Market Participation and Farmers’ Adoption of Green Control Techniques: Evidence from China

Abstract

As a standard clean production technology, the wide use of green control techniques (GCT) helps improve the quality of agricultural products and protect the environment. However, the level of application of GCT by Chinese farmers is not high. The market, as the ultimate place to realize the value of grain and the returns of grain farmers, plays an essential part in promoting GCT. Based on survey data from grain farmers in Sichuan Province, China, this study used a conditional mixed process (CMP) model to examine the effect of farmers’ market participation on their GCT adoption behavior and a mediated effects model to test the impact mechanism. The study further explored the effect of farmers’ market participation capability on their GCT adoption behavior. The results showed that (1) farmers’ market participation could significantly increase the likelihood of adopting GCT. (2) Market participation could influence farmers’ adoption behavior through agricultural income, market information, and green cognition. (3) Further research found that farmers were more likely to adopt GCT if their market participation capability was strong. These findings highlight the fact that developing countries have increased their efforts to encourage farmers to participate in markets and to enhance their market participation capability, thereby facilitating the adoption of GCT by farmers.

1. Introduction

The significant rise in farm production in recent years has relied heavily on applying chemical pesticides [1], particularly in developing nations like China. However, Chinese farmers arbitrarily increase the dosage and frequency of chemical pesticide use and shorten the use interval, leading to low pesticide efficiency in China. The Department of Science, Technology, and Education of the Ministry of Agriculture and Rural Development reported that the pesticide utilization rate of major crops in China in 2023 was 41%, which is much lower than the pesticide utilization rate of wheat, corn, and other grain crops in developed nations. Pesticide misuse has caused severe pollution to China’s air, water, and soil [2,3]. A reduction in chemical pesticide usage is imperative. In 2022, the Chinese government implemented the ‘Action Program for Chemical Pesticide Reduction by 2025’ to address the issues of excessive chemical pesticide usage, improve the quality and security of agricultural products, and promote a sustainable ecological environment. This program intends to promote GCT and implement integrated pest and sustainable management techniques. Although pesticide reduction and use programs have produced significant benefits in the past year, the farmers’ adoption rate of GCT has yet to improve.

GCT, characterized by cleaner production, resource conservation, and environmental friendliness, is considered an adequate replacement for chemical pesticides [4]. It includes physical and biological prevention and control, ecological regulation, scientific application of pesticides, and other measures [5,6]. It has been pointed out that GCT can increase crop yields [7,8], enhance farmers’ incomes [9,10], improve the ecological environment [11,12], and safeguard the health of humans [13,14]. For example, based on survey data from farmers growing bitter gourd in Bangladesh, Rahman and Norton [8] found that GCT could safeguard grain yield while reducing farmers’ dependence on chemical pesticides. How can we promote and increase their adoption?

Academics believe farmers’ adoption of GCT is associated with farm characteristics and external factors such as policies and markets. As far as farm characteristics are concerned, individual characteristics such as gender [15], age [16], education level [17,18], technical training received [19], cognition of GCT [20,21], and whether or not they have membership in a cooperative [22,23] have been found to impact farmers’ GCT adoption significantly. For example, based on survey data, Sharma and Peshin [21] found that green cognition positively impacted the willingness of vegetable farmers in India to embrace integrated pest management (IPM) technology. Furthermore, the geographic environment in which the farm is located [24], farm size [5,25], farm income [26], and number of laborers [27] can also significantly affect the promotion of GCT. For example, based on survey data from six prefecture-level cities in Jiangsu Province, China, Li et al. [26] found that farmers’ agricultural income in previous years promoted GCT adoption. Farmers’ adoption of GCT is associated with external factors, such as the degree of governmental regulation [28,29], the intensity of market regulation [30], and the conditions of market access [31]. Xie et al. [31] surveyed 645 kiwifruit producers in Shaanxi Province, China. They employed the IV-probit model and found that enhancing farmers’ access to more valuable marketplaces stimulated them to adopt environmentally friendly production technology. However, the number of farmers using GCT in China remains low [5]. The low adoption rate of GCT has been attributed to the high cost of technology, the lack of technology promotion, and farmers’ subjective application awareness [32]. However, these explanations do not consider the role of market mechanisms.

As limited rational “economic beings” who seek to maximize household utility, Chinese farmers engage in agricultural production for two purposes [33]. One is to satisfy their demand for agricultural products, and the other is to obtain agricultural income. As far as the goal of obtaining agricultural income is concerned, farmers need to produce in the face of market demand [34,35]. Consumer preference for green agricultural products increases daily with economic growth and rising incomes [36,37]. However, most developing countries have insufficient market mechanisms and information asymmetry problems [38,39]. The green agricultural products farmers produce are difficult to recognize in the market and struggle to gain the corresponding income. This discourages farmers from using green production technologies and producing green agricultural products. Some studies have shown that farmers’ market participation is an inexhaustible driving force for agricultural transformation worldwide [40]. The participation of farmers in market transactions is considered an essential engine for safeguarding grain supply [41], achieving food security [42], and enhancing productivity [40]. Can farmers’ market participation affect their green control technology adoption behavior? What are the underlying mechanisms?

Although there is extensive research on the factors influencing farmers’ adoption of GCT and the significance of engaging in markets, there is still room for improvement and expansion. Firstly, although some scholars have noted that market participation can promote farmers’ adoption behavior for green production technologies, they have not yet thoroughly analyzed the intrinsic mechanisms through which market participation affects farmers’ GCT adoption behavior. Secondly, previous studies have neglected the endogeneity of market participation when examining its effect on the adoption behavior of green production technologies. For example, using an IV-probit model to investigate the relationship between organizational support, market access, and farmers’ use of green production technologies, Xie et al. [31] successfully accounted for endogeneity in organizational backing. However, they neglected the endogeneity problem in market participation. Lastly, previous studies have tended to examine only whether farmers participate in the market as a dichotomous decision-making process, neglecting the capability of farmers to participate in the market.

Based on the above analysis, the main objective of this study is as follows: Firstly, this study used the CMP model to address the endogeneity of market participation. Secondly, this research uses field research data from grain farmers in Sichuan Province, China, in 2023, employing the CMP model to analyze how farmers’ market participation impacts the adoption of GCT and their mechanisms of impact. Lastly, this study examined the effect of farmers’ market participation capability on their GCT adoption behavior. The results of this study aim to serve as a guide for developing nations in creating policies to advance GCT.

