Can the Digital Economy Enhance the Export Competitiveness of Agricultural Products?—An Empirical Analysis Based on Panel Data from 30 Chinese Provinces

Abstract

As a new engine of economic development, the digital economy is profoundly reshaping the governance structure and production models of modern agriculture and has increasingly become a significant force driving the high-quality development of agricultural export trade. Based on provincial panel data from 30 Chinese provinces spanning from 2014 to 2023, this paper employs a two-way fixed effects model to systematically examine the statistical association between the digital economy and agricultural export competitiveness. Furthermore, a mediation effects model is adopted to explore the underlying transmission pathways, and a threshold effects model is applied to reveal the nonlinear characteristics of this association, thereby providing a multidimensional analysis of the underlying mechanisms. The empirical results indicate that: (1) there is a significant positive statistical association between the digital economy and agricultural export competitiveness, and this finding remains robust across a series of robustness checks; (2) upgrading of agricultural industrial structure, agricultural technological innovation, and agricultural labor transfer serve as three important transmission pathways through which the digital economy is associated with enhanced export competitiveness; (3) this positive association exhibits regional heterogeneity, with stronger effects observed in central regions and major grain-producing areas; and (4) there is a nonlinear relationship between the digital economy and agricultural export competitiveness. Finally, based on these empirical findings, several policy implications are proposed, including strengthening digital infrastructure construction, enhancing trade openness, cultivating agricultural digital talents, optimizing factor allocation structures, and improving the digital economy governance system, so as to fully unlock the potential space of the positive association between the digital economy and agricultural export competitiveness.

1. Introduction

As the new generation of information technology revolution—represented by big data, artificial intelligence, and the Internet of Things—sweeps across the globe, the digital economy is reshaping the landscape and trajectory of the world economy, becoming a core engine driving global economic recovery and growth [1]. In recent years, China has attached great importance to the enabling role of the digital economy in agriculture, successively issuing guiding documents such as the Strategic Outline for the Development of Digital Villages and the Digital Rural Agricultural Development Plan (2022–2025). Against this backdrop, China’s agricultural economy has achieved leapfrog development. However, according to data published in the China Agricultural Products Import and Export Annual Statistical Report, China’s agricultural product imports reached as high as USD 215.16 billion in 2024, while exports amounted to only USD 103.01 billion, resulting in a trade deficit of USD 112.15 billion—nearly double that of 2018. This indicates that China’s agricultural products have a high degree of external dependence, and the prospects for export trade and international competitiveness remain far from optimistic. Moreover, as modern agricultural production transitions from a phase prioritizing output to one prioritizing quality, people have placed higher demands on the quality of agricultural products. The contradiction between the traditional low-value-added agricultural production structure and the international consumer market has become increasingly prominent [2], severely hindering the realization of China’s strategic goal of transforming from a “large trading nation” to a “strong trading nation.” Therefore, how to harness the disruptive force of the digital economy to enable the high-quality development of agricultural product exports—and to solve the long-standing problems of low value addition, information asymmetry, and tightening environmental constraints, thereby building sustained and stable international competitive advantages—has become a major issue of both theoretical value and practical significance.

2. Literature Review

2.1. Research on the Digital Economy

Since Tapscott (1996) [3] introduced the concept of the “digital economy,” its core has consistently revolved around the restructuring of information generation, transmission, and processing methods through digital forms, aiming to enhance efficiency and reduce transaction costs (Ren and He, 2022) [4]. The digital economy is essentially a structural transformation resulting from the deep embedding of digital technologies into various sectors of the national economy, manifesting as a new economic form characterized by digital infrastructure as the foundation, digital platforms as the carrier, and data elements as the core driving force (Xu and Zhang, 2020) [5]. The G20 Digital Economy Development and Cooperation Initiative identifies digitized knowledge and information as key factors of production, a definition that has been widely adopted.

Due to differences in statistical calibers and the virtuality and high permeability of the digital economy (Wen et al., 2019) [6], there is no unified methodology for its measurement, and it faces technical constraints. Existing research primarily follows two paradigms: calculating value added based on national economic accounting, and constructing multi-dimensional indicator systems for comprehensive evaluation. Domestic studies predominantly adopt the comprehensive index method, such as Guan et al. (2020) constructing a provincial-level evaluation system [7]; the Digital Financial Inclusion Index (Guo et al., 2020) is also widely used [8]. Indicator weighting generally employs the entropy method or principal component analysis (Zhang and Ye, 2017) [9]. This paper, referencing existing studies and considering the specific implementation pathways of agricultural digitization, constructs an indicator system from three dimensions: digital infrastructure, digital industrialization, and industrial digitization.

2.2. Research on the Export Competitiveness of Agricultural Products

The measurement methods for the export competitiveness of agricultural products mainly include the Diamond Model, the Constant Market Share Model, and comprehensive evaluation indicator systems. International market share, Revealed Comparative Advantage index, and Trade Competitiveness index are widely adopted (Pan et al., 2021) [10]. In terms of measurement results, Yu et al. (2016) found that China’s exports to South Asia are at a disadvantage [11]; Yang and Tian (2022), through a comparison among China, Japan, and South Korea, found that China’s exports are primarily low- and medium-technology products [12]; Shuai et al. (2003) pointed out that while exports are growing, competitiveness is weakening [13]; in terms of product categories, vegetables and fruits outperform grains (Mao and Feng, 2007) [14]. Regarding influencing factors, the marginal contribution of traditional factors is diminishing, while the enabling effects of modern factors such as technology, information, and data are becoming prominent (Wu et al., 2023) [15]. Trade policy adjustments exert significant impacts (Narayan and Bhattacharya, 2019) [16], and technological progress is a key pathway (Li and Ding, 2016) [17]. Although it may increase production costs in the short term, it has a significant positive promoting effect in the long run (Zhang and Hu, 2013) [18]. Upon reviewing the relevant literature, it is evident that existing studies primarily analyze the impact mechanisms of agricultural product exports through their constituent elements, lacking research that incorporates the digital economy, influencing factors, and conditional factors into a unified analytical framework.

2.3. Research on the Impact of the Digital Economy on the Export Competitiveness of Agricultural Products

Existing studies have focused on the promoting role of digital tools in agricultural trade. Fan (2020) found that Internet popularization can shorten customs clearance time and reduce costs [19]; Tang and Li (2023) found that digital trade platforms can generate positive incentives [20]; Carlos Arnade (2017) found that digital platforms can reduce fixed entry costs for exports [21]; Qi et al. (2023) found that networked sales platforms can enhance production and sales efficiency [22]. However, most of these studies analyze the digital economy as a whole variable, with fewer conducting differentiated internal structural tests. In terms of transmission pathways, Zhu et al. (2019) confirmed that informatization levels have a positive impact on agricultural total factor productivity [23]; Ma et al. (2019) confirmed that digital platforms can reduce agricultural product transaction costs [24]; Bai and Huang (2021) argued that digital platforms can generate quality premiums for agricultural products [25].

Although the existing literature on the digital economy enabling agricultural product exports provides valuable insights, certain limitations remain: a lack of research incorporating the digital economy, mediation mechanisms, and threshold effects into a unified framework; and insufficient systematic testing of regional differences and conditional factors. Future research needs to integrate threshold effect models with regional heterogeneity analysis to deeply reveal the differentiated characteristics and threshold conditions of the digital economy in enabling the export competitiveness of agricultural products.

A review of the existing literature reveals that previous studies have left room for further exploration in the following aspects: (1) Most studies have focused on the linear association between the digital economy and agricultural trade, with few embedding it within a unified analytical framework of “influencing factors—transmission pathways—boundary conditions” to examine its indirect mechanisms and nonlinear characteristics. (2) Discussions on regional heterogeneity have largely been confined to mean comparisons across the eastern, central, and western macro-regions, lacking a nuanced examination of structural dimensions in agricultural production, such as grain functional zones. (3) Constrained by the narrowness of research perspectives, few studies have integrated mediation mechanisms and threshold effects into a coherent empirical framework, making it difficult to comprehensively uncover the “black box” of how the digital economy is associated with agricultural export competitiveness.

Accordingly, the marginal contributions of this paper are threefold. First, we construct a multiple-mediation framework of “digital economy → (industrial structure upgrading/technological innovation/labor transfer) → agricultural export competitiveness,” thereby opening the “black box” of the transmission pathways through which the digital economy is associated with export competitiveness. Second, we introduce trade openness as a threshold variable to reveal the “nonlinear leap” in this association as external market conditions evolve. Third, we conduct heterogeneity analyses from both geographic regional and grain functional zone perspectives, providing more granular empirical evidence for differentiated policy design. These three contributions together constitute the incremental value of this paper relative to the existing literature, helping to deepen the systematic understanding of the relationship between the digital economy and agricultural trade, while offering an extensible analytical framework for future research.

