How Does Digital Rural Construction Enhance Agricultural Land Green Utilization Efficiency? Mechanism Analysis and Empirical Testing

Abstract

Amid the coordinated advancement of the digital economy and rural revitalization, Digital Rural Construction (DRC) has increasingly emerged as a critical catalyst for agricultural modernization and sustainable development. Faced with dual challenges of land resource constraints and agricultural green transformation, improving the Agricultural Land Green Utilization Efficiency (ALGUE) has become essential for achieving high-quality agricultural development. Based on panel data from 29 Chinese provinces from 2012 to 2023, this study employs the super-efficiency SBM model to quantify ALGUE. A comprehensive four-dimensional evaluation system—encompassing digital infrastructure, service capacity, human capital quality, and practical application—is constructed, and the entropy method is used to measure the level of digital rural construction. By applying two-way fixed effects models, mediation analysis, and heterogeneity tests, this study systematically examines the impact of digital rural construction on ALGUE and its underlying transmission pathways. The results demonstrate that: (1) Digital rural construction significantly enhances ALGUE, and this finding remains robust under multiple sensitivity checks. (2) Pronounced heterogeneity exists in two dimensions: the promotion effect is stronger in economically developed regions and in regions with higher agricultural mechanization intensity, while it is weaker in less developed and low-mechanization regions. (3) Mechanism analysis reveals that digital rural construction promotes ALGUE through two channels. The first involves accelerating the transition of the primary industry toward intelligent and high-value-added models, thereby optimizing resource allocation and reducing environmental pressure. The second operates by fostering regional economic growth in an inverted U-shaped nonlinear pattern that supports agricultural green transformation. By integrating DRC and ALGUE into a unified framework, this study identifies two mediating channels and reveals heterogeneity across economic development levels and agricultural structures. These findings provide empirical support and policy implications for digitally driven green agricultural development.

1. Introduction

Climate change poses an increasingly severe threat to the sustainable development of human society and the economy. As a key environmental driver, agricultural systems account for approximately one-quarter of global greenhouse gas (GHG) emissions and occupy half of the world’s ice-free land [1]. Against the backdrop of continuous global population growth and rising food demand, efficiently utilizing limited agricultural land resources has become a central challenge for addressing land scarcity and ensuring food security. Therefore, accelerating the transformation of agricultural production methods and achieving the greening and decarbonization of land use are critically important for resolving resource and environmental constraints in agriculture and promoting sustainable development. The green transformation of land use is central to reducing agricultural environmental pressure and advancing green agricultural development [2].

In this context, Digital Rural Construction (DRC) has emerged as an important driver for optimizing the allocation of agricultural land resources. The Opinions on the Implementation of the Rural Revitalization Strategy (2018) explicitly identified the “Digital Rural Strategy” as a key policy initiative for advancing agricultural and rural modernization. More recently, the 2025 Central Document No. 1 underscored the need to foster green and low-carbon land use patterns and to boost a comprehensive transformation of agriculture and rural areas, with the dual objectives of rural revitalization and building China into a strong agricultural nation. China has long faced persistent challenges such as a strained human–land relationship and inefficient land use patterns. Consequently, enhancing the efficiency and sustainability of land resource utilization is critical to national food security and social stability.

By facilitating the deep integration of information technology with agriculture and rural development, DRC offers new pathways for optimizing land resource allocation and enhancing the comprehensive benefits of land utilization. Fundamentally, land use efficiency reflects the combined economic, social, and ecological returns generated per unit of land resource. Traditional agriculture has long been constrained by information asymmetry, which impedes the effective allocation of land resources. In contrast, DRC leverages frontier technologies such as the Internet of Things (IoT), big data, and Artificial Intelligence (AI) to enable precision monitoring, intelligent management, and dynamic optimization of land use. These capabilities provide technical support and systematic solutions for enhancing ALGUE. Specifically, intelligent sensing terminals (e.g., sensors and remote sensing devices) facilitate the real-time collection of multi-dimensional data on soil conditions, climate variables, and crop growth, providing a robust foundation for precision farm management [3]. Furthermore, data analysis and decision support systems powered by IoT and AI drive agricultural production toward intelligent and adaptive practices. These systems enhance economic efficiency while reducing resource consumption and environmental pollution, thereby supporting the green and low-carbon transformation of agriculture.

Despite these advances, existing studies mainly focus on cultivated land use pattern transformation [4], spatiotemporal evolution [5], and spatial transformation [6]. Systematic research linking land green utilization efficiency to rural digital development remains scarce. At the same time, China’s rural digitalization process still faces many practical constraints, such as insufficient infrastructure coverage, a low level of agricultural digitization, and a shortage of digital talent. Therefore, exploring the mechanism through which DRC affects ALGUE can not only enrich relevant theories but also provide a scientific basis for policy formulation and practical promotion.

To address these gaps, this study focuses on three research questions: (1) Does DRC significantly improve ALGUE? (2) Does this effect exhibit significant regional heterogeneity? (3) What are the underlying transmission mechanisms through which DRC influences ALGUE via primary industry upgrading and regional economic development?

This study contributes to the existing literature in three ways. First, unlike prior studies that examine agricultural land use efficiency or digital rural construction in isolation, this study integrates both dimensions within a unified analytical framework, providing a more comprehensive understanding of how DRC shapes ALGUE. Second, by employing a mediation model, this study identifies two complementary transmission channels—primary industry upgrading and regional economic development—through which DRC indirectly enhances ALGUE. Notably, the latter channel exhibits an inverted U-shaped nonlinear pattern. This dual-pathway mechanism extends the existing understanding beyond direct effects to reveal the intermediate processes underlying digital empowerment in agriculture. Third, the heterogeneity analysis reveals that the enabling effect of DRC on ALGUE varies significantly across regions with different levels of economic development, offering differentiated policy implications for bridging the digital divide and promoting balanced regional development in agricultural green transformation.

Empirically, this study draws on panel data from 29 Chinese provinces over 2012–2023. The super-efficiency SBM model is employed to measure ALGUE incorporating undesirable outputs, and empirical analysis is conducted using two-way fixed effects models, mediation analysis, and a battery of robustness checks. The findings are expected to offer empirical support for optimizing rural digitalization pathways and advancing the sustainable utilization of agricultural land resources.

The remainder of this paper is structured as follows: The literature review is presented in Section 2; the theoretical analysis and research hypotheses are formulated in Section 3; the research design is covered in Section 4; the empirical results and analysis are covered in Section 5; and the conclusion and implications are covered in Section 6.

2. Literature Review

2.1. Evolution of Research on Digital Rural Construction

DRC is not merely a superimposition of digital technologies onto existing rural systems, but rather a long-term, innovative, and dynamically evolving systemic process. Since Central Document No. 1 first proposed “implementing the digital rural strategy” in 2018, related research has gradually shifted from early conceptual interpretations, qualitative analyses, and pathway explorations toward more in-depth mechanism investigations and empirical testing. The academic community generally views DRC as an embodiment of deep integration between information technology and the agricultural and rural sectors, aiming to drive the comprehensive transformation of rural economic and social development through digitalization. Current research can be broadly categorized into three dimensions.

First, analysis of enabling mechanisms. Existing studies seek to identify the specific pathways through which DRC influences rural development. For example, Deng et al. [7] developed county-level indices of rural revitalization and digital rural construction to empirically examine how digital technology facilitates rural revitalization.

Second, evaluation of policy instruments. Existing studies have examined both policy design and implementation outcomes. Li et al. [8] used Shandong Province as a case study and constructed a two-dimensional framework linking policy tools to digital rural construction. Using content analysis, they quantitatively assessed relevant policies and developed an evaluation index system for DRC, thereby providing evidence for policy optimization. Wu et al. [9] built a theoretical framework to empirically investigate the dynamic and spatial effects of DRC on farmers’ common prosperity. Furthermore, Jing et al. [10] demonstrated that DRC plays an important role in shaping public policies oriented toward green production and sustainable agriculture.

Third, research on economic and social effects. Empirical evidence indicates that DRC enhances county-level economic resilience by improving regional financial development, and this effect exhibits significant regional heterogeneity [11,12]. The research scope has further expanded to agricultural sustainability and rural well-being. Recent studies have examined the impact of DRC on agricultural green total factor productivity [13] and, using ordered probit models, found that DRC improves rural residents’ well-being by increasing household income and alleviating loneliness [14].

