Reserve regulation and multidimensional relative poverty of farmers: Evidence from the Panda Nature Reserves in China

The accurate evaluation of the relationship between nature reserves and poverty is highly significant for the harmonious coexistence between human and nature. It is widely recognized that the establishment of nature reserves is of great importance to the income poverty of farmers, but less attention has been paid to the impact of different reserves on the multidimensional relative poverty of farmers. Based on the survey data of Panda Nature Reserves in China, we analyze the influence of reserve regulation (or not) and regulation intensity on the multidimensional relative poverty of farmers and its mechanism. Results show that farmers in reserves are more likely to fall into multidimensional relative poverty than those outside the reserves, and there is a U‐shaped relationship between regulation intensity and multidimensional relative poverty. Further, the mechanism analysis show that, on average, the establishment of reserves has no significant impact on farmers' resource utilization capability, but too high or too low regulation intensity will affect farmers' resource utilization capacity, and aggravate their multidimensional relative poverty. The conclusions of this paper are not only conducive to the expansion of theoretical research on regulation and poverty, but also provide policy implications for realizing the coordinated development between biodiversity conservation of nature reserves and rural livelihood.

theoretical research on regulation and poverty, but also provide policy implications for realizing the coordinated development between biodiversity conservation of nature reserves and rural livelihood.

Recommendations for Resource Managers
The effects of the regulation of nature reserves on the relative poverty of farmers is analyzed in this paper.
We find that there is a "quantity" problem of the impact of reserve regulation on the farmers' relative poverty. The following implications could be realized based on the observations: • When identifying the relative poverty of farmers, it is necessary to make an in-depth analysis from the perspectives of economic poverty and multidimensional poverty, and provide targeted and accurate assistance policies for different types of farmers.
• The formulation and implementation of the reserve regulations should consider the difficulty of farmers' livelihood transformation to balance the relationship between the community development and ecological protection.

K E Y W O R D S
relative poverty, reserve regulation, resource utilization

| INTRODUCTION
Human beings are closely related to forests (Roberts, 2019). The world has a total forest area of 4.06 billion hectares, accounting for 31% of the global land area, which is equivalent to 0.52 hectares per capita (FAO, 2020). Forests can provide shelter for most terrestrial biological diversity. Moreover, a wide variety of animals and plants in the forest can also provide food, fuel, and other important raw materials for human beings to support their livelihood activities.
The establishment of nature reserves (NRs) has always been the most common governance measure to protect natural resources and biodiversity. By the end of 2018, China had established 2750 NRs, with a total area of 1.4733 million square kilometers, accounting for 14.88% of the land area. The NRs have effectively protected at least 90% of the terrestrial ecosystem, 85% of the wild animal population, and 65% of higher plant, enabling the restoration of some rare and endangered animal and plant populations, among which the national treasure giant panda is a typical representative (Wang, 2017).
However, NRs and poverty-stricken areas often overlap geographically. Most NRs are mostly located in remote mountainous areas with poor infrastructure and underdeveloped economies. Generally, inconvenient transportation means that people living in the communities surrounding the NRs stay away from the market, which is not conducive to obtaining nonagricultural employment opportunities (Angelsen & Wunder, 2003). Also, weak infrastructure and insufficient social security lead to low physical capital and human capital. Poor rural families have no choice but to depend highly on natural resources to survive (Camara-Leret et al., 2019). However, in the process of protecting natural resources and biodiversity, farmers are restricted in their use of resources without equal compensation, which exacerbates their poverty (Wilkie et al., 2006). Therefore, the conservation of natural resources and development of community economies have always the academic foci.
