Place Attachment, Self-Efficacy, and Farmers’ Farmland Quality Protection Behavior: Evidence from China

Abstract

Farmland pollution severely threatens humanity’s sustainable development. Exploring farmland quality protection behavior (FQPB) from the farmers’ perspective is considered one of the best ways to solve the farmland pollution problem. This study develops a theoretical framework for farmers’ FQPB from the perspectives of place attachment (consisting of place identity, dependency, and affection) and self-efficacy. We conducted a primary survey of 412 corn farmers from the northwestern Chinese province of Gansu and empirically examined the effects of place attachment and self-efficacy on farmers’ FQPB and verified the moderating effects that self-efficacy exerts on the influence of place attachment on FQPB by using hierarchical regression and propensity score matching models. The results indicate that: (1) among the three dimensions of place attachment, farmers with stronger place identity and place affection are more likely to implement FQPB; conversely, farmers who exhibit stronger place dependency are less likely to engage FQPB; and (2) self-efficacy not only effectively promotes farmers’ FQPB but also has an enhancing effect on the influence of place identity and place affection on FQPB. Our results suggest that policymakers should encourage farmers to maintain a place’s image and guide farmers to participate in place construction; thus, farmers’ place identity and place affection can be fostered. Meanwhile, the government should diversify the income sources of farmers to reduce their dependency on a single source. The finding that self-efficacy effectively promotes FQPB also implied that the formulation of farmland quality protection policies should shift from traditional command-based policies to participatory approaches, utilizing the initiative of farmers to enhance the policy’s effectiveness, which can not only promote farmers’ FQPB through self-efficacy but also strengthen the positive influence of place identity and place affection on FQPB.

1. Introduction

Farmland can fundamentally guarantee human survival; however, it is also characterized by ecological vulnerability. Due to the irresponsible utilization of farmland resources, approximately 80% of farmland worldwide has been damaged to differing degrees [1], which has led to substantial ecological costs [2]. According to the report “Global Assessment of Soil Pollution” published by the Food and Agriculture Organization (FAO) and the United Nations Environment Program (UNEP), approximately one-third of the world’s farmland has been degraded as a result of the improper utilization of agricultural production materials such as pesticide and fertilizer. Farmland pollution is a severe problem [3], and stakeholders worldwide are focusing on devising methods of mitigating it. Numerous nations have formulated farmland protection policies to ensure the sustainable development of farmland, such as the EU Biodiversity Strategy 2030 [4], the Working Land Conservation Programs and the Farmland Protection Program in the United States, and France’s Ecophyto National Plan. Generally, these policies have not yet yielded the anticipated results [5,6]. Farmers, who are subjects of agricultural production and operation, can crucially influence farmland quality protection, which is the primary reason why many studies focus on farmers’ farmland quality protection behavior (FQPB).

Existing research has suggested that farmers’ FQPB may undergo three development phases. The first phase is the empirical exploration phase, during which researchers investigate the impact of demographic and socioeconomic characteristics on farmers’ FQPB. Studies have found that male and younger farmers were more likely to adopt FQPB than female and elderly farmers [7]. Moreover, farmers are more likely to exhibit FQPB in agricultural production if their education levels and annual household incomes are relatively high [8]. Nonetheless, several studies have demonstrated that farmers’ gender status and educational level do not affect FQPB [9]. The aforementioned contradiction among the empirical evidence has prompted researchers to consider FQPB from the human–land relations perspective. Thus, understanding based on the fragmentation of the human–land relationship is the first stage of investigating farmers’ FQPB from the human–land relationship perspective; this stage mainly emphasizes the decisive role of human subjective initiative in the human–land relationship and views human subjective ability as the key to farmland quality protection [10]. Studies from this viewpoint have indicated that farmers’ abilities may affect their FQPB [11], and farmers with stronger self-efficacy are likelier to adopt FQPB [12]. Studies based on the planned behavior theory have shown that perceived behavioral control (self-efficacy) could significantly facilitate farmers’ FQPB [13,14]. Due to the increasing prominence of human–land conflicts, such as the global ecological crisis and farmland pollution, people’s epistemology has been expanding, and their understanding of the fragmented human–land relationship has gradually evolved into an understanding of its interdependence. In the interdependence phase of the human–land relationship, studies have indicated that both place identity [15] and place affection [16,17] can increase farmers’ FQPB. Also, place attachment, which represents multiple aggregates of identity-based, functional, and affective attachments, is an essential part for human–land interdependence [18] and can significantly affect FQPB [19]. However, some studies have demonstrated that a stronger place attachment on farmlands may negatively affect farmers’ FQPB [20]. The contradictory research findings regarding human–land interdependence have created barriers to comprehending farmers’ FQPB.

Over the past century, farmland pollution has caused significantly negative impacts on the agricultural environment in many countries and regions of the world [21,22]. In China, where the population is quite high and land is limited, the strong economic demand for farmland has exacerbated farmland pollution [23]. In fact, the recognition of the interdependence of the human–land relationship, in which farmers value their connection to the land, makes it impossible to detach FQPB from the influence of place attachment [19]. Moreover, self-efficacy, which reflects the subjective initiative of producers, cannot be overlooked either [12]. However, in the current research on FQPB, place attachment and self-efficacy are at different stages of the epistemology of human–land relations, and they do not enter into a unified analytical framework in the existing research, which makes it difficult for us to understand FQPB in a more detailed way. Thus, incorporating place attachment and self-efficacy into the same analytical framework for analyzing farmers’ FQPB is crucial for agricultural managers and policymakers to better understand farmers’ FQPB and formulate more suitable farmland quality protection policies.

