Does the Winner Take All in E-Commerce of Agricultural Products under the Background of Platform Monopoly?

Abstract

This paper explores the impact of e-commerce on profit margin from the perspective of scale using nation-level survey data from China. The results show that e-commerce can increase the profit margin of cooperatives, and that a higher proportion of sales via e-commerce strengthens profitability. Secondly, the effects of agri-e-commerce on cooperative profit margin is highly dependent on young talents with rich experience and high education level; in particular, female leaders have a significantly stronger effect on improving profit margins by using e-commerce than male leaders. Thirdly, the operating scale of cooperatives does not have a direct impact on the profit margins, but a bigger operating scale can significantly enhance the effect of e-commerce on profit margins. Lastly, a higher degree of standardization of cooperatives and products more clearly evidences the role played by e-commerce, including the number of brands and certification. Overall, this research provides a micro-foundation for cooperatives to better incorporate the key role of e-commerce under the background of platform monopoly and sheds light on how the government can formulate relevant policies to better support China’s e-commerce development.

1. Introduction

In 2008, China introduced its first Anti-Monopoly Law; in the same year, China’s agri-e-commerce entered a stage of rapid development, which continues to this day. Currently, regardless of the e-commerce sales platforms of agricultural products or the sellers, China’s agri-e-commerce operation has obvious characteristics of scale and specialization, which also leads to the situation of monopoly operation in the later period.

The e-commerce sales platforms of agricultural products have formed an obvious multi-oligopoly situation. Alibaba has 75% of the market share of agricultural e-commerce in China, followed by JingDong and PinDuoduo with 23.77% and 1.72% (Ministry of Agriculture and Rural Affairs Information Center, 2019). Sellers have extremely limited options to sell agricultural products through e-commerce platforms.

On 1 August 2022, China enacted the latest Anti-Monopoly Law, and the area of platform economy was the focus of this revision. Article 9 of the Anti-Monopoly Law provides that operators shall not use data, algorithms, technology, capital advantages, and platform rules to engage in monopolistic practices prohibited by this law. Article 22 provides that an operator with a dominant market position shall not use data and algorithms, technology, and platform rules to engage in the abuse of dominant market position, as provided in the preceding paragraph.

In fact, the Chinese government has already conducted many anti-monopoly investigations prior to the latest Anti-Monopoly Law amendments, and 2021 was dubbed the “year of anti-monopoly” in China.

In February 2021, the Anti-Monopoly Committee of the State Council of China issued the Anti-Monopoly Guidelines on the Platform Economy in accordance with the Anti-Monopoly Law of the People’s Republic of China, aiming to prevent and stop monopolies in the platform economy, as well as safeguard consumers and the public interest of society. In April of the same year, the domestic e-commerce giant Alibaba Group was administratively punished for its “two-for-one” monopolistic practices, which had widespread social impact.

In 2021, Internet companies accounted for 75.42% of China’s antitrust cases, and well-known Chinese Internet companies such as Alibaba (Hangzhou, China), MeiTuan (Beijing, China), and Tencent (Shenzhen, China) were subject to severe antitrust penalties for using their monopoly positions and interfering with fair competition in the market.

Chinese agricultural e-commerce sellers who want to better promote their products must accept marketing programs from e-commerce platforms, which are paid and, in some cases, very expensive and not friendly to some small-scale operators.

In terms of sellers, larger and more professional sellers have more obvious comparative advantages in agricultural e-commerce sales, and agricultural e-commerce sales do not form a long-tail effect. Just like Google, Amazon, and other Internet giants in the early stage of development, by significantly lowering the threshold of advertising placement, making advertising placement no longer expensive and unattainable, many small advertisers eventually accounted for nearly half of Google’s advertising revenue.

Similarly, by lowering the entry threshold for selling agricultural products through e-commerce, more small-scale operators can break through the geographical limitation through e-commerce and sell their products to a broader market. Eventually, many small-scale farmers accumulate a huge sales scale through the accumulation of small quantities of sales.

With the policy and financial support from the Chinese government, more and more small-scale farmers are trying to sell their agricultural products through e-commerce. In China, the threshold for farmers to sell agricultural products through e-commerce is very low; they only need to spend a small deposit to register a personal store to sell fresh produce (these produces are not industrially processed, which involves food processing permits and some food safety issues).

However, it is difficult for many individually registered stores to receive promotional support from e-commerce platforms, and most operators cannot afford the high marketing costs. For example, when consumers search for apples on e-commerce platforms, priority is given to merchants with “paid sellers” or sellers with larger sales volumes.

Another disadvantage of small-scale operators is that they have higher marginal costs. For example, by signing agreements with logistics companies, large-scale sellers reduce costs by about half or more, and this cost difference is also reflected in multiple sales processes such as packaging and storage.

Small-scale farmers have found an effective way to reduce costs and enhance competitiveness through cooperative management, but the actual effects of cooperative management have yet to be tested.

Cooperative societies are important organizations established by Chinese farmers to conduct cooperative business, and their service scope covers nearly half of Chinese farmers. According to official statistics from the Chinese government, as of 2020, 2.24 million cooperatives have been established in China, with 66.83 million cooperative members served. Of these cooperatives, 268 thousand have established processing, distribution, and marketing entities, and 40 thousand cooperatives sell products through e-commerce; these cooperatives help about 110 million farmers sell their produce through e-commerce [1].

