The Influence of Regional Specialization in Agriculture on Farmers’ Pest Control Behaviors Based on a Dual Examination of Control Strategies and Control Costs

Abstract

The pest control behaviors of agricultural producers are crucial not only for effectively managing pests and diseases but also for ensuring the sustainable development of agricultural production and the environment. Regional specialization in agriculture, as a typical method to optimize planting structure, influences farmers’ control behaviors through dual mechanisms. However, existing research has largely overlooked this issue. This paper systematically examines the influence and mechanisms of agricultural regional specialization on farmers’ pest control strategies and costs. Using village tracking research data and farmer data from the Jiangsu, Sichuan, and Jilin provinces of China over 15 years, the study employs fixed-effects models and the instrumental variable method to provide empirical evidence. The research results indicate that by reducing biological diversity, which exacerbates the occurrence and damage of pests and diseases, and increasing the homogeneity of planting structures, the development of regional specialization in agriculture prompts farmers to shift their control strategy from “ex post treatment” to “ex ante prevention” and has increased the control costs. The policy implication is that farmers should be encouraged to practice moderate crop rotation in rapidly increasing regional specialization. Additionally, strengthening real-time monitoring of pest and disease dynamics is essential to guide farmers in early prevention and timely control.

1. Introduction

Pests and diseases are one of the most important causes of agricultural production losses, and globally about 20–40% of crop production losses are caused by pests, diseases and phytophagous animals [1,2]. In China, the frequent and severe outbreaks of crop pests and diseases have resulted not only in substantial economic losses but also in threats to national food security [3,4]. The longstanding battle between humans and pests and diseases has shown that scientific and effective control measures are crucial for maintaining agricultural production, increasing farmers’ income, reducing losses, and enhancing efficiency [5,6], which can offer a competitive advantage in agriculture [7]. Nonetheless, irrational use of control methods, such as excessive pesticide application, can worsen agricultural non-point source pollution, harm ecological environments, disrupt stable agricultural production, compromise the quality and safety of agricultural products, and threaten the sustainable development of agricultural resources and the ecological environment of the country [8].

Inappropriate control behaviors for pests and diseases among farmers are prevalent worldwide, with pesticide overuse being the most severe issue [9,10]. As a traditional agricultural country, China continues to face serious crop pest and disease challenges, and producers primarily rely on pesticides for control [11,12,13]. This reliance has resulted in China’s pesticide usage exceeding the global average during the same period, making it a significant contributor to the global increase in pesticide usage [14,15]. In this context, the Chinese government has established various regulations to standardize and guide producers’ pest and disease control behaviors, striving to promote green development of the agricultural sector. China has issued several key documents, such as the “Opinions on Innovative Institutional Mechanisms to Promote Green Development of the Agricultural Sector”, the “Rural Revitalization Strategy (2018–2022)”, the “National Plan for Green Agricultural Development in the 14th Five-Year Plan”, and the “Action Plan for Reducing the Quantity of Chemical Pesticides by 2025”. These initiatives aim to address the high consumption of agricultural production inputs and to continue promoting the reduction and increased efficiency of chemical fertilizers and pesticides. With the proactive guidance of policies, China’s pesticide usage has been steadily decreasing overall [16], and green pest control technologies have gradually developed [17]. Unlike traditional methods that rely on chemical agents to eliminate pests and diseases, green pest control technology is a comprehensive system that integrates measures such as integrated ecological regulation, agronomic cultivation practices, biological control, physical control, and emergency precision chemical control [18]. For major grain crops in China, the current green pest control technologies for corn primarily include the soybean–corn strip intercropping technique [19], biological control of trichogramma and beauveria bassiana, pheromone trap control, etc. [20]. The green pest control technologies for rice mainly include isolated seedling cultivation, breeding resistant varieties, promoting the reproduction of natural enemies, etc. [21]. It is worth noting that the adoption rate of green pest control technologies in China is still relatively low [22,23,24], and the use of pesticides remains the primary method for farmers to control pests and diseases [11,12,13]. Additionally, farmers have a low adoption rate of new, simplified application methods that help reduce pesticide usage. According to the statistical data from this research group’s survey in three provinces in China, from 2003 to 2018, the proportion of crop protection from pests and diseases using machinery for corn increased from 10.55% to 32.37%, and the proportion for rice rose from 11.80% to 35.33%. However, in 2018, the proportion of drone usage for crop protection in corn production in sample villages across the three provinces was only 6.32%, and for rice, this proportion was just 3.27%. The slow development of green pest control technologies has not fundamentally changed farmers’ strong reliance on pesticides. The amount of pesticide used in China still significantly exceeds the economically optimal level [25]. Consequently, it is crucial to further examine the economic rationale behind farmers’ pest control behaviors. Regulating and guiding farmers’ control measures and strategy choices is essential for completing pest control tasks with minimal harm to the ecology. Through these efforts, China can strive to implement the concept of green agricultural development and achieve agricultural modernization transformation.

The behavior of agricultural producers in controlling pests and diseases and the underlying logic have consistently attracted academic attention. Existing studies indicate that farmers’ pesticide application behaviors, including application intensity, types and brands, are influenced by external environmental factors like climate change, natural disasters, and public policies [26,27]. These behaviors are also affected by personal characteristics such as farmers’ risk preferences [28,29] and by other agricultural factors, as well as agricultural organization and management methods [30,31,32]. Additionally, scholars have noted that crop diversity within agroecosystems influences agrochemical inputs [33]. Many existing studies agree that the diversified planting structures provide stability for agroecosystems. An increase in monoculture within these systems raises the risk of pests and diseases, leading to higher inputs of chemical fertilizers and pesticides. This escalation can result in significant issues, such as increased agricultural surface pollution and reduced agroecological efficiency [34,35,36].

As a typical method of adjusting planting structures, the development of regional specialization in agriculture will enable ecosystems to be artificially and purposefully selected and modified. This process affects farmers’ pest control behaviors by altering the characteristics of pests and diseases and changing the plant protection conditions. Agricultural regional specialization involves a regional division of labor, transitioning from traditional dispersed agriculture to more centralized agricultural production models [37]. This transformation is influenced by ecological, technological, and socio-economic factors. Within this framework, agricultural producers focus their resources on producing specific agricultural products that have comparative advantages. This approach has been shown to lead to economies of scale and agglomeration effects [38,39], thereby promoting agricultural economic development and transforming traditional agriculture into modern agriculture [40]. Since the early years of its founding, China has been exploring paths to develop regional specialization through agricultural zoning. Driven by strong policy support, the spatial layout of China’s agricultural production has witnessed gradual and significant changes. The geographic agglomeration of crop production has intensified, making the specialized division of labor in agriculture more pronounced [41].