3. Theoretical Analysis and Research Hypotheses

3.1. The Direct Impact of the Digital Economy on the Export Competitiveness of Agricultural Products

Against the backdrop of the rapid penetration of agricultural digitization, the digital economy has become an endogenous driving force for rural economic development. Its direct impact on the export competitiveness of agricultural products is primarily reflected in the following three aspects: (1) In agricultural production, the digital economy can promote the intelligent, intensive, and precise development of traditional agriculture [26]. Internet of Things technologies, characterized by information perception, can accurately and promptly collect agricultural production data, enabling farmers to extract effective information from production data and achieve precise regulation of agricultural production. Meanwhile, the digital economy enhances the level of agricultural mechanization and automation, greatly improving production efficiency, reducing labor intensity and production costs, and ensuring the stability and consistency of agricultural product quality. (2) In agricultural capital circulation, the digital economy effectively alleviates the difficulties of costly and inaccessible financing faced by agricultural production through inclusive finance, mobile payments, and online credit. This enables agricultural enterprises to invest in advanced technologies, equipment, and high-quality inputs, breaking through capital constraints and avoiding agricultural cyclical problems, thereby providing the necessary financial support for enhancing the export competitiveness of agricultural products [27]. (3) In agricultural product export circulation, the digital economy breaks through traditional temporal and spatial constraints, enabling exporting enterprises to directly reach global consumers and realize a series of trade processes such as online precision marketing, trade negotiations, contract signing, export customs declaration, and inspection and quarantine. Simultaneously, the development of the digital economy enhances market transparency, facilitating farmers and agricultural exporting enterprises to adjust production strategies in a timely manner, changing their status as “price takers” and enhancing their bargaining power in factor and product markets, thereby improving the export competitiveness of agricultural products.

Furthermore, the enhancing effect of the digital economy on the export competitiveness of agricultural products may exhibit heterogeneity. First, provinces (municipalities) in different geographical locations differ in terms of policy support and resource allocation. The eastern and central regions tend to receive more policy support and resource allocation, helping them establish more comprehensive infrastructure and service systems in the digital economy field, thereby enhancing the quality and quantity of agricultural products. In contrast, the western region may be at a disadvantage in resource acquisition, facing issues such as digital technology monopolies, imperfect infrastructure, and information trade barriers. However, it may also exhibit unique advantages in agricultural production, operation, and circulation due to shorter decision-making chains and more flexible policy implementation [28]. Second, different provinces (municipalities) differ in terms of economic development levels and agricultural production modes. Major grain-producing areas typically possess abundant cultivated land resources, comprehensive infrastructure, and mature agricultural production systems, and also frequently benefit from key national policy support, further promoting the enhancing effect of the digital economy on the export competitiveness of agricultural products. Non-major grain-producing areas, constrained by geographical location and other factors, are at a disadvantage in terms of production scale, industrial structure, and product quality [29]. Based on this, this paper proposes the following hypothesis:

H1. 

The digital economy can enhance the export competitiveness of agricultural products, and this enhancing effect exhibits regional heterogeneity.

3.2. Transmission Mechanisms of the Digital Economy’s Impact on the Export Competitiveness of Agricultural Products

The digital economy can indirectly enhance the export competitiveness of agricultural products through three pathways: upgrading of the agricultural industrial structure, agricultural technological innovation, and agricultural labor transfer.

First, there is a positive synergistic relationship with the upgrading of agricultural industrial structure. The digital economy enhances the efficiency of information allocation, strengthens the market’s regulatory capacity over agricultural factors, alleviates information gaps between producers and consumers, reduces factor misallocation, and promotes the coordinated development of various sub-sectors within agriculture, laying the foundation for the rationalization of the industrial structure. Meanwhile, the deep application of technologies such as the Internet of Things and big data drives the entire agricultural industry chain toward digital and intelligent evolution, breaks down traditional industrial barriers, fosters the development of new business forms, and promotes the upgrading of agriculture toward high value addition. The rationalization and upgrading of the industrial structure jointly provide solid support for the export competitiveness of agricultural products.

Second, there is a statistically positive association with the enhancement of agricultural technological innovation. The digital economy reduces the costs of information acquisition and transactions for technological innovation, enabling innovation entities to more conveniently access cutting-edge technologies and market dynamics. Digital village construction enhances the digital literacy of agricultural practitioners and promotes the implementation and transformation of technological innovation outcomes. Technological innovation, on the one hand, enhances the level of agricultural mechanization, automation, and intelligence, strengthens the synergy among various industries, and promotes the evolution of agricultural product exports toward higher quality; on the other hand, it extends the industrial chain, promotes the transformation of agricultural products toward deep processing and high value addition, and significantly enhances their international market competitiveness.

Third, there is a statistically positive association with the expansion of agricultural labor transfer. Digital technologies break down information barriers in the labor market and reduce job search costs. Flexible employment opportunities provided by Internet platforms accelerate the transfer of agricultural labor to non-agricultural sectors. Meanwhile, new business forms fostered by the digital economy create abundant employment demand, promote the optimal allocation of production factors, and enhance agricultural production efficiency. The income growth effect brought by labor transfer, through the income elasticity mechanism, stimulates the upgrading of agricultural product consumption, indirectly promoting the evolution of the agricultural industrial structure toward high value addition and enhancing export competitiveness.

In summary, this paper proposes hypothesis H2: The digital economy indirectly enhances the export competitiveness of agricultural products by promoting the upgrading of the agricultural industrial structure, stimulating agricultural technological innovation, and facilitating labor transfer. Based on this, this paper proposes the following hypothesis:

H2. 

The digital economy enhances the export competitiveness of agricultural products by promoting the upgrading of the agricultural industrial structure, stimulating agricultural technological innovation, and facilitating labor transfer.

3.3. Nonlinear Impact of the Digital Economy on the Export Competitiveness of Agricultural Products

The improvement of China’s agricultural product export competitiveness is not only related to the level of digital economy development but also closely associated with trade openness. Ren (2014) pointed out that in the context of increasingly sophisticated international division of labor, a country must have global cooperation opportunities to gain competitive advantages [30]. The digital economy must be closely integrated with export trade to maximize its competitiveness in agricultural exports. When the level of trade openness is low, the production and export of agricultural products face numerous problems such as distorted price mechanisms, restricted market access, weak technological innovation, and insufficient human capital accumulation, making it difficult to fully leverage the enhancing effect of the digital economy on the export competitiveness of agricultural products. Moreover, developing the digital economy requires extremely high costs. In regions with low trade openness, information is isolated, infrastructure is imperfect, and they are trapped in the dilemma of “digital depression” and “digital divide” for a long time. Developing the digital economy not only fails to enhance the export competitiveness of agricultural products but may instead cause negative impacts, trapping agricultural enterprises and agricultural product processing enterprises in situations where capital cannot circulate. In regions with higher trade openness, the digital economy provides a key platform for agricultural enterprises to access global agricultural technology and R&D networks, not only accelerating local human capital accumulation and technological learning processes but also helping to break through innovation bottlenecks and optimize factor structures, thereby more effectively translating into export competitiveness of agricultural products. This suggests that the impact of the digital economy on the export competitiveness of agricultural products may be moderated by regional trade openness levels, potentially exhibiting a nonlinear relationship—that is, a threshold effect based on trade openness. Based on this, this paper proposes the following hypothesis:

H3. 

Trade openness exhibits a threshold effect in the enhancing role of the digital economy on the export competitiveness of agricultural products.

4. Research Design

Based on the analysis of the above theoretical mechanisms, this chapter focuses on examining the impact of the digital economy on China’s agricultural product export competitiveness, constructing relevant empirical models to test the three hypotheses proposed in Section 3.

4.1. Model Specification

4.1.1. Benchmark Regression Model

To investigate the direct impact of the digital economy on the export competitiveness of agricultural products, the following benchmark model is constructed:

$${\mathrm{T}\mathrm{c}}_{it}={\partial }_0+{\partial }_1{Dige}_{it}+{\partial }_{2}contro{l}_{it}+{\gamma }_i+{\rho }_i+{ϵ}_{it}$$

(1)

where ${\mathrm{T}\mathrm{c}}_{it}$ represents the export competitiveness of agricultural products in province (municipality) i at time t; ${Dige}_{it}$ represents the level of digital economy development in province (municipality) i at time t; $contro{l}_{it}$ represents a series of control variables; ${\gamma }_i$ and ${\rho }_i$ represent province (municipality) fixed effects and time fixed effects, respectively; and ${ϵ}_{it}$ is the random disturbance term.

4.1.2. Mediation Effect Model

To further test the mechanisms through which the digital economy enables the export competitiveness of agricultural products, the following models are specified:

$${\mathrm{T}\mathrm{c}}_{it}={\partial }_0+{\partial }_1{Dige}_{it}+{\partial }_{2}contro{l}_{it}+{\gamma }_i+{\rho }_i+{ϵ}_{it}$$

(2)

$${ M}_{it}={\partial }_2+{\partial }_3{Dige}_{it}+{\partial }_{4}contro{l}_{it}+{\gamma }_i+{\rho }_i+{ϵ}_{it}$$

(3)

$${\mathrm{T}\mathrm{c}}_{it}={\partial }_5+{\partial }_6{Dige}_{it}+{\partial }_7{M}_{it}+{\partial }_{8}contro{l}_{it}+{\gamma }_i+{\rho }_i+{ϵ}_{it}$$

(4)

where $M_{it}$ represents a series of mediating variables, respectively denoting the upgrading of the agricultural industrial structure, agricultural technological innovation, and agricultural labor transfer; ${\partial }_1$ represents the total effect of the digital economy on the export competitiveness of agricultural products; ${\partial }_3$ represents the effect of the digital economy on the mediating variables; ${\partial }_6$ represents the direct effect of the digital economy on the export competitiveness of agricultural products; and the meanings of other symbols are the same as in Equation (1).

