Does Service Utilization Improve Members’ Welfare? Evidence from Citrus Cooperatives in China

Abstract

Farmers’ professional cooperatives (hereinafter referred to as “cooperatives”) are an important carrier for small farmers to organically connect to the big market. Under the background of frequent phenomena such as generalization of cooperatives, cooperative alienation, empty shelled cooperatives and involution of cooperatives in China, whether cooperatives can improve members’ welfare has attracted the attention of all walks of life. Service is the essential attribute of cooperatives, and the key to the functioning of cooperatives is the services utilization by members (hereinafter referred to as “service utilization”). Therefore, examining the impact of service utilization on members’ welfare has important theoretical and practical significance, which helps to scientifically evaluate of the current value of cooperatives in China and then objectively respond to questions about the development of cooperatives. Thus, this study takes the micro-survey data of 74 citrus cooperatives and 524 citrus members in Citrus County, Sichuan Province, China as examples, and uses the endogenous switching model to empirically analyze the impact of service utilization on members’ welfare. The results show that the level of service utilization is not high, and only half of the members use cooperative services. However, service utilization can significantly improve members’ welfare. Specifically, the average treatment effect of service utilization on citrus yields, net returns and household income is respectively 285.446 kg/mu, 1290 yuan/mu and 4980 yuan/person. Simultaneously, service utilization can increase citrus yields, net returns and household income by 13.49%, 18.32% and 17.99% on average. Finally, some countermeasures and suggestions are put forward, such as actively publicizing cooperative’s services, guiding members to use cooperative’s services, improving the standardization level of cooperatives and strengthening policy support for cooperatives.

1. Introduction

Agricultural productive services (hereinafter referred to as “service”), also known as agricultural socialized service, are an effective way for small farmers to connect with the big market [1,2,3], and their suppliers mainly include cooperatives, leading enterprises, the government, etc. [4,5]. As one of the most important new agricultural operation subjects in China [6], cooperatives have unique advantages such as specific organizational goals, strong organizational basis of farmers, and membership system integrating owners, promoters and users [7,8,9]. Therefore, cooperatives have become an important part of China’s new socialized agricultural service system. Agricultural countries in the world generally attach great importance to the development of cooperatives. For example, cooperatives in European agriculture trade accounting for 40–60% of the share [10]. The Chinese government also vigorously supports the development of cooperatives, and president Xi Jinping also stressed that China should develop cooperatives in light of local conditions and explore ways to develop more cooperatives. Under the joint efforts of the society from all walks of life, there were more than 2.259 million legally registered cooperatives in China by the end of April 2021, according to official data from the Ministry of Agriculture and Rural Affairs. Furthermore, cooperatives radiated to nearly half of the country’s farmers, with an average of more than 3 cooperatives per village.

However, the development quality of cooperatives is questioned meanwhile, such as generalization of cooperatives [11], cooperative alienation [12], empty shelled cooperatives [13] and involution of cooperatives [14]. Specifically, the generalization of cooperatives means that all kinds of farmers’ organizations are regarded as cooperatives, and these organizations may only have the name of cooperation but have no real cooperation. The cooperative alienation means that the cooperative is controlled by the core members, and without democratic management and scientific distribution, which deviates from the essence of the cooperative. The empty shelled cooperatives mean that the cooperative has no real name and does not provide substantive services. The involution of cooperatives refers to that although cooperatives have the institutional form of cooperative economy, they do not play the due function of cooperatives, and the degree of farmers’ management organization has not changed from loose to close, from low-level to high-level. So, whether the cooperative still improve members’ welfare, which has attracted extensive attention from academia, business circles and political circles in China. Service is the essential attribute of cooperatives [15,16], more importantly, and the key to the functioning of cooperatives lies in the utilization of their services by members. Therefore, it is of great theoretical and practical significance to scientifically evaluate the value of current Chinese cooperatives, and objectively respond to the questioning of the development of cooperatives when investigate the impact of service utilization on members’ welfare.

Cooperative services mainly include agricultural materials, sales, capital, technology, information and other services [17,18,19]. By serving its members, cooperatives enable participants to obtain maximum benefits from cooperative operations [7]. Specifically, cooperative membership is conducive to improving the yield, productivity and quality of dairy farmers’ milk, and thereby increases their competitiveness both in international and domestic markets [20]. Furthermore, dairy cooperative can increase dairy farmers’ milk yield by 40 percent, net return by 38 percent on average [21]. Verhofstadt and Maertens (2015) pointed out that the membership in cooperatives was able to increase farmers’ income and reduces poverty [22]. Moreover, farmers with larger farms and in more remote areas benefit largest. Ma and Abdulai (2016) found that joining cooperatives had a statistically and positive significant effect on farmers’ apple yields, farm net returns and household income. In addition, farmers with small-scale farms can benefit more, while farmers with medium and large farms can benefit less [23]. Similarly, although cooperatives can improve the economic status of members, the treatment effect from agricultural cooperatives is heterogeneous for different members, and the economic benefits produced by cooperatives are significant only for small-scale farms [24]. Hoken and Su (2018) thought that joining a cooperative was conducive to increasing the family income of members with relatively high transaction costs [25]. The research of Mojo et al. (2017) shows that that participation in cooperatives positively affects household income. what’s more, members economically perform significantly better than if they had not been members and nonmembers would have even performed better than members if they had joined cooperatives [26]. Bachke (2019) pointed out that cooperative membership had a positive effect on the value of agricultural production (18%) and on total income (15%) [27]. In addition, cooperatives can increase the productivity of coffee [28].

To sum up, members’ welfare is mainly reflected in the yield, agricultural income and household income. Many studies have examined the impact of joining cooperatives on members’ welfare, but farmers did not necessarily use cooperative services after joining cooperatives [29,30], and the key to the functioning of cooperatives lies in the services utilization. Therefore, if we directly investigate the impact of joining cooperatives on members’ welfare, we cannot scientifically and reasonably evaluate the function and value of cooperatives. Meanwhile, in order to reduce the impact of industrial heterogeneity, many scholars focused on apple cooperatives [31,32,33], rice cooperatives [9,25,34], etc. for empirical research, but there are few achievements in the study of citrus cooperatives [35,36], which is seriously inconsistent with China’s status as the big citrus country. Therefore, this study takes Sichuan citrus cooperatives and citrus members as examples to empirically investigate the impact of service utilization on members’ welfare, which is innovative to a certain extent.

Furthermore, based on the previous relevant research, this study puts forward the following three hypotheses:

Hypotheses 1.

Service utilization is positively related to members’ citrus yields.

Hypotheses 2.

Service utilization is positively related to members’ net returns.

Hypotheses 3.

Service utilization is positively related to members’ household income.

The remaining structure of this study is as follows: The second part is an overview of the development status of citrus industry and cooperatives in Sichuan, China. The third part is data sources and descriptive statistics. The fourth part is model setting. The fifth part is empirical results and discussion and the last part is conclusions and policy suggestions.

