Can New Digital Infrastructure Promote Agricultural Carbon Reduction: Mechanisms and Impact Assessment

Abstract

The development of new digital infrastructure enables the construction of intelligent agricultural production systems, enhances agricultural sustainability, and supports the national “dual-carbon” goals. Based on a theoretical analysis and using panel data for 31 Chinese provinces during 2011–2023, this study constructs a two-way fixed-effects model to empirically test the impact of new digital infrastructure on agricultural carbon emissions, and provides insights for differentiating provincial heterogeneity, as well as impact mechanism. The main findings are as follows: (1) New digital infrastructure is negatively correlated with agricultural carbon emissions, and this conclusion holds after a series of robustness and endogeneity tests. (2) Heterogeneity analysis reveals that, by geographic location, new digital infrastructure has a significant carbon reduction effect in eastern provinces but increases emissions in central provinces. By the digital development level, this study highlights the dual importance of digital infrastructure and financial supports. Contrary to those provinces leading in digital infrastructure development, there is a positive correlation in lagging areas. By policy support level, the significant carbon reduction effect is only observed in provinces with advanced digital-inclusive finance support. For green development policy support, it significantly reduces agricultural carbon emissions in pioneer regions but increases emissions in follower regions. (3) Mechanism tests indicate that new digital infrastructure promotes agricultural carbon reduction mainly through scale-economy effects and energy efficiency upgrading effects. Policy implications to improve agricultural digital infrastructure development and accelerate carbon emission reductions are finally suggested.

1. Introduction

With the advancement and implementation of the dual-carbon goals, China’s carbon reduction practices have gradually extended to the agricultural sector, attracting widespread attention. President Xi has stated that “a strong agricultural nation is the foundation of a strong socialist modern country” and that “advancing high-quality agricultural development is pivotal to national modernization and sustainable development strategy.” However, China’s agriculture has long relied on extensive production methods and inefficient energy use, resulting in both low economic and environmental performance [1]. According to the 2023 China Agricultural and Rural Low-Carbon Development Report, total agricultural carbon emissions reached 830 million tons of CO₂ equivalents, about seven percent of China’s total emissions. While lower than industrial emissions, agricultural emissions grew markedly from 1980 to 2020, and their complex and stochastic sources make monitoring and control challenging. To meet the “dual-carbon” goals, in May 2022, the Ministry of Agriculture and Rural Affairs and the National Development and Reform Commission issued the Agricultural and Rural Emission Reduction and Carbon Sequestration Implementation Plan to guide systematic carbon reduction efforts. Yet, mandatory cuts risk undermining normal agricultural growth and farmers’ incomes, and the rising food demand limits drastic reductions in fertilizer and pesticide use, constraining mitigation effectiveness [2]. Thus, new drivers of low-carbon, green agriculture warrant greater attention in China’s high-quality agricultural transformation.

Under the Digital Rural Development Strategy Outline and the National Smart Agriculture Action Plan (2024–2028), the construction of digital villages has accelerated. Optimized digital infrastructure, governance, and services now support agricultural intelligence and low-carbon transformation [3].

Among them, new digital infrastructure plays a fundamental role. In this study, the new digital infrastructure is defined as an infrastructure system with the core support of modern information technology, covering 5G communication networks, data centers, artificial intelligence, the Internet of Things, industrial internet, blockchain, and other key areas. It is different from traditional infrastructure represented by transportation, energy, water conservancy, etc. The latter focuses more on physical resource allocation and economic growth foundation, while new digital infrastructure emphasizes the efficient flow of information elements, the integrated use of data resources, and the construction of intelligent systems.

The digital economy, with new digital infrastructure as its core carrier, can effectively leverage its connectivity and low-carbon nature. It links technology, labor, and resources efficiently in agricultural production, and upgrades digital management across pre-, mid-, and post-production stages. This promotes a shift toward low-input, low-emission, and high-output agriculture [4]. The Chinese government emphasizes leveraging digital agriculture and rural construction to modernize the sector. Therefore, clarifying how new digital infrastructure affects agricultural carbon reduction and its mechanisms is essential to unlock reform dividends and deepen green low-carbon transformation in agriculture.

Research on agricultural carbon emission measurement and its drivers has expanded. Major emission sources include chemical inputs (fertilizers, pesticides, and films), soil carbon loss from tillage, fossil fuel consumption by farm machinery (direct and indirect), and greenhouse gases from livestock digestion and manure. Measurement methods encompass emission factor approaches, model simulations, and field measurements [5]. Given the wide-ranging and diffuse nature of agricultural emissions and the complexity of model calibration, the emission factor method is generally most applicable. It estimates emissions by multiplying activity data by corresponding factors [6].

Studies using these methods show that China’s agricultural emissions have risen in phases since 1993, with significant regional differences—higher in the eastern provinces [7]. Some employ LMDI decomposition to analyze economic, efficiency, and labor influences [8]; others use econometric models to test the effects of technological progress [9], industrial clustering [10], and arable land transfer [11].

Recent interest in the digital empowerment of agriculture has focused mainly on economic benefits, with less emphasis on environmental dimensions. On the economic side, digital–technology integration enhances intelligence, precision [12], and scale in production; boosts efficiency and farmer incomes [13]; and expands the marketing and traceability channels to raise product value [14], thus driving high-quality development and rural revitalization [15]. Environmentally, some empirical research finds the optimization effects of digital transformation on carbon emission-related elements. For example, Zhong et al. show that digitalization fosters technological progress, reducing carbon emission intensity; Xu et al. develop digitalization indices to quantify rural transformation’s impact on the planting sector’s carbon emission rate, identifying industrial structure upgrading and scale optimization as key channels [16]. Some researchers find that digitalization promotes factor and efficiency improvements via innovation, scale operations, and clustering, with notable spatial spillovers, and some find that digital transformation improves farmers’ capital endowment through education and information, aiding the green low-carbon transition.

However, some studies suggest that the effectiveness of the digital economy is phased, such as the “green paradox” in the manufacturing sector. This view has been less explored in the field of agricultural carbon reduction, and empirical research has rarely considered the impact of regional differences on the effectiveness of digital transformation. On the one hand, digital infrastructure and digital products themselves are also important sources of carbon emissions, and they have the characteristics of long cycles and high energy consumption. In the early stage of the development of the digital economy, carbon emissions may be exacerbated. For instance, Australia’s long-term data show digitalization spurred electricity consumption and raised local emissions [17]. China’s current digital economy is both capital- and technology-intensive; while it can lower emissions via efficiency gains, it also expands scale, raising operational costs and energy use [18]. This may produce an “N-shaped” non-linearity in the digital economy’s impacts on agricultural carbon emission efficiency. On the other hand, considering regional heterogeneity, China’s eastern, central, and western provinces differ markedly in economic and agricultural development. Measured digital infrastructure levels vary across regions, as do economic bases and digital-inclusive finance support, leading to east–west disparities in informatization progress. Likewise, regional differences in green-development foundations affect digitalization carbon reduction efficacy. However, few studies explore these regional nuances in agricultural emission research.