Existing studies have made valuable explorations into farmers’ behaviors in pest control and its underlying logic, yet several deficiencies remain. Firstly, while the literature describes changes in agricultural factor inputs due to planting structure alterations, it lacks the theological analysis of farmers’ adaptive behavioral choices based on farmer behavior theory, resulting in insufficient theoretical depth. Secondly, current studies focus on changes in pesticide application behavior from the perspective of biological diversity reduction, neglecting the objective law that the development of regional specialization in agriculture will affect the efficiency of agricultural factor output through the expansion of economies of scale, leading farmers to change their factor inputs and production methods. This oversight leads to an inadequate analysis of the impact mechanism. Thirdly, when examining the influence of planting structure on farmers’ control behavior (i.e., factor inputs), existing research either employs controlled experiments for specific crops in natural sciences or relies on provincial or national data, which hinders an objective reflection and testing of the micro-influence mechanism. Lastly, current measurements of regional specialization in agriculture overlook the unique aspects of agricultural production, namely double cropping. Mixing crops at different maturity stages within a year has clearly resulted in measurement bias. Consequently, there is a need to enhance both the accuracy and generalizability of the results in related studies.

This study focuses on the influence of agricultural regional specialization development on farmers’ pest control behaviors, within the context of China’s efforts to optimize and adjust agricultural productivity layout and vigorously promote the green development of the agricultural sector. Firstly, using farmer behavior theory, it explores how regional specialization affects farmers’ pest control strategies and clarifies the economic logic behind these changes. Secondly, it examines the effects and mechanisms of specialization development on control costs for pests and diseases. Finally, theoretical analyses are empirically tested using panel data from four time periods (2003, 2008, 2013, and 2018) of 285 villages across Jiangsu, Sichuan, and Jilin provinces of China and production data of 975 farm households in these villages. Sample areas exhibit significant differences in agricultural development. The fixed-effects model and instrumental variable method are employed for this analysis. This exploration aims to deepen the understanding of farmers’ behavioral logic, guiding them toward effective pest control practices in agriculture for achieving green production. Additionally, the findings enrich interpretations of related behavioral theories and offer theoretical references for similar adaptive behaviors among farmers.

This paper contributes to existing research in three key areas. First, it examines the impact of agricultural regional specialization on farmers’ pest control behaviors, offering a forward-looking analysis grounded in the theory of farmers’ behavior. Second, it simultaneously incorporates both farmers’ pest control strategies and costs into the analytical framework, aiming to elucidate multiple mechanisms such as the reduction in biological (crop) diversity and the improvement in production efficiency. Third, it innovatively includes agricultural double cropping in the measurement indicators, addressing measurement bias in existing studies and providing robust data and technical support to ensure the accuracy and reliability of the estimation results.

2. Theoretical Analysis and Research Hypotheses

Agricultural pests and diseases refer to the combination of crop diseases and pest infestations in agricultural production. This term describes the adverse effects harmful organisms or unfavorable environmental conditions have on crops. These effects interfere with normal metabolism of crops, leading to a series of changes from physiological functions to tissue structure, manifested in abnormal morphological symptoms like wilting and rot. As China’s major grain crops, rice and corn are significantly negatively affected in terms of both yield and quality due to pests and diseases [42,43], which have been aggravated in recent years [44]. In the process of corn production in China, common pests include the Asian corn borer, corn locust, corn weevil, and leaf mite; common diseases include gray mildew, stalk rot, and brown spot disease. For rice, prevalent pests include brown planthopper, rice leafroller, and rice borer; common diseases are rice sheath blight and rice blast. Due to differences in accumulated temperatures and hydrothermal conditions, the predominant types of pests and diseases vary across regions. For instance, rice sheath blight is more common in the warmer and more humid rice-growing areas of southern China. These pests and diseases cause tremendous economic losses and also threaten national food security [45]. Thus, finding and implementing efficient pest and disease control methods have always been important demands for agricultural producers.

Regional specialization in agriculture is one of the important factors influencing farmers’ pest and disease control behaviors. The development of agricultural regional specialization is the process of purposeful selection and modification of the agricultural ecosystem by humans. At its core, it represents the deepening of the division of labor in agricultural production, embodying the theory of comparative advantage in this field. While higher levels of specialization bring economic benefits like economies of scale and agglomeration effects, they also lead to negative consequences such as reduced biodiversity in the region [46]. Consequently, the development of regional specialization in agriculture influences farmers’ pest control behaviors and associated costs by altering biological relationships within agroecosystems and prompting shifts in the allocation of production factors. Given that pesticide application remains the primary method of pest control among agricultural producers in China, this study will analyze and examine how the development of agricultural regional specialization within village areas impacts farmers’ pesticide application strategies and costs during food production. The mechanism of influence is illustrated in Figure 1.

2.1. Influence of Regional Specialization in Agriculture on Farmers’ Pest Control Strategies

The development of regional specialization in agriculture will lead farmers to adjust their pest control strategies through two main mechanisms, resulting in increased pesticide use for preventive purposes before pests and diseases occur. First, alterations in the characteristics and severity of pest and disease outbreaks will enhance the expected return of preventive measures, thereby boosting farmers’ motivation to control pests and diseases proactively. Second, a more uniform planting structure within the region will reduce the various obstacles that farmers face in applying pesticides preemptively.

At this stage in China, pesticide application is still the primary method of controlling pests and diseases. Farmers’ strategies for pest control can be categorized into “ex ante prevention” and “ex post treatment” based on the timing relative to pest and disease outbreaks. “Ex ante prevention” involves applying pesticides before pests and diseases occur, guided by early warning forecasts from agricultural authorities and affiliated plant protection organizations. The main purpose of this approach is to eliminate larvae or disease foci before large-scale, high-intensity outbreaks occur. On the other hand, “ex post treatment” involves applying pesticides after a pest or disease has appeared, aiming to contain the damage and recover yield losses.

2.1.1. Increasing the Expected Returns from Ex Ante Preventive Pesticide Application

The core of farmer behavior theory is rational choice. This involves making optimal decisions through cost–benefit analysis of various response behaviors with the aim of maximizing expected returns while considering both personal and external constraints [47,48]. The development of agricultural regional specialization can influence the characteristics of pests and diseases, leading to more frequent outbreaks and increased damage, which translates to greater crop yield losses. With this in mind, and drawing from Meijia Li and Guanghua Lin’s (2023) analysis of farmers’ response behavior based on disaster experience and loss controllability [49], a decision framework is constructed. This framework assesses how the development of agricultural regional specialization impacts farmers’ control strategies for pests and diseases.

Farmers have two primary strategies for dealing with pests and diseases: “ex ante prevention” and “ex post treatment”. Each strategy presents two possible scenarios: Scenario 1, where pests and diseases occur with probability $p$, and Scenario 2, where they do not occur with probability $(1-p)$. According to the expected utility theory in economics, farmers will evaluate and compare the subjective expected returns of these four scenarios under the two strategies. Assuming all other conditions remain constant, farmers will opt to apply pesticides in advance to prevent pests and diseases only if the subjective expected net benefit of doing so exceeds that of applying pesticides for treatment after pests and diseases have occurred.