2. Citrus Industry and Cooperatives in Sichuan

Citrus is a plant of the Rutaceae family, and its commodity code in the International Convention on the Harmonized Commodity Description and Coding System is 0805. The Food and Agriculture Organization of the United Nations (FAO) divides citrus into oranges, broad-skinned citrus, lemons and limes, grapefruit and other citrus fruits, while China usually divides them into 6 types, namely mandarin, tangerines, oranges, grapefruits, lemons and Kumquat. Sichuan citrus mainly include five categories: mandarin, tangerine, orange, pomelo and lemon. Citrus is the world’s first major fruit, and China’s citrus area and output ranks first in the world. Data shows that in 2019 China’s citrus harvested area was 2,854,159 hectares, accounting for about 22.34% of the world’s total area, and citrus output was 43,539,916 tons, which accounting for about 21.55% of the total global production. Sichuan Province is located in the southwest of China, with no frost damage in winter, slow recovery in spring, long citrus development cycle and tree retention time, and the market time is concentrated from January to May, which has natural climatic advantages and incomparable late maturity advantages in other regions. Furthermore, some citrus varieties in Sichuan can be sold as early as October of that year. Therefore, Sichuan adapts measures to local conditions and vigorously develops late-ripening citrus. The area of citrus orchards in Sichuan was 323.1 thousand hectares, ranking fourth in the country, and the output of citrus was 11.367 million tons, ranking eighth in the country in 2019. In addition, Sichuan Fruit Industry Revitalization Promotion Plan clearly states that by 2022, Sichuan’s late-ripening citrus industry scale, output, market share and brand influence will be the first in China.

Cooperatives are one of the most important new agricultural business entities. Since the promulgation of the Law of the People’s Republic of China on Farmers’ Professional Cooperatives (LPRCFPC), Chinese cooperatives have achieved rapid growth, and more than 2,259,000 cooperatives had been registered in accordance with the LPRCFPC by the end of April 2021, 86 times that of 2007. However, generalization of cooperatives, cooperative alienation, empty shelled cooperatives and involution of cooperatives have aroused widespread concern in society, Therefore, in order to improve the quality and efficiency of cooperatives, the Chinese government has actively introduced policies, such as the Implementation Plan for Jointly Promoting the Quality Improvement of Farmers’ Professional Cooperatives, the Action for Improving the Quality of Farmers’ Professional Cooperatives, and the Special Clean-up Work Plan for Empty Shelled Cooperatives of Farmers’ Professional Cooperatives, etc. What’s more, one of the key points of all these policies is to improve the services of cooperatives. Sichuan Province has vigorously developed various types of cooperatives in accordance with the policy of “Active Development, Gradual Regulation, Strengthening Support, and Improving Quality”. As of the end of 2018, the total number reached 99,553, and 3,821,101 farmer households joined the cooperative, accounting for about 40% of the total number of agricultural households in Sichuan Province.

Due to the low transaction frequency of citrus, and the high asset specificity and transaction uncertainty, based on Williamson’s transaction cost theory, the channel and method suitable for China’s citrus fresh fruit transaction is the intervention of a third-party organization [37], which provides a theoretical basis for the embedding of cooperatives in the citrus industry. In practice, committed to the development of the province’s late-ripening citrus industry, Sichuan has issued a series of planning and layout policies, factor allocation policies, organizational management policies, market system policies, etc. Among them, the organizational management policy emphasizes the development of citrus cooperatives. Thus, citrus cooperatives have developed rapidly in recent years in Sichuan Province. The same is true of other major citrus provinces in China.

3. Data Sources and Descriptive Statistics

The research objects of this study are Sichuan citrus cooperatives and their members. Due to the lack of statistical data specifically for citrus cooperatives, major citrus counties were mainly considered in the selection of sample areas to ensure that there were enough samples of citrus cooperatives. Based on this, the author calculated the average citrus yield in 2018 and 2019 in citrus-producing counties in Sichuan Province, and obtained the ranking of the top ten citrus counties (see

Table 1. List of major citrus counties in Sichuan Province (Unit: ton).

Serial Number	County (District)	2019	2018	Mean
1	Anyue County	471,115	440,857	455,986
2	Renshou County	403,000	386,888	394,944
3	Pujiang County	284,929	263,902	274,415.5
4	Zizhong County	236,427	222,761	229,594
5	Jintang County	182,151	182,718	182,434.5
6	Yanjiang District	179,797	179,021	179,409
7	Jiang’an County	157,006	149,558	153,282
8	Dongpo District	151,486	143,853	147,669.5
9	Rong County	152,973	141,322	147,147.5
10	Danling County	148,185	135,283	141,734
Total	——	2,367,069	2,246,163	——
Proportion	——	52.34%	49.67%

Source: Sichuan Provincial Department of agriculture and rural development.

). Statistics show that in 2019, 130 counties (cities, districts) in Sichuan Province produced citrus with a total output of 4,522,264 tons, while the total output of citrus in the top 10 citrus counties was 2,367,069 tons, accounting for as high as 52.34% in Sichuan Province. Obviously, the top ten citrus counties are representative of the overall situation of Sichuan citrus. Therefore, this study finally chose to conduct research in 10 counties including Anyue County, Renshou County, Pujiang County, Zizhong County, Jintang County, Yanjiang District, Jiang’an County, Dongpo District, Rong County and Danling County. In particular, Jintang County conducted a pre-investigation, and the formal investigation was carried out in the remaining 9 counties (district).

The research is divided into pre-investigation and formal investigation. The research team conducted a pre-investigation in Jintang County, Sichuan Province on 15 July 2020. On the one hand, the research learned and collected relevant information about the citrus industry and cooperatives in Jintang County through discussions. On the other hand, field research was conducted with the assistance of the Bureau of Agriculture and Rural Affairs, focusing on testing the scientific rationality of the questionnaire items. At last, a total of 4 questionnaires for the chairman and 12 questionnaires for members were collected in the pre-investigation. Based on the field feedback from the pre-investigation, the investigation plan and questionnaires were improved, and the final formal investigation plan and questionnaire were formed. The formal investigation was carried out from 20 July to 1 August 2020, and was carried out by two teams in 9 sample counties (districts). Each sample county (district) had a discussion first in Bureau of Agriculture and Rural Affairs and conduct a questionnaire survey in the field. The purpose of the discussion was to understand and obtain county-level citrus industry and cooperatives, and other relevant statistical information and materials, what’s more, to seek research guidance and assistance. The sample selection method of cooperatives adopts a combination of typical sampling and random sampling.

First, the agricultural and rural bureau of the sample county (city) provides a list of citrus cooperatives, and the research team randomly selected 5 to 12 citrus cooperatives, whose director or other council members who have a better understanding of the cooperative operation would be investigated. At the same time, each cooperative randomly investigated 5–15 members. Since the head of the household has a better understanding of the production and operation of the family, in order to ensure the quality of the member sample, the member questionnaire is mainly conducted for the head of the household. Specifically, due to different varieties, the production cycle of citrus generally spans two years, from which the growth is mainly in that year, and sales from October that year to May next year. Therefore, the survey data focuses on the citrus production cycle from 2019 to 2020 in fact. Obviously, this is cross-sectional data. Finally, 90 questionnaires for cooperatives were distributed and 80 were recovered, from which 74 were valid for cooperatives, with an effective rate of 92.50%. 650 questionnaires for members were distributed and 580 were recovered, from which 524 were valid, with an effective rate of 90.34%. The sample distribution is shown in

Table 2. Distribution of valid samples.

