The Spatial Characteristics of Sustainable Development for Agricultural Products E-Commerce at County-Level: Based on the Empirical Analysis of China

Abstract

This paper used the sectional data of Chinese counties to analyze the spatial distribution characteristics of sustainable development of e-commerce for agricultural products in China at the county-level. The standard deviation ellipses and Moran’s index were used to research this subject. Then, by constructing spatial measurement models, the spatial spillover effects and influencing factors of the development of county-level agricultural products e-commerce were analyzed from economic development, economic structure, economic vitality, and agricultural development. The results show that: (1) the top 100 counties of agricultural products e-commerce mainly concentrate in southeastern China, spreading along the northeast and southwest; (2) the county-level agricultural products e-commerce shows significant negative spatial spillover effects; (3) the level of economic development and public services have a positive impact on the development of county-level agricultural products e-commerce, while the level of industrial development shows a negative correlation; (4) the level of agricultural development and industrial development have a significant negative external effect on the development of agricultural products e-commerce. This study is of great significance to promote the sustainable development of agricultural products e-commerce, the process of rural urbanization and the optimization of county industrial patterns.

1. Introduction

Nowadays, e-commerce has become an important channel for the circulation of global commodities. Almost all companies are trying to expand their customers by trading online. In 2019, the sales volume of agricultural products e-commerce in China was 397.5 billion yuan, up 27% year-on-year and nearly eight times that of 2013. The huge population of China provides enough consumers to assist the standardized development of the agricultural products e-commerce market. Chinese e-commerce of agricultural products is in a period of stable growth, which can be used for reference by other countries, especially countries dominated by agriculture.

At present, the Chinese economy is transiting from the traditional to the new. In the process of implementing the supply-side structural reform, e-commerce plays an important role as the link of industrial integration [1]. Since 2015, the Chinese government has issued several policies to promote the development of rural e-commerce [2]. Encouraged by various measures, e-commerce of agricultural products has developed rapidly and contributed to poverty alleviation. In 2019, the online retail sales of agricultural products in China’s counties was 269.31 billion yuan that is a 28.5% increase compared with the last year. Among them, the total sales amount of 823 poverty-stricken counties is 19.08 billion yuan. In China, the per capita average area of cultivated land is too small and the modernization of agriculture has not had a comprehensive implementation yet. Through the development of e-commerce, the added value of agricultural products has increased, which helps to improve agricultural productivity.

E-commerce is being used increasingly by consumers to buy high value-added agricultural products with geographical indication, green, and organic labels [3]. E-commerce can control product quality by controlling the source, information tracing and customizing [4]. It makes the sale and purchase of agricultural products more convenient and provides new business opportunities for agriculture [5]. During COVID-19, the online food supply chain, which provided convenience for home segregation, reduced unnecessary population gathering and exposure [6,7]. E-commerce for agricultural products helps to build a green environment and promotes the sustainable development of agriculture [8,9].

E-commerce lets farmers obtain market information easier, expanding the sale space of agricultural products. It has changed the status that farmers are forced to accept prices, eased the periodic fluctuation of the price of agricultural products, and helped to sell agricultural products from rural areas to cities [10,11,12]. The development of e-commerce for agricultural products has promoted the expectation price and marginal revenue of pollution-free agricultural products, thus promoting the development of green agriculture [13,14]. Although the development of e-commerce provides more opportunities for the cross-border trade of agricultural products, it does not change the deficit of trade [15,16]. At present, the e-commerce business of agricultural products’ leading enterprises has not ushered in explosive growth yet. With the gradual perfection of e-commerce departments, the enterprise profits will sustain growth [17]. Only with agricultural products as the core, will the sustainable development of rural e-commerce be achieved [18,19].

According to the results of recent research, the expected return of e-commerce will be improved with the increasing freshness of agricultural products. Only safe, high-quality, and trustworthy agricultural products can obtain better prices and higher sales volume through e-commerce. Improving the seller’s reputation and providing high-quality after-sales service can help to enhance the loyalty of consumers in e-commerce [20,21,22,23,24,25]. Agricultural products with unique cultural connotations and the added value of being ecologically friendly are more suitable for the development of e-commerce, so the characteristic agricultural production areas have the obvious advantages in the development of e-commerce [21,26,27]. Increasing the type of packaged products and improving the brand image of agricultural products can be used to enhance the added value of agricultural products [28].

As most farmers tend to avoid risks and in a disadvantageous position in the contractual relationship of e-commerce, government intervention is needed to help its development [19,29,30]. The development of agricultural products e-commerce can be hindered by underdeveloped industry, weak processing capacity of agricultural products, the lack of agricultural products for online sales, remote geographical location, high entry barriers for B2C platforms and difficulties marketing in C2C platforms [31,32]. The over-investment in information technology will not necessarily lead to the development of e-commerce, but the opening of county-level e-government is conducive [33,34].