When addressing pest and disease challenges, farmers rely on their control experience and available information, such as weather forecasts and agricultural notices, to make subjective judgments about the likelihood and severity of different outbreaks. Let us denote the subjectively assessed probability of pest and disease outbreaks as $p(o_1,o_2)$ in Scenario 1, where $o_1$ is the frequency of past occurrences ever faced by farmers, and $o_2$ represents the damage level, reflected in the crop yield reduction proportion. Here, $o_1$, $o_2$ is an increasing function of $p(o_1,o_2)$, which means $p(o_1,o_2)$ increases with higher $o_1$ and $o_2$ values. In contrast, the likelihood of no outbreak, as judged by farmers, is $1-p(o_1,o_2)$ in Scenario 2. Assuming the farmer’s operational scale is m and productivity remains constant, $f\left(m\right)$ represents the annual output in a normal year, applicable to Scenario 2. If pests and diseases occur with probability $p(o_1,o_2)$, the perceived loss degree by farmers is $ol(o_1,o_2)$, satisfying $ol(o_1,o_2)<1$ and its increasing function of $o_1$ and $o_2$. Therefore, the final output in Scenario 1 is $f\left(m\right)(1-ol(o_1,o_2\left)\right)$. Additionally, assume market prices remain stable with the agricultural product pricing at $P_0$.

Years of agricultural practice in China have shown that while the misuse of pesticides and other agrochemicals can significantly harm human health and the environment, their judicious use plays a crucial role in controlling pest and disease issues and mitigating crop yield losses [6,50]. Building on this understanding, it is assumed that preventive pesticide application before the onset of pests and diseases can reduce yield loss with an impairment factor $s_1(e_1,e_2)$, where $s_1(e_1,e_2)<1$. Similarly, remedial treatment application after the occurrence has an impairment factor $s_2(e_1,e_2)$, where $s_2\left(e_1,e_2\right)<1$. Here, $e_1$ refers to different types of pests and diseases, while $e_2$ refers to the various types of pesticides and application methods used by farmers.

Combined with the stated assumptions, the subjective expected returns to farmers facing pests and diseases for ex ante prevention can be expressed as follows:

$$p\left(o_1,o_2\right)P_{0}f\left(m\right)\left(1-s_1\left(e_1,e_2\right)ol\left(o_1,o_2\right)\right)+(1-p\left(o_1,o_2\right))P_{0}f\left(m\right)$$

(1)

The subjective expected return of a farmer facing a pest or disease for ex post treatment can be expressed as

$$p\left(o_1,o_2\right)P_{0}f\left(m\right)\left(1-s_2\left(e_1,e_2\right)ol\left(o_1,o_2\right)\right)+\left(1-p\left(o_1,o_2\right)\right)P_{0}f\left(m\right)$$

(2)

Farmers need to pay costs for controlling pests and diseases both ex ante and ex post. Assuming that $r\left(e_1,e_2\right)$ is the total cost for controlling pests and diseases by farmers, then it can be expressed as

$$c=r\left(e_1,e_2\right)m$$

(3)

In summary, the subjective expected net benefit of preventive pesticide application by farmers before the occurrence of pests and diseases can be expressed as the difference between Equations (1)–(3), and then the equilibrium condition is

$$\pi =p\left(o_1,o_2\right)P_{0}f\left(m\right)\left(s_1\left(e_1,e_2\right)-s_2\left(e_1,e_2\right)\right)ol\left(o_1,o_2\right)-r\left(e_1,e_2\right)m=0$$

(4)

From Equation (4), it can be inferred that the behavioral decisions made by farmers on ex ante preventive or ex post treatment in the face of pest and disease problems will be influenced by the following two factors:

The first factor is the characteristics of pest and disease outbreaks, namely $o_1,o_2$. The frequency of pests and diseases is represented by $o_1$ and the degree of damage (yield reduction ratio) is represented by $o_2$. The relationship between the two and the farmers’ control strategy (i.e., whether farmers choose to dose beforehand or afterward) is $\partial \pi /\partial o_1>0$, indicating that the more frequent the pests and diseases are, the more farmers tend to apply pesticides in advance to prevent pests and diseases; $\partial \pi /\partial o_2>0$ indicates that the more serious the damage degree of the pests and diseases is, the more farmers tend to apply pesticides in advance for prevention.

The second factor is the difference in loss recovery between ex ante prevention and ex post treatment for pests and diseases, namely $s_1\left(e_1,e_2\right)-s_2\left(e_1,e_2\right)$. Since there are many types of crops and the types of diseases and pests are also complex and diverse, as well as the fact that the optimal control time for different types of diseases and pests has differences, it is impossible to generalize the specific loss-recovery differences between ex ante prevention and ex post treatment of pesticide application. However, it has been found that in the context of increased homogeneity of crop species in the region, compared with the diversified planting structure, a more homogeneous planting structure will lead to the occurrence of pests and diseases with a heavy degree of damage and fast spreading characteristics [51]. Therefore, at this time, once the pests and diseases occur and form a scale, this will result in more serious agricultural losses, and the economic costs and time of treatment and rescue are also higher. Therefore, farmers will prefer “prevention” rather than “treatment”, i.e., farmers will tend to carry out more preventive spraying of pesticides before the occurrence of pests and diseases.

2.1.2. Mitigating Operational Barriers to Ex Ante Preventive Pesticide Application

Before the full development of regional specialization in agriculture, farmers face three unfavorable factors that hinder the advanced prevention of pests and diseases. First, farmers’ application behavior has externalities. When farmers cultivate multiple crops, pesticides used for one crop’s pests or diseases may harm non-target crops [52], potentially reducing farm income and causing disputes among farmers. Second, under a diversified planting structure, farmers must prepare and apply pesticides separately for each crop at different times, leading to high economic and labor costs for preventive pesticide application. Third, compared to monoculture, a diversified planting structure reduces the availability of outsourcing services for pest control and results in higher service fees if farmers opt for outsourced preventive plant protection services.

When regional specialization in agriculture is developed, the planting structures within the region become more homogeneous, and the three main disadvantages above are mitigated to varying degrees. First, the negative externality of pesticide application on target crops, which can affect the growth and development of other crops, is reduced under a homogenized cropping pattern. Second, regions with more homogeneous planting structures are more likely to establish unified plant protection operations, significantly lowering the economic costs and time associated with preparation and implementation, which is particularly beneficial for operations using agricultural machinery. Finally, a homogeneous planting structure facilitates collective action among farmers, enhancing their bargaining power when purchasing outsourced services.

Based on the above theoretical analysis, the following research hypothesis is proposed:

H1. 

The development of regional specialization in agriculture will lead to a shift in farmers’ control strategies for pests and diseases, and farmers’ ex ante pesticide application for preventive purposes will increase.

2.2. Influence of Regional Specialization in Agriculture on Farmers’ Pest Control Costs

The development of regional specialization in agriculture affects the cost of pest and disease control for farmers through two main mechanisms. First, it alters the characteristics and severity of pest and disease outbreaks. Second, it increases the homogeneity of planting structures within the region, allowing for more efficient and uniform plant protection operations.

First, existing studies have shown that the decline in biological diversity within farmland ecosystems contributes to ecological imbalances, leading to more frequent occurrences and severe damage of pests and diseases [53,54,55]. This increases farmers’ need for plant protection and, consequently, their pesticide use. Second, a more homogenized planting structure in the region allows for greater consistency in planting times and the quality of planting operations, which enhances control effectiveness. This consistency can significantly reduce the recurrence of pests and diseases and the over-application of pesticides caused by inconsistent plant protection practices [56,57]. As a result, it decreases the repeated use of chemicals by farmers and lowers their pest control costs.

Based on the above theoretical analysis, the following research hypothesis is proposed:

H2. 