Economic Zone	Major Citrus Counties	Cooperatives		Members
		Quantity	Proportion	Quantity	Proportion
Chengdu Plain	Pujiang County	9	12.16%	60	11.45%
	Dongpo District	5	6.76%	60	11.45%
	Renshou County	9	12.16%	46	8.78%
	Danlian County	8	10.81%	59	11.26%
	Anyue County	12	16.22%	71	13.55%
South Sichuan	Zizhong County	9	12.16%	78	14.89%
	Jiang’an District	6	8.11%	50	9.54%
	Yanjiang District	6	8.11%	47	8.97%
Northeast Sichuan	Nanbu County	10	13.51%	53	10.11%
Total		74	100.00%	524	100.00%

.

4. Methodology and Descriptive Analysis

4.1. Econometrics Model

Refer to the Mincer income Equation [38], the members’ welfare Equation is constructed following:

$$Y_i=X_i{\beta }_i+\delta U_i+{\epsilon }_i$$

(1)

The dependent variable $Y_i$ is members’ welfare, which is specifically described by the member’s citrus yields, net returns and household income. $X_i$ is the control variable, including the individual characteristics of members, the characteristics of household management, the basic characteristics of cooperatives and the characteristics of external environment. See

Table 3. Descriptive statistics of variables.

Variables	Definition	Mean	SD
The dependent variable
Citrus yields	Total citrus production divided by planting area (kg/mu 1)	1998.874	566.765
Net returns	Total revenue of citrus minus total cost (ten thousand yuan 2/mu)	0.653	0.327
Household income	Annual household income divided by total number of people (ten thousand yuan)	2.451	1.575
Core independent variable
Service utilization	1 = utilization, 0 = non-utilization	0.500	0.500
Control variables
The individual characteristics of members
Age	Actual value (years)	55.269	9.931
Health status	Five categorical variables, 1–5 in ascending order	3.933	0.788
Education level	Actual years of education (years)	7.435	3.553
Special experience 3	1 = yes, 0= no	0.170	0.376
Mobile phone 4	1 = yes, 0 = no	0.739	0.440
Understanding level of cooperatives	Five categorical variables, 1–5 in ascending order	3.015	1.144
The characteristics of household management
Family population	Actual number(person)	4.160	1.676
Planting area	Actual value (mu)	6.137	4.822
Planting time	Actual years (years)	13.008	9.833
Sales risk	Five categorical variables, 1–5 in ascending order	0.393	0.489
Proportion of citrus income	Citrus income divided by annual household income (%)	62.389	31.226
The basic characteristics of cooperatives
Demonstration grades	1 = non-demonstration, 2 = county demonstration, 3 = municipal demonstration, 4 = provincial demonstration, 5 = national demonstration	2.212	1.314
Voting method in the membership meeting	1 = “one person, one vote”, 0 = other	0.653	0.477
Surplus distribution method	1 = according to trading volume or shares, 0= other	0.456	0.499
The characteristics of external environment
The development level of external service market	Five categorical variables, 1–5 in ascending order	3.292	0.750
The economic region	1 = Chengdu Plain, 2 = Northeast Sichuan, 3 = South Sichuan	1.590	0.856
Instrument variable
Whether the old and new cooperatives can be distinguished	1 = yes; 0 = no	0.727	0.446

Note: ¹ 1 mu = 1/15 hectare; ² yuan is Chinese currency unit ($1 = 6.37 yuan); ³ Special experience refers to that whether the member has ever belonged to communists or village cadres; ⁴ Mobile phone refers to that whether the member has the experience of using mobile phone.

for details.

If the member’s service utilization decision ($U_i$) is random, then the Equation (1) can get an unbiased estimate. However, in reality, the decision for members whether to use cooperative services is not random. Furthermore, Members’ decision whether to use cooperative services may be affected by some unobservable factors, and these unobservable factors may also affect the member’s welfare, which leads to that $U_i$ and ${\epsilon }_i$ are relevant. If Equation (1) is estimated directly, the estimation results may be biased due to the existence of the sample self-selection problem.

In order to solve the problem of sample selection deviation, some scholars tried to use measurement methods such as PSM [39,40,41]. However, PSM can only solve the selection deviation caused by observable variables, while Endogenous Switching Regression (ESR) has been widely used to solve the selection deviation caused by both observable and unobservable variables [42,43,44]. Specifically, the maximum likelihood method is used to estimate the decision and result Equation at the same time in ESR, in which Equation (1) is the result Equation, and the decision Equation is described below.

Refer to the stochastic utility decision model [45], whether members use cooperative services or not depends on the difference between the utility of the use of cooperative services ($U_{1i}$) and the utility of not using cooperative services ($U_{0i}$), that is, if $U_i^{*}=U_{1i}-U_{0i}>0$, the members use cooperative services. Otherwise, members do not use cooperative services. Therefore, this study defines the decision equation of whether members use cooperative services as:

$$U_i^{*}=\Phi \left(Z_i\right)+Z{\mu }_i, \mathrm{if} U_i^{*}>0,U_i=1. \mathrm{Otherwise}, U_i=0$$

(2)

$U_i^{*}$ is the latent variable. $U_i$ = 1 represents members using cooperative services, while $U_i$ = 0 represents members not using cooperatives’ services. $Z_i$ is a vector of exogenous independent variables, including the individual characteristics of members, the characteristics of household management, the basic characteristics of cooperatives and the characteristics of external environment, etc. ${\mu }_i$ is the random disturbance term.

It should be emphasized that, in order to enhance the identifiability of the model, at least one of the independent variables ($X_i$) from the decision Equation (2) is not in the independent variable ($Z_i$) from the welfare Equation (1), that is, there is an instrumental variable, $U_i$ represents the behavioral decision of member i whether to use cooperatives’ services, and ${\epsilon }_i$ is the random disturbance term.

Furthermore, the welfare Equation corresponding to members who use and do not use

$$Y_{iu}=X_{iu}{\beta }_{iu}+{\sigma }_{\mu u}{\lambda }_{iu}+{\epsilon }_{iu}, if U_i=1$$

(3)

$$Y_{in}=X_{in}{\beta }_{in}+{\sigma }_{\mu n}{\lambda }_{in}+{\epsilon }_{in}, if U_i=0$$

(4)

$Y_{iu}$ and $Y_{in}$ represent the benefits of members who use and do not use cooperative services correspondingly. $X_{iu} \mathrm{and} X_{in}$ represent the influencing factors of membership welfare. ${\epsilon }_{iu}$ and ${\epsilon }_{in}$ are the random disturbance term. In order to overcome the problem of sample selection bias, the inverse mills ratios ${\lambda }_{iu}$ and ${\lambda }_{in}$, covariance ${\sigma }_{\mu u}=\mathrm{cov}\left({\mu }_i,{\epsilon }_{iu}\right)$ and ${\sigma }_{\mu n}=\mathrm{cov}\left({\mu }_i,{\epsilon }_{in}\right)$ are introduced. Finally, the maximum likelihood estimation method is used for joint estimation of Equations (3) and (4).

