Research on the Efficiency Measurement and Spatial Spillover Effect of China’s Regional E-Commerce Poverty Alleviation from the Perspective of Sustainable Development

Abstract

The development of e-commerce plays a very important role in changing the production and operation mode, optimizing the allocation of market resources, promoting sustainable development, and ultimately achieving the goal of e-commerce poverty alleviation. Therefore, the efficiency of e-commerce poverty alleviation has become a focus of attention for both the government and academia. The authors of this paper selected the panel data of 30 provinces and cities in China from 2010 to 2021, in order to measure the poverty alleviation efficiency of e-commerce in each province and city. We used the Moran’s I index to measure its spatial correlation to verify the existence of its spatial effect; we then used the spatial Durbin model to analyze the spatial spillover effect in the efficiency of e-commerce poverty alleviation. The conclusions are as follows: First, there is a significant positive spatial correlation of the efficiency of e-commerce poverty alleviation among different regions in China. Moran’s I index exceeds 0.5, indicating that there is a significant spatial effect in the efficiency of e-commerce poverty alleviation, and the existence of its spatial effect is unavoidable in the empirical analysis. Secondly, from the perspective of the efficiency of e-commerce poverty alleviation in various regions of the country, the overall e-commerce poverty alleviation efficiency is not high, and there are large differences among regions. The regions in which efficiency is higher include Tianjin, Beijing, and Shanghai. Regionally, the highest are in the east and the lowest are in the west. Secondly, from the decomposition of spatial spillover effects, the direct effects of each influencing factor are all positive. Only the financial development environment is less significant, and the indirect effects indicate that only four indicators have significant spatial spillover effects, of which the most significant is industrial agglomeration. The level of industrial agglomeration is not significantly related to the level of human capital, and there is a negative correlation between it and the efficiency of e-commerce poverty alleviation. The authors studied the poverty alleviation efficiency and spatial spillover effect of China’s regional e-commerce from the perspective of sustainable development, which is beneficial to China’s regional poverty alleviation results, providing practical guidance and decision-making reference for implementing differentiated coping strategies in different regions. The research complements, improves, and expands the research content in this field.

1. Introduction

Since sustainable development involves many aspects, such as nature, the environment, society, economy, science and technology, politics, etc., the definition of sustainable development changes due to the different perspectives of researchers. Economists believe that sustainable development means the maximization of the net benefits of economic development while maintaining the quality of natural resources and the services they provide. This paper explores the research on the poverty alleviation efficiency of e-commerce in the context of sustainable development. E-commerce, through intensive management, takes social and economic benefits as the fundamental purpose, achieves the greatest return on investment with the least cost, and effectively reduces business costs so that the product producer becomes the biggest beneficiary; the sustainable development of the regional economy is effectively realized. With the rapid rise of e-commerce, the role of e-commerce in promoting regional economic development, driving entrepreneurship and employment, and achieving poverty alleviation in poverty-stricken areas has become more and more prominent. The results of one study [1] were officially incorporated into the poverty alleviation policy system by the State Council Poverty Alleviation Office in 2014.

The economics community has always been concerned about the spatial distribution of poverty. At the end of the 20th century, the World Bank first drew the spatial distribution map of the world’s poor population, and the poverty space began to widely concern and be recognized by experts. Minot and Baulch (2005) [2] believed that poverty has obvious spatial distribution characteristics. Regional poverty is closely related to uneven spatial distribution and geographic location. Xing (2013) [3] combined spatial economics and geography to form a new perspective and a new method of spatial poverty research and applied it to poor rural areas to effectively implement poverty reduction policies by combining spatial capital. The overall poverty problem in China is spatially uneven, especially in the central and western regions. The contiguous problem of destitute areas still exists. Poverty in these areas is greatly affected by the natural geographical environment [4]. Therefore, the solution to rural poverty should start from a spatial perspective. With the development of spatial economics and geographic technology, in fact, throughout the existing research at home and abroad, the research results on e-commerce poverty alleviation have been relatively fruitful, but they are mostly concentrated in the qualitative stage. The research content mainly focuses on the current situation, problems, and countermeasures of e-commerce poverty alleviation, and quantitative research is relatively rare [5,6]. In addition, as a new means of targeted poverty alleviation, e-commerce poverty alleviation is different from traditional poverty alleviation methods. It needs to take an intensive poverty alleviation and development path and focus on the efficiency of poverty alleviation. At the same time, e-commerce poverty alleviation itself has a strong regionality. The characteristics also need to be properly considered from the geospatial perspective, but these issues have not yet attracted enough attention from the academic community [7].

Although scholars have conducted in-depth research on e-commerce poverty alleviation, there are still several propositions worthy of in-depth exploration, such as the efficiency of e-commerce poverty alleviation, whether there are differences among regions, and e-commerce development, in addition to e-commerce as directly driving the development of the e-commerce industry. In addition to the direct effect of improving rural poverty, is there any indirect effect of e-commerce development on the development of non-e-commerce industries? In addition, it can be seen from the previous description that both e-commerce and poverty have obvious spatial agglomerations. E-commerce poverty alleviation, whether there is a spatial effect on the direct and indirect effects of efficiency, that is, a spatial spillover effect, the spillover effect of e-commerce poverty alleviation efficiency is largely reflected in the virtual space formed by the Internet. In addition, the maturity and improvement of measurement methods, spatial measurements, dynamic panel data, quantile regression, and other methods provide empirical tools for studying the diffusion effect of e-commerce poverty alleviation efficiency. For reference, please see the research of Zhuang Ziyin, and Hua Rui (2017) [8] and Gui Huang Bao (2014) [9]. The authors have, in this paper, attempted to construct a spatial Durbin model to measure the direct and indirect effects of the poverty alleviation efficiency of e-commerce in various provinces and conducted a spatial analysis of the two effects to discuss their spatial spillover effects. The super-efficiency DEA model was used to evaluate the efficiency of e-commerce poverty alleviation in China, and a spatial Durbin model was constructed to decompose and calculate the direct and indirect effects of e-commerce poverty alleviation in different provinces and cities in order to provide theoretical reference for the solving of poverty problems in China.

The main contributions of this paper include the following: (1) Against the background of the COVID-19 pandemic, the sustainable development of the economy and society was severely altered, and exploring the poverty alleviation efficiency of e-commerce has important practical significance for realizing the sustainable development of the regional economy. (2) The direct and indirect effects of e-commerce poverty alleviation efficiency were measured, and its spatial spillover benefits were analyzed on this basis; this is the main expansion of and improvement to the existing research. The improvement of the poverty alleviation efficiency of e-commerce has important practical significance.

