Rural E-commerce and air pollution: evidence of dual mechanisms from production inputs and supply-chain expansion

Introduction: Rural e-commerce is reshaping agricultural production and circulation in China, improving market access but also raising concerns about its environmental impact. Whether its rapid expansion improves or worsens air quality remains unclear.

Methods: Using county-level data from 2011 to 2021, this study exploits the staggered rollout of the E-commerce into Rural Areas Comprehensive Demonstration (RECD) program and applies a multi-period Difference-in-Differences model to estimate the policy’s effect on PM2.5 concentrations.

Results: The RECD policy significantly increases PM2.5 levels. Further analysis shows two channels: intensified agricultural inputs and expanded logistics and warehousing. The effect is strongest in regions with higher agricultural dependence, better terrain accessibility, and higher initial levels of economic development and e-commerce.

Discussion: Rural e-commerce improves market efficiency but also introduces environmental risks by altering production and circulation patterns. Strengthening green technologies, clean logistics and environmental regulation is essential to ensure a sustainable rural digital transformation.

1 Introduction

The diffusion of e-commerce in rural areas is rapidly reshaping the link between agricultural production and markets worldwide. It has also become a focal topic in development economics and agricultural policy research. According to the Food and Agriculture Organization of the United Nations (FAO), the development of rural e-commerce not only significantly reduces production and circulation costs and improves resource allocation efficiency, but also broadens market access for farmers, alleviates poverty, reduces resource waste in rural areas, and fosters the transition toward green agriculture (Fao and Zju, 2021). In this context, countries have adopted strategic approaches to rural e-commerce that differ in form yet share common objectives. In India, the construction of a unified nationwide Electronic National Agriculture Market (e-NAM) serves as the core initiative to break down regional market barriers, enhance price transparency, and strengthen farmers’ bargaining power (Manogna, 2025). In the European Union, the “Smart Villages” initiative focuses on strengthening digital infrastructure and community-based innovation, promoting deep integration between rural e-commerce and local value chains (Hlaváček et al., 2023). In China, the systematic “E-commerce into Rural Areas Comprehensive Demonstration” program has simultaneously advanced county–township–village logistics systems, service stations, and digital skills training, contributing to the construction of a comprehensive rural e-commerce ecosystem (Li and Gan, 2025).

However, behind this rapid boom, there may be environmental consequences that have not received sufficient attention, giving rise to new discussions on sustainability. On the one hand, some scholars emphasize the positive environmental effects of e-commerce and argue that it can significantly reduce pollution by improving production efficiency and optimizing resource allocation (Abukhader, 2008; Chen, 2019; Du et al., 2023). Existing evidence indicates that the carbon emission intensity of online retail is generally lower than that of offline retail (Edwards et al., 2010; Zhao et al., 2019). With the widespread adoption of e-commerce, the digitalization of services and the optimization of supply chains can reduce the excessive use of chemical fertilizers and pesticides in agricultural production (Wang et al., 2022; Ji et al., 2023). They also enhance the production and marketing efficiency of traditional agricultural products, promote product standardization and quality upgrading, and thus help facilitate agricultural green transformation (Li et al., 2021). On the other hand, another strand of research highlights the environmental costs of e-commerce, suggesting that it may increase carbon emissions and, in some cases, even offset its potential emission-reduction benefits (Pålsson et al., 2017; Hidayatno et al., 2019; Cheba et al., 2021). In the logistics segment, high-frequency and fragmented transportation operations may lead to increased freight emissions (Jaller and Pahwa, 2020; Siragusa et al., 2022), and in terms of waste generation, e-commerce also results in large volumes of packaging waste (Kim et al., 2022; Pinos et al., 2022; Syed Ali et al., 2024).

Although extensive research has examined the environmental impacts of rural e-commerce, important gaps remain. Existing studies have not sufficiently grounded their analyses in the distinctive characteristics of the agricultural and agri-food sectors. As a result, they largely overlook environmental effects related to changes such as accelerated agricultural mechanization and the expansion of cold-chain warehousing. In addition, the literature provides only limited explanations for the pronounced regional heterogeneity in environmental outcomes. Whether rural e-commerce promotes green transformation or leads to pollution accumulation depends heavily on development paths and regional contexts, yet key moderating mechanisms such as institutional environmental constraints, market incentives for green products, and the adoption of green technologies have not been systematically incorporated into analytical frameworks or rigorously tested with causal methods.

Against this backdrop, the present study makes several contributions to the literature. First, exploiting the staggered rollout of the E-commerce into Rural Areas Comprehensive Demonstration (RECD) program as a quasi-natural experiment and constructing a county-level panel dataset covering 2011–2021 linked with PM₂.₅ concentration data, we employ a difference-in-differences (DID) framework to rigorously identify the causal impact of rural e-commerce development on local environmental quality, thereby providing credible causal evidence. Second, by jointly incorporating agricultural production input adjustments and the expansion of supply-chain logistics into a unified analytical framework, we identify two distinct environmental pathways—production and circulation—through which rural e-commerce affects pollution, enabling a more comprehensive understanding of its environmental impact. Third, by introducing green technology adoption, market incentives, and institutional environmental constraints as moderating factors, we further uncover the sources of regional heterogeneity in environmental effects and provide policy-relevant insights for guiding rural e-commerce toward a development model that balances economic efficiency with environmental sustainability.

The remainder of this paper is structured as follows. Section 2 presents the theoretical framework and develops the research hypotheses. Section 3 introduces the research design, including data sources, variable construction, and econometric models. Section 4 reports the empirical results, including baseline estimates, mechanism tests, and moderation analyses. Section 5 concludes with key findings, policy implications, and directions for future research.

2 Theoretical analysis and hypothesis

The RECD program is a major policy initiative to promote the rural digital economy. It aims to strengthen the rural e-commerce ecosystem and improve market conditions for farmers’ entrepreneurship and income growth. Its primary objectives focus on enhancing agricultural performance and revitalizing county economies through the construction of online trading platforms, the optimization of supply chain organization, and the upgrading of rural industries. However, the policy does not place environmental protection at its core; its institutional design emphasizes economic outcomes while largely overlooking potential environmental externalities. Against the backdrop of the rapid expansion of rural e-commerce, an important question arises: while facilitating the restructuring of agricultural production and distribution systems, might the RECD program also generate environmental impacts by altering agricultural input structures and increasing logistics demand? Building on existing research, this study identifies two potential channels. First, e-commerce-driven order agriculture, large-scale production and standardization may increase the use of fertilizers, pesticides and mechanical power. This intensifies agricultural non-point source pollution. Second, the surge in e-commerce transactions may stimulate the expansion of warehousing, transportation, and cold-chain logistics, increasing energy consumption and transportation emissions and putting pressure on regional environmental quality.

2.1 Agricultural input intensification mechanism

Rural e-commerce influences agricultural production and its environmental outcomes mainly through two opposing effects. On the one hand, the theory of division of labor suggests that the expansion of market scale deepens specialization and pushes production systems toward larger and more standardized operations. With rural e-commerce continuously expanding market boundaries, stabilizing orders and forming persistent market constraints, agricultural production is no longer oriented only toward fragmented and seasonal demand (Li et al., 2024). Instead, producers need to achieve larger output scale, more stable supply, and more uniform product specifications. Under such market pressure, they tend to increase the use of fertilizers and pesticides and accelerate mechanization in order to secure yield, appearance, and supply continuity. This strengthens input intensity, increases resource consumption, and raises environmental burdens (Lin and Li, 2023). As a result, a scale effect characterized by production expansion and intensified input emerges, leading to an increase in pollution emissions (Jiang et al., 2020; Yang et al., 2024).

On the other hand, the theory of induced technological change emphasizes that technological progress does not occur automatically. It is induced by market demand, factor constraints, and institutional incentives. When rural e-commerce connects producers to premium markets for high-value agricultural products, producers have stronger incentives to optimize production structures and adopt more efficient and greener technologies (Qiu et al., 2024; Xie et al., 2025). Product upgrading and technological improvement may reduce pollution intensity per unit of output and help mitigate environmental pressure during production (Gamage et al., 2023).

However, according to the World of Organic Agriculture and Trends 2025 report, certified organic farmland in China accounts for only about 0.5 percent of total arable land. Organic food accounts for less than 2 percent of the total food market, and per capita organic consumption is only about 10.7 euros, far below the global average (FiBL and Of, 2025). These facts indicate that the scale and intensity of green and organic product demand in China remain limited. At the current stage, the technology and structural upgrading effects driven by market demand are unlikely to dominate. Their overall influence is weaker than the scale effect generated by production expansion.