In this study, we first develop a theoretical analysis framework of FQPB that both considers place attachment and self-efficacy. Specifically, using primary survey data of corn farmers from the northwestern Chinese province of Gansu, hierarchical regression and propensity score matching models were utilized to investigate the effects of place attachment and self-efficacy on farmers’ FQPB and the potential moderating effects of self-efficacy on the relationship between place attachment and farmers’ FQPB. Endogenous tests and robustness tests were further conducted for the validity and reliability of the results. Thus, the major contributions of this study are tripartite: (1) place attachment and self-efficacy are simultaneously incorporated into the theoretical analytical framework of farmers’ FQPB, so that the objective relationship of interdependence between people and land is combined with farmers’ subjective initiative, which can further deepen our understanding of farmers’ FQPB, and is more conducive to the formulation of the government’s FQPB policy; (2) we investigate how self-efficacy, as a manifestation of farmers’ subjective initiative, could moderate the effects of place attachment on farmers’ FQPB, which provides a novel research perspective for academic research on farmers’ FQPB; and (3) this study focuses on smallholders in China, where the agricultural land per capita is minimal, to provide insight into farmland pollution control for agricultural managers in China and other developing countries with similar situations.

The rest of the paper consists of five sections. Section 2 presents the theoretical framework and research hypotheses; Section 3 presents the research methodology; Section 4 presents the empirical results; Section 5 contains a discussion; and we offer conclusions and policy recommendations at the end of this paper.

2. Theoretical Framework and Hypothesis

2.1. The Effect of Place Attachment on Farmers’ FQPB

Place attachment, as an important variable reflecting human–place relations in environmental psychology, describes the identity, functional, and emotional attachment attitudes of an individual or a group to a specific place [24], and this portrayal of human–place relations undoubtedly acts on farmers and influences their FQPB [19]. From the perspective of environmental psychology’s dimensional division on place attachment, place attachment includes three dimensions: place identity, place dependency, and place affection [24,25], in which place identity is used to reflect an individual’s or a group’s identification with a specific place, place dependency is used to characterize an individual’s or a group’s functional attachment to the land due to productive needs, and place affection is used to embody an individual’s or a group’s emotional connection. At present, place attachment has been widely used in the field of pro-environmental behavior research [15].

Place identity is an important concept in environmental psychology, which refers to the consistency between place and self in terms of image and value [26]. Environmental psychology believes that there is a dynamic relationship between the individual and the physical environment, in which the individual creates an environment that “reveals self-image”, and the environment feeds back “image information” to the individual, thus forming the individual’s place identity [27], so individuals with a strong sense of place identity tend to maintain place identity for the purpose of preserving the homogeneity of their self-image [28]. In the case of farmers, smallholder farmers in developing countries generally have lower incomes, which often makes their relationship with the land stronger, thus making place identity an important part of the internal logic of farmers’ behavior [29]. Farmers with strong place identity frequently exhibit a philosophy that entails aligning their place image with their self-image, which is ingrained in their behavioral logic and influences their behavior [30]. For example, if farmlands are polluted by pesticides, fertilizers, or contaminated water, this may cause a bad image of the place, resulting in an identity contradiction between place image and self-image. Thus, farmers who have strong place identity may recognize the identity contradiction and intend to maintain the consistency of place identity and self-image, and, thereby, are more likely to implement FQPB. Based on these, we propose the following hypothesis:

Hypothesis 1 (H1).

Place identity can promote FQPB.

Place dependency, which refers to the functional attachment to land, affects individuals whose productive needs can be satisfied locally [24], reflecting the ability of a location to provide the appropriate conditions or support for an individual’s desired goals [31]. In feudal private ownership of land-dominated traditional Chinese agrarian society, farmland was the basis of farmers’ livelihoods, which led to an interdependent relationship between farmers and the land. However, due to the foundation of the modern Chinese state and the abolition of feudal land ownership, farmers now possess the freedom to choose their occupations. Diversified income sources [32], such as income from part-time work and labor, have contributed, to a certain degree, to distinctions in farmers’ place dependency, which has led to the differentiation of agricultural production behavior to a certain degree. Currently, place dependency often profoundly influences farm household behavior, and it affects both economic goals and behavioral inertia [33,34]. On one hand, producers whose primary source of income is agriculture may exhibit high-level place dependency. Even if those farmers are strongly motivated to preserve the farmland quality, they may be constrained by the need to survive and earn a living, which may result in the excessive utilization of pesticide and fertilizer to increase crop yield [35]. On the other hand, even if farmers have less economic pressure for living, they may develop path dependence due to inertia behaviors, such as the irrational utilization of pesticide and fertilizer in the past [36], which may reduce the likelihood of employing FQPB. Therefore, the following hypothesis is proposed:

Hypothesis 2 (H2).

Place dependency exerts a disincentive effect on FQPB.

Place affection refers to the emotional bond between humans and places [24]. Academic theoretical studies on individual place attachment have shown that affective motivation is an important prerequisite for individual behavior [37], and that place affection, as an intrinsic driving force, has a significant impact on individual behavior [38,39]. On this basis, Western countries frequently rely on religious traditions to maintain affection between farmers, communities, and land [40]; by contrast, in China, where religious traditions are absent, the “home village” has been the source of affection that has preserved the relationship between humans and places since antiquity [41]. In China, place affection bears a more localized meaning: agrestic complex. The agrestic complex originates from the phrase “fallen leaves return to the roots”, which was developed during China’s millennia-old agricultural civilization and has been internalized into the cultural psychology of Chinese farmers, and the concept is the value logic that underpins farmers’ behavioral choices [42]. Since ancient times, place affection has been ingrained in the behavioral logic of Chinese farmers due to their close relationship with the land, prompting them to protect their farmlands [43]. Therefore, the following hypothesis is proposed:

Hypothesis 3 (H3).

Place affection can promote FQPB.

2.2. The Effect of Self-Efficacy on Farmers’ FQPB

Self-efficacy is an important concept in social cognitive theory, which refers to an individual’s beliefs and judgments regarding their capability to perform a specific behavior in a given scenario [44], and theoretical studies have shown that individuals with weak self-efficacy perceive themselves as incapable of accomplishing a target task [12]. Self-efficacy has been shown to be an important factor influencing farmers’ pro-environmental behaviors and sustainable agricultural development behaviors [45,46]. Since the reform and liberalization, China’s economy has progressively transformed into a market-based economy, which has led to significant changes to the social structure of rural areas [47]. Farmers have transitioned from relatively confined villages to an open modern society, and the enormous changes at the economic and social levels have resulted in changes in their abilities and the formation of behavioral differentiation. The difference in self-efficacy, which indicates whether a person believes he or she is capable or not, is the basis of this behavioral differentiation—not just the intensity of actual ability. Self-efficacy, a crucial basis of subjective motivation, profoundly influences the FQPB of farmers from the belief’s perspective. Farmers with relatively weak self-efficacy tend to believe that they are unable to address the farmland pollution problem. Consequently, they treat the management of farmland pollution negatively [12]. However, producers with relatively strong self-efficacy believe they are accountable for farmland quality protection, and they possess the ability to mitigate farmland pollution. They will, therefore, take the initiative to implement FQPB. The following hypothesis is therefore proposed:

Hypothesis 4 (H4).

