Do Chinese Farmers Misuse Pesticide Intentionally or Not?

Abstract

Nonstandard pesticide-application behavior leads to excessive pesticide residue and even affects the quality and safety of agricultural products and agricultural sustainability. Based on 968 valid samples randomly selected in Jiangsu Province of China, it focuses on the impact of incident shock and yield fluctuation avoidance on the pesticide-application behavior of farmers. Then, it investigated the impact of intentional factors, such as insufficient cognition and lack of knowledge, on their improper pesticide-application behavior. This study shows that, besides the pursuit of improper income, inadequate awareness and preventive actions to avoid operational risks are also important factors in farmers’ nonstandard pesticide application. In addition, the study also shows that farmers who understand the responsibility unit of agricultural product quality and safety supervision are more inclined to choose standardized application of pesticides. The higher the education level of farmers, the higher the probability of standardized application of pesticides. Therefore, farmers’ nonstandard pesticide-application behavior is largely due to the farmers’ insufficient awareness of the harm of pesticide residues or the lack of trust in the efficacy of pesticides. Moreover, the study also shows that adverse selection phenomenon exists in pesticide-application training.

1. Introduction

The nonstandard behavior of farmers in the process of pesticide application has posed quality and safety issues for agricultural products and decreased agricultural sustainability in China. Generally, it is believed that the farmers’ motivation for nonstandard pesticide-application behavior is intentional and interest oriented. However, our survey shows that inadequate awareness and preventive actions to avoid operational risks are also important reasons of farmers’ nonstandard application of pesticides, besides the pursuit of improper income on purpose. It is common sense that pesticide is harmful, but there exist different perceptions of the severity of harm. Some farmers are not aware of the harm of pesticide residue, and some even think that consumers in the city are too sensitive to pesticide residue. Additionally, agricultural operation income is relatively low and unstable. In order to reduce the risks in agricultural operations and to stabilize agricultural income, farmers may tend to use excessive pesticides or adopt pesticide varieties with higher efficacy. In this situation, it is of great significance to study the status and reasons for nonstandard pesticide application, especially unintentional reasons. It helps improve farmers’ pesticide-application behavior and thereby provide corresponding support to reduce the risks to the quality and safety of agricultural products and improve agricultural sustainability in China.

There exists abundant literature about nonstandard pesticide-application behavior and its influencing factors [1,2,3,4,5]. Based on economic theory, the decision-making of farmers’ pesticide application is mainly affected by cost and benefit. For rational farmers, the marginal benefit from the use of pesticides should not be less than the marginal cost [6]. In addition, farmers’ behaviors regarding pesticide application are affected by other factors, such as farmers’ individual characteristics [7], land ownership [8], historical habits [9], technology [10], farmers’ organization forms [11], and farmers’ livelihood differentiation [12]. For example, Zadjali et al. examined how agricultural association participation impacts the pesticide-application behavior of farmers in Oman. It was found that farmers participating in an agricultural association were prone to use less banned pesticides [13].

Moreover, multiple studies have also been conducted to study how to change farmers’ nonstandard pesticide-application behavior.

Law and regulation are the most direct ways to change the behavior of farmers [14,15]. However, Lichtenberg showed that the impact of government regulation on farmers’ pesticide use behavior is relatively limited due to the dispersion of farmers. The impact of inadequate cognition on nonstandard pesticide-application behavior has also been tested [16]. For instance, Jallow et al. investigated farmers in Kuwait and found that farmers’ cognition of pesticide hazards had a significant impact on the excessive use of pesticide [17]. Waichman et al. showed that the irrational design of pesticide instructions led to farmers’ ignorance and excessive application of pesticides through a survey of farmers in Brazil [18]. Education and training are the main methods to correct cognition, and they have become the main target variables for researchers to investigate the factors improving farmers’ behavior [19]. For example, Goodhue et al. examined changes in the pesticide-application behavior of California farmers in educational programs [20]. The results showed that education can significantly reduce the amount of pesticide application. However, Schreinemachers et al. found that farmers with education and training were inclined to increase pesticide application, according to the survey of off-season potato growers in Bangladesh [21]. In addition, some research showed that risk perception was also an important factor affecting pesticide inputs [2]. From the perspective of China, as Chinese are paying more attention to the quality and safety of agricultural products and agricultural sustainability, empirical studies on the application of pesticides by Chinese farmers in recent years also started to increase with microdata [22,23,24]. The situation of pesticide application in China was not optimistic. According to the survey conducted by Ma and Huo with 1086 apple growers, only 17.24% of the respondents were fully aware of the pesticide safety interval [25]. The pesticide-application behavior of Chinese farmers was also affected by many factors [5,12,26,27]. Some research found that income and risk preference of farmers impacted pesticide-application behavior [28,29]. Huang et al., through the study of 232 cotton growers, showed farmers with risk aversion tended to apply more pesticide [30]. In addition, some scholars studied the impact of government regulation and income incentive on farmers’ pesticide application [31,32]. For example, Wang and Gu conducted a survey on the pesticide-application behavior of vegetable farmers and found that market incentives affect pesticide-application behavior more significantly than government regulation [33].

Overall, although extensive research has been carried out on pesticide-application behavior, few studies clearly distinguish intentional and interest-oriented excessive pesticide-application behavior from unintentional behavior. According to the theoretical analysis of farmers’ pesticide-application behavior, this paper constructs a measurement variable reflecting farmers’ unintentional nonstandard pesticide-application behavior. Based on the survey data of 968 sample farmers, the effects of main independent variables on three dimensions of illegal behaviors, such as excessive application of pesticide, banned pesticide application, and ignoring the preharvest interval (PHI), are systematically tested.

2. Material and Methods

Farmers’ behaviors are rational based on neoclassical economics theory, and utility maximization is the main target of decision. However, human behavior is also affected by subjective social norms [34], which means the actual decision process should also be analyzed from the perspective of behavior externality. The reduction of negative externality of decision-making behavior will also improve the farmers’ pesticide-application efficiency.

