Spatiotemporal Evolution and Driving Factors of Agricultural Digital Transformation in China

Abstract

With the digital economy continuing to integrate deeply into the agricultural sector, agricultural digital transformation has emerged as a pivotal driver of rural revitalization and the development of a robust agricultural economy. Although existing studies have affirmed the positive role of agricultural digital transformation in promoting rural development and enhancing agricultural efficiency, its spatiotemporal evolution patterns, regional disparities, and underlying driving factors have not yet been systematically and thoroughly investigated. This study seeks to fill that gap. Based on provincial panel data from China spanning 2011 to 2023, this study employs the Theil index, kernel density estimation, Moran’s index, and quantile regression to systematically assess the spatiotemporal dynamics and driving factors of agricultural digital transformation at both national and regional levels. The results reveal a steady overall improvement in agricultural digital transformation, yet regional development imbalances remain prominent, with a shift from inter-regional disparities to intra-regional disparities over time. The four major regions exhibit a stratified evolutionary trajectory marked by internal differentiation: the eastern region retains its lead, while central and western regions show potential for catch-up, and the northeastern region faces a “balance trap.” Economic development foundation, human capital quality, and policy environment support are identified as the core driving forces of transformation, while other factors demonstrate pronounced regional and phase-specific variability. This study not only deepens theoretical understanding of the uneven development and driving logic of agricultural digital transformation but also provides empirical evidence to support policy optimization and promote more balanced and sustainable development in the agricultural sector.

1. Introduction

According to the World Bank’s projections, the global population is expected to reach 9.7 billion by 2050, leading to a dramatic surge in food demand [1]. However, due to constraints such as limited resources, climate change, and the restructuring of the labor force, traditional agricultural production models are increasingly unable to meet the multidimensional requirements of modern agricultural development in terms of production efficiency, environmental sustainability, and technological innovation [2]. In response, emerging digital technologies—including artificial intelligence (AI), the Internet of Things (IoT) [3], blockchain [4], and big data analytics [5]—are profoundly reshaping the global agricultural value chain. These innovations enable precision planting through AI, real-time field monitoring via IoT, quality assurance through blockchain-enabled traceability systems, and supply chain optimization using big data. Collectively, they are accelerating the transition toward a fourth agricultural production revolution [6], which is restructuring the entire “soil-to-table” ecosystem and promoting a paradigm shift toward enhanced productivity, ecological sustainability, and systemic resilience [7,8].

As the world’s largest producer and consumer of agricultural products, China presents a unique empirical context for studying agricultural digital transformation. In recent years, under the guidance of the national “Digital Countryside” strategy, rural digital infrastructure has undergone substantial improvements: 5G coverage in administrative villages now exceeds 90%, and the agricultural digital economy penetration rate has surpassed 35% [9]. Applications of digital technologies are also expanding from foundational infrastructure toward deeper integration in areas such as elite seed sharing, BeiDou-enabled smart farmland surveillance, and precision livestock farming. Nonetheless, compared to global frontrunners in digital agriculture—such as the Netherlands and the United States—China continues to face numerous challenges, including escalating resource-environmental constraints, pressures on maintaining the arable land red line, lagging technological iteration, institutional barriers to industrial structure upgrading, and a growing shortage of specialized talent [10]. These structural constraints underscore the urgent need for a systematic, multidimensional investigation into the spatiotemporal dynamics and driving mechanisms underpinning China’s agricultural digital transformation.

Existing research on agricultural digitalization predominantly addresses three key dimensions. First, technological effect studies have consistently demonstrated that the deep integration of digital technologies can optimize agricultural resource allocation [11], enhanced production efficiency [12], and mitigated environmental impacts [13,14]. These outcomes collectively contribute to the advancement of sustainable agricultural practices [15]. Second, in the domain of spatiotemporal evolution analysis, scholars have employed quantitative methods such as kernel density estimation, Markov chain modeling, panel quantile regression, entropy-weighted TOPSIS, and the Dagum Gini coefficient to evaluate the role of digital elements in enhancing agricultural economic performance and promoting regional coordination. However, most existing studies are confined to single-province analyses, thereby limiting the generalizability of findings [16,17]. Third, research on driving mechanisms generally identifies policy support [18], digital infrastructure [19,20], economic development levels [21], technological supply [22,23], and human capital availability [24] as core determinants influencing digital agricultural transformation. Nevertheless, insufficient attention has been paid to the heterogeneity of these drivers across regions at different development stages, which presents a notable gap in the literature.