It is necessary to examine the constant term, value of ${\sigma }_{\mu u}$ and its significance from the constant term of memberships’ welfare equation. First of all, if the constant term passes the significance test and is positive, the expected value of the welfare of the members who take advantage of the service of the cooperative is greater, indicating that the use of the cooperative service is beneficial to the improvement of the membership welfare; Secondly, if ${\sigma }_{\mu u}$ passes the significance test, it indicates that the model has a selectivity bias, and the classical OLS regression cannot be estimated effectively. Thirdly, if ${\sigma }_{\mu u}$ < 0, it indicates that the welfare level of the members who use cooperative services choose not to use will be reduced.

The average treatment effect of service utilization on the welfare of the members is obtained by comparing the expected welfare of the members who use and do not use the service of the cooperative under real and virtual conditions.

Expected benefits of members using cooperative services:

$$E[Y_{iu}|U_i=1]=X_{iu}{\beta }_u+{\sigma }_{\mu u}{\lambda }_{iu}$$

(5)

Expected benefits of members who do not take advantage of cooperative services:

$$E[Y_{in}|U_i=0]=X_{in}{\beta }_n+{\sigma }_{\mu n}{\lambda }_{in}$$

(6)

In the counterfactual scenario, the expected benefits of members who use cooperative services without using cooperative services:

$$E[Y_{in}|U_i=1]=X_{iu}{\beta }_n+{\sigma }_{\mu n}{\lambda }_{iu}$$

(7)

Similarly, the expected benefits of the members who do not use cooperative services when they use cooperative service are as follows:

$$E[Y_{iu}|U_i=0]=X_{in}{\beta }_u+{\sigma }_{\mu u}{\lambda }_{in}$$

(8)

By making the difference between Equations (5) and (7), the average treatment effect, hereinafter referred to as ATT, of service utilization on member welfare is:

$$ATT=E\left[Y_{iu}\left|U_i=1\left]-E\right[Y_{in}\right|U_i=1\right]=X_{iu}\left({\beta }_u-{\beta }_n\right)+{\lambda }_{iu}\left({\sigma }_{\mu u}-{\sigma }_{\mu n}\right)$$

(9)

Similarly, the Average Treatment Effect of Untreated, hereinafter referred to as ATU, for members who do not use cooperative services is:

$$ATU=E\left[Y_{in}\left|U_i=0\left]-E\right[Y_{iu}\right|U_i=0\right]=X_{in}\left({\beta }_u-{\beta }_n\right)+{\lambda }_{in}\left({\sigma }_{\mu u}-{\sigma }_{\mu n}\right)$$

(10)

Therefore, the average treatment effect of the treatment group and the control group can be calculated respectively, namely ATT and ATU. Since ATU focus on the effect of samples not affected by intervention, their estimation results have little significance for intervention evaluation [46], so the most important is ATT. Therefore, this study focuses on ATT to examine the impact of service utilization on members’ welfare.

4.2. Descriptive Analysis

The purpose of this study is to explore the impact of service utilization on members’ welfare. The core independent variable is service utilization, that is, whether members use cooperative services or not. In this study, cooperative services include agricultural materials, sales, capital, technology and information. As long as members use one or more services, it means that members use cooperative services. On the contrary, it means that members do not use cooperative services. It can be seen from Table 3 that only half of the members use cooperative services. It shows that members do not actively use cooperative services, which further verifies that joining a cooperative does not mean using cooperative services. Therefore, when examining the function and value of cooperatives, it is more scientific and reasonable to analyze the impact of service consumption on members’ welfare than to analyze the impact of joining cooperatives on members’ welfare.

Members’ welfare is represented by citrus yields, net returns and household income, and the net returns is replaced by the difference between the output value and the variable input cost of citrus per mu. Moreover, the variable input cost mainly includes chemical fertilizer, pesticide, bagging, film, insect killing lamp, irrigation and hired labor, etc. Statistics show that the average yield of citrus is 1998.874 kg per mu, and the average net returns is 6530 yuan per mu, and the average household income is 24,510 yuan per person per year.

Clearly, in addition to the potential impact of the core independent variable service utilization, the members’ welfare may also be affected by the individual characteristics of members, the characteristics of household management, the basic characteristics of cooperatives and the characteristics of external environment [47,48,49], which are treated as control variables. In this study, age, health status, education level, special experience, mobile phone and understanding level of cooperatives are used to characterize the individual characteristics of members. The characteristics of household management is represented by family population, planting area, planting time, proportion of citrus income and sales risk. The basic characteristics of cooperatives is replaced by demonstration grades, voting method in the membership meeting and surplus distribution method. The characteristics of external environment is replaced by the development level of external service market and the economic region of members. Furthermore, whether the old and new cooperatives can be distinguished is instrument variable. The descriptive statistics of each variable are shown in Table 3.

Comparing two groups of members who used and did not use cooperative services, there are significant differences between them in some characteristics, mainly in the following aspects: Compared to members who did not use cooperative services, their citrus yields, net returns and household income were higher for those who took advantage of cooperative services. Moreover, members using cooperative services had small age, better health status, higher level of education, special experience, mobile phone experience, higher understanding degree of cooperatives, larger citrus planting area, longer citrus planting time, relatively high citrus income proportion and with underdeveloped external services market. In addition, they were more likely to distinguish the old and new cooperatives. The detailed data are shown in

Table 4. Mean difference between groups of variables.

Variables	Non-Utilization		Utilization		Differences between Groups
	Mean	SD	Mean	SD	Mean	SD
Citrus yields	1882.462	35.738	2115.286	32.799	232.824 ***	48.507
Net returns	0.580	0.019	0.709	0.021	0.111 ***	0.028
Household income	2.134	0.085	2.768	0.105	0.634 ***	0.135
Age	56.237	0.610	54.302	0.613	−1.935 **	0.864
Health status	3.874	0.047	3.992	0.050	0.118 *	0.069
Education level	6.985	0.223	7.885	0.213	0.901 ***	0.308
Special experience	0.198	0.025	0.141	0.022	0.057 *	0.033
Mobile phone	0.786	0.025	0.690	0.029	0.095 **	0.038
Understanding level of cooperatives	2.798	0.070	3.233	0.069	0.435 ***	0.098
Family population	4.130	0.109	4.191	0.097	4.160	0.073
Planting area	4.896	0.272	7.377	0.303	2.480 ***	0.407
Planting time	12.061	0.581	13.954	0.628	1.893 **	0.856
Sales risk	0.397	0.030	0.389	0.030	−0.008	0.043
Proportion of citrus income	57.233	1.979	67.546	1.827	10.313 ***	2.693
Demonstration grades	2.206	0.081	2.212	0.082	0.011	0.115
Voting method in the membership meeting	0.656	0.029	0.649	0.030	0.008	0.042
Surplus distribution method	0.466	0.031	0.447	0.031	−0.019	0.044
The development level of external service market	3.302	0.047	3.282	0.046	−0.019 **	0.066
Northeast Sichuan	0.088	0.018	0.115	0.020	0.027	0.026
Chengdu	0.641	0.030	0.668	0.641	0.027	0.042
South Sichuan	0.271	0.028	0.218	0.026	−0.053 **	0.038
Whether the old and new cooperatives can be distinguished	0.515	0.031	0.939	0.015	0.424 ***	0.034

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

.