The main structure of this paper is as follows: The first part introduces the research background of this paper and describes its research motivation and significance. The second part, with a review of the domestic and foreign literature, presents the contributions of scholars in this field and the research innovation of this paper. The third part expounds on the method used in this paper and points out the applicability of the method. The fourth part measures the poverty alleviation efficiency of China’s regional e-commerce through empirical analysis and explores its spatial spillover effect on this basis. The fifth part is the conclusion of this paper and summarizes the research results and provides relevant policy recommendations.

2. Literature Review

2.1. Research on E-Commerce Development in Urban and Rural Areas

Over the past several decades, with the continuous improvement of the Internet environment and of various e-commerce technologies, e-commerce has been widely used in various sectors of the economy and society [10]. The huge market and potential business opportunities it brings have accelerated economic development and social progress, fundamentally changed the traditional business model, and promoted the process of economic globalization. As the basis of national economic growth, cities have become an important carrier for the development of e-commerce in today’s economic globalization and networking. In China, due to geographical and historical factors, the economic levels and degrees of informatization in China’s cities and rural areas are different. The development of e-commerce is extremely dependent on the construction of urban infrastructure and the level of informatization, which also leads to a significant difference in the level of e-commerce between urban and rural areas in China [11]; this is mainly reflected in the following aspects: (1) The growth of the urban e-commerce system is slow, and the development time of urban e-commerce is also longer than that of the rural mall system. Therefore, although the development of urban e-commerce is relatively mature, the growth rate is relatively slow. Compared to the development of urban e-commerce, the development potential of rural e-commerce is greater, and there will be substantial growth in the future [12]. (2) The network coverage of the urban system is large. The network is the infrastructure for the development of the e-commerce system. Where there is no network, whether in urban or rural areas, there is no possibility of the emergence of a mall system. The development of the urban mall system is larger than the rural mall system in terms of network coverage. If the rural e-commerce system wants to develop in the future, the first step would be to establish a network that can rival that of the cities. (3) The logistics system of the rural shopping mall system is not perfect. Because the populations in rural areas are not compact and dense, and rural e-commerce is only in its infancy, the logistical capability of the rural shopping mall system does not match that of the urban e-commerce system. Building a sound and complete logistics platform is also an important channel for rural e-commerce to catch up with urban e-commerce [13].

During the COVID-19 pandemic, the number of face-to-face transactions decreased significantly, and e-commerce achieved rapid growth. Due to the lack of market demand and economic strength in rural areas, the early development of e-commerce was mostly concentrated in large and medium-sized cities, and the development of the rural market was not given top priority [14]. However, based on the actual situation, the information needs of rural areas and the vast number of farmers in the country are very high, and some rural areas that are dominated by agriculture, industry, tourism and culture, and commercial trade have particularly significant demands for e-commerce. At the same time, in recent years, policies have also attached great importance to the rural market, and the continuous development of the rural economy has also opened up a broad market space for the development of e-commerce. Therefore, the rural e-commerce market has great potential and is a strategic market for the future. The infrastructure conditions of rural e-commerce will determine the battle against poverty in rural China, and e-commerce poverty alleviation has become the main supporting measure for the sustainable development of the rural economy.

2.2. Research on E-Commerce Poverty Alleviation

At present, e-commerce poverty alleviation studies focus on three aspects: The first aspect is its characteristics and methods. Teng et al. [15] pointed out that the main reason for poverty in most areas is that, in the operation and sales of agricultural and forestry products, they are unable to reach larger market environments and consumer groups. E-commerce poverty alleviation can help poor farmers directly connect to the larger Internet market and promote their goods. Supply and demand seamlessly connect, increasing the product price and income by de-intermediation and helping the poor to become richer. Lin and Kang [16] pointed out that helping the poor to move out of poverty can not only allow the poor to participate directly in the entrepreneurship or employment of rural e-commerce, but it also benefits the local e-commerce industry. The second aspect is the poverty alleviation effect and mechanism. E-commerce facilitates rural industrial development by promoting information sharing, factor flow, and resource docking [17], and has an obvious positive effect on the increase in the income of enterprises in poverty-stricken areas [18]; this is the “endogenous driving force” to promote economic development in impoverished areas [19]. At the same time, e-commerce has effectively alleviated the problem of low-price sales caused by slow sales and information asymmetry. The development of e-commerce enterprises not only provides employment and entrepreneurship opportunities for the poor but also indirectly drives the integration of production and villages in poverty-stricken areas, providing transportation, logistics, and information. The development of diversified industries, such as entertainment and tourism, has created conditions that effectively bridge the digital divide in poverty-stricken areas and help the poor to achieve an “increase in income” [20,21].

2.3. Research on the Spatial Effect of E-Commerce Poverty Alleviation

Unlike the analysis of poverty spaces, rural e-commerce research introduced spatial analysis factors very early. In reality, rural e-commerce often presents spatial aggregation characteristics, and major e-commerce platforms have infiltrated rural areas. The rural Taobao project of the rural Taobao and Xinnong people, which is popular in rural areas and promoted by e-commerce and poverty alleviation initiatives, has a significant step characteristic, with Jiangsu, Zhejiang, and Shanghai as the core regions of its spread [22]. From the perspective of spatial clustering of the population, e-commerce reduces the information asymmetry of residents through knowledge and technology spillovers and increases the consumption level of poverty-stricken areas. Meanwhile, rural residents’ information consumption has a certain positive spatial effect overflow among regions [23]. Regardless of the situation of the rural e-commerce spatial distribution, scholars believe that e-commerce has a positive impact on poverty alleviation. By breaking the regional and time constraints in traditional business activities, farmers in poverty-stricken areas rely on e-commerce platforms to realize new agricultural and commercial relationships between joint products, joint facilities, joint standards, joint data, and joint markets, ultimately achieving cross-regional and inter-time production. In addition, e-commerce platforms have greatly increased the flow of agricultural products into the cities by virtue of lower customer acquisition costs. The development of the rural e-commerce industry has brought about a rapid gathering of upstream and downstream industrial chains and supporting service industries. At the same time, e-commerce has also been promoted. Other industries, especially those related to agriculture, manufacturing, and services, promote economic growth [24]. Economic growth is the main avenue for poverty reduction. Under the nonlinear poverty alleviation paradigm, financial poverty alleviation, with supportive loans and agricultural aid services as its main forms, also relies on rural e-commerce platforms to achieve poverty reduction [25,26,27]. Yang and Shi (2019) used the non-radial super-efficiency DEA model to measure the poverty alleviation efficiency of e-commerce in 36 counties and cities in the Dabie Mountains, one of the typically contiguous destitute areas. Spatial statistical analysis methods, including correlation and hotspot analysis, were used to analyze the spatial characteristics of e-commerce poverty alleviation efficiency in the Dabie Mountains [28]. Zhang and Tang (2019) analyzed the impact of e-commerce on poverty alleviation efficiency into direct and indirect effects and discussed the spatial effect of e-commerce poverty alleviation efficiency by constructing a modified Feder model [29].

