Will the Experience of Human–Wildlife Conflict Affect Farmers’ Cultivated Land Use Behaviour? Evidence from China

Abstract

Effectively managing and mitigating “human–wildlife conflict” (HWC) by adjusting the use of cultivated land to realise the coexistence of humans and wildlife plays an important role in protecting biodiversity, ensuring food security, improving cultivated land use efficiency, and improving the livelihoods of community residents in nature reserves. Based on the “harmful experience–expectation change–behavioural adjustment” theoretical analysis framework and survey data on 1008 farmers in China’s Giant Panda National Park, this paper uses a binary logit model and a mediating utility model to analyse the effect of HWC experience on the cultivated land use of farmers and the action mechanisms. The results show the following: (1) HWC experience increases the probability that farmers’ cultivated land use behaviour will be adjusted, which will not only increase the probability that farmers’ planting structure of cultivated land will be adjusted but also increase the probability that farmers will abandon cultivated land. (2) Farmers in the national park have a higher probability of adjusting their cultivated land use behaviour after experiencing HWCs than those outside the national park. Farmers in the national park mainly prefer to adjust the planting structure, while farmers outside the park are more inclined to abandon cultivated land. (3) Low-income farmers are more likely to adjust their cultivated land use behaviour after experiencing HWCs than high-income farmers. The low-income group mainly tends to adjust the planting structure of cultivated land, while the high-income group is more inclined to directly abandon cultivated land. (4) The farmers in the group with a high degree of dependence on cultivated land for their livelihood have a higher probability of adjusting their cultivated land use behaviour after experiencing HWC than those in the low-dependency group, and they tend to adjust the planting structure. (5) HWC experience mainly improves farmers’ adjustment of cultivated land use behaviour by improving their awareness of future risks of HWC. In other words, “HWC” will change the behavioural decision-making of farmers and the differences in constraints, such as different regions and different income levels, will lead to differences in “HWC” affecting farmers’ behaviours. The conclusions of this paper not only help in understanding the adjustment mechanism of farmers’ cultivated land use behaviour in the context of HWCs but also provide a decision-making reference for actively changing cultivated land use methods to address HWCs.

1. Introduction

Human–wildlife conflict (HWC) is a global phenomenon that occurs under the constraints of limited natural resources and it affects the livelihoods of millions of people worldwide [1,2]. In many countries, especially developing countries with abundant wildlife resources, the increasingly intensified HWCs have caused serious problems regarding ecology, food security, social governance, and the sustainable development of farmers’ livelihoods as well as other issues [1,3]. HWCs not only pose an enormous obstacle to wildlife conservation but also bring great economic or welfare losses to humans [4,5,6]. For example, in 2005, wild boars led to crop losses of USD 800 million in the United States [7]. In the 2005–2012 period, compensation expenditures for wildlife damage in Europe amounted to USD 41.38 million/year [8]. In October 2020, nearly 8000 farmers in northeastern Nigeria were displaced by the destruction of crops by a herd of approximately 250 elephants [9]. Therefore, handling the relationship and building a harmonious relationship of coexistence between humans and wildlife have become the primary tasks of global biodiversity conservation [10,11,12].

Farmers often show self-adjustment behaviours after experiencing HWCs [13]. For example, previous studies have pointed out that to reduce HWC-related losses to their families, farmers will independently adjust their living and production methods [14]. Such adjustments include independently setting up protective facilities and adjusting the planting structure. In particular, they expand the planting area of economic crops that are not easily damaged [15,16] and even directly abandon the cultivation of seriously damaged land [17,18,19,20]. Although the self-regulation behaviour of farmers has alleviated HWCs to a certain extent [14,21,22,23], it is limited by wildlife protection, land use control, and the strong dependence of farmers on natural resources. Farmers’ voluntary adjustment of cultivated land use behaviour cannot effectively alleviate HWCs, and it is not sustainable because the livelihood results of farmers are damaged [24,25,26,27]. In response to this dilemma, government departments are actively building protective infrastructure [28,29], directly providing economic compensation to farmers [26,30,31], and carrying out the timely implementation of population control [32] and other administrative measures. However, government control measures are short term due to budget constraints [33], and they cannot achieve a long-term sustainable solution to HWC. Therefore, there is an urgent need to adopt policies to effectively intervene in the self-adjustment behaviours of farmers. In particular, agricultural industrial policies should be used to improve the effectiveness of farmers’ self-adaptation behaviours.

Clarifying the impact of HWC on the cultivated land use behaviour of farmers is the premise of introducing targeted governance policies. The impact of HWC on farmers’ cultivated land use behaviour has been discussed. For example, some studies have pointed out that the HWC will promote the adjustment of the planting structure of cultivated land. The planting area of grain and oil crops will be reduced, and the planting area of economic crops that are not easily damaged will be increased [15,16]. Severe HWCs can lead to soil fertility degradation [19] and cultivated land abandonment [17,18,20]. However, studies have been mostly empirical observations and analyses, neither discussing the heterogeneity of the effect of HWC on farmers’ cultivated land use behaviour nor effectively verifying the mechanism of the impact of HWC on farmers’ cultivated land use behaviour. Although some studies have analysed the impact of HWC on farmers’ cultivated land use behaviour, they address only farmers’ cultivated land abandonment behaviour and cannot effectively analyse the heterogeneity and mechanism of impact.

The driving factors of the nongrainisation and abandonment of cultivated land by farmers are also important topics of academic attention. Previous studies have identified socioeconomic and natural environments as the main driving factors [34,35]. For example, some studies have discussed urbanisation [36], rising agricultural factor prices and globalisation (foreign trade development) [35,37], social institutions [38], and other socioeconomic factors. Other studies have discussed the impact of natural conditions such as topography [39], climate, and soil [40,41] on cultivated land use behaviour, especially cultivated land abandonment. Notably, only a few studies have analysed the impact of HWC on cultivated land use behaviour, mainly investigating the impact of HWC on cultivated land abandonment of farmers in mountainous areas [17,18]. Existing research neither discusses the influence of HWC on the adjustment of farmers’ planting structure of cultivated land, especially the nongrainisation of farmers’ cultivated land, nor effectively identifies the mechanism through which HWC affects farmers’ adjustment of their cultivated land use behaviour. Therefore, there is an urgent need to further discuss the impact of HWCs on farmers’ cultivated land use behaviour and its mechanism in the context of the increasingly fierce global conflict between humans and animals.