Long before the popularity of e-commerce for agricultural products, cooperatives were an important channel for the sale of agricultural products in China, helping hundreds of millions of farmers sell their produce [2]. Therefore, Chinese farmers have the incentive to make up for their resource endowment deficiency through cooperatives and reduce marginal costs through cooperative management [3]. However, whether cooperative management can improve profitability remains to be tested. Previous studies on scale management in agricultural production provide a reference for our analysis.

China is a typical small-scale peasant economy; the problem of scale management of agricultural production has been widely and deeply studied. Research on the scale economy of agricultural production has been very sufficient. According to the theory of division of labor, division of labor and specialization are the keys to the realization of increasing returns to scale [4]. In the field of agricultural production, there are significant differences in the management objectives, production management methods, and economic benefits of farmers under different operation scales [5]; some scholars also quantitatively measured the impact of scale management on production cost and factor input [6,7], while the relationship between operation scale and labor productivity is also a widely discussed topic [8,9]. Scale management is not only beneficial to the sustainable development of agriculture [10]; it can also improve agricultural production efficiency [11], leading to an increase in farmers’ income [12].

In stark contrast to agricultural production, our current theoretical research and empirical testing of the e-commerce market are insufficient. There is no doubt that e-commerce can increase the income of agricultural product sellers [13,14,15]; this is considered not only a solution to distribution challenges in rural area [16] but also an important factor for the general development of rural areas [17]. E-commerce also has positive externality; it has been proven that e-commerce can increase information availability [18], the possibility of tech-adoption [19], and the welfare of smallholders [20] and consumers [21]. Some classical studies talked about the scale and efficiency of cooperatives but not e-commerce [22,23]; further research is needed on the role of agricultural e-commerce.

It needs to be determined whether large-scale and professional operation of agricultural e-commerce sale improve the market competitiveness of sellers, as concluded by some studies in the field of agricultural production. The answer to the following question has important implications for policymaking: Should there be direct support for small-scale farmers to sell their agricultural products through e-commerce, or support organizations such cooperatives to unite small-scale farmers to help them achieve scale and professional development?

Compared with existing studies, this paper makes three academic contributions.

First, we built a profit function model to analyze how factor inputs in the sales process affect cooperatives’ profit margins. This theoretical framework reveals how resource endowments restrict cooperatives from fully utilizing e-commerce.

Second, we used national-level data to discuss scale operations in agricultural e-commerce, and we extended the research field from agricultural production to sales market in line with previous studies.

Third, we analyzed factors that influence the efficiency of e-commerce through econometric models, which provides some experience for operators who want to use e-commerce to sell agricultural products in the future.

The remainder of the paper is organized as follows: Section 2 explains the theoretical analysis framework; Section 3 presents the data and methodology; Section 4 provides the results and discussion based on scale; Section 5 presents the conclusions and policy recommendations.

2. Theoretical Analysis

In this section, we establish a theory analysis framework based on transaction cost theory [24,25,26]. E-commerce provides a new sales channel for cooperatives, but it also increases cost input. This paper constructs a profit function model to analyze the impact of e-commerce on the quantity and cost of cooperative product sales. We make the following assumptions about the sales process:

Precondition 1: Cooperatives obtain products in two ways: one is through the production and processing of cooperatives; the other is through purchases by members or other channels. Since this study focuses on analyzing the cost–benefit situation of the sales process, the model simplifies the product acquisition process, and uses the unit product price $h$ to represent the average acquisition cost of cooperative products.

Precondition 2: Cooperative products are sold in two ways: one is through e-commerce channels; the other is through traditional channels.

Precondition 3: The price of cooperative products is determined exogenously and is not influenced by factors other than the differences in product characteristics, such as changes in cooperative sales strategies. When cooperatives need to reduce product prices for marketing, this can be regarded as increasing marketing costs on the basis of exogenous prices.

On the basis of the above preconditions, the following research hypotheses are drawn for the theoretical analysis:

Hypothesis 1:

E-commerce sales products are priced higher than traditional sales channels.

Hypothesis 2:

The price per unit factor input of agricultural products sold through e-commerce is higher.

Hypothesis 3:

The expected increase in sales brought about by the marginal cost input of e-commerce sales is greater than that of traditional sales channels.

Assuming that the cooperative needs to sell the total product is $Q$, and that $H$ is the cost of obtaining these products, then $H=Q\times h$. The function $f$ is the cost of the cooperative to sell products. The cost of selling products includes two parts: human capital investment and material capital investment. The amount of products to sell is $Q$, the total human capital needed is $L$, and the total material capital needed is $K$; then,

$$Q=f\left(L,K\right).$$

(1)

Assuming that the number of products sold through e-commerce is $Q_1$, and that the number of products sold through traditional channels is $Q_0$, then $Q=Q_1+Q_0$, where $L_1$, $L_0$, $K_1$, $K_0$ are the human capital and material capital investment of different sales channels. The functions of different sales channels are as follows:

$$Q_1=f_1\left(L_1,K_1\right),$$

(2)

$$Q_0=f_0\left(L_0,K_0\right).$$

(3)

