Impact of Digital Villages on Agricultural Green Growth Based on Empirical Analysis of Chinese Provincial Data

Abstract

The construction of digital villages has progressed in tandem with the transformation of traditional production methods, offering new perspectives for agricultural green growth and sustainable development. This study employs the entropy value method alongside the super-efficient global SBM (Slacks-Based Measure) mixed function model, which assesses efficiency by accounting for both inputs and outputs, thereby facilitating a comprehensive evaluation of agricultural green growth. This methodology facilitates the examination of the correlation between digital villages and agricultural green growth, as well as the influence of digital villages on this growth. Furthermore, the utilization of financial resources is employed as a mediating variable to elucidate the mechanism of action. The utilization of green finance and agricultural insurance can be facilitated by the establishment of digital villages, and that has been shown to promote agricultural green growth. Additionally, the promotion of agricultural green growth by digital village construction is stronger in middle-altitude regions, non-grain-producing regions, and regions where the digital literacy of the rural labor force is higher than average, as well as areas where the use of agricultural film is higher than average. Accelerating the construction of digital villages and promoting the utilization of rural financial resources while adapting the digital village development to local conditions are crucial for effectively fostering agricultural green growth and sustainable agricultural development.

1. Introduction

The promotion of green growth in agriculture is a substantial endeavor that can expedite the establishment of a resilient agricultural nation, enhance agricultural modernization, and foster sustainable, high-quality agricultural development. The 20th CPC National Congress Report explicitly underscores the imperative of prioritizing agricultural and rural development. Agricultural production has substantially enhanced food security and farmers’ incomes. Nonetheless, conventional agricultural production techniques are unsustainable [1]. The advancement of agriculture in a vast nation with smallholder farmers is limited by natural resources. China sustains 20% of the global population utilizing merely 7% of the world’s arable land and limited water resources [2]. The per capita arable land area in China is 0.09 hm², which is less than 40% of the world’s average [3]. Therefore, the scale of agricultural production cannot be enlarged. Furthermore, the increase in food production depends on inputs from production factors. The sloppy development and excessive reliance on pesticides and chemical fertilizers have caused environmental pollution and soil degradation in farming areas. This has brought a serious burden on the ecological environment [4] and restricted further development of agriculture. In 2024, China Central Government issued a primary directive addressing agricultural development, emphasizing the need for “greening agriculture”, while the government’s work report stressed the significance of “speeding up the environmentally conscious transformation of the agricultural model”. This highlights the urgency and necessity for a fundamental transformation in agricultural economic growth and a reconfiguration of the existing system to facilitate sustainable agricultural development. This raises a question of both theoretical and practical importance: how can sustainable growth in agriculture be encouraged?

The digital economy, driven by developments in information and communication technology (ICT), has become a key accelerator for global social-economic growth [5]. Its expansion has promoted the application of digital technologies across various sectors, consequently enhancing production efficiency and optimizing industrial structures [6,7]. In agriculture, the digital economy has enabled solutions for sustainability challenges while boosting output [8]. In this context, the concept of digital villages has emerged as a localized implementation of the digital economy within rural regions [9]. By incorporating technologies such as smart systems [10] and data-driven approaches [11], digital villages improve farmers’ access to markets and information networks, thus advancing the rural digital economy. This transformation fosters agricultural modernization and sustainability as rural areas increasingly adopt new technologies, shifting from traditional to more efficient, eco-friendly farming practices [10]. The digital economy provides the technological backbone for digital villages, enhancing resource efficiency, market connectivity, and rural production. Moreover, through precision agriculture, automation systems, digital platforms, and real-time agricultural data, digital technologies facilitate the sustainable transformation of rural economies [12,13]. Many developed nations have adopted strategies to promote digital villages. For instance, Germany’s Digital Agenda for Agriculture, launched by the Federal Ministry of Food, seeks to boost agricultural productivity and sustainability through digital technologies [14]. Similarly, China’s Central Document No. 1 (2018, 2024) [15,16] emphasizes building digital villages, advancing smart agriculture, and shifting agricultural practices towards digitalization, intelligence, and environmental sustainability.

The establishment of digital villages offers innovative concepts for the green growth of agriculture. These concepts encompass the enhancement of precision agriculture, invigorating production factors, integrating with the digital marketplace, and facilitating the dissemination of agricultural technology. Establishing digital villages involves using digital technology, artificial intelligence, and advancements in smart and precision agriculture, encompassing automation technology, agrarian information systems, physical information systems, agricultural data, and additional modalities [17]. Digital technology contributes to the growth of agriculture in three main ways. First, precision agriculture and smart agriculture use agricultural production data based on user interaction [18] to improve the efficiency of pesticides, fertilizers, water, and other resources used throughout agricultural cultivation. Furthermore, producers are going to be capable of promptly modifying agricultural strategies in response to information regarding climate change and natural disasters. For example, using remote sensing data and satellite imagery in agriculture can facilitate the monitoring of crop growth patterns with high precision. The integration of precipitation sensors with irrigation systems can also enable the implementation of water-saving irrigation strategies [19]. Second, information and communication technology (ICT) is active in agricultural and rural development [20]. Farmers and agribusinesses have access to information on weather, prices, and marketing through the Internet, increasing income growth opportunities and improving welfare [21]. The application of information technology has the potential to improve agricultural productivity. As a result, farmers are more inclined to adopt smart agriculture and green production technologies [22] to improve agricultural productivity [23], facilitating the transition from traditional to digital agriculture and promoting the growth and inclusiveness of the agricultural economy [21]. Third, the implementation of agricultural automation has the potential to reduce labor costs, increase the efficiency of agricultural cultivation, reduce the risk of labor-related fatigue and health hazards faced by farmers, and stimulate the growth of the agricultural sector [24]. Small robots utilized for weeding, fertilizing, and seeding in precision agriculture can autonomously eliminate weeds without manual oversight or harm to crop development [25], thereby diminishing labor requirements and enhancing the productivity of farming planting. Consequently, examining the trend of digital village building and thoroughly analyzing the determinants of agricultural green growth is a crucial subject for achieving sustainable agricultural development.

Current research on digital villages revolves around three primary domains. Firstly, the development and measurement of digital villages are explored. Faxon (2022) integrated agricultural research with digital geography, introducing the concept of digital villages as networked social spaces [26]. Sampetoding (2024) conducted a systematic literature review, elucidating the characteristics and developmental stages of digital transformation and its implications for smart villages [27]. Li (2022) quantified the development level of digital villages and spatial regional disparities by constructing a digital village system, identifying population density, industrial structure, and economic development as key determinants influencing the level of digital villages [28]. Secondly, the practical logic and challenges inherent in digital village construction are examined. Brynjolfsson (2019) [29] believed that labor productivity gains from technology adoption are unevenly distributed and concentrated among a few beneficiaries; Li (2023) [30] emphasized that existing development of digital villages remains susceptible to enhancement and that there are practical misconceptions such as placing too much emphasis on hardware and equipment, the spread of data formalism, the irrational allocation of public resources, and over-reliance on operators; and Lv (2020) [31] argued that for digital villages to play their role, rural residents need to have the ability to pay, the ability to collect and process information, and the willingness to improve production and life. Thirdly, the construction of digital villages plays a crucial role in advancing rural economic development. Digital villages contribute to achieving the common prosperity of farmers and rural communities [32], while also driving rural economic growth through various pathways, such as promoting rural industrial development [33], fostering urban–rural integration [34], and supporting rural revitalization [35]. Additionally, digital villages stimulate entrepreneurship [36] and help raise household incomes [37].

Regarding green growth in agriculture, Li and Xu (2022) [38] contended that green growth within agriculture is characterized by the diminishment of chemical inputs, including pesticides and fertilizers, the reduction in carbon dioxide emissions, the enhancement of production and income efficiency, and the exploration of the synergistic interplay between “emission reduction” and “efficiency increase”. The concept of green growth in agriculture is both comprehensive and systematic, encompassing a range of influencing factors including economic, social, and environmental considerations [39]. From the existing research, scholars have, respectively, discussed the digital villages [40], digital financial inclusion [41,42], agricultural insurance [43], agricultural socialization services [44], and human capital [45]. Liu (2019) suggested that China’s agricultural green growth reached a turning point in 2010, largely due to advancements in agricultural technology, energy utilization, and pollution control technologies [39]. In a study by Khanh Chi (2022), the factors influencing farmers’ adoption of environmentally sustainable practices were analyzed, using structural equation modeling to assess the effects of farm utilization capacity and technological spillover on green production and economic growth. The results revealed that both factors positively impacted green production and contributed to economic growth [46].