Moran’s index and the e-commerce development index, analytic hierarchy process, social network analysis, entropy method, spatial econometric model, and data envelopment analysis are widely used in the existing research on e-commerce of agricultural products [35,36,37,38,39,40,41,42,43]. At present, the correlation study of e-commerce focuses on the use of robust models, such as theory of planned behavior (TPB), and statistical analysis is generally completed by the structural equation model [5,44].

Because the lack of official statistical data on agricultural products e-commerce in China, the data used by scholars in correlation study are generally from the Alibaba platform, including the county e-commerce development index released by Alibaba Group in 2015 and the number of shops major managed in the agricultural products sections of Taobao and T-mall platforms in various regions [42,43,45,46]. Therefore, the research on the development of agricultural products e-commerce in China is mostly qualitative analysis, with less quantitative research.

Based on the top 100 counties for agricultural products e-commerce in China, this study makes a quantitative analysis of the development of agricultural products e-commerce in China at the county-level. By drawing the standard deviation ellipses, which depend on the geographic coordinates of the counties, the spatial distribution characteristics of agricultural products e-commerce development in China at the county-level in the past two years were analyzed. Using data from the China County Statistical Yearbook, the spatial econometric models, based on economic gap and geographical gap, have been constructed to work out the influencing factors which can influence the development of agricultural products e-commerce in the county.

2. Materials and Methods

2.1. The Research Methods

2.1.1. Standard Deviational Ellipse

The standard deviation ellipse (SDE) was proposed by Lefever in 1926, is mainly used to show the spatial distribution characteristics of geographical elements, and has been widely used in geography, ecology, economy, and other fields. The SDE quantitatively describes the basic characteristics of geographical elements’ spatial distribution which include the dispersal direction, discrete degree, the distribution of subject area, and other information, through its center, major and minor axes, azimuth, and other parameters [47,48,49]. In this paper, the SDE was used to analyze the spatial distribution of the top 100 counties in China. The SDE is only drawn based on the geographical coordinates of each county, the unweighted method is adopted here to calculate the parameters of SDE. The concrete calculating method is as follows:

Mean central coordinate:

$$SD{E}_x=\frac{1}{n}{\displaystyle \sum }_{i=1}^n{x}_i$$

(1)

$$SD{E}_y=\frac{1}{n}{\displaystyle \sum }_{i=1}^n{y}_i$$

(2)

Azimuth:

$$\tan \theta =\frac{A+B}{C}$$

(3)

$$A={\displaystyle \sum }_{i=1}^n{\tilde{x}}_i^2-{\displaystyle \sum }_{i=1}^n{\tilde{y}}_i^2$$

(4)

$$B=\sqrt{{\left({\displaystyle \sum }_{i=1}^n{\tilde{x}}_i^2-{\displaystyle \sum }_{i=1}^n{\tilde{y}}_i^2\right)}^2-4{\displaystyle \sum }_{i=1}^n{\tilde{x}}_i^2{\tilde{y}}_i^2}$$

(5)

$$C=2{\displaystyle \sum }_{i=1}^n{\tilde{x}}_i{\tilde{y}}_i$$

(6)

X-axis standard deviation:

$${\sigma }_x=\sqrt{\frac{1}{n}{\displaystyle \sum }_{i=1}^n{\left({\tilde{x}}_i\cos \theta -{\tilde{y}}_i\sin \theta \right)}^2}$$

(7)

$${\sigma }_y=\sqrt{\frac{1}{n}{\displaystyle \sum }_{i=1}^n{\left({\tilde{x}}_i\sin \theta -{\tilde{y}}_i\cos \theta \right)}^2}$$

(8)

In the formula $\left(x_i,y_i\right)$ is the geographical coordinates of the research subject; ${\tilde{x}}_i$ and ${\tilde{y}}_i$ are the deviations from the geographical coordinates of the research subject to the average central coordinates, ${\sigma }_x$ and ${\sigma }_y$ are the standard deviations of the X-axes and Y-axes.

2.1.2. Moran’s Index

In this study, Moran’s index was used as a preliminary test of the spatial relationship of agricultural products e-commerce in counties. It is the most popular measure mode of spatial autocorrelation. The value of Moran’s index is between −1 and 1. If Moran’s I > 0 it means that there is positive correlation, and the high value is adjacent to the high value, the higher the value is, the stronger the spatial correlation will be; if Moran’s I < 0 it means that there is negative correlation, and the high value is adjacent to the low value, the smaller the value, the stronger the spatial difference will be; if Moran’s I = 0 it means that there is no spatial autocorrelation [50,51,52]. In this paper, the global Moran’s index was used to test the spatial correlation of the development degree of agricultural products e-commerce among counties. The formula of the global Moran’s index is as follows:

$${\mathrm{Moran}}^{’}\mathrm{s}I=\frac{{{\displaystyle \sum }}_{i=1}^n{{\displaystyle \sum }}_{j=1}^n{W}_{ij}\left(x_i-\overline{x}\right)(x_j-\overline{x})}{s^2{{\displaystyle \sum }}_{i=1}^n{{\displaystyle \sum }}_{j=1}^n{W}_{ij}}$$