4.1.3. Threshold Effect Model

Based on the preceding theoretical analysis, the mechanism through which the digital economy enables the export competitiveness of agricultural products may be constrained by the level of trade openness. Therefore, this paper, referencing the study of Yu and Zhang (2024) [31], specifies the following model to test the nonlinear relationship between the two:

$${\mathrm{T}\mathrm{c}}_{it}={\partial }_0+{\partial }_1{Dige}_{it}·I({To}_{it}\le \theta )+{\partial }_2{Dige}_{it}({To}_{it}>\theta )+{\partial }_{3}contro{l}_{it}+{\gamma }_i+{\rho }_i+{ϵ}_{it}$$

(5)

where ${To}_{it}$ is the threshold variable; $\theta$ is the threshold value; I is the indicator function, taking the value of 1 when the threshold variable is greater than $\theta$, and 0 otherwise; and the meanings of other variables and parameters are consistent with the preceding text.

4.2. Variable Selection

4.2.1. Dependent Variable

Export Competitiveness of Agricultural Products (TC). To measure the export competitiveness of agricultural products, this paper selects the Trade Competitiveness (TC) index as the dependent variable. This index was first proposed by Grubel and Lloyd (1975) [32], measuring the proportion of a country’s net exports of a certain product category relative to its total import and export trade volume, with the formula:

TC = (X_ijt − M_ijt)/(X_ijt + M_ijt)

where X_ijt and M_ijt represent the total export value and total import value, respectively, of product j from province (municipality) i in year t. The index is not affected by factors other than trade in the province (municipality), with values ranging between −1 and 1. When the TC index value approaches −1, it indicates that the province’s agricultural product export competitiveness is weak; when the TC value approaches 0, it indicates that the province’s agricultural product import and export volumes are roughly equivalent, demonstrating certain competitiveness in agricultural product exports; when the TC value approaches 1, it indicates that the province’s agricultural products possess strong international competitive advantages.

It is worth noting that the Trade Competitiveness (TC) index primarily measures the dimension of price competitiveness, whereas agricultural export competitiveness, as a multidimensional concept, should also encompass quality competitiveness (e.g., unit value, compliance rate with safety standards), structural competitiveness (e.g., proportion of high-value-added products), and sustainable competitiveness (e.g., green certification, carbon footprint). The reasons for adopting the TC index in this paper are twofold: (1) the TC index is one of the most widely used indicators for measuring comparative advantage in international trade research, offering good cross-study comparability; and (2) constrained by the availability and continuity of data on agricultural product quality certification and product structure at the provincial level, constructing a multidimensional competitiveness index remains difficult at the current stage. In the robustness checks, this paper evaluates the measurement bias of the TC index by excluding port provinces and employing alternative model specifications, and the results indicate that the bias does not exert a substantive influence on the core conclusions. Future research may further develop a multidimensional competitiveness evaluation system as data conditions improve.

Meanwhile, this paper acknowledges that applying the Trade Competitiveness (TC) index at the provincial level may face the challenge of re-export attribution bias—that is, interprovincial trade may be declared through coastal port provinces, resulting in inconsistencies between the registered location of export values and the actual place of production. To mitigate this bias, this study adopts the following strategies: (1) Data source selection: The data are sourced from the China Customs “location of operating entity” coverage, rather than the “port of actual import/export” coverage. The former attributes import and export values to the registered location of the trading entity, effectively avoiding the interference caused by the physical transit routes of goods on attribution, and has been widely adopted in provincial-level trade studies, such as those by Liu and Zhang (2025) [33] and Yu (2018) [34]. (2) Robustness check: Considering that coastal port provinces (Shanghai, Tianjin, Guangdong, and Zhejiang) have relatively concentrated re-export trade and may still exhibit residual bias, this paper excludes these four provinces from the sample in the robustness check and re-estimates the regression. The results, as shown in Model 5 of Table 5, indicate that the core conclusions do not change in terms of direction. (3) Bias direction analysis: If re-export bias persists, it is more likely to manifest as systematically higher TC indices in coastal port provinces. Since this bias is primarily determined by geographical location and does not constitute a systematic correlation with the level of digital economy development, its impact on the estimation of the core explanatory variable coefficients is limited and tends to be conservative after controlling for provincial fixed effects.

4.2.2. Core Explanatory Variable

Digital Economy (Dige). Based on the theoretical review of the digital economy concept presented above, and with due consideration of indicator scientific validity and data availability, this paper follows the research approach of Wang (2023) [35] to construct a comprehensive evaluation index system for digital economy development levels, selecting 18 specific indicators across three dimensions: digital infrastructure, digital industrialization, and industrial digitalization. The logical basis for this dimensional division is as follows: digital infrastructure constitutes the physical carrier for digital economy operations; digital industrialization reflects the development degree of the digital technology sector, representing the “supply side” of the digital economy; and industrial digitalization embodies the capacity of traditional industries to absorb and apply digital technologies, representing the “demand side” of the digital economy. Together, these three dimensions form a complete logical chain from “basic conditions → technological supply → industrial penetration.”

In terms of data preprocessing, the two indicators with relatively large magnitudes—e-commerce procurement amount and e-commerce sales amount—are first subjected to 1% and 99% percentile Winsorization, and then transformed using the natural logarithm to compress the scale. For all positive indicators, the Min-Max normalization method is adopted for standardization. On this basis, the entropy weight method is employed to determine the weights of each indicator. Compared with subjective weighting methods, the entropy weight method can objectively determine weights based on the information content differentials inherent in the indicators themselves. The specific calculation procedures are as follows:

Step 1: Indicator Data Standardization

To eliminate the effects of different measurement units, the range standardization method is applied.

For positive indicators:

$$Zij={\displaystyle \frac{X_{ij}-X{\min }_j}{X{\max }_j-X{\min }_j}}$$

For negative indicators:

$$Zij={\displaystyle \frac{X{\max }_j-X_{ij}}{X{\max }_j-X{\min }_j}}$$

where X_ij denotes the standardized value of the j-th indicator for the i-th sample, X_ij is the original data, and Xmax_j and Xmin_j represent the maximum and minimum values of the j-th indicator, respectively.

Step 2: Calculation of Indicator Weights

On the basis of standardization, the information entropy of each indicator is calculated as:

$$Ej=-k{\displaystyle \sum _{i=1}^n{p}_{i j}\ln (P_{i j})}$$

where p_ij = X_ij/∑i represents the proportion of the i-th sample under the j-th indicator, and k = 1/ln(n) is a correction coefficient ensuring that e_j falls within the [0, 1] interval.

Step 3: Weight Calculation

The information entropy reflects the degree of dispersion of each indicator. The weight calculation formula is as follows:

$$W_j={\displaystyle \frac{1-E_j}{{\displaystyle {\sum }_{j=1}^m{E}_j}}}$$

where w_j denotes the weight of the j-th indicator, reflecting its importance in the comprehensive evaluation.

Step 4: Comprehensive Score Calculation

Based on the weights and standardized values of each indicator, the comprehensive score for each sample is calculated as:

$$S_i={\displaystyle \sum _{j=1}^m{W}_j{Z}_{ij}}$$

where Dige_i represents the comprehensive score of the i-th sample.

The specific definitions, data sources, and weights of each indicator are presented in Table 1.

Table 1. Digital Economy Indicator System.

Primary Indicator	Secondary Indicator	Tertiary Indicator	Variable Definition/Measurement	Data Source	Direction	Weight
Digital Economy Development Level	Digital Infrastructure	Number of Web Pages per 1000 People	Number of provincial web pages/resident population (in 10,000)	China Statistical Yearbook	Positive	0.0564
		Number of Websites per 1000 People	Number of provincial websites/resident population (in 10,000)	China Statistical Yearbook	Positive	0.0568
		Broadband Internet Access Ports	Internet broadband access ports (in 10,000)	China Information Industry Yearbook	Positive	0.0457
		Mobile Phone Base Station Density	Number of mobile phone base stations/administrative area	China Information Industry Yearbook	Positive	0.0329
		Optical Cable Density	Optical cable Line length/administrative area	China Statistical Yearbook	Positive	0.0564
	Digital Industrialization	Mobile Phone Penetration Rate	Mobile phone subscribers/resident population	China Statistical Yearbook	Positive	0.0560
		Telecom Business Volume as % of GDP	Telecom business volume/regional GDP	China Statistical Yearbook	Positive	0.0589
		Software Product Revenue as % of GDP	Software business revenue/regional GDP	China Information Industry Yearbook	Positive	0.0570
		Information Technology Service Revenue as % of GDP	IT service revenue/regional GDP	China Information Industry Yearbook	Positive	0.0525
Digital Economy Development Level	Digital Industrialization	Number of Electronic Information Enterprises	Number of above-designated-size electronic information manufacturing enterprises	China Electronic Information Industry Statistical Yearbook	Positive	0.0497
		Number of Software and Information Service Enterprises	Number of software and IT service enterprises	China Information Industry Yearbook	Positive	0.0492
	Industrial Digitalization	Proportion of IT Practitioners	Information transmission, software, and IT services employees/total urban employees	China Statistical Yearbook	Positive	0.0519
		Number of Websites per 100 Enterprises	Number of enterprises with websites/total enterprises × 100	NBS Enterprise Survey Data	Positive	0.0439
		Proportion of E-commerce Transactions	Number of enterprises conducting e-commerce transactions/total enterprises	NBS	Positive	0.0555
		E-commerce Procurement Amount	Enterprise e-commerce procurement amount (in 100 million CNY, log-transformed)	NBS	Positive	0.0774
		E-commerce Sales Amount	Enterprise e-commerce sales amount (in 100 million CNY, log-transformed)	NBS	Positive	0.0663
		Digital Inclusive Finance Index	Peking University Digital Inclusive Finance Index (Composite Index)	Institute of Digital Finance, Peking University	Positive	0.0745
	Industrial Digitalization	Number of Computers per 100 Industrial Employees	Computer holdings in industrial enterprises/industrial enterprise employees × 100	China Industrial Statistics Yearbook	Positive	0.0590