The remainder of this paper is structured as follows: Section 2 presents the analytical framework. Section 3 details this study’s research methodology, data sources, and variable selection. Section 4 presents the estimated empirical results. Section 5 discusses the results of this study. Section 6 summarizes the paper and proposes relevant strategies.

2. Analytical Framework

2.1. Characteristics of GCT and Adoption Obstacles

GCT has been proven to provide better economic, ecological, and social benefits in practice [43,44,45]. For example, Gurr and Lu [43], based on four years of rice field experimental data in Thailand, China, and Vietnam, analyzed and found that GCT could reduce insecticide use by 70%, increase rice yield by 5%, and enhance market economic benefits by 7.5%. Liu et al. [44] utilized farmer survey data from Changsha, Hunan Province, China, and discovered that pesticide residues in agricultural products in green control and prevention areas were lower than pesticide residue detection standards, effectively improving agricultural product quality.

Although GCT has favorable economic, ecological, and social benefits, its application requires higher input costs, longer investment cycles, and a higher level of practical operation [46]. As far as Chinese farmers are concerned, most of the obstacles to the adoption of GCT include farmers not having a good mastery of the techniques [32], limited cognition of the impacts of the adoption of GCT [46], a preference for short- versus long-term benefits, and fear that the outputs would not realize high quality and prices [47]. As a result, Chinese farmers prefer traditional chemical pesticides to decrease the risk associated with high inputs. Thus, whether GCT can improve the actual economic benefits to farmers, help them to obtain more market price information, and enhance farmers’ levels of green cognition are the keys to promoting the application of GCT.

2.2. Obstacles to the Mechanisms of Market Participation Promoting Farmers’ Adoption of GCT

Farmers’ market participation in the early stages can effectively help them overcome this dilemma and actively adopt GCT. Participating in farmers’ markets can lead to increased income from agriculture [48,49] and provide valuable information on market preferences and sales channels for green agricultural products [50,51]. This can be further interpreted as increased income, unblocking information on market prices, and an enhanced cognitive value of green production.

The first is the income mechanism. Farmers’ involvement in markets in the early stages can increase agricultural income and provide financial capital for agricultural production in the next stage. The high initial input and long payback period of GCT have become important factors that hinder farmers from adopting them [46]. Instead, farmers can earn agricultural income by selling agricultural products to the market. Higher household income can be used to improve agricultural production practices, reduce the financial limitations on farmers’ adoption of GCT, and ultimately increase farmers’ likelihood of adopting GCT.

The second is the information mechanism. Involvement in market transactions can assist farmers in reducing the danger of encountering difficulties in realizing the worth of environmentally friendly agricultural products due to asymmetrical information access by enhancing the accessibility of market price information. Farmers in developing countries have fewer information channels and are disadvantaged [52]. Simultaneously, producers and buyers have a more significant disparity in information on environmentally friendly agricultural goods [38,39]. When farmers cannot effectively grasp the price information of agricultural products, they are more inclined to use traditional chemical pesticides at lower costs to reduce the risk associated with high inputs. According to the existing literature [51], farmers’ market participation in the early stage allows them to have more market information in negotiating with other market participants, strengthening their confidence in selling. When farmers can sell their agricultural products at a premium because of their high quality, they are more likely to adjust their production practices to improve product quality. Concerning pest control techniques, they were more inclined to use GCT.

The third mechanism is cognitive. Market participation increases income expectations of GCT by enhancing the value cognition of green production. According to Chayanov [53], farmers are finite rational “economic beings”. Farmers tend to reduce their input costs in case of uncertainty in the sales price to improve the net income of agricultural operations. Income expectations influence farmers’ production inputs. Farmers’ market participation in the early stage enhances farmers’ knowledge of the prospects of selling green agricultural products. It increases their positive expectations of using green production techniques to gain income, thus increasing the likelihood of farmers adopting GCT [54,55]. In particular, farmers are more likely to believe that adopting ecologically sustainable techniques would increase the market value of agricultural products, mainly because they have achieved high quality and prices when selling their products in the early stages. This would strengthen farmers’ confidence in using GCT and increase their income expectations from using GCT [56]. Driven by higher selling prices, farmers choose to adopt GCT actively [57].

In conclusion, this section reveals the mechanism through which market participation promotes farmers’ GCT adoption behavior by gaining agricultural income, unblocking market price information, and enhancing the cognitive value of green production. Based on the above discussion, this research proposes the following four hypotheses.

H1.

Farmers’ market participation positively impacts their GCT adoption behavior.

H1a.

Farmers’ market participation promotes GCT adoption by increasing agricultural income.

H1b.

Farmers’ market participation promotes GCT adoption by unblocking market information.

H1c.

Farmers’ market participation promotes GCT adoption by enhancing their value cognition of green production.

Figure 1 illustrates how farmers’ market participation influences farmers’ adoption behavior based on the theoretical analysis provided earlier.

3. Materials and Methods

3.1. Research Methodology

3.1.1. Empirical Specifications

According to the theory of farmer behavior [53], as limited rational “economic beings”, farmers seek to maximize household utility. When making decisions about market participation, farmers maximize household utility as the basis for decision-making. Taking grain farmers as an example, they choose to participate in the market when the household utility gained from selling part or all of the grain they produce to the market is greater than the household utility gained from using all the grain made for their consumption. Referring to Li et al. [58], this study assumes that grain farmers can gain utility from selling their grain and not selling their grain, referred to as $U_s^{*}$ and $U_n^{*}$. Additionally, ${ M}_i^{*}$ refers to the anticipated difference in the net gain between participating and not participating in the market. If a grain farmer decides to participate in market transactions, the expected net gain between grain farmer participation and no market is greater than zero (i.e., $M_i^{*}$ $=U_s^{*}$ − $U_n^{*}$ > 0). Although $M_i^{*}$ may not be directly observable, it can still be represented by a latent variable function, as previously described.

$$M_i^{*}=\mathsf{\alpha }{X}_n+{\epsilon }_i,M_i=\left\{\begin{array}{c}1,if M_i^{*}>0\\ 0,if M_i^{*} \le 0\end{array}\right.$$

(1)

The variable ${ M}_i^{*}$ represents a latent variable indicating a farmer i’s propensity to participate in the market. Although direct observation of ${ M}_i^{*}$ is not possible, it can be approximated by $M_i$. More precisely, if the farmer i sold grains to the market in 2022, $M_i$ is equal to 1, otherwise 0. $X_n$ depicts an array of chosen control factors, including age, gender, and cooperative membership. The symbol $\mathsf{\alpha }$ represents the parameter that needs to be estimated. The symbol ${\epsilon }_i$ represents the error term.

