With the implementation of the “Broadband China” and “Digital Rural” strategies, digital technology has become a significant input factor in agricultural production in the new era. Based on the theory of the digital economy, the construction of a digital rural construction can leverage the substitution and integration effects of the digital economy to alleviate the mismatch of agricultural resource factors and improve agriculture’s total green factor productivity (Honghai and Xinmin, 2023). Digital rural construction can reduce farmers’ transaction costs and alleviate market information asymmetry, effectively lowering agricultural production costs and environmental pollution (Rolandi et al., 2021). This construction has accelerated the widespread dissemination of advanced technologies in rural areas (Popescu et al., 2020), and technological innovation is a crucial measure to promote agricultural carbon reduction. It facilitates the recombination and efficient allocation of various production factors, thereby improving production efficiency (Acemoglu and Restrepo, 2018). Due to specific differences in resource endowments among different regions, the impact of digital rural construction on agricultural carbon emission intensity varies across areas.

Hypothesis 1:

Digital rural construction is beneficial for reducing agricultural carbon emission intensity.

Hypothesis 2:

The impact of digital rural construction on agricultural carbon emission intensity has regional differences.

The government can disseminate information on agricultural low-carbon production technologies to farmers through rural digital platforms. Farmers can utilize the Internet and social media to access high-quality educational resources, enabling them to acquire advanced green production technology quickly to enhance their labor skills, knowledge level, and environmental awareness. This can influence rural labor behavior, encouraging farmers to adopt green production practices beneficial for agricultural carbon emission reduction (Ma et al., 2022). The construction of digital rural areas can overcome the limitations of traditional rural areas in terms of talent, resources, time, and space, thereby improving resource allocation efficiency, promoting the enhancement of green total factor productivity in agriculture, and exerting a substitution effect on the rural labor force, alleviating issues related to insufficient rural labor. Digital technology, as a fundamental driver in digital rural area construction, can catalyze changes in agricultural workers’ skills and other elements, forming digital agrarian productivity (Junge and Qinmei, 2023). In promoting digital rural construction, digital technology can lower the cost of farmers’ access to market information (Mary George et al., 2016; Song et al., 2020), enhance human capital by focusing on farmers’ digital literacy, reinforce environmental awareness, and encourage green production practices (Aldieri et al., 2019). The proliferation of rural digitization has paved new pathways for developing low-carbon agriculture.

Hypothesis 3:

The construction of digital agriculture can reduce the intensity of agricultural carbon emissions by promoting the level of rural human capital.

Currently, agricultural management entities in China predominantly consist of small-scale operations, and the construction of digital rural areas constitutes projects with significant externalities and essential public welfare attributes. It is a comprehensive undertaking involving various components, such as network infrastructure construction and the enhancement of rural human capital, which necessitate financial backing. However, farmers typically need more financial resources to undertake such endeavors. The level of economic development and the availability of financial support are critical factors influencing the effectiveness of digital rural construction (Xing et al., 2023). Moreover, financial support for agriculture can facilitate the aggregation of information, technology, and talent, thereby aiding in the high-level development of digital rural areas (Chunlin et al., 2024).

Hypothesis 4:

financial support for agriculture can positively influence the impact of digital rural construction on agricultural carbon emission intensity. Based on the above theoretical analysis, the theoretical framework diagram of the impact of digital rural construction on agricultural carbon emission intensity was constructed, as shown in Figure 1.

AHP and Delphi methods are commonly employed to determine indicator weights, yet they entail significant subjectivity. Therefore, the entropy method is utilized to ascertain the weights of each indicator, enhancing the scientific validity of the evaluation results, which boast high credibility and accuracy. The specific application methods of the entropy method can be found in some literature (Chen and Zhang, 2023).

①Dimensionless treatment of indicators

converting the indicator values of each indicator to 0–1, which can avoid the impact of different indicator dimensions.

The dimensionless processed indicators are standardized and translated to obtain the standardized indicator values: X t i j ″ = 0.99 × X t i j ′ + 0.01

③Calculate the proportion of indicator value

The carbon emission measurement method based on IPCC (Intergovernmental Panel on Climate Change) carbon emission coefficient is the common method to estimate carbon emission at present with the advantages of simple calculation process, easy popularization and less data demand. In china there are six primary sources of agricultural carbon emissions (Li et al., 2011), including pesticides, chemical fertilizers, agricultural fuel, agricultural plastic film, agricultural sowing areas, and agricultural irrigation areas. The corresponding carbon emission coefficients are as follows (Li et al., 2011; West and Marland, 2002; Zhi and Gao, 2009; Wu et al., 2007; Dubey and Lal, 2009): 4.9341 kg/kg,0.8956 kg/kg, 0.5927 kg/kg, 5.180 kg/kg, 312.60 kg/km², 25 kg/hm². The total agricultural carbon emissions are calculated as follows: ac = Σ ac i = Σ T i × δ i (1)

Where ac represents the total agricultural carbon emissions, ac_i represents the carbon emissions of various carbon sources, T_i and δ_i respectively represent the actual consumption of multiple carbon sources and their corresponding carbon emission coefficients.

To test the impact of digital rural construction on agricultural carbon emission intensity, the benchmark regression model constructed is as follows: acintensity it = a 0 + a 1 digcounty it + a 2 controls it + μ i + λ t + ε it (2) Where acintensity_it is the agricultural carbon emission intensity of Province i in year t, digcounty_it is the digital rural construction level of Province i in year t, and controls is a series of control variables. μ_i and λ_t refer to the fixed province effect and time effect, respectively, and ε_it refers to the random disturbance term.

