Forest Biological Disaster Control Behaviors of Forest Farmers and Their Spatial Heterogeneity in China

Abstract

With more and more loss caused by forest biological disasters (FBDs) in China, forest farmers, as one of the most important stakeholders, are participating in the control. In this study, the ordinary least squares model, simultaneous equation model, and propensity score matching method were used with the data from 818 surveys conducted in the typical FBD outbreak provinces, to reveal the FBD control behaviors of forest farmers and their differences between western and eastern regions of China. The results indicated the following. (1) Household factors: forest farmers could timely take control measures. An increase of 1 ha in the area of occurrence would increase the control measures by 3.26 ha. However, the control measures can only reduce 50% of the economic loss caused by FBDs and cannot effectively control the spread trend of FBDs. There are issues, including an insufficient and old labor force, insufficient technology support, and low consciousness of ecology protection. (2) External factors: forest farmers would increase control when the temperature rises, and reduce control when rainfall increases. After village committees unify organizing the control, their participation enthusiasm would increase, which would have a substitution relationship with the household investment. (3) Regional difference: the eastern region of China has higher figures than the western in terms of outbreak area, economic losses, control measures, and pesticide cost. If the western forest farmers have the control funds as the eastern forest farmers have, and the eastern forest farmers have the control intensity as the western forest farmers have, the overall FBD control effect would be better.

1. Introduction

Forest biological disasters (FBDs) are situations where the outbreak scale of diseases, pests, rodent or hare infestations, and harmful plants exceeds the natural resistance of the forest ecosystem, causing losses in ecosystem services and other economic outputs [1,2,3]. Across the world, invasive insects cost a minimum of US$ (hereafter “$”) 70.0 billion per year, while associated health costs exceed $6.9 billion per year [4,5]. Along with the global ecological environment degradation, climate changes, international trade carrying invasive organisms, and China’s large area of planted forests with weak resistance to biological disasters, the scale of domestic FBDs continues to expand in China [6]. The average growth rate has been 1.3% per year in the past decade, and the outbreak area of FBDs reached 12.55 million ha in 2021. Among them, diseases accounted for 22.68%, insect pests accounted for 61.87%, rabbit disasters accounted for 13.91%, and harmful plants accounted for 1.51% [7]. The average economic loss caused by FBDs exceeds $21.54 billion per year [8]. There are more than 8000 species of forest pests in China, of which more than 300 are serious threats. Among them, pine wood nematode (Bursaphelenchus xylophilus), American white moth (Hyphantria cunea), forestry rodent (hare) infestations, and South American Climber (Mikania micrantha) are the key threats [9,10]. In 2021, the outbreak area of Bursaphelenchus xylophilus disease alone reached 1.72 million ha, causing economic losses of around $3 billion. Mikania micrantha disease alone caused ecological losses of up to $123.08 million in the Pearl River Delta region, while the losses caused by Hyphantria cunea and forestry rodent (hare) infestations were as high as $32.76 billion and $95.37 billion, respectively [11,12,13].

According to the Forest Law of China, the control of FBDs follows the guideline of “whoever manages is responsible for control [14].” Collective forest land accounts for 61.34% of the national forest area and is mainly managed by individual farmers or cooperative or shareholding organizations, while forest farmers are the important part [15,16]. While forest farmers are encouraged to participate in forestry activities, they also shoulder the responsibility of FBD control [17,18]. Because FBDs spread, the untimely control behaviors of forest farmers can lead to an expansion of the outbreak area, leading to economic loss [19]. In addition, focusing only on their own economic interest, forest farmers are more likely to use highly toxic but low-cost pesticides [20,21]. This kind of behavior could cause non-point source pollution and endanger ecological environmental safety. Even if pests develop antibodies to pesticides, pesticides could also lead to ineffective input from forest farmers and endanger the health of humans and animals [22,23]. Therefore, it is urgent to understand the status quo of forest farmers’ control behaviors using the integrated method of survey study and empirical analysis.

Thus far, research on the participation of forest farmers in FBD control has been conducted mainly focusing on three aspects. (1) The first is about the participation and willingness to pay of forest farmers. Stallman et al. (2015) explored the willingness of 229 farmers in Missouri, USA, to cooperate in pest control, and found that 91% of farmers are willing to cooperate [24]. Sheremet et al. (2017) analyzed the willingness of farmers to participate in pest control in Finland through a choice experiment method and found that households are willing to pay £14.9 per hectare for the control measures taken in commercial forests [25]. Zhu et al. (2015) analyzed the data of farmers in Zhenjiang Province, China, and found that 66.23% of the farmers are willing to participate in control by professional companies [26]. Han et al. (2019) found that the larger the area of commercial forests, the greater the demand for FBDs’ control from farmers [27]. The greater the labor force in the household, the higher willingness they have to accept the control services to keep the quality of their commercial forests. (2) The second is about the support policies for forest farmers in FBD control. Jones et al. (2010) found that forest farmers can reduce losses by $16 million through technical training and control information early-warning systems based on a pesticide database [28]. In this way, forest farmers could be informed in advance of FBD outbreak trends and suitable prevention measures, as the environmental externality generated by participating in FBD control is higher than forest farmers’ economic benefits. Hanley et al. (2012) proposed the development of an ecological compensation mechanism to encourage farmers to participate in forest protection through tax benefits, which can alleviate the economic burden of forest farmers caused by the spillover effect [29,30]. (3) The third aspect is about the forms of forest farmer participation in FBD control. Ayres and Lombardero (2017) constructed couplings between forest, forest management, and socioeconomic systems to explore the pest control cooperative model to increase the commercialization level of government services so that forest managers are encouraged to participate in FBD control with government support [31]. Adinoyi (2014) proposed that, for state-owned forest land, governments should finance the formation of socialized pest control (including forest insurance, control companies, a control specialized team, etc.), and individual forest farmers should be responsible for FBD control [32]. Liao et al. (2022) used an endogenous transformation model and found that control companies can significantly improve the input of production factors from forest farmers, and there is a complementary effect between capital input and labor input [33]. Moreover, forest farmers need to increase their capital and labor input to match the requirements of socialized services.