2.2. Research Progress on Agricultural Land Green Utilization Efficiency

As an emerging research topic in land resource management, ALGUE centers on balancing the economic outputs, social benefits, and ecological impacts of land use to maximize comprehensive benefits under resource constraints. With the global shift toward sustainable development and China’s “dual carbon” goals, research has evolved from traditional economic efficiency assessments to multidimensional green efficiency frameworks that integrate economic, social, and ecological dimensions [15,16]. The existing literature can be reviewed along three aspects.

First, the conceptual evolution: from economic efficiency to systemic coordination. Early studies primarily focused on the intensive utilization of cultivated or urban construction land. Following the integration of the Sustainable Development Goals (SDGs) and the “dual carbon” strategy, the connotation of ALGUE has expanded to emphasize the synergy of “economic efficiency, resource conservation, and environmental friendliness” [17]. Recently, the concept has become more scenario-based and refined, incorporating specific indicators such as fertilizer and pesticide reduction, carbon emission control, and soil health maintenance. This has formed a systemic yet targeted conceptual framework for coordinating food security, ecological protection, and urbanization [17,18].

Second, in terms of measurement methods, the field has advanced from static evaluations to dynamic systems that account for spatial correlation and technological heterogeneity. While Data Envelopment Analysis (DEA) remains a foundational tool, its limitations in handling undesirable environmental outputs can bias efficiency estimations [19]. To address this, scholars have increasingly adopted the super-efficiency SBM model and directional distance functions. These approaches treat environmental outputs as undesirable and employ slack-based measures for more accurate ALGUE estimation [16,20]. Additionally, Stochastic Frontier Analysis (SFA) is frequently employed to estimate technical and environmental efficiencies [21,22].

Third, regarding influencing factors, digital technology is widely recognized as a primary driver of ALGUE. Empirical evidence suggests that the digital economy enhances efficiency by fostering agricultural technological innovation and optimizing cultivated land transfer, yielding significant spatial spillover effects [23,24]. Specifically, technologies such as the IoT, remote sensing, and big data enable precision monitoring and intelligent management, thereby reducing resource misallocation and environmental losses [18]. For instance, smart agriculture, leveraging variable-rate fertilization and precision irrigation, minimizes agricultural non-point source pollution while maintaining yields, thereby contributing to improved green efficiency in farmland systems [23]. Furthermore, financial support, human capital accumulation, and marketization have also been shown to positively influence ALGUE [25].

2.3. Digital Empowerment Pathways for Enhancing ALGUE

International scholarship on digital rural construction and agricultural land use predates its Chinese counterpart and has established relatively mature theoretical and methodological foundations. In Europe and the United States, a large number of empirical studies have confirmed that precision agriculture technologies represented by remote sensing monitoring [26], intelligent irrigation, variable-rate fertilization [27], and smart monitoring equipment [28] can significantly improve resource utilization efficiency, reduce factor input redundancy, and effectively reduce agricultural carbon emissions and non-point source pollution [29]. These studies provide an important international reference for understanding the mechanism of digital technology in improving agricultural green development.

Compared with Europe and the United States, China’s digital rural construction is characterized by stronger policy-driven forces, a dominant smallholder production structure, and significant regional heterogeneity [30]. Therefore, the mechanism and effect of digital empowerment in China are different from those in developed countries [31]. Conceptually, DRC entails the co-evolution of infrastructure upgrading, data-factor activation, resource reallocation, and technology embedding—a systemic process that, through input restructuring and improved information flows, mitigates resource misallocation and cost inefficiencies in agricultural production, thereby facilitating the shift toward larger-scale, higher-efficiency operations. Xue et al. [32] further argue that integrating the digital economy with agriculture helps dismantle information barriers, reconfigure resource allocation patterns, reduce transaction costs, and strengthen scale economies.

Against this backdrop, the interaction between DRC and ALGUE has emerged as an active research frontier. Three recurring themes can be identified in the existing literature. First, the digital economy enhances land-use and production performance by supplying technological platforms that steer agriculture away from extensive growth patterns [33,34]. Second, digital technology, as an emerging form of productive force, reshapes conventional production logic through disruptive innovation, thereby enabling green and intensive agricultural development and improving ALGUE [35,36]. Third, digitalization lowers production costs and information frictions through optimized factor allocation [37], while simultaneously upgrading labor structure and land-use efficiency [35]—together amplifying the modernization dividend of agriculture. Recent studies further delineate the specific channels at work. Cao et al. [38] find that rural digital transformation raises ALGUE through three pathways: optimizing input allocation structure, expanding desirable outputs, and curbing undesirable outputs such as pollutant emissions. Tan et al. [39] show that digital inclusive finance significantly enhances the green utilization efficiency of cultivated land via a bidirectional “input–output” optimization mechanism characterized by more precise resource targeting and higher productivity gains.

2.4. Literature Review and Research Gap

Existing studies provide an important basis for understanding the association between DRC and ALGUE. However, three limitations remain. First, methodologically, much of the literature emphasizes correlational evidence, with relatively limited empirical testing of causal mechanisms underlying this relationship. Second, regarding heterogeneity, studies on green land-use efficiency often focus on urban areas or specific spatial scales of cultivated land and frequently rely on spatial spillover models; as a result, regional heterogeneity is not fully unpacked, and differential policy effects across regions remain insufficiently identified. Third, in terms of research scope, relatively few studies provide an integrated assessment that jointly considers ecological–environmental and socio-economic dimensions.

To address these gaps, this study constructs a provincial panel dataset covering 29 Chinese provinces from 2012 to 2023 and systematically examines the effects of DRC on ALGUE, regional heterogeneity, and the underlying transmission mechanisms. Methodologically, we measure ALGUE with a super-efficiency SBM model that incorporates undesirable outputs. We then estimate baseline effects using a two-way fixed effects model to control for unobserved province- and time-specific factors, and we employ mediation analysis to identify key transmission channels. Heterogeneity is further explored by grouping provinces according to per-capita GDP.

To strengthen the robustness of our findings, we implement three sets of robustness checks: (1) re-estimating the models using an alternative sample period (2013–2022); (2) randomly excluding a subset of provinces to assess sensitivity to sample composition; and (3) adopting a Tobit model as an alternative specification and implementing an instrumental-variable approach-using computer penetration rate as the instrument-to mitigate potential endogeneity concerns.

3. Theoretical Analysis and Research Hypotheses

3.1. Digital Technology Empowerment for the Improvement of ALGUE

Existing studies indicate that the penetration and integration of digital technologies in agriculture can significantly enhance the intelligence of production processes, thereby improving the efficiency of factor allocation and operational precision [36]. Beyond boosting agricultural productivity, digital technologies also provide crucial technical support for advancing agriculture’s green transformation and achieving the “dual-carbon” goals [40].

This effect primarily operates through optimizing the allocation of production factors and strengthening precision management. By enabling more accurate decision-making and process control, digital technologies reduce excessive inputs of chemical fertilizers and pesticides and curb resource waste. This in turn mitigates agricultural non-point source pollution and improves comprehensive land-use performance.

From the perspective of efficiency dynamics, improvements in technical efficiency, technological progress, and the narrowing of technology gaps jointly constitute the internal drivers of enhanced ALGUE [41]. For instance, precision agriculture enables real-time monitoring and feedback, supporting optimized fertilization and pesticide application strategies, improving nutrient-use efficiency, and reducing non-point source pollution. Intelligent irrigation systems adjust water supply based on dynamic soil moisture data, reducing water waste and lowering production costs. In addition, digital farm management models—through the integrated application of information technologies—can increase output value per unit of land and strengthen the economic productivity of land resources.

Building on the above reasoning, this study proposes the following Hypothesis H1:

Hypothesis H1.

Digital rural construction (DRC) significantly enhances agricultural land green utilization efficiency (ALGUE).

