2.1 Study area

Yajiang is on the eastern edge of the Tibetan Plateau in Southwest China (Figure 1). This region lies within one of the 36 global biodiversity hotspots—the Mountains of Southwest China. The mountainous terrain of the region harbors the most botanically diverse temperate forest ecosystems globally, with 12,000 plant species, 300 mammal species, and 686 bird species—many of which are endemic to the area (Mittermeier et al., 2011). The milder temperatures of the Yarlung River valley, a tributary of the Yangtze River, make it one of the few agricultural areas on the Tibetan Plateau (d'Alpoim Guedes et al., 2014). The questionnaire-based interviews were conducted in five villages: Tangzu, Reri, Malangcuo, Tarihe, and Marihe. These five villages are situated in proximity to the renowned Zhaga Monastery in Yajiang, a significant center of Buddhist worship. Owing to religious and traditional taboos against any form of destruction on the sacred mountain of the monastery, coupled with regular conservation patrol activities organized by the monastic community, the primary forest has been preserved. Consequently, apex predators, endemic ungulates, and birds are all recorded here. The altitude of the villages ranges from 2500 to 3700 m.a.s.l., restricting agriculture to subsistence crops such as Tibetan barley, wheat, potato, and maize. Unlike pastoralist communities, people raise a small number of livestock in close proximity to their dwellings, and as is the case for communities at the edge of the forests, there is no distinct threshold between forest and cropland, making it easy for wild animals to forage in village areas, thereby heightening the susceptibility of the local people to HWC. Between 2010 and 2019, a compensation scheme operated in these five villages for people suffering crop damage or loss. The damage was quantified according to the area affected and mitigation measures, and data were recorded by village representatives. The local forestry bureau compensated part of the recorded crop loss on an annual basis, and the monastery, as an independent third party, supervised both the villagers and the forestry bureau to guarantee transparency across the whole process (Yang et al., 2019).

2.2 Tolerance questionnaire

Targeted sampling was conducted based on the participant list of the compensation scheme, and haphazard sampling was also used to collect answers to general questions from any respondents who happened to be at home and agreed to be interviewed. The interviews were conducted in the local Tibetan language, with local forestry staff acting as translators. After excluding unquantifiable interview responses due to answer reliability and translation difficulties, data from 83 households in total were used: 55% of the total number of households (n = 150) across the five villages. A total of 76 households, out of the 83 surveyed, had joined the compensation scheme, and they represented 56% of all participants of the compensation scheme (n = 136). The villagers were asked five questions relating to their tolerance of wildlife species (Questions 1–5; Table 1). The questions were developed based on previous research (Bruskotter et al., 2015; Cerri et al., 2017; Kansky, 2015; Kansky & Knight, 2014).

Responses to Questions 1 and 2 were coded as 1 (yes), 0 (yes conditional on some species being excluded or in low numbers), or −1 (no or do not know). Question 3 asked people whether it was right to kill animals. If respondents answered it was right to kill all crop-raiding animals, the answer was coded as −1. The “yes, kill some of them” answer was coded as 0; “no, it is wrong to kill” was coded as 1. Questions 4 and 5 were posed specifically regarding eight different species: the leopard (Panthera pardus), Lord Derby's parakeet (Psittacula derbiana), the porcupine (Hystrix brachyura hodgsoni), the wild boar (Sus scrofa), the Tibetan macaque (Macaca thibetana), the white-eared pheasant (Crossoptilon crossoptilon), the Asiatic black bear (Ursus thibetanus), and the wolf (Canis lupus). The responses to these two questions were coded by calculating the mean value from the scores attributed to the eight species. For each question, scores were aggregated, resulting in an overall tolerance score for each respondent that ranged from 5 (high tolerance) to −5 (low tolerance).

