Pathways Through Which Digital Technology Use Facilitates Farmers’ Adoption of Green Agricultural Technologies: A Comprehensive Study Based on Grounded Theory and Empirical Testing

Abstract

The use of digital technologies can break down information barriers in rural areas, thereby creating crucial conditions for the widespread adoption of green agricultural technologies (GATs) among farmers. To explore the relationship between digital technology use (DTU) and farmers’ adoption of GATs, this study draws on 18 in-depth interviews and 608 survey responses collected from rice farmers in Sichuan Province, China. By adopting a mixed-methods design, it offers a comprehensive examination of the mechanisms through which digital technology use (DTU) promotes the adoption of green agricultural technologies (GATs) among farmers. Grounded theory analysis reveals that the DTU–GATs adoption pathway can be conceptualized within a “condition–process–outcome” framework. Specifically, digital infrastructure, farmers’ capital endowment, and practical needs constitute the foundational conditions, while technology perception and the regional soft environment act as key mediating processes. The ultimate outcomes include improvements in economic performance, social well-being, and ecological sustainability. Empirical evidence confirms that DTU significantly promotes the adoption of GATs, primarily by enhancing farmers’ perceptions of technology and improving the agricultural soft environment at the regional level. Moreover, the effects of DTU display substantial heterogeneity across different types of green technologies and among various farmer groups. These findings highlight the importance of strengthening digital infrastructure in rural areas, enhancing farmers’ digital literacy and capacity, and leveraging digital tools to tailor the dissemination and guidance of GATs. Such efforts are essential to raise farmers’ awareness, foster a supportive soft environment for sustainable agriculture, and ultimately advance the adoption of GATs.

1. Introduction

Green agricultural development is a crucial approach to addressing the increasing tension in global agricultural resources and the environment [1,2]. Currently, the insufficient total supply of agricultural products remains a challenge for many developing countries. To increase agricultural yields, chemical inputs such as pesticides, fertilizers, and plastic films are being excessively used in many developing countries, leading to severe agricultural non-point source pollution [3,4]. As the largest developing country in the world, China’s agricultural development has long been focused on “pursuing increased agricultural product quantity,” resulting in severe misuse of agricultural chemicals [5]. Taking fertilizers as an example, compared to 1996, the amount of fertilizer applied in China in 2015 increased by 21.95 million tons, with a fertilizer application intensity of 446 kg/ha, approximately twice the international safety threshold [6]. In fact, during this period, the overall fertilizer utilization rate in China was less than 30%, with a large amount of unused fertilizer entering water bodies or soils through surface runoff, severely polluting the ecological environment [7,8]. In this context, the Chinese government has gradually recognized the urgency of green agricultural development and introduced a series of policies, such as the “Zero Growth Action Plan for Fertilizer Use by 2020” and the “Zero Growth Action Plan for Pesticide Use by 2020.” These policies aim to guide farmers in adopting GATs through a combination of incentives and penalties (e.g., organic fertilizer subsidy policies, straw burning prohibition policies, etc.). Unfortunately, the implementation of these policies has not yielded ideal results, and the adoption of GATs by Chinese farmers remains relatively low. As of 2022, China’s fertilizer usage intensity remained well above the global average, while its pesticide consumption continued to rank first worldwide [9].

The lack of information is often considered the primary reason for the difficulty in promoting green agricultural production technologies [10,11]. The Household Contract Responsibility System implemented in China has made small-scale farming households the main agricultural operators. These small farmers have long been embedded in traditional rural social networks based on kinship and locality, where agricultural information is primarily concentrated within a small elite group (e.g., village leaders) [5,12,13]. The majority of small farmers, however, are distant from information hubs, making it difficult for them to access timely information on GATs, thus hindering the promotion and wide-spread adoption of such technologies. Therefore, breaking down the barriers to information dissemination in rural areas is key to promoting the widespread adoption of GATs in China and other developing countries.

The rapid advancement of digital technologies has fundamentally reshaped traditional modes of economic development, giving rise to new forms of industry and driving China’s economic expansion. Official statistics indicate that in 2023, the scale of China’s digital economy reached 53.9 trillion yuan, with its growth rate outpacing nominal GDP growth by 2.76 percentage points and contributing 66.45% to overall GDP growth [14]. Despite this remarkable progress, the development of the digital economy remains uneven. It is predominantly concentrated in urban areas, where digital technologies are more extensively applied in manufacturing and service sectors, while rural regions continue to lag far behind [15]. This urban–rural divide is largely attributable to the underdeveloped digital infrastructure in rural areas, where issues such as poor network quality and low internet speeds persist, thereby constraining farmers’ incentives to participate in digital economic activities [16].

In response, the Chinese government has increasingly prioritized the promotion of digital technologies in agriculture and rural development. Policy initiatives such as the Digital Agriculture and Rural Development Plan (2019–2025) and the Digital Countryside Development Strategy Outline have provided important institutional support for advancing rural digitalization. As a result, more than 99% of Chinese villages are now covered by 4G networks, the rural Internet user base has expanded to 300 million, and the informatization rate of agricultural extension services has reached 61.3% [17]. The diffusion of digital technologies in the agricultural sector has accelerated the digital transformation of production, processing, distribution, and management, thereby laying a foundation for meeting increasingly diverse and higher-quality demands for agricultural products [18].

Against this backdrop, some scholars have begun to focus on the impact of DTU on the promotion of GATs in China. They generally believe that DTU expands farmers’ information channels, helping them gain a deeper understanding of the advantages and effectiveness of GATs, thereby enhancing their awareness and willingness to adopt such technologies [19]. At the same time, digital technologies can provide farmers with efficient, cost-effective, and convenient technical services such as online training, remote education, and expert consultations, reducing the difficulty of adopting GATs and facilitating their adoption [20]. However, the aforementioned studies primarily conduct empirical analysis through econometric models, and the measurement of relevant variables is not comprehensive (e.g., measuring DTU based on whether farmers use the Internet), which may affect the reliability of the research findings. In addition, the promotion and application of digital technologies in rural China is relatively recent, and theoretical research combining digital technology with farmer behavior is still in its early stages. Relying solely on quantitative research makes it difficult to deeply analyze the impact mechanisms of DTU on farmers’ adoption of GATs. In fact, quantitative and qualitative research are not mutually exclusive; their combined use can provide a more comprehensive understanding of the research findings [21,22]. Against the backdrop of accelerating agricultural digitalization and growing demand for green agricultural products in China, further investigation into the relationship between DTU and farmers’ adoption of green technologies is of great significance.