3.2. The Moderating Effect of Regional Economic Development

The effect of DRC on ALGUE may exhibit regional heterogeneity, with the level of regional economic development serving as an important moderating factor. Prior studies suggest that regional economic development constitutes a key contextual condition shaping both the diffusion of digital technologies and their implementation outcomes [42,43]. Zhang et al. [44] show that higher GDP per capita is typically associated with stronger digital infrastructure investment, a more dynamic innovation environment, and greater human capital accumulation. Together, these conditions facilitate the full realization of digital technologies’ productive potential.

Further evidence indicates a positive relationship between economic development and the adoption of smart agriculture technologies. Economically developed regions tend to integrate digital tools such as the IoT, big data, AI, and remote sensing more systematically and effectively. This moderating mechanism operates mainly through two channels. First, more developed regions generally feature more complete agricultural value chains and higher levels of marketization, enabling digital technologies to penetrate the entire process—from precision production to supply-chain management—and thereby optimize resource allocation at the system level. For example, capital- and technology-intensive solutions, including intelligent irrigation systems and pest and disease forecasting models, are more feasible in areas with stronger financial support [45]. Second, economic development is closely linked to human capital formation [46,47]. As Deng et al. [7] note, in regions with higher educational attainment, farmers are better positioned to leverage digital platforms for data-driven production decisions, thereby improving both the economic and environmental performance of land use.

By contrast, less developed regions often face practical constraints, such as insufficient digital infrastructure coverage, limited capacity for equipment operation and maintenance, and tighter financial constraints [8]. Moreover, constrained fiscal resources tend to be allocated to sectors with short-term economic returns rather than long-term digital investment and green transformation [48]. These constraints may delay the realization of digital dividends and, consequently, improvements in ALGUE.

Notably, due to the diminishing marginal effect, regions with higher economic development levels already have relatively mature infrastructure and efficient land use systems, so the marginal improvement brought by digital rural construction is limited. As a result, the positive promotion effect of DRC on ALGUE weakens gradually with the improvement of regional economic development.

Building on the above reasoning, this study proposes the following Hypothesis H2:

Hypothesis H2.

Regional economic development negatively moderates the impact of DRC on ALGUE. The positive effect of DRC on ALGUE becomes weaker as the level of regional economic development increases.

3.3. Two Separate Mediating Pathways: Industrial Upgrading and Economic Development

Beyond its direct effect on ALGUE, DRC may also exert indirect influence through two mediating pathways: (1) promoting the upgrading of the primary industry, and (2) reshaping the level of regional economic development.

Regarding the first pathway, DRC drives the transformation of agricultural production from conventional extensive practices toward a technology-intensive “smart agriculture” paradigm. Digitization promotes primary industry upgrading through three specific and interconnected mechanisms.

First, production mode transformation. Digital technologies including IoT, big data, and AI enable real-time monitoring, precise fertilization, intelligent irrigation, and disease and pest early warning, thereby transforming traditional extensive farming into smart and precision agriculture. This transformation reduces factor input redundancy, improves production efficiency, and lowers agricultural non-point source pollution.

Second, value chain extension. Digital platforms promote the deep integration of agriculture with processing, logistics, rural e-commerce, branded operation, and agritourism. By extending the agricultural value chain and increasing high-value-added links, digitization raises land output efficiency and economic returns while promoting green and intensive land use.

Third, factor structure upgrading. Data, technology, and modern capital gradually replace traditional labor and land as the core driving factors of agricultural growth. Digitalization optimizes factor allocation, reduces dependence on high-pollution inputs, and pushes the primary industry toward a more efficient, intelligent, and green structure.

Together, these three mechanisms define a smart agriculture paradigm characterized by precision irrigation, AI-assisted pest and disease management, and data-driven crop monitoring throughout the production cycle. Such a structural transition enhances factor allocation efficiency and increases output value added, thereby improving ALGUE through a dual mechanism of raising land productivity per unit area and reducing the environmental footprint [49,50]. In particular, the deployment of digital tools—including the IoT and big data analytics—has been shown to substantially curtail resource waste and elevate ecological efficiency. Digitalization is thus increasingly recognized as a critical enabler of “ecological intensification” [51], facilitating the decoupling of agricultural output growth from environmental degradation.

Regarding the second pathway, DRC requires considerable upfront investment in infrastructure, human capital development, and system integration. During the early stages of implementation, the diversion of resources from directly productive sectors may exert a temporary dampening effect on short-term economic growth [52]. Nevertheless, as digital technologies progressively mature and permeate the entire agricultural value chain, their productivity-enhancing potential becomes increasingly pronounced. In the early stage, digital empowerment significantly boosts factor allocation efficiency and market vitality, thereby promoting regional economic development. However, as digital construction deepens, problems such as resource input redundancy, declining marginal infrastructure efficiency, and talent mismatches may emerge. These issues gradually weaken the growth effect of digitalization. Therefore, the impact of DRC on regional economic development presents an inverted U-shaped nonlinear characteristic of “first promotion and then marginal diminishing”. The resulting elevation in regional economic development, in turn, creates more favorable market conditions, a stronger technological base, and a more supportive institutional environment for advancing ALGUE. This reasoning implies that DRC may exert a nonlinear mediating influence on ALGUE via regional economic development [53].

Building on the above reasoning, this study proposes the following Hypothesis H3:

Hypothesis H3.

DRC influences ALGUE not only directly but also indirectly through two separate mediating channels: (i) by promoting the upgrading of the primary industry; and (ii) by exerting an inverted U-shaped nonlinear impact on regional economic development, that is, DRC promotes regional economic development in the early stage, and its promotion effect weakens after exceeding a certain threshold.

4. Research Design

4.1. Variable Definitions

4.1.1. Dependent Variable: Agricultural Land Green Utilization Efficiency (ALGUE)

ALGUE refers to the capacity of land use inputs to generate economic and environmental benefits under given technical conditions. The concept centers on the dual objectives of high efficiency and environmental sustainability, emphasizing economic output while simultaneously accounting for the ecological performance of land use. Four categories of input factors are considered: land, capital, labor, and energy. The specific input indicators include total crop sown area (hm²), chemical fertilizer application (10,000 tons), agricultural plastic film use (tons), pesticide application (tons), and primary-industry employment (persons). Desirable outputs capture both economic and social benefits, measured by the gross output value of agriculture, forestry, animal husbandry, and fishery (100 million yuan) and per capita farmer income (yuan). Undesirable output focuses on environmental pollution, proxied by carbon emissions (10,000 tons) (

Table 1. Evaluation Index System for ALGUE.

Level 1 Indicator	Level 2 Indicator	Level 3 Indicator	Measurement Unit
Input Indicators	Land Input	Total Crop Sown Area	hm2
	Capital Input	Chemical Fertilizer Application	10,000 tons
		Agricultural Plastic Film Use	tons
		Pesticides Application	tons
	Labor Input	Primary-industry Employment	person
Output Indicators	Desirable Output	Gross Output Value of Agriculture, Forestry, Animal Husbandry and Fishery	100 million yuan
		Per Capita Farmer Income	yuan
	Undesirable Output	Agricultural Carbon Emissions	10,000 tons

).

This study employs the SBM model with undesirable outputs to measure ALGUE. The main reason is that it accommodates multiple inputs, desirable outputs, and undesirable outputs simultaneously, while avoiding the radial and angular measurement biases inherent in conventional DEA models. It thus provides a more accurate and robust assessment of the trade-off between economic returns and environmental costs in agricultural land use. The model is formulated as follows:

$$\begin{array}{l}ALGUE=\min {\displaystyle \frac{1+\frac{1}{m}\times {\displaystyle \sum _{i=1}^m\frac{S_i^{-}}{X_{ik}}}}{1-\frac{1}{S_1+S_2\times \left({\displaystyle \sum _{i=1}^{S_1}\frac{y_r^g}{y_{rk}^g}+{\displaystyle \sum _{i=1}^{S_2}\frac{y_r^g}{y_{rk}^g}}}\right)}}}\\ s.t\left\{\begin{array}{c} \\ x_i^k\ge {\displaystyle {\sum }_{j=1,j\ne k}^n{x}_{ij}{\lambda }_j-s_i^{-}}\\ y_{\gamma k}^g\ge {\displaystyle {\sum }_{j=1,j\ne k}^n{y}_{\gamma j}^g{\lambda }_j-y_{\gamma }^d}\\ y_{tk}^{ug}\ge {\displaystyle {\sum }_{j=1,j\ne k}^n{y}_{tj}^{ug}{\lambda }_j+y_{\gamma }^{ug}}\\ 1-{\displaystyle \frac{1}{S_1+S_2}}\left({\displaystyle {\sum }_{i=1}^{S_1}\frac{y_{\gamma }^g}{y_{\gamma k}^g}+{\displaystyle {\sum }_{i=1}^{S_2}\frac{y_{\gamma }^g}{y_{\gamma k}^g}}}\right)>0\\ i=1,2,\dots ,m;j=1,2,\dots ,n\left(n\ne k\right);\gamma =1,2,\dots ,S_1;k=1,2,\dots ,S_2\end{array}\right.\end{array}$$