2.3 Attitudes toward wildlife

We used Likert-scale questions (five categories, ranging from −2 to 2) to gauge attitudes toward the eight species under consideration. This approach was based on the premise that species preference significantly informs attitudes, which in turn are indicative of tolerance levels. The overall attitude score was derived from the mean value of the preference scores of the eight species. Species with sufficient data points were then further grouped into different categories: crop-raiding animals (porcupine, wild boar, and white-eared pheasant) and carnivores (leopard, Asiatic black bear, and wolf). To avoid confusion, images of the animals were shown to the respondents. Information was also collected for the explanatory factors of tolerance by asking questions about wildlife experiences (i.e., How many species have you seen in the wild before?), mitigation effectiveness (i.e., Do you think your fence is useful for preventing the crop-raiding animals?), financial investment on mitigation (i.e., How much money did you spend on fencing?), wildlife ownership (i.e., Who do you think the wildlife belongs to?), awareness of protection laws (i.e., Do you know the wildlife protection laws?), and religious practices to evaluate the level of belief (i.e., How many times have you recited the six-words mantra? How often do you attend the assembly organized by the Zhaga Monastery? Did you participate in the construction of the monastery? Did you take part in the conservation patrols (such as antipoaching and waste collection)? Demographic data, including age, gender, education, and annual income, were collected as other important explanatory factors. Project records of compensation payments and the cropland size of each household from local forestry bureau records were also included. These explanatory variables were in line with previous review studies (Bruskotter et al., 2015; Dickman, 2010; Kansky & Knight, 2014; Naughton-Treves & Treves, 2005). See Supporting Information S2 for the full questionnaire.

2.4 Mitigation measures

Mitigation measures (fencing) indicated the willingness of villagers to invest in preventing damage from wildlife and likely reflected tolerance to some extent. However, as an explanatory factor of tolerance, the efficacy—or lack thereof—of fencing and the resultant frustrations from failed mitigation measures may actually decrease people's tolerance toward wildlife, especially “problem wildlife” (Kansky & Knight, 2014). Given that mitigation measures are fundamental in conflict management, we conducted a simple mitigation effectiveness assessment to explore the mitigation measures in these five villages relating to fence type, financial investment, time investment, and mitigation action effectiveness. The perceived effectiveness was derived from the interviews with the respondents, and the actual effectiveness was calculated based on the recorded damage data from the compensation scheme (see Section 2.5).

2.5 Wildlife damage

The five villages participated in the compensation scheme for several years, and damage and compensation payment data were recorded and updated every year. All recorded damage data, compensation payments, and cropland size of each household were acquired from the local forestry bureau. The total cropland of each household was measured in the field by the local forestry bureau staff and a local conservation NGO between 2011 and 2014 (Yajiang Forestry Bureau, 2017). Since logging and reclaiming new cropland have been strictly restricted in this region since the late 1990s, the size of cropland remains stable.

With respect to the perceived damage from each crop-raiding animal, we asked respondents to rank the crop-raiding animals in order of problems they caused and to specify the level of damage caused by each crop-raiding animal. The perceived damage of each species was categorized as low (20%–40% of their cropland area), medium (40%–60%), and high (60%–100%).

2.6 Data analysis

IBM SPSS Statistics 24 was used to analyze the data. Kruskal–Wallis tests with pairwise comparison were used to test the difference between groups, including the preference and damage perception between different crop-raiding animals. Mann–Whitney U tests were used to test for differences in wildlife preference between the crop-raiding animal group and the carnivore group, wildlife encounter experience at different locations (wild and cropland), differences in perceived mitigation effectiveness between fence investment categories, and for the mantra recital difference between age groups. Spearman Rank Order Correlation was used to test the correlations between cropland size and fence cost, and with actual damage; and Person correlation coefficient and χ² test were used to test the mitigation effectiveness for different aspects of fencing, such as financial investment, time, and combinations of fence types.

Generalized linear models (GLMs) were conducted to test the associations between tolerance, as the dependent variable, and its potential explanatory factors. The GLMs included categorical data such as age, gender, and perceived damage as fixed factors and continuous data such as wildlife preference, wildlife experience (number of species people encountered in the local area), damage area, compensation payments, and financial investment in mitigation measures, as covariables.

