq et al., 2025), accelerates economic expansion (Erumban and Das, 2016), facilitates green transitions (Wu et al., 2022), and advances global sustainability goals (Zheng and Wang, 2022). It also exhibits spillover effects that foster regional coordination (Tranos et al., 2021). Nonetheless, certain researchers contend that the digital economy might exacerbate disparities between nations, thereby intensifying global imbalances (Drahokoupil and Piasna, 2017). On a micro scale, research generally concentrates on specific sectors. In non-agricultural industries, it has been shown to reduce production costs, improve transaction efficiency (Yu et al., 2024), and substantially improve total factor productivity in industry. Moreover, Wu et al. (2024) discovered that the digital economy facilitates optimal allocation of skilled human resources. However, it also results in the displacement of numerous low-skilled workers by digital technologies, thereby exacerbating income inequality. Within the agricultural sector, the digital economy optimizes agricultural product circulation, drives superior growth (Wang et al., 2024), and fosters innovations in rural digital governance (Wang et al., 2023). Additionally, it boosts farmers’ productivity and income. Nevertheless, the integration between the digital economy and agriculture remains nascent. The persistent digital divide continues to hinder the full realization of its potential benefits (Qi et al., 2025). The third domain focuses on the correlation between digitization and FSR. This area of inquiry remains nascent, with existing studies primarily examining how digital infrastructure (Dong et al., 2025), digital inclusive finance (Wang et al., 2023), emerging digital technologies (Wang et al., 2023; Zhu et al., 2024), and agricultural e-commerce platforms (Suali et al., 2024) influence FSR.

In summary, while existing studies have yielded meaningful insights into both the digital economy and FSR, several critical research gaps remain. First, the comprehensive study on the interplay between the digital economy and agriculture, its particular impact on FSR, is little examined and requires further examination. Second, most current assessments of the digital economy’s impact depend on conventional econometric approaches such as DID (Wang et al., 2023) and fixed effects models (Zhu et al., 2024). But, given the multidimensional nature of both the digital economy and FSR, these models fall short in capturing intricate causal mechanisms. Third, current research on the mechanisms focuses primarily on macro-level indicators, lacking a systematic analysis of specific transmission pathways.

Therefore, this study employs panel data from 2003 to 2022, obtained from 276 prefecture-level cities in China. It leverages the SC policy as a quasi-natural experiment and employs a double machine learning (DLM) model to assess the effects of the digital economy on FSR. This study presents three primary contributions in comparison to prior research: First, a more rigorous and comprehensive evaluation index system is constructed to assess FSR. Second, the mechanisms are clarified from the standpoint of the food industry chain, while the spatial spillover effects are also investigated, thereby extending the scope of related research. Third, by leveraging the DML causal inference framework, this study effectively addresses the challenges of the “curse of dimensionality” and multicollinearity. By capitalizing on the advantages of high-dimensional and nonparametric estimation, it precisely evaluates the impact, thus demonstrating methodological promotion.

2 Theoretical analysis and research hypotheses

2.1 Definition of FSR

As defined by the FAO in 2021, FSR denotes the capacity to endure disruptions while ensuring continuous availability of sufficient, safe, and nutritious food for the population and protecting the livelihoods of agricultural stakeholders. In contrast, China places a stronger emphasis on stable food production and supply as the cornerstone of its national food security strategy. Based on this perspective, this study adopts a supply-side approach. FSR is thereby described as the capacity to sustain consistent food production and supply amid uncertainty, encompassing absorptive, adaptive, and transformative capacity. More precisely, absorptive capacity describes the ability to withstand diverse shocks while maintaining functional stability. Adaptive capacity describes the capability to promptly respond to disruptions, facilitating rapid recovery and a return to a trajectory of stable and sustainable growth. Transformative capacity is defined as the system’s potential to undergo structural transformation and innovation in response to long-term uncertainties, thereby strengthening its future shock-resistance.

This study identifies two core differences in the definition of FSR compared to existing literature. On one hand, the definition focuses on the food supply side. This perspective is rooted in the fact that food production and management in developing countries like China are still in an exploratory stage, with their ability to ensure stable production and supply needing further improvement. It thus aligns more closely with the concept of food system resilience centered on agricultural production. On the other hand, compared to the framework proposed by Tendall et al. (2015), which defines resilience as the ability to “absorb, recover from, learn from, and adapt to shocks,” this study dissects FSR into three dimensions: absorptive, adaptive, and transformative capacity. By excluding the recovery dimension and adjusting to place greater emphasis on the transformative capacity, this aligns with the core theme of the study, which focuses on the empowerment of FSR through the transformation of system structure in the context of the digital economy.

2.2 The impact of the digital economy on FSR

Network economics posits that the digital economy exhibits strong penetration, broad reach, and high innovation capacity (Zhu et al., 2024). These features directly affect the production, distribution, and consumption of food, giving rise to scale effects, cost efficiencies, and precise matching, thereby enhancing the food system’s absorptive, adaptive, and transformative capacity in the face of shocks. First, the scale effect. Digital technologies help reduce resource inefficiencies caused by uncontrolled land expansion. With the support of agricultural data platforms integrating soil, climate, and market information, producers can optimize planting scales, lower per-unit input costs, and realize economies of scale. These efficiencies allow resources to be directed toward improved seed development and the adoption of advanced agricultural technologies (Jin et al., 2024), thereby fostering innovation across the food industry and enhancing the food system’s transformative capacity. Second, the cost efficiency. The theory of information asymmetry posits that unequal access to information among market participants results in suboptimal decisions and heightened risks. The digital economy helps address these asymmetries by dismantling information barriers, supporting accurate forecasting and real-time tracking of supply and demand. This allows farmers to optimize planting decisions, minimize production losses, and enhance the quality and efficiency. Moreover, enhanced early risk detection through digital technologies facilitates timely responses across the food system, reducing exposure and improving both absorptive and adaptive capacities. Third, the precision matching effect. The digital economy has fostered new sales models for agricultural products, including e-commerce and livestreaming, which integrate online and offline channels to align producers more precisely with consumer demand (Xiao and Abula, 2023). In addition, Internet of Things technologies such as climate-controlled greenhouses, soil testing devices, and intelligent irrigation systems enable effective data collection and feedback, allowing resource inputs to be better aligned with crop needs. This reduces uncertainty throughout the entire production and marketing process, thereby reinforcing food systems against external shocks.

Hypothesis 1. The digital economy improves FSR.