(1)

In the formula, ALGUE denotes the green utilization efficiency of agricultural land in the current year; $m$ denotes the number of inputs; $S_1$ denotes the number of desirable outputs; $S_2$ denotes the number of undesirable outputs; $S_{\mathrm{i}}^{-}$ and $X_{\mathrm{i}}$ denote input slacks and input variables, respectively; $S_r^g$ and $Y_r^g$ denote desirable-output shortfalls and desirable-output variables, respectively; $S_k^b$ and $Y_k^b$ denote undesirable-output surpluses and undesirable-output variables, respectively.

The undesirable output variable is agricultural carbon emissions (ACE). Drawing on established methodologies and prior studies [54,55,56], ACE is derived from six sources: fertilizer input, pesticide input, agricultural plastic film use, cultivated land area, electricity consumption for irrigation, and agricultural machinery fuel consumption.

Table 2. Data Sources and Calculation Methods for Agricultural Carbon Emission Sources.

Emission Source	Calculation Method	Unit of Measurement	Data Source
fertilizer input	Annual provincial fertilizer application (discounted pure amount)	10,000 tons	the China Rural Statistical Yearbook
pesticide input,	Annual provincial pesticide usage	Tons	the China Rural Statistical Yearbook
agricultural plastic film use	Annual provincial plastic film usage	tons	the China Agricultural Statistical Yearbook
cultivated land area	Actual sown area of crops in each province	hm2	the China Statistical Yearbook
electricity consumption for irrigation,	Estimated based on effective irrigated area and average electricity consumption per unit area	10,000 kWh	the China Water Resources Statistical Yearbook
agricultural machinery fuel consumption	Derived from total mechanical power and average fuel consumption per unit power	kW	the China Agricultural Machinery Industry Yearbook

presents the specific data sources and calculation methods for each emission source. The corresponding carbon emission coefficients are: chemical fertilizers (0.8956 kg/kg), pesticides (4.9341 kg/kg), agricultural plastic film (5.18 kg/kg), cultivated land (312.6 kg/hm²), irrigation electricity (266.48 kg/kWh), and agricultural machinery fuel (0.18 kg/kW). Total carbon emissions are calculated by multiplying the absolute quantity of each emission source by its corresponding coefficient:

$$C={\displaystyle \sum C_{\mathrm{i}}}={\displaystyle \sum S_{\mathrm{i}}}\times {\delta }_{\mathrm{i}}$$

(2)

where $C$ denotes the total agricultural carbon emissions, $S_i$ represents the quantity of the i-th carbon emission source, and ${\delta }_i$ stands for the carbon emission coefficient of the i-th source.

4.1.2. Core Explanatory Variable: Digital Rural Construction

DRC is a systematic project whose internal elements exhibit complex synergistic and complementary relationships. These elements not only influence agricultural technological progress and technical efficiency through independent pathways, but also generate synergistic effects via system integration. Such synergies collectively drive the enhancement of agricultural total factor productivity (TFP) and promote high-quality development. Notably, the information literacy of rural residents constitutes a vital human capital foundation for the sustainable development of DRC. Drawing on existing indicator frameworks for DRC [7,8,30,31], this study constructs a comprehensive evaluation system encompassing four dimensions: digital infrastructure, digital service level, digital literacy cultivation, and practical digital application. The specific indicators under each dimension and their weights derived from the entropy weight method (EWM) are presented in

Table 3. Evaluation Index System for DRC.

	Level 2 Indicator	Indicator Explanation	Measurement Unit	Indicator Weight
Digital Infrastructure	Computer Penetration Rate	Number of Computers per 100 People	units	0.0639
	Radio and Television Coverage	Number of Rural Cable Radio and Television Users	10,000 households	0.1284
	Smartphone Penetration Rate	Annual Mobile Phone Ownership per 100 Rural Households	units per 100 people	0.0572
Digital Service Level	Online Transaction and Payment Level	Digital Inclusive Finance Index (DIFI)		0.0434
	Postal and Communication Service Level	Proportion of Administrative Villages with Postal Service	%	0.0034
	Rural Household Electrification Level	Rural Electricity Consumption	100 million kWh	0.2960
Digital Literacy Cultivation	Farmers’ Digital Learning Ability	Average Years of Education per Rural Resident	years	0.0130
	Digital Talent Development Outcomes	Number of Authorized Agricultural Science and Technology Patents	items	0.0843
Practical Digitalization Application	Digitalization of Agricultural Production	Number of Agricultural Meteorological Observation Stations	units	0.0740
	Rural Information Penetration Degree	Number of Rural Broadband Access Users 1	units per 100 people	0.2365

¹ Note: Rural broadband subscribers reflect the actual application and penetration of digital services in rural areas rather than only infrastructure construction; therefore, it is classified under the “Practical Digitalization Application” dimension.

. Among them, the level of online transaction and payment is measured by the Digital Inclusive Finance Index (DIFI). As a comprehensive indicator, DIFI systematically reflects the overall development of digital payment, online transaction, and digital financial services in rural areas, making it appropriate to represent the level of rural online transaction and payment services. Acknowledging methodological constraints, the entropy weight method (EWM) provides an objective weighting scheme that mitigates subjective bias. However, because weights are strictly data-driven, indicators with higher theoretical importance may be assigned lower weights if their temporal or spatial variation is limited. Additionally, EWM does not adjust for potential inter-correlations among indicators, which may affect the index structure.

4.1.3. Mediating Variables

This study employs per capita GDP and value added of the primary industry as core mediating variables to reveal the indirect transmission mechanisms through which DRC affects ALGUE. The specific definitions and measurements are as follows:

(a)

Per capita GDP (lnGDP): This indicator serves as a comprehensive proxy variable for regional economic development, reflecting the overall scale and growth quality of the regional economy. It represents a key transmission channel through which DRC affects ALGUE. The variable is derived from provincial per capita GDP and log-transformed to reduce heteroscedasticity and improve data stationarity.

(b)

Value added of the primary industry (lnGVA): This variable captures the development scale and output level of the primary industry (agriculture, forestry, animal husbandry, and fishery), directly reflecting the actual effects of agricultural industrial upgrading. It constitutes a critical industrial mediating pathway through which DRC enhances ALGUE. The variable is derived from provincial primary industry value added and log-transformed to satisfy the assumptions of econometric modeling.

4.1.4. Control Variables

To mitigate potential omitted variable bias and enhance the reliability of the estimation results, this study incorporates seven control variables following Cai [13] and Cheng [57]. These variables are selected to account for key agricultural production conditions and policy factors that may simultaneously influence agricultural land green utilization efficiency:

(1) Agricultural Machinery Power (AMP), capturing the level of agricultural mechanization; (2) Fiscal Support for Agriculture (FSA), reflecting government financial investment in the agricultural sector; (3) Effective Irrigation Rate (EIR), measuring the adequacy of irrigation infrastructure; (4) Technological Progress (lnRSF), proxied by research and development input in logarithmic form to reduce skewness; (5) Farmland Scale Operation (FSO), representing the degree of land consolidation and scale management; (6) Natural Disaster Severity (NDS), measured as the ratio of crop affected area to total crop sown area in each province and year, controlling for exogenous climate and disaster shocks to agricultural production; and (7) Per Capita Cultivated Land Area (PCCLA), accounting for land resource endowment differences across provinces.

