1. Introduction
Since the beginning of reform and opening up, China’s rural economic construction has made great achievements while rural ecological pollution has become increasingly severe. Urban pollution has been transferred to the countryside in the process of urbanization. The increase in the consumption level of farmers has led to a rapid increase in the amount of domestic garbage, while its composition has become more complex. However, compared with urban areas, rural environmental management facilities are relatively backward, and farmers’ environmental awareness is relatively weak. Consequently rural environmental pollution has become a stumbling block for the construction of a livable and beautiful countryside, and it is urgent to solve the problem of rural environmental pollution. Although the construction of a new and beautiful countryside has promoted the improvement of the rural ecological environment to a certain extent, ecological problems accumulated over a long period are fundamentally challenging to solve through actions taken over a short period. Rural ecological problems are still the weak point of the policy of comprehensively promoting rural revitalization in the new stage of development [
1]. The report of the 20th National Congress of the Communist Party of China puts forward the need to “promote green development and harmonious coexistence between human beings and nature”, “accelerate the construction of a waste recycling system”, and “promote urban and rural habitat improvement” [
2]. Reduction, reuse, and resourcing, as characteristics of a circular economy in line with the requirements of green development and the construction of ecological civilization, are crucial ways to overcome the contradiction between economic development and environmental protection.
Farmers are the main body of rural production and life and the direct bearer of rural environmental damage, but they are also the direct beneficiaries of rural environmental governance. Therefore, the rural environment relies on the majority of farmers living in rural areas to protect it, and the breadth and depth of their participation largely determine the level of rural environmental governance [
3], thus affecting the development process of the circular economy. With the current village “hollowing out” phenomenon becoming increasingly severe, most young farmers go out to work; the age of the farmers left behind and their degree of education, environmental awareness, and other personal qualities are generally at a lower level. In the case of limited investment in public services in rural areas by government departments, how to promote the construction of a circular economy through the in-depth participation of villagers in rural environmental management is a question we need to think deeply about. According to the social interaction theory, the public is not an independent agent in society, and its behavior is influenced by that of the members of society, resulting in peer effects. The rural community is a traditional “acquaintance society”; farmers’ social interaction is based on their obtained information or the village norms [
4]. The social interaction effect of behavioral decision-making has important policy implications, and the social impact of behavioral spillovers has a multiplier effect, which can effectively enhance policy effects. Therefore, an issue that needs to be further considered is how policymakers can strengthen the interaction among farmers from the external rural environment as well as the internal factors of individual farmers and enhance the interaction effect.
Therefore, based on the collection of first-hand information and data, the following study takes the most widely used household waste disposal behavior in rural environmental management as an example and uses neighbors and neighboring villages as reference groups to study the farmers’ interaction behavior, as well as the peer effects generated by the interaction, in depth. The study results show that farmers’ interaction effects are different in reference groups of different sizes. There are endogenous interaction effects, situational interaction effects, and association effects in neighboring groups. The endogenous interaction effect disappears when the scope of the reference group is expanded to the village. On this basis, effective countermeasures for rural environmental management are proposed from a circular economy perspective.
2. Review of the Literature
The circular economy is a sustainable model of economic development centered on the efficient use and recycling of resources and based on the principles of reduction, reuse, and recycling [
5]. Under increasing environmental pollution and resource consumption, solid waste, if not properly managed, can pose multiple threats to the environment and public health, including water pollution, soil degradation, air pollution, and disease transmission [
6]. Therefore, there is a need for proper waste management to reduce, reuse, recycle, and properly dispose of solid waste, develop circular agriculture [
7], and expand the availability of natural resources to sustain life on earth [
8]. At present, the world’s significant economies generally take the development of a circular economy as the primary path to overcome resource and environmental constraints, cope with climate change, and cultivate new economic growth points; additionally, the development of a circular economy is necessary to promote ecological priority, conservation and intensification, and green transformation of the development model [
9]. Some scholars have proposed that the circular economy is a complex systematic project that requires the participation and collaboration of the central and local governments, enterprises, the public, and many other subjects. In China, because the promotion of circular economy focuses on top-down approaches, government departments play a leading role, meaning the public, enterprises, and other stakeholders do not participate in the main decision-making body enough [
10]. In particular, the participation of farmers is even more insufficient.