Self-efficacy can promote FQPB.

2.3. Moderating Effect of Self-Efficacy

The influence of place identity on farmers’ FQPB essentially maintains the consistency of farmers’ place image with their self-image. Farmers with strong self-efficacy believe that they are capable of adopting FQPB, which can maintain the consistency between place image and farmer’s self-image [26]. Therefore, self-efficacy can strengthen the influence of place identity on FQPB. For economic goals and behavioral inertia, farmers with stronger place dependency tend to prioritize yield maximization over farmland quality protection in agricultural production. However, farmers with high self-efficacy are more likely to incorporate farmland quality protection into their behavioral decision-making system. In this scenario, the effect of profit-seeking on their behavior will be weakened [48]. Thus, self-efficacy reduces the negative impact of place dependency on farmland protection behavior. Moreover, farmers with stronger place affection, influenced by agrestic complex, are more likely to reduce the damage on farmland environment. As a strong belief support [12], self-efficacy can exert a catalytic role in transforming farmers’ place affection into FQPB. Consequently, the following hypotheses are formulated:

Hypothesis 5 (H5).

Self-efficacy can strengthen the role of place identity in promoting farmers’ FQPB.

Hypothesis 6 (H6).

Self-efficacy can weaken the negative influence of place dependency on farmers’ FQPB.

Hypothesis 7 (H7).

Self-efficacy exerts a positive moderating effect on the influence of place affection on farmers’ FQPB.

Based on the preceding analysis, the theoretical framework of this study is depicted in Figure 1.

3. Materials and Methods

3.1. Study Area and Data Source

The data utilized herein are obtained from a primary survey of corn farmers in Gansu Province, northwestern China. In Gansu, the main crops cultivated are corn, wheat, potato, vegetables, Chinese herbal medicine, and oilseeds, with corn having the largest planted area, accounting for 26.3% of the total agricultural land in 2021. Therefore, we selected corn farmers as our survey subjects and conducted the survey in Lanzhou, Dingxi City, Tianshui City, Pingliang City, and Baiyin City (see Figure 2). We selected Gansu province as the study area primarily due to the province’s more severe agricultural pollution [49]. After the implementation of the Returning Farmland to Forest Program in China, the northwest region has become the leading area of pesticide application in China, where Gansu Province exhibits the highest pesticide utilization per unit farming area (i.e., 8.04 kg/ha in 2019) among the entire five northwestern provinces. Also, the problem of farmland pollution in Gansu Province resulting from increased pesticide inputs is one of the main factors that will potentially limit the development of farmland quality for the next decade [50]. In addition, the National Bureau of Statistics of China reports that in 2021, the average amount of agricultural fertilizer applied per unit of farmland area in Gansu Province was 462.02 kg/ha, which considerably exceeds the internationally accepted upper limit for safe fertilizer application (225 kg/ha). The pollution of farmland in Gansu Province has become a crucial problem, and it urgently requires research attention.

Before the formal survey, 30 farm households in Maying village, Yuzhong County, Lanzhou City, Gansu Province, were selected for a preliminary survey, in March 2023. To ensure the reliability and validity of the questionnaire, we revised it based on the preliminary survey results and farmers’ feedback. The formal survey was conducted from March to April 2023 and the questionnaire was comprised of three sections, namely, respondents’ demographic and socioeconomic characteristics, farmers’ FQPB, and farmers’ place attachment and self-efficacy. Stratified random sampling was employed in our study area. Specifically, we chose five cities in Gansu province that exhibited relatively large maize sowing areas. Subsequently, we randomly selected two to three counties (districts) in each city, followed by two to three townships (towns), and two to three villages in each county (district). According to the actual number of farm households in each village, 10 to 15 questionnaires were randomly distributed. Professionally trained investigators were hired to conduct in-depth, face-to-face interviews that lasted an average of 20 min per household to collect data. A total of 480 questionnaires were distributed in this survey. After removing invalid questionnaires, such as those containing incomplete information, 412 valid questionnaires were collected, resulting in an effective response rate of 85.83%.

3.2. Variable Selection

3.2.1. Dependent Variable

Herein, farmland quality protection behavior (FQPB) is the dependent variable. FQPB refer to the behaviors taken by farmers to maintain or improve the quality of farmland during the agricultural production process, including chemical reduction behaviors, such as reduced application of pesticides and fertilizers, conservation farming behaviors, such as returning straw to the field and fallow rotations, and waste disposal behaviors, such as recycling of agricultural films [51,52]. In fact, the excessive use of pesticides and fertilizers by farmers in agricultural production has led to a serious problem of farmland pollution in China [53,54], which is an important reason why the no. 1 document of the Chinese central government in recent years explicitly mentioned the need to reduce the amount of pesticides and chemical fertilizers so as to improve the quality of farmland. In the case of our study population, the rates of straw return to the field and agricultural film use were low, and the adoption of FQPB, such as conservation tillage and harmless treatment of wastes, were rare. Therefore, combining the connotation of farmers’ FQPB, China’s main concerns about farmers’ farmland pollution control, and the actual situation of our research subjects, we focus on the FQPB of farmers’ pesticide and fertilizer reduction and application behaviors. The specific questions in the questionnaire were “I reduced the application amount of pesticide per mu in 2022, compared with 2021” (1 hectare = 15 mu) and “I reduced the application amount of fertilizer per mu in 2022, compared with 2021”. Respondents’ responses for these two questions were measured using a 5-point Likert-type scale, with options from 1 to 5 (strongly disagree to strongly agree). The scores of the two questions are added together and averaged as the farmers’ FQPB.

