The Impacts of Rural E-Commerce on County Economic Development: Evidence from National Rural E-Commerce Comprehensive Demonstration Policy in China

Abstract

County economies are essential drivers of national economic development, acting as critical engines for growth and regional equilibrium. This study uses the National Rural E-commerce Comprehensive Demonstration policy, initiated by the Ministry of Finance and the Ministry of Commerce, China to empirically investigate the impact of rural e-commerce on county economic development and inequality, based on economic and night light data from 2000 to 2021. By applying a staggered difference-in-differences (DID) model, we find that rural e-commerce significantly boosts county economic development. This result remains robust after a series of robustness tests. The impact is stronger on the primary sector compared to the secondary and tertiary sectors, and the effect is more pronounced in the central and western regions than in the eastern regions. Furthermore, rural e-commerce effectively reduces economic inequality, contributing to inclusive development. Mechanistically, e-commerce into rural demonstration policy fosters county economic development by enhancing human capital mobility, accelerating logistics development, and promoting the growth of local enterprises.

1. Introduction

County economies play a crucial role in driving high-quality national economic development and serve as a key link in the integration of urban and rural areas in China [1,2]. As the fundamental spatial unit of economic development, counties encompass nearly 70% of the population and account for 50% of retail sales in China [3]. Therefore, county economies can be seen as both the foundational support for urban centers and the strategic hub for rural regions. Reese (1994) found that compared to cities, counties fulfill a coordinating role and are more likely to implement demand-side incentives in local economic development [1]. In terms of spatial structure, county economies are a multi-layered regional system with the county seat as its center, towns serving as connecting nodes, and rural areas as its core. The development of county economies could effectively promote the spatial connection between urban and rural economies, breaking the dualistic separation between the two [1]. It also facilitates the free flow of factors such as labor and capital across regions, contributing to the integrated development of urban and rural areas [4].

However, county economic development still faces numerous challenges. At the macro level, county economies are primarily driven by agricultural industries, and many regions grapple with common issues such as industrial homogeneity, the outflow of productive factors, and regional disparities [5,6]. These shared challenges exacerbate competition between counties, further intensifying inequality. At the micro level, key factors such as technology, labor, and capital are crucial for economic development, but the flow of these factors at the county level faces two major dilemmas. On the one hand, urban development has often been achieved at the expense of rural areas, resulting in a concentration of resources in cities. This one-way flow of factors from rural to urban areas leads to the depletion of resources and the idle use of resources in counties [7]. On the other hand, to protect the basic rights of relatively vulnerable farmers, the government has been reluctant to fully open rural areas to urban capital. Restrictions on the capitalization, monetization, and marketization of rural resources have limited investment in county economies, particularly in central and western regions, where development in the secondary and tertiary sectors remains sluggish [8]. Therefore, the focus of county economic development still lies in fully utilizing the region’s resource endowments, facilitating the flow of factors, and stimulating endogenous drivers of growth.

To retain resources within counties, it is essential to address the information barriers and market fragmentation caused by spatial and distance constraints. With the rapid development of the digital economy, modern models of commodity trading and circulation, exemplified by e-commerce, have emerged as a new driving force in breaking down market inefficiencies and transforming the existing structure of the global economy [9]. The development of rural e-commerce could effectively eliminate the spatial barriers to market transactions between urban and rural areas, significantly reducing the transaction costs [3]. In addition, e-commerce could substantially dismantle market barriers, expanding the sales reach of regional products to a national scale. The increased sales generated through e-commerce also attract the return of previously outflowed capital, thereby boosting local employment opportunities [10]. In Africa, e-commerce has changed rural economies in Kenya, most notably in the Rift Valley and western Kenya by allowing farmers to bypass the ‘middleman’ system and improve their profit margins [11].

Numerous studies have investigated the impact of e-commerce on economic development worldwide and have noted that e-commerce is a useful tool for promoting efficiency improvements, better asset utilization, and income growth. This is particularly relevant as rural economic development remains a critical policy challenge, especially in the face of persistent urban–rural disparities, inadequate infrastructure, and limited market access for rural producers [12,13,14]. However, some studies outline major challenges faced by less developed countries, including limited physical infrastructure, poor access to goods and services, constrained economic opportunities, and a lack of human capital.

Some studies have also investigated the impact of e-commerce on the rural economy and the development of modern agriculture in China [10]. A key area of focus is its impact on rural incomes. For example, Couture et al. (2021) estimated the effect of China’s first nationwide e-commerce expansion program on rural household income and consumption [15]. Peng et al. (2021) [16] argued that e-commerce relaxes the restriction that employees need to be physically present in stores, making it a highly accessible form of off-farm employment for individuals constrained by agricultural or caregiving responsibilities. As a result, e-commerce has a significant positive effect on rural incomes [16]. However, Liu and Zhou (2023) highlighted that urban residents have a higher ability to share rural e-commerce than rural residents, potentially exacerbating the urban–rural income gap [17]. Hong and Su (2024) further found that the construction of rural e-commerce platforms promotes industrial restructuring [18]. While most existing studies focus on rural households, relatively few explore the potential impact of rural e-commerce on county economic development. Moreover, Wei et al. (2025) and Lv et al. (2025) focused on income inequality and found that rural e-commerce development could significantly reduce internal income inequality explained by non-farm employment and entrepreneurial activities [19,20]. The closest study is by Qin et al. (2023), who found that rural e-commerce has a positive impact on county economy, with industrial restructuring and non-agricultural employment serving as the main channels [3]. However, they do not examine its potential effects on economic inequality.

Therefore, this study makes three contributions to the existing literature. First, we adopt a staggered difference-in-differences (DID) model to re-estimate the relationship between rural e-commerce and county economic development based on a case of implementation of the National Rural E-commerce Comprehensive Demonstration Project (NRECDP) since 2014. We extend the study by Qin et al. (2023) by incorporating an updated dataset of 2679 counties from 2000 to 2021 and enhancing the measurement of county economic development using night light data [3]. We conduct a series of robusness checks, including parallel trend test, placebo test, and propensity score matching difference-in-differences model (PSM-DID), to ensure the validity and reliability of our findings on the impact of rural e-commerce.