In summary, although existing research affirms the positive role of agricultural digital transformation in promoting rural development and improving agricultural efficiency, its spatiotemporal evolution patterns, regional disparities, and underlying driving factors remain insufficiently and systematically explored. To address this gap, this study constructs a dual-dimensional analytical framework, “spatiotemporal evolution–driving identification,” aimed at systematically characterizing the evolutionary trajectory and motivational structure of agricultural digital transformation in China at a macro scale. From the spatiotemporal perspective, the study employs Theil index decomposition to reveal intra- and inter-regional disparities across the four major regions. It combines kernel density estimation to depict distributional changes and applies both global and local Moran’s I indices to analyze the spatial clustering and diffusion characteristics of agricultural digital transformation, thereby tracing its spatiotemporal evolution throughout China. In the driving forces dimension, the study utilizes a random effects model and panel quantile regression to identify the impact and contribution differences of driving factors across various development stages, thereby uncovering the heterogeneity of internal mechanisms within distinct geographic units. This study hypothesizes that China’s agricultural digital transformation exhibits significant spatial heterogeneity and that regional development paths follow the hierarchical diffusion model. These paths are primarily influenced by economic foundations, human capital structures, and policy environments. Ultimately, this research seeks to enhance the spatiotemporal insight and causal understanding of agricultural digitalization and to provide theoretical support for regionally differentiated policy formulation, facilitating China’s agriculture to achieve high-quality advancement and strategic leapfrogging under the wave of digital transformation.

2. Material and Methods

2.1. Study Area and Data Sources

This study selected 30 provinces (including autonomous regions and municipalities directly under the Central Government, excluding Hong Kong, Macao, Taiwan, and Tibet) on the Chinese mainland as the research regions, and the 30 provinces are categorized into four major regions, namely, the eastern, central, western, and northeastern regions, according to the criteria of the National Bureau of Statistics (Table S1). The research covered a time span from 2011 to 2023. The original data were primarily sourced from multiple authoritative statistical channels, including public data from the National Bureau of Statistics [25], the China Statistical Yearbook [26], the China Rural Statistical Yearbook [27], the China Population and Employment Statistical Yearbook [28], the China Science and Technology Statistical Yearbook [29], individual provinces statistical yearbooks, and relevant datasets such as the China Taobao Village Research Report published by the Alibaba Research Institute [30]. For missing data, linear regression interpolation was employed for completion. To deeply investigate the spatio-temporal evolution characteristics of agricultural digital transformation, the years 2011, 2017, and 2023 were selected as critical observation years to analyze its spatio-temporal evolution trends.

2.2. Construction of the Indicator System

This study fully incorporated the research experience of predecessors [11,31,32,33,34,35,36] and selected a total of 21 indicators across five dimensions to construct the indicator system for agricultural digital transformation: digital basic support capabilities, digitalization level of agricultural production, digital efficiency of agricultural operations, digital circulation system, and level of rural digital livelihood services. The specific indicator system is shown in

Table 1. Indicator system for agricultural digital transformation level.

First-Level Indicator	Second-Level Indicator	Third-Level Indicator	Measurement Method	Weight
Agricultural digital transformation	Digital basic support capability	Coverage rate of rural mobile communication network	Number of mobile phones per 100 rural households	0.0080
		Comprehensive coverage rate of rural digital infrastructure	Internet penetration rate × 0.6 + optical fiber cable line coverage rate × 0.4	0.0390
		Penetration rate of rural intelligent equipment	Number of intelligent terminal devices per 100 rural households	0.0307
		Fiscal support intensity for agricultural digital transformation	Proportion of fiscal expenditure on agriculture, forestry, and water affairs in general public budget expenditure	0.0249
		R & D investment intensity in agricultural science and technology	Number of patent authorizations in the agricultural field/rural permanent population	0.1202
	Digital level of agricultural production	Agricultural power output efficiency	Added value of agriculture, forestry, animal husbandry, and fishery/rural electricity consumption	0.0507
		Demonstration bases for rural digital economy	Number of taobao villages	0.1380
		Density of national digital agriculture parks	The Ministry of Agriculture and Rural Affairs has approved the designated demonstration zones and industrial parks	0.0457
		Application rate of agricultural machinery and equipment	Total agricultural machinery power per unit of cultivated land	0.0440
	Digital efficiency of agricultural operation	Digital display level of agricultural operation entities	Proportion of agricultural enterprises with official websites	0.0095
		Penetration rate of enterprise e-commerce	Proportion of agricultural enterprises engaging in e-commerce transactions	0.0255
		Digital retail scale of agricultural products	Digital retail sales of agricultural products per rural resident.	0.1872
		Development index of rural digital inclusive finance	Adopting the “China digital Inclusive finance index” by the digital finance research center of Peking University	0.0051
		Scale of rural online consumption market	Proportion of rural online retail sales in total social online retail sales	0.0233
	Digital circulation system	Coverage rate of rural smart logistics network	Length of delivery routes for rural users on delivery sections	0.0475
		Frequency of rural logistics services	Per capital weekly express delivery frequency in rural areas	0.0122
		Efficiency index of rural logistics services	Number of postal outlets/rural permanent population	0.0700
		Breadth of rural digital technology application	Average number of people served per postal and telecommunications business outlet	0.0152
	Level of rural digital Livelihood services	Level of rural mobile payment	Adopting the mobile payment sub-index in the “County digital finance index” by the digital finance research center of Peking University	0.0054
		Degree of digitalization	Adopting the digitalization dimension score in the “Peking University digital inclusive finance index”	0.0018
		Density of agricultural digital technology talents	Full-time equivalent of research and development personnel in each region × proportion coefficient by province type × proportion coefficient by year	0.0960

