Impact of Digital Infrastructure on Farm Households’ Scale Management

Abstract

The construction and development of digital infrastructure have emerged as a crucial indicator of national competitiveness, which holds significant importance in driving the sustained growth of the national economy and the comprehensive advancement of society. This paper explores the impact of digital infrastructure on farm households’ scale management, aiming to reveal the role and potential of digital technology in agricultural modernization. Additionally, it seeks to offer a scientific foundation for the government to formulate agricultural policies and advance agricultural modernization. Using the OLS (Ordinary Least Squares) model, moderating effect model, and other methods, this study investigates how digital infrastructure affects farm households’ scale management based on micro-level research data of 2510 farm households from the CRRS (China Rural Revitalization Survey). The following conclusions are drawn: Firstly, the enhancement of digital infrastructure can motivate farm households to expand the land management area and increase the unit output of land. Secondly, farm households’ digital literacy positively moderates the effect of digital infrastructure on their land unit output; moreover, digital skills training for farm households exhibits a positive moderating effect on the influence of digital infrastructure on their management area. Finally, there is a heterogeneity in the impact of digital infrastructure on farm households’ scale management. Specifically, the promotion of farm households’ scale management is stronger in plain areas and weaker in hilly and mountainous areas; stronger for middle-aged and older and small-scale farm households; and weaker for youth groups and large-scale farm households. Based on this, this paper suggests increasing the investment in digital infrastructure construction, improving farm households’ digital literacy, carrying out digital skills training, and formulating differentiated regional policies for reference.

1. Introduction

In the current era where globalization and informatization are intertwined, digital infrastructure has become the core carrier to drive the development of the digital economy, and its popularity is directly related to the balance of regional development and national competitiveness [1]. The Chinese government attaches great importance to the digital transformation of rural areas and has continued to invest in the construction of rural digital infrastructure through a series of policies and plans, such as the Outline of the Digital Rural Development Strategy (2019) and the 14th Five-Year Plan for the Development of the Digital Economy (2021). By the end of 2023, China had built the world’s largest 5G network, the number of rural broadband access users exceeded 190 million, the informatization rate of agricultural production reached 26.5%, and the e-tail sales of agricultural products exceeded 580 billion yuan. This series of achievements has laid the technical foundation for agricultural modernization, and also heralds its far-reaching impact on agricultural production and management.

Academics generally agree that digital infrastructure positively contributes to rural revitalization agricultural production [2,3]. In terms of efficiency improvement, precision agriculture, agricultural digital twin simulation systems, and the predictive functions of machine learning can improve the efficiency and rationality of agricultural factor allocation, reduce resource wastage and production costs, and thus increase agricultural production and the sustainability of agricultural development [4,5,6]. Simultaneously, applying digital infrastructure promotes the change of agricultural production mode and the development of technology-intensive modern agriculture [7,8]. In terms of supply chain reconstruction, digital infrastructure can bridge the “digital divide” between urban and rural areas and lead to the digitization of the agricultural supply chain [9]. It can also make the agricultural supply chain more efficient and stable by enhancing the synergy of agricultural production, processing, and marketing [10,11]. From a different perspective, it also improves the transparency of agricultural prices, helps farm households obtain fairer market prices, and minimizes unnecessary economic losses due to farm households’ bargaining power disadvantage [12]. Nevertheless, some research reveals potential contradictions. Lioutas et al. (2021) emphasized that rapid digitalization in rural agriculture has unintended environmental consequences [13]. In particular, the operation of energy-intensive facilities using smart agriculture platforms poses a challenge to the green development of agriculture [14,15]. Recent research by Liu et al. (2024) emphasizes that the spread of digital infrastructure significantly reduced grain production and sown area [16]. The mechanism lies in the fact that the Internet has enhanced the flow of information in the labor market, prompted farmers to switch to non-farm employment, and increased the cultivation of cash crops. Accordingly, the complex role of digital infrastructure in agricultural production is highlighted, which can improve quality and efficiency through technological upgrades, or impact the environment and food security due to changes in resource allocation.

Adam Smith’s concepts of rising compensation and the division of labor are the theoretical foundations of economies of scale, which are the main route to agricultural modernization [17]. Traditionally, the study of scale economy has mainly focused on the level of manufacturers and firms, which has been greatly enriched by the theoretical contributions of scholars [18,19,20]. Scale management has also received academic attention over the years, resulting in a number of studies on the dynamic measurement of scale economy in agriculture within a certain country or region [21,22]. In recent years, some studies have discussed the scale economy in agriculture in terms of the farm households’ scale management and have proposed a series of factors that may affect the scale management of farm households, such as social capital, Agricultural Socialized Services (ASS), market participation, and agricultural mechanization [23,24,25,26].

Digital technologies such as the Internet and big data have promoted the development of precision agriculture and smart agriculture, as well as the collection and dissemination of agricultural data and information, providing support for continuously empowering the large-scale development in the agricultural sector [27,28,29]. Digital infrastructure, as a physical vehicle for data collection and analysis, has a significant role in scaling up agricultural management. Specifically, digital infrastructure, with its more efficient rate of information flow and technological innovation, promotes the development of agricultural productivity and production efficiency, which can effectively increase farm households’ income [30,31]. This can increase farm households’ inputs in expanding agricultural reproduction, thus increasing the agricultural scale management [32,33]. Furthermore, with the extensive construction and application of digital infrastructure in rural areas, farm households would be trained in relevant digital skills, and their digital literacy would be significantly enhanced, all of which could amplify the positive impact of digital infrastructure on farm households’ scale management.