Regional specialization in agriculture alters the characteristics of pests and diseases, which may lead farmers to increase pesticide application. However, it also facilitates the unified operation of the plant protection chain, potentially reducing the need for repeated pesticide applications. The ultimate impact on the cost of pest and disease control for farmers remains an empirical question.

3. Econometric Model, Data, and Variables

3.1. Model

3.1.1. Baseline Test

The following fixed-effects model was set up to test the effect of regional specialization in agriculture on farmers’ strategies and costs for pest control:

$${Pesfoc}_{it}={\alpha }_0+{\alpha }_1{ARS}_{it-1}+{\theta }_1{X}_{it}+{\mu }_i+v_t+{\epsilon }_{it}$$

(5)

$${Pescid}_{it}={\beta }_0+{\beta }_1{ARS}_{it-1}+{\theta }_2{X}_{it}+{\mu }_i+v_t+{\epsilon }_{it}$$

(6)

The explanatory variable ${Pesfoc}_{it}$ in Equation (5) is the proportion of the number of pesticides applied by farmers to cope with pests and diseases for the purpose of ex ante prevention in the total number of pesticide applications in the village $i$ planting the main food crops (corn or rice) in period $t$, in order to determine if the focus of pesticide application by farmers in the management of pests and diseases is to prevent the pests and diseases before their occurrence or to treat them after their occur. The explanatory variable ${Pescid}_{it}$ in Equation (6) is the total number of pesticides applied by farmers to deal with pests and diseases in village $i$ when they planted the main food crops (corn or rice) in period $t$, which includes both preventive pesticide application and post-treatment pesticide application.

The key explanatory variable in Equations (5) and (6) is the agricultural regional specialization index ${ARS}_{it-1}$ for village $i$ in period $t-1$. This index represents the extent of specialization in crop types within agricultural production in the sample village. Drawing on previous research methods, we use the Herfindahl–Hirschman Index (HHI), derived from industrial economics for measuring industrial concentration, alongside the Shannon–Wiener Information Index (Shannon), an ecological measure of biological diversity, to assess the regional specialization in agriculture within the village [46,58,59]. Additionally, because the selection of varieties and the development degree of agricultural regional specialization can be influenced by the characteristics of pests and diseases and the effectiveness of control, there exists a theoretical endogeneity issue. Consequently, using the lag of one period for the key explanatory variables can mitigate endogeneity’s impact on parameter estimation to some extent.

It is important to note that previous scholars’ methods for measuring the development of agricultural regional specialization often overlook the unique issue of double cropping in agriculture. The double or triple cropping index is an important concept of agricultural cropping systems, which can be understood as the degree and type of multiple tillage in a given area during a year. However, existing studies tend to blend crops with different maturity periods within a single year, resulting in measurement bias and biased empirical estimates. To address this problem, our research introduces an innovative approach by incorporating double cropping into the index. Specifically, we calculate the degree of regional specialization for crops based on their maturity periods and then aggregate these calculations using the weighted proportion of the sown area. This advancement in measurement makes a significant marginal contribution.

The innovative Herfindahl Index (FC_HHI), considering double cropping, is shown below:

$$\begin{gathered} FC\_HHI=\left(X_S/X\right)\times {\sum }_{i=1}^n{\left(X_{Si}/X_S\right)}^2+\left(X_A/X\right)\times {\sum }_{i=1}^n{\left(X_{Ai}/X_A\right)}^2 \\ =\left(X_S/X\right)\sum _{i=1}^n{S_{Si}}^2+\left(X_A/X\right)\sum _{i=1}^n{S_{Ai}}^2 \end{gathered}$$

(7)

In Equation (7), $X$ represents the total sown area for all crop types, including both summer and autumn maturity periods. Specifically, $X_S$ refers to the total sown area of crops with summer maturity, while $X_A$ pertains to the total sown area of crops with autumn maturity. $X_{Si}$ represents the sown area of the $i\_th$ crop with summer maturity, and $X_{Ai}$ represents the sown area of the $i\_th$ crop with autumn maturity. Therefore, $S_{Si}$ indicates the proportion of the sown area of the $i\_th$ crop for summer maturity relative to the village’s total sown area for that season. Similarly, $S_{Ai}$ denotes the proportion of the sown area of the $i\_th$ crop for autumn maturity relative to the total for the village in autumn. The variable $n$ is the number of distinct crops sown in the village. The FC_HHI ranges between (0,1], where a larger value signifies a more homogeneous planting structure, indicating a higher degree of specialization.

The innovative Shannon Information Index (FC_Shannon), considering double cropping, is shown below:

$$FC\_Shannon=-[\left({\displaystyle \frac{X_S}{X}}\right)\times \sum _{i=1}^n{P}_{Si}\times ln{P}_{Si}+\left({\displaystyle \frac{X_A}{X}}\right)\times \sum _{i=1}^n{P}_{Ai}\times ln{P}_{Ai}]$$

(8)

In Equation (8), $X$, $X_S$, $X_A$, $n$ have the same meaning as in Equation (7). The variable $P_{Si}$ represents the proportion of the $i\_th$ crop’s sown area relative to the total for summer maturity, whereas $P_{Ai}$ represents the proportion for autumn maturity; the stipulation In0 = 0 is applied here. To align with the calculation direction of the innovative Herfindahl Index (FC_HHI), the measure is taken as negative. Therefore, a larger FC_Shannon value indicates a higher degree of regional specialization in agricultural production among the sample villages, reflecting a more homogeneous planting structure within the village area.

In addition to agricultural regional specialization, farmers’ pest and disease control behaviors are influenced by a combination of natural and social factors, due to the dual natural and social attributes of agriculture. In the existing research, a series of analyses have been conducted on this issue. Existing studies point out that the behavior of pesticides in the soil is governed by a variety of complex processes, including climatic conditions [60,61]. Therefore, natural factors such as temperature and precipitation directly affect the effectiveness of chemicals in pesticides in suppressing pests and diseases in the farmland environment. When high temperatures lead to excessive transpiration and heavy rain causes strong wash-off effects, the effectiveness of pesticides is reduced. Consequently, rational farmers will choose the appropriate time to apply pesticides based on weather conditions or increase the frequency of application in adverse climatic conditions to ensure that the chemicals used to kill pests and diseases are sufficient and effective. Existing studies have indicated that factors such as farmers’ gender, education level [62], risk preferences [63], training [64], household population [62], and household income [65], as well as the external market environment and policies [66], all have varying degrees of influence on farmers’ pest and disease control behaviors.

Based on the above discussion, this paper selects a series of factors that may affect farmers’ pesticide application as control variables $X_{it}$ in this study. For the model in Equations (5) and (6), these include village characteristics such as the variety of grain crops and agricultural labor structure within the village, as well as external environmental factors like temperature, precipitation, and sunlight duration. The household and individual circumstances of farmers are controlled in Equation (11), estimated using farmers’ household data. The term ${\mu }_i$ represents unobservable factors that account for village-level characteristics that remain constant over time but differ between villages. $V_t$ captures temporal effects that are consistent across regions, while ${\epsilon }_{it}$ is the idiosyncratic perturbation term that accounts for unobservable factors at the village level that vary over time. The parameters $\alpha$, $\beta$, and $\theta$ are to be estimated in the model.

