Research on the Impact of Rural Digital Economy on Agricultural Total Factor Productivity: A Dual Perspective of Human Capital and Scale Operations

Abstract

The advancement of the rural digital economy (RDE) has created new opportunities to alleviate resource and environmental constraints in agriculture and to accelerate the green transformation of agricultural production. However, the existing literature provides limited analysis of the underlying mechanisms and insufficient identification of the multidimensional pathways through which the RDE promotes agricultural green total factor productivity (AGTFP). Using provincial panel data from 2013 to 2022, this study measures the level of the RDE with the entropy method and calculates AGTFP using the SBM-GML model. It further employs a two-way fixed effects model, a mediation model, and a threshold model to examine the specific mechanisms through which the RDE affects AGTFP. The results show that a 1% increase in the RDE is associated with a 3.7% increase in AGTFP, with significant positive effects on its decomposed indices. These findings remain robust after a series of robustness and endogeneity tests. Further analysis indicates that the RDE enhances AGTFP through improvements in rural human capital and agricultural-scale operations, with a significant threshold effect demonstrating increasing marginal returns beyond the threshold. The mediating effects are more pronounced in major grain-producing regions. Policy implications emphasize integrating the RDE with green agricultural production, promoting digital talent development and moderate-scale operations, and reducing regional disparities. These findings provide empirical evidence for fully leveraging the dividends of the RDE and advancing agricultural modernization.

1. Introduction

China cultivates 7% of the world’s arable land to feed 22% of the global population, with total agricultural output increasing from 139.7 billion yuan in 1978 to 8975.5 billion yuan in 2023, a growth of approximately 64 times. This expansion has largely relied on a traditional production model focused on increasing factor inputs. However, the input structure under this model is often inefficient, particularly due to the overuse of chemical inputs such as fertilizers and pesticides, which has contributed to the gradual depletion of high-quality arable land, declining groundwater levels, and rising agricultural non-point source pollution and carbon emissions. In response, the government has placed considerable emphasis on addressing agricultural resource and environmental challenges. In recent years, the Central Government’s No. 1 Policy Document has repeatedly highlighted the importance of reducing fertilizer and pesticide use while improving efficiency, thereby promoting green development in agriculture and rural areas. Similarly, the report of the 20th National Congress emphasized concepts such as “clear waters and green mountains,” green development, and low-carbon practices. Against this backdrop, a central challenge for China is how to reconcile agricultural output growth with the sustainable use of resources and environmental protection under these constraints [1]. Optimizing AGTFP has emerged as a key strategy for achieving this objective [2].

The digital economy, driven by next-generation information and communication technologies and leveraging diverse digital resources, has emerged as a key engine of high-quality economic growth [3]. As a major agricultural country, China emphasizes the digital economy’s role in enhancing agricultural production [4,5]. The White Paper on Digital Rural Development (2024) highlights that the digitalization of rural industries, with a focus on smart agriculture and the digital economy, is fundamental to building digital villages, with green agricultural production as an essential requirement. This study thus examines how to fully harness the development benefits of the RDE to empower agricultural production, reduce reliance on traditional production models, and accelerate the green transformation of agriculture.

The RDE is a key expression of the technological revolution in agriculture and is gradually reshaping patterns of value creation. Human capital serves as the primary agent for applying digital technologies, and its level directly influences the development of the RDE [6]. The widespread adoption of new agricultural models, including digitalized farming, integrated cultivation and harvesting, and smart agriculture, places higher demands on rural human capital. Evidence indicates that regions with higher rural human capital are more likely to adopt green production technologies, which positively correlates with green technological efficiency [7].

By enabling the rapid exchange and dissemination of information, the RDE provides a critical foundation for developing and improving the rural land transfer market, thereby promoting the expansion of agricultural-scale operations [8]. Compared with fragmented smallholder production, scale operations allow for the more efficient allocation of inputs, facilitating the precise use of fertilizers and pesticides and reducing the negative environmental impacts of input redundancy [9]. Furthermore, larger-scale operations lower the marginal cost of adopting new digital technologies, thereby supporting improvements in AGTFP [10]. Therefore, the roles of rural human capital and agricultural-scale operations in facilitating the green transformation of agricultural production under the empowerment of the RDE represent another key focus of this study.

In light of the above, as the RDE matures, it is crucial to examine how its benefits can be harnessed to enhance AGTFP and advance sustainable agricultural development. This study seeks to clarify the specific effects of the RDE on AGTFP and to identify the mechanisms through which rural human capital and agricultural-scale operations contribute. The results aim to offer new insights for mitigating environmental constraints, promoting the green transformation of agricultural production, and supporting the development of a modern and resilient agricultural sector.

2. Literature Review

Against the backdrop of shrinking resource stocks and mounting environmental pressures, transforming the economic development model to balance output with environmental protection and achieve sustainable development has become a central concern in the literature. Numerous studies have highlighted that the digital economy, as an emerging economic form, can fully exploit the comparative advantages of digital resources to drive technological progress, control environmental pollution, and enhance green total factor productivity, thereby accelerating the transition to a greener economy [11,12]. As a major agricultural nation, China has naturally become a focal point for research on the RDE and green agricultural production. The literature most closely related to this study can be grouped into three main strands.