The existing literature offers valuable insights but leaves gaps. First, most work examines the digital economy as a whole rather than focusing on new digital infrastructure, which is still expanding unevenly across rural areas. Second, the theoretical mechanisms by which new digital infrastructure drives carbon reduction remain underdeveloped, especially regarding the empirical validation of multiple pathways. Third, while many studies consider east–central–west heterogeneity, they seldom account for other regional digital supports or green development differences.

To address these gaps, this study measures agricultural emissions based on planting production processes across 31 provinces from 2011 to 2023. Planting is emphasized because digital infrastructure applications are more extensive there, while livestock emissions involve biogenic processes less amenable to digital management, and feed is planting-derived. Using panel data and two-way fixed-effects models, this study comprehensively evaluates the impact of new digital infrastructure on agricultural carbon reduction. It makes three marginal contributions: (1) focusing on new digital infrastructure to assess its effect during a critical period of high-quality transformation; (2) clarifying mechanisms through scale economy and energy efficiency upgrading effects and empirically testing these pathways; and (3) conducting a multi-dimensional heterogeneity analysis of “geographical location–digital development level–green development” comprehensively to effectively identify the scenarios where new digital infrastructure plays a better role in agricultural “carbon reduction” and support the formulation of differentiated agricultural policies at the provincial level.

2. Theoretical Analysis and Research Hypotheses

This section focuses on a theoretical analysis of the direct effects of the new digital infrastructure on agricultural carbon reduction, as well as the dual pathways of scale economy effects and energy efficiency upgrading effects. Figure 1 presents the theoretical mechanism by which new digital infrastructure development influences agricultural carbon reduction.

2.1. Direct Effects of New Digital Infrastructure on Agricultural Carbon Reduction

The development and improvement of new digital infrastructure bring about digital elements and digital technologies that can be deeply integrated into the entire process of agricultural production, accelerating the low-carbon transformation of agriculture. Its main functions are reflected in the following three aspects: (1) Supporting the construction of a modern agricultural production management system. Scientific data monitoring and analysis improve crop scheduling and sustainable land use [19]. For example, smart greenhouses use multifunctional sensor networks to monitor seeding, tillage, and climate data, enabling intelligent sowing, fertilization, and irrigation. (2) Enabling precise input management in production. Full-cycle data analysis via smart equipment, IoT, and big data allows more accurate management of inputs such as fertilizers and machinery. In the high-emission fertilizer and pesticide stage, centralized application of fertilizers and integrated pest management reduce waste from individual operations, improve input efficiency [20], and overcome “the more fertilizer is applied, the more fertilizer the land will need” in extensive farming [21], thus optimizing resource allocation and enhancing carbon mitigation and sequestration capability. (3) Facilitating scientific supervision on green agricultural governance platforms. Big data, remote sensing, blockchain, and other digital technologies support the development of systems for monitoring and forecasting agricultural emissions and sinks. These systems enable end-to-end data supervision from production to consumption, allowing governments to measure and manage emissions and carbon sinks precisely and dynamically, and to base policy decisions on solid scientific evidence [4].

More deeply, considering the complexities of modern agricultural development practice, the carbon reduction effect of new digital infrastructure may be influenced by some external conditions. First, the regional economic development level would significantly influence its digitalization progress as well as in the agricultural area. Economically advanced regions, with superior infrastructure and greater digital investment capacity, can more effectively adopt precision farming and smart irrigation technologies, enhancing resource efficiency and reducing agricultural emissions [22]. Second, some studies proposed that there may exist a “green paradox” or “rebound effect” between new digital infrastructure development and carbon-related issues [23]. While digital infrastructure (e.g., 5G and data centers) enables emission reduction through smart technologies, it requires large amounts of resources and emission-intensive industrial inputs in the early stage of fundamental construction, creating a “first increase, then decrease” emission pattern that constitutes an environmental paradox [24]. Third, government policy support is also a critical influencing factor, especially policies in the digital transformation and carbon reduction fields. For example, some incentive policies, such as green power subsidies for data centers and 5G base station efficiency standards, directly reduce the carbon footprint. Some carbon regulatory policies like Carbon Emissions Trading Systems push enterprises to adopt more clean technologies [25]. Some coordinated initiatives like China’s “East Data West Computing” project could balance computing growth with renewable energy utilization. Such policy portfolios can transform digital infrastructure from an “energy consumer” to “emission mitigator”.

In summary, as foundational support, new digital infrastructure effectively promotes digitalization and smart management upgrades, standardizes agricultural operations, and aids in the establishment of a scientific governance framework for agricultural carbon reduction. However, considering the complex market condition, the carbon reduction effect of new digital infrastructure would be influenced by diverse external factors, such as the regional economic level, development stage of new digital infrastructure, as well as policy support. Based on the above analysis, this study proposes the following hypothesis:

Hypothesis 1.

The development of new digital infrastructure would generally contribute to agricultural carbon reduction, yet this effect may be influenced by some external conditions.

2.2. Indirect Mechanisms of New Digital Infrastructure on Agricultural Carbon Reduction

2.2.1. Scale Economy Effect

Marshall and other economists define the scale economy effect as the tendency for firms or industries to reduce costs and increase efficiency by expanding the production scale and concentration, which refines divisions of labor and specialization [26]. Modern production theory and practice further find that digital integration and intelligence can accelerate this process [27]. In the green transformation of agriculture, new digital infrastructure supports large-scale and standardized production, optimizes resource allocation, and enhances green production efficiency.

First, the development of new digital infrastructure would significantly accelerate the formation of the scale economy in agricultural productions. (1) It enhances production efficiency. IoT technologies, sensors, wireless networks, and data analysis enable the long-term, systematic monitoring, collection, and analysis of production data. Artificial intelligence algorithms can then precisely diagnose production conditions and optimize planning, thereby improving productivity and enabling large-scale management. (2) It accelerates productivity gains. New digital infrastructure supports the development and use of next-generation smart and automated agricultural equipment, such as intelligent machinery, automated irrigation systems, drones, and smart monitoring platforms. Standardized workflows improve quality and labor efficiency, generating surplus labor capacity for larger-scale collaborative management. (3) It expands consumer markets. With digital village policies like “digital commerce for agriculture” and the “Internet + agricultural products” initiative, rural e-commerce has grown rapidly. E-commerce reduces transaction steps. Real-time data analysis and market intelligence allow producers to respond faster to demand changes, adjust production and sales plans, foster industrial clusters, and improve competitiveness [28].