3.2.2. Independent Variables

Herein, place attachment and self-efficacy are the independent variables. Place attachment includes three dimensions: place identity, place dependency, and place affection [25]. This study referred to the place attachment scale developed by William and Vaske [55] for measuring the three dimensions of place attachment (i.e., place identity, independency, and affection), building on the previous definitions. Specifically, respondents were asked the following two questions to measure place identity: “I identify with and accept the traditional cultural practices of my village” and “I share the same values as other members of my village”. Place independency was measured by the two questions of “This is the best place for me to perform my activities” and “My current residence meets my needs”. The three question items utilized to measure place affection were “I like the place I live in”, “The place I live bears immense significance to me”, and “The place I live is very special to me”. In regard to another core variable of self-efficacy, we referred to the study of Chen et al. [56] to measure farmers’ self-efficacy, and it includes the following three questions: “I believe that my adoption of pesticide and fertilizer application reduction behaviors is effective in protecting farmland”, “I believe that I am capable of contributing to the goal of farmland quality protection”, and “It is easier for me to adopt pesticide and fertilizer application reduction practices”. Both place attachment and self-efficacy were also measured on a 5-point Likert scale.

3.2.3. Control Variables

Referring to the existing research findings on farmers’ FQPB [57,58], this study utilized farmers’ demographic characteristics and socioeconomic characteristics as control variables, with farmers’ demographic characteristics including gender, age, political profile, education level, and knowledge of FQPB, and with the farmers’ socioeconomic characteristics, including annual agricultural income, annual household income, years of agricultural production, planting scale, and distance of the largest area of farmland from the town.

3.3. Methodology

3.3.1. Hierarchical Regression

The dependent variable, herein, is a continuous variable; therefore, we utilized a hierarchical regression model to explore the effects of place attachment and self-efficacy on farmers’ FQPB and the moderating effect that self-efficacy exerts on the effects of place attachment on farmers’ FQPB. To analyze the effect of place attachment on farmers’ FQPB, we utilized the following equation:

$$Y=\alpha +{\beta }_{1}PI+{\beta }_{2}PD+{\beta }_{3}PA+\eta X+\theta$$

(1)

where Y represents farmers’ FQPB; PI represents place identity; PD represents place dependency; PA represents place affection; X denotes a control variable; α denotes a constant term; β₁, β₂, β₃, and η denote regression coefficients; and θ denotes a random disturbance term.

Based on Equation (1), to explore the effect of self-efficacy on farmers’ FQPB, we replaced place identity, place dependency, and place affection with self-efficacy (SE), and the following equation was applied:

$$Y=\alpha +{\beta }_{4}SE+\eta X+\theta$$

(2)

where β₄ denotes the estimated coefficient. To further analyze the moderating effect that self-efficacy exerts on the influence of place attachment on farmers’ FQPB, this study added the interaction terms of place attachment and self-efficacy to the model based on the inclusion of both place attachment and self-efficacy, and the following equations were applied:

$$\begin{array}{l}Y=\alpha +{\beta }_{1}PI+{\beta }_{2}PD+{\beta }_{3}PA+{\beta }_{4}SE+\delta PI\times SE+\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\rho PD\times SE+\sigma PA\times SE+\eta X+\theta \end{array}$$

(3)

where PI × SE, PD × SE, and PA × SE denote the interaction terms of place identity and self-efficacy, place dependency and self-efficacy, and place affection and self-efficacy, respectively, and δ, ρ, and σ denote the regression coefficients. We have centralized place identity, place dependency, place affection, and self-efficacy before introducing interaction terms to avoid multicollinearity.

3.3.2. Propensity Score Matching

The strength of farmers’ place attachment and self-efficacy is usually associated with their demographic and socioeconomic characteristics [12,59], which may lead to endogeneity in model estimation due to self-selection bias (e.g., farmers with larger planting scale tend to have stronger place dependency, and, therefore, they may be reluctant to adopt FQPB) and reduce the accuracy of regression results. This study utilizes the propensity score matching method to control for demographic and socioeconomic characteristics associated with place attachment and self-efficacy; thus, sample self-selection bias is eliminated. The propensity score matching method mainly measures the average treatment effect on the treated (ATT) in the case of the farmer implementing FQPB, which represents the difference between the observed outcome of sample i in the intervention state [60] and its counterfactual illustrated by the place identity example. The following formula is applied:

$$ATT=E\left(Y_{1i}|D_i=1\right)-E\left(Y_{0i}|D_i=1\right)$$

(4)

where ATT represents the difference between the implementation of FQPB by farmer i in the stronger place identity scenario and that in the weaker place identity scenario, D_i = 1 denotes a farmer with stronger place identity, Y_1i denotes the FQPB behavior of a farmer with stronger place identity, and Y_0i denotes the FQPB behavior of a farmer with weaker place identity. E(Y_1i|D_i = 1) denotes the expected value of FQPB implemented by farmers with stronger place identity, and E(Y_0i|D_i = 1) denotes the expected value of FQPB implemented by those farmers who have weaker place identity, if they exhibit stronger place identity.

Since E(Y_0i|D_i = 1) cannot be directly observed in reality, we can verify the influence of place identity on farmers’ FQPB by using the propensity score matching method (PSM) to select samples from farmers with weaker place identity who are matched with farmers with stronger place identity in the scenario of similar propensity scores; thus, a new sample pair can be formed, and sample self-selection bias can be eliminated.

4. Results

4.1. Descriptive Statistics

The descriptive statistical results from our survey samples are presented in

Table 1. Descriptive Statistics of Variables.