The first strand of the literature extensively investigates the concept, indicator systems, and measurement methods of the RDE. Conceptually, there is broad consensus that applying next-generation information and communication technologies across the entire agricultural value chain, with digital information as a core production factor, enhances efficiency at all production stages and fosters new agricultural business models and development pathways [4,13,14,15]. Regarding indicator systems, the European Union’s Digital Economy and Society Index (DESI) serves as a reference framework for constructing RDE indicators [16,17]. In China, scholars have developed composite indicator systems based on digital infrastructure, socio-economic environment, digital services, and industrial digitalization to evaluate the level of RDE development [18,19]. With respect to measurement, most studies employ principal component analysis or the entropy method. Because principal component analysis requires subjective weighting during the design process, it may introduce bias. As a result, many researchers prefer the entropy method, which offers a more objective assessment of the RDE [20,21].

The second strand of the literature examines the concept and measurement of agricultural green total factor productivity (AGTFP). Traditional production factors often have limited marginal contributions, and agriculture faces substantial resource and environmental constraints [1]. Conventional total factor productivity ignores environmental impacts, limiting its ability to assess performance objectively [1,2]. By accounting for agriculture’s effects on resources and the environment, AGTFP provides a more comprehensive and accurate measure of productivity [2]. Measurement methods for AGTFP generally follow those for total factor productivity, including parametric approaches based on stochastic frontier production functions and non-parametric approaches using data envelopment analysis [22,23]. Among these, the non-parametric method combining the slack-based measure (SBM) model with the global Malmquist–Luenberger (GML) model is widely applied, as it does not require the specification of a particular production function [24,25,26].

The third strand of the literature examines the impact of the RDE on AGTFP. Evidence from multiple perspectives consistently indicates that the rapid development of the RDE strengthens AGTFP. At the macro-level, Hong et al. (2022) employ an OLS model with instrumental variables to address endogeneity and find that the expansion of the RDE enhances rural digital financial inclusion, providing crucial support for AGTFP optimization [27,28]. At the meso-level, the study using the generalized 2-SLS method and system GMM model demonstrate that internet penetration and digital technologies promote green and sustainable agricultural development [23]. At the micro-level, Li et al. (2024) apply an endogenous switching model and show that digital agriculture encourages farmers to adopt organic fertilizers by improving ecological awareness and widening access to information [29].

The existing literature offers valuable guidance on indicator selection and measurement, model specification, and the treatment of endogeneity. As the RDE reshapes agricultural production, key input factors such as labor and land may be affected. Evidence suggests that mismatches in rural human capital and land allocation can significantly reduce agricultural productivity and output [30]. Studies further indicate that higher levels of rural human capital facilitate the adoption of green production practices and technologies, thereby enhancing AGTFP [31]. Land transfers are critical for developing agricultural-scale operations, which alleviate resource misallocation, improve green technological efficiency, and reduce agricultural carbon emissions [32]. Yet, few studies simultaneously examine rural human capital and agricultural-scale operations within the framework of the RDE’s impact on AGTFP or explicitly identify their underlying mechanisms. To fill this gap, this study uses provincial panel data to estimate two-way fixed effects and threshold models to assess the specific impact of the RDE on AGTFP, and applies a mediation analysis to clarify the mechanisms through which rural human capital and scale operations influence outcomes.

This study offers several contributions. First, it examines the effect of the RDE on AGTFP, highlighting the mediating roles of rural human capital and agricultural-scale operations, and validates these mechanisms using a two-step mediation model, thereby strengthening the robustness of the empirical results. Second, it applies a threshold model to explore the non-linear relationship between RDE and AGTFP, providing insights for harnessing the long-term potential of the RDE to facilitate the green transformation of agriculture.

3. Theoretical Analysis and Research Hypotheses

3.1. The Direct Impact of the RDE on AGTFP

The traditional production model relies on increasing factor inputs, particularly the excessive use of fertilizers and the inefficient application of pesticides, generating significant negative externalities that reduce environmental carrying capacity. According to the Environmental Kuznets Curve theory, in the early stages of economic development, gains in production efficiency often entail substantial environmental costs; only with technological progress and the transformation of production models can economic growth and environmental protection be achieved simultaneously. AGTFP incorporates the environmental impacts of agricultural production into the assessment of efficiency, capturing both economic and ecological outputs. Amid mounting constraints on agricultural resources and the environment, enhancing AGTFP and overcoming the path dependence of traditional production models is critical for sustainable agricultural development. The RDE, driven by new digital technologies and digital resources, reshapes agricultural production methods and optimizes factor allocation, providing a concrete foundation for improving AGTFP. This study estimates AGTFP using the SBM-GML model and decomposes it into green technological change (GTC) and green efficiency change (GEC). The following provides a detailed analysis.