China is one of the representative countries that analyses the adjustment of “HWC” to farmers’ cultivated land use behaviour. First, China has abundant wildlife resources and implements the strictest wildlife protection system. China’s wildlife protection law does not allow any unauthorised hunting of wildlife and rarely open planned hunting indicators [42]. Second, the forest margins of nature reserves in China are facing increasingly intensified HWCs. According to incomplete statistics reported by 34 provincial-level administrative regions, from 2017 to 2020, there were approximately 3,811,300 incidents of various types of wildlife accidents nationwide (including damage to crops, damage to houses, causing traffic accidents, and attacking people), resulting in approximately 2,432,600 hm² of fields being destroyed or abandoned due to wildlife damage. More than 24,000 people were injured or killed (the wild animals that kill people mainly include Sus scrofa, Ursus thibetanus, and Budorcas taxicolour), resulting in a direct economic loss of more than USD 2.2562 billion. Additionally, the number of incidents and the amount of damage caused by accidents show an increasing trend [43]. Farmers in areas with deep poverty and a concentration of biodiversity are highly dependent on natural resources and face the most severe conflict between livelihood development and the most stringent ecological protection [44]. Finally, to ensure food security, China has implemented a strict system of control over the use of cultivated land. This system does not allow the arbitrary nongrain conversion or “discarding” of cultivated land [45,46]. This policy factor has also led to the intensification of the contradiction between Chinese farmers and wildlife, and the constraints on the adjustment of farmers’ cultivated land use behaviour after experiencing HWCs are more complicated. Therefore, analysing the impact of HWC in China on farmers’ cultivated land use behaviour has important reference value not only for China’s HWC governance but also for the HWC governance of other countries worldwide.

In the context of China’s accelerated implementation of the dual strategy of rural revitalisation and ecological civilisation, this study uses survey data on 1008 farmers from a typical sample of six counties in China’s Giant Panda National Park (aboriginal settlement and agricultural use are also permitted in general control areas) to study the impact of HWC experience on farmers’ cultivated land use behaviour. This study can enhance our understanding of the impact of HWC and expand the framework for analysing the driving factors of farmers’ adjustment of cultivated land use behaviour. Finally, it can help to effectively manage HWCs from the perspective of cultivated land use adjustment and ensure the sustainable livelihood of community residents in nature reserves.

2. Literature Review and Hypothesis

2.1. Literature Review

At the 2004 World Parks Congress held by the International Union for Conservation of Nature (IUCN) in Durban, South Africa, the issue of HWC was first brought to the global stage [47]. HWC has also become the focus of international research in conservation biology [48,49], with continuously expanding research comprehensively covering disciplines such as economics, ecology, sociology, biology, and management, demonstrating that HWC is a comprehensive issue of “nature–society–culture” interrelatedness [50,51].

HWC is the result of a combination of factors. The rapid increase in the global population has led to the expansion of the space of human activity, especially dramatic changes in land use, resulting in the crowding of wildlife habitats. Coupled with the increasingly perfect wildlife protection system, the number of some wildlife has been reduced. Recovery, the combined effect of the two, is the main reason for the intensification of HWC [52,53,54]. Additionally, imperfect governance institutions [27] and climate change [55] are important factors leading to HWC.

The impact of HWCs on human beings includes direct negative impacts and indirect behavioural impacts and has always been a key issue of academic concern. The direct negative effects of HWCs on human life include the destruction of crops, including commercial crops (reducing the productivity of cultivated land), the killing of livestock, the destruction of houses, and attacks on humans [1,56,57]. However, from the perspective of impact heterogeneity, HWC has a greater direct impact on low-income groups [1,44]. There are many indirect effects of HWCs on human life. For example, (1) increased damage to ecosystems, especially the hunting of wildlife [5,6], occurs because for farmers who experience HWCs, protection is more passive [13,44]. (2) The increase in the labour intensity and risks of agricultural operations makes farmers more psychologically and physically vulnerable [57,58,59]. (3) Changes in social systems, especially ecological protection systems [60,61] and changes in family gender roles [1,59], have been promoted.

HWC also affects farmers’ cultivated land use behaviour. For example, previous researchers found that the intensification of HWC will encourage farmers, especially those in hilly areas or nature reserves, to adjust the planting structure of cultivated land [15,16,20] or directly abandon it [17,18,20]. However, this response behaviour is affected by the dependence of farmers’ livelihood capital on cultivated land [24,25]. Farmers who are more dependent on crop income are more inclined to increase their investment in crop protection and are more concerned about crop losses [22,24].

2.2. Hypothesis

Previous studies have shown that HWC will directly affect the output of cultivated land of farmers and indirectly affect the psychological status of farmers. In particular, HWC will strengthen the perception of the risk of HWC and affect farmers’ cultivated land use behaviour to a certain extent. The theory of planned behaviour (TPB) holds that people’s behaviour changes are affected by expectations [62] and that the disaster experience of people or families will affect people’s behaviour by affecting their psychology [63]. On this basis, this study establishes a “harmful experience–expectation change–behavioural adjustment” theoretical analysis framework. This means that HWC experience may affect farmers’ independent adjustment of cultivated land use behaviour by reducing cultivated land output efficiency and increasing their perception of HWC risk. At the same time, the influence of HWC experience on farmers’ cultivated land use behaviour will be regulated by other individuals, families, villages, and macro factors. (Figure 1 ). Based on this theoretical framework, we propose the following hypothesis.

Hypothesis 1.

Compared with farmers without HWC experience, farmers with HWC experience will have an increased probability of adjusting their original cultivated land use behaviour. For farmers with HWC experience, not only the probability of adjusting the planting structure of cultivated land but also the probability of abandoning cultivated land will increase. We divided the samples into two groups with “HWC” experience and without “HWC” experience and used a binary logit selection model to verify the difference in the effect of “HWC” experience on the probability of adjusting farmers’ cultivated land use behaviour.

Hypothesis 2.

Compared with farmers outside national parks who are less regulated by ecological space, farmers in national parks are more dependent on cultivated land for their livelihoods, which will cause farmers with HWC experience to have a higher probability of adjusting their cultivated land use behaviour and be more inclined to adjust the planting structure of cultivated land. Farmers outside national parks will be more inclined to directly abandon farming due to their weak dependence on cultivated land for their livelihoods. We divided the samples into two groups, “inside the park” and “outside the park” and used the binary logit selection model to verify and analyse the differences in the adjustment probabilities of farmers’ cultivated land use behaviour inside and outside the park.

Hypothesis 3.

Compared with high-income farmers, low-income farmers are more likely to adjust their cultivated land use behaviour after experiencing HWCs because their livelihoods are more dependent on cultivated land, and they are more inclined to adjust the planting structure of cultivated land. However, high-income farmers are more inclined to directly abandon farming because of their weak dependence on cultivated land for their livelihoods. We divided the samples into “high-income” and “low-income” groups and used a binary logit selection model to verify the difference in the adjusted probabilities of cultivated land use behaviour after farmers experienced “HWC”.

Hypothesis 4.

Compared with farmers with a low dependence on cultivated land, farmers with a high dependence on cultivated land for their livelihoods have a higher probability of adjusting their cultivated land use behaviour after experiencing HWCs. They also have a higher probability of adjusting the planting structure of cultivated land, while cultivated land abandonment is less likely. We divided the samples into two groups of “high dependence” and “low dependence” and used the binary logit selection model to perform group regression to verify the difference between farmers’ cultivated land dependence on the adjustment of farmers’ cultivated land use behaviour after HWC.

Hypothesis 5.

HWC experience mainly increases the probability that farmers will adjust their cultivated land use behaviour by reducing the cultivated land output efficiency of farmer households and improving their awareness of the risk of damage by wildlife. We will use the mediation effect model to verify the mechanism of how “HWC” experience affects farmers’ cultivated land use behaviour adjustment.

Figure 1. Theoretical framework underlying the impact of HWC experience on cultivated land use behaviour [15,16,17,18,20,22,24,25,62,63].