In this study, physical pest control techniques and biological pest control techniques were selected to represent GCT for measurement. In 2023, if grain farmers implemented any of the above methods, it is equal to one; otherwise, it will be zero. Therefore, the correlation between farmers’ market participation and GCT adoption behavior can be mathematically represented by the subsequent equation:

$$P\left(Y_i=1\right)=P\left(Y_i^{*}>0\right)=\mathrm{\varnothing }({\beta }_0+{\beta }_1{M}_i+\sum _{n=1}^N{\beta }_i{X}_n+{\mu }_i)$$

(2)

where P indicates farmers’ probability of adopting GCT and $Y_i$ is a latent variable denoting the GCT adoption behavior of farmer i. $M_i$ and $X_n$ are defined above. The symbols ${\beta }_0$ and ${\beta }_1$ represent the parameter that needs to be estimated. The symbol ${\mu }_i$ represents the error term.

3.1.2. Model Selection

Drawing on the analytical approaches of different authors on behavioral decision-making for GCT adoption [8,31], this research uses the probit model as the baseline regression model to estimate Equation (2).

This study’s empirical design aims to tackle endogeneity problems when estimating how farmers’ market participation impacts GCT adoption behavior. This endogeneity problem originates from two sources. One is an omitted variable. Besides the factors noted (such as age, gender, and cooperative membership), there are also unseen factors (such as farmers’ management abilities and drive) that impact farmers’ participation in markets and their adoption of GCT. Unfortunately, empirical analyses do not account for these hidden variables. Another important consideration is the concept of reverse causality or simultaneity. While this study assumes that farmers’ involvement in the market determines their decision to adopt GCT, it is plausible that farmers with higher levels of green cognition would prioritize the adoption of GCT and commercialize their green agricultural products through market participation. Failure to adequately consider these problems may result in biased estimations of the correlation between market participation and GCT adoption behavior.

We can use ESP, CMP, and other comparable methods to estimate the impact of market participation on GCT adoption. This study references the research of He et al. [59] and Zhou et al. [60] and utilizes Roodman’s [61] CMP model for estimation purposes to estimate the impact of market participation on GCT adoption. The analysis in this study was conducted using Stata 17.0.

3.1.3. Conditional Mixed Process Model

The CMP model predicts Equations (1) and (2) simultaneously and uses a maximum likelihood (ML) estimate to quantify the unbiased impact of market participation on the adoption behavior of GCT, according to Roodman [61]. This strategy has the advantage of being able to meticulously select the most suitable model based on the features of the variables. Specifically, the CMP model is segmented into two main stages. The initial stage involves identifying external or instrumental factors and calculating the correlations with the external variables. In the subsequent stage, the outcomes of the initial step are included in the model to estimate the combined probability. Meanwhile, the ML estimator produces a correlation coefficient ${\rho }_{\mu \epsilon }$ (i.e. atanhrho_12) between the error terms of the two equations. If ${\rho }_{\mu \epsilon }$ is significant, it will indicate an endogeneity problem with the market participation variable, proving the validity of the estimation of the CMP model [61].

It was necessary to select a valid instrumental variable to ensure the validity of the CMP model. An instrumental variable is considered valid if it is connected to the endogenous variable and has no connection to the dependent variable [62]. For this reason, utilizing previous research and peer group effects theory, this research employed the ‘average value of market participation of other farmers in the village where the respondent farmer lives’ as an instrumental variable. This ‘average value of market participation of other farmers in the village where the respondent farmer lives’ was chosen for the following reasons: First, the market involvement of other farmers from the village in which the interviewed farmers live affected the market participation of the interviewed farmers. Typically, the higher the average value of market participation of other farmers in the village where the interviewed farmer lives, the higher the likelihood that the interviewed farmer will participate in the market. Hence, it is anticipated that the average market participation value of other farmers will have a positive connection with the respondents’ market involvement. In other words, the selected instrumental variable satisfies the connection assumption. Furthermore, the average market participation value of other farmers does not directly influence respondents’ adoption of GCT. Thus, the selected ‘average value of market participation of other farmers in the village where the respondent farmer lives’ satisfied the assumption of homogeneity. This research refers to Adhvaryu [63], and Li [58] used the two-stage least squares method to verify the statistical validity of the instrumental variable.

The results presented in

Table 1. Results of instrumental variable tests.

Variable Name	Two-Stage Least Squares Method
Market participation in the GCT adoption behavior	0.367 *** (0.106)
Instrument variable’s impact on market participation	0.699 *** (0.057)
Control variable	Yes
Phase I F-value	151.51 ***
Observations	819

Note: *** p <0.01; standard errors in parentheses.

suggest that, in the first stage, the coefficient of regression of the instrumental variables on farmers’ market participation was 0.367. GCT adoption behavior had a regression coefficient of 0.699 for farmers’ market participation during the second stage. Both coefficients are statistically significant at a level of 1% and exhibit a positive correlation. Thus, the instrumental variable satisfies the correlation requirements. In addition, the F-value of the initial phase was 151.51; based on general guidelines, a value greater than 10 suggests the absence of a weak instrumental variable [64]. Utilizing the “average value of market participation of other farmers in the village where the respondent farmer lives” is a justifiable instrumental variable.

3.1.4. Mediation Effect Model

This study aimed to examine the mechanisms through which farmers’ market participation impacts their GCT adoption behavior. This study refers to Wen and Ye’s [65] mediation effect test model and uses a stepwise regression method to examine the impact mechanism. It is assumed that the treatment variable (market participation) of farmer i affects the dependent variable (GCT adoption behavior) through the variable (mediator variable). The treatment variable had a dual effect on the dependent variable. The first component concerns the immediate impact of market participation on the adoption behavior of GCT, taking into account the mediator variable, which is frequently referred to as the direct effect. The second component pertains to the impact exerted by the mediator variable, which is referred to as the indirect effect. This can be calculated by Equations (3)–(5) for the total effect ($a_1$), indirect effect ($b_1\times c_2$), and direct effect ($c_1$), respectively. The equations used are as follows:

$$Y_i=a_0+a_1{M}_i+a_2{X }_i+{\epsilon }_1$$

(3)

$$C_i=b_0+b_1{M}_i+b_2{X}_i+{\epsilon }_2$$

(4)

$$Y_i=c_0+c_1{M}_i+c_2{C}_i+c_3{X}_i+{\epsilon }_3$$

(5)

where $C_i$ represents the selected mediator variables: agricultural income, market information, and green cognition. $M_i$ and $X_n$ are as defined above and denote the estimated parameters. The symbols $a_1$, $b_2$, and $c_3$ represent the parameters that must be estimated. The symbols ${\epsilon }_1$, ${\epsilon }_2$, and ${\epsilon }_3$ represent the error terms.