Existing studies show the positive effects of digital infrastructure but often lack empirical validation and ignore regional and moderating factors. First, the inadequacy of empirical validation and path analysis. Although it is generally agreed that digital infrastructure promotes the expansion of land scale management, this conclusion is mainly based on qualitative theoretical analysis. It lacks sufficient verification at the empirical level [5]. And its specific impact path and mechanism still need to be revealed. Second, the singularity of measurement. Most studies tend to directly measure the scale of agricultural management through the land management area, such as the average household cultivated area, the per capita cultivated area of rural households, and the per capita sown area of crops [34,35]. In recent years, some studies have also chosen to reflect the agricultural scale management by calculating the logarithm of the ratio of cultivated land area to the number of agricultural employees [36]. However, these methods are still limited to a single dimension of the land management area, ignoring its reflection in the average yield and other areas of “production efficiency”. Third, regional analysis and resource endowment heterogeneity are missing. Some studies have analyzed the agricultural scale management in specific regions [36], but there are fewer studies on cross-regional development and resource endowment heterogeneity. Given the problem of unbalanced and insufficient development in China, there is a need to systematically explore the differential impacts of the development status and resource endowment of different regions on agricultural scale management from the perspective of coordinated regional development.

Consequently, this paper analyzes the impact of digital infrastructure on agricultural scale management and its internal mechanism by utilizing a linear regression model, moderating effects model, 2SLS regression, and other methods with 2150 research data obtained from 10 randomly selected sample provinces in eastern, central, western, and northeastern China, and proposes the following marginal contributions: Firstly, this study constructs a transmission path through which digital infrastructure affects agricultural scale management via online sales of agricultural products, and fully illustrates the moderating role of digital literacy accumulation, public affairs participation, and digital skills training. This not only empirically demonstrates the impact of digital infrastructure on agricultural scale management, but also improves the mechanism of this impact, filling the gap in research on agricultural scale management. Furthermore, it provides a theoretical reference for the application of moderating effects in subsequent related studies. Secondly, this paper provides a new insight on measuring the agricultural scale management. It comprehensively assesses agricultural scale management from the dual dimensions of total quantity and efficiency, which provides specific research ideas for the subsequent related studies. Thirdly, this paper analyzes the heterogeneity of the impact of digital infrastructure on agricultural scale management in the region and different resource endowments. This not only provides ideas for subsequent studies on heterogeneity but also responds to the United Nations’ Sustainable Development Goals and China’s regional coordinated development strategy, offering insights to address regional development imbalances.

Based on the above, this study aims to answer the following questions: How does digital infrastructure influence the expansion of agricultural scale management among rural households? What role do digital literacy and digital skills training play in this relationship?

2. Theoretical Analysis and Research Hypotheses

2.1. The Direct Effects of Digital Infrastructure on the Farm Households’ Scale Management

The application and innovation of digital technology have significantly improved the precision and efficiency of agricultural production [37]. The introduction of smart agricultural equipment and the integration and application of big data, Internet of Things and other digital technologies have promoted the transformation of agriculture from traditional extensive management to refined and smart management. It not only optimizes the allocation of resources but also significantly improves the total factor productivity of agriculture [38], allowing for farm households to obtain higher economic output with limited resources, thus laying a solid foundation for expanding land scale management [39]. Improving agricultural total factor productivity not only signals the vitality of agricultural technological innovation but also accelerates the innovation of agricultural management models, creating favorable conditions for forming land scale management. Furthermore, the rise of digital inclusive finance provides a new path for farm households to break traditional financial constraints. The traditional financial system in agriculture is a kind of credit constraint, seriously restricting the ability of farm households to obtain financial support and expand the scale of management. The improvement to digital infrastructure, especially the popularization of digital inclusive finance, effectively reduces the transaction costs of rural financial services and broadens the financing channels of farm households [40]. Digital inclusive finance not only improves the financial accessibility of farm households but also alleviates the problem of information asymmetry through technical means such as big data, promotes the smooth transfer of land [41], and provides a strong financial guarantee for farm households to expand their land scale management [42]. In addition, digital infrastructure further promotes the expansion of agricultural scale management by optimizing the allocation of agricultural production factors. Digital construction has enabled agricultural production to grasp key information such as market demand and climate change more accurately and has promoted the upgrading and transformation of agricultural industrial structures [43,44]. The application of digital technology improves the agricultural decision-making scientificity and the level of smart management, reduces production costs, and enhances the efficiency of agricultural management [45,46]. Through digital platforms, farm households have better access to market dynamics, technical guidance, and climate data to make more precise and efficient production decisions, enhancing agricultural production’s overall competitiveness and sustainability [47,48].

H1. 

Digital infrastructure can promote the expansion of farm households’ land scale management.

2.2. The Moderating Effects of Digital Literacy

With the rapid development of information technology, the application of digital infrastructure in agriculture has become increasingly widespread, providing unprecedented opportunities for farm households. However, simply having advanced digital infrastructure is not enough to ensure that farm households can fully utilize their potential to realize the expansion of their scale management. Farm households with high digital literacy, due to their profound understanding and efficient application of digital technologies, are better able to seize these opportunities. Generally equipped with strong capabilities in understanding and applying digital technologies, they can make better use of tools such as the Internet, Internet of Things, and big data analysis to properly solve the technical problems faced in modern agricultural production [49]. With the wide application of digital technology in agriculture, the significance of digital infrastructure for farm households to increase agricultural income and production is expanding, which puts forward higher requirements for farm households’ digital literacy. As farm households’ digital literacy increases, it can amplify the efficacy of digital infrastructure in solving the information asymmetry problem that farm households have long suffered from in agricultural production [50]. This can help farm households in many aspects, such as preventing pests and diseases, accessing more advanced agricultural technologies, reducing the acquisition costs of related production materials, adapting to climate change, and accessing quality financial credit services [4,5,51,52,53]. The expansion of land scale management, however, requires financial support, sufficient land size, and long-term stability in agricultural production, which is similar to the effect of digital literacy on farm households’ gains in agricultural production. More importantly, digital literacy also promotes innovation in distributing and marketing agricultural products. Highly digitally literate farm households can utilize online platforms for online sales, live streaming, and smart agriculture promotion, thereby broadening sales channels, increasing product awareness, and adding value to production [54]. This can encourage farm households to increase their income, further incentivizing farm households to expand their land scale management with the help of digital infrastructure and the data elements and digital technologies it contains. Based on the above analysis, this paper proposes hypothesis 2:

H2. 

