Impact of digital economy development on agricultural carbon emissions: evidence from scale and efficiency effects

Abstract

Agricultural digitalization is increasingly viewed as a key pathway to modernizing agriculture and advancing low-carbon transformation. Using provincial panel data from China spanning 2011–2022, this study examines how agricultural digitalization affects agricultural carbon emissions and carbon emission efficiency from the perspectives of scale effect and efficiency effect. A two-way fixed-effects model is employed as the main analytical approach. The results show that agricultural digitalization significantly improves agricultural carbon performance. Specifically, a 1% increase in the level of agricultural digitalization leads to a 15.72% reduction in total agricultural carbon emissions and a 3.97% improvement in carbon emission efficiency, indicating that both scale and efficiency effects are operative. Mechanism evidence suggests that, at the current stage, the mitigation effect primarily works through incremental innovation rather than breakthrough innovation. The carbon-reduction effect is stronger in southern China and in provinces with more advanced digital development, and digitalization also exhibits spatial spillover effects. These findings suggest that accelerating agricultural digitalization, while strengthening region-specific support and innovation incentives, can contribute to the transition toward low-carbon agriculture.

1 Introduction

In the context of intensifying global climate change and strengthened low-carbon transition objectives, agriculture, as a major source of greenhouse gas emissions, has become a central topic in research on mitigation pathways and green development strategies. According to the Intergovernmental Panel on Climate Change (IPCC), agricultural production activities constitute a major source of anthropogenic greenhouse gas emissions (), the global food system accounts for more than one third of total anthropogenic greenhouse gas emissions worldwide (). The United Nations Environment Programme reports that approximately 40% of anthropogenic CH 4 emissions originate from agricultural production activities. Moreover, the greenhouse effect of CH 4 is more than 26 times that of CO 2 (; ). As one of the world's major agricultural producers, China's agricultural production activities constitute the largest anthropogenic greenhouse gas emission source aside from energy use and industrial processes (). Compared with the industrial sector, agricultural carbon emissions are characterized by dispersed sources, complex production processes, and strong constraints on technological substitution. Accordingly, achieving agricultural emission-reduction targets under the constraints of safeguarding food security and ensuring a stable supply of agricultural products remains a formidable practical challenge.

In recent years, the digital economy, represented by big data, cloud computing, artificial intelligence, and the platform economy, has developed rapidly and profoundly reshaped agricultural factor allocation patterns, industrial organization structures, and production decision-making mechanisms. Existing research on the impact of digitalization on agricultural carbon emissions has yielded two distinct sets of findings. One perspective argues that digitalization can reduce carbon emission intensity by promoting the precise allocation of inputs, intelligent management, and the diffusion of green technologies (; ; ). Another perspective emphasizes expansion mechanisms, arguing that digitalization leads to stronger market connectivity, large-scale mechanization, and rapid expansion of digital infrastructure, which may stimulate output growth and energy demand, thereby increasing carbon emissions under certain conditions (; ). However, the key issue in existing research is not whether digitalization has a positive or negative impact on agricultural carbon emissions, but rather that digitalization may operate through different pathways and exert varying effects on agricultural carbon emissions.

To systematically elucidate the differentiated mechanisms through which digitalization affects agricultural carbon emissions, it is first necessary to clearly define two core concepts: the scale effect and the efficiency effect. The scale effect refers to how digitalization alters total agricultural carbon emissions by influencing output scale, production structure, and the demand for high-carbon inputs at the extensive margin. Efficiency effects refer to how digitalization alters the carbon emission efficiency per unit of agricultural output by improving the allocation and utilization of input factors such as land, labor, machinery, fertilizers, pesticides, and irrigation at the intensive margin. In the agricultural production process of a given region, it is possible for carbon emissions per unit of output to decrease while total carbon emissions ultimately rise due to production expansion; conversely, if the production structure changes, total agricultural carbon emissions may decrease even if improvements in carbon emission efficiency are limited. Unfortunately, most existing literature examines total agricultural carbon emissions (; ) and agricultural carbon emission efficiency () in isolation, making it difficult to determine whether the observed trends are driving a genuine low-carbon transformation in agriculture or merely shifting the balance between extensive expansion and intensive improvements.