Through the literature review, it can be concluded that the status quota of farmers’ participation in FBD control mainly focuses on participation willingness, participation methods, and willingness to pay. Overall, forest farmers are willing to participate in control to reduce the economic losses of their commercial forests. With government support, their willingness to participate would increase. Furthermore, the pest control cooperative model and control companies are developing trends of forest farmer participation in FBD control. However, from a theoretical perspective, there is a relatively limited in-depth analysis of influencing factors and effects of forest farmers’ control behaviors. Moreover, although China’s forestry development plays a big role in global ecological development, it is very limited to study the control behaviors of forest farmers in China at both the national and regional scales.

Therefore, the research objectives of this study are: (1) to analyze the impact of household and external factors on FBD control behaviors of forest farmers using developed OLS and SEM models; (2) to analyze the spatial heterogeneity (between the eastern and western regions) of forest farmers’ control behaviors using the propensity score matching method.

2. Study Area and Method

2.1. Study Area

The study area is the mainland of China. The concept of the “Hu Line”, a well-known demographic-geographic line in China, was introduced to divide the country into two regions: western and eastern [34,35,36]. The western region is sparsely populated and the economic development is relatively lagging, while the eastern region is relatively economically developed with a dense population [37,38]. The differences in population, environment, and economy between the western and eastern regions of China, including the differences in forest farmers’ cognition, natural conditions, government support, etc., may result in regional differences in the perception of FBD outbreaks and the resulting losses, the behaviors of measures taken, the investment, etc. Thus, two representative provinces were selected from both the western and eastern regions, and a comparative analysis was conducted for the spatial heterogeneity of control behaviors of forest farmers.

As shown in Figure 1, the colors from dark to light represent the outbreak areas of FBDs by province. In the western region, the provinces of Xinjiang and Inner Mongolia were selected for the highest area of FBDs. Both are also the main epidemic areas for rodents (hares). In the east, Shandong Province, with a high outbreak area, was chosen. Shandong is the main epidemic area for Hyphantria cunea, while Bursaphelenchus xylophilus has also occurred frequently there in recent years. Guangdong Province was selected due to its high incidence of Bursaphelenchus xylophilus. It is also a major epidemic area for Mikania micrantha [39]. This research mainly focuses on the outbreak of overall FBDs in all collective forest land owned by farmers.

2.2. Data Collection

A questionnaire survey was conducted in the provinces of Shandong, Guangdong, Inner Mongolia, and Xinjiang from January to December 2019. Specifically, the questionnaire was developed based on previous studies by Ke and Wen (2014) and Duan and Wen (2016), as well as the authors’ experience and expertise [40,41]. Interviews were then conducted with province-, prefecture-, and county-level officials to identify typical communities/villages with representative FBD occurrence and control measures [42]. Five types of forest management models, including management by a regular individual household, management by large households, management by joint households, management by cooperatives, and management by village groups, were investigated. The interviews were conducted by research personnel based on the questionnaire content, asking questions of the forest managers and filling in the questionnaire based on the answers. After nearly one year of questionnaire surveys, 20 villages from 10 counties of five cities in Shandong, 20 villages from 10 counties of five cities in Guangdong, 20 villages from 10 counties of five cities in Inner Mongolia, and 18 villages from six counties of two cities in Xinjiang, were surveyed. Overall, 818 of 890 questionnaires were effective, with a validity rate of 91.91%. Among them, a total of 437 sets of data were obtained from the eastern region, where Guangdong and Shandong are located, and 381 sets of data were obtained from the western region, where Xinjiang and Inner Mongolia are located. In the form of an interview, researchers ask forest operators questions according to the questionnaire content, and the researchers fill in the questionnaire according to the answer content.

The details of the questionnaire are discussed below.

2.3. Indicators

The questionnaire consisted of three parts: basic information about forest farmers, their perception of FBDs, and their perception of FBD control measures. For each part, the specific indicators selected are shown in

Table 1. Contents and specific indicators of the forest farmers’ control behavior questionnaire.

Contents	Indicators
Basic information about forest farmers	Gender, age, years of education, health, family size, management models, equality status of women, total household income, and total forestry income.
Perception of FBDs	The identification of the FBDs, loss types caused by FBDs, the impact of pesticides on life, the way to obtain control knowledge, the preferred way to get support, the entity to take the control measures for commercial forest and public welfare forests, and the source of pesticide acquisition.
Perception of FBD control	Control area, pesticide input, labor day input, proportion of forestry income in the overall household income, area of planted forests, outbreak area, economic loss, control knowledge cognition, temperature-rise perception, rainfall-increase perception, and perception of the organizational ability of village committee for FBD control.

.

In the part on basic information and control behaviors of interviewees, we described the basic information of the interviewed forest managers, referencing the related literature [43,44,45]. Descriptive statistics were used to describe the gender, age, education level, health status, family size, management form, women’s equality status, total household income, and income from forestry, to understand the household characteristics. Statistical analysis was used to investigate the perceptions of forest farmers regarding the impact of FBD, ways to acquire control knowledge, the entity to take control measures, preferred support for better control, sources of pesticide acquisition, and the impact of these on their daily lives. This was done to gain a general understanding of forest farmers’ basic attitudes towards control measures, as well as the impact of FBDs and corresponding control measures on their livelihoods [46].

Forest farmers’ control behavior not only depends on the personal characteristics of the interviewed individuals but is mainly influenced by the overall situation of the household, such as the amount of labor, livelihood assets, etc. [47]. Based on the Cobb–Douglas (C–D) production function, labor and capital are important factors affecting output [48]. Forest farmers’ control behavior meets the input–output relationship, and the control behavior adopted by forest farmers is an important output variable in the function [16,49]. The input of labor hours is used as an indicator for labor in the control, and the amount of capital invested in pesticide purchases is the capital variable in the C–D function [50]. Therefore, we chose the control area, input of labor days, and pesticide purchasing costs as the indicators to be analyzed.

The income that forest farmers receive from forestry production is their main motivation to take control measures to mitigate the loss of FBDs [51]. The larger the area of the monoculture plantation, the higher the risk of the occurrence of FBDs and the related loss [52,53,54]. Therefore, the factors of the proportion of forestry income in the overall household income, the area of planted forest, the area of FBDs, and the amount of economic loss are selected. In addition, forest farmers’ awareness of the control could affect the adoption of specific control measures and enthusiasm for the control of forest farmers [55]. Therefore, the indicator of the forest farmers’ knowledge about FBD control is selected in the questionnaire.