Digital literacy plays a positive moderating role in the impact of digital infrastructure on the expansion of land scale management.

2.3. Moderating Effects of Digital Skills Training

With the continuous advancement of digital infrastructure construction, the application of derivative products and services based on digital technology in agriculture is becoming increasingly widespread, which puts forward higher requirements for the digital skills of farm households. In this context, digital skills training has emerged as a crucial pathway to enhance farmers’ digital capabilities and facilitate the digital transformation of agriculture. Through targeted teaching content and hands-on practice, such training helps farmers gain in-depth understanding of the principles, functions, and application scenarios of digital technologies and services. This training approach based on practical experience can significantly improve farmers’ cognitive level and digital technology application ability, while enhancing their recognition of and willingness to learn about a series of agricultural digital infrastructure rooted in digital agricultural technologies [55,56]. This can increase farm households’ frequency of digital infrastructure use so that the impact of digital infrastructure on expanding land scale management can be fully realized. Moreover, digital skills training can cultivate farm households with high digital skills and digital literacy, and such farm households can promote the effectiveness of digital technologies to other farm households, which increases the overall acceptance of digital infrastructure among farm households [57]. Rural areas, as traditional “acquaintance societies”, are highly influenced by the experiences and advice of family members, friends, neighbors, and other farm households [58,59]. Therefore, digital basic training plays a significant role in increasing overall farm household acceptance of digital infrastructure, which can further amplify the positive moderating role of digital infrastructure in the impact of land scale management expansion. Through digital skills training, farm households will have a deeper understanding of the great potential of digital technology in improving agricultural productivity, reducing production costs, and broadening marketing channels. This cognitive shift can prompt farm households to more actively adopt a range of digital agriculture and smart agriculture production tools driven by digital infrastructure [60,61]. Farm households will gradually believe that using digital technologies can bring tangible income-increasing effects, thereby enhancing their confidence and motivation to expand land scale management based on digital infrastructure construction. Digital skills training is not only an enhancement in farm household capability but also a key component in promoting the digital transformation of agriculture. As more and more farm households master and apply digital technologies, agricultural production methods will undergo profound changes, and the efficiency and quality of agricultural production can be significantly improved.

H3. 

Digital skills training plays a positive moderating role in the impact of digital infrastructure on the expansion of land scale management.

In summary, Figure 1 shows the theoretical framework of this study.

3. Research Design

3.1. Data Sources

The data in this paper are derived from the China Rural Revitalization Survey (CRRS) released by the Institute of Rural Development, Chinese Academy of Social Sciences in 2020.

Based on the principle of multi-level random sampling and considering factors such as economic development level, regional location, and agricultural development status, the CRRS randomly selected 10 sample provinces in the eastern, central, western, and northeastern regions of China according to a proportion of 1/3 of the number of provinces in each region. These include Zhejiang, Shandong, and Guangdong in the eastern region; Anhui and Henan in the central region; Guizhou, Sichuan, Shaanxi, and Ningxia Hui Autonomous Region in the western region; and Heilongjiang in the northeastern region. Within the selected 10 provinces, the CRRS further deepened the sampling process. First, all counties (cities, districts) were divided into 5 equal groups according to their per capita GDP levels, with 1 county (city, district) randomly selected from each group to ensure a balanced geographical distribution. Second, in each county (city, district), 3 towns were selected using a similar random sampling principle, and in each town, 1 administrative village with higher economic development level and 1 with lower economic development level were, respectively, selected. Finally, based on the list of rural households provided by village committees, an equal-interval sampling method was used to select 12 to 14 rural households in each administrative village to ensure the representativeness and diversity of the sample. The selected samples cover 50 counties (cities, districts), 150 towns, and 300 administrative villages, with a baseline survey sample size of more than 3800 households and over 15,000 people, on which follow-up investigations were carried out. After removing outliers, addressing missing values, and cleaning the data, a total of 2510 valid rural household samples were finally used.

3.2. Variable Selection

3.2.1. Explained Variable

The explained variable in this paper is agricultural scale management. Agricultural scale management involves the combination ratio of production factors, that is, the rationality of the scale’s internal structure and the number of scopes (the size of the occupied space). It is an important indicator of the scale and efficiency of agricultural production. This paper draws on the practice of Cui and Liu (2024) to assess the agricultural scale management from the dual dimensions of total agricultural management and efficiency, which is measured by the unit output (tons/hectare) at the efficiency level, and by the land management area (hectares) at the total level [62]. Then, the two variables are logarithmically processed to reduce the range of absolute values of the data and eliminate heteroskedasticity.

3.2.2. Core Explanatory Variable

The core explanatory variable in this paper is digital infrastructure. The digital infrastructure of a farm household refers to the digitized equipment and network facilities equipped in the farm household for information access, communication, production, and living. These facilities include, but are not limited to, broadband networks, smart phones, computers, smart TVs, and so on, which provide farm households with more convenient and efficient information access and communication. Based on the research of Hu et al. (2023), this paper also refers to the 2020 China Rural Revitalization Questionnaire (CRRS) and characterizes the responses to the question “What is the condition of the network at your home?”, which measures the network condition of the farm households, reflecting the digital infrastructure of the farm households at the micro level [1]. After that, this paper treats the answer to this question as a dummy variable, assigning a value of 1 if the answer is “perfect” and 0 otherwise.