3.2. Data Sources

Sichuan Province is China’s traditional grain-growing province and the only grain-exporting province in western China. Selecting grain farmers in Sichuan to analyze grain production behavior is appropriate. Sichuan, situated on the periphery of the Tibetan Plateau, has consistently served as a crucial ecological region in China and has a significant function in maintaining the water quality and climate in the Yangtze River Basin. Therefore, the popularization of GCT in Sichuan is not only beneficial to the improvement of Sichuan’s ecological environment but is also beneficial to the water environment of the entire Yangtze River Basin.

This research relied on survey data collected through questionnaires in Chongzhou, Dayi, Renshou, Luzhou, Zhongjiang, and Jiange, six regions of Sichuan Province (shown in Figure 2), China, between August and October 2023. The survey covered the fundamental characteristics of the respondents, family situation, market participation, and GCT adoption status. We used one-on-one, face-to-face interviews to conduct surveys, and the length of each questionnaire was 1–1.5 h. The sample was selected using a stratified random sampling method. Initially, six counties (cities) were chosen randomly, considering economic development, grain cultivation area, and topographic variations. Next, towns in the counties above were categorized into three different groups, taking into account their financial status and the extent of grain cultivation. Each group randomly selected one town, resulting in 18 towns. Subsequently, three villages were chosen randomly in each town, resulting in 54 villages. Finally, 15–20 grain farmers were randomly selected for the survey in each selected village based on the area of grain they cultivated. Finally, 819 valid questionnaires were obtained from the grain farmers.

3.3. Variable Selection

3.3.1. Dependent Variable

GCT is a product of localized development and the practice of IMP in China, which includes physical prevention and control, biological prevention and control, ecological regulation, scientific application of pesticides, and other measures [5,6]. Physical pest control techniques are the oldest means of dealing directly with pests by killing them, interrupting their habitual behavior, or modifying the environment to discourage their activities [66]. Meanwhile, biological pest control techniques are recognized as the most effective potential alternatives to chemical pesticides, which improve the environmental sustainability of agricultural production and can also increase the efficiency of the technique [67]. Both techniques have little or no negative impact on the environment and leave no residue on agricultural products. Therefore, this study refers to Adhikari et al. [66] and Rodrigues et al. [67] and chose physical and biological pest control techniques to represent the GCT for measurement. In 2023, if grain farmers implemented any of the aforementioned techniques, then Y is equal to 1; otherwise, it is 0.

3.3.2. Core Independent Variable

This study examines the independent variables of farmers’ market participation and market participation capability. Referring to Li et al. [59] and Burke et al. [67], we defined farmers’ market participation as the behavior of farmers who sell some or all of the grain they produce to the market and receive income directly from grain sales. This research also uses dummy variables to portray grain farmers’ market participation. When farmers sold grain to the market in 2022, they are seen as actively engaging in the market and are given a value of 1. Otherwise, they are given a value of 0. Using a binary variable to quantify market participation only accounts for specific effects on farmers’ GCT adoption behavior while neglecting the impact on their capability to participate in the market. Therefore, in this study, the survey data of 658 grain farmers who participated in the market in 2022 were selected from the data of 819 grain farmers to examine further the effect of farmers’ market participation capability on GCT adoption behavior. This study uses whether or not grain farmers who participate in the market achieve good quality and price for their produce to capture their market participation capability. More specifically, a farmer who achieved high quality and a high price for agricultural products in the 2022 market is given a value of 1; otherwise, the farmer is given a value of 0.

3.3.3. Control Variables

Attempting to improve the precision of the model’s estimation, we carefully referred to the existing studies [5,17,22,23] and selected the variables that might affect the farmers’ GCT adoption behavior as control variables from the three dimensions of farmers’ individual characteristics, household characteristics, and village characteristics. More specifically, individual farmer characteristics include age, gender, level of education, status of household head, and part-time employment; household characteristics include household size, cultivated area, cooperative membership, and household assets; and village characteristics include the level of market development and village topography.

3.3.4. Mediating Variable

This study assumes that farmers’ market participation promotes their GCT adoption behavior by gaining agricultural income, unblocking market price information, and enhancing the value cognition level of green production. Therefore, this study selected three mediating variables: agricultural income, market information, and green perceptions. Specifically, agricultural income is expressed using the logarithm of farmers’ income from grain sales in 2022 (10,000 RMB); market information is expressed using the level of farmers’ view that they have easy-to-obtain information on market prices; and green cognition is expressed using the level of farmers’ opinions that green production raises the price of agricultural products.

4. Results

4.1. Descriptive Statistics of Characteristic Variables

Table 2. Defining variables and providing statistical descriptions.

Variable	Definition and Measure	Mean	S.D.
Dependent variable
GCT adoption behavior	Whether farmers adopted physical pest control techniques or biological pest control techniques in 2023, yes = 1; no = 0	0.308	0.462
Physical pest control techniques	Whether farmers adopted physical pest control techniques such as yellow boards and insecticide lamps in 2023, yes = 1; no = 0	0.242	0.428
Biological pest control techniques	Whether farmers adopted biological pest control techniques in 2023, yes = 1; no = 0	0.148	0.355
Independent variable
Market participation	Whether farmers sold grain in 2022, yes = 1; no = 0	0.803	0.398
Market participation capability	Whether agricultural products achieved high quality and prices in 2022, yes = 1; no = 0	0.118	0.323
Mediating variable
Agricultural income	Agricultural income is expressed using the logarithm of farmers’ income from grain sales in 2022 (10,000 RMB)	−0.017	0.776
Market information	Level of farmers’ views that they have easy-to-obtain information on market price, self-assessed 1 to 5	3.695	0. 948
Green cognition	Level of farmers’ views that green production raises the price of agricultural products, self-assessed from 1 to 5	2.932	1.162
Control variables
Age	Age of farmers (years)	53.437	10.180
Gender	Sex of farmer, male = 1; no = 0	0.680	0.467
Level of education	Farmers’ number of years of education (years)	8.416	4.147
Part-time employment	Whether or not farmers are full-time, full-time farmers = 1; otherwise = 0	0.723	0.448
Status of the head of the household	Status of former head of household, Chinese Communist Party member, or village cadre = 1; otherwise = 0	0.241	0.428
Cooperative membership	Farmers’ cooperative membership, membership in cooperative = 1; no = 0	0.265	0.442
Cultivated area	Farmers’ actual acreage in 2022 (acres)	46.739	198.639
Household size	Total number of persons in peasant households	4.446	1.674
Household assets	Farmers’ total household assets such as houses, cars, etc., as of 2022 (10,000 RMB)	41.024	51.170
Village topography	Topography of the village, 1 = plain; 2 = hilly; 3 = mountainous	1.951	0.335
Level of market development	Level of local market development, self-assessment 1 to 5	3.396	0.742
Instrumental variable	Average market participation value of fellow farmers in the same village as the respondents in 2022	0.804	0.226