Therefore, to systematically and comprehensively explore the pathways and mechanisms through which DTU promotes farmers’ adoption of green technologies, this study applies grounded theory coding techniques to analyze the interview data from 18 rice farmers, extracting and synthesizing the pathways through which DTU facilitates the adoption of GATs by farmers, based on their practical cases. At the same time, this paper uses 608 farmer questionnaire survey data, adopts the Ordered Probit model and Control Function Method (CFM) for empirical testing, and further verifies the results of the case analysis, thereby forming a mutual confirmation between quantitative and qualitative research and ensuring the reliability of the research conclusions of this paper. This study makes three main contributions: First, based on grounded theory, it employs a mixed-methods approach combining in-depth interviews and surveys to analyze the pathways through which DTU promotes the adoption of GATs by farmers from both qualitative and quantitative perspectives. This addresses the gap in the existing literature regarding qualitative analysis and enriches the research on DTU and farmer behavior. Second, this study comprehensively measures farmers’ use of digital technologies based on their application in areas such as agricultural input purchasing, agricultural product sales, technical learning, information access, and social interactions. It also assesses farmers’ adoption of GATs through a composite measure of five key technologies: green cultivation, pest and disease control, fertilization, irrigation, and waste management. Therefore, this study provides a more comprehensive variable measurement, enhancing the reliability of the empirical analysis. Third, this study explores the mechanisms through which DTU influences farmers’ adoption of GATs from the perspectives of technology perception and regional soft environment, providing targeted recommendations for further promoting GATs among Chinese farmers.

2. Literature Review

With the development of China’s socio-economic landscape, the demand for agricultural products among Chinese consumers has shifted from meeting basic needs to a focus on green, healthy, and safe products [5]. The traditional agricultural supply model, which relied on increased pesticide and fertilizer use, has become outdated in response to the upgrading of consumer demand [7]. As the main participants in China’s agricultural production, how to promote farmers’ adoption of GATs and advancing green agricultural development has attracted significant scholarly attention. Currently, scholars primarily focus on researching the factors influencing farmers’ adoption of GATs and their effects [3,23,24,25,26]. Behavioral economics theory suggests that human behavior is characterized by bounded rationality, where, under certain environmental influences, individuals make relatively satisfactory behavioral decisions based on their limited cognitive abilities [27]. Therefore, existing research primarily analyzes the factors influencing farmers’ adoption of GATs from two perspectives: farmers’ internal characteristics and the external environment. Relevant studies indicate that individual and household attributes are the fundamental internal factors influencing farmers’ adoption of technologies, affecting their adoption of green technologies to varying degrees. Relevant studies indicate that individual and household attributes are the fundamental internal factors influencing farmers’ technology adoption, with a significant impact on the adoption of GATs [28,29,30,31]. Among these factors, farmers’ perceptions of technology play a crucial role in shaping their adoption of GATs. Generally, the easier farmers perceive the technology to use, the higher the perceived benefits, and the lower the perceived risks, the stronger their willingness to adopt GATs [32,33,34,35]. Some studies categorize the external environmental factors influencing farmers’ adoption of GATs into the regional soft environment and the regional hard environment. Research indicates that regional soft environment factors, such as government actions [36,37], market conditions [38], and social capital [39,40], have a significant impact on farmers’ adoption of GATs. At the same time, regional hard environment factors such as geographical location, climate, terrain, and infrastructure form the foundation of agricultural development and also have a significant impact on farmers’ GATs [41]. Additionally, some studies have found that farmers’ adoption of GATs not only has economic effects but also generates social and ecological impacts [42,43,44].

Digital technology refers to modern technologies that enable the digitalization of production and management through hardware devices such as computers, mobile phones, and sensors, supported by digital software [45,46]. The widespread application of digital technologies has brought extensive, continuous, and profound transformations to traditional economic systems, fundamentally reshaping people’s modes of production and daily life [47,48]. With the ongoing development of the global digital economy, the information dividends generated by digital technologies have gradually diffused into the agricultural and rural sectors. Consequently, an increasing number of scholars have begun to examine the effects of DTU on farmers’ production behaviors [49,50]. Among these behaviors, farmers’ adoption of GATs is directly linked to the sustainability of agricultural production, and the relationship between DTU and such adoption has received growing academic attention [51,52]. Existing research indicates that digital technologies can effectively reduce the costs, risks, and complexity associated with adopting GATs [53,54,55,56] while increasing farmers’ expected returns from GATs adoption [57], thereby motivating farmers to adopt GATs. Specifically, digital technologies can provide farmers with more transparent, convenient, and low-cost information and technical support, which not only significantly lowers the barriers to adopting GATs but also enhances farmers’ awareness of these technologies, thereby promoting their adoption [19,58,59,60]. Although some research has attempted to explore new pathways for promoting farmers’ adoption of GATs from the perspective of DTU, most of these studies rely primarily on micro-level data for empirical testing and lack in-depth analysis based on typical case studies. Research combining both qualitative and quantitative approaches is even rarer.

Overall, current academic research on farmers’ adoption of GATs is relatively extensive, with scholars exploring various pathways to promote this adoption from different perspectives. Some researchers have also attempted to analyze farmers’ adoption of GATs is relatively extensive, with scholars exploring various pathways to promote this adoption from the perspective of DTU, but these studies are still in the early stages, and the related theories and analytical methods need further development and refinement. To explore in greater depth the pathways through which DTU facilitates farmers’ adoption of GATs, this study combines qualitative and quantitative analysis methods, employing a grounded theory framework. It derives a theoretical model of the relationship between DTU and farmers’ adoption of GATs from real-world case studies, and empirically tests this model using econometric analysis. This approach combines case study analysis with empirical testing to provide a detailed examination of the important role digital technology plays in the adoption process, offering valuable insights for further promoting the adoption of GATs among Chinese farmers.