Several studies have been undertaken to evaluate the impact of digital villages on agricultural green growth. The prevailing consensus is that the development of digital villages, and the digital economy more broadly, facilitates agricultural green growth [33,40,47]. This perspective is reinforced by research, including that of Jiang (2022), who believed that the positive effects may be strengthened over time [48]. The mechanism of the effect is as follows: firstly, digital villages can optimize the allocation of resources and mitigate the mismatch [40,49] of land, capital, labor, and other factors to foster green growth in agriculture, especially in areas with high human capital [50]. Secondly, digital villages can promote the scale operation of the agricultural management main body, leading to the optimization of the agricultural industrial structure, the division of labor refinement, and the specialization of production [47], which will in turn promote agricultural green growth. Thirdly, the digital countryside promotes agricultural green growth through the utilization of the Internet and the dissemination of digital technologies [12]. Fourthly, the development of the digital countryside exhibits spatial relevance, with its spillover effects on agricultural green growth being either positive or negative [48,51].

China’s formal proposal for digital village construction is relatively recent. While some of the existing literature has empirically explored the role of digital village construction in enhancing agricultural production technologies and driving green growth, several gaps remain. First, no studies have yet examined the use of financial resources as an intermediary mechanism in analyzing the impact of digital villages on agricultural green growth. Financial support can alleviate farmers’ financial constraints, promote the adoption of green technologies and the scaling up of operations, enhance production efficiency, and optimize resource allocation. In addition, financial innovation amplifies the inclusive effects of digital technology, facilitating the coordinated development of rural finance and regional economies. Moreover, financial resources, through risk management tools such as agricultural insurance, help farmers cope with uncertainties, thereby ensuring the sustainability of the green transition. Second, heterogeneity analysis has not adequately identified the existing variations in the impact of digital village construction on agricultural green growth. Differences in regional, resource, and socio-economic conditions may result in varying outcomes of digital village construction across different areas. Overlooking these disparities may oversimplify the policy effects, leading to an inaccurate assessment of the full scope and potential benefits of digital village construction. A thorough investigation of heterogeneity can provide a more comprehensive understanding of its mechanisms, revealing the diversity and variability in the influence of digital village construction on agricultural green growth, thereby offering scientific support for the formulation of tailored policies.

This study seeks to develop a regression model utilizing provincial-level panel data from China for the years 2021 to 2022 to empirically examine the mechanisms through which digital villages impact agricultural green growth and to assess the mediating effect of financial resource utilization. Distinguishing from previous studies, this study employs the inclusive impacts of digital technologies [52] and knowledge spillover resulting from digital village construction to foster green agricultural development. It examines the influence of digital village initiatives on agricultural sustainable growth, green technological advancement, and technological efficiency. Additionally, this research investigates the mediating effect of financial resource utilization on promoting agricultural green growth. It also examines the heterogeneity of the impact of digital village construction on agricultural green growth by considering diverse perspectives, such as natural conditions, the digital literacy of rural labor, and the means of agricultural production.

This research seeks to address the following critical questions:

How does digital village construction influence agricultural green growth?

What role does utilizing financial resources play in facilitating agricultural green growth?

What are the key heterogeneities in the influence of digital villages on sustainable agricultural growth?

The principal aim for this study is to furnish thorough responses to these inquiries. The anticipated research outcomes are as follows: First, the findings will offer a scientific basis for government policymakers to devise strategies that promote agricultural green growth. Second, we will offer pragmatic guidance for farmers and agribusinesses in implementing eco-friendly production practices, thus aiding in the sustainable advancement of agriculture. Third, this study will enhance the deployment of financial resources, including agricultural insurance and green finance, while fostering growth as well as innovation in agricultural sustainable technologies.

2. Theoretical Analysis and Research Hypotheses

2.1. The Immediate Effect of Digital Villages on Green Growth in Agriculture

Digital villages leverage digital technology as an instrument, and advances in digital technology promote the modernization of agriculture and lay the foundation for green growth in agriculture. Firstly, considering the greening requirements and the attributes of digital technology, the utilization of digital technology aligns to the fundamental essence for green development [53]. In particular, digital technology can facilitate the transition of agriculture from the conventional model of high input, high consumption, and high environmental impact to a modern model of low carbon, energy-efficiency, and high yield at both the input and output levels. On the one hand, the advancement of digital technology facilitates the advancement of agricultural development enhances the efficiency of green agricultural production. The utilization of sensors, remote sensing technology, and big data analysis provide the necessary information technology support for modern agriculture [54]. On the other hand, digital technology is employed to prevent and control agro-ecological environmental pollution, construct pollution detection systems for the atmosphere, soil, and agricultural surface sources, track and trace the pollution of agricultural production, and respond rapidly to sudden environmental problems to repair the ecological environment. Secondly, digital villages can effectively stimulate the vitality of production factors. The incorporation of digital villages introduces a new production factor, namely data, which is integrated into the existing structure of factor endowments. The enhancement of factor endowments leads to a novel resource allocation system, optimizes productive as well as unlimited movement through manufacturing variables [33] along with increases in the efficiency of production, distribution, circulation, and consumption. As a result, agricultural green growth is encouraged. Thirdly, the digital market, which relies on the Internet, the Internet of Things, blockchain, and other technologies can effectively reduce transaction costs and the degree of information asymmetry between economic agents. Furthermore, it can play a regulating role in the price mechanism [55]: the effective market was enhanced, and agricultural products were sold via e-commerce, expediting market circulation and offering consumers green agriculture. The sale of agricultural products through e-commerce platforms improved market efficiency and provided consumers with green agricultural product purchasing channels, thereby supporting green agricultural growth. Fourthly, digital villages facilitate the dissemination of agricultural technology and enhance the information literacy and competencies of rural residents [56]. Farmers can access agricultural cultivation-related information through various media, including television, the Internet, and short videos, thereby increasing the accessibility of information for farmers, improving their understanding of environmental pollution, and transitioning from a conventional production system to a more sustainable one.

Green growth in agriculture comprises two fundamental components: technological advancement in sustainable agricultural production and enhancement of the effectiveness of using sustainable technologies in agriculture. Endogenous technological progress is a critical determinant of sustained economic growth [57,58]. Technological progress enables the production possibility boundary curve to shift outward [59], allowing for the realization of greater outputs under the same resource conditions. For example, the implementation of an automatic irrigation system can facilitate enhanced efficiency in water utilization, thereby promoting a more balanced growth of crops. Similarly, the deployment of genetic enhancement technology can bolster the resilience of crops to pests and diseases, curtail the reliance on pesticides, and ultimately enhance crop yield. The establishment of digital villages is thus correlated with the progression into agricultural green manufacturing techniques and improving green technology efficiency.

In summary, this study proposes Hypothesis 1.

H1: 

Digital villages construction contributes to green growth in agriculture.

H1.1: 

Digital villages construction promotes the development of agricultural green technology.

H1.2: 

Digital villages construction promotes the efficiency of agricultural technology.

2.2. Mechanisms of the Impact of Digital Villages on Green Growth in Agriculture

The role of finance is to facilitate economic growth [60]. The following is accomplished by optimizing how to allocate resources and achieving a balance between risk and return [61]. Green finance serves as a conduit between the financial sector, environmental enhancement, and green growth [60]. It can facilitate the dissemination of technology and the development of eco-friendly infrastructure. The central tenet of green finance is the internalization of environmental externalities, which serves to effectively promote green development. From an economic perspective, enterprises that create environmental benefits can obtain a green product premium, thus obtaining direct or indirect economic returns. This subsequently directs the allocation of financial resources to efficient and environmentally sustainable industries, fostering industrial advancement and the enhancement of environmental quality [62]. Conversely, enterprises that generate environmental pollution must bear the associated costs, which acts as a disincentive for polluting investments on the supply side of capital. This has the effect of alleviating the traditional structural resource mismatch problem of finance in supporting the development of micro-enterprises [63].