2.1. Theoretical Analysis

A farmer’s efficiency is the sum of the direct efficiency from a behavior’s economic benefits and the indirect efficiency from the externalities originated by the behavior. In terms of farmers’ pesticide application, farmers choose the application method that they think can most directly increase individual direct efficiency. On the other hand, the externality influence due to pesticide application will also have an impact on efficiency, such as the quality and safety risks of agricultural products. Assuming $\mathrm{U}$ is efficiency, which comes from farmers’ pesticide-application behavior:

$$\mathrm{U}=\mathrm{U}\left(\mathrm{W}\right(b_j),{\mathsf{\theta }}_i\mathrm{S}(b_j\left)\right)$$

where W is the direct efficiency of pesticide application, S is the impact of pesticide application externality on the farmers’ efficiency, and $b_j$ is pesticide-application behavior. It is assumed that the pesticide-application behavior is a continuous variable, and the larger the $b_j$, the stronger the negative externality. It should be noted that the impact of the externality of pesticide-application behavior on farmers’ efficiency will be affected by the external cognitive status ${\mathsf{\theta }}_i$ of farmer $i$. When ${\mathsf{\theta }}_i$ is 1, that is, the externality of pesticide-application behavior can be fully recognized, $\mathrm{U}=\mathrm{U}\left(\mathrm{W}\right(b_j),\mathrm{S}(b_j\left)\right)$. When farmers are not fully aware of the externality of pesticide-application behavior, then $\mathrm{U}=\mathrm{U}\left(\mathrm{W}\left(b_j\right)\right)$. The optimal decision condition for farmers is marginal utility MU equals marginal cost MC when farmers fully understand the externality of pesticide-application behavior, the decision-making condition of maximizing utility is $\frac{\partial \mathrm{U}}{\partial b_j}=\frac{\partial \mathrm{U}}{\partial \mathrm{W}}\frac{\partial \mathrm{W}}{\partial b_j}+\frac{\partial \mathrm{U}}{\partial \mathrm{S}}\frac{\partial \mathrm{S}}{\partial b_j}=\mathrm{M}\mathrm{C}$ and suppose the equilibrium pesticide-application behavior is $b_j^1$. When ${0<\mathsf{\theta }}_i<1$, the decision-making condition is $\frac{\partial \mathrm{U}}{\partial b_j}=\frac{\partial \mathrm{U}}{\partial \mathrm{W}}\frac{\partial \mathrm{W}}{\partial b_j}+{\mathsf{\theta }}_i\frac{\partial \mathrm{U}}{\partial \mathrm{S}}\frac{\partial \mathrm{S}}{\partial b_j}=\mathrm{M}\mathrm{C}$ and suppose the equilibrium pesticide-application behavior is $b_j^2$. Since $\frac{\partial \mathrm{U}}{\partial \mathrm{S}}<0$, so, $b_j^2<b_j^1$ and the optimal $b_j$ will increase when ${\mathsf{\theta }}_i$ decreases. That is, the more inadequate cognition of the negative externality of pesticide-application behavior, the more likely the negative externality will increase.

The above analysis shows that farmers’ nonstandard pesticide-application behavior is influenced not only by the intentional behavior of farmers to save costs or improve improper income but also by farmers’ external cognition of pesticide application, that is, the impact of pesticide residue hazards. Nonstandard behavior reflects a farmer’s unintentional actions to some extent under the condition of insufficient cognition. Theoretically, it is clear that “unintentional” behavior is also an important reason for nonstandard pesticide application. Farmers’ unintentional behavior partly comes from the psychology of avoiding yield fluctuation in addition to limited cognition of the pesticide residue hazards. For example, farmers may think that if the standard pesticide-application dosage cannot achieve the desired effect, only by increasing the pesticide dosage can ensure the effectiveness of the pesticide and keep a stable agricultural yield. It seems that excessive pesticide application is partly due to the incorrect cognition of pesticide efficacy and preventive input, which is not the farmers’ intentional violation. Inadequate cognition is the main factor that leads to farmers’ “unintentional” nonstandard behavior application; so, the impact of incident variables on farmers’ pesticide-application behavior is mainly introduced to investigate whether the “unintentional” behavior affects the nonstandard pesticide-application behavior from the empirical level. The underlying logic is that if farmers do not fully understand the quality and safety hazards of agricultural products caused by pesticide nonstandard application, their cognition will be somewhat corrected and their nonstandard decision behavior should be affected when receiving relevant information about the quality and safety events of agricultural products. If the information’s impact on farmers’ pesticide-application behavior is not significant, it shows that farmers’ nonstandard behavior originates from intentional violation to a large extent. In addition, the cognition variable, standard dosage of pesticide application influencing the yield fluctuation of agricultural products, is introduced to measure the impact of the standardization of pesticide application. The underlying logic is that if farmers think that pesticide application with a standard dosage cannot guarantee agricultural product yield, that will affect pesticide application behavior, which indicates that farmers have a strong awareness of agricultural operation risks to a certain extent and hope to ensure agricultural products yield by excessive pesticide application. The cognition of yield influence reflects the motivation to avoid fluctuation risk of agricultural yield to a certain extent. In addition, the regulatory variable will be introduced to focus on as a contrast to measure whether government regulation can correct the farmers’ pesticide-application behavior to a certain extent. In general, if intentional illegal pesticide application is relatively common, farmers’ behavior will be more affected by government regulation.