provides definitions and descriptive statistics for the selected variables. This part uses the summarize command of stata17.0 to calculate the mean and standard deviation of individual variables. The results showed that the percentage of grain farmers adopting GCT was low at 30.77%. However, there are major obstacles to the large-scale promotion and application of GCT in China [11]. This is reasonable because physical and biological pest control techniques require higher initial input costs and practical operational requirements. Therefore, farmers are more inclined to use traditional chemical pesticides, which are less expensive and more efficient in controlling pests and diseases. This study found that the rate at which physical pest control techniques were adopted was greater than the rate at which biological pest control techniques were adopted. This is possible because physical pest control techniques are simpler to operate, more varied, and adaptable.

In the sample, the percentage of grain farmers participating in the market is high, with more than 80% selling grain in 2022. Of the grain farmers involved in the market, 32.98% embraced GCT, a significantly higher rate than the 21.74% of those not in the market. This can be seen as a correlation between farmers’ market participation and GCT adoption behavior. In addition, the percentage of farmers with stronger market participation capability (i.e., farmers who have achieved high quality and prices for their grain among the grain farmers participating in the market) adopting GCT was 53.49%. In contrast, the percentage of farmers with weaker market participation capability (i.e., farmers who have not achieved high quality and prices for their grain among the grain farmers participating in the market) was only 29.90%. This indicates a positive correlation between farmers’ market participation capability and GCT adoption behavior.

As detailed in

Table 3. Univariate test.

Group	Market Participation	Non-Market Participation	Tests of Differences	High Quality and Prices	No Quality and Good Price	Tests of Differences
N	658	161		86	572
GCT Adoption Behavior	0.330	0.217	−0.112 ***	0.535	0.299	−0.236 ***

Note: *** denotes significance at the 1% levels; values in parentheses are standard errors.

, this study executed a univariate test by grouping grain farmers based on whether they participated in the market or not. The data showed that the mean value of farmers’ GCT adoption behavior in the sample of the participating market group was 0.330, which was significantly higher than that of the sample of the non-participating market group, which was 0.217, and the t-test was significant at the 1% level. Similarly, a univariate test was executed by grouping the grain farmers participating in the market based on whether they realized high quality and price. The mean value of the GCT adoption behavior of grain farmers in the sample of the group realizing high quality and price was 0.535, which was much higher than 0.299 in the sample of the group not realizing high quality and price and passed the t-test of significance at the 1% level. This data tentatively suggest that both farmers’ market participation and market participation capability positively influence their GCT adoption behavior. However, further estimation using econometric modeling is required to exclude the effect of other factors.

Table 2 provides the basic information on the control variables. For example, concerning individual characteristics, the farmers in the sample were predominantly older and less educated males, and approximately 24% of the household heads were Chinese Communist Party members or village cadres. Concerning household characteristics, the mean household size is 4.446, which suggests that the sample consists predominantly of small-scale farmers. Concerning village characteristics, the topography of the villages was mainly hilly.

4.2. Determinants of Farmers’ Market Participation

The findings from the initial baseline and CMP models are shown in Models (1), (2), and (3) in

Table 4. Estimates of market participation and GCT adoption behavior.

Variables	Probit	CMP
	(1)	(2)	(3)
Market participation	0.355 *** (0.131)		0.954 *** (0.281)
Age	−0.007 (0.006)	−0.006 (0.007)	−0.006 (0.005)
Gender	0.178 * (0.107)	0.106 (0.124)	0.146 (0.106)
Level of education	0.016 (0.016)	0.018 (0.019)	0.013 (0.015)
Part-time employment	0.156 (0.118)	0.324 ** (0.140)	0.103 (0.119)
Status of household head	0.261 ** (0.117)	−0.120 (0.143)	0.287 ** (0.116)
Cooperative membership	0.189 * (0.110)	0.410 *** (0.152)	0.091 (0.119)
Cultivated area	0.000 * (0.000)	0.004 *** (0.002)	0.000 (0.000)
Household size	0.039 (0.029)	0.001 (0.034)	0.038 (0.029)
Household assets	0.001 (0.001)	0.002 (0.002)	0.001 (0.001)
Village topography	−0.173 (0.147)	0.644 *** (0.189)	−0.245 (0.149)
Level of market development	0.160 ** (0.068)	−0.366 *** (0.089)	0.233 *** (0.075)
Constant	0.355 *** (0.131)	−1.166 (0.718)	−1.846 *** (0.571)
Instrumental variable		2.238 *** (0.254)
${\rho }_{\mu \epsilon }$			−0.479 ** (0.242)
Pearson’s chi-square	57.500 ***		259.710 ***
Log-likelihood	−476.771		−777.551
Pseudo R-squared	0.057
Observations	819	819	819

Note: * p <0.1; ** p <0.05; *** p <0.01; standard errors in parentheses.

. As shown in Model (3), the coefficient of ${\rho }_{\mu \epsilon }$ is negative and statistically significant at 5%. This suggests that there is unobserved endogeneity in the market participation. Thus, the CMP model estimates outperformed the initial baseline model estimates. Hence, the CMP model is a valid approach for analyzing the relationship between grain farmers’ involvement in markets and their GCT adoption behavior.