2.3 Mechanisms of digital economy on FSR

The theory of factor allocation and endogenous growth theory reveal that the efficiency of factor allocation directly impacts the resilience and long-term growth of production systems. The enhancement of food system resilience depends on the optimized allocation of key elements across various stages of the supply chain (Flaschel et al., 2013; Xie et al., 2025). The theory of division of labor further supports the rationality of the segmentation and coordination of supply chain stages. The food system, as a complete system encompassing pre-production, in-production, and post-production stages, exhibits significant differences in the core elements and functions across its stages. The pre-production stage, which includes activities such as the purchase of agricultural inputs, heavily relies on financial support (Jia, 2023). The development of digital inclusive finance directly determines the accessibility and adequacy of funds, thus impacting the resilience of this stage. The in-production stage focuses on agricultural labor as the core input, and the level of agricultural labor productivity is directly related to production efficiency and supply stability (Frank et al., 2019), making it critical for maintaining system resilience. The post-production stage enhances the value chain through food processing. According to the specialization and collaboration logic of the division of labor theory, deeper processing can effectively stabilize market supply and strengthen the supply chain’s risk resilience (Dong et al., 2025), thereby enhancing the resilience of the food system. Therefore, this study, based on the characteristics of the food supply chain stages and core elements, and in combination with relevant theories, divides the mechanisms through which the digital economy influences food system resilience into three categories: digital inclusive finance, agricultural labor productivity, and agricultural product processing. These categories correspond to the key impact pathways at the pre-production, in-production, and post-production stages, respectively.

2.3.1 Digital inclusive finance development

In the pre-production stage, limited access to finance remains a critical constraint on agricultural development. The digital economy develops digital inclusive finance, enhancing FSR. On one hand, digital platforms facilitate data sharing, allowing financial institutions to reduce information asymmetries. Through AI-based risk assessment and internet-enabled channels, financial services are extended to rural residents, low-income groups (Li, 2024), and other underserved populations. On the other hand, digital inclusive finance creates financial solutions tailored to the cycles of agricultural output. An example is credit lines designated for particular crops, which are linked to their sowing timelines. And it narrows market information gaps, aligns capital supply with demand, and improves the efficiency of financial allocation, thereby easing farmers’ financing constraints. This enables farmers to invest in machinery and expand production (Zhao et al., 2023) while also supporting knowledge acquisition, ultimately strengthening FSR.

Hypothesis 2. The digital economy can improve FSR by developing digital inclusive finance.

2.3.2 Agricultural labor productivity enhancement

In the in-production stage, rural areas face structural challenges such as an aging and feminized labor force, along with inefficiencies in traditional farming practices. The digital economy drives agricultural innovation and boosts labor productivity, strengthening FSR. First, agricultural big data platforms integrate key data on crop genetics, soil conditions, and pest dynamics, facilitating targeted breeding and technological advancement (Lu et al., 2025). Meanwhile, information and communication technologies accelerate the diffusion and adoption of these innovations, promoting the integration of digital tools throughout food production. Smart agricultural machinery enables more precise breeding, fertilization, and cultivation, effectively alleviating labor shortages. These advancements improve both the quantity and quality of food production while also boosting labor productivity (Jin et al., 2024). Second, increased labor productivity stabilizes food production and reduces output volatility. The widespread adoption of digital technologies enhances farmers’ digital literacy, thereby providing essential human capital for food system transformation. This supports a more effective distribution of agricultural labor during various production phases, reducing the waste of human resources during both peak and off-peak seasons. Lastly, greater labor productivity frees up time and resources, enabling producers to invest in risk-resilient infrastructure and better respond to market and environmental shocks.

Hypothesis 3. The digital economy can improve FSR by enhancing agricultural labor productivity.

2.3.3 Agricultural product processing deepening

In the post-production stage, the processing and marketing of food products represent a pivotal segment of the chain. The digital economy deepens the processing of agricultural products, promoting FSR. For one thing, the digital economy stimulates the establishment of smart factories, enabling agricultural processing enterprises to adopt intelligent and automated production systems that reduce post-harvest losses and lower processing costs. These systems, powered by real-time data analytics, also refine processing procedures to preserve nutritional value and prolong the shelf life of products. Simultaneously, it leverages platforms such as “Taobao Villages,” “Taobao Towns,” and third-party e-commerce networks (Mondejar et al., 2021) to integrate regional characteristics into product design, effectively meeting diversified market demands and spurring the expansion of the agricultural processing industry. For another, deepened agricultural processing enhances brand recognition and product value-added, boosts income for both enterprises and farmers, and strengthens linkages among large-scale grain producers, processing enterprises, and consumers. This stabilizes food production (Qi et al., 2025) and supports benefit-sharing across the processing chain. It improves supply chain coordination, enhances market responsiveness, reduces disruption risks, and reinforces the foundations of food system stability.

Hypothesis 4. The digital economy can improve FSR by deepening agricultural product processing.

2.4 Spatial spillover effects of the digital economy on FSR

Spatial economics posits that as production factors move across regions, economic activities often display spatial clustering and mutual dependence. The digital economy compresses time and erodes spatial boundaries by facilitating rapid information transmission (Deng et al., 2022). Existing studies indicate that the digital economy exerts dual spatial impacts on neighboring regions. On one hand, it may generate a “spillover effect” by accelerating knowledge dissemination and boosting local development. On the other hand, excessive disparities in digital development can drive the agglomeration of advantageous resources toward areas with superior digital infrastructure, further widening regional development gaps (Zhu et al., 2024). By contrast, this study focuses on the intrinsic attributes of digital economic technologies, arguing that their robust network diffusion capacity and inherent externalities give rise to positive spatial spillovers, which in turn contribute to the enhancement of FSR. As an example, digital system supports real-time monitoring of agricultural conditions in neighboring regions, facilitating coordinated efforts in pest management and disaster prevention. Geographically, neighboring cities often share similar topography, soil types, and climatic conditions. By removing spatial–temporal barriers and promoting data sharing, the digital economy enables the fluid transfer of agricultural talent, capital, and equipment across regions (Hong et al., 2023). This promotes the effective deployment of resources toward food production and strengthens system-wide resilience. Industrially, the digital economy stimulates innovation in food cultivation, while the rapid dissemination of technology accelerates its broader application. Moreover, regions leverage their resource advantages to develop specialized agricultural clusters. Digital platforms promote cross-regional technological exchange among agricultural firms, enhancing cooperation and complementarity. This promotes the coordinated upgrading of the food industry, builds a resilience network based on risk-sharing, and strengthens interregional FSR. Figure 1 depicts the conceptual framework.

Figure 1

Hypothesis 5. The digital economy exerts positive spatial spillover effects on FSR.

3 Materials and methods

3.1 Data sources

To ensure data availability and continuity, this study employs panel data including 276 prefecture-level cities from 2003 to 2022. The primary data sources include regional statistical bureaus and various official statistical yearbooks, such as the China Urban Statistical Yearbook and China Social Statistical Yearbook. Linear interpolation was employed to handle missing values, while Table 1 provides descriptive statistics.