Second, the scale economy would significantly curb the agricultural carbon emissions through three channels. (1) Regional agglomeration enables the pooling of land, machinery, and inputs across farms, cutting redundant capital and energy expenditures per unit of output and immediately lowering carbon intensity [29]. (2) Specialization within these clusters induces more shared machinery platforms, and those centralized fertilizer distribution patterns would promote more energy-saving production. (3) Profits generated by agglomeration could further support the diffusion of digital and low-carbon technologies, further reducing fossil fuel and fertilizer use [30]. However, government policy may play an important role in leading the digital and agricultural industrial agglomeration in practice, which would bring about some potential ecological risks from excessive scaling.

Thus, government policy could pay more attention to steer expansion toward higher total factor productivity instead of mere acreage growth [31]. Some regulation indicators and real-time monitoring then act as automatic brakes, ensuring that any increase in scale remains within scientifically defined ecological limits. Therefore, steering agricultural clusters toward genuine economies of scale, rather than mere acreage expansion, can serve as a pivotal lever for carbon mitigation in China’s farming sector.

Overall, the development of new digital infrastructure would support large-scale agricultural production to raise the input intensity; improve fertilizer, pesticide and machinery efficiency; and lower marginal costs, thus boosting green production efficiency. Then, the clustering of upstream and downstream industries would help to build more shared economic patterns and promote advanced technologies’ diffusion in the whole production chain [32]. Based on the above analysis, this study proposes the following hypothesis:

Hypothesis H2a.

The scale economy effect represents a critical mechanism enabling the new digital infrastructure to achieve carbon mitigation outcomes.

2.2.2. Energy Efficiency Upgrading Effect

Improving energy efficiency has long been a key pathway for countries to develop low-carbon economies. It involves reducing energy consumption in production and consumption through technological, managerial, behavioral, and policy measures to achieve sustainable economic, environmental, and social development. In agriculture—where energy efficiency is relatively low—this challenge requires digital technologies to further optimize energy production, distribution, management, and use, enabling systematic energy management to reduce consumption and environmental impacts and drive sustainability.

First, new digital infrastructure development is a benefit for improving energy efficiency. (1) It supports green technological progress in agriculture. New growth theories and China’s practice of high-quality agricultural transformation confirm that core technology upgrades are the main driver of growth [33]. New digital infrastructure facilitates experiential learning and breakthrough transformations in green production and management technologies. On the energy consumption side, agricultural digital management can effectively collect and process agricultural production management and crop information and promote energy conservation and consumption reduction with the support of new technologies [34]. (2) It enables the construction of high-efficiency energy-management networks. Digital infrastructure supports collaborative, intelligent, and networked systems for agricultural production and resource inputs, enhancing the transparency and monitoring of energy audits and usage as well as improving energy management and utilization efficiency [35]. (3) It fosters low-carbon awareness and behavior. Supported by information networks, agricultural practitioners can easily access national low-carbon policies online. Simultaneously, governments can use digital technologies to promote key innovations, precisely and broadly disseminating quality policies and training to target farmers and ensuring the rational use of resources and energy [36].

Second, improving energy efficiency can curb agriculture emissions via three pathways. (1) At the input stage, more application of energy-saving agricultural machinery and equipment could directly cut fossil fuel use and related emissions [37]. (2) During the operation period, more precise production management, such as in irrigation and fertilization, could prevent over-application and lower energy use per unit of output while maintaining or even raising yields. (3) At the system level, integrating modern energy supply systems, such as rooftop solar and biomass CHP, allows farms to integrate renewable energy into production processes on site, creating additional carbon reduction benefits.

Overall, rural new digital infrastructure can significantly enhance agricultural energy efficiency, directly reducing resource and energy input needs in production. Continuous improvements in energy efficiency enable the same agricultural output to be produced with less and cleaner energy, delivering simultaneous reductions in both absolute and intensity-based carbon emissions from crop production. Based on the above analysis, this study proposes the following hypothesis:

Hypothesis H2b.

The energy-efficiency upgrading effect is another critical mechanism enabling the new digital infrastructure to achieve carbon mitigation outcomes.

3. Model Design and Data Processing

3.1. Model Construction and Variable Descriptions

3.1.1. Baseline Model Construction

To examine the effect of new digital infrastructure on agricultural carbon reduction, this study follows related studies and constructs a two-way fixed-effects model for empirical testing. The model specification is as follows:

$$carbo{n}_{it}={\alpha }_0+{\alpha }_{1}dig{i}_{it}+{\alpha }_{2}Contro{l}_{it}+{\mu }_i+{\nu }_t+{\epsilon }_{it}$$

(1)

where $i$ and $t$ denote the province and year, respectively; $carbo{n}_{it}$ is the dependent variable measuring agricultural carbon emissions; $dig{i}_{it}$ is the core explanatory variable representing the level of new digital infrastructure; $Contro{l}_{it}$ is a vector of control variables; ${\mu }_i$ captures individual (province) fixed effects; ${\nu }_t$ captures time fixed effects; and ${\epsilon }_{it}$ is the random error term.

3.1.2. Variable Selection and Descriptions

• Core Explanatory Variable and Dependent Variable

The core explanatory variable in this study is the level of rural new digital infrastructure. Based on the current status of digital infrastructure development in rural China, drawing on relevant studies [38], and considering the data availability, this study selects two positive indicators, the fiber optic cable length per square kilometer and the number of rural broadband access users. These dual indicators effectively characterize the digital foundation of agriculture through both physical infrastructure support and knowledge capacity dimensions. Among them the indicator “fiber optic cable length per square kilometer” directly quantifies digital infrastructure density, which is also the fundamental support to apply a series of digital agricultural equipment to build a smart agriculture system. The indicator “the number of rural broadband access users” could describe the digital literacy of rural residents to some extent. Higher internet adoption rates indicate basic competency in utilizing digital agricultural infrastructure to leverage new digital infrastructure for agricultural productivity. And this study uses the entropy method to construct a composite index representing the overall level of rural new digital infrastructure ($dig{i}_{it}$).