Model (2) of Table 4 shows the factors impacting grain farmers’ market participation. According to the regression analysis of the factors impacting market participation decisions, the regression coefficient for farmers’ part-time employment is 0.323, which is positive at a level of significance of 5%. This shows that farmers’ part-time employment significantly and positively affects grain farmers’ market participation. Full-time farmers focus on investing additional human and material resources in grain cultivation to enhance the productivity of grain production. Consequently, they rely heavily on agricultural income and must enhance their financial status by selling their harvested grains. Consequently, full-time farmers who rely solely on farming are more inclined to engage in market activities.

The regression coefficient for cooperative membership was 0.410, which is significantly positive at the 1% level of significance, indicating that cooperative membership positively influences grain farmers’ market involvement. The primary reason for this might be that cooperatives unite more farmers to interface with the large market and improve their bargaining power. Thus, cooperative membership can help farmers compete more effectively in the market [68].

The regression coefficient for the cultivated area was 0.004, indicating a statistically significant positive relationship at the 5% significance level. This finding indicates that farmers’ cultivated area is significantly and positively correlated with market involvement. Cultivated areas are essential determinants of the commercialization of agricultural products. Grains are large field crops; the larger the cultivated area, the more economies of scale can be realized. When large-scale farmers produce grain, their initial purpose is to use it for marketing and to earn farm business income. Therefore, the greater the size of a farmer’s operations, the more grain they use for marketing; that is, the more easily they can participate in the market. This positive correlation has also been observed in countries such as Ghana [69] and Zambia [70].

The regression coefficient for the market development level of −0.366 shows a statistically significant negative relationship at 1%. In other words, the higher the level of local market development, the less likely farmers are to participate in agricultural market transactions. This is understandable because farmers in areas with higher levels of market development are predominantly part-time households whose household income is dependent mainly on secondary and tertiary industry employment and who produce grain for food.

The regression coefficient for village topography was 0.644, indicating a statistically significant negative relationship at the 1% significance level. The results suggest that the topography of farmers’ villages is significantly and positively correlated with market participation. In terms of the value assigned to the village topography, the larger the coefficient, the less flat the topography. Plains have larger cities and higher land prices. On the one hand, fewer farmers are involved in grain production; on the other hand, most of those involved in grain production are for their consumption. Therefore, the flatter the terrain of a village, the lower the chance of farmers participating in the market, which is in line with the fact that Sichuan Province is primarily mountainous.

Furthermore, the average value of market participation of other farmers in the village where the respondent farmer lives has a notable positive impact on grain farmers’ market participation.

4.3. Market Participation Impacts on GCT Adoption Behavior

Table 4 shows the regression findings on how grain farmers’ involvement in the market influences their adoption of GCT. Model (3) of Table 4 reports the coefficients of the variables that impact GCT adoption behavior. The regression coefficient for grain farmers’ market participation was 0.954, which is significantly positive at the 1% significance level. This finding suggests grain farmers participating in the market show a notable increase in GCT adoption. Thus, H1 was validated.

Model (3) of Table 4 also indicates a significant correlation between the status of household head, market development level, and farmers’ adoption of GCT. Specifically, CCP members or village cadres can keep abreast of national policies related to green agricultural production and adopt GCT in response to national guidelines. As local market development increases, agricultural price mechanisms become more refined, making it easier for green agricultural products to achieve high quality and prices. Hence, the likelihood of GCT implementation is higher in regions with more advanced local markets.

4.4. Robustness Tests

Using propensity score matching (PSM) to test the robustness of empirical results, this study follows He et al. [59]. By creating a “counterfactual” causal inference framework, the PSM approach can provide a more precise analysis of how market participation impacts farmers’ GCT adoption behavior. The framework successfully distinguished market participation from other impacts on farmers’ GCT adoption behavior. Using nearest-neighbor, caliper, and kernel matching, we examine how farmers’ market participation impacts their GCT adoption behavior. In

Table 5. Results of substituting independent variables in PSM.

Matching Algorithms	Treated	Controls	Standard Error	The Average Treatment Effect
	(4)	(5)	(6)	(7)
Nearest-neighbor matching (1:4)	0.318	0.178	0.055	0.140 *** (2.52)
Caliper matching (with a caliper of 0.02)	0.318	0.180	0.055	0.138 *** (2.49)
Kernel matching (with a bandwidth of 0.06)	0.318	0.176	0.053	0.141 *** (2.66)

Note: *** p <0.01; the number of t-values is shown in parentheses. The sample size is 819.

, we show the estimated results of the average treatment effects for the three matching methods. The results indicate that all the corresponding outcomes related to participation in the market are statistically significant at a significance level of 1%. The impact of market participation on farmers’ GCT adoption behavior appears robust, supporting the study’s findings.

4.5. Mechanism Analysis

Farmers’ GCT adoption behavior is significantly influenced by market participation, but its specific mechanism remains unclear. Based on the above theoretical analysis, this research assumes that grain farmers’ market participation affects their GCT adoption behavior through agricultural income, market information, and green cognition. Following He et al. [59], a stepwise regression method was employed to examine how farmers’ market participation influences their behavior in adopting GCT, with the endogeneity issue being addressed in all models using the CMP command. Based on the stepwise regression analysis of the CMP model,

Table 6. Estimates of the mediating effect of market participation and GCT adoption behavior.

Variables	(Path 1)		(Path 2)		(Path 3)
	(8)	(9)	(10)	(11)	(12)	(13)
Market participation	0.549 *** (0.156)	0.797 *** (0.275)	0.549 *** (0.199)	0.907 *** (0.294)	0.790 *** (0.201)	0.883 *** (0.299)
Agricultural income		0.339 *** (0.076)
Market information				0.089 * (0.051)
Green cognition						0.085 ** (0.043)
Control variable	Yes	Yes	Yes	Yes	Yes	Yes
lnsig_2	−0.425 *** (0.033)		−0.057 ** (0.028)		0.122 *** (0.029)
${\rho }_{\mu \epsilon }$	−0.477 *** (0.170)	−0.359 * (0.209)	−0.281 ** (0.137)	−0.448 * (0.246)	−0.443 *** (0.118)	−0.419 * (0.244)
Pearson’s chi-square	597.590 ***	280.710 ***	229.790 ***	262.900 ***	303.470 ***	263.770 ***
Log-likelihood	−1089.346	−767.049	−1406.430	−775.954	−1536.041	−775.523
Observations	819	819	819	819	819	819

Note: * p < 0.1; ** p < 0.05; *** p < 0.01; standard errors in parentheses.

shows the results.