3.2 Model settings

3.2.1 Double machine learning model

This study employs the DML model introduced by Chernozhukov et al. (2018) to empirically evaluate the influence of the digital economy on FSR. In comparison to conventional causal inference methodologies such as DID, propensity score matching, and synthetic control, the DML model offers several advantages and has been increasingly adopted in causal analysis (Yang et al., 2021). First, the DML model identifies covariates with strong predictive power from high-dimensional datasets, enabling richer control variables while alleviating the challenges of dimensionality and multicollinearity, thus improving the accuracy of causal inference. Second, by utilizing pre-treatment data from the control group, the DML model constructs counterfactual outcome functions through machine learning algorithms, ensuring the validity of counterfactual predictions and improving the overall efficiency of the estimation. Finally, as a nonparametric regression framework, the DML model captures complex nonlinear relationships among variables while regularizing the nuisance parameter functions, thus overcoming the limitations of linearity assumptions in conventional causal models. Based on this, the study sets the subsequent partly linear DML model.

In Equations 1, 2, denotes the level of FSR. is a binary indicator for the implementation of the digital economy. represents treatment effect of the policy. stands for a high-dimensional vector of control variables that may jointly influence both and . The function , whose exact form is unspecified, is estimated via machine learning techniques. represents the error term, presumed to have a conditional mean of zero.

When estimating the unknown function using machine learning with regularization, the resulting regularization bias may violate the unbiasedness condition of the treatment effect estimator. Specifically, the slow convergence of to hinders convergence of to . To speed up the convergence rate and guarantee unbiasedness in finite samples, we implement the following auxiliary regressions.

In Equations 3, 4, denotes the regression function of the treatment variable concerning a collection of high-dimensional control variables, whose exact functional form is unknown and inferred by machine learning. The residual term is presumed to possess a conditional mean of zero. First, the auxiliary regression is estimated via machine learning, and the residuals are constructed as . Next, the outcome regression is estimated, producing the partial residual equation: . Using as an instrumental variable for , Equation 5 yields an unbiased estimator.

At this stage, the double application of machine learning estimators accelerates the convergence of and converge more rapidly to and , respectively, allowing for an unbiased estimation of the treatment effect in finite samples.

3.2.2 Mechanism model

This study adopts the analytical framework of Huang et al. (2023) to examine the mechanism, which addresses the endogeneity bias inherent in traditional mediation analysis methods. The analysis first explores, at a theoretical level, the influence of mechanism variables on the outcome, followed by an empirical examination of how the explanatory variable affects the mechanism pathways.

In Equations 6, 7, represents the mechanism variables, encompassing digital inclusive finance, agricultural labor productivity, and agricultural product processing. If is statistically significant, this would show the presence of the corresponding transmission mechanism.

3.2.3 Spatial econometric model

To further examine the spatial spillover impacts, this study follows the standard procedure for spatial econometric model selection. Specifically, spatial autocorrelation is first tested using Moran’s I index, followed by LM, LR, and Wald tests. The model is specified as follows.

In Equation 8, denotes the level of FSR, and is a dummy variable indicating whether the digital economy has been implemented in city . represents the spatial autocorrelation coefficient, and is the spatial weight matrix. and represent the coefficients for the spatial lags of the digital economy variable and the control variables.

3.3 Variable selection

3.3.1 Dependent variable

Food system resilience (FSR). On the supply side, it encompasses not just food production but also the actions of production and management organizations. Building upon the frameworks of Zurek et al. (2022) and Wang et al. (2023), we employ an index-based evaluation method to thoroughly assess the level. Based on resilience evolution theory and the FSR framework presented in this study, FSR is decomposed into three dimensions: absorptive capacity, adaptive capacity, and transformative capacity. Specifically, absorptive capacity includes intrinsic stability and supply stability; adaptive capacity includes sustainability and recoverability; and transformative capacity comprises collaborative diversity and technological advancement. Accordingly, a three-level indicator system with 16 indicators is constructed to assess FSR. The precise structure of the indicators is shown in Table 2. Furthermore, the entropy method is utilized to compute overall FSR scores for each prefecture-level city.

3.3.2 Explanatory variable

Digital economy (Event). Following Wang and Zhong (2023), this study treats the SC pilot policy as a quasi-natural experiment to identify the causal effect of the digital economy on FSR. The variable Event is assigned a value of 1 for prefecture-level cities in the year of and subsequent years following the implementation of the SC policy, and 0 otherwise.

3.3.3 Mechanism variables

• Digital Inclusive Finance (Df). Following Hu et al. (2023), it is assessed utilizing the composite index created by the Digital Finance Research Center.

• Agricultural Labor Productivity (Alp). It is calculated as the ratio of primary industry value added to the workforce employed in that sector.

• Agricultural Product Processing (Sap). It is measured by the main income of large-scale enterprises engaged in agricultural product processing.

3.3.4 Control variables

Guided by the frameworks of Kc et al. (2024), Hirth et al. (2025), and Wang et al. (2023), this study incorporates an extensive range of control variables. (1) Farmers’ income level (Fil), gauged by the year-end per capita net income of rural residents. (2) Rural infrastructure (Ri), assessed by the proportion of country roads to overall rural road area. (3) Human capital (Hc), calculated based on the average length of schooling for rural residents. (4) Regional industrial structure (Ris), measured by the ratio of the total output value of the tertiary industry to that of the secondary industry. (5) Degree of openness (Ou), assessed by the total volume of highway freight transportation. (6) Urban–rural economic disparity (Urd), assessed by the ratio of urban to rural per capita net income at year-end. (7) Living standard of farmers (Lsf), measured by the per capita consumption expenditure of rural residents. (8) Urbanization level (Urb), captured by the share of urban residents in the total population. (9) Digital infrastructure in rural areas (Dig), quantified by the percentage of administrative villages with broadband internet access. (10) Regional population density (Rp), assessed by dividing the land area by the number of residents. Furthermore, to improve model precision and account for potential nonlinearities, squared terms of control variables are incorporated, along with city and year fixed effects captured via dummy variables.

4 Results

4.1 Baseline regression results

This study adopts the DML model to assess the impact, applying a 1:4 sample-splitting strategy and a more interpretable debiased lasso regression algorithm. Regression results are reported in Table 3. Column (2) extends Column (1) by including the first-order terms of the control variables. Column (3) further incorporates the second-order terms of the control variables. The coefficient remains significantly positive, suggesting that the digital economy enhances FSR. Controlling for other factors, the digital economy improves resilience by approximately 0.0028 units on average, thereby supporting Hypothesis 1.

4.2 Robustness tests

4.2.1 Substitution of the main variables

The dependent variable, FSR, is typically assessed using either a composite index or a core variable approach. The core variable method simulates the gap between the actual and expected outcomes in the presence of shocks, enabling a comparison of different units’ responses to disruptions. Following the approach proposed by Martin and Sunley (2015), this study uses grain output as the core indicator to assess FSR. The specific equation is as follows: . Here, denotes the FSR, is the actual increase rate of grain output, while stands for the expected growth rate based on the provincial average. The deviation between actual and expected growth rates is used to recalibrate FSR at the city level. On the other hand, we replace the explanatory variable. Drawing on Jiang et al. (2022) and Wang and Shao (2023), we construct a city-level composite index of the digital economy, encompassing four dimensions: internet penetration, digital industry employment, digital output, and mobile phone usage. These are measured, respectively, as broadband internet subscribers per 100 people, software workers as a share of total urban employment, per capita telecom service volume, and mobile phone users. As reported in Table 4, the results confirm the robustness.