(9)

$$S^2=\frac{1}{n}{\displaystyle \sum }_{i=1}^n{\left(x_i-\overline{x}\right)}^2$$

(10)

In the formula W_ij is the element of spatial weight matrix, considering that the development of agricultural products e-commerce in different areas may be affected by both geographical and economic aspects. In this study, the nested weight matrix of spatial economy and geography is constructed by using the geographical distance and the per capita GDP [53]. The formula of the spatial weight matrix is as follows:

$$W{1}_{ij}=\{\begin{array}{cc}0& \left(i=j\right)\\ {\left(\frac{GD{P}_i}{GD{P}_j}\right)}^{\frac{1}{2}}& \left(i\ne j\right)\end{array}$$

(11)

$$W{2}_{ij}=\{\begin{array}{cc}0& \left(i=j\right)\\ \frac{1}{d_{ij}}& \left(i\ne j\right)\end{array}$$

(12)

$$W_{ij}=W{1}_{ij}\times W{2}_{ij}$$

(13)

In the formula GDP_i and GDP_j are the per capita GDP of county i and county j in 2018; d_ij is the distance between the centers of county i and county j [54,55,56].

2.1.3. Spatial Econometric Model

In this study, the spatial effect of the development of agricultural products e-commerce in the county is analyzed by constructing the weighted least squares (WLS), spatial lag model (SLM) and spatial Durbin model (SDM). The SLM model mainly examines the influence of the spatial lag term, while the SDM model examines the influence of the spatial lag term and the dependent of the dependent variable on the explanatory variables in other regions. The general structure of the two models is as follows:

WLS Model:

$$y={\beta }_0+\beta X+\epsilon$$

(14)

SLM Model:

$$y=\rho Wy+\beta X+{\beta }_0+\epsilon$$

(15)

SDM Model:

$$y=\rho Wy+\beta X+\delta WX+{\beta }_0+\epsilon$$

(16)

In the formula, $\rho$ is spatial auto regressive coefficient, ɛ is random error, W is spatial weight matrix, X is explanatory variables, $\delta , \beta$ is coefficient vector [57,58].

2.2. Variable Selection and Data Description

2.2.1. Variable Selection

In this study, the ratio of e-commerce sales volume of agricultural products and the added value of primary industry is used to measure the development degree of agricultural products e-commerce in each county. Considering that agricultural products sold by e-commerce are generally from the crops of the previous year, so the ratio of the retail sales of agricultural products e-commerce in 2019 and the added value of the primary industry in 2018 was used as the dependent variable. Because of the limitations of county data acquisition, the explanatory variables are mainly selected from four aspects that include economic development level, industrial structure, economic vitality, and agricultural conditions [59]. The influencing factors and their descriptions are shown in

Table 1. Variable selection and measurement indicators description.

Primary Indicators	Secondary Indicators	Symbol	Calculation Formula
Economic development level	Per capita GDP (yuan/household)	X1	GDP/household
	Per capita savings (yuan/household)	X2	Savings deposits/household
	Public service level	X3	Public budget expenditure/GDP
	Financial development level	X4	Financial institutions at the end of the year/GDP
Industrial structure	Primary industry development level	X5	Primary industry’s added value/GDP
	Secondary industry development level	X6	Secondary industry’ added value/GDP
	The proportion of employees in secondary industry	X7	Secondary industry’s employees/household
	The proportion of employees in tertiary industry	X8	Tertiary industry’s employees/household
Economic vitality	Industrial development level	X9	The number of industrial enterprises above designated size
	Telecommunications development level	X10	The number of fixed telephone/registered households
Agricultural conditions	The number of characteristic agricultural products	X11	Originated from China’s Agricultural Products business information public service platform
	The agricultural mechanization level (kW/ha)	X12	Total power of agricultural machinery/the crop sown area
	The facility agriculture development level	X13	Facility agricultural area/the crop sown area
	Cultivated land coverage	X14	Crop sown area/administrative area

.

• Economic Development Level

Generally, economic level is positively correlated with transportation development level, science, and education level, human resources level, and other factors restricting the development of agricultural products e-commerce [60,61,62]. The economic level dimension mainly selects four indicators: per capita GDP, per capita savings, public service level, and financial development level. The per capita GDP is calculated by dividing the GDP by the household registered population; the per capita savings are calculated by dividing the balance of savings deposits of residents by the household registered; the level of public service is measured by the proportion of public budget expenditure in regional gross domestic product, which is calculated by dividing public budget expenditure by regional gross domestic product. The level of financial development is measured by the proportion of the loan balance of financial institutions at the end of the year in the GDP, which is calculated by dividing the loan balance of financial institutions at the end of the year by the GDP.