The utilization of technologies such as big data, artificial intelligence, and cloud computing enables digital villages to facilitate green financial projects, offering the benefits of convenient access to information, rich interaction, and a reduced cost of information interaction [64]. Firstly, digital villages provide a means of overcoming information asymmetry in product markets. Agricultural products and food are considered trust goods, and information asymmetry can result in adverse selection that is detrimental to all parties involved. Those consumers who are at a disadvantage in terms of the information they possess will elect to suppress the price to avoid the loss that would otherwise be incurred as a result of the information asymmetry. Furthermore, prices that are too low may result in producers being reluctant to supply high-quality products, thereby impeding the attainment of both superior quality and a competitive price. In the field of information economics, the processes of signaling and information screening serve to facilitate the realization of good quality and good price. The digital villages facilitate the monitoring and traceability of the production process through the flow of information and enable the effective matching of production and marketing information through an information-matching mechanism [52]. This allows for the fulfillment of the purchase demands of green consumers. The premium attached to green products will result in an expansion of the production scale of enterprises and farmers, thereby generating the financial demand for green development. Secondly, the digital villages facilitate the resolution of the information asymmetry between financial institutions and green agricultural business entities. The deployment of digital technology enables enterprises and farmers to establish and accumulate credit records, thereby enhancing their capacity to obtain green financial support, reducing the cost of acquiring “green knowledge”, and effectively managing the risk of bad debt [65]. The utilization of blockchain technology and the construction of big data platforms enables financial institutions to mitigate information asymmetry, enhance the efficacy of green project assessment, circumvent the practice of “greenwashing” by enterprises, and monitor the progress of projects in real-time, thereby ensuring that the projects are promoted by the established plan. To illustrate, financial institutions such as the Chongqing Agricultural and Commercial Bank and the Huzhou Bank have implemented online green credit management systems, with intelligent identification of green credit, automatic measurement of environmental benefits, and online management of the entire process. This has notably enhanced the efficiency of green financial utilization. Thirdly, the digital villages have resulted in a reduction in transaction costs. The utilization of technologies such as mobile payments and digital wallets has enhanced the scope of financial services in rural areas, facilitating the transfer of funds and stimulating the enthusiasm of enterprises and farmers to engage in green financial projects. Fourthly, the implementation of innovative green finance regulatory tools serves to prevent the emergence of new types of financial risks and to reduce compliance burdens. The utilization of big data platforms to facilitate cross-regional and cross-financial institution green information sharing provides the underlying data support necessary for the effective supervision of financial activities. The deployment of machine learning and other advanced methods for the optimization of financial risk early warning models to enhance the timeliness and accuracy of early warning, and the use of satellite remote sensing and other sophisticated technologies for the tracking and supervision of green financial projects, are also contributing factors. Furthermore, the realization of environmental and social risk reviews and the implementation of fine-grained supervision of green financial business represent additional key elements. The specific logic diagram is shown in Figure 1.

With a balance of RMB 30.08 trillion at total conclusion for 2023, China’s domestic and foreign currency green loans had grown by 36.5% year over year and made up 13.4% of the total loan balance. Green loan funds were primarily directed towards the green upgrading of infrastructure, the clean energy industry, and the energy conservation and environmental protection industry, with respective loan balances of RMB 13.09 trillion, RMB 7.87 trillion, and RMB 4.21 trillion [66]. Green finance, through various channels such as bonds, funds, and insurance, has provided financial backing as well as risk management for the green growth of agriculture, thereby promoting the flow of capital into socially beneficial green agricultural projects and fostering sustainable agricultural growth.

In summary, this study proposes Hypothesis 2.1

H2.1: 

Digital villages promote green growth in agriculture through green finance.

Agricultural insurance represents an effective risk-sharing mechanism, whereby risks are transferred from farmers to the government and insurance companies. To encourage farmers to purchase agricultural insurance, the government bears a significant proportion of the associated subsidy costs [67]. The development of agricultural insurance in China is shaped by a combination of policy-based insurance attributes and institutional inducements [68]. The establishment of digital villages can facilitate the popularization and advancement of policy-based agricultural insurance. Firstly, it mitigates the information asymmetry that exists between insurance companies and farmers. Farmers possess a more comprehensive understanding of the risks associated with agricultural production. In areas where disasters occur more frequently, there is an increase in enthusiasm among farmers for agricultural insurance, which leads to adverse selection. Farmers may relax their precautions in agricultural production due to the insurance, and in the event of an insured loss, they falsify the extent of the loss to obtain a higher level of compensation, thereby generating moral hazard. The occurrence of adverse selection and moral hazard will result in a high payout rate, which in turn will bring risks to the operation of insurance companies and thus impede agricultural insurance from developing. Digital villages can be built with information platforms, blockchain technology, remote sensing technology, and other effective constraints on farmers to ensure and adopt disaster mitigation measures. This can mitigate the moral hazard and adverse selection resulting from information asymmetry in the claims process of insurance companies [69], thereby fostering the robust development and expansion of the agricultural insurance market. Secondly, it enhances the operational efficacy of the agricultural insurance market. The application process for agricultural insurance can be simplified, the threshold for farmers’ participation reduced, and digital platforms developed to facilitate the purchase and claims operations of insurance, thereby reducing the operating costs of insurance companies and the transaction costs of farmers [70]. The utilization of digital platforms serves to reinforce the promotion of agricultural insurance, facilitate the uptake of insurance plan proposals by farmers through the establishment of trust mechanisms, and stimulate farmers’ willingness to pay for the out-of-pocket portion of policy insurance. Thirdly, it optimizes the agricultural insurance rate setting. The establishment of digital villages facilitates the detection of alterations in the external environment of agricultural production, while agricultural output may be monitored and evaluated through making use of various devices like satellite imagery [71], thereby supplying essential data for rate formulation. For example, Shaanxi took the lead in implementing a dynamic rate mechanism for agricultural insurance in 2022 to accurately control agricultural production risks.

Agricultural insurance can promote green agricultural growth through two main mechanisms. First, it encourages farmers to adopt green production technologies to produce higher value-added green agricultural products. Generally, farmers in developing countries tend to adopt safety-first production behaviors, basing their production decisions on survival strategies to minimize the probability of income falling below a certain threshold [72]. Agricultural insurance can facilitate farmers’ investment in more profitable green products, as it transfers excess risk to a third party, thereby reducing the perceived risk associated with improving production methods and guiding farmers to transform traditional production practices [43]. This, in turn, stimulates farmers to purchase agricultural machinery and equipment, thereby promoting green agricultural growth.

Second, agricultural insurance can improve farmers’ access to credit. Due to the relatively small average loan amounts for farmers, the seasonal nature of agricultural production, and the increased exposure to natural disaster risks, the likelihood of default risk increases, leading to higher transaction costs for financial intermediaries, making it challenging for formal financial institutions to provide loans. Agricultural insurance reduces the default risk of farmers, thereby contributing to indirect benefits in the supply of credit [73]. Reduced credit constraints enable farmers to improve agricultural production technologies [74], promoting green agricultural growth. For example, to support the development of the corn industry in Heilongjiang Province, China Pacific Insurance (Group) Co., Ltd. collaborated with relevant banks and futures companies to pilot corn price index insurance. Credit inclusion provided farmers with the liquidity needed for planting, supporting agricultural production and farmers’ financing needs, thus facilitating the expansion of production scale and the adoption of green production technologies. The particular logic diagram is illustrated in Figure 2.

In 2023, agricultural insurance in China provided risk protection for agricultural development amounting to RMB 4.98 trillion, with premiums reaching RMB 143 billion, representing a year-on-year growth of over 17% [75]. A comprehensive agricultural insurance product supply system has been established, covering costs, yields, income, price indices, and weather indices and spanning a wide range of farm products in sectors such as farming, forestry, animal husbandry, and fisheries. Technological advancements have enhanced the precision of loss assessment and expedited claim settlements, improving underwriting and claim efficiency and strengthening agricultural risk management capabilities. As of the end of June 2023, the balance of insurance funds directed towards green development-related industries stood at RMB 1.67 trillion, an increase of 36% year on year [75]. Insurance companies have consistently expanded green investment, primarily in green technology, clean energy, energy conservation, and environmental protection, thereby promoting green growth.

In summary, this study proposes Hypothesis 2.2.

H2.2: 

Digital villages promote green growth in agriculture through agricultural insurance.

3. Research Methodology

3.1. Data Sources

This research investigates the impact of digital village development on green agricultural growth across 31 provinces, autonomous regions, and municipalities in China, using data from 2011 to 2022. This study integrates variables concerning agricultural inputs and outputs and various control and mediating factors. These data were sourced from databases such as EPS and Wind. In instances where data were unavailable, the values were estimated through interpolation and the application of an average growth rate. The digital financial inclusion index, which serves as a measure of digital villages, was derived from the Peking University Financial Inclusion Index [76].

3.2. Variable Description

3.2.1. Explained Variable

Agricultural green growth, as determined by agricultural green total factor productivity (AGTFP), is the primary explanatory variable. Following previous scholars’ methods [49,77,78], input variables were quantified using three primary factors: labor, land, and capital. The output variables were categorized into desired and non-desired outputs. Desired output was represented by the total agricultural output value, with nominal values adjusted to real values using the agricultural output value index, with 2010 as the base period [78]. Non-desired output was measured through environmental pollutants, including total carbon emissions, total nitrogen emissions, and total phosphorus emissions. Total carbon emissions $E=\sum E_i=\sum T_i\times {\delta }_i$, $E_i$ is emissions from carbon sources such as crops, fertilizers, pesticides, agricultural films, irrigation, diesel, etc.; ${\delta }_i$ is the carbon emission factor, calculated with reference to the relevant data in the text of Liu (2021) [77]. Total phosphorus as well as nitrogen emissions combined $F=\sum F_i=\sum G_i\times {\phi }_i, F_i$ is total nitrogen and total phosphorus; ${\phi }_i$ is a nitrogen and phosphorus emission factor [79]. Details are shown in

Table 1. Indicators for measuring AGTFP.