In addition to household-related factors, external factors such as climate conditions and policies may affect the control behaviors of forest farmers [56,57,58]. Therefore, the indicators of climate and rainfall increase perception are selected. The government organizes forest farmers to participate in control efforts, distributes ecological pesticides, and hires control organizations to improve the popularity of forest farmers’ ecological control [33]. In China, the village committee is a basic government organization that directly deals with forest farmers [59,60]. Therefore, the ability of the village committee to organize control efforts is selected to analyze the impact of external factors on their control behaviors.

2.4. Analytical Methods

In this study, the ordinary least square method (OLS) was first employed and then updated to a simultaneous equation model (SEM) overcoming the endogenous problems of OLS, to analyze the influencing factors of forest farmers’ control behaviors. Second, a propensity score matching model (PSM) was developed to further analyze the spatial heterogeneity of forest farmers’ control behaviors between the eastern and western regions. The details are as follows.

2.4.1. OLS Model

The OLS model was first developed to preliminarily analyze the influencing factors of the control behaviors of forest farmers. OLS, the most commonly used estimation method for single-equation linear regression models, minimizes the error sum of squares with respect to the regression parameters [61]. OLS models are based on the assumption that the relationship between the study variable and explanatory variables is linear, at least approximately, and the error term has a zero mean and constant variance and is uncorrelated and normally distributed [62]. According to the C–D production function, the control area was used as the output variable of the control behaviors, while the other indicators were the independent variables. Specifically, the OLS model was developed as below.

$${CA}_i={\omega }_0+{\omega }_1{PC}_i+{\omega }_2{ID}_i+{\omega }_3{PO}_i+{\omega }_4{CK}_i+{\omega }_{5}V{C}_i+{\omega }_6{PA}_i{+{\omega }_7{TI}_i+{\omega }_8{RI}_i+{\omega }_9{FI}_i+{\omega }_{10}{ES}_i+u}_i$$

where $\omega$ is the coefficient, $\mu$ is the random error, CA is the control area, PC is the pesticide cost, ID is input days for the control work, PO is the outbreak area of FBDs, CK is control knowledge, VC is the ability of the village committee organizing the FBD control, PA is the area of planted forests, TI is the perception of temperature increment, RI is the perception of rainfall increment, FI is the percentage of forestry income in the overall household income, and ES is the economic loss caused by FBDs. Of note, TI, RI, and VC are the forest farmers’ perceptions of the degree of increment or strength. The perception response is a Likert scale divided into five points, with 5 points for the strongest perception and 1 point for the weakest one.

2.4.2. SEM

The control area, investment, and working hours are part of the control measures. Regarding the control area, it could be affected by other indicators. Specifically, the proportion of forestry income in the household income could directly affect the pesticide purchase of forest farmers, hence indirectly affecting the control area [33,63]. The area of planted forests is the direct factor leading to an increase in the FBD outbreak area; as a result, forest farmers would take control measures [64]. Meanwhile, according to the study by Cai et al. (2019), there is a mutual causal relationship between control measures, outbreak area of FBDs, and economic loss [63]. As the OLS model has endogenous issues, we further employed the SEM model to overcome the issues and improve the regression efficiency of the model [42,65].

Combined with a two-stage least squares method with seemingly unrelated regression, the method of three-stage least squares is the most commonly used estimation in SEM [65] to establish four partial equations of control area, pesticide cost, outbreak area, and economic loss. These four indicators are endogenous variables, and the others are corresponding control variables. Thus, in the function of control area, the coefficient of pesticide cost means the efficiency of the control funds, and pest outbreak area means the timeliness of control measures, and the factors of ID, CK, VC, TI, and RI could also influence the control measures. In the function of pesticide cost, the coefficient of control measures taken area means the timeliness of the control fund investment, and the factors of FI, CK, and VC could also influence the pesticide cost input willingness of farmers. In the function of pest outbreak area, the coefficient of control area means the efficiency of control measures, and the plantation area also would influence the pest outbreak trends. In the function of economic losses, the pest outbreak area would increase the losses, and the control measures could reduce the losses.

The specific formula was finally determined after repeated testing of the endogeneity, stability, and significance of each index, where $\alpha$,$\beta$,$\lambda$, and $\theta$ represent the corresponding coefficients, respectively. $\mathsf{\mu }$ are error terms. The details are as follows:

$${CA}_i={\alpha }_0+{\alpha }_1{PC}_i+{\alpha }_2{ID}_i+{\alpha }_3{PO}_i+{\alpha }_4{CK}_i+{\alpha }_5{VC}_i+{\alpha }_6{{TI}_i+{\alpha }_7{RI}_i+u}_{1i}$$

$${PC}_i={\lambda }_0+{\lambda }_1{FI}_i+{\lambda }_2{CA}_i+{\lambda }_3{CK}_i+{\lambda }_4{VC}_i+u_{2i}$$

$${PO}_i={\beta }_0+{\beta }_1{PA}_i+{\beta }_2{CA}_i+u_{3i}$$

$${ES}_i={\theta }_0+{\theta }_1{CA}_i+{\theta }_2{PO}_i+{\mu }_{4i}$$

2.4.3. PSM Model

As forest farmers are affected by different external factors, such as education environment, natural environment, and economic development status, they would have regional differences in the perception of FBD outbreak and its control. Therefore, the PSM method was selected to analyze the perception differences in FBD outbreak area, economic losses, control area, and pesticide cost, after excluding other external factors, if the farmers in the eastern region are in the environment of farmers in the western region.

Specifically, Shandong and Guangdong in the eastern region are assigned as 1 and set as the treatment group; Inner Mongolia and Xinjiang in the western region were assigned as 0 and set as the control group. The factors of control area, outbreak area, pesticide cost, and economic losses are used as matching dependent variables, and the selected control variables are consistent with the dependent variables of each regression model in the SEM equation.