The dependent variable is measured by agricultural carbon emissions ($carbo{n}_{it}$) mainly from planting process, without considering the livestock. This is mainly because the new digital infrastructure is more widely adopted in the planting process in China. Additionally, the livestock-related greenhouse-gas emission sources (e.g., enteric methane and manure CH₄ and N₂O) are difficult to measure along with the biological activities [39]. Based on existing research, this study divides the sources of agricultural carbon emissions into four main categories:

• Emissions from the production and use of agricultural inputs such as chemical fertilizers, pesticides, and plastic films;
• Direct and indirect fossil fuel consumption (mainly diesel) from the use of agricultural machinery;
• Carbon losses caused by soil disturbance during tillage and plowing;
• Indirect fossil fuel-induced emissions from electricity use during irrigation.

Agricultural carbon emissions for each province are estimated using the emission factor method, with the calculation formula as follows:

$$C={\displaystyle \sum C_i={\displaystyle \sum S_i\cdot {\beta }_i}}$$

(2)

In the equation, $i$ represents different agricultural activities; $C$ represents the total agricultural carbon emissions; $C_i$ denotes the carbon emissions from different agricultural activities; and $S_i$ is the scale of each type of agricultural activity. Specifically, fertilizer, pesticides, plastic films, and diesel are measured by actual usage; tillage is measured by the actual sown area of crops; and agricultural irrigation is measured by the actual irrigated area. ${\beta }_i$ refers to the emission factors for each agricultural activity. The emission factors of different agricultural activities are set according to the relevant literature and are reported in

Table 1. The emission factors.

Source of Carbon	Coefficient	Units	Source
Fertilizer	0.8956	kg·kg−1	Oak Ridge National Laboratory, USA
Pesticides	4.934	kg·kg−1	Oak Ridge National Laboratory, USA
Plastic films	5.18	kg·kg−1	Nanjing Agricultural University, China
Diesel	0.5927	kg·kg−1	Intergovernmental Panel on Climate Change
Tillage	312.6	kg·km−2	China Agricultural University
Agricultural irrigation	25	kg·ha−1	Dubey [40]

.

2.

Control Variables

Based on relevant studies and agricultural production practices, agricultural carbon emissions are influenced by multiple factors. This study selects control variables from four dimensions: overall rural development level, traditional rural infrastructure, support for the rural green transition, and agricultural production conditions.

In terms of overall rural development, four control variables are selected: rural population (pop), employment in the primary industry (worker), number of farming households (agrnum), and rural per capita disposable income (perinc). In general, rural population and the number of farming households reflect the size and agglomeration of the rural population. A larger rural population can provide abundant labor, thus supporting the expansion of agricultural production. Employment in the primary industry directly reflects the scale of the population engaged in agricultural production and operations—a larger value indicates a higher level of agricultural development and greater labor absorption. Rural per capita disposable income reflects the economic returns of agriculture; higher income levels suggest better agricultural and related industrial development, generating stronger economic benefits. Collectively, these indicators reflect different aspects of rural economic development and are important determinants of agricultural carbon emissions. However, the relationship between agricultural development and carbon emissions remains uncertain. On one hand, a larger agricultural sector requires more energy and resource inputs for production and operations, which may lead to higher carbon emissions. On the other hand, with continuous rural and agricultural development, production experience and technologies may improve, enhancing agricultural efficiency and enabling emission reductions. Therefore, the effects of these variables require empirical validation.

In terms of traditional rural infrastructure, two control variables are selected: traditional infrastructure and per capita electricity consumption in rural areas (ele). The level of traditional infrastructure is measured by rural road mileage per capita (trafun). Studies show that improved rural transportation infrastructure can shorten spatial and temporal distances between rural areas, markets, and labor mobility, supporting more large-scale and standardized agricultural production. It also attracts social capital investment and promotes green rural development projects, contributing to emission reduction [41]. Greater electricity infrastructure and usage in rural areas suggest increased use of electricity-powered equipment in both rural life and agricultural production and transport, indirectly driving higher energy and resource consumption and thus more carbon emissions [42].

Regarding support for the rural green transition, two control variables are selected: the proportion of environmental protection fiscal expenditure to total public fiscal expenditure (envexp) and average years of education among rural laborers (aveedu). Overall, in the process of the green transformation of agriculture, government policy support for the environment is a key factor. Higher environmental expenditure helps improve rural living conditions, waste treatment systems, and the efficiency and structure of power generation and consumption, thereby supporting rural low-carbon and green development [43]. On another front, higher educational attainment among rural laborers helps enhance farmers’ awareness and learning capacity in production and management, promotes a shift away from traditional practices, and indirectly affects the input use of agricultural resources and energy, thus influencing carbon emissions [44].

Regarding agricultural production conditions, two control variables are included: fertilizer use (nutri) and number of large livestock at year end (anima). According to the literature, the use of fertilizers and pesticides during agricultural production, emissions from animal digestion during livestock farming, and the inputs of various energy and resource-consuming activities are all major contributors to agricultural carbon emissions.

3.2. Data Sources and Processing

This study uses provincial panel data from 31 provinces in China for the period 2011–2023. Due to data availability, the regions of Hong Kong, Macao, and Taiwan are excluded from the analysis. Data related to agricultural production activities for estimating carbon emissions, measures of new digital infrastructure, and control variables are mainly obtained from the China Rural Statistical Yearbook, China Energy Statistical Yearbook, China Statistical Yearbook, and the Statistical Yearbooks of Individual Provinces. Missing values are primarily addressed through interpolation methods, with the sensitivity analysis is reported in Appendix A. In addition, to minimize the impact of large data discrepancies on the regression results, logarithmic transformation is applied to absolute-value variables.

4. Empirical Results Analysis

4.1. Baseline Regression Analysis

Based on Model (1), this study empirically examines the effect of new digital infrastructure on agricultural carbon reduction. The regression results are presented in

Table 2. Baseline regression results.

	(1)	(2)	(3)	(4)
digi	−0.018	−0.022 *	−0.028 **	−0.010 *
	(0.013)	(0.011)	(0.011)	(0.005)
pop	−1.233 ***	−1.390 ***	−1.347 ***	−0.904 ***
	(0.234)	(0.257)	(0.257)	(0.286)
worker	0.385 ***	0.189 ***	0.182 ***	0.091
	(0.121)	(0.050)	(0.052)	(0.059)
agrnum	0.299 **	0.147	0.155	0.028
	(0.135)	(0.195)	(0.189)	(0.135)
perinc	1.088 ***	1.072 ***	1.207 ***	0.859 ***
	(0.148)	(0.193)	(0.190)	(0.179)
ele		0.059 **	0.056 **	0.058 **
		(0.025)	(0.022)	(0.023)
trafun		−0.100 **	−0.093 **	−0.057 **
		(0.037)	(0.036)	(0.019)
envexp			−3.339 ***	−1.459 ***
			(0.867)	(0.342)
aveedu			−0.193	0.065
			(0.341)	(0.203)
nutri				0.613 ***
				(0.050)
anima				−0.010
				(0.022)
_cons	−0.439	2.935	1.827	−0.227
	(1.413)	(2.108)	(1.812)	(1.487)
Province	YES	YES	YES	YES
Year	YES	YES	YES	YES
N	403	403	403	403
R2	0.993	0.995	0.996	0.997

Note: robust standard errors are reported in parentheses. * p < 0.1, ** p < 0.05, and *** p < 0.01. The same notation applies hereafter *.