Currently, there have been many academic discussions on China’s rural environmental governance, mainly focusing on the current situation of rural environmental governance [
11,
12], the governance model [
13,
14,
15], problems of rural environmental governance [
16,
17], and rural environmental governance strategies [
18,
19,
20]. Many scholars have already explored how the government should play its role in environmental governance [
14,
21], and some scholars believe that government-led “regulatory governance” is no longer suitable for the current social development and ecological environment. The multi-body “interactive governance” model of the government, enterprises, and the public is a governance model that meets the current situation [
15,
17,
22,
23]. Since farmers’ awareness of environmental participation is still weak, some scholars have proposed a “beneficiary subject” approach based on government compensation or incentives to achieve “internal incentives-external compatibility” [
24,
25] in order to mobilize farmers’ motivation to participate. Some other scholars advocate the model of autonomous rural governance [
13]. Most of the above literature explores the model of rural environmental governance from a “macro” perspective. However, in rural environmental governance, farmers are the direct bearers of rural environmental damage and the main body of rural environmental governance [
26]. Therefore, studying the interaction effect of environmental governance from the “micro” perspective of farmers will effectively improve the efficiency of rural environmental governance.
Current research on interaction effects has been explored in the areas of corporate decision-making behavior [
27], individual economic decision-making behavior [
28,
29], consumer spending behavior [
30], student health [
31], crime [
32], and learning [
33,
34]. However, there are few quantitative studies on the interaction effects of farmers, which are mainly focused on agricultural production technology and economic decision-making. In terms of agricultural production, in the 1990s, Case tested the interaction effect with a study of regions adopting new agricultural technologies [
35]. Later, many scholars also tested the interaction effect among farmers. Foster, Munshi, et al. tested the interaction effect of adopting new agricultural technologies by taking villages adopting new technologies as their research subject [
36,
37]. Matuschke I proposed that social network interaction could spread information effectively among farmers [
38]. Luo Qing examined the interaction effects of agricultural technologies among neighborhood, kinship, and plot-adjacent clusters based on field survey data from Mengzhai Village in Henan Province, one of the few studies to analyze the interaction effects [
39] quantitatively. Skevas T (2022) investigated the role of peer effects on farmers’ decision to adopt drones [
40]. Ziheng Niu (2022) analyzed the role of peer effects among different groups of farmers based on survey data of farmers in the North China Plain region [
41]. In terms of economic decision-making, Durlauf, in his discussion on the mechanism of social interaction’s influence on residents’ financial decision-making behaviors, classified the channels of social interaction’s influence on residents’ insurance purchasing decisions into endogenous and situational interactions. Endogenous interaction emphasizes that residents’ decision-making behavior is influenced by others while theirs, in turn, influences others. In situational interaction, residents’ decision-making behavior is only influenced by others and does not, in turn, influence others, which is a kind of “demonstration effect” [
42]. Moran J R et al., in a study of the impact of social interaction on the choice of medical insurance for the elderly, found that social interaction hurts the purchase of commercial medical insurance behavior of residents. That is, the higher the level of social interaction, the lower the likelihood of purchasing commercial medical insurance; less educated people are more likely to be affected by social interaction, thus reducing their likelihood of purchasing commercial medical insurance [
43]. Regarding environmental governance, He Xingbang studied the effect of social interaction on individual environmental behavior and found that social interaction is heterogeneous for different environmental behaviors. Social interactions significantly impact daily environmental behaviors and minorly impact non-daily environmental behaviors [
44]. Social networks as a carrier of residents’ interaction can significantly improve the level of residents’ domestic waste classification, and the more frequent the network interaction, the higher the level of domestic waste classification. Network interaction as an information carrier can promote information dissemination and cultivate residents’ environmental awareness and behavior. Shi Hengtong also believes that all social network variables positively affect farmers’ participation in watershed ecological management behavior [
45]. Farmers with higher levels of social networks can access more information resources on watershed ecological management, and specific monitoring and constraining mechanisms for watershed ecological management among these farmers will increase the willingness of farmers to participate in collective rural action.