.

To overcome the subjective bias of subjective weighting methods and ensure the objectivity of weight allocation [37], an objective weighting method based on data characteristics (i.e., the entropy weight method) was used for preliminary processing of the original data. The entropy weight method can reflect the information volume of indicators through data dispersion, so it is widely used to measure the indicator system. Since the measurement units of each indicator are different, it is necessary to standardize the original data to eliminate the influence of dimensions.

2.2.1. Data Standardization

$$\mathrm{Positive} \mathrm{indicators}: X_{ij}^{\prime }={\displaystyle \frac{X_{ij}-max\left(X_{ij}\right)}{max\left(X_{ij}\right)-min\left(X_{ij}\right)}}$$

(1)

where $X_{ij}$ represents the original data of the $j$ year in the $i$ province, and $max\left(X_{ij}\right)$ and $min\left(X_{ij}\right)$ are the maximum and minimum values of each indicator across all provinces and the study period, respectively. $X_{ij}^{\prime }$ is the standardized value.

2.2.2. The Proportion Matrix

$$y_{ij}={\displaystyle \frac{x_{ij}^{\prime }}{{\sum }_{i=1}^m x_{ij}^{\prime }}}$$

(2)

where $y_{ij}$ represents the proportion of the $i$ province in the $j$ indicator.

2.2.3. The Information Entropy

$$e_j=-{\displaystyle \frac{1}{ln n}}{\sum }_{i=1}^m y_{ij}ln y_{ij}$$

(3)

where $e_j$ represents the information entropy of the $j$ indicator. $n$ represents the number of provinces.

2.2.4. The Weights

$$w_j={\displaystyle \frac{1-e_j}{{\sum }_{j=1}^m (1-e_j)}}$$

(4)

where $w_j$ represents the weight of the $j$ indicator.$m$ represents the total number of indicators.

2.2.5. The Comprehensive Evaluation Index

$$S_i={\sum }_{j=1}^m w_j\times y_{ij}$$

(5)

where $S_i$ represents the comprehensive score of the $i$ province.

2.3. Theil Index

The Theil index was employed to quantify regional disparities in agricultural digital transformation. Based on the principle of information entropy, overall regional differences were decomposed into within-regional and between-regional components to measure the degree of disparities, analyze structural characteristics, and identify the key driving factors of these differences [38].

$$T={\displaystyle \frac{1}{n}}{\sum }_{i=1}^n\left({\displaystyle \frac{x_i}{\overline{x}}}ln\left({\displaystyle \frac{x_i}{\overline{x}}}\right)\right)$$

(6)

where $T$ is the overall Theil index of the agricultural digital transformation level. $x_i$ is the indicator of the agricultural digital transformation degree of the $i$ province. $\overline{x}$ is the average value of the agricultural digital transformation level. $n$ is the number of provinces. The overall Theil index difference can be decomposed into within-region differences and between-region differences.

$$T=T_b+T_w={\sum }_{k=1}^K Y_{k}ln\left({\displaystyle \frac{Y_k}{n_k/n}}\right)+{\sum }_{k=1}^K Y_k\left[{\sum }_{i\in g_k} {\displaystyle \frac{y_i}{Y_k}}ln\left({\displaystyle \frac{y_i/Y_k}{1/n_k}}\right)\right]$$

(7)

where $T_b$ represents the between-region difference. $T_w$ represents the within-region difference. $K$ represents the number of regional groups. $Y_k$ is the proportion of the $K$ regional indicator in the national total value. $y_i$ is the proportion of the $i$ province in the overall. $n_k$ represents the number of provinces in the k region. $T_k=\sum _{i\in g_k} {\displaystyle \frac{y_i}{Y_k}}\mathit{ln}\left({\displaystyle \frac{\frac{y_i}{Y_k}}{\frac{1}{n_k}}}\right)$ is the Theil index of the $k$ region.