As it is not possible to obtain the cognitive status of forest farmers in the eastern region if they live in the western region, it is necessary to construct a counterfactual framework; that is, given a set of covariates, to score each farmer, calculate the probability of entering the treatment group, and record it as the tendency score. In fact, it is the dimensionality reduction projected on the disposal dimension of a set of covariables that can match the farmers on the east side with those on the west side in multiple dimensions [66]. Therefore, that makes the distribution of the covariates of the two paired farmers the same; only one is assigned to the eastern group and the other to the western group. This is equivalent to a random experiment. After all samples are matched, the difference in the outcome variable between the paired sample groups is calculated to obtain an average treatment effect [63]. The specific formula is as follows:

$$\mathrm{A}\mathrm{T}\mathrm{T}=\mathrm{E}\left({\mathrm{Y}}_{\mathrm{i}}-{\mathrm{Y}}_0|\mathrm{P}=1\right)=\mathrm{E}\left({\mathrm{Y}}_{\mathrm{i}}|\mathrm{P}=1\right)-\mathrm{E}\left({\mathrm{R}}_0|\mathrm{P}=1\right)$$

where P is the region of the farmer, (P = 1 means the farmer is located in the eastern region, P = 0 in the western region), and ${\mathrm{Y}}_{\mathrm{i}}$ represents the output variables of outbreak area, control area, pesticide cost, and economic losses in the eastern region. ${\mathrm{Y}}_0$ represents the output variables of outbreak area, control area, pesticide cost, and economic losses in the west region. To account for unobservable $\mathrm{E}\left({\mathrm{Y}}_0|\mathrm{P}=1\right)$ variables, the control group in the western region and the treatment group in the eastern region were used to compare treatment effects and standard errors, etc. Mahalanobis matching, K-nearest neighbor matching, caliper-neighbor matching, radius matching, kernel matching, and local linear regression matching are used to estimate and compare with the coefficients before matching.

3. Results

3.1. The Status Quo of Control Behaviors

3.1.1. Sample Demographics of the Interviewees

The sample demographics of the interviewees are shown in

Table 2. Sample Demographics of the Interviewees.

Indicator		Amounts	Proportion (%)	Mean	Standard Deviation
Gender	Male = 1	717	87.65	0.91	0.28
	Female = 0	101	12.35
Age	Under 35	117	14.3	51.93	12.30
	35 to 45 years old	152	18.58
	46 to 60 years old	358	43.77
	over 60 years old	191	23.35
Years of education	6 years and below (elementary school)	519	63.45	5.09	4.21
	6–9 years (middle school)	212	25.92
	10–12 years (high school)	75	9.17
	13+ years (college)	12	1.47
Family size	2 people or less	212	25.92	3.70	1.63
	3–5 people	489	59.78
	6 people or more	117	14.3
Health degree	Serious illness = 1	35	4.28	3.88	1.18
	Mild illness = 2	98	11.98
	Generally healthy = 3	121	14.79
	Healthier = 4	249	30.44
	Very healthy = 5	315	38.51
Operation form	Single household = 1	453	55.38	1.35	0.97
	Joint household = 2	141	17.24
	Shared cooperation = 3	93	11.37
	Village group management = 4	82	10.24
	Forestry cooperative = 5	49	5.99
Status of Women	Lower than men = 1	145	17.73	1.90	0.46
	Equality = 2	602	73.59
	Higher than men = 3	71	8.68
Annual household income	Below $2000	211	25.79	9545.60	51,172.37
	$2001–5000	234	28.61
	$5001–10,000	189	23.11
	$10,001–20,000	124	15.16
	Over $20,001	60	7.33
Annual forestry income	Below $500	637	77.87	1518.57	16,808.78
	$501–2000	112	13.69
	$2001–20,000	58	7.09
	$20,001	11	1.34

Note: 1 US dollar is converted to 6.5 Chinese yuan.

. Most of them are healthy men aged 46 to 60 (with an average age of 52) and have an elementary school education of around 5 years. The family size is mostly between 3 and 5 (59.78%), and 73.59% of families state that males and females are equal. The main forms of forestry management are single households (55.38%), followed by joint households (17.24%), shared cooperation (11.37%), village group management (10.24%), and forestry cooperatives (5.99%). The annual household income is mostly between $2001 and $5000, with an average of $9546. There are 11 large-scale forestry households with annual forestry income exceeding $20,000, while the group with forestry income below $500 accounts for the highest proportion (77.87%), with an average overall household income of $1518.57. Overall, the forestry management of collective forests is still dominated by single-household management with lower forestry income, while the education level and family livelihood welfare are relatively low.

3.1.2. The Descriptive Statistics of Control Behaviors

In terms of the household, as shown in

Table 3. The descriptive statistics of control behaviors.

Indicator		Mean	Standard Deviation	Mean (Eastern Region)	Mean (Western Region)
Proportion of forestry income (FI)	Forestry income to total household income (%)	0.09	0.21	0.08	0.09
Plantation area (PA)	Continuous variable (ha)	1.76	5.92	0.62	2.04
Outbreak area of FBDs (FO)	Continuous variable (ha)	0.68	3.23	1.00	0.32
Economic losses induced by FBDs (ES)	Continuous variable (US dollar)	153.77	874.64	249.19	44.32
Area with control measures (CA)	Continuous variable (ha)	0.94	5.73	1.29	0.53
Pesticide cost (PC)	Continuous variable (US dollar)	75.93	381.46	105.43	42.10
Labor days for control work (ID)	Continuous variable (days)	5.47	15.53	7.26	3.42
The perception of temperature increase (TI)	5–1 is very large–very small	2.89	1.11	2.97	2.80
The perception of rainfall increase (RI)	5–1 is very large–very small	3.17	1.19	3.43	2.87
Understanding of control knowledge (CK)	5–1 is know very well–very ignorant of	3.32	0.98	3.68	2.99
The ability of village committee to organize control work (VC)	5–1 is very strong–very weak	4.03	0.99	4.14	3.91

Note: Forestry income includes income from forestry operations and subsidies related to forestry; household income includes wage income, operating income (planting, forestry, gathering, sideline, breeding, and self-employment), subsidies, capital income, and other incomes.