Second, we further explore the impact of rural e-commerce on economic inequality, providing valuable insights into the promotion of inclusive economic development in China. Addressing economic inequality has become a key priority in the global pursuit of inclusive and sustainable development. To the best of our knowledge, there is limited research focusing on the impact of rural e-commerce on economic inequality. With the exception of some studies focusing on income inequality [19,20,21], this area remains largely underexplored. To fill this gap, we measure economic inequality at the county level using various distributional statistics, such as inter-quantile ranges, variance, and Gini coefficients derived from recentered influence function (RIF) regressions. This clarifies the relationship between rural e-commerce and inclusive economic development at the county level.

Third, we investigate the mechanisms through which rural e-commerce affects county economic development, focusing on factor flow, logistics transportation, and enterprise growth. As Li and Qin (2022) noted, China is home to a large number of mountains which make the flows of factors and logistics transportation difficult [22]. Rural e-commerce has the potential to bridge these gaps by connecting farmers with external markets. Therefore, this study absorbs passenger volume, freight volume, and the number of industrial enterprises above designated size to offer a new analytical framework that explains the underlying dynamics linking rural e-commerce to economic development at the county level. The findings may also provide valuable policy implications for other developing countries facing similar geographic and infrastructural constraints, highlighting the potential of digital platforms to enhance rural connectivity and promote inclusive growth.

2. Theoretical Analysis and Research Hypothesis

2.1. Background of National Rural E-Commerce Comprehensive Demonstration Policy in China

In 2014, the Chinese government launched the e-commerce into rural demonstration policy to promote rural consumption and e-commerce-driven poverty alleviation, focusing on developing rural infrastructure and providing training for rural e-commerce professionals. Since then, a total of 1489 counties nationwide have benefited from this initiative. Unlike traditional distribution models, e-commerce involves various business activities—such as production, sales, and distribution—facilitated by internet-based technologies. And the key features include digital payment methods, supply chain management, and the separation of production and sales. E-commerce integrates with the real economy, offering services such as information aggregation, supply-demand matching, price discovery, transaction facilitation, and credit tracking. In essence, it expands market functions through digital networks like the internet [23,24].

Compared to traditional business models, e-commerce shifts some of the physical markets to the internet, enabling the rapid and accurate dissemination of related information to locations nationwide. Consumers could also track order and logistics information in real time via the internet, breaking the spatial limitations inherent in traditional business operations [25]. Within the context of county economic development, e-commerce typically operates in two primary models. The first is the e-commerce platform model, supported by both government and private enterprises, such as third-party operational platforms developed by companies like Alibaba. This model, backed by government support, reduces the e-commerce investment burden for county participants, helps attract customers, and boosts sales performance. The second model involves participants leveraging their own industrial structures and resource endowments to establish self-operated e-commerce platforms [26].

2.2. The Relationship Between Rural E-Commerce and County Economic Development

The role of e-commerce in driving county economic development is closely tied to the functional dynamics of the county economy. As a regional development system, the county economy integrates multiple geographic units—namely, the county seat, towns, and rural areas—and serves as a critical link in urban–rural integration [4,10,27]. However, many counties face significant challenges due to inefficient industrial structures and weak levels of factor agglomeration, which hinders economic development by limiting the effective utilization of resources and reducing the competitiveness of local industries [22]. Addressing these challenges needs improving resource allocation efficiency and facilitating the flow of production factors, which helps optimize inter-industrial structures and foster industrial agglomeration. And counties could unlock their economic potential and better integrate into broader development frameworks by tackling structural imbalances and enhancing the clustering of factors.

In terms of industrial structure, most county economies are primarily driven by the primary sector, with relatively underdeveloped secondary and tertiary sectors. Therefore, the key to boosting county economic development lies in advancing agricultural industries, which will in turn facilitate structural upgrades. This involves promoting modern agriculture to increase both the value and efficiency of the primary sector, expanding the scale of the secondary and tertiary sectors, and driving the integration of these sectors—through initiatives like deep processing of agricultural products and leisure agriculture [9]. The demand behind industrial upgrading hinges on the clustering of development factors around the county economy, with spillover effects into towns and rural areas.

In addition to promoting county economic development, rural e-commerce also contributes to reducing economic inequality. Previously, studies have extensively explored the impact of e-commerce on income disparities, particularly between urban and rural areas. For example, Lin et al. (2023) and Zhang et al. (2024) found that e-commerce effectively matches market supply and demand, helping to narrow urban–rural income gaps [28,29]. Similarly, Liu et al. (2021) and Li and Qin (2022) argued that rural e-commerce reduces capital outflows from rural areas, attracts population inflows, and generates endogenous development momentum [17,22]. Yin and Choi (2022) identified an inverted-U relationship between e-commerce development and the urban–rural income gap [30]. In the initial stages, urban areas benefit more from e-commerce due to superior infrastructure and resource availability, leading to a widening income disparity. However, as e-commerce further develops and penetrates rural regions, it delivers greater income benefits to rural residents, thereby narrowing the income gap. However, Liu and Zhou (2023) observed that the benefits of e-commerce are not evenly distributed [17]. They noted that rural residents often gain less than urban residents, and in some cases, an “elite capture” effect may emerge, where the advantages of e-commerce are concentrated among urban residents.

Under traditional models, county commercial systems have remained underdeveloped, largely dependent on offline markets that fail to meet the growing consumption needs of rural populations [31]. By contrast, e-commerce platforms, operating through online transactions, help overcome spatial limitations, enhance transaction efficiency between consumers and producers, and optimize the allocation of urban and rural resources [22]. Consequently, the emergence of rural e-commerce promotes direct and rapid communication and trade between rural farmers and consumers, facilitating the sale of agricultural products and increasing their value. E-commerce positively influences household consumption growth, with the effects being more pronounced in rural areas and among low-income households, reducing spatial consumption inequality [32]. Additionally, the development of rural e-commerce has also stimulated the growth of related industries, such as logistics and packaging, further enhancing the economic vitality of agricultural counties and narrowing the gap with other counties [3]. Therefore, the following hypothesis is proposed:

Hypothesis 1.

Rural e-commerce contributes to promoting county economic development and reducing economic inequality.

2.3. Theoretical Mechanisms of Rural E-Commerce Promotes County Economic Development

The role of e-commerce, supported by internet and digital technologies, in promoting county economic development could be realized through three primary mechanisms: enhancing factor flow between urban and rural areas, improving logistics transportation systems, and driving enterprise growth. Based on these mechanisms, we develop a general conceptual framework to systematically illustrate how rural e-commerce influences county economic development (Figure 1).