The RDE for GTC: Digital technologies inject new momentum into agricultural production, enabling iterative upgrades of equipment and infrastructure and increasing the marginal contribution of technological progress to AGTFP [33]. Technologies such as smart agricultural machinery, precision fertilization, plant protection, and drone-based pesticide application illustrate the widespread integration of digital tools across production stages. In particular, the precise application of chemical inputs reduces non-point source pollution and mitigates the negative externalities of agricultural production, thereby lowering undesired outputs [29]. According to induced innovation theory, under resource constraints, technological advancements are driven by external demand, as farmers seeking higher output increase their demand for improved and innovative production techniques, thereby accelerating technological progress [34].

The RDE for GEC: Information generated through the application of digital technologies in production is collected, filtered, and integrated by universities, research institutes, and other institutions, allowing R&D departments to refine technologies based on actual production conditions, thereby enhancing research efficiency and avoiding mismatches between new technologies and production practices. Simultaneously, the rapid development of the RDE accelerates information flows and reduces information asymmetry, lowering the costs of technology dissemination. This expands the geographic and demographic reach of new technologies, strengthens their diffusion, and improves technology transfer efficiency [35]. In particular, the widespread adoption of smart communication devices in rural areas increases smallholders’ access to new technologies, reduces learning costs, and positively influences traditional production practices, ultimately optimizing factor allocation [36].

Therefore, the following hypothesis is developed:

H1: 

The RDE enhances AGTFP.

H1a: 

The RDE enhances GTC.

H1b: 

The RDE enhances GEC.

3.2. The Mediating Effect of Rural Human Capital

The Chicago School views human capital as the primary driver of economic growth, with improvements in population literacy enhancing individuals’ ability to manage uncertainty [37]. As China advances toward high-quality development, policymakers have increasingly recognized the critical role of human capital. The related policy documents noted that the shortage of professionals skilled in both agriculture and information technology presents a major challenge for the informatization of rural areas. In this study, the RDE fosters new business models, production methods, and practices, gradually enhancing the intelligence and scientific rigor of agricultural processes, increasing the comparative returns of agricultural activities, and attracting talent from across society, thereby strengthening overall rural human capital. Human capital not only serves as a key input in agricultural production but also influences AGTFP by shaping the effectiveness of digital technologies applied within production processes [38].

High-quality human capital allows farmers to fully leverage digital dividends by efficiently capturing and integrating information on agricultural production and market prices, anticipating risks, and minimizing losses [28]. Moreover, skilled labor can quickly learn and adopt to apply digital tools in agricultural practices [34], optimizing the technological benefit while amplifying the positive effects of green innovations on AGTFP.

Inadequate human capital limits farmers’ capacity to process the diverse information provided by the RDE, which can impair production decisions, exacerbate resource misallocation and risk exposure, reduce the effectiveness of digital agricultural technologies [39], and hinder both the diffusion of new technologies and improvements in GEC.

Therefore, the following hypothesis is developed:

H2: 

The RDE optimizes AGTFP by increasing rural human capital levels.

3.3. The Mediating Effect of Agricultural Scale Operations

Next-generation information and communication technologies are a key driver of the RDE, enabling household farms, large-scale specialized farmers, and other emerging operators to access land transfer market information and providing technological support for land consolidation, thereby facilitating the expansion of agricultural-scale operations. As rational economic agents, these operators are motivated to adopt digital and green technologies to maximize returns [40], which in turn expands the coverage and impact of these innovations. The broad application of digital technologies in scale operations enhances both resource and environmental efficiency while supporting output growth [8], contributing to the optimization of AGTFP. The mediating role of scale operations is further examined through production information services, scale effects, and the integration of farmers within agricultural networks.

First, the RDE promotes the adoption of smart and digital agriculture, providing comprehensive production services for scale operations; technologies such as early warning systems for disasters and pests, along with crop growth monitoring, not only enhance production efficiency for large-scale operators but also strengthen their capacity to manage external risks, thereby improving AGTFP. Second, the widespread integration of digital technologies across agricultural production stages generates scale effects from new input factors, which partially substitute for traditional inputs and enhance production efficiency [36]. Finally, the rapid digitalization and informatization of agriculture have increasingly reinforced linkages between operators and smallholders. Cooperatives, leading enterprises, and other operators use communication technologies, such as the internet and big data platforms, to disseminate green technologies at lower costs, thereby broadening their reach, fostering the adoption of sustainable practices, improving traditional production methods, and accelerating the green transformation of agriculture [8,41].

Therefore, the following hypothesis is developed:

H3: 

The RDE optimizes AGTFP by developing agricultural-scale operations.

3.4. Non-Linear Relationship Between the RDE and AGTFP

Based on Hypothesis 1, the rapid development of the RDE introduces advanced digital inputs into agricultural production, improving traditional practices and enhancing AGTFP. However, the effect of the RDE on AGTFP operates through a coupled and coordinated process and is constrained by objective factors such as geography, resource endowments, and the level of economic development, thereby generating a nonlinear impact on AGTFP [23]. In the early stages, policies, funding, technologies, and infrastructure remain exploratory, outcomes vary across regions, and the application of digital inputs often suffers inefficiencies, slowing improvements to traditional practices and limiting AGTFP [42]. As RDE matures, digital inputs integrate organically with agricultural production, amplifying marginal effects on green technology innovation, technology diffusion, and output growth, and increasingly driving AGTFP. Consequently, the RDE exhibits a threshold effect, whereby its enabling impact on AGTFP strengthens and displays increasing marginal returns once the threshold is surpassed [43].