5. Empirical Results

This study empirically analyzes the impact of service utilization on the members’ welfare by ESR model. The results of multicollinearity test between independent variables show that the maximum VIF is 69.49 and the average VIF is 8.09, meeting the requirement that the maximum VIF is greater than 10 and the average VIF is greater than 1, indicating the existence of significant multicollinearity. The detailed results are shown in

Table 5. Results of independent variable multicollinearity test.

Variables	VIF	1/VIF
Service utilization	1.39	0.718
Age	68.60	0.015
Age squared	69.49	0.014
Health status	1.18	0.850
Education level	1.63	0.614
Special experience	1.29	0.776
Mobile phone	1.56	0.642
Understanding level of cooperatives	1.38	0.727
Family population	1.16	0.859
Planting area	1.46	0.687
Planting time	1.36	0.733
Sales risk	1.07	0.938
Proportion of citrus income	1.40	0.715
Demonstration grades	1.18	0.848
Voting method in the membership meeting	1.26	0.796
Surplus distribution method	1.11	0.900
The development level of external service market	1.12	0.893
Northeast Sichuan	1.40	0.712
South Sichuan	1.40	0.716
Whether the old and new cooperatives can be distinguished	1.34	0.745
Mean VIF	8.09

. Earlier in the article, the members’ welfare Equation is defined based on Mincer income Equation, and the typical feature of Mincer income Equation is that it includes age and the square term of age, which leads to multicollinearity. For such problems, learning from the research of relevant scholars [50], although not dealing with it is inaccurate, as long as polynomial regression can get a better prediction model, there is no need to consider too much multicollinearity. Therefore, this study does not accurately process multicollinearity. At the same time, in order to correct the potential heteroscedasticity, Robust estimation is adopted in all empirical models. STATA15.1 software is used for all models, and the results are as follows:

5.1. Applicability Test of the Model

Before analyzing the estimation results of ESR model, it is necessary to test the applicability of using ESR model to investigate the impact of service utilization on members’ citrus yields, net returns and household income. Taking the impact of service utilization on members’ citrus yields as an example, the results of endogeneity test (see

Table 6. Regression results of the impact of service utilization on members’ citrus yields.

Variables	Decision Equation		Non-Utilization		Utilization
	Coefficient	SE	Coefficient	SE	Coefficient	SE
Age	0.039	0.050	−7.027	23.452	−4.482	21.245
Age squared	0.000	0.000	0.036	0.213	0.047	0.196
Health status	0.075	0.085	11.895	39.190	−46.506	34.627
Education level	0.002	0.022	−6.962	10.023	3.851	9.021
Special experience	−0.061	0.181	105.052	91.184	2.235	71.616
Mobile phone	−0.077	0.177	83.070	71.657	174.351 **	80.280
Understanding level of cooperatives	0.066	0.063	−38.423	30.657	−17.272	26.829
Family population	−0.004	0.039	12.673	17.365	16.362	17.219
Planting area	0.055 ***	0.015	−33.328 ***	8.315	−30.673 ***	6.340
Planting time	0.007	0.007	35.615 ***	3.396	29.622 ***	3.016
Sales risk	0.047	0.130	−61.019	59.043	−11.535	54.167
Proportion of citrus income	0.003	0.002	1.643 *	0.991	2.748 ***	1.056
Demonstration grades	−0.027	0.051	3.129	22.877	−17.758	21.280
Voting method in the membership meeting	−0.043	0.144	54.932	65.170	14.787	59.747
Surplus distribution method	0.025	0.131	−44.793	56.399	44.853	56.133
The development level of external service market	0.015	0.087	89.075 **	38.086	23.486	36.278
Northeast Sichuan	0.498 **	0.243	−77.424	122.367	−131.430	99.849
South Sichuan	0.011	0.169	−98.306	72.979	−143.160 **	72.811
Whether the old and new cooperatives can be distinguished 1	1.553 ***	0.158
constant	−3.100 **	1.543	1500.579 **	706.570	1685.549 **	651.970
ln${\sigma }_{\mu u}$	—	—	—	—	6.059 ***	0.073
${\mathsf{\rho }}_{\mu u}$	—	—	—	—	0.570 **	0.169
$ln {\sigma }_{\mu n}$	—	—	6.079 ***	0.045	—	—
${\mathsf{\rho }}_{\mu n}$	—	—	−0.106	0.242	—	—
Wald chi2	194.91 ***
Log likelihood	−4176.823
LR test of indep. eqns.	6.79 ***

Note: *, ** and *** are significant at 10%, 5% and 1% levels respectively; ¹ whether the old and new cooperatives can be distinguished is an instrumental variable, which is related to service utilization, and not related to members’ citrus yields.

) show that the LR test of indep. eqns. is 6.79, and the value of Wald Chi-square test is 194.91, both of which rejects the null hypothesis of mutual independence between decision Equation and welfare Equation at 1% level. In addition, ${\mathsf{\rho }}_{\mu u}$ also passes the significance test at 1% level, indicating that unobserved factors and observable factors affect members’ service utilization decisions and citrus yields at the same time, that is, there is selection bias. If not corrected, biased and inconsistent estimation parameters will be generated [51]. Therefore, it is appropriate to use endogenous transformation model to investigate the impact of service utilization on members’ citrus yields. Similarly, it is also appropriate to use ESR to investigate the impact of service utilization on members’ net returns and household income.

5.2. Analysis of Estimation Results of Decision Equation

As mentioned earlier, ESR model jointly estimates decision Equations and welfare Equations, and decision Equations that represent whether members use cooperative services or not are given in the second columns of Table 6,

Table 7. Regression results of the effect of service utilization on members’ citrus net returns.