4.2.2 Exclusion of other policy interventions

This study introduces dummy variables for “Broadband China,” “National Big Data Comprehensive Pilot Zones,” “Digital Government,” and “Digital Villages” policy into the models. Table 5 reveals that the baseline regression findings are resilient and not significantly influenced by other policies.

4.2.3 Re-specification of the DML model

To avoid model specification errors in machine learning from affecting the estimation results, this study re-specifies the DML model from four aspects. First, the sample-splitting ratio is modified from 1:4 to 1:2 and 1:7, and the regressions are re-estimated accordingly. Second, the machine learning algorithms employed in the main and auxiliary regressions are replaced with support vector machines (SVM) and neural networks (NNET). Third, the number of resampling iterations in the DML model is increased to 101, with the results presented in Table 6. Fourth, we employ the generalized random forest algorithm proposed by Athey et al. (2019), which has been widely applied in the effect testing of policy evaluation for double machine learning (Davis and Heller, 2019). As shown in Figure 2, as the number of trees gradually increases, the proportion of the distribution in the positive interval continuously rises, further verifying the robustness of the results.

Figure 2

4.2.4 Adjustment of research sample

Given that Beijing, Shanghai, Chongqing, and Tianjin are directly governed municipalities, their samples are excluded from the analysis. Second, considering that the COVID-19 pandemic in 2020 severely disrupted production and operations across all sectors, the 2020 samples are removed, and the model is re-estimated. Finally, to reduce potential distortions caused by outliers, the data are winsorized and trimmed at the 1 and 5% thresholds. Table 7 demonstrates the robustness.

4.3 Endogeneity tests

4.3.1 Instrumental variable

Regions with higher levels of FSR often have more advanced digital infrastructure, potentially indicating a bidirectional causal relationship (Zhu et al., 2024). This study adopts the methodological framework proposed by Chernozhukov et al. (2018) and constructs a partially linear instrumental variable model using a DML approach to mitigate the endogeneity issue. The variables and parameter settings in Equations 9, 10 align with those utilized in the baseline regression. The particular model is described as follows.

Following Deng et al. (2022), this research takes the 1984 fixed-line telephone counts across regions as the instrumental variable. For one thing, internet technology development arises from advances in communication tools, with fixed-line telephones representing the earliest form. Regions with a higher number of fixed-line telephones typically had more developed communication infrastructure, which later facilitated digital advancement. Moreover, the transformative progression of the digital economy stemmed from innovations in fixed-line communication (Zhu et al., 2024), highlighting the close linkage between fixed telephone prevalence and digital economic development. This satisfies the relevance condition for a valid instrumental variable. For another, fixed-line phones are impossible to exert a direct impact on FSR because they are no longer a prominent digital device owing to technology advancements. This satisfies the exogeneity condition for a valid instrument. Furthermore, as the 1984 fixed-line telephone data are cross-sectional and cannot be directly applied in a panel framework, we employ the methods established by Nunn and Qian (2014). Specifically, the instrumental variable is created by interacting the 1984 fixed-line telephone count with the quantity of broadband subscribers. Table 8 corroborates the robustness.

4.3.2 Difference-in-differences (DID) model

The DID approach is effective in mitigating endogeneity issues arising from omitted confounders or collider bias, thereby alleviating potential biases in coefficient estimation. Accordingly, this study further employs a multi-period DID model.

In Equation 11, denotes the degree of FSR. The interaction term represents the implementation of digital economy through the SC policy. denotes the intercept term. denotes the collection of control variables. captures the estimated coefficients associated with the control variables. is the central focus of this study. As shown in Table 9, the coefficient on remains significantly positive, thereby confirming the robustness.

Second, satisfying the parallel trends assumption is essential for credible policy evaluation. As noted by Wen et al. (2024), testing the parallel trends assumption remains necessary even when using the DML model. Accordingly, this study conducts a parallel trends test for heterogeneous treatment effects based on the multi-period DID model. The results, presented in Figure 3, validate the parallel trends assumption. In addition, to eliminate the impact of unobservable confounding factors, a placebo test is conducted. Figure 4 illustrates the dependability of the baseline findings.

Figure 3

Figure 4

Finally, applying a two-way fixed effects multi-period DID model may lead to aggregation bias, as treatment effects across different groups and periods are averaged, potentially distorting the true causal effect (Goodman-Bacon, 2021). We employ the Bacon decomposition method to resolve this issue, which categorizes the staggered implementation of digital economy policies into three types of comparison groups. Among them, the comparison between “Later-Treated vs. Early-Treated Smart City Pilot Policy” generates biased treatment effects, which can compromise the validity (Wing et al., 2024). According to the literature, if the weight of these biased comparisons is below 10%, the estimation results are generally considered reliable. As shown in Figure 5 and Table 10, the predicted value of the adverse treatment effect is −0.004, with a corresponding weight of only 2.2%, suggesting that the baseline estimates are robust.

Figure 5

4.4 Mechanism test

The foregoing theoretical study indicates that the digital economy can influence FSR through digital inclusive finance in the pre-production stage, agricultural labor productivity in the in-production stage, and agricultural product processing in the post-production stage. Table 11 indicates that the estimated coefficients for the digital economy’s effects on digital inclusive finance, agricultural labor productivity, and agricultural product sales are significantly positive. It suggests that the digital economy has beneficially impacted the entire food industry chain, thereby strengthening FSR and providing empirical support for Hypotheses 2,3, and 4.

4.5 Spatial spillover effect analysis

4.5.1 Spatial autocorrelation test

Table 12 presents Global Moran’s I. The p-values for FSR from 2003 to 2022 are all below 0.1, indicating significant spatial clustering. Therefore, employing spatial econometric modeling is an appropriate approach.

4.5.2 Spatial model selection

Table 13 presents a series of tests used to identify the two-way fixed effects Spatial Durbin Model (SDM). First, the LM test results indicate that spatial econometric models outperform the OLS model. Subsequently, the LR and Wald tests were applied to refine model selection. Both yielded statistically significant results, indicating that the SDM is the most suitable option.

4.5.3 Spatial spillover effect test

This study employs an inverse distance spatial weight matrix, with the estimated findings presented in Table 14. The spatial autoregressive coefficient () for FSR is 0.6875 and is significantly positive, demonstrating a significant regional correlation in FSR. Moreover, the spatial interaction term for the digital economy in neighboring regions is markedly positive, suggesting substantial spillover effects, thereby supporting Hypothesis 5. Furthermore, the results remain robust when replacing the SDM with SEM and SAR specifications, as both ρ and the digital economy coefficient remain significantly positive.