Although the academic community has carried out a number of studies on e-commerce poverty alleviation, there is still room for further expansion, as follows: (1) In the past, scholars have mainly studied the current situation and problems of e-commerce poverty alleviation, but few scholars have studied the poverty alleviation efficiency of e-commerce. This research expands and improves upon the existing research. (2) Previous studies have explored e-commerce poverty alleviation from the perspective of space. The authors of this paper analyzed its direct effect, indirect effect, and spillover effect from the perspective of the spatial effect, which can provide a theoretical reference for e-commerce poverty alleviation.

3. Methods

At present, there are two main tools for efficiency evaluation: stochastic frontier analysis (SFA) and data envelope analysis (DEA). SFA is a parameter method that uses the stochastic frontier production function for efficiency estimation. The advantage of SFA is that it considers the influence of the existence of random errors on the results, determines the form of the production function in advance, and then studies the production process of the enterprise, which can improve the accuracy of the computational technical efficiency. It can also analyze the correlation between efficiency and influencing factors, but SFA cannot solve a multi-output problem. Compared to SFA, DEA can handle the efficiency measurements among multiple inputs and multiple outputs of different measurement units. It does not need to assume the expression of the production function relationship in advance to avoid the problem of parameter estimation, and it can avoid the problem of subjective weighting. Efficiency is a comprehensive indicator, which is more suitable for the comparison of efficiency among evaluation units, and it can provide managerial decision-makers with useful information on efficiency improvement. At the same time, DEA-SBM (a method based on slack measurement) can solve the problem of undesired outputs; therefore, DEA is more suitable for the calculation of the poverty alleviation efficiency of e-commerce.

3.1. Super-Efficiency DEA Evaluation Method

(1)

Model construction

The DEA model is a data envelopment analysis method to measure the operational efficiency of multiple entities. This method finds the frontier of efficiency with the possible set of production in the economic system, and then uses the efficiency front as an indicator of efficiency. The production possible set is a set of various input and output combinations, and the efficiency front is the set of the maximum outputs that all possible input combinations can form. The data envelopment analysis can be linearly planned. The method calculates the efficiency frontier and evaluates the production efficiency of decision-making units (DMU) through the efficiency frontier. The decision-making unit on the efficiency front is relatively effective.

If a decision unit has an input vector of X = (x₁, x₂,…, x_i,…, x_m) in an economic activity, where xi is the i-th input, the output vector is Y = (y₁, y₂,…, y_r,..., y_s), where y_r represents the r-th output and (x_j, y_j) represents the input vector and output vector of the j-th decision unit, and (X₀, Y₀) is the corresponding indicator of the decision unit. Therefore, (X, Y) is the economic activity of this decision-making unit. n decision-making units are derived from an input matrix of an order of n × m, and the corresponding output set can constitute an n × s order production out of the matrix. Chaens et al. [30] proposed a CCR model for data envelopment analysis to evaluate the overall effectiveness of the technology.

$$\mathrm{s}.\mathrm{t}.\{\begin{array}{l}{\displaystyle {\sum }_{j=1}^n{\lambda }_j{X}_j}+s^{-}={\theta }_i{X}_i,i=1,2,\dots ,n\\ {\displaystyle {\sum }_{j=1}^n{\lambda }_j{X}_j}-s^{+}=Y_i,i=1,2,\dots ,n\\ {\lambda }_j\ge 0,j=1,2,\dots ,n\\ s^{+}\ge 0,s^{-}\ge 0\end{array}$$

(1)

Since data envelopment analysis has the potential to derive multiple effective decision-making units that result in no further comparisons, Anderson and Christian [31] proposed a super-efficiency model for data envelopment analysis that can be used to compare each decision-making unit. The basic principle of the super-efficiency model is to keep the existing technology level unchanged, and the input elements of the decision-making unit are scaled up proportionally. The efficiency value is constant for both the effective and ineffective decision-making units. The super-efficiency model of data envelopment analysis can be expressed as follows:

$$\{\begin{array}{l}\mathrm{s}.\mathrm{t}.{\displaystyle {\sum }_{j=1}^n{\lambda }_j{X}_j}+s^{-}=\theta t^{super}t{x}_0\\ {\displaystyle {\sum }_{j=1}^n{\lambda }_j{X}_j}-s^{+}=Y_0\\ {\displaystyle {\sum }_{j=1}^n{\lambda }_j}=1\\ \forall {\lambda }_j\ge 0,j=1,2,\dots ,n,\\ s^{+}\ge 0,s^{-}\ge 0.\end{array}$$

(2)

where $\theta t^{super}$ represents the super-efficiency value, which is the new decision unit combination ratio reconstructed from the original decision unit.

(2)

Selection of input and output indicators

When choosing to use data, including analysis methods, it is necessary to determine the input and output indicators. The authors of the article consulted many literature materials, experts, and scholars in the field [28,32]. The input indicators were mainly used to evaluate the input of various production factors. Regarding the variables, the authors chose the degree of the mechanization of agricultural production, the number of pension institutions, tourism income, total agricultural output value, village clinics, and number of cultural stations. In terms of the output indicators, the purpose of poverty alleviation to promote local economic development and help farmers to move out of poverty and become wealthier is a variable to enable the evaluation of efficiency. Therefore, the non-poverty incidence and per capita GDP were selected. A total of 30 domestic regions in China were selected, and the sample interval was 2010–2021. The data were taken from the “China Rural Statistical Yearbook”, the “China Cultural Relics Statistical Yearbook”, the “China Rural Poverty Monitoring Report”, and the “China Statistical Yearbook”.

3.2. Spatial Measurement

3.2.1. Spatial Correlation Analysis

The spatial correlation between the observation points needed to be verified first. If it existed, the appropriate spatial measurement model was used to carry out the analysis; otherwise, the traditional measurement model was used for estimation. The spatial correlation test mainly includes two methods: global and local [33,34]. The purpose of the global spatial correlation test is to study the spatial distribution properties of spatial samples in the whole spatial geographic system, and it can be verified by the global Moran’s I index. The local spatial correlation test is used to study the spatial distribution properties of the observation points in the spatial geographic unit subsystem, which is usually tested by the Moran’s I scatter plot. The authors selected Moran’s I index for testing. The mathematical expression is as follows:

$$Moran’sI=\frac{{{\displaystyle \sum }}_{i=1}^n{{\displaystyle \sum }}_{j=1}^n{W}_{ij}\left(Y_i-\overline{Y}\right)\left(Y_j-\overline{Y}\right)}{S^2{{\displaystyle \sum }}_{i=1}^n{{\displaystyle \sum }}_{j=1}^n{W}_{ij}}$$

(3)

The Moran’s I index value is from −1 to 1. If the value is less than 0, the spatial sample observation points show a negative correlation. The closer the value is to −1, the greater the difference between spatial units; if Moran’s I is greater than 0, the spatial geographic unit exhibits a positive correlation. If the value is closer to 1, the correlation between spatial units will be high. If the value is equal to 0, this means that the observation points are randomly distributed in space. When the value is closer to 0, the correlation between the samples will be lower.