2.2 Cold-chain logistics expansion mechanism

In addition to the production stage, rural e-commerce also generates environmental externalities through supply-chain expansion. By reshaping transaction modes and enlarging market coverage, rural e-commerce transforms the agricultural circulation system and produces new environmental impacts. From the perspective of transaction cost theory, e-commerce platforms significantly reduce information search, bargaining, and monitoring costs. This stimulates consumers to engage in more frequent and diversified consumption activities, so that transactions once occasional and intermittent become regular and high-frequency (Malone et al., 1987; Teo and Yu, 2005). From the perspective of spatial economics, rural e-commerce weakens geographical constraints and significantly improves market accessibility, effectively mitigating the traditional distance effect (Fan et al., 2018). It extends sales boundaries from local to cross-regional and even national markets, thereby connecting producers with a much larger and more widely distributed consumer base (Wei et al., 2020). The decline in transaction costs and the expansion of market scope jointly increase transaction frequency, shifting agricultural circulation from low-frequency, batch-based transactions to more continuous and intensive market operations.

In this process, environmental externalities at the supply-chain level gradually emerge. As the market space expands and order density increases, the demand for logistics and warehousing rises, and transportation activities become more frequent and larger in scale (Hanlin and Fan, 2025). For fresh agricultural products, strong perishability and strict time requirements create a rigid dependence on cold-chain logistics (Chen et al., 2025b). This leads to the rapid construction and continuous operation of origin-based pre-cooling facilities, regional cold-chain distribution centers, and terminal cold-storage warehouses within a wider spatial scope. A large and continuously operating low-temperature logistics network is thus formed. Under current technological conditions, this network is highly energy-intensive. When combined with increasingly frequent transportation activities, it becomes an important source of pollution within the rural e-commerce supply chain. According to relevant statistics, in 2020, transportation and product returns associated with e-commerce accounted for 37% of total greenhouse gas emissions. By 2030, the number of delivery vehicles worldwide is projected to increase by 36% to approximately 7.2 million, which will not only lead to an additional six million tons of carbon dioxide emissions but also extend urban commuting time by 21% (Igini, 2024).

2.3 Moderating role of institutions, green technology, and market incentives

In fact, the environmental effects of rural e-commerce exhibit complex non-linear patterns (Liang et al., 2021), and their specific manifestations are deeply contingent on the surrounding development context. Technological innovation provides the basis for reducing pollution intensity. Market transformation reshapes producers’ incentives. Institutional innovation helps clarify the relationship between government and market, strengthens regulation when necessary, and supports the effective functioning of market mechanisms. This logic is consistent with Ecological Modernization Theory (Gibbs, 2017), which emphasizes that environmental improvement relies on the joint functioning of markets, technology, and institutions. Therefore, a systematic analytical framework can be constructed from three dimensions: markets, technology, and institutions, to explain how different conditions moderate the environmental consequences of rural e-commerce.

First, the preference of the demand side acts as an endogenous moderating mechanism that conditions how production expansion translates into environmental outcomes. When consumer markets show strong preferences for green, organic and high-quality agricultural products and are willing to pay a premium, the role of e-commerce platforms changes. Instead of simply amplifying scale, they begin to drive quality upgrading (Li et al., 2024; Qiu et al., 2025). This change reshapes producers’ incentive structures, encouraging them to reduce dependence on chemical inputs and instead adopt greener production practices to capture quality premiums, thereby mitigating potential environmental externalities on the production side (Bold et al., 2022; Fusillo et al., 2025). Accordingly, this study uses the number of green food enterprises to reflect the breadth of the green market development, and green food output to capture the depth and intensity of green market development.

Second, green technological innovation is a key enabling condition for mitigating the environmental effects of rural e-commerce. In the production stage, precision agriculture technologies can substantially increase the efficiency of fertilizer and pesticide use. They enable higher yields without a corresponding rise in pollution and can even reduce pollution while improving productivity (Getahun et al., 2024; Cai et al., 2025). At the same time, new-energy electric agricultural machinery raises the level of mechanization while reducing energy consumption and greenhouse gas emissions per unit of output (Yang et al., 2025). This allows the expansion of mechanization to remain compatible with green development. In the circulation stage, green logistics technologies can significantly reduce energy consumption and emission intensity per unit of cargo. These technologies help relieve the environmental pressures created by the expansion of cold-chain and transportation networks (Jayarathna et al., 2023; Tetteh et al., 2024).

Finally, the environmental sustainability of rural e-commerce is not inherent in the technologies themselves but is shaped by their design, use, and regulatory context (Fichter, 2002). Strict emission standards and environmental supervision can effectively curb high-pollution production and logistics models (Borowiec et al., 2024). At the same time, subsidies and tax incentives can reduce the cost of adopting green agricultural technologies and clean logistics infrastructure (Liu et al., 2025). These incentives guide production factors to shift from high-carbon paths toward low-carbon activities and thus improve the environmental performance of rural e-commerce in a systemic way.

Accordingly, this study proposes the following research hypotheses.

H1

H1. Rural e-commerce development increases PM2.5 concentrations.

H2

H2. Agricultural input intensification and logistics expansion are transmission channels through which rural e-commerce increases PM2.5 concentrations.

H3

H3. Institutional constraints, green technological innovation, and market incentives jointly mitigate the pollution effect of rural e-commerce.

3 Research design

3.1 Model specification

Considering the gradual rollout of the REDC policy across counties, this study employs a multi-period Difference-in-Differences (DID) model to identify its net effect on PM₂.₅ concentrations. The core idea is to place counties that adopt the policy at different times into the same estimation framework. The model then generates two differences: between treated and control counties, and between pre- and post-treatment periods. This approach mitigates concerns related to omitted variable bias, reverse causality, and other endogeneity issues. Accordingly, the multi-period DID model is specified as follows:

$Y_{it}={\alpha }_0+{\alpha }_1{RECD}_{it}+X_{it}^{\prime }\beta +{\lambda }_i+{\gamma }_t+{\epsilon }_{it}(1)$

$Y_{it}$ denotes the PM2.5 concentration in county $i$ and year $t$, reflecting the level of fine particulate pollution. The core explanatory variable ${RECD}_{it}$ measures rural e-commerce development. It equals 1 if county $i$ is included in the national Rural E-commerce Comprehensive Demonstration program in yeari is included in the national Rural E-commerce Comprehensive Demonstration program in year $t$ and 0 otherwise. The coefficient ${\alpha }_1$ captures the net effect of rural e-commerce development on PM2.5 pollution. $X_{it}^{\prime }$ is a vector of control variables accounting for socioeconomic and climatic conditions. ${\lambda }_i$ and ${\gamma }_t$ denote county and year fixed effects, respectively, controlling for time-invariant county characteristics and common time shocks. ${\epsilon }_{it}$ is the error term.

To verify that the treatment and control groups followed comparable pre-policy trends, we expand Equation 1 into a dynamic DID framework in line with Beck et al. (2010). The dynamic DID model is specified as follows:

$Y_{it}={\alpha }_0+{\displaystyle \sum _{k\ne -1}}{\alpha }_{2}Even{t}_{i,t}^k+X_{it}^{\prime }\beta +{\lambda }_i+{\gamma }_t+{\epsilon }_{it}(2)$

In the dynamic DID specification presented in Equation 2, we introduce an event-time variable $Even{t}_{i,t}^k$ to capture the relative timing of the REDC policy. Specifically, $Even{t}_{i,t}^k=1$ if year $t$ is $k$ years away from the first year in which county $i$ was included in the REDC program, and 0 otherwise. The year immediately preceding the policy implementation ($k=-1$) is omitted from the regression and serves as the reference period. Accordingly, the set of event-time indicators allows us to track both pre-treatment leads and post-treatment lags of the policy effect over time. The coefficients ${\alpha }_2$ on the event-time indicators $Even{t}_{i,t}^k$ reveal the temporal evolution of the REDC policy effect. In particular, the coefficients associated with the pre-treatment periods ($k<0$) indicate whether treated and control counties exhibited similar trends prior to the implementation of the REDC program. If these pre-treatment coefficients are statistically insignificant and close to zero, the parallel trend assumption is supported. All other variables and notations remain consistent with those defined in Equation 1.

According to the above theoretical analysis, to investigate the mechanism through which the REDC policy influences the allocation of agricultural production factors and the expansion of logistics and warehousing infrastructure, we select five mediating variables: the proportion of agricultural labor, agricultural machinery input, chemical fertilizer input, road network length, and logistics warehousing land area. Following the step-wise mediation testing framework of Baron and Kenny (1986) (Baron and Kenny, 1986), we test whether rural e-commerce affects these mediators and, in turn, whether changes in agricultural input structure and logistics-warehousing demand serve as two potential transmission channels. The specific empirical models are constructed as follows:

${Med}_{it}={\alpha }_0+{\alpha }_3{RECD}_{it}+X_{it}^{\prime }\beta +{\lambda }_i+{\gamma }_t+{\epsilon }_{it}(3)$

$Y_{it}={\alpha }_0+{\alpha }_4{RECD}_{it}+{\alpha }_5{Med}_{it}+X_{it}^{\prime }\beta +{\lambda }_i+{\gamma }_t+{\epsilon }_{it}(4)$

In Equation 3, ${Med}_{it}$ denotes any of the five mediating variables. A significant ${\alpha }_3$ indicates that the REDC policy affects the mediator. In Equation 4, a significant ${\alpha }_5$ suggests that the mediator exerts an impact on the dependent variable after controlling for $REC{D}_{it}$. If ${\alpha }_4$ becomes insignificant after adding the mediator, a complete mediating effect is implied; if ${\alpha }_4$ remains significant, the sign consistency between ${\alpha }_3\times {\alpha }_5$ and ${\alpha }_4$ determines whether a partial mediating effect exists. All other variables and notations remain consistent with those defined in Equation 1.