Variables (Abbreviation)	Definition	Scale/Units	Mean	S.D.
Dependent Variable
Pesticide	The situation of farmers’ reduction of pesticide application	5-point Likert scale (1 = strongly disagree, 5 = strongly agree)	2.62	1.20
Fertilizer	The situation of farmers’ reduction of fertilizer application	5-point Likert scale	2.43	1.09
Independent Variables
Place identity	The level of farmers’ identity with their place	5-point Likert scale	3.93	0.78
Place dependency	The level of farmers’ dependency on their place	5-point Likert scale	4.02	0.60
Place affection	The level of farmers’ affection with their place	5-point Likert scale	3.31	0.89
Self-efficacy	The degree of belief in one’s own ability	5-point Likert scale	3.18	1.15
Control Variables
Gender	Gender of householder	0 = female and 1 = male	0.54	0.51
Age	Age of householder	year	59.62	11.18
Political profile	Whether subject is a Communist Party member or not	0 = no and 1 = yes	0.28	3.47
Edu	Education level of householder	1 = illiterate, 2 = primary school, 3 = junior high school, 4 = high school, 5 = bachelor’s degree or above	2.16	1.07
Knowledge	Knowledge level of farmland quality protection	5-point Likert scale	2.14	1.30
Agricultural income	The annual income from farming	10,000 yuan	1.31	2.05
Household income	The annual income of households	10,000 yuan	1.59	2.42
Production year	Years in agricultural production	year	39.58	13.66
Scale	Planting scale	ha	0.70	0.57
Distance	Distance of the largest area of farmland from the town	km	7.17	10.53

.

From the results depicted in Table 1, it can be observed that, in regard to farmers’ FQPB, the mean values of farmers’ pesticide application reduction behaviors and fertilizer application reduction behaviors are 2.62 and 2.43, respectively, which are relatively low. The cumulative percentages of respondents who agreed and strongly agreed with the application reduction behaviors of pesticide and fertilizer were 32.76% and 28.16%, respectively, which indicates that farmers’ adoption of FQPB remains at the lower level in our study area and can be enhanced to a large extent.

Regarding independent variables, the mean values of place identity, place dependency, and place affection of farmers were 3.93, 4.02, and 3.31, respectively, all of which exceeded 3; these results indicate that the place attachment of our respondents was generally strong, which might be related to the agrestic complex of Chinese farmers. Moreover, the mean self-efficacy value of our respondents was 3.18, which was somewhat lower compared to place attachment.

For the demographic and socioeconomic characteristics of the respondents in our survey, there were 223 male respondents, who accounted for 54.1% of the total, and the average age was 59.6 years, which is consistent with the fundamental characteristics of “elderly farming” in rural China. The respondents’ education level was relatively low, mainly at junior high school and below (89.7%), and their knowledge of farmland quality protection was low, with only 18.7% of the respondents choosing “agree” and “strongly agree”. Respondents’ annual agricultural income and annual household income were both relatively low, with mean values of 13,100 yuan (approximately US $1869) and 15,900 yuan (approximately US $2269), respectively, and annual agricultural income was still the primary source of income for the respondents in our survey. In addition, respondents have been engaged in agricultural production for a relatively long period, with an average of 39.6 years and an average planting scale of 0.7 hectares, which is consistent with the characteristics of smallholder production and operation in rural China. Thus, our sample data are representative to a certain degree.

4.2. Hierarchical Regression

This study utilizes a hierarchical regression model to analyze the effects of place attachment and self-efficacy on farmers’ FQPB (

Table 2. Regression Results pertaining to the Effects of Place Attachment and Self-Efficacy on Farmers’ FQPB.

Variables	Model 1		Model 2		Model 3
	Coef.	Robust Std. Err.	Coef.	Robust Std. Err.	Coef.	Robust Std. Err.
Place identity	0.216 ***	0.070			0.204 ***	0.068
Place dependency	−0.317 ***	0.091			−0.320 ***	0.091
Place affection	0.222 ***	0.068			0.212 ***	0.066
Self-efficacy			0.148 ***	0.050	0.134 ***	0.046
Gender	−0.192	0.120	−0.256 **	0.123	−0.216 *	0.118
Age	0.007	0.006	0.006	0.007	0.006	0.006
Political profile	0.030 ***	0.006	0.032 ***	0.006	0.031 ***	0.006
Edu	0.016	0.071	0.052	0.069	0.010	0.070
Knowledge	0.046	0.047	0.048	0.047	0.053	0.046
Agricultural income	0.004	0.032	0.008	0.036	0.009	0.032
Household income	0.060 **	0.030	0.073 **	0.031	0.064 **	0.029
Production year	−0.014 **	0.006	−0.012 *	0.006	−0.013 **	0.006
Scale	−0.001	0.008	−0.000	0.008	−0.002	0.008
Distance	−0.001	0.005	0.003	0.006	0.002	0.005
Prob > chi2	0.000		0.000		0.000

Note: ***, **, and * denote the 1%, 5%, and 10% significance levels, respectively.

). The independent variables of Model 1 include only place attachment and control variables; the independent variables of Model 2 are self-efficacy and control variables; and the independent variables of Model 3 include place attachment, self-efficacy, and control variables.

The baseline regression results in Table 2 indicate that in Model 1, place identity, place dependency, and place affection all pass the test at the 1% significance level. Among these three variables, place identity and place affection exert a significantly positive effect on farmers’ FQPB, which indicates that those who have stronger place identity and place affection are more inclined to adopt FQPB. Additionally, place dependency exerts a significantly negative effect on farmers’ FQPB, which indicates that when the place dependency of farmers is stronger, they are less likely to adopt FQPB.

In Model 2, self-efficacy passes the test at the 1% significance level, which indicates that self-efficacy can promote farmers’ FQPB, and when the self-efficacy is stronger, the farmers are more likely to implement farmland pollution control.

The regression results of Model 3 also confirm the results of Model 1 and Model 2. Place identity, place dependency, and place affection still pass the significance test after including all independent and control variables. Based on these results, hypotheses H1, H2, H3, and H4 are initially confirmed.

Additionally, the status of farmers’ gender, political profile, annual household income, and annual agricultural production significantly affects farmers’ FQPB. Specifically, farmers who are female, possess party membership, and possess higher annual household income are more likely to adopt FQPB. However, our results also show that the respondents in our survey become less likely to adopt FQPB when the number of agricultural production years is greater.

4.3. Moderating Effects

To verify the moderating effect of self-efficacy in the influence of place attachment on farmers’ FQPB, this study further adds the interaction terms of place identity, place dependency, place affection, and self-efficacy on farmers’ FQPB on the basis of Model 3 shown in Table 2, and the results are illustrated in

Table 3. Results of the Moderating Effect of Self-efficacy on FQPB.