3.2.2. Panel Spatial Econometric Model

The spatial econometric models include the basic spatial lag model (SLM), the spatial error model (SEM), and the comprehensive spatial Durbin model (SDM). SLM is used to identify the spillover effect or diffusion effect of the explained variable in space, while SEM measures the degree to which the explained variable is affected when the spatial factor caused by the explanatory variable is omitted. SDM is a combination of SLM and SEM, allowing spatial dependencies to exist in all variables at the same time. First, a standard panel data model was built:

$$Y_{it}=\alpha +X_{it}\beta +c_i+{\mu }_i+{\epsilon }_{it}$$

(4)

In the formula, i represents the cross-sectional dimension, t represents the time dimension, $X_{it}$ represents the explanatory variable matrix of $n\times k$, and $\beta$ is the k dimension-free coefficient vector. $c_i$ is the individual effect, ${\mu }_i$ is the time effect, and ${\epsilon }_{it}$ represents the random error term.

(1)

Spatial lag model (SLM)

The right side of the equation of the standard model is transformed into SLM by adding the spatial lag term of the explanatory variable:

$$Y_{it}=\alpha +X_{it}\beta +\rho {{\displaystyle \sum }}_{j=1}^N{w}_{ij}{Y}_{jt}+c_i+{\mu }_i+{\epsilon }_{it}$$

(5)

where $w_{ij}$ represents the spatial weight coefficient, which can be in the $\mathrm{n}\times \mathrm{n}$ dimensional matrix form, $\rho$ indicates the coefficient of the spatial lag term of the explained variable, and $N$ indicates the lag period.

(2)

Spatial Error Model

When the spatial effect due to the explanatory variables cannot be represented in the model, this spatial effect will be reflected in the error term, forming the SEM:

$$Y_{it}=\alpha +X_{it}\beta +c_i+{\mu }_i+v_{it}$$

(6)

$$v_{it}=\lambda {{\displaystyle \sum }}_{j=1}^N{w}_{ij}{v}_{jt}+{\epsilon }_{it}$$

(7)

where $\lambda$ represents the coefficient of the spatial error lag term. The other symbols have the same meanings as above.

(3)

Spatial Durbin model (SDM)

The spatial spillover effect of poverty alleviation efficiency leads to obvious spatial agglomeration properties of poverty alleviation efficiency, that is, the poverty alleviation efficiency of a certain region will affect the poverty alleviation efficiency of neighboring regions. At the same time, the production activities among different provinces do not exist independently. Economic activities in a certain area have an impact on neighboring areas, which in turn has led to changes in poverty alleviation efficiency. The spatial Durbin model includes the spatial correlation of the dependent variable and the independent variable. Pace and Lesage [32] found that the spatial Durbin model can well reflect the externalities and spatial spillover effects caused by different influencing factors.

The panel data model mainly includes two forms: the fixed effect and the random effect. Before the panel data model was constructed, it was necessary to determine which of these forms to use. In the fixed effect model, the intercept between the cross-sections differs, and the slope coefficients are the same. In the random effect model, the difference between individuals appears as an intercept variation, but the variation is random. If the observation points are randomly selected to represent a larger population, a random impact model should be used. The authors explored the spatial spillover effect of e-commerce poverty alleviation efficiency in 30 provinces (cities and districts) in China, so a fixed effect model was adopted, and the selection was determined by the Hausman test. The Durbin model is as follows:

$$Y_{it}={\eta }_i+{\alpha }_t+\delta W_{ij}{Y}_{it}+{\beta }_1{X}_{it}+{\beta }_2{W}_{ij}{X}_{it}+{\epsilon }_{it}$$

(8)

where W_ij (N × N) is a geospatial weight matrix, ij represents a spatial unit, t represents the year, Y_it represents the N × 1 vector, X_it represents the N × K matrix, N represents the number of spatial observation points, and K is the explanation. ${\eta }_i$ and ${\alpha }_t$ are the spatially fixed and time-fixed variables, respectively, ε_it is a random disturbance term, δ represents the degree of change in the surrounding area caused by the change of the explained variable in a certain area, β₁ represents the degree of change of the explained variable in the same area caused by the change of the explanatory variable in a certain area, and β₂ represents the degree of change of the explanatory variable in a certain area to the explained variable in the surrounding area. If β₂ = 0, the SLM model was selected. If β₂ + δβ₁ = 0, the SEM model was selected. In this study, the parameters of the maximum likelihood estimation model were used, and the Wald test and LR test were used to verify whether the spatial panel Durbin model can be divided into a spatial lag model and a spatial error model. Direct and indirect effects were used to explain the results of the spatial econometric model estimates. Direct effects refer to the average variation in the province’s dependent variables caused by changes in the province’s independent variables. Indirect effects refer to the average changes in the neighborhood dependent variables caused by changes in the independent variables in the region.

3.2.3. Spatial Spillover Effect Decomposition

The total effect is the sum of all the effects. The improvements to the original model are as follows:

$$\left(I_n-\rho W\right)Y=\alpha l_n+\beta X+\theta WX+\epsilon$$

(9)

$$\mathrm{Y}={{\displaystyle \sum }}_{r=1}^k{S}_r\left(W\right)X_r+V\left(W\right)l_n\alpha +V\left(w\right)\epsilon$$

(10)

$$S_r\left(W\right)=V\left(W\right)\left(I_n{\beta }_r+W{\theta }_r\right)$$

(11)

$$\mathrm{V}\left(W\right)={\left(I_n-\rho W\right)}^{-1}=I_n+\mathsf{\rho }\mathrm{W}+{\rho }^2{W}^2+{\rho }^3{W}^3+\cdots$$