2.4. Kernel Density Estimation Method

Kernel density estimation is a non-parametric estimation method that does not require specific assumptions about the data distribution, exhibits weak model dependence [39], and demonstrates strong robustness. It intuitively represents the distribution and characteristics of a random variable through a continuous probability density curve. The calculation formula is as follows:

$$f(x)={\displaystyle \frac{1}{Nh}}{\sum }_{i=1}^N k\left({\displaystyle \frac{X_i-x}{h}}\right)$$

(8)

$$k(x)={\displaystyle \frac{1}{\sqrt{2\pi }}}exp\left(-{\displaystyle \frac{x^2}{2}}\right)$$

(9)

where $f\left(x\right)$ represents the kernel density estimate value at $x$. $N$ represents the number of observations in the region (i.e., the number of provinces), and $X_i$ represents the observation of the $i$ province. $k\left(x\right)$ represents the Gaussian kernel function. $h$ is the bandwidth. Our study employs the Gaussian kernel function to estimate the distribution dynamics of agricultural digital transformation at both the national and regional levels.

2.5. Spatial Autocorrelation Analysis

Moran’s index is an essential method for analyzing spatial heterogeneity and spatial autocorrelation. It is divided into global Moran’s I and local Moran’s index. Global Moran’s index measures overall spatial distribution characteristics and correlation levels, while local Moran’s index reveals local spatial association characteristics between individual provinces and their neighboring provinces. The calculation formula is as follows.

2.5.1. Global Moran’s Index

$$I={\displaystyle \frac{n{\sum }_{i=1}^n {\sum }_{j=1}^n w_{ij}\left(x_i-\overline{x}\right)\left(x_j-\overline{x}\right)}{({\sum }_{i=1}^n {\sum }_{j=1}^n w_{ij}){\sum }_{i=1}^n {\left(x_i-\overline{x}\right)}^2}}$$

(10)

2.5.2. Local Moran’s Index

$$I_i={\displaystyle \frac{x_i-\overline{x}}{\frac{1}{n}{\sum }_{i=1}^n {\left(x_i-\overline{x}\right)}^2}}{\sum }_{j=1}^n w_{ij\left(x_j-\overline{x}\right)}$$

(11)

where $I$ represents the global Moran’s index. $n$ represents the number of provinces.${ x}_i$ represents the level of agricultural digital transformation of the $i$ province.$\overline{x}$ represents the average value of the provinces. $w_{ij}$ represents the elements of the weight matrix based on spatial adjacency relationships.$I_i$ is the local Moran’s index, which classifies regions into five types: no significant clustering, high-high clustering, high-low clustering, low-high clustering, and low-low clustering.

2.6. Analysis of Driving Factors

2.6.1. Construction of the Driving Factor System

Based on the principle that indicators in the evaluation system should not be redundantly included in the analysis of influencing factors, and considering existing research and practical conditions [23,40,41,42,43,44], this study constructed a driving factor indicator system across six dimensions: economic development foundation (EDF), technological support strength (TSS), human capital accumulation (HCA), human capital quality (HCQ), policy environment support (PES), and market vitality level (MVL) (

Table 2. Indicator system of driving factors.

Driving Factors	Variable Explanation	Abbreviation
Economic development foundation	Per capita disposable income of rural residents	EDF
Technological support strength	Proportion of agricultural technology contract amount	TSS
Human capital accumulation	Agricultural technicians per 10,000 farmers	HCA
Human capital quality	Proportion of the rural labor force with high school education or above	HCQ
Policy environment support	Frequency of “digital village + agricultural digitalization” related terms in provincial government work reports	PES
Market vitality level	Proportion of agricultural products online retail sales to total agricultural output value	MVL

). Further analysis was conducted to examine their impact on the level of agricultural digital transformation.

2.6.2. Regression Analysis of Driving Factors

Our study employed panel data regression to systematically explore the variations in driving factors across both cross-sectional and temporal dimensions. To ensure model robustness, the variance inflation factor (VIF) test was conducted to examine multicollinearity, with results indicating that all driving factors have VIF values below 10 (Table S2), confirming the absence of multicollinearity. The Hausman test result (p = 0.15, Table S3) suggested that the random effects model is appropriate, with the following estimation equation:

$$y_{it}={\beta }_0+{\beta }_{1}EDF+{\beta }_{2}TSS+{\beta }_{3}HCA+{\beta }_{4}HCQ+{\beta }_{5}PES{+\beta }_{6}MVL+{\mu }_i+{\epsilon }_{it}$$

(12)

where $y_{it}$ represents the dependent variable for individual i at period t. ${\beta }_0$ denotes the intercept term. ${\beta }_X$ represents the coefficient to be estimated. ${\mu }_i$ represents the individual-specific random effect, and ${ \epsilon }_{it}$ represents the stochastic error term.