Category	Variable	Indicator	Evaluation	Unit
Input	Labor	Labor inputs	Number of people employed in agriculture	104
	Land	Land inputs	Total sown area of crops	103 hm2
	Capital	Agricultural machinery power inputs	Total power of agricultural machinery	104 kW/h
		Fertilizer	Fertilizer use	104 kg
		Pesticide	Pesticide use	104 kg
		Film	Agricultural plastic film use	104 kg
	Resource and energy	Water	Effective irrigation area	103 hm2
		Electricity consumption	Rural electricity consumption	105 kW/h
Output	Desired output	Gross agricultural product	Agricultural value added	billion
	Undesired output	Carbon emission	Agricultural carbon emission	104 kg
		Surface pollution	Agricultural surface pollution	104 kg

.

This study employs the super-efficient global Slacks-Based Measure mixed function model along with the Malmquist productivity index to assess AGTFP. The SBM model assumes that all DMUs operate under similar production technologies and environmental conditions, with production technology sets being additive and input and output data being accurate, reliable, non-negative, and comparable. The model is typically based on the assumption of CRS, although VRS can be applied in specific cases. Despite the model’s computational complexity and sensitivity to outliers, which may lead to high computational costs when handling large datasets, it may fail to accurately reflect efficiency in cases of extreme efficiency values. However, compared with traditional SFA and DEA models, the SBM model can simultaneously accommodate both radial and non-radial distance functions, providing more flexibility within measuring. This advantage allows it to address situations where both input and output exhibit radial and non-radial characteristics, which conventional SFA and DEA models cannot resolve. In the context of agricultural production, the method is capable of fully considering both desired and undesired outputs, and the index is more adaptable. Furthermore, the Malmquist index can be decomposed into technical progress (TC) and technical efficiency (TE), which allows for a comprehensive exploration of the growth components of AGTFP. Concerning methods of Zhou and Yin (2024) [80] and Lee and Lee (2022) [62], the SBM model is thus set in.

$${min\rho }^{\ast }={\displaystyle \frac{1+\frac{1}{m}{\sum }_{i=1}^m\frac{s_i^x}{x_{i0}}}{1-\frac{1}{s_1+s_2}\left({\sum }_{k=1}^{s_1}\frac{s_k^y}{y_{k0}}+{\sum }_{l=1}^{s_2}\frac{s_l^z}{z_{t0}}\right)}}$$

(1)

$$s.t.x_{i0}^t\ge \sum _{j=1}^n{\lambda }_j{x}_j-s_i^x,\forall i;$$

$$y_{k0}^t\le \sum _{j=1}^n{\lambda }_j{x}_j+s_k^y,\forall k;$$

$$z_{t0}^t\ge \sum _{j=1}^n{\lambda }_j{z}_j-s_l^z,\forall l;$$

$$1-{\displaystyle \frac{1}{s_1+s_2}}\left({\sum }_{k=1}^{s_1}{\displaystyle \frac{s_k^y}{y_{k0}}}+{\sum }_{l=1}^{s_2}{\displaystyle \frac{s_l^z}{z_{t0}}}\right)>0$$

$$S_i^x\ge 0,S_k^y\ge 0,S_l^z\ge 0,{\lambda }_j\ge 0, \forall i,j,k,l$$

In Equation (1), the variable ρ represents the efficiency value of the decision unit. The slack variables for inputs and outputs (desired outputs, undesired outputs) are represented by $s^x,{ s}^y,s^z$, and the weight vector is represented by λ.

The Malmquist index (Equation (2)) is used to measure AGTFP. The efficiency value of inputs and outputs (desired outputs, undesired outputs) in period t under the current period’s lower production frontier is denoted by $E_c^t\left(x^t,y^t,z^t\right)$. This value is derived from the super-efficient global SBM model.

$$\begin{array}{c}{ML}_c^{t+1}\left(x^t{,y}^t,z^t,x^{t+1},y^{t+1},z^{t+1}\right)\\ ={[{\displaystyle \frac{E_c^t\left(x^{t+1},y^{t+1},z^{t+1}\right)}{E_c^t\left(x^t,y^t,z^t\right)}}\times {\displaystyle \frac{E_c^{t+1}\left(x^{t+1},y^{t+1},z^{t+1}\right)}{E_c^{t+1}\left(x^t,y^t,z^t\right)}}]}^{1/2}\end{array}$$

(2)

According to the Malmquist index decomposition, AGTFP can be expressed as the product of TE and TC.

$${AGTFP}_i^{t,t+1}={ECTFP}_i^{t,t+1}+{TCTFP}_i^{t,t+1}$$

(3)

Since the calculated AGTFP represents dynamic changes and reflects variations compared to the previous year, this study converts the calculated AGTFP chain indices into fixed-base indices, setting the 2010 AGTFP as 1, thereby obtaining the actual AGTFP for the sample.

$${AGTFP}_i={\prod }_{t+1}^t{agtfp}_t$$

(4)

3.2.2. Explanatory Variables

This study employs the digital villages index (DVI) as its core explanatory variable. The development of digital villages is evaluated using the DVI, concerning previous scholars [81,82]. The index is constructed of four Tier 1 indicators and ten Tier 2 indicators, as detailed in

Table 2. Indicators for measuring the DVI.

Tier 1 Indicators	Tier 2 Indicators	Nature	Weights
Digital Industry Development	Proportion of Taobao villages among administrative villages	positive	0.3475
	E-commerce sales and purchases	positive	0.1477
	Digital Inclusive Finance Index	positive	0.0294
Digital Information Foundations	Rural mobile phone ownership per 100 households	positive	0.0173
	Rural broadband access users	positive	0.0990
	Number of rural meteorological observation stations	positive	0.0334
Digital Service Level	Rural delivery routes	positive	0.0437
	Rural household spending per capita: amount spent on transportation and communication	positive	0.0429
Digitization of Agriculture	Rural per capita electricity consumption	positive	0.1904
	Mechanization of agricultural production	positive	0.0486

. To conduct the computation, the entropy value approach was employed.

The entropy approach was employed to compute the DVI as follows.

$$X_{ijt}={\displaystyle \frac{x_{ijt}-min{x}_{jt}}{max{x}_{jt}-min{x}_{jt}}}$$

(5)

In Equation (5), $X_{ijt}$ represents the standardized value, where i represents provinces (i = 1, 2, …, m; m = 31 for a total of 31 provinces), t represents years (t = 2011, 2022, …, 2022; for a total of 12 years), and j represents the indicators of the digital countryside (j = 1, 2, …, n; n = 12 for a total of 12 indicators); $max{x}_{jt}$ is the maximum value of the jth indicator for all years and $min{x}_{jt}$ is the minimum value.

$$P_{ijt}={\displaystyle \raisebox{1ex}{$X_{ijt}$}\!\left/ \!\raisebox{-1ex}{$\sum _{i=1}^m{X}_{ijt}$}\right.}$$

(6)

In Equation (6), $P_{ijt}$ is the weight of the jth indicator in province i in year t.

$$E_{jt}=-{\displaystyle \frac{1}{lnm}}{\sum }_{i=1}^m{P}_{ijt}ln{P}_{ijt}$$

(7)

In Equation (7), $E_{jt}$ is the entropy value of the jth indicator in year t.

$$d_{jt}=1-E_{jt}$$

(8)

In Equation (8), $d_{jt}$ calculates the information utility value. The information utility value of an indicator depends on the difference between its information entropy $E_{jt}$ and 1. Its value directly affects the size of the weights: the greater the information utility value, the greater the importance of the evaluation and the greater the weights.

$$W_{jt}={\displaystyle \raisebox{1ex}{$d_{jt}$}\!\left/ \!\raisebox{-1ex}{$\sum _{j=1}^n{d}_{jt}$}\right.}$$

(9)

In Equation (9), $W_{jt}$ calculates t year jth indicator’s weight.

$${DVI}_{it}={\sum }_{j=1}^n{W}_{jt}\times X_{ijt}$$

(10)

In Equation (10), ${DVI}_{it}$ calculates the digital villages development index (DVI).

3.2.3. Control Variables

Referring to previous studies, this study selected the following control variables: (i) industrial structure was quantified by the ratio of the added value of the secondary and tertiary sectors to regional GDP; (ii) natural disasters, measured by the proportion of the affected area in the sown area of grain crops; (iii) rural per capita electricity consumption, using a logarithm of rural electricity consumption to rural population size; (iv) per capita regional GDP, measured by the logarithm of the ratio of regional GDP to regional population; (v) government intervention, measured by the proportion of local fiscal expenditure in regional GDP; (vi) urbanization level, the percentage of the people living in cities compared to those living in rural; and (vii) the total power of agricultural machinery per unit, calculated by dividing the total power on farming equipment by the total planted crop area. Details are shown in

Table 3. Descriptive statistics of the main variables.