Accordingly, this study takes the level of agricultural digitalization as the core explanatory variable and characterizes agricultural carbon emission performance from two dimensions: total agricultural carbon emissions and agricultural carbon emission efficiency. From a differentiated innovation perspective, the paper applies threshold models, spatial models, and other methods to systematically examine the impact of agricultural digitalization on agricultural carbon emissions and the associated mechanisms. The primary contributions of this paper can be summarized in three aspects. First, unlike existing studies that largely focus on general tests of the correlation between digitalization and agricultural carbon emissions, this paper systematically examines the differentiated impacts of digitalization on agricultural carbon performance from the dual perspectives of total carbon emissions and carbon emission efficiency. It thereby reveals whether digitalization drives the low-carbon transition in agriculture through adjustments in the scale of emissions at the extensive margin or through efficiency improvements at the intensive margin. Second, this study incorporates both the phased characteristics of digitalization and regional interdependencies into the agricultural carbon performance analysis framework, systematically examining the nonlinear effects of agricultural digitalization on total carbon emissions and carbon emission efficiency, as well as its spatial spillover effects. Unlike existing studies that discuss threshold effects and spatial effects in isolation, this paper not only examines the differential roles of digitalization in agricultural low-carbon transformation across different development stages but also analyzes whether the local low-carbon benefits it generates can spread to surrounding areas, thereby identifying the stage-specific conditions and spatial boundaries under which digitalization delivers environmental performance. Third, unlike existing studies that treat innovation as a single transmission channel, this paper distinguishes between incremental innovation and breakthrough innovation. It examines the differentiated roles of these innovation types in the process by which agricultural digitalization influences carbon performance, thereby providing a deeper understanding of the underlying mechanisms through which digitalization empowers low-carbon transformation in agriculture.

The structure of the paper is organized as follows. Section 2 develops the theoretical framework and formulates the research hypotheses. Section 3 introduces the research design, data sources, and the measurement of variables. Section 4 presents the empirical findings, robustness tests, and mechanism analyses. Section 5 provides further analysis of threshold effects and spatial effects. Section 6 concludes the study and offers relevant policy implications.

2 Theoretical analysis and research hypotheses

2.1 Analytical framework

From the perspective of production function theory, agricultural digitization can be understood as a technology-enhancing factor that not only shifts the production frontier but also influences the generation of agricultural carbon emissions. In other words, since agricultural output (Y) is jointly determined by land (L), labor (N), capital and machinery (K), intermediate inputs such as fertilizers and pesticides (M), energy inputs (E), and the level of digitization (D), the agricultural production process can be expressed as:

$Y = A \left(\right. D \left.\right) \cdot F \left(\right. L , \textrm{ } N , \textrm{ } K , \textrm{ } M , \textrm{ } E \left.\right)$

In this context, A(D) represents the technology-enhancing effect resulting from agricultural digitization. As the level of digitization increases, improvements are made in information acquisition, factor allocation, timing of input application, and management practices during the production process, thereby driving growth in agricultural output and altering the mix of inputs.

Agricultural carbon emissions, as an unintended byproduct, are the aggregate result of emissions from various agricultural subsectors at the provincial level. These include emissions from crop production activities, emissions from major livestock and poultry farming activities, as well as emissions associated with the use of agricultural inputs.

$C = C^{\text{crop}} + C^{\text{livestock}} + C^{\text{input}}$

Specifically, C crop refers to carbon emissions from crop production, C livestock refers to emissions from livestock and poultry farming, and C input refers to carbon emissions associated with agricultural inputs.

The impact of digitalization on agricultural carbon emissions does not necessarily operate through the same micro-level mechanisms across all subsectors; rather, it can exert its influence through sector-specific channels and manifest as a comprehensive net effect at the provincial level. Specifically, in crop production, digitalization primarily affects input allocation and emission intensity through precision fertilization, smart irrigation, optimized scheduling of agricultural machinery, and crop condition monitoring; in livestock farming, digitalization primarily influences methane and nitrous oxide emissions through precision feeding, monitoring of rearing environments, disease early warning systems, and digital manure management. Therefore, the digitalization indicators used in this study are best understood as measures of the level of digital development in agriculture or the digital enablement environment, rather than as a uniform adoption rate of a single production technology.

At the same time, agricultural carbon emissions depend not only on the scale of production itself but also on the emission intensity per unit of output. Let the total agricultural carbon emissions be C; then, this can be further expressed as:

$C = Y \cdot \theta \left(\right. D , \textrm{ } Z \left.\right)$

Here, θ(D, Z) represents the carbon emission intensity per unit of output, and Z denotes other conditional variables that influence emission intensity, such as agricultural structure, resource endowments, institutional environment, and technological adoption capacity. Digitalization may reduce carbon emission intensity per unit of output that is, decrease θ through precision fertilization, smart irrigation, and optimized scheduling of agricultural machinery; alternatively, it may increase total output Y by expanding operational scale, enhancing market connectivity, and increasing the intensity of machinery and energy use. Therefore, the impact of digitalization on total agricultural carbon emissions depends on the combined direction and relative magnitude of its effects on output scale and emission intensity. Consequently, the analysis of agricultural carbon performance can be divided into two interrelated yet distinct theoretical dimensions.

The scale effect refers to the impact of digitalization on total agricultural carbon emissions by altering the scale of agricultural output, the organization of production, and the absolute volume of carbon-related inputs. As shown in the above equation, when digitalization significantly increases the scale of agricultural output Y, total carbon emissions C may still rise even if the emission intensity per unit of output θ decreases. For example, if digitalization expands market access, leads to larger-scale operations, and increases the use of machinery, electricity, cold chain logistics, and other energy-intensive infrastructure, total carbon emissions may rise. However, scale expansion does not necessarily imply an increase in carbon emissions per unit of output, as economies of scale, improved organizational methods, and more efficient coordination mechanisms may also reduce emission intensity. Therefore, the scale effect focuses on the marginal impact on carbon performance that is, whether digitalization ultimately expands or curbs total carbon emissions.