Figure 1. Theoretical framework underlying the impact of HWC experience on cultivated land use behaviour [15,16,17,18,20,22,24,25,62,63].

3. Data, Variables, and Methods

3.1. Data

HWCs mostly occur around nature reserves with rich wildlife resources [51]. Giant Panda National Park is one of the most typical areas with rich biodiversity in China. It has abundant wildlife resources and numerous indigenous inhabitants [64]. According to the “Giant Panda National Park Master Plan”, the national park spans Sichuan, Shanxi, and Gansu Provinces, involving 152 townships and 120,800 aboriginal residents. Among them, the area in Sichuan involves 119 townships and 89,900 (74.42%) indigenous residents, most of whom belong to China’s contiguous special areas and key national-level poverty alleviation and development areas. The overall income level of the population is lower than the national average. Since the official pilot construction of the Giant Panda National Park in 2018, the most stringent ecological protection policy has been implemented in this area, resulting in the extremely prominent HWCs in this area and seriously affecting the sustainable development of the livelihoods of local residents [65]. Therefore, Giant Panda National Park is somewhat representative. On this basis, this study selected Giant Panda National Park (Sichuan area) as the survey area and selected 6 typical counties as specific survey counties based on the distribution of mountains, ethnic characteristics, and socioeconomic conditions. The data used in this study come from a microsampling survey on the livelihood status of farmers in 6 sample counties that was conducted from July to October 2021. After eliminating invalid samples with serious missing basic information, sample data on 1008 farmers were finally obtained (questionnaire effective response rate of 98.9%). In the information obtained from the questionnaire responses, household income and expenditure, resources, assets, damage caused by wildlife, etc., refer to the situation of farmer households in 2020, while the adjustment of cultivated land use behaviour refers to the situation of farmer households in the first half of 2021. The geographical location of the survey area and the distribution of samples are shown in Figure 2.

3.2. Variables

3.2.1. Dependent Variable

China’s Land Management Law (2020 Revision) defines cultivated land as land directly used for growing crops. In this study, cultivated land specifically refers to land mainly used by farmers to grow grain and oil crops (including corn, wheat, rapeseed, and rice). Additionally, cultivated land use behaviour mainly refers to the different interest behaviour decisions made by farmers based on their own economic interests, the market economic environment, natural environmental factors, and other common constraints. It mainly includes three behaviours: cultivation, circulation, and abandonment [66,67,68]. In this study, cultivated land use behaviour mainly refers to the planting structure and abandonment. The dependent variable is whether farmers’ cultivated land use behaviour has changed. It is measured using a binary selection variable, that is, whether the pattern of farmers’ cultivated land use was adjusted in 2021 from the previous year’s use pattern (1 = yes; 0 = no). Regarding planting structure adjustment, the variable is whether farmers changed from grain and oil crops to cash crops that are less susceptible to wildlife (1 = yes; 0 = no). In regard to abandoning farming, the variable is whether farmers directly abandoned farming on cultivated land damaged in 2021 (1 = yes; 0 = no).

3.2.2. Focal Variables

HWC is defined in a broad sense internationally as a negative impact on any party, even direct or indirect harm to the relationship between humans and wildlife [42,56,69,70,71]. The greatest direct negative impact of wildlife on human beings is the destruction of crops [56]. In this study, the HWC experience refers specifically to the experience of loss to crops of farmer households caused by wildlife (Figure 3), (mainly Sus scrofa, Macaca thibetana, Ursus thibetanus, and Budorcas taxicolour). The focal variable is whether the grain and oil crops grown on the cultivated land of a farmer suffered damage from wildlife in 2020 (1 = yes; 0 = no). There are two replacement variables for HWC experience. (1) First, we use the value of the loss of the cultivated land area per unit of farmer household (total loss value of grain and oil crops/total area of cultivated land actually operated by the family) to measure HWC experience. (2) Second, we use whether the county where the family is located is a pilot county for implementing the wild animal damage insurance policy and wild boar population control policy (1 = yes; 0 = no).

3.2.3. Control Variables

To ensure the accuracy of the estimation results, we followed similar previous studies [18,35,37,66,67], controlling for the main personal characteristics of decision makers (e.g., gender, age, educational level, political status, these individual characteristics will affect the decision-making of farmers’ cultivated land use), family characteristics (e.g., income level, human capital level, cultivated land livelihood dependence, support burden, agricultural production equipment assets, these family characteristics will affect the decision-making of the interviewed farmers on the use of cultivated land), village characteristics (e.g., terrain, geological hazards), and other socioeconomic policy characteristics (e.g., distance from the central county, traffic conditions in production areas, intensity of ecological regulation, intensity of land use regulation, regional and economic policy factors will also affect the output of arable land, thereby affecting farmers’ attitudes towards arable land use). The design, meaning, and statistics of the dependent, focal, and control variables are shown in

Table 1. Empirical variables’ design, variables’ meaning or calculation method, and variables’ basic statistics.

Variable Type	Variable Name	Variable Meaning	Mean	Sd	Min	Max
Dependent variables	Adjustment	Will cultivated land change the original farming method in 2021? (1 = yes; 0 = no)	0.16	0.36	0	1
	Adjustment structure	Will grain and oil crops be adjusted to other cash crops in 2021? (1 = yes; 0 = no)	0.07	0.25	0	1
	Abandoned	Will there be abandoned cultivated land in 2021? (1 = yes; 0 = no)	0.1	0.3	0	1
	Abandoned area	Actual abandoned planting area of cultivated land in 2021 (hm2)	0.077	0.232	0	3.33
Focal variables	HWC_exp	Will farmers experience “HWC” in 2020? (1 = yes; 0 = no)	0.49	0.50	0	1
	Degree of damage	Total value of grain and oil damage/actual cultivated land management area (USD)	55.89	322.14	0	7334.7
	Policy pilot	Is there a pilot county for animal damage insurance? (Implemented from 2020) (1 = yes; 0 = no)	0.700	0.46	0	1
Control variables	Age	year	54.56	11.53	20	89
	Gender	1 = female; 2 = male	1.32	0.47	1	2
	Education level	1 = illiterate; 2 = primary school; 3 = junior high school; 4 = high school and technical secondary school; 5 = college and above	2.43	0.90	1	5
	Party	Whether the respondent is a Communist Party member (1 = yes; 0 = no)	0.13	0.33	0	1
	Per_income	Total household income/total population (USD, logarithm of regression)	3143.79	2298.1	97.84	19,070.2
	Dep_agriculture	Total agricultural income/total household income (%)	0.33	0.31	0	2.170
	Capital	The level of family human capital (education, health, gender, labour force, total by entropy weight method)	1.69	0.49	0.45	3.790
	Machinery	Agricultural Machinery Fixed Assets Status (sum of quantities, tractor = 1; small tiller = 0.5)	0.27	0.23	0	1
	Par_burden	Number of household labour force/total household population	0.2	0.41	0	1
	Knowledge	Respondents’ knowledge of eco-compensation policies (1 = do not understand at all; 2= just heard about it; 3 = know roughly; 4 = understand a little; 5 = understand very well)	1.59	0.72	1	5
	Terrain	The terrain of the village (1 = plateau; 2 = mountainous; 3 = deep hill; 4 = shallow hill; 5 = plain)	2.17	0.50	2	4
	Distance	The distance from the residence of famers to the central county seat (actual kilometres, KM)	42.22	26.84	1	120
	Fre_wildlife damage	Has the frequency of local wildlife damage increased in recent years? (1 = yes; 0 = no)	2.87	0.85	1	5
	Agricultural_control	Agricultural production and operation are subject to the intensity of ecological regulation (1 = no control; 2 = operate to expand the input of operation control elements; 3 = maintain the original; 4 = prohibit operation)	1.59	0.91	1	4
	Cultivated land control	Intensity of cultivated land use control (1 = no regulation; 2 = only grain and oil crops; 3 = prohibited use by sector; 4 = most prohibited use)	3.01	0.87	1	4
	Traffic	Residential traffic conditions (1 = very bad, muddy road; 2 = poor, rural cement trail; 3 = average, near the cement country road; 4 = good, near the county road; 5 = very good, near the national road)	2.93	0.81	1	5
Other variables	Cultivated land area	Actual area (hm2)	0.34	0.42	0	4
	Sum_losses	The actual amount of damage to the family caused by all wildlife (USD, estimated according to the price of the year)	991.21	2521.9	0	38,140.4
	Grain_losses	The loss of grain and oil caused by wildlife to farmers’ households (USD, estimated according to the price of the year)	149.33	425.7	0	8801.6