Firstly, in terms of enhancing the factor flow between urban and rural areas, the development of rural e-commerce breaks the traditional relationship between industrial development and production factors, thereby facilitating the circulation across regions. Serving as a platform and medium for business operations, e-commerce uses existing county resources, such as geographical indication products, to establish both local and national networks. This enables local products to access nationwide markets. The core of this mode lies in producing goods within the county and delivering them through logistics systems to broader markets. By shifting county economies from basic product processing to full industrial chain development, this mode significantly enhances the value-added processing of agricultural products and increases farmers’ incomes.

The e-commerce development allows county economies to capitalize on advantages such as low production costs for local labor and land, offsetting transaction costs associated with spatial and distance constraints. It promotes local industrialization and encourages the return of resources, such as entrepreneurs moving back to hometowns, creating a virtuous cycle of economic growth within the county. Furthermore, the big data and technological capabilities of e-commerce act as crucial drivers of economic development. By integrating these capabilities with other resources, e-commerce reshapes the foundational conditions, mitigating the limitations and delays caused by geographic distance and informational barriers [24,33]. Wei et al. (2020) analyzed rural e-commerce from the perspective of factor flow and observed that it initially forms clusters in villages, towns, or industrial parks [34]. These clusters attract resources, which gradually flow into rural areas, creating agglomeration effects that disrupt the traditional unidirectional migration of rural labor to cities. Additionally, with the support of e-commerce into rural demonstration policy, skilled e-commerce professionals from cities are increasingly returning to counties, particularly villages and towns. These professionals serve as critical assets for driving digital economic and stimulating county economic development.

Secondly, rural e-commerce improves logistics transportation. Under traditional mode, county economic development primarily targeted urban residents within the county, with products largely circulating locally. However, agricultural products produced by smallholder farmers often struggled to reach markets beyond the county due to limited sales channels and inadequate access to market information. The rise in rural e-commerce, supported by the construction of rural logistics and delivery infrastructure, has expanded market reach for county economies. It improves market accessibility and facilitates the circulation of goods, enabling the upward flow of agricultural products to broader markets and the downward flow of industrial goods to rural areas. This dynamic not only boosts consumption but also drives overall economic development at the county level [35].

E-commerce further enhances market connectivity by reducing transaction costs and increasing the efficiency of product exchanges. This platform enables seamless communication between upstream producers and downstream consumers, breaking the geographical constraints that traditionally hindered trade. Moreover, it helps businesses connect with remote consumers and suppliers, expanding the market reach beyond local boundaries [17]. From a cost and pricing perspective, Liu et al. (2022) [17] found that although e-commerce may increase transaction costs, it yields even greater gains by raising product sales prices, thereby increasing farmers’ revenues. Beyond higher prices, participation in rural e-commerce provides farmers with access to new collaborators, clients, and suppliers, further expanding sales volumes and improving overall market performance [17].

Finally, e-commerce fosters county economic development by supporting the expansion of local businesses. The construction of rural e-commerce systems is inherently systemic, with the establishment of e-commerce technology platforms driving the modernization of regional distribution networks and related infrastructure. This drives the rapid development of local industries such as express delivery, cold chain storage, and information technology services. The growth of these sectors not only increases the scale of business activity but also facilitates the transformation and upgrading of local enterprises. Additionally, rural e-commerce, supported by infrastructure improvements and policy incentives, creates a favorable environment for business development. This environment attracts high-quality enterprises to e-commerce demonstration counties, further stimulating local economic activity and contributing to regional economic development [36]. Therefore, we propose the following hypothesis:

Hypothesis 2.

Rural e-commerce promotes county economic development through three key mechanisms: enhancing factor flow between urban and rural areas, improving logistics transportation, and driving enterprise growth.

3. Empirical Model

3.1. Empirical Model Construction

To evaluate the impact of rural e-commerce on county economic development, this study constructs a quasi-natural experiment based on the e-commerce into rural demonstration policy launched by the Ministry of Finance and the Ministry of Commerce. A difference-in-differences (DID) model is applied to estimate effects of the policy on county economic development. Specifically, in July 2014, Chinese government issued the “National Rural E-commerce Comprehensive Demonstration Project (NRECDP)” to harness the internet as a tool for promoting rural economic development and accelerating the growth of rural e-commerce. This initiative outlined a comprehensive demonstration plan targeting rural counties, with a focus on strengthening public service systems for rural e-commerce, improving supply chain systems, and creating conducive development. Initially, 56 demonstration counties were selected as pilot regions, each receiving a dedicated fund of 20 million RMB to support the project. Over time, the scope of the initiative expanded. From 2015 to 2021, an additional 200 demonstration counties were added annually, accompanied by performance evaluations to ensure effective implementation.

Considering that the policy implementation occurred in multiple phases over time, this study employs a staggered difference-in-differences (staggered DID) method to quantitatively assess the impact of the e-commerce into rural demonstration policy. This approach treats the policy as an exogenous variable, similarly to a “natural experiment” or “quasi-natural experiment” in economic systems, enabling the identification of causal relationships by comparing differences in outcomes between treatment and control groups. And the baseline DID model is specified as follows:

$$y_{it}={\alpha }_0+{\alpha }_1{DID}_{it}+\sum {\beta }_j{Z}_{jit}+{\lambda }_i+{\mu }_t+{\epsilon }_{it}$$

(1)

where $y_{it }$ represents the economic development level of county $\mathit{i}$ in year $\mathit{t}$, measured by gross regional product (GRP). ${DID}_{it}$ is the binary variable for the implementation of the e-commerce into rural demonstration policy, equal to 1 if county $\mathit{i}$ implemented the policy in year $t$, and 0 otherwise. $Z_{jit}$ represents a set of control variables that influence economic development, including administrative area, population size, informatization level, savings rate, industrial structure, public fiscal expenditure, annual average temperature, and annual total precipitation, based on studies by Carleton et al. (2016) and Zhang et al. (2024) [29,37]. ${\alpha }_0$ is the intercept term, and ${\alpha }_1$ and ${\beta }_j$ are the coefficients to be estimated. If ${\alpha }_1$ is positive and statistically significant, Hypothesis 1 would be supported from the mean value perspective. ${\lambda }_i$ and ${\mu }_t$ are county and time fixed effects, respectively, controlling for unobserved factors specific to counties and time periods. ${\epsilon }_{it}$ is the error term. To address potential autocorrelation in the error term that could lead to biased estimates, robust standard errors clustered at the county level are used in all regression analyses.