2.2. Empirical Analysis

2.2.1. Questionnaire Survey

The survey was conducted at random in the villages of Jiangsu Province of China in 2014. Before the survey, we conducted a preliminary investigation and investigators were trained to understand the questionnaire well. At last, 968 valid questionnaires were obtained. Jiangsu Province is the largest producer of japonica rice in southern China. In this survey, the sample size of Southern, Middle, and Northern Jiangsu accounts for 22%, 24%, and 54%, respectively. The average age of survey participants is 46.81 years old. Participants over 50 years old account for 36.47%, while participants younger than or equal to 35 years old account for 13.43% of the sample. The proportion of women is 46.80%, and male participants account for 53.20%. For education level, the percentage of farmers with junior middle school education is the highest, which accounts for 41.94%, followed by those with primary education, which accounts for 32.23%. Participants with high school, college, or a higher education level account for 14.15% and 11.67% of the sample, respectively. It shows that the education level is not high for the sample. From the perspective of annual income, participants with an annual income level in the range of 10,000–30,000 RMB account for 45.66% of the sample, which is the largest proportion. Participants with annual income less than 10,000 RMB, in the range of 30,000–60,000 RMB, 60,000–100,000 RMB, and more than 100,000 RMB account for 20.87%, 23.24%, 6.92%, and 3.31%, respectively. From the perspective of the main source of household income, migrant work is the main source of income for most farmers (44.83%) followed by farming (36.57%), and participants doing business other than farming account for 18.60%. It shows that although migrant work is the main income source for most participants, agriculture is still the main source of income for some participants. The average operation area for a farming household is 9.01 mu. However, the average operation scale will drop to 5.53 mu if large agricultural management households with farmland areas exceeding 100 mu are excluded, which indicates that the scale of farmland operation for the sample is not large. Demographic characteristics of Survey Participants are shown in

Table 1. Demographic characteristics of Survey Participants.

Characteristic Variables	Definition	Assignment	Mean	Standard Deviation
Gender	Sample gender	Male = 1; Female = 0	0.532	0.499
Age	sample age	Actual value	46.812	11.624
Family size	the number of families	Actual value	4.409	1.337
Education level	Education level of samples	Primary school or below = 1; Junior high school = 2; Senior high school = 3; College or above = 4	2.053	0.964
Income level	Annual income level of samples	<=10,000 = 1; 10,000–30,000 = 2; 30,000–60,000 = 3; 60,000–100,000 = 4; >1000,000 = 5	2.261	0.974
Income source	Main income source of family	Farming = 1; Others = 0	0.366	0.482

.

2.2.2. Model Setup

We use the following econometric model to investigate whether farmers’ nonstandard pesticide-application behavior is unintentional and to test the influence of related factors on farmers’ pesticide-application behavior.

$${\mathrm{f}(\mathrm{y}}_i)={\mathsf{\alpha }}_i+{\mathrm{X}}_{\mathrm{i}\mathrm{j}}{\mathsf{\beta }}_j+{\mathrm{Z}}_{ij}{\mathsf{\gamma }}_j+{\mathrm{W}}_{ij}{\mathsf{\delta }}_j+{\mathsf{\epsilon }}_i$$

(1)

where ${\mathrm{y}}_i$ denotes pesticide-application behavior of individual farmer i, ${\mathrm{X}}_{\mathrm{i}\mathrm{j}}$ is a vector of the main independent variables, ${\mathrm{Z}}_{ij}$ is a vector of control variables of related cognition, ${\mathrm{W}}_{ij}$ is the vector of the control variables of individual characteristics, ${\mathsf{\beta }}_j,{\mathsf{\gamma }}_j$ and ${\mathsf{\delta }}_j$ are the estimated coefficient vectors, and ${\mathsf{\alpha }}_i$ and ${\mathsf{\epsilon }}_i$ are the constant term and random error, respectively.

(1)

Dependent variable

Different pesticide-application behaviors will bring different levels of quality and safety risk for agricultural products. The effects of pesticide dosage, pesticide variety, and PHI on pesticide-application behavior are investigated in this study, respectively.

In recent years, the dosage per mu of pesticide application was far more than the reasonable level in China. The choice of pesticide dosage will have a great impact on the pesticide residue of agricultural products, which will lead to quality and safety risks for agricultural products. Regarding the question “what is the choice of pesticide dosage when using pesticide”, the survey results show that 61.88% of farmers chose “according to prescribe dosage of the instructions”, and 7.23% of farmers chose “less than prescribed in the instructions.” Participants who choose “a little more than prescribed in the instructions” and “more casually” account for 21.90% and 7.23%, respectively, and that will increase the quality and safety risks of agricultural products due to the nonstandard pesticide-application dosage.

From the perspective of the choice of pesticide variety, the impact of a pesticide on the quality and safety of agricultural products is not only shown in terms of pesticide dosage but also the toxicity of the pesticide. The pesticide variety is continuously updated and medium or high toxicity pesticides are phased out with societal development, advances in technology, and increased awareness of sustainability. Low toxicity or biological pesticides have become the mainstream of pesticide application. However, the application of a highly toxic banned pesticide often happens in pursuit of efficacy or cost savings in the actual pesticide application, which causes many quality and safety outbreaks in agricultural products. Regarding the question “will you choose the pesticide banned by the country when choosing the pesticide?”, 60.43% of participants chose “no”, and participants who chose “(probably) will, because I don’t know what is banned pesticide in the country” account for 28.10%. There are still 11.47% of participants who chose “yes, as long as no one caught it”. Although there might be some bias in the hypothetical surveys, the results still show that some farmers do not know how to choose pesticide variety and there are large potential risks. In addition, 28.10% of participants are not clear about which pesticide is banned and may choose a banned pesticide. It also preliminarily shows that some farmers’ illegal pesticide applications are unintentional behaviors to some extent.

In addition, if the pesticide applied is not fully degraded, it will still likely cause excessive pesticide residues and further cause quality and safety risks to agricultural products. Regarding the question “do you consider preharvest interval?”, the results show that participants who consider PHI account for 69.83%, and those who do not consider PHI account for 30.17%. It shows that there are still high risks in the specific operation of PHI in China.