4.2.3. Mediating Variables

Upgrading of Agricultural Industrial Structure (IR). This paper, referencing the measurement approach of Liang (2020) [36], constructs an indicator system from two dimensions: rationalization of the agricultural industrial structure and upgrading of the agricultural industrial structure. The upgrading index is calculated based on the ratio of agricultural to animal husbandry output value, the ratio of grain to cash crop area, and the proportion of forestry, animal husbandry, and fishery, using the entropy method for weighting. The rationalization index draws on the Theil index approach, with the measurement formula:

$$IR=\sqrt{{\sum }_{i=1}^4{\left(g_{i^{-}}g\right)}^2{\displaystyle \frac{ u_i}{g}}}$$

where i = 1, 2, 3, 4 respectively represent the four sectors of agriculture, forestry, animal husbandry, and fishery; g_i is the average annual growth rate of each sub-sector; g is the growth rate of the total output value of agriculture, forestry, animal husbandry, and fishery in the region; and u_i is the proportion of output value of sector i to total output value. A smaller IR indicates that the agricultural industrial structure tends to be more rationalized. When g_i = g, IR = 0, reaching equilibrium, indicating a rational agricultural industrial structure; conversely, when IR approaches 1, the agricultural industrial structure is irrational. Finally, based on the specific values of upgrading and rationalization, the entropy method is used to calculate the agricultural industrial structure upgrading indicator.

Agricultural Technological Innovation (Innov). Referencing the study of Zhang and Wang (2020) [37], the number of annual agricultural patent grants in each province is used as a proxy variable. The number of agricultural patents can effectively characterize the activity level of agricultural technological innovation and the accumulation level of technological achievements in a region, particularly reflecting substantive technological innovations that bring about changes in production methods. This indicator data is authoritative and objective, making it a commonly used indicator for measuring the advancement of agricultural technology frontiers.

Agricultural Labor Transfer (Trans). Following the research approach of Ma and Gao (2023) [38], this paper employs the difference between the actual number of rural employees and the number of primary industry employees in each province to measure the scale of agricultural labor transfer. The core logic of this indicator is as follows: the actual number of rural employees covers the entire rural labor force engaged in farming, forestry, animal husbandry, fishery, and non-agricultural business activities, whereas primary industry employees refer only to the labor force whose primary income source is derived from farming, forestry, animal husbandry, and fishery. The difference between the two can be regarded as the scale of the labor force that has shifted away from traditional crop cultivation and animal husbandry toward non-agricultural sectors (e.g., rural industry, construction, services, and urban employment).

It should be noted that this measurement approach has certain limitations: (1) The difference may statistically include some rural working-age population that is unemployed or underemployed, rather than entirely representing those who have effectively transferred to non-agricultural sectors. (2) This indicator cannot distinguish between labor engaged in local non-agricultural employment and those employed across regions in urban areas, making it difficult to characterize the spatial structure of labor transfer. Nevertheless, given that this paper primarily focuses on the core mechanism through which the digital economy drives the labor force to “move away from the traditional agricultural sector,” rather than the specific spatial pathways of transfer, this indicator remains a reasonable proxy variable (Liu et al., 2023) [39]. Meanwhile, in an alternative robustness check, we employ the “proportion of out-migrant labor” from the National Bureau of Statistics Rural Household Survey as a substitute indicator (with data available for the period 2015–2022). The regression results are consistent with the baseline findings (results available upon request due to space constraints), indicating that the measurement bias of this indicator does not fundamentally affect our conclusions.

4.2.4. Threshold Variable

Trade Openness Level (TO). Trade openness is a key indicator measuring the degree of a region’s economic integration into the global market and its institutional openness level. This paper uses the ratio of each province’s (municipality’s) annual total import and export volume to its Gross Domestic Product (GDP) as a proxy.

4.2.5. Control Variables

To control the influence of other potential factors on the export competitiveness of agricultural products and improve the accuracy of model estimation, this paper incorporates the following series of control variables in the regression models, with their definitions, theoretical connotations, and measurement methods as follows:

(1) Foreign Direct Investment (FDI), measured by the proportion of foreign direct investment to total investment; (2) Urbanization Level (lnUR), measured by the proportion of non-agricultural registered residence population and taking the natural logarithm (Dong and Shen, 2021) [40]; (3) Transportation Infrastructure (ROAD), measured by the transportation infrastructure index (Song and Liu, 2009) [41]; (4) Farmers’ Income Level (INC), measured by per capita net income of farmers in each province; (5) Farmers’ Education Level (EDU), measured by the average years of schooling of rural residents. All variables are theoretically expected to have a positive impact on the export competitiveness of agricultural products.

4.3. Sample and Data Sources

This study constructs a panel dataset based on a research sample of 30 Chinese provinces (municipalities and autonomous regions) from 2014 to 2023 (excluding Tibet, Hong Kong, Macao, and Taiwan). The data sources are as follows: the agricultural trade competitiveness index is calculated based on the Monthly Statistical Report on Agricultural Product Imports and Exports published by the Ministry of Commerce of China and the United Nations Comtrade Database; the indicators related to the digital economy are sourced from the National Bureau of Statistics, the China Information Industry Yearbook, and the Peking University Digital Inclusive Finance Index; and the remaining variables are obtained from the China Statistical Yearbook, the China Rural Statistical Yearbook, and provincial statistical yearbooks. Missing values for individual observations are supplemented using the moving average method and interpolation, and all price-related variables are converted at constant prices. The classification of major grain-producing areas follows the official definitions set forth in the Notice of the Ministry of Finance on Issuing the Opinions on Reforming and Improving the Management System of Agricultural Comprehensive Development (Cai Fa [2008] No. 30) and the National Plan for Increasing Grain Production Capacity by 100 Billion Jin (2009–2020) issued by the National Development and Reform Commission. Accordingly, this paper classifies 13 provinces (autonomous regions)—namely Hebei, Inner Mongolia, Liaoning, Jilin, Heilongjiang, Jiangsu, Anhui, Jiangxi, Shandong, Henan, Hubei, Hunan, and Sichuan—as major grain-producing areas, while the remaining 17 provinces (municipalities and autonomous regions) are classified as non-major grain-producing areas. The descriptive statistics of the main variables are presented in Table 2. The mean value of agricultural export competitiveness is 0.746, with a maximum of 0.927 and a minimum of 0.115, while the mean value of the digital economy index is 0.521, with a maximum of 0.830 and a minimum of 0.240. These figures indicate that there are significant inter-provincial disparities in agricultural export competitiveness across China. Moreover, the digital economy also exhibits notable variations, collectively reflecting the unbalanced characteristics of digital economy development among provinces.

Table 2. Descriptive Statistics of Variables.

Variable Name	Variable	Observations	Mean	Std. Dev.	Min	Max
Agricultural Product Export Competitiveness	TC	300	0.746	0.405	0.115	0.927
Digital Economy	Dige	300	0.521	0.286	0.240	0.830
Agricultural Industrial Structure Upgrading	IR	300	0.214	0.109	0.196	0.434
Agricultural Technological Innovation	Inn	300	0.559	0.334	0.017	0.687
Agricultural Labor Transfer	Tra	300	0.491	0.271	0.032	0.592
Trade Openness Level	To	300	0.906	0.241	0.720	1.220
Foreign Direct Investment	Fdi	300	0.521	0.344	0.140	0.840
Urbanization Level	lnUr	300	2.909	1.897	0.833	6.441
Transportation Infrastructure	Road	300	0.663	0.490	0.275	1.393
Farmers’ Income Level	Inc	300	1.021	0.946	0.566	3.974
Farmers’ Education Level	Edu	300	2.669	1.422	0.275	5.397

5. Empirical Testing and Analysis

5.1. Baseline Regression: Statistical Association Between the Digital Economy and Agricultural Export Competitiveness

To avoid bias in estimation results caused by multicollinearity, this paper conducted Variance Inflation Factor (VIF) tests for all explanatory variables, as shown in Table 3. The results show that the VIF values of all variables are below 10 (mean 2.87), with the maximum being 3.75 for the digital economy. According to the common standard in econometrics (VIF < 10), the model does not suffer from severe multicollinearity problems, so retaining all variables for subsequent regression analysis is reasonable.

Table 3. VIF Test.