• Economic Development Level Industrial Structure

The industrial structure refers to the internal structure of the economic system, which reflects the economic development stage and growth level, which can be used to reflect the development differences between regions [63,64]. The dimensions of the industrial structure include four indicators: the development level of the primary industry, the development level of the secondary industry, the proportion of employees in the secondary industry, and the proportion of employees in the tertiary industry. The development level of the primary industry is calculated by dividing the primary industry’s added value by the regional GDP; the development level of the secondary industry is calculated by dividing the added value of the secondary industry by the regional GDP; the proportion of employees in the secondary industry is calculated by dividing the number of employees in the secondary industry by the registered population; the proportion of employees in the tertiary industry is calculated by dividing the number of employees in the tertiary industry by the registered population.

• Economic Vitality

The industrial development level will have an impact on the finishing of agricultural products, thus indirectly affecting the development of agricultural products e-commerce. The development of e-commerce is inseparable from the network, so the development level of telecommunications will also have an impact on the development of agricultural products e-commerce. In this study, the development level of industrial and telecommunications are classified as the dimension of economic vitality. The industrial development level is measured by the number of industrial enterprises above a designated size; the telecommunications development level is calculated by dividing the number of fixed telephones by the registered households.

• Agricultural Conditions

Agricultural products are different from industrial products. It has a strong dependence on natural conditions, so the level of agricultural development will also have some influences on the development of agricultural products e-commerce [65]. The agricultural conditions dimension mainly selects four indicators: the number of characteristic agricultural products, the agricultural mechanization level, the facility agriculture development level and the cultivated land coverage. The number of characteristic agricultural products originated from China’s Agricultural Products business information public service platform; the agricultural mechanization level is measured by the ratio of the total power of agricultural machinery to the crop sown area; the facility agriculture development level is measured by the ratio of facility agricultural area to the crop sown area; the cultivated land coverage is measured by the ratio of crop sown area to administrative area.

2.2.2. Data Description

The data of agricultural products e-commerce sales volume in counties used in this study are from the Report on the Development of Nationwide Digital Agriculture and Rural E-commerce in Counties in 2020, which was jointly issued by the Ministry of Agriculture and Rural Information Center and the Institute of International Electronic Commerce Center of China on 30 April 2020. The data are mainly from

Table 2. The results of multicollinearity test between measurement indicators data.

Variable	VIF	1/VIF
X1	6.23	0.1606
X2	5.87	0.170286
X3	4.49	0.222646
X4	4.21	0.237441
X5	3.47	0.288364
X6	3.33	0.30074
X7	2.87	0.348337
X8	2.77	0.360363
X9	2.65	0.37687
X10	2.38	0.420581
X11	2.17	0.459961
X12	1.36	0.737177
X13	1.14	0.877965
X14	1.09	0.917358
Mean VIF	3.15

,

Table 3. The factors contribution rate and eigenvalue after factor analysis.

Factor	Eigenvalue	Variance Contribution (%)	The Contribution Rate of Accumulative Total of Variance (%)
f1	1.682	15.390%	15.390%
f2	1.621	14.830%	30.220%
f3	3.652	33.410%	63.630%
f4	1.323	12.100%	75.730%
f5	1.232	11.270%	87.000%
f6	0.547	5.010%	92.010%
f7	0.428	3.920%	95.930%
f8	0.178	1.630%	97.560%
f9	0.267	2.440%	100.000%

,

Table 4. The load of each factor after rotation.

Variable	f1	f2	f3	f4	f5
X1	0.756	0.003	0.144	0.361	−0.151
X2	0.819	−0.054	−0.032	0.226	−0.079
X3	−0.067	−0.012	0.003	−0.145	0.987
X4	0.499	−0.115	−0.128	0.137	0.011
X5	−0.588	0.012	−0.548	−0.129	0.106
X6	0.071	0.032	0.989	0.106	0.018
X7	0.643	−0.047	0.218	0.581	−0.189
X8	0.951	0.091	0.098	0.121	0.055
X9	0.474	−0.036	0.143	0.821	−0.200
X10	0.740	−0.112	0.043	0.243	−0.056
X11	−0.192	−0.036	−0.094	−0.103	−0.059
X12	0.032	0.784	0.099	−0.064	−0.059
X13	−0.021	0.998	−0.018	−0.004	0.017
X14	−0.076	−0.048	−0.038	−0.075	−0.108

and

Table 5. Variables of factor interpretation.

Symbol	Factor	High Load Variable	Variable Symbol
f1	Economic Development Level	Per capita GDP	X1
		Per capita savings	X2
		Financial development level	X4
		The proportion of employees in secondary industry	X7
		The proportion of employees in tertiary industry	X8
		Telecommunications development level	X10
f2	Agricultural Development Level	The agricultural mechanization level	X12
		The facility agriculture development level	X13
f3	Industrial Structure	Primary industry development level	X5
		Secondary industry development level	X6
f4	Industrial Development Level	Industrial development level	X9
f5	Public Service Level	Public service level	X3

, the top 100 counties of agricultural products in online retail in 2019. The date of characteristic agricultural products are counted from China’s Agricultural Products business information public service platform. The data involved in other indicators came from the China County Statistical Yearbook 2018 (County and City Volume) and the county’s e-government platform.