Variables	Coef.	Robust Std. Err.
Place identity	0.236 ***	0.071
Place dependency	−0.317 ***	0.090
Place affection	0.194 ***	0.070
Self-efficacy	0.140 ***	0.045
Place identity × self-efficacy	0.094 **	0.045
Place dependency × self-efficacy	−0.050	0.070
Place affection × self-efficacy	0.106 **	0.052
Control variables	Controlled
Prob > chi2	0.000

Note: *** and ** denotes the 1% and 5% significance levels, respectively.

.

Compared with the results of Model 3 contained in Table 2, the goodness of fit (i.e., R²) of the results in Table 3 increases by 0.110 (p < 0.001) after adding the interaction terms of place identity, place dependency, place affection, and self-efficacy, which indicates the existence of the self-efficacy moderating effect. From the regression results contained in Table 3, the interaction terms of place identity, place affection, and self-efficacy significantly influence farmers’ FQPB. Since self-efficacy significantly contributes to the positive effect of place identity and place affection on farmers’ FQPB, hypotheses H5 and H7 are confirmed; however, our results show that self-efficacy does not inhibit the negative effect of place dependency on farmers’ FQPB. Hypothesis H6 is not confirmed.

4.4. Propensity Score Matching

This study, which aims to avoid the influence of self-selection bias on the results, divides the sample into two groups, namely, the stronger group (above the mean) and the weaker group (below the mean), as per the means of four variables: place identity, place dependency, place affection, and self-efficacy. Furthermore, we assess the differences in farmers’ FQPB under the stronger and weaker scenarios of place identity, place dependency, place affection, and self-efficacy after balancing the differences in farmers’ demographic and socioeconomic characteristics using a propensity score matching model (PSM).

4.4.1. Binomial Logistic Regression of Propensity Matching Scores

Herein, referring to the study conducted by Zhang et al. [61], the four core independent variables are first considered as the dependent variables, and farmers’ demographic and socioeconomic characteristics are utilized as independent variables in a binomial logistic regression (

Table 4. Binomial Logistic Regression Results for Propensity Matching Score.

Variables	Place Identity		Place Dependency		Place Affection		Self-Efficacy
	Coef.	Robust Std. Err.	Coef.	Robust Std. Err.	Coef.	Robust Std. Err.	Coef.	Robust Std. Err.
Gender	0.424 *	0.244	−0.143	0.245	−0.488 **	0.246	0.030	0.243
Age	−0.010	0.017	0.023	0.016	0.004	0.017	0.019	0.017
Political profile	0.038	0.043	0.072 ***	0.025	0.055 **	0.024	−0.346	0.353
Edu	0.070	0.109	−0.132	0.102	0.298 ***	0.097	−0.053	0.094
Knowledge	−0.091	0.110	0.009	0.108	−0.153	0.160	0.013	0.112
Agricultural income	0.005	0.051	−0.036	0.061	0.091	0.070	−0.019	0.050
Household income	−0.011	0.059	−0.031	0.061	0.192 **	0.084	−0.027	0.062
Production year	0.011	0.013	−0.019	0.012	0.029 **	0.014	−0.014	0.014
Scale	0.015	0.015	0.039 ***	0.015	0.011	0.016	0.033 **	0.015
Distance	−0.001	0.012	−0.021 *	0.012	0.015	0.010	−0.049 ***	0.012
Prob > chi2	0.678		0.012		0.000		0.003

Note: ***, **, and * indicate the 1%, 5%, and 10% significance levels, respectively.

). The variable with significant regression results is selected as the PSM matching variable.

4.4.2. The Average Treatment Effect

Referring to the study conducted by Wu et al. [60], the nearest neighbor matching, radius matching, and kernel matching of PSM are utilized to estimate the average treatment effect on the treated and to assess the effects of place attachment and self-efficacy on farmers’ FQPB. The results are presented in

Table 5. The Estimates of the Average Treatment Effects.

Group	Matching Method	Average Treatment Effect	Standard Deviation	t Value
Place identity	1:3 nearest neighbor matching	0.344 **	0.140	2.45
	Radius matching (caliper 0.01)	0.344 **	0.140	2.45
	Nuclear matching	0.232 **	0.111	2.09
	ATT mean	0.307	-	-
Place dependency	1:1 nearest neighbor matching	−0.273 **	0.130	−2.09
	Radius matching (caliper 0.02)	−0.262 **	0.131	−2.00
	Nuclear matching	−0.293 ***	0.111	−2.65
	ATT mean	−0.276	-	-
Place affection	1:3 nearest neighbor matching	0.432 ***	0.135	3.19
	Radius matching (caliper 0.02)	0.412 ***	0.137	3.02
	Nuclear matching	0.371 ***	0.115	3.21
	ATT mean	0.405	-	-
Self-efficacy	1:1 nearest neighbor matching	0.274 **	0.126	2.17
	Radius matching (caliper 0.02)	0.257 **	0.127	2.03
	Nuclear matching	0.197 *	0.109	1.80
	ATT mean	0.243	-	-

Note: ***, **, and * indicate the 1%, 5%, and 10% significance levels, respectively.

.

The results in Table 5 indicate that the average treatment effect on the treated (ATT) of farmers’ FQPB, grouped by place identity, is significant at the 5% level. The ATT results corresponding to the three methods are 0.344, 0.344, and 0.232, respectively, with a 0.307 mean value, which indicates that after eliminating the self-selection bias occasioned by demographic and socioeconomic characteristics, farmers with stronger place identity are more likely to implement FQPB, and that the likeliness of this group exceeds that of farmers with weaker place identity by 30.7%. In regard to place dependency, the ATT results obtained using the three propensity matching methods are significant after grouping, with magnitudes of −0.273, −0.262, and −0.293, respectively. The results of the three matching methods are comparable, and farmers with stronger place dependency are 27.6% less likely to implement FQPB than farmers with weaker place dependency. Moreover, after grouping place affection, the ATT results obtained using the three propensity matching methods are all significant, with magnitudes of 0.432, 0.412, and 0.371, respectively, and with a mean value of 0.405, which indicates farmers with stronger place affection are 40.5% more likely to implement FQPB compared to farmers with weaker place affection. Finally, regarding self-efficacy group, the ATT results estimated by the three propensity matching methods are also significant, with ATT results of 0.274, 0.257, and 0.197, respectively, and with a 0.243 mean value, which indicates that farmers with stronger self-efficacy are 24.3% more likely to implement FQPB than those with weaker self-efficacy. The aforementioned results reconfirm hypotheses H1 to H4.