There is enough empirical evidence indicating that NRs are effective in protecting biodiversity, but the socioeconomic impact of establishing NRs on local communities, especially the livelihood of farmers, is not clear. Researchers generally believe that the poverty of farmers is aggravated due to the establishment of NRs. The reason is that NRs lead to the regulation of utilization of natural resources by farmers, forcing them to change their traditional livelihoods and thus falling into poverty Walelign et al., 2016, Wang & Liu, 2021. Meanwhile, the protection policies that ignore the needs of farmers further increase the vulnerability of their livelihood and the incidence of poverty (Wang, 2017). For example, farmers are forced to move out of their original addresses and deprived of land use rights and management rights, and so forth. But some other researchers point out that the establishment of NRs can alleviate farmers' poverty because NRs contain great ecological value (Wilson et al., 2010). Although farmers bear the cost of protection, they also obtain the benefits from protection. The direct benefits mainly include that farmers can enjoy diverse ecosystem services, obtain ecological compensation, eco-tourism income, employment in the NRs, and reasonable collection of resources (Ferraro & Hanauer, 2014;Karanth et al., 2012). Moreover, farmers can also gain indirect benefits, such as the improvement of community environment and infrastructure, training in skills, and other support policies (Miranda et al., 2016). The direct and indirect benefits provided by the NRs to farmers can improve their ability to deal with vulnerabilities and crises, thereby reducing the incidence of poverty (Bennett et al., 2012). Therefore, the establishment of NRs is conducive to alleviating the poverty of farmers, at least it will not exacerbate the poverty (Andam et al., 2010;Clements et al., 2014).
The inconsistent research conclusions above have led to a fierce debate on the impact of NRs on poverty, and make very useful explorations. However, the existing studies still have some deficiencies. First, the "protection" of NRs are regarded as the same in most research, and the analysis is carried out only from the perspective that farmers inside the reserves are regulated and those outside are not, which ignores the difference in the implementation of regulation in different NRs. Therefore, the impact of regulation intensity on farmers' poverty is rarely distinguished. Second, the existing literature pays more attention to absolute income poverty and less to multidimensional relative poverty of farmers, that is, the relatively weak or deprivation of health, education, social participation, self-development, and other capabilities available for farmers. With the elimination of absolute poverty, the issue of relative poverty has become norm and prominent as long as there are social differentiation and inequality (Luo, 2020). In addition, compared with income poverty at one dimension, the measurement of farmers' relative poverty from multidimension is more helpful in exploring the essence of poverty. Third, previous research focuses on evaluating the impact direction, and degree of the NRs on poverty, but the exploration of its mechanism is insufficient. In this paper, using the data of Panda NRs in China, we analyze the influence of reserve regulation, including regulated or not and regulation intensity, on the multidimensional relative poverty of farmers and its mechanism. This study expands the research on regulation and poverty. Moreover, it not only provides policy suggestions for promoting the construction of China's ecological civilization and realizing common prosperity, but also provides enlightenment for other developing countries to improve the management of NRs.