Variables	Decision Equation		Non-Utilization		Utilization
	Coefficient	SE	Coefficient	SE	Coefficient	SE
Age	0.044	0.049	−0.005	0.015	0.006	0.017
Age squared	0.000	0.000	0.000	0.000	0.000	0.000
Health status	0.063	0.081	−0.031	0.025	0.018	0.027
Education level	0.003	0.021	0.005	0.006	−0.010	0.007
Special experience	−0.035	0.174	−0.012	0.057	0.084	0.057
Mobile phone	−0.064	0.167	0.007	0.045	0.045	0.062
Understanding level of cooperatives	0.034	0.062	−0.027	0.019	−0.022	0.022
Family population	0.006	0.038	0.005	0.013	0.005	0.013
Planting area	0.054 ***	0.015	−0.010 ***	0.005	−0.010 *	0.005
Planting time	0.010	0.007	0.014 ***	0.002	0.014 ***	0.002
Sales risk	0.039	0.126	0.046	0.043	0.046	0.043
Proportion of citrus income	0.003	0.002	0.003	0.001	0.003 ***	0.001
Demonstration grades	−0.020	0.049	0.015	0.017	0.015	0.017
Voting method in the membership meeting	−0.021	0.137	−0.044	0.047	−0.044	0.047
Surplus distribution method	0.049	0.125	0.039	0.043	0.039	0.043
The development level of external service market	0.017	0.083	0.011	0.029	0.011	0.029
Northeast Sichuan	0.466 **	0.230	0.051	0.077	0.051	0.077
South Sichuan	0.045	0.165	−0.032	0.057	−0.032	0.057
Whether the old and new cooperatives can be distinguished 1	1.329 ***	0.165
constant	−3.033 **	1.488	0.041 **	0.506	0.041	0.506
ln${\sigma }_{\mu u}$	—	—	—	—	−1.017 ***	0.072
${\mathsf{\rho }}_{\mu u}$	—	—	—	—	0.854 ***	0.074
$ln {\sigma }_{\mu n}$	—	—	−1.295 ***	0.044	—	—
${\mathsf{\rho }}_{\mu n}$	—	—	0.020	0.216	—	—
Wald chi2	77.30 ***
Log likelihood	−357.410
LR test of indep. eqns.	7.89 ***

Note: *, ** and *** are significant at 10%, 5% and 1% levels respectively; ¹ whether the old and new cooperatives can be distinguished is an instrumental variable, which is related to service utilization, and not related to members’ citrus net returns.

and

Table 8. Regression results of the effect of service utilization on members’ household income.

Variables	Decision Equation		Non-Utilization		Utilization
	Coefficient	SE	Coefficient	SE	Coefficient	SE
Age	0.036	0.048	−0.037	0.051	−0.124 ***	0.063
Age squared	0.000	0.000	0.000	0.000	0.001 ***	0.001
Health status	0.031	0.080	−0.019	0.084	−0.055	0.103
Education level	−0.001	0.021	−0.002	0.022	−0.013	0.027
Special experience	−0.043	0.171	0.143	0.196	−0.036	0.214
Mobile phone	−0.039	0.167	0.001	0.155	0.210	0.233
Understanding level of cooperatives	0.078	0.061	−0.102	0.067	−0.012	0.080
Family population	−0.048	0.041	−0.292 ***	0.039	−0.511 ***	0.052
Planting area	0.068 ***	0.014	0.209 ***	0.020	0.234 ***	0.018
Planting time	0.008	0.007	0.023 ***	0.007	0.038 ***	0.009
Sales risk	0.027	0.124	−0.241 *	0.127	0.180	0.161
Proportion of citrus income	0.002	0.002	−0.020 ***	0.002	−0.011 ***	0.003
Demonstration grades	0.001	0.049	0.025	0.049	0.103	0.063
Voting method in the membership meeting	−0.027	0.138	0.115	0.140	0.301 *	0.178
Surplus distribution method	0.057	0.123	−0.038	0.121	−0.057	0.163
The development level of external service market	0.011	0.083	0.061	0.082	0.038	0.107
Northeast Sichuan	0.392 *	0.234	−0.138	0.262	0.339	0.295
South Sichuan	0.033	0.162	−0.239	0.158	0.046	0.215
Whether the old and new cooperatives can be distinguished 1	1.187 ***	0.174
constant	−2.633 *	1.484	4.599 ***	1.522	5.936 ***	1.918
ln${\sigma }_{\mu u}$	—	—	—	—	0.312 ***	0.063
${\mathsf{\rho }}_{\mu u}$	—	—	—	—	−0.059 ***	0.048
$ln {\sigma }_{\mu n}$	—	—	0.861	0.058	—	—
${\mathsf{\rho }}_{\mu n}$	—	—	−0.161	0.265	—	—
Wald chi2	315.09 ***
Log likelihood	−1026.830
LR test of indep. eqns.	13.76 ***

Note: * and *** are significant at 10%, 5% and 1% levels respectively; ¹ whether the old and new cooperatives can be distinguished is an instrumental variable, which is related to service utilization, and not related to members’ household income.

. It can be seen that the significant variables affecting whether members use cooperative services are consistent in three models. Specifically, the planting area passes the significance test at the 1% level, and the coefficient is positive, indicating that the larger the citrus planting area is, the more members tend to use cooperative services, which is similar to the research results of Ma et al., who pointed out that the larger the farm size, the more likely farmers were to join cooperatives [23], so the higher the possibility of using cooperative services, because the larger the operating area, the higher the demand for cooperative services. Among the members of three economic zones, only the members of Northeast Sichuan Economic Zone pass the significance test at the level of 5% or 10%, and the coefficient is positive, indicating that the members of Northeast Sichuan economic zone are more inclined to use cooperative services than those of Chengdu Plain Economic Zone. Therefore, regional characteristics have a significant impact on members’ service utilization decision [52].

Whether the old and new cooperatives can be distinguished passes the significance test at the 1% level, and the coefficient is positive, indicating that the members who can distinguish the old and new cooperatives are more likely to use cooperatives’ services, which is consistent with the research conclusion of Ito et al. [24]. For members who can distinguish the old and new cooperatives, they have a better understanding of the advantages of new cooperatives, such as helping farmers decrease agricultural risks, reduce transaction costs, solve information asymmetry and improve market negotiation ability, etc. Therefore, they are more likely to use cooperatives’ services. However, statistics show that members’ understanding level of cooperatives is generally low, with an average of only 3.015 (see Table 3). Furthermore, the understanding level of cooperatives has not passed the significance test at the 10% level. Therefore, it is very important to strengthen the knowledge publicity of cooperatives and promote members’ understanding level of cooperatives.

More importantly, the validity test of the instrumental variable is carried out. the Kleibergen-Paap rk LM statistic is 96.345, and it passes the significance test at the 1% level, indicating that the unidentifiable hypothesis is rejected at the 1% level, which means that the model is identifiable after using instrumental variables. Moreover, the F statistic in first-stage regressions is 16.86, greater than 10, and passes the significance test at the 1% level, indicating that whether the old and new cooperatives can be distinguished is not a weak instrumental variable. In addition, due to space limitations, this study only examines the endogeneity of the core independent variable service utilization. At the same time, the relevant tables are not presented, and interested readers can request them from authors.

5.3. Analysis of Estimation Results of Welfare Equation

Welfare Equations that represent the impact of service utilization on members’ citrus yields, net returns and household income are shown in the fourth and sixth columns of Table 6, Table 7 and Table 8, and the specific analysis is as follows.