Variable	Explanation	Mean	Standard Deviation
Explained Variables
AGTFP	Refer to Equations (1) to (4) for specific calculations	0.170	0.380
Explanatory Variables
Digital villages index	Calculated by entropy method	0.025	0.046
Control Variables
Industrial structure	Value added of secondary and tertiary industries/Gross regional product	0.903	0.052
Natural disasters	Affected area/area sown with food crops	0.216	0.204
Rural electricity consumption	Ln (Rural electricity consumption/rural population)	6.793	1.090
GDP per capita	Ln (Gross regional product/regional population)	10.857	0.463
Government intervention	Local fiscal expenditure/Gross regional product	0.291	0.205
Urbanization level	urban population/(urban population + rural population)	0.592	0.130
Gross power per unit of agricultural machinery	Total power of agricultural machinery/total sown area of crops	7.026	3.636
Agricultural structure	Sown area of grain crops/total sown area of crops	0.660	0.146
Irrigated area	Effective irrigated area/total sown area of crops	0.460	0.207
Mediating Variables
Green bonds	Number of green bond issues	12.797	25.255
Agricultural insurance	Agricultural insurance premiums/value added of primary sector	0.014	0.018

.

3.2.4. Mediating Variables

(i) Green finance, using green bonds as a proxy variable, as proposed by Wang (2023) [1], whereby green bond amount issues is utilized for calculating green bond development; and (ii) agricultural insurance, as put forth by Zhu et al. (2021) [83], with the depth of agricultural insurance serving as the metric for measuring the level of agricultural insurance development. Agricultural insurance density is equal to agricultural insurance premiums divided by the value added of the primary sector.

3.3. Model Setting

3.3.1. Benchmark Regression Model

This study employed a two-way fixed effects model to assess influences of the DVI on agricultural green growth. The two-way fixed effects model was selected for the following reasons: (i) Due to differences in infrastructure levels, government support, and natural condition factors, digital villages may exhibit distinct characteristics. This subsequently influences estimation for AGTFP. Controlling for individual fixed effects mitigates this potential source of estimation bias. (ii) Agricultural cultivation is affected by seasonality, climate change, and policy changes, and controlling for time fixed effects can reduce this part of the estimation bias. (iii) It is possible that there is a bidirectional correlation between digital rural development and AGTFP. The use of bidirectional fixation can help to reduce the endogeneity problem. Clustered standard errors were used during the regression to control for regions, mitigating estimation bias caused by heteroskedasticity and serial correlation.

$${lnAGTFP}_{it}={\alpha }_0+{\alpha }_1{DVI}_{it}+\lambda {Control}_{it}+{\delta }_t+{\phi }_i+{\epsilon }_{it}$$

(11)

In Equation (11), AGTFP is the explanatory variable agricultural green total factor productivity; DVI is the digital villages index; Control is the control variable; ${\delta }_t$ represents area fixed effects, and ${\phi }_i$ represents time fixed effects; ${\alpha }_0$ is the intercept term; ${\alpha }_1$ and λ are the coefficients to be estimated; ${\epsilon }_{it}$ is the random perturbation term; and the subscripts i represents the province and t represents the year.

3.3.2. Mediation Effects Modeling

The mediated effects model also uses a two-way fixed effects model, utilizing Cluster control areas.

$$M_{it}={\beta }_0+{\beta }_1{DVI}_{it}+\lambda {Control}_{it}+{\delta }_t+{\phi }_i+{\sigma }_{it}$$

(12)

In Equation (12), $M_{it}$ is the mediating variable; DVI is the digital villages index; Control is the control variable; ${\delta }_t$ represents the area fixed effect; ${\phi }_i$ represents the time fixed effect; ${\beta }_0$ is the intercept term; ${\beta }_1$ and λ are the coefficients to be estimated; ${\sigma }_{it}$ is the random perturbation term; and the subscripts i represents the province and t represents the year.

4. Results and Analyses

This section initially examines the temporal progression of the DVI and AGTFP. Next, the correlation of each explanatory variable is studied. Then, benchmark regression models are utilized to identify the positive impact of digital villages on agricultural green growth, with the conclusions further validated through endogeneity tests and robustness checks. Subsequently, the mediation effect models reveal that digital villages promote agricultural green growth through green finance and agricultural insurance. Finally, the variability in the influence of digital villages on green growth under different objective situations is analyzed.

4.1. Temporal and Spatial Evolution

Figure 3 illustrates the spatial distribution of the digital villages index (DVI) across Chinese provinces for 2011 and 2022. In 2011, most western and northern areas in China exhibited low or moderate DVI levels, while coastal regions such as Shanghai, Jiangsu, and Guangdong had higher DVIs, highlighting a significant regional imbalance in digital village development. In contrast, by 2022, the DVIs in western and inland regions, including Yunnan, Xinjiang, and Gansu, significantly improved. Meanwhile, the high development levels in the eastern areas were sustained, indicating a continued advantage in digital village development. Overall, the gap in digital village development among regions in China narrowed by 2022, with significant improvement in the relative growth levels in western as well as inland regions.

Figure 4 presents the geographical spread of AGTFP among Chinese provinces for the years 2011 and 2022. In 2011, AGTFP was generally higher in the eastern coastal and northeastern regions, reflecting their advantages in agricultural technological advancements and resource allocation efficiency, while western regions, particularly Xinjiang, Tibet, and Gansu, exhibited relatively low AGTFP levels. By 2022, AGTFP levels in the eastern and central regions further increased, and some western provinces, such as Xinjiang and Gansu, also showed improvement. Overall, there was a widespread regional enhancement in total factor productivity in China by 2022; while the eastern coastal regions maintained a clear leading advantage, the western and central regions gradually caught up.

The spatiotemporal evolution of the DVI and AGTFP reveals a specific correlation between the two. As digital village development progresses, particularly in western and inland areas, the increase in the DVI appears to align with improvements in AGTFP. Therefore, the relationship between digital village development and green agricultural growth warrants further empirical research to explore how digital technologies can influence agricultural production methods and achieve sustainable development goals.

4.2. Correlation Analysis

4.2.1. Multicollinearity Test

A VIF was employed to check for multicollinearity amongst the explanatory variables prior to model estimation. According to findings, the benchmark regression’s explanatory factors have a mean VIF of 3.37, with a maximum value of 7.15, both below the threshold of 10, as shown in

Table 4. Multicollinearity test results.

Variable	Variable	Variable
GDP per capita	7.15	0.140
Urbanization level	6.47	0.155
Rural electricity consumption	3.24	0.308
Government intervention	2.5	0.401
Digital villages index	2.34	0.428
Gross power per unit of agricultural machinery	2.05	0.487
Industrial structure	2.05	0.488
Natural disasters	1.15	0.866
Mean VIF	3.37

. This implies that the regression model does not have a serious multicollinearity problem.

4.2.2. Correlation Matrix

The correlation matrix between explanatory variables as well as explained variables is presented in

Table 5. Matrix of variable correlations.

	AGTFP	Digital Villages Index	Industrial Structure	Natural Disasters	Rural Electricity Consumption	GDP Per Capita	Government Intervention	Urbanization Level	Gross Power Per Unit of Agricultural Machinery
AGTFP	1
Digital villages index	0.229 ***	1
Industrial structure	−0.071	0.453 ***	1
Natural disasters	−0.134 ***	−0.308 ***	−0.204 ***	1
Rural electricity consumption	0.040	0.540 ***	0.533 ***	−0.207 ***	1
GDP per capita	0.111 **	0.684 ***	0.661 ***	−0.319 ***	0.694 ***	1
Government intervention	−0.196 ***	−0.340 ***	−0.202 ***	0.067	−0.597 ***	−0.363***	1
Urbanization level	0.076	0.476 ***	0.598 ***	−0.196 ***	0.769 ***	0.853 ***	−0.498 ***	1
Gross power per unit of agricultural machinery	−0.174 ***	0.0420	0.214 ***	−0.112 **	−0.187 ***	0.0390	0.583 ***	−0.188 ***	1

Note: ** denotes p < 0.05, *** denotes p < 0.001.

. Government intervention and natural disasters have adverse effects on AGTFP, while economic development and DVI levels have a positive impact. Urbanization and agricultural mechanization also significantly promote agricultural green development and economic growth.

4.3. Benchmark Regression

Table 6. Impact of digital villages on AGTFP.