Although the random effects model effectively captures mean effects, it has limitations in characterizing the heterogeneity of variable impacts. Given that the comprehensive evaluation index of agricultural digital transformation exhibited non-normal distribution (p < 0.05, Table S4), and the Breusch-Pagan test confirmed the presence of heteroscedasticity (p < 0.05, Table S5), quantile regression was further employed in this study. Three quantiles—0.25 (low quantile), 0.5 (median quantile), and 0.75 (high quantile)—were selected for analysis, with the following estimation equation:

$$min\beta {\sum }_{i=1}^N{\sum }_{t=1}^T{\rho }_{\tau }(Y_{it}-X_{it}\beta )$$

(13)

where $\beta$ is the vector of regression coefficients to be estimated. N is the number of individuals. $T$ is the number of time periods. $X_{it}$ represents the explanatory variables for individual $i$ at period $\mathrm{t}. \tau$ denotes the quantile level. ${\rho }_{\tau \left(e\right)}$ represents the quantile loss function, which is defined as:${ \rho }_{\tau \left(e\right)}=\left\{\begin{array}{c}\tau e,\\ (\tau -1)e\end{array}\right.{\displaystyle \genfrac{}{}{0pt}{}{e\ge 0}{e<0}}$.

2.7. Data Analysis and Software

All statistical analyses were conducted using R v4.5.0 (R Core Team, 2025). To assess spatial autocorrelation at both global and local levels, global Moran’s I and local Moran’s I were computed using the moran.test and localmoran functions from the “spdep” package, respectively. The phtest and plm functions from the “plm” package were used to conduct the Hausman test and construct the random effects model, respectively. The bptest function from the “lmtest” package was used to diagnose heteroscedasticity. The rq function from the “quantreg” package was used to perform panel quantile regression. Data visualization was achieved using the “ggplot2” package in R and ArcGIS version 10.2.

3. Results

3.1. Spatial Pattern Evolution of Agricultural Digital Transformation

From 2011 to 2023, the national agricultural digital transformation index exhibited trends of regional differentiation, overall improvement, and dynamic optimization. In 2011, the national transformation index fluctuated within the interval of 3.08 to 6.30, which indicated a relatively low starting point for agricultural digital transformation nationwide. Only a few prominent eastern and coastal provinces exhibited leading performance during this period. By 2017, the index had ascended to the 5.08~8.81 interval, with observable advancements in economically developed regions such as the Yangtze River Delta and Pearl River Delta (Figure 1b). By 2023, the level of agricultural digital transformation had undergone further elevation, with the overall index achieving relatively high values clustered within the 6.34~11.61 range (Figure 1c). When synthesizing annual average data, agricultural digital transformation revealed persistent regional disparities. Eastern coastal provinces including Hebei, Shandong, Jiangsu, Zhejiang, and Guangdong have consistently held leading positions, while concurrently widening the developmental gap with western provinces (Figure 1d).

3.2. Analysis of Regional Disparities in Agricultural Digital Transformation

From a national perspective, the overall Theil index, within-regional Theil index, and between-regional Theil index of agricultural digital transformation exhibited synergistic trends from 2011 to 2023, following an initial decline and subsequent rebound. A higher index value indicates more pronounced regional disparities. Notably, the inflection points of the overall Theil index and within-regional Theil index were observed in 2015, while that of the between-regional Theil index was identified in 2016 (Figure 2, Table S6). Before 2015, the contribution rate was primarily driven by the between-regional Theil index, reaching a maximum value of 58%. After that, the within-regional Theil index became the dominant factor. Overall, the Theil index of agricultural digital transformation demonstrated a fluctuating pattern throughout 2011–2023: the initial decline reflected a narrowing of development gaps between regions, while the later rebound signaled intensified differentiation within regions. By 2023, the index had returned to its 2011 level, indicating that regional disparities remain unresolved—particularly the imbalanced development among provinces within the same region.

From the perspective of the four major regions, the between-regional Theil indices of the eastern and central regions exhibited similar upward trends, though the eastern region’s between-regional Theil index has consistently been higher than that of the central region across all years. In contrast, the western region displayed an overall downward trend in its between-regional Theil index. Although this region recorded higher between-regional inequality than other regions in 2011 and 2012, its index has remained only slightly above that of the northeastern region since 2019. Meanwhile, the between-regional Theil index of the northeastern region exhibited remarkable stability throughout the study period (Figure 3).