3.2.3. Moderating Variables

The moderating variables in this paper are farm households’ digital literacy level and digital skills training. For the farmer group, their digital literacy is mainly reflected in the comprehensive ability to obtain information, integrate information, participate in community collaboration, and apply basic digital technologies by using basic digital tools represented by smartphones in daily production and living scenarios. To achieve an accurate and reasonable assessment of farmers’ digital literacy level, this study integrates the core competency dimensions of the EU DigComp 2.2 framework and the research of Shatila et al. (2025) [63,64]; constructs an indicator system from three aspects: information and data literacy, communication and collaboration, and use of digital technology; and uses the CRITIC weighting method for empowerment. The specific measurement items and weights are shown in the

Table 1. The CRITIC weight index system of farmers’ digital literacy.

Primary Indicator	Secondary Indicator	Measurement Items	Weights
Farmers’ digital literacy	Information and data literacy	Do you use a 4G/5G mobile phone? (yes = 1; no = 0)	0.174
		How easy it is to get information through a mobile phone or Internet? (1 = difficult; 2 = normal; 3 = easy)	0.127
		How timely is it to obtain important information through mobile phones? (1 = not timely; 2 = average; 3 = timely)	0.129
		Can network information meet daily needs such as production and life? (1 = Completely can not meet; 2 = Not very meet; 3 = average; 4 = basically meet; 5 = completely meet)	0.119
	Communication and collaboration	Do you often communicate with other villagers about important public affairs through WeChat groups? (1 = Never; 2 = rarely; 3 = sometimes; 4 = often)	0.163
	Use of digital technology	Do you use mobile payment to buy agricultural products such as pesticides and fertilizers? (yes = 1; no = 0)	0.284

.

In exploring the impacts of digital infrastructures, digital device training can effectively improve farmers’ ability to use digital technology and have a direct impact on the role of digital infrastructure in promoting farm household scale management [65]. In this paper, digital device skills training is characterized by the question “Have you received computer or smart phone Internet access training?” which is 1 if you have received it, and 0 if you have not.

3.2.4. Control Variables

In order to enhance the explanatory strength and predictive accuracy of the model on the phenomenon, to ensure the robustness of the model results, and to be closer to the actual situation, a total of 13 variables are selected as the control variables of this paper from the three aspects of the individual characteristics, family characteristics, and village characteristics of the primary decision-makers in agricultural production. Drawing on Xu et al.’s (2019) study, four variables of gender, age, marital status, and education level were selected as control variables regarding household head characteristics and endowment characteristics [66]. Drawing on Sina et al.’s (2019) study, we selected four variables: the number of agricultural laborers, the share of agricultural income, whether to join cooperatives, and whether to transfer land [67]. Drawing on Chen and Wang et al.’s (2019) study at the level of village characteristics, four variables were selected: distance to the county town, poverty-stricken villages, number of village households, and e-commerce service stations or product resale points in the village [68]. Finally, the definition of each variable used in this paper is shown in

Table 2. Variable definition and assignment.

Variable	Variable Name	Definition and Assignment	Mean	S.D.
Explained variables	Land management area	Natural logarithm of land area (hectares)	1.787	4.998
	Unit output	Unit output (tons/ha)	6.540	9.672
Core explanatory variables	Digital infrastructure	How good is the condition of the internet at your home (1 = very good; 0 = other)	0.433	0.496
Moderating variables	Digital literacy	Calculate using CRITIC weighting method	0.606	0.243
	Digital skills training	Have you received training in accessing the Internet via computer or smart phone (1 = yes; 0 = no)	0.081	0.272
Household head characteristics variables	Gender of household head	1 = male; 0 = female	0.949	0.220
	Age of household head	Actual age (years)	4.060	0.185
	Marital status of household head	1 = married; 2 = unmarried; 3 = divorced; 4 = widowed	1.143	0.596
	Educational level of household head	1 = Not attending school; 2 = Elementary school; 3 = Junior high school; 4 = High school; 5 = Secondary school; 6 = Vocational high school and technical school; 7 = University college; 8 = University undergraduate; 9 = Graduate student	2.746	1.016
	Political status of the household head	1 = Ordinary people; 2 = Communist Party members; 3 = Youth Communist League members; 4 = Democratic parties	1.225	0.429
Family characteristics	Number of agricultural laborers	Number of laborers in the household engaged in agriculture (persons)	1.363	0.399
	Share of agricultural income	Share of agricultural income in total household income (%)	10.484	1.788
	Whether to join a cooperative	Does your household join a cooperative? (1 = yes; 0 = no)	0.236	0.425
	Whether to transfer land	Does your household participate in land transfer? (1 = yes; 0 = no)	0.467	0.499
	Mechanization level	The average mechanization level of the five stages of cultivation, sowing, pesticide application, fertilization, and harvesting	0.374	0.990
Village characteristics	Distance to county town	Distance of the village from the nearest county town (kilometers)	3.014	0.722
	Poverty-stricken village	Is your village a poor village? (1 = yes; 0 = no)	0.316	0.465
	Number of households in village	Number of households in village (households)	6.322	0.695
	E-commerce service station or product outlet in the village	Whether there is an e-commerce service station or product outlet in the village.	0.483	0.500

.

3.3. Model Construction

3.3.1. Linear Regression Model

The study examines how digital infrastructure impacts farm households’ scale management. It considers multiple independent variables that influence this relationship. Farm household scale management is continuous, while the independent variables include continuous and dichotomous types. Therefore, this study employs multiple linear stepwise regression to analyze these factors, sequentially introducing the variables into the model. The specific settings for this analysis are detailed as follows:

$$Lnlan{d}_i={\alpha }_0+{\alpha }_{1}DI+\sum controls+{\mu }_i$$

(1)

$$Lnoutpu{t}_i={\beta }_0+{\beta }_{1}DI+\sum controls+{\mu }_i$$

(2)

where $Lnland$ denotes the land management area for logarithmic treatment, $Lnoutput$ denotes the unit output for logarithmic treatment, $DI$ is the numerical infrastructure for 0–1 variables, and $controls$ denotes a series of control variables. ${\alpha }_0$ and ${\beta }_0$ are intercept terms, ${\alpha }_1$ and ${\beta }_1$ are regression coefficients, and ${\mu }_i$ is a random disturbance term.