To summarize, there have been many studies on the circular economy and China’s rural environmental governance. However, most are from the “macro” perspective, exploring the mode of rural environmental governance and the participating subjects. There needs to be more quantitative literature on the effects of farmers’ participation in environmental governance from a “micro” perspective. Research on the interaction effect between farmers primarily focuses on agricultural production and farmers’ economic decision-making. The research field is relatively single, the research object is mainly confined to a specific region or village, the reference group is not precise enough, the research subject is rough, and there is a lack of empirical examination of the “micro” subjects. Whether the precision of the research subjects and reference groups will affect the identification of the interaction effect is yet to be proved. Based on the shortcomings of related academic studies, the following is a quantitative empirical test of the interaction effects of domestic waste disposal in rural environmental management from a circular economy perspective, with individual farmers and villages as a whole as the research subjects.
3. Theoretical Analysis and Research Hypothesis
Social interaction theory suggests that individuals’ behavioral decisions are not independent; they depend not only on their characteristics and the environment in which they live but are also directly influenced by the behavior and characteristics of other individuals in the reference group. This influence occurs in three main ways. The first form of influence is social norms. Behavioral decisions of the reference group are likely to be regarded as some social norms, especially in relatively closed rural societies, and individuals follow such social norms to form social interaction effects. In the same village, if the surrounding neighbors adopt a more environmentally friendly way of disposing of household waste, then the farmers themselves will adopt a more environmentally friendly way of disposing of household waste for the sake of face or due to the influence of the “circle doctrine.” The second form of influence is strategy complementarity. The behavioral decisions of the reference group may affect the marginal benefits of individual behavioral decisions, thus forming a social interaction effect [
46]. The rural environment is a public good for villagers in the same village. If neighbors adopt more environmentally friendly ways to dispose of household waste, it will also bring a better living environment for farmers themselves. The third form of influence is information exchange. The daily communication between the reference group and the actors will generate information transfer and learning, and the information output by others and self-learning will also affect the actors’ decision-making choices. Neighbors may convey the benefits of environmentally friendly behaviors to farmers in their daily communication, thus improving farmers’ waste disposal behaviors.
Manski (2000) classified social interaction theory, categorizing interaction effects into endogenous, situational, and association effects [
47]. Endogenous interaction effect refers to an individual’s behavior being affected by the behavior of the reference group members and their own decisions, in turn, affecting the behavioral decisions of the reference group members. In other words, endogenous interaction means that the individual and the reference group members are influenced by each other. Several scholars have used the effect of the average performance of cohort members on individual output to identify endogenous effects [
48]. However, because social effects have a lag in the transmission process, some scholars identify endogenous interaction effects by assuming that the behavior of their contemporaries influences individuals; some authors assume that there is a time lag of several years, while others assume that social effects operate across generations. Contextual interaction effects are when an individual’s behavior changes in response to the exogenous characteristics of the people around them, and this influence is unidirectional. The correlation effect refers to the fact that the behavior of individuals within the same group is consistent because they share similar individual characteristics, institutional environments, and other objective conditions. While endogenous and situational interactions express how a behavioral agent may be influenced by his or her social environment, correlation effects express a non-social phenomenon. Social influences affect individuals with a lag, while the effects of non-social forces are simultaneous. Taking farmers’ domestic waste disposal behavior as an example, there is an endogenous interaction effect if, other things being equal, farmers’ domestic waste disposal behavior changes with the domestic waste disposal behavior of their neighbors, other villagers, or other reference groups. A contextual interaction effect exists if farmers’ domestic waste disposal behavior varies with the exogenous characteristics of the cluster. A correlation effect exists if farmers in the same village adopt similar domestic waste disposal behaviors because they live within the same social norms or have similar upbringings.