To further explore whether the environmental impact of the REDC policy depends on county-level external contexts, we examine the heterogeneous effects of the policy along three dimensions: green market incentives, institutional constraints, and the green technology environment. Following prior research (Liu et al., 2023), we extend the baseline model by interacting the REDC indicator with the contextual variable and estimate the following specification:

$Y_{it}={\alpha }_0+{\alpha }_6{RECD}_{it}+{\alpha }_7{H}_{it}+{\alpha }_8{RECD}_{it}*H_{it}+X_{it}^{\prime }\beta +{\lambda }_i+{\gamma }_t+{\epsilon }_{it}(5)$

where $H_{it}$ represents county-level external contexts as specified in Equation 5. Specifically, green market incentives are captured by the number of certified green-product enterprises and the output of certified green products; institutional constraints are measured by the number of environmental penalty cases; and the green technology environment is proxied by the number of green patents. In this setting, ${\alpha }_8$ identifies whether the marginal effect of rural e-commerce development on PM2.5 concentrations varies across counties with different levels of market incentives, regulatory pressure, and technological support. A significantly negative (positive) ${\alpha }_8$ implies that a stronger contextual environment reinforces (weakens) the pollution-mitigation effect of the REDC policy. All other variables and notations remain consistent with those defined in Equation 1.

3.2 Variable definitions

3.2.1 Key independent variable

Considering that the REDC policy was implemented in different years across counties, the indicator variable is defined as follows: it takes the value of 1 in the year when a county is first designated as a demonstration county and remains 1 thereafter, and 0 otherwise. The data for this variable are obtained from the official list of “E-commerce Entering Rural Areas” demonstration counties released by the Ministry of Commerce of China.

3.2.2 Dependent variable

In this study, air pollution is measured by the annual average concentration of PM2.5 (μg/m³). Specifically, the PM2.5 data are sourced from the Tibetan Plateau Scientific Data Center and the Atmospheric Composition Analysis Group (ACAG) at the University of Washington, both of which provide nationally consistent gridded PM2.5 estimates that have been widely adopted in empirical research.

The gridded PM2.5 data are spatially matched to county administrative boundaries, and the weighted averages within each county are computed to obtain monthly values. These values are then aggregated to calculate annual average PM2.5 concentrations for each county and year. 3.2.3 Mediator Variables.

The five mediating variables are designed to reveal how rural e-commerce development may influence air quality. The proportion of agricultural labor, agricultural machinery input, and chemical fertilizer input reflect shifts in the agricultural production factor structure, which can alter pollutant emissions arising from agricultural activities. In parallel, road network length and logistics warehousing land area capture the expansion of logistics and warehousing infrastructure, potentially increasing freight transportation intensity and logistics-related emissions. By incorporating these mediators, we are able to assess whether the impact of rural e-commerce on PM2.5 concentrations operates through changes in agricultural input patterns and the growing demand for logistics and warehousing facilities.

3.2.3 Moderator variables

This study incorporates three moderator variables. First, institutional constraints are measured by the number of environmental administrative penalty cases at the county level, capturing the intensity of environmental regulatory enforcement. Second, green technological innovation is measured by the number of authorized green patents, reflecting the level of regional green technology accumulation. Third, market strength is represented by the number of certified green-product enterprises and the output of certified green products, indicating the degree of market incentives and enterprise vitality.

3.2.4 Control variables

To address other determinants of air quality, we include a comprehensive set of control variables at the county level. Following previous studies (Zhao et al., 2018; Feng et al., 2019; Guo et al., 2023; Guo et al., 2024), vegetation, climate, socioeconomic development, industrial structure, and fiscal conditions are incorporated to avoid confounding effects. Specifically, the vegetation normalization index (NDVI), annual temperature, annual average wind speed, sunshine duration, relative humidity, and sea level pressure and annual precipitation capture ecological and climatic characteristics that affect the dispersion, transformation, and deposition of air pollutants. Population density measures the intensity of human activities and potential emissions associated with residential energy consumption and transportation. The proportion of industrial output value reflects the level of industrialization, which is closely associated with production-related emissions. Economic development is controlled for using the logarithm of GDP to account for differences in economic scale and energy demand. Furthermore, electricity consumption per unit of GDP reflects the energy-use efficiency of local economic activities. The logarithms of general government fiscal income and fiscal expenditure indicate local fiscal capacity and the potential intensity of public intervention, including infrastructure provision and environmental governance. To further account for industrial environmental pressure, we include the logarithm of the aggregated industrial emissions of sulfur dioxide, nitrogen oxides, and industrial smoke and dust. Finally, the fixed asset investment-to-GDP ratio is included to capture investment-driven industrial and infrastructure expansion that may be associated with higher energy consumption and pollution emissions.

After implementing data interpolation and outlier removal procedures, the final dataset covers 1,698 counties between 2011 and 2021, forming a balanced panel with 18,678 county-year observations. Among them, 766 counties were designated as pilot areas, including 178 in the eastern region, 175 in the central region, 345 in the western region, and 68 in the northeastern region. The descriptive statistics and detailed definitions of all variables used in the analysis are reported in Table 1.

Table 1

Table 1. Summary statistics and variable definition.

Table 2 reports mean comparisons between treated and control counties before and after the implementation of the pilot policy. Although overall PM₂.₅ concentrations decline over time, the pollution gap between the two groups narrows markedly in the post-policy period, indicating a weakening of the pre-existing air quality advantage in treated areas. At the same time, production- and logistics-related variables show clear post-policy divergence: agricultural mechanization and road network length increase substantially in treated counties, while agricultural labor share, fertilizer use intensity, and warehousing land area exhibit no systematic post-policy differences. Overall, the relative rise in PM₂.₅ concentrations in treated counties coincides with intensified production and expanded logistics, providing descriptive support for the subsequent difference-in-differences analysis of the policy’s environmental effects.

Table 2

Table 2. Comparison between treatment and control groups.

3.3 Data sources

The empirical analysis combines multiple authoritative data sources to construct a county-level panel dataset. The core explanatory variable—rural e-commerce comprehensive demonstration policy—is obtained from official documents issued by the Ministry of Commerce of China. PM2.5 concentrations are sourced from the Tibetan Plateau Scientific Data Center and the Atmospheric Composition Analysis Group (ACAG) at the University of Washington. Economic and demographic control variables are drawn from the China County Statistical Yearbooks. Climate information is collected from the China Meteorological Data Service Center (surface climate dataset) and NASA’s Normalized Difference Vegetation Index (NDVI). Data on certified green agricultural products are obtained from the China Certification and Accreditation Database (CCAD) developed by Zhejiang University. Logistics and warehousing indicators are sourced from the China County Construction Statistical Yearbooks. Agricultural input variables—including agricultural labor, machinery, and fertilizer use—are taken from the China County Statistical Yearbooks. Environmental penalty records are collected from official announcements released by the Ministry of Ecology and Environment, and supplemented with relevant enforcement notices from the Ministry of Natural Resources. Green patent information is obtained from the China National Intellectual Property Administration. The data on Taobao Villages are obtained from the China Taobao Village Development Report, which is released annually by the Alibaba Research Institute. The list of national poverty-stricken counties is obtained from the official documents issued and approved by the Leading Group Office of Poverty Alleviation and Development of the State Council of China. The classification of major grain-producing regions, major grain-marketing regions, and grain production–consumption balanced regions is based on official documents issued by the Ministry of Agriculture and Rural Affairs of China. County-level terrain ruggedness and elevation data are obtained from the FABDEM V1-2 dataset. Provincial initial e-commerce development levels are derived from the China E-commerce Development Index Report (2014–2015) jointly released by the National Engineering Laboratory for E-commerce Technologies at Tsinghua University and several authoritative institutions. Key environmental governance regions are identified according to the Air Pollution Prevention and Control Action Plan issued by the State Council of China in 2013 and the Three-Year Action Plan for Winning the Blue Sky Defense Battle released in 2018. The lists of National Rural Industrial Integration Demonstration Parks (RIDP), Low-Carbon City Pilot (LCCP), and Rural Returnee Entrepreneurship Pilot (RREP) are obtained from officially approved documents issued by the National Development and Reform Commission of China. Information on the opening time of high-speed rail services is manually collected and verified using records from the official China Railway ticketing platform and relevant county-level news reports.