3.1.2. Mechanism Test

In order to verify the mechanism that the development of agricultural regional specialization influences the control behavior of farmers by exacerbating the frequency and damage of pests and diseases, the following model was set up for testing:

$${OCC}_{it}={\gamma }_0+{\gamma }_1{ARS}_{it-1}+{\theta }_3{X}_{it}+{\mu }_i+v_t+{\epsilon }_{it}$$

(9)

$${Loss}_{it}={\delta }_0+{\delta }_1{ARS}_{it-1}+{\theta }_4{X}_{it}+{\mu }_i+v_t+{\epsilon }_{it}$$

(10)

The explanatory variables include the frequency of pests and diseases and the level of damage, respectively. They are proxied by the frequency of pests and diseases (${OCC}_{it}$) and the proportion of food production reduction caused (${Loss}_{it}$) when the main food crop (corn or rice) was grown in village $i$ in period $t$, respectively. The key explanatory variable remains the village agricultural regional specialization index lagged one period to control potential endogeneity. Control variables include village characteristics and external climatic factors that influence the pests and diseases. The meanings of ${\mu }_i$, $V_t$, ${\epsilon }_{it}$ remain the same as in Equation (5). The parameters $\gamma$, $\delta$, and $\theta$ are to be estimated in the model.

3.1.3. Robustness Test

We used two methods for assessing the robustness of the empirical results.

1. Test using farm household data. We used farm household data and constructed the following cross-sectional data model to test the influence of the development of agricultural regional specialization in a village on the pesticide application behavior of individual farmers to control pests and diseases:

$${ipcid}_i={\vartheta }_0+{\vartheta }_1{ARS}_i+{\theta }_5{X}_i+{\epsilon }_i$$

(11)

$${icost}_i={\tau }_0+{\tau }_1{ARS}_i+{\theta }_6{X}_i+{\epsilon }_i$$

(12)

The first explanatory variable $i{pcid}_i$ is the number of pesticide applications carried out by the individual farmer on the largest plot operated by the $i\_th$ farmer growing the main food crop (corn or rice) in 2018. The second explanatory variable $i{cost}_i$ is the average pesticide cost per mu (unit of area equal one fifteenth of a hectare) for pest control by the $i\_th$ farmer on the same plot. The key explanatory variable ${ARS}_i$ represents the level of agricultural regional specialization within the village where the $i\_th$ farmer located in 2018. Based on the discussion of control variables in Section 3.1.1, this section selects a series of characteristic variables at different levels affecting each farmer’s pesticide use, denoted $X_i$. These variables include personal attributes of individual farmers (such as gender, age, education level, and training), household characteristics (e.g., number of laborers in the household, proportion of agricultural labor force), plot features (like area, slope, and soil quality), and village attributes (for instance, the level of economic development and clan power) and include climatic factors like temperature, precipitation, and hours of sunshine that influence pest management. ${\epsilon }_i$ is the residual term, while $\vartheta$, $\tau$ and $\theta$ are the parameters to be estimated in the model.

2. Test using the instrumental variable method (IV). To further control for the endogeneity of the key explanatory variable “regional specialization in agriculture”, the instrumental variable method is employed using “the level of agricultural regional specialization of other villages in the township (town) except the sample village” as the instrumental variable (IV) for key explanatory variables of the sample village. The two-stage least squares method is then used for estimation. This variable satisfies the two necessary conditions for being an instrumental variable: relevance and exogeneity. Regarding relevance, due to the influence of climate, hydrology, and other natural factors, agricultural production exhibits geographical similarity, and the planting structure has spatial spillover effects [67]. Regarding exogeneity, the choice of pest and disease control behavior is determined by the farmers’ production mode based on the actual agricultural production of the village and individual characteristics. The cultivation structure of other villages within the township cannot directly affect the control behavior of farmers in the sample village. The following models (13), (14), (15) are still estimated using village survey data.

In the first-stage regression of the two-stage least squares method (2SLS), the instrumental variable ${IV}_{it}$, which satisfies both relevance and exogeneity, can isolate the exogenous part of the endogenous variable ${ARS}_{it}$, as shown in Equation (13). In the second stage, the fitted values $\widehat{{ARS}_{it}}$ obtained from Equation (13) are substituted into Equations (14) and (15) for instrumental variable regression. At this point, consistent estimators for the coefficients $\omega$ and $\pi$ of the endogenous variable ${ARS}_{it}$ can be obtained. The meanings of ${\mu }_i$, $V_t$, ${\epsilon }_{it}$ remain the same as in Equation (5), while $\rho$, $\omega$, $\pi$ and $\theta$ are the parameters to be estimated in the model.

$${ARS}_{it}={\rho }_0+{\rho }_1{IV}_{it}+{\theta }_7{X}_{it}+{\mu }_i+v_t+{\epsilon }_{it}$$

(13)

$${Pesfoc}_{it}={\omega }_0+{\omega }_1\widehat{{ARS}_{it}}+{\theta }_8{X}_{it}+{\mu }_i+v_t+{\epsilon }_{it}$$

(14)

$${Pescid}_{it}={\pi }_0+{\pi }_1\widehat{{ARS}_{it}}+{\theta }_9{X}_{it}+{\mu }_i+v_t+{\epsilon }_{it}$$

(15)

3.2. Data and Variables

The data used in the analysis of this study came from two sections:

The first dataset is a four-period balanced panel dataset spanning 15 years (2003, 2008, 2013, and 2018), derived from a continuous tracking survey of 285 villages in the provinces of Jiangsu, Jilin, and Sichuan in China. This dataset includes a total of 1140 village-level samples and 975 farmer samples obtained from production surveys conducted by the research group involved in this study. The study contains detailed information on agricultural cultivation in the sampled villages, incorporating data dating back to 1997. It also includes comprehensive details on agricultural pests and diseases, such as the number of occurrences, the grain yield, and pesticide usage for prevention and treatment. Additionally, the dataset captures various village characteristics, including land resource endowment, topography, agricultural labor force, and geographical location.

The three sample provinces cover a wide geographical area and are diverse and representative. Jiangsu Province is densely populated, with well-developed non-agricultural industries and a robust market for land transfer and agricultural machinery services. Jilin Province, as one of China’s major agricultural production areas, boasts abundant per capita arable land and other agricultural resources. In contrast, Sichuan Province faces a scarcity of arable land, is predominantly hilly and mountainous, exports more labor, and has an underdeveloped economy. Additionally, the long-term tracking survey effectively captures dynamic changes in the villages and helps mitigate the challenge of observing small, gradual changes caused by the inertia of agricultural practices.

The second dataset consists of daily values from the Chinese Surface Climate Data (V3.0). The occurrence and management of pests and diseases are strongly influenced by regional climatic conditions [68,69,70]. Thus, this study utilized the daily climate data collated by the Climatic Data Center, National Meteorological Information Center, China Meteorological Administration. In this study, climate data from daily values were matched to the county level of the sample village tracking survey data in the three provinces based on the distribution of weather station locations to control for the influence of climate factors on farmers’ pest control behaviors.

Table 1. Descriptive statistics of variables for data of village.