Equation (1) controls for the impact of other potential factors on county economic development to compare the differences in economic development before and after the implementation of the policy between treatment and control groups. This approach requires strict assumptions regarding the randomness and exogeneity of the policy intervention. However, the selection of e-commerce demonstration counties by the government tends to prioritize national poverty counties. This selection bias, influenced by specific criteria or other unobserved factors, could potentially result in biased estimates.

To address this issue, four robustness checks are employed in this study. First, a parallel trend test is conducted on the difference-in-differences model to assess whether the parallel trend assumption holds. Namely, prior to the policy implementation, the economic development of pilot and non-pilot counties should have followed parallel trends or shown no significant difference. Specifically, based on Equation (1), an event study method is used to examine whether there is a parallel trend in county economic development before and after the policy implementation for both the treatment and control groups. Given the limited number of observations before the eighth period and after the fourth period, which reduces the degrees of freedom for model estimation, this study specifies the following model for the parallel trend test:

$$y_{it}={\alpha }_0+{\alpha }_1{DID}_{it}+{\sum }_{-8}^4{{\rho }_{S}P}_S+\sum {\beta }_j{Z}_{jit}+{\lambda }_i+{\mu }_t+{\epsilon }_{it}$$

(2)

where ${\rho }_S$ represents estimated coefficients, and $P_s$ is the dummy variable for e-commerce into rural demonstration policy. The remaining variables are defined and estimated in the same way as in Equation (1). If the policy effects are not statistically significant before the initiation of the pilot program, this would indicate that the parallel trend assumption holds.

Additionally, based on the parallel trend test, placebo tests are conducted to further evaluate the robustness of the policy effects. Using Stata 17 software, a random sample of counties is selected from the entire sample of counties to serve as the treatment group, with the starting time of the e-commerce into rural demonstration policy also chosen randomly. The model specified in Equation (1) is then re-estimated, and this procedure is repeated 1000 times to obtain the estimated coefficients. Kernel density estimation is subsequently applied to present distribution of the estimated coefficients and the corresponding p-values.

Moreover, given that previous studies have employed both GRP per capita and night light data at the county level as proxy variables for county economic development [38], this study further replaces the dependent variable with these proxies and re-estimates the model to evaluate the robustness of the results. As DID model helps mitigate time-invariant unobserved heterogeneity, time-varying unobserved factors may still bias the results. We also employed panel instrumental variable approach (fixed effect) and used the Number of Communication Facilities as instruments for e-commerce adoption to re-estimate Equation (1).

Finally, this study further employs a multiple-period PSM-DID model for robustness checks. To enhance the randomness between the treatment and control groups, propensity score matching (PSM) is used to address potential self-selection bias in the sample. Specifically, the control variables from Equation (1) are treated as covariates, with the dependent variable being the e-commerce into rural demonstration policy. A 1:1 nearest-neighbor matching is applied to construct a cross-sectional PSM match, ensuring that the dataset satisfies the common support condition. A multiple-period DID model is then used to re-estimate the effects of the e-commerce into rural demonstration policy on county economic development.

By estimating the impact of e-commerce into rural demonstration policy on county economic development, this study further explores its effect on economic inequality. Generally, Gini coefficient, Theil index, and Variation coefficient are adopted to measure inequality. However, OLS regression cannot directly estimate the impact of e-commerce into rural demonstration policy on aggregated index. Firpo et al. (2009) and Rios-Avila et al. (2020) developed Recentered Influence Functions (RIF) to analyze how small changes in the distribution of key explanatory variables influence the unconditional quantiles of the dependent variable $\mathit{y}$ [39,40]. The basic RIF estimation equation is

$$RIF\left(q_{\tau },y, F_y\right)=q_{\tau }+{\displaystyle \frac{\tau -1(y\le q_{\tau })}{f_y\left(q_{\tau }\right)}}$$

(3)

where $RIF\left(q_{\tau },y,F_y\right)$ is the recentered influence function of $F_y$, $q_{\tau }$ is the unconditional quantile of $y$, and $f_y$ is the density function.

Based on Recentered Influence Functions (RIF), there are three approaches that could be extended to estimate the effects of e-commerce into rural demonstration policy on economic inequality to test our Hypothesis 1 from the perspective of inequality. First, inter-quantile range estimates are constructed by comparing the 10th and 90th percentiles to determine whether significant differences in economic development exist under the same policy. And the equation can be written as

$$RIF\left(y, {iqr}_y(q_1-q_2)\right)=RIF\left(q_{y1},y, F_y\right)-RIF\left(q_{y2},y, F_y\right)$$

(4)

Second, a variance-based RIF regression equation (RIFVAR) is employed to evaluate inequality. The equation can be written as

$$RIF\left(y, {\sigma }_y^2\right)={(y-{\mu }_y)}^2$$

(5)

where ${\sigma }_y^2$ is the variance of dependent variable $y$, and ${\mu }_y$ is the mean of dependent variable $y$. Third, the Gini coefficient is calculated using the Lorenz curve to analyze regional disparities in economic development. The equation can be written as

$$RIF\left(y, {Gini}_y\right)=1+{\displaystyle \frac{2}{{\mu }_y^2}}{R}_y-{\displaystyle \frac{2}{R_y}}\left[y\left\{1-F_y\right\}\right]$$

(6)

where ${Gini}_y$ is calculated the Gini coefficient of dependent variable. Compared to other methods, the advantage of RIF regression in estimating inequality is that it allows for the study of the impact of explanatory variables on the distribution statistics of the dependent variable within the OLS framework, providing more economically meaningful estimates [35].

Once Hypothesis 1 is validated, we further construct an econometric model to examine the mechanism through which rural e-commerce influences county-level economic development, thereby providing a test of Hypothesis 2. The equation can be written as

$$M_{it}={\theta }_0+{\theta }_1{DID}_{it}+\sum {\phi }_j{Z}_{jit}+{\lambda }_i+{\mu }_t+{\epsilon }_{it}$$

(7)

where $M_{it}$ is the theoretical mechanisms of rural e-commerce promoting county economic development, including factor flow between urban and rural areas, logistics transportation, and enterprise growth; and ${\theta }_1$ is the core coefficients to be estimated providing a statistical evidence of validity of Hypothesis 2.