Therefore, the following hypothesis is developed:

H4: 

There is a threshold effect on the effectiveness of the RDE in raising AGTFP.

4. Methodology

4.1. Variables

4.1.1. Independent Variable

The independent variable in this study is the RDE. Drawing on the existing literature, four primary dimensions are selected, namely infrastructure, funding, service, and consumption [3,19,44,45], with corresponding secondary indicators specified under each dimension. The development level of the RDE is measured using the entropy weighting method. The detailed indicator system is presented in Table S1 of Supplementary Materials A.

4.1.2. Dependent Variables

This study applies the SBM-GML model to estimate AGTFP [25,26]. This study constructs an SBM-GML model based on the assumptions of input orientation, non-radial measurement, and increasing returns to scale. The detailed specification of the SBM-GML model is provided on Equations (S8) and (S9) of Supplementary Materials A. Building on the conventional TFP framework, GTFP integrates the negative environmental impacts of production as undesirable outputs within the measurement system. The input variables in this study include labor, land, and other factors; total output value serves as the desirable output, while carbon emissions represent undesirable output [2,12,27,43]. The detailed input and output indicators are presented in Tables S2 and S3 of Supplementary Materials A. The specific calculation procedure for undesired outputs is provided on Equation (S7) of Supplementary Materials A.

4.1.3. Control Variables

Drawing on the existing literature, this study selects control variables across five dimensions: resource endowments, agricultural fiscal support, the urban–rural income gap, urbanization, and rural consumption [7,12,43,46]. Resource endowments, proxied by the disaster rate (disa), capture the natural conditions for agricultural production, as frequent disasters can damage output and limit AGTFP growth. Agricultural fiscal support, measured by the share of fiscal expenditure allocated to agriculture, forestry, and water affairs (fin-a) and to environmental protection (fin-e), provides critical funding for green agricultural production and rural digital economy development. The urban–rural income gap (urig), defined as the ratio of urban to rural disposable income, influences the allocation and mobility of resources. Urbanization (urba), measured by the share of urban population in total population, reflects residents’ likelihood of adopting advanced technologies and open-minded practices, facilitating the diffusion of digital innovations. Rural Engel’s coefficient (reng), the share of food expenditure in total household expenditure, captures rural living standards and consumption behavior.

4.1.4. Mediating Variables

In the theoretical framework, rural human capital and agricultural-scale operations are posited to mediate the effect of the RDE on AGTFP. Accordingly, this study treats rural human capital (educ) and agricultural-scale operations (scal) as mediating variables. Rural human capital is measured by per capita years of education, calculated by multiplying the number of workers at each educational level (illiterate or semi-literate, primary, junior high, senior high, and tertiary or above) by the corresponding years of schooling (1, 6, 9, 12, and 15.5 years) [47]. Agricultural scale operations are proxied by the land transfer rate, defined as the proportion of transferred land to total land area, since land transfers are a fundamental prerequisite for developing large-scale agricultural operations [9]. The names, definitions, and units of all variables are provided in Table S4 of Supplementary Material A.

4.2. Empirical Models

4.2.1. Two-Way Fixed Effects Model

This study constructs Equation (1) to examine the impact of the RDE on AGTFP.

$$AGTF{P}_{i,t}, GE{C}_{i,t}, GT{C}_{i,t}={\beta }_0+{\beta }_{1}RD{E}_{i,t}+{\beta }_2{X}_{i,t}+{\beta }_{3}co{n}_{i,t}+{\mu }_i+{\eta }_t+{\epsilon }_{i,t}$$

(1)

In Equation (1), $i$ and $t$ denote provinces and time, respectively. $AGTF{P}_{i,t}$, $GE{C}_{i,t}$ and $GT{C}_{i,t}$ are the dependent variables. $RD{E}_{i,t}$ is the independent variable. $X_{i,t}$ represents the set of control variables. $co{n}_{i,t}$ is the constant term. ${\mu }_i$ and ${\eta }_t$ denote individual and time fixed effects, respectively. ${\epsilon }_{i,t}$ is the random error term. The coefficient ${\beta }_1$ is the main parameter of interest in this study.

4.2.2. Mediation Effect Model

The theoretical analysis explores the mediating effects of rural human capital and agricultural-scale management in the process through which RDE enhances AGTFP. To further test H2 and H3, a two-step mediation model is employed to identify the transmission pathways of these mediators [12]. To improve estimation accuracy, the model incorporates province and time fixed effects, reducing the influence of time-invariant regional characteristics and external shocks. The control variables are also included to address other potential confounding factors. The identification assumptions of the mediation effect model are detailed in Supplementary Materials B. The results of the regression analysis on dynamic mediating effect are presented in Table S5 of Supplementary Material B.

$$me{d}_{i,t}={\beta }_0+{\gamma }_{1}RD{E}_{i,t}+{\beta }_2{X}_{i,t}+{\beta }_{3}co{n}_{i,t}+{\mu }_i+{\eta }_t+{\epsilon }_{i,t}$$

(2)

$$AGTF{P}_{i,t}={\beta }_0+{\gamma }_{2}me{d}_{i,t}+{\beta }_2{X}_{i,t}+{\beta }_{3}co{n}_{i,t}+{\mu }_i+{\eta }_t+{\epsilon }_{i,t}$$

(3)

This study focuses on the estimated coefficients ${\gamma }_1$ and ${\gamma }_2$. In Equations (2) and (3), $me{d}_{i,t}$ denotes the mediating variables. All other variables are defined as in Equation (1). Equation (2) examines the effect of the RDE on the mediating variables. Equation (3) assesses the impact of the mediating variables on AGTFP.