| THEORETICAL ANALYSIS
Based on the theory of relative poverty and the context of NRs, we construct a theoretical framework to discuss the impact of reserve regulation on the multidimensional relative poverty of farmers from three dimensions: capital accumulation, risk shock, and social equity ( Figure 1).
The reserve regulation, in fact, tends to strictly restrict the development and utilization of natural resources by local farmers (Wang et al., 2013). The reserve regulation that only focuses on the protection of natural resources and ignores the interests of farmers leads to poverty (Wunder, 2001). For example, the Liangshan, Minshan, and Qinling Mountains, where the Panda NRs are located, remain contiguous areas of extreme poverty in China. In general, therefore, reserve regulation will change the rights attached to resources, leading farmers have to transform the mode of production and life, and aggravate multidimensional relative poverty of them. Specifically, first, reserve regulation can reduce the accumulation of human and social capital of farmers, leading to the reduction of their feasible capability, and thus increasing multidimensional relative poverty of them. Inadequate accumulation of human capital and social capital is an important cause of poverty and decreasing capital inequality plays a key role in the governance of relative poverty (M. W. Cheng et al., 2014). Human capital mainly includes health status, education level, labor skills, and learning ability of individuals. The NRs, especially those with strict regulations, limits the rights of farmers to develop and utilize nature resources. Before farmers can successfully transfer their traditional livelihoods that are highly dependent on natural resources, the welfare of them will be seriously damaged, which is not conducive to the accumulation of human capital and poverty reduction (Walelign et al., 2016). In addition, social capital mainly includes social network, social trust, and social participation (Putnam et al., 1993). Reserve regulation may destroy the original social structure, weaken F I G U R E 1 The mechanism of reserve regulation affecting the multidimensional relative poverty of farmers social capital (Duan & Ouyang, 2020) and has a negative impact on the accumulation of social capital. As a result, such regulation reduces the availability of social opportunities for farmers, especially the poverty-stricken farmers, increases the inequality of opportunities, and weakens their resource utilization capacity, thereby aggravating their multidimensional relative poverty.
Second, reserve regulation increases the risk shock, which hinders the improvement of farmers' capability, and aggravates the multidimensional relative poverty. Exogenous risk shock is an important reason for farmers to fall into poverty and the poverty path is direct and destructive . Reserve regulation leads to huge changes in the environment surrounding farmers, and environmental factors deprive their feasible capability through risk shock, which cause them to fall into relative poverty for a long time. Meanwhile, reserve regulation increases the risk of wild animal damage and other risk shocks that farmers need to bear (Li et al., 2006). On the one hand, these risk shocks directly cause farmers to lose their sources of income and they may select credit channels to deal with the risks. As a result, poverty-stricken farmers have great restrictions on their income mobility and fall easier into poverty (Wang et al., 2019). On the other hand, risk shock will cause the loss of livelihood capital and further lead to insufficient investment into the feasible capability of povertystricken farmers, which increases their livelihood vulnerability (Fang, 2019). Moreover, risk shock will cause a negative psychological impact, which will not only attack the self-confidence of poverty-stricken farmers but also their production and living activities (Bandiera et al., 2017). Therefore, reserve regulation may increase the risk shocks, reduce the resource utilization ability of farmers and aggravate their multidimensional relative poverty.
Third, on average, reserve regulation increases social inequity, causes inequality of developing opportunities and rights, leads to the lack and insufficiency of farmers' feasible capability, and thus exacerbates multidimensional relative poverty of them. Based on the theory of feasible capability, social inequity is mainly reflected in the deficiency of social opportunities, political rights and interests, transparency guarantee, and protective guarantee. These rights and opportunities will directly expand the feasible capability of farmers, help farmers live more freely, and improve their overall ability (Sen, 2012). The loss of social rights, political rights, and production and development rights brought about by the reserve regulation will further increase the inequity of communities. In addition, according to the view of modern welfare economics, the fundamental way to realize equality of economic conditions is to maintain the equality of opportunity. Such equality mainly depends on market mechanism, separation of property rights, and governance mechanism in alleviating relative poverty (Luo, 2020). Reserve regulation will reduce the degree of marketization, the clarity of the definition of resource ownership and the community participation of farmers, such as the collective income distribution mechanism is not perfect and the favorable policy is not reasonable, which increase social inequity, are unfavorable to the formation and improvement of farmers' feasible ability and lead farmers to fall into multidimensional relative poverty.
The following two hypotheses to be tested are proposed.
H1. Reserve regulation may aggravate the multidimensional relative poverty of farmers. H2. Reserve regulation aggravates the multidimensional relative poverty through resource utilization capacity of farmers.
However, with the rapid growth of NRs, the contradiction between protection and development has become increasingly intensified. To alleviate the contradiction between the two, some NRs allow the reasonable use of natural resources without affecting ecological protection and attempt to improve the livelihood of farmers by attracting tourism, developing green industries, promoting infrastructure development or increasing the economic benefits of environmental services (Angelsen et al., 2014;. For example, China's Wolong and Tangjiahe NRs achieve inclusive development of ecology and livelihoods by developing ecotourism and green agriculture. It can be seen that the different implementation of the reserve regulation in different NRs will lead to major differences in the livelihood development of local farmers. Under the strict regulation of NRs, farmers are prohibited from using natural resources in any form, leading to serious damage to human and social capitals and greatly increasing the risk shock and social inequity faced by farmers. As a result, the multidimensional relative poverty is increased. Does it mean that more deregulation is better for alleviating poverty? The answer is no. If the regulation intensity is too weak, the resource utilization ability of farmers will be close to the state without the reserve regulation. The overuse or extensive use of natural resources is not conducive to the sustainability of farmers' livelihood, nor can it make up for the negative impact of reserve regulation on the feasible capability of farmers, and it is also unfavorable to get rid of the multidimensional relative poverty. However, if the regulation intensity is moderate, the positive impact of reserve regulation on the resource utilization ability of farmers will be enough to make up for the negative impact, that is, the (direct and indirect) benefits that farmers obtain from protection outweigh costs, and then reserve regulation can alleviate the multidimensional relative poverty. Therefore, when further considering the regulation intensity, too high or too low regulation intensity of NRs will aggravate the multidimensional relative poverty of farmers. And then, Hypothesis 3 is proposed.
H3. A U-shaped relationship may exist between regulation intensity and the multidimensional relative poverty of farmers.