3.3. Dynamic Evolution Characteristics of Agricultural Digital Transformation

From 2011 to 2023, the peak values and variation intervals of the kernel density curves for the nation and its four major regions exhibited a consistent rightward shift, reflecting distinct regional characteristics (Figure 4). Nationally, the kernel density curve transitioned from a high, narrow single peak in 2011 to a low, wide distribution between 2017 and 2023 (Figure 4a), indicating an overall improvement in agricultural digital transformation nationwide, yet with increasingly pronounced regional disparities. The western region displayed an asymmetric evolutionary pattern characterized by a “wide-narrow-wide” fluctuation and persistent multimodality (Figure 4d), suggesting significant internal variation in transformation levels and a complex pattern of differentiation. With the exception of the western region, the other three regions demonstrated similar trends of decreasing peak height and increasing peak width, albeit with notable differences. The eastern and central regions developed multimodal structures in the later years (Figure 4b,c), indicating the formation of groups at different development levels within these regions and a widening of internal disparities. The northeastern region maintained stability and multimodality in the early period, followed by a decline in peak height and an expansion in peak width after 2017 (Figure 4e), reflecting a growing internal disparity over time.

3.4. Spatial Correlation Characteristics of Agricultural Digital Transformation

The global Moran’s I exhibited fluctuating trends from 2011 to 2023 (Figure 5). Between 2011 and 2014, the index remained at a relatively high level, consistently exceeding 0.3. A notable decline occurred in 2015–2016, reaching a recent low of 0.23 in 2016. From 2017 onward, the index rebounded, maintaining a moderate level, though slightly below the earlier values.

From the perspective of local Moran’s I, 21 out of 30 provinces exhibited stable spatial clustering patterns in 2011, 2017, and 2023 (Figure 6). Specifically, Inner Mongolia, Liaoning, Jilin, Heilongjiang, Guizhou, Yunnan, Shaanxi, Gansu, Qinghai, Ningxia, and Xinjiang consistently exhibited low-low clustering. These provinces, predominantly located along the western and northeastern peripheries, have experienced prolonged lags in agricultural digital transformation due to complex topography, inadequate infrastructure investment, and limited technology adoption. In contrast, Shanghai, Jiangsu, Zhejiang, Anhui, Shandong, and Henan consistently showed high-high clustering, reflecting strong agricultural digitization supported by favorable geographic conditions and robust digital economies. Their advanced development levels not only benefit local growth but also positively influence adjacent regions. Hebei, Hubei, Hunan, and Sichuan remained in the high-low cluster group throughout the study period, while Tianjin consistently exhibited a low-high clustering pattern—highlighting a spatial mismatch between local transformation capacity and neighboring environments. These provinces could benefit from adopting best practices from leading regions to foster agricultural digital development in lagging areas. Moreover, provinces such as Beijing, Fujian, Jiangxi, Guangdong, and Hainan experienced transitions from high-high clustering to either high-low or low-high clustering. This shift reflects structural changes driven by accelerated urbanization and the declining relative importance of agriculture. Lastly, Shanxi, Guangxi, and Chongqing continuously shifted between low-high and low-low clustering, indicating unstable digital foundations and heightened sensitivity to policy changes or market volatility.

3.5. Analysis of the Driving Factors of Agricultural Digital Transformation

Panel Data Regression Results

The agricultural digital transformation index is shaped by six key driving factors, each exerting significant effects across the random effects model and various quantile levels, reflecting dynamic variation patterns. The economic development foundation (EDF) exhibits a stable and consistently positive effect across all quantile models, with the strongest impact observed in the high quantile range, confirming its role as the core driving factor for agricultural digital transformation. Human capital quality (HCQ) shows the most significant positive influence in the low and median quantile ranges, demonstrating a “moderate matching” pattern, indicating that skill enhancement is crucial for transformation efficiency during the early and intermediate stages, thereby positioning it as a secondary core driver. Policy environment support (PES) displays a stage-specific positive effect in the lower quantile ranges, suggesting that policy intervention remains an important driver in regions with weaker foundational conditions. In contrast, technological support strength (TSS) and human capital accumulation (HCA) show negative effects in the high quantile range, implying that structural challenges such as the “disconnection between input and output” in technology investment and “skill mismatch” continue to hinder digital transformation. Of particular concern is the market vitality level (MVL), which exhibits a significantly negative effect across all models—most prominently in high quantile regions. This highlights how structural barriers, resource misallocation, and platform monopolies have become critical obstacles restricting the deep advancement of agricultural digitalization, necessitating institutional restructuring and mechanism optimization to address these issues (

Table 3. Results of random effects model and panel quantile regression model.