It is essential to distinguish between the role of endogenous interaction effects, situational interaction effects, and correlation effects because these assumptions imply different predictions of the impact of public policy [
47]. Social multiplier effects can only occur if endogenous interaction effects are present. As an example of a policy intervention on farmers’ environmental behavior, some villagers in a village are educated about the environment and persuaded about environmental behavior, but not other villagers. If the per capita annual centralized collection of domestic waste of the villagers being educated improves as the per capita annual centralized collection of domestic waste of the villagers in the village improves, then the environmental publicity and education of this government department not only restrains the environmental behavior of individual villagers but also indirectly regulates the environmental behavior of all the villagers in the village as their environmental behaviors are regulated and feed back into the environmental behaviors of individual villagers being educated and persuaded to further enhancement. Contextual interaction and association effects imply no such feedback.
Figure 1 illustrates the mechanism of action of the interaction effect.
Based on the above analysis, the following hypotheses are proposed:
Hypothesis 1. There are endogenous interaction effects on farmers’ household waste disposal behavior.
The endogenous interaction effect is a two-way influence with a time lag. When farmers choose to dispose of their household waste, they are influenced by their neighbors and change their household waste disposal behavior, which affects the average level of the whole reference group and feeds back to the farmers to further regulate their household waste disposal behavior more rationally. Suppose the surrounding neighbors collect and dump their garbage at the centralized location. In that case, the farmers will also tend to collect and dump their garbage at the centralized location, increasing the amount of centralized garbage disposal in the village as a whole. In the next stage, the amount of garbage disposal in the village will continue to increase along with the increase in the average amount of garbage disposal in the village. In daily life, it is challenging to avoid communication and interaction between farmers; the dissemination of information will affect the farmers’ decision-making and judgment because the choice of individual action will be affected by the awareness of the effect of their behavior on others and the expectations of others on their behavior [
49]. Farmers choose to centralize the dumping of domestic waste because others expect them to do so, and their neighbors also centralize the dumping of domestic waste because they expect them to do so.
Hypothesis 2. There are contextual interaction effects on farmers’ household waste disposal behavior.
The contextual interaction effect is a unidirectional influence where farmers’ habits are created in interaction with the external environment [
50]. Farmers will only be influenced by their surrounding neighbors when choosing their domestic waste disposal behavior, but their own behavioral decisions will not affect their surrounding neighbors in reverse. If their neighbors collect and dump their domestic waste, the farmers will be influenced by their neighbors to collect and dump their domestic waste. However, the farmers’ behavior will not, in turn, affect their neighbors. This behavior is the result of normative social influence. Normative social influence refers to an individual’s behavior remaining consistent with the rest of the group to maintain his or her relationship with the social group, better adapt to the group, and not be rejected by the group. This behavior is sometimes pressured rather than voluntary [
49]. Therefore, if farmers are surrounded by neighbors who collect and dump their household waste at a centralized location, they will adjust their behavior and choose to dispose of their waste in the same way as their neighbors to adapt to the group.
Hypothesis 3. There are correlated effects between farmers’ household waste disposal behavior.
The correlation effect refers to individuals within the same group behaving similarly because they share similar personal characteristics, institutional environments, and other objective conditions. If all farmers in the same village choose to collect and dump household garbage at a centralized location, it may be because the village rules and regulations are stricter or there are stronger rewards and penalties for environmental behavior; on the other hand, if all farmers in the same village dispose of garbage randomly, it may be because the village rules and regulations are very loose. The relevant environmental policy constraints must be implemented because the villagers generally have a low literacy level. The villagers’ low level of education reflects the institutional environment’s significant influence on individual behavioral choices.