(12)

where r represents the number of independent variables, X represents the r-th independent variable value, and θ represents the r-th lag variable coefficient of the lag variable Wr. Then, the following matrix relationship is as follows:

$$\left(\begin{array}{c}{Y}_1\\ Y_2\\ M\\ Y_n\end{array}\right)={{\displaystyle \sum }}_{r=1}^k\left(\begin{array}{cccc}{S}_r(W{)}_{11}& S_r(W{)}_{12}& \dots & S_r(W{)}_{1n}\\ S_r(W{)}_{21}& S_r(W{)}_{22}& \dots & S_r(W{)}_{2n}\\ \dots & \ddots & & ⋮\\ S_r(W{)}_{n1}& S_r(W{)}_{22}& \dots & S_r(W{)}_{kn}\end{array}\right)\left(\begin{array}{c}{X}_{1r}\\ X_{2r}\\ ⋮\\ X_{kr}\end{array}\right)+V(W)l_n\alpha +V(W)\epsilon$$

(13)

$$Y_i={{\displaystyle \sum }}_{r=1}^k\left[S_r{\left(W\right)}_{i1}{X}_{1r}+S_r{\left(W\right)}_{i2}{X}_{2r}+\cdots S_r{\left(W\right)}_{ik}{X}_{kr}\right]+V\left(W\right)l_n\alpha +V\left(W\right)\epsilon$$

(14)

$$\frac{\partial Y_i}{\partial X_{jr}}=S_r{\left(W\right)}_{ij}$$

(15)

$$\frac{\partial Y_i}{\partial X_{ir}}=S_r{\left(W\right)}_{ii}$$

(16)

where S_r(W)_ij evaluates the impact of the r-th explanatory variable in the region j on the explained variable in the region i, while S_r(W)_ii evaluates the impact of the rth explanatory variable in region i on the local explained variable impact. If j and r are not equal, then Y_i to X_jr is usually not 0, and the i-th and j-th elements contained in the S_r(W) matrix will affect it. At the same time, the partial derivative will usually not be βr. In addition to the obvious correlation between the regional explanatory and explained variables in this region, there is also an internal correlation with the explained variables in the surrounding regions. The theory proposed by Pace and Lesage considers that these variables are the direct effect and the indirect effect, and the sum of the two is the total effect. In short, direct effects represent the intra-regional spillover effects, while indirect effects represent the inter-regional spillover effects.

4. Results

4.1. Results of Poverty Alleviation Efficiency of E-Commerce in Various Regions of China

In this study, the data were substituted into the DEA model, and the results were obtained, as shown in

Table 1. Results of e-commerce poverty alleviation efficiency in various regions of China.

Region		2010	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020	2021	Mean
East	Beijing	2.012	2.015	2.018	2.112	2.132	2.146	2.157	2.188	2.167	2.233	2.256	2.269	2.142
	Tianjin	1.793	1.799	1.813	1.823	1.848	1.986	1.995	2.018	2.117	2.165	2.167	2.178	1.975
	Hebei	0.489	0.493	0.499	0.503	0.501	0.511	0.516	0.522	0.518	0.533	0.536	0.522	0.512
	Liaoning	0.401	0.413	0.424	0.436	0.441	0.447	0.449	0.458	0.467	0.472	0.478	0.385	0.439
	Shanghai	1.986	1.994	1.999	2.012	2.018	2.119	2.134	2.144	2.156	2.167	2.189	2.199	2.093
	Jiangsu	0.811	0.815	0.819	0.826	0.838	0.849	0.864	0.877	0.889	0.895	0.915	0.933	0.861
	Zhejiang	0.798	0.803	0.816	0.828	0.837	0.854	0.869	0.889	0.898	0.901	0.905	0.916	0.860
	Fujian	0.765	0.789	0.797	0.805	0.811	0.818	0.829	0.833	0.852	0.866	0.883	0.898	0.829
	Sahndong	0.465	0.469	0.471	0.477	0.487	0.495	0.501	0.513	0.522	0.534	0.544	0.556	0.503
	Guangdong	0.891	0.899	0.911	0.927	0.936	0.949	0.959	0.969	0.989	0.995	0.999	1.101	0.960
	Hainan	0.613	0.615	0.618	0.624	0.627	0.635	0.647	0.688	0.701	0.733	0.762	0.782	0.670
Central	Shanxi	0.412	0.415	0.425	0.433	0.437	0.453	0.476	0.488	0.498	0.517	0.542	0.573	0.472
	Jilin	0.501	0.506	0.511	0.514	0.516	0.515	0.517	0.521	0.524	0.528	0.533	0.542	0.519
	Heilong jiang	0.572	0.573	0.582	0.587	0.586	0.590	0.592	0.598	0.603	0.608	0.612	0.619	0.594
	Anhui	0.611	0.617	0.623	0.632	0.639	0.648	0.655	0.667	0.678	0.682	0.687	0.693	0.653
	Jinagxi	0.711	0.719	0.714	0.721	0.722	0.719	0.725	0.729	0.737	0.753	0.765	0.768	0.732
	Henan	0.501	0.513	0.515	0.517	0.519	0.527	0.524	0.529	0.535	0.538	0.544	0.553	0.526
	Hubei	0.476	0.481	0.488	0.491	0.494	0.498	0.503	0.508	0.514	0.519	0.517	0.525	0.501
	Hunan	0.651	0.655	0.664	0.674	0.672	0.683	0.689	0.691	0.693	0.698	0.707	0.718	0.683
West	Neimenggu	0.578	0.581	0.589	0.584	0.588	0.593	0.596	0.603	0.608	0.609	0.612	0.616	0.596
	Guangxi	0.603	0.609	0.611	0.614	0.619	0.617	0.618	0.619	0.624	0.627	0.626	0.635	0.619
	Chonhqing	0.503	0.505	0.511	0.513	0.519	0.518	0.522	0.529	0.527	0.535	0.538	0.544	0.522
	Sichaun	0.515	0.518	0.523	0.528	0.531	0.536	0.534	0.541	0.547	0.552	0.558	0.566	0.537
	Guizhou	0.316	0.319	0.326	0.336	0.341	0.348	0.364	0.373	0.383	0.389	0.394	0.399	0.357
	Yunnan	0.502	0.505	0.509	0.513	0.519	0.526	0.527	0.534	0.536	0.535	0.538	0.547	0.524
	Shanxi	0.442	0.445	0.451	0.456	0.464	0.476	0.481	0.487	0.495	0.498	0.501	0.507	0.475
	Gansu	0.209	0.213	0.215	0.216	0.223	0.226	0.231	0.237	0.239	0.241	0.352	0.254	0.238
	Qinghai	0.363	0.365	0.369	0.373	0.379	0.386	0.388	0.389	0.391	0.393	0.392	0.398	0.382
	Ningxia	0.423	0.427	0.435	0.436	0.438	0.445	0.463	0.477	0.489	0.498	0.534	0.565	0.469
	Xinjiang	0.341	0.345	0.349	0.354	0.361	0.369	0.375	0.381	0.388	0.387	0.394	0.397	0.370

.