Variables	Definition of Variables	Mean	Std. Dev.	Min	Max
Proportion of preventive behaviors	Proportion for preventive purposes out of the total number of pesticide applications (%)	55.92	35	0	100
Total number of pesticide applications	Total number of pesticide applications (including ex ante prevention and ex post treatment)	3.17	2.03	0	12
FC_HHI	HHI for double cropping	0.63	0.24	0.17	1
HHI	HHI for mixed	0.47	0.25	0.13	1
FC_Shannon	Shannon for double cropping	−0.69	0.46	−1.85	0
Shannon	Shannon for mixed	−1.06	0.56	−2.18	0
Percentage of early-maturing varieties	Percentage of area sown with early maturing grain varieties (%)	33.59	34.29	0	100
Double cropping index 1	The total area sown with crops at each maturity period divided by cultivated area	1.56	0.59	0.72	3.92
Number of main grain species	Number of varieties of corn or rice grown in the sample villages	7.28	8.35	1	55
Agronomic harmonization index	Discrete coefficients at the commune (township) level for agricultural practices used in growing major food crops	0.58	1.25	0	7.35
Soil quality differences	Difference in standing yield between the best and the worst land in the sample villages for growing corn or rice (kg/mu)	2.99	2.54	0	30
Percentage of sloping arable land	Proportion of cultivated land area with slope of 15° and above to total cultivated land area (%)	22.63	22.93	0	100
Aging of agricultural workforce	Percentage of the agricultural workforce over 60 years old (%)	32.21	19.49	0	95
Feminization of agricultural workforce	Percentage of women in the village agricultural workforce (%)	44.65	13.60	0	80
Percentage of non-farm employment	Percentage of non-farm employment to total employment in the village (%)	40.18	19.60	0	95.45
Power of clan	Proportion of the total population of the village owned by the most owners of the same family name in the village (%)	20.46	17.16	0.50	95
Location condition 1	The distance from the village council of the sample village to the county government (km)	27.88	21	0.50	200
Location condition 2	The distance from the village council of the sample village to the entrance of the nearest motorway (km)	28.85	29.13	0.10	350
Economic conditions	Number of businesses in the village	1.29	5.38	0	115
Accessibility of factor markets	Number of shops for agricultural supplies	1.07	2.59	0	35
Sunshine hours	Sunshine hours (hundred hours)	18.89	3.81	13.24	24.98
Average temperature	Average annual temperature (°C)	13.12	4.10	5.68	16.97
Annual precipitation	Annual precipitation (m)	9.39	1.89	5.11	13.25
Observations of village		1140

Note: 1. The data were obtained from the thematic tracking survey of villages in Jiangsu, Sichuan, and Jilin provinces; 2. The meteorological data were compiled from the National Meteorological Science Data Center, Beijing, China. ¹ In the sample provinces of this study, except Jilin, which can only grow crops for one season due to its high latitude and insufficient accumulated temperature, Jiangsu and Sichuan all have a double cropping situation.

presents the descriptive statistics of the variables employed in the regression analysis using village data. On average, farmers in the sample villages applied pesticides about 3 times during a food crop growing season. Additionally, 55.92% of pesticide applications were preemptive to prevent the occurrence of pests and diseases.

Table 2. Descriptive statistics of variables for data of farmer households.

Variables	Definition of Variables	Mean	Std. Dev.	Min	Max
Total number of pesticide applications	Total number of pesticide applications on sample farmer’s largest plot when growing the main food crop (including ex ante prevention and ex post treatment)	3.32	2.09	0	12
Pesticide cost per mu	Average pesticide cost per mu spent on the largest plot for pest control (CNY/mu)	177.64	186.30	0	846
FC_HHI	HHI for double cropping	0.68	0.22	0.20	1
FC_Shannon	Shannon for double cropping	−0.68	0.37	−1.45	0
Number of family laborers	Number of laborers in sample farm households (persons)	3.01	1.27	1	8
Proportion of agricultural labor force	Proportion of agricultural laborers in farm households to total labor force (%)	58.12	31.63	0	100
Gender	Sex of the farmer interviewed (1 = male; 0 = female)	0.78	0.41	0	1
Age	Age of the farmer interviewed	55.98	10.75	27	89
Education	Total number of years of education for the farmer interviewed (years)	6.52	3.26	0	15
Physical health	Physical health of the farmer interviewed (1 = incapacitated; 0 = healthy)	0.07	0.26	0	1
Non-agricultural employment	Whether the farmer interviewed was involved in off-farm work (1 = yes; 0 = no)	0.47	0.50	0	1
Agricultural experience	Length of time (years) that the farmer interviewed has been involved in agriculture (specifically farming)	35.53	13.23	0	74
Agricultural training	Number of agricultural technology trainings/lectures attended by the farmer interviewed in the last 3 years (times)	0.63	1.42	0	15
Income	Net income of farm household in 2018 (CNY ten thousand)	5.81	6.23	−4	68.10
Appetite for risk	The amount of pesticide used by the farmer interviewed compared to the notification or package instructions (1 = higher; 2 = about the same; 3 = lower)	2.32	0.93	1	3
Plot size	Size of the largest plot (mu) when the farmer interviewed grows the main food crop (corn or rice)	2.99	16.18	0.05	348
Plot location	Distance (in miles) of the largest plot from the farmer interviewed house	1.46	2.06	0	32
Soil type of the plot	Soil type of largest plot (1 = sandy; 2 = loam; 3 = clay; 4 = other)	2.06	0.91	1	4
Irrigation of the plot	Can the largest plots be irrigated (1 = yes; 0 = no)	0.65	0.48	0	1
Fertility of the plot	How fertile is the largest plot (1 = good; 2 = moderate; 3 = poor)	1.72	0.60	1	3
Slope of the plot	Slope of the largest plot (1 = flat; 2 = sloping; 3 = depressed)	1.28	0.52	1	3
Double cropping of the plot	How many seasons has the largest plot matured for	1.46	0.52	1	5
Varieties of fall crops for the plot	Fall-ripening food crops of the largest plot (1 = rice; 0 = corn)	0.57	0.50	0	1
Natural disasters	Whether the largest plot was affected in 2018 (1 = yes; 0 = no)	0.44	0.50	0	1
Power of clan	Proportion of the total population of the village owned by the most owners of the same family name in the village (%)	19.95	15.51	1	70
Economic conditions	Number of businesses in the village	3.31	12.43	0	115
Accessibility of factor markets	Number of shops for agricultural supplies	1.40	2.15	0	35
Sunshine hours	Sunshine hours (hundred hours)	20.03	3.43	14.83	24.77
Average temperature	Average annual temperature (°C)	12.31	4.36	6.54	16.97
Annual precipitation	Annual precipitation (m)	9.24	1.82	5.75	13.16
Observations of farmer household		975

Note: 1. Data on farm households were collected through interviews with farmers during follow-up surveys conducted in villages across Jiangsu, Sichuan, and Jilin provinces; 2. The meteorological data were compiled from the National Meteorological Science Data Center.

presents the descriptive statistics of the variables employed in the regression analysis using farmer household data. The total pesticide applications performed by the sample farmers interviewed was also about 3 and the average cost of pesticides on one mu of land was 177.64 CNY during a food crop growing season.