4. Study Area
The areas investigated by the research team include Fujian, Anhui, and Shaanxi provinces, which are representative of the eastern, central, and western regions of China. Among them, Fujian Province is located in the southeast coastal area of China, belongs to the eastern region, and has a relatively developed economy. The specific research areas in Fujian Province include Minqing County, Gutian County, Sha County, the Yanping District of Nanping City, Wuyishan City, the Xinluo District of Longyan City, and the Yongding District. Minqing County is a county under the jurisdiction of Fuzhou City, Fujian Province. It is located in the east of Fujian Province, northwest of Fuzhou City and the lower reaches of the Minjiang River. Gutian County is a county under the jurisdiction of Ningde City. It is located in the northeast of central Fujian Province, connecting with Ningde and Luoyuan to the east, Nanping and Jianou to the west, Minqing and Minhou to the south, and Pingnan to the north. Shaxian County belongs to Sanming City, Fujian Province. It is located between the Wuyi Mountains and the Daiyun Mountains. The Shaxi River runs through the whole area. The terrain slopes from the two sides to the middle. Yanping District is a district under the jurisdiction of Nanping City, Fujian Province. It is located in the north-central part of Fujian Province, at the confluence of Jianxi and Xixi and at the lowest point of the central Fujian valley. Wuyishan City is affiliated with Nanping City, located in the northwest of Fujian and at the junction of Fujian and Jiangxi. It is the location of a famous scenic spot, Wuyishan. Xinluo District is a district under the jurisdiction of Longyan City, Fujian Province, located in the west of Fujian Province. Yongding District is located in the southwest of Fujian Province, bordering Nanjing County in the east, Pinghe County in the southeast, Dapu County and Mei County in Guangdong Province in the southwest, Shanghang County in the northwest, and Xinluo District in the northeast. It belongs to Longyan City, Fujian Province.
Anhui Province is located in the central and eastern parts of China and is the most dynamic part of the Yangtze River Delta. It belongs to the central region, and its economy is relatively underdeveloped. The research areas in Anhui Province include Changfeng County, Feidong County, Funan County, and Jieshou City. Changfeng County is affiliated with Hefei City. The county is located in the north of Hefei City, the provincial capital, and connects with Hefei, Huainan, and Bengbu. Feidong County is affiliated with Hefei City and is located in the central part of Anhui Province. It is connected to county-level Chaohu City, the Nanqiao District of Chuzhou City, and Quanjiao County in the east, Chaohu Lake in the south, Yaohai District, Baohe District, and Changfeng County in the west, and Dingyuan County in the north. Funan County is a county under the jurisdiction of Fuyang City, Anhui Province. It is located in the northwest of Anhui Province, on the north bank of the junction of the upper and middle reaches of the Huaihe River and across the river from Gushi County and Huaibin County in Xinyang City, Henan Province to the south. Jieshou City is a county-level city under the jurisdiction of Anhui Province. It is managed by Fuyang City. It is located in the northwest of Anhui Province, at the border of the northwest of Anhui Province and the junction of Henan and Anhui provinces.
Shaanxi Province is located in the hinterland of China, along the middle reaches of the Yellow River. It belongs to the Western region, and its economy is relatively underdeveloped. The research areas include Lintong District, Xi’an City, Baota District, Yan’an City, Jingbian County, and Yulin City. Lintong District is under the jurisdiction of Xi’an City, Shaanxi Province. It is located in the middle of the Central Plains, the east gate of the ancient capital of Xi’an. It is adjacent to Lishan Mountain in the south, Weinan High-tech Industrial Development Zone in the east, Chanba Ecological Zone and Xinzhu International Port Area in the west, and Yanliang National Aviation Industry Base in the north. Baota District belongs to Yan’an City, Shaanxi Province, and is located in the northern part of Shaanxi Province. It is adjacent to Yanchang in the east, Ansai in the west, Ganquan, Yichuan, Fuxian in the south, and Yanchuan and Zichang in the north. Jingbian County belongs to Yulin City, Shaanxi Province. It is located in the northwest of Shaanxi Province, on the southern edge of the Mu Us Desert. According to the topography, it is divided into three areas: a windy beach area in the north, the Liangmaojian area in the middle, and a hilly and gully area in the south. The temperature is high, the precipitation is basically the same, and the sunshine is less.