3. Exploratory Research Based on Grounded Theory

3.1. Research Design

3.1.1. Research Methods

Grounded theory, proposed by Glaser and Strauss [61], is a qualitative research method that generates abstract theories from empirical data and real-world observations. Subsequently, through the contributions of Strauss [62] and Charmaz [63], grounded theory evolved into three major schools: classic grounded theory, Straussian (or procedural) grounded theory, and constructivist grounded theory. Among them, Straussian grounded theory places greater emphasis on human agency, arguing that data contain implicit causal relationships that require researchers’ informed interpretation and systematic organization. Its analytical process consists of three stages: open coding, axial coding, and selective coding [64]. Therefore, this study follows the data analysis procedures of Straussian grounded theory—open coding, axial coding, and selective coding—to systematically analyze the interview data. Through this step-by-step process, labels, concepts, and categories are extracted, causal relationships among variables at different levels are identified, and an abstract theoretical framework is ultimately constructed.

3.1.2. Data Collection

Compared to quantitative research, which relies on large-scale random sampling, case studies place greater emphasis on purposive sampling—that is, selecting cases that are most likely to provide rich and relevant information for addressing the research question [65]. The selection of case farmers in this study was closely aligned with the research objectives, while also considering their use of digital technologies and green production technologies. The interview materials were collected through field interviews conducted by the research team in major rice-producing areas of Sichuan Province, China, from July to September 2022. Prior to the interviews, an interview outline was developed based on the research needs, and the interview team received training to ensure familiarity with the questions and proficiency in interview techniques, thereby enhancing the quality of the interviews. During the fieldwork, 20 farmers who met the selection criteria were chosen for one-on-one in-depth interviews, each lasting approximately 30 min and recorded with consent. A total of 18 valid interview transcripts were obtained. After transcription, 15 interviews were randomly selected for coding and model construction, while the remaining 3 were used for theoretical saturation testing.

3.1.3. Research Process

To enhance the robustness of the research, extensive efforts were made to collect and supplement the data. Following the coding procedures of Straussian grounded theory, the data were analyzed step by step, with continuous refinement and abstraction of concepts and their relationships, ultimately leading to the construction of a substantive theory. The specific coding strategy is illustrated in Figure 1.

To minimize analytical bias caused by individual subjectivity, the case coding in this study was conducted collaboratively by three researchers experienced in grounded theory coding. Prior to the coding process, all team members received standardized technical training. After a thorough discussion to ensure the operability of the coding procedure, each member independently coded five cases. Final labels were determined through group discussion based on the individual coding results. During the processes of labeling, conceptualization, categorization, and theorization, the coding team conducted follow-up phone interviews with respondents as needed to gather additional information and continuously refine the emerging labels, concepts, categories, and relationships. Finally, the three reserved cases were re-coded and analyzed to test the theoretical saturation of the coding process.

3.2. Data Coding

3.2.1. Open Coding

This study adheres to the fundamental principle of remaining maximally open to all potential theoretical insights and conducts open coding around the core theme of “the pathways through which DTU facilitates farmers’ adoption of green technologies.” First, the coding team extracted original statements closely related to the research theme from the interview transcripts and conducted initial conceptualization, identifying 70 substantively meaningful initial concepts (prefixed with “A”). These concepts were then grouped and abstracted into categories based on similarity in meaning or attributes, resulting in 30 initial categories. The correspondence between concepts and categories is presented in

Table 1. Open coding categorization.

Initial Categories	Initial Concepts
Social Capital Endowment	A01 Cadre Experience A02 Political Identity
Digitalization of Information Sharing	A03 Online Message Transmission A30 Expansion of Online Information Channels A37 Online information search A38 Online Information Acquisition
Human Capital Endowment	A04 Age A09 Familiarity with GATs A56 Education level
Digitalization of Technology Sharing	A05 Online Technical Training A27 Online Technical Learning A31 Online Technical Consultation
Digitization of Knowledge Sharing	A06 Online Video-Based Learning A54 Online Experience-Based Learning
Digitalization of Agricultural Input Sharing	A07 Online Land Leasing
Digitalization of Product Sharing	A36 Online Sales of Green Agricultural Products
Economic Benefits	A08 Reduce Land Leasing Costs A21 Improve Agricultural Product Quality A24 Reduce Agricultural Input Costs A39 Reduce Labor Costs A41 Stable and Increased Agricultural Production A47 Increase in Agricultural Product Prices
Digitalization of Agricultural Machinery Sharing	A10 Online Rental of Agricultural Machinery
Green Pest Management Technology	A11 Physical Pest Control Technology A62 Biological Pest Control Technology
Green Fertilization Technology	A13 Application of Organic (Farmyard) Manure A67 Application of Commercial Organic Fertilizers
Improvement of the policy environment	A17 Online Policy Promotion A28 Online Access to Policy Information
Green Waste Treatment Technologies	A15 Straw Crushing and Return to Field A18 Agricultural film recycling and processing
Perception of Economic Benefits	A12 Reduction in pesticide use A16 Reduction in fertilizer use A23 Green agricultural products command higher prices A64Increase in agricultural output
Perception of Environmental Benefits	A14 Organic fertilizers cause less environmental pollution A19 Chemical pesticides and fertilizers pollute the environment A48 GATs are beneficial for soil protection
Environmental benefits	A25 GATs are beneficial for environmental protection A53 GATs are beneficial for ecological conservation
Social benefits	A55 Beneficial for promoting green production among other farmers A68 Beneficial for meeting consumer demand for green agricultural products
Demand for improving agricultural product quality	A20 The quality of agricultural products in the market remains low A46 Family members and friends have a demand for green agricultural products
Demand for product sales	A22 Green agricultural products are of high quality but are undervalued in the market A34 Green agricultural products suffer from insufficient consumer trust A35 Green consumption remains weak in the local market
Improvement of the market environment	A26 Online promotion of agricultural products A29 Engage with consumers online
Demand for cost reduction and income enhancement	A32 The expectation of income improvement among farmers is relatively high A40 High labor costs
Improvement of the social environment	A33 Online communication among farmers A43 Learning about peers’ farming models via digital channels
Demand for safety and health	A44 Farmers place a high priority on food safety A50 Farmers care about the village environment
Perception of social benefits	A51 Green production contributes to human health A52 Green production contributes to ensuring food security
digital devices	A57 Feature phones used by elderly users often lack Internet access A60 smartphones and computers
Perceived ease of use	A58 Green production technologies are easy to use
Digital networks	A59 Rural areas have an extensive network coverage
Perception of technological risks	A61 The adoption of GATs entails a risk of failure
Perception of market risk	A66 Green agricultural products may involve sales risks
Green cultivation technologies	A69 Cultivating during appropriate seasons helps reduce pest infestations A70 Deep tillage with machinery

.