The efficiency effect refers to how digitalization influences agricultural carbon emission efficiency at the intensive margin by altering the allocation of factors and the generation of emissions within the production function. If agricultural carbon emission efficiency is defined as the output level corresponding to a unit of carbon emissions, or more generally understood as the ability to control unwanted outputs under given input-output conditions, then digitalization through the use of precision monitoring technologies, intelligent decision support systems, and smart equipment can improve the allocation and timing of inputs such as fertilizers, pesticides, irrigation, machinery, and energy. This reduces redundant inputs and lowers carbon emissions per unit of output, i.e., reduces θ (; ; ). In the empirical analysis that follows, the efficiency dimension is represented by agricultural carbon emission efficiency as a proxy variable. This indicator reflects performance determined by both desired and undesired outputs under given input conditions, thereby corresponding to the concept of intensive margins in the aforementioned theory. It should be noted that improved efficiency does not necessarily imply a reduction in total carbon emissions, as the same efficiency gains may stimulate output expansion or lead to structural adjustments. Conversely, even with limited efficiency gains, total carbon emissions may still decrease if digitalization reduces production scale or drives a shift toward low-carbon-intensive activities. Therefore, total carbon emissions and carbon emission efficiency do not represent a simple substitution relationship in terms of variable structure; rather, they respectively characterize different theoretical dimensions of carbon performance.

Furthermore, taking the logarithm of the above equation and evaluating the partial derivative yields:

$\frac{\partial ln \textrm{ } C}{\partial D} = \frac{\partial ln \textrm{ } Y}{\partial D} + \frac{\partial ln \theta}{\partial D}$

This equation indicates that the impact of digitalization on total carbon emissions can be broken down into two components: one is the change in output scale resulting from digitalization, which reflects the scale effect; the other is the change in emissions intensity per unit of output resulting from digitalization, which reflects the efficiency effect. When the emission reduction effect of the former exceeds that of the latter, total agricultural carbon emissions may increase; conversely, they may decrease. Consequently, the theoretical impact of digitalization on total agricultural carbon emissions is uncertain, whereas its effect on improving the efficiency of agricultural carbon emissions has a clearer direction.

In addition to crop production, agricultural digitalization may also affect emissions from livestock production. Through precision feeding, environmental monitoring, disease surveillance, and digital manure management, digital technologies can improve feed use efficiency, reduce waste, and mitigate methane and nitrous oxide emissions from animal husbandry. Therefore, the carbon consequences of digitalization are not limited to crop-related inputs, but may also extend to livestock-related emission sources.

Based on the above analysis, this study proposes the following hypothesis:

Hypothesis 1. Agricultural digitization is expected to improve the carbon efficiency of agriculture by optimizing the allocation of factors of production, production decisions, and process control; however, its impact on total agricultural carbon emissions remains theoretically uncertain, as efficiency gains may occur simultaneously with an expansion in production scale.

2.2 Innovation heterogeneity and carbon transfer mechanisms

Innovation is a key mechanism through which digitalization influences carbon performance, but different types of innovation affect economies of scale and efficiency in distinct ways. Digitalization can reduce information costs, strengthen data feedback, promote collaboration, and improve the allocation of innovation resources, thereby stimulating green technological innovation in agriculture. However, the impact of innovation on agricultural carbon emissions varies depending on whether it manifests primarily as incremental innovation or breakthrough innovation. Incremental innovation refers to improvements made within existing technological frameworks, such as more precise fertilization, optimized water-saving irrigation, and smarter agricultural machinery scheduling. The impact of incremental innovation on carbon emissions lies in improving agricultural carbon emissions by reducing emission intensity within the existing production system. Breakthrough innovation, on the other hand, may reshape the production frontier through new organizational forms, service models, or production technologies. Consequently, breakthrough innovation may influence carbon emission intensity not only by pushing the technological frontier outward but also by altering production organization, market coverage, and the feasible boundaries of agricultural operations (; ). Consequently, if digitalization primarily induces incremental innovation, its impact on agricultural carbon emissions is more likely to manifest as optimization within the existing system that is, improved carbon efficiency. If, however, digitalization induces breakthrough innovation, its impact on agricultural carbon emissions may reshape production methods, significantly enhancing efficiency while potentially giving rise to new production models, thereby simultaneously influencing both economies of scale and efficiency gains.

Based on the above analysis, this study proposes the following hypothesis:

Hypothesis 2. Agricultural digitization influences agricultural carbon emissions through heterogeneous innovation channels; incremental innovations primarily enhance efficiency gains, while breakthrough innovations may simultaneously influence both efficiency gains and scale effects.

2.3 Threshold effects of agricultural digitalization on carbon emissions

The impact of agricultural digitalization on agricultural carbon emissions may exhibit nonlinear and threshold-dependent characteristics. This can be attributed to the early stage of agricultural digitalization, during which inadequate digital infrastructure and low technology adoption rates limit its impact on agricultural carbon emissions. Once agricultural digi