.

3.3. Methods

Based on the characteristics of the dependent variable, the benchmark model is a binary logit model. The benchmark model is designed as follows.

$$\log (\frac{{{\displaystyle p}}_i}{1-{{\displaystyle p}}_i})={{\displaystyle \lambda }}_1*{{\displaystyle hwc\_\exp }}_i+{\displaystyle \sum _{k=1}^n{{\displaystyle \lambda }}_k*{{\displaystyle control}}_i^n}+\epsilon$$

(1)

In Equation (1), p_i is the probability that farmer i will adjust the use behaviour of the cultivated land that he or she manages after experiencing HWC. hwc_expi is a dummy variable representing whether the farmer experienced HWC in 2020 (1 = yes; 0 = no), and ${{\displaystyle \lambda }}_1$ represents the estimated coefficient of the focal variable and represents the value of the average marginal effect. Control i represents other control variables that affect farmers’ adjustment of cultivated land use behaviour, such as household head characteristics, family characteristics, and village characteristics. ${{\displaystyle \lambda }}_k$ represents the value of the average marginal effect of the control variables. ε is the random disturbance term.

4. Results

4.1. Descriptive Statistical Results

4.1.1. Basic Statistics

Table 1 reports the basic statistical results of the main variables. The results show that among the 1008 farm households surveyed, 49% experienced HWCs in their grain and oil crops. The average loss caused by wildlife to farm households was USD 991.21/household, accounting for 6.9% of total household income. For grain and oil crops, the loss was USD 149.3/household, accounting for 1.03% of total household income, while for grain and oil crops, the loss was USD 305.85/household, accounting for 2.12% of total household income. Among the sample farmers, 16% changed their cultivated land use behaviour, 6% changed their planting structure, and 10% abandoned their cultivated land. The adjustment direction of the planting structure was mainly replanting tea, traditional Chinese medicinal materials, and other leguminous crops that are not easily damaged. The average area of abandoned cultivated land was 0.077 hm², accounting for 21.7% of the total cultivated area per household. After suffering losses, only 26 households (accounting for 5.3% of the damaged households) received compensation from the government. The minimum compensation was only USD 7.33, the highest was USD 220.04, and the average was only USD 33.59, a difference of USD 276.5 from the average loss. This result shows that wildlife has an enormous negative impact on the livelihood of local farmers. The statistics of the variables are presented in Table 1.

4.1.2. Mean t Test

First, we clarify the differences in cultivated land use behaviour and household income between farmers with and without HWC experience. We grouped based on the presence or absence of HWC experience and performed mean t tests on the key variables.

Table 2. Group mean t test of cultivated land use behaviour, loss to cultivated land by HWC and household income characteristics without “HWC” experience and with “HWC” experience.

Variables	G1 (No)	Mean 1	G2 (Yes)	Mean 2	Mean Diff
Adjustment	516	0.0780	492	0.240	−0.162 ***
Adjustment structure	492	0.0330	419	0.107	−0.075 ***
Abandoned	500	0.0480	447	0.163	−0.115 ***
Abandoned area (hm2)	516	0.052	492	0.102	−0.048 ***
Per_cultivated land_loss (USD)	516	0	492	114.49	−114.49 ***
Policy pilot	516	0.539	492	0.862	−0.323 ***
Per_income (USD)	516	3373.52	492	2902.78	470.78 ***

Note: statistics in parentheses *** p < 0.01.

reports the mean t test results, which are as follows: (1) The probability of cultivated land use adjustment, the planting of undamaged cash crops, and cultivated land abandonment by farmer households with HWC experience is significantly higher than that by farmer households without HWC experience. (2) The value of damage to grain and oil crops for farmers with HWC experience is significantly higher than that for farmers without HWC experience. Additionally, farmers with HWC experience are mostly located in wildlife damage insurance pilot counties. For farmers with HWC experience, the level of household income is significantly lower than that of farmer households without HWC experience.

4.2. Empirical Results

4.2.1. Basic Regression Results

The results reported in Table 3 are the mean marginal effect values, and the numbers in parentheses are the t test statistics. Models (1)–(2) represent the empirical results of whether the interviewed households adjusted their cultivated land use in 2021. Model (1) does not add the control variables and Model (2) adds the control variables, including those related to the individual farmer household head, the family economy, village characteristics, and other geographical and policy factors. The results show that the value of the average marginal effect of damage due to wildlife is 0.137, and this result is significant at the 1% level. This finding means that when a farmer’s family experiences HWC, the probability that the farmer’s family will adjust its cultivated land use behaviour increases by 13.7%. The values of the average marginal effect of age, household per capita income, household livelihood dependence on agriculture, and ecological compensation policy cognition on the adjustment of cultivated land use behaviour are −0.002, −0.031, −0.129, and −0.036, respectively. These results are significant at the 1% and 5% levels. These findings mean that when age, the per capita income of the family, the dependence of the family’s livelihood on agriculture, and ecological compensation policy cognition increase by one unit, the probability that the farmer’s family will adjust its cultivated land use method decreases by 0.2%, 3.1%, 12.9%, and 3.6%, respectively. In addition, the value of the average marginal effect of residential traffic conditions on the adjustment of cultivated land use behaviour is 0.051, and this result is significant at the 1% level. That is, if the traffic conditions in the local residential area increase by one unit, the probability that farmer households will adjust their cultivated land use behaviour increases by 5.1%. The practical economic implication of these results is that when a farmer’s family experiences damage from wildlife and the traffic conditions are better, the farmer’s family has a higher probability of adjusting its cultivated land use behaviour. The more the livelihood depends on agriculture and the higher the ecological compensation policy, the lower the probability of adjusting cultivated land use behaviour will be.