3.3.2. Moderating Effect Model

This paper constructs a moderating effect test model to explore the moderating role of farm households’ digital literacy and digital skills training in the relationship between digital infrastructure and farm households’ scale management. The specific settings for this analysis are detailed as follows:

$$Lnlan{d}_i={\alpha }_0+{\alpha }_{1}DI+{\alpha }_{2}DI\times Z+\sum controls+{\mu }_i$$

(3)

$$Lnoutpu{t}_i={\beta }_0+{\beta }_{1}DI+{\beta }_{2}DI\times Z+\sum controls+{\mu }_i$$

(4)

where $Z$ denotes the moderating variables farm household digital literacy and digital skills training, $DI\times Z$ denotes the interaction term between digital infrastructure and the moderating variables, and the meanings of the remaining letters are consistent with the meanings of the benchmark regression. The moderating effect model tests whether ${\alpha }_2$ and ${\beta }_2$ are significant, and if they are significant, the moderating effect is valid.

4. Empirical Analysis

4.1. Benchmark Regression

Before conducting formal regression analysis, we first test the linearity and multicollinearity of the model, as well as the assumptions of normal distribution and homoscedasticity of the error term. The Spearman correlation analysis results show that the correlation coefficients between land management area, unit yield, independent variables, and control variables are in the weak correlation interval of 0.05–0.25, indicating that the relationship between variables is suitable for subsequent linear regression analysis. Regarding the issue of multicollinearity, we found through calculating the variance inflation factor (VIF) that the VIF values of all variables are less than 10, indicating the absence of multicollinearity among variables. In terms of testing the normality of error terms, we use the sktest command of STATA17.0 software to test the skewness and kurtosis of residuals. The results reject the null hypothesis at a significance level of 5%, confirming that residuals can be considered to follow a normal distribution to a certain extent. For the homoscedasticity hypothesis, we use the Breusch Pagan test to obtain a p-value of 0.067, indicating that the model has a certain degree of heteroscedasticity, which may affect the standard error of the regression coefficients. To solve this problem, we use cluster robust standard error regression for subsequent analysis, and all subsequent regressions are the results of cluster robust standard error regression.

Table 3. Baseline regression.

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
	Land Area	Unit Output	Land Area	Unit Output	Land Area	Unit Output
Digital infrastructure	0.065 ***	0.386 ***	0.042 *	0.377 ***	0.049 **	0.381 ***
	(0.024)	(0.033)	(0.022)	(0.033)	(0.022)	(0.033)
Gender of household head	0.182 ***	0.179 **	0.166 ***	0.181 **	0.148 ***	0.154 *
	(0.045)	(0.083)	(0.044)	(0.083)	(0.045)	(0.082)
Age of household head	−0.445 ***	−0.055	−0.366 ***	−0.012	−0.330 ***	−0.007
	(0.066)	(0.093)	(0.062)	(0.094)	(0.061)	(0.094)
Marital status	−0.028	−0.035	−0.017	−0.031	−0.019	−0.034
	(0.018)	(0.028)	(0.019)	(0.029)	(0.019)	(0.029)
Educational level	−0.029 **	−0.021	−0.030 **	−0.023	−0.021 *	−0.022
	(0.013)	(0.018)	(0.012)	(0.018)	(0.012)	(0.018)
Political status	−0.026	−0.034	−0.037	−0.043	−0.032	−0.023
	(0.027)	(0.040)	(0.026)	(0.040)	(0.026)	(0.040)
Labor force			0.005	−0.004	0.021	0.015
			(0.027)	(0.043)	(0.028)	(0.043)
Share of agricultural income			0.040 ***	0.029 ***	0.041 ***	0.032 ***
			(0.009)	(0.010)	(0.009)	(0.010)
Whether join a cooperative			0.049 *	0.040	0.031	0.043
			(0.029)	(0.040)	(0.029)	(0.041)
Whether land is transferred			0.305 ***	0.070 **	0.306 ***	0.053
			(0.022)	(0.034)	(0.022)	(0.033)
Mechanization level			0.039	0.075	0.038 ***	0.071
			(0.024)	(0.047)	(0.011)	(0.046)
Distance to County					0.126 ***	0.092 ***
					(0.016)	(0.025)
Poverty-stricken village					0.091 ***	−0.003
					(0.025)	(0.035)
Number of village households					−0.000	0.142 ***
					(0.017)	(0.027)
E-commerce in the village service station					0.030	−0.102 ***
					(0.022)	(0.033)
Constant	2.383 ***	1.804 ***	1.254 ***	0.744 *	0.621 *	−0.488
	(0.292)	(0.397)	(0.294)	(0.437)	(0.320)	(0.478)
Provincial fixed effects	Fixed	Fixed	Fixed	Fixed	Fixed	Fixed
N	2510	2510	2510	2510	2510	2510
R2	0.338	0.143	0.400	0.149	0.423	0.164

Standard errors in parentheses, * p < 0.1, ** p < 0.05, *** p < 0.01.

reports the benchmark regression results of digital infrastructure on farm households’ scale management. Models 1–6 employ land management area and unit output as dependent variables and are used to conduct regression analysis by gradually incorporating household head characteristic variables, family characteristic variables, and village characteristic variables. The regression results show that the estimation coefficient of digital infrastructure for land management area is stable at around 0.050, and the estimation coefficient for unit output is stable at around 0.380, both of which remain significantly positive. This indicates that digital infrastructure can significantly expand the scale of farmers’ operations, validating research hypothesis 1. From the results of Model 5 and Model 6, it can be seen that, while keeping the control variables constant, farmers with complete digital infrastructure have an average land management area 4.9% higher than those with incomplete infrastructure, and a yield per hectare 38.1% higher. Due to the strengthening of digital infrastructure, the precision and efficiency of agricultural production are significantly improved, and the optimal allocation of resources also improves agricultural total factor productivity, which in turn enhances land scale management. Additionally, strengthening digital infrastructure reduces the transaction costs of rural financial services by promoting the development of digital inclusive finance, enabling farmers to access credit, insurance, and other financial services more efficiently. This enhances their capacity for agricultural investment and risk management, thereby promoting land scale management [34].