3.2. Data Sources

To effectively estimate Equation (1), it is necessary to construct both the treatment group and the control group for the difference-in-differences (DID) analysis. The treatment group consists of counties listed as e-commerce into rural comprehensive demonstration counties that implemented the policy during the specified period. And the control group includes counties that were not listed as demonstration counties or did not fall within the policy implementation period. The e-commerce into rural demonstration policy has been implemented progressively since 2014, starting with 56 counties in the first batch, and expanding to over 200 counties annually in subsequent years.

This study includes 1336 demonstration counties in the treatment group and 1343 other counties in the control group, with the sample period spanning from 2000 to 2021, based on data availability. The e-commerce into rural demonstration policy data is derived from the “National Rural E-commerce Comprehensive Demonstration Project” published by the Ministry of Commerce of China from 2014 to 2021. And the economic development of the counties is measured using the logarithm of county-level GRP and GRP per capita to examine the impact of e-commerce on county economic development. The relevant data are obtained from the China County Statistical Yearbook. In addition, we adopt the average value of night light data, which is widely recognized as a reliable proxy for local economic activity, to conduct a robustness check. Henderson et al. (2012) emphasized that night light data serve as a useful proxy for economic activity at temporal and geographic scales for which traditional data are of poor quality or are unavailable [41]. This dataset is obtained from National Earth System Science Data Center, National Science & Technology Infrastructure of China (http://www.geodata.cn, accessed on 30 January 2025).

The control variables in this study are measured using the following proxies: population size is represented by the number of registered residents in the county; informatization level is calculated by the number of fixed telephone lines; savings level is measured by urban and rural household savings; industrial structure is captured by the proportion of primary industry value-added in total GRP; and public fiscal expenditure is represented by the logarithm of general public budget expenditure. All these data are obtained from the China County Statistical Yearbook.

Additionally, the annual average temperature and annual total precipitation data are sourced from the National Climate Data Center (NCDC) of the U.S. National Oceanic and Atmospheric Administration (NOAA), which provides global climate data from 1901 to the present at county level. To measure factor flow at the county level, this study uses road passenger volume and rail passenger volume, while road and rail freight volume are used to measure logistics transportation at the county level. These data are retrieved from the China Urban Statistical Yearbook. Since passenger and freight data are reported at the city level, county-level estimates are derived using a weighted index based on the proportion of county economic activity within the city’s total economy. Furthermore, the number of industrial enterprises above designated size in the county is used as the measurement of enterprise growth, to explore the mechanism of e-commerce into rural demonstration policy on county economic development. Data for this variable are also sourced from the China County Statistical Yearbook.

4. Empirical Results Analysis

4.1. Baseline Estimation Results

Table 1. Estimation results of the impact of rural e-commerce on county economic development.

Variable	(1)	(2)	(3)	(4)	(5)
Demonstration County ∗ Time	0.073 *** (0.011)	0.092 *** (0.011)	0.066 *** (0.009)	0.063 *** (0.009)	0.057 *** (0.004)
Population Size	--	0.636 *** (0.055)	0.241 *** (0.047)	0.239 *** (0.047)	0.153 *** (0.013)
Administrative Area	--	0.170 * (0.044)	0.132 *** (0.045)	0.134 *** (0.045)	0.316 *** (0.016)
Industrial Structure			−1.913 *** (0.064)	−1.900 *** (0.063)	−1.884 *** (0.018)
Informatization Level	--	--	0.020 *** (0.005)	0.018 *** (0.005)	0.031 *** (0.002)
Savings Level	--	--	0.048 *** (0.011)	0.050 *** (0.011)	0.059 *** (0.004)
Public Fiscal Expenditure	--	--	0.195 *** (0.013)	0.192 *** (0.013)	0.210 *** (0.005)
Annual Average Temperature				−0.539 *** (0.052)	−0.523 *** (0.027)
Annual Total Precipitation				−0.008 (0.006)	−0.002 (0.004)
Intercept	13.331 *** (0.014)	9.695 *** (0.376)	8.733 *** (0.413)	10.378 *** (0.456)	9.578 *** (0.148)
Individual Fixed Effects	Yes	Yes	Yes	Yes	Yes
Time Fixed Effects	Yes	Yes	Yes	Yes	Yes
Adjusted R2	0.970	0.972	0.981	0.982	0.980

Notes: *** 1% level; * 10% level. Standard errors are in parentheses.

presents the estimation results of the impact of the e-commerce comprehensive demonstration project on county economic development. In Column (1), where only individual and time fixed effects are controlled, the policy shows a positive and statistically significant effect on county economic development, with an estimated coefficient of 0.073 at the 1% significance level. This indicates that the implementation of the e-commerce comprehensive demonstration project significantly promotes county economic development.

When various control variables and both individual and time fixed effects are taken into account (Columns 2 to 4 of Table 1), the policy dummy variable still shows a robust and significant positive effect on GRP, with an estimated coefficient of 0.063 after controlling for potential factors. Consistent with our baseline results, the FGLS estimation (Column 5 of Table 1) indicates a significant impact, with a coefficient of 0.057. This reinforces the conclusion that the e-commerce comprehensive demonstration policy has significantly contributed to county economic development, further driving broader regional growth. Therefore, Hypothesis 1 is partially supported. Our results are also consistent with the findings of Anvari and Norouzi (2016) and Qin et al. (2023) that e-commerce had a positive effect on county-level economic development [3,42].

Among the control variables, population size, administrative area, informatization level, savings level, and public fiscal expenditure all show significant positive effects on county economic development, while industrial structure has a significant negative effect.

4.2. Robustness Test

To verify the robustness of the baseline estimation results, alternative measures of the dependent variable are employed, and the empirical analysis is re-estimated. Following previous studies such as Xu et al. (2021) [38], we substitute the dependent variable with the logarithms of GRP per capita, the average value of night lights, and the maximum value of night lights to measure county economic development. The results, presented in

Table 2. Robustness Tests.