4.2. Model Specification

4.2.1. Baseline Regression Model

To examine the direct effect of DRC on ALGUE, this study employs a two-way fixed effects model specified as follows:

$${ALGUE}_{it}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{DRC}}_{\mathrm{it}}+{\mathsf{\alpha }}_n{\mathit{Control}}_{it}+{\mathsf{\mu }}_i+{\mathsf{\lambda }}_t+{\mathsf{\epsilon }}_{it}$$

(3)

To test the moderating effect of regional economic development, we further introduce an interaction term DRC × lnGDP into the baseline model:

$$\begin{array}{l}{ALGUE}_{it}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{DRC}}_{\mathrm{it}}+{\mathsf{\alpha }}_2\ln GD{P}_{\mathrm{it}}+{\mathsf{\alpha }}_3{\left(\mathrm{DRC}\times \ln GDP\right)}_{\mathrm{it}}\\ +{\mathsf{\alpha }}_n{\mathit{Control}}_{it}+{\mathsf{\mu }}_i+{\mathsf{\lambda }}_t+{\mathsf{\epsilon }}_{it}\end{array}$$

(4)

where ${ALGUE}_{it}$ denotes the agricultural land green utilization efficiency of province $i$ in year $t$; ${\mathit{DRC}}_{it}$ denotes the level of digital rural construction in province $i$ in year $t$; ${(DRC\times \ln GDP)}_{it}$ is the interaction term used to examine the moderating role of regional economic development; ${\mathit{Control}}_{it}$ denotes a vector of control variables; ${\mathsf{\mu }}_i$ and ${\mathsf{\lambda }}_t$ capture province and year fixed effects, respectively; and ${\mathsf{\epsilon }}_{\mathit{it}}$ is the idiosyncratic error term.

4.2.2. Mediation Model

To investigate the indirect effect of DRC on ALGUE through mediating variables, this study specifies the following mediation models:

$${\ln GDP}_{\mathit{it}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{DRC}}_{\mathit{it}}+{\mathsf{\alpha }}_n{\mathit{Control}}_{it}+{\mathsf{\mu }}_i+{\mathsf{\lambda }}_t+{\mathsf{\epsilon }}_{it}$$

(5)

$${\ln \mathit{GVA}}_{\mathit{it}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{DRC}}_{\mathit{it}}+{\mathsf{\alpha }}_n{\mathit{Control}}_{it}+{\mathsf{\mu }}_i+{\mathsf{\lambda }}_t+{\mathsf{\epsilon }}_{\mathit{it}}$$

(6)

$$ALGU{E}_{\mathrm{it}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{DRC}}_{\mathit{it}}+{\mathsf{\alpha }}_2{Mediator}_{\mathit{it}}+{\mathsf{\alpha }}_n{\mathit{Control}}_{it}+{\mathsf{\mu }}_i+{\mathsf{\lambda }}_t+{\mathsf{\epsilon }}_{\mathit{it}}$$

(7)

$${\ln GDP}_{\mathit{it}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{DRC}}_{\mathit{it}}+{\mathsf{\alpha }}_2{{\mathrm{DRC}}_{\mathit{it}}}^2+{\mathsf{\alpha }}_n{\mathit{Control}}_{it}+{\mathsf{\mu }}_i+{\mathsf{\lambda }}_t+{\mathsf{\epsilon }}_{\mathit{it}}$$

(8)

where ${\mathrm{lnGDP}}_{\mathrm{it}}$ denotes the natural logarithm of per capita GDP for province $i$ in year $t$; ${\mathrm{lnGVA}}_{\mathrm{it}}$ denotes the natural logarithm of primary industry value added for province $i$ in year $t$; ${DRC}_{it}$ denotes the level of digital rural construction in province $i$ in year $t$; ${Mediator}_{\mathrm{it}}$ denotes lnGDP or lnGVA in turn; ${{\mathrm{DRC}}_{\mathit{it}}}^2$ is the quadratic term of digital rural construction; ${Control}_{\mathit{it}}$ denotes a vector of control variables; ${\mathsf{\mu }}_i$ and ${\mathsf{\lambda }}_t$ capture province and year fixed effects, respectively; and ${\mathsf{\epsilon }}_{\mathit{it}}$ is the idiosyncratic error term.

4.3. Data Sources and Descriptive Statistics

This study uses provincial-level panel data covering 29 provinces in China from 2012 to 2023, excluding Beijing, Xizang, and the Hong Kong, Macao, and Taiwan regions. Beijing is excluded due to its ultra-high urbanization rate and extremely low proportion of agricultural land and primary industry. Xizang is excluded because of severe missing data on key variables during the sample period. Hong Kong, Macao, and Taiwan are excluded due to their inconsistent statistical standards and data sources that cannot be unified with mainland China. Data for measuring DRC and ALGUE are drawn from multiple official sources, including the China Statistical Yearbook, China Household Survey Yearbook, China Science and Technology Statistical Yearbook, China Rural Statistical Yearbook, China Agricultural Machinery Industry Yearbook, and the EPS Data Platform. The Digital Inclusive Finance Index is sourced from the Digital Finance Research Center of Peking University.

In calculating agricultural carbon emissions and resource inputs, the following data processing strategies are adopted: cultivated land data are measured by the actual sown area of crops in the corresponding year; carbon emissions from irrigation electricity are estimated based on the effective irrigated area of each province; agricultural plastic film usage is derived from per capita per unit area consumption at the provincial level; and carbon emissions from agricultural machinery fuel are computed based on the total mechanical power of each province. Missing values are imputed using linear interpolation. Descriptive statistics for all variables are presented in

Table 4. Descriptive Statistics of Variables.

	VarName	Obs	Mean	SD	Min	Median	Max
Dependent Variable	ALGUE	348	0.862	0.450	0.171	1.038	3.846
Core Explanatory Variable	DRC	348	0.263	0.138	0.070	0.223	0.770
Control Variables	AMP	348	0.721	0.356	0.327	0.629	2.698
	FSA	348	0.115	0.034	0.040	0.120	0.200
	PCCLA	348	3.245	2.155	0.920	2.786	15.198
	FSO	348	8.452	4.506	2.516	7.138	32.424
	NDS	348	0.125	0.105	0.000	0.100	0.620
	EIR	348	0.421	0.150	0.172	0.385	0.994
	lnRSF	348	7.646	1.028	3.951	7.681	9.923
Mediating Variables	lnGDP	348	10.948	0.419	9.889	10.922	12.156
	lnGVA	348	7.438	0.996	4.565	7.690	8.781

.

5. Empirical Results and Analysis

5.1. Baseline Regression Results

Baseline regression estimates may suffer from endogeneity biases caused by reverse causality and omitted variable bias. To address this issue, this study adopts the instrumental variable (IV) approach combined with two-stage least squares (2SLS) for correction. Because the computer penetration rate is a constituent indicator of the DRC evaluation system (Table 3) and thus violates the exclusion restriction, we instead use the number of fixed telephones per 100 rural residents in the base year (2012) as the instrumental variable for DRC. This indicator is highly relevant to DRC as it reflects the initial foundation of rural information infrastructure, and is strictly exogenous since it is a historical variable that only affects ALGUE through the channel of DRC without direct or indirect non-digital effects on the explained variable.

The first-stage regression diagnostic result shows that the F-statistic of the IV is 38.62, well above the conventional critical value of 10, ruling out the problem of weak instrumental variables. The second-stage regression results in

Table 5. Baseline Regression Results.

	(1)		(2)
	ALGUE		ALGUE
DRC	0.8124 ***	NDS	0.1100
	(3.8133)		(0.6275)
AMP	−0.4345 *	EIR	0.0426
	(−1.8543)		(0.2190)
FSA	1.5245 **	lnRSF	0.0591 *
	(2.2374)		(1.9207)
PCCLA	−0.0421 ***	_cons	0.3087
	(−4.0527)		(1.1339)
FSO	0.0155 *
	(1.7043)
F	3.581
r2_a	0.893
N	348

Note: t-statistics in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

indicate that after accounting for endogeneity, the estimated coefficient of DRC on ALGUE remains positively significant at the 5% level (coefficient = 2.9035, t = 0.7864). This confirms that the baseline findings are not driven by reverse causality or omitted variable bias, and the positive promoting effect of DRC on ALGUE is robust and reliable. Moreover, the larger coefficient of DRC in the 2SLS regression implies that the baseline model may have underestimated the actual marginal effect of DRC on ALGUE due to endogeneity.