3.1 Demographic profile

Among the 83 respondents, each representing one household, there were 47 (56.6%) male and 36 (43.4%) female respondents, and over half were in the age range 41–60 years (53%). There was a high illiteracy rate; 85.5% of respondents had no formal school education. The main family annual income source was from collecting caterpillar fungus (Ophiocordyceps sinensis) and matsutake mushroom (Tricholoma matsutake), and 65% of respondents made more than 10,000 (CNY) in 1 year (Table 2). All the respondents were Tibetans and self-described as “absolute Buddhist.” A total of 87.9% of respondents attended the Buddhist assembly organized by the monastery every time, 98.8% of respondents participated in the voluntary reconstruction of the monastery, and 83.1% of respondents or their family members took part in the sacred mountain conservation patrol activities every year, typically three–four times a year, which were organized by the monastery and self-funded. The age or gender limitations were most apparent in patrol activities: only young, physically fit men were allowed to participate. The only significant difference in mantra recitals was between age groups (Table 3). People over 60 years old recited the religious mantra (six true words) more times every day than other age groups. Therefore, the only differences in the answers to the questions about religious practices were due to age or gender restrictions.

3.2 People's attitude toward different animal groups and in different contexts

People's preferences between animal groups differed significantly: crop-raiding animals were less preferred than carnivores (Table 3). Between the crop-raiding animal groups, there was a statistically significant difference in preference between white-eared pheasants and the other two, wild boars (p < 0.001**) and porcupines (p < 0.001**). Of these three crop-raiding species, people favored white-eared pheasants, which were most frequently seen in the fields: 90.4% of respondents reported encountering white-eared pheasants on their cropland. There was no significant difference in preference between wild boars and porcupines (p = 0.915) (Figure 2). People in the five villages reported a high frequency of wildlife encounters. On average, each respondent observed 10 different species during the previous year. There was a statistically significant difference in feelings about wildlife experiences at different locations between those in the wild and those in cropland (Table 3). People tended to have positive feelings (76.1%, n = 142 encounters) when they encountered pheasants, ungulates, and other animals in the wild, with fewer instances of negative feelings (15.4%, n = 142 encounters) and no feelings (8%, n = 142 encounters). Conversely, when they encountered crop-raiding animals in cropland, people tended to have negative feelings (56.1%, n = 157 encounters) rather than positive feelings (39.5%, n = 157 encounters) or no feelings (4%, n = 157 encounters). Most (78.3%, n = 65) people believed wildlife ownership should lie with the government. However, when asked about who should take responsibility for protecting these animals, 51.8% of respondents (n = 43) chose “everyone.”

3.3 Perceived and actual damage

Using data available for the 83 household respondents, the average size of cropland owned by each household was 0.33 ± 0.19 ha, and the average area of damage per household was 0.04 ± 0.04 ha. Approximately 2.6% (n = 2) recorded damage that could be categorized as “high damage” (60%–100% of the cropland area of a household), and 88.3% (n = 73) recorded damage categorized as “low damage” (0%–39% of the cropland area). There was a statistically significant positive correlation between the damage loss and the cropland size of each household; the more cropland cultivated, the more damage was recorded (Table 4). The average compensation payment was CNY516.1 ± 311.33 per household for loss during the years before the survey was conducted. A total of 75.9% of respondents (n = 63) ranked wild boar as the worst crop-raiding animal. There was a significantly different perception of damage caused by the crop-raiding animals (Table 3 and Figure 2). Wild boar was perceived to cause more damage than porcupine (p = 0.002*) and white-eared pheasant (p < 0.001**). And between white-eared pheasant and porcupine, less damage was perceived from white-eared pheasant (p < 0.001**; Figure 2).