Variable	(1)	(2)	(3)
	GRP per Capita	Average Night Lights	Maximum Night Lights
Demonstration County ∗ Time	0.068 *** (0.010)	0.033 ** (0.015)	0.061 *** (0.011)
Control Variables	Yes	Yes	Yes
Individual Fixed Effects	Yes	Yes	Yes
Time Fixed Effects	Yes	Yes	Yes
Adjusted R2	0.969	0.966	0.834

Notes: *** 1% level; ** 5% level. Standard errors are in parentheses.

, show that e-commerce into rural demonstration policy still has a significant positive impact on county economic development, with the estimated coefficients remaining statistically significant at the 1% level. Furthermore, the coefficients for GRP per capita and total GRP are very close, suggesting that the development of rural e-commerce effectively drives economic development, whether analyzed from a total economic output perspective or on a per capita basis. Moreover, the implementation of the rural e-commerce policy is associated with an increase of 0.033 in average night lights and 0.061 in maximum night lights, providing additional evidence of the policy’s effectiveness.

Although the estimates support a positive effect of rural e-commerce on county economic development, the validity of this conclusion depends on the fulfillment of the parallel trend assumption. This means that in the absence of rural e-commerce policy shock, the economic development has the same trend in the treatment and control groups. In other words, prior to the policy implementation, the economic development of pilot and non-pilot counties should have followed parallel trends or shown no significant difference. Figure 2 presents the results of the parallel trend test. Relative to the base year, the estimated coefficients for the eight periods before the policy implementation in both the pilot and non-pilot counties are not statistically significant. This indicates no significant difference in economic development between the pilot and non-pilot counties prior to the policy intervention, thereby satisfying the parallel trend assumption.

After the implementation of the policy in the pilot counties, the effect of the policy shows a clear upward trend over time. The 95% confidence intervals for the estimated coefficients remain consistently above zero, indicating that, following the implementation of the policy, the GRP of pilot counties is significantly higher than that of non-pilot counties. Moreover, the magnitude of the policy’s impact strengthens progressively over time. This pattern likely reflects the delayed but cumulative effects of rural e-commerce development, such as the construction of logistics infrastructure, which requires time to achieve full operational capacity. Consequently, the policy’s influence accumulates over time, yielding a sustained and growing impact on county economic development.

To further validate the effect of the e-commerce into rural demonstration policy on county economic development, a placebo test was conducted. Placebo test entails assigning a hypothetical treatment period or group and re-estimating the DID model, and an insignificant estimated effect provides evidence that the original findings are not driven by spurious variation, thereby strengthening the causal interpretation. Specifically, the policy implementation was randomized across counties and time periods 1000 times, and the difference-in-differences estimates were recalculated for each iteration. The kernel density plot of the core estimated coefficients reveals that the distribution is centered around zero and follows a normal distribution (Figure 3). This result is significantly different from the coefficient of 0.063 in the baseline estimation result. Additionally, most of the placebo test estimates have significance levels exceeding 0.05, suggesting no systematic relationship in the randomized scenarios. These findings further emphasize the robustness of positive effect of the e-commerce into rural demonstration policy on county economic development, and support Hypothesis 1.

DID model helps mitigate time-invariant unobserved heterogeneity, but time-varying unobserved factors may still bias the results. To further address this issue, we employed panel instrumental variable approach (fixed effect). Specifically, according to research by Huang et al. (2019) [43], we used the Number of Communication Facilities as instruments for e-commerce adoption. The estimation results using this IV approach remain robust (coefficient = 0.014, standard error = 0.020, Cragg-Donald Wald F statistic = 1396.04) and consistent with our main findings (The Cragg-Donald Wald F statistic measures the strength of the instruments in an IV regression).

Given that the e-commerce into rural comprehensive demonstration policy is not a strictly natural experiment, the possibility of sample selection bias cannot be ruled out. For example, State Council Office of Poverty Alleviation joined the project in 2016, emphasizing the policy’s poverty alleviation objectives. This raises concerns about potential biases in the selection of demonstration counties. To address this issue, we further employ the propensity score matching difference-in-differences (PSM-DID) approach as a robustness check. The PSM method is based on the “conditional independence assumption”, which assumes no significant differences in observable characteristics between the treatment and control groups after matching. Before using PSM-DID to estimate the policy effect, we first test the common support test and balance test of control group and treated group. To ensure robustness, this study utilized both nearest neighbor matching and kernel matching to match the two groups. Then, the results of the PSM-DID estimation, presented in

Table 3. PSM-DID estimation results.

Variable	GRP		GRP per Capita		Average Night Lights
	(1)	(2)	(3)	(4)	(5)	(6)
Demonstration County ∗ Time	0.022 *** (0.005)	0.058 *** (0.004)	0.017 *** (0.006)	0.066 *** (0.005)	0.040 *** (0.010)	0.016 ** (0.007)
Control Variables	Yes	Yes	Yes	Yes	Yes	Yes
Individual Fixed Effects	Yes	Yes	Yes	Yes	Yes	Yes
Time Fixed Effects	Yes	Yes	Yes	Yes	Yes	Yes
Adjusted R2	0.988	0.983	0.967	0.970	0.982	0.969

Notes: *** 1% level; ** 5% level. The nearest neighbor matching and kernel matching were employed, respectively, with standard errors reported in parentheses.

, use both GRP, GRP per capita and average night lights as dependent variables. After applying PSM, the DID model results consistently support that rural e-commerce significantly promotes county economic development. This conclusion still holds whether economic development is measured by GRP, GRP per capita or average night lights, reinforcing the robustness of the Hypothesis 1.

4.3. Heterogeneity Analysis

We further explore the heterogeneous effects of rural e-commerce on county economic development across different industries, with the estimated results presented in

Table 4. Heterogeneity results by industries.

Variable	Value Added of the Primary Industry		Value Added of the Secondary Industry		Value Added of the Tertiary Industry
	Total	per Capita	Total	per Capita	Total	per Capita
Demonstration County ∗ Time	0.139 *** (0.009)	0.177 *** (0.012)	0.051 *** (0.013)	0.103 *** (0.014)	0.049 *** (0.009)	0.02 *** (0.010)
Control Variables	Yes	Yes	Yes	Yes	Yes	Yes
Individual Fixed Effects	Yes	Yes	Yes	Yes	Yes	Yes
Time Fixed Effects	Yes	Yes	Yes	Yes	Yes	Yes
Adjusted R2	0.975	0.934	0.969	0.950	0.980	0.963

Notes: *** 1% level. Standard errors are in parentheses.