The average variance inflation factor of the original data is 3.15 more than 1, indicating that there is multicollinearity. The calculated results are shown in Table 2. So, this study planned to use factor analysis to extract common factors.

The KMO value and Bartlett’s ball test were worked out with Stata software. The KMO value of the data is 0.798, which is more than 0.5. The Bartlett’s ball test strongly rejected the original hypothesis, indicating that the data accord with condition of factor analysis. In this study, five main factors were extracted from 14 independent variables by maximum likelihood factor method. The contribution rate and characteristic values of the factors are shown in Table 4. As can be seen from Table 3, there were five factors with an eigenvalue more than 1, and the cumulative contribution rate of variance is 87%. It means that the five factors can represent most of the information of the original data. The component matrix after rotation is shown in Table 4.

According to Table 4, it can be seen that f₁ mainly explains the five variables of per capita GDP, per capita savings, secondary industry scale, tertiary industry scale, and telecommunication development level, which is defined as economic development level factors; f₂ mainly explains the two variables of the level of agricultural mechanization and the facility agriculture development level, which are defined as agricultural development level factors; f₃ mainly explains the two variables of the primary industry development level and the secondary industry development level, which is defined as industrial structure factors; f₄ mainly explains the industrial development level, which is defined as the industrial development level factor; f₅ mainly explains the public service level, which is defined as the public service-level factor. See Table 5 for details. The two indicators of the number of characteristic agricultural products and cultivated land coverage are not reflected in the above five factors. This study tried to delete the two indicators or regression with them as independent factors but did not get a better regression model. Therefore, the subsequent regression calculations are based on the above five factors as the basic data.

Table 6. The component score coefficient matrix.

Variable	f1	f2	f3	f4	f5
	Economic Development Level	Agricultural Development Level	Industrial Structure	Industrial Development Level	Public Service Level
X1	0.246	0.026	0.656	−0.237	0.416
X2	0.039	−0.004	0.754	−0.146	0.345
X3	0.000	0.006	−0.034	0.999	0.000
X4	−0.108	−0.075	0.545	−0.023	0.196
X5	−0.592	0.008	−0.488	0.147	−0.145
X6	1.000	−0.005	−0.016	0.000	0.000
X7	0.320	−0.018	0.648	−0.293	0.485
X8	0.197	0.158	0.968	0.008	0.000
X9	0.262	−0.012	0.511	−0.331	0.706
X10	0.120	−0.054	0.681	−0.116	0.308
X11	−0.123	−0.049	−0.115	−0.030	−0.196
X12	0.124	0.768	−0.041	−0.067	−0.088
X13	0.017	0.995	−0.095	−0.002	0.000
X14	−0.051	−0.056	−0.105	−0.101	−0.077

is the matrix of the factor score coefficient, and then the values of each factor were calculated according to Table 6 and the original data.

3. Results

3.1. The Results of Standard Deviation Ellipse

Based on the geographic coordinates of the top 100 counties of agricultural products e-commerce sales volume which were published in 2018 and 2019 in the Report on the Development of Nationwide Digital Agriculture and Rural E-commerce in Counties, the standard deviation ellipse of the top 100 counties’ distribution in the past two years were drawn by ArcGis. They are shown in Figure 1; and the specific parameters of the standard deviation ellipses are shown in

Table 7. The standard deviation ellipse parameters of county distribution in 2018 and 2019.

Year	Shape_Length	Shape_Area	Center X	Center Y	XStd. Dist	YStd. Dist	Rotation
2018	51.054	190.691	115.591	30.914	9.896	6.134	57.115
2019	50.038	186.069	116.201	31.323	9.545	6.206	58.778

. Because the ellipse was drawn based on geographical coordinates, the values of the perimeter, area, and XY axis length of the ellipse listed in Table 7 only play a role of trend comparison, and there is no practical significance.

In Figure 1, the color of the Chinese map was filled according to the number of the top 100 counties of agricultural products e-commerce in each province in 2019. According to this picture, we can see clearly that most of the top 100 counties are concentrated in the southeast coastal areas, and then spread to the northwest. Xinjiang, Qinghai, and Gansu are connected in the northwest of China, and there are no top 100 agricultural products e-commerce counties. The picture shows that agricultural products e-commerce in the east is developed more than the west.

The standard deviation ellipse in Figure 1 was drawn based on the geographic coordinate of the top 100 counties of agricultural products e-commerce in the past two years. The elliptical area contains 68% of the top 100 counties. It can be seen from Figure 1 that in the past two years, the top 100 counties for agricultural products e-commerce in China have shown a spatial distribution pattern of “northeast–southwest”. The main distribution is in southeast of China, mostly concentrated in coastal provinces. The distribution directions in 2018 and 2019 are basically the same, both extending along the direction of Guangxi–Jiangsu.