Table 1 shows the basic situation of the surveyed rural areas, including population and land area. It can be seen that there are three towns with a population of less than 10,000, nine towns with a population of 10,000–20,000, seven towns with a population of 20,000–30,000, five towns with a population of 30,000–40,000, three towns with a population of 40,000–50,000, two towns with a population of 50,000–60,000, and two townships with a population of more than 60,000. There are six townships with a land area of less than 10,000 hectares, twelve with a land area of 10,000–20,000 hectares, and eleven with a land area of 20,000–30,000 hectares. There are two townships with a land area of more than 40,000 hectares.
5. Data Source, Variable Selection, and Model Setting
5.1. Data Source
The article relies on the National Social Science Fund Project, obtains first-hand information through the field research of the research team members, and uses quantitative statistical methods for statistical analysis. The research time was from July 2017 to August 2018, lasting one year. The research locations were Fujian, Anhui, and Shaanxi provinces, representing the east, middle, and west of China.
The survey methods are: 1. Province selection: first, select Fujian, Anhui, and Shaanxi provinces to represent the eastern, central, and western parts of China; 2. City Selection: According to the list of areas provided by the government departments with rural environmental contiguous improvement or rural environmental comprehensive improvement projects, randomly select 2–5 cities in each province; 3. Counties (districts) selection: randomly select 1–2 counties or districts with this project in the selected cities; 4. Township (street) selection: select 1–7 townships in the selected counties or districts; 5. Village selection: Randomly select 1–6 administrative villages in the selected townships or streets; 6: Research object selection: Randomly select 1–6 farmers in the selected administrative villages to conduct face-to-face questionnaire surveys.
The survey content is as follows: Field surveys are used, and oral questionnaires are used to obtain the basic situation of rural households, the situation of domestic waste disposal of rural households, the subjective cognition of rural households, the willingness of rural households to protect the environment, and the situation of rural social capital. Among them, the household waste treatment part includes two stages: taking 2013 as the node, including the situation before 2013 and the current stage, and assessing the domestic waste disposal of farmers before 2013 (denoted as T1) and at the current stage (denoted as T2). The final research area includes three provinces, ten prefecture-level cities, and one hundred and three administrative villages. Nearly 530 households were surveyed, and 529 questionnaires were obtained. According to the research purpose of this paper, after excluding invalid questionnaires, a total of 525 valid questionnaires were obtained. The sample distribution is shown in
Table 2.
The results of the statistical analysis of the samples are shown in
Table 3. Regarding gender, males accounted for 52.20 percent of the survey sample and females for 47.80 percent, with the proportion of males slightly higher than that of females, which aligns with the current situation in rural areas. Regarding marital status, 95.22% of the survey respondents were married. In terms of age level, the older the age, the higher the proportion of survey respondents, with the highest proportion (42.45%) being over 60 years old. In rural areas, most young and middle-aged people go out to work, leaving only the elderly and children in the villages, so more than 40% of the survey respondents are elderly. 92.93% of the respondents are not village cadres, and the proportion of village cadres is only 7.07%. The proportion of party members is only 9.64%, indicating that the number of party members in rural areas is still relatively small. In terms of education level, most of the respondent’s education level is concentrated in elementary school and junior high schools, with the highest proportion of elementary school education (35.76%), followed by junior high schools (34.80%) and illiteracy (18.93%), while the proportion of senior high school education and above is only 10.51%. Combined with the survey respondents’ age distribution, most were older and generally had a lower level of education, so the education level was mainly concentrated in junior high school and below.