. The results show that e-commerce into rural comprehensive demonstration policy has positive influences on both the total and average value added across the primary, secondary, and tertiary industries. Specifically, the estimated coefficients for the total value-added of the primary, secondary, and tertiary industries are 0.139, 0.051, and 0.049, respectively, with the largest impact observed in the primary industry. This finding suggests that the demonstration policy of promoting e-commerce in rural areas has a more significant effect on the development of the primary sector, aligning with the policy’s original intention. Specifically, the focus of the project is to utilize the “Internet +” model to strengthen the rural e-commerce infrastructure, establish a supportive ecosystem for e-commerce, and optimize supply chain systems to boost the agricultural market and rural economic development. Therefore, the primary sector benefits the most. This also supports the idea that the adoption of e-commerce can help narrow urban–rural disparities and achieve the goal of rural revitalization by promoting the development of the agricultural sector (Qin et al., 2023) [3].

Moreover, given the predominantly rural nature of county economies, agriculture remains the primary focus of industrial development in most counties. And the positive effect of e-commerce on the agricultural sector is undoubtedly crucial to county economic development [44]. Notably, while the policy directly promotes agricultural development, its positive impacts on the value-added of the secondary and tertiary sectors indicate that its influence extends beyond the primary sector. This spillover effect fosters balanced development across industries, contributing to the coordinated growth of county economies. These findings also support that e-commerce would promote the modernization of agriculture by enhancing both the value and efficiency of the primary sector, while simultaneously facilitating the integration of all three sectors—through initiatives such as value-added processing of agricultural products and the development of leisure agriculture.

Regional differences in the impact of rural e-commerce on county economic development are shown in

Table 5. Heterogeneity results by regions.

Variable	Western Regions		Central Regions		Eastern Regions
	GRP	GRP per Capita	GRP	GRP per Capita	GRP	GRP per Capita
Demonstration County ∗ Time	0.036 *** (0.013)	0.044 *** (0.013)	0.037 *** (0.014)	0.044 *** (0.017)	−0.003 (0.017)	0.018 (0.023)
Control Variables	Yes	Yes	Yes	Yes	Yes	Yes
Individual Fixed Effects	Yes	Yes	Yes	Yes	Yes	Yes
Time Fixed Effects	Yes	Yes	Yes	Yes	Yes	Yes
Adjusted R2	0.980	0.972	0.972	0.962	0.983	0.980

Notes: *** 1% level. Standard errors are in parentheses.

. Specifically, the e-commerce into rural demonstration policy has a statistically significant positive effect in the western and central regions, whereas its impact in the eastern region is not significant. In the western region, the coefficients for GRP and GRP per capita are 0.036 and 0.044, respectively, with similar magnitudes observed in the central region. These regional differences can be attributed to variations in resource endowments and the economic environments that shape e-commerce development. The eastern region, characterized by higher levels of economic development, well-established industrial infrastructure, and advanced e-commerce penetration, has already reaped substantial benefits from rural e-commerce. Consequently, the marginal effect of rural e-commerce policies on economic development in this region is relatively limited, and the policy’s impact is less pronounced. In contrast, the western and central regions, which lag behind in e-commerce infrastructure, present greater developmental potential. Therefore, these regions have experienced more significant economic benefits, highlighting the transformative potential of e-commerce policies in less developed areas.

4.4. The Impacts on Economic Inequality at County Level

The preceding analysis indicates that the e-commerce into rural comprehensive demonstration policy significantly promotes county economic development. However, does this development lead to county inequalities?

Table 6. Estimation results of economic inequality.

Variable	GRP			GRP per Capita			Standard Deviation of Night Lights
	Inter-Quantile Ranges	VAR	Gini Coefficients	Inter-Quantile Ranges	VAR	Gini Coefficients
Demonstration County ∗ Time	−0.114 *** (0.010)	−0.925 *** (0.092)	−0.175 *** (0.001)	−0.149 *** (0.013)	−0.698 *** (0.092)	−0.022 *** (0.002)	−0.671 *** (0.088)
Control Variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Individual Fixed Effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time Fixed Effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Adjusted R2	0.404	0.573	0.591	0.326	0.442	0.467	0.870

Notes: *** 1% level. Standard errors are in parentheses. Economic equality is calculated based on GRP (Gross Regional Product) and GRP per capita, respectively.

shows the estimated results of the impact of the policy on economic inequality at county level. We find that regardless of whether GRP or GRP per capita is used as the dependent variable to calculate economic inequality index, the policy shows a significant negative effect across three unconditional quantile models. This indicates that the rural e-commerce policy effectively reduces economic inequality at the county level, thereby fostering more inclusive economic development, which is consistent with our theoretical hypothesis. Moreover, it suggests that the development of rural e-commerce can help narrow regional development gaps and promote common prosperity.

Additionally, the estimates based on the standard deviation of night lights reveal that the policy not only mitigates disparities in economies across counties but also reduces internal economic inequality within individual counties. These findings are consistent with the conclusions of Tang and Zhu (2020) [21], who argue that e-commerce development could narrow the urban–rural income gap. It further highlights the positive role of e-commerce in promoting rural economic development [21].

5. Mechanism Analysis

On the basis of theoretical analysis, we further introduce three variables, namely factor flow, logistics transportation, and enterprise growth, to explore the mechanisms through which rural e-commerce influences county economic development. The results are presented in

Table 7. The mechanism of the e-commerce into rural comprehensive demonstration policy on county economic development.

Variable	Factor Flow		Logistics Transportation		Enterprise Growth
	Road Passenger Volume	Rail Passenger Volume	Road Freight Volume	Rail Freight Volume	Number of Industrial Enterprises Above Designated Size
Demonstration County ∗ Time	0.142 *** (0.027)	0.132 *** (0.044)	0.123 *** (0.027)	−0.022 (0.084)	0.050 *** (0.016)
Control Variables	Yes	Yes	Yes	Yes	Yes
Individual Fixed Effects	Yes	Yes	Yes	Yes	Yes
Time Fixed Effects	Yes	Yes	Yes	Yes	Yes
Adjusted R2	0.835	0.853	0.840	0.816	0.911

Notes: *** 1% level. Standard errors are in parentheses.