Models (3)–(4) show the empirical results of whether the interviewed households adjusted the original grain and oil planting varieties to other cash crop varieties that are less susceptible to wildlife damage in 2021. Model (3) is the result without adding the control variables and Model (4) is the estimation result with the control variables added. The results show that the estimated coefficient of the adjustment of crop varieties due to wildlife is 0.054, and this result is significant at the 1% level. This finding means that when a farmer’s family experienced HWC, it adjusted the amount of crop varieties that it cultivated, with the probability increasing by 5.4%. The estimated coefficients of traffic conditions in production areas and the ecological regulation of agricultural management are 0.03 and 0.025, respectively, and they are both significant at the 1% level. That is, when traffic conditions and ecological regulation increase by one unit, the probability that farmers’ planting structure of cultivated land will be adjusted increases by 3% and 2.5%, respectively. Additionally, the value of the average marginal effect of ecological compensation policy cognition is −0.044, and this result is significant at the 1% level. This finding means that when farmer households’ awareness of local ecological compensation policies increases by 1 unit, the probability that these households will adjust the planting structure of cultivated land decreases by 4.4%. These results show that damage due to wildlife, better traffic conditions and stronger ecological regulation will significantly increase the probability of cultivated planting structure adjustment. In contrast, awareness of ecological compensation policies will reduce the probability of cultivated planting structure adjustment.

Models (5) and (6) show the empirical results of whether farmer households abandoned cultivated land in 2021. Model (5) is the result without adding the control variables and Model (6) is the estimation result with the control variables added. The results show that the value of the average marginal effect of HWC experience on cultivated land abandonment is 0.109, and this result is significant at the 1% level. That is, when a farmer’s family’s HWC experience increases by 1 unit, the rate of cultivated land abandonment is 0.109, and the probability of abandonment increases by 10.9%. The value of the average marginal effect of the age of the household head on cultivated land abandonment is 0.002, and this result is significant at the 5% level. That is, if the age of the household head increases by 1 year, the probability of abandoning cultivated land decreases by 0.2%. This result may be because the older the farmer is, the lower the chances that the farmer will go out to work and the greater the extent to which the farmer can only stay home to work in agriculture. Additionally, the value of the average marginal effect of residential traffic conditions on cultivated land abandonment is 0.041, and this result is significant at the 1% level. That is, for every 1 unit increase in the traffic conditions in the residential area of a farmer, the probability of cultivated land abandonment increases by 4.1%.

Furthermore, the wild animal damage insurance adopted by county-level governments (including the wild boar population regulation policy implemented in early 2021) and the self-protection measures taken by farmer households have positive and significant effects on the adjustment of farmers’ cultivated land use behaviour. This result shows that the government’s current wildlife damage control policies and farmers’ self-adjustment policies have not effectively reduced the adjustment of farmers’ cultivated land use behaviour. There are two main reasons for this result. (1) The government’s wildlife damage control policy is less effective. Only 26 farmers (accounting for 5.3% of the damaged households) received insurance compensation after suffering losses caused by wildlife. The minimum compensation amount was only USD 7.3, the highest was USD 219.02, and the average was only USD 33.43, which was USD 275.24 lower than the average loss. Additionally, the wild boar population control policy implemented by the Sichuan provincial government started in early May 2021. The method of hunting wild boars only with hunting dogs is limited, and the actual hunting efficiency is extremely poor. (2) The prevention and control measures taken by farmers themselves cannot fundamentally solve HWC. The survey statistics show that only 222 farm households (accounting for 45% of the sample of farmer households who experienced damage) took self-control measures among the farmer households that suffered damage from wildlife, of which 157 (accounting for 70.7% of the farmer households who took measures) took control measures. For these households, there is only a slight control effect, and for the rest, there is no actual control effect.

Table 3. Basic regression results of the influence of HWC on the adjustment of cultivated land use behaviour of farmers.

	Adjustment		Adjustment Structure		Abandoned
	Model (1)	Model (2)	Model (3)	Model (4)	Model (5)	Model (6)
HWC_exp	0.166 ***	0.137 ***	0.078 ***	0.054 ***	0.120 ***	0.109 ***
	(6.88)	(5.155)	(3.978)	(2.707)	(5.319)	(4.303)
Age		−0.002 *		−0.001		−0.002 **
		(−1.869)		(−0.948)		(−2.264)
Gender		0.002		−0.003		0.013
		(0.074)		(−0.158)		(0.627)
Education		0.001		−0.002		0.006
		(0.036)		(−0.1586)		(0.477)
Party		0.006		0.031		−0.017
		(0.165)		(1.206)		(−0.513)
Per_income		−0.031 *		−0.003		−0.021
		(−1.936)		(−0.238)		(−1.502)
Dep_agriculture		−0.129 ***		−0.021		−0.041
		(−3.100)		(−0.714)		(−0.36)
Capital		−0.011		−0.019		−0.015
		(−0.297)		(−0.695)		(−0.71)
Par_burden		0.020		−0.038
		(0.253)		(−0.664)
Machinery		0.034		0.007		0.028
		(1.351)		(0.388)		(1.38)
Knowledge		−0.036 **		−0.044 ***		−0.016
		(−2.142)		(−3.028)		(−1.10)
Terrain		0.016
		(0.749)
Geological disaster		0.001				0.004
		(0.074)				(0.37)
Cultivated land control		0.004				0.017
		(0.309)				(1.62)
Traffic		0.051 ***		0.030 ***		0.041 ***
		(3.743)		(2.719)		(3.38)
Distance				−0.000
				(−0.737)
Agricultural_control				0.025 ***
				(2.916)
Bank				−0.028
				(−1.604)
Policy pilot		0.077 **		0.026		0.078 **
		(2.37)		(1.10)		(2.509)
Self-prevention		0.061 **		0.044 **		0.036
		(2.31)		(2.18)		(1.561)
N	1008	1008	911	911	947	947
R2	0.059	0.128	0.046	0.142	0.056	0.107

Note: statistics in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01. In addition, the estimated coefficients are the mean boundary effect values, while the t values are in parentheses.

Table 3. Basic regression results of the influence of HWC on the adjustment of cultivated land use behaviour of farmers.

	Adjustment		Adjustment Structure		Abandoned
	Model (1)	Model (2)	Model (3)	Model (4)	Model (5)	Model (6)
HWC_exp	0.166 ***	0.137 ***	0.078 ***	0.054 ***	0.120 ***	0.109 ***
	(6.88)	(5.155)	(3.978)	(2.707)	(5.319)	(4.303)
Age		−0.002 *		−0.001		−0.002 **
		(−1.869)		(−0.948)		(−2.264)
Gender		0.002		−0.003		0.013
		(0.074)		(−0.158)		(0.627)
Education		0.001		−0.002		0.006
		(0.036)		(−0.1586)		(0.477)
Party		0.006		0.031		−0.017
		(0.165)		(1.206)		(−0.513)
Per_income		−0.031 *		−0.003		−0.021
		(−1.936)		(−0.238)		(−1.502)
Dep_agriculture		−0.129 ***		−0.021		−0.041
		(−3.100)		(−0.714)		(−0.36)
Capital		−0.011		−0.019		−0.015
		(−0.297)		(−0.695)		(−0.71)
Par_burden		0.020		−0.038
		(0.253)		(−0.664)
Machinery		0.034		0.007		0.028
		(1.351)		(0.388)		(1.38)
Knowledge		−0.036 **		−0.044 ***		−0.016
		(−2.142)		(−3.028)		(−1.10)
Terrain		0.016
		(0.749)
Geological disaster		0.001				0.004
		(0.074)				(0.37)
Cultivated land control		0.004				0.017
		(0.309)				(1.62)
Traffic		0.051 ***		0.030 ***		0.041 ***
		(3.743)		(2.719)		(3.38)
Distance				−0.000
				(−0.737)
Agricultural_control				0.025 ***
				(2.916)
Bank				−0.028
				(−1.604)
Policy pilot		0.077 **		0.026		0.078 **
		(2.37)		(1.10)		(2.509)
Self-prevention		0.061 **		0.044 **		0.036
		(2.31)		(2.18)		(1.561)
N	1008	1008	911	911	947	947
R2	0.059	0.128	0.046	0.142	0.056	0.107

Note: statistics in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01. In addition, the estimated coefficients are the mean boundary effect values, while the t values are in parentheses.