, which also reports the outcomes under different sets of control variables to enhance the robustness of the conclusions.

According to the regression results in Table 2, after progressively incorporating control variables from various dimensions, the coefficient of new digital infrastructure becomes significantly negative with respect to agricultural carbon emissions. After controlling for time and individual fixed effects, the coefficient of $digi$ in Column (4) is significantly negative at −0.010, indicating that new digital infrastructure significantly reduces agricultural carbon emissions and promotes carbon reduction in agriculture. Therefore, the carbon reduction effect of new digital infrastructure proposed in Hypothesis 1 is proved in the baseline result, while the influence of external conditions will be further discussed in the heterogeneity analysis part.

A closer analysis of how new digital infrastructure affects agricultural development reveals that, as a key support for smart rural construction, it plays an important role in promoting the transformation and upgrading of agriculture. Intelligent management systems greatly improve the efficiency of energy and resource use in agriculture. Moreover, the technological innovations and applications it facilitates in energy-saving and low-carbon practices significantly optimize the structure of agricultural energy inputs and enhance low-carbon efficiency, thereby generating a notable carbon reduction effect.

Overall, the new digital infrastructure has not only enhanced agricultural production efficiency but also promoted the conservation of resources and the reduction of emissions. The application of digital technology in agriculture not only reduces production costs but also avoids the excessive use of chemical fertilizers and pesticides, alleviating the pressure of agricultural production on the environment. Although the digital transformation of agricultural production may require a relatively high investment cost in the short term, in the long run, this transformation not only helps to significantly reduce carbon emissions but also enhances the sustainable development capacity of agriculture. The application of low-carbon technologies driven by new digital infrastructure not only enhances the competitiveness of agriculture but also prompts farmers and enterprises to pay more attention to environmental protection and sustainability, thus forming a positive economic cycle.

4.2. Robustness and Endogeneity Analyses

4.2.1. Robustness Tests

To verify the robustness of the baseline regression results, this study conducts several robustness checks. The empirical results are reported in

Table 3. Robustness test results.

	(1)	(2)	(3)	(4)
digi	−0.012 *		−0.051 *	−0.011 *
	(0.006)		(0.024)	(0.006)
repdig		−0.030 ***
		(0.009)
npknut			0.006
			(0.006)
pop	−1.992 ***	−0.893 ***	−1.212 ***	−0.947 ***
	(0.321)	(0.276)	(0.365)	(0.299)
worker	0.102	0.098	0.190 **	0.111 *
	(0.066)	(0.061)	(0.073)	(0.054)
agrnum	0.049	0.070	0.106	0.066
	(0.148)	(0.136)	(0.210)	(0.137)
perinc	0.910 ***	0.959 ***	1.443 ***	0.887 ***
	(0.200)	(0.200)	(0.307)	(0.184)
ele	0.063 **	0.056 **	0.057 **	0.055 **
	(0.026)	(0.022)	(0.022)	(0.023)
trafun	−0.063 **	−0.059 **	−0.090 **	−0.054 **
	(0.023)	(0.020)	(0.033)	(0.020)
envexp	−1.511 ***	−1.492 ***	−3.582 ***	−1.463 ***
	(0.371)	(0.352)	(0.899)	(0.358)
aveedu	0.061	0.008	−0.223	0.043
	(0.220)	(0.196)	(0.419)	(0.207)
nutri	0.627 ***	0.602 ***		0.608 ***
	(0.052)	(0.053)		(0.051)
anima	−0.016	0.002	0.011	−0.012
	(0.025)	(0.022)	(0.023)	(0.023)
_cons	−0.337	−1.462	−0.937	−0.437
	(1.627)	(1.540)	(1.688)	(1.513)
Province	YES	YES	YES	YES
Year	YES	YES	YES	YES
N	403	403	377	372
R2	0.984	0.997	0.994	0.997

.

• Replacing the dependent variable: The rural per capita carbon emissions (percar) indicator is used as an alternative dependent variable to assess the impact of new digital infrastructure on agricultural carbon reduction from a per capita emissions perspective. The result is shown in Column (1).
• Replacing the core independent variable: A new digital infrastructure index (repdig) is constructed using the number of rural broadband access users and the total length of fiber optic cables through the entropy weight method. This replaces the original core explanatory variable to re-estimate its effect on agricultural carbon emissions. The result is presented in Column (2).
• Replacing a control variable: The total amount of fertilizer used in the original model is replaced by the use of nitrogen, phosphorus, and potassium fertilizers (npknut), which account for a large share of agricultural input and are major contributors to agricultural carbon emissions. The result is shown in Column (3).
• Excluding special years: Considering that economic fluctuations in certain years may affect the development of rural digital infrastructure and agricultural production activities, this study excludes the year 2020, a year marked by significant socio-economic volatility, and re-estimates the baseline model. The result is shown in Column (4). Overall, under all robustness checks, the coefficient of new digital infrastructure on agricultural carbon emissions remains significantly negative and consistent with the baseline regression results. This confirms the robustness of the main findings, suggesting that the development of new digital infrastructure can effectively reduce agricultural carbon emissions.

4.2.2. Endogeneity Analysis

Although robustness tests can partly mitigate endogeneity issues, they cannot completely eliminate them. In the process of agricultural development, there may exist a bidirectional causal relationship between carbon emissions and new digital infrastructure, or omitted variable bias due to missing control variables, all of which can lead to endogeneity. To address this, this study re-estimates the baseline results using the two-stage least squares (2SLS) method.