.

In terms of factor flow, the e-commerce into rural demonstration policy significantly promotes both road and rail passenger volume, with estimated coefficients of 0.142 and 0.132, respectively. This suggests that rural e-commerce has effectively reduced spatial barriers to human capital mobility, facilitating the movement of labor between urban and rural areas, and thereby contributing to county economic development.

For accelerating the flow of factors between urban and rural areas, rural e-commerce has broken the traditional connection between industrial growth and the factors of production, fostering the circulation of factors between urban and rural regions. As a platform and medium for business, e-commerce leverages existing local resources and endowments, such as geographic indication products, to establish both local and national operational platforms within the county. This expands the market reach of local products to a national scale. The core of this mode is that products need only be produced within the county and then distributed nationwide through the logistics network. Therefore, the county economy could upgrade from a primary product processing chain to a full industrial chain development model, playing a significant role in adding value to agricultural products and increasing farmers’ incomes. Within this industrial framework, county economies can utilize advantages such as low costs of local human resources, land, and other production factors to offset transaction costs caused by spatial and distance barriers. This, in turn, promotes the localization of industries and attracts the return of factors (such as entrepreneurial returnees), creating a virtuous cycle for county economic development.

At the same time, the powerful data, information, and technology embedded in e-commerce can also serve as crucial factors of production for economic development. By integrating with other resources, it reshapes the foundational conditions for county economic development, alleviating the limitations imposed by distance and information delays (Zhu et al., 2021; Wang et al., 2023) [24,33]. From the perspective of factor flow, Wei et al. (2020) and Zhang et al. (2022) found that the rural e-commerce industry, after initially establishing clusters in villages, towns, or industrial parks, attracts additional resources, such as talent, which gradually flow into rural areas, creating a convergence effect [10,27]. This challenges the traditional outflow pattern of rural human capital to urban areas. Meanwhile, rural e-commerce development policy has encouraged the return of experienced professionals from urban centers to counties. These returnees serve as key drivers of the digital economy in rural regions, thereby promoting sustained county economic growth. Therefore, the theoretical pathway linking rural e-commerce to economic development via factor flow is empirically supported.

With regard to logistics transportation, the rural e-commerce demonstration policy has a significant positive effect on road freight volume at the county level, with an estimated coefficient of 0.123. However, the impact of the e-commerce into rural project on county-level rail freight volume is not statistically significant. This suggests that rural e-commerce predominantly enhances county-level logistics, particularly in rural areas, by facilitating the efficient movement of industrial and agricultural products between urban and rural regions, thus driving economic development at the county level. This is consistent with both theoretical expectations and practical observations, where rural e-commerce development primarily supports logistics through road transport rather than rail.

Furthermore, we find that rural e-commerce development can stimulate county economic development by fostering the growth of enterprises. The estimated results show that, in the pilot counties of the “E-Commerce into Rural” initiative, the number of industrial enterprises above designated size increased by 5% compared to non-pilot counties, providing significant support and incentives for the core drivers of county economic development. The findings suggest that rural e-commerce plays a key role in nurturing the growth of enterprises at the county level, thereby contributing to overall economic development. Based on the above analysis, Hypothesis 2 is supported.

6. Conclusions and Implications

6.1. Conclusions

County economies are crucial drivers of high-quality economic development and serve as key links in the urban–rural integration in China. However, the development of county economies, especially in the central and western regions, still faces significant challenges. Given the rapid development of the digital economy, the question of whether modern model of commodity circulation, represented by e-commerce, can become a new driver of county economic development is of great importance. Therefore, this study empirically explores the impact of the e-commerce into rural comprehensive demonstration project on county economic development by constructing a quasi-natural experiment using county economic data and night light data from 2000 to 2021.

The estimated results show that the implementation of the e-commerce into rural comprehensive demonstration project has a significant positive impact on county economic development. And this conclusion remains robust even after conducting various tests, including parallel trend test, placebo test, and PSM-DID estimations. In addition, the results of heterogeneity analysis by industries reveal that the rural e-commerce positively impacts economic development across all three sectors—primary, secondary, and tertiary industries—with the greatest effect observed in the primary industry. Given that agriculture is the dominant sector in county economies, it benefits the most from the e-commerce into rural demonstration project. And from the perspective of regional heterogeneity analysis, we find that the project has a significant positive impact on county economic development in the western and central regions, while its effect in the eastern region is not significant, possible due to the already established infrastructure and more advanced e-commerce development in those areas.

Furthermore, for mechanisms, we find that the e-commerce into rural comprehensive demonstration project can achieve county economic development through three mechanisms: enhancing factor flow between urban and rural areas, improving logistics transportation, and driving enterprise growth. Lastly, the results show that the policy contributes to reducing county economic inequality, achieving inclusive economic development at the county level.

6.2. Implications

Based on the findings of this study, the following policy implications are proposed. Firstly, given the positive impact of rural e-commerce on economic development, it is essential to broaden the scope of the e-commerce into rural comprehensive demonstration project. This expansion can be achieved by leveraging e-commerce to accelerate the development of sales, logistics, and information systems, particularly in rural regions. The primary focus should be on overcoming barriers to the “outflow of agricultural products” and further accelerating the construction of rural logistics systems. This involves enhancing rural logistics service networks, improving infrastructure such as highways and internet connectivity, and promoting seamless economic integration between urban and rural areas, as well as across counties. E-commerce can leverage local resources and geographic indicator products to drive poverty alleviation by attracting returning agricultural professionals, fostering entrepreneurship, and creating jobs. This revitalizes underdeveloped regions and fuels inclusive county economic development.

Secondly, it is essential to further activate the role of e-commerce in integrating agriculture with other industries, advancing the modernization of the agricultural sector. By leveraging the advantages of e-commerce in information technology and tailoring efforts to local conditions, the focus could be on gathering market information, promoting the deep processing of agricultural products, and increasing their industrial added value. E-commerce tools such as live streaming and new media technologies can help create a brand network for agricultural products, expanding the reach to broader markets and enhancing brand development. Special attention could be given to nurturing e-commerce-driven agricultural enterprises, with support measures such as investment in professionals, tax incentives, and the creation of a favorable business environment for e-commerce. These efforts will foster the growth of county-level enterprises, improving county economies.