Table 1 shows that the change trends of poverty alleviation efficiency in different regions are quite different. Some regions have a faster growth rate, such as Shanxi and Hainan. Although some regions have a growth trend, the changes are relatively stable and there are no major changes. From the perspective of the spatial dimension, the differences among regions are large, and the distribution in the country also has typical differences, which are closely related to the geographical locations and resource endowment of the regions. The overall poverty alleviation efficiency is the highest in the eastern part of the country, including Tianjin, Beijing, and Shanghai, and the lowest in the western part of the country.

4.2. Spatial Correlation Analysis Results

The data were derived from the results of the previous DEA model calculation and substituted into the GeoDa software to obtain the results (

Table 2. E-commerce poverty alleviation efficiency Moran’s I index statistics (2010–2021).

Year	Poverty Alleviation Efficiency		Year	Poverty Alleviation Efficiency
	Moran	p-Value		Moran	p-Value
2010	0.5034	0.0003	2016	0.5233	0.0004
2011	0.5512	0.0011	2017	0.5245	0.0013
2012	0.5034	0.0009	2018	0.5278	0.0004
2013	0.5316	0.0012	2019	0.5363	0.0004
2014	0.5019	0.0014	2020	0.5378	0.0009
2015	0.5176	0.0011	2021	0.5344	0.0016

).

In Table 2, the Moran’s I index values all exceed 0.5, indicating that the spatial effect is significant. Therefore, the existence of spatial effects cannot be ignored.

4.3. Results of Spatial Spillover Effect Decomposition of China’s E-Commerce Poverty Alleviation Efficiency

4.3.1. Model Checking

(1)

Model selection test

In this study, using MATLAB software, the Wald and LR tests were used to verify the spatial Durbin model. The results are shown in

Table 3. Spatial panel data Wald and LR test results.

Test Method	Statistical Value	p-Value
Wald spatial lag	77.091	0.001
Wald spatial error	75.223	0.001
LR spatial lag	66.129	0.002
LR spatial error	63.339	0.001

, including the Wald spatial lag (H0: no spatial autocorrelation), Wald spatial error (H0: no spatial error), LR spatial lag (H0: no spatial autocorrelation), and LR spatial lag (H0: no spatial error). The test rejected the null hypothesis. Therefore, the spatial Durbin model cannot be divided into the spatial lag model and the spatial error model. Therefore, the spatial Durbin model is the most suitable model for analyzing the problems in this chapter.

(2)

Model Form Test

The authors explored the spatial effects of poverty alleviation efficiency in various provinces in China. The time span used was 12 years. Due to the short period of time and the many cross-sections, short panel data were used, so the unit root test and cointegration test were not performed on the data. In this study, using EViews software, the random or the fixed effect was selected by the Hausman test. The results are shown in

Table 4. Spatial panel data Hausman test results.

Hausman Test	Chi-Sq. Statistic	Chi-Sq. d.f.	Prob.
Cross-section random	21.31	10	0.022

.

In Table 3, the Hausman Test statistic value of the poverty alleviation efficiency of various provinces in China is 21.31, and p is 0.022, rejecting the null hypothesis. Therefore, the spatial Durbin model adopted a fixed effect model.

4.3.2. Spatial Durbin Model Estimation Results

(1)

Indicator selection

According to the principles of reliability and availability, and drawing from relevant scholars [35], the authors analyzed the influencing factors on China’s regional e-commerce poverty alleviation efficiency from the transportation infrastructure level [28], communication facilities level [36], financial support [37], industrial agglomeration [38], human capital level [39], and financial environment level [40]. Among these, the level of county transportation infrastructure was mainly measured by the mileage of county-level highways (national highways and provincial highway traffic lengths). The level of communication facilities was mainly measured by the number of fixed-line telephone users, and the measurement of financial support used was the public finance expenditure of the local government. The economic benefits of agglomeration were mainly measured by the population density [41,42]. At the same time, considering that there are few higher education institutions in the county areas, the practice of measuring the level of human capital per 10,000 students, used in the past, is not accurate. Therefore, according to the practices of Zhang and Chen [43], the number of students per school was considered to be 10,000 students for the purpose measuring the level of human capital. Finally, the level of the financial environment was measured by the balance of various loans of financial institutions.

(2)

Model construction

The authors introduced the spatial lag variable to construct the spatial measurement model, including the “spatial effect”. The data on e-commerce poverty alleviation efficiency were obtained from the previous calculation results. The model is constructed as follows:

$$\begin{gathered} LnEF{F}_{i,t}={\alpha }_i+{\beta }_{1}LnJ{T}_{i,t}+{\beta }_{2}LnT{X}_{i,t}+{\beta }_{3}LnC{Z}_{i,t}+{\beta }_{4}LnJ{J}_{i,t}+{\beta }_{5}LnR{L}_{i,t}+ \\ {\beta }_{6}LnJ{R}_{i,t}+\gamma W{X}_{-i,t}+{\gamma }_t+{\mu }_{i,t}, {\mu }_{i,t}=\mathsf{\rho }W{\mu }_{-i,t}+{\epsilon }_{i,t},\left|\rho \right|\le 1 \end{gathered}$$

(17)

where t is the year, i is the region, ${\alpha }_i$ and ${\gamma }_t$ represent the regional and temporal disturbances, ε_i,t is the disturbance term, and ρ represents the spatial lag coefficient. The model examines the “spatial effect” of each influencing factor on the regional e-commerce poverty alleviation efficiency by introducing the spatial lag term $W{\mu }_{-i,t}$ of all explanatory variables. If the coefficients of all spatial lag variables are not significant, then there is no “spatial effect”. Conversely, if the coefficients of one or more spatial lag variables are significant, then the “spatial effect” is not negligible. The spatial lag variable coefficient and significance level reflect the specific direction and effect of the “spatial effect” of each influencing factor.

(3)

Estimation results

In this study, the spatial panel software program written by Elhorst [44] was used to perform the operation of the spatial panel Durbin model through the MATLAB software. The spatial Durbin model regression results are shown in

Table 5. Spatial panel Durbin model regression results.

Variable	Regression Coefficient	Variable	Regression Coefficient
LnJT	0.112 ***	W × LnJT	0.145 ***
LnTX	0.331 ***	W × LnTX	0.361 ***
LnCZ	0.034 ***	W × LnCZ	0.112 ***
LnJJ	0.219 *	W × LnJJ	0.204 **
LnRL	0.003 ***	W × LnRL	0.012 ***
LnJR	0.009	W × LnJR	0.013 *
W*dep.var	0.1765 ***
R-squared	0.9877
log-likelihood	461.092

Note: *, **, and *** are significant at 10%, 5%, and 1%.