| Data collection
We applied the questionnaire survey of farmers (face to face interview) to survey the local communities around the 5 and 12 Panda NRs in Shaanxi and Sichuan in July and October 2018 and January and May 2019, respectively. The communities in every NR were sequenced based on their economic development level and per capita annual income and divided into two groups to randomly select one community from each group. The communities outside the NRs were processed in the same way. Then, four communities (two inside and two outside) were selected in one Panda NR. Finally, 15-20 households were selected randomly from each community. Altogether 60 communities were surveyed as some NRs do not have four communities. The questionnaires collected the data on two levels: villages and families. The information on the villages mainly included the natural environment, geographical features, and economic development. Meanwhile, family members, resource endowment, the status of production and operation, and issues related to the NRs and particular subjects are also gained. After the samples with missing key information were weeded out, 864 valid samples were obtained.

| Dependent variable: Multidimensional relative poverty
Based on the previous research (Duan & Ouyang, 2020;Fang & Zhou, 2021) and the context of NRs, health, education, and living standard are selected as the three dimensions of multidimensional relative poverty in this paper. Further, setting indicator weight is one of the important links to accurately identify multidimensional relative poverty. According to the data characteristics, the equal weight method is used to assign the indicator weight of multidimensional relative poverty. The indicators and their assignment are shown in Table 1. Furthermore, the multidimensional relative poverty of farmers is measured following the methods of Fang and Zhou (2021). First of all, the absolute poverty line is defined and the double boundary method of Alkire-Foster is used to identify the poverty-stricken farmers. Set up a deprivation matrix g g = [ ] α ij α , and g ij α meets: (1)

Dimension Indicator Critical value Assignment
Education Years of education The labor with an average education level lower than 6 years is assigned as 1; otherwise, it is assigned as 0.

1/6
Entering School The household with children aged 6-16 years who have not received education is assigned as 1; otherwise, it is assigned as 0.

1/6
Health Incidence of serious disease and disability The household with disabilities or major diseases is assigned as 1; otherwise, it is assigned as 0.

1/6
Timely medical treatment The household with patients who failed to receive medical treatment in a regular medical institution is assigned as 1; otherwise, it is assigned as 0.

Sources of drinking water
The household using non-tap water/non-barreled water/non-purified water/non-filtered water as cooking water is assigned as 1; otherwise, it is assigned 0.

1/15
Fuel type The household using only firewood and straw for heating and cooking is assigned as 1; otherwise, it is assigned as 0.

1/15
Washroom The household without flush toilet is assigned as 1; otherwise, it is assigned as 0.

1/15
Housing structure The house with civil structure is assigned as 1, otherwise it is assigned as 0.

1/15
Waste disposal The household without a fixed/classified garbage point nearby is assigned as 1; otherwise it is assigned as 0.
1/15 LIN AND GAO Natural Resource Modeling | 7 of 25 y ij is the value of farmer i on the j indicator, z j represents the critical value at which farmers are deprived of the j indicator. When α = 0, g ij α represents deprivation status of farmer i on the j indicator. That is, when g = 1 ij α , it means that the farmer is deprived and is in poverty on the indicator. g = 0 ij α means that the farmer is not deprived and is not in poverty on the indicator.
Second, the weighted sum value of the farmer who is deprived in each indicator is calculated.
(2) c i is the weighted total deprivation score of farmer i, ω j represents the weight of the indicator j. If the degree of deprivation of the farmer i is greater than the critical value, the farmer i is in a state of multidimensional poverty. Otherwise, the farmer i is not in a state of multidimensional poverty.
In addition, m j is calculated to obtain the measurement of multidimensional relative poverty. c j is the weighted total deprivation score of indicator j, and a is denoted the particular proportion, so m j can be expressed as Then the deprivation matrix is redefined. m j is the relative poverty deprivation critical value of indicator j. In this case, the vaule of g ij α is Similarly, g ij α represents deprivation state of farmer i on the j indicator. g = 1 ij α means that the farmer is deprived and is in multidimensional relative poverty on the indicator. g = 0 ij α represents the farmer is not in multidimensional relative poverty on the indicator.
Finally, on the basis of calculating the relative deprivation at one indicator, we can identify whether the farmer is multidimensional relative poverty according to the indicator k. In terms of the setting of indicator k, with reference to the practices of Alkire and Santos (2014) and Xing and Li (2019), k = 3 is selected for analysis in this paper.