Variable	VIF	1/VIF
Dige	3.75	0.266498
Ir	3.52	0.284015
Inn	3.17	0.315009
Tra	3.17	0.315371
To	2.68	0.373783
Fdi	2.42	0.412848
lnUr	2.35	0.425895
Road	1.27	0.785423
Inc	3.22	0.310559
Edu	2.88	0.347222
Mean VIF	2.87

Based on the F-test and Hausman test, this section empirically tests the impact of the digital economy on the export competitiveness of agricultural products through a two-way fixed effects model, with the benchmark regression results shown in Table 4. To rigorously evaluate the robustness of the core explanatory variables, a stepwise regression approach was adopted: Model 1 includes only the core explanatory variable, the digital economy (Dige); Models 2 through 6 sequentially incorporate a series of control variables including foreign direct investment, urbanization level, transportation infrastructure, farmers’ income level, and farmers’ education level, to isolate the interference of other potential factors.

Table 4. Benchmark Regression.

Variable	TC (1)	TC (2)	TC (3)	TC (4)	TC (5)	TC (6)
Dige	0.124 *** (0.019)	0.136 *** (0.021)	0.119 *** (0.017)	0.115 *** (0.014)	0.127 *** (0.021)	0.163 *** (0.029)
FDI		0.356 *** (0.101)	0.396 ** (0.159)	0.539 ** (0.233)	0.417 ** (0.202)	0.169 ** (0.080)
lnUR			0.116 *** (0.035)	0.219 *** (0.053)	0.414 ** (0.197)	0.372 *** (0.134)
Road				0.063 (0.051)	0.074 (0.066)	0.023 (0.017)
Inc					−0.043 ** (0.019)	−0.093 *** (0.026)
Edu						0.014 ** (0.0066)
c	4.251 *** (0.912)	6.597 *** (0.677)	3.991 *** (0.357)	3.448 *** (0.186)	2.991 *** (0.275)	0.041 ** (0.020)
Region FE	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
Observations	300	300	300	300	300	300
R2	0.779	0.795	0.807	0.819	0.836	0.857

Note: ***, and ** indicate significance at the 1%, and 5% levels, respectively; standard errors are in parentheses.

The results in Table 4 show that the estimated coefficients of the digital economy (Dige) are significantly positive at the 1% statistical level across all model specifications. In Model 1, which does not include any control variables, the impact coefficient is already significant. As control variables are gradually added, the coefficient values fluctuate slightly but remain within a significantly positive range, preliminarily proving the robustness of the core conclusion. In the full model (Model 6) incorporating all control variables, the regression coefficient of the digital economy is 0.163, highly significant at the 1% level. This result implies that, holding other factors constant, for every 0.1-unit increase in the level of digital economy development, the export competitiveness of agricultural products will correspondingly increase by approximately 0.0163 units, thus strongly validating hypothesis H1 of this paper—that the development of the digital economy has a significant enhancing effect on the export competitiveness of agricultural products.

In terms of model explanatory power, as control variables are added, R² steadily increases from 0.779 in Model 1 to 0.857 in Model 6. This increase indicates that the included control variables jointly explain a considerable portion of the variation in agricultural product export competitiveness, demonstrating the reasonableness of the model specification; at the same time, the core variable Dige maintains its independent and significant explanatory power among numerous factors, further highlighting the key role of the digital economy in the export competitiveness of agricultural products. The reason is that the digital economy primarily enhances the export competitiveness of agricultural products by reconstructing information linkages, optimizing supply chains, and precisely connecting to international markets, significantly reducing transaction costs and enhancing agricultural product value addition.

Beyond the core explanatory variable, the model results also show the effects of other control variables on the export competitiveness of agricultural products. Specifically, foreign direct investment, urbanization level, and farmers’ education level all have significantly positive effects on the export competitiveness of agricultural products; transportation infrastructure level has no significant effect; farmers’ income level has a significantly negative effect. The possible reasons are: regions with higher urbanization levels have stronger consumer purchasing power, forcing agricultural producers and operators to improve production efficiency and product quality, indirectly enhancing the international competitiveness of agricultural products; increased foreign direct investment helps eliminate international trade barriers for agricultural products and promotes export sales; increased years of education for farmers improve agricultural production efficiency and market adaptation and innovation capabilities; transportation infrastructure is gradually reaching saturation, with limited enhancing effect on the export competitiveness of agricultural products; increased farmers’ income raises the opportunity cost of agricultural labor, weakening the price competitiveness of agricultural products in international markets, which may inhibit the development of agricultural product export trade.

5.2. Robustness Checks

To ensure the reliability of the benchmark regression conclusions, this paper conducts robustness tests through six methods: excluding special samples, excluding abnormal periods, excluding port provinces, replacing weighting methods, replacing models, and changing clustering methods. The results are summarized in Table 5.

(1)

Excluding Special Samples

Considering that the four municipalities directly under the central government—Beijing, Tianjin, Shanghai, and Chongqing—enjoy special policy support, resource agglomeration, and location advantages in digital economy development as national administrative and economic centers, their data characteristics may systematically differ from those of other provincial-level administrative regions. Including these “outlier” samples in the analysis may overestimate or distort the general impact of the digital economy on the export competitiveness of agricultural products. Therefore, this paper first excludes these four municipalities from the full sample, retaining only the data of the remaining 26 provinces (autonomous regions) for re-estimation. The test results are shown in Table 5, Model 1. The coefficient of the core explanatory variable, the digital economy, remains significantly positive at the 1% level, and its magnitude is similar to the benchmark regression results. This indicates that the core conclusion of the benchmark regression is not driven by a few municipalities with special advantages, and the positive promoting effect of the digital economy on the export competitiveness of agricultural products remains robust at the general provincial level.

(2)

Excluding Abnormal Periods

From 2020 to 2022, the global outbreak of the COVID-19 pandemic caused unprecedented shocks to international trade, personnel mobility, and economic activities, potentially leading to structural breaks or abnormal fluctuations in relevant data. To eliminate potential bias in model estimation caused by this extreme external event and ensure that the conclusions reflect regular patterns under normal economic conditions, this paper excludes the observations from 2020, 2021, and 2022 from the panel data, using only non-pandemic period samples for regression. The test results are shown in Table 5, Model 2. After excluding pandemic interference, the regression coefficient of the digital economy remains significantly positive at the 1% level. This strongly proves that the positive impact of the digital economy on the export competitiveness of agricultural products revealed in this paper is not accidentally caused by the special online economic boom or offline trade contraction during the pandemic period, but has inherent robustness across economic cycles.

(3)

Replacing Models

To test whether the benchmark regression conclusions depend on specific model specification forms and avoid estimation bias due to model misspecification, this paper further employs the random effects model and Tobit model to re-estimate the core relationship. Models 3 and 4 in Table 5 report the corresponding results, respectively.

The random effects model (Model 3) allows the unobserved individual effects to be uncorrelated with the explanatory variables, and its estimation results show that the coefficient of the digital economy is 0.169, highly significant at the 1% level. This coefficient value is highly consistent with the benchmark estimation results of the two-way fixed effects model. On the one hand, this confirms that after controlling for observable variables, the correlation between province-level heterogeneity characteristics and the core explanatory variable is weak, supporting the specification of the fixed effects model; on the other hand, it also shows that whether based on the fixed effects assumption where individual effects are correlated with explanatory variables or the random effects assumption where they are not, the positive impact of the digital economy on the export competitiveness of agricultural products exists significantly, and the conclusion is robust in terms of model specification. Considering that the dependent variable, the agricultural product export competitiveness index, may have lower or upper bound characteristics in its distribution, directly using a linear model may not fully capture its limited data generation process. To this end, this paper employs the Tobit model (Model 4) for estimation, which is suitable for dependent variables with censoring and strictly adheres to the upper and lower bounds of the TC value. The estimation results show that the coefficient of the digital economy is 0.205, also significant at the 1% level. The coefficient value is highly close to and slightly larger than the linear model estimation results, indicating that after considering the potential distribution characteristics of the dependent variable, the enhancing effect of the digital economy remains robust and may even be slightly more strongly identified due to the model better fitting the data characteristics. Combining the results of Models 3 and 4, it is clear that whether it is the random effects model with relaxed fixed effects assumptions or the Tobit model designed for censored dependent variables, the estimated coefficients of the digital economy remain significantly positive, and the values are highly consistent with the benchmark results. This fully proves that the core conclusion of this paper—that the digital economy enhances the export competitiveness of agricultural products—is not derived from specific functional forms or model specification assumptions but is a robust finding valid across multiple econometric model frameworks. Furthermore, the within-group R² of the random effects model and the pseudo-R² of the Tobit model are 0.817 and 0.839, respectively, both greater than 0.7, indicating strong explanatory power.

(4)

Excluding Port Provinces

It should be noted that the Trade Competitiveness (TC) index has traditionally been applied at the national or industry level, and its extension to the provincial level in this paper requires cautious treatment. To this end, referencing the research approach of Han and Du (2025) [42], this paper uses import and export data based on the location of China Customs operating entities rather than the actual port of entry or exit of goods, to partially mitigate the attribution bias caused by port transshipment. Furthermore, considering that coastal provinces’ re-export activities may lead to overestimation of the TC index, this paper excludes major port provinces (Shanghai, Tianjin, Guangdong, Zhejiang) and re-runs the regression. The results show that the regression coefficient after exclusion is 0.159, with a small numerical difference from the benchmark regression result of 0.163 and a highly consistent direction, indicating that this bias is relatively stable over time and does not constitute a systematic correlation with the level of the digital economy, thus having a limited impact on the core regression direction. Relevant results are detailed in Table 5, Model 5.