3.2.2. Axial Coding

Axial coding involves reorganizing and integrating the preliminary categories generated during open coding to identify interconnections among them and develop higher-level core categories [66]. To further explore the relationships among the initial categories, this study reorganized and clustered 30 initial categories, resulting in 8 higher-level axial categories. The relationships among the axial categories can be interpreted using the “conditions–process–outcome” model. The detailed results of axial coding are presented in

Table 2. The main category of axial coding forms.

Correspondence	Main Categories	Initial Categories
Conditions	Digital Infrastructure	Digital networks; digital devices
	Capital Endowment	Social Capital Endowment; Human Capital Endowment
	Practical Needs	Demand for improving agricultural product quality; Demand for product sales; Demand for cost reduction and income enhancement; Demand for safety and health
Process	Digital Technology Use	Digitalization of Information Sharing; Digitalization of Technology Sharing; Digitization of Knowledge Sharing; Digitalization of Agricultural Input Sharing; Digitalization of Agricultural Machinery Sharing; Digitalization of Product Sharing
	Regional Soft Environment	Improvement of the policy environment; Improvement of the market environment; Improvement of the social environment
	perception of technology	Perception of Economic Benefits; Perception of Environmental Benefits; Perception of social benefits; Perception of technological risks; Perception of market risk; Perceived ease of use
	Adoption of GATs	Green Pest Management Technology; Green Fertilization Technology; Green Waste Treatment Technologies; Green cultivation technologies
Outcome	Benefits of Adopting GATs	Economic Benefits; Environmental benefits; Social benefits

.

3.2.3. Selective Coding

Selective coding abstracts the core category from all identified categories and links it with the main categories to form a coherent “storyline,” thereby constructing a theoretical model that reflects the full narrative. Based on the analysis and synthesis of original case data, key concepts, and hierarchical categories, this study develops a theoretical model illustrating how the use of digital technologies facilitates farmers’ adoption of GATs, as shown in Figure 2.

3.2.4. Theoretical Saturation Test

To further verify theoretical saturation and ensure that no new concepts emerge, this study conducted additional coding analysis using three reserved interview transcripts. The results indicate that no new concepts were identified in the remaining data. Therefore, the categories and relationships presented in the theoretical model (Figure 2) are considered theoretically saturated and constitute a well-grounded framework based on empirical data and contextual analysis.

4. Empirical Examination Based on Questionnaire Survey

4.1. Research Hypotheses

The theoretical framework of this study is illustrated in Figure 3. The diffusion of digital technologies in rural China has profoundly transformed farmers’ cognitive and behavioral patterns, thereby providing new momentum for advancing the adoption of GATs. First, DTU can reduce the costs associated with adopting GATs. By virtualizing these technologies, digital platforms enable farmers to engage in learning, training, and information exchange related to green agriculture at a lower cost, thereby reducing the expenses involved in acquiring such technologies [58]. Moreover, digital trading platforms help farmers strengthen their connections with input suppliers, effectively streamlining the supply chain and lowering the cost of purchasing green pro-duction inputs [67]. Second, DTU can effectively reduce the risks associated with adopting GATs. Existing research indicates that green technologies often require targeted application and involve complex operations; improper use by farmers can lead to yield reductions and technical risks [68]. In addition, the market for green agricultural products is characterized by information asymmetry, which may result in the phenomenon of “high quality but low price,” posing significant market risks [69]. With the information-sharing capabilities of digital technology, farmers can access detailed guidance on the proper use of green technologies anytime and anywhere, thereby minimizing technical risks. Furthermore, digital platforms enable direct interaction with consumers, allowing farmers to align production with green consumer demand, reduce information asymmetry, and effectively mitigate market risks [70]. Finally, digital technologies can reduce the perceived difficulty of adopting GATs. By breaking the spatial and temporal constraints of technology dissemination, digital tools expand farmers’ access to green agricultural knowledge. Through mobile Internet platforms, farmers can search for, learn about, exchange, and consult on GATs anytime and anywhere, significantly lowering the barriers to accessing and using such technologies [54,55]. Therefore, this study proposes the following hypotheses:

H1. 

DTU has a significant positive effect on farmers’ adoption of GATs.

Farmers’ intrinsic motivation to adopt GATs originates from their internal perceptions of such technologies, which primarily includes perceptions of benefits, risks, and ease of use [69,71]. The fundamental logic behind farmers’ production decisions is to maximize benefits and minimize risks. While the pursuit of benefit maximization serves as the core motivation for adopting GATs, the presence of risk introduces uncertainty into the potential gains, making risk minimization a crucial precondition for adoption [72]. In addition, adoption is influenced by the perceived ease of use: the learning and application of GATs require investments of time, money, and effort. If farmers perceive these technologies as difficult to adopt, they are generally less willing to try them [73]. Based on the technical information they possess, farmers make comprehensive evaluations regarding the benefits, risks, and ease of adoption of GATs. These evaluations shape their overall technological cognition, which ultimately influences their decision on whether to adopt such technologies. On the one hand, digital technologies enhance farmers’ ability to access information, enabling them to acquire more comprehensive and systematic knowledge about GATs [74]. This deepens their cognition regarding the benefits, risks, and ease of adoption of such technologies, and helps them develop rational strategies for adopting green practices. On the other hand, by breaking down information barriers, improving the efficiency of information dissemination, and reducing the cost of information acquisition, digital technologies help lower farmers’ perceived risks, while enhancing their expected benefits and perceived ease of use, thereby encouraging the proactive adoption of GATs [20]. Therefore, this study proposes the following hypotheses:

H2. 

Technological cognition mediates the relationship between DTU and farmers’ adoption of GATs.