4.2.2. Heterogeneity Analysis

(1) Differences between inside and outside the national park. To further understand the differences in the adjustment of cultivated land use behaviour by farmers inside and outside the national park, based on the location relationship between the residence of farmers and Giant Panda National Park, we divided the sample groups into two groups: the group inside the national park and the group outside the national park. The intensity of ecological regulation faced by farmer households inside the national park is far greater than that faced by farmer households outside the park.

Table 4. Heterogeneity analysis of the influence of HWC on cultivated land use behaviour adjustment of farmers inside and outside of the Giant Panda National Park.

	Adjustment		Adjustment Structure		Abandoned
	Model (In)	Model (Out)	Model (In)	Model (Out)	Model (In)	Model (Out)
HWC_exp	0.145 ***	0.103 ***	0.085 ***	0.015	0.053	0.126 ***
	(3.934)	(2.721)	(2.729)	(0.698)	(1.431)	(3.211)
Control	Yes	Yes	Yes	Yes	Yes	Yes
N	543	465	494	417	498	449
R2	0.146	0.145	0.171	0.181	0.199	0.143

Note: statistics in parentheses *** p < 0.01. In addition, the estimated coefficients are the mean boundary effect values, while the t values are in parentheses.

reports the difference in the impact of HWC experiences on the cultivated land use behaviour of farmers under the constraints of different ecological space regulation intensities. The results show that the probability of adjustment of cultivated land use behaviour after HWC experience will be significantly increased, regardless of whether the farmer household is inside or outside the park. The probability of adjustment for farmer households inside the park is 4.2% higher than that for farmer households outside the park. However, there are differences in the direction of adjustment. Farmers in the park have more choices to adjust the planting structure of their cultivated land after experiencing HWCs, while farmers outside the park more often choose to directly abandon cultivated land. These results show that different ecological space regulatory constraints lead to differences in the cultivated land use behaviours of farmer households caused by wildlife. However, the main reason for these differences in impacts may be the differences in the livelihood dependence of farmers on cultivated land (cultivated grain and oil income/household total income). Comparing the dependence of household livelihoods on cultivated land and agriculture inside and outside the national park, we find that the average dependence of household livelihoods on cultivated land in the sample group outside the national park is 0.036, while the average dependence on cultivated land for farmers inside the park is 0.044. Farmers outside most national parks have more opportunities to engage in off-farm employment because they are closer to central towns in China. As a result, most farmers outside national parks are less dependent on cultivated land than farmers in national parks. Clearly, the livelihoods of farmer households are more dependent on cultivated land; thus, they have more motivation to choose to adjust the planting structure rather than directly abandoning cultivated land. If farmers inside the park choose to directly abandon farming, their livelihood will be more damaged. In contrast, farmers outside the national park are less dependent on cultivated land. Thus, the negative impact of wildlife damage on their livelihoods is relatively small, and the probability of directly choosing to abandon cultivated land is higher.

(2) Differences in income levels. The estimation results in Table 3 show that the higher the income level is, the lower the probability of adjusting cultivated land use behaviour. This finding shows that there may be differences in the adjustment of cultivated land use behaviours by farmers with different income levels. Therefore, we further analysed the heterogeneity of the cultivated land use behaviours of farmer households after experiencing HWCs under the constraints of different income levels. We classify high and low categories based on the percentile distribution of household per capita income levels. The high-income group is the sample above the 50% percentile of household per capita income, while the low-income group is the sample below the 50%.

Table 5. Income differences.

	Adjustment		Adjustment Structure		Abandoned
	Model (high)	Model (low)	Model (high)	Model (low)	Model (high)	Model (low)
HWC_exp	0.092 **	0.192 ***	0.033	0.085 ***	0.096 ***	0.068
	(2.536)	(4.927)	(1.264)	(2.723)	(2.873)	(1.425)
Control	Yes	Yes	Yes	Yes	Yes	Yes
N	504	504	465	446	475	472
R2	0.108	0.145	0.113	0.225	0.133	0.114

Note: statistics in parentheses ** p < 0.05, *** p < 0.01. In addition, the estimated coefficients are the mean boundary effect values, while the t values are in parentheses.

reports the estimation results for the two income groups. The results show that the experience of damage caused by wild animals will increase the probability of farmers adjusting their cultivated land use behaviour. However, the probability of farmer households adjusting their cultivated land use behaviour in the low-income group is 10% higher than that in the high-income group. There is also a significant difference in the direction of adjustment between the high-income and low-income groups. The high-income group tends to abandon cultivated land, while the low-income group chooses to adjust the planting structure.

(3) Differences in the level of cultivated land livelihood dependence. To further verify that the dependence of farmer households’ livelihood on cultivated land is the fundamental reason affecting the adjustment direction of the cultivated land use behaviour of high- and low-income groups inside and outside the park, we grouped farmers in terms of high and low dependency based on the 50% percentile of dependency levels. The group above the 50% quantile is the high-dependency group, and the group below the 50% quantile is the low-dependency group.

Table 6. Differences in the dependence rate of livelihood on cultivated land.

	Adjustment		Adjustment Structure		Abandoned
	Model (high_d)	Model (low_d)	Model (high_d)	Model (low_d)	Model (high_d)	Model (low_d)
HWC_exp	0.174 ***	0.139 ***	0.090 **	0.055 **	0.102 ***	0.139 ***
	(3.502)	(4.176)	(2.242)	(2.212)	(3.192)	(3.025)
Control	Yes	Yes	Yes	Yes	Yes	Yes
N	535	473	477	434	501	446
R2	0.142	0.129	0.171	0.153	0.167	0.114

Note: statistics in parentheses ** p < 0.05, *** p < 0.01. In addition, the estimated coefficients are the mean boundary effect values, while the t values are in parentheses.

reports the regression results of the differences between groups with different degrees of dependence. The results show that the high-dependency group had a 3.5% higher probability of adjusting its cultivated land after experiencing damage from wild animals than the low-dependency group. The probability of adjusting the planting structure was also 3.5% higher, but the probability of abandoning cultivated land was 3.7% lower. This result further proves that the degree of dependence of farmer households’ livelihood on cultivated land will significantly affect the characteristics of farmers’ cultivated land use behaviour after experiencing HWC. Additionally, the higher the degree of dependence is, the stronger the adaptive behaviour.