Model (8) and Model (9) show a direct and indirect impact of market participation and agricultural income on farmers’ GCT adoption behavior that is statistically significant and positive. Therefore, H1a was verified. This outcome suggests that farmers engaged in the market can generate income from agriculture, ultimately boosting their likelihood of adopting GCT. The magnitude of this mediating effect was determined to be 0.186 (the market participation coefficient in Model (8) of Table 6 is multiplied by the agricultural income coefficient in Model (9)), representing approximately 19.497% (calculated by dividing the above mediating effects by the market participation coefficients of Model (3) in Table 4) of the overall influence of market participation on farmers’ adoption behavior of GCT. Among the three mediating variables, agricultural income had the most substantial mediating effect. This result suggests that encouraging farmers to sell their agricultural products for higher agricultural income is the most direct and effective measure to promote the adoption of GCT by farmers.

According to Model (10) and Model (11), the direct impact of market participation and the indirect effect of market information on farmers’ GCT adoption behaviors are statistically significant and positive. Therefore, H1b was verified. This result indicates that farmers’ market participation can unblock market price information, thus effectively enhancing the likelihood of farmers adopting GCT. The magnitude of this mediating effect was calculated to be 0.049 (the market participation coefficient in Model (10) of Table 5 is multiplied by the market information coefficient in Model (11)), which accounts for approximately 5.136% (calculated by dividing the above mediating effects by the market participation coefficients of Model (3) in Table 4) of the total impact of market participation on farmers’ adoption behavior of GCT. The finding suggests that market participation can reduce the risk of the unrealized value of green agricultural products due to information asymmetry by unblocking market price information, which, to a certain extent, can promote farmers’ adoption of GCT.

The results from Model (12) and Model (13) indicate that the direct impact of market participation and the indirect effect of green cognition on farmers’ GCT adoption behavior is positive and statistically significant. Therefore, H1c was verified. The results indicate that farmers’ market participation can enhance their value cognition and thus increase their likelihood of adopting GCT. It was estimated that the size of this mediating effect on farmers’ adoption behavior of GCT was 0.067 (the market participation coefficient in Model (12) of Table 6 is multiplied by the green cognition coefficient in Model (13)), or approximately 5.136% (calculated by dividing the above mediating effects by the market participation coefficients of Model (3) in Table 4) of its total effect. The result suggests that market participation can increase farmers’ income expectations of GCT by enhancing their value cognition of green production and promoting their GCT adoption.

In conclusion, this section econometrically validates the mechanism of market participation’s role in facilitating farmers’ GCT adoption behavior through gaining agricultural income, unblocking market price information, and enhancing the level of value cognition of green production.

4.6. Further Discussion

Farmers’ participation in market transactions was measured as a binary variable in the above analysis; however, the impact of farmers’ market participation capability on their GCT adoption behavior was neglected. Hence, this section delves deeper into how farmers’ market involvement capability affects their adoption of GCT. It uses survey data from 658 farmers who participated in the market in 2022, chosen from a pool of 819 farmers, and used whether or not grain farmers who participate in the market achieve good quality and price for their produce to capture their market participation capability. This offers a broader understanding of how farmers’ market participation decisions and market participation capability impact the adoption of GCT. This contributes to understanding the intricate and diverse nature of market transactions, as opposed to previous research that oversimplified farmers’ market involvement as a binary choice and overlooked the influence of farmers’ market participation capability. Meanwhile, farmers’ market participation capability is also measured using a binary variable. Therefore, we continue to use the probit model as the baseline regression in this part of the study and use the CMP model to control for endogeneity. In addition, to address endogeneity, ‘the average value of realized high quality and prices other farmers in the village where the respondent farmer lives’ was chosen as an instrumental variable.

The findings from the initial baseline and CMP models are shown in Models (14), (15), and (16) of

Table 7. Estimates of market participation capability and GCT adoption behavior.

Variables	Probit	CMP
	(14)	(15)	(16)
High quality and prices	0.572 *** (0.155)		2.053 *** (0.259)
Age	−0.008 (0.006)	−0.006 (0.008)	−0.004 (0.006)
Gender	0.202 * (0.119)	−0.035 (0.145)	0.142 (0.116)
Level of education	0.014 (0.017)	−0.053 (0.021)	0.033 ** (0.017)
Part-time employment	0.215 (0.131)	−0.035 (0.156)	0.162 (0.126)
Status of household head	0.292 ** (0.132)	0.106 (0.165)	0.214 * (0.127)
Cooperative membership	0.109 (0.119)	0.391 *** (0.140)	−0.020 (0.114)
Cultivated area	0.000 (0.000)	0.000 (0.000)	0.000 (0.001)
Household size	0.021 (0.032)	0.053 (0.039)	−0.001 (0.030)
Household assets	0.000 (0.001)	0.003 *** (0.001)	−0.001 (0.000)
Village topography	−0.338 * (0.168)	−0.484 (0.200)	−0.138 (0.163)
Level of market development	0.125 (0.072)	−0.200 ** (0.082)	0.1458 ** (0.068)
Constant	−0.515 (0.606)	0.566 (0.702)	−1.169 ** (0.572)
Instrumental variable		0.860 ** (0.428)
${\rho }_{\mu \epsilon }$			−1.350 ** (0.646)
Pearson’s chi-square	57.150		90.730
Log-likelihood	−388.615		−618.055
Pseudo R-squared	0.069
Observations	658	658	658

Note: * p < 0.1; ** p < 0.05; *** p < 0.01; standard errors in parentheses.

. In Model (16), the coefficient of ${\rho }_{\mu \epsilon }$ (i.e., atanhrho_12) is negative and statistically significant at the 5% level. This implies unobserved endogeneity in market participation capability. Considering this, the CMP model estimation was better than the original baseline model estimation. The CMP model estimation indicated that grain farmers’ market participation capability coefficient was 2.053, with a statistically significant positive effect at the 1% level. Grain farmers who actively participate in the market tend to adopt GCT. Meanwhile, the comparison found that the coefficient of farmers’ market participation capability is much larger than that of the farmers’ market participation coefficient (0.954), indicating that farmers are more prone to adopt GCT if they have strong market participation capability. This may be because market participation is often subject to high transaction costs and unanticipated risks, which limit the contribution of farmers’ market participation to their GCT adoption behavior. This discovery emphasizes that encouraging farmers to use GCT is not something that happens once but rather is an ongoing effort. Therefore, while enabling farmers to participate in the market, more attention should be paid to improving their market participation capability to ensure that farmers who participate in the market can realize high quality and prices.