. The average proportion of forestry income to total household income is 9%, and the average area of planted forests is 1.76 ha. The average forest area affected by FBDs is 0.68 ha per year, resulting in an annual average loss of nearly $153.8. The average area for taking control measures is 0.94 ha per year, with an average annual investment of $75.9 in pesticides and 5.5 days of labor. The perception of the increase in rainfall is slightly stronger than that of temperature rise, and the respondents have a relatively good understanding of FBD control. They also believe that the village committee has relatively strong capabilities to organize FBD control. In terms of regional differences, the proportion of forestry income and the area of planted forests are higher in the western region than in the east. However, the perception of the affected area, economic losses, area of taking control measures, pesticide use, and time spent on control, as well as the perception of the increase in temperature and rainfall, knowledge of FBD control, and evaluation of the village committee’s ability to organize FBD control, are much lower in the western region than in the eastern region.

Forest farmers believe that FBDs mainly damage forest timber (80.8%) while having little impact on house damage, and human and animal infection. The knowledge of FBD control is mainly learned through life experience (43.2%), government-hosted technical training (41.3%), and introduction when purchasing pesticides (30.5%). The pesticides are mainly acquired from self-purchasing (68.7%). They believe that the application of pesticides pollutes the environment (54.4%) and increases the cost of FBD control (39.7%). Furthermore, 48.6% of respondents think that commercial forests should control against FBDs by themselves, while 50.8% of farmers believe that public-welfare forests should be controlled by the local forestry department because forest farmers cannot directly obtain economic benefits from public-welfare forests. They hope to receive support through financial subsidies (41.4%), government organizations (36.8%), ecological pesticide acquisition (29.4%), and technical training (27.6%).

Regarding the FBD type, rat and rabbit disasters, aphids, red spiders (Kanazawa spider mites), and economic crop diseases frequently occur in Xinjiang, and pesticides such as propargite, abamectin, and lime sulfur are usually used. Rat/rabbit disasters, sea-buckthorn moth (Holcocerus hippophaeclus), Hyphantria cunea, cryptorrhynchus lapathi (Phytoecdysteroid), Kanazawa spider mites, etc. are always found in forestland in Inner Mongolia, while the main pesticides used are diflubenzuron, matrine, cypermethrin, pyrethroid, etc. Shandong forest farmers are mostly faced with Hyphantria cunea, Bursaphelenchus xylophilus, rotten disease, altosis sulcus (Eucryptorrhynchus brandti), alternaria leaf spot, and drosophila, while pesticides such as diflubenzuron, methylurea, dichlorvos, and imidacloprid are mainly used. Guangdong forest farmers are mainly faced with threats from Mikania micrantha, Bursaphelenchus xylophilus, and yellow bamboo locust (Ceracris kiangsu), etc., while they mainly use diflubenzuron, dichlorvos, dimethoate and high-efficiency chlorolipin (Beta-cypermethrin). Among these provinces, dichlorvos, dimethoate, and cypermethrin are pesticides with restrictions, and unreasonable application would lead to serious non-point source pollution of pesticides at the rural level.

Therefore, in terms of the impact of pesticides on farmers’ lives, 56.22% and 49.38% of the respondents from the eastern and western regions, respectively, believed that the use of pesticides would damage the environment and cause air and water pollution, increasing the cost of household input (52.21% in the east, 31.12% in the west). Additionally, more of the west side believes that pesticides would reduce food safety (19.09%) than the east side (11.24%). Moreover, a small number of forest farmers believe that pesticides would affect the quality and sales of forest products (10.84% in the east, 6.22% in the west), make themselves or their families sick (9.64% in the east, 4.15% in the west), and even cause livestock to get sick and other hazards (4.02% in the east, 0.83% in the west).

3.2. The Analysis of Control Behaviors

In order to have a clearer understanding of the impact of funds, outbreak, and loss, as well as plantation area, climate perception, labor input, and the organizational ability of village committees, on forest farmers’ prevention and control behaviors, the OLS model was first used. First, the collinearity test was conducted on the model. The VIF coefficient was between 1.14 and 1.94, with an average of 1.15; thus, there was no collinearity issue. Second, the heteroskedasticity issue was tested using the Breusch–Pagan and Cook–Weisberg tests. As a result, the p-value is 0.001, at the 1% significance level to refuse the hypothesis of constant variance, so we can conclude that there is no heteroscedasticity issue in the model. The regression results are shown in

Table 4. Analysis of control behaviors and the influencing factors.

	OLS		SEM
Variables	Control Area	Std. Err.	Control Area	Std. Err.	Pesticide Cost	Std. Err.	Pest Area	Std. Err.	Economic Losses	Std. Err.
Plantation area (PA)	−0.02	0.01 (−1.55)					−0.01	0.01 (−1.02)
The perception of temperature increase (TI)	−0.04	0.06 (−0.59)	0.13 **	0.07 (1.99)
The perception of rainfall increase (RI)	−0.02	0.06 (−0.39)	−0.56 ***	0.16 (−3.52)
Area with control measures (CA)					191.46 ***	28.98 (6.61)	1.11 ***	0.08 (13.13)	−751.31 *	400.62 (−1.88)
Proportion of forestry income (FI)	0.43 *	0.26 (1.65)			267.67 ***	75.10 (3.56)
Outbreak area of FBDs (FO)	0.94 ***	0.04 (22.79)	3.26 ***	0.86 (3.78)					1354.14 ***	364.52 (3.71)
Economic losses induced by FBDs (ES)	0.00004	0.00 (0.57)
Pesticide cost (PC)	0.0003	0.00 (1.39)	−0.003	0.00 (−1.55)
Labor days for control work (ID)	−0.003	0.004 (−0.58)	−0.08 ***	0.02 (−3.52)
Understanding of control knowledge (CK)	0.005	0.04 (0.13)	−0.01	0.04 (−0.31)	2.58	8.64 (0.30)
The organizational control ability of the village committee (VC)	0.07	0.06 (1.14)	0.33 ***	0.08 (4.23)	−26.18 *	14.91 (−1.76)
constant	−0.03	0.34 (−0.09)	−0.51	0.34 (−1.49)	44.99	81.40 (0.55)	−0.21 ***	0.08 (−2.62)	−29.34	109.11 (−0.27)
p-value	***		***		***		***		***

Note: The signs of “***” “**” “*” represent the significance at the 1%, 5%, and 10% levels, respectively. The numbers within the brackets are the statistical z value.

.