Driving Factors	Random Effects Regression Model	Panel Quantile Regression Model
		τ = 0.25	τ = 0.50	τ = 0.75
EDF	0.000163	0.00018	0.00016	0.00024
TSS	−0.0134	0.0269	−0.032	−0.0804
HCA	0.0157	−0.0108	−0.00496	−0.042
HCQ	4.7926	5.0223	6.2696	4.9703
PES	0.0183	0.03007	0.03069	0.0169
MVL	−0.1176	−0.199	−0.1879	−0.2212

Note: The numerical values in the table represent the regression coefficients of each independent variable in the respective regression models. Bolded values indicate variables (driving factors) that have a statistically significant impact on the dependent variable (composite index of agricultural digitalization transformation). This composite index was constructed based on an evaluation indicator system and synthesized using the entropy weighting method to measure the level of agricultural digital transformation. For specific methodological details regarding its construction, please refer to Section 2.2.

).

4. Discussion

From 2011 to 2023, the national agricultural digital transformation in China demonstrated an unbalanced evolutionary feature of “simultaneous overall improvement and regional differentiation.” The growth of the national agricultural digital transformation index and the rightward shift of the kernel density curve center indicate that all regions achieved digital upgrades to varying degrees during the study period. Spatially, the development level presented a stepped distribution pattern across the eastern–central–western–northeastern regions, revealing significant regional disparities—consistent with the findings of Zhang and Liu [45,46,47]. From a regional perspective, agricultural digital transformation showed distinct heterogeneity characterized by “layered progression and internal differentiation.” The eastern region formed a digital core zone by virtue of its economic advantages, advanced digital infrastructure, and strong market consumption capacity. However, the multi-peak structure of its kernel density curve reveals disparities among provinces, which is consistent with the research findings of Acemoglu [48]. The transformation trajectory in the central region was more volatile, with the overall Theil Index following a “rise–decline–rise” trend, reflecting the joint influence of policy and market forces. While early-stage coordination led to notable progress, subsequent divergence emerged due to locational and industrial heterogeneity. The western region demonstrated a typical “policy-driven, market-constrained” transformation path. Although targeted fiscal support fostered partial breakthroughs in the early stage, improvements in digital productivity remained limited, and infrastructure investments failed to translate into substantive output, resulting in evident polarization. This pattern is consistent with the assessment of Li [49]. The northeastern region has long been trapped in a state of “equilibrium stagnation.” Although state-owned farms possess relatively strong capacities for technological adoption, smallholders remain constrained by financial and skill-related barriers, leading to low adoption rates and marked structural differentiation.

The analysis of driving factors further reveals the structural mechanisms underpinning agricultural digital transformation. This study identifies three core drivers—economic development foundation (EDF), human capital quality (HCQ), and policy environment support (PES). Economic development foundation exhibits the strongest marginal effect in high-quantile models, indicating that high-income regions possess greater capacity for resource integration and technological embedding [50]. This aligns with existing research that regards per capita GDP and household consumer expenditure as key transformative forces [51,52], reaffirming the foundational role of economic strength as the base layer of transformation. Human capital quality demonstrates stage-dependent variability and a “moderate matching” effect, suggesting that agricultural digital transformation relies more on improvements in the quality rather than the quantity of human capital [53], in line with the arguments presented by Wang [43]. In contrast, human capital accumulation (HCA) shows negative coefficients in most models, indicating that skill mismatches and population aging may suppress transformation outcomes [54]. Technological support strength (TSS), while a critical input dimension, shows negative effects in certain regional models, reflecting a disconnect between resource investment and productivity outcomes. This underscores the need to enhance local adaptation mechanisms and innovation ecosystems. This finding also indirectly supports the empirical conclusion by Kitole, which highlights the decisive role of service quality in technology adoption behavior [55]. The role of policy environment support (PES) exhibits significant stage-based variation. In early transformation phases, infrastructure-oriented policies effectively mitigated regional disparities; in later stages, capital markets and technological innovation became primary drivers, forming an evolutionary path of “policy-driven-market-led” development [56]. Meanwhile, market vitality level (MVL) consistently acts as a negative constraint, with structural barriers—such as market fragmentation, competitive imbalance, and platform monopolies—posing deep challenges to the equitable diffusion and coordinated development of digital technologies. These findings corroborate the concerns expressed by Abiri, regarding agricultural market segmentation and digital oligopolies [13]. In broader developing country contexts, Mhlanga identifies poor infrastructure, low technological awareness, high costs, and privacy concerns as critical barriers to digital transformation in African regions, further emphasizing the importance of institutional governance [57]. While this study primarily focuses on structural variables, existing literature also highlights the role of behavioral response mechanisms. Blasch emphasize that social networks and trust mechanisms offer stronger explanatory power in farmers’ technology adoption decisions [58]. Kitole identifies functional complementarity between education levels and extension services [55]. Zhao proposes that transformation costs, expected returns, and policy incentive structures are decisive in shaping transformation pathways [59]. In contrast to behavioral research, this study contributes by examining the regional heterogeneity of structural driving forces and extends the understanding of spatial sensitivity in behavioral response through quantile analysis.