4.2.3. Robustness Check

To test the robustness of the estimation results of the basic model, we conducted robustness tests by finding substitute variables for the focal variables and changing the size of the estimated sample. We used an artificial method (selecting only samples with HWC experience) and propensity score matching (PSM) [72] to select subsamples for robustness estimation. In the pilot area of the wildlife damage insurance policy, the area’s HWC level is higher than that of the nonpilot area. For this reason, we used the animal harm insurance policy pilot as a replacement for whether farmers experienced HWC. We also took the loss per unit of cultivated land (the greater the loss, the more serious the damage) as a replacement variable for HWC experience, and only the samples of farmer households with HWC experience were selected for the regression.

Table 7. Robustness test.

	M1 (Substitute Core Explanatory Variables)			M2 (Change the Sample Size)
	Adjustment	Adjustment structure	Abandoned	Adjustment	Adjustment structure	Abandoned
	Model (1)	Model (2)	Model (3)	Model (4)	Model (5)	Model (6)
Insurance_test	0.128 ***	0.048 **	0.085 ***
	(4.142)	(2.06)	(3.265)
Degree of damage				0.033 *	0.006	0.038 **
				(1.863)	(0.392)	(2.277)
Control	Yes	Yes	Yes	Yes	Yes	Yes
N	1008	911	947	492	419	447
R2	0.103	0.119	0.093	0.112	0.111	0.125
PSM_logit:				Model (7)	Model (8)	Model (9)
HWC_exp				0.111 ***	0.031	0.072 *
				(3.08)	(1.152)	(1.801)
control				Yes	Yes	Yes
N				398	359	372
R2				0.109	0.251	0.167

Note: statistics in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01. In addition, the estimated coefficients are the mean boundary effect values, while the t values are in parentheses.

reports the estimation results of the robustness test. Models (1)–(3) of M1 are the estimation results from using the dummy variable of the wildlife damage insurance pilot policy to replace HWC experience, while Models (4)–(6) of M2 are the estimation results from using damage per unit area of cultivated land. The degree replaces HWC experience and estimates the results only for the sample of farmers with HWC experience. Models (7)–(9) of M2 are the estimation results of selecting samples using PSM. The results show that the adjustment of the planting structure in the estimation of the results based on the reduced sample size is not significant. However, all other robustness tests show that the experience of wildlife damage increases the probability that farmers will adjust their cultivated land use behaviour. Clearly, the estimation results of this study are robust overall.

4.3. Analysis of the Action Mechanism

The theoretical analysis points out that the main mechanism of the impact of HWC on farmers’ cultivated land use behaviour may be to promote the adjustment of farmers’ cultivated land use behaviour by reducing farmers’ value expectations of the output of cultivated land. There are two main channels through which farmers’ expectations of cultivated land output are reduced. (1) Reducing the output per unit area of cultivated land leads to direct economic losses, and (2) increasing farmer households’ awareness of future HWC experience risks indirectly increases farmers’ risk awareness. We use the output value per unit of cultivated land in 2020 as the representative value of the output efficiency of cultivated land, and we take the respondents’ perception of the frequency of future damage by wildlife as the representative value of the risk of damage by wild animals. The mediating effect model can identify the mechanism of action between two variables well [73,74]. Therefore, this study adopts the principle of the mediating effect model to estimate the mechanism of the impact of HWC on the adjustment of farmers’ cultivated land use behaviour. The specific equations of the mediating effect are simultaneously designed as follows:

$$\log n\frac{{{\displaystyle p}}_i}{1-{{\displaystyle p}}_i}n={{\displaystyle \lambda }}_1*{{\displaystyle hwc\_\exp }}_i+{\displaystyle \sum _{k=1}^n{{\displaystyle \lambda }}_k*{{\displaystyle control}}_i^n}+\epsilon$$

(2)

$${{\displaystyle \mathrm{M}}}_{\mathrm{i}}=\alpha +{{\displaystyle \lambda }}_2*{{\displaystyle hwc\_\exp }}_i+{\displaystyle \sum _{h=1}^n{{\displaystyle \lambda }}_{\mathrm{h}}*{{\displaystyle control}}_i^n+\epsilon }$$

(3)

$$\log n\frac{{{\displaystyle p}}_i}{1-{{\displaystyle p}}_i}n={{\displaystyle \lambda }}_3*{{\displaystyle hwc\_\exp }}_i+{{\displaystyle \lambda }}_4*{{\displaystyle M}}_i+{\displaystyle \sum _{k=1}^n{{\displaystyle \lambda }}_k*{{\displaystyle control}}_i^n}+\epsilon$$

(4)

Equation (2) is the benchmark model for this study, which is the same as Equation (1), and the meaning of the coefficients will not be repeated here. Equation (3) is the estimation model of the influence of HWC experience on the mediating variable. Mi represents the mediating variable, α is the intercept term, and ${{\displaystyle \lambda }}_2$ is the estimated coefficient of the mediating variable. Equation (4) is the effect of HWC experience and the intermediary variables on the adjustment of farmers’ cultivated land use behaviour. ${{\displaystyle \lambda }}_3$ denotes the estimated coefficient of HWC experience, and ${{\displaystyle \lambda }}_4$ denotes the estimated coefficient of the mediator variable. According to the basic principle of the mediating effect model, if the mediating effect of mediating variable Mi exists, then ${{\displaystyle \lambda }}_1$, ${{\displaystyle \lambda }}_2$, ${{\displaystyle \lambda }}_3$, and ${{\displaystyle \lambda }}_4$ will be significant, and the estimated value of ${{\displaystyle \lambda }}_3$ will decrease due to the addition of the mediating variable, proving that there is a partial mediating effect of Mi.

Table 8. Results of the mediation effect test.

	Adjustment	Per_Effect_Land	Fre_Wildlife Damage	Adjustment	Adjustment
	Model (1)	Model (2)	Model (3)	Model (4)	Model (5)
HWC_exp	0.137 ***	−146.697	0.401 ***	1.49 ***	0.111 ***
	(5.155)	(−0.79)	(15.461)	(4.593)	(3.703)
per_effect_land				0.000
				(0.855)
Fre_wildlife damage					0.087 **
					(2.262)
Control variable	Yes	Yes	Yes	Yes	Yes
N	1008	829	1008	829	1008
R2	0.059	0.046	0.205	0.131	0.113
Sobel (T)		Coef	Std Err	Z	P > |Z|
Adjustment		8.57	3.734	2.296	0.021

Note: statistics in parentheses ** p < 0.05, *** p < 0.01. In addition, the estimated coefficients are the mean boundary effect values, while the t values are in parentheses.

reports the mechanism of the impact of HWC experience on the adjustment of farmers’ cultivated land use behaviour. The results show that only the mediating effect of HWC risk perception is significant, while the mediating effect of reducing the output efficiency per unit area of cultivated land is not significant. Therefore, we conducted the Sobel test [75] on the role of the HWC risk perception channel. The results of the Sobel test (T) show that the mediating effect of HWC risk perception is significant at the 5% level. This result also shows that HWC experience increases the probability that farmers will adjust their cultivated use behaviour mainly by improving their perception of HWC risks.