4.2.3. Threshold Effect Model

To test Hypothesis 4, this study applies a threshold effect model [48], with the RDE as the threshold variable. By determining the threshold value, the model further investigates whether the effect of the RDE on AGTFP is nonlinear.

$$\begin{array}{l}AGTF{P}_{i,t}={\alpha }_0+{\alpha }_{1}RD{E}_{i,t}\times I(RD{E}_{i,t}\le {\theta }_1)+{\alpha }_{2}RD{E}_{i,t}\times I({\theta }_1<RD{E}_{i,t}){\beta }_2{X}_{i,t}+\\ {\beta }_{3}co{n}_{i,t}+{\mu }_i+{\eta }_t+{\epsilon }_{i,t}\end{array}$$

(4)

In Equation (4), ${\alpha }_1$ and ${\alpha }_2$ represent the coefficients of the RDE on AGTFP for different ranges of the threshold variable, while $\theta$ and $I$ denote the threshold value and the indicator function, respectively. All other variables are defined as in Equation (1). This study focuses primarily on coefficients ${\alpha }_1$ and ${\alpha }_2$.

4.3. Data Sources

The study sample includes 26 provinces and 4 municipalities over the period 2013–2022. Tibet, as well as the Hong Kong, Macao, and Taiwan regions, are excluded due to data incompleteness. Missing values for a small portion of the data were imputed using linear interpolation, resulting in a 10-year provincial panel dataset. All raw data for AGTFP, the secondary indicators of the RDE, the mediating variables, and the control variables were obtained from the China Statistical Yearbook, China Rural Statistical Yearbook, China Science and Technology Statistical Yearbook, China Financial Yearbook, China Financial Analysis Report, China Rural Management Statistical Yearbook, and the National Bureau of Statistics. These data do not require special access and can be directly downloaded from the official website of the National Bureau of Statistics. To ensure accuracy and consistency, the data were cross-checked with the CEIC database, which requires an institutional or individual subscription. For the heterogeneity analysis, the list of grain-producing regions is based on the classification of major production, major consumption, and balanced provinces published by the Ministry of Agriculture and Rural Affairs. To reduce the impact of extreme values on the estimation results, all control variables were log-transformed. The descriptive statistics for all variables are reported in

Table 1. Descriptive statistics of variables.

Types	Variables	N	Mean	Sd
Dependent variables	AGTFP	300	1.338	0.558
	GEC	300	1.015	0.233
	GTC	300	1.320	0.454
Independent variable	RDE	300	0.509	0.111
Control variables	disa	300	0.115	0.035
	fin-a	300	0.029	0.010
	fin-e	300	0.13	0.113
	urig	300	2.511	0.362
	urba	300	0.614	0.114
	reng	300	0.321	0.057
Mediating variables	educ	300	7.927	0.621
	scal	300	35.000	16.550

.

5. Empirical Results and Analysis

5.1. Baseline Regression Analysis

Columns (1) to (3) of

Table 2. Baseline regression results.

Variable	AGTFP	GEC	GTC	AGTFP	GEC	GTC
	(1)	(2)	(3)	(4)	(5)	(6)
RDE	0.182 *** (0.025)	0.046 ** (0.022)	0.017 *** (0.002)	0.037 *** (0.004)	0.036 ** (0.018)	0.120 ** (0.054)
Control variables	No	No	No	Yes	Yes	Yes
Constant	0.041 *** (0.012)	0.971 *** (0.059)	0.046 *** (0.009)	0.077 *** (0.019)	0.065 *** (0.018)	0.230 *** (0.035)
Time-fixed	No	No	No	Yes	Yes	Yes
Province-fixed	No	No	No	Yes	Yes	Yes
N	300	300	300	300	300	300
R2	0.338	0.701	0.459	0.409	0.714	0.518

Note: *, **, *** denote significance level at 10%, 5% and 1% respectively, with robust standard errors in parentheses, as shown below. The variance inflation factors (VIF) for all explanatory variables remain below the conventional threshold of 10, with a mean VIF of 1.720, indicating the absence of severe multicollinearity concerns in our empirical specification.

show the regression results without control variables and time and individual effects, while columns (4) to (6) show the regression results with control variables and with time and individual effects. The baseline results in columns (4)–(6) demonstrate statistically significant positive effects of RDE on AGTFP and its components. The baseline regression results show that RDE increases AGTFP by 3.7% at the 1% level, raises GEC by 3.6% at the 5% level, and enhances GTC by 12%, providing empirical support for H1, H1a, and H1b.