.

From the regression results, R² = 0.9877 shows that the explanatory variables explain the vast majority of the explained variables. The spatial autoregressive coefficient in Table 5 is 0.1765 and passed the significance test, which means that the e-commerce poverty alleviation efficiency has a positive spillover effect on its neighborhood, that is, for every 1% increase in a region’s e-commerce poverty alleviation efficiency, there will be a resultant 0.1765% increase in poverty alleviation efficiency, meaning that there is a significant spatial effect of the efficiency of China’s e-commerce poverty alleviation. It also can be seen that at least four of the six variables, after the inclusion of spatial effects, are significant, indicating that the model fits more scientifically and accurately after the inclusion of spatial effects.

(4)

Spatial spillover effect decomposition results of e-commerce poverty alleviation efficiency

In the normal panel data model, if there is no influence of the spatial lag term, the model regression coefficient can be considered to reflect the influence of the independent variable on the dependent variable. However, in the spatial Durbin model, due to the existence of correlation, the spatial lag term is added, and the regression coefficients of the model cannot simply describe the influence of the explanatory variables on the interpreted variables. The authors drew on the effects used by Lesage and Pace in 1999 to analyze the spatial effects of e-commerce poverty alleviation efficiency in 30 regions. Direct effects refer to those of the explanatory variables on the interpretative variables in the region. The role of indirect effects is the effect of explanatory variables in the region on the interpretative variables in the neighborhood. The total effect refers to the average effect of the independent variables on all regions. The direct and indirect effects of China’s e-commerce poverty alleviation efficiency are shown in

Table 6. Direct and indirect effect estimates.

Variable	Direct Effect	Indirect Effect	Total Effect
LnJT	0.2312 ***	0.1113 ***	0.3425 ***
lnTX	0.2933 ***	0.0528 ***	0.3461 ***
lnCZ	0.0431 ***	0.0127	0.0558 ***
lnJJ	0.0899 ***	0.2134 ***	0.3033 *
lnRL	0.1198 ***	−0.0132	0.1066 ***
LnJR	0.2312	0.0146	0.2458 *

Note: *, *** are significant at 10% and 1%, respectively.

.

From Table 6, the following results can be obtained:

The following results are from the perspective of the transportation infrastructure level. The direct effect of the level of transport infrastructure is positive. For every 1% increase in the level of transport infrastructure, the impact on e-commerce poverty alleviation efficiency in the region is 0.2312%. The level of transport infrastructure is the key to e-commerce achieving its external logistics. The support of transportation is important, especially for rural areas and their various topographical features, although the construction of transportation infrastructure is relatively behind and unable to meet increasing logistical demands, greatly restricting the improvement of e-commerce poverty alleviation efficiency. Therefore, the improvement of transportation infrastructure can improve the efficiency of logistics transportation, promote rural economic development, and improve e-commerce poverty alleviation efficiency. The indirect effect is also positively correlated with the efficiency of e-commerce poverty alleviation. It tested at the significance level of 1%, indicating that the improvement of the transportation infrastructure level can not only promote the poverty alleviation efficiency of a region, but it can also promote the economic development of neighboring areas. As the improvement of transportation facilities is conducive to the flow of resources and labor, the efficiency of poverty alleviation in neighboring areas will be promoted through the diffusion effect.

The following results are from the perspective of financial support. The direct effect of financial support is 0.0431, the indirect effect is 0.0127, and the total effect is 0.0558. The greater the government’s financial support, the higher the efficiency of e-commerce poverty alleviation. The efficiency of e-commerce poverty alleviation will increase by 0.0558%. The direct effect is positive, meaning that the strong financial support of the region can not only accelerate the construction, incubation, growth, and e-commerce enterprise development in the region, but it can also help to optimize the external environment for e-commerce development in the region. This can improve regional poverty as internal e-commerce helps the poor. The indirect effect is also positive, meaning that with financial support, the e-commerce industry in the region will develop, which can promote the industrial development of neighboring regions as well as economic growth, thus realizing an increase in the disposable income of residents.

The following results are from the perspective of the human capital level. The direct effect of the human capital level is 0.1198, indicating that the improvement of the human capital level provides protection for the reserve and training of regional electric merchants and can provide intellectual support for regional e-commerce development. The development potential endorses the positive role of e-commerce and poverty alleviation, as well as the motivation of talent for the development of the e-commerce industry; therefore, regions should become attractive to talent. The indirect effect is negative but does not pass the significance test, indicating that an increase in human capital investment in the region improves the efficiency of e-commerce poverty alleviation in the region, but it has less impact on e-commerce poverty alleviation efficiency in neighboring regions, mainly because of the talent pool. The talent will flow to the regions with better e-commerce industry, crowding out the talent pool and resulting in the loss of talent in the adjacent areas, which is not conducive to the efficiency of poverty alleviation in adjacent areas.

The following results are from the perspective of the communication facilities level. The direct effect of the level of communication facilities is positive. For every 1% increase in the level of communication facilities, the impact on e-commerce poverty alleviation efficiency in the region is 0.2933%. Communication facilities are the basis for e-commerce. Based on complete transportation facilities, the network communication environment, and other supporting infrastructure construction, creating a perfect network space is fundamental for e-commerce poverty alleviation to work with a stable development platform. Therefore, the direct effect of the level of communication facilities is relatively large and can promote e-commerce industry development to a greater extent, thereby improving the efficiency of poverty alleviation. The indirect effect is also positive but passes the test at the 10% significance level, which indicates that the improvement of the communication facilities in the region has less impact on the adjacent areas because e-commerce development depends on the network; therefore, the impact of network construction is more direct in the immediate area, and has less impact on the surrounding areas.

The following results are from the perspective of industrial agglomeration. The direct effect of industrial agglomeration is positive, and an increase of 1% of the industrial agglomeration level has an impact on the e-commerce poverty alleviation efficiency of 0.0899%. Industrial agglomeration is the organic combination of the e-commerce industry in the region. Companies are associated with their surrounding businesses. Industrial agglomeration can attract the flow of talent, capital, technology, and other resources, which can drive regional economic development and improve e-commerce poverty alleviation efficiency. The indirect effect is positive, but its elasticity coefficient is smaller than the direct effect. Due to the formation of regional industrial agglomeration areas, industry-related enterprises will be brought together near the center, which can drive the development of neighboring regional industries and promote their economic growth and improvement. The disposable income of residents will increase the efficiency of the poverty alleviation of e-commerce.