5.3.1. Analysis of the Impact of Service Utilization on Members’ Citrus Yields

The effects of planting area, planting time and the proportion of citrus income on citrus yields of the two types of members are consistent. Among them, the planting area has passed the significance test at the 1% level in the welfare Equation of the two types of members, and the coefficient is negative, indicating that the smaller the citrus planting area, the higher the citrus yields of the two types of members, which is in line with Abdulai and Huffman [53]. Obviously, this reflects the law of diminishing marginal product to a certain extent. Citrus planting time in the welfare Equation of the two types of members passes the significance test at the 1% level, and the coefficient is positive, indicating that the longer the citrus planting time, the higher the citrus yields of the two types of members, which is in response to the research of Ortega et al. [28]. Because the longer the planting time, the richer the human capital accumulated by the members, and human capital plays an important role in improving the citrus yields. The proportion of citrus income in the welfare Equation of the two types of members passes the significance test at the level of 10% and 1% respectively, and the coefficient is positive, indicating that the higher the proportion of Citrus income, the higher members’ citrus yields. Because the higher the proportion of citrus income, the higher the specialization level of members’ citrus production, and specialization is conducive to improving efficiency.

Mobile phones, the development level of external service market, southern Sichuan and other variables have a certain heterogeneity in the impact on citrus yields of the two types of members. For members who do not use cooperative services, the development level of the external service market passes the significance test at the 5% level, and the coefficient is positive, indicating that the higher the development level of the external service market, the higher the citrus yields of such members. It can be seen that the external service market is of great significance to the citrus production of members who do not use the cooperative services. Cooperatives and external service market are the core components of the agricultural socialized service system. Therefore, when improving the welfare of members, we should not only strengthen the services of cooperatives, but also pay attention to improving the external service market.

For members who use cooperative services, mobile phone passes the significance test at the 5% level, and the coefficient is positive, indicating that the use of mobile phones helps to improve the citrus yield of such members. Because mobile phone can effectively reduce information asymmetry and promote the promotion of advanced production technology [54]. Southern Sichuan Economic Zone passed the significance test at the level of 5% and the coefficient is negative, indicating that compared with Chengdu Plain Economic Zone, the members of southern Sichuan economic zone are less likely to increase the citrus yields. Obviously, regional characteristics have an important impact on the output of agricultural products of members. Therefore, when investigating the welfare effect of cooperatives, we should pay attention to regional differences. The detailed results are shown in Table 6.

5.3.2. Analysis of the Impact of Service Utilization on Members’ Net Returns

The impact of planting area and planting time on the members’ citrus net returns of the two types of members is consistent. Among them, the citrus planting area has passed the significance test at the level of 1% and 10% respectively in the welfare Equation of the two types of members, and the coefficients are all negative, indicating that the smaller the citrus planting area is, the greater the increase of the members’ citrus net returns is, which is similar to the findings of [23]. Therefore, the production scale of citrus is not as big as possible, and it is necessary to find the optimal scale. The citrus planting time in the welfare Equation of the two types of members passed the significance test at the 1% level, and the coefficient is all positive, indicating that the longer the citrus planting time, the higher the members’ citrus net returns. Generally, planting experience is conducive to improving farmers’ agricultural income, and the longer the planting time, the more experience the members have.

The impact of the proportion of citrus income on members’ citrus net returns is heterogeneous. For members who use cooperative services, the proportion of citrus income passes the significance test at the level of 1%, and the coefficient is positive, indicating that the higher the proportion of Citrus income, the higher the possibility of increasing members’ citrus net returns of such members. Because the higher the proportion of citrus income, the higher the level of specialization of members, and specialization can increase efficiency [21]. However, the proportion of citrus income will not significantly affect the citrus net returns of members who do not use cooperative services. The detailed results are shown in Table 7.

5.3.3. Analysis of the Impact of Service Utilization on Members’ Household Income

Family population, planting area, planting time and the proportion of citrus income have consistent effects on the household income of the two types of members. All four variables pass the significance test at the 1% level, and the coefficients were negative, positive, positive, and negative, indicating that the smaller the family population, the longer the citrus planting time, the larger the citrus planting area, and the lower the proportion of citrus income, the more likely the household income of the two types of members will increase. Generally speaking, the larger the family population, the more the population far away from agriculture, so the household income is lower. The longer the planting time of members, the richer the citrus production experience they have accumulated, which is conducive to improving the yield and quality of citrus and the natural increase of household income. Large scale is conducive to the formation of economies of scale. Therefore, the larger the planting scale of members, the higher their household income, which is consistent with the research conclusions of [55]. Compared with non-agricultural industries, the comparative income from engaging in agriculture is low [56]. Therefore, the lower the proportion of citrus income, the greater the possibility of increasing the household income of the two types of members.

Age, sales risk and the voting method in the membership meeting have a certain heterogeneity on the household income of the two types of members. For members who have not used cooperative services, the sales risk passes the significance test at the 10% level, and the coefficient is negative, indicating that the lower the sales risk, the higher the household income of this type of members. However, the impact of sales risk on the household income of members who use cooperative services has not passed the significance test at the 10% level, which can be seen that the use of cooperative sales services can reduce the sales risk of members to a certain extent [33], and then improve the farmers’ market access ability [27]. For members who have used cooperative services, both the age and the square of the age pass the significance test at the 10% level, and the former coefficient is negative, the latter coefficient is positive, indicating that the household income of this type members and their own age show an “inverted U-shaped” feature. The voting method in the membership meeting passed the significance test at the 10% level, and the coefficient was positive, indicating that the voting method of “one person, one vote” helps to increase the household income of this type members. Therefore, improving the distribution system of cooperatives is very important to improve the welfare of members [57]. In the practice of improving the quality and efficiency of cooperatives, we should focus on improving the distribution system of cooperatives. The detailed results are shown in Table 8.

5.4. Analysis of Treatment Effect

The average treatment effect of service utilization on the citrus yields, net returns and household income of members is calculated respectively by Equation (9), and the detailed results are shown in

Table 9. Average treatment effect of service utilization on members’ welfare.

Members’ Welfare	Utilization	Non-Utilization	ATT	Change (%)
Citrus yields	2115.459 (22.516)	1830.013 (25.070)	285.446 *** (33.697)	13.49
Net returns	0.704 (0.011)	0.575 (0.009)	0.129 *** (0.014)	18.32
Household income	2.769 (0.075)	2.270 (0.068)	0.498 *** (0.101)	17.99

Note: *** is significant at 1% level.

. Taking the average treatment effect of service utilization on members’ citrus yields as an example, for members using cooperative services, the expected value of citrus yields of them is 2115.459 kg, and if they had not used cooperative services, their expected value of the citrus yields should have been 1830.013 kg. Furthermore, it is calculated that the ATT of service utilization on citrus yields is 285.446 kg. ATT passes the significance test at the 1% level, which shows that after controlling the observable and unobservable factors, the citrus yields is increased by 13.49% on average.