4.5.4 Spatial effects decomposition

This study adopts the method of Lesage and Pace (2008), applying partial differential decomposition in the SDM to separate total spatial effects into direct and indirect components. Table 15 indicates that the digital economy not only bolsters FSR in the local region but also has a more pronounced positive impact on neighboring regions.

4.6 Heterogeneity analysis

4.6.1 FSR dimensions heterogeneity

As shown in Table 16, the effect on absorptive capacity is not statistically significant, but it significantly enhances both adaptive and transformative capacities. This means the digital economy mainly boosts the adaptive and transformative aspects of FSR.

4.6.2 Functional grain zones heterogeneity

The government has classified regions into major and non-major grain-producing zones based on their functional roles and implemented differentiated resource allocation strategies. This study establishes a dummy variable (Fun) that is assigned a value of 1 for prefecture-level cities classified as important grain-producing regions and 0 for all others. Table 17 illustrates that the digital economy exerts a more beneficial impact on FSR in major grain-producing regions.

4.6.3 Farming households heterogeneity

Variations in household income levels directly influence farmers’ decisions regarding the allocation of labor, capital, and technology in grain production, thereby resulting in differentiated impacts on FSR. Based on rural per capita net income, this study classifies regions into high- and low-income farming areas and defines a dummy variable (Inc), which equals 1 if the income is above average and 0 otherwise. Column (2) of Table 17 indicates that the digital economy significantly enhances FSR in affluent agricultural regions.

4.6.4 Regional heterogeneity

Regional differences in economic development and urbanization have progressively shaped the integration of the digital economy into the food sector, leading to varied impacts on FSR. On one hand, cities are divided into high- and low-economic-level groups by average total and per capita GDP. Dummy variables for total GDP (Gdp) and per capita GDP (Agdp) are defined, with a value of 1 indicating cities above the average, and 0 otherwise. Columns (3) and (4) of Table 17 illustrate that cities with higher economic levels, either by total GDP or per capita GDP, benefit more from the digital economy.

For assessing the role of urbanization, cities are categorized into high and low urbanization categories based on the average urbanization rate. A dummy variable (Urr), is defined as 1 for cities exceeding the average and 0 otherwise, is interacted with the digital economy variable (Event) and included in the DML model. Column (5) of Table 17 indicates that regions with elevated urbanization levels observe a more significant influence from the digital economy.

5 Discussion

5.1 Main findings

This study investigates the impact of the digital economy on food system resilience by treating wheat, rice, maize, and other food crops as a homogeneous group, finding that the digital economy significantly improves FSR. This is primarily because of the modernization and digital transformation of the food system, thereby strengthening its overall resilience. For absorptive capacity, improvements in digital infrastructure and technology facilitate disaster risk prediction and visualization, thereby reducing uncertainty. For adaptive capacity, the growth of digital industries supports integration across the food supply chain, enabling timely responses and rapid adjustments to external shocks. In terms of transformative capacity, digital technologies transcend traditional constraints and drive innovation in smart governance, providing new momentum for the food system. The expanding influence of the digital economy on the food sector has drawn increasing academic interest (Kapoor and Vij, 2018), and the findings correspond with those of Zhu et al. (2024) and Wang et al. (2023).

However, existing studies offer diverse perspectives on how the digital economy affects FSR. Xie et al. (2025) examine the issue from the standpoint of resource optimization, while Taghikhah et al. (2025) consider the role of digital economy workers. This study adopts a food industry chain perspective, revealing that the digital economy enhances FSR through the development of digital inclusive finance, enhancement of agricultural labor productivity, and the deepening of agricultural product processing. In the pre-production stage, the development of digital inclusive finance reduces the geographical and informational barriers associated with conventional financial services (Zhang et al., 2023). It eases credit constraints, increases financial access, and addresses financing challenges for food producers. This allows farmers to invest in machinery, inputs like fertilizers and pesticides, and scaling-up costs, thereby strengthening food system stability (Jia, 2023). In the in-production stage, rising agricultural labor productivity enhances food production efficiency. It also facilitates the reallocation of more resources and time to the innovative development of the food system. This supports the digital transformation of food production, and reinforces systemic resilience. In the post-production stage, deeper processing of agricultural products extends the food value chain and increases value-added output. It strengthens links among producers, processors, and consumers, enhances branding, and improves supply–demand alignment. These developments help the food system better withstand market uncertainties and boost overall resilience.

This study further reveals that the digital economy has a positive spillover impact on FSR, in contrast to the siphoning effect identified by Lin et al. (2025) in their study on industrial structure. Focusing on the agricultural sector, this study highlights that, unlike traditional non-agricultural industries, food production often occurs under similar ecological and production conditions across neighboring regions. Both the reproducibility of agricultural technology and the widespread accessibility of digital infrastructure serve to strengthen the spillover impacts of the digital economy. This finding is consistent with Jiang et al. (2022). However, when spatial effects are decomposed, our findings diverge from those of Dong et al. (2025), who reported a stronger local impact. In contrast, we observe a stronger enhancing impact in neighboring regions. This discrepancy arises because the emergence of the digital economy in local regions often starts from scratch, requiring substantial investment in capital and labor to establish digital infrastructure. In the short term, this crowds out resources originally allocated to grain production, storage, and logistics, thereby weakening FSR and partially offsetting the benefits of digital advancement. In contrast, neighboring regions benefit from the digital spillover without bearing the initial infrastructure costs. By leveraging existing networks, platforms, and technologies, they are better able to translate digital empowerment into stronger FSR.

A further heterogeneity analysis is conducted in this study. Specifically, regarding FSR, the digital economy does not have a significant impact on the resilience’s absorptive capacity but significantly enhances the adaptive and transformative capacity of the food system. Absorptive capacity focuses on the initial shock effects. It relies on the status of infrastructure hardware. Meanwhile, the digital economy transforms digital information technology combined with infrastructure into actual resistance capacity. This transformation has not yet been completed in some regions, however. At the same time, macro measures such as agricultural policies and food reserve adjustments play a significant role in resistance, mitigating the actual effects of the digital economy. In contrast, adaptive and transformative capacity depends on information efficiency and resource integration, which are more compatible with digitalization. Specifically, the digital economy breaks down information barriers. It facilitates the transmission of real-time market information through digital platforms, optimizing food production input decisions. This enables rapid response and recovery during shocks. Additionally, the digital economy fosters innovative models such as smart agriculture by reconfiguring production factors. It thereby drives the digital transformation of the entire supply chain. For another, at the household level, high-income households are more likely to adopt smart machinery, IoT devices, and digital agricultural services. This enhances production stability and strengthens risk resilience through digital process optimization. Third, at the regional level, cities boasting higher economic status are equipped with mature infrastructure and policy support, enhancing FSR. High urbanization levels accelerate resource flows, promoting the incorporation of the digital economy into agriculture. Strong market demand and investment momentum optimize resource allocation, drive industrial digitalization, and reinforce FSR. These findings align with the conclusions of Hu et al. (2025).