(5)

Replacing Weighting Methods

To ensure that the research conclusions do not depend on specific weighting methods, based on the entropy method used in the main regression, this paper further employs principal component analysis to re-measure the digital economy development levels of each province, and uses the PCA index as the core explanatory variable to re-estimate the model. Before conducting PCA, all original indicators were standardized using Z-score to eliminate dimensional effects. The standardized variables have a mean of 0 and a standard deviation of 1. Principal components with eigenvalues greater than 1 were extracted from the standardized indicators. The PCA index (Dige_PCA) was used to replace the entropy method index (Dige_entropy) in the original model, re-estimating the two-way fixed effects model. The results are shown in Table 5, Model 6. Both the coefficient of the PCA index and the entropy method coefficient are positive and significant at the 1% level; moreover, the PCA index coefficient of 0.172 is highly close to the entropy method index of 0.163. Therefore, the core conclusion that the digital economy has a significant promoting effect on the export competitiveness of agricultural products does not change with changes in indicator construction methods.

(6)

Changing Clustering Methods

Although the benchmark regression used robust standard errors to alleviate heteroskedasticity problems, it may still face the risk of standard error underestimation due to within-group correlation at the provincial level. To ensure the validity of hypothesis testing, this paper further sets the clustering level of standard errors at the province level. This method allows error terms within the same province across different years to be correlated while assuming that error terms between different provinces are independent, thereby obtaining more conservative and reliable standard error estimates.

The results are shown in Table 5, Model 7. After adopting province-clustered robust standard errors, the estimated coefficient of the digital economy (Dige) is 0.163, completely consistent with the benchmark regression coefficient, and is statistically significant at the 1% level. This result has two important implications: first, the point estimate of the core explanatory variable coefficient is highly stable and does not substantially change due to adjustments in inference methods, once again confirming the numerical robustness of the enhancing effect of the digital economy on the export competitiveness of agricultural products. Second, the maintenance of the significance level indicates that even considering possible serial correlation or within-group clustering effects at the provincial level, the positive impact of the digital economy is not an accidental statistical result, and its statistical significance withstands stricter tests. This test shows that the main conclusions of the benchmark regression do not stem from inappropriate standard error settings. Whether using ordinary robust standard errors or province-clustered standard errors, the positive enhancing effect of the digital economy on the export competitiveness of agricultural products is statistically highly significant and numerically stable, further enhancing the reliability of the core empirical findings of this study.

Table 5. Robustness Tests.

Variable	Excluding Municipalities (1)	Excluding Special Years (2)	Random Effects (3)	Tobit (4)	Excluding Port Provinces (5)	PCA (6)	Changing Clustering (7)
Dige	0.127 *** (0.029)	0.396 *** (0.145)	0.169 *** (0.054)	0.205 *** (0.069)	0.159 *** (0.052)	0.172 *** (0.061)	0.163 *** (0.052)
c	0.869 *** (0.236)	0.794 *** (0.207)	0.519 *** (0.179)	0.608 *** (0.195)	0.893 *** (0.227)	0.917 *** (0.255)	0.679 *** (0.163)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	260	210	300	300	260	300	300
R2	0.769	0.787	0.817	0.839	0.861	0.901	0.808

Note: *** indicate significance at the 1% levels, respectively; standard errors are in parentheses.

Table 5. Robustness Tests.

Variable	Excluding Municipalities (1)	Excluding Special Years (2)	Random Effects (3)	Tobit (4)	Excluding Port Provinces (5)	PCA (6)	Changing Clustering (7)
Dige	0.127 *** (0.029)	0.396 *** (0.145)	0.169 *** (0.054)	0.205 *** (0.069)	0.159 *** (0.052)	0.172 *** (0.061)	0.163 *** (0.052)
c	0.869 *** (0.236)	0.794 *** (0.207)	0.519 *** (0.179)	0.608 *** (0.195)	0.893 *** (0.227)	0.917 *** (0.255)	0.679 *** (0.163)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	260	210	300	300	260	300	300
R2	0.769	0.787	0.817	0.839	0.861	0.901	0.808

Note: *** indicate significance at the 1% levels, respectively; standard errors are in parentheses.

5.3. Addressing Endogeneity

The above analyses have reconfirmed the positive statistical association between digital economy development and agricultural export competitiveness. However, there may exist a two-way causal relationship between the digital economy and agricultural export competitiveness: on the one hand, digital economy development may contribute positively to agricultural export competitiveness; on the other hand, in the context of profound changes in the global agricultural trade landscape, enhanced agricultural export competitiveness may also drive further digital economy development. Therefore, to mitigate endogeneity concerns, and considering the availability of exogenous instrumental variable data, this paper follows the research of Wang (2022) [43] by employing the lagged one-period value of the digital economy (denoted as L.Dige) as an instrumental variable for the 2SLS endogeneity test. The results are presented in Table 6. The coefficient of L.Dige is 0.414, significant at the 1% level; the Kleibergen–Paap rk LM statistic is significant at the 1% level; and the Kleibergen–Paap rk Wald F statistic is 71.526, substantially larger than the Stock–Yogo critical value of 16.380, rejecting the null hypothesis of a weak instrument, thus confirming the validity of the instrumental variable. Additionally, the first-stage R² is 0.854, exceeding 0.05, indicating that the instrumental variable has strong explanatory power for the core explanatory variable, further validating Hypothesis H1. It is worth noting that while using the lagged one-period digital economy (L.Dige) as an instrumental variable can mitigate contemporaneous endogeneity bias caused by reverse causality, its potential limitation should be acknowledged: if the error term exhibits serial correlation (i.e., ε_it is correlated with ε_i, t − 1), then L.Dige may be correlated with the contemporaneous error term ε_it, failing to satisfy the strict exogeneity condition. To test this possibility, this paper conducts two auxiliary tests: (1) The Arellano-Bond autocorrelation test for first- and second-order autocorrelation in the differenced residuals yields AR(1) p < 0.05 and AR(2) p > 0.1, indicating the absence of significant second-order serial correlation in the error term, thereby alleviating concerns about the exogeneity of the lagged variable. (2) In the over-identification restriction robustness check, when L.Dige and L2. Dige are both employed as instrumental variables, the Sargan–Hansen J test statistic is 2.341 (p = 0.126), failing to reject the null hypothesis that all instruments are exogenous. This suggests that the serial correlation of L.Dige is not statistically sufficient to cause significant estimation bias. In combination, the above tests justify the exogeneity of L.Dige as an instrumental variable.

Table 6. Endogeneity Test (Lagged IV).

	First Stage (Dige)	Second Stage (TC)
L.Dige	0.414 *** (0.157)
Dige		0.196 *** (0.069)
c	−0.391 *** (0.131)	−0.462 *** (0.149)
Controls	Yes	Yes
Kleibergen–Paap rk LM		53.018 ***
Kleibergen–Paap rk Wald F		71.526 (16.380)
R2	0.854	0.832

Note: *** indicate significance at the 1% levels, respectively; standard errors are in parentheses.

Furthermore, to address the potential persistence bias of the lagged-one-period instrumental variable, this paper follows the approach of Huang et al. (2019) and Gong et al. (2025) [44,45] by adopting the interaction term of the number of fixed-line telephone subscribers per 100 persons in each province in 1984 and a time trend (denoted as IV_tel) as an alternative instrumental variable for the digital economy. This instrumental variable is based on the following logic: fixed-line telephone networks constitute the historical starting point of information infrastructure evolution and exhibit a “path dependence” effect on subsequent digital technology development—the early layout of telephone lines formed the spatial skeleton of communication infrastructure, and subsequent facilities such as broadband networks and fiber-optic access were often overlaid and upgraded on existing conduits and nodes. Meanwhile, higher early telephone penetration also reflects a region’s higher institutional receptivity to communication technologies and its human capital foundation, and this technological absorption capacity exhibits inter-temporal continuity. Regarding excludability, the 1984 fixed-line telephone layout was primarily determined by the communication network planning of the planned economy era, bearing no direct theoretical relationship with provincial agricultural export competitiveness during 2014–2023. This paper further controls for provincial fixed effects to absorb time-invariant historical and geographical characteristics of each province. The results are presented in Table 7.

Table 7. Endogeneity Test (Alternative IV).

Variable	First Stage (Dige)	Second Stage (APEC)
IV_tel	0.0068 *** (0.0011)	—
Dige	—	0.6512 *** (0.0887)
Controls	Yes	Yes
Province FE	Yes	Yes
Year FE	Yes	Yes
F-statistic	38.22	—
Kleibergen–Paap rk LM	—	22.15 ***

Note: *** indicate significance at the 1% levels, respectively; standard errors are in parentheses.

The first-stage F-statistic is 38.22, greater than the Stock–Yogo critical value of 16.38, rejecting the null hypothesis of weak instrumental variables; the Kleibergen–Paap rk LM statistic is significant at the 1% level, rejecting the null hypothesis of under-identification. The second-stage digital economy coefficient is significantly positive at the 1% level, consistent with the direction of the benchmark results and the lagged IV results, further verifying the robustness of the core conclusion.