From a practical perspective, the rapid development of the RDE improves AGTFP through two primary channels. On the one hand, digital technologies enable more precise management of production factor inputs, optimize the input structure, reduce redundancy, and enhance green technological progress. On the other hand, they streamline farmers’ learning and operational processes, expand channels for the dissemination and adoption of new technologies, lower the costs of technology promotion, and increase the efficiency of technology diffusion, thereby improving AGTFP.

5.2. Robustness Tests

To assess the robustness of the baseline regression results, this study conducts a series of robustness tests. The results show that the RDE continues to have a statistically significant positive effect on AGTFP and its decomposed indices. Detailed results are reported in Tables S6–S8 of Supplementary Materials C.

5.3. Endogenous Discussion

The two-way fixed effects model mitigates endogeneity arising from time-invariant unobserved heterogeneity, yet potential biases may persist due to reverse causality and omitted variables. This study employs instrumental variables a two-stage least squares (IV-2SLS) approach to address potential endogeneity concerns and strengthen causal inference.

Since the data in this study consist of balanced panel data, we have chosen the cross-multiplier terms of the historical characteristic variables and the time series variables as IV. This approach aligns with the methods described in the relevant literature [11,19]. This study constructs IV from the interaction of 1984 postal service density (post offices per million population) and one-period lagged rural landline penetration to address potential endogeneity concerns. The growth of digital technology depends significantly on landline telephony to ensure universal coverage. Regions with high landline penetration often also experience better network access. Therefore, using landline telephony as a foundation for developing digital technology meets the requirement for the relevance of an instrumental variable. As communication technology continues to evolve, landline telephones struggle to meet the increasing demands of various agricultural production segments. Consequently, there is virtually no direct impact on the AGTFP, which indicates that the exogeneity requirement is satisfied.

Although the preceding analysis addresses the relevance and exogeneity of the instrumental variable, historical variables may still influence AGTFP through alternative channels and thus may not fully satisfy the exclusion restriction. Accordingly, to further address endogeneity, this study evaluates the validity of the IV by examining the exclusion restriction through interactions between time trends and control variables and by conducting placebo tests. The results confirm that the IV satisfies the exclusion restriction. The detailed findings are reported in Table S9 of Supplementary Materials D.

Table 3. Results of IV-2SLS.

Variable	RDE	AGTFP	RDE	GEC	RDE	GTC
	(7)	(8)	(9)	(10)	(11)	(12)
RDE		0.038 *** (0.005)		0.107 ** (0.047)		0.015 ** (0.007)
IV	0.013 *** (0.003)		0.051 *** (0.019)		0.048 *** (0.0003)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes
Constant	0.016 *** (0.005)	0.021 (0.033)	0.015 ** (0.007)	0.012 (0.021)	0.071 *** (0.025)	0.013 (0.021)
Time-fixed	Yes	Yes	Yes	Yes	Yes	Yes
Province-fixed	Yes	Yes	Yes	Yes	Yes	Yes
Kleibergen–Paap rk LM	46.734
Cragg–Donald Wald F statistic	53.882
N	300	300	300	300	300	300
R2	0.676	0.633	0.698	0.772	0.705	0.824

Notes*, **, *** denote significance at the 10%, 5%, and 1% levels, respectively, with robust standard errors reported in parentheses.

reports the IV-2SLS regression results. The IV pass both the weak instrument and under identification tests. The RDE exerts a significant positive effect on AGTFP and its decomposition indices.

The empirical findings demonstrate that the baseline regression results remain statistically and economically significant across various robustness checks and endogeneity tests; thereby H1, H1a, and H1b are proven.

5.4. Analysis of Mediating, Effect

5.4.1. Mediating Effect of the Rural Human Capital

Columns (13) and (14) of Table 4 show that the RDE has a significant positive effect on rural human capital at the 1% level, and that rural human capital, in turn, has a significant positive effect on AGTFP at the 1% level. Therefore, H2 is supported. The analysis indicates that increasing human capital in rural areas primarily enhances traditional production techniques and strengthens farmers’ capacity to adopt innovations and optimize production decisions. Rather than relying solely on increased factor inputs to raise yields, agricultural producers are adopting modern technologies and equipment to improve AGTFP.

5.4.2. Mediating Effect of the Agricultural Scale Operations

Columns (15) and (16) of Table 4 show that the RDE has a significant positive effect on agricultural-scale operations at the 1% level, and that agricultural-scale operations, in turn, have a significant positive effect on AGTFP at the 5% level. Therefore, H3 is supported. In practice, land transfer platforms enhance operating entities’ access to information, improving supply–demand matching in land transactions. This increases the land transfer rate and promotes the efficient use of idle rural land, thereby facilitating large-scale agricultural operations [49]. By exploiting economies of scale, these new operating entities can encourage smallholder farmers to adopt innovative technologies, increasing adoption rates and enhancing smallholders’ AGTFP.

Table 4. The regression results of mediating effect.