5.2. Heterogeneity Analysis

First, to formally test the moderating effect of regional economic development, we introduce the interaction term DRC × lnGDP into the baseline model (

Table 6. Moderating Effect of Regional Economic Development.

	(1)		(2)
	ALGUE		ALGUE
DRC	8.3505 ***	PCCLA	−0.0187
	(2.9367)		(−1.0945)
lnGDP	0.3372 ***	FSO	0.0058
	(4.0776)		(0.6562)
DRC × lnGDP	−0.7244 ***	NDS	0.0453
	(−2.8843)		(0.4177)
AMP	−0.4261 ***	EIR	−0.1676
	(−8.1044)		(−0.6822)
FSA	−0.2750	lnRSF	−0.0628 ***
	(−0.3583)		(−2.6381)
F	9.91
r2_a	0.2429
N	348

Note: t-statistics in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

). The estimated coefficient of the interaction term is −0.7244, which is statistically significant at the 1% level. This result indicates that regional economic development negatively moderates the impact of DRC on ALGUE. Specifically, the positive promotion effect of DRC on ALGUE gradually weakens with the improvement of regional economic development, showing a typical diminishing marginal effect. This finding supports Hypothesis H2.

To further investigate the potential heterogeneous impacts of digital rural construction (DRC) on agricultural land green utilization efficiency (ALGUE), this study conducts subsample regressions using the median-split method. The choice of the median as the grouping threshold is supported by both theoretical and statistical rationales. First, in line with neoclassical growth theory and club convergence, per capita GDP and agricultural mechanization intensity reflect regional developmental stages and structural differences, and median partitioning effectively classifies observations into relatively homogeneous groups. Second, the median mitigates the distortion caused by extreme values and ensures balanced sample sizes across groups, thereby improving the robustness of estimation results.

Regarding regional economic development, provinces are split by the median of per capita GDP. The results show that the positive effect of DRC on ALGUE is statistically significant and much stronger in the high-development group, whereas it is weaker and less significant in the low-development group. Economically advanced regions possess more complete digital ecosystems, better resource endowments, and mature application scenarios, which enable fuller release of digital dividends.

From the perspective of agricultural structure, we group samples by the median of per-unit sown area mechanical power (total agricultural machinery power divided by total sown area) to capture the modernization and intensification of agricultural production. The estimates reveal that DRC significantly improves ALGUE only in the high mechanization intensity group, while the effect is insignificant in the low mechanization group. Regions with higher mechanization feature more standardized and large-scale production, which provides a favorable foundation for digital technologies to optimize factor allocation, reduce resource waste, and alleviate environmental pressure. In contrast, regions dominated by small-scale and extensive farming lack sufficient capacity to absorb digital empowerment, thus limiting the improvement of ALGUE driven by DRC.

Overall, the impact of DRC on ALGUE exhibits significant regional heterogeneity and depends on regional economic development and agricultural structural conditions (

Table 7. Heterogeneity Analysis.

	(1) Low-Level Group	(2) High-Level Group	(3) High-Mechanization Group	(4) Low-Mechanization Group
	ALGUE	ALGUE	ALGUE	ALGUE
DRC	0.3253 *	1.2317 **	0.9183 ***	0.3041
	(1.6992)	(2.2201)	(3.1087)	(1.5732)
AMP	−0.0013	−0.6941 ***	−0.5896 *	−0.0018
	(−0.0228)	(−8.6130)	(−1.9259)	(−0.0319)
FSA	1.1801	0.8841	1.8830 *	−0.1390
	(1.4175)	(0.6121)	(1.8123)	(−0.1171)
PCCLA	−0.0486 ***	0.1338	−0.0376 *	−0.0197
	(−3.8056)	(1.4793)	(−1.7856)	(−0.2709)
NDS	0.0943	−0.1653	0.1205	0.1206
	(0.8829)	(−0.8368)	(0.5139)	(0.5459)
EIR	0.3147	−0.0756	0.0376	0.5372 *
	(1.4933)	(−0.1535)	(0.1172)	(1.7634)
lnRSF	0.0702 **	0.0170	0.1580 ***	−0.0580
	(2.3773)	(0.3476)	(2.9108)	(−1.6019)
_cons	0.1707	0.4811	−0.1789	0.6538 **
	(0.7014)	(1.0553)	(−0.4110)	0.1206
province	control	control	control	control
year	control	control	control	control
F	73.192	44.287	5.642	2.935
r2_a	0.932	0.892	0.916	0.898
N	174	174	169	178

Note: t-statistics in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

).

5.3. Mediation Analysis

5.3.1. The Mediating Effect of Economic Development

Table 8. Mediation Analysis: Economic Development.

	(1)	(2)	(3)
	ALGUE	lnGDP	ALGUE
DRC	0.8124 ***	2.0445 **	0.5171 **
	(3.8133)	(2.1450)	(2.4215)
DRC2		−1.5897 *
		(−1.8445)
lnGDP			0.3622 ***
			(3.3132)
AMP	−0.4345 *	−0.0935 **	−0.4008 *
	(−1.8543)	(−2.3354)	(−1.6780)
FSA	1.5245 **	−0.5255	1.5139 **
	(2.2374)	(−0.6758)	(2.2749)
PCCLA	−0.0421 ***	0.0071	−0.0228 *
	(−4.0527)	(0.5057)	(−1.8459)
FSO	0.0155 *	−0.0549 **	0.0143
	(1.7043)	(−2.2187)	(1.6041)
NDS	0.1100	0.0521	0.0893
	(0.6275)	(0.8545)	(0.5188)
EIR	0.0426	0.3777 **	−0.0898
	(0.2190)	(2.1807)	(−0.4618)
lnRSF	0.0591 *	0.0293	0.0456
	(1.9207)	(0.7812)	(1.5802)
_cons	0.3087	10.1539 ***	−3.4937 ***
	(1.1339)	(35.7532)	(−2.8211)
F	3.581	209.268	5.114
r2_a	0.893	0.883	0.898
N	348	348	348

Note: t-statistics in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

reports the mediation analysis results, with regional economic development (measured as lnGDP) serving as the mediating variable. Column (1) replicates the baseline regression of DRC on ALGUE. The coefficient of DRC is positive and statistically significant, indicating that digital rural construction directly promotes agricultural land green utilization efficiency and thereby providing the empirical basis for the subsequent mechanism analysis.

Column (2) shows that the estimated coefficient of DRC on economic development (lnGDP) is significantly negative at the 5% level, although the magnitude is relatively small. This seemingly counterintuitive result likely reflects short-term structural adjustment costs during the early phase of rural digital transformation. In particular, the “creative destruction” induced by digital infrastructure investment may temporarily suppress traditional rural economic activities, including smallholder production and informal employment. During this transitional period, productivity gains from digitalization may not yet be fully reflected in aggregate output. Instead, the exit of backward production capacity and transitional labor reallocation can generate a short-run contraction in conventional economic indicators. In addition, rigidities in factor market adjustment may slow the reallocation of released land and labor toward higher value-added green sectors, thereby producing a modest short-term drag on measured economic performance. To verify the nonlinear relationship suggested in Hypothesis H3, we further introduce the quadratic term of DRC (DRC²) into the model. The results show that the primary coefficient of DRC is significantly positive and the quadratic coefficient is significantly negative, presenting a typical inverted U-shaped pattern between DRC and regional economic development. This indicates that digital rural construction promotes economic development in the early stage, but the marginal effect gradually weakens and even declines after reaching a certain threshold.

Column (3) incorporates both DRC and lnGDP into the regression framework. The results indicate that the coefficient of DRC remains positive and statistically significant at the 5% level, although its magnitude declines relative to Column (1). Meanwhile, lnGDP exhibits a positive and highly significant coefficient at the 1% level, suggesting that higher levels of regional economic development provide important support for ALGUE. This pattern is consistent with the Environmental Kuznets Curve (EKC) hypothesis, whereby capital deepening, technological upgrading, and rising environmental demand associated with economic maturity jointly create the material conditions and incentive structure necessary for the diffusion of green agricultural technologies.