Assuming that the human capital element prices of e-commerce and traditional channels are $w_1$ and $w_0$, the material capital element prices are $r_1$ and $r_0$, and the average selling prices are $P_1$ and $P_0$. Although the profit distribution mechanism of cooperatives and enterprises is different, their ultimate operating goal is to maximize profits. Cooperatives have an incentive to sell products through e-commerce only if the price of agricultural products sold through e-commerce channels is higher than that of traditional channels. Since there are many types of agricultural products in the sample data, we did not add the variable “prices” into the model in the large sample data analysis. However, we compared the prices of different products in the sample cooperatives; the average sales price of e-commerce channels was 64.5% higher than that of other channels, implying that $P_1>P_0$; therefore, Hypothesis 1 is proven. $\left(P_1-h\right)f\left(L_1,K_1\right)$ and $\left(P_0-h\right)f\left(L_0,K_0\right)$ represent the total sales revenue of different sales channels after deducting product acquisition costs; $\left(w_1{L}_1+r_1{K}_1\right)$ and $\left(w_0{L}_0+r_0{K}_0\right)$ represent the total cost of the sales process. The sales profit of different sales channels as a function of human capital and material capital can be expressed as

$${\pi }_1\left(L_1,K_1\right)=\left(P_1-h\right)f_1\left(L_1,K_1\right)-\left(w_1{L}_1+r_1{K}_1\right),$$

(4)

$${\pi }_0\left(L_0,K_0\right)=\left(P_0-h\right)f_0\left(L_0,K_0\right)-\left(w_0{L}_0+r_0{K}_0\right).$$

(5)

Then, the first-order condition for profit maximization of e-commerce sales channels is

$$\frac{\partial \pi }{\partial L_1}=\left(P_1-h\right)\frac{\partial f_1}{\partial L_1}-w_1=0,$$

(6)

$$\frac{\partial \pi }{\partial K_1}=\left(P_1-h\right)\frac{\partial f_1}{\partial K_1}-r_1=0.$$

(7)

When the marginal income increase brought by the factor input portfolio is equal to the factor price, it can lead to the maximization of profit. The sales quantity $Q_1$ corresponding to this factor input portfolio is the e-commerce sales quantity that maximizes profits. Similarly, the first-order conditions for the profit maximization of traditional sales channels are the following:

$$\frac{\partial \pi }{\partial L_0}=\left(P_0-h\right)\frac{\partial f_0}{\partial L_0}-w_0=0,$$

(8)

$$\frac{\partial \pi }{\partial K_0}=\left(P_0-h\right)\frac{\partial f_0}{\partial K_0}-r_0=0.$$

(9)

Generally, selling products through e-commerce has higher requirements for equipment and human capital quality. For example, e-commerce sales requirements for the comprehensive quality requirements of sales personnel are higher; they must understand how to operate the computer and have rich marketing experience. Thus, the unit factor price of e-commerce channels is higher, whereby $w_1>w_0$ and $r_1>r_0$; therefore, Hypothesis 2 is proven. Just as in the previous analysis, e-commerce sales generally provide a higher selling price, whereby $\left(P_1-h\right)>\left(P_0-h\right)$. The operation goal of cooperatives is to maximize profits, and the leader of cooperative is a rational person. Accordingly, if the return on unit factors input is higher by e-commerce, only then will cooperatives have the motivation to use it; therefore, Hypothesis 3 is proven, as expressed in Equations (10) and (11).

$$\frac{\partial f_1}{\partial L_1}>\frac{\partial f_0}{\partial L_0}.$$

(10)

$$\frac{\partial f_1}{\partial K_1}>\frac{\partial f_0}{\partial K_0}.$$

(11)

Consider the combination of the two sales channels together. Suppose that $L_1$ is the human capital investment of the e-commerce channel; then, $L_1=aL$. Suppose that $K_1$ is the material capital investment of the e-commerce channel; then, $K_1=bK$. Suppose that $L_0$ is the human capital investment of the traditional sale channel; then, $L_0=cL$. Suppose that $K_0$ is the material capital investment of the traditional sale channel; then, $K_0=dK$. Furthermore, $a+b=1$, and $c+d=1$.

Assuming that the number of products sold by the cooperative is $Q$, the profit $\pi$ of the cooperative can be expressed as the sum of the profits of the e-commerce and traditional channels.

$$\pi \left(L,K\right)={\pi }_1\left(aL,bK\right)+{\pi }_0\left(cL,dK\right).$$

(12)

Then, the first-order condition for the cooperative to maximize profit is

$$\frac{\partial \pi }{\partial L}=\left(P_1-h\right)a\frac{\partial f_1\left(aL,bK\right)}{\partial L}+\left(P_0-h\right)c\frac{\partial f_0\left(cL,dK\right)}{\partial L}-w_{1}a-w_{0}c,$$

(13)

$$\frac{\partial \pi }{\partial K}=\left(P_1-h\right)b\frac{\partial f_1\left(aL,bK\right)}{\partial K}+\left(P_0-h\right)d\frac{\partial f_0\left(cL,dK\right)}{\partial K}-r_{1}b-r_{0}d.$$

(14)

In theory, there exists a set of factor inputs $X\left(a,b,c,d\right)$ through which cooperatives can maximize profits. Meanwhile, factors inputs in profit equations have a direct influence on most of the variables analyzed later, such as the number of brands, standardized production, and number of employees, which can accurately represent the degree of specialization of cooperatives. Obviously, all these variables rely on human capital ($L$) and material capital ($K$) investment. However, there are uncertainties in the sales market, and the allocation of elements of cooperatives is also restricted by many factors, such as the scarcity of e-commerce talents in many rural areas of China, which seriously limits the development of e-commerce. On the basis of the theoretical analysis, considering Hypotheses 1, 2, and 3, we propose the following research hypotheses and test them using survey data:

Hypothesis 4:

Selling products through e-commerce can improve the profit margins of cooperatives.