Variable	educ	AGTFP	scal	AGTFP
	(13)	(14)	(15)	(16)
RDE	0.114 *** (0.032)		0.278 *** (0.093)
educ		0.067 *** (0.014)
scal				0.032 ** (0.015)
Control variables	Yes	Yes	Yes	Yes
Constant	0.862 *** (0.077)	0.319 *** (0.017)	0.061 *** (0.023)	0.015 (0.025)
Time-fixed	Yes	Yes	Yes	Yes
Province-fixed	Yes	Yes	Yes	Yes
N	300	300	300	300
R2	0.757	0.835	0.743	0.817

Notes:*, **, *** denote significance at the 10%, 5%, and 1% levels, respectively, with robust standard errors reported in parentheses.

Table 4. The regression results of mediating effect.

Variable	educ	AGTFP	scal	AGTFP
	(13)	(14)	(15)	(16)
RDE	0.114 *** (0.032)		0.278 *** (0.093)
educ		0.067 *** (0.014)
scal				0.032 ** (0.015)
Control variables	Yes	Yes	Yes	Yes
Constant	0.862 *** (0.077)	0.319 *** (0.017)	0.061 *** (0.023)	0.015 (0.025)
Time-fixed	Yes	Yes	Yes	Yes
Province-fixed	Yes	Yes	Yes	Yes
N	300	300	300	300
R2	0.757	0.835	0.743	0.817

Notes:*, **, *** denote significance at the 10%, 5%, and 1% levels, respectively, with robust standard errors reported in parentheses.

6. Extensibility Analysis

6.1. Threshold Effect Test

This study uses the bootstrap method with 500 replications to determine the number and value of the threshold variable.

Table 5. Results of the threshold effect test.

Number of Thresholds	F-Value	p-Value	Bootstrap Times	1%	5%	10%
(17)	(18)	(19)	(20)	(21)	(22)	(23)
single	26.96	0.002	500	22.306	16.953	12.777
double	4.67	0.726	500	22.540	14.912	12.185
triple	6.29	0.758	500	23.863	28.764	38.291

presents the bootstrap results, with the p-value for a single threshold below 0.01, indicating the presence of a single threshold.

Table 6. Results of threshold model estimation.

Variable	AGTFP
	(24)
RDE ≤ 0.534	0.177 *** (0.054)
RDE > 0.534	0.258 *** (0.044)
Control variables	Yes
Constant	0.105 ** (0.042)
Time-fixed	Yes
Province-fixed	Yes
N	300
R2	0.761

Notes: *, **, *** denote significance at the 10%, 5%, and 1% levels, respectively, with robust standard errors reported in parentheses.

presents the regression results of the single-threshold model. The estimated threshold value of 0.534 for the RDE shows statistically significant positive effects on AGTFP on both sides of this critical point. Notably, the coefficient increases from 0.177 to 0.258, indicating a nonlinear enhancement effect, where exceeding this development threshold generates larger marginal gains in AGTFP.

In the early stage of RDE development, the integration of new production factors, such as digital technologies, into agricultural systems requires gradual adaptation and assimilation. Constrained by factors such as geographical heterogeneity and cultivation practices, this period results in relatively slow improvements in AGTFP. With strengthened policy support and increased capital investment from relevant authorities, the RDE has expanded substantially in both depth and scope, facilitating the deeper integration of digital technologies into agricultural production systems and generating progressively larger marginal contributions to AGTFP. These findings collectively indicate that RDE development exerts a nonlinear enhancement effect on AGTFP. Thus, H4 is proven.

6.2. Heterogeneity Analysis

Table 7. Results of heterogeneity analysis.

Variable	Main Grain Production Areas			Main Marketing Areas			Balancing Areas
	educ	scal	AGTFP	educ	scal	AGTFP	educ	scal	AGTFP
	(25)	(26)	(27)	(28)	(29)	(30)	(31)	(32)	(33)
RDE	0.045 *** (0.016)	0.058 *** (0.013)		0.120 *** (0.045)	0.105 (0.220)		0.062 ** (0.028)	0.072 * (0.040)
educ			0.108 ** (0.046)			0.096 *** (0.035)			0.041 * (0.025)
scal			0.029 *** (0.009)			0.029 (0.034)			0.118 ** (0.056)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Constant	0.062 *** (0.008)	0.015 (0.281)	0.056 ** (0.028)	0.163 (0.176)	0.399 (0.249)	0.151 *** (0.045)	0.165 *** (0.012)	0.092 * (0.052)	0.046 *** (0.009)
Time- fixed	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Province- fixed	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	130	130	130	70	70	70	100	100	100
R2	0.892	0.771	0.708	0.608	0.475	0.638	0.605	0.691	0.780

Notes: *, **, *** denote significance at the 10%, 5%, and 1% levels, respectively, with robust standard errors reported in parentheses. There are 13 major grain-producing areas in the country, namely Heilongjiang, Henan, Shandong, Sichuan, Jiangsu, Hebei, Jilin, Anhui, Hunan, Hubei, Inner Mongolia, Jiangxi and Liaoning; the main marketing areas are Beijing, Tianjin, Shanghai, Zhejiang, Fujian, Guangdong and Hainan; and the production and marketing balance areas are Shanxi, Ningxia, Qinghai, Gansu, Yunnan, Guizhou, Chongqing, Guangxi, Shaanxi and Xinjiang.

shows that the RDE has statistically significant positive effects on rural human capital at the 1% level in the main grain-producing and marketing zones, and at the 5% level in the balancing zones. This analysis explores regional variations in the impact of rural human capital on AGTFP. The regression results indicate that rural human capital exerts the strongest positive effect on AGTFP improvement in the main marketing areas. This outcome reflects the economic development and policy advantages present in these regions, which are typically more developed provinces. These advantages facilitate rapid the RDE development and attract digitally skilled agricultural talent, thereby enhancing the mediating role of rural human capital in the main marketing areas.