5.2. Variable Selection and Reference Groups Classification
Combined with the existing literature and actual research, in the data processing, only villages with dedicated personnel responsible for managing household waste disposal were selected as the sample data; food waste, pesticide bottles, plastic bottles, used cartons, and used plastic bags were selected as the range of household waste, and direct throwing into garbage collection points, sorting into garbage collection points, and selling were selected as the centralized household waste disposal behavior. The explanatory variables were selected as the annual centralized disposal amount of household waste per capita and the sum of the annual centralized disposal weight per capita for food waste, pesticide bottles, plastic bottles, used cartons, and used plastic bags. Among them, the weights of food waste and used cartons can be directly derived from the questionnaire. Pesticide bottles and plastic bottles are reflected as quantities in the questionnaire, so it is necessary to convert pesticide bottles and plastic bottles into weights, and the weights of used plastic bags need to be light and are therefore neglected. Feng Chengyu (2011) calculated the weight of waste pesticide packaging plastic bags and plastic bottles in his paper. It concluded that the average weight of waste pesticide packaging plastic bags and plastic bottles was 7.5 g, while the average weight of plastic bottles was 18 g [
51,
52].
In order to test the endogenous interaction effects, contextual interaction effects, and correlated effects of farmers’ behavior regarding household waste disposal, the following aspects were selected as explanatory variables based on the availability of survey data and concerning the existing literature for the neighborhood reference groups. The first is the average annual centralized household waste disposal per capita in the reference groups in 2013. This indicator is further explained here by assuming that five farmers A, B, C, D, and E, were surveyed in a village. If A is the leading group, the reference groups of A are B, C, D, and E. If B is the leading group, the reference groups of B are A, C, D, E, and so on. The year 2013 was chosen because it takes time for the interaction effect to occur, especially for a habitual behavior such as household waste disposal, which needs time to change gradually, and the year 2013 was the completion date of the first batch of rural environmental improvement projects. The second explanatory variable is the average characteristics of individual farmers’ groups. The average age, number of members of the reference groups, average years of education, average health status, and the average length of time living at the home of the group they belong to are calculated in the same way as the annual centralized household waste disposal amount per capita of the reference groups. The third explanatory variable is the individual characteristics of farmers. These include their age, number of household members, years of education, health status, and length of time living at home. The explanatory variables for the annual centralized household waste disposal per capita of individual farmers in 2017 are shown in
Table 4.
According to the above theoretical analysis, we selected neighbors and neighboring villages as reference groups to distinguish the size of the reference groups, as shown in
Table 5. The neighboring group in this study refers to other surveyed subjects within the same village, except for the farmers, which are within 1000 m of each other. Neighboring villages in this study refer to other surveyed villages in the same township or district, except for the surveyed villages. Due to the actual situation of the research, the number of neighbors of each farmer and the number of neighboring villages are sometimes different. Therefore, we selected a number of neighbors ranging from one to twelve. In other words, the number of reference groups for each farmer is at least one and at most twelve, and the neighboring villages are all neighboring villages in the same township or district. The number of reference groups is at least one and at most five.
In processing the data of the reference groups of neighboring villages, the individual data of survey respondents in the same village are averaged and the final result is used as the data of that village. Suppose a total of four villages, A, B, C, and D, are surveyed in a town; then the reference groups for village A are B, C, and D. Suppose a total of five farmers, A, B, C, D, and E, were surveyed in village A. The data of each group of these five farmers were summed and averaged as the data of village A. From this, the age, number of household members, years of education, health status, and length of residence at a home in village A could be obtained. By analogy, the data of B, C, and D can be obtained.
5.3. Model Setting
According to the requirements of the Manski interaction effect identification strategy in the above theoretical analysis, the different variables of endogenous interaction effects, contextual interaction effects, and correlated effects are included in the model, which is able to overcome the measurement problems of simultaneity, endogeneity, and omitted variable bias in the following manner:
In the above equation, is the annual centralized household waste disposal per capita of individual farmers at ; refers to the information of individual farmers at ; refers to the exogenous characteristics of the reference groups of farmers at , such as the average age and education level of the reference groups’ members; is the average annual centralized household waste disposal per capita of the reference groups’ members who constitute the individual farmers at at , which is also a proxy variable for the treatment effect of the reference groups in the rural household waste treatment. Where is the exogenous eigenvector of the reference groups of farmers, , excluding itself, and , denotes the reference groups of . Correlated effects exist if , endogenous interaction effects exist in the group if , and contextual interaction effects exist in the group if .