For the instrumental variables of the new digital infrastructure indicator, this study selects the number of fixed-line telephones per hundred people in 1984 (tel) and the number of postal and telecommunications offices at the end of the year (post). These are interacted with year variables to construct panel data-appropriate instrumental variables. These two instrumental variables are “strictly exogenous” and “strongly correlated”. On the one hand, the application of fixed-line telephones and the development of postal and telecommunications services in 1984 marked the initial stage of China’s digital infrastructure. The current technological path greatly depends on the current level of digital economic development and provides the foundational facilities for internet access. Thus, the relevance condition is satisfied. On the other hand, as traditional communication tools, the usage frequency and influence of these instruments have declined greatly with technological advancement, indicating no direct impact on contemporary economic activities [45]. Being historical cross-sectional data, they no longer directly affect current economic outcomes, thereby satisfying the exogeneity requirement.

According to the empirical results in

Table 4. Endogeneity test results.

	(1)	(2)
	digi	carbon
tel	0.019 **
	(2.015)
post	−0.050 ***
	(−5.336)
digi		−0.044 **
		(−2.30)
pop	−2.026 ***	−0.898 ***
	(−7.673)	(−3.468)
worker	0.041	0.108 *
	(0.335)	(1.756)
agrnum	−0.325	0.076
	(−0.691)	(0.624)
perinc	7.069 ***	1.203 ***
	(4.602)	(5.497)
ele	−0.017	0.057 ***
	(−0.701)	(2.773)
trafun	−0.142	−0.067 ***
	(−1.100)	(−2.966)
envexp	−6.972 ***	−1.777 ***
	(−2.776)	(−5.417)
aveedu	−1.275	0.016
	(−0.630)	(0.090)
nutri	−0.277	0.590 ***
	(−0.891)	(9.282)
anima	0.061	0.009
	(0.739)	(0.413)
Constant	508.983 ***	−5.621 ***
	(5.367)	(−3.935)
Province	YES	YES
Year	YES	YES
N	403	403
R2		0.997

, the first-stage instrumental variables significantly affect new digital infrastructure in Column (1), and the Cragg–Donald Wald F test rejects the weak instrument hypothesis, indicating the validity of the instruments. In Column (2), the coefficient of new digital infrastructure in the second stage is significantly negative and consistent with the baseline regression results.

Overall, the endogeneity test results further confirm the validity of the baseline regression findings in this study.

4.3. Heterogeneity Analysis

The baseline empirical results show that new digital infrastructure exerts a significant “carbon reduction” effect on agriculture. However, considering the considerable differences among provinces in terms of geographic location, digital development, and greening progress, it remains to be explored whether these differences affect the effectiveness of new digital infrastructure in promoting agricultural carbon reduction. Accordingly, this section systematically conducts a multidimensional heterogeneity analysis based on provincial “geographic location–digital development level–green development” to deeply investigate the favorable conditions under which new digital infrastructure contributes to emission reduction, supporting comprehensive green agricultural policy formulation.

4.3.1. Geographic Location Heterogeneity

China’s vast territory exhibits significant regional disparities in economic development, agricultural resources and environment, policy support, and investment, which in turn affect agricultural development and digitalization progress. Existing studies generally find significant regional differences in the impact of digital economy development on agricultural issues [7]. As a foundational and core component of the digital economy, whether new digital infrastructure exhibits geographic heterogeneity in its effects is worth further exploration.

In response, this study divides the 31 provinces into three groups—eastern, central, and western regions—according to traditional geographic standards and performs grouped regressions. The results are shown in Columns (1)–(3) of

Table 5. Heterogeneity analysis results.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
	Geographic Location			New Digital Infrastructure		Digital Inclusive Finance		Green Development
	East China	Central	West China	Leading	Lagging	Developed	Developing	Pioneer	Follower
digi	−0.115 **	0.019 *	0.002	−0.093 *	0.007 **	−0.037 ***	0.006	−0.241 **	0.008 *
	(0.047)	(0.010)	(0.004)	(0.044)	(0.003)	(0.012)	(0.007)	(0.079)	(0.004)
Control	YES	YES	YES	YES	YES	YES	YES	YES	YES
_cons	−8.318 **	3.244 ***	1.781 **	−2.747	3.401 ***	−1.826	2.264 ***	−10.561 *	2.900 ***
	(2.802)	(0.635)	(0.809)	(2.078)	(0.346)	(2.179)	(0.600)	(5.475)	(0.484)
Province	YES	YES	YES	YES	YES	YES	YES	YES	YES
Year	YES	YES	YES	YES	YES	YES	YES	YES	YES
N	143	104	156	208	195	195	208	143	260
R2	0.996	1.000	0.999	0.996	1.000	0.997	0.999	0.996	0.999

. The results indicate that new digital infrastructure significantly reduces agricultural carbon emissions in the eastern region (coefficient = −0.115), slightly increases emissions in the central region (coefficient = 0.019), and has no significant effect in the western region. This suggests that new digital infrastructure effectively promotes agricultural carbon reduction in the east but may stimulate carbon emissions in the central region, with an insignificant impact in the west.

Specifically, the eastern provinces enjoy pronounced geographic advantages and a legacy of policy leadership; their higher level of economic development has made them frontrunners in digital transformation [22]. As a result, new digital infrastructure has advanced swiftly, accelerating agricultural modernization and smarter management, and thereby fostering low-carbon, green farming practices. In contrast, the central region—despite its expansive plains and favorable agronomic conditions that underpin its role as the country’s core grain belt—faces persistent natural constraints and comparatively delayed policy reforms [46]. Slower economic growth has left the rural digital infrastructure underdeveloped. During the expansion phase, the carbon savings typically expected from digital upgrades may be eclipsed by the surge in energy required to build and operate new facilities, potentially causing net agricultural emissions in the central provinces to rise.

4.3.2. Heterogeneity in the Level of New Digital Infrastructure Development

Rural digital development requires not only the hardware support of new digital infrastructure but also coordinated software support, such as digital finance and digital technologies. This study focuses on new digital infrastructure but recognizes that average effects may mask province-level differences in digital infrastructure development.

Similar to traditional infrastructure development, new digital infrastructure involves a long cycle from investment and construction to usage and effectiveness. Related research has found that the economic and environmental effects of digital development may exhibit stage characteristics [45]. At different stages—early construction, rapid growth, and mature expansion—digital infrastructure of varying levels faces different cost and efficiency profiles. Therefore, exploring how varying digital infrastructure levels differentially impact agricultural carbon reduction is necessary to provide a more comprehensive understanding of scale and efficiency.