In conclusion, the driving mechanisms and evolutionary pathways of agricultural digital transformation are not universally applicable but instead require a multi-tiered and diversified framework tailored to regional resource endowments, institutional configurations, and stakeholder compositions. Future research may integrate international experiences and micro-level tracking methodologies to further enrich theoretical models and enhance the precision and contextual adaptability of policy design.

5. Conclusions, Suggestions, and Limitations

5.1. Conclusions

This study systematically examines the spatiotemporal evolution and driving factors of agricultural digital transformation in China from 2011 to 2023. The main conclusions are as follows:

(1)

Agricultural digitalization has steadily advanced, accompanied by pronounced spatial differentiation.

During the study period, the overall level of agricultural digital transformation increased significantly nationwide, as evidenced by the continuous growth of the transformation index. This indicates that digital upgrading in agriculture has been widely promoted across the country. However, regional disparities remain prominent. The transformation pathway has undergone a phase-wise shift from being dominated by interregional differences to being increasingly shaped by intraregional heterogeneity. Four major regions formed a spatial pattern characterized by “layered progression and internal divergence,” following a stepped distribution from east to central, west, and northeast China. This reflects pronounced spatial fragmentation and region-specific development paths.

(2)

Economic foundation, human capital, and policy environment are the core drivers of transformation.

Economic development foundation exerts a significant positive influence on agricultural digital transformation, particularly in higher quantile regions, where stronger economic capacity enables more effective resource integration and technological embedding. Human capital quality outweighs mere quantity accumulation, displaying a “moderate matching” effect. Policy environment support exhibits a stage-specific evolution, transitioning from initial infrastructure-based interventions aimed at closing foundational gaps to a later phase driven by market dynamics and technological innovation. Meanwhile, both technological investment and market mechanisms reveal regional sensitivity. In some areas, the conversion efficiency of technological resources remains low, indicating a mismatch between input and output. Structural barriers such as market segmentation and platform monopolization further constrain the equitable diffusion and coordinated development of digital technologies. These findings suggest that agricultural digital transformation continues to face dual challenges: improving resource allocation efficiency and optimizing the institutional environment. Breakthroughs will require enhancements in local technological adaptability and more robust market governance frameworks.

5.2. Suggestions

To address the developmental barriers encountered by regions and stakeholders in agricultural digital transformation and achieve the transition from regional divergence to holistic coordination, this paper proposes the following policy recommendations:

First, provinces should systematically assess the driving forces and constraints of agricultural digital transformation and develop differentiated strategies based on regional resource endowments. Specifically, the eastern region should leverage its role as a “digital core zone” to facilitate the diffusion of digital technologies to surrounding areas. The central region should focus on enhancing human capital quality through targeted education and training programs. The western region should receive increased fiscal support and technology promotion to overcome geographical and infrastructural bottlenecks. The northeastern region should prioritize digital integration between state-owned farms and smallholder farmers to bridge the implementation gap between large-scale and small-scale agriculture. Additionally, between-regional exchanges and cooperation should be strengthened to establish a cross-regional collaborative development mechanism for agricultural digitalization. This mechanism should facilitate the sharing of best practices, technological resources, and policy insights, thereby enhancing overall digitalization levels and achieving high-quality agricultural modernization.

Second, a long-term and stable policy framework should be established to ensure the sustainability of agricultural digital transformation. This necessitates consistent policy support to mitigate disruptions caused by short-term policy fluctuations. Additionally, enhanced monitoring and evaluation of policy implementation effects are essential, enabling timely adjustments and refinements to ensure the effective realization of policy objectives.

Finally, the uneven progression of agricultural digital transformation stems from the interplay of policy cycles, economic foundations, and human capital. Moving forward, efforts should adopt a regionally tailored approach of “innovation-driven radiation in the East, institutional activation in the Central region, infrastructure-based support in the West, and collaborative breakthroughs in the Northeast” to facilitate the shift from regional divergence to holistic coordination, thereby fostering high-quality agricultural modernization.

5.3. Limitations

This study innovatively integrates long-term provincial panel data with multidimensional analytical tools to reveal the spatiotemporal evolution, regional heterogeneity, and driving factors of agricultural digital transformation in China, contributing a macro–regional–mechanism analytical framework that advances the understanding of digital agriculture’s uneven development and its driving dynamics. While these insights offer theoretical depth and practical relevance, certain limitations remain. Due to data accessibility constraints, county-level assessments could not be conducted, thereby limiting the precision and applicability of the findings. Additionally, restrictions imposed by statistical methodologies have resulted in a lack of data on digital technology adoption at the individual farmer level, making it difficult to accurately capture grassroots transformation dynamics. Future research could incorporate remote sensing monitoring and farmer survey data to enhance analytical depth and robustness.