Similarly, the ATT of service utilization on members’ net returns is 1290 yuan, and service utilization can increase members’ net returns by 18.32% on average. The ATT of service utilization on members’ household income is 4980 yuan, and service utilization can increase members’ household income by 17.99% on average.

5.5. Robustness Analysis

In this part, OLS is used to estimate the impact of service utilization on members’ citrus yields, net returns and household income, and the results are shown in

Table 10. Robustness test results of the impact of service utilization on members’ welfare.

Members’ Welfare	OLS		ESR
	Coefficient	SE	ATT	SE
Citrus yields	239.045 ***	39.533	285.446 ***	33.697
Net returns	0.131 ***	0.027	0.129 ***	0.014
Household income	0.265 ***	0.100	0.498 ***	0.101

Note: *** is significant at 1% level.

. Taking the effect of service utilization on members’ citrus yields as an example, the results show that the marginal impact of service utilization on members’ citrus yields is 239.045 kg, and the significance test at the 1% level shows that service utilization has a significant positive impact on members’ citrus yields. Compared with ATT calculated by ESR, the influence direction of service utilization in two models are the same, but the result value of OLS is smaller. Obviously, the model estimation results of OLS do not consider the endogeneity of the model caused by observable variables and unobservable variables, so it biases the impact of service utilization on members’ citrus yields to a certain extent. Therefore, OLS estimation can only be regarded as a rough estimation, and the result of ESR is relatively stable.

Similarly, using ESR to estimate the impact of service utilization on members’ net returns and household income is also relatively robust.

6. Conclusions and Discussion

6.1. Conclusions

The purpose of this study is to investigate the impact of service utilization on members’ welfare. Therefore, using the micro survey data of 74 citrus cooperatives and 524 citrus members in the citrus counties (district) in Sichuan Province, the ESR is used to analyze the impact of service utilization on members’ citrus yields, net returns and household income. The main conclusions are as follows:

Firstly, service utilization can significantly improve members’ welfare. Specifically, the average treatment effect of service utilization on members’ citrus yields is 285.446 kg/mu, which can increase members’ citrus yield by 13.49% on average; The average treatment effect of service utilization on members’ net returns is 1290 yuan/mu, which can increase members’ net returns by 18.32% on average; The average treatment effect of service utilization on members’ household income is 4980 yuan/person, which can increase members’ household income by 17.99% on average. Therefore, this study proves the hypothesis 1–3 proposed above, and provides a theoretical basis for objectively responding to the query of the development of Chinese cooperatives.

Secondly, for members’ citrus yields, the planting area significantly reduces the citrus yields of the two types of members, while the citrus planting time and the proportion of citrus income significantly increase the citrus yields of the two types of members. In addition, the development level of the external service market has significantly increased the citrus yields of members who do not use cooperative services, and mobile phones has significantly increased the citrus yields of members who use cooperative services. For members’ net returns, the planting area significantly reduces the net returns of two types of members, while the planting time significantly increases the net returns of two types of members, and the proportion of citrus income only significantly increases the net returns of members using cooperative services. In terms of members’ household income, both the family population and the proportion of citrus income significantly reduce the members’ household income, while the planting area and planting time significantly increased the members’ household income. Moreover, the sales risk significantly reduces the members’ household income who do not use cooperative services, and the “one person, one vote” voting method helps to increase the household income of members who use cooperative services.

Thirdly, the planting area, whether the old and new cooperatives can be distinguished and the Northeast Sichuan Economic Zone will significantly promote the utilization of cooperative services by members.

6.2. Discussion

When evaluating the function and value of cooperatives, most studies directly investigated the impact of joining cooperatives on members’ welfare [21,22,23]. Obviously, this estimation is biased to a certain extent [30], because farmers do not necessarily use cooperatives’ services after joining cooperatives, and the sample data showed that only half of members used citrus cooperatives’ services. In addition, there are also some differences in the degree to which members use cooperative services. For example, some members use one service, and some members use multiple services. As for the same service, some members use it in a higher proportion, while others use less. Service is the essential attribute of cooperatives, and the key to the functioning of cooperatives lies in service utilization. Therefore, this study attempts to evaluate the function and value of cooperatives from the perspective of service utilization, which seems more reasonable. At the same time, this study draws on relevant research results [21] and uses ESR model for estimation, which effectively solves the problem of selective deviation caused by observable variables and unobservable variables. Thus, the model estimation results are more scientific and reasonable. Finally, the results show that service utilization can significantly improve members’ welfare, which responds to the research conclusion of [23] to a certain extent, that is, joining cooperatives can improve members’ welfare.

There are still some limitations, and future research can be improved from following aspects:

(1)

A limitation of this study is the potential impact of COVID-19 on the findings. The COVID-19 broke out in early 2020, which mainly affects the sales of citrus, but with the help of the Chinese government, the plight of cooperatives’ citrus sales has been effectively alleviated. According to the feedback from majority cooperatives’ directors, affected by the sales inertia, there was no significant change in the members who used the cooperative’s sales services before and after the COVID-19. Further analysis can be done using data previous years before the COVID-19 and compared with the results of this study, which can check the robustness of the results.

(2)

This study points out that members do not necessarily use cooperative’s services, combined with previous research, both joining cooperatives and service utilization all can improve members’ welfare to a certain extent. Therefore, how different are the effects between joining cooperatives and service utilization on members’ welfare? Which is worth further study.

(3)

This study takes citrus cooperatives as an example, and there are various types of cooperatives in China. It can be studied by taking rice, apple or other cooperatives as examples to check whether the conclusions of this study are consistent with them.

6.3. Recommendations

Based on the above conclusions and discussions, following policy recommendations are put forward:

Firstly, actively publicizing cooperative’s services. Governments are supposed to firmly stand for the development of cooperatives. In order to eliminate the negative impact of public opinion on cooperatives in China, cooperatives can take LPRCFPC and other materials as carriers to vigorously publicize the functions and values of cooperatives. What’s more, the advantages and characteristics of cooperatives’ agricultural materials, sales, capital, technology and information services shall be publicized, which is conducive to guiding members to have an in-depth understanding of cooperative services.

Secondly, guiding members to use cooperative’s services. Promote members to plant on an appropriate scale. guide members who do not use cooperative services to use cooperative services, especially for members with large citrus planting area and long planting years. Moreover, take the utilization level of cooperative services by members as an important assessment index for the government to evaluate the development of citrus cooperatives.

Thirdly, improving the standardization level of cooperatives. Cooperatives should not only enrich the types of their services, but also improve the quality of their services, timely to meet the diversified needs of members. At the same time, focusing on the management system and distribution system, improve the system construction of cooperatives, and then improve the democratic management level of cooperatives.

Fourthly, strengthening policy support for cooperatives. Integrate multi support policy resources, and increase financial support towards cooperatives. Especially, strengthening policy support for land, roads, irrigation and other infrastructure, actively building a citrus sales platform, innovating financial products and characteristic insurance varieties.