Digital agriculture's impact on carbon dioxide emissions varies with the economic development of Chinese provinces

Achieving China's dual carbon objectives (carbon peak by 2030 and carbon neutrality by 2060) necessitates significant reductions in agricultural carbon emissions. Digital empowerment, defined as the complementary integration of digital technology with production factors, relations, and methods, offers a potential pathway towards sustainable agricultural practices. This study examines the complex relationship between digital empowerment and agricultural carbon dioxide (CO2) emissions in China. Previous research suggests that digital technologies can enhance efficiency and resource allocation, potentially leading to carbon emission reductions. However, the impact of digitalization on agriculture remains debated, particularly considering potential initial increases in resource use before efficiencies are realized. This paper seeks to address this gap by analyzing the nonlinear relationship between digital empowerment and agricultural CO2 emissions across diverse economic and agricultural contexts within China, specifically focusing on the impact of different levels of digital technology penetration. Understanding this relationship is crucial for policymakers aiming to leverage digital technologies to promote sustainable agriculture while also addressing concerns related to potential unintended environmental consequences, specifically in terms of carbon emissions.

Existing literature highlights the potential of digital technologies to improve agricultural productivity and sustainability. Studies have shown that digital agriculture can optimize resource use, reduce waste, and improve decision-making processes related to fertilizer application, pesticide use, and energy consumption. For instance, precision agriculture techniques enabled by digital technologies can lead to more targeted input application, minimizing environmental impact. Some studies have indicated a positive correlation between digital technology adoption and reduced carbon emissions in agriculture. However, others have cautioned against a simplistic view, pointing out potential trade-offs. Initial increases in technology-related energy consumption or intensification of production could lead to temporarily higher emissions before longer-term efficiency gains are realized. The literature also lacks a comprehensive analysis of the heterogeneous impact of digital agriculture across diverse economic contexts. This paper builds on existing literature by investigating the nonlinear nature of this relationship and considering the moderating role of economic development and regional agricultural characteristics.

This study utilizes panel data from 30 provincial-level regions in China from 2012 to 2021. The key explanatory variable is digital empowerment, measured using the input-output method. This method captures both direct and indirect linkages between the digital economy and the agricultural sector, providing a more comprehensive assessment compared to approaches focusing solely on direct impacts. The fully distributed consumption coefficient is used to calculate the digital empowerment index for each province and year, reflecting the complex interdependencies between digital industries and agriculture. The data for the digital empowerment index are obtained from the China Interregional Input-Output Table and the China Statistical Yearbook. The dependent variable is agricultural carbon emissions, calculated systematically from four primary sources: the inputs of agricultural production materials (fertilizers, pesticides, plastic films); rice planting (considering early, middle, and late rice); livestock and poultry breeding (using IPCC emission factors); and agricultural energy utilization. Data for these calculations are sourced from the China Statistical Yearbook, China Rural Statistical Yearbook, China Energy Statistics Yearbook, and China Agricultural Machinery Industry Yearbook. Several control variables are included: agricultural disaster rate, crop structure, industrial structure, urbanization rate, and rural population size. This study employs a two-way fixed-effects model to analyze the relationship between digital empowerment and agricultural carbon emissions, accounting for both regional and time-specific effects. Given the potential for nonlinearity, a quadratic term of the digital empowerment index is included in the model. Moreover, a mediating effects model is employed to examine the roles of factor allocation efficiency and carbon-intensive factor inputs in the overall relationship. Factor allocation efficiency is measured using the stochastic frontier analysis method applied to a translog production function. Carbon-intensive factor input is measured as the proportion of various cost components in the total agricultural input cost. An instrumental variable approach is used to address potential endogeneity issues, employing mobile phone penetration rates as an instrumental variable for digital empowerment. Furthermore, spatial econometric models are employed to analyze the spatial spillover effects of digital empowerment on agricultural carbon emissions. Robustness checks are conducted using various methods, including the SYS-GMM model, excluding key regions, and winsorizing the data. Heterogeneity analysis is performed by dividing provinces into subgroups based on economic development level and grain production.

The study's key findings support the hypothesis of a nonlinear, inverted U-shaped relationship between digital empowerment and agricultural carbon emissions. The two-way fixed-effects model reveals a significant positive coefficient for the linear term of digital empowerment and a significant negative coefficient for the quadratic term, confirming the inverted U-curve. The inflection point of this curve is calculated to be 0.0862. This indicates that, while initially, increasing digital empowerment might lead to higher emissions due to factors such as increased use of inputs, further increases beyond the inflection point eventually result in emission reductions, likely due to improved efficiency and resource optimization. The mediating effects analysis reveals that both factor allocation efficiency and carbon-intensive factor inputs play significant roles in the mechanism through which digital empowerment impacts carbon emissions. Improved factor allocation efficiency leads to lower emissions, while higher carbon-intensive factor inputs lead to higher emissions. The instrumental variable regression, designed to address potential endogeneity, confirms the robustness of the inverted U-curve relationship. Heterogeneity analysis reveals that the impact of digital empowerment on agricultural carbon emissions is significantly stronger in economically developed regions and major grain-producing areas than in less developed regions or non-grain-producing regions. Spatial econometric models reveal significant positive spatial spillover effects, implying that the impact of digital empowerment in one region can influence neighboring regions, which can either increase or decrease emissions depending on factors such as the level of development in the surrounding regions. Robustness checks employing different methodologies (SYS-GMM, excluding key regions, winsorizing) consistently support the inverted U-curve relationship. The findings are robust across various analytical techniques.

This research provides strong evidence of a nuanced relationship between digital empowerment and agricultural carbon emissions in China. The inverted U-curve relationship underscores the importance of considering the stage of digital development when implementing policies to promote sustainable agriculture. While initial increases in digitalization might temporarily increase emissions, sufficient investment and technology adoption beyond a certain threshold can lead to significant carbon reductions. The mediating mechanisms highlighted—improved factor allocation efficiency and reduced carbon-intensive factor inputs—offer valuable insights for targeted policy interventions. The findings regarding heterogeneity suggest that policies should be tailored to the specific economic and agricultural contexts of different regions. Economically developed regions and major grain-producing areas may be more receptive to the carbon-reducing effects of digital empowerment, while regions with less developed digital infrastructure or different agricultural structures may require different strategies. Spatial spillover effects suggest that policies aimed at promoting digital agriculture should consider regional interdependencies and potential diffusion or resonance effects.

This study demonstrates a significant nonlinear relationship between digital empowerment and agricultural carbon emissions in China, following an inverted U-curve pattern. The findings underscore the importance of considering the stage of digital development and regional heterogeneity when formulating policies to achieve carbon emission reduction targets in the agricultural sector. Further research could explore the specific policy instruments that would effectively promote digital agriculture beyond the inflection point of the inverted U-curve, focusing on strategies to mitigate any potential negative environmental consequences. Further investigating the long-term effects of digital agriculture on emissions, particularly in less developed regions, would also be highly valuable.

This study uses data from 2012-2021. Future studies should use more up-to-date data to verify the findings. The study relies on certain assumptions in the calculation of agricultural carbon emissions, such as the use of emission coefficients from other studies. The accuracy of the findings depends on the accuracy of the underlying data sources. While addressing potential endogeneity issues through instrumental variables, other unobserved factors might influence the relationship between digital empowerment and carbon emissions, hence the results reflect the contextual aspects of the selected time frame and regions within China.