(2)

Main Independent Variables

According to the theoretical analysis, the impact of information about agricultural products’ quality and safety outbreaks, yield effect, and government regulation is mainly included as the independent variables. Agricultural product risk and safety outbreak information will affect farmers’ cognition of the harm of pesticide residue to some extent, which is the main variable to measure whether farmers have nonstandard pesticide application unintentionally. During the investigation of quality and safety incidents of agricultural products such as “toxic cowpeas” and “poisonous ginger” caused by pesticide residue, participants who choose “have heard of” account for 75.21%, and participants who never heard of such events account for 24.79%, shown in Figure 1. In recent years, events of vegetable quality and safety problems, such as toxic cowpeas (Xinhua, 2010, http://www.chinadaily.com.cn/china/2010-02/27/content_9514301.htm (accessed on 27 February 2010)) and poisonous ginger (China Daily, 2013, http://www.chinadaily.com.cn/opinion/2013-05/09/content_16486816.htm (accessed on 9 May 2013)) have occurred in China frequently. It shows although farmers are the main producers of agricultural products, nearly 1/4 of the farmers have not heard about the related quality and safety events of agricultural products, which also reflects poor access to information on the quality and safety of agricultural products to some extent. As a realistic education, the quality and safety events information of agricultural products have not been heard of, which may affect the farmers’ attention to the quality and safety of agricultural products, and further affect pesticide-application behavior. From the farmers’ cognition of the impact of pesticide-application status on the agricultural products yield, high intensity and nonstandard application behavior of farmers is largely due to doubts of the pesticide’s effectiveness. Theoretically speaking, pesticide application with a standard dosage can guarantee the pesticide’s effectiveness. However, some farmers have a strong risk perception and expect to use excessive input or a highly toxic pesticide to ensure effectiveness. Figure 2 shows that 20.66% of participants think a standard pesticide dosage cannot guarantee the pesticide’s effect and largely affects agricultural product yields. Participants who believe standard pesticide dosage has a certain influence on yield account for 50.41% and 28.93% of participants who think that the standard pesticide dosage can achieve the desired effect and guarantee agricultural product yield. In addition, the government’s supervision of the quality and safety of agricultural products is intended to regulate the farmers’ quality and safety behaviors of agricultural products, which should also affect pesticide-application behavior. As for the question “whether there is local supervision on the quality and safety of agricultural products?” the survey shows that 39.05% of participants chose “hardly any”, 53.51% chose “occasionally”, and only 7.44% chose “often”. Therefore, the level of quality supervision of agricultural products is not high in China.

Based on the previous literature, such as Asmare et al. (2022) [1] and Wang et al. (2017) [4], other relevant variables are considered as control variables in the study. Specifically, the recognition of the regulatory department, understanding of the relevant laws, and the status of agricultural product certification are used as control variables for farmers’ awareness of government regulation of agricultural products. Whether agriculture can be irrigated with wastewater is used as a control variable for the farmers’ cognition of common sense. The judgments on the quality and safety statuses of agricultural products are used as control variables for farmers’ cognition of the quality and safety environment of agricultural products. The summary statistics for these variables are shown in

Table 2. The definition and summary statistics of independent variables.

Type	Variables	Definition	Assignment	Mean	Standard Deviation
Main independent variables	Incidents shock	Have you heard of events such as “poison bean sprouts” and “poisonous ginger”?	Yes = 1; No = 0	0.752	0.432
	Yield effect	The effect of standard pesticide application on yield	No influence = 1; Some influence = 2; Great influence = 3	1.917	0.700
	Regulation situation	Whether to carry out the quality and safety supervision of agricultural products	Almost no = 1; Occasionally = 2; Almost yes = 3	1.684	0.604
Main control variables	Regulation cognition	Which department regulates the quality of agricultural products	Agricultural department = 1; Others = 0	0.574	0.495
	Legal knowledge	Do you know the <<Law on the Quality and Safety of Agricultural products>>	Never heard of = 1; Others = 0	0.274	0.446
	Certification Cognition	which is the most strict standard in China’s agricultural certification	Organic food = 1, Others = 0	0.154	0.361
	Wastewater irrigation	Is it okay to irrigate farmland with factory wastewater?	Yes = 1; Do not know or NO = 0	0.881	0.324
	Security status	Cognition of the quality and safety of agricultural products in China	Very good 5-4-3-2-1 Very poor	2.960	0.970

.

3. Results

The effects of incident shock, production impact, and government regulation on farmers’ pesticide-application behavior are investigated without including control variables. Dependent variables are defined as follows: for pesticide usage, when participants choose more than the standard dosage while applying pesticide, the pesticide dosage variable is defined as one, and zero otherwise; for the banned pesticide, when participants choose “will, as long as no one caught it”, is defined as one, and zero otherwise; and for the PHI variable, when not considering the preharvest interval, PHI variable is defined as one, and zero otherwise. The regression results are shown in

Table 3. Estimation results on influencing factors of pesticide application without control variables.

	Excessive Application		Using Banned Pesticide		Ignoring PHI
	(1)	(2)	(1)	(2)	(1)	(2)
Yield effect	0.066 *** (0.019)	0.225 *** (0.065)	0.103 *** (0.015)	0.569 *** (0.083)	0.070 *** (0.021)	0.198 *** (0.062)
Incidents shock	−0.087 *** (0.032)	−0.279 *** (0.101)	−0.113 *** (0.026)	−0.516 *** (0.116)	−0.110 *** (0.035)	−0.305 *** (0.096)
Regulation situation	−0.028 (0.021)	−0.101 (0.075)	−0.026 * (0.015)	−0.177 * (0.096)	−0.012 (0.024)	−0.037 (0.070)
Constant	0.204 *** (0.057)	−0.844 *** (0.197)	0.046 (0.041)	−1.740 *** (0.249)	−0.728 *** (0.065)	−0.616 *** (0.188)
R2/Pseudo R2	0.024	0.022	0.080	0.110	0.023	0.018
N	968	968	968	968	968	968

Note: *** and * denote significance at 1% and 10% levels, respectively, and numbers in parentheses mean standard deviation.