5. Discussion

The results of this study indicate that farmers’ market participation has a positive effect on their GCT adoption behavior. This is similar to what Xie et al. [31] found in their study, but they have not yet had a systematic study of their influence mechanisms. In this study, it was found that market participation can promote farmers’ GCT adoption behavior through agricultural income, market information, and green cognition mechanisms. More specifically, market participation in the early stage would increase farmers’ household income, help farmers understand the market price information of green agricultural products, and enhance farmers’ cognition of the value of green production, thus effectively solving the dilemma of farmers’ adoption of GCT. It is worth noting that this study found agricultural income to be the variable with the highest mediating effect in the mediating effect analysis. This is understandable because financial resources, as the most significant factor of production, are the primary consideration for the majority of farmers in adopting GCT. The high input costs and long payback period of GCT are currently the most significant obstacles to the adoption of GCT by the majority of smallholder farmers. Whereas market participation in the early stages helps in acquiring farm income. More household income can be used to improve agricultural production practices, alleviating the financial constraints on the adoption of GCT, and thus increasing the likelihood of farmers adopting GCT.

In addition, this study further discussed the effect of farmers’ market participation capability on their GCT adoption behavior, and the coefficient of farmers’ market participation capability (2.053) was much larger than the coefficient of farmers’ market participation (0.954). This indicates that if the market participation capability of farmers is strong, then farmers are more likely to adopt GCT. This may be because market participation is often subject to high transaction costs and unanticipated risks, limiting the contribution of farmers’ market participation to their GCT adoption behavior. Farmers with strong market participation capability realized the achievement of high quality and price for their agricultural products during the pre-market participation process. Meanwhile, these farmers have accumulated more capital financial resources and market information and have mastered more agricultural product market sales channels. This helps farmers to effectively reduce the effect of transaction costs and unexpected risks in participating in the market and makes farmers more likely to believe that the implementation of green production can bring about an increase in the selling price of agricultural products. As a result, farmers with strong market participation are more likely to adopt GCT.

Compared to the existing literature on farmers’ green production techniques adoption behavior and market participation, the marginal contribution of this paper lies in the following three aspects. Firstly, this study not only examines the effect of farmers’ market participation on their GCT adoption behavior but also considers the effect in respect of farmers’ market participation capability. This helps us to gain a more comprehensive insight into the effects of farmers’ market participation decisions and market participation capability on their GCT adoption behavior. Secondly, this study has thoroughly discussed the effect mechanisms from the aspects of income, information, and cognition. This is conducive to enriching the theoretical and empirical analysis of the role of market mechanisms in farmers’ production behavior and providing theoretical and practical support for solving the problem of insufficient application of GCT. Thirdly, this study uses the CMP model to effectively control the endogeneity problem between the relationship between market participation and farmers’ GCT adoption behavior, which makes the estimation results more reliable. The relevant findings of this study are intended to inform policy formulation for the promotion of GCT in developing countries.

Nevertheless, this study has some limitations. For example, this study specifically examined how grain farmers’ involvement in markets relates to their adoption of GCT in Sichuan Province, China, and the generalizability of these findings to other areas requires additional confirmation. Meanwhile, farmers’ GCT adoption behavior was not static and unchanging, but this research explored the correlation between farmers’ market involvement and their GCT adoption behavior using only 2023 farmer survey data. To enhance the comprehension of farmers’ GCT adoption behavior, future research could analyze dynamic panel data from various regions to investigate the dynamic process more thoroughly.

6. Conclusions

Based on survey data from 819 grain farmers in Sichuan Province, China, this study examines the impact of grain farmers’ participation in the market on their adoption of GCT. Specifically, the analysis utilized the CMP and mediated effect models to explore this impact’s relationship and underlying mechanism. The primary findings indicate that various factors influence farmers’ decisions to participate in markets. For example, farmers’ part-time employment can affect farmers’ market participation, with full-time farmers more likely to participate in the market. Furthermore, the topography of farmers’ villages and the level of market development are significantly and positively associated with their market participation. Secondly, the mediating effect analysis shows that farmers’ market participation can influence their GCT adoption behavior through agricultural income, market price information, and green cognition. Lastly, further research found that farmers with strong market participation capability were more likely to adopt GCT.

Using the empirical results, we suggest policies encouraging farmers to use GCT, support sustainable agricultural growth, and preserve their environment. Policy recommendations can help inform decisions regarding the advancement of GCT in developing countries.

Firstly, encouraging farmers to participate in markets gives full play to the positive contribution of market participation to farmers’ GCT adoption behavior. Studies have found that farmers’ part-time employment affects their participation in markets and full-time farmers are more likely to participate in agricultural markets. It is therefore crucial for governments to develop targeted support programs to improve market access for part-time farmers. This could include the provision of flexible market hours or mobile markets to match the schedules of part-time farmers, and financial incentives or subsidies to encourage them to participate in markets. In addition, supporting part-time farmers through training programs on efficient farming methods can help them maximize their limited farming time, potentially increasing their profitability and interest in participating fully in markets. At the same time, the government should encourage full-time farmers to participate in the market on a sustained basis to increase the market participation capacity of full-time farmers.

Secondly, improving farmers’ income, unlocking market information channels, and boosting competence increase green perceptions. This study found that farmers’ market participation promotes farmers’ GCT adoption behaviors through agricultural income, market information, and green cognition. Therefore, the government should implement agricultural green production subsidy policies to increase farmers’ relative income from producing green agricultural products and enhance their ability to adopt GCT. At the same time, the government should reduce the information asymmetry between farmers and consumers of agricultural products through the establishment of an agricultural information service-sharing platform to reduce the market risk of adopting green production. In addition, the government should actively publicize knowledge and policies related to green production to raise farmers’ green production consciousness.

Finally, improving farmers’ market participation capability would give full play to the positive contribution of market participation capability to farmers’ GCT adoption behavior. This study found that farmers with strong market participation capability are more likely to adopt GCT. Therefore, the government should actively improve farmers’ market participation capability to promote the adoption of GCT. For example, the government can help farmers acquire knowledge of market operations and improve their professional skills and marketing capability by increasing support for farmers’ vocational skills training. The government should further improve the system of co-operatives, encourage farmers to participate in co-operatives, and improve their bargaining power in the market. The government can also encourage decentralized farmers to sell their agricultural products through voluntary associations to improve their bargaining power.