According to the data in Table 7, the azimuth of the 2019 ellipse is 1.663° larger than that of 2018. It indicates that compared with 2018, the ellipse in 2019 rotated clockwise by 1.663°, which means that the overall distribution direction also rotated clockwise by 1.663°. The geographical coordinate of the distribution center of the top 100 counties in 2018 is (115.591, 30.914), which is in Luotian County, Huanggang City, Hubei Province. The geographical coordinate of the distribution center of the top 100 counties in 2019 is (116.201, 31.323), which is in Huoshan County, Lu’an City, Anhui Province.

Comparing the standard deviation ellipse parameters of two years, the standard deviation ellipse area of the top 100 counties in 2019 is slightly smaller than that in 2018, indicating that the distribution of the top 100 counties is more intensive. The center and distribution region of the ellipse moved to the northeast in 2019. This shows that the development of agricultural products e-commerce in the northeast of China is stronger than the northwest in 2019. In the past two years, the width of the distribution ellipse, that is, the length of the short axis, is almost the same; but the length, that is, the length of the long axis is slightly shortened. It shows that the direction of the distribution of the top 100 counties for agricultural products e-commerce development in the county is slightly weakened.

3.2. The Results of Moran’s Index

Using a nested weight matrix that includes both economic distance and geographic distance, the global Moran’s index of the development degree of agricultural products e-commerce in each county in 2019 is calculated.

The result shows that: Moran’s $I=-0.124$ p < 0.05, |Z| > 1.96, Moran’s index is significantly less than 0. It shows that there is a negative correlation between the development degree of agricultural products e-commerce in each county, the spatial distribution of the top 100 counties is scattered, and there are spatial differences. The county with a high level of agricultural products e-commerce development and a county with a low level of agricultural products development are adjacent in geographical and economic distance. The counties with close geographical distance and similar per capita GDP have different levels of agricultural products e-commerce development. Moran’s index is significantly less than 0, it means that it has passed the spatial correlation test, and the spatial econometric analysis can be carried out further.

3.3. The Results of Regression Analysis

Taking the ratio of agricultural products e-commerce sales volume of each county to the value added of the primary industry as the dependent variable, the economic development level factor f₁, agricultural development level factor f₂, industrial structure factor f₃, industrial development level factor f₄, and public service-level factor f₅ are the independent variables. WLS, SLM, and SDM are constructed, and the measurement model is as follows:

WLS Model:

$$y_i={\beta }_0+\beta f_i+\epsilon$$

(17)

SLM Model:

$$y_i=\rho {\displaystyle \sum }_{j=1}^n{W}_{ij}{y}_j+\beta f_i+{\beta }_0+\epsilon$$

(18)

SDM Model:

$$y_i=\rho {\displaystyle \sum }_{j=1}^n{W}_{ij}{y}_i+\beta f_i+\delta {\displaystyle \sum }_{j=1}^n{W}_{ij}{f}_i+{\beta }_0+\epsilon$$

(19)

In the formula, i and j represent different counties; W represents the spatial matrix weight; y is the explained variable; f represents the explanatory factors; $\rho$is the spatial autocorrelation regression coefficient; δ and β are the coefficients of the explanatory factor and the explanatory factor spatial lag; ɛ is random error term.

The White statistic of the data is $63.58>{\mathsf{\chi }}_{0.05}^2\left(20\right)=40.0$, strongly rejecting the null hypothesis, indicating that the data has heteroscedasticity. So, the weighted least squares method is used to eliminate the influence of heteroscedasticity. The results of the WLS estimation, the SLM estimation, and the SDM estimation are shown in

Table 8. The results of WLS estimation, SLM estimation and SDM estimation.

Factor	WLS	SLM	SDM
f1	0.251 *	1.057 ***	1.075 ***
f2	−0.014	−0.004	−0.024
f3	0.124 ***	−0.158	−0.321
f4	−0.180 ***	−0.575 ***	−0.364 *
f5	0.385 ***	0.308 **	0.286 ***
w_f1			−0.431
w_f2			−3.465 **
w_f3			3.197 *
w_f4			−1.092 *
w_f5			0.656
C	0.304 ***	0.707 ***	0.656
Rho		−0.517 ***	−1.045 **
Lambda			0.513 *
R-squared	0.667
Prob > chi2		0.000	0.000

Note: Single, double, and triple asterisks (*, **, ***) indicate significance at the 20%, 10%, and 5% level.

.