Hypothesis 5:

The higher the proportion of e-commerce sales, the higher the cooperative profit margin.

Hypothesis 6:

The higher the degree of specialization of cooperatives, the more obvious the effect of e-commerce on increasing the profit margin of cooperatives.

3. Materials and Methods

3.1. Data Collection

The data in this paper were drawn from a survey of cooperatives by the Ministry of Agriculture and Rural Affairs of the People’s Republic of China. The final sample size was 635. A total of 372 cooperatives used e-commerce while 263 did not, with an e-commerce utilization rate of 58%. The regional distribution of the population is detailed in

Table 1. Regional distribution of sample data.

Eastern Region		Central Region		Western Region
Region	Number of Samples	Region	Number of Samples	Region	Number of Samples
Beijing	7	Shanxi Province	19	Chongqing City	21
Tianjin	15	Inner Mongolia	17	Sichuan Province	43
Hebei Province	21	Anhui Province	30	Guizhou Province	18
Liaoning Province	20	Heilongjiang Province	26	Yunnan Province	19
Shanghai	10	Jilin Province	37	Gansu Province	31
Jiangsu Province	12	Jiangxi Province	20	Shaanxi Province	48
Zhejiang Province	19	Henan Province	28	Qinghai Province	13
Fujian Province	16	Hubei Province	28	Ningxia Hui nationality	14
Shandong Province	25	Hunan Province	24
Guangdong Province	21
Zhuang Nationality in Guangxi	23
Hainan Province	10
Total	199	Total	229	Total	207

.

3.2. Variable Selection and Descriptive Statistical Analysis

Our study used the profit margin as a measure of cooperatives’ profitability. The profit margin was obtained by dividing total profit by total costs; data on total profit and total costs were obtained by means of investigation. On the basis of the theoretical analysis and literature review, control variables were classified according to the characteristics of the cooperative leaders and their experience, as well as the characteristics of the cooperative and their products.

There were four groups of control variables. Firstly, we considered the characteristics of the cooperative leaders, including gender, age, and education level. Secondly, we considered the experiences of cooperative leaders. Most of these variables embody distinctive Chinese characteristics. For example, 37.8% of the cooperative leaders had migrant work experience, and 35.3% of the cooperative leaders had civil service experience, whereby they once worked for a government agency at the village level or above. According to our previous research, these experiences have a significant influence on the management of cooperatives. Furthermore, we also used entrepreneurial training, technology promotion, and management experience as control variables, which can accurately reflect the human capital situation of cooperator leaders. Thirdly, we considered the characteristics of cooperatives, including variables about standards, such as the amount of brand, standard production, computer management system, and make-to-order production, as well as variables about the scale operation, such as the number of employees and operating land area. Lastly, we considered the characteristics of product, including product certification and the degree of product processing.

Table 2. Summary and description of control variables.

Variable Name	Variable Meaning	Variable Assignment	Average Value	Standard Deviation
E-commerce adoption (EA)	Cooperatives that use e-commerce to sell products	Yes = 1; no = 0	0.586	0.493
E-commerce percentage (EP)	The proportion of agricultural products sold by cooperatives through e-commerce in all products (based on sales amount)	Original value	0.355	0.521
Cooperative leader’s characteristics
Gender	Gender of cooperative leaders	Male = 1; female = 0	0.888	0.315
Age	Age of cooperative leaders	Original value	48.903	8.578
Education	Education level of cooperative leaders	Primary school and below = 1; junior high school = 2; senior high school = 3; junior college = 4; undergraduate and above = 5	3.217	0.930
Cooperative leader’s experience
Migrant workers	Decision-makers of cooperatives have experience with migrant work	Yes = 1; no = 0	0.378	0.485
Entrepreneurship training	Cooperative decision-makers have participated in entrepreneurship training	Yes = 1; no = 0	0.729	0.445
Civil servant	Cooperative decision-makers have the experience of civil servants above the village level	Yes = 1; no = 0	0.353	0.478
Agricultural extension worker	Cooperative decision-makers have experience as agricultural extension workers	Yes = 1; no = 0	0.181	0.385
Enterprise manager	Cooperative decision-makers have the experience of enterprise managers	Yes = 1; no = 0	0.354	0.479
Characteristics of cooperatives
Number of brands	Number of brands owned by cooperatives	(Number of pieces)	2.027	1.050
Standardized production	Cooperative production process follows clear production standards and has a supervision mechanism	Yes = 1; no = 0	0.920	0.272
Number of employees	Total number of workers employed by cooperatives, including short-term and temporary workers	(In thousands)	0.231	1.296
Operating land area	Includes land owned by cooperatives and transferred land	(Per 10,000 mu)	0.398	1.831
Contract farming	Signs production order with supermarket and performs production processes according to orders	Yes = 1; no = 0	0.597	0.491
Computerized office	Uses a computer to manage and record daily business data	Yes = 1; no = 0	0.532	0.499
Product features
Pollution-free certification	The product is certified as pollution-free	Yes = 1; no = 0	0.436	0.496
Green food certification	The product is certified as green	Yes = 1; no = 0	0.244	0.430
Organic food certification	The product is certified as organic food	Yes = 1; no = 0	0.110	0.313
Agro-product geographical indications	The products have agro-product geographical indications	Yes = 1; no = 0	0.148	0.355
Preliminarily processed products	Seed removal, purification, classification, sun drying, peeling, or bulk packaging of new agricultural products to provide preliminary market services	Yes = 1; no = 0	0.658	0.475
Deeply processed products	After the initial processing, the products are further processed for the purpose of pursuing greater benefits	Yes = 1; no = 0	0.117	0.321
Product standard level	Certification standard confirming the level of product quality	National standard = 1; industry standard = 2; local standard = 3; enterprise standard = 4; lower standard = 5	2.685	1.645