. Column (1) shows the regression results using the least squares, and column (2) is the probit regression result. The regression result shows that incidents’ shock and yield effect both have a significant impact on the three dimensions of pesticide-application behaviors, including excessive application of a pesticide, using a banned pesticide, and ignoring PHI. The incident shock coefficient is negative and significant, which means participants who did not hear of events such as poisonous bean sprouts and poisonous ginger tend to apply excessive pesticides, use a banned pesticide, and ignore PHI. The positive coefficient of the yield effect variable means that the participants believe that the standard pesticide application has little impact on agricultural product yield fluctuation and tend to apply a pesticide normatively. From the perspective of government regulation, the negative coefficient is consistent with the expectation. However, the negative effect is significant in the regression of using a banned pesticide only. Participants supervised by the government are less likely to use a banned pesticide.

We add control variables in regression, and the regression results are shown in

Table 4. Estimation results with control variables.

Variable	Excessive Application		Using Banned Pesticide		Ignoring PHI
	(1)	(2)	(1)	(2)	(1)	(2)
Yield effect	0.059 *** (0.019)	0.202 *** (0.066)	0.094 *** (0.015)	0.550 *** (0.089)	0.061 *** (0.021)	0.179 *** (0.063)
Incidents shock	−0.075 ** (0.033)	−0.247 ** (0.104)	−0.096 *** (0.026)	−0.456 *** (0.123)	−0.099 *** (0.036)	−0.293 *** (0.101)
Regulation situation	−0.014(0.021)	−0.051 (0.077)	−0.021 (0.016)	−0.149 (0.106)	−0.004 (0.025)	−0.006 (0.075)
Regulation cognition	−0.055 ** (0.027)	−0.180 ** (0.093)	−0.074 *** (0.021)	−0.439 *** (0.118)	−0.043 (0.030)	−0.128 (0.089)
Legal knowledge	0.058 * (0.032)	0.203 * (0.105)	−0.039 * (0.021)	−0.192 (0.147)	0.017 (0.035)	0.055 (0.102)
Security status	−0.003 (0.014)	−0.011 (0.048)	−0.008 (0.011)	−0.045 (0.061)	−0.001 (0.015)	−0.006 (0.046)
Certification Cognition	−0.001 (0.037)	0.003 (0.127)	−0.007 (0.028)	−0.026 (0.155)	−0.099 *** (0.038)	−0.303 ** (0.127)
Wastewater irrigation	−0.007 (0.043)	−0.020 (0.142)	−0.166 *** (0.040)	−0.662 *** (0.147)	−0.170 *** (0.050)	−0.461 *** (0.132)
Income source	−0.052 * (0.030)	−0.189 * (0.106)	0.043 ** (0.022)	0.273 ** (0.130)	−0.015 (0.032)	−0.045 (0.097)
Gender	−0.005 (0.027)	−0.022 (0.095)	0.012 (0.020)	0.056 (0.121)	0.004 (0.030)	0.007 (0.089)
Age	−0.002 (0.001)	−0.007 (0.005)	−0.003 *** (0.001)	−0.016 *** (0.006)	−0.005 *** (0.001)	−0.015 *** (0.005)
Household population	0.027 *** (0.010)	0.098 *** (0.034)	−0.002 (0.007)	−0.024 (0.044)	0.009 (0.011)	0.026 (0.033)
Education	−0.014 (0.016)	−0.047 (0.057)	−0.014 (0.011)	−0.109 (0.069)	−0.055 *** (0.017)	−0.178 *** (0.055)
Income	0.000 (0.014)	0.008 (0.050)	0.016 (0.011)	0.110 * (0.060)	0.039 ** (0.016)	0.121 ** (0.048)
Land area	0.000 * (0.000)	0.001 (0.000)	0.000 ** (0.000)	0.001 (0.000)	0.000 (0.000)	0.001 (0.001)
Constant	0.235 *** (0.129)	−0.796 * (0.435)	0.376 *** (0.091)	−0.143 (0.498)	0.675 *** (0.135)	0.572 (0.403)
R2/Pseudo R2	0.045	0.043	0.143	0.185	0.068	0.056
N	968	968	968	968	968	968

Note: ***, **, and * denote significance at 1%, 5%, and 10% levels, respectively, and numbers in parentheses mean standard deviation.

. Table 4 shows that an incident’s shock and production influence variables are significant for three dimensions of pesticide-application behaviors. It means participants who have not heard of quality and safety incidents of agricultural products caused by pesticides and consider a pesticide dosage’s effect on agricultural production prefer excessive pesticide application, use a banned pesticide with higher probability, and are also more likely to ignore PHI. The sign of the coefficient for the influence of government regulation on nonstandard behavior is expected, but not significant. Among the control variables, knowing about an agricultural product’s regulatory agency helps prevent excessive pesticide application and the choice of banned pesticide. The legal cognition variable plays a significant role in promoting the normalization of pesticide dosage. Participants who think that plant wastewater cannot be used to irrigate farmland tend to avoid banned pesticides and consider PHI; income sources have significant effects on pesticide dosage and banned pesticide application. Participants whose main income sources come from farming are more inclined to apply a standard dosage of pesticide and use a banned pesticide. They pay more attention to pesticide efficiency but neglect negative externality. Age has a significant effect on banned pesticide application and PHI consideration. Younger participants tend to apply a banned pesticide and ignore PHI more than the older ones. Participants with a larger household size tend to apply excessive pesticides. Participants with higher education and lower income are also inclined to pay attention to PHI compared to those with lower education and higher income.