Overall, the model is significant at the 1% significance level. However, only two variables of eleven passed the significance test in the OLS model, which verifies the hypothesis that there is an endogenous issue in this simple OLS model. Furthermore, the control measures taken would also have an impact on the capital investment and its outbreak, and there may be an endogeneity issue between the variables, which affects regression efficiency. Hence, to overcome this issue, the SEM model was developed. As a result, all of the four sub-models of control area, pesticide cost, FBD outbreak area, and economic loss, which are endogenous variables, passed the significance test (1%).

Specifically, in the sub-model of the control measures taken area, the perceived increase in temperature and rainfall passed the significance test. Moreover, for each additional unit of temperature and rainfall, the control area would increase by 0.13 ha and decrease by 0.56 ha, respectively. In addition, the outbreak area (ha), input labor days, and ability of community committees to organize control also passed the significance test. For an additional unit of each of them, the control area would increase by 3.26 ha, decrease by 0.08 ha, and increase by 0.33 ha, respectively.

In the sub-model of investment to pesticide purchase, the control area, the percentage of forestry income in household income, and the organizational ability of community committees passed the significance test. The percentage of forestry income in household income can greatly stimulate forest farmers to purchase pesticides. With every 1% increase, the investment of forest farmers to purchase pesticides would increase by 267.7 US dollars. The control cost per hectare would be 191.5 US dollars. The organization of community committees would reduce the willingness of forest farmers to purchase pesticides individually. With a one-unit increment of the organizational ability of the community committees, the individual investment would decrease by 26.18 US dollars.

In the sub-model of the FBD outbreak area, the CA passed the significance test. With the 1-ha increment of the control area, the FO would increase by 1.11 ha. This indicates the employed control measures mitigate the occurrence but they still cannot totally control the outbreak area of FBDs.

In the sub-model of economic loss resulting from FBDs, both the variables of FO and CA passed the significance test. For a 1-ha increase in FO, the economic loss would increase by 1354.1 US dollars. However, for a 1-ha increase in CA, the economic loss would decrease by 751.3 dollars. This indicated that the employed control measures can reduce the economic loss resulting from FBDs to some degree.

3.3. The Spatial Heterogeneity Analysis

In order to analyze the spatial heterogeneity in control behaviors and economic-loss perception of forest farmers, the PSM method was used to explore the differences in control measures, capital investment, FBD outbreak area, and economic losses between both sides of the “Hu Line” while removing the differences of external conditions such as nature and environment. PSM could more clearly correct the influence of regional deviation on control behavior. By taking different matching methods, the maximum loss was 19 samples in the experimental group and 11 samples in the control group, with 766–818 valid matching samples remaining. This indicates the matching effect is good. In terms of the balancing test, the values of pseudo R² and LR chi² after matching were smaller than those before matching. This indicates that the model after matching was more stable.

Specifically, in the sub-model of the FBD outbreak area, as shown in

Table 5. Matching quality indicators and PSM results of FBD outbreak area and the induced economic losses.

	FBD Outbreak Area						Economic Losses
Matching Method	Ps R2	LR chi2	ATT	Diff.	S.E.	t-Test	Ps R2	LR chi2	ATT	Diff.	S.E.	t-Test
Before	0.03	21.95	0.62	0.26	0.13	1.97 **	0.013	14.69	249.19	204.86	60.92	3.36 ***
Mahalanobis	0.00	0.25	0.60	0.40	0.17	2.37 **	0.001	0.81	246.16	166.64	58.29	2.86 ***
K-nearest neighbor	0.01	4.31	0.60	0.43	0.14	3.00 ***	0.001	0.67	246.16	158.04	60.68	2.60 ***
Caliper-neighbor	0.002	2.05	0.47	0.31	0.14	2.26 **	0.003	3.26	191.57	110.91	43.62	2.54 **
Radius	0.002	1.41	0.47	0.30	0.12	2.44 **	0.009	10.04	191.57	134.74	42.21	3.19 ***
Kernel	0.003	2.38	0.60	0.35	0.13	2.64 ***	0.00	0.20	246.16	170.29	59.20	2.88 ***
Local linear regression	0.00	0.25	0.60	0.40	0.17	2.35 **	0.001	0.81	246.16	166.33	58.29	2.85 ***

Note: The signs of “***” “**” indicate statistical significance levels at 1% and 5%, respectively.

. All results passed the t-test and significance test. Before matching, forest farmers in the eastern region believed that the average outbreak area of their forest land was 0.62 ha per year, while this figure was 0.36 ha in the western region. However, after matching to exclude the difference in external conditions between both sides, forest farmers in the eastern region would think that their actual outbreak areas would be reduced to 0.47–0.6 ha, while forest farmers in the western region would think this would be about 0.17 ha. This result indicates that the outbreak risk of FBDs in the eastern region is higher than in the western region before and after matching. After matching, the outbreak area in both regions decreases, while the outbreak area in the eastern region is still larger than in the western region.

In terms of economic loss, all results passed the t-test and significance test. Before matching, the forest farmers in the eastern region believed that the average annual loss would be 249.2 dollars, while those in the western region believed that would be 44.3 dollars. However, after matching to exclude the difference in the control area and outbreak area between both regions, the forest farmers in the eastern region would believe that the economic loss is reduced to 191.6–246.2 US dollars, while those in the western region would believe that the economic loss is increased to 80 US dollars.

In terms of the control area, the caliper-neighbor and radius-matching methods passed the significance test, as shown in

Table 6. Matching quality indicators and PSM results of control area and pesticide cost.

	Control Area						Pesticide Cost
Matching Method	Ps R2	LR chi2	ATT	Diff.	S.E.	t-Test	Ps R2	LR chi2	ATT	Diff.	S.E.	t-Test
Before	0.13	120.16	1.11	0.52	0.27	1.92 *	0.08	75.08	99.16	51.80	25.81	2.01 **
Mahalanobis	0.01	12.58	0.62	−0.08	0.19	−0.39	0.01	8.68	77.46	56.68	16.95	3.34 ***
K-nearest neighbor	0.01	9.34	0.62	−0.08	0.21	−0.39	0.003	3.08	77.46	42.25	35.28	1.20
Caliper-neighbor	0.01	11.73	0.46	−0.31	0.20	−1.54 *	0.003	2.81	76.07	41.16	35.37	1.16
Radius	0.01	7.91	0.46	−0.37	0.19	−1.96 **	0.002	1.69	76.07	38.20	31.74	1.20
Kernel	0.003	2.51	0.62	−0.18	0.19	−0.97	0.01	5.52	77.46	46.04	29.79	1.55
Local linear regression	0.01	12.58	0.62	−0.14	0.19	−0.70	0.01	8.68	77.46	31.77	16.96	1.87 *

*, **, and *** indicate statistical significance levels at 10, 5, and 1%, respectively.