| Main independent variable: Reserve regulation
The main independent variable is reserve regulation. First, the influence of the existence of reserve regulation (i.e., regulation or not) on the multidimensional relative poverty of farmers is discussed. Without considering the spillover effect of reserve regulation, referring to the existing studies (Duan et al., 2021;Oldekop et al., 2015), whether the farmer is located inside the NRs or not is taken as a proxy variable showing whether the farmer is regulated or not. If the farmer is located in NRs, the reserve regulation is assigned as 1; otherwise, it is assigned as 0.
Second, the different regulation intensities are further analyzed. Due to differences in the implementation of the regulations in different NRs, we believe that the regulation intensity includes two levels, namely, objective expression and subjective perception. Objective expression refers to the intensity of objective implementation of the reserve regulation in different zones. Based on the institution and geographical division of NRs, the regulation of the core zone is the most strict, followed by the buffer zone and experimental zone, and the final one is the surrounding area outside the NRs. Subjective perception refers to the feelings of local farmers, as regulated objects, regarding the enforcement intensity of reserve regulation. Different farmers have different cognition and compliance to the regulation, resulting in differences in the perception of behavior constraints (Yergeau, 2020).
In terms of the construction of farmers' subjective regulation intensity, we mainly calculate from two dimensions, agricultural production regulation, and pollutant regulation. For households that are highly dependent on natural resources, their livelihoods are mainly maintained through the direct use of natural resources, including operating land, breeding livestock, and collecting timber and non-timber forest products (NTFPs). Moreover, there is empirical evidence suggesting that the regulation of agricultural production and the regulation of pollutants can be the factors that lead to the poverty of farmers (Lin & Gao, 2021;Luo, 2020). In the context of NRs, a total of seven indicators, including application of pesticide and chemical fertilizer, wood logging, firewood collection, wild medicinal material collection, grazing, waste water discharge, and human and animal manure treatment, are selected. Farmers score the seven indicators from 0 to 3 points according to their own subjective feelings. 0 represents no regulation, 1 represents loose regulation, 2 represents strict regulation, and 3 represents complete ban. Then, considering the data characteristics of this study, the entropy method is used to assign values to these seven indicators and Table 2 reports the specific assignment of each indicator.
To alleviate the subjective cognitive bias and get the comprehensive indicator of the regulation intensity, we further using the objective intensity of NRs to modify the subjective intensity. The calculation of the regulation intensity is as follows:

| Mediating variables
Based on the above theoretical analysis, resource utilization capacity mainly includes four aspects: human capital, social capital, risk shock, and social equity. The inequality of human capital and social capital will lead to unequal development opportunities (F. Cheng et al., 2002;Mogues & Carter, 2005), risk shocks will increase the livelihood vulnerability of farmers (Bandiera et al., 2017), and social inequity may reinforce class divides (Bowles & Gintis, 2002), thus leading to the multidimensional relative poverty of farmers. Specifically, due to the 9-year compulsory education, whether farmers participated in skill training activities can better reflect the difference in human capital caused by NRs, so it is selected to measure human capital in this paper. Following the existing literature (Duan & Ouyang, 2020;Liu et al., 2014), the logarithm of annual expenditure on gifts for human relationship is used to measure social capital. In addition, the questionnaire recorded the risk shocks faced by farmers to investigate the losses that farmers have to bear due to the establishment of NRs, including land reduction, resource utilization restriction, livestock loss, and other property loss. Therefore, the number of risk shocks faced by farmers is adopted to measure risk shock. Based on the new political economy, democracy can cause economic growth (Acemoglu et al., 2019). From the microview, the fairness of the village cadre election can reflect the democracy and equity of village (Tang, 2019), so the equity of election activities in the village is adopted to measure social equity, valued 0-4 from low to high.

| Control variables
Following the existing research (Duan & Ouyang, 2020;Ma et al., 2020), the control variables mainly include the householder, family and village characteristics. Specifically, householder characteristics include age, gender, years of education, and cadre status; family characteristics include the number of labor force, dependency burden, proportion of nonagricultural employment, area of agricultural land, area of forest land; village characteristics include distance between residence and town market, and distance from village committee to cement road. The definitions and descriptive statistics of relevant variables are shown in Table 3.

| Model specification
The dependent variable is whether the farmer is multidimensional relative poverty, which is a binary variable. Thus, the Probit model is used to analyze the impact of reserve regulation on the multidimensional relative poverty of farmers. The model specification is as follows: Y is is the multidimensional relative poverty of farmers i in region s. If the farmer is multidimensional relative poverty, Y = 1 is ; otherwise, Y = 0 is . X is represent the regulation of NRs, including reserve regulation and regulation intensity (RI). Control is are the control variables, including householder characteristics, family characteristics, and village characteristics. district is is the regional fixed effect to control the missing variables at the regional level. ε is is the error term.
Moreover, based on the U-shaped relation hypothesis about the impact of the regulation intensity on the multidimensional relative poverty of farmers, it is necessary to further explore the dynamic relationship between regulation intensity and multidimensional relative poverty. The quadratic term of regulation intensity (RI2) is introduced into Equation (6) to construct the model shown in the following equation: Finally, to explore the mechanism of reserve regulation on the multidimensional relative poverty of farmers, the following mediating effect model is constructed: M is refers to human capital, social capital, risk shock, and social equity, other variables are the same as above. ε ε − 1 3 are the error term. Table 4 shows the results of reserve regulation on the multidimensional relative poverty of farmers. It can be seen that the reserve regulation is significant at the statistical level of 1%, and the marginal effect is about 0.084. That is, reserve regulation has a depriving effect on the feasible capability of farmers, which increases the probability of farmers falling into multidimensional relative poverty by 8.4%.