The following results are from the perspective of the financial environment level. The direct effect of the financial environment is positive but has not been verified with significant testing. Although the financial environment can provide the necessary development environment for e-commerce development, the development of rural e-commerce is lagging; many regions are relatively behind and in need of more e-commercial traffic. Regarding the construction of facilities and communication facilities, the short-term effect of the financial environment is not obvious, but the impact is positive. The indirect effects are also positive but fail to pass the significance test, indicating that the improvement of the financial environment has little impact on the poverty alleviation efficiency of the region and adjacent regions in the short term, and attention should be paid to the degree and direction of its impact in the long term.

5. Conclusions and Applications

5.1. Conclusions

The authors used a super-DEA model to measure the e-commerce poverty alleviation efficiency and the Moran’s I index to measure the spatial autocorrelation and the transportation infrastructure at different levels, including the communication facilities level, financial support, industrial agglomeration, human capital level, financial environment level, etc. The spatial Durbin panel model was used to explore the factors affecting e-commerce poverty alleviation efficiency and the spatial spillover effect. The conclusions are as follows:

(1)

From the perspective of space, the efficiency of e-commerce poverty alleviation varies greatly among regions, with Tianjin, Beijing, and Shanghai being the most efficient regions. In general, the efficiency of e-commerce poverty alleviation is the highest in the eastern region of China and the lowest in the western region of China.

(2)

There is a significant spatial autocorrelation effect in e-commerce poverty alleviation efficiency. The Moran’s I index values are all greater than 0.5; that is, e-commerce poverty alleviation efficiency among neighboring regions is high. Mutual influences have positive spatial correlations, while economic factors contribute to this spatial correlation.

(3)

From the regression results of influencing factors, the influence of the level of communication facilities is the most significant, and the influence coefficient reaches 0.331, which is the basic premise for the rapid development of e-commerce. At the same time, the level of transportation infrastructure is also at the core of promoting the rapid development of e-commerce. The elasticity coefficient factor had values up to 0.112. If the other factors remain fixed, the impact of the financial environment on the efficiency of poverty alleviation is minimal, but it is worth noting that it has a positive impact. Only by starting research in the medium and long terms can it be clear that these factors have a positive effect on poverty alleviation efficiency.

(4)

From the regression results of the influencing factors, the influence of the level of communication facilities is the most significant, and the influence coefficient reaches 0.331, which is the basic premise for the rapid development of e-commerce. At the same time, the level of transportation infrastructure is also at the core of promoting the rapid development of e-commerce. The elasticity coefficient factor has values up to 0.112. If the other factors remain fixed, the impact of the financial environment on the efficiency of poverty alleviation is minimal, but it is worth noting that it has a positive impact.

(5)

From the decomposition of the spillover effects, the three most significant factors from the perspective of direct effects are the level of communication facilities, transportation infrastructure, and the level of human capital. It can be seen that the levels of communication facilities, transportation infrastructure, and human capital are major factors affecting the future development of e-commerce and are the basic premise for its rapid development. It is worth noting that the direct effect of financial conditions is the least significant factor.

The analysis of the indirect effects shows that the most significant factors are the levels of communication facilities, transportation infrastructure, and industrial agglomeration. Industrial agglomeration is the most significant, which can provide assistance for the development of surrounding areas and effectively improve the efficiency of poverty alleviation in surrounding areas. It is worth noting that the indirect effect of human capital is not positive. The local area must attach great importance to the cultivation of human capital to avoid the loss of outstanding talents, and to improve the efficiency of local poverty alleviation.

5.2. Suggestions

In order to fully recognize the significance of the spatial diffusion effect of e-commerce poverty alleviation efficiency and improve rural poverty alleviation efficiency to promote rural revitalization, the government should mainly focus on the following four aspects:

First, support the Internet, e-commerce development environment, and improve entrepreneurs’ and farmers’ understanding of e-commerce, making full use of the diffusion of the direct effects of e-commerce and driving the development of e-commerce-related industries [45,46].

Second, strengthen the construction of rural logistics infrastructure, establish rural e-commerce sites, continue to improve e-commerce in rural comprehensive demonstration counties, and provide other support policies, driving all poverty-stricken areas to actively cooperate with Taobao, e-mail, and other third-party e-commerce service platforms to promote the construction of rural e-commerce public service systems [47,48,49].

Third, invest in poverty-stricken areas, seizing the opportunities brought about by the spatial diffusion effect of the Internet’s economic development, making full use of the Internet technology spillover effect and the demonstration effect of the externalities of e-commerce networks, cultivating endogenous hematopoietic capacity in poverty-stricken areas, improving poverty alleviation efficiency, and narrowing the regions’ differences in economic growth [50].

Fourth, increase publicity efforts. On the one hand, the investment in propaganda funds should be increased so that the propaganda of e-commerce poverty alleviation can cover the entire county, and even neighboring counties, as much as possible. In addition, part of the publicity funds should also be invested in the brand-building of local agricultural products to create agricultural products with local characteristics and make them irreproducible. On the other hand, in addition to focusing on the early publicity of e-commerce poverty alleviation, it is also necessary to carry out the middle and late publicity work of the policy [51,52]. The government should carry out in-depth and detailed theoretical publicity, explaining in detail the operations of e-commerce, and tell villagers about the flexibility and simplicity of e-commerce so that those living in poverty can more easily accept the concept of e-commerce.

Fifth, the targeted training of e-commerce talent should promote the integration of education poverty alleviation and e-commerce poverty alleviation in terms of talent construction. First, realize the orientation training of professional talents. For children from poor families with financial difficulties, selecting high school graduates who are interested in serving their hometowns and interested in e-commerce, and signing targeted training agreements can not only meet the needs of e-commerce talent but also retain talent. Secondly, a create a clear division of labor and improve management efficiency. In the e-commerce poverty alleviation cycle system, the personnel requirements for each link are different, and the allocation of functional personnel in the production link, platform operation and maintenance, logistics and transportation, and after-sales service should be reasonable. The e-commerce poverty alleviation model of “every household opens an online store” will not only cause inefficient sales due to the low cultural level of farmers but also make poor people put a lot of energy into sales and ignore production, which greatly reduces the efficiency of poverty reduction [11,53,54]. The whole process, from production to after-sale, should be based on the actual situation and rational allocation, and training should be carried out for specialized personnel and posts to improve the operation efficiency of the circulation system.

5.3. Limitations and Prospects

First, regarding the research on the theoretical model in this paper, although a certain expansion analysis has been carried out on the basis of relevant literature and theories, it needs to be further improved in terms of in-depth and systematic research. In addition, because some factors cannot be quantified, there are some shortcomings.

Second, the characteristics of the industry distribution of e-commerce are very obvious, but due to the limited number of observations of industry-level data, this paper does not discuss the poverty alleviation efficiency of e-commerce at the industry level.

Third, because the data mainly come from the provincial level, the variables comprising this data cannot play an effective role in the research at the city level.