Behavioral economics posits that human rationality is bounded, resulting from the joint influence of internal characteristics and external environmental factors. Under such bounded rationality, individuals tend to adopt satisficing strategies rather than pursuing absolute optimality [75]. In this context, farmers’ adoption of GATs is influenced not only by internal factors such as cognitive perceptions and personal abilities, but also by external conditions, including regional hard environments—such as geography, climate, and infrastructure—and soft environments, such as policies, markets, and social norms [76]. Among these, the hard environment serves as the foundational constraint for agricultural development, typically stable and difficult to change in the short term. In contrast, the soft environment is more flexible and diverse, playing a guiding role in shaping farmers’ production behaviors and technological choices [77]. Therefore, elements of the soft environment—such as regional green development policies, green technology extension efforts, and the dissemination of green concepts—can significantly influence farmers’ decisions regarding the adoption of GATs. Rural China is characterized by a social network rooted in kinship and geographic proximity, within which information dissemination follows a hierarchical and segmented pattern [12]. Information related to green agricultural development—such as policy measures, promotional activities, and advocacy of green concepts—is typically first accessed by a small group of rural elites and then gradually diffused to the broader farming population. However, farmers located on the periphery of information networks often experience restricted access to digital channels and tend to have relatively low levels of information literacy. As a result, they may lack comprehensive and detailed knowledge of local green development initiatives. Consequently, the influence of the regional soft environment on these farmers’ adoption of GATs tends to be limited. The advent of modern digital technologies has fundamentally transformed the modes of information dissemination in rural areas. These technologies have enabled the decentralization of rural information flows, breaking through the traditional spatial and temporal constraints of information transmission. As a result, information channels have been significantly expanded and the efficiency of communication has improved. This transformation facilitates farmers’ access to timely, diversified, and comprehensive in-formation regarding the social, policy, and market aspects of green agricultural development, thereby increasing the likelihood that they will adopt GATs. Therefore, this study proposes the following hypotheses:

H3. 

The regional soft environment mediates the relationship between DTU and farmers’ adoption of GATs.

4.2. Data Source

The research team conducted field interviews with rice farmers in Sichuan Province, China, from July to September 2022. Sichuan Province is one of China’s major rice-producing regions, with rice cultivation primarily concentrated in eight cities located in the central, northeastern, and southeastern parts of the province. This study first employed a random sampling method to select five sample cities: Chengdu, Mianyang, Guang’an, Luzhou, and Yibin. Following the administrative hierarchy and using stratified random sampling, one to three sample counties were randomly selected within each city, followed by one to three townships within each county, and then one to three villages within each township. Finally, 15 to 25 rice-farming households were randomly selected in each village to complete the questionnaire survey. After removing invalid responses, a total of 608 valid questionnaires were obtained. The sample distribution is illustrated in Figure 4.

4.3. Model Specification and Variable Selection

4.3.1. Model Specification

The degree of adoption of GATs by farmers is a ranking variable and can be estimated using the Ordered Probit model. Therefore, this paper constructs the following model:

$${GTA}_i^{\mathrm{*}}={\beta }_0+{\beta }_1{DTU}_i+{\beta }_{2}Control+{\epsilon }_i$$

(1)

In Equation (1), ${GTA}_i^{*}$ represents the extent to which farmer i adopts GATs; ${DTU}_i$ represents the digital technology usage of farmer i; $Control$ refers to the control variables; ${\beta }_0$ denotes the constant term; ${\epsilon }_i$ denotes the random error term.

${GTA}_i^{*}$ is a latent variable. Assume that $r_0$, $r_1,r_2$, $r_3$, $r_4$ are unknown segmentation points of farmers’ GATs adoption behavior, and $r_0<r_1<r_2<r_3<r_4$. Equation (2) can be defined as follows:

$${GTA}_i=\left\{\begin{array}{c}0,{GTA}_i^{*}\le r_0\\ 1,r_0<{GTA}_i^{*}\le r_1\\ 2,r_1<{GTA}_i^{*}\le r_2\\ 3,r_2<{GTA}_i^{*}\le r_3\\ 4,r_3<{GTA}_i^{*}\le r_4\\ 5,{GTA}_i^{*}>r_4\end{array}\right.$$

(2)

Accordingly, the conditional probabilities of farmers adopting different numbers of GATs can be expressed as:

$$P({GTA}_i=0)=\mathsf{\Phi }(r_0-{\beta }_1{DTU}_i-{\beta }_{2}Control)$$

(3)

$$P\left({GTA}_i=1\right)=\mathsf{\Phi }\left(r_1-{\beta }_1{DTU}_i-{\beta }_{2}Control\right)-\mathsf{\Phi }(r_0-{\beta }_1{DTU}_i-{\beta }_{2}Control)$$

$$P\left({GTA}_i=2\right)=\mathsf{\Phi }\left(r_2-{\beta }_1{DTU}_i-{\beta }_{2}Control\right)-\mathsf{\Phi }(r_1-{\beta }_1{DTU}_i-{\beta }_{2}Control)$$

$$P\left({GTA}_i=3\right)=\mathsf{\Phi }\left(r_3-{\beta }_1{DTU}_i-{\beta }_{2}Control\right)-\mathsf{\Phi }(r_2-{\beta }_1{DTU}_i-{\beta }_{2}Control)$$

$$P\left({GTA}_i=4\right)=\mathsf{\Phi }\left(r_4-{\beta }_1{DTU}_i-{\beta }_{2}Control\right)-\mathsf{\Phi }(r_3-{\beta }_1{DTU}_i-{\beta }_{2}Control)$$

$$P\left({GTA}_i=5\right)=1-\mathsf{\Phi }(r_4-{\beta }_1{DTU}_i-{\beta }_{2}Control)$$

In Equation (3), Φ(·) denotes the cumulative distribution function of the standard normal distribution. The parameters of the model are estimated using the method of maximum likelihood.

Meanwhile, the Control Function Method (CFM) is applied in this study to mitigate potential endogeneity issues within the model [78]. This approach consists of two steps. First, the DTU variable is regressed on its instrumental variables and control variables to estimate the generalized residuals. Second, the predicted residuals are included as an additional regressor in the second-stage estimation. If the residual term is statistically significant, it indicates that DTU is endogenous, and incorporating the residual helps correct estimation bias. Conversely, if the residual is not significant, it suggests that DTU is exogenous, and excluding the residual leads to more efficient and unbiased estimation results. Accordingly, the model is constructed as follows:

$${DGT}_i={\delta }_0+{\delta }_1{HCE}_i+{\delta }_{2}Control+{\omega }_i$$

(4)

$${GTA}_i=b_0+b_1{DTU}_i+b_2{\sigma }_i+b_{3}Control+{ϵ}_i$$

(5)

In Equations (4) and (5), ${HCE}_i$ denotes the household Internet communication expenditure of farmer i, and ${\sigma }_i$ represents the generalized residual obtained from the regression in Equation (4). ${\delta }_0$ and $b_0$ are constant terms, ${\delta }_1$, ${\delta }_2$, $b_1$, $b_2,\mathrm{and}{b}_3$ are parameters to be estimated. ${\omega }_i\mathrm{and}{ϵ}_i$ denote random error terms.