| Impact of reserve regulation on multidimensional relative poverty
T A B L E 4 Impact of reserve regulation on the multidimensional relative poverty of farmers 4.2 | Impact of regulation intensity on multidimensional relative poverty The influence of regulation intensity on the multidimensional relative poverty of farmers is further analyzed and the results are shown in Table 5. The regression coefficients of regulation intensity are negative, but none of them pass the significance test at the 10% statistical level. It is expected that there may be a nonlinear relationship between the regulation intensity and multidimensional relative poverty of farmers. The U-shaped relationship between the regulation intensity and the multidimensional relative poverty can be seen in Table 6. The coefficient of RI (regulation intensity) is significantly negative, while the coefficient of the RI2 is significantly positive. The result shows that a U-shaped relationship exists between the regulation intensity and the multidimensional relative poverty of farmers. In other words, there is an inflection point in the impact of regulation intensity on the multidimensional relative poverty of farmers: on the left of the inflection point, the enhancement of reserve regulation will alleviate the multidimensional relative poverty of farmers; on the right of the inflection point, the enhancement of reserve regulation will improve the probability of farmers falling into multidimensional relative poverty.
To verify whether the U-shaped relationship exists between regulation intensity and farmers' multidimensional relative poverty, Utest is further conducted. The Utest results show T A B L E 5 Impact of regulation intensity on the multidimensional relative poverty of farmers that there is a U-shaped relationship with the inflection point of the regulation is approximately 1.835, and the original hypothesis is rejected at the 5% statistical level.

| Reserve regulation, resource utilization capacity, and multidimensional relative poverty
The mechanism of the impact of reserve regulation on farmers' multidimensional relative poverty is discussed. Table 7 shows whether the existence of reserve regulation affects the multidimensional relative poverty of farmers through resource utilization capacity. It can be seen that reserve regulation has no significant impact on the resource utilization capacity of farmers after other control variables are added. The path of reserve regulation affecting multidimensional relative poverty through resource utilization capacity may not exist. We further use the Bootsrap method to retest the findings, and the results are shown in Table 8. The mechanism that reserve regulation affects multidimensional relative poverty through resource utilization capacity is also untested. The reason may be that reserve regulation has a U-shaped impact on the multidimensional relative poverty of farmers, the effects will cancel each other when the average marginal effect is analyzed under the situation of regulated or not.
T A B L E 7 Relations of reserve regulation, resource utilization capacity, and multidimensional relative poverty

| Regulation intensity, resource utilization capacity, and multidimensional relative poverty
To test the relationship of regulation intensity, resource utilization capacity, and multidimensional relative poverty, the inflection point of regulation intensity (1.835) is used to group for comparative verification. Tables 9 and 10 show the empirical results when the regulation intensity does not reach the inflection point (<1.835) and exceeds the inflection point (>1.835). Table 9 reveals that when regulation intensity does not reach the inflection point, it can alleviate multidimensional relative poverty by improving the resource utilization capacity of farmers. Regulation intensity has a significantly positive impact on human capital, social capital, risk shock, and social equity, with estimated coefficients of 0.012, 0.102, 0.048, and 0.124, respectively. The results show that if the regulation is stronger, it will be more conducive to human and social capital accumulation and social equity, but at the same time it will increase risk shock. Moreover, the results in Column (5) show that besides the significantly positive impact of risk shock on multidimensional relative poverty, human capital, social capital, and social equity can significantly reduce the multidimensional relative poverty of farmers.
The results when the regulation intensity exceeds the inflection point (>1.835) are shown in Table 10. When regulation intensity exceeds the inflection point, it significantly lowers the human capital, social capital, and social equity of farmers at the statistical level of 1%, with regression coefficients of −0.010, −0.060, and −0.083, respectively. Meanwhile, regulation intensity has a significantly positive impact on risk shock. These findings show that in the T A B L E 9 Relations of regulation intensity (<1.835), resource utilization capacity, and multidimensional relative poverty group with high regulation intensity, the stronger the regulation, the lower the human capital, social capital, and social equity of farmers, and the greater risk shock. Further, it can be seen from Column (5), if the regulation is stronger, farmers are more likely to fall into multidimensional relative poverty.