4. Results

4.1. Estimated Results of Hausman Test for Model Setting

To address the potential endogeneity between agricultural regional specialization and farmers’ pest control behaviors, this paper employs two strategies. The first strategy involves lagging the key explanatory variable, the “agricultural regional specialization index”, by one period. Since the dataset includes agricultural planting structure information from 1997, this approach does not lead to any loss of samples. Additionally, given that the data are from longitudinal surveys conducted at five-year intervals, using lagged variables over a larger time span is more effective. Finally, endogenous problems can be addressed at the technical level by controlling for other village-level variables, which helps to exclude interfering mechanisms. The second strategy is to use an instrumental variables approach to control endogeneity problems.

According to econometric theory, the testing logic in this paper involves using Hausman test to compare the estimation results of the fixed-effects model (FE-OLS) and the random-effects model (RE-GLS), both of which initially do not address the endogeneity problem of the Agricultural Regional Specialization index. The fixed-effects model primarily relies on within-group dynamic information for parameter estimation, whereas the random-effects model incorporates both within-group dynamics and between-group cross-sectional heterogeneity. However, the random-effects model is less effective than the fixed-effects model in handling endogeneity issues. Subsequently, Hausman test is employed to compare the differences between the OLS estimation results and the instrumental variable estimation results, aiding in the final decision on the appropriate model and estimator [71].

Table 3. Results of Hausman test for model setting.

Differences in Estimates	Explained Variable: Percentage of Pesticide Applications for Prevention	Explained Variable: Total Number of Pesticide Applications
FE-OLS vs. RE-GLS	40.03 ***	39.30 ***
FE-OLS vs. FE-IV	0.00	3.57

Note: *** represents statistical significance at the 1% levels.

presents the Hausman test results for the model set up analyzed using village tracking survey data. The test results indicate that using lagged variables to address endogeneity is effective, and the FE-OLS estimates are unbiased. Consequently, this study will utilize the FE-OLS estimation results to report and discuss the influence of regional specialization in agriculture on farmers’ pest control strategies and costs. Additionally, the estimated results from the instrumental variables method will be presented in the robustness test section.

4.2. Estimated Results of the Influence of Regional Specialization in Agriculture on Farmers’ Control Behaviors for Pests and Diseases

Table 4. Estimated results of the influence of regional specialization in agriculture on farmers’ pest control strategies and costs.

	Proportion of Preventive Behaviors (%)		Total Number of Pesticide Applications (Times)
	FC_HHI	FC_Shannon	FC_HHI	FC_Shannon
FC_HHI (lagging one period)	14.008 ***		1.105 ***
	(3.462)		(4.221)
FC_Shannon (lagging one period)		6.936 ***		0.425 ***
		(3.380)		(3.337)
Control variables
Double cropping index	2.010 *	2.333 **	−0.060	−0.059
	(1.853)	(2.052)	(−0.854)	(−0.835)
Number of main grain species	−0.192	−0.207	−0.001	−0.001
	(−1.184)	(−1.223)	(−0.125)	(−0.114)
Percentage of early-maturing varieties	0.163 **	0.111	−0.003	−0.003
	(2.304)	(1.503)	(−0.731)	(−0.750)
Agronomic harmonization index	0.255	0.226	0.046 **	0.040 *
	(0.806)	(0.688)	(2.254)	(1.946)
Soil quality differences	−0.319	−0.299	0.033	0.033
	(−0.760)	(−0.680)	(1.217)	(1.212)
Percentage of sloping arable land	−0.004	−0.010	0.001	0.001
	(−0.120)	(−0.285)	(0.591)	(0.467)
Aging of agricultural workforce	0.031	0.037	0.001	0.001
	(0.973)	(1.107)	(0.591)	(0.467)
Feminization of agricultural workforce	−0.032	−0.043	0.001	0.001
	(−0.821)	(−1.032)	(0.591)	(0.467)
Percentage of non-farm employment	0.025	0.024	0.001	0.001
	(0.703)	(0.658)	(0.591)	(0.467)
Power of clan	−0.142 ***	−0.120 **	−0.007 **	−0.006 *
	(−2.649)	(−2.138)	(−1.986)	(−1.811)
Location condition 1	0.045	0.067	−0.010 ***	−0.010 ***
	(0.952)	(1.367)	(−3.257)	(−3.219)
Location condition 2	0.012	0.012	0.001	0.001
	(0.633)	(0.627)	(0.545)	(0.702)
Economic conditions	−0.073	−0.094	−0.003	−0.004
	(−0.845)	(−1.048)	(−0.484)	(−0.663)
Accessibility of factor markets	0.014	0.113	0.012	0.011
	(0.056)	(0.441)	(0.780)	(0.709)
Sunshine hours	1.908 *	2.165 **	−0.009	−0.004
	(1.905)	(2.066)	(−0.142)	(−0.056)
Average temperature	−2.023	−2.153	−0.040	−0.052
	(−1.139)	(−1.159)	(−0.347)	(−0.454)
Annual precipitation	0.145	0.136	0.021	0.020
	(0.354)	(0.318)	(0.785)	(0.754)
Dummy variable for crop species	--	--	--	--
	--	--	--	--
Constant	31.185 *	42.194 **	3.009 ***	4.055 ***
	(1.872)	(2.480)	(2.790)	(3.839)
Year dummy	control	control	control	control
Province fixed-effect	control	control	control	control
Observations	1140	1140	1140	1140
R2	0.069	0.069	0.056	0.049
Number of sample villages	285	285	285	285

Note: t-statistics reported in parentheses; *, **, and *** represent statistical significance at the 10%, 5%, and 1% levels.

presents the estimated results of how agricultural regional specialization development affects farmers’ pest control behaviors, including control strategies and costs. The results indicate that, after adjusting for measurement bias in the indicators by including agricultural double cropping, both the Herfindahl Index (FC_HHI) and the Shannon Information Index (FC_Shannon), which measure the level of development of regional specialization in agriculture, show a positive effect on the proportion of preventive behaviors as well as the number of pesticide applications. These effects were significant at the 1% statistical level, thereby supporting hypotheses 1 and 2 of this paper.

Taking the fixed-effect model estimation results (FE-OLS) with the Herfindahl Index (FC_HHI) as the key explanatory variable as an example, the marginal effect on the proportion of preventive behaviors is 14.008, and on the total number of pesticide applications, it is 1.105. This suggests that as regional agricultural specialization develops, which means the planting structure in given area becomes more homogeneous, this will result in a 14% increase in the proportion of preventive behavior by farmers against pests and diseases and a 1.1 time increase in the total number of pesticide applications.

Other control variables align with theoretical expectations. For instance, village clan power significantly reduces the proportion of preventive behavior by farm households at the 1% significance level. This finding reflects the importance of mutual aid behavior among farm households, based on blood ties, in rural China. Such behavior aids in the timely observation, prevention, and control of pests and diseases, partially compensating for the scarcity of agricultural laborers. Consequently, the need for early prevention decreases, leading to a relatively lower percentage of preventive pesticide applications. Additionally, timely control of pests and diseases have effectively alleviated the issue of recurring outbreaks, reducing the total number of pesticide applications by farmers.