4.3.2. Variable Selection

The dependent variable in this study is farmers’ adoption of GATs. Following the approaches of Kuang et al. (2023) and Liu et al. (2024) [3,79], we consider five categories of GATs to comprehensively capture adoption behavior: green cultivation, green pest and disease control, green fertilization, green irrigation, and green waste management. Each category is measured as a binary variable, where a value of 1 indicates adoption and 0 otherwise. The overall adoption level is represented by the cumulative number of technologies adopted, ranging from 0 to 5.

The core explanatory variable is the use of digital technologies. Following the methodology of Zheng et al. (2022) and Kai et al. (2023) [80,81], this study assesses farmers’ digital technology usage across various aspects of production and daily life, including social interaction, information access, agricultural input procurement, and agricultural product sales. During the survey, farmers were asked to rate their level of DTU in these areas using five-point Likert scales. An entropy-weighted average was then applied to construct a composite index measuring the overall extent of DTU.

Technology perception and the regional soft environment are identified as mediating variables in this study. Farmers’ perceptions of GATs are primarily reflected in three dimensions: perceived usefulness, perceived risk, and perceived ease of use. Following the approach of Huang et al. (2022) [82], perceived usefulness is measured by the degree to which farmers recognize the economic benefits (e.g., increasing income, reducing input costs, and saving labor) and social benefits (e.g., improving product quality and promoting social development) of green technologies. Perceived ease of use is assessed by asking farmers about the level of difficulty they experience when adopting GATs. Perceived risk is captured by farmers’ concerns regarding potential technical risks (e.g., yield reduction) and market risks (e.g., failure to achieve premium pricing despite quality improvements). The regional soft environment primarily consists of three dimensions: the social environment, policy environment, and market environment. The social environment is measured by the frequency of communication among farmers, the prevalence of GATs adoption among peers, and the farmer’s sense of community identity. The policy environment is assessed across three aspects: government publicity, incentives, and constraints. The market environment is evaluated based on market recognition and expectations. All responses to the related questions are rated using five-point Likert scales, ranging from 1 (strongly disagree) to 5 (strongly agree). A comprehensive index is then constructed using the entropy weighting method.

Following the approach of Luo et al. (2022) [83], this study uses household expenditure on Internet and communication services as an instrumental variable for DTU. Such expenditure is closely related to the extent of farmers’ digital technology adoption but does not directly influence their decisions regarding GATs’ adoption. Therefore, it serves as a theoretically sound and appropriate instrumental variable.

To ensure reliable estimation, the analysis accounts for individual, household, and environmental characteristics as control variables. Individual characteristics include the respondent’s gender, age, education level, political affiliation, off-farm employment status, and risk preference. Household operational characteristics encompass total household size, number of migrant workers, highest education level among household members, political affiliation of the main household member, and farm size. Environmental characteristics include the distance to the county seat, topographical features, and regional dummy variables. Detailed variable definitions are presented in

Table 3. Variable selection and assignment description.

Variables	Definition and Assignment	Mean	S.D.
Dependent variable
Green cultivation technologies	Whether to adopt green cultivation techniques, No = 0, Yes = 1	0.610	0.488
Green pest control technologies	Whether to adopt green pest control techniques, No = 0, Yes = 1	0.497	0.500
Green fertilization technologies	Whether to adopt green fertilization techniques, No = 0, Yes = 1	0.594	0.492
Green waste utilization technologies	Whether to adopt green waste utilization techniques, No = 0, Yes = 1	0.770	0.421
Green irrigation technologies	Whether to adopt green irrigation techniques, No = 0, Yes = 1	0.015	0.121
Green agricultural technologies (GATs)	Number of green production technologies adopted (items)	2.485	1.413
Core explanatory variable
Digital technology use (DTU)	Entropy-based composite score	0.285	0.300
Mediating variables
Technology perception	Entropy-based composite score	0.455	0.218
Regional soft environment	Entropy-based composite score	0.453	0.205
Control variables
Gender	Gender of farmer, Female = 0, Male = 1	0.877	0.329
Age	Age of farmer (years)	54.984	10.291
Education	Years of Education (years)	8.498	3.191
Village cadre	Whether serving as a village cadre, No = 0, Yes = 1	0.148	0.355
Part-time farming	Whether engaged in part-time farming, No = 0, Yes = 1	0.286	0.452
Risk preference	Risk preference of farmers, Low = 1, Medium = 2, High = 3	2.125	0.832
Total household size	Number of all household members	4.684	1.802
Number of household members working away from home	Calculated in real terms, in number of persons	1.053	1.192
Highest education level in household	Highest education level among household members (years)	11.928	3.461
Number of cadre members	Number of cadre members in the household	0.194	0.420
Planting scale	Logarithmic value of rice planting area	3.264	1.884
Distance to county seat	Measured by actual distance (km)	23.681	12.769
Topography	Plain = 1, Hilly = 2, Mountainous = 3	1.691	0.507
Region	Chengdu Plain Economic Zone = 1, Northeastern Sichuan Economic Zone = 2, Southern Sichuan Economic Zone = 3	2.031	0.903
Instrumental variable
Household communication and Internet expenditure	Logarithm of household communication and Internet expenditure	7.025	0.882

.

4.4. Results

4.4.1. Benchmark Regression Results

As shown in

Table 4. Benchmark regression results.