Dependent variables are considered only as dummy variables in Table 4. In our survey, there are options such as “casual” and “don’t know what pesticide are banned by the state” for pesticide dosage and banned pesticides. Although participants who choose these options have no subjective violation intention, their behavior influences the quality and safety of agricultural products. To comprehensively evaluate farmer behavior and use the survey data effectively, further analysis is performed on two normative behaviors, pesticide dosage and banned pesticide. Due to the uncertainty of nonintentional violation, the effect on the quality and safety risk of agricultural products is between standardized application and deterministic illegal pesticide application, and we make the following transformation of pesticide dosage and banned pesticide variable: for pesticide dosage, the dependent variable is one when participants choose more than the standard amount or less than the standard dosage, two for casual pesticide application, and three when pesticide applications exceed the standard; for banned pesticide, the dependent variable is one when participants choose to not apply a banned pesticide, two when choosing not clear what pesticide is banned, and three when choosing to apply banned pesticide as long as not being caught. The regression results are shown in

Table 5. Estimation results from transformation of the dependent variables.

Variable	Excessive Application			Using Banned Pesticide
	(1)	(2)	(3)	(1)	(2)	(3)
Yield effect	0.241 *** (0.059)	0.219 *** (0.060)	0.235 *** (0.069)	0.415 *** (0.056)	0.380 *** (0.057)	0.647 *** (0.100)
Incidents shock	−0.304 *** (0.092)	−0.258 *** (0.095)	−0.285 *** (0.108)	−0.324 *** (0.091)	−0.279 *** (0.095)	−0.334 ** (0.138)
Regulation situation	−0.127 * (0.069)	−0.053 (0.071)	−0.053 (0.079)	−0.176 *** (0.066)	−0.140 ** (0.069)	−0.167 (0.115)
Regulation cognition		−0.204 ** (0.085)	−0.226 ** (0.097)		−0.259 *** (0.080)	−0.482 *** (0.131)
Legal knowledge		0.298 *** (0.093)	0.276 ** (0.110)		−0.030 (0.090)	−0.157 (0.162)
Security status		−0.044 (0.044)	−0.031 (0.049)		−0.056 (0.042)	−0.067 (0.068)
Certification Cognition		−0.028 (0.118)	−0.010 (0.132)		0.028 (0.108)	0.000 (0.171)
Wastewater irrigation		−0.093 (0.124)	−0.071 (0.150)		−0.630 *** (0.118)	−0.874 *** (0.176)
Income source		−0.241 ** (0.096)	−0.224 ** (0.109)		0.155 * (0.089)	0.258 * (0.147)
Gender		0.024 (0.087)	−0.007 (0.098)		0.110 (0.082)	0.059 (0.134)
Age		−0.006 (0.004)	−0.007 (0.005)		−0.013 *** (0.004)	−0.020 *** (0.007)
Household population		0.086 *** (0.031)	0.103 *** (0.035)		0.041 (0.031)	0.002 (0.052)
Education		−0.088 * (0.053)	−0.085 (0.060)		−0.117 ** (0.049)	−0.162 ** (0.077)
Income		−0.013 (0.047)	0.000 (0.053)		0.098 ** (0.042)	0.157 ** (0.070)
Land area		0.001 (0.000)	0.001 (0.000)		0.001 *** (0.000)	0.006 (0.006)
Constant			−0.548 (0.453)			0.133 (0.545)
cut1	0.532 (0.178)	0.245 (0.391)		0.528 (0.171)	−0.651 (0.365)
cut2	0.816 (0.178)	0.537 (0.391)		1.527 (0.173)	0.405 (0.366)
R2/Pseudo R2	0.021	0.043	0.057	0.044	0.081	0.233
N	968	968	881	968	968	696

Note: ***, **, and * denote significance at 1%, 5%, and 10% levels, respectively, and numbers in parentheses mean standard deviation.

.

Column (2) shows the results by including control variables. In addition, to study the effects of relevant factors on behavior in pesticide application, the regression of pesticide dosage excludes participants choosing casual pesticide application in column (3), and banned pesticide regression excludes participants choosing not clear what pesticide is banned. Columns (1) and (2) use ordered selection regression, and column (3) adopts probit regression. From the regression results, production and incident shock have a significant effect on pesticide dosage and further indicate that the nonstandard pesticide application of farmers partially originates from the response to income uncertainty or inadequate and incorrect cognition of harmful information. In addition, the regression results also show that government supervision has an inhibitory effect on the use of a banned pesticide. For the control variables, correct recognition of regulatory agencies and understanding of the law could promote normal pesticide-application behavior for farmers to a certain extent; correct recognition of regulatory agencies and correct understanding of wastewater irrigation could promote farmers’ normal pesticide variety use to a certain extent. Increased education levels could promote normal pesticide application and variety use behavior.

The regression results show that the normalization of pesticide application is affected by many factors. Overall, incident shock and yield effect variables have significant effects on three dimensions of behaviors. Farmers impacted by quality safety incidents shock on agricultural products and who think standard dosage will not decrease the yield for agricultural products are inclined to apply a pesticide according to standard, are less likely to choose a banned pesticide, and have relatively stronger consciousness to consider PHI. Thus, unintentional behavior, such as insufficient information cognition and avoiding agricultural production risk, leads to farmers’ nonstandard pesticide-application behavior to some extent. In addition, the above analysis also shows that, although government regulation has a certain effect on the normalization of pesticide applications, the impact is relatively low.

Robustness Test

We define a new pesticide-application behavior variable behavior, which is the summation of the three dimensions of the pesticide-application behavior value. When behavior is equal to three, it means that all three dimensions of pesticide-application behavior, including application dosage, pesticide types, and PHI are nonstandard. When behavior value is zero, it means that all three dimensions of pesticide-application behaviors are normative. As the value of behavior increases, farmers’ pesticide-application behavior gets worse. Of all the participants, 532 (54.96% of the sample) do not have any of the three dimensions of nonstandard pesticide-application behaviors, and their values of the variable behavior are zero; 32 participants (3.31% of the sample) have the behavior value as three, indicating these participants have all the three dimensions of nonstandard pesticide-application behaviors; participants with a behavior value of one and two account for 29.86% and 11.88% of the sample, respectively. Overall, although the percentage of participants who have the three dimensions of nonstandard pesticide-application behaviors is not high, the percentage of participants who do not have any of the three dimensions of nonstandard pesticide-application behaviors is less than 60%. In addition, the quality and safety of agricultural products is a major issue related to the health of people and any nonstandard pesticide-application behavior is likely to lead to quality and safety problems. Thus, the overall pesticide-application behavior of the sampled farmers still needs to be further improved.