Taken together, the evidence suggests that regional economic development functions as an inconsistent mediator linking DRC to ALGUE. After introducing the quadratic term of DRC, we find that the primary coefficient of DRC is significantly positive and the quadratic coefficient is significantly negative, indicating an inverted U-shaped nonlinear effect of DRC on regional economic development. This means that DRC promotes economic development in the early stage, and its promotion effect gradually weakens after exceeding a certain threshold. The above results show that DRC enhances ALGUE partly by promoting regional economic development in an inverted U-shaped manner, which provides partial support for Hypothesis H3.

5.3.2. The Mediating Effect of Primary Industry Upgrading

Table 9. Mediation Effect Test of Agricultural Industrial Upgrading.

	(1)	(2)	(3)
	ALGUE	lnGVA	ALGUE
DRC	0.8124 ***	0.9706 ***	0.4282 *
	(−3.8133)	(−5.6027)	(−1.8637)
lnGVA			0.3958 ***
			(−3.394)
AMP	−0.4345 *	−0.048	−0.4155 *
	(−1.8543)	(−1.5303)	(−1.7830)
FSA	1.5245 **	1.0635 **	1.1036
	(−2.2374)	(−2.3225)	(−1.6257)
PCCLA	−0.0421 ***	0.011	−0.0464 ***
	(−4.0527)	(−0.8682)	(−4.4830)
FSO	0.0155 *	−0.0206 ***	0.0236 ***
	(−1.7043)	(−2.7015)	(−2.6391)
NDS	0.11	−0.1105 *	0.1537
	(−0.6275)	(−1.6560)	(−0.9257)
EIR	0.0426	−0.2338	0.1351
	(−0.219)	(−1.2637)	(−0.6978)
lnRSF	0.0591 *	0.0860 ***	0.025
	(−1.9207)	(−3.9586)	(−0.8846)
_cons	0.3087	6.6870 ***	−2.3379 ***
	(−1.1339)	(−37.0994)	(−2.7577)
F	3.581	8.662	4.951
r2_a	0.893	0.991	0.9
N	348	348	348

Note: t-statistics in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

reports the mediation analysis results, with the value added of the primary industry (log-transformed as lnGVA) serving as the mediating variable. Column (1) replicates the baseline regression of DRC on ALGUE. Column (2) shows that the coefficient of DRC on lnGVA is statistically significant and positive at the 1% level, indicating that digital rural construction significantly promotes primary industry upgrading.

This effect likely operates through the digital restructuring of the agricultural production system. In particular, the adoption of IoT and big data technologies improves real-time production monitoring and factor allocation efficiency, thereby enhancing resource utilization and output performance. Moreover, the expansion of digital services—such as e-commerce and smart logistics—extends the agricultural value chain and increases the market accessibility and value added of agricultural products, promoting a shift toward higher value-added production.

Column (3) incorporates both DRC and lnGVA into the regression framework. The results indicate that the coefficients of both variables remain statistically significant at the 10% and 1% levels, respectively, confirming that primary industry upgrading serves as a partial mediator between DRC and ALGUE. Mechanistically, the digital upgrading of the primary industry is reflected not only in output expansion but also in the transition toward a more intensive, resource-efficient, and environmentally sustainable production mode. Intelligent management reduces excessive fertilizer and pesticide inputs, while precision operations lower energy consumption and mitigate agricultural non-point source pollution, thereby jointly improving the economic and ecological performance of land use.

To further validate the mediating channel and improve the rigor of mechanism analysis, we conduct a bootstrap mediation analysis with 1000 replications and calculate the mediation proportion (indirect effect/total effect). The results are reported in

Table 10. Bootstrap Mediation Effect Test of Agricultural Industrial Upgrading.

	Observed Coef.	Std. Err.	z	p > |z|	95% Conf. Interval
_bs_1(Indirect)	−0.3365	0.1160	−2.90	0.004	[−0.5638, −0.1091]
_bs_2(Total)	0.5667	0.2115	2.68	0.007	[0.1522, 0.9812]
_bs_3(Direct)	0.9032	0.2329	3.88	0.000	[0.4468, 1.3596]

Note: Bootstrap replications = 1000. _bs_1 = indirect effect; _bs_2 = total effect; _bs_3 = direct effect.

. The bootstrap test shows that the indirect effect of DRC on ALGUE through primary industry upgrading is −0.3365. Notably, given that the total effect of DRC on ALGUE is positive, this negative indirect effect indicates an inconsistent (suppressive) mediation pattern, wherein primary industry upgrading partially offsets rather than reinforces the direct positive effect. The 95% confidence interval [−0.5638, −0.1091] does not include zero, confirming that the indirect effect is statistically significant. The mediation proportion accounts for 37.3% of the total effect in absolute terms, confirming a substantial suppressive mediation.

Taken together, primary industry upgrading plays a significant negative partial mediating role between DRC and ALGUE. Although DRC promotes the scale expansion of the primary industry, such expansion may increase factor input and environmental pressure in the short run, resulting in a negative indirect effect. Meanwhile, the direct positive effect of DRC remains dominant, so the overall impact of digital rural construction on agricultural land green utilization efficiency is significantly positive.

5.4. Robustness Checks

To ensure the reliability and robustness of the baseline results, this study conducts three complementary robustness checks that progressively address potential concerns related to sample selection and model specification.

(1)

Sample period adjustment.

To mitigate potential boundary effects associated with specific years, the observations for 2012 and 2023 are excluded, and the baseline model is re-estimated using the truncated panel covering 2013–2022.

(2)

Provincial sample reduction.

Considering the substantial regional heterogeneity in economic development, resource endowments, and policy implementation across Chinese provinces, we adopt stratified random sampling stratified by eastern, central, and western regions (retaining the original proportional structure of the three regions) and set a fixed random seed (123456) to randomly reduce the number of provinces while ensuring replicability. This test examines whether the core findings are sensitive to the provincial sample composition.

(3)

Alternative estimation method.

To address potential specification bias arising from the bounded nature of the dependent variable, a Tobit model is further employed for re-estimation. This approach appropriately accommodates the truncated or censored characteristics of the efficiency measure, thereby providing a more accurate assessment of the marginal effect of DRC on ALGUE.

The results of the three robustness checks are reported in

Table 11. Robustness Checks.

	(1) Adjusted Sample Period	(2) Reduced Provincial Sample	(3) Tobit Model
	ALGUE	ALGUE	ALGUE
DRC	0.6312 ***	1.0051 ***	0.5066 **
	(−3.6952)	(−3.9628)	(−2.5451)
AMP	0.0288	−0.4861 *	−0.3873 ***
	(−0.4448)	(−1.9490)	(−7.3911)
FSA	0.9927 *	1.1261	0.0852
	(−1.741)	(−1.3837)	(−0.1243)
PCCLA	−0.0368 ***	−0.0462 ***	−0.026
	(−4.0553)	(−3.8782)	(−1.5572)
FSO	0.0235 **	0.0217 **	0.0129 *
	(−2.5289)	(−2.2439)	(−1.6648)
NDS	0.1238	0.1319	0.0349
	(−0.6676)	(−0.6057)	(−0.3193)
EIR	0.192	0.1561	−0.2
	(−1.2279)	(−0.6456)	(−0.8770)
lnRSF	0.0434	0.0392	−0.019
	(−1.327)	(−1.0277)	(−0.8888)
_cons	0.035	0.3838	1.1993 ***
	(−0.1194)	(−1.1405)	(−6.3345)
F	3.734	3.992
r2_a	0.929	0.89
N	290	287	348

Note: t-statistics in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

. Across all specifications, the estimated coefficient of DRC remains positive and statistically significant, with no substantive change in either magnitude or significance level. These consistent findings confirm that the baseline results are statistically robust and not sensitive to variations in the sample period, provincial composition, or model specification.

5.5. Endogeneity Test

To address potential endogeneity concerns arising from measurement error, omitted variables, and reverse causality, we employ the two-stage least squares (2SLS) method using the number of fixed telephones per 100 rural residents in 2012 as the instrumental variable for DRC. This instrument satisfies both relevance and exogeneity conditions: it reflects the early-stage foundation of rural digital infrastructure and thus is highly correlated with DRC, while it does not directly affect ALGUE through other channels as a historical variable.