Variable Name	(1)	(2)	(3)	(4)
	Ordered Probit	Ordered Probit	CFM	CFM
DTU	2.179 ***	2.223 ***	1.864 ***	1.902 ***
	(0.285)	(0.287)	(0.286)	(0.288)
Residual			9.576 ***	9.796 ***
			(1.428)	(1.492)
Gender	−0.063	−0.123	0.152	0.093
	(0.121)	(0.115)	(0.124)	(0.117)
Age	−0.017 **	−0.018 **	0.103 ***	0.105 ***
	(0.007)	(0.008)	(0.019)	(0.019)
Education	0.041 *	0.056 **	−0.186 ***	−0.176 ***
	(0.022)	(0.022)	(0.039)	(0.040)
Village Cadre	0.242	0.217	1.311 ***	1.309 ***
	(0.191)	(0.191)	(0.230)	(0.236)
Part-time farming	−0.171 *	−0.120	−0.429 ***	−0.382 ***
	(0.103)	(0.103)	(0.108)	(0.107)
Risk preference	0.136 *	0.164 **	−0.295 ***	−0.275 ***
	(0.075)	(0.075)	(0.101)	(0.102)
Total household size	0.122 ***	0.100 ***	0.104 ***	0.080 **
	(0.030)	(0.032)	(0.031)	(0.032)
Number of household members	0.056	0.074	0.023	0.041
working away from home	(0.050)	(0.052)	(0.053)	(0.054)
Highest education level in household	0.023	0.028 *	0.067 ***	0.073 ***
	(0.017)	(0.017)	(0.017)	(0.017)
Number of cadre members	0.073	0.111	−0.602 ***	−0.576 ***
	(0.160)	(0.163)	(0.187)	(0.190)
Planting scale	0.086 ***	0.183 ***	0.070 **	0.173 ***
	(0.032)	(0.039)	(0.033)	(0.039)
Distance to county seat	−0.001	−0.000	0.030 ***	0.032 ***
	(0.004)	(0.004)	(0.006)	(0.007)
Topography	0.063	−0.040	0.351 ***	0.251 **
	(0.108)	(0.109)	(0.116)	(0.114)
Region	No	Yes	No	Yes
Pseudo R2	0.2087	0.2196	0.2375	0.2495
Wald test	354.48 ***	336.02 ***	406.43 ***	360.91 ***
Observations	608	608	608	608

Note: *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

, Models (1) and (2) report estimates based on the Ordered Probit model, while Models (3) and (4) address potential endogeneity using the CFM. Regional dummy variables are included in Models (2) and (4). In Models (3) and (4), the residual terms are statistically significant at the 1% level, indicating the presence of endogeneity in DTU. Incorporating the residuals into the regression helps correct estimation bias. The results from Model (2) show that DTU has a significantly positive effect on farmers’ adoption of GATs at the 1% level, suggesting that DTU significantly promotes adoption. This finding is consistent with the conclusions of Zhang et al. (2024) and Musajan et al. (2024) [58,84]. The estimates in Model (4), which correct for endogeneity, remain consistent with those in Model (2), confirming the robustness of the baseline findings and supporting Hypothesis 1.

4.4.2. Analysis of the Mechanism Pathways

The grounded theory analysis indicates that technology perception and the regional soft environment are key pathways through which DTU promotes the adoption of GATs among farmers. To further validate this finding, we quantify both technology perception and the regional soft environment using micro-level data and conduct an empirical analysis using a mediation effect model. The results are presented in

Table 5. Results of the mediation effect test.

Variable Name	(1)	(2)	(3)	(4)	(5)	(6)
	GATs	Technology Perception	GATs	GATs	Regional Soft Environment	GATs
DTU	2.223 ***	0.303 ***	1.245 ***	2.223 ***	0.263 ***	1.409 ***
	(0.287)	(0.030)	(0.317)	(0.287)	(0.031)	(0.309)
Technology perception			3.959 ***
			(0.394)
Regional soft environment						4.014 ***
						(0.388)
Control variable	Yes	Yes	Yes	Yes	Yes	Yes
Region	Yes	Yes	Yes	Yes	Yes	Yes
Constant term		0.297 ***			0.277 ***
		(0.088)			(0.090)
R2		0.5540			0.5217
Pseudo R2	0.2196		0.2955	0.2196		0.2942
F value		60.40 ***			52.88 ***
Wald test	336.02 ***		353.97 ***	336.02 ***		396.30 ***
Observations	608	608	608	608	608	608

Note: *** indicates statistical significance at the 1% level.

.

Models 1 to 3 examine the mediating role of technology perception, while Models 4 to 6 test the mediating effect of the regional soft environment. Model 2 shows that DTU has a significantly positive effect on farmers’ technology perception at the 1% significance level, indicating that DTU can significantly enhance farmers’ awareness and understanding of agricultural technologies. Model 3 reveals that both DTU and technology perception positively influence the adoption of GATs at the 1% level. This suggests that DTU promotes adoption by improving farmers’ perception of technology, providing empirical support for Hypothesis 2. This finding is consistent with the conclusions of Yu et al. (2024) [85].

Similarly, Models 5 and 6 demonstrate that DTU significantly improves the regional soft environment at the 1% level, and that both DTU and the regional soft environment have significantly positive effects on farmers’ adoption of GATs. These findings confirm that DTU facilitates adoption by enhancing the regional soft environment, thereby supporting Hypothesis 3. This finding aligns with the perspective of Musajan et al. (2024) [84], despite notable differences in study subjects, methods, and regions between the two studies.

4.4.3. Further Analysis

To examine how DTU influences farmers’ adoption of different types of GATs, following the approach of Xiong et al. (2025) [86], this study categorizes the GATs involved in rice production into five types: green cultivation, green pest control, green fertilization, green waste utilization, and green irrigation. A series of regression analyses are conducted accordingly. As shown in

Table 6. Effects of DTU on Farmers’ Adoption of Different GATs.

Variable Name	(1)	(2)	(3)	(4)	(5)
	Green Cultivation	Green Pest Control	Green Fertilization	Green Waste Utilization	Green Irrigation
DTU	1.224 ***	2.524 ***	2.519 ***	2.319 ***	−1.564
	(0.372)	(0.339)	(0.379)	(0.499)	(1.189)
Control variable	Yes	Yes	Yes	Yes	Yes
Region	Yes	Yes	Yes	Yes	Yes
Constant term	−2.101 **	−3.241 ***	−1.157	0.439	−5.574 ***
	(0.963)	(0.857)	(0.857)	(1.091)	(1.875)
Pseudo R2	0.3026	0.3205	0.2715	0.3252	0.3721
Wald test	167.42 ***	221.02 ***	159.24 ***	122.35 ***	48.01 ***
Observations	608	608	608	608	608

Note: ** and *** indicate statistical significance at the 5% and 1% levels, respectively.