Variables	AGTFP	Technical Efficiency	Technical Progress
	(1)	(2)	(3)
Digital villages index	1.592 ***	−0.241	1.833 ***
	(0.433)	(0.361)	(0.391)
Industrial structure	−9.988 ***	−6.712 **	−3.276
	(1.837)	(2.186)	(2.266)
Natural disasters	−0.137 **	−0.143 **	0.006
	(0.057)	(0.058)	(0.066)
Rural electricity consumption	−0.142 ***	0.041	−0.183 ***
	(0.030)	(0.025)	(0.021)
GDP per capita	0.459	−0.176	0.636 *
	(0.435)	(0.371)	(0.363)
Government intervention	−0.766	−0.662	−0.104
	(0.523)	(0.553)	(0.644)
Urbanization level	5.203 **	−1.758	6.961 ***
	(2.544)	(1.121)	(1.826)
Gross power per unit of agricultural machinery	−0.046 *	−0.051 **	0.005
	(0.024)	(0.020)	(0.016)
Constant term	2.853	9.070 **	−6.217 *
	(3.579)	(4.023)	(3.346)
Year	Control	Control	Control
Region	Control	Control	Control
Sample size	372	372	372
r2	0.603	0.268	0.573

Note: * denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.001; robust standard errors in parentheses; regressions are clustered using Cluster for provincial areas. The following table is the same.

(1) contains the benchmark regression results. The regression outcomes demonstrated a significant positive correlation among the DVI and AGTFP. This indicates that the establishment of digital villages improves AGTFP and fosters agricultural green growth. Thus, Hypothesis 1 is verified. Among the control variables, industrial structure had a negative impact on AGTFP. This likely occurred because the expansion of the secondary and tertiary sectors drew labor away from agriculture, thereby slowing the growth rate of the agricultural sector. This is because labor resources for agricultural production and secondary and tertiary industries are more dependent on technological advancement and innovation, which do not necessarily benefit agriculture. The effects of GDP per capita and government intervention were not significant. Thus, Hypothesis 1 is validated.

A decomposition of AGTFP into technical progress and technical efficiency, followed by regression, demonstrates that the DVI has a statistically significant positive promotion impact on technical strides in the 1% level. This suggests that inputs from biochemistry, agro-mechanical farming, and other cutting-edge technologies are necessary for the expansion of AGTFP [77]. Additionally, the development of digital villages makes it easier for farms to implement green production technology. Furthermore, the regression results for the digital villages index and technical efficiency were not statistically significant. This was primarily due to two key factors. Firstly, the current digital village construction process tends to prioritize the provision of hardware and equipment, digital formalism, and the irrational allocation of public resources. Secondly, there is an over-reliance on operation service providers [30]. Furthermore, the differences in information investment, consumption, and capacity in digital village construction exacerbate the information divide, which is detrimental to the enhancement of technical efficiency [84]. Consequently, Hypothesis 1.1 is validated, whereas Hypothesis 1.2 is not.

4.4. Endogenous Discussion

To further mitigate the issue of endogeneity resulting from omitted variables and reverse causation, this study employs an IV-GMM model with instrumental variables to validate the robustness of the regression outcomes through generalized moment estimation. The endogenous instrumental variable utilized was the DVI with a one-period lag. The previous period’s value of the DVI not only provided the necessary basis for the development of the digital villages in the current period and ensured the correlation between the instrumental variable and the explanatory variables but also had a negligible effect on the current period’s AGTFP. This satisfied the exogeneity condition for the instrumental variable. Following the approach of Huang et al. (2019) [85], the exogenous instrumental variable selected was the number of fixed-line telephones per 100 people in 2004, as the early development of the Internet relied heavily on telephone dialing technology.

Consequently, the initial coverage of the Internet was higher in areas with a higher fixed-line telephone penetration. This phenomenon exhibited a high correlation with the DVI, thus satisfying the correlation condition for instrumental variables. Nevertheless, the number of fixed-line telephones had a negligible impact on AGTFP, thereby satisfying the exogeneity condition. Since the fixed-line telephones per 100 people in 2004 is a fixed value, the method proposed by Ji and Yang (2020) [86] was used to create an interaction term between this value and the year dummy variable (2011–2022) to assemble a time-series exogenous instrumental variable. According to the preliminary regression results, the number of fixed-line phones in 2004 and the lagged term of digital villages both significantly positively affected the DVI. Moreover, the instrumental variables effectively captured the development level of the digital villages. After addressing endogeneity through the introduction of these instrumental variables, the subsequent regression results confirmed that the influence of the DVI on AGTFP remained statistically significant and positive. Additionally, the regression results exhibited robustness. Therefore, the weak instrumental variable test was accepted, as was the unidentifiable test. The Hansen J test statistic was 1.054 with a p-value of 0.305, indicating that the instrumental variable exogeneity test was also passed. Details are shown in

Table 7. Instrumental variables regression.

Variables	Phase I: Digital Villages Index	Phase II: AGTFP
Digital villages index		1.493 ** (0.490)
Digital villages index lag term	0.939 *** (0.129)
Number of fixed-line telephones per 100 people in 2004 × year	0.0002 * (0.0001)
Industrial structure	0.110	−10.414 ***
	(0.120)	(1.898)
Natural disasters	0.002	−0.117 **
	(0.002)	(0.059)
Rural electricity consumption	0.007 ***	−0.153 ***
	(0.001)	(0.026)
GDP per capita	0.028 *	0.521
	(0.015)	(0.413)
Government intervention	0.059	−0.627
	(0.037)	(0.461)
Urbanization level	0.387	3.963 **
	(0.261)	(1.921)
Gross power per unit of agricultural machinery	−0.001	−0.041 *
	(0.001)	(0.022)
Year	Control	Control
Region	Control	Control
Kleibergen–Paap rk LM test	7.049 **
Kleibergen–Paap rk Wald F test	129.650 ***
Cragg–Donald Wald F test	737.732 ***
Hansen J test	1.054
Sample size	341

Note: * denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.001. The regression reduces the amount of data by 1 year when the digital villages index lag was added, as the 2022 data were not available due to the absence of a lag term.

.

4.5. Robustness Test

In light of the potential for residual bias in the aforementioned regression results, this study proceeds to undertake further tests from the following perspectives. (i) The core explanatory variable, namely the digital villages index (DVI) per 100, was replaced. (ii) The explanatory variable was replaced with AGTFP calculated by the EBM method. (iii) The control variables of agricultural structure and irrigated area were incorporated. (iv) The four municipalities of Beijing, Tianjin, Shanghai, and Chongqing were excluded from the analysis. The regression results for the digital villages index and AGTFP were all significantly positively correlated, thereby demonstrating the robustness of the conclusions. Details are shown in

Table 8. Robustness test results.

Variables	AGTFP	AGTFP-EBM	AGTFP	AGTFP
	(1)	(2)	(3)	(4)
Digital villages index		0.979 **	1.546 **	1.475 ***
		(0.322)	(0.435)	(0.306)
Digital villages index/100	159.173 ***
	(43.312)
Industrial structure	−9.988 ***	−6.525 ***	−10.123 ***	−10.780 ***
	(1.837)	(1.040)	(1.913)	(1.790)
Natural disasters	−0.137 **	−0.072 *	−0.139 **	−0.117 **
	(0.057)	(0.037)	(0.059)	(0.050)
Rural electricity consumption	−0.142 ***	−0.083 **	−0.138 ***	−0.258 ***
	(0.030)	(0.024)	(0.032)	(0.063)
GDP per capita	0.460	0.263	0.522	1.075 **
	(0.435)	(0.281)	(0.429)	(0.318)
Government intervention	−0.766	−0.459	−0.748	−0.488
	(0.523)	(0.348)	(0.522)	(0.439)
Urbanization level	5.203 **	3.122 *	4.824 *	2.275
	(2.544)	(1.597)	(2.544)	(1.637)
Gross power per unit of agricultural machinery	−0.046 *	−0.030 *	−0.042 **	−0.028 *
	(0.024)	(0.017)	(0.019)	(0.015)
Agricultural structure			0.263
			(0.782)
Irrigated area			−0.248
			(0.417)
Constant term	2.853	2.318	2.393	−0.765
	(3.579)	(2.455)	(3.301)	(2.892)
Year	Control	Control	Control	Control
Region	Control	Control	Control	Control
Sample size	372	372	372	324
r2	0.603	0.603	0.605	0.712

Note: * denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.001. Two control variables, agricultural structure and irrigated area, were added to the regression in column (3).

.

4.6. Mechanism of Action Test

This study used a two-way fixed effects model to assess Hypotheses 2.1 and 2.2.

Table 9. Mechanism of the digital villages index on green growth in agriculture.