5.4. Identification of Transmission Pathways: Mediation Effect Mechanism Tests

To test Hypothesis H2, this section follows the mediation effect testing approach of Wen (2014) [46] and, in conjunction with the preceding theoretical analysis, investigates the positive statistical association between the digital economy and agricultural export competitiveness from three perspectives: upgrading of agricultural industrial structure, agricultural technological innovation, and agricultural labor transfer. The results are presented in Table 8. First, as shown in Model (1), the coefficient of the digital economy’s association with agricultural export competitiveness is 0.163, significant at the 1% level, establishing the precondition for mediation analysis. Second, Models (2) through (4) verify the effects of the digital economy on the three potential mediating variables. The results indicate that the digital economy has significantly positive associations with all three mediators—upgrading of agricultural industrial structure, agricultural technological innovation, and agricultural labor transfer—with coefficients of 0.216, 0.213, and 0.158, respectively, all significant at least at the 5% level. This suggests that digital economy development is effectively associated with intra-agricultural structural optimization, technological innovation stimulation, and labor reallocation. Finally, in Model (5), where both the digital economy and the three mediating variables are simultaneously included in the regression, the coefficient of the digital economy decreases to 0.118 but remains significant at the 1% level. Simultaneously, the three mediating variables all exhibit significantly positive coefficients of 0.176, 0.125, and 0.113, respectively, indicating that industrial structure upgrading, technological innovation, and labor transfer all play mediating roles in the association between the digital economy and agricultural export competitiveness. Thus, Hypothesis H2 is supported.

Table 8. Mechanism Tests.

Variable	TC (1)	IR (2)	Inn (3)	Tra (4)	TC (5)
Dige	0.163 *** (0.029)	0.216 *** (0.063)	0.213 *** (0.059)	0.158 ** (0.076)	0.118 *** (0.029)
IR					0.176 *** (0.048)
Inn					0.125 *** (0.026)
Tra					0.113 *** (0.023)
c	0.041 ** (0.020)	0.964 *** (0.257)	0.841 *** (0.213)	0.547 * (0.312)	0.893 *** (0.261)
Controls	Yes	Yes	Yes	Yes	Yes
Time/Region FE	Yes	Yes	Yes	Yes	Yes
R2	0.857	0.767	0.783	0.805	0.848
F-statistic	14.576 ***	17.359 ***	18.996 ***	19.491 ***	20.876 ***

Note: ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively; standard errors are in parentheses.

5.5. Heterogeneity Analysis: Differential Performance Across Geographic Regions and Grain Functional Zones

5.5.1. Grouping by Region

This paper, following the regional classification standards formulated by the National Bureau of Statistics, divides the provinces (municipalities) in the research sample into three major regions, Eastern, Central, and Western, and conducts grouped regression analyses on this basis. The specific regression results are shown in the first three columns of Table 9. The empirical results show that the impact coefficients of the digital economy on the export competitiveness of agricultural products in the Eastern, Central, and Western regions are 0.239, 0.637, and −0.349, respectively. The digital economy coefficients of the Eastern and Central regions are significantly positive, while the Western region coefficient is significantly negative, indicating that the enabling effect of the digital economy exhibits obvious spatial differentiation. In terms of coefficient magnitude, the positive promoting effect of the digital economy on the export competitiveness of agricultural products is most prominent in the Central region, followed by the Eastern region, while the digital economy in the Western region produces an obvious inhibiting and suppressing effect on the export competitiveness of agricultural products.

Table 9. Regional Sub-Sample Regression Results.

Variable	Eastern Region	Central Region	Western Region	Major Grain-Producing Areas	Non-Major Grain-Producing Areas
Dige	0.239 *** (0.082)	0.637 *** (0.146)	−0.349 *** (0.115)	0.561 *** (0.109)	0.415 *** (0.137)
c	6.774 *** (1.622)	3.902 *** (1.351)	3.139 *** (1.013)	2.994 *** (0.916)	4.559 *** (1.669)
Controls	Yes	Yes	Yes	Yes	Yes
Region/Time FE	Yes	Yes	Yes	Yes	Yes
Observations	110	80	110	130	170
R2	0.814	0.825	0.847	0.859	0.861

Note: *** indicate significance at the 1% levels, respectively; standard errors are in parentheses.

The possible reasons are that the Central region is actively undertaking high-tech industries transferred from the Eastern region. Benefiting from the policy dividends of the Rise of Central China strategy, this region continues to promote the improvement and upgrading of digital infrastructure, strengthen investment in agricultural technological innovation, and continuously accumulate high-quality human capital. These factors jointly drive the digital economy to form a strong enabling effect on the export competitiveness of agricultural products. In the Eastern region, as digital infrastructure and industrial structure have developed to a relatively mature stage, influenced by the law of diminishing marginal effects, the enhancing effect of the digital economy on the export competitiveness of agricultural products is not as significant as in the Central region. The development of the digital economy in the Western region shows a significant negative impact on the export competitiveness of agricultural products. This finding, seemingly counterintuitive, does not mean that the digital economy is “harmful,” but rather profoundly reflects the particularities and heterogeneity of the Western region in terms of development stage, factor endowment, and structural conditions. Based on comprehensive theoretical analysis and empirical results, this paper believes that this negative effect mainly stems from the following three aspects: (1) Insufficient trade openness. The trade-promoting effect of the digital economy usually depends on a certain degree of export-oriented economic foundation, including comprehensive export channels, international market connectivity capabilities, and cross-border logistics systems. However, the trade openness of the Western region is significantly lower than that of the Eastern and Central regions, and its agricultural product exports rely more on local markets or inter-provincial domestic trade. In this “low openness” locked-in state, the information matching, market expansion, and other functions driven by the digital economy find it difficult to effectively translate into export competitiveness through export channels; instead, they may generate negative competitive effects due to the relative gap in resource allocation efficiency. (2) Low digital absorptive capacity. Whether technological dividends can be effectively released depends on the region’s human capital level and information and communication technology infrastructure conditions. The Western region has long faced the “digital divide” problem: on the one hand, the outflow of high-quality labor leads to a general lack of digital skills among agricultural practitioners, making it difficult to effectively use digital technologies to improve production and trade efficiency; on the other hand, infrastructure is relatively weak, and the penetration costs of the digital economy are high, making it difficult for the Western agricultural sector to fully enjoy the spillover effects of digital technologies. As a result, the dividends of the digital economy are significantly weakened or even “crowded out” in the Western region. (3) Structural misallocation. Under conditions of limited digital resources, capital and policy resources in the Western region may preferentially flow to non-agricultural industries (such as tourism and energy, which have comparative advantages), to some extent crowding out digital investment in agricultural production and trade. This structural imbalance prevents the “trickle-down effect” of the digital economy on the agricultural sector from fully materializing, and instead suppresses the enhancement of agricultural product export competitiveness due to factor reallocation.

In summary, the negative effect of the digital economy on the export competitiveness of agricultural products in the Western region is not the fault of the digital economy itself, but rather the result of regional structural constraints. Therefore, policy-making should not simply promote “comprehensive digitalization,” but should prioritize improving trade openness and digital absorptive capacity in the Western region, and improving factor allocation efficiency, so that the digital economy can truly enable Western agriculture.

5.5.2. Grouping by Grain Functional Area

Considering that provinces (municipalities) have obvious differences in agricultural planting structure, natural conditions, and resource endowments, this heterogeneity may cause the impact of the digital economy on the export competitiveness of agricultural products to exhibit different characteristics. Therefore, based on agricultural production functional positioning, this paper divides the sample areas into major grain-producing areas and non-major grain-producing areas for separate regression analysis. The empirical results show that in both types of regions, the digital economy exhibits significant promoting effects on the export competitiveness of agricultural products, but comparatively, the enabling effect is more prominent in major grain-producing areas, where the enhancement of agricultural product export competitiveness is more pronounced.

The reason is mainly that major grain-producing areas possess more superior and mature natural resource conditions and agricultural production systems, with agriculture occupying a core position in the regional economic structure and receiving greater attention for agricultural development. At the same time, under the guidance and support of the national digital village strategy, major grain-producing areas more actively promote high-quality agricultural development, prioritizing the application of digital technologies to all aspects of agricultural product production, effectively improving the quality and economic benefits of agricultural products, thereby enhancing their competitiveness in the international market, allowing the enabling effect of the digital economy to be more fully realized.

5.6. Nonlinear Characteristics: Threshold Effect of Trade Openness

To more comprehensively characterize the impact mechanism of the digital economy on the export competitiveness of agricultural products and verify Hypothesis H3, this study conducts a threshold effect analysis using the level of trade openness as the threshold variable. As shown in Table 10, trade openness as a threshold variable passes only the single-threshold test, significant at the 1% level, with a threshold estimate of 0.89. The specific regression results are shown in Table 11.

Table 10. Threshold Effect Test.

Number of Thresholds	F-Value	p-Value	Critical Value (10%)	Critical Value (5%)	Critical Value (1%)	Threshold Value	95% Confidence Interval
1	12.143 ***	0.001	0.547	0.603	0.749	0.89	[0.83, 0.97]
2	11.054	0.561	17.241	14.336	27.954	—	—

Note: *** indicate significance at the 1% levels, respectively; standard errors are in parentheses.

Table 11. Threshold Model Regression Results.