4.2. Robustness Test

In order to further ensure the stability of the regression results and enhance the credibility and general applicability of the research results, this paper adopts a 1% Winsorization of the dependent variable before and after, a reduction in the overall regression sample size, as well as the propensity score matching (PSM) method to conduct a robustness test of the model in order to comprehensively verify the robustness and reliability of the benchmark regression conclusions. Among them, PSM uses 1:1 nearest neighbor matching without replacement, sets the clamp value to 0.01, and performs regression again using the matched samples. The results of the balance test are shown in

Table 4. Balance test results for matching quality.

Sample	Ps R2	LR chi2	p > chi2	MeanBias	MedBias	B	R	%Var
Unmatched	0.048	164.180	0.000	8.400	6.600	51.900	1.550	22.000
Matched	0.007	16.270	0.844	3.100	3.000	19.900	1.05	33.000

. After matching, the likelihood ratio is no longer significant, and both the mean bias and median bias show a significant decrease. The balance coefficient B after matching is below 25%, indicating that all matching variables are generally balanced after matching. The ATT results show that the treatment group is significantly higher than the control group by 0.031 and is significant at the 10% level, preliminarily validating the stability of the regression results. The results of the three different robustness tests are shown in

Table 5. Robustness test.

	Model 7	Model 8	Model 9	Model 10	Model 11	Model 12
	Winsorization		90% Sample Regression		PSM
	Land Area	Unit Output	Land Area	Unit Output	Land Area	Unit Output
Digital infrastructure	0.047 **	0.379 ***	0.058 **	0.384 ***	0.066 **	0.357 ***
	(0.021)	(0.031)	(0.024)	(0.036)	(0.029)	(0.040)
Constant	0.633 **	−0.513	0.929 ***	0.490	0.672 *	0.631
	(0.302)	(0.467)	(0.341)	(0.505)	(0.427)	(0.562)
Controlled variables	Controlled	Controlled	Controlled	Controlled	Controlled	Controlled
Provincial fixed effects	Fixed	Fixed	Fixed	Fixed	Fixed	Fixed
N	2510	2510	2259	2259	1617	1617
R2	0.440	0.175	0.434	0.163	0.377	0.123

Standard errors in parentheses, * p < 0.1, ** p < 0.05, *** p < 0.01.

. The regression coefficient of digital infrastructure on farm operating scale remains statistically significant at the 5% level, consistent with the results of the baseline regression in the preceding text, further validating the hypothesis.

4.3. Endogeneity Test

In exploring the impact of digital infrastructure on farm household scale management, there may exist issues such as omitted variables and reverse causality; in order to ensure the reliability of the findings, this paper uses the instrumental variable method to test the model again. Drawing on the Reference Zhang (2024), the mean value of the digital infrastructure situation of other farm households within the same village is used as an instrumental variable [69]. Because the level of people’s digital infrastructure within a village can be influenced by the people around them, and there is a certain correlation between the digital infrastructure situation of the people around them and their own digital infrastructure situation, while the digital infrastructure situation of the people around them does not have a direct impact on the scale of their own business and is not correlated with the other control variables in the model and the random perturbation term, the selection of this instrumental variable is more reasonable. The results of 2SLS regression based on instrumental variables are demonstrated in

Table 6. Instrumental variable regression.

	Model 13	Model 14	Model 15
	First Stage	Second Stage
	Digital Infrastructure	Land Area	Unit Output
Digital infrastructure		0.303 **	1.696 ***
		(0.149)	(0.263)
IV	0.412 ***
	(0.049)
Constant	0.307	0.556 *	−0.911
	(0.276)	(0.330)	(0.610)
Control Variables	Controlled	Controlled	Controlled
Provincial fixed effects	Fixed	Fixed	Fixed
N	2510	2510	2510
R2	0.088	0.366	0.348
Cragg–Donald Wald F	65.957
LM (p-value)	65.039
	(0.000)

Standard errors in parentheses, * p < 0.1, ** p < 0.05, *** p < 0.01.

; from the results, it can be seen that the coefficient of IV is significant at 1% level, and the hypothesis of positive correlation of instrumental variables is valid. In addition, after the test, the F-value is greater than 10, indicating that the instrumental variable passes the weak instrumental variable test. In the results of the second stage of regression, it can be seen that the regression coefficients of digital infrastructure on management area and unit output are still significantly positive at the 1% level, which is consistent with the results of the previous benchmark regression. To summarize, the effect of digital infrastructure in promoting farm households’ scale management still exists after excluding possible endogeneity problems.

4.4. Mechanism Test

The theoretical analysis in the previous section suggests that farm households’ digital literacy and digital skills training may play a moderating role in the impact of digital infrastructure on farm households’ scale management. In order to test the hypothesis, this part will test the moderating role played by farm households’ digital literacy and digital skills training. Firstly, the three variables of digital infrastructure, digital literacy, and digital skills training are centrally processed, and then regression analysis is conducted by generating interaction terms. The variance inflation factor (VIF) test results showed that the VIF values of each variable were all less than 10, indicating that the moderation effect model does not have multicollinearity issues. The final regression results are shown in

Table 7. Regulating effect test.