The first-stage regression result shows that the instrumental variable (number of landline telephones per 100 rural residents in 2012) is positively and significantly correlated with digital rural construction, which is consistent with theoretical expectations. Regions with better early communication infrastructure are more likely to achieve higher levels of digital rural construction in later periods. The F-statistic of the IV is 38.62, well above the conventional critical value of 10, ruling out the problem of weak instrumental variables.

The second-stage results are reported in

Table 12. 2SLS Second-Stage Results.

	(1)		(2)
	ALGUE		ALGUE
DRC	2.9035 ***	NDS	−0.3800
	(0.7864)		(0.2325)
AMP	0.6053 ***	EIR	−0.5562 ***
	(0.0718)		(0.1618)
FSA	0.8910	lnRSF	−0.3086 ***
	(0.8928)		(0.0903)
FSO	0.0724 ***	_cons	1.9575 ***
	(0.0127)		(0.5007)
PCCLA	−0.1091 ***
	(0.0254)
F	38.62
r2_a	0.092
N	348

Note: t-statistics in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

. The coefficient of DRC is 2.9035 and significant at the 1% level, which is notably larger than the baseline estimate of 0.8124. This indicates that the baseline model is downward biased due to endogeneity. The downward bias mainly arises from two sources: first, measurement error in the DRC composite index attenuates the OLS coefficient toward zero; second, omitted factors (e.g., short-term agricultural structural adjustment costs) are positively correlated with DRC but negatively restrict ALGUE, leading to further underestimation. After addressing endogeneity, the true and stronger promoting effect of DRC on ALGUE is accurately identified.

The changes in the coefficients and significance of control variables such as FSO and PCCLA between the baseline regression and 2SLS regression are mainly caused by the mitigation of endogeneity bias, which helps obtain more accurate estimation results.

To further verify the exogeneity of the instrumental variable and rule out the interference of unobserved historical factors such as historical agricultural development level, we conduct a placebo test (

Table 13. Placebo Test.

	(1)		(2)
	ALGUE		ALGUE
placebo_DRC	86.723	NDS	−16.263
	(719.610)		(134.324)
AMP	1.898	EIR	1.518
	(10.631)		(15.592)
FSA	−0.634	lnRSF	−0.317
	(29.353)		(2.915)
FSO	−0.142	_cons	−10.164
	(1.835)		(86.450)
PCCLA	0.473
	(5.102)
F
r2_a
N	348

Note: t-statistics in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01

). We randomly generate a placebo treatment variable (placebo_DRC) and regress ALGUE on this false treatment using the same instrumental variable. If the IV is valid, it should have no predictive power for the randomly assigned placebo. The results in Table 10 show that the coefficient of the instrumental variable is not significant (p > 0.1), which confirms that the instrumental variable is not correlated with unobserved factors affecting ALGUE and that the baseline results are not driven by omitted variables. The validity of the instrumental variable is therefore verified.

6. Conclusions and Discussion

6.1. Main Conclusions

Drawing on panel data from 29 Chinese provinces over the period 2012–2023, this study employs two-way fixed effects models and mediation models to examine the effect of DRC on ALGUE and to investigate the underlying transmission mechanisms. The principal findings are as follows.

First, DRC exerts a statistically significant positive effect on ALGUE, which is consistent with the mainstream view that digitalization effectively supports agricultural green development. This result remains robust across multiple sensitivity checks, including sample period truncation, stratified random reduction in provincial samples, and alternative model specification. The mechanism operates primarily through the diffusion of digital technologies, which improve the allocative efficiency of agricultural production factors and facilitate the green transformation of production practices, thereby enhancing land utilization performance in both economic and ecological dimensions.

Second, the effect of DRC on ALGUE exhibits dual-dimensional heterogeneity. The positive effect is substantially stronger in economically advanced regions, which benefit from systematic advantages in digital infrastructure, human capital, and technology application ecosystems that enable fuller absorption of DRC policy dividends. By contrast, less developed regions face constraints including inadequate digital access, limited technological absorptive capacity, and structural talent shortages, which attenuate the enabling effect of DRC.

Third, DRC indirectly enhances ALGUE through two parallel mediating pathways. The primary industry upgrading pathway operates through digital technologies that optimize factor allocation and extend the agricultural value chain. Intelligent production practices further reduce resource consumption and environmental pollution, improving the ecological performance of land utilization. The regional economic development pathway functions with an inverted U-shaped nonlinear characteristic—DRC promotes economic development in the early stage, but its marginal effect weakens after exceeding a certain threshold, with only short-term adjustment costs in the initial stage. Together, these pathways constitute a coherent transmission chain of “digital diffusion → structural upgrading and economic development → efficiency improvement.”

This study contributes by integrating DRC and ALGUE into a unified analytical framework, identifying two parallel mediating channels, and revealing dual-dimensional heterogeneity, thus providing targeted empirical evidence and policy implications for digital-enabled agricultural green development.

6.2. Policy Implications Discussion

6.2.1. Policy Implications

Based on the empirical findings of significant regional heterogeneity and dual structural differences, this study proposes classified, differentiated, and precise policy implications to promote DRC and enhance ALGUE in line with local realities.

First, implement regionally differentiated digital infrastructure investment strategies. For economically developed regions, priority should be given to deploying cutting-edge digital infrastructure such as 5G networks, IoT systems, big data centers, and satellite remote sensing platforms. These regions should also support the R&D and adoption of intelligent agricultural equipment, building a full-chain digital agriculture system spanning precision planting, intelligent processing, and digital marketing. For less developed regions, efforts should focus on consolidating basic digital facilities, including rural broadband, mobile communications, and digital service stations. Targeted infrastructure subsidies should be provided to lower the threshold for digital technology access and gradually narrow the regional digital divide.

Second, formulate classified digital empowerment paths matched to local agricultural structure. In high-mechanization regions characterized by large-scale and intensive production, digital technologies should be deeply integrated into farmland management, fertilizer and pesticide reduction, and clean production processes. Intelligent monitoring of agricultural carbon emissions should be strengthened to improve green land-use efficiency through precise factor allocation. In low-mechanization regions dominated by smallholder farmers, the focus should be on promoting low-cost, easy-to-operate digital tools such as mobile agricultural applications. Green production technologies, including soil testing and formulated fertilization, should be extended to guide the transformation from extensive land use toward green and intensive practices.

Third, establish differentiated digital talent cultivation and recruitment mechanisms. Developed regions can leverage universities and research institutions to train high-end digital agriculture professionals, with an emphasis on interdisciplinary talents skilled in big data analysis and intelligent agricultural decision-making. Less developed regions should focus on grassroots digital skills training for farmers, popularizing basic digital operations and green production knowledge. These regions should also support the return of skilled professionals and rural revitalization volunteers to participate in digital rural construction.

Fourth, build cross-regional digital coordination and green benefit-sharing mechanisms. Economically developed regions and highly mechanized major agricultural provinces should be encouraged to provide digital technology assistance and share practical experience with less developed regions, facilitating the cross-regional flow of digital technologies, management models, and green solutions. In addition, trading mechanisms for agricultural carbon emissions and ecological benefits should be improved, converting green land-use efficiency gains across regions into tradable economic value. This would create a long-term incentive structure for agricultural green development.

6.2.2. Limitations and Future Research

This study has several limitations that warrant acknowledgement and suggest directions for future research. First, due to data constraints, both ALGUE and DRC are measured using proxy indicators, and certain environmental pressure variables rely on parametric assumptions. As more granular micro-level and remote sensing data become available, future studies can refine these measurements to improve precision. Second, the reliance on provincial-level panel data may mask within-province heterogeneity and regional spatial interactions. Digital development and agricultural land use patterns often vary markedly across counties and townships. Future research could employ sub-provincial data and spatial econometric models to identify diffusion pathways and spillover effects at finer geographic scales. Third, the mediation analysis in this study focuses on primary industry upgrading and regional economic development, potentially overlooking other relevant transmission channels. Future work could extend the analytical framework to incorporate mechanisms such as land transfer and scale operation, green technology innovation, and digital finance, and further examine potential nonlinear or threshold effects in the relationship between DRC and ALGUE.