4.4.3. Balance Test

To ensure the reliability of the PSM matching results, a balance test is conducted, and the results are illustrated in

Table 6. Results of Balance Test.

Group	Matching Stage	Matching Method	Pseudo-R2	LR	Mean Bias	Med. Bias	B
Place identity	Before matching		0.005	2.58	8.3	8.2	16.7
	After matching	1:3 nearest neighbor matching	0.000	0.11	0.9	0.1	2.8
		Radius matching (caliper 0.01)	0.000	0.11	0.9	0.1	2.8
		Nuclear matching	0.003	2.04	1.5	1.5	12.1
Place dependency	Before matching		0.015	8.00 **	15.5	15.3	27.6 +
	After matching	1:1 nearest neighbor matching	0.008	3.26	2.4	1.9	20.4
		Radius matching (caliper 0.02)	0.008	3.05	2.3	1.8	19.8
		Nuclear matching	0.004	1.80	4.1	4.5	15.4
Place affection	Before matching		0.031	16.37 ***	20.0	18.7	41.2 +
	After matching	1:3 nearest neighbor matching	0.004	1.61	8.4	7.3	13.9
		Radius matching (caliper 0.02)	0.003	1.46	7.8	5.6	13.3
		Nuclear matching	0.001	0.27	3.4	3.4	5.7
Self-efficacy	Before matching		0.030	16.06 ***	26.6	26.6	39.6 +
	After matching	1:1 nearest neighbor matching	0.001	0.40	3.8	3.8	6.8
		Radius matching (caliper 0.02)	0.000	0.24	2.8	2.8	5.2
		Nuclear matching	0.001	0.25	3.2	3.2	5.3

Note: *** and ** indicate the 1% and 5% significance levels, respectively. Pseudo-R² denotes the pseudo-determination coefficient. LR denotes the likelihood ratio statistic. Mean bias represents the mean deviation, and med. bias represents the median deviation. B denotes the absolute standardized deviation. ⁺ denotes a B value that exceeds the 25% critical value.

.

The matching results, which are contained in Table 6, indicate that the Pseudo-R² value, LR statistic, mean deviation, median deviation, and B value decrease after using the three methods of group matching for the four core variables (i.e., place identity, place dependency, place affection, and self-efficacy), and the PSM results all pass the balance test, which indicates that the PSM matching effect presented in this study is practical and reliable.

4.5. Robustness Test

To ensure the robustness of the results, the value of the dependent variable (i.e., FQPB) utilized in the hierarchical regression model is further replaced by summing up the scores of pesticide and fertilizer application reduction behaviors rather than the average scores. The effects of place attachment and self-efficacy on farmers’ FQPB and the moderating effect of self-efficacy are analyzed again, and the results are depicted in

Table 7. Results of Robustness Tests of the Effects of Place Attachment and Self-efficacy on Farmers’ FQPB.

Variables	Model 4		Model 5		Model 6		Model 7
	Coef.	Robust Std. Err.	Coef.	Robust Std. Err.	Coef.	Robust Std. Err.	Coef.	Robust Std. Err.
Place identity	0.433 ***	0.140			0.409 ***	0.136	0.473 ***	0.142
Place dependency	−0.635 ***	0.182			−0.640 ***	0.183	−0.634 ***	0.180
Place affection	0.444 ***	0.135			0.424 ***	0.132	0.387 ***	0.139
Self-efficacy			0.297 ***	0.099	0.267 ***	0.092	0.280 ***	0.089
Place identity × self-efficacy							0.188 **	0.089
Place dependency × self-efficacy							−0.101	0.139
Place affection × self-efficacy							0.213 **	0.103
Gender	−0.385	0.241	−0.513 **	0.247	−0.433 *	0.235	−0.449 *	0.233
Age	0.015	0.012	0.013	0.014	0.013	0.013	0.014	0.013
Political profile	0.060 ***	0.012	0.065 ***	0.013	0.061 ***	0.013	0.061 ***	0.012
Edu	0.032	0.143	0.104	0.138	0.020	0.140	0.016	0.137
Knowledge	0.092	0.094	0.114	0.094	0.107	0.093	0.144	0.095
Agricultural income	0.008	0.064	0.016	0.071	0.018	0.064	0.027	0.064
Household income	0.121 **	0.060	0.146 **	0.062	0.128 **	0.058	0.126 **	0.058
Production year	−0.028 **	0.012	−0.025 *	0.013	−0.027 **	0.012	−0.026 **	0.012
Scale	−0.001	0.015	−0.001	0.016	−0.004	0.015	−0.005	0.016
Distance	−0.002	0.010	0.005	0.012	0.003	0.010	0.005	0.009
Prob > chi2	0.000		0.000		0.000		0.000

Note: ***, **, and * indicate the 1%, 5%, and 10% significance levels, respectively.

.

The results of Table 7 are consistent with the results in Table 2 and Table 3 that the core independent variables of place identity, place dependency, place affection, and self-efficacy all have significant effects on farmers’ FQPB. In regard to control variables, gender, political profile, annual household income, and years of agricultural production remain the key factors influencing farmers’ FQPB. These results indicate that the findings of our study are valid and reliable.

5. Discussion

This study develops a theoretical framework that analyzes the effects of place attachment and self-efficacy on farmers’ FQPB and investigates the potential moderating effects of self-efficacy on the influence of place attachment on farmers’ FQPB. We found several interesting and important results.