Accordingly, the 31 provinces are classified into “leading” and “lagging” new digital infrastructure development groups based on whether their digital infrastructure index is above or below the mean. The grouped regression results are presented in Columns (4) and (5) of Table 5. The results show significant differences: for the leading provinces, the digital infrastructure coefficient is significantly negative (−0.093), indicating that well-developed digital infrastructure better supports smart agriculture development, management technology, and efficiency improvements, yielding favorable carbon reduction. Conversely, in lagging provinces, the coefficient is significantly positive (0.007), suggesting that digital infrastructure development may increase agricultural carbon emissions. This may be because early-stage infrastructure expansion requires high capital and energy resource consumption, and immature infrastructure systems cannot yet optimally support the green agricultural transformation, leading to net carbon emission increases.

4.3.3. Policy Support Heterogeneity

Furthermore, considering the significant policy-intensive features in the digital and green transformation practice, this study further explores the heterogeneity roles considering the digital and carbon-related policy supports in different regions.

• Heterogeneity in the Digital-Inclusive Finance Development Level

Digital-inclusive finance is a key financing channel for agricultural development and a critical digital complement to digital infrastructure. Studies show that inclusive finance offers diverse financial products that lower financing barriers for farmers’ expansion and entrepreneurship, facilitating the development of agricultural industrial chains and leading enterprises, and synergizing with digital infrastructure to promote high-quality agricultural transformation [47]. However, significant inter-provincial disparities exist in digital inclusive finance development, and whether this affects digital infrastructure’s agricultural carbon reduction effect warrants further quantification.

Based on the Peking University Digital Inclusive Finance Index, provinces are grouped into “developed” and “developing” digital-inclusive finance regions by whether their index is above or below the mean. Regression results are reported in Columns (6) and (7) of Table 5. The results show that new digital infrastructure significantly reduces agricultural carbon emissions in digital-inclusive finance developed regions (coefficient = −0.037) but has no significant effect in developing regions. This suggests that digital finance services such as mobile payments and online banking effectively broaden financing and improve financial service efficiency for small and micro enterprises and agriculture-related sectors, providing funding support for rural digital infrastructure and green agricultural industries, thereby better promoting green modern agriculture.

2.

Green Development Heterogeneity

The level of green, high-quality agricultural transformation is closely related to the overall green development level of each province. Differences in resource and environmental carrying capacity and green economic policies result in substantial provincial variation in green production and development. From this perspective, this study uses the Economic Greening Index from the China Green Development Index Report to classify provinces into “green development pioneers” and “followers” by whether their index exceeds the mean. The results are shown in Columns (8) and (9) of Table 5.

The empirical results indicate that new digital infrastructure significantly reduces agricultural carbon emissions in green development pioneer provinces (coefficient = −0.241) but increases emissions in follower provinces (coefficient = 0.008). Specifically, pioneer provinces pursuing “dual-carbon” goals usually implement comprehensive policies and funding supporting industry, energy, and technology greening. Their stronger policy inclination and support for green agriculture, combined with digital infrastructure development, more effectively advance high-quality agricultural transformation and accelerate emission reduction. In contrast, follower provinces with relatively lagging economic development provide weaker policy support for green development, limiting its spillover to green agriculture. The rural digital infrastructure remains underdeveloped, and its potential to empower green low-carbon agriculture awaits further realization.

5. Further Analysis: Testing the Influencing Mechanisms

This study empirically finds that new digital infrastructure significantly promotes agricultural “decarbonization”. Based on the theoretical mechanism analysis in the previous section, this part further empirically explores how new digital infrastructure exerts agricultural “decarbonization” benefits through scale economy effects and energy efficiency improvement effects. Using the typically adopted method that is widely used to discuss the mediation effect [48], this study mainly tests the relationship between the independent variable of “new digital infrastructure” and the different mediator variables using the two-way fixed-effect model, and the relationships between the mediator variables and the dependent variable are mainly interpreted by the theoretical and literature explanations.

5.1. Mechanism of the Scale Economy Effect

The development of new digital infrastructure, empowered by intelligence, digitization, and networking, can influence agricultural production costs and labor inputs to some extent, thereby affecting farmers’ production and management capacities as well as agricultural production expansion. To examine this, this study selects the crop planting area per farmer household (perarea) as a proxy variable for agricultural scale economy effects to empirically verify the impact of new digital infrastructure on scale economy benefits. Based on this, through theoretical analysis, the mechanism by which the scale economy influences agricultural carbon emissions is further explored. The empirical results are shown in Column (1) of

Table 6. Results of tests of the influencing mechanism.

	(1)	(2)
	Perarea	Eneffien
digi	0.051 ***	0.017 ***
	(0.011)	(0.003)
Control	YES	YES
_cons	11.423 ***	0.312
	(1.207)	(0.408)
Province	YES	YES
Year	YES	YES
N	403	175
R2	0.998	0.967

.

The empirical results indicate that the coefficient of the effect of new digital infrastructure on scale economy benefits is significantly positive at 0.051, suggesting that the development of new digital infrastructure effectively enhances farmers’ capacity to expand production and management, generating a significant scale economy effect. Related research and practical production experience show that the expansion of the farmland operating scale can significantly reduce agricultural carbon emissions [49]. Enhanced digital infrastructure and the widespread adoption of agricultural machinery could better improve the farmers’ production expertise and accelerate the uptake of low-carbon and more efficient farming technologies. These advances could reduce the marginal demand for agrochemicals and related energy inputs as production scales expand [50]. Meanwhile, broader regional growth in agricultural output not only raises the overall capacity but also invigorates upstream and downstream segments, from input production to processing, logistics, and retail, creating a more integrated and standardized value chain [51]. The resulting economies of scale and efficiency would better contribute to agricultural carbon emission control. In summary, Hypothesis 2a is supported.

5.2. Mechanism of the Energy Efficiency Improvement Effect

The development of new digital infrastructure brings the wider application of digital technologies such as big data and the internet, which can better assist modern agricultural management and affect the efficiency of agricultural resource and energy use. To test this, this study selects energy consumption per unit of output value in agriculture, forestry, animal husbandry, and fishery (eneffien) as a proxy variable for agricultural energy efficiency, empirically testing the impact of new digital infrastructure on agricultural energy efficiency. The empirical results are shown in Column (2) of Table 6.

The coefficient of the effect of new digital infrastructure on energy efficiency is significantly positive at 0.017, indicating that with the improvement of new digital infrastructure, it can better promote the development and application of smart grids, the Internet of Things (IoT), and new energy technologies in agricultural production and living, thereby enhancing energy efficiency. In the agricultural transformation, improving agricultural energy efficiency is a key task for China’s rural emission reduction and green agriculture development. Energy efficiency improvement in the overall agricultural production process can effectively optimize energy use and promote the use of clean energy and advanced energy-efficient equipment, such as solar dryers and energy-saving greenhouses [52]. In the long term, it could significantly reduce the energy input requirements, effectively cutting the use of fossil fuels and agricultural fertilizers, and thus promoting agricultural carbon reduction [53]. In summary, Hypothesis 2b is supported.