	Model 16	Model 17	Model 18	Model 19
	Land Area	Unit Output	Land Area	Unit Output
Digital infrastructure	0.013	0.139 *	0.048 **	0.384 ***
	(0.055)	(0.082)	(0.023)	(0.033)
Digital literacy	0.237 ***	−0.104
	(0.052)	(0.077)
Digital infrastructure× digital literacy	0.049	0.398 ***
	(0.092)	(0.130)
Digital infrastructure× digital skills training			0.229 **	0.140
			(0.092)	(0.134)
Digital skills training			−0.009	−0.096
			(0.045)	(0.074)
Constant	0.226	−0.165	0.859 ***	0.299
	(0.336)	(0.508)	(0.324)	(0.470)
Control variables	Controlled	Controlled	Controlled	Controlled
Provincial fixed effects	Fixed	Fixed	Fixed	Fixed
N	2510	2510	2510	2510
R2	0.402	0.173	0.399	0.171

Standard errors in parentheses, * p < 0.1, ** p < 0.05, *** p < 0.01.

, Models 16–19.

The interaction term of digital infrastructure and farm households’ digital literacy is significantly positive for land unit output and insignificant for operating areas. In contrast, the regression coefficient of the interaction term of digital infrastructure and digital skills training is significantly positive for the operating area and insignificant for land unit output. The regression coefficient of digital infrastructure continues to be significant and is significant at a significance level of over 5%. Overall, the above indicates that farmers’ digital literacy has a positive moderating effect on the impact of digital infrastructure on farmers’ land unit yield. The training of digital skills for farmers shows a positive moderating effect on the impact of digital infrastructure on the operating area of farmers.

4.5. Heterogeneity Analysis

4.5.1. Regional Heterogeneity

Yin et al. (2025) argue that the low population density and complex terrain of hilly and mountainous areas result in high costs for the construction of digital infrastructure such as optical cable laying and base station construction [70], which may limit their effectiveness. To verify this mechanism, we group villages based on terrain and conducts empirical tests using group regression, dividing villages into plain groups and hilly and mountainous groups.

Table 8. Regional heterogeneity test.

	Model 20	Model 21	Model 22	Model 23
	Plain Area	Hilly and Mountainous Areas	Plain Area	Hilly and Mountainous Areas
	Land Area	Land Area	Unit Output	Unit Output
Digital infrastructure	0.050 *	0.032	0.404 ***	0.328 ***
	(0.027)	(0.039)	(0.041)	(0.050)
Constant	1.386 ***	0.868	−0.975	1.400
	(0.377)	(0.593)	(0.627)	(0.873)
Permutation test	——		0.076 *
Control variables	Controlled	Controlled	Controlled	Controlled
Provincial fixed effects	Not fixed	Not fixed	Not fixed	Not fixed
N	1028	1482	1028	1482
R2	0.355	0.466	0.182	0.158

Standard errors in parentheses, * p < 0.1, *** p < 0.01.

regression results of Models 20–21 show that, under the condition of unchanged control variables, compared with households with inadequate digital infrastructure, households in the plain group with well-developed digital infrastructure have a significantly higher operating scale than the 5% level, and this difference is statistically significant at the 10% level, while there is no significant effect in the hilly and mountainous group; Models 22–23 show that both groups with well-developed digital infrastructure in the two regions are significantly higher than the control group. However, further testing of the regression coefficients of the two groups using a permutation test indicates that the promotional effect on unit output in the plain region is 7.6% higher than that in the hilly and mountainous region.

Based on the above analysis, we find that well-developed digital infrastructure in the plain region can significantly expand farmland operational scale and enhance unit output; while hilly and mountainous regions exhibit yield-enhancing effects, their overall promotional impact is significantly weaker than that of plain regions. The complex terrain conditions in hilly and mountainous areas increase the construction costs of digital infrastructure, significantly reducing the input/output ratio of farmers’ investments in infrastructure optimization. When the increased costs of optimizing digital infrastructure exceed the benefits derived from its promotional effects, farmers’ motivation to expand their scale diminishes, thereby rendering the promotional effect of digital infrastructure on operational scale no longer significant. On the other hand, terrain undulations and land fragmentation constraints limit the application of digital technologies such as drone pest control and water-saving irrigation, ultimately weakening the marginal promotional effect of digital infrastructure on unit output.

4.5.2. Age Heterogeneity

Different age groups may have differences in learning ability, adaptability, and the future direction of development, making the impact of digital infrastructure on the farm households’ scale management in different age groups heterogeneous. Hence, this study draws on Meng et al. (2023) to take 35 years as the node and divide the sample into the youth group and middle-aged and old-aged groups to explore the heterogeneity of the impact of digital infrastructure on the scale of farming in different age groups [71]. The results, as shown in

Table 9. Age heterogeneity test.

	Model 24	Model 25	Model 26	Model 27
	Youth Group	Middle-Aged and Old-Aged Group	Youth Group	Middle-Aged and Old-Aged Group
	Land Area	Land Area	Unit Output	Unit Output
Digital infrastructure	0.057	0.051 **	0.589	0.376 ***
	(0.535)	(0.023)	(0.677)	(0.033)
Constant	−2.595	1.181 ***	3.375	0.444
	(19.407)	(0.332)	(20.100)	(0.496)
Controlled variables	Controlled	Controlled	Controlled	Controlled
Provincial fixed effects	Fixed	Fixed	Fixed	Fixed
N	612	1898	612	1898
R2	0.724	0.400	0.860	0.167

Standard errors in parentheses, ** p < 0.05, *** p < 0.01.

, show that in the youth group, there is no significant contribution of the impact of digital infrastructure on the farm households’ scale management. In contrast, the regression results in the middle-aged and old-aged groups is still valid. This may be due to the difference in future directions between the youth and middle-aged and old-aged groups. The youth group is more inclined to pursue employment opportunities in non-agricultural fields, and they may be attracted by urban life, high-paying positions or other non-agricultural positions and have relatively low interest and willingness to engage in agriculture and thus are not willing to continue to expand the agricultural scale management. On the other hand, the middle-aged and old-aged group has been deeply engaged in the agricultural field for many years and accumulated rich experience and resources. They have deep feelings and reliance on agriculture, so they are more inclined to improve the efficiency of agricultural production further and expand the scale management by improving the digital infrastructure to realize the sustainable development of agriculture.