To investigate the relevant factors affecting the overall pesticide-application behavior and further examine whether unintentional behavior is the cause of nonstandard pesticide application, behavior is used as the dependent variable for regressions. In

Table 6. Robustness test.

	(1)	(2)	(3)	(4)	(5)	(6)
Yield effect	0.239 *** (0.037)	0.336 *** (0.053)	0.213 *** (0.037)	0.304 *** (0.054)	0.321 *** (0.049)	0.445 *** (0.071)
Incidents shock	−0.309 *** (0.064)	−0.415 *** (0.085)	−0.270 *** (0.064)	−0.376 *** (0.087)	−0.281 *** (0.083)	−0.364 *** (0.112)
Regulation situation	−0.066 * (0.040)	−0.091 (0.061)	−0.039 (0.040)	−0.043 (0.063)	−0.034 (0.051)	−0.014 (0.079)
Regulation cognition			−0.173 *** (0.051)	−0.265 *** (0.077)	−0.241 *** (0.067)	−0.337 *** (0.099)
Legal knowledge			0.036 (0.057)	0.084 (0.084)	0.029 (0.074)	0.090 (0.109)
Security status			−0.011 (0.026)	−0.023 (0.039)	−0.013 (0.034)	−0.031 (0.050)
Certification Cognition			−0.106 (0.066)	−0.139 (0.103)	−0.102 (0.089)	−0.117 (0.134)
Wastewater irrigation			−0.342 *** (0.082)	−0.465 *** (0.105)	−0.405 *** (0.110)	−0.530 *** (0.137)
Income source			−0.025 (0.056)	−0.058 (0.085)	0.008 (0.071)	0.002 (0.107)
Gender			0.011 (0.051)	0.006 (0.078)	0.084 (0.066)	0.092 (0.100)
Age			−0.010 *** (0.003)	−0.015 *** (0.004)	−0.014 *** (0.003)	−0.021 *** (0.005)
Household population			0.034 * (0.019)	0.061 ** (0.029)	0.031 (0.024)	0.059 (0.037)
Education			−0.084 *** (0.030)	−0.139 *** (0.047)	−0.144 *** (0.039)	−0.235 *** (0.063)
Income			0.055 ** (0.027)	0.090 ** (0.041)	0.099 *** (0.035)	0.167 ** (0.053)
Land area			0.001 * (0.000)	0.001 (0.001)	0.001 *** (0.000)	0.002 *** (0.001)
Constant	0.521 (0.111)		1.286 *** (0.233)		1.384 *** (0.302)
cut1		0.304 (0.166)		−0.774 (0.344)		−0.769 (0.444)
cut2		1.250 (0.170)		0.215 (0.344)		0.104 (0.445)
cut3		2.118 (0.172)		1.116 (0.340)		0.907 (0.440)
R2/Pseudo R2	0.074	0.033	0.132	0.063	0.205	0.098
N	968	968	968	968	633	633

Note: ***, **, and * denote significance at 1%, 5%, and 10% levels, respectively, and numbers in parentheses mean standard deviation.

, columns (1)~(4) are the regression results of the full sample, and columns (5)~(6) are the regression results, excluding participants who choose pesticide dosage randomly or are unclear about a banned pesticide. Columns (1), (3), and (5) are the least squares regression results, and columns (2), (4), and (6) are the regression results using the PROBIT model.

The results show that the yields and impact of events are statistically significant at a 1% significance level, and the sign of the coefficients is expected. Participants who think using a standard dosage of pesticide cannot ensure agricultural products’ yield and have not heard of quality and safety events, such as “poison bean sprouts” and “poison ginger”, are more likely to have nonstandard pesticide-application behavior. Although the sign of the influence of government regulation on participants’ pesticide-application behavior is expected, it is not significant. For control variables, those participants who have the correct cognition of regulations are more likely to have standard pesticide-application behavior, and participants who regard the use of factory wastewater for irrigation as inappropriate are more likely to have standard pesticide-application behavior. Younger participants and those with a lower education level and higher income levels are more likely to have inappropriate pesticide-application behavior.

4. Discussion

The inappropriate use of pesticides by farmers is common in developing countries. For example, there are studies that show that 100% of the sampled farmers in Vietnam, 73% in Cambodia, and 59% in Laos overused pesticides [35]. Zhao et al. (2018) found that many farmers still use pesticide 666 [36], which has been banned in China since the 1980s. The inappropriate use of pesticides can have a significant impact on the environment, food safety, and the health of farmers. In order to improve farmers’ behavior and reduce the inappropriate use of pesticides, it is necessary to investigate the causes of farmers’ improper use of pesticides. There have been many studies on the influencing factors of farmers’ pesticide-application behavior. For example, the influence of attitudes, subjective norms, moral norms, perceived behavioral control, and knowledge of farmers’ pesticide-application behavior was analyzed based on the theory of planned behavior [37]. According to Schreinemachers et al.’s (2019) investigation and research on farmers in Southeast Asian countries, Pesticide overuse was positively associated with men in charge of pest-management decisions, farmers seeking advice from pesticide sellers, and a strong belief that pesticides are effective [35]. Cai et al. (2021) use data collected from 452 apple farms in Shaanxi and Shandong provinces in 2017 and found that 70.6% of apple farms engage in excessive pesticide use, with farmers that have higher risk aversion more likely to overuse pesticides [38]. Some studies have also focused on the intentional improper use of pesticides by farmers in pursuit of improper benefits [39]. Few literature divides the influence factors on farmers’ pesticide application into two aspects: unintentional and intentional. It is of great significance to clarify the subjective intention or non-intention of pesticide application for farmers to put forward corresponding countermeasures.