The analysis of regional heterogeneity shows that the RDE has statistically significant positive effects on agricultural-scale operations in both main grain-producing and balancing areas. The main grain-producing areas bear the responsibility of ensuring stable production and supply. Expanding agricultural operations, as a key measure to increase grain output, contributes positively to AGTFP by optimizing factor inputs, improving their efficiency, and reducing the negative environmental impacts of redundancy, thereby balancing economic returns with environmental protection.

7. Conclusions and Suggestions

This study uses panel data from 30 provinces and municipalities over the period 2013–2022 to examine the specific impacts and transmission mechanisms through which the RDE influences AGTFP. Using econometric models, the analysis investigates the mediating roles of rural human capital accumulation and agricultural-scale operations, while identifying threshold effects in the relationship between RDE and AGTFP. The main findings are summarized as follows.

First, baseline regressions show that the RDE has statistically significant positive effects on AGTFP and its component indices, with the results robust to both robustness checks and endogeneity tests. Second, the mediation analysis indicates that enhancing rural human capital and expanding agricultural-scale operations are key channels through which the RDE improves AGTFP. Third, threshold regression results show that the effect of RDE on AGTFP follows a nonlinear pattern, with marginal gains increasing once the threshold is exceeded. Finally, regional heterogeneity analysis demonstrates that rural human capital and agricultural-scale operations play a more pronounced mediating role in primary grain producing areas.

Based on the aforementioned findings, this paper makes several policy proposals.

First, policy coordination should align the development of the RDE with the objectives of agricultural green transformation. This requires the systematic application of digital technologies, including drone monitoring systems, 5G networks, and cloud computing, across agricultural value chains, with particular emphasis on building agricultural information platforms to improve production decision-making. Public investment should prioritize the adoption of digital technologies in green agricultural practices, with a focus on innovation in germplasm resource management, precision cultivation, and energy efficient agricultural machinery, in order to fully realize digital productivity gains.

Second, the development dividends of the RDE should be fully utilized to strengthen rural human capital and promote appropriately scaled agricultural operations. Local governments should prioritize key regions and establish well-structured digital training systems to cultivate rural digital talent. Emphasis should be placed on improving farmers’ digital literacy, particularly their ability to apply digital agricultural technologies and their awareness of green production practices. Information and communication technologies should be used to develop and improve regional land transfer platforms, thereby alleviating information constraints in traditional land markets and enhancing matching efficiency among market participants. By facilitating land transfer and consolidating fragmented land resources, scaled agricultural operations can be further promoted. Additionally, new agricultural business entities should be encouraged to disseminate green production technologies and environmental practices to smallholders, thereby supporting their transition toward green production.

Third, the policy framework should seek to narrow regional development disparities and prevent potential polarization effects in the advancement of the RDE. This requires establishing coordinated governance mechanisms for the allocation of digital dividends to strengthen RDE’s contribution to AGTFP in core production regions, while implementing targeted provincial support measures to promote knowledge diffusion, improve supply chain efficiency, and enhance market integration in both marketing and balanced regions. Such efforts help address regional digital gaps and optimize the agricultural productivity gains from digital transformation across different regional contexts.

8. Research Limitations

Sample Selection: This study uses provincial panel data from China, which ensures strong regional representativeness and relatively complete coverage. However, it may not fully reflect differences at the prefecture or county levels. In particular, during rural digital development and the green transformation of agricultural production, micro-level variations exist in digital infrastructure, production patterns, and related factors. Future research could adopt a county-level perspective to provide a more granular analysis of how the RDE affects AGTFP and employ micro-level survey data to examine the mediating roles of human capital and agricultural-scale operations, thereby enhancing both explanatory depth and policy relevance.

Variables Measurement and Model Construction: The RDE is measured using the entropy method, which mitigates potential weighting bias inherent in subjective approaches, while AGTFP is calculated using the SBM-GML model. Nevertheless, both remain composite indicators, and the selection and construction of their components may introduce measurement bias that could influence the conclusions. Future research could perform cross-validation using both single indicators and composite indices, or employ quasi-natural experimental designs to further address endogeneity and improve the robustness of findings.

Mechanism Analysis: Adjustments in rural human capital and agricultural-scale operations often involve time lags. This study focuses primarily on their indirect effects. Future research could incorporate lagged mechanism variables or develop dynamic models to better capture the long-term transmission pathways of these mediating effects.