. Before matching, forest farmers in the eastern region believed that the average control area per year was 1.11 ha, while those in the western region thought it was 0.59 ha. The control area of the eastern region is larger than in the western region. However, after matching to exclude the difference in outbreak area, investment, labor input, perception of the increment of the temperature and precipitation, control knowledge, and organizational ability of the community committees, the forest farmers in the eastern region believe that the control area would be reduced to 0.46 ha, while those in the western region think the control area would be around 0.8 ha. This result indicates that, after matching, the control area of the eastern region would be decreased, and the western control area would be increased and even larger than in the eastern region.

In terms of the investment in pesticide purchase, before the matching, the forest farmers in the eastern region spent 99.2 dollars per year, while those in the western region spent 47.4 dollars per year. However, after matching to exclude the difference in control area, the percentage of forestry income, control knowledge, and the organizational ability of the community committees, the investment of the forest farmers in the eastern region would be reduced to 77.46 dollars, being 31.7–56.7 dollars more than in the western region. This result indicated that the control investment of the eastern region is higher than that of the western region before and after matching, and after matching, the investment from the farmers in both regions would be reduced.

4. Discussion

In this paper, a survey study was conducted to understand FBD outbreak, control, and economic loss from the perspective of forest farmers. Furthermore, the OLS and SEM models were developed based on the Cobb–Douglas production function to investigate the influencing factors of the control behaviors of forest farmers in this paper. Overall, this study has a marginal contribution to understanding the control behaviors of forest farmers and their spatial heterogeneity in China, which is informative for the decision-making process regarding how to develop policy instruments to motivate the control behaviors of forest farmers.

(1) In terms of control behaviors, outbreak area and economic loss would significantly affect the control behaviors of forest farmers, which is consistent with what is stated in the previous study: “if forest farmers realize the area of FBDs is larger and the resulting economic losses are more severe, they would be more conscious of taking control measures [53,55]”.

For an increase of 1 ha in the occurrence area, the control area would increase by 3.26 ha, and for an increase of 1 ha in the control area, it would reduce the loss of economic value by $751.31. However, the existing control measures cannot completely stop the spread/outbreak trend of FBDs. Increasing FBDs by 1 ha would still cause an economic loss of $1354.14. These findings are generally consistent with the conclusion that national control is “timely but inefficient” given by [63]. In addition, based on the survey, it was revealed that, in control activities, single-household control is still dominant (55.38%), with the issues of lack of funds, technical support, and labor input negatively affecting control efficiency.

Moreover, it was indicated that there are issues of lacking labor forces and input labor hours. According to the survey of forest farmers, most household sizes are around 4 people, and most of the people who play the major role in taking control measures are 50+ years old and not well educated. This is consistent with the current situation in the rural areas of China. With rapid urbanization, most young laborers move to urban areas [36]. As a result, there is a shortage of young people in rural areas to engage in production and management activities.

The investment is relatively timely, with an average of 191 dollars per hectare for control. Households whose main income source is from forestry would increase the pesticide input by $267.67. This conclusion is very consistent with the statement that “if the households’ livelihood mainly relies on forestry income, they are more conscious of adopting management measures to obtain higher income” in the previous study [51]. However, the average income from forestry is just around 1500 dollars per year for forest farmers. Of the respondents, 54.4% agreed that pesticides cause environmental pollution, and 39.7% thought they would increase the economic cost; 68.7% of them purchase pesticides on their own, and regarding their knowledge regarding pesticides, 43.2% gained it from life experience, and 30.5% from business introduction. As a result, considering the limited income, forest farmers usually use cheap but high-pollution chemical pesticides despite being aware of the environmental concerns of these pesticides, resulting in non-point source pollution of pesticides.

(2) The weather conditions have influences on control behaviors. The perceived climate warming stimulates control behaviors, while the perceived precipitation increase reduces the willingness to take control measures. According to Rupert et al. (2017), precipitation and temperature are two important indicators of climate change [67]. Forest farmers perceive the increase in temperature and then predict that FBDs will increase based on their experience. As a result, they would enhance the control behaviors for FBDs. The increased precipitation would negatively affect the survival of pests to inhibit the outbreak of biological disasters to a certain extent, and it would dilute pesticides and negatively affect the control effect. So, when precipitation increases, forest farmers would reduce their willingness to take control measures [68].

(3) Forest farmers think that the organization of the community committee in FBDs’ control is helpful. According to the analysis of the SEM model, the control measures of forest farmers would increase by 0.33 ha if the organization capacity of the village committee increased by one unit. However, it has a substitution relationship with the willingness of forest farmers to purchase pesticides. That increase would lead to a decrease in the fund input from the forest farmers by $26.18. The survey indicated that 41.4% of forest farmers hope the government can provide funds, organization and distribution of ecological pesticides, technical training (27.6%), and other forms of support. In order to alleviate the issues of forest land abandonment and the lack of labor and control technology caused by urbanization [50], community committees organize forest farmers to outsource control work to professional companies to improve the control effect [33]. Under such the organization of the committees, forest farmers’ enthusiasm for control would increase by leveraging the scale economy. Ecological compensation funds from the government are commonly used funding sources by the community committee to organize control work, which can reduce the amount of forest farmers’ investment in purchasing pesticides [60]. However, such an organization being formed by community committees is still in the small-scale pilot stage in China, and further policy support still needs to be developed.