4 Empirical analyses

4.1 Baseline estimation

Table 3 reports the baseline regression results on the impact of the Rural E-commerce Demonstration County (RECD) policy on county-level PM₂.₅ concentrations. The estimation is based on a multi-period difference-in-differences model. We adopt a stepwise approach by sequentially adding different groups of control variables, while controlling for county fixed effects and year fixed effects to ensure the robustness of the core estimates. Specifically, Column (1) includes only the policy variable and fixed effects; Column (2) further adds economic control variables; Column (3) incorporates climatic control variables; and Column (4) includes both economic and climatic controls, representing the most comprehensive specification. All regressions cluster standard errors at the county level.

Table 3

Table 3. Benchmark regression results.

Across all specifications, the RECD coefficient remains significantly positive at the 1% level, indicating that the policy consistently increases PM₂.₅ concentrations. Taking the full model in Column (4) as an example, the estimated coefficient of RECD is 2.677. This means that, ceteris paribus, designation as a demonstration county increases the average PM₂.₅ concentration by about 2.677 μg/m³, in line with previous findings (Ji et al., 2023).

4.2 Robustness checks

4.2.1 Parallel trend test

The validity of the DID identification strategy relies on the crucial assumption that the treatment and control groups follow similar trends prior to the implementation of the policy. The event-study approach provides a direct test of this assumption by plotting the estimated coefficients for each period before and after the policy intervention.

Figure 1 presents the dynamic effects of the RECD policy by displaying the estimated coefficients and their 95% confidence intervals over time. Before the implementation of the policy, all coefficients fluctuate closely around zero and their confidence intervals include zero. This pattern indicates that there is no systematic difference between the treatment and control groups in the pre-policy period, confirming that the parallel trend assumption holds and ensuring the credibility of the subsequent causal inference. After the policy implementation, the estimated coefficients experience a pronounced upward shift and remain significantly positive for at least 5 years. The persistently elevated and statistically significant coefficients imply that the RECD policy has an immediate and lasting adverse effect on local air quality. Overall, the dynamic trajectory provides robust evidence that the RECD policy leads to a sustained increase in PM₂.₅ concentrations following policy adoption.

Figure 1

Figure 1. Parallel trend test.

4.2.2 Placebo test

To examine whether the baseline DID results may be driven by unobservable factors or random shocks rather than the true policy intervention, we conduct a placebo test by constructing fictitious treatment groups. Specifically, we randomly select the same number of counties as in the actual treatment group and assign them a “pseudo” RECD policy shock. We then repeat the estimation procedure 500 times to generate the distribution of estimated coefficients under the null scenario in which the RECD policy has no effect.

Figure 2 presents the placebo test results. The coefficients generated from the randomized pseudo-treatment assignments are centered closely around zero, and the peak of the kernel density estimate lies far below the true policy effect (${\alpha }_1$ = 2.937). Most placebo coefficients are statistically insignificant. These patterns indicate that the estimated policy effect in the baseline model is unlikely to be driven by random variation or model misspecification. Overall, the placebo test provides strong counterfactual evidence supporting the robustness of our causal inference, reinforcing the conclusion that the RECD policy indeed increases county-level PM₂.₅ concentrations.

Figure 2

Figure 2. Placebo test.

4.2.3 PSM-DID

To address potential selection bias arising from the non-random designation of REDC, we re-estimate the baseline results using the Propensity Score Matching Difference-in-Differences (PSM-DID) approach. The selection of demonstration counties is likely influenced by observable characteristics such as industrial foundation, geographic location and development potential, which may themselves be associated with air pollution. If these systematic differences are ignored, the conventional DID estimates may confound pre-existing advantages with the true policy effect. We construct a counterfactual control group using four matching techniques, namely, nearest neighbor matching, kernel matching, local linear regression matching and Mahalanobis distance matching, in order to approximate a randomized experimental setting and more rigorously identify the net environmental effect of the RECD policy.

Table 4 reports the PSM-DID estimation results. Although the four matching methods differ in sample selection mechanisms and weighting strategies, the direction and significance of the RECD coefficient remain highly consistent across all specifications. The coefficients are significantly positive at least at the 10% level, confirming the robustness of the conclusion that the RECD policy exerts a detrimental effect on county-level PM₂.₅ concentrations. In general, despite variations in matching algorithms, the estimated coefficients for RECD remain consistently positive and statistically significant across all PSM-DID models. This confirms that, even after mitigating sample self-selection concerns, the core finding of the baseline regressions is robust—that the RECD policy exerts significant environmental pressure by increasing county-level PM₂.₅ concentrations.

Table 4

Table 4. PSM-DID estimates.

4.2.4 Ruling out alternative policy explanations

To isolate the effect of rural e-commerce on air pollution, it is necessary to account for other concurrent policy shocks that may also influence local air quality, including environmental regulations, industrial restructuring, energy transition initiatives, transport infrastructure expansion, and rural development programs. Accordingly, this study incorporates a set of relevant policy shocks into the empirical design, with the results reported in Table 5, to mitigate potential confounding effects.

Table 5

Table 5. Ruling out alternative policy explanations.

Table 5 reports the regression results after accounting for competing policy interventions. In Column (1), we combine the key governance regions designated by the Air Pollution Prevention and Control Action Plan (2013–2017) and the Three-Year Action Plan for Winning the Blue Sky Defense Battle (2018–2020), and include them in the model as a policy dummy variable. These policies jointly capture the strong environmental regulations imposed on key regions in different periods, which substantially promoted industrial restructuring and energy transition in these areas (Yu et al., 2022; Huang et al., 2024). The regression results show that the coefficient of this policy group is significantly negative, indicating that these stringent environmental policies effectively reduced pollution and absorbed potential bias from policy shocks. After controlling for these effects, the positive impact of the RECD policy on PM₂.₅ remains robust, with consistent magnitude and significance, which strengthens the credibility of our core conclusion.

Column (2) further controls for the potential confounding effects of regional industrial integration and upgrading by treating counties approved to establish National Rural Industrial Integration Demonstration Parks (RIDP) during the sample period as the policy group and adding a corresponding dummy variable to the regression model. This approach helps exclude the potential impact of such policies on local air pollution, which may arise from promoting industrial integration and upgrading through expanding agricultural production or reducing fertilizer use intensity (Chen et al., 2025a; Li et al., 2025). The regression results show that the estimated coefficient of this dummy variable is not statistically significant, indicating that the policy effect of the demonstration parks is independent of the impact of the rural e-commerce demonstration policy on PM₂.₅ concentrations. This finding further suggests that the environmental effect identified in this study is not driven by other major rural industrial policies implemented during the same period, thereby enhancing the robustness of the conclusions and the validity of the identification strategy.

Column (3) incorporates the Low-Carbon City Pilot Policy (LCCP) as a key control variable. This policy is included because, as a comprehensive environmental regulation aimed at promoting energy structure optimization and low-carbon industrial transformation at the city level, its implementation may systematically affect local energy consumption and associated emissions (Yu and Zhang, 2021; Wang et al., 2023), thereby posing a potential competing explanation for the findings of this study. The regression results show that the estimated coefficient of the LCCP variable is significantly negative, indicating that the policy effectively reduces PM₂.₅ concentrations in pilot areas. After controlling for this policy, the positive impact of the RECD policy on PM₂.₅ concentrations remains robust and statistically significant. This suggests that even in regions subject to strong low-carbon transition policies, the pollution-increasing effect induced by rural e-commerce development through intensified production and logistics activities still exists independently and is not fully offset by broader emission-reduction policies.

Column (4) incorporates the county-level high-speed rail opening time as a key control variable this adjustment is made for two main reasons. First, high-speed rail may produce environmental effects by partially replacing road transport and improving transport efficiency, which could help reduce emissions and improve air quality (Zhu et al., 2022; Liu et al., 2024). Second, the expansion of high-speed rail may interact with the production and logistics channels examined in this paper by reshaping factor flows, industrial distribution, and local economic activities, potentially affecting logistics intensity and regional agglomeration (Chi et al., 2023; Chen et al., 2024). After controlling for these potential influences, the positive effect of the RECD policy on PM₂.₅ concentrations remains robust and significant. This indicates that the estimated environmental effect is not driven by high-speed rail development, and that the pollution-increasing mechanism of rural e-commerce through strengthened production and logistics activities is independent and credible.