4.5.3. Scale Management Heterogeneity

Sheng et al. (2019) argues that groups with different farmland scale managements may have differences in resource endowment, business strategies, and technology adoption capabilities [72]. Thus, farm households of different scales may differ in their ability and effectiveness in coping with digital infrastructure. Using the 50% quartile of management area as the cut-off point, the sample is divided into small-scale and large-scale business groups to explore the heterogeneity of the facilitating effect of digital infrastructure on farm households with different scales. The results are shown in

Table 10. Heterogeneity test of scale management.

	Model 24	Model 25	Model 26	Model 27
	Large Scale	Small Scale	Large Scale	Small Scale
	Land Area	Land Area	Unit Output	Unit Output
Digital infrastructure	−0.018	0.026 *	0.078	0.392 ***
	(0.115)	(0.016)	(0.112)	(0.034)
Constant	1.503	0.445 **	−3.085 *	0.277
	(1.603)	(0.219)	(1.724)	(0.485)
Control variables	Controlled	Controlled	Controlled	Controlled
Provincial fixed effects	Fixed	Fixed	Fixed	Fixed
N	1267	1243	1267	1243
R2	0.334	0.388	0.385	0.169

Standard errors in parentheses, * p < 0.1, ** p < 0.05, *** p < 0.01.

; the promotion effect of digital infrastructure on farm households’ land management area and unit output is not significant in the large-scale management group but is significant in the small-scale group. This may be because small-scale farm households have relatively limited resource endowments and low productivity and land utilization, and the introduction of digital infrastructure can effectively compensate for these shortcomings and significantly enhance their operational efficiency and output. In contrast, large-scale operators may face multiple constraints, such as diminishing marginal benefits and increasing complexity of resource coordination when further expanding their land management area or increasing their unit output, as they already have a high level of resource utilization and production efficiency. Therefore, there is relatively little room for digital infrastructure to enhance its role, and it is challenging to produce a significant effect as it does for small-scale farm households.

5. Discussion and Conclusions

5.1. Discussion

In the face of increasingly severe food security threats, improving agricultural production efficiency and promoting scale managements are inevitable trends. Existing studies primarily focus on the impact of digital finance and the digital economy on agricultural production but pays less attention to infrastructure, particularly the digital component. To address this gap, this study examines the positive role of digital infrastructure in expanding farm scale and enhancing land productivity. This conclusion confirms hypothesis 1 and aligns with existing research [30]. As a key enabler of agricultural transformation through the digital economy, analyzing the impact of digital infrastructure on scale agriculture management provides a theoretical basis for governments to formulate rural development strategies.

Additionally, this study reveals the moderating effects of digital literacy and digital skills training on the relationship between digital infrastructure and agricultural scale management, offering a new perspective that differs from existing research, which primarily focuses on the mediating effects of technological innovation [73,74]. The results indicate that digital literacy exerts a positive moderating effect on the process of enhancing unit output through digital infrastructure, while digital skills training exerts a positive moderating effect on the process of expanding operational scale through digital infrastructure. The moderating effects of digital literacy and digital skills training indicate that a one-size-fits-all approach should not be adopted when promoting digital infrastructure. Instead, differentiated strategies should be developed based on the characteristics of different regions and individuals.

This study also delineates the geographical boundaries and individual differences of digital infrastructure on scale managements, identifying its impact in plain regions. Furthermore, the positive effects of digital infrastructure are more pronounced among middle-aged and elderly farmers and small-scale farmers. Under resource constraints, the conclusions of this paper provide theoretical support for factor allocation, suggesting that to maximize the positive impact of digital infrastructure on farm scale, priority should be given to developing plain regions and increasing support for middle-aged and elderly farmers and small-scale farmers.

Although this study contributes to a better understanding of the relationship between digital infrastructure and farm household scale management, some shortcomings still need to be further optimized in future studies. First, this study only focuses on data from a sample of selected provinces in China, and the findings may be quite different from those in other regions with distinct agricultural characteristics. Since the sample was selected from eastern, central, western, and northeastern regions, future studies have to pay further attention to other regions with different agricultural resources, environmental conditions, and demographic and institutional characteristics. Second, digital infrastructure is a broad concept that encompasses many other aspects, such as broadband access, mobile network coverage, and digital platforms. To fully understand the impact of digital infrastructure on farm household scale management, would be worthwhile to further explore the impact of these digital infrastructures and other elements on agricultural production and efficiency in future studies.

5.2. Conclusions

Exploring the impact of digital infrastructure on the scale management of farm households not only promotes the process of agricultural modernization, boosts the income of farm households, optimizes the allocation of agricultural resources, and promotes the development of the rural economy but also provides a scientific basis for the government to formulate targeted policies. Based on the CRRS research data, this paper investigates the impact of digital infrastructure on the scale management of farm households using the OLS model and moderating effect model. It draws the following conclusions: (1) Strengthening digital infrastructure can encourage farm households to expand the land management area and increase the unit output of land. (2) Farm households’ digital literacy shows a positive moderating effect in the role of digital infrastructure in influencing the unit output of farm households’ land; farm households’ digital skills training shows a positive moderating effect in the role of digital infrastructure in influencing the farm households’ management area. (3) The impact of digital infrastructure on farm households’ scale management is heterogeneous, with a stronger promoting effect on farm households’ scale management in plain areas and a weaker effect in hilly and mountainous areas; the effect is stronger for middle-aged and elderly farmers and small-scale farmers, and weaker for young groups and large-scale farmers.