According to Table 8, three models are obtained as follows:

WLS Model:

$$y=0.251f_1-0.014f_2+0.124f_3-0.180f_4+0.385f_5+0.304+\epsilon$$

(20)

SLM Model:

$$y=-0.517{\displaystyle \sum }_{j=1}^n{W}_{ij}{y}_i+1.057f_1-0.004f_2-0.158f_3-0.575f_4+0.308f_5+0.707+\epsilon$$

(21)

SDM Model:

$$\begin{gathered} y=-1.105{\displaystyle \sum }_{j=1}^n{W}_{ij}{y}_i+1.075f_1-0.024f_2-0.321f_3-0.364f_4+0.286f_5 \\ -0.431w\_f_1-3.465w\_f_2+3.197w\_f_3-1.092w\_f_4+0.656w\_f_5+0.656+\epsilon \end{gathered}$$

(22)

According to the results in Table 8, the estimated results in WLS of industrial structure factor f₃, industrial development level factor f₄, public service-level factor f₅, and constant term regression are all significant at the 5% level; economic development level factor f₁ is significant at the 20% level. In the SLM model, the estimated results of economic development level factor f₁, industrial development level factor f₄, and the constant term are all significant at the 5% level, and the public service-level factor f₅ is significant at the 10% level. In the SDM model, the economic development level factor f₁ and the public service-level factor f₅ are significant at the 5% level, the spatial lag term of the agricultural development level factor w_f₂ is significant at the 10% level, and the spatial lag term of the industrial structure factor w_f₃ and the spatial lag term of industrial development level factor w_f₄ is significant at the 20% level. The spatial autocorrelation coefficient ρ is significant negative at the 100% level.

Comprehensively, economic development level factor f₁, public service-level factor f₅, and the spatial lag of industrial structure factor w_f₃ have a significant positive effect on agricultural products e-commerce. However, industrial development level factor f₄, the spatial lag of agricultural development level factor w_f₂, and the spatial lag of industrial development level factor w_f₄ have a significant negative effect on it.

Table 5 and Table 8 show per capita GDP, per capita savings, financial development level, the proportion of employees in secondary, the proportion of employees in tertiary, telecommunications development level, and public service level are positively correlated with the development of county-level agricultural products e-commerce. Industrial development level is negatively correlated with the development of county-level agricultural products e-commerce. The agricultural mechanization level, the facility agriculture development level, primary industry development level, secondary industry development level, and industrial development level are negatively correlated with the development of county-level agricultural products e-commerce from a spatial perspective. The spatial autocorrelation coefficient $\rho$ is significantly less than 0, indicating that there is a significant negative spatial spillover effect of the county-level agricultural products e-commerce.

4. Discussion

The standard deviation ellipse result shows that the top 100 counties of agricultural products e-commerce in China are mainly concentrated in the cities of the southeast coast, which also happens to be the most economically developed region in China. The regions with developed economy have perfect infrastructure, abundant human resources, and efficient information transmission, which can respond to market changes in a timely manner, so economically developed regions can lead the development of the industry better. The clustering scope of the top 100 counties moved to the northeast, that indicates the rapid development of agricultural e-commerce in northeast of China in 2019. Relying on the advantages of geographical resources, the northeast region is the big granary of China. With the nourishing of the black soil, the crops are developing well [66], so agricultural products are favored by more people. Because of the underdeveloped economy and the late start of e-commerce development, the development of agricultural products e-commerce has been slow in Northeast China.

In recent years, the Chinese government has paid more and more attention to agricultural products e-commerce. The Chinese e-commerce market is increasingly active, and e-commerce outsourcing services have become more and more perfect. Many county-level governments in northeast China cooperate with e-commerce operation companies in the south, which complement each other and promote the e-commerce development of county agricultural products in northeast China [65]. Because of the early start of e-commerce and the perfect supporting technology, the coastal areas have taken the lead in the development of e-commerce. However, with the continuous improvement of e-commerce technology, the development speed of agricultural products e-commerce in northeast China has been steadily improving.

These indicators, which include per capita GDP, per capita savings, financial development level, the proportion of employees in secondary, the proportion of employees in tertiary, telecommunications development level, and public service level are the symbols of regional economic prosperity. E-commerce will help to promote the green and sustainable development of agriculture [67,68]. It has played a positive role in environmental and social development [69]. More and more consumers prefer shopping online, which provides an incentive for farmers to open online sales channels [70,71]. Developed areas pay more attention to the green development of the social, economic, and ecological environment, so agricultural products e-commerce is improving better [72].

The measurement results show that the level of industrial development has a negative impact on the development of county-level agricultural products e-commerce, which is the opposite of the assumption at the beginning of the model design. It was originally assumed that the level of industrial development would affect the finishing of agricultural products, and the development of county-level agricultural products e-commerce by affecting the types of packaged agricultural products. However, the finishing of agricultural products only needs to be equipped with corresponding production equipment, and the general industrial development level can be satisfied. Therefore, the advanced industrial level cannot bring a significant positive impact on county-level agricultural products e-commerce [73].