shows the precise definition and summary of control variables.

Before the econometric analysis, the samples were divided into two groups according to whether the products were sold by e-commerce, and the differences of the control variables between the two groups were investigated. According to the mean t-test results, most of the control variables were significantly different between the two groups.

Table 3. Statistical description of the differences between variable groups.

Variable	E-Commerce Cooperatives		Non-E-Commerce Cooperatives		Mean Difference
	Mean Value	Standard Deviation	Mean Value	Standard Deviation
Characteristics of decision-makers
Gender	0.85	0.36	0.94	0.24	−0.09 ***
Age	48.67	8.29	49.24	8.98	−0.57
Education	3.38	0.88	2.99	0.95	0.38 ***
Decision-maker experience
Migrant workers	0.39	0.49	0.36	0.48	0.04
Entrepreneurship training	0.79	0.41	0.64	0.48	0.15 ***
Teacher	0.05	0.21	0.01	0.11	0.04 **
Civil servant	0.33	0.47	0.39	0.49	−0.06
Agricultural extension worker	0.22	0.41	0.13	0.34	0.09 **
Enterprise manager	0.41	0.49	0.27	0.45	0.14 ***
Characteristics of cooperatives
Number of brands	2.23	1.1	1.74	0.9	0.49 ***
Standardized production	0.95	0.21	0.87	0.33	0.08 ***
Number of employees	0.29	1.49	0.15	0.95	0.14
Operating land area	0.24	0.47	0.62	2.78	0.38 **
Contract farming	0.72	0.45	0.43	0.5	0.29 ***
Computerized office	0.63	0.48	0.4	0.49	0.23 ***
Product features
Pollution-free certification	0.48	0.5	0.37	0.48	0.12 ***
Green food certification	0.3	0.46	0.16	0.37	0.14 ***
Organic food certification	0.13	0.34	0.08	0.27	0.06 **
Certification of geographical indications for agricultural products	0.17	0.38	0.11	0.32	0.06 **
Preliminarily processed products	0.72	0.45	0.57	0.5	0.15 ***
Deeply processed products	0.13	0.34	0.09	0.29	0.04 *
Product standard level	2.45	1.43	3.01	1.86	−0.56 ***

Notes: * significant at the 10% level; ** significant at the 5% level; *** significant at the 1% level.

provides the statistical description information and differences of the two groups of samples.

3.3. Model Selection

(1)

Ordinary Least Square (OLS)

The baseline model used ordinary least squares regression; in Equation (15), $y$ is the profit margin of the cooperatives, $x_1$ denotes the virtual variables for whether cooperatives use e-commerce to sell products, and $Z$ represents the set of control variables.

$$y=f\left(X\right)+\mu ={\beta }_0+{\beta }_1{x}_1+{\beta }_{2}Z+\mu .$$

(15)

Considering that a higher profit margin of the cooperative denotes stronger profitability and, thus, a greater probability of using e-commerce, the explanatory variable and the explained variable may be mutually causal. The number of local express service providers was used as an instrumental variable to solve this problem. To better solve the endogeneity problem, we used the endogenous switching regression model (ESR) to test robustness. Therefore, we simplified the analysis process in OLS regression, and we only show the results of the econometric model.

In Equation (16), $y$ is the profit margin of the cooperatives, $x_2$ is the proportion of agricultural products sold by cooperatives through e-commerce with respect to all products (based on sales amount), and $Z$ represents the set of control variables.

$$y=f\left(X\right)+\mu ={\beta }_0+{\beta }_1{x}_2+{\beta }_{2}Z+\mu .$$

(16)

(2)

Endogenous Switching Regression Model (ESR)

If cooperatives are randomly assigned to use e-commerce for sales, then Equation (15) can accurately estimate the impact of using e-commerce on the cooperative’s profit margin. However, there is a “self-selection” problem in whether to use e-commerce sales, and the decision may be affected by unobservable factors, which in turn have an impact on the cooperative’s profit margin. In addition, there may be some other unobservable factors, which will affect the profit margin of cooperatives and whether to use e-commerce sales. This leads to a bias in the estimate of Equation (15).

The ESR model can incorporate the bias caused by unobservable variables into the model, examine the factors influencing the profit margin of cooperatives sold through e-commerce and cooperatives sold through non-e-commerce separately, which can effectively solve the sample selection bias.