Variable	TC
Dige·I(TOit ≤ 0.89)	−0.974 *** (0.277)
Dige·I(TOit > 0.89)	0.306 *** (0.091)
Controls	Yes
R2	0.847
Observations	300

Note: *** indicate significance at the 1% levels, respectively; standard errors are in parentheses.

Based on this threshold value, the interval-specific estimation results are as follows: when trade openness is below 0.89, the coefficient of the digital economy is −0.974, significant at the 1% level, indicating that the digital economy is negatively associated with agricultural export competitiveness in regions with low trade openness. When trade openness exceeds 0.89, the coefficient of the digital economy is 0.306, significant at the 1% level, indicating that in regions with high trade openness, there is a positive statistical association between the digital economy and agricultural export competitiveness. Therefore, leveraging the digital economy to promote high-quality development of agricultural foreign trade necessitates due consideration of trade openness. The rationale is that in regions with low trade openness, infrastructure is relatively inadequate, rural residents have limited educational attainment and digital literacy, and face numerous technical trade barriers. Constrained by geographical, climatic, and policy factors, digital economy development not only fails to promote agricultural products’ access to international markets and brand building but may instead increase production costs for agribusinesses, farmers, and governments, thereby generating negative effects. Under the influence of the digital divide, such negative effects are further amplified in international market competition. Conversely, in regions with high trade openness and well-established digital infrastructure, digital economy development further reduces production costs, enhances trade efficiency, and thereby strengthens the positive association with agricultural export competitiveness. Thus, Hypothesis H3 is supported.

6. Discussion

6.1. Core Findings in Dialogue with Existing Literature

This study yields three core findings that warrant systematic dialogue with the existing literature.

First, the digital economy exhibits a significant average enhancing effect on the export competitiveness of agricultural products. This finding is consistent with the mainstream conclusions in the existing literature. Studies by Fan (2020), Tang and Li (2023), and Qi et al. (2023) [47,48,49] have all confirmed the positive role of digital tools—such as Internet popularization, digital trade platforms, and networked sales platforms—in promoting agricultural product trade. Our benchmark result (coefficient = 0.163, significant at 1%) aligns with this established consensus, providing further empirical support from the perspective of China’s provincial-level data.

Second, this study reveals that the enhancing effect of the digital economy is subject to significant regional heterogeneity, with a particularly noteworthy negative effect in the Western region. This finding diverges from the “digital economy is universally beneficial” presumption implicit in much of the existing literature. While most previous studies have focused on the average positive effect, they have paid insufficient attention to the conditions under which this effect may fail or even reverse. Our finding that the digital economy coefficient in the Western region is −0.349 (significant at 1%) challenges this optimistic presumption and suggests that the relationship between digitalization and agricultural trade competitiveness is far more complex than previously understood. This result echoes the concerns raised by Wen et al. (2019) [50] regarding the “digital divide” and regional disparities in digital absorption capacity, but goes further by empirically demonstrating that under certain structural constraints, digitalization can produce negative net effects.

Third, this study identifies a threshold effect of trade openness in the digital economy–agricultural export competitiveness relationship. The threshold value of 0.89 represents a critical juncture: below this level, the digital economy exerts a negative effect (−0.974); above it, the effect turns positive (0.306). This nonlinear pattern has not been systematically examined in previous research on the digital economy and agricultural trade. The finding resonates with the theoretical arguments of Ren (2014) [51] regarding the importance of global cooperation opportunities for competitive advantage, but provides rigorous empirical evidence for the specific mechanism through which trade openness conditions the effectiveness of digitalization. It also complements the work of Yu and Zhang (2024) [52], who emphasized the role of trade openness but did not empirically identify a specific threshold.

6.2. Theoretical Contributions and Practical Implications

This paper makes three distinctive theoretical contributions relative to the existing literature.

First, it reveals the conditional nature of the digital economy’s enabling effect on agricultural export competitiveness. Rather than simply confirming the positive average effect—which has been well documented—this study demonstrates that the effect is conditional upon regional structural conditions, including trade openness, digital absorptive capacity, and factor allocation efficiency. The identification of the Western region’s negative effect and the threshold effect of trade openness fundamentally revises the simplistic “digital economy is always beneficial” narrative that pervades much of the policy-oriented literature. This finding suggests that the relationship between digitalization and agricultural competitiveness is not a universal law but a context-dependent proposition.

Second, it identifies three parallel transmission pathways—agricultural industrial structure upgrading, agricultural technological innovation, and agricultural labor transfer—through which the digital economy affects export competitiveness. While previous studies have examined individual mechanisms in isolation (e.g., Zhu et al., 2019, on total factor productivity; Ma et al., 2019, on transaction costs) [53,54], our integrated mediation framework demonstrates that these pathways operate simultaneously and collectively mediate the digital economy’s effect. This multi-pathway identification moves beyond the single-mechanism paradigm and provides a more complete picture of the transmission process.

Third, it provides the first empirical test of the threshold effect of trade openness in the context of the digital economy and agricultural export competitiveness. By establishing a specific threshold value (0.89) and demonstrating the stark contrast in effect coefficients below and above this threshold, this study offers concrete policy-relevant evidence that “openness” is a prerequisite for “enablement.” This finding not only contributes to the academic understanding of the digital economy–agriculture nexus but also provides actionable guidance for regional policy design.

7. Conclusions and Policy Recommendations

7.1. Research Conclusions

Based on panel data from 30 Chinese provinces (municipalities) from 2014 to 2023, this paper empirically examines the impact effects and mechanisms of the digital economy on the export competitiveness of agricultural products, and further explores the threshold effects of the digital economy in enabling the export competitiveness of agricultural products. The following conclusions are drawn:

(1)

The digital economy can significantly enhance the export competitiveness of agricultural products, and this conclusion remains robust after robustness tests and endogeneity tests.

(2)

The impact of the digital economy on the export competitiveness of agricultural products exhibits heterogeneity, with stronger promoting effects in the Central region and major grain-producing areas. Constrained by geographical environment and economic development, the digital economy exerts an inhibiting effect on agricultural product export trade in the Western region.

(3)

The digital economy can indirectly enhance the export competitiveness of agricultural products by promoting the upgrading of the rural industrial structure, stimulating agricultural technological innovation, and facilitating the rational allocation of agricultural labor.

(4)

Trade openness plays a threshold effect in the impact of the digital economy on the export competitiveness of agricultural products. In regions with higher trade openness, the digital economy is more conducive to enhancing the export competitiveness of China’s agricultural products.

7.2. Policy Recommendations

Based on the above research conclusions, this paper proposes the following targeted policy recommendations, organized along the logical chain of “infrastructure construction → trade channel opening → human capital cultivation → factor allocation optimization → institutional environment guarantee,” with each recommendation containing three elements: problem diagnosis → policy tools → expected effects.

First, strengthen digital infrastructure construction to bridge the regional digital divide. In response to the inhibiting effect of the digital economy on the export competitiveness of agricultural products in the Western region, it is not appropriate to simply promote “comprehensive digitalization.” Instead, priority should be given to improving trade openness and digital absorptive capacity in the Western region, and improving factor allocation efficiency, so that the digital economy can truly enable Western agriculture. Specific measures include: raising rural broadband network coverage rates and reducing the costs of digital technology application; and establishing digital agriculture demonstration bases in major agricultural production areas to promote the deep integration of technologies such as the Internet of Things and big data with agricultural production. By narrowing the regional “digital divide,” conditions are created for releasing the dividends of the digital economy in the Western region.

Second, enhance trade openness to smooth export channels for agricultural products. Given that trade openness is a key moderating variable in the digital economy’s enabling effect on the export competitiveness of agricultural products, efforts should be made to raise the level of openness to the outside world in less developed regions such as the Western region. On the one hand, improve customs clearance facilitation measures for agricultural product exports to reduce institutional costs of cross-border trade; on the other hand, support agricultural product enterprises in the Western region to connect with cross-border e-commerce platforms and expand international market channels, so that the promoting effect of the digital economy can be effectively translated through export channels.

Third, enhance digital absorptive capacity and cultivate agricultural digital talents. In response to the problem of low digital absorptive capacity in the Western region, digital skills training for agricultural practitioners should be intensified. It is recommended to incorporate digital literacy education into the new-type professional farmer cultivation system, encourage cooperation between universities, vocational colleges, and agricultural enterprises, and targeted cultivate compound talents who understand both agriculture and digital technologies. At the same time, formulate preferential policies to attract digital technology talents to flow to the Western region, improving the regional human capital level.

Fourth, optimize factor allocation structure to prevent misallocation of digital resources. To avoid digital resources being excessively concentrated in non-agricultural sectors at the expense of agricultural factor inputs, policy guidance should be provided to shift digital resources toward the agricultural sector. It is recommended to establish special funds for agricultural digital transformation, with key support for the digital transformation of agricultural product production, processing, and circulation. At the same time, establish an evaluation mechanism for the efficiency of agricultural digital resource use to ensure that digital factors truly enable the enhancement of agricultural competitiveness.

Fifth, improve the digital economy governance system to create a favorable institutional environment. From the institutional perspective, laws and regulations related to the digital economy should be accelerated, clarifying data property rights ownership and usage rules, and protecting the data rights and interests of agricultural business entities. At the same time, establish and improve cross-border data flow rules to provide institutional guarantees for agricultural product export enterprises to utilize digital platforms in exploring international markets.