| Bootstrap test of mediating effects
Further, the mediating effect of resource utilization capacity is verified by bootstrap. The results of regulation intensity are shown in Tables 11 and 12. Table 11 reveals that when the regulation intensity does not reach the inflection point, the path of "regulation intensity-resource utilization capacity-multidimensional relative poverty" has been verified. Specifically, the mediating effect of human capital, social capital, risk shock, and social equity accounts for 52.75%, 51.83%, −28.44%, and 23.39% of the total effect, respectively.
First, in the group with low regulation intensity, the role of regulation intensity in alleviating multidimensional relative poverty through human capital accounted for the largest proportion, indicating that human capital played an important role. The stronger the regulation of NRs, the more it promoted the accumulation of farmers' human capital and the improvement of resource utilization ability, thus alleviating multidimensional relative poverty. Second, the social capital is consistent with the mechanism of human capital. Third, the mediating effect of risk shock is positive. The risk shocks lead to farmers' loss of income, insufficient investment in feasible capacity, and discourage their enthusiasm for production, thus increasing the possibility of multidimensional relative poverty. Fourth, the mediating T A B L E 10 Relations of regulation intensity (>1.835), resource utilization capacity and multidimensional relative poverty  Note: Standard errors are in brackets. ***, **, and * refer to the significance at the levels of 1%, 5%, and 10%, respectively. effect of social equity can significantly reduce multidimensional relative poverty of farmers. Therefore, in the group with high regulation intensity, the regulation of NRs can increase community equity, provide equal opportunities for farmers' development, and help alleviate multidimensional relative poverty of farmers. Table 12 shows that when the regulation intensity exceeds the inflection point, the path of "regulation intensity-resource utilization capacity-multidimensional relative poverty" has also been verified. Specifically, the mediating effect of human capital, social capital, risk shock, and social equity accounts for 22.78%, 31.67%, 14.44%, and 31.67% of total effect, respectively.
First, in the group with high regulation intensity, the mediating effect of human capital significantly increased multidimensional relative poverty of farmers. The stronger the regulation of NRs, the more restrictions of human capital accumulation of farmers, leading to their multidimensional relative poverty. Second, the regulation intensity in aggravating multidimensional relative poverty through social capital is large. This indicates that the stronger the regulation of NRs, the more difficult it is for farmers to accumulate social capital. Third, the mediating effect of risk shock will significantly increase the multidimensional relative poverty of farmers. If the regulation of NRs is stronger, farmers will not only suffer greater exogenous shock, but also reduce the expectation of income stability, thus leading to multidimensional relative poverty. Fourth, the proportion of social equity is the same as the mediating effect of social capital. As a result, in the group with high regulation intensity, the stronger the regulation of NRs is, the more unfair the community is, and the more unequal the development opportunities are for farmers.
To sum up, the influence path of reserve regulation (i.e. regulated or not) on multidimensional relative poverty through resource utilization capacity did not pass the test. Regulation intensity can affect multidimensional relative poverty through the resource utilization capacity of farmers. Specifically, when regulation intensity does not reach the inflection point, if it is stronger, farmers are more likely to improve their own human capital and social capital. Meanwhile, it is more conducive to social equity, thereby providing equal development opportunities for farmers and alleviating multidimensional relative poverty, but it may cause farmers to be more likely to face great risk shock. When regulation intensity exceeds the inflection point, if it is stronger, it is more difficult for farmers to accumulate human capital and social capital. Moreover, risk shock is greater and the society is more unfair, leading farmers to fall easily into multidimensional relative poverty.