5.2 Key contributions

Research on FSR remains in its infancy. Existing studies primarily employ conventional models such as fixed effects, DID, and synthetic control, which may inadequately encompass the complex effects of the digital economy on FSR (Yang et al., 2021). By employing a DML causal inference model, this study tackles “curse of dimensionality” and multicollinearity challenges, leveraging its strengths in high-dimensional and nonparametric estimation to accurately assess the impact of the digital economy. Moreover, the digital economy exerts a broad and far-reaching influence (Chinoracky and Corejova, 2021), and its transformative nature offers development opportunities for both developed and developing countries. This study presents a Chinese paradigm of digital economy development, offering valuable insights for countries aiming to integrate digitalization into agriculture and move beyond traditional growth models (Sachs, 2019).

5.3 Limitations

This study possesses certain limitations. First, the panel data included in this study encompasses 276 prefecture-level cities from 2003 to 2022, excluding those that adopted digital economy policies after 2022.

Second, due to severe data deficiencies in certain cities, the analysis includes only a subset of prefecture-level cities. This approach may introduce potential selection bias, thereby affecting the generalizability of the research findings. Regarding sample characteristics, the excluded prefecture-level cities are primarily concentrated in economically underdeveloped regions. These areas often face challenges such as low digital infrastructure coverage and delayed agricultural digital transformation, making them key regions that urgently require the digital economy to enhance food system resilience. Regarding the differences in impact pathways, on one hand, the low level of digital economy infrastructure may result in an increasing marginal effect on food system resilience as the digital economy develops. Its constraining effect may be minimal. The sample composition may have prevented the study from exploring this differential impact, thus leading to an estimation bias towards regions with better digital infrastructure and making it difficult to generalize the findings to economically underdeveloped areas. On the other hand, the agricultural development model of the excluded prefecture-level cities relies on traditional agricultural inputs. The mechanisms identified in this study, such as digital inclusive finance, may not be suitable for these regions, leading to an overestimation or underestimation of their applicability. It should also be noted that while the DML model can partially mitigate bias caused by observed variables, its ability to address sample self-selection issues arising from data missingness is limited. Therefore, future research should continue to improve the dataset and incorporate data from more regions into the analytical framework.

Third, due to data availability limitations, this study did not disaggregate food crops, treating rice, wheat, and maize as a homogeneous group. This may have masked the differentiated impacts of the digital economy on the resilience of different crop systems. Different food crops have varying preferences for digital technologies in different cultivation stages. For instance, rice requires more intelligent irrigation systems and climate monitoring technologies, while wheat demands drone-based plant protection technologies. At the same time, the resilience needs of different crops vary. For example, rice requires greater flood resilience, while wheat needs higher cold tolerance. Therefore, future research could disaggregate crop types, construct resilience indicators for the food system specific to each crop, and re-examine the differentiated impacts of the digital economy on them. Additionally, focusing on the entire supply chain of a single crop, future studies should explore the differentiated effects at various growth stages. Such refinement would yield important practical implications for promoting differentiated development within the food sector.

6 Conclusion

This study explores the impact of the digital economy on FSR. The principal conclusions are as follows: First, the digital economy significantly improves FSR. Second, it improves resilience through the development of digital inclusive finance, enhancement of agricultural labor productivity, and deepening of agricultural product processing. Third, the digital economy’s effects are heterogeneous. Considering the dimensions of FSR, it primarily affects adaptive and transformative capacities. With regard to functional grain zones, major grain-producing areas experience greater improvements, facilitating policy synergy. At the household level, the effect is more prominent among higher-income farmers. At the regional level, economically developed and highly urbanized areas have a better possibility of increasing FSR. Fourth, the digital economy produces spatial spillover effects. It not only strengthens local FSR but also promotes resilience in neighboring regions.

These findings emphasize the necessity of prioritizing the digital economy, particularly by improving its supporting infrastructure. Efforts should focus on accelerating rural digitalization, especially by prioritizing infrastructure deployment in underdeveloped and remote areas. This will help bridge the regional digital divide and establish a robust, wide-reaching, and high-quality digital infrastructure network. Simultaneously, greater emphasis should be placed on digitally transforming the entire grain supply chain. This includes achieving precise planning before production, intelligent management during production, and efficient post-production circulation and traceability. Through digital coordination across all segments, production efficiency and value creation can be significantly enhanced. This will reinforce system resilience against a variety of risks. Finally, a coordinated digital economy strategy is essential, including efforts to break down “data silos.” A cross-regional food data-sharing mechanism needs to be constructed to enhance the flow of information and resources, broadening the coverage and reach of digital economy benefits.

References

• 1

  AtheyS.TibshiraniJ.WagerS. (2019). Generalized random forests. Ann. Stat.47, 1148–1178. doi: 10.1214/18-AOS1709

  • CrossRef
  • Google Scholar

• 2

  ChernozhukovV.ChetverikovD.DemirerM.DufloE.HansenC.NeweyW.et al. (2018). Double/debiased machine learning for treatment and structural parameters. Econometrics J.21, C1–C68. doi: 10.1111/ectj.12097

  • CrossRef
  • Google Scholar

• 3

  ChinorackyR.CorejovaT. (2021). How to evaluate the digital economy scale and potential?Entrep. Sustain. Issues8, 536–552. doi: 10.9770/jesi.2021.8.4(32)

  • CrossRef
  • Google Scholar

• 4

  DavisJ. M. V.HellerS. B. (2019). Rethinking the benefits of youth employment programs: the heterogeneous effects of summer jobs. Rev. Econ. Stat.102, 1–47. doi: 10.1162/rest_a_00850

  • CrossRef
  • Google Scholar

• 5

  DengH. Y.BaiG.ShenZ. Y.XiaL. Q. (2022). Digital economy and its spatial effect on green productivity gains in manufacturing: evidence from China. J. Clean. Prod.378:134539. doi: 10.1016/j.jclepro.2022.134539

  • CrossRef
  • Google Scholar

• 6

  DongY. N.QiC. J.GuiC. J.YangY. Y. (2025). Spatial spillover effects of digital infrastructure on food system resilience: an analysis incorporating threshold effects and spatial decay boundaries. Foods14:1484. doi: 10.3390/foods14091484

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 7

  DrahokoupilJ.PiasnaA. (2017). Work in the platform economy: beyond lower transaction costs. Intereconomics52, 335–340. doi: 10.1007/s10272-017-0700-9

  • CrossRef
  • Google Scholar

• 8

  ErumbanA. A.DasD. K. (2016). Information and communication technology and economic growth in India. Telecomm Policy40, 412–431. doi: 10.1016/j.telpol.2015.08.006

  • CrossRef
  • Google Scholar

• 9

  FlaschelP.FrankeR.VenezianiR. (2013). Labour productivity and the law of decreasing labour content. Camb. J. Econ.37, 379–402. doi: 10.1093/cje/bes025