First, our results indicated that increasing farmers’ place identity and place affection could significantly facilitate farmers’ adoption of FQPB, which is consistent with the findings of Wang et al. [19], but different from the findings of Mullendore et al. [20]. The results obtained by Mullendore et al. indicate that place identity has no effect on farmers’ FQPB. Wang et al.’s sampling location is Ningbo, China, which is different from ours, but they are all small farmers with small per capita cultivated land area, and the similarity of cultural backgrounds can form similar behavioral logics of the respondents [62]. Mullendore et al.’s study is on American farmers, which is different from ours to some extent, and who are more likely to rent farmland. Therefore, it is possible that U.S. farmers’ place identity may be only limited to their community where they live rather than to the location of the rented farmland [63]. This potential reason may motivate U.S. farmers to maximize the economic value of the rented farmland during the rental period while ignoring the quality protection of the rented farmland. By contrast, Chinese agricultural production is predominantly smallholder-based, with the majority of farmers owning contracted land, which offers farmers a closer relationship with the land [64] and a stronger sense of place identity, thereby promoting FQPB.

The result that farmers with greater place dependency are less likely to adopt FQPB is consistent with the findings of Cross et al. [65] and Li et al. [66]. Cross et al.’s study of farmers in Wyoming and Colorado in the western U.S. showed that farmers with higher place dependency may be reluctant to relinquish potential benefits due to economic considerations, and, thus, be less inclined to pursue farmland quality protection [65]. Li et al.’s study of Chinese farmers found that lower place dependency resulted in a lower likelihood of farmers’ implementation of FQPB, and even in farmland abandonment [66]. In fact, farmers with greater place dependency may be less likely to adopt FQPB due to economic goals and behavioral inertia. Due to survival pressure, farmers with strong place dependency tend to possess a larger proportion of annual agricultural income, and they may overutilize pesticides and fertilizer to ensure agricultural output [35]. From the perspective of farmers’ behavioral inertia, it is difficult for them to change their production behaviors associated with the irrational application of pesticide and fertilizer over the years, and, therefore, they may still heavily rely on pesticide and fertilizer application in agricultural production [34]. It is important to note that this kind of behavioral inertia for fertilizer or pesticides over application is detrimental to the implementation of farmers’ FQPB. Economic goals and behavioral inertia in farmers’ behavior exist in both developing and developed countries [34,67], which indicates that farmers’ economic objectives, behavioral inertia, and a combination of both may reduce the likelihood for farmers with strong place dependency to adopt FQPB.

Our results also revealed that self-efficacy significantly contributes to farmers’ FQPB, which is consistent with existing studies [12,46,68]. Perry’s study showed that farmers with high self-efficacy have better outcome expectations, tend to set more challenging goals, and are more likely to engage in conservation agriculture behaviors than farmers with low self-efficacy [12]. Specifically, farmers with greater self-efficacy are more likely to believe they are accountable for farmland quality protection and capable of mitigating farmland pollution [44]. These farmers are willing to consider the farmland quality protection goal into their decision-making process, and, thereby, are more likely to implement FQPB to avoid farmland pollution with the belief that they have the ability to do so [12].

In regard to the moderating effect, we also found that self-efficacy could facilitate and have a positive effect on the influence of place identity on farmers’ FQPB. This result suggests that farmers with higher self-efficacy typically believe they are capable of adopting FQPB and are able to maintain consistency between their place image and self-image [30], thereby reinforcing the effect of place identity on FQPB. Interestingly, there is no moderating effect of self-efficacy in the negative effect of place dependency on farmers’ FQPB, which is inconsistent with the hypothesis. The results may be attributable to the fact that smallholders in China earn a generally low income and that economic goals are frequently prioritized over other motivations [69]. Affected by economic goals, farmers tend to overapply pesticides and fertilizer to maximize agricultural output and profits [35]. Even if farmers exhibit a high degree of self-efficacy, they may neglect the protection of their farmlands. Finally, our results indicate that self-efficacy can reinforce the effect of place affection on farmers’ FQPB. Therefore, farmers with higher self-efficacy believe they can adopt FQPB, which can lead to the transformation of their place affection into FQPB behavior.

6. Conclusions and Policy Implications

Using primary survey data of corn farmers in Gansu Province, this study empirically investigates the effects of place attachment and self-efficacy on farmers’ FQPB and the moderating effect of self-efficacy on the relationship between place attachment and farmers’ FQPB. The findings bear several important policy implications. First, place identity and place affection positively affect farmers’ FQPB, whereas place dependency exerts a negative effect. This finding indicates that when formulating farmland quality protection policies, policymakers should consider the vital role of place attachment in farmers’ FQPB, actively build/reinforce farmers’ place identity and place affection on their neighborhood, and weaken their place dependency. Effective measures such as encouraging farmers to maintain the place image actively can enhance farmers’ place identity, and guiding farmers to participate in local construction can help maintain farmers’ place affection. Moreover, to reduce farmers’ place dependency, agricultural managers and policymakers should help farmers increase farmland productivity and their income sources through the integrated development of primary, secondary, and tertiary industries. Meanwhile, the government can utilize agricultural production demonstrations or pilot projects to guide farmers from the inertial behavior of excessive pesticide and fertilizer application to that of reduced pesticide and fertilizer application.

In addition, our results indicate that self-efficacy not only effectively promotes farmers’ FQPB but also has an enhancing effect on the influences of place identity and place affection on FQPB. This important finding may imply that the government should change the traditional top–down command-based policy logic in promoting farmland quality protection by allowing farmers to participate in policy formulation as a means of implementing their subjective initiative, thereby increasing the enforceability of the policy, which not only increases the promoting effect of self-efficacy on farmers’ FQPB but also strengthens the positive influences of place identity and place affection on farmers’ FQPB.

This study bears some limitations. Due to budget and time constraints, our survey only focused on corn farmers and was not able to investigate the potential heterogeneous effects of different types of farmers on their FQPB. Moreover, the place dependency herein represents functional attachment; functional attachment can be further subdivided into economic dependency and social relationship dependency, which may affect farmers’ FQPB differently. In addition, place attachment, as a subjective psychological factor, is characterized by contextual dependence, which may vary depending on the gender and age of the respondents, etc. Refinements of these issues above are the authors’ intentions for future research.