(4) As for the spatial heterogeneity of control behaviors, the values of the eastern region are higher than in the western region, both before and after matching in terms of FBD outbreak area, economic loss, and pesticide cost. The reasons could be more international trade transportation carrying invasive species such as Hyphantria cunea and Bursaphelenchus xylophilus in the eastern region [69], and the more suitable weather and conditions (moister and hotter) for pests [70]. Therefore, the eastern forest farmers thought the FBD outbreak area was larger. Additionally, the forests in the eastern region are mainly for economic benefits, under programs such as the Development of Fast-growing and High-yield Timber Plantation Bases, which are aimed at relieving domestic timber demand pressure and better protecting natural forests [6]. These forestlands mostly plant single-stand tree species with weak resistance to disasters, making the control work more difficult [64]. Thus, the eastern forest farmers thought the economic losses were more serious in the eastern region than in the western region. Eastern forest farmers have a higher education level and are more sensitive to forest biological disasters [71], hence paying more attention and investing more funds, technologies, and control measures to reduce economic losses [35].

(5) The regional perception bias of control behaviors shows that, after matching, the FBD outbreak area, pesticide cost, economic losses, and control area perceived by the eastern farmers would be slightly reduced if the eastern farmers were in the environment of the western forest farmers, as the western region has a large area and sparse population, and large-scale forestlands such as the Three-North (west-north-northeast China) networks of the shelterbelt program offer more ecological services. Thus, if the eastern forest farmers were in the western farmers’ conditions, they would reduce their economic loss perception. In addition, the economic development in the western region is lagging, and the fragile ecological environment overlaps with economic poverty [35,72]. As a result, eastern forest farmers would reduce their pesticide costs and control technology investment if they were under the western conditions. Therefore, all indicators in the eastern region would decline after matching.

After matching, in the western region, FBD outbreak area and pesticide cost would slightly decrease, while economic losses and control area would slightly increase. The control area is higher than in the eastern region if the difference between both regions is excluded (in other words, the western farmers have the same conditions as the eastern farmers). The reason for this is, firstly, that the arid western region with less precipitation, multiple plateaus, and mountains is not conducive to the reproduction and survival of forest pests [36]. Secondly, the most serious FBDs in the western region have strong regularity, and the main challenge is the lack of sufficient funds. Thus, the major control work in the western region is to control disasters within a reasonable range and prevent large-scale disasters [64]. Meanwhile, farmers in the eastern region, with the more difficult control work, need higher control technology and investment funds, as they need to control invasive species such as Hyphantria cunea and Bursaphelenchus xylophilus [73]. Thus, if the western farmers are in the control difficulty of the eastern region, the farmers would think the actual control area and control funds could be moderately reduced in the western region. However, if the western farmers have the same education level and loss assessment ability as the eastern farmers, they would think the economic loss is higher than before matching. In addition, in the western region, the plantation area owned by forest farmers is higher, and the family is more dependent on forestry income. After matching, under the eastern control standard, the control area of western farmers would actually be larger, and far larger than that of eastern farmers.

5. Conclusions

In this paper, the control behaviors and their spatial heterogeneity of forest farmers are investigated via the development of the Cobb-Douglas production function-based econometric models. The major conclusions are summarized below.

First, the impact of household factors on control behaviors: the control behaviors and investment for pesticides of forest farmers are relatively timely. Households whose incomes are mainly from forestry are more willing to invest in pesticide purchases, and the control activities can reduce the economic loss by around 50%. However, there are still issues. The overall control efficiency is low, and the labor input is insufficient. Most of the control knowledge of forest farmers comes from their life experience and business advertisements, while the awareness of the merits of taking control measures in an environmentally friendly manner is weak. Thus, in the case of limited income, forest farmers tend to use cheap but efficient chemical pesticides, while resulting in serious pesticide non-point source pollution.

Second, the impact of external factors on control behaviors: the perception of the temperature rise would encourage forest farmers’ control behaviors, while the perception of rainfall rise would discourage the control behaviors by forest farmers. The organization of community committees in FBD control would improve the enthusiasm for the control work by forest farmers, while it would weaken the willingness of forest farmers to the investment in pesticide purchases.

Third, the regional difference of control behaviors: the outbreak area, economic value loss, prevention, and control area, capital investment measures, temperature and rainfall increase perception, prevention and control knowledge, and village committee organizational ability evaluation on the east side are all higher than those in the western region. After the PSM matching, it is indicated that the control effect would be better if the control intensity of forest farmers in the eastern region is as same as that of farmers in the western region, and the funding and technology support of forest farmers in the western region is as same as that of farmers in the eastern region.

The Suggestions and Study Limitations

The business model of public (government-led)-private (market-driven)-participation (forest farmers) (PPP) was proposed to improve control efficiency. For example, the village committee as the basement executive organization could manage the collective forests through forestry cooperatives and shareholding systems. Forest farmers could become shareholders by inputting forestland or trees. Besides the proportional dividends to forest farmers, the village committee could use part of the forest income as forest control costs for hiring forest rangers or control companies. As for the commercial forest controlled by forest farmers, the village committee could improve control knowledge and efficiency by providing ecological pesticide and technical training.

To enrich the allocation of production factors, including labor, funds, technology, etc., it was proposed that a control warning and monitoring mechanism and control system at the village level should be established to provide disaster information to farmers after summarizing disaster outbreak regulations and meteorological factors. Leveraging the market encouragement mechanism, control companies could be introduced to select young and poor farmers as forest rangers to join the control team and receive control skill training. To reduce the economic risks of forest farmers, the compensation for infected trees and forest insurance could be improved. Combining FBDC with rural revitalization policies could improve the livelihood of farmers and ensure the ecological safety of forestland.

To alleviate regional heterogeneity, underdeveloped areas should be taken seriously. The support could be financial subsidies, special regulation concessions, project investments, etc., from central to local governments. To formulate differentiated control measures, more attention should be given to poverty-stricken areas, key ecological protection areas, and demonstration areas. Additionally, to strengthen the emergency prevention and quarantine capacity at the grass-roots level, emergency control material reserves such as emergency control central stations, emergency pharmaceutical and drug storage areas, etc., should be built at different levels and regions. Under suitable policy adjustments, western forest farmers should have the same control funds as eastern forest farmers, and eastern forest farmers should have the same control efforts and intensities as western forest farmers.

Regarding the limitations of this research, the control difficulties and investments are different for different FBDs. For example, the main control measure for Bursaphelenchus xylophilus is cutting and burning epidemic wood, which leads to a big economic loss. However, it is very challenging to include the differentiated control difficulties of different FBDs in this control behavior study, and there is no literature thus far. Therefore, in the next step, we would like to quantify the control difficulties of different FBDs to further this study.