  • CrossRef
  • Google Scholar

• 10

  FrankM. R.AutorD.BessenJ. E.BrynjolfssonE.CebrianM.DemingD. J.et al. (2019). Toward understanding the impact of artificial intelligence on labor. Proc. Natl. Acad. Sci. USA116, 6531–6539. doi: 10.1073/pnas.1900949116

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 11

  Goodman-BaconA. (2021). Difference-in-differences with variation in treatment timing. J. Econom.225, 254–277. doi: 10.1016/j.jeconom.2021.03.014

  • CrossRef
  • Google Scholar

• 12

  HanX.LyuK.NieF. Y.ChenY. Q. (2024). Resilience effects for household food expenditure and dietary diversity in rural western China. J. Integr. Agric.23, 384–396. doi: 10.1016/j.jia.2023.12.010

  • CrossRef
  • Google Scholar

• 13

  HirthS.MorganE.KaptanG.SourdR. C. D.TallontireA.YoungW.et al. (2025). Restoring food system resilience in a turbulent world: supply chain actors’ shared responsibility. Bus. Strateg. Environ.34, 6007–6023. doi: 10.1002/bse.4287

  • CrossRef
  • Google Scholar

• 14

  HongM. Y.TianM. J.WangJ. (2023). The impact of digital economy on green development of agriculture and its spatial spillover effect. China Agric. Econ. Rev.15, 708–726. doi: 10.1108/CAER-01-2023-0004

  • CrossRef
  • Google Scholar

• 15

  HuD. B.GuoF.ZhaiC. Z. (2023). Digital finance, entrepreneurship and the household income gap: evidence from China. Inf. Process. Manag.60:103478. doi: 10.1016/j.ipm.2023.103478

  • CrossRef
  • Google Scholar

• 16

  HuM. J.ZhengY. J.ChenG.LiZ. J. (2025). Spatio-temporal synergies of digital economy and green finance: catalyzing green low-carbon transition in the Yangtze River Delta region. J. Environ. Manag.390:126199. doi: 10.1016/j.jenvman.2025.126199

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 17

  HuangG. Y.ShenC. H. H.WuZ. X. (2023). Firm-level political risk and debt choice. J. Corp. Finan.78:102332. doi: 10.1016/j.jcorpfin.2022.102332

  • CrossRef
  • Google Scholar

• 18

  JensenP. D.OrfilaC. (2021). Mapping the production-consumption gap of an urban food system: an empirical case study of food security and resilience. Food Secur.13, 551–570. doi: 10.1007/s12571-021-01142-2

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 19

  JiaX. M. (2023). Digital economy, factor allocation, and sustainable agricultural development. Sustainability.15:4418. doi: 10.3390/su15054418

  • CrossRef
  • Google Scholar

• 20

  JiangQ.LiJ. Z.SiH. Y.SuY. Y. (2022). The impact of the digital economy on agricultural green development: evidence from China. Agriculture12:1107. doi: 10.3390/agriculture12081107

  • CrossRef
  • Google Scholar

• 21

  JinM. M.FengY.WangS. K.ChenN.CaoF. P. (2024). Can the development of the rural digital economy reduce agricultural carbon emissions? A spatiotemporal empirical study based on China’s provinces. Sci. Total Environ.939:173437. doi: 10.1016/j.scitotenv.2024.173437

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 22

  KapoorA. P.VijM. (2018). Technology at the dinner table: ordering food online through mobile apps. J. Retail. Consum. Serv.43, 342–351. doi: 10.1016/j.jretconser.2018.04.001

  • CrossRef
  • Google Scholar

• 23

  KaranE. P.AsgariS.AsadiS. (2023). Resilience assessment of centralized and distributed food systems. Food Secur.15, 59–75. doi: 10.1007/s12571-022-01321-9

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 24

  KcU.Campbell-RossH.GoddeC.FriedmanR.Lim-CamachoL.CrimpS. (2024). A systematic review of the evolution of food system resilience assessment. Glob. Food Secur.40:100744. doi: 10.1016/j.gfs.2024.100744

  • CrossRef
  • Google Scholar

• 25

  LeSageJ. P.PaceR. K. (2008). Spatial econometric modeling of origin-destination flows. J. Reg. Sci.48, 941–967. doi: 10.1111/j.1467-9787.2008.00573.x

  • CrossRef
  • Google Scholar

• 26

  LiY. (2024). How does the development of rural broadband in China affect agricultural total factor productivity? Evidence from agriculture-related loans. Front. Sustain. Food Syst.8:1332494. doi: 10.3389/fsufs.2024.1332494

  • CrossRef
  • Google Scholar

• 27

  LinZ. H.YangP. L.ZhuZ. N.XiaM. K. (2025). Evaluating the impact of digital technology development on green total factor productivity: empirical evidence from Chinese cities. Inf. Technol. Dev.4, 1672–1694. doi: 10.1080/02681102.2025.2521273

  • CrossRef
  • Google Scholar

• 28

  LuZ. Y.YangL. Q.GouD.WuZ. Y. (2025). Promotion of rural industrial revitalization through the development of the rural digital economy. Front. Sustain. Food Syst.9:1598461. doi: 10.3389/fsufs.2025.1598461

  • CrossRef
  • Google Scholar

• 29

  MartinR.SunleyP. (2015). On the notion of regional economic resilience: conceptualization and explanation. J. Econ. Geogr.15, 1–42. doi: 10.1093/jeg/lbu015

  • CrossRef
  • Google Scholar

• 30

  MondejarM. E.AvtarR.DiazH. L. B.DubeyR. K.EstebanJ.Gómez-MoralesA.et al. (2021). Digitalization to achieve sustainable development goals: steps towards a smart green planet. Sci. Total Environ.794:148539. doi: 10.1016/j.scitotenv.2021.148539

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 31

  NunnN.QianN. (2014). Us food aid and civil conflict. Amrican Economic Review104, 1630–1666. doi: 10.1257/aer.104.6.1630

  • CrossRef
  • Google Scholar

• 32

  PooM. C. P.WangT. N.YangZ. L. (2024). Global food supply chain resilience assessment: a case in the United Kingdom. Trans. Res. Part A181. doi: 10.1016/j.tra.2024.104018

  • CrossRef
  • Google Scholar

• 33

  QiJ. L.XuJ.JinJ.ZhangS. T. (2025). Digital economy-agriculture integration empowers low-carbon transformation of agriculture: theory and empirical evidence. Sustainability17:2184. doi: 10.3390/su17052183

  • CrossRef
  • Google Scholar

• 34

  SachsJ. D. (2019). Some brief reflections on digital technologies and economic development. Ethics Int. Aff.33, 159–167. doi: 10.1017/S0892679419000133

  • CrossRef
  • Google Scholar

• 35

  SadiqM.ChienF. S.LeongM. K.VermaS.MirazM. H. (2025). Toward a circular path: integrating knowledge, open innovation, green HRM, entrepreneurship, and digital orientation in Chinese fashion industry. Corp. Soc. Responsib. Environ. Manag.32, 4432–4447. doi: 10.1002/csr.3194