3.4 Mitigation measures and effectiveness

The mitigation measures in our study sites were fences. A total of 92.8% (n = 77) of respondents replied that they built fences around cropland. There were five types of fences to prevent crop-raiding animals: thorny branches (22%), wooden board (1.3%), stone (1.3%), hand-made wire fencing (3.9%), and commercial iron fencing (14.3%). Most respondents (57.1%) used a combination fence with at least two fence types, which differed from single material fences in terms of structure and stability. Only the commercial iron fence and wire fence required financial outlay; the average money spent on fencing per household was CNY377.7 ± 454.6, and 54.2% of households (n = 45) spent 1 week building their fence. The fence cost was positively correlated with the cropland size (Table 4). Approximately 61.4% of the households (n = 51) believed the fences were not useful. Thus, perceptions of mitigation effectiveness were significantly different between those who had made a financial investment and those who had not (Table 3). People who spent more money on fencing felt that the fence was useful, while those who spent less money believed the fence was not at all useful. However, no significant correlation existed between actual damage and mitigation (fencing) investment, time spent building fences, and the number of fence types used in combination fencing (Table 4).

3.5 Tolerance toward crop-raiding animals and its driving factors

Our quantitative surveys across 83 households in five villages showed that the average tolerance score toward animals was 1.02 ± 1.41, denoting a positive preference (n = 83; range from −5 to 5; Figure 3). The GLMs result showed that there was significant correlation between tolerance and the explanatory variable wildlife preference (p < 0.001**). There was no significant correlation between tolerance and wildlife experience, perceived/real damage, compensation payments, financial investment into fencing, and other demographic factors (Table 5). People with a higher wildlife preference had a higher tolerance than others.

4.1 The importance of recognizing local and cultural value systems in tolerance attitudes

The cognitive hierarchy framework, which integrates values, value orientations, attitudes and norms, behavior intention, and behavior, has been applied to explain and predict human response when conflict occurs (Cerri et al., 2017; IUCN, 2023; Manfredo et al., 2009). In our study, overall wildlife preference was the key factor affecting people's tolerance. We also found that people's preference for wildlife was different between animal groups and that resentment toward crop-raiding animals did not affect feelings toward other animals. Wildlife experiences played an important role in people's preference toward wildlife at different locations: our respondents had more positive experiences in the wild and more negative experiences in croplands. The white-eared pheasant in our study was a good example of how cultural values and value orientation influenced people's tolerance toward wildlife. Although it was ranked as one of the top three crop-raiding animals, and 85% of respondents reported daily encounters with it in cropland, people still had a higher positive attitude toward it than other crop-raiding animals, perhaps because this species is considered sacred in Tibetan culture: its conspicuous white is a symbolic color for Tibetan Buddhists (Wang et al., 2012).

Previous studies showed that people on the east Tibetan Plateau orientated themselves with nature by following the Buddhist teachings of sacred nature sites, sin, cycles of karma, and empathy and equity of all beings (Woodhouse et al., 2015). These values mean that people hold non-utilitarian feelings toward animals, and consider animals part of their society (Manfredo et al., 2009). In line with this, Bhatia et al. (2017) found positive associations between religiosity and attitude toward two carnivores in Buddhist villages. Similarly, Gogoi (2018) reported that people's tolerance was high despite stressful experiences with elephants because the elephant is a culturally important animal in India. While the highly homogeneous religiosity in the five villages in our study meant that we could not say definitively whether religion was a key factor driving people's tolerance, it is likely that it is important. There are few studies on how people's traditional values for wildlife, especially regarding culture and religion, contribute to wildlife tolerance in this region; more research is needed on this subject. Also, for conservation practitioners addressing HWC in the region, negative experiences and retaliatory behavior may not be directly correlated, due to the complexity of the local context: assumptions around the drivers of behavior may be incorrect (cf. cognitive hierarchy framework). Thus, we must consider attitudinal factors—how people consider their relationship with the natural world and other species—in different parts of the conflict management process (Dickman, 2010; Manfredo et al., 2009).