| DISCUSSION AND CONCLUSIONS
Since poverty reduction became one of the Millennium Development Goals of the United Nations and with the implementation of reducing emissions from deforestation and forest degradation in developing countries, the discussion on how NRs affect the poverty of local farmers is becoming more and more intense. Conservationists believe that the NRs is very important to stabilize the global environment (Bruner et al., 2001;Sanderson & Redford, 2003). Antipoverty advocates point out that NRs have obvious positive externalities, so the protection may be at the expense of the interests of local communities (Wilkie et al., 2006). Other scholars find that while the utilization of natural resources is restricted in NRs, the protection can provide multiple economic values to help improve the welfare of farmers (Bakkegaard et al., 2017;Thondhlana & Muchapondwa, 2014). Therefore, the net impact of NRs on poverty may be positive or negative, and the objectives of poverty eradication and environmental protection are not necessarily incompatible (Adams et al., 2004;Oldekop et al., 2015). In addition, the implementation of the regulation of different NRs is heterogeneous, leading to significant differences in the impact of reserve regulation on farmers' poverty (Ferraro et al., 2013). On the one hand, the mandatory one-size-fits-all government regulation behavior is bound to increase the production and living costs of farmers, which will weaken their livelihoods and lead to the lack of feasible capacity. Therefore, reserve regulation has an obvious poverty-causing effect (Angelsen & Wunder, 2003). On the other hand, reserve regulation with compulsion replaces the market to make up for the limitations of market competition mechanism and realize the optimal allocation of resources (especially natural and labor resources), which increases the choices of farmers and improves the resource utilization capacity of farmers, thereby alleviating their relative poverty (Ferraro & Hanauer, 2014).
However, poverty is a dynamic and historical comprehensive concept. With the evolution of time, space, and socioeconomic development, it has evolved from absolute poverty to relative poverty and from single dimension to multi-dimension. The multidimensional relative poverty proposed by Sen (1999) has gradually become a hot spot in the field of poverty study. Sen believes that poverty is not only manifested in the state that is unable to meet basic needs due to low income but also contains the deprivation of individual relative rights owing to the lack and insufficiency of feasible ability. As a dimension for measuring poverty, income is very important, but it cannot be used to reflect other dimensions of poverty (Ye & Wang, 2010). As long as there are social differentiation and inequality, there will be relative poverty and it will become the norm in the process of social and economic development. Exploring the causes of multidimensional relative poverty of farmers in NRs is conducive to alleviating the contradiction between protection and development.
In this paper, a theoretical analysis framework "Reserve regulation, Resource utilization ability, and Multidimensional relative poverty" is constructed. Based on the survey data on 864 households of 17 Panda NRs in Sichuan and Shaanxi, the impact of reserve regulation (i.e. regulated or not) and regulation intensity on multidimensional relative poverty of farmers is evaluated by using Probit model and the mediating effect model. We found that reserve regulation has significantly positive impact on the multidimensional relative poverty of farmers. The existence of reserve regulation makes farmers fall easily into multidimensional relative poverty. Furthermore, through distinguishing different regulation intensities, a U-shaped relationship is found between the regulation intensity and the multidimensional relative poverty of farmers. When regulation intensity is less than the inflection point, strengthening regulation can alleviate multidimensional relative poverty; when regulation intensity is more than the inflection point, strengthening regulation will aggravate the multidimensional relative poverty of farmers.
From the perspective of resource utilization capacity, we empirically test the mechanism of reserve regulation and regulation intensity on the relative poverty of farmers. The result shows that, first, the path of reserve regulation affecting multidimensional relative poverty through resource utilization capacity does not pass the test. The reason may be that reserve regulation has a U-shaped relationship with the multidimensional relative poverty, so the effects will cancel out when the average marginal effect is carried out only through regulation or not. Second, when regulation intensity is less than the inflection point, except the mediating effect of risk shock is significantly positive, the mediating effects of human capital, social capital, and social equity are significantly negative. When regulation intensity is greater than the inflection point, the mediating effects of human capital, social capital, risk shock, and social equity are all significantly positive.
Generally speaking, reserve regulation will have a negative impact on the livelihood of local farmers, which aggravates the multidimensional relative poverty of farmers. However, differences in the implementation of reserve regulation in different NRs lead to differences in the livelihood development of farmers. When regulation intensity is lower than the threshold, increasing regulation intensity is conducive to the improvement of farmers' feasible capacity, and then alleviate their multidimensional relative poverty. Once the threshold is exceeded, its influential direction is just the opposite. This shows that the impact of reserve regulation on farmers' poverty is "too much is as bad as too little." AUTHOR CONTRIBUTIONS Chao Lin: Conceptualization (lead); data curation (equal); formal analysis (equal); investigation (equal); methodology (equal); software (equal); validation (equal); visualization (equal); writingoriginal draft (lead); writingreview & editing (equal). Lan Gao: Data curation (equal); formal analysis (equal); funding acquisition (equal); investigation (equal); methodology (equal); project administration (equal); resources (equal); supervision (equal); validation (equal); visualization (equal); writingreview & editing (equal).