Column (5) introduces the Rural Returnee Entrepreneurship Pilot policy (RREP) as an additional policy shock, considering the complex relationship between entrepreneurship and sustainable development. Entrepreneurship, especially green and knowledge-intensive entrepreneurship, may foster technological innovation, industrial upgrading, and more efficient resource use, bringing potential long-term environmental benefits (He et al., 2020). However, entrepreneurial activities can also generate new production, consumption, and logistics demands in the short term, which may increase local energy use and emissions (Omri and Afi, 2020). The results show that, after controlling for this policy, the positive effect of the RECD policy on PM₂.₅ remains robust and significant, while the RREP policy itself is not statistically significant. This suggests that our main conclusion remains valid after accounting for the potential impact of the RREP policy.

4.2.5 Additional robustness checks

To further validate the robustness of the baseline findings, we conduct a series of robustness checks, and the results are reported in Table 6.

Table 6

Table 6. Additional robustness checks.

Replacing the explanatory variable. Column (1) replaces the binary RECD indicator with a continuous measure of rural e-commerce penetration, measured by the number of Taobao Villages. The coefficient of RECD remains significantly positive, confirming that the main conclusion is robust. This robustness check is consistent with Zhang et al. (2024) (Zhang et al., 2024), who evaluate the effects of rural e-commerce by substituting the policy dummy with a continuous indicator of e-commerce development.

Replacing the explained variable. Column (2) uses the ACAG PM₂.₅ dataset from the University of Washington as an alternative outcome measure. The estimated coefficient remains positive and statistically significant, indicating that the findings are not driven by a particular dataset, consistent with the approach of Tang et al. (2025) (Tang et al., 2025).

Controlling for high-dimensional fixed effects. Columns (3) and (4) incorporate province–year and city–year interaction fixed effects, respectively, to absorb higher-level time-varying confounders (Jiang, 2025). Although the coefficient magnitude decreases slightly, it remains significant at the 1% level, suggesting that the estimated policy effect persists even under more demanding identification conditions.

Replacing the estimation method. Column (5) reports estimates based on the System GMM approach, which replaces the baseline fixed-effects specification to account for potential endogeneity and dynamic persistence in PM₂.₅ concentrations. This approach follows the existing literature on air pollution, which commonly employs dynamic panel GMM estimators to address potential endogeneity and sources of bias (Wu et al., 2023). The coefficient on RECD remains positive and statistically significant, and the AR (2) and Hansen tests support the validity of the GMM specification.

4.3 Mediation analysis

Table 7 presents the mechanism analysis and reveals how the RECD policy affects county-level PM₂.₅ concentrations via reallocation of agricultural production factors. The first channel is labor reallocation. Column (1) shows the RECD policy significantly reduces agricultural labor input, implying that rural e-commerce expansion creates non-agricultural jobs and shifts labor away from farming. Though Column (2) indicates that agricultural labor itself has no statistically significant direct effect on PM₂.₅, the shift in labor still sets the stage for input substitution.

Table 7

Table 7. Changes in production factor inputs.

The second channel is mechanization. Column (3) finds that the RECD policy significantly increases total agricultural machinery power; Column (4) shows mechanization has a significant positive effect on PM₂.₅ concentrations, suggesting that emissions climb not only via machinery use but also via the energy consumption of large-scale mechanized farming.

The third channel is chemical fertilizer use. Column (5) shows the RECD policy raises fertilizer application by 0.095 units—reflecting greater input to meet enlarged market demand via e-commerce. Column (6) confirms that higher fertilizer use significantly increases PM₂.₅ concentrations, consistent with the fact that ammonia volatilisation from fertilizers is a key precursor of fine particulate matter.

Taken together, the results show that the RECD policy affects air quality by reshaping agricultural input structure. The expansion of market demand shifts factor allocation away from labor and toward machinery and fertilizers. Mechanization increases fossil fuel consumption and direct emissions, while fertilizer use contributes to PM₂.₅ through ammonia volatilisation (Wen et al., 2024).

Table 8 examines the transmission pathway through which the RECD policy affects air quality via the expansion of the logistics system. The empirical results show that the e-commerce policy significantly increases county-level highway mileage and logistics warehousing capacity. The environmental effects of these two mediating variables are statistically significant: highway length exerts a significant positive impact on PM₂.₅ concentrations, reflecting the pollution contribution from transportation emissions; similarly, warehousing capacity also exhibits a promoting effect on PM₂.₅, capturing the environmental costs associated with the construction and operation of warehousing facilities.

Table 8

Table 8. Increase in logistics and warehousing demand.

On the one hand, e-commerce platforms lower market entry barriers for agricultural products, enabling produce that was previously confined to local markets to enter broader circulation networks, thereby substantially increasing the logistics demand for outward transportation of agricultural goods. On the other hand, the inflow of manufactured industrial products into rural areas expands simultaneously, generating bidirectional logistics pressure. This reconfiguration of supply chains manifests directly at two levels: the sharp rise in highway freight volume leads to substantial transportation-related emissions, while the large-scale construction of supporting warehousing facilities generates additional construction dust and operational energy consumption (Zhang et al., 2022).

4.4 Heterogeneous effects analysis

To capture regional dependence on agricultural production, we classify the sample into major grain-producing, major grain-consuming, and grain production–consumption balance regions following the literature (Ji et al., 2023; He et al., 2024). Columns (1)–(3) of Table 9 show clear heterogeneity across these regions. The RECD policy significantly increases PM₂.₅ concentrations only in major grain-producing regions, while no significant effects are found in the other regions. This result supports the production-side mechanism proposed in this study. In major grain-producing regions, rural e-commerce is more likely to induce production expansion and input intensification, leading to higher mechanization, energy use, and emissions. In contrast, rural e-commerce has limited effects on local production intensity and air pollution in grain-consuming and balanced regions.

Table 9

Table 9. Heterogeneous effects analysis.

To further explore heterogeneity from the perspective of economic development, the sample is divided into poverty and non-poverty counties according to the official list of national poverty-stricken counties. Columns (4) and (5) of Table 9 show that the REDC policy has a significant environmental effect in non-poverty counties, while its impact in poverty counties is negligible and statistically insignificant. The stronger effect in non-poverty counties can be explained by their mature industrial base and larger market scale. In these areas, the policy facilitates labor reallocation and encourages the substitution of machinery for labor, while also increasing logistics and warehousing demand. These changes raise freight intensity and energy consumption, resulting in higher pollution (Ma, 2010). In contrast, the weak effect in poverty counties suggests that economic development acts as an important boundary condition for the transmission of environmental consequences. Low initial economic density and an underdeveloped industrial base constrain changes in factor allocation and logistics demand. Continuous labor out-migration further limits the expansion of economic activities. As a result, the REDC policy does not generate noticeable environmental impacts through the same channels in poverty counties.

Topography is a key factor shaping population distribution and the spatial concentration of economic activities (Zhang et al., 2019). Based on regional elevation and terrain ruggedness, we divide the sample into plains-and-hills regions and plateau-and-mountain regions. Columns (1)–(2) of Table 10 show pronounced heterogeneity across these terrain types. The rural e-commerce comprehensive demonstration policy significantly increases county-level PM₂.₅ concentrations only in plains and hills, while no significant effect is observed in plateau and mountainous areas. This result is consistent with the strong dependence of production and logistics activities on terrain conditions. Plains and hills feature better transport accessibility and higher concentrations of agricultural and logistics activities. As a result, rural e-commerce development is more likely to induce production expansion and higher logistics frequency, leading to increased emissions. In contrast, natural constraints and weaker infrastructure in plateau and mountainous areas limit the expansion of production and logistics, resulting in a more muted environmental impact.

Table 10

Table 10. Heterogeneous effects analysis.

Initial e-commerce endowments shape both the starting point of rural e-commerce development and the intensity of economic responses to policy shocks. Drawing on the China E-commerce Development Index Report (2014–2015) released by the National Engineering Laboratory for E-commerce Technologies at Tsinghua University, we use the 2014 e-commerce development index to classify counties into groups with high and low initial e-commerce development. Columns (3)–(4) of Table 10 show significant heterogeneity in the effect of rural e-commerce on air pollution across these groups. The RECD policy significantly increases PM₂.₅ concentrations in both low- and high-penetration counties, with a larger estimated effect in counties with higher initial e-commerce penetration. This suggests that where e-commerce foundations are more developed, the policy more strongly amplifies existing production and logistics networks, leading to greater emission increases. In counties with lower initial penetration, although the policy promotes e-commerce development, limited baseline conditions constrain its environmental impact.

4.5 The moderating effects of institutional, technological, and market contexts

Table 11 examines the moderating roles of institutional constraints, technological foundations, and market conditions in shaping the environmental effects of the REDC policy. By incorporating interaction terms between the policy indicator and the three moderator variables, the estimation results show that all three dimensions significantly condition the policy’s environmental impact.

Table 11

Table 11. The moderating effects of institutional, technological, and market contexts.