The ESR model also estimates the impact of e-commerce application on the cooperative’s profit margin under two different decisions; it includes a choice equation and an income equation. The estimation of ESR model is divided into two steps. First, the choice equation of whether the cooperative uses e-commerce is estimated through the binary choice model, and then the income determination equation is established to estimate the real impact of using e-commerce on the profit margin of the cooperative.

$$Y_{\mathrm{i}}=\alpha X_{\mathrm{i}}+S_{\mathrm{i}}+e_{\mathrm{i}}.$$

(17)

$$S_{\mathrm{i}}={\beta }_{\mathrm{i}}{Z}_{\mathrm{i}}+{\mu }_{\mathrm{i}}.$$

(18)

In Equation (17), $S_{\mathrm{i}}$ is decision variables of whether cooperatives use e-commerce, where $S_{\mathrm{i}}=1$ indicates that cooperatives use e-commerce. $Y_{\mathrm{i}}$ is the outcome variable for cooperative profitability, $X_{\mathrm{i}}$ is series of observable variables, and $e_{\mathrm{i}}$ is the residual term. In Equation (18), ${\beta }_{\mathrm{i}}$ is the variable vector to be estimated, $Z_{\mathrm{i}}$ is the observable nonrandom vector, ${\mu }_{\mathrm{i}}$ is the residual term.

As in the previous analysis, we used the endogenous switching model to construct counterfactual scenarios to measure the true impact of the use of e-commerce on cooperative profitability. Equations (19) and (20) were used to define the profit rate equation for the cooperative.

$$Y_{1\mathrm{i}}={\lambda }_1{X}_{1\mathrm{i}}+{\epsilon }_{1\mathrm{i}},S_{\mathrm{i}}=1.$$

(19)

$$Y_{2\mathrm{i}}={\lambda }_2{X}_{2\mathrm{i}}+{\epsilon }_{2\mathrm{i}},S_{\mathrm{i}}=1.$$

(20)

Equations (18)–(20) were simultaneously estimated using the full information maximum likelihood method (FIML). The equations below can estimate the conditional expectation of the profit rate under the two scenarios of cooperatives using e-commerce and not using e-commerce to sell agricultural products. When $j=1$, $k=2$, this represents the expected values of the actual situations observed in the sample; when $j=2$, $k=1$, this represents the conditional expectation under the counterfactual condition.

$$E\left(Y_{\mathrm{ji}}|G_{\mathrm{i}}=1,X_{1\mathrm{i}}\right)={\beta }_1{X}_{1\mathrm{i}}+\frac{{\sigma }_1{\rho }_{1}f\left(\alpha Z_{\mathrm{i}}\right)}{F\left(\alpha Z_{\mathrm{i}}\right)}.$$

(21)

$$E\left(Y_{\mathrm{ki}}|G_{\mathrm{i}}=0,X_{2\mathrm{i}}\right)={\beta }_2{X}_{2\mathrm{i}}+\frac{{\sigma }_2{\rho }_{2}f\left(\alpha Z_{\mathrm{i}}\right)}{\left[\left(1-F\alpha Z_{\mathrm{i}}\right)\right]}.$$

(22)

By comparing the expected value under the real condition and the counterfactual condition, the average treatment effect (ATE) of any cooperative using e-commerce is a weighted average of the treated group (ATT) and the untreated group (ATU).

$$\mathrm{ATT}=E\left(Y_{1\mathrm{i}}|G_{\mathrm{i}}=1,X_{1\mathrm{i}}\right)-E\left(Y_{2\mathrm{i}}|G_{\mathrm{i}}=1,X_{2\mathrm{i}}\right).$$

(23)

$$\mathrm{ATU}=E\left(Y_{2\mathrm{i}}|G_{\mathrm{i}}=1,X_{2\mathrm{i}}\right)-E\left(Y_{1\mathrm{i}}|G_{\mathrm{i}}=1,X_{1\mathrm{i}}\right).$$

(24)

4. Results and Discussion

4.1. Baseline Regression

According to the results shown in

Table 4. Baseline regression result.

Var	2SLS (1)	LIMI (2)	OLS (3)	OLS (4)
E-commerce adoption (EA)	1.623 **	1.623 **		0.158 ***
	(0.766)	(0.766)		(0.027)
E-commerce percentage (EP)			0.282 **
			(0.126)
Control	Yes	Yes	Yes	Yes
_cons	1.022 ***	1.022 ***	0.253 ***	0.225 ***
	(0.598)	(0.598)	(0.035)	(0.078)
N	635	635	372	635
Hausman test	5.66 **
DWH test	5.98 **
F-statistic	26.51

Note: Robust errors are in parentheses; *** p < 0.01, ** p < 0.05.

, the Hausman test was significant at the 5% level, and the DWH test was significant at the 5% level, indicating that there is an “endogenous” problem in the application of cooperative e-commerce. The F-statistic was 26.51 (larger than the empirical value of 10); hence, there is no weak instrumental variable problem.

In order to verify the validity of instrumental variable selection, the limited information maximum likelihood method (LIMI) was used for estimation, and the estimated result was significantly positive at the 5% level. The results of the instrumental variable method show that, after solving the endogenous problem, the previous conclusions about the effect of e-commerce on the increase in cooperative profit margins was still robust. Therefore, the Hypothesis 4 is proven.