At the same time, the productivity of industry is significantly higher than that of agriculture. So, counties with high levels of industrial development will focus on developing industry, and resources such as labor and financial would tend to shift toward industry [74]. At present, the grain–labor coupling model of agricultural production in China is in the stage of simultaneous growth of grain and labor [75]. Industrially developed counties have high levels of urbanization, low land utilization, and fewer laborers engaged in agricultural production. The development of county-level agricultural products e-commerce would be inevitable [76,77]. Conversely, counties with backward industrial development will receive more agricultural production support and hope to get rid of poverty through the development of agricultural products e-commerce. Therefore, the industrial development level factor has a negative impact on the development of county-level agricultural products e-commerce.

The agricultural mechanization level, the facility agriculture development level, and primary industry development level represent the agricultural maturity. Secondary industry development level and industrial development level represent the industrial maturity. The regression coefficients of their spatial lag are negative, which shows a negative external effect on the development of county-level agricultural products e-commerce. The Moran’s index is consistent with the spatial autoregressive coefficient, which is significantly less than 0. That is, the development of agricultural products e-commerce of neighboring counties causes reciprocal inhibition, and it shows a negative spatial spillover effect. The development of agricultural products e-commerce will plunder resources in neighboring counties that should be the reason of reciprocal inhibition. Some scholars have studied the spatial characteristics of county retail e-commerce in Zhejiang Province of China and found a similar conclusion [78].

The growth of crops is greatly affected by geographical conditions, so there is a serious homogenization of agricultural products produced in neighboring counties, which also makes competition more intense. Counties with well-known products are more likely to be favored by consumers and get more turnover. Therefore, the development of county-level agricultural products e-commerce shows a negative spatial spillover effect. The distribution of developed counties is scattered. Additionally, the developed counties become growth poles that show a polarization effect. Because the weight matrix used in spatial measurement also includes the economic distance, the Moran’s index, and the spatial autocorrelation coefficient are significantly less than zero, also showing that the development of agricultural products e-commerce is different in counties with similar per capita GDP.

5. Conclusions and Limitations

5.1. Conclusions

According to the results of this paper, the following conclusions can be drawn. First, areas with a developed agricultural foundation can improve the development speed of e-commerce by purchasing mature e-commerce services. Second, advanced e-commerce technology can rapidly improve the development of agricultural products e-commerce in the initial stage, but it is not obvious in the mature stage. Third, economic development, economic vitality, and infrastructure construction have a positive effect on the development of county-level agricultural products e-commerce. However, the industrial development level will have a negative effect on it. Agricultural development level and industrial development level inhibit the development of county-level agricultural products e-commerce in surrounding counties. Fourth, the county-level agricultural products e-commerce has a negative spatial spillover effect, and there are differences in neighboring counties with similar geographical location and per capita GDP. The distribution of the counties which are developed in agricultural products e-commerce is scattered. Additionally, the developed counties become the growth poles in surrounding counties.

To promote the development of agricultural products e-commerce, the government should strengthen the construction of e-commerce demonstration centers and make it a growth pole in the region, through the strategic cooperation with the surrounding counties which are developed in agricultural products e-commerce to weaken the inhibitory effect of neighboring counties. This can be done by improving the finish machining capacity of agricultural products, increasing the categories of commodities on sale, building agricultural products traceability system, upgrading agricultural transportation and preservation technology to segment the market and capture greater market share. Of course, it is also necessary to strengthen the construction of rural infrastructure, increase the reserve of rural e-commerce talent pool, and improve the efficiency of rural logistics. Internet companies should provide streamlined online sales software that can run at high speed on low configuration mobile phones, to make online transactions easier for people with lower education.

With the continuous development of the internet and digital financial technology, new commodity trading models have been formed, and agricultural products are also changing from traditional trading models to electronic trading models. The combination of e-commerce and commodity traceability technology ensures that consumers can easily and quickly obtain safer and more assured food. At the same time, it can also increase the added value of agricultural products and increase the total value of the entire industrial chain of agricultural products. In many developing countries, agriculture has not yet been fully modernized. The development of agricultural products e-commerce can promote the intensive and modern production of agriculture. According to the practical experience of the Chinese people, it is found that the sustainable development of agricultural products e-commerce can not only help the underdeveloped areas out of poverty, but also help to promote the process of local urbanization. It is hoped that this study can provide some references for the sustainable development of county-level agricultural products e-commerce, agricultural modernization, and rural urbanization in regions dominated by agriculture.

5.2. Limitations

This paper studies the development of agricultural products e-commerce based on the county-level, so there are some difficulties in data acquisition. Therefore, this study has the following limitations: first, the model only takes the top 100 agricultural products e-commerce counties in China as the research subject, and fails to include more counties into the research object; second, this article lacks the research on the spatial heterogeneity and evolution of spatial distribution in the development of county-level agricultural products e-commerce; third, because the statistical data at the county level in China is not perfect, this study has not considered the factors that can influence the development of agricultural products e-commerce, such as informatization and transportation.

In future research, it will be necessary to expand the sample data and incorporate into the model factors that influence the development of e-commerce, such as the level of information and supply chains. Additionally, according to the different geographical locations and conditions, it is also a significant direction to study the spatial heterogeneity of agricultural products e-commerce development at the county-level.