  • CrossRef
  • Google Scholar

• 36

  SualiA. S.SraiJ. S.TsolakisN. (2024). The role of digital platforms in e-commerce food supply chain resilience under exogenous disruptions. Supply Chain Manag.29, 573–601. doi: 10.1108/SCM-02-2023-0064

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 37

  TaghikhahF. R.PriorD.HafeziR.BakerD.MatousP. (2025). Understanding digital capabilities and their impacts on Australian Agri-food supply chain resilience: engineering vs. socio-ecological thinking. Technol. Forecast. Soc. Change218:124191. doi: 10.1016/j.techfore.2025.124191

  • CrossRef
  • Google Scholar

• 38

  TendallD. M.JoerinJ.KopainskyB.EdwardsP.EdwardsP.ShreckA.et al. (2015). Food system resilience: defining the concept. Glob. Food Secur.6, 17–23. doi: 10.1016/j.gfs.2015.08.001

  • CrossRef
  • Google Scholar

• 39

  TranosE.KitsosT.Ortega-ArgilesR. (2021). Digital economy in the UK: regional productivity effects of early adoption. Reg. Stud.55, 1924–1938. doi: 10.1080/00343404.2020.1826420

  • CrossRef
  • Google Scholar

• 40

  WangH. F.LiG. S.HuY. Z. (2023). The impact of the digital economy on food system resilience: insights from a study across 190 Chinese towns. Sustainability15:16898. doi: 10.3390/su152416898

  • CrossRef
  • Google Scholar

• 41

  WangJ. Y.LinQ. N.ZhangX. B. (2024). How does digital economy promote agricultural development? Evidence from sub-Saharan Africa. Agriculture-Basel.14:63. doi: 10.3390/agriculture14010063

  • CrossRef
  • Google Scholar

• 42

  WangL. H.ShaoJ. (2023). Digital economy, entrepreneurship and energy efficiency. Energy269:126801. doi: 10.1016/j.energy.2023.126801

  • CrossRef
  • Google Scholar

• 43

  WangY.WangY.ShahbazM. (2023). How does digital economy affect energy poverty? Analysis from the global perspective. Energy282:128692. doi: 10.1016/j.energy.2023.128692

  • CrossRef
  • Google Scholar

• 44

  WangX. A.ZhongM. (2023). Can digital economy reduce carbon emission intensity? Empirical evidence from China’s smart city pilot policies. Environ. Sci. Pollut. Res.30, 51749–51769. doi: 10.1007/s11356-023-26038-w

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 45

  WenH. W.HuK. Y.NghiemX. H.AcheampongA. O. (2024). Urban climate adaptability and green total-factor productivity: evidence from double dual machine learning and differences-in-differences techniques. J. Environ. Manag.350:14. doi: 10.1016/j.jenvman.2023.119588

  • CrossRef
  • Google Scholar

• 46

  WingC.YozwiakM.HollingsworthA.FreedmanS.SimonK. (2024). Designing difference-in-difference studies with staggered treatment adoption: key concepts and practical guidelines. Annu. Rev. Public Health45, 485–505. doi: 10.1146/annurev-publhealth-061022-050825

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 47

  WuK. P.FuY. M.KongD. M. (2022). Does the digital transformation of enterprises affect stock price crash risk?Financ. Res. Lett.48:102888. doi: 10.1016/j.frl.2022.102888

  • CrossRef
  • Google Scholar

• 48

  WuM.MaY.GaoY.JiZ. H. (2024). The impact of digital economy on income inequality from the perspective of technological progress-biased transformation: evidence from China. Empir. Econ.67, 567–607. doi: 10.1007/s00181-024-02563-6

  • CrossRef
  • Google Scholar

• 49

  XiaoY. T.AbulaB. (2023). Examining the impact of digital economy on agricultural trade efficiency in RCEP region: a perspective based on spatial spillover effects. J. Knowl. Econ.15, 9907–9934. doi: 10.1007/s13132-023-01484-6

  • CrossRef
  • Google Scholar

• 50

  XieY.YaoR. K.WuH. T.LiM. D. (2025). Digital economy, factor allocation, and resilience of food production. Land14:139. doi: 10.3390/land14010139

  • CrossRef
  • Google Scholar

• 51

  YangC. F.JiX.ChengC. M.LiaoS.BrightO.ZhangY. F. (2024). Digital economy empowers sustainable agriculture: implications for farmers’ adoption of ecological agricultural technologies. Ecol. Indic.159:111723. doi: 10.1016/j.ecolind.2024.111723

  • CrossRef
  • Google Scholar

• 52

  YangX. D.WuH. T.RenS. Y.RanQ. Y.ZhangJ. N. (2021). Does the development of the internet contribute to air pollution control in China? Mechanism discussion and empirical test. Struct. Change Econ. Dyn.56, 207–224. doi: 10.1016/j.strueco.2020.12.001

  • CrossRef
  • Google Scholar

• 53

  YuL. Z.WangJ. N.LouS. Y.WeiX. H. (2024). To leave or to stay: digital economy development and migrant workers’ location. J. Asian Econ.94:101792. doi: 10.1016/j.asieco.2024.101792

  • CrossRef
  • Google Scholar

• 54

  ZengH. S.ZhangB. H.YanY.HuangC. Y. (2025). How pilot free trade zones affect food system resilience: quasi-natural experiment evidence from China. Front. Sustain. Food Syst.9:1460485. doi: 10.3389/fsufs.2025.1460485

  • CrossRef
  • Google Scholar

• 55

  ZhangB. C.DongW. H.YaoJ.ChengX. L. (2023). Digital economy, factor allocation efficiency of dual-economy and urban-rural income gap. Sustainability15:13514. doi: 10.3390/su151813514

  • CrossRef
  • Google Scholar

• 56

  ZhaoK.YuB. T.YangX. T. (2023). The agricultural-ecological benefit of digital inclusive finance development: evidence from straw burning in China. Sustainability15:3242. doi: 10.3390/su15043242

  • CrossRef
  • Google Scholar

• 57

  ZhengJ. J.WangX. W. (2022). Impacts on human development index due to combinations of renewables and ICTs - new evidence from 26 countries. Renew. Energy191, 330–344. doi: 10.1016/j.renene.2022.04.033

  • CrossRef
  • Google Scholar

• 58

  ZhuZ.WangZ.YuS.TangZ. M.LiuB. (2024). The impact of the digital economy on food system resilience: evidence from China. PLoS One19:16898. doi: 10.1371/journal.pone.0311689

  • CrossRef
  • Google Scholar

• 59

  ZurekM.IngramJ.BellamyA. S.GooldC.LyonC.AlexanderP.et al. (2022). Food system resilience: concepts, issues, and challenges. Annu. Rev. Environ. Resour.47, 511–534. doi: 10.1146/annurev-environ-112320-050744

  • CrossRef
  • Google Scholar