In Model (4), the OLS model shows that EA can increase the profit margins by 12% (absolute value). In Model (3), where the explanatory variable is replaced with EP, the OLS model shows that every 1% increase in the proportion of e-commerce sales (EP) can increase the cooperative profit margin by 0.282% (absolute value). Therefore, Hypothesis 5 is proven.

4.2. Robustness Test

In order to further verify the accuracy of the research conclusions, the ESR model was used to test robustness. According to the results shown in

Table 5. Results of ESR model.

Group	Decision-Making Stage		ATE
	EA	No-EA	ATT	ATU
EA group	0.444	0.321	0.123 ***
No-EA group	0.417	0.357		0.060 ***

Note: Robust errors are in parentheses; *** p < 0.01.

, e-commerce has significantly increased the profit margin of cooperatives. Under the counterfactual assumption, if the cooperatives that use e-commerce do not use e-commerce, the profit margin would drop from 44.4% to 32.1%, a decrease of 12.3%; if the cooperatives that do not use e-commerce use e-commerce, the profit margin would increase from 35.7% to 41.7%, an increase of 6.0%. This shows that the application of e-commerce has a significant effect on improving the profit margin of cooperatives. Compared with the results of ESR model, the OLS regression slightly exaggerated the effect of e-commerce application, but the bias was still within a reasonable range.

4.3. Discussion about Scale

Samples were grouped by EP as follows: “EP > 10%”, “EP > 20%”, “EP > 30%”, “EP > 40%”, and “EP > 50%”. The results of the OLS model are shown in

Table 6. Impact of EP on profit margin.

Group	>10%	>20%	>30%	>40%	>50%
EP	0.330 *	0.355 **	0.780 *	0.844 **	2.661
	(0.193)	(0.139)	(0.467)	(0.360)	(0.165)
Control	Yes	Yes	Yes	Yes	Yes
p-Value	0.090	0.021	0.085	0.026	0.125
N	242	147	79	45	27

Note: Robust errors are in parentheses; * p < 0.10, ** p < 0.05.

. Expect for the “>50%” group, EP had a significant positive effect of cooperative’s profit margin; this was most likely due to an insufficient sample size for the “>50%” group. In different groups, a higher EP led to a stronger effect of EP on profit margin.

According to coefficients of regression model, as the PE increased, the effect of e-commerce sales on improving profit margins showed a clear upward trend, especially in the “>50%” group, where the p-value was 0.125, indicating that the regression coefficients were basically reliable. This provides a reasonable explanation for the behavior of cooperatives investing heavily in the early stages of e-commerce development to capture market share.

4.4. Discussion about Influencing Factors

In order to further analyze the factors influencing the function of e-commerce, we added the interaction items of each variable with EA into the model for OLS regression estimation. Experience was defined as the experience of cooperative leaders, including work as a migrant worker, enterprise manager, civil servant, etc. Total assets were defined as the scale of fixed assets of cooperatives.

According to the result shown in

Table 7. Factors influencing the effect of E-commerce.

Var	Full Sample	Fresh Agri-Product	Primary Agri-Product	Deep Processing Agri-Product
Gender	2.896	4.589	−3.295	2.682
	(2.690)	(8.897)	(2.996)	(4.031)
Gender × EA	−3.453 *	1.447	−3.961 **	1.478
	(1.858)	(3.154)	(1.548)	(3.179)
Age	0.263	−0.599	0.045	0.896 *
	(0.235)	(0.389)	(0.057)	(0.540)
Age × EA	−0.389 *	−0.298	−0.877	0.376 *
	(0.206)	(0.286)	(0.547)	(0.204)
Education	2.784 **	1.845	2.926	1.661 ***
	(1.084)	(1.748)	(3.850)	(0.309)
Education × EA	1.695 ***	1.141 *	2.002 ***	3.688 ***
	(0.511)	(0.652)	(0.365)	(0.579)
Experience	2.241 *	1.847	3.451 *	3.967 *
	(1.209)	(2.478)	(1.916)	(2.234)
Experience × EA	0.958 *	1.254	2.514	0.425 ***
	(0.536)	(1.114)	(2.155)	(0.158)
Number of brands (NB)	2.265 *	3.368	1.754 *	2.661 *
	(1.307)	(4.974)	(1.034)	(1.374)
NB × EA	2.755 ***	1.263 *	3.145 **	2.882
	(0.857)	(0.672)	(1.259)	(1.956)
Number of certifications (NC)	3.339 ***	1.074 *	2.471 **	4.879 *
	(1.026)	(0.607)	(1.023)	(2.915)
NC × EA	3.525 **	1.789	3.859 ***	3.478
	(1.607)	(1.837)	(0.855)	(4.217)
Total assets	5.771	6.141 *	4.154	6.385
	(4.654)	(3.232)	(6.968)	(5.298)
Total assets × EA	2.147 **	2.014 ***	2.657 **	0.587 *
	(0.855)	(0.410)	(1.236)	(0.328)
Control	Yes	Yes	Yes	Yes
_cons	12.74 ***	9.77 ***	13.67 ***	10.69 ***
	(2.34)	(0.69)	(1.51)	(3.11)
N	635	78	418	121

Note: Robust errors are in parentheses; * p < 0.10, ** p < 0.05, *** p < 0.01.