4.3. Mechanism Test: Estimated Results of the Influence of Regional Specialization in Agriculture on the Frequency and Damage of Pests and Diseases

Table 5. Estimated results of the influence of regional specialization in agriculture on frequency and damage degree of pests and diseases.

	Number of Occurrences (Times)		Proportion of Grain Yield Reduction (%)
	FC_HHI	FC_Shannon	FC_HHI	FC_Shannon
FC_HHI (lagging one period)	2.795 ***		10.790 ***
	(9.704)		(7.151)
FC_Shannon (lagging one period)		1.306 ***		4.912 ***
		(9.309)		(6.682)
Control variables	control		control
Year dummy	control		control
Province fixed-effect	control		control
R2	0.179	0.172	0.117	0.110
Observations	1140		1140

Note: t-statistics reported in parentheses; *** represent statistical significance at the 1% levels.

reports the estimated results of the effect of regional specialization in agriculture on the frequency of pests and diseases and the level of damage. The results show that the development of agricultural regional specialization has a positive effect on the frequency of pests and diseases and the proportion of grain yield reduction caused by them, and the results are significant at the 1% statistical level. Thus, the theoretical mechanism that agricultural regional specialization affects farmers’ control strategies and control costs by exacerbating the negative impacts of pests and diseases in the theoretical analysis is confirmed.

4.4. Robustness Test

4.4.1. Estimated Results of Instrumental Variable Method (IV)

The estimated results in

Table 6. Results of IV estimation of the influence of regional specialization in agriculture on farmers’ pest control strategies and costs.

	Proportion of Preventive Behaviors (%)		Total Number of Pesticide Applications (Times)
	FC_HHI	FC_Shannon	FC_HHI	FC_Shannon
FC_HHI (current period)	12.85 *		1.170 **
	(1.685)		(2.437)
FC_Shannon (current period)		7.943 **		0.456 **
		(2.159)		(2.042)
Control variables	control		control
Year dummy	control		control
Province fixed-effect	control		control
Anderson LM test	21.68 ***	21.44 ***	26.24 ***	21.40 ***
Wald F-statistic	10.46 (7.03)	19.13 (11.65)	10.55 (7.77)	19.25 (11.59)
Observations	975		975

Note: t-statistics reported in parentheses; *, **, and *** represent statistical significance at the 10%, 5%, and 1% levels.

indicate that after using the instrumental variable method, regional specialization in agriculture still shows a positive effect on the proportion of preventive behaviors as well as the number of pesticide applications, and the results are statistically significant at the 10% and 5% statistical levels, respectively. This suggests that the technical operation of using lagged terms to control potential endogeneity problems is feasible and the estimates in this study are more reliable. Due to space limitations, the results of other control variables are not reported.

4.4.2. Estimated Results of the Influence of Regional Specialization in Agriculture on the Pest Control Behaviors of Individual Farmer Households

The above estimation results reflect the average influence of agricultural regional specialization on the strategies and costs of pest control by farm households within sample villages using village-level data. Based on the actual situation of agricultural production in rural China, which primarily relies on decentralized, small family units, this study argues that individual farmers’ pest control behaviors are aimed at maximizing family farm income. Therefore, we discussed the change in strategies and costs of farmers in pest and disease control on a household level.

Table 7. Estimated results of the influence of regional specialization in agriculture on the pest control behaviors of individual farmer households.

	Total Number of Pesticide Applications by Farmer Household		Average Pesticide Cost for Farmer Household (CNY/mu)
	FC_HHI	FC_Shannon	FC_HHI	FC_Shannon
FC_HHI (lagging one period)	0.536 *		65.981 **
	(1.918)		(2.125)
FC_Shannon (lagging one period)		0.371 **		33.653 **
		(2.231)		(2.052)
Control variables	control		control
Province dummy	control		control
R2	0.623	0.624	0.480	0.480
Observations	975		975

Note: t-statistics reported in parentheses; * and ** represent statistical significance at the 10% and 5% levels.

reports the estimated results of the robustness test using farmer household data. The estimation results show that the development of agricultural regional specialization within the village area has a significant positive impact on the total number of pesticide applications and the average pesticide cost per mu of individual farmer household, and is significant at the 10% and 5% statistical levels. It indicates that the research conclusion drawn from this study that agricultural regional specialization will lead to an increase in the pest control costs of farmers is robust.

5. Discussion

Pests and diseases are one of the major causes of agricultural production losses. Effective control measures by agricultural producers not only mitigate these damages and ensure normal agricultural production but also contribute positively to the sustainable development of agricultural resources and the ecological environment. Regional specialization in agriculture, a typical method for adjusting and optimizing planting structures, influences farmers’ control behaviors through two mechanisms: altering pest and disease characteristics and modifying operation conditions for agricultural plant protection. Although this issue is crucial, existing research has seriously neglected it, let alone analyzed it using systematic theoretical analysis and rigorous empirical testing. Neglecting this issue will lead to a lack of academic analysis on farmers’ control behaviors. In agricultural production practice, it will also expose regions with developed agricultural specialization or homogeneous planting structures to risks stemming from inappropriate pest and disease control behaviors by farmers.

Therefore, this paper systematically analyzes the influences of agricultural regional specialization on farmers’ pest control strategies and costs, based on the theory of farmers’ behavior, and clarifies the different impact mechanisms. Empirical evidence is provided using data from four periods of village tracking survey and production data from 975 farmers across 285 administrative villages in Jiangsu, Sichuan, and Jilin provinces of China. The study’s main conclusions are as follows: First, regional specialization in agriculture significantly impacts farmers’ pest control strategies, shifting their focus from ex post treatment to ex ante prevention. Second, it has notably increased the frequency of pesticide application by farmers and raised their costs for plant protection. Third, the impact of regional specialization on farmers’ control strategies and costs is mediated by two mechanisms: reducing regional biodiversity, which increases the frequency and damage level of pests and diseases and strengthens farmers’ demand for control, and enhancing planting structure homogeneity, which leverages the advantages of unified plant protection operations.

The conclusions of this study carry significant policy implications: Firstly, it is essential to rationally assess the impact of increased pest control costs resulting from the development of agricultural regional specialization. In regions where the level of specialization is rapidly advancing and the planting structure is highly homogeneous, it is crucial to encourage farmers to implement moderate crop rotation based on local natural conditions. This practice can enhance the soil microbial community structure and improve soil fertility, thereby mitigating the rise in control costs caused by intensified pest and disease damage due to decreased biodiversity. Secondly, attention should be paid to the impact of the development of regional specialization in agriculture, which has led to a change in farmers’ control strategies. By leveraging the trend of regional specialization, farmers should be actively guided to focus on early prevention and timely control, employing green control technologies to curb pest and disease damage with minimal environmental and resource costs. Thirdly, theoretical research should concentrate on analyzing and testing the various mechanisms that influence farmers’ pest control behaviors. This focus will allow for proactive guidance in adjusting farmers’ prevention and control strategies within agricultural practices. Fourthly, agricultural departments of regional government should enhance dynamic real-time monitoring of pests and diseases. They should leverage mega data platforms to disseminate forecasts and alerts through multiple channels, thereby improving the accuracy and relevance of the information. This approach will provide more timely and reliable data for the preemptive prevention of pests and diseases.