Regarding the institutional environment, environmental regulation exerts a significant negative moderating effect, reflecting its disciplinary role. In regions with stronger regulatory enforcement, producers face higher compliance pressure, which limits their ability to respond to the market expansion brought by rural e-commerce through pollution-intensive input increases. As a result, the escalation of agricultural non-point source pollution is effectively contained, indicating that the regulatory system functions primarily as a constraint on environmentally harmful production expansion under the REDC policy.

In terms of the technological background, green technological innovation also shows a significant negative moderating effect. This implies that the accumulation and diffusion of green technologies enhance the capability of regions to decouple agricultural production from environmental damage. Through precision agriculture, eco-friendly inputs, and the recycling and resource utilization of agricultural waste, technologically advanced regions are able to maintain production efficiency while reducing the environmental cost per unit of output, thus easing the ecological pressure associated with e-commerce development.

With respect to market conditions, both certified green-product output and the number of certified green-product enterprises display significant negative moderating effects, highlighting the incentive role of market forces. When the breadth and depth of regional green-product development are higher, producers become more inclined to replace quantity-driven growth with quality- and sustainability-oriented production strategies. The price premium and reputational incentives associated with certified green products motivate producers to voluntarily reduce the use of pollution-intensive inputs, thereby dampening the environmental deterioration risk induced by the REDC policy.

5 Conclusion and policy implications

This study investigates the environmental impact and mechanisms of the RECD policy. The results show that the policy significantly increases PM₂.₅ concentrations through production-factor reallocation and logistics-warehousing expansion. The environmental effect is not spatially uniform: it is more pronounced in regions with stronger agricultural dependence, better terrain accessibility, and higher initial economic and e-commerce development, while relatively weak or insignificant in areas with poorer economic foundations or stricter geographical constraints. The heterogeneous effects further reinforce the proposed mechanisms, as stronger pollution responses are precisely observed in regions where production expansion and logistics intensification are more likely to occur.

In addition to identifying the environmental effects of rural e-commerce, this study offers two substantive contributions. It shows within a unified analytical framework that rural e-commerce can deteriorate air quality through both agricultural production adjustments and supply-chain expansion, revealing a dual pollution pathway. This reveals a dual pollution pathway. It also shows that the environmental impact of rural e-commerce is highly context-dependent. Market incentives can moderate pollution outcomes in a manner comparable to environmental regulation and green technological innovation, highlighting the importance of a co-governance approach that aligns regulatory enforcement, technological upgrading, and market-based incentives to facilitate the joint advancement of rural digital development and environmental sustainability.

Taken together, the results provide a practical foundation for developing policies that mitigate the environmental costs associated with rural e-commerce. First, strengthening environmental regulation and establishing environmental responsibility systems for e-commerce enterprises are essential. Policymakers should develop green operation standards for rural e-commerce platforms and logistics enterprises. These standards should cover packaging, warehousing and transportation, and aim to integrate full-chain environmental management into business operations. This can help reduce the environmental burden of e-commerce on the circulation side. Second, accelerating the diffusion of green technologies can facilitate the green transition of both agricultural production and logistics distribution. Environment-friendly practices such as soil testing–based fertilization and precision pesticide application should be prioritized in the agricultural sector, while the adoption of new energy vehicles and energy-efficient warehousing facilities can help lower environmental impacts throughout the supply chain. Third, improving market-based sustainability incentives is key to building a robust green agricultural product market. Simplifying certification procedures, introducing third-party certification agencies, and providing financial support for certification can reduce participation costs for producers; meanwhile, establishing a credible price-premium transmission mechanism can motivate producers to voluntarily reduce chemical input use and shift toward more ecological and intensive production models. Finally, differentiated environmental regulation is needed. Regions where the RECD policy generates stronger environmental impacts should implement stricter emission controls and promote green technologies in production and logistics. By contrast, regions with lower agricultural dependence and weaker pollution responses should integrate pollution-control planning early to prevent a “pollution-first, governance-later” trajectory.

This study is not without limitations. Due to the lack of suitable micro-level data, it is not possible to disentangle the precise contributions of individual pollution sources within agricultural production and logistics activities, nor to directly observe the behavioral responses of farmers, logistics enterprises, and e-commerce platforms to the RECD Policy. Future research could benefit from integrating micro-scale emission measurements or survey-based and firm-level datasets to more accurately characterize how environmental outcomes evolve through production and supply-chain decisions under rural e-commerce development. Nonetheless, this limitation does not challenge the principal conclusion of the paper that rural e-commerce entails non-negligible environmental costs and that its sustainability critically depends on institutional, technological, and market conditions.

Data availability statement

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary Material.

Author contributions

JH: Formal Analysis, Conceptualization, Software, Methodology, Writing – original draft. GY: Project administration, Validation, Visualization, Data curation, Funding acquisition, Supervision, Writing – review and editing, Investigation, Resources.

Funding

The author(s) declared that financial support was received for this work and/or its publication. This research was funded by Key projects of the National Social Science Fundation of China (No. 17AGL008).

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fenvs.2026.1754204/full#supplementary-material

References

Abukhader, S. M. (2008). Eco-efficiency in the era of electronic commerce – should ‘eco-effectiveness’ approach be adopted? J. Clean. Prod. 16, 801–808. doi:10.1016/j.jclepro.2007.04.001

CrossRef Full Text | Google Scholar

Baron, R. M., and Kenny, D. A. (1986). The moderator–mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. J. Pers. Soc. Psychol. 51, 1173–1182. doi:10.1037/0022-3514.51.6.1173

PubMed Abstract | CrossRef Full Text | Google Scholar

Beck, T., Levine, R., and Levkov, A. (2010). Big bad banks? The winners and losers from bank deregulation in the United States. J. Finance 65, 1637–1667. doi:10.1111/j.1540-6261.2010.01589.x

CrossRef Full Text | Google Scholar

Bold, T., Ghisolfi, S., Nsonzi, F., and Svensson, J. (2022). Market access and quality upgrading: evidence from four field experiments. Am. Econ. Rev. 112, 2518–2552. doi:10.1257/aer.20210122

CrossRef Full Text | Google Scholar

Borowiec, J., Papież, M., and Śmiech, S. (2024). The impact of environmental regulations on carbon emissions in countries with different levels of emissions. Environ. Sci. Pollut. Res. 31, 66759–66779. doi:10.1007/s11356-024-35702-8

PubMed Abstract | CrossRef Full Text | Google Scholar

Cai, B., Shi, F., Geremew, B. A., Addis, A. K., Abate, M. C., Dessie, W., et al. (2025). Precision agriculture techniques for optimizing chemical fertilizer use and environmental sustainability: a systematic review. Front. Agron. 7, 1665444. doi:10.3389/fagro.2025.1665444

CrossRef Full Text | Google Scholar

Cheba, K., Kiba-Janiak, M., Baraniecka, A., and Kołakowski, T. (2021). Impact of external factors on E-commerce market in cities and its implications on environment. Sustain. Cities Soc. 72, 103032. doi:10.1016/j.scs.2021.103032

CrossRef Full Text | Google Scholar

Chen, L.-F. (2019). Green certification, e-commerce, and low-carbon economy for international tourist hotels. Environ. Sci. Pollut. Res. 26, 17965–17973. doi:10.1007/s11356-018-2161-5

PubMed Abstract | CrossRef Full Text | Google Scholar

Chen, X., Gong, B., and Qin, Z. (2024). High-speed railroads and local agricultural development. doi:10.2139/ssrn.5054961

CrossRef Full Text | Google Scholar

Chen, H., Liu, M., Xie, S., and Chen, L. (2025a). A study on the impact of rural industrial integration on food production: empirical evidence from 2,571 counties in China. Front. Sustain. Food Syst. 9, 1679453. doi:10.3389/fsufs.2025.1679453

CrossRef Full Text | Google Scholar

Chen, Y., Puah, C.-H., and Arip, M. A. (2025b). Forecasting and spatial distribution analysis of cold chain logistics demand for fresh agricultural products in China. Environ. Dev. Sustain. doi:10.1007/s10668-025-06969-9

CrossRef Full Text | Google Scholar

Chi, F., Han, H., Chi, F., and Han, H. (2023). The impact of high-speed rail on economic development: a county-level analysis. Land 12, 874. doi:10.3390/land12040874

CrossRef Full Text | Google Scholar

Du, L., Wang, X., and Zhang, S. (2023). Can rural e-commerce contribute to carbon reduction? A quasi-natural experiment based on China’s e-commerce demonstration counties. Environ. Sci. Pollut. Res. 30, 104336–104349. doi:10.1007/s11356-023-29792-z

PubMed Abstract | CrossRef Full Text | Google Scholar

Edwards, J. B., McKinnon, A. C., and Cullinane, S. L. (2010). Comparative analysis of the carbon footprints of conventional and online retailing: a “last mile” perspective. Int. J. Phys. Distrib. Logist. Manag. 40, 103–123. doi:10.1108/09600031011018055

CrossRef Full Text | Google Scholar

Fan, J., Tang, L., Zhu, W., and Zou, B. (2018). The Alibaba effect: spatial consumption inequality and the welfare gains from e-commerce. J. Int. Econ. 114, 203–220. doi:10.1016/j.jinteco.2018.07.002

CrossRef Full Text | Google Scholar

FAO and ZJU (2021). Digital agriculture report: rural e-commerce development experience from China. 1st Edn. Rome, Italy: FAO. doi:10.4060/cb4960en

CrossRef Full Text | Google Scholar

Feng, H., Zou, B., Wang, J., and Gu, X. (2019). Dominant variables of global air pollution-climate interaction: geographic insight. Ecol. Indic. 99, 251–260. doi:10.1016/J.ECOLIND.2018.12.038

CrossRef Full Text | Google Scholar

FiBL, R. I., and Of, O. A. (2025). Latest organic data at the BIOFACH congress 2025. Available online at: https://www.organic-world.net/index/news-organic-world/article/3575.html (Accessed December 25, 2025).