Variables	Green Bonds	Agricultural Insurance
	(1)	(2)
Digital villages index	249.370 **	0.109 *
	(72.087)	(0.062)
Industrial structure	110.732	−0.052
	(281.317)	(0.042)
Natural disasters	−3.030	−0.002
	(6.169)	(0.001)
Rural electricity consumption	−1.747	−0.005 **
	(3.245)	(0.001)
GDP per capita	−13.656	0.001
	(34.063)	(0.012)
Government intervention	24.959	0.023
	(47.264)	(0.025)
Urbanization level	−74.231	−0.215 ***
	(185.037)	(0.056)
Gross power per unit of agricultural machinery	−0.741	−0.001 **
	(3.053)	(0.001)
Agricultural structure	−23.778	−0.097 ***
	(131.596)	(0.025)
Irrigated area	−54.838	0.014
	(35.241)	(0.015)
Constant term	129.082	0.238 *
	(437.245)	(0.129)
Year	Control	Control
Regions	Control	Control
Sample size	217	312
r2	0.516	0.720

Note: * denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.001. Additional controls for agricultural structure and irrigated area factors; the start of the green bond issuance was in 2016, so the observed variable is 217; the agricultural insurance data are affected by the extremes, so it has been subjected to a 10% shrinkage at both ends.

displays the regression results.

In Table 9’s (1) column, the mediating variable is green bonds. The DVI coefficient was significant at 249.37. This indicates that the growth of digital villages facilitates the issuance of green bonds. Building digital infrastructure and creating smart agriculture requires a lot of money in the process of creating digital villages. Consequently, digital village development projects seek a variety of financing options. This is coupled with societal concern for environmental preservation and sustainable development, which raises awareness as well as support for green bonds. Investors are more inclined to invest in environmental protection and sustainable development projects, and digital village development projects align with this requirement. Therefore, with the construction of digital villages, green bonds develop rapidly. Thus, Hypothesis 2.1 is confirmed.

In Table 9’s (2) column, the mediating variable is agricultural insurance. The coefficient was significant at the 10% statistical level. This indicates that the development of digital villages has a facilitating effect on agriculture insurance. The digital villages provide the fundamental conditions that enable the government to promote the popularity of agricultural insurance. Farmers have become aware of the significance of digital agriculture and agricultural insurance for the protection of agricultural production through a variety of sources, including television news, mobile phone videos, government propaganda, and neighborhood exchanges. As a result, they have adopted digital agriculture to enhance production efficiency and purchased agricultural insurance to mitigate the risk of agricultural business. Thus, Hypothesis 2.2 is confirmed.

4.7. Heterogeneity Analysis

4.7.1. Heterogeneity in Natural Conditions and Food Production Layout

The traditional approach to agricultural development is contingent upon the natural conditions that prevail in a given area. The formation of different agricultural production patterns is influenced by a range of factors, including altitude, which gives rise to variations in temperature, precipitation, sunshine duration, soil texture, and other characteristics. Additionally, the construction of communication infrastructures and the utilization of the internet by users also exhibit altitude-related differences [87]. These factors contribute to the heterogeneity observed in the AGTFP across different regions. In the context of modern agricultural development, China has implemented a series of strategies to enhance the comprehensive output capacity of food. These include the aggregation of resource factor inputs, the establishment of 13 [88] grain-producing regions, the development of facility-based agriculture, and the investment in green production technologies. The current status of these initiatives is as follows: The proportion of total grain production accounted for by grain-producing regions is more than 78% [89]. Consequently, there may be a discrepancy between AGTFP in grain and non-grain-producing regions. Accordingly, this study categorizes the regions according to altitude [90] and grain production layout. (The average digital villages index for major food-producing and non-food-producing regions were 0.140 and 0.127, respectively.) The altitudes are classified as low, medium, and high, while the grain production layout is divided into grain and non-grain-producing regions. The objective is to evaluate how the digital villages affect AGTFP in various environmental settings and grain production pattern areas. The specific results are presented in

Table 10. Heterogeneity test of the digital villages index for AGTFP: natural conditions and food production layout.

Variables	Explained Variable: AGTFP
	Low Altitude	Medium Altitude	High Altitude	Grain-Producing Regions	Non-Grain-Producing Regions
	(1)	(2)	(3)	(5)	(6)
Digital villages index	0.823	1.713 ***	0.050	0.223	1.926 **
	(0.722)	(0.339)	(1.325)	(0.492)	(0.544)
Industrial structure	−7.554 **	−17.078 ***	−15.117 ***	0.223	1.926 **
	(2.199)	(2.507)	(2.702)	(0.492)	(0.544)
Natural disasters	−0.147	−0.051	0.056	−11.815 ***	−7.686 *
	(0.085)	(0.067)	(0.113)	(1.438)	(3.785)
Rural electricity consumption	−0.097 **	−0.552 **	−0.229 *	−0.145	−0.087
	(0.028)	(0.120)	(0.114)	(0.138)	(0.067)
GDP per capita	−0.976	1.762 **	0.946 *	−0.286 **	−0.135 ***
	(0.639)	(0.583)	(0.474)	(0.074)	(0.020)
Government intervention	−2.967 **	1.058	−0.766	0.778	0.489
	(0.766)	(1.742)	(0.453)	(0.470)	(0.490)
Urbanization level	7.407 *	7.601 ***	3.050	−1.277	−0.680
	(3.947)	(1.565)	(3.118)	(1.460)	(0.575)
Gross power per unit of agricultural machinery	−0.076 *	0.020	−0.018	5.202 *	6.060 *
	(0.036)	(0.020)	(0.017)	(2.483)	(3.076)
Constant term	14.545 **	−4.153	4.216	−0.0456 **	−0.051
	(5.047)	(5.625)	(5.802)	(0.019)	(0.035)
Regions	Control	Control	Control	Control	Control
Year	Control	Control	Control	Control	Control
Sample size	132	120	120	156	216
r2	0.653	0.806	0.754	0.689	0.611

Note: * denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.001.

. The results of the regressions in columns (1) to (3) were grouped according to different altitudes, and the digital villages index in the middle-altitude regions had a positive effect on AGTFP, with a coefficient of 1.713, which was significant at the 1% level of statistics; meanwhile, the positive effect of the low- and high-altitude regions was not significant, although it was positive. The digital villages index was the highest in the low-altitude regions; digital agriculture had developed earlier, and the space for further improvement was limited, while the construction of digital infrastructure in the high-altitude region was costly and difficult, so digital villages were not effective in improving AGTFP at low and high altitudes.

Columns (5) and (6) present the regression results, grouped according to different grain production layouts. In the non-grain production regions, the digital villages index had a positive effect on AGTFP, with a coefficient of 1.926, which was significant at the 5% statistical level. However, the regression results in the grain production regions were not statistically significant. The primary reasons for this are as follows: firstly, the expansion of production in grain-producing regions was occurring at the expense of the ecological environment [91], and it still relied on factor inputs to maintain grain production [92], which was not conducive to AGTFP. Secondly, the non-grain-producing regions were situated primarily in the western region, which lacked the advantages offered by the natural environment and geographical location of the eastern coast. With the advent of digital villages, the government has allocated greater resources to infrastructure, intelligent agricultural equipment, and plant protection technology, which can be applied in the western region. Precision agriculture is gradually being adopted by small farmers, which significantly advances AGTFP.

4.7.2. Heterogeneity in Digital Literacy Among Rural Labor Force

The ongoing advancement of digital village construction in China may lead to an evolution in the digital divide, shifting from Internet access to usage disparities [93]. The digital literacy of the rural labor force (calculated using the average years of education of the labor force in rural areas, average years of education = ∑literacy level × corresponding years of education × the proportion of the labor force under this literacy level, of which 0 years of no schooling, 6 years of primary schooling, 9 years of middle schooling, 12 years of high schooling, and 16 years of college education and above; the average years of education is 7.923) has emerged as a pivotal factor in the sustainable growth of agriculture. In this study, we adopted a classification of the digital literacy of the rural labor force based on the level of education, dividing the rural labor force into two groups: those with above-average and below-average digital literacy. The regression of above-average group results indicated that the coefficient of the digital villages index was 1.315, which was statistically significant at the 10% level. However, the regression results for the below-average group were not statistically significant. It can therefore be concluded that those in the rural labor force with a higher level of education will demonstrate a higher level of digital literacy, superior information reception and processing capabilities, and an enhanced ability to master green production technologies, which leads to an improvement in AGTFP. Details are shown in

Table 11. Heterogeneity test of the digital villages index on AGTFP: digital literacy of the rural labor force.

Variables	Explained Variable: AGTFP
	Above-Average	Below-Average
	(1)	(2)
Digital villages index	1.315 *	1.845
	(0.322)	(0.836)
Industrial structure	−8.938 **	−15.162 **
	(1.556)	(1.682)
Natural disasters	−0.162	−0.050
	(0.063)	(0.023)
Rural electricity consumption	−0.135 **	−0.3570
	(0.030)	(0.151)
GDP per capita	−0.168	1.507 **
	(0.458)	(0.132)
Government intervention	−2.206	−0.837 **
	(0.800)	(0.178)
Urbanization level	5.454 ***	2.204
	(0.120)	(1.834)
Gross power per unit of agricultural machinery	−0.045	−0.033 *
	(0.0167)	(0.010)
Constant term	8.417	−0.475
	(3.751)	(1.649)
Regions	Control	Control
Year	Control	Control
Sample size	193	179
r2	0.564	0.758

Note: * denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.001. Regressions were clustered using Cluster for east, central, and west regions.