4.2 The limitations of traditional HWC management for conservation

Traditionally, a wide range of economic schemes, such as compensation, insurance, and conservation payments, are applied in conflict management with the aim of both offsetting the costs to those affected and increasing tolerance for wildlife (Naughton-Treves & Treves, 2005; Nyhus, 2016; Treves & Bruskotter, 2014). Between 1999 and 2015, US$222 million was paid in compensation to offset the damage caused by wildlife globally (Ravenelle & Nyhus, 2017). However, tangible cost was less important than intangible cost in explaining people's attitudes toward wildlife (Kansky & Knight, 2014). This was in line with our study, where neither actual crop damage caused by animals, nor the insurance compensation payments for the damage, were correlated with people's tolerance. Paradoxically, previous studies looking at economic payments have shown that if the compensation is far less than the market value (e.g., due to lack of funds, as in the case of this study), tolerance for wildlife might not be associated with compensation payments (Dickman et al., 2011). In contrast, where compensation is high enough to cover the damage costs and opportunity costs, it may then lead to “moral hazards”—where people intentionally make false damage claims—resulting in negative impacts on conservation. High subsidies might even encourage agricultural expansion and habitat conversion, which could further push the local communities into a “poverty trap” due to intensified competition for shared and limited natural resources (Bulte & Rondeau, 2005; Ravenelle & Nyhus, 2017).

Another typical component of conflict management is various mitigation measures, such as physical barriers, guarding, early-warning systems, deterrents, and aversions (IUCN, 2023; Treves et al., 2009). While the mitigation measures in our study (fences) did not effectively prevent damage from animals, they had a positive outcome in another way: people's perception of fence effectiveness increased their tolerance toward wildlife. The villagers who spent more money on fencing were likely to perceive it as useful and likely to have higher tolerance. This emphasizes the necessity to integrate approaches that consider the human dimension, which should be informed by affected stakeholders within the local context (IUCN, 2023).

4.3 The importance of engaging all stakeholders in conflict management

Madden and McQuinn (2017) posited that there were three types of conflict that formed a conflict framework. At the top was dispute conflict, which was clearly evident and easily articulated. However, dispute conflict was just the tip of the iceberg, and the more important types of conflict were underlying conflicts—those unresolved or unsatisfactorily solved historical conservation conflicts between different stakeholders—and deep-rooted conflicts that challenge social identity and values. For instance, once wildlife with a severe negative interaction becomes characterized as a vicious species in local culture, tolerance would remain low even with the conflict removed. As part of biodiversity conservation conflicts, HWC can be presented in different ways between different interest groups (Hill, 2017). This is attributable to the involvement of multiple stakeholders who sometimes hold conflicting standpoints (Peterson et al., 2010; Redpath et al., 2013; Young et al., 2010), bringing about confrontations on how best to manage wildlife (Draheim et al., 2015). Thus, engaging all stakeholders in the process, generating trust, and enhancing acceptance between them are key aims in conservation conflict management (Carpenter et al., 2000; Young et al., 2016). In this study, the stakeholders were five local communities, the monastery, an NGO, and the local forestry bureau. The local forestry bureau decentralized power to local communities by allowing them to assess the damage loss themselves. The monastery played an important role in helping to generate trust among the stakeholders, acting as a neutral stakeholder to supervise the legitimacy of the loss assessment in compensation schemes, and adding authority to local communities when they negotiated with the local forestry bureau. The NGO played a facilitator role and helped to connect funding resources. Additionally, the monastery's Buddhist teaching determined local people's values and, in addition to the sacred mountain patrols every year, facilitated the raising of local people's conservation awareness, and promoted conservation as the mutual goal among all stakeholders. Therefore, despite ongoing crop-raiding conflicts, people's tolerance in this region was generally high, and their attitudes toward wildlife positive. Despite most people experiencing crop-raiding and perceiving that ownership of wildlife lies with the government, half of the respondents considered that it should be everyone's responsibility to protect wildlife.

4.4 The influential factors of demography

Our result was in line with the previous finding that socio-demographic variables were poor predictors of attitudes toward animals (Kansky & Knight, 2014). However, other studies have shown strong relationships between socio-demography and these attitudes (Bhatia et al., 2017; Kleiven et al., 2004; Ogra, 2008; Suryawanshi et al., 2014). We should not simply assume that socio-demographic factors are only useful for describing populations because tolerance could be region-specific. Therefore, demographic variables should be included in assessments of tolerance in each case. They can be useful for developing contextually relevant local adaptive strategies for wildlife conflict management.