This study shows that some farmers believe that the standard application of pesticides does not guarantee the proper yield of agricultural products. This makes these farmers choose to overapply pesticides. Some farmers have not heard of the agricultural product quality and safety problems caused by pesticide residues, such as “toxic bean sprouts” and “toxic ginger”, and have not been directly impacted by the corresponding agricultural product quality and safety problems. This makes the cognition of agricultural product quality and the safety risk inadequate, resulting in nonstandard pesticide-application behavior. In addition, the study also shows that farmers who understand the responsibility unit of agricultural product quality and safety supervision are more inclined to choose standardized application of pesticides. The higher the education level of farmers, the higher the probability of standardized application of pesticides. Therefore, farmers’ nonstandard pesticide-application behavior is largely due to the farmers’ insufficient awareness of the harm of pesticide residues or the lack of trust in the efficacy of pesticides. This study enriched the research on the influencing factors of farmers’ improper use of pesticides from different perspectives and provided research support for putting forward corresponding policy recommendations.

Moreover, similar to this paper, Wang and Liu (2021) used this data to compare the influence of farmers’ attitudes and government supervision on the application of pesticides by farmers [40]. Compared with the previous article, this paper changed the research perspective, focusing on the impact of event information shock and yield fluctuation avoidance on the pesticide-application behavior of farmers, and then investigated the impact of subjective and intentional violation factors, such as insufficient cognition and lack of knowledge, on their improper pesticide-application behavior. This paper can be regarded as a further supplement and extension to the study of Wang and Liu (2021).

According to this study, lack of knowledge is an important cause of improper pesticide-application behavior of farmers. Therefore, in order to improve the pesticide-application behavior of farmers and reduce the impact of pesticide abuse, information and knowledge should be provided to farmers, such as strengthening relevant training. Strengthening training is the main measure to correct farmers’ pesticide-application behavior and alleviate the risk of agricultural product quality and safety caused by unintentional factors. The willingness of farmers to participate in training was also investigated in the survey, but the results were not satisfactory. Farmers who chose to participate in free agricultural product planting safety training accounted for 54.55% of the total sample, farmers who chose to see the situation accounted for 28.20%, and 17.25% of the sample farmers chose not to participate. Farmers’ participation in quality and safety training of agricultural products is not high. Therefore, in order to better correct the improper application behavior of farmers, in addition to training farmers, it is necessary to improve the participation rate of training through incentives and other methods, so that farmers can obtain relevant information about pesticide application and its impact.

5. Conclusions

The quality and safety problems of agricultural products are largely due to farmers’ improper pesticide use behavior. Therefore, the improvement of the quality and safety of agricultural products and agricultural sustainability mainly depends on the improvement of farmer’s behavior. Pesticide is one of the major quality and safety risk sources for agricultural products, which makes regulation of farmers’ pesticide-application behavior the key to reducing the quality and safety risk of agricultural products. On one hand, improper pesticide-application behavior is for pursuing improper income; on the other hand, some farmers have inadequate cognition and avoid the risk of agricultural management, which leads to improper pesticide-application behavior, and the “unintentional” actions increase the quality and safety risks for agricultural products. To test this hypothesis, the survey data of farmers in Jiangsu Province of China are used to investigate the standardization of pesticide application from the aspect of pesticide-application dosage, pesticide variety, and PHI. How factors such as incidents shock and yield fluctuation avoidance impact farmers’ pesticide applications are further studied. The results show that the cognition of a standard pesticide application’s impact on product yield and the awareness of agricultural product quality and safety incidents have significant effects on all three dimensions of pesticide-application behavior. Specifically, the farmers who think pesticide application with prescribed dosage cannot guarantee agricultural product yield and the farmers who have not heard about the quality and safety incidents of agricultural products tend to overuse pesticides, use a banned pesticide, and ignore pesticide withdrawal before the agricultural product’s harvest. The result shows that the nonstandard pesticide-application behavior is influenced by cognition of quality and safety hazards of agricultural products and uncertainty of pesticide efficacy. It indicates that “ignorance” about the quality and safety hazards of agricultural products and the “unintentional” violation to ensure a stable agricultural product yield are also important reasons for nonstandard pesticide application besides the intentional violation to get improper profits.

Therefore, to improve agricultural sustainability and reduce the quality and safety risks of agricultural products in China when regulating nonstandard pesticide-application behavior, it needs not only to strengthen supervision to prevent the occurrence of intentional misconduct but also to provide education to improve farmers’ knowledge and, thus, correct their unintentional wrong pesticide-application behaviors. Based on these findings, we provide the following implications and conclusions.

First, it is necessary to improve farmers’ cognition. The survey shows that Chinese farmers have inadequate cognition about the quality and safety of agricultural products, and many farmers do not know the varieties of banned pesticides, as well as the PHI. At the same time, some farmers believe that pesticide application with a standard dosage will decrease yields, which leads to nonstandard pesticide-application behavior to some extent and impacts the quality and safety of agricultural products in China. Thus, the farmers’ cognition should be improved through clear labeling, education, or training.

Second, it is vital to make farmers aware of food safety incidents. The survey indicates that some farmers have never heard about the incidents about the quality and safety of agricultural products in China, which may affect their cognition of the serious consequences of nonstandard pesticide application. Our results indicate that awareness of such incidents significantly affects farmers’ pesticide-application behavior. Therefore, making incident information available to farmers can help improve farmers’ pesticide-application behavior and agricultural sustainability.

Finally, the risk from agriculture should be further reduced. Farmers are facing relatively high risks from an unstable yield and they need to bear most of the risks. Considering these risks can lead to nonstandard pesticide-application behavior, it is necessary to reduce the risk of agricultural production through agricultural insurance.