.

4.7.3. Heterogeneity in the Utilization of Materials for Agricultural Production

Pesticides, fertilizers, and agricultural films are all significant means of production and boosters of agricultural production. However, their excessive use can result in rural environmental pollution, compromise food quality and safety, and hinder the pursuit of green agricultural growth. This study presents a methodology for constructing production material using indicators from the three dimensions of pesticides, fertilizers, and agricultural films. It then examines the differences in the extent to which the digital villages index affects AGTFP under different levels of use. In line with the approach outlined by Tang et al. (2022) [74], an interaction term between the digital villages index and “High” has been constructed and regressed, with the results presented in

Table 12. Heterogeneity test of the digital villages index on AGTFP: use of agricultural production materials.

Variables	Explained Variable: AGTFP
	Pesticides	Agricultural Film	Fertilizer
	(1)	(2)	(3)
Digital villages index * High	0.229	−1.028 **	−0.471
	(0.541)	(0.404)	(0.626)
Digital villages index	1.463 **	1.822 ***	1.803 ***
	(0.538)	(0.393)	(0.479)
High	−0.077	−0.003	−0.321 ***
	(0.110)	(0.091)	(0.061)
Industrial structure	−0.141 **	−0.120 **	−0.124 **
	(0.059)	(0.057)	(0.059)
Natural disasters	−9.898 ***	−9.355 ***	−9.686 ***
	(1.939)	(1.730)	(1.817)
Rural electricity consumption	−0.139 ***	−0.154 ***	−0.144 ***
	(0.030)	(0.033)	(0.031)
GDP per capita	0.495	0.372	0.401
	(0.399)	(0.437)	(0.409)
Government intervention	−0.808	−0.827	−0.773
	(0.522)	(0.545)	(0.534)
Urbanization level	4.988 **	5.994 **	5.496 **
	(2.389)	(2.626)	(2.496)
Gross power per unit of agricultural machinery	−0.047 *	−0.043	−0.043 *
	(0.025)	(0.025)	(0.025)
Constant term	2.555	2.859	3.191
	(3.335)	(3.579)	(3.394)
Year	Control	Control	Control
Regions	Control	Control	Control
Sample size	372	372	372
r2	0.613	0.624	0.619

Note: * denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.001.

. (The assignment of values is based on whether the indicator is greater than the sample mean for that year. A value of 1 is assigned if the indicator is higher than the mean, while a value of 0 is assigned in the event that the indicator is lower than the mean.) With a coefficient of −1.767, the interaction term’s result in column (2) was significant. This suggests that in regions where agricultural films are used more often, the development of digital villages has a greater impact on the improvement of AGTFP. The impact of digital villages construction was not statistically significant in regions exhibiting disparities in pesticide and fertilizer utilization. This conclusion is in line with non-significant regression findings shown in large grain-producing regions, where considerable levels of fertilizer and pesticide are still applied to grain fields to guarantee a steady supply of grain [94,95].

5. Conclusions and Policy Implications

This study uses Chinese province panel data between 2011 and 2022 to create a digital villages index using the entropy value approach. The super-efficient global SBM mixed function model and the Malmquist productivity index were used to estimate AGTFP, which was used to evaluate agricultural green growth. To examine the influence of digital village development on agricultural green growth, as well as the mechanisms underlying this relationship, the bidirectional fixed effect model, mediated effect model, and IV-GMM model were applied.

The key findings are as follows. Firstly, the establishment of digital villages facilitates the progress of sustainable agricultural expansion and green growth. After controlling for other influencing factors, the growth of the digital villages index is associated with an increase in AGTFP. These findings are corroborated by a range of robustness testing methods. Secondly, the development of digital villages facilitates the utilization of financial resources, enabling the growth of green finance and agricultural insurance, which in turn drives agricultural green growth. Thirdly, the influence of digital village development on agricultural green growth is particularly significant in middle-altitude regions, non-grain-producing regions, areas with above-average digital literacy among the rural workforce, and a higher-than-average use of agricultural films.

In light of the aforementioned findings, three key policy insights can be derived. The first aspect pertains to accelerating the rate of development of digital villages. To achieve sustainable rural development, it is essential to facilitate the implementation of environmentally friendly agricultural practices, including green planting and green production technologies. This will not only promote green growth in agriculture but also enhance farmers’ productivity and income. The construction of a green and intelligent village will facilitate improvements in green agricultural production and the protection of rural ecological environments. It is essential to address the shortcomings of a construction-centric approach and a platform-centric approach, which have been prevalent in the digital village construction process.

The second aspect is the promotion of the utilization of rural financial resources to guide the high-quality development of green finance and agricultural insurance. Firstly, it is necessary to reinforce the financial support for the development of green agriculture, increase the credit support for the pilot zones engaged in green agricultural development, and encourage investors to invest in green funds and green bonds. Furthermore, developing organic farms and other green projects must be accelerated to achieve tangible green agricultural growth. Secondly, strengthening green protection in critical areas of agriculture is to be completed by the enrichment of the categories of agricultural insurance for small farmers. These categories include but are not limited to income insurance for unique agricultural products, index insurance, regional yield insurance, and comprehensive insurance for agricultural machinery. The strengthening of the function of insurance protection and the enhancement of the ability of farmers to withstand market price risks of agricultural products are to be achieved by this measure. Thirdly, the rural digital inclusive financial service system must be improved, the radius of financial services expanded to remote rural and mountainous areas, and cooperation between financial services and express logistics, e-commerce sales, and public service platforms promoted. This will form a “four-flow” rural village in which capital flow, logistics, business flow, and information flow are integrated.

The third aspect involves tailoring the building of digital villages to the unique conditions of each local area. The initial phase focuses on developing digital villages in alignment with the region’s specific geographical and climatic characteristics. It is recommended that infrastructure investment in high-altitude areas is upgraded and that regional resource advantages be leveraged to cultivate specialized digital agriculture and digital tourism industries. Enterprises should be encouraged to invest in the digital industry and establish a presence in high-altitude areas, to improve the industrial chain. Enhancing farmers’ digital literacy may be achieved through the implementation of the “government driven, societal involvement and collaborative construction” strategy. It is essential to cultivate farmers’ digital awareness and ability, reduce the threshold for using digital technology, and encourage farmers to develop smart agriculture through the use of information technology. Furthermore, it is vital to cultivate agricultural digital talents and encourage college students from agriculture-related universities to provide farmers with guidance on digital technology. Additionally, there has to be a set of standards for highly digitally literate farmers, and farmers will be encouraged to take part in the development of digital villages. Secondly, the advancement of agricultural production techniques through intelligent agricultural production will be encouraged in main food production regions with high inputs of production materials. Utilizing arable land and causing surface pollution will be monitored, and intelligent irrigation soil testing and formulation technologies will be implemented to reduce the use of pesticides and chemical fertilizers. The Internet platform will be used to smooth production, supply, and marketing channels, stabilize the prices of factors of production, improve the yield of agricultural production and the quality of agricultural products, and safeguard the quality and safety of agricultural products. Ultimately, this will promote green growth in agriculture.

Digital village development and green agricultural growth jointly promote the sustainable development of rural economies, societies, and environments. Digital technologies enhance the efficiency of resources such as labor, land, and water, reduce environmental impacts, and improve the total factor productivity of green agriculture. Information technology fosters social equity by enabling financial resources to be utilized in rural areas, promoting green agricultural growth and aiding in achieving the United Nations’ Sustainable Development Goals (SDGs). The development of digital villages is crucial for current agricultural and rural development and future sustainability.

Despite confirming the facilitative role of digital villages in green agricultural growth and examining the mediating effect of financial resource utilization, this study has certain limitations that require further exploration in future research. For example, it examines how the development of digital villages affects the expansion of green agriculture at the provincial level but does not address municipal or micro-level perspectives of farmers. Will digital village development exacerbate the digital divide and widen income disparities among farmers? These issues need ongoing investigation. Secondly, the transmission mechanisms of digital village development on green agricultural growth are complex. Financial resource utilization promotes factor mobility; the role of these factors in economic growth is significant, and this study does not provide a detailed analysis. Future research plans to explore these dynamics from the perspective of factor mobility. Lastly, given China’s large population, complex terrain, and significant regional differences in resource endowments, the sustainability of green agricultural growth cannot be generalized, necessitating a careful consideration of these factors.