, DTU is significantly positively associated with farmers’ adoption of green cultivation techniques, green pest and disease management, green fertilization techniques, and green waste utilization, with coefficients of 1.224, 2.524, 2.519, and 2.319, respectively, all significant at the 1% level. This indicates that DTU has a stronger effect on the adoption of green pest and disease management, green fertilization, and green waste utilization techniques. A possible explanation is that China’s agricultural development has long relied heavily on conventional chemical pesticides and fertilizers, resulting in severe agricultural non-point source pollution [5,7,8], while agricultural waste is often burned by farmers, causing serious air pollution [87]. Consequently, the Chinese government has placed particular emphasis on promoting green pest and disease management, green fertilization, and green waste utilization technologies. Digital technology facilitates the rapid and widespread dissemination of relevant policies among farmers, thereby incentivizing the adoption of these green production practices. In contrast, adopting green cultivation techniques requires investment in new farmland machinery, which entails high costs that some smallholder farmers cannot afford, resulting in a relatively lower effect of DTU on this adoption. Interestingly, DTU does not have a significant effect on farmers’ adoption of green irrigation techniques. One possible explanation is that the surveyed area lies in a subtropical monsoon climate zone with abundant rainfall throughout the year. Rice production in this region typically relies on natural precipitation, resulting in relatively low irrigation demand. Additionally, the high investment and maintenance costs of water-saving irrigation facilities may further discourage farmers from adopting green irrigation technologies.

Considering the significant heterogeneity among sampled farmers, this study follows the approach of Guo et al. (2022) [88] and Chen et al. (2025) [89] to conduct subgroup regressions based on farmers’ average age and average education level. Farmers with age less than or equal to the average are classified as the younger group, while those above the average are classified as the older group. Similarly, farmers with education level less than or equal to the average are grouped as the lower-education group, and those above the average as the higher-education group. The regression results are presented in

Table 7. Effects of DTU on GATs Adoption Among Different Farmer Groups.

Variable Name	Age		Education
	(3)	(4)	(7)	(8)
	Younger Farmers	Older Farmers	Lower Education Group	Higher Education Group
DTU	2.947 ***	1.306 **	2.484 ***	2.756 ***
	(0.342)	(0.539)	(0.514)	(0.352)
Control variable	Yes	Yes	Yes	Yes
Region	Yes	Yes	Yes	Yes
Pseudo R2	0.2504	0.1302	0.1627	0.2463
Wald test	153.62 ***	125.96 ***	140.27 ***	175.94 ***
Observations	303	305	248	360

Note: ** and *** indicate statistical significance at the 5% and 1% levels, respectively.

. DTU has a significantly positive effect on the adoption of GATs among younger and more educated farmers, with larger coefficients. This may be because younger and more educated farmers tend to use digital technologies more intensively and are more capable of learning and adopting new agricultural technologies.

5. Conclusions and Policy Implications

The emergence of digital technologies has exerted a subtle yet profound influence on farmers’ production and daily lives. Some scholars even argue that digital technologies may usher in a new wave of industrial revolution in agriculture [90]. Against this backdrop, this study combines qualitative and quantitative methods to investigate the pathways through which DTU facilitates farmers’ adoption of GATs. Based on grounded theory analysis, the study identifies that DTU significantly affects the adoption of GATs among farmers. Specifically, digital infrastructure, farmers’ capital endowment, and practical needs serve as prerequisite conditions; technology perception and the regional soft environment represent two key transmission pathways; and achieving economic, social, and ecological benefits constitutes the ultimate goal and outcome. To further verify the findings from the grounded theory analysis, this study conducts an empirical test using survey data from 608 rice farmers in Sichuan Province, China. Employing an Ordered Probit model and the CFM, the results show that DTU has a significantly positive impact on farmers’ adoption of GATs. This effect is primarily mediated by improvements in farmers’ technology perception and enhancements in the regional agricultural soft environment—findings that are consistent with the grounded theory analysis. Moreover, the study finds considerable heterogeneity in the effects of DTU across different types of GATs and farmer groups.

Based on the above analysis, DTU can effectively promote farmers’ adoption of GATs. Driven by strong government support, the digitalization of agriculture and rural areas in China is advancing rapidly. However, the study area in western China has relatively lower socio-economic development, digital infrastructure, and farmers’ digital literacy. To further facilitate farmers’ adoption of GATs in this region, this study proposes four policy recommendations: (1) Provide digital skills training for farmers to enhance their awareness and application of digital technologies, guiding them to access and learn more about green agricultural production knowledge and techniques, thereby improving their understanding of GATs. (2) Strengthen government leadership and support by increasing financial investment in the research and application of agricultural digital technologies. Support agricultural research institutions in developing technologies in areas such as the Internet of Things, big data, and artificial intelligence, and promote the broad application of advanced technologies in agricultural production, further improving the regional soft environment for farmers’ adoption of green technologies GATs. (3) Establish dedicated platforms for promoting GATs to provide more targeted guidance and publicity, leveraging digital technologies to disseminate practical, easily adoptable green production techniques that meet farmers’ immediate needs. (4) Give priority to digitally marginalized rural groups by offering customized digital support and training. Emerging technologies, such as smartphone applications and agricultural IoT systems, can be targeted at younger farmers, while simplified devices or applications combined with individualized guidance should be provided for older or less-educated farmers.

This study adopts a mixed-methods approach to provide a more comprehensive and in-depth analysis, addressing some of the gaps in existing research. It offers useful insights for future studies on how DTU influences farmer behavior. Nevertheless, the study has certain limitations. First, both the case materials and survey data were collected exclusively from rice farmers in Sichuan Province, without nationwide data or consideration of other typical crops. Given the considerable variation in agricultural production environments across regions and the differing GATs required for various crops, the effect of DTU on farmers’ adoption of green technologies may vary across regions and crop types. Second, the level of agricultural digitalization within the study area is heterogeneous, which may affect the precision of the estimates. For instance, Chengdu and Mianyang exhibit higher overall economic and technological development and more advanced digital infrastructure, resulting in relatively higher agricultural digitalization and a stronger effect of DTU on farmers’ adoption of green technologies. Third, while grounded theory is employed to explore the antecedents and consequences of DTU in promoting farmers’ adoption of GATs, empirical testing is limited to the effect of DTU itself due to constraints in data availability and article length. The causal mechanisms related to the antecedents and consequences identified through qualitative analysis have not yet been empirically verified. This represents a promising direction for future research.