, the education level and personal experience of cooperative leaders, as well as the number of brands and certifications, all had a significant positive and direct impact on the profit margin of cooperatives. Furthermore, EA could bring additional gains to the influence of these factors on profit margin. In contrast, gender, age, and total assets had a significant positive effect only under the premise of using e-commerce. Therefore, Hypothesis 6 is proven; a detailed analysis is provided below.

(1)

Variables related to the leaders

First, we found gender differences in how e-commerce works, whereby women were proven to play a better role in e-commerce management, especially in the group of primary agri-product selling, as shown in Table 7. A possible reason is that women have more experience in buying products through e-commerce than men. By imitating other people’s e-commerce marketing methods, they can improve e-commerce marketing skills and the application effect of e-commerce. The survey data also support this inference, where the proportion of women buying products through e-commerce was 39.8% higher than that of men.

Second, overall, younger leaders had more advantages, but there were some differences across groups, with elder leaders being better at selling deep processing agri-products by e-commerce.

Third, in line with most previous studies about e-commerce adoption and marketing channel selection of agricultural products [27,28], the impact of education and experience was also positive; in each group and the full sample, a higher education level led to a more obvious effect of e-commerce, and, in the full sample analysis, leaders with richer experience were better at using e-commerce.

(2)

Variables related to the cooperatives

First, similar to most studies, the numbers of brands and certifications could directly increase the cooperatives’ rate of profit [29] and influence the decision of e-commerce adoption [30,31], as well as enhance the effect of e-commerce. Brand and product certifications are important endorsements of the quality of agricultural products, as well as an important basis for product selection in the process of consumer e-commerce purchases. Agricultural products are typically graded according to experience and trust. E-commerce sales prevent consumers from judging product quality through common methods such as touch, which exacerbates the information asymmetry of consumers in product selection. Therefore, brands and certifications, as important standards to measure the standardization degree of products and producers, can effectively reduce consumers’ search costs for product quality information, which is conducive to online sales of products.

Second, in line with previous studies that found no direct correlation between scale and profit [32], the total assets of cooperatives had no significant impact on profit margins. However, for the EA cooperatives, total assets played an important role in increasing profit margins, with evidence that e-commerce can affect the farmland scales of new agricultural business entities (including cooperatives) in China [33].

5. Conclusions and Implications

5.1. Conclusions

On the basis of national-level micro-data of 635 cooperatives in China, this study proved that there are obvious scale advantages in e-commerce sales of agricultural products. This paper differs from most previous studies in that it did not focus on the typical effects of e-commerce, instead putting forward proposals on how to better develop e-commerce of agricultural products.

First, we found huge differences in the profitability of e-commerce for cooperatives of different scales, which implies that joint operation is a feasible pathway for further development. Second, we found some factors that influence the effects of e-commerce on profit rate. Third, according to the analyses about influencing factors, we propound some targeted policy recommendations below, which can increase the efficiency of policymaking and fund use.

5.2. Policy Implication

(1)

Give priority to supporting the development of large-scale business entities

According to the results in Section 4.3, large-scale cooperatives have an obvious advantage in e-commerce management; this conclusion can be extended to other kinds of agricultural business entities. Although the threshold for the use of e-commerce for agricultural products has been continuously lowered, the investment in operation and management is still beyond the affordability of most small-scale operators. On the one hand, capital is profit-seeking, and it is difficult to change the monopoly situation of various e-commerce platforms and e-commerce marketing models in a short period of time. In order to obtain more promotion, necessary marketing investment is indispensable; therefore, large-scale cooperatives have more advantages. On the other hand, most agricultural products are not resistant to bumps and are perishable and deteriorated, which puts forward higher requirements for cold storage facilities. Large-scale operation entities can reduce daily management costs through large-scale operations. Compared with small farmers, they have more cost advantages and competitiveness.

(2)

Strengthen talent support

The discussion about influencing factors on cooperative leaders highlighted the importance of talent support. According to the results, young people with a high level of education and rich experience are essential for the successful e-commerce of agricultural products. However, at present, the leaders of most cooperatives are middle-aged and elderly people who lack the necessary Internet knowledge and skills. Some cooperatives want to recruit young college students to develop e-commerce business, but they cannot provide competitive salaries. Moreover, the development space and environment in rural areas are much worse than those in cities. In the future, necessary policy support should be provided for young talents to work in rural areas, including more competitive wages, affordable housing, and vocational skills training.

(3)

Improve the standardization level of cooperatives

According to the sample data, some cooperatives have an operating income of tens of millions, but most cooperatives operate on a small scale and have a low degree of standardization. Therefore, the quality of the products of these cooperatives is uneven, which limits the sustainable development of e-commerce. The results in Table 7 on brands and certifications show that they not only directly increase the profit rate of cooperatives but also have a lasting positive impact through e-commerce adoption. Even if not for the development of e-commerce, the government should take the lead in certifying cooperatives and products, improving the standardization of products, and improving the supervision system to ensure the social credibility of these certifications.

(4)

Reduce costs through joint operations

Now that we have demonstrated the advantages of scale management, it makes sense to promote it. China has established a complete joint development system for cooperatives, called “Unite Cooperative”, including a federation of cooperatives in a certain field. Therefore, we should take advantage of the organizational system of cooperatives to allow more small-scale cooperatives to unite, thus reducing sales costs and enhancing competitiveness.