Utilizing the stepwise regression method (Wen and Ye, 2014) for testing, the model is constructed as follows: acintensity it = a 0 + cdigcounty it + a 2 controls it + μ i + λ t + ε it (3) lnhuman it = b 0 + adigcounty it + b 2 controls it + μ i + λ t + ε it (4) acintensity it = c 0 + c ′   digcounty it + blnhuman it + c 3 controls it + μ i + λ t + ε it (5) Where rural human capital is the intermediary variable, c is the total effect coefficient of digital rural construction affecting agricultural carbon emission intensity; a is the effect of the core explanatory variable of digital rural construction on the intermediary variable of rural human capital; c' is the direct effect of digital rural construction on agricultural carbon emission intensity after controlling the influence of rural human capital; b is the indirect effect of the intermediary variable rural human capital on agricultural carbon emission intensity after controlling the impact of the core explanatory variable digital rural construction.

To further analyze the regulatory effect of financial support for agriculture on the impact of digital rural construction on agricultural carbon emission intensity, the regulatory effect model is developed as follows: acintensity it = m 0 + m 1 digcounty it + m 2 financialsupport it + m 3 digcounty it financialsupport it + m 4 controls it + μ i + λ t + ε it (6)

(1) Replace Core Explanatory Variables. Carbon emissions from food production are a significant component of agricultural carbon emissions. To test robustness, the intensity of agricultural carbon emissions is replaced with the carbon emission intensity from food production, and the baseline regression is performed again. The results are presented in Table 4 (1).

The findings from these three robustness tests consistently show a negative and significant coefficient for digital rural construction, affirming the robustness of the baseline regression conclusion: the construction of digital countryside can significantly reduce agricultural carbon emission intensity.

Using the second lag of digital rural construction (L2. digcounty) as an instrumental variable helps avoid endogenous estimation bias caused by two-way causality. The initial phase of digital rural construction creates favorable conditions for subsequent stages, satisfying the relevance requirement. Early-stage digital countryside construction reduces agricultural carbon emission intensity in the current period, meeting the homogeneity requirement. The two-stage least squares method (IV_2SLS) was employed to re-examine the impact of digital rural construction on agricultural carbon emission intensity (see Table 5). In the first-stage regression results, the instrumental variable L2. digcounty is positively correlated with the endogenous variable (discount) at the 1% level, indicating that the chosen instrumental variable is sufficiently associated with the endogenous variable. In the under identification test, the LM value is 112.230, passing the 1% significance test, which confirms the absence of under identification issues. The Cragg-Donald Wald F-value is 4,276.309, exceeding the critical value of 16.38 at the 10% significance level for the Stock-Yogo weak instrumental variable test, indicating no weak instrumental variable issues. The Hansen J statistic indicates no over identification problem, affirming the appropriateness of the selected instrumental variable. After addressing the endogeneity problem using the instrumental variable method, digital rural construction continues to show a significant negative correlation with agricultural carbon emission intensity at the 1% level, confirming the robustness of the result.

Divide the country into two categories based on whether it is a major grain-producing area. The regression results are shown in Table 6. The regression coefficients for digital rural construction are negative and pass the significance tests at 1% and 5%. However, the degree of influence varies across different regions. Digital rural construction significantly inhibits agricultural carbon emission intensity in major grain-producing areas. This effect is due to the country’s favorable resource allocation towards these grain-producing areas. On the one hand, the major grain-producing areas bear the heavy responsibility of maintaining national food security, the state attaches great importance to the development of major grain-producing areas, and has introduced a series of targeted policies to promote the construction of digital countryside, such as providing a number of financial funds for the construction of digital countryside in major grain-producing areas for infrastructure construction, digital technology research and development and application promotion. The main grain-producing areas have relatively good network coverage and perfect infrastructure, which provides the necessary basic conditions for the construction of digital countryside. On the other hand, major grain-producing areas usually have large areas of farmland which facilitate the large-scale application of digital technology in agricultural production. The observed regional differences, attributed to varying resource endowments, confirm hypothesis 2.

This part analyzes the impact of the sub-dimension of digital rural construction on agricultural carbon emission intensity (see Table 7). Rural broadband access users and the average mobile telephone ownership per 100 rural households significantly reduce agricultural carbon emission intensity, each passing the 1% significance test. The total index of digital inclusive finance within the construction of digital rural finance and the length of rural delivery routes within the construction of digital rural service platforms show no significant effect on carbon emissions. But when the level of digital inclusive finance is above the 0.25 quartile value (lndigfi ≥ 5.106) and the length of rural delivery route is above the 0.75 quartile value (lndelroute ≥ 12.154), the agricultural carbon emission intensity is also significantly reduced, and both pass the significance test of 10%. In other words, with the continuous improvement of digital inclusive finance and the length of rural delivery routes, the effect on agricultural carbon emission reduction will gradually increase, and there is a nonlinear relationship between them. The sub-dimensional construction level is related to the overall level of digital rural construction. The mobile phones and the Internet enables farmers to obtain weather information and information of agricultural materials products earlier, so they can implement scientific and reasonable irrigation and fertilization, and avoid environmental problems caused by excessive use of agricultural materials. The popularization of digital inclusive finance and the construction of express routes require a large amount of investment in the early stage, then the carbon reduction effect will gradually appear in the later stage, so as to achieve the purpose of reducing agricultural carbon emission intensity.