6. Research Conclusions and Policy Implications

6.1. Research Conclusions

Developing green agriculture is a key part of promoting the high-quality transformation of agriculture. Facing the era of the digital economy and “dual-carbon” constraints, it is essential to fully recognize the role of new digital infrastructure as a new quality that productivity empowers agricultural carbon reduction. Based on panel data from 31 Chinese provinces from 2011 to 2023, this study empirically investigates the effect of new digital infrastructure on agricultural carbon reduction using a fixed-effect model after analyzing the theoretical mechanisms. It also analyzes heterogeneity based on provincial “geographical location–digitalization level–policy support”, and finally discusses the mechanistic paths through economies of scale and energy efficiency improvements. The main findings are as follows:

• New digital infrastructure has a significant negative impact on agricultural carbon emissions, showing a clear carbon reduction effect. This finding remains robust after a series of robustness and endogeneity tests, fully reflecting the significant role of rural new digital infrastructure development in promoting modern and green agriculture.
• The effect of new digital infrastructure on agricultural carbon reduction shows significant heterogeneity. In terms of geographical location, new digital infrastructure development effectively promotes carbon reduction in eastern provinces but increases agricultural carbon emissions in central regions, with no significant effect in western regions. Regarding the digitalization level, it significantly suppresses agricultural carbon emissions in provinces with high-level new digital infrastructure but raises emissions in lagging provinces, reflecting the stage differences in energy input and emission reduction efficiency during different development phases. Considering the policy support level, new digital infrastructure development effectively promotes carbon reduction in provinces with advanced digital-inclusive finance but has no significant effect in less developed areas, indicating that other digital supporting measures facilitate the carbon reduction effect of new digital infrastructure. Regarding green development policy support, new digital infrastructure significantly reduces carbon emissions in green development pioneer provinces but increases emissions in follower provinces, reflecting how regional green development progress and policy focus influence the effectiveness of new digital infrastructure.
• New digital infrastructure mainly drives agricultural carbon reduction through two paths: stimulating economies of scale in agriculture and improving energy efficiency, thus promoting agriculture’s transition toward greater intelligence and greening.

6.2. Policy Implications

Based on the research findings, the following policy recommendations are proposed:

• The development of new digital infrastructure is a key foundational support for promoting low-carbon green agriculture. During the critical period of a new round of digital rural construction and smart agriculture development, efforts should accelerate the extension of new digital infrastructure to rural areas and promote broadband network popularization and 5G network coverage, achieving “5G in every county”. At the same time, the empowering role of new digital infrastructure should be fully leveraged, with technologies such as big data, the Internet of Things, and artificial intelligence comprehensively applied to the digital transformation of agricultural production, enhancing standardized and refined management and improving the efficiency of agricultural inputs.
• Tailor key strategies to local conditions by formulating differentiated provincial new digital infrastructure development and cooperation policies. Correctly recognize and accelerate bridging the “digital divide” between regions. Economically and digitally developed eastern provinces can further leverage their advantages to accelerate advanced technology research and development for the integration of digital technology and agriculture. For backward provinces, establishing a three-in-one policy support system of “technological empowerment + institutional innovation + ecological compensation” is necessary. The central government should increase its special transfer payments, with a focus on supporting the construction of digital infrastructure such as 5G base stations, Internet of Things sensors, and satellite remote sensing monitoring. In addition, innovation of the mechanism for realizing the value of agricultural carbon sinks is necessary. Pilot agricultural carbon sink projects in backward provinces should be included in the national voluntary emission reduction trading market. Referring to the revenue distribution model of forestry carbon sinks, part of the carbon trading revenue should be returned to farmers, and market incentives should be used to promote pollution reduction and carbon emission reduction. At the same time, a differentiated ecological compensation mechanism should be established to provide subsidies to farmers who adopt carbon sequestration measures such as conservation tillage and green manure planting.
• Strengthen the coordination of digital economy and green transformation policies to fully unleash the emission reduction dividends of new digital infrastructure. To this end, further promote the development of digital-inclusive finance, expand diversified financing channels and financial product systems, provide more convenient financing support for agricultural infrastructure projects, assist farmers and agricultural enterprises in production operations and technological innovation, and stimulate economies of scale in agriculture. Meanwhile, promote the implementation of supporting green policies related to industry, energy, and consumption; guide the orderly completion of new digital infrastructure projects; and especially focus on “carbon reduction” management during the early expansion phase, planning in advance the organic integration of digital infrastructure with the agricultural production chain to improve agricultural management and the efficient use of energy and resources.

7. Limitations and Further Studies

This study mainly investigates the impact of new digital infrastructure on agricultural carbon emissions from theoretical and empirical aspects. Some limitations still remain in this study.

From the indicator aspect, first, constrained by the data availability, the core explanatory variable of “new digital infrastructure” is constructed and mainly considers two aspects of physical infrastructure and knowledge capacity supports. To fully capture the “new” characteristics of digital infrastructure, more indicators reflecting contemporary features (e.g. 5G, data centers, and IoT) could be applied in further studies. Second, for the explained variable, this study mainly focuses on the carbon footprint in the planting process; additionally, a uniform set of emission factors was adopted in the same agricultural activities without accounting for technological or regional heterogeneity. With the improvement of the carbon emission calculation method, increasingly disaggregated emission factors can be extended to province-level agricultural inventories, as well as incorporating the livestock-related greenhouse gas emissions sources, such as enteric methane and manure CH₄ and N₂O.

At the methodological level, first, this study discussed the heterogeneity of the effect of new digital infrastructure development level using the grouped regression method. To comprehensively identify the “rebound effect” or “green paradox” of digital transformation, some non-linear regression methods such as the square term or threshold model could further be applied in the following studies. Second, in the mechanism testing the empirical method design, the method adopted in this study is more applied to investigate the underlying mediation paths, which could not effectively measure the scale or efficiency of the mediators. To fulfill this gap, other typical method such as the Sobel test or SEM model could be also applied in future studies.

From the policy assessment aspect, first, the empirical results found the heterogeneity of the effect of new digital infrastructure development, and the scale economy effect would be a significant transformation path. However, the potential ecological risks brought about by the excessive scale that existed in some industrial sectors indicates that the investigation of the optimal scale of the digital economy and agricultural agglomeration is also a worthy work. Second, this study discussed how those selected policies would influence the patterns between new digital infrastructure and carbon emissions, while more targeted policy effect assessment works are also of great significance.