Google Scholar

Fichter, K. (2002). E-commerce: sorting out the environmental consequences. J. Ind. Ecol. 6, 25–41. doi:10.1162/108819802763471762

CrossRef Full Text | Google Scholar

Fusillo, F., Orsatti, G., and Scandura, A. (2025). Public green demand and green innovation: evidence from US firms. Eurasian Bus. Rev. 15, 647–679. doi:10.1007/s40821-025-00304-y

CrossRef Full Text | Google Scholar

Gamage, A., Gangahagedara, R., Gamage, J., Jayasinghe, N., Kodikara, N., Suraweera, P., et al. (2023). Role of organic farming for achieving sustainability in agriculture. Farming Syst. 1, 100005. doi:10.1016/j.farsys.2023.100005

CrossRef Full Text | Google Scholar

Getahun, S., Kefale, H., and Gelaye, Y. (2024). Application of precision agriculture technologies for sustainable crop production and environmental sustainability: a systematic review. Sci. World J. 2024, 2126734. doi:10.1155/2024/2126734

PubMed Abstract | CrossRef Full Text | Google Scholar

Gibbs, D. (2017). “Ecological modernization,” in International encyclopedia of geography (John Wiley & Sons, Ltd), 1–13. doi:10.1002/9781118786352.wbieg0434

CrossRef Full Text | Google Scholar

Guo, S., Tao, X., Liang, L., Guo, S., Tao, X., and Liang, L. (2023). Exploring natural and anthropogenic drivers of PM2.5 concentrations based on random forest model: beijing–tianjin–hebei urban agglomeration, China. Atmosphere 14, 381. doi:10.3390/atmos14020381

CrossRef Full Text | Google Scholar

Guo, A., Wu, G., Miao, Z., and Chen, X. (2024). E-commerce and air pollution: evidence from China. J. Clean. Prod. 480, 144030. doi:10.1016/j.jclepro.2024.144030

CrossRef Full Text | Google Scholar

Hanlin, J., and Fan, Y. (2025). Research on the development strategy of rural E-commerce empowered by new productive force. J. Econ. Manag. Sci. 8, p108. doi:10.30560/jems.v8n1p108

CrossRef Full Text | Google Scholar

He, J., Nazari, M., Zhang, Y., and Cai, N. (2020). Opportunity-based entrepreneurship and environmental quality of sustainable development: a resource and institutional perspective. J. Clean. Prod. 256, 120390. doi:10.1016/j.jclepro.2020.120390

CrossRef Full Text | Google Scholar

He, X., Liu, W., He, X., and Liu, W. (2024). Coupling coordination between agricultural eco-efficiency and urbanization in China considering food security. Agriculture 14, 781. doi:10.3390/agriculture14050781

CrossRef Full Text | Google Scholar

Hidayatno, A., Destyanto, A. R., and Fadhil, M. (2019). Model conceptualization on e-commerce growth impact to emissions generated from urban logistics transportation: 5th international conference on power and energy systems engineering, CPESE 2018. Energy Procedia 156, 144–148. doi:10.1016/j.egypro.2018.11.119

CrossRef Full Text | Google Scholar

Hlaváček, P., Kopáček, M., Kopáčková, L., and Hruška, V. (2023). Barriers for and standpoints of key actors in the implementation of smart village projects as a tool for the development of rural areas. J. Rural. Stud. 103, 103098. doi:10.1016/j.jrurstud.2023.103098

CrossRef Full Text | Google Scholar

Huang, Z., Li, X., Chen, J., and Zhang, D. (2024). Does the three-year action plan for winning the blue sky defense battle implementation enabling the high-quality development of regional energy? A quasi-natural experiment from China. Energy 312, 133405. doi:10.1016/j.energy.2024.133405

CrossRef Full Text | Google Scholar

Igini, M. (2024). The environmental impact of online shopping. Earth. Org. Available online at: https://earth.org/online-shopping-and-its-environmental-impact/(Accessed November 23, 2025).

Google Scholar

Jaller, M., and Pahwa, A. (2020). Evaluating the environmental impacts of online shopping: a behavioral and transportation approach. Transp. Res. Part D. Transp. Environ. 80, 102223. doi:10.1016/j.trd.2020.102223

CrossRef Full Text | Google Scholar

Jayarathna, C. P., Agdas, D., and Dawes, L. (2023). Exploring sustainable logistics practices toward a circular economy: a value creation perspective. Bus. Strategy Environ. 32, 704–720. doi:10.1002/bse.3170

CrossRef Full Text | Google Scholar

Ji, X., Xu, J., and Zhang, H. (2023). Environmental effects of rural e-commerce: a case study of chemical fertilizer reduction in China. J. Environ. Manage. 326, 116713. doi:10.1016/j.jenvman.2022.116713

PubMed Abstract | CrossRef Full Text | Google Scholar

Jiang, X. (2025). From burning to clean: how china’s heating transition reduces pollution and enhances land-use sustainability. Front. Environ. Sci. 13, 1598729. doi:10.3389/fenvs.2025.1598729

CrossRef Full Text | Google Scholar

Jiang, M., Hu, X., Chunga, J., Lin, Z., and Fei, R. (2020). Does the popularization of agricultural mechanization improve energy-environment performance in China’s agricultural sector? J. Clean. Prod. 276, 124210. doi:10.1016/j.jclepro.2020.124210

CrossRef Full Text | Google Scholar

Kim, Y., Kang, J., and Chun, H. (2022). Is online shopping packaging waste a threat to the environment? Econom. Lett. 214, 110398. doi:10.1016/j.econlet.2022.110398

CrossRef Full Text | Google Scholar

Li, F., and Gan, Y. (2025). Research on the sustainable development capability of chinese rural E-commerce based on multidimensional perspective. Sci. Rep. 15, 11547. doi:10.1038/s41598-025-95653-z

PubMed Abstract | CrossRef Full Text | Google Scholar

Li, B., Xu, C., Zhu, Z., and Kong, F. (2021). Does e-commerce drive rural households engaged in non-timber forest product operations to adopt green production behaviors? J. Clean. Prod. 320, 128855. doi:10.1016/j.jclepro.2021.128855

CrossRef Full Text | Google Scholar

Li, J., Shi, J., Cao, R., Wu, J., and Liu, J. (2024). How e-commerce can boost china’s high-quality agricultural exports. Front. Sustain. Food Syst. 8, 1372129. doi:10.3389/fsufs.2024.1372129

CrossRef Full Text | Google Scholar

Li, X., Huang, W., Liu, J., Li, X., Huang, W., and Liu, J. (2025). The impact of China’s national modern agricultural industrial parks on fertilizer use from the perspective of food security. Sustainability 17, 11227. doi:10.3390/su172411227

CrossRef Full Text | Google Scholar

Liang, C., Liu, Z., and Geng, Z. (2021). Assessing e-commerce impacts on China’s CO2 emissions: testing the CKC hypothesis. Environ. Sci. Pollut. Res. 28, 56966–56983. doi:10.1007/s11356-021-14257-y

PubMed Abstract | CrossRef Full Text | Google Scholar

Lin, Y., and Li, C. (2023). Does rural e-commerce agglomeration help expand family farms’ size? Evidence from taobao villages in China’s yangtze river delta. Electron. Commer. Res. 23, 1731–1752. doi:10.1007/s10660-023-09735-z

CrossRef Full Text | Google Scholar

Liu, N., Liu, Y., and Yu, X. (2023). The impact of public environmental concern on environmental pollution: the moderating effect of government environmental regulation. PLOS One 18, e0290255. doi:10.1371/journal.pone.0290255

PubMed Abstract | CrossRef Full Text | Google Scholar

Liu, Z., Zhu, S., and He, C. (2024). Reducing industrial pollution and interregional environmental inequality via the world’s largest high-speed railway network. PNAS Nexus 3, pgae507. doi:10.1093/pnasnexus/pgae507