Thus far, the HWC risk perception channel in hypotheses H1, H2, H3, H4, and H5 has been confirmed. Why is there no mechanism for reducing the output efficiency per unit area of cultivated land? Although the HWC experience reduces the output efficiency per unit area of cultivated land (average reduction of USD 318.2/hm²), this difference is not significant. We think that the possible reason is that farmer households will alleviate the loss of cultivated land by adjusting the planting structure. According to the survey statistics, 61 farmer households (accounting for 38.6% of adjustment behaviours) experienced damage due to wildlife. Clearly, since farmers will adjust the planting varieties after facing this harmful experience, such adjustment may also narrow the difference in the output level per unit area of cultivated land for households with and without wildlife damage.

5. Discussion

Based on microsurvey data on 1008 farmer households in six counties in China’s Giant Panda National Park and on the basis of the “harmful experience–expectation change–behavioural adjustment” theoretical analysis framework, this study focused on the effect of HWC experience on the cultivated land use behaviour of farmer households under the constraints of different land spatial locations, different levels of livelihood dependence on cultivated land, and different income levels. (1) This study constructed a “harmful experience–expectation change–behavioural adjustment” theoretical analysis framework and focused on identifying the mechanism through which HWC risk perceptions affect farmers’ adjustment of cultivated land use behaviour. (2) Compared with previous empirical observational studies, this study used microsurvey data on 1008 farmers in six counties in China with typical representatives to verify the impact of HWC on farmers’ cultivated land use behaviour and the impact mechanism, and the conclusions are more representative and have effective reference value. (3) This study focused on analysing the differences in cultivated land use behaviours under different ecological spatial locations, different levels of cultivated land livelihood dependency, and different levels of income for farmers after experiencing HWCs.

In contrast to previous studies, this study found that HWC will lead to the adjustment of the planting structure of cultivated land [15,16,20] or the direct abandonment of cultivated land [17,18,20]. The results of this study also further support the conclusion that HWC affects farmers’ cultivated land use behaviour. However, different from previous studies, we further verified the mediating role of wildlife damage risk perception in the impact of HWC on farmers’ cultivated land use behaviour. Furthermore, this study analysed the differences in the direction of the adjustment of cultivated land use behaviour among farmers with different locations, levels of dependency, and incomes.

This study concluded that HWC will increase the probability that farmers will adjust their cultivated land use behaviour. Furthermore, it proved the idea of previous researchers that farmers have a certain degree of initiative in HWC governance [13,22,24], and this difference in subjective proactive behaviour is affected by the degree of dependence on cultivated land or agriculture for household livelihoods [24,25]. In addition, previous researchers have found that only farmers who are more dependent on crop income are more inclined to increase their investment in crop protection and are more concerned about crop losses [22,24]. This study further verified the subjective initiative of farmers under the constraint of HWC, especially the specific content of the independent adjustment of cultivated land use behaviour and the differences in independent behaviour among different groups. At the same time, we found that farmers whose household livelihoods have low dependence on cultivated land make adaptive adjustments to the planting structure of cultivated land, while farmers with high dependence directly abandon farming. Clearly, these conclusions have important reference value for improving cultivated land use policy and provide a useful reference for effectively governing HWC from the perspective of adjusting the mode of cultivated land use.

For a long period of time, there has been an increasing trend of nongrainisation of cultivated land [76,77] and abandonment of cultivated land [18,78]. This trend will seriously threaten the population and cultivated land resources as well as the food security of China, which is dominated by hilly areas [45,46]. This study draws the conclusion that HWC will significantly increase the probability of nongrainisation and abandonment of cultivated land, and it further expands the driving factors of the nongrainisation and abandonment of cultivated land in hilly areas. This study holds great reference value for China and other developing countries to come to a new understanding of national food security when HWCs become increasingly severe.

There are still some limitations in this study that future research can address. (1) Most HWCs have existed for a long period of time; thus, the impact may also exist for a long period of time. However, this study discussed the changes in the cultivated land use behaviour of farmers only over a one-year period, and it only performed research on farmers with and without HWC experience. Due to the differences in grouping, cultivated land use behaviours were divided into only two categories based on the adjustment of the planting structure and the abandonment of cultivated land. Future research can refine cultivated land use behaviour and further identify the impact of HWC on cultivated land use behaviour over a longer period of time. (2) The adjustment of farmer households’ cultivated land use behaviour may also be affected by the expansion of agricultural technology and the neighbourhood demonstration effect. However, due to insufficient consideration in the research design, the impact of these two key variables on the adjustment of farmers’ cultivated land use behaviour was not effectively considered for the time being. Future research can further explore the regulatory role of agricultural technology expansion and the neighbourhood demonstration effect after wildlife damage occurs. (3) The ownership of wildlife in China belongs to the state, but in many countries worldwide, it belongs to private landowners. Clearly, differences in property rights to wildlife lead to differences in impact and governance. Future research can further compare the differences in the impact of different allocations of wildlife property rights on farmers’ cultivated land use behaviour.

6. Conclusions and Implications

The main conclusions of this study are as follows.

(1) HWC experience will increase the probability that farmers will adjust their cultivated land use behaviour, which will not only increase the probability that farmers will adjust the planting structure of cultivated land but also increase the probability that they will abandon cultivated land. (2) Farmer households in the national park have a higher probability of adjusting their cultivated land use behaviour after experiencing HWCs than those outside the national park. However, there are differences in the direction of adjustment. Among them, farmers in the national park mainly prefer to adjust the planting structure, while outside the national park, they prefer to abandon farming altogether. (3) Low-income farmers have a higher probability of adjusting their cultivated land use behaviour after experiencing HWCs than high-income farmers. However, there are also significant differences in the direction of adjustment. Among them, the low-income group is mainly inclined to adjust the planting structure of cultivated land, while the high-income group is inclined to directly abandon cultivated land. (4) For farmers in the high-dependence group, the probability of adjusting cultivated land use behaviour after experiencing HWC is higher than that for farmers in the low-dependence group, and they mainly tend to adjust the planting structure. (5) HWC experience mainly increases the probability that farmers will adjust their cultivated use behaviour by improving their awareness of HWC risks.

Our findings have several policy implications. The intensified HWC will not only directly affect the livelihood of farmers and have a greater negative impact on low-income groups but also indirectly affect farmers’ use of cultivated land through changes in their risk perceptions. Cultivated land will be abandoned or subject to nongrainisation to a certain extent. Therefore, the government should be involved in the HWC governance process in a timely manner. First, the government should incorporate HWC as a category in government disaster management and incorporate HWC governance into its disaster prevention management budget to prevent HWC from causing irreversible losses to farmers’ livelihoods, food security, and biodiversity conservation. Second, on the basis of the “blocking”-based governance strategy of population regulation, direct economic compensation, and barrier infrastructure construction, government management departments should respect the independent adaptability of farmers and adopt a “sparse” governance strategy. They should plan and improve agricultural industrial policies to adapt to the self-regulating process of farmers’ adjustment of their cultivated land use behaviour. Finally, the intervention of government policies should be based on the different characteristics of locations and income groups, and targeted governance policies should be introduced, especially to target nature reserves, which cause serious damage, and low-income groups, which suffer greater losses.