PubMed Abstract | CrossRef Full Text | Google Scholar

Liu, L., Tan, Y., and Deng, Y. (2025). The impact of green subsidies on farmers’ low-carbon production transition. Int. J. Clim. Change Strateg. Manage. 17, 595–613. doi:10.1108/IJCCSM-08-2024-0146

CrossRef Full Text | Google Scholar

Ma, C. (2010). Who bears the environmental burden in China--an analysis of the distribution of industrial pollution sources? Ecol. Econ. 69, 1869–1876. doi:10.1016/j.ecolecon.2010.05.005

CrossRef Full Text | Google Scholar

Malone, T., Yates, J., and Benjamin, R. (1987). Electronic markets and electronic hierarchies. Commun. ACM 30, 484–497. doi:10.1145/214762.214766

CrossRef Full Text | Google Scholar

Manogna, R. L. (2025). Determinants of small farmers access to agricultural markets: evidence from South India. Discov. Sustain. 6, 962. doi:10.1007/s43621-025-01412-5

CrossRef Full Text | Google Scholar

Omri, A., and Afi, H. (2020). How can entrepreneurship and educational capital lead to environmental sustainability? Struct. Change Econ. Dyn. 54, 1–10. doi:10.1016/j.strueco.2020.03.007

CrossRef Full Text | Google Scholar

Pålsson, H., Pettersson, F., and Winslott Hiselius, L. (2017). Energy consumption in e-commerce versus conventional trade channels - insights into packaging, the last mile, unsold products and product returns. J. Clean. Prod. 164, 765–778. doi:10.1016/j.jclepro.2017.06.242

CrossRef Full Text | Google Scholar

Pinos, J., Hahladakis, J. N., and Chen, H. (2022). Why is the generation of packaging waste from express deliveries a major problem? Sci. Total Environ. 830, 154759. doi:10.1016/j.scitotenv.2022.154759

PubMed Abstract | CrossRef Full Text | Google Scholar

Qiu, H., Tang, W., Huang, Y., Deng, H., Liao, W., and Ye, F. (2024). E-commerce operations and technology perceptions in promoting farmers’ adoption of green production technologies: evidence from rural China. J. Environ. Manage. 370, 122628. doi:10.1016/j.jenvman.2024.122628

PubMed Abstract | CrossRef Full Text | Google Scholar

Qiu, H., Tang, W., Deng, H., Hu, L., Wang, J., Sheng, B., et al. (2025). E-commerce adoption and farmers’ pesticide reduction behavior: implications for food system security. Front. Nutr. 12, 1665628. doi:10.3389/fnut.2025.1665628

PubMed Abstract | CrossRef Full Text | Google Scholar

Siragusa, C., Tumino, A., Mangiaracina, R., and Perego, A. (2022). Electric vehicles performing last-mile delivery in B2C e-commerce: an economic and environmental assessment. Int. J. Sustain. Transp. 16, 22–33. doi:10.1080/15568318.2020.1847367

CrossRef Full Text | Google Scholar

Syed Ali, S. A., Ilankoon, I. M. S. K., Zhang, L., and Tan, J. (2024). Packaging plastic waste from e-commerce sector: the indian scenario and a multi-faceted cleaner production solution towards waste minimisation. J. Clean. Prod. 447, 141444. doi:10.1016/j.jclepro.2024.141444

CrossRef Full Text | Google Scholar

Tang, J., Zhang, N., and Wang, L. (2025). Rural E-commerce policy and environmental pollution: evidence from China’s poverty-stricken counties. J. Environ. Manage. 393, 127061. doi:10.1016/j.jenvman.2025.127061

PubMed Abstract | CrossRef Full Text | Google Scholar

Teo, T. S. H., and Yu, Y. (2005). Online buying behavior: a transaction cost economics perspective. Omega 33, 451–465. doi:10.1016/j.omega.2004.06.002

CrossRef Full Text | Google Scholar

Tetteh, F. K., Owusu Kwateng, K., and Mensah, J. (2024). Green logistics practices: a bibliometric and systematic methodological review and future research opportunities. J. Clean. Prod. 476, 143735. doi:10.1016/j.jclepro.2024.143735

CrossRef Full Text | Google Scholar

Wang, H., Fang, L., Mao, H., and Chen, S. (2022). Can e-commerce alleviate agricultural non-point source pollution? — a quasi-natural experiment based on a China’s E-commerce demonstration city. Sci. Total Environ. 846, 157423. doi:10.1016/j.scitotenv.2022.157423

PubMed Abstract | CrossRef Full Text | Google Scholar

Wang, L., Shao, J., and Ma, Y. (2023). Does china’s low-carbon city pilot policy improve energy efficiency? Energy 283, 129048. doi:10.1016/j.energy.2023.129048

CrossRef Full Text | Google Scholar

Wei, Y. D., Lin, J., and Zhang, L. (2020). E-commerce, taobao villages and regional development in China. Geogr. Rev. 110, 380–405. doi:10.1111/gere.12367

CrossRef Full Text | Google Scholar

Wen, H., Huang, Y., and Shi, J. (2024). Revitalizing agricultural economy through rural E-commerce? Experience from China’s revolutionary old areas. Agriculture 14, 1990. doi:10.3390/agriculture14111990

CrossRef Full Text | Google Scholar

Wu, Q., Wang, R., Wu, Q., and Wang, R. (2023). Do environmental regulation and foreign direct investment drive regional air pollution in China? Sustainability 15, 1567. doi:10.3390/su15021567

CrossRef Full Text | Google Scholar

Xie, X., Luo, Z., and Chen, L. (2025). How can rural e-commerce promote green production on family farms in China? Evidence from hunan province. J. Rural. Stud. 120, 103843. doi:10.1016/j.jrurstud.2025.103843

CrossRef Full Text | Google Scholar

Yang, S., Chen, H., Li, Z., Ruan, Y., and Yang, Q. (2024). Temporal and spatial analysis of fertilizer application intensity and its environmental risks in China from 1978 to 2022. Environ. Sci. Eur. 36, 188. doi:10.1186/s12302-024-01011-7

CrossRef Full Text | Google Scholar

Yang, H., Ding, F., Gu, F., Wu, F., Yu, Z., Zhang, P., et al. (2025). Achieving the sustainable agricultural development goals by adopting the new energy electric agricultural machinery: an analysis of opportunities and challenges of China. Energies 18, 4211. doi:10.3390/en18164211

CrossRef Full Text | Google Scholar

Yu, Y., and Zhang, N. (2021). Low-carbon city pilot and carbon emission efficiency: quasi-experimental evidence from China. Energy Econ. 96, 105125. doi:10.1016/j.eneco.2021.105125

CrossRef Full Text | Google Scholar

Yu, Y., Dai, C., Wei, Y., Ren, H., and Zhou, J. (2022). Air pollution prevention and control action plan substantially reduced PM2.5 concentration in China. Energy Econ. 113, 106206. doi:10.1016/j.eneco.2022.106206

CrossRef Full Text | Google Scholar

Zhang, J., Zhu, W., Zhu, L., Cui, Y., He, S., and Ren, H. (2019). Topographical relief characteristics and its impact on population and economy: a case study of the mountainous area in western henan, China. J. Geogr. Sci. 29, 598–612. doi:10.1007/s11442-019-1617-y

CrossRef Full Text | Google Scholar

Zhang, Y., Fan, X., Ma, Y., and Sun, Y. (2022). How does cold chain logistics of fresh agricultural products affect carbon emissions? SSRN J. doi:10.2139/ssrn.4183326

CrossRef Full Text | Google Scholar

Zhang, N., Yang, W., and Ke, H. (2024). Does rural e-commerce drive up incomes for rural residents? Evidence from taobao villages in China. Econ. Anal. Policy 82, 976–998. doi:10.1016/j.eap.2024.04.023

CrossRef Full Text | Google Scholar

Zhao, H., Guo, S., Zhao, H., Zhao, H., Guo, S., and Zhao, H. (2018). Characterizing the influences of economic development, energy consumption, urbanization, industrialization, and vehicles amount on PM2.5 concentrations of China. Sustainability 10, 2574. doi:10.3390/su10072574

CrossRef Full Text | Google Scholar

Zhao, Y.-B., Wu, G.-Z., Gong, Y.-X., Yang, M.-Z., and Ni, H.-G. (2019). Environmental benefits of electronic commerce over the conventional retail trade? A case study in shenzhen, China. Sci. Total Environ. 679, 378–386. doi:10.1016/j.scitotenv.2019.05.081

PubMed Abstract | CrossRef Full Text | Google Scholar

Zhu, X., Dai, C., and Wei, Y. (2022). Does the opening of high-speed railway improve air quality? Evidence from China. Socioecon. Plann. Sci. 84, 101381. doi:10.1016/j.